**Photodynamic Therapy of Ovarian Carcinoma Cells with Curcumin-Loaded Biodegradable Polymeric Nanoparticles**

**Lili Duse 1,**†**, Michael Rene Agel 1,**†**, Shashank Reddy Pinnapireddy 1, Jens Schäfer 1, Mohammed A. Selo 2,3, Carsten Ehrhardt <sup>2</sup> and Udo Bakowsky 1,\***


Received: 20 May 2019; Accepted: 13 June 2019; Published: 15 June 2019

**Abstract:** Accumulation of photosensitisers in photodynamic therapy in healthy tissues is often the cause of unwanted side effects. Using nanoparticles, improved bioavailability and site-specific drug uptake can be achieved. In this study, curcumin, a natural product with anticancer properties, albeit with poor aqueous solubility, was encapsulated in biodegradable polymeric poly(lactic-*co*-glycolic acid) (PLGA) nanoparticles (CUR-NP). Dynamic light scattering, laser Doppler anemometry and atomic force microscopy were used to characterise the formulations. Using haemolysis, serum stability and activated partial thromboplastin time tests, the biocompatibility of CUR-NP was assessed. Particle uptake and accumulation were determined by confocal laser scanning microscopy. Therapeutic efficacy of the formulation was tested in SK-OV-3 human ovarian adenocarcinoma cells post low level LED irradiation by determining the generation of reactive oxygen species and cytotoxicity. Pharmacologic inhibitors of cellular uptake pathways were used to identify the particle uptake mechanism. CUR-NP exhibited better physicochemical properties such as stability in the presence of light and improved serum stability compared to free curcumin. In addition, the novel nanoformulation facilitated the use of higher amounts of curcumin and showed strong apoptotic effects on tumour cells.

**Keywords:** PLGA; nanoscaled drug delivery; LED; cancer; serum stability; reactive oxygen species; cellular uptake

#### **1. Introduction**

Light has long been used in the treatment of diseases as different as psoriasis, rickets or cancer [1]. With an alarming mortality rate, cancer is one of the most prevalent causes of death worldwide [2]. Hence, tremendous efforts have been made to develop novel safe, selective and effective cancer therapies. One such treatment, photodynamic therapy, exemplifies a novel minimally invasive tumour targeting therapy, in which tumour tissues can be selectively destroyed by three main mechanisms [3,4]. In the first case there is a combination of three individually inert entities viz. a photoactive drug molecule (photosensitiser; PS), oxygen and a light source of a specific wavelength, which upon combining, activate the photosensitiser and exhibit toxicity towards cancer cells and tissues [5,6]. This process of photosensitisation hugely relies upon the presence of molecular oxygen and forms

the basis for a successful photodynamic therapy (PDT) [7]. After absorbing energy from the light source, the PS interacts with molecular oxygen via energy transfer process or electron transfer process and generates reactive oxygen species (ROS), which oxidise cellular and subcellular organelles to induce either apoptosis or necrosis leading to cell death [8,9]. Since cytotoxic ROS occurs only after irradiation of the PS, tumour tissues can be targeted with high precision. Due to its high reactivity and short half-life, ROS generation only targets cells lying in close proximity of the site of irradiation [1]. The half-life of singlet oxygen in biological systems is ~0.01–0.04 μs, with a restricted site of action of a radius spanning 0.01–0.02 μm [5,10]. PDT can also damage the tumour-associated vasculature, leading to tumour infarction. Finally, PDT can activate the immune response against tumour cells. All three mentioned mechanisms can also influence each other and act synergistically. PDT has an added advantage of almost no or minimal effect towards healthy tissues, making this therapy site specific [11]. It could also be combined with other therapies or immune-stimulatory agents like microbial adjuvants or cytokine for T-cell therapy and adoptive cellular therapy, respectively [1]. Another advantage of PDT is its effectiveness against otherwise chemoresistant cell types [12]. Furthermore, the photosensitiser can be administered by various means, such as systemically, locally or topically [1].

For the success of any PDT, it is important to choose an optimal PS. Despite many studies performed using different PS, only a few have reached the stage of advanced human clinical trials or even U.S. Food and Drug Administration (FDA) approval for clinical use [13]. A very promising photosensitiser for photodynamic treatment of tumours is curcumin. Curcumin (diferuloylmethane), is a naturally occurring yellowish polyphenol extract from the rhizomes of turmeric (*Curcuma longa*), which is cultivated widely in south and southeast tropical Asia [14]. Curcumin is the most active component of turmeric, making up approx. 0.5–3.14% of the dry weight of turmeric powder. Curcumin is used for a variety of therapeutic activities against many different diseases and conditions such as skin diseases, pulmonary and gastrointestinal systems, aches, pains, wounds, sprains, and liver disorders [15].

Offering a variety of potential applications, curcumin has been commonly used as a food-colouring agent (E100) as well as in pharmaceutical research due to its anti-inflammatory, anti-oxidative, anti-carcinogenic and anti-microbial effects [16,17]. It is a proven anticancer agent and therefore a perfect choice for a photosensitiser [18]. However, curcumin exhibits a very low level of bioavailability, since it is a highly lipophilic substance (solubility 11 ng/mL in phosphate buffer saline, PBS; pH 5) [19]. Systemically absorbed curcumin is prone to an extensive first-pass metabolism and undergoes a fast metabolic reduction with biliary excretion [20]. In 1978, Wahlstrom and Blennow observed the uptake of curcumin using Sprague–Dawley rats. After oral administration of 1 g/kg curcumin only negligible amounts of curcumin could be found in blood plasma [21].

These limitations can be overcome by the use of polymeric nanoparticles such as carrier systems, wherein the active substance is protected from degradation in the physiological environment and the bio-membrane permeability and cellular uptake of highly hydrophobic molecules are facilitated [17]. Several studies have revealed a remarkable enhancement of curcumin's phototoxicity through nanoparticle encapsulation [22–26]. One such polymer is PLGA, which is an FDA and European Medicines Agency (EMA) approved biocompatible and biodegradable polymer [27,28]. In the body, PLGA undergoes hydrolysis wherein the endogenous metabolite monomers lactic acid and glycolic acid are and easily metabolised by the body (to water and carbon dioxide) via Krebs cycle [27,29,30]. The degradation time of PLGA and the release of drug from the nanoparticles mainly depends on the nature of copolymer composition [31]. In this study, PLGA 50:50 was chosen due to its ability to hydrolyse much faster than other types of PLGA with a half-life (50% loss of molecular weight) of 15 days for microspheres [32,33].

Another important aspect in PDT is the choice of the irradiation device. There is an array of radiation devices available for PDT such as lasers or lamps [13]. In the current study, with an aim to make PDT more cost effective, efficient and safe, we have utilised a custom manufactured prototype low level light emitting diode (LED) array for the irradiation of PS [34]. LEDs are energy efficient, generate less thermal energy in the form of heat and are versatile in terms of structural form. They have the ability to emit light over a wide area with a relatively homogenous output, which enables treatment of larger lesions in fewer therapy sessions. Using a combination of curcumin loaded PLGA nanoparticles and LED-based PDT, superficial tumours (melanomas and lymphomas) and accessible adenocarcinomas (such as ovarian and cervical) could be treated effectively. The PLGA nanoparticles can be administered intratumourally or could be surface modified to enable systemic application.

In the present study, we have exploited the combination of biodegradable PLGA nanoparticles and LED-based PDT using curcumin as a photosensitiser. The nanoparticles have been characterised for their size and surface charge using Dynamic light scattering (DLS) and laser Doppler anemometry (LDA), respectively. Structural morphology was analysed using atomic force microscopy (AFM) and transmission electron microscopy (TEM). The photo-destructive effects have been evaluated in vitro by cytotoxicity assays and confocal laser scanning microscopy (CLSM). The cellular uptake in SK-OV-3 cells was also analysed using three different uptake inhibitors. Irradiation experiments were carried out in SK-OV-3 tumour cells and the oxidative stress induced during the irradiation was analysed by determining the inhibition of ROS. Furthermore, haemolysis assays using fresh blood, serum stability using serum and activated partial thromboplastin time tests (aPTT) using plasma were used to demonstrate the biocompatibility of the curcumin-loaded polymeric nanoparticles.

#### **2. Materials and Methods**

#### *2.1. Materials*

Curcumin (95% purity) was obtained from Alfa Aesar (Ward Hill, MA, USA); PLGA (Resomer RG 503 H) was supplied by Evonik (Essen, Germany); poly(vinyl alcohol) (PVA, Mowiol 4-88) was purchased from Kuraray (Hattersheim, Germany); polysorbate 80 (Tween 80), 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT), 2',7'-dichlorofluorescin diacetate (DCFDA) and tert-butyl hydroperoxide (TBHP) were obtained from Sigma-Aldrich Chemie (Taufkirchen, Germany). Inhibitors of cellular uptake, viz. Filipin III, chlorpromazine and dynasore were also obtained from Sigma Aldrich. Ultrapure water, generated by a PURELAB flex 4 device (ELGA LabWater, High Wycombe, UK) was used for all experiments. For cell culture studies, ultrapure water was additionally autoclaved and filter-sterilised using 0.2 μm polyethersulphone membrane filters (Sarstedt, Nümbrecht, Germany) prior to use. HPLC-grade ethyl acetate (VWR, Darmstadt, Germany) was used to prepare the organic phases. All other chemicals used were of analytical grade. All buffers used in this study were prepared in the laboratory, unless stated otherwise.

#### *2.2. Light Source*

The LED device used in this study was custom made by Lumundus (Eisenach, Germany) to be usable with microtiter plates. The device is equipped with an array of two different LEDs, capable of emitting light at wavelengths of 457 nm and 620 nm, respectively. Irradiation time, current (i.e., 20, 40, 60, 80, and 100 mA) and wavelength settings are adjustable as per the energy requirement. Radiation intensity was calculated based on the current and irradiation time.

#### *2.3. Cell culture*

The human ovarian adenocarcinoma cell line SK-OV-3 was procured from American Type Culture Collection (ATCC, Manassas, VA, USA). SK-OV-3 cells were cultivated in Iscove's modified Dulbecco's medium (Capricorn Scientific, Ebsdorfergrund, Germany) supplemented with 10% foetal bovine serum (Sigma-Aldrich, Taufkirchen, Germany) at 37 ◦C and 7% CO2 in humidified atmosphere. The medium was replaced every other day and the cells were sub-cultured upon reaching 80% confluency.

#### *2.4. Formulation of Nanoparticles*

Curcumin-loaded nanoparticles (CUR-NP) and unloaded PLGA nanoparticles (NP) were prepared from PLGA by the emulsion–diffusion–evaporation technique as previously described by Kumar et al. [35]. In a pilot study, the method was optimised by varying formulation parameters (e.g., concentration of PLGA, PVA and curcumin; homogenisation speed and time). Briefly, 200 mg of PLGA were dissolved in 5 mL ethyl acetate at room temperature. Curcumin stock solution was prepared by dissolving curcumin in ethyl acetate (2 mg/mL). Equal volumes of PLGA and curcumin stock solutions were mixed and filtered through a 0.45 μm nylon syringe filter (Pall Corporation, New York, NY, USA). The organic solution was added dropwise to an aqueous solution of 2% PVA (*w*/*w*) and the resulting emulsion was stirred in a sealed tube for about 3 h, before homogenising for 10 min at 13,400 rpm using an Ultra-Turrax T25 homogeniser (IKA-Werke, Staufen, Germany). Nanoprecipitation was induced by adding water at a constant rate of 120 mL/h using a syringe pump (Perfusor, B. Braun, Melsungen, Germany). Finally, the organic solvent was evaporated by stirring overnight at room temperature in a dark environment. The nanosuspension was adjusted to a final volume of 50 mL and a concentration of 0.1 mg/mL curcumin. Unloaded nanoparticles were prepared by the same protocol, except that pure ethyl acetate was used instead of the curcumin stock solution. To remove agglomerates, the nanoformulations were centrifuged at 2000× *g* for 45 s using an Eppendorf Centrifuge 5418 (Eppendorf, Hamburg, Germany). The resulting pellet was discarded, and the supernatant was washed three times with water to separate the nanospheres from non-encapsulated curcumin. Between each washing step, centrifugation at 16,000× *g* for 45 min was performed. For prolonged storage, the CUR-NP were freeze-dried using an Alpha 1-4LSC lyophiliser (Martin Christ Gefriertrocknungsanlagen, Osterode am Harz, Germany) using 0.5% PVA as a cryoprotectant. Afterwards the nanoformulations were stored at 4 ◦C protected from light.

#### *2.5. Dynamic Light Scattering and Laser Doppler Anemometry*

To determine particle size distribution and zeta (ζ)-potential of the particles, DLS and LDA were used, respectively (Zetasizer Nano ZS, Malvern Instruments, Malvern, UK). A viscosity of 0.88 mPa × s and a refractive index of 1.33 of water at 25 ◦C were assumed for data interpretation. Measurement position and laser attenuation were automatically adjusted by the instrument. The instrument performs 15 size runs per measurement with each lasting 10 s. For ζ-potential measurements, the instrument automatically performs 15–100 runs per measurement, depending upon the sample. All samples were diluted with filtered PBS buffer (1:100) [36]. Data from at least three independent experiments were measured for both DLS and LDA analysis.

#### *2.6. Morphological Characterisation*

To investigate the morphology of CUR-NP and to confirm size data obtained by DLS measurements, AFM was performed using a Nanowizard 3®(JPK Instruments, Berlin, Germany) and a Digital Nanoscope IV Bioscope (Veeco Instruments, Santa Barbara, CA, USA), respectively. Formulations were diluted 1:100 with water and 20 μL of the diluted sample were placed onto a silica wafer or an untreated microscopic glass slide. The samples were left to settle onto the surface for a few minutes and the remaining fluid was absorbed by a lint-free wipe (Kimtech Precision Wipes, Kimberly-Clark, Fullerton, CA, USA). Measurements were performed in tapping mode, in which the cantilever oscillated with determined amplitude close to its resonance frequency, with scan rates from 0.5 to 1 Hz. A HQ:NSC16/AL\_BS (Anfatec Instruments, Oelsnitz, Germany) cantilever was used [37]. Data were processed using JPKSPM data processing software (v. 5.1.8, JPK instruments).

For the TEM analysis, the nanoparticle suspension was applied onto 300-mesh copper grids. Samples were then negative stained thrice with 2% uranyl acetate (pH 4.2), which was alternated by washing steps with water. The samples were then allowed to dry overnight before being examined under the TEM (LEO 912 AB, Carl Zeiss, Jena, Germany) [22].

#### *2.7. Yield, Encapsulation E*ffi*ciency and Loading Capacity*

The freeze-dried nanoparticles were weighted, and the percentage yield was calculated using the following equation

$$\% \text{Yield} = \frac{\text{Dry weight of nanopparticles}}{\text{Weight of drug and polymer used for NP preparation}} \times 100\tag{1}$$

The encapsulation efficiency (%EE) was determined by extracting curcumin from nanospheres that were previously washed and separated from free curcumin, as previously described in Section 2.4. The nanosuspension was mixed with an equal volume of acetonitrile (a solvent for PLGA and curcumin) and sonicated for 20 min. The absorbance of this solution was measured spectrophotometrically at 425 nm using a Multiskan GO micro plate reader (Thermo Fischer Scientific, Waltham, MA, USA), and the amount of drug quantified using a calibration curve recorded with known curcumin concentrations. For determination of the total drug content, nanoparticles without any washing procedure were measured in the same way. The %EE (*w*/*w*) was calculated as follows:

$$\% \text{EE} = \frac{\text{Drug}\_{\text{encapusulated}}}{\text{Drug}\_{\text{total}}} \times 100\tag{2}$$

In addition, the loading capacity (LC) was calculated using the equation:

$$\% \text{LC} = \frac{\text{Drug}\_{\text{encapsolated}}}{\text{Dry weight of nanoppartitions}} \times 100\tag{3}$$

#### *2.8. In Vitro Drug Release*

For determination of drug release, 10 mg lyophilised nanoparticles were redispersed in 30 mL PBS buffer (pH 5.5 and pH 7.4) containing 1% polysorbate 80 (*v*/*v*) to assure sink conditions. This dispersion was stored under absence of light in a shaking incubator (TH 15/KS 15, Edmund Bühler, Bodelshausen, Germany) at 37 ◦C under slight agitation (50 rpm). For the following seven days, 500 μL samples were drawn and replaced with fresh medium. The samples were centrifuged for 10 min at 16,000× *g* (Centrifuge 5418, Eppendorf AG) and the amount of curcumin present in the supernatant was determined spectrophotometrically at 425 nm, using Multiskan GO micro plate reader (Thermo Fischer Scientific).

#### *2.9. Serum Stability*

To study the stability of the formulations in physiologically relevant conditions, 2 mL serum was diluted with 20 mM HEPES (pH 7.4) to make a 60% serum solution. Curcumin-loaded PLGA nanoparticles were added in a ratio of 1:5 (*v*/*v*). The mixture was incubated at 37 ◦C in a shaking incubator at 100 rpm for 24 h. Samples were further diluted to a ratio of 1:20 (*v*/*v*) with 20 mM HEPES (pH 7.4) prior to DLS and LDA analyses at different time points. All measurements were carried out in triplicates [38].

#### *2.10. In Vitro Irradiation Experiments*

Cells were seeded onto in 96-well plates (10,000 cells/0.35 cm<sup>2</sup> (per well); Nunclon Delta, Thermo Fisher Scientific, Waltham, MA, USA) and were allowed to adhere overnight. Various concentrations of CUR-NP and free curcumin (dissolved in DMSO) were added to the wells and incubated for 4 h. Irradiation was performed at a wavelength of 457 nm with a radiation fluence of 8.6 J/cm2 and the plates were incubated overnight. Dark (un-irradiated) plates were used as control [39]. Subsequently, the medium was replaced with fresh medium containing MTT dye and cells were incubated for 4 h. Absorbance of the resulting formazan crystals dissolved in DMSO was measured using a plate reader

(FLUOstar Optima; BMG Labtech, Offenburg, Germany) at 570 nm. Viability of untreated cells was considered as 100%.

#### *2.11. Cellular Uptake Studies*

To determine the cellular uptake mechanism, SK-OV-3 cells were seeded at a density of 10,000 cells/0.35 cm2 (per well) in 96-well plates (Nunclon Delta, Nunc/Thermo Fischer Scientific, Waltham, MA, USA) and incubated overnight. The next day, cells were washed with PBS containing Ca2<sup>+</sup> and Mg2<sup>+</sup> (pH 7.4) and were incubated for 30 min with pharmacological inhibitors of different endocytic mechanisms viz. dynasore (80 μM), chlorpromazine (14 μM) and Filipin III (8 mM) [40]. After incubation with the uptake inhibitors, cells were washed again with PBS and exposed to 50 μM CUR-NP or free curcumin (dissolved in DMSO) for 4 h. After irradiation at 457 nm with a radiation fluence of 8.6 J/cm2, cells were incubated with MTT and the absorbance from resulting formazan crystals (dissolved in DMSO) was determined at 570 nm as described above.

#### *2.12. Reactive Oxygen Species*

ROS were determined using 2',7'-dichlorofluorescin diacetate (DCFDA, Abcam, Cambridge, UK) according to the supplier's protocol with slight modifications [34]. Briefly, SK-OV-3 cells were seeded in 96-well plates as mentioned above. The next day, cells were incubated with CUR-NP for 4 h. Tert-butyl hydroperoxide (TBHP, 50 μM) was used as positive control. Cells were subsequently washed using PBS containing Ca2<sup>+</sup> and Mg2<sup>+</sup> (pH 7.4) and supplemented with fresh medium. Irradiation was carried out at 457 nm with a radiation fluence of 8.6 J/cm2. The cells were washed again with PBS and incubated with culture medium (IMDM without phenol red) containing 25 μM DCFDA for 1 h. Cell culture lysis reagent (Promega, Mannheim, Germany) was used to lyse the cell and fluorescence was recorded using a FLUOstar Optima plate reader (λex 480 nm/λem 520 nm).

#### *2.13. Intracellular Visualisation of PDT*

Cells were seeded onto 12-well plates (90,000 cells/3.5 cm<sup>2</sup> (per well); Nunclon Delta) containing cover slips (15 mm in diameter). The plates were incubated for 24 h, before being used for the experiments. CUR-NP suspension was added dropwise to each well. After 4 h, the supernatant was removed and cells were washed twice with PBS containing Ca2<sup>+</sup> and Mg2<sup>+</sup> (pH 7.4). Cell were treated with a 4% formaldehyde solution for 20 min to fix them. The nucleus was counterstained with 0.1 μg/mL 4',6-diamidino-2-phenylindole (DAPI) for 20 min. Finally, the cells were washed again with PBS (pH 7.4) and the cover slips were mounted onto slides. Samples were examined under a LSM700 confocal laser-scanning microscope (Carl Zeiss Microscopy, Jena, Germany).

#### *2.14. Activated Partial Thromboplastin Time Test*

An aPTT test was performed as previously described, using a Coatron M1 coagulation analyser (Teco, Neufahrn, Germany) to determine the effect of the formulations on blood coagulation [41]. Briefly, 25 μL of plasma was mixed with 25 μL each of sample and of aPTT reagent followed by addition of an equal volume of pre-warmed 0.025 M calcium chloride solution to activate coagulation. Coagulation was determined spectrophotometrically.

#### *2.15. Haemolysis Assay*

To determine the effect of the formulations on blood, human erythrocytes were isolated from fresh blood as described previously [42]. Briefly, following prior consent from the donor, whole blood was collected into EDTA tubes as centrifuged to separate the red blood cell (RBC) pellet. The RBC pellet was washed thrice with PBS buffer (pH 7.4) and diluted to a ratio of 1:50 with PBS. The erythrocytes were incubated together with the formulations in V-bottom microtiter plates placed in an orbital shaker KS4000 IC (IKA Werke) for 1 h at 37 ◦C. Supernatants were collected from the plates following centrifugation and the absorbance was determined at 540 nm using a plate reader (FLUOstar Optima). PBS (pH 7.4) and 1% Triton X-100 were used as controls.

#### *2.16. Statistical Analysis*

All experiments were performed in triplicates and data are presented as mean ± standard deviation, unless otherwise stated. Two-tailed Student's *t*-test was performed to identify statistical significance differences.

#### **3. Results**

#### *3.1. Physicochemical Characterisation*

Preliminary experiments were performed to determine an ideal stabiliser concentration, homogenisation speed and curcumin content. In emulsion–diffusion–evaporation technique, PVA acts as a stabiliser of the nanodroplets and is one of the widely used emulsifier for polymeric nanoparticle preparations. Both, particle size and size distribution are affected by the concentration of PVA in the aqueous phase during particle preparation [35]. After testing different PVA concentrations, we narrowed down to 2% PVA considering the physicochemical properties of the resultant nanoformulations. The Z-average data from DLS as well as LDA results are summarised in Table 1. CUR-NP and NP showed a narrow size distribution indicating a high reproducibility. No extensive change in particle size was noticeable after curcumin loading. This was also the case for lyophilised samples, thereby showing that lyophilisation has no deleterious effects on the physicochemical characteristics. Furthermore, LDA revealed a slightly negative ζ-potential at pH 7.4, regardless of curcumin loading for both formulations.

**Table 1.** Hydrodynamic diameter, polydispersity index (PDI) and ζ-potential of curcumin-loaded poly(lactic-*co*-glycolic acid) (PLGA) nanoparticles (CUR-NP) and unloaded PLGA nanoparticles (NP). Lyophilised samples represent samples resuspended after lyophilisation. Hydrodynamic diameters are expressed as a measure of particle size distribution by intensity. All samples were measured in triplicates (*n* = 9, independent formulations) and results are expressed as means ± SD.


PLGA in an aqueous environment undergoes hydrolysis spanning over several weeks, depending on the ratio of lactic acid to glycolic acid. The PLGA 50:50 used in this study therefore exhibits the fastest degradation [43]. To increase the storage stability the nanoparticles were lyophilised, using PVA as cryoprotectant. Freeze-dried and resuspended nanoparticles showed no extensive change in particle size (CUR-NP +0.3 nm, NP −1.8 nm) or polydispersity index (data not shown).

For the determination of encapsulation efficiency, a direct method of dissolving the purified particles and measuring the actual amount of curcumin available in the nanoparticle formulation was used. Since curcumin is near to insoluble in water, un-encapsulated free curcumin should exist in crystalline form. In this case the crystals would sediment during ultracentrifugation and be present in the pellet. Hence discarding the supernatant would not lead to sufficient separation of nanoparticles from crystalline curcumin. Therefore (in contrast to NP), a first separation step of centrifugation with a relatively low force of 2000× *g* is necessary to sediment the larger crystals. The optimised formulation of CUR-NP showed a relatively high encapsulation efficiency (EE) with an actual loading capacity of 2% (Table 2). The good EE results can be explained by the fact that PLGA and curcumin are both soluble in ethyl acetate, which was employed for the nanoparticle preparation process. Moreover, PLGA's ability to effectively encapsulate curcumin has been demonstrated in previous studies [44–46].

**Table 2.** Theoretical load, percentage of yield, encapsulation efficiency and loading capacity of curcumin-loaded poly(lactic-*co*-glycolic acid) (PLGA) nanoparticles (CUR-NP) and unloaded PLGA nanoparticles (NP). All samples were measured in triplicates (*n* = 9, independent formulations) and results are expressed as means ± SD.


<sup>a</sup> % *w*/*w* loading of curcumin to polymer; EE, encapsulation efficiency; LC, loading capacity.

#### *3.2. Morphological Characterisation*

In all images (Figure 1), round-shaped CUR-NP with a monomodal size distribution are clearly visible. Due to dilution of pure nanoparticle samples, even single CUR-NPs with a smooth surface indicating that the curcumin is completely incorporated could be visualised. The AFM size analysis was in agreement with the DLS measurements. It should however be noted that the differences in size arising from the DLS and AFM measurements is because the hydrodynamic diameter is obtained in aqueous conditions whereas the latter is measured under atmospheric conditions [47]. Furthermore, a PVA corona was noticeable in TEM micrographs (Figure 1C). While free PVA was removed through centrifugation and washing of the redispersed nanoparticles prior to each experiment, it is a known fact, that small quantities of PVA still remain on the surface of PLGA nanoparticles, which cannot be removed even by extensive washing. This strong adsorption of PVA may be caused by hydrophobic bonding of PVA's hydroxyl groups to the acetyl groups of PLGA [48].

**Figure 1.** Morphology of representative curcumin-loaded poly(lactic-*co*-glycolic acid) (PLGA) nanoparticles (CUR-NP) shown by atomic force microscopy (AFM) (**A**,**B**) and transmission electron microscopy (TEM) (**C**). (**A**) Freshly prepared nanoparticles, (**B**,**C**) nanoparticles post lyophilisation. All AFM micrographs are presented in height mode. Samples were negatively stained with 2% uranyl acetate prior to TEM measurement. Scale bars in TEM micrographs represent 200 nm.

#### *3.3. In Vitro Drug Release*

Different mechanisms were reported for the drug release from PLGA nanoparticles: (i) desorption of drug absorbed on the particles surface, (ii) diffusion through the polymer matrix, (iii) erosion of the polymer matrix, and (iv) a combination of erosion and diffusion processes [29,49].

The cumulative drug release of curcumin from CUR-NP at pH 5.5 and 7.4 is shown in Figure 2. Due to curcumin's low water solubility (~0.01 μg/mL at pH 5 and ~0.4 μg/mL at pH 7.3) and low stability at neutral to basic pH, the addition of a solubility-enhancing component was necessary to assure sink conditions and to achieve UV/VIS detectable concentrations [50]. We chose polysorbate 80 for this purpose, since it is reported, that it is capable of increasing the solubility of hydrophobic drugs and protecting them against degradation through the formation of micelles [51].

**Figure 2.** Cumulative in vitro drug release of curcumin from CUR-NP in phosphate buffer (pH 5.5 and 7.4) containing 1% polysorbate 80 to assure sink conditions. The inset represents the first 12 h of the release study in a different scale (hours instead of days).

CUR-NP showed a typical biphasic release pattern of PLGA nanoparticles with a burst release of around 20% loading within the initial 4 h, followed by controlled release of 90% drug loading over the following 7 days. Burst release is most likely attributed to the drug absorbed to the particle's surface, while controlled release might be induced through a combination of erosion and diffusion processes [52,53]. Furthermore, drug release was much faster in acidic medium compared to pH 7.4 which could be caused by a faster degradation rate of PLGA under acidic conditions [33].

Tumour cells often exhibit an extracellular pH of around 6.5, which would lead to a faster release of curcumin in this acidic environment and therefore an increased cytotoxicity compared to healthy tissue [54,55]. Since the pH of endosomes is even lower (pH ~5.5), drug release of internalised nanoparticles would proceed even faster, leading to higher intracellular curcumin concentration and enhanced tumour damage [56]. Both mechanisms could be used for passive tumour-targeting strategy.

#### *3.4. Serum Stability*

Serum stability assay was performed to evaluate the susceptibility of the formulations. For this purpose, DLS and LDA analyses were performed after incubating the nanoparticles in serum. The experiments were carried at 37 ◦C in a shaking incubator to simulate physiological conditions. The results show shrinkage in the nanoparticle size after 24 h incubation time in serum (Table 3). This could be related to a release of curcumin bound to the particle surface, which leads to particle erosion [57]. Furthermore, the decrease in particle size could be caused by osmotic forces, as reported previously [58]. The increase in the PDI of the nanoparticles could be attributed to the decrease in homogeneity in the presence of serum. After an initial increase (as compared to DLS measurements from native particles), the ζ-potential decreased to more negative values. The initial increase could be due to the accumulation of plasma proteins on the surface of the NP and the subsequent decrease could be attributed to the drug release (curcumin) onto the surface of the NP [59].

**Table 3.** Physicochemical changes of poly(lactic-*co*-glycolic acid) (PLGA) nanoparticles containing 0.1 mg/mL curcumin. Particles were incubated for 24 h in serum at a volume ratio of 5:1. The hydrodynamic diameter is expressed as a measure of particle size distribution by intensity. All samples were measured in triplicates (*n* = 9, independent formulations) and data are expressed as means ± SD.


#### *3.5. Irradiation Experiments*

For the photo destructive effect of PDT, SK-OV-3 ovarian carcinoma cells were incubated with different concentrations of CUR-NP und free curcumin for 4 h and were irradiated with different radiation fluence. The efficacy of the irradiation was determined by MTT assay. As shown in Figure 3A, after treatment of cells, curcumin-loaded biodegradable nanoparticles and free curcumin dissolved in DMSO show a high photocytotoxic effect. There is a remarkable difference between the dark (non-irradiated) and irradiated plates. An incubation time of 4 h was found to be ideal among all the formulations, since burst release of around 20% curcumin loading from CUR-NP was found within 4 h of drug release studies (Figure 2). In the initial studies, a radiation fluence of 1.4, 4.3, 8.6 and 13.2 J/cm<sup>2</sup> were tested using CUR-NP and for the subsequent studies, the fluence was narrowed down to 8.6 J/cm2 for the LED induced PDT (data not shown). Beyond a curcumin concentration of 50 μM, there was no further difference in the effect of PDT. It can also be seen that free curcumin dissolved in DMSO, induced a higher photocytotoxicity effect, which could be attributed to controlled release properties of PLGA nanoparticles, as this might be a rate limiting step for the drug being available for light activation. Nevertheless, since DMSO-dissolved curcumin is not suitable for therapeutic applications and due to its hydrophobic nature, we have used a biodegradable polymeric PLGA nanoformulation to increase the bioavailability of curcumin and to make it suitable for therapeutic applications [60]. The internalisation and subsequent localisation of the photosensitiser within the tumour determine the outcome of the therapy.

#### *3.6. Subcellular Localisation of Curcumin*

To visualise the effect of PDT on tumour cells, qualitative fluorescence microscopic analysis was performed after irradiation (457 nm; 8.6 J/cm2) of the cells with 50 μM CUR-NP and free curcumin dissolved in DMSO. The cells were fixed with 4% formaldehyde solution and the cell nucleus was counterstained with 300 nM DAPI. Substantial intracellular localisation of curcumin near the cell's nucleus could be seen in both dark and irradiated samples, as shown in Figure 3B. Upon irradiation, photo-destruction induced by PDT could be clearly observed in the micrographs, which is evident from the nuclear perforation. This might be due to the chromatin condensation and DNA fragmentation caused by the curcumin induced PDT [8]. Comparing free curcumin dissolved in DMSO with curcumin-loaded nanoparticles reveals that CUR-NPs indeed damage the nucleus while only free curcumin alone is not as damaging although the micrographs show a cellular uptake of curcumin by both the samples. This confirms the hypothesis that curcumin particles are internalised by uptake mechanisms such as endocytosis and free curcumin penetrates the cells by diffusion and loses its effect (Section 3.6). It could also be seen that upon irradiation, the fluorescence of free curcumin increased

and was more distributed than the CUR-NP. Furthermore, the irradiated blank/untreated cells show no photo-destruction.

**Figure 3.** Phototoxic effect of curcumin-loaded poly(lactic-*co*-glycolic acid) PLGA nanoparticles (CUR-NP) and free curcumin dissolved in DMSO (CUR) on SK-OV-3 ovarian carcinoma cells: (**A**) for the MTT-assay, either the nanoformulation or free curcumin was incubated for 4 h at 37 ◦C and were irradiated at 457 nm with a radiation fluence of 8.6 J/cm2. Dark was used as negative control and represents cells without irradiation. All samples contain 0.1 mg/mL curcumin and were measured in triplicates (*n* = 9, independent formulations). Results are expressed as means ± SD. (**B**) CLSM micrographs of SK-OV-3 cells incubated with CUR-NP or free curcumin for 4 h at 37 ◦C and subsequent irradiation (457 nm, 8.6 J/cm2). The cell nucleus was counterstained with 0.1 μg/mL DAPI and was fixed with 4% formaldehyde solution. The curcumin concentration in CUR-NP was 50 μM. Nuclear damage is clearly witnessed in the irradiated samples whereas in the dark, the nucleus is intact. Dark was used as negative control and represents cells without irradiation. Scale bars denote 20 μm.

#### *3.7. Cellular Uptake*

Different cells use different endocytic pathways for the internalisation of nanoparticles. To study cellular uptake, several different inhibitors of endocytic pathways were utilised. For this purpose, the dynamin-, clathrin- and clathrin-independent endocytosis pathways were considered. Dynasore, a GTPase inhibitor rapidly and reversibly inhibits dynamin activity, which is essential for membrane fission during clathrin-mediated endocytosis (CME). Chlorpromazine also causes a block in CME by inducing the assembly of adaptor proteins and clathrin, leading to the formation of endosomes which then fuse with lysosomes [61]. Chlorpromazine also interferes with the intracellular clathrin processing [62]. Filipin III derived from *Streptomyces filipensis* is a polyene macrolide antibiotic, which inhibits the raft/caveolae endocytosis pathway. It interacts with cholesterol whose presence and state influence endocytic functions [63]. Figure 4 shows a substantial inhibition of CUR-NP by Filipin III and chlorpromazine. The uptake mechanism is dependent on the cell line and hence the data should be regarded cautiously. In our case, chlorpromazine exhibited an 80–90% inhibition of clathrin-dependent endocytosis of CUR-NP and Filipin III a marked 100% inhibition of caveolae-mediated and lipid-raft mediated endocytosis but both could not inhibit free curcumin. Free curcumin dissolved in DMSO does not seem to be internalised by any of these endocytic pathways suggesting a different internalisation mechanism such as diffusion [64]. The inhibition by dynamin dependant endocytosis by Dynasore showed no effect on the CUR-NP.

**Figure 4.** Cellular uptake of curcumin-loaded poly(lactic-*co*-glycolic acid) (PLGA) nanoparticles (CUR-NP) and free curcumin (dissolved in DMSO) in SK-OV-3 cells in presence of three specific inhibitors. After 30 min incubation time with different inhibitors (i.e., 8 μM Filipin III, 80 μM Dynasore and 14 μM chlorpromazine), cells were incubated with 50 μM of either CUR-NP or free curcumin dissolved in DMSO for 4 h at 37 ◦C and were irradiated at 457 nm with a radiation fluence of 8.6 J/cm2 (IR). The MTT assay was performed at the end of the experiment to determine the effect of the pathway inhibits on internalisation. Viability of untreated cells was considered as 100%. Dark was used as negative control and represents unirradiated samples. Blank represents cells without any inhibitor. All samples were measured in triplicates (*n* = 9, independent formulations) and results are expressed as means ± SD.

#### *3.8. Reactive Oxygen Species*

ROS generation is the backbone of PDT and is a determining factor in the efficacy of the therapy. As shown in Figure 5, curcumin induced the production of cytotoxic reactive oxygen species after

cells were irradiated (50 μM curcumin; 8.6 J/cm<sup>2</sup> radiation fluence). ROS was determined by the measurement of the fluorescence from the lysed cells. Upon incubation with cells, a deacetylated form of DCFDA is oxidised by the resulting ROS to form fluorescent DCF [65]. As expected, curcumin-loaded biodegradable PLGA polymeric nanoparticles exhibited the highest ROS generation followed by free curcumin dissolved in DMSO thereby corresponding to the results of the PDT. The fact that CUR-NP caused the maximum damage to the tumour cells can be explained by the ROS generation, which in this case was intracellular and could be only detected from the cells, which have internalised. Since free curcumin was not internalised by the endocytic pathways, mentioned in Figure 4, the uptake from CUR-NP seems to be higher. As expected, blank (negative control) did not show any ROS generation.

**Figure 5.** Production of cellular reactive oxygen species (ROS) in response to curcumin-loaded poly(lactic-*co*-glycolic acid) (PLGA) nanoparticles (CUR-NP) or free curcumin (0.1 mg/mL). ROS was measured in SK-OV-3 cells after exposure (for 45 min) to 2',7'-dichlorofluorescin diacetate (25 μM) and incubation with CUR-NP or free curcumin for 4 h at 37 ◦C with subsequent irradiation (457 nm, 8.6 J/cm2). Tert-butyl hydroperoxide (TBHP) was used as positive control and non-irradiated cells served as negative control (Dark), whereas Blank represents untreated cells. Values are represented in relative fluorescence units (RFU). All samples were measured in triplicates (*n* = 9, independent formulations) and results are expressed as means ± SD.

#### *3.9. Haemocompatibility*

To study the effect of the nanoformulation on human erythrocytes, haemolysis assay, which determines the amount of haemoglobin released from the erythrocytes upon exposure to NPs (Figure 6A) was performed. To determine the change in coagulation time upon addition of NPs, aPTT was determined (Figure 6B). Upon its release from erythrocytes, haemoglobin reacts with the atmospheric oxygen to form oxyhaemoglobin [66]. The CUR-NP used in the study showed a minimal haemolytic potential. An increase in the aPTT time was also found to be minimal in the case of CUR-NP. The aPTT analysis revealed that the encapsulated curcumin increased the coagulation time by only 4.01 s, whereas upon addition of free curcumin alone, the coagulation time increased to 120.13 s suggesting an interaction of curcumin with the intrinsic proteins or coagulation factors. The normal coagulation time for the plasma was tested to be 32.2 ± 0.1 s and was found to be in the normal range [67]. Coagulation time between 30 and 40 s is considered within acceptable range and coagulation time above 70 s denotes spontaneous and continuous bleeding leaving the patients with the risk of haemorrhage [67–69].

**Figure 6.** Blood compatibility of curcumin-loaded poly(lactic-*co*-glycolic acid) (PLGA) nanoparticles (CUR-NP), unloaded PLGA nanoparticles (NP) or free curcumin: (**A**) haemolysis assay and (**B**) aPTT test were performed. Where applicable, curcumin was used at a concentration of 0.1 mg/mL. Blank represents erythrocytes in the haemolysis assay and blood plasma in the aPTT test, respectively. Triton X-100 was used as positive control in (**A**) and is considered as 100% haemolysis. All samples were measured in triplicates (*n* = 9, independent formulations) and results are expressed as means ± SD. Dotted lines indicate the tolerance/threshold limits.

#### **4. Conclusions**

In this study, curcumin-loaded biodegradable PLGA nanoparticles (CUR-NP) were formulated for photodynamic therapy using a custom manufactured prototype low power LED device. CUR-NP exhibited better physicochemical properties compared to free curcumin, such as improved serum stability. Furthermore, haemocompatibility of curcumin was improved through polymeric encapsulation, minimising the risk of haemorrhages in patients. Successful cellular uptake of CUR-NP was demonstrated with the aid of confocal laser scanning microscopy, which showed photo-destruction and nuclear perforation. Nanoformulation facilitated the use of higher amounts of curcumin, which would otherwise be impossible due to its poor solubility and stability in aqueous media. Curcumin, which elsewise is safe for normal cells, showed cytotoxic effects on tumour cells upon irradiation at a low intensity therefore selectively inhibiting tumour growth. Since production of ROS occurs only upon irradiation of the intracellular photosensitiser, tumour tissues can be targeted with precision. Using LED as an irradiation source offers a number of advantages such as portability, durability and economical compared to lasers. Moreover, LED's have the ability to irradiate a relatively larger area, which could be beneficial for the treatment of large tumours or even skin cancer. The use of visible light of low-energy makes this therapy safer for both the patient and the operator. Finally, the good biocompatibility of this well-documented and highly reproducible formulation process of polymeric nanoparticles are significant advantages of this anti-tumour therapy. Transforming this novel strategy into a safe and effective therapy by combining PDT with other therapies would be our prime research interest in the future.

**Author Contributions:** Study conceptualization, U.B.; methodology, L.D. and M.R.A.; validation, L.D. and M.R.A.; investigation, L.D. and M.R.A.; data curation, L.D. and M.R.A.; visualisation, S.R.P.; manuscript drafting, L.D. and M.R.A.; manuscript review and editing, U.B., C.E., S.R.P., and M.A.S.; project supervision, J.S.

**Funding:** This research received no external funding.

**Acknowledgments:** The authors would like to thank Eva Maria Mohr for her help and technical assistance, Thomas Anders for performing the serum stability studies and Elias Baghdan for his support and motivation. The authors are also grateful to Lumundus GmbH for developing the prototype LED device used in this study and would like to express their gratitude towards Martin Hardt and Sabine Agel from imaging unit of Justus-Liebig-University Gießen.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Factorial Design as a Tool for the Optimization of PLGA Nanoparticles for the Co-Delivery of Temozolomide and O6-Benzylguanine**

#### **Maria João Ramalho, Joana A. Loureiro, Manuel A. N. Coelho and Maria Carmo Pereira \***

LEPABE—Laboratory for Process Engineering, Environment, Biotechnology and Energy, Faculty of Engineering, University of Porto, Rua Dr. Roberto Frias, 4200-465 Porto, Portugal

**\*** Correspondence: mcsp@fe.up.pt

Received: 22 June 2019; Accepted: 8 August 2019; Published: 10 August 2019

**Abstract:** Poly(d,l-lactic-*co*-glycolic) (PLGA) nanoparticles (NPs) have been widely studied for several applications due to their advantageous properties, such as biocompatibility and biodegradability. Therefore, these nanocarriers could be a suitable approach for glioblastoma multiforme (GBM) therapy. The treatment of this type of tumours remains a challenge due to intrinsic resistance mechanisms. Thus, new approaches must be envisaged to target GBM tumour cells potentially providing an efficient treatment. Co-delivery of temozolomide (TMZ) and O6-benzylguanine (O6BG), an inhibitor of DNA repair, could provide good therapeutic outcomes. In this work, a fractional factorial design (FFD) was employed to produce an optimal PLGA-based nanoformulation for the co-loading of both molecules, using a reduced number of observations. The developed NPs exhibited optimal physicochemical properties for brain delivery (dimensions below 200 nm and negative zeta potential), high encapsulation efficiencies (EE) for both drugs, and showed a sustained drug release for several days. Therefore, the use of an FFD allowed for the development of a nanoformulation with optimal properties for the co-delivery of TMZ and O6BG to the brain.

**Keywords:** drug delivery; experimental design; fractional factorial design; O6-methylguanine DNA methyltransferase (MGMT) protein; glioblastoma multiforme

#### **1. Introduction**

Nanomedicine has been arousing increasing interest, since it allows the early diagnosis and monitoring of several diseases and also can increase the efficacy of conventional pharmacological treatments by enabling their controlled delivery [1–4]. Several materials have been widely studied for nanoparticles (NPs) development. Among polymeric materials, poly(d,l-lactic-*co*-glycolic) (PLGA) is perhaps the most used, due to being FDA-approved, biocompatible, biodegradable, and having tunable physicochemical properties [5]. Since these polymeric NPs allow the release of drugs in a controlled and sustained manner for long periods, the required drug doses and administration frequency can be minimized, decreasing the toxicity of the encapsulated drug [6]. Additionally, PLGA NPs are up-taken by endocytic mechanism, circumventing the p-glycoprotein-mediated cellular efflux, enabling drug accumulation in the target cells. Also, since this polymer can be easily functionalized with different materials, the design of NPs with diverse targeting moieties can be achieved [7].

Therefore, several efforts have been conducted in seeking the development of PLGA NPs for novel therapies with several therapeutic drugs without reducing their bioavailability or activity. The entrapment of therapeutic drugs in PLGA NPs also allow a reduction in their toxicity, and therefore these NPs have been extensively studied for cancer treatment. Most particularly, several attempts to encapsulate temozolomide (TMZ) in PLGA NPs for brain delivery have been conducted [8]. TMZ, the gold standard treatment for glioblastoma multiforme (GBM), presents several limitations, as it

has low bioavailability and high toxicity, reducing its pharmacological activity [9]. Therefore, TMZ encapsulation in a nanocarrier emerges as a suitable approach to increase its chemotherapeutic efficacy, since it will avoid drug elimination after administration and increase its accumulation in the target tissues. GBM is the brain cancer with a high mortality rate, since its classical therapy fails to effectively cure the disease [10]. Thus, it is urgent to find new approaches for its treatment.

In fact, different PLGA delivery nanosystems were developed for TMZ in the last decade. Jain et al., 2014 developed PLGA NPs for TMZ entrapment. Although exhibiting promising results, modifying the NPs' surface with a targeting moiety to improve the specificity of the nanocarriers could be a suitable strategy to enhance its efficiency [11]. Later, Ananta et al., 2016 prepared PLGA NPs that were not able to maintain a favorable sustained release of TMZ [12]. Also, some authors used active targeting strategies to increase the NPs specificity to GBM cells. In fact, Jain et al., 2011 modified the surface of PLGA NPs with transferrin molecules to promote transport across the blood–brain barrier (BBB). The developed nanocarriers enhanced the accumulation and cytotoxic effect of TMZ in mice's brains [13]. Lee et al., 2016 also functionalized PLGA NPs with folate molecules to increase the NPs accumulation in target cells, but the prepared nanosystems did not exhibit good encapsulation efficiency (EE) values [14]. Our group (2018) also developed anti-transferrin receptor monoclonal antibody modified PLGA NPs for the delivery of TMZ. The obtained results showed that PLGA NPs are a promising approach for GBM treatment with TMZ [15].

However, GBM patients usually exhibit low sensitivity to therapy due to intrinsic resistance mechanisms. The O6-methylguanine DNA methyltransferase (MGMT) protein is highlighted as one of the main causes of therapeutic failure of TMZ [16]. Therefore, new strategies to decrease resistance to therapy are necessary. A promising approach is the use of molecules that are able to revert or inhibit these intrinsic resistance mechanisms. Most recently, co-treatment with O6-benzylguanine (O6BG) has been explored to decrease the resistance to TMZ's therapy by binding to the MGMT protein, leading to its inhibition, and consequently hindering the repair of the damaged DNA [17,18]. Nanoencapsulation of O6BG may reduce its toxicity in healthy tissues by targeted delivery. Accumulation of O6BG in healthy tissues is undesirable to avoid inactivation of the MGMT protein in these tissues, and consequently exacerbates the toxicity of alkylating agents, such as TMZ [19].

Also, systemic administration of two free drugs usually leads to infective pharmacological activity and, consequently, treatment failure, due to differences in the biodistribution profile of each drug [20]. Therefore, the entrapment of both drugs in NPs should address this problem. In fact, to increase therapeutic efficiency, the co-loading of TMZ with other molecules in PLGA NPs has also already been studied for glioblastoma therapy. Xu et al., 2016 proposed the co-delivery of TMZ with paclitaxel using PLGA NPs [21]. Until this moment, to the best of our knowledge, the co-encapsulation of TMZ and O6BG in a nanosystem was not reported. Thus, the aim of this work was to prepare a nanocarrier for the co-loading of both drugs.

However, the use of PLGA NPs faces a few limitations because of their poor loading capacity and the typical initial burst release. Additionally, the production of PLGA NPs requires different stages that can present high costs and be difficult to scale-up, such as centrifugation and dialysis. Also, it can be challenging to entrap hydrophilic drugs, since those exhibit a high partition into the aqueous phase during NPs preparation [6]. Therefore, is necessary to optimize the preparation methods. As well, usually the entrapment of two distinct molecules, with different physicochemical properties such as hydrophilicity and molecular weight, can be a challenging task. Though, high encapsulation of both drugs is desirable to increase the nanosystem efficiency and to reduce the amount of administered polymer. Also, it is essential to control the physicochemical properties of the developed NPs, such as size and surface charge, since these parameters control their biological fate, biodistribution, and toxicity, therefore affecting their therapeutic potential [22,23].

Therefore, experimental design could be a suitable approach to optimize the nanoformulation. In fact, in the last years, experimental design has been used for the optimization of drug-loaded NPs [24,25]. This is because a nanoformulation design requires full knowledge of the correlation between

the experimental factors and the obtained NPs properties. To obtain an optimized nanoformulation using a conventional screening method (evaluating the effect of one experimental variable at a time) is expensive and time-consuming. Experimental design is therefore a validated and useful tool for the development and optimization of experimental procedures with a lower number of observations while still providing the desired information on the correlation between the experimental and the response variables. The obtained model can then be used for predicting future observations within the original design range [26]. Different types of experimental designs can be used, and in this work, a fractional factorial design (FFD) is proposed. When several experimental variables are being studied, FFD is a rapid and reliable tool, allowing the exploration of a maximum number of variables, while requiring less experimental observations, and still obtaining the desired information. However, since it uses only a partial combination of factors, some information about possible interactions can be lost [27].

Thus, the effect of five experimental variables on the physicochemical properties of the developed PLGA NPs and encapsulation efficiency (EE) of both drugs was studied in this work. The studied experimental factors were the amounts of TMZ, O6BG, surfactant, organic solvent, and polymer.

#### **2. Materials and Methods**

#### *2.1. Materials*

TMZ (MW 194.15, purity ≥99%) was purchased from Selleck Chemicals (Munich, Germany). O6BG (MW 241.25, purity <sup>≥</sup>98%) was obtained from Abcam (Cambridge, UK). PLGA Resomer® RG503H (50:50; MW 24,000–38,000), poly(vinyl alcohol) (PVA) Mowiol® 4-88 (MW 31,000), phosphate buffer saline (PBS) and dichloromethane were acquired from Sigma–Aldrich (St. Louis, MO, USA). Uranyl acetate was provided by electron microscopy sciences (Hatfield, UK).

#### *2.2. Preparation of TMZ*+*O6BG-Loaded PLGA NPs*

For the synthesis of PLGA NPs loading, both TMZ and O6BG, a variation of the single emulsion–solvent evaporation method, were used [28]. Known amounts of PLGA, TMZ, and O6BG were dissolved in a dichloromethane solution. A 2% (*w*/*v*) PVA solution was added drop-by-drop to the prepared organic solution. Then, the solution was vortexed (Genius 3, ika®vortex, Germany) and placed in an ultrasonic bath at an ultrasonic frequency of 45 kHz (Ultrasonic cleaner, VWRTM, Kuala Lumpur, Malaysia) to yield an oil-in-water (o-in-w) emulsion.

The emulsion was then poured into a 0.2% (*w*/*v*) PVA solution and maintained in unremitting agitation in a magnetic stirrer (800 rpm, Colorsquid, ika®, Staufen, Germany) until complete organic solvent evaporation (6 h). The suspension was then filtered using a membrane with a pore size of 200 nm, (polyethersulfone membrane syringe filter, VWR, Radnor, PA, USA) and stored at 4 ◦C overnight to increase the NPs' stability, avoiding their aggregation. After, the samples were centrifuged for 30 min at 14100<sup>×</sup> *<sup>g</sup>* (MiniSpin®plus, Eppendorf, Hamburg, Germany), to separate NPs from the non-encapsulated drug. The supernatant containing the non-encapsulated drug was saved for analysis.

#### *2.3. Experimental Design and Data Analysis*

High and comparable EE values for each drug are a prerequisite for the co-loading in the same NPs. In preliminary studies, it was verified that the co-encapsulation of TMZ and O6BG reduced the encapsulation of each drug alone. Thus, it was necessary to optimize the entrapment of both drugs. Also, it was important to control the physicochemical properties of the developed NPs, since these parameters influence the therapeutic efficacy and toxicity of the NPs. To obtain an optimized formulation, it is useful to study how the several experimental parameters influence the entire production process.

Therefore, a 25-2 FFD was implemented using the Minitab Statistical Software (Minitab Inc., State college, PA, USA) to determine the effect of different experimental factors on the PLGA NPs features. The studied independent variables were the amount of both used drugs, the quantity of surfactant and

organic solvent, and the amount of polymer. A variation of a full factorial design, in which only a subset of the total runs was performed, took place. The chosen factors of interest were varied on two levels (determined in preliminary studies) according to the experimental plan presented in Table 1. Two replicates were conducted for each combination and for center levels, and the order of the experiments was randomly sorted to avoid any bias.


Note: mTMZ—Mass of TMZ; mO6BG—Mass of O6BG; %PVA—percent weight per volume of PVA; VDCM—Volume of dichloromethane; mPLGA—mass of PLGA.

The observed response dependent variables were the NPs size, PdI, zeta potential values, and EE of both TMZ and O6BG. For the experimental design, 18 formulations were prepared, and an outline of the experimental plan and its results is shown in Table 2.


**Table 2.** Outline of the experimental design and results. The experimental levels (low, centre, and high) are represented by the coded values of −1, 0, and +1, respectively.

The applied experimental design accounts for main terms and two-factor interactions terms. The latter refers to two different variables that interact with each other, creating a combined effect on the response that independently would not occur. Therefore, the main effects and the two-factor interactions are accounted for in the used regression model. Thus, regression equations were obtained for each studied dependent variable to quantify the relationship between these and all the experimental independent variables. The experimental data was then fitted to the following polynomial regression Equation (1) [29]:

$$Y = \beta\_0 + \sum\_{i=1}^{5} \beta\_i X\_i + \sum \beta\_{ij} X\_i X\_j \tag{1}$$

in which *Y* is the predicted response; β*<sup>0</sup>* is the intercept term and the remaining term; *Xi,j* is the studied levels of the independent variables; and βj,i is the fitted coefficients for *Xi,j*.

The statistical regression models for the different dependent variables were fitted independently. Additionally, the polynomial equations were statistically validated using ANOVA (Analysis of Variance) by statistical significance of coefficients, *R2* values, and normal distribution of the residues. Minitab Statistical Software was also used for the statistical analysis of the data.

#### *2.4. TMZ*+*O6BG-Loaded PLGA NPs Physicochemical Characterization*

#### 2.4.1. Dynamic Light Scattering for Size Determination

The mean diameter and size distribution of the prepared NPs were evaluated by dynamic light scattering (DLS). The measurements were performed in a ZetaSizer Nano ZS (Malvern Instruments, Worcestershire, UK). The attained data is given in intensity distribution. The intensity-weighted mean diameter (Z-average) is given. For the optimized formulation, at least three independent measurements were performed, and obtained results are expressed as the mean and standard deviation (SD). Statistical analysis was performed using the *t*-student test, and *p*-values ≤ 0.05 were considered significant.

#### 2.4.2. Laser Doppler Velocimetry Method for Zeta Potential Determination

The zeta potential values of the prepared NPs were determined by laser doppler velocimetry method. The measurements were also performed in a ZetaSizer Nano ZS (Malvern Instruments, Worcestershire, UK). The analysis was performed using the dielectric constant of water. For the optimized formulation, at least three independent measurements were performed, and obtained results are expressed as the mean and SD. Statistical analysis was performed using the *t*-student test.

#### 2.4.3. Transmission Electron Microscopy for Morphological Analysis

The morphological analysis of the NPs was obtained by transmission electron microscopy (TEM). The NPs were prepared on copper grids (Formvar/Carbon-400 mesh Copper, Agar Scientific, Essex, UK) and negatively stained. For that, 10 μL of samples were stained with 2% (*v*/*v*) uranyl acetate for 45 s, and air-dried. This is a heavy metal salt able to scatter electrons, enhancing the contrast to better visualize the samples [30]. Then, the NPs were visualized using a JEM 1400 electron microscope (Jeol, Tokyo, Japan) at an accelerating voltage of 80 kV.

#### *2.5. TMZ*+*O6BG-Loaded PLGA NPs Stability Studies*

The stability of the prepared PLGA NPs was analysed through size and zeta potential variations. PLGA NPs' dispersions in ultrapure water were stored at 4 ◦C and DLS measurements were performed at different timepoints to evaluate variations in PLGA NPs size and zeta potential values. These measurements were performed weekly, for 6 weeks. For optimized nanoformulation, three independent samples were used.

#### *2.6. Drug Encapsulation E*ffi*ciency of TMZ*+*O6BG-PLGA NPs*

The TMZ and O6BG EE values of the prepared PLGA NPs were assessed by UV–Vis spectrophotometry, using the following Equation (2):

$$EE = \frac{\text{total amount of drug} - \text{amount of free drug}}{\text{total amount of drug}} \times 100\tag{2}$$

Non-encapsulated TMZ and O6BG molecules were separated from the NPs colloidal suspension by centrifugation (30 min, 14100<sup>×</sup> *<sup>g</sup>*, MiniSpin®plus, Eppendorf, Germany) and quantified (UV-1700 PharmaSpec UV-Vis spectrophotometer, Shimadzu, Kyoto, Japan) at 240 nm and 329 nm for O6BG and TMZ, respectively. The results were correlated to control samples corresponding to total amount of drug. For the optimized formulation, three independent experiments were conducted, and obtained results are expressed as the mean and SD. Statistical analysis was performed using the *t*-student test.

#### *2.7. In Vitro Release of TMZ and O6BG from PLGA NPs*

To assess the drug release profile of the developed NPs, in vitro dialysis studies were performed for 20 days at 37 ◦C. For that, a cellulose dialysis membrane (Float-A-Lyzer G2, CE, 10KDa, SpectrumLabs, Los Angles, CA, USA) was rinsed in ultrapure water for 24 h before the beginning of the experiments and equilibrated with release medium 1 h before the dialysis.

A known amount of TMZ+O6BG-PLGA NPs diluted in 2 mL of release medium was placed into the inner space of the dialysis membrane. The outside space was filled with a known volume of release medium to ensure sink conditions. PBS (pH 7.4, 0.01 M) was used as the release buffer to mimic the physiological salt concentrations and pH of blood plasma. The dialysis membrane was kept in continuous stirring at 200 rpm at 37 ◦C, simulating the physiological temperature. The amount of drug release at predetermined timepoints (0, 6, and 24 h; and day 3, 6, 8, 9, 11, 13, 16, and 20) was quantified by UV-Vis spectrophotometry (UV-1700 PharmaSpec UV-Vis spectrophotometer, Shimadzu, Kyoto, Japan). A solution of TMZ and O6BG in PBS was used as control. For the optimized formulation, three independent experiments were conducted, and obtained results are expressed as the mean and SD. Statistical analysis was performed using the *t*-student test.

The TMZ and O6BG release curves, representing the percentage of drug released in function of time, were then plotted by the following Equation (3):

$$1\,\%\,\text{drug\ release} = \frac{amount\,\,of\,\,drug\,\,released\,\,at\,\,time\,\,t}{amount\,\,of\,\,encapsulated\,\,d\,rug} \times 100\tag{3}$$

#### **3. Results and Discussion**

#### *3.1. Statistical Analysis of Experimental Data*

The applied experimental design allowed the identification of the experimental factors influencing the physicochemical properties of the NPs, astheir size and the encapsulation efficiency of both drugs on the polymeric matrix. Since zeta potential (≤−20 mV) and PdI values (<0.2) were always inside the desired range, they were not considered for this model.

Previous nanoformulation studies were conducted to identify the five major experimental variables that affect the PLGA NPs properties, such as size and EE of both drugs. The amount of TMZ, O6BG, surfactant, organic solvent, and polymer were chosen and varied in two levels. The levels choice was based on the knowledge acquired in the preliminary experiments. All other parameters, such as type of surfactant and organic solvent, aqueous to organic phase ratio, time of sonication, ultrasonic frequency of sonication, process temperature, and emulsification and evaporation processes, were maintained constantly.

The mathematical model (Equation (1)) was fitted to the data and statistical analysis was performed using ANOVA (Table 3). As Table 3 shows, the model is statistically significant (*p* < 0.05) with insignificant lack of fit (*p* > 0.05) for all the chosen responses. Hence, the linear model was acceptable for response prediction within the range of experimental variables.



as

blue.

**Table**

Figure 1 also shows a satisfactory agreement between experimental observations and predicted response, proving that the applied regression model is suitable for the determination of the optimal experimental settings for the preparation of NPs.

**Figure 1.** Graphical plots of the experimental values versus the model predicted values for (**A**) mean size, (**B**) EE for TMZ, and (**C**) EE for O6BG. Red dots indicate high values and blue dots low values, respectively.

The statistical analysis of the obtained results also allowed the determination of the regression coefficient (RC) values that describes the direction of the relationship between an experimental variable and the variable response. A negative sign before a factor indicates that the response decreases, whereas a positive sign shows that the response increases with the experimental factor. With the RC values, it was possible to obtain the polynomial regression equation for each studied response, quantifying the relationship between each of the studied experimental variables and the response. The non-significant factors (*p* > 0.05) were excluded from the mathematical model, allowing the reduction in complexity of the attained equations.

Response surface analysis and contour graphs were plotted based on the determined model polynomial function in a three and a two-dimensional model, respectively, illustrating the effect of the chosen significant independent factors on each observed response. The effect of all the independent variables on each dependent variable was studied and the effects of the most significant variables are discussed in detail below.

#### 3.1.1. Effect on NPs' Size

The size of the NPs is a critical property, influencing its half-life, biodistribution, and cellular internalization. The size of the NPs ranged from 162 nm (sample 7) to 205 nm (sample 12). Almost all the studied variables significantly affected the size of the NPs (*p* < 0.05), as shown in Table 3.

The polynomial Equation (4) describes the relationship between the significant studied experimental variables (*p* < 0.05) and the size of the NPs:

$$\begin{array}{ll} \text{Size} = & 134.9 - 9.67 \,\text{amount} \,\text{TMZ} \,(\text{mg}) + 83.9 \,\text{amount} \,\text{O6BG} \,(\text{mg}) \\ & + \,\text{15.05} \,\text{amount} \,\text{PVA} \,(\%) + \,\text{1.644 } \,\text{amount} \,\text{PLGA} \,(\text{mg}) \\ & - \,\text{13.41 } \,\text{amount} \,\text{O6BG} \,(\text{mg}) \times \,\text{amount} \,\text{PA} \,(\%) \\ & - \,\text{2.541 } \,\text{amount} \,\text{O6BG} \,(\text{mg}) \times \,\text{amount} \,\text{PLGA} \,(\text{mg}) \end{array} \tag{4}$$

The experimental factor that most significantly affected NPs size was the amount of surfactant, which exhibited a positive effect on the size of the NPs. Thus, an increase in PVA concentration led to an increase in NP size. Although using a high quantity of surfactant can induce the creation and stabilization of smaller NPs due to a reduction of the interfacial tension between the polymer and the external aqueous phase [31], the opposite effect was verified. This may be explained by the increased

viscosity of the aqueous phase when increasing the PVA concentration. A higher viscosity decreases the shear stress, originating emulsion droplets with larger sizes. Also, higher amounts of surfactant can promote the coalescence of the NPs, yielding NPs with larger diameters [32]. Also, some studies report that residual PVA remains at the surface, contributing to the size increase [33].

Additionally, the amount of PLGA also exhibited a significantly positive effect on the size of the NPs (*p* < 0.05). Increasing the polymer amount increases the viscosity of the organic phase, decreasing the shear stress, as mentioned above. Also, the augmented viscosity hampers the diffusion of the organic solvent into the aqueous phase, leading to the formation of larger emulsion droplets, originating larger NPs after solvent evaporation [34].

Higher initial loading of O6BG positively affected the NPs size (*p* < 0.05), since it will result in higher drug loading, as will be discussed later. The contrary was observed for TMZ amount (*p* < 0.05), since higher amounts of TMZ will decrease its entrapment, as will also be discussed later.

Response surface analysis and contour graphs were plotted (Figure 2) based on this model polynomial function. Contour and surface response plots allowed visual identification of the optimal levels of each factor, to choose the most suitable values for the development of an optimal formulation. Both response surface and contour plots (Figure 2A,B) showed that the lower the amount of PVA and PLGA, the lower the NPs size, as already predicted by the calculated positive RC values (Equation (4)). Thus, optimal nanoformulation with small dimensions would fall into the low and central levels of both factors.

**Figure 2.** (**A**) Response surface plot and (**B**) contour plot showing the effect of two factors (amount of PVA and PLGA) on the resulting NPs size. (**C**) Response surface plot and (**D**) contour plot showing the effect of two factors (amount of TMZ and O6BG) on the resulting NPs size.

In addition to all the five factors significantly affecting the NPs size, some significant two-factor interactions were observed (Table 3). Interaction plots on Figure 3 show high interaction between two factors. Each point in the interaction plot shows the mean size values at different combinations of factor levels. As the lines are not parallel, with different slopes, the plot indicates that there is an interaction between the two factors [35]. The same is verified in all the attained interaction plots.

**Figure 3.** Interaction plots for size data means showing significant two-way interaction terms for the independent variables. Solid blue lines display factor at low level, whereas dashed green lines are the high level of the factors. (**A**) interaction term X2X3: amount of O6BG/amount of surfactant, (**B**) X2X5: amount of O6BG/amount of PLGA.

The main effects can be confused with the two-factor interactions, which explains the results obtained in Figure 2. Although Equation (4) predicts that the lower the amount of TMZ and the higher the amount of O6BG, the higher the NPs size, optimal values fall into high levels of both TMZ and O6BG. As shown in Figure 3A,B, at low levels of both PLGA and PVA amounts, O6BG positively affects the size of the NPs. On the other hand, at high levels of PLGA and PVA, increasing the O6BG amount decreases the size of the NPs.

#### 3.1.2. Effect on the Encapsulation Efficiency of TMZ

High encapsulation of the drug is desirable to increase the nanosystem efficiency and to reduce the amount of administered polymer. The EE values for TMZ in the prepared NPs ranged from 27.2 (sample 10) to 56.7% (sample 4).

The polynomial Equation (5) describes the relationship between the significant studied experimental variables (*p* < 0.05) and the EE of TMZ:

$$\begin{array}{ll} \text{EE\% } TMZ &=& 11.84-8.80 \,amount \, TMZ \, (mg)-3.40 \, amount \, PVA \, (\%) \\ &+& 7.95 \, amount \, DCM \, (mL)+1.461 \, amount \, PLGA \, (mg) \\ &-& 13.87 \, amount \, O6BG \, (mg) \times amount \, PVA \, (\%) \\ &-& 2.153 \, amount \, O6BG \, (mg) \times amount \, PLGA \, (mg) \end{array} \tag{5}$$

EE values were significantly influenced by almost all the studied independent variables, as shown by the calculated *p*-values on Table 3. All the experimental variables show a positive effect on the EE of TMZ, except PVA concentration and TMZ amount. PLGA had a positive effect on the EE of TMZ. As already mentioned, a higher amount of PLGA polymer results in larger NPs and, consequently, higher encapsulation of the drug. Also, the increased viscosity caused by higher PLGA concentration mentioned above could complicate the diffusion of TMZ molecules into the aqueous phase, enhancing the drug's entrapment into the NPs' polymeric matrix [36]. On the other hand, increased volumes of organic solvent increased drug solubility, therefore increasing drug entrapment.

Also, the encapsulation of TMZ decreased with PVA concentration due to a higher partition of TMZ molecules into the aqueous phase during emulsification, decreasing the EE values. It is reported that drug molecules can diffuse out from the oil nanodroplets and solubilize in PVA micelles at the aqueous phase [37]. Increasing the TMZ amount may cause saturation of the organic phase, leading to the partition of the drug molecules to the aqueous phase, lowering its entrapment in the polymeric matrix.

Response surface analysis and contour graphs (Figure 4) show that the optimal formulation would be prepared with low amounts of PVA and TMZ, and high amounts of PLGA and organic solvent.

**Figure 4.** (**A**) Response surface plot and (**B**) contour plot showing the effect of two factors (amount of PVA and PLGA, respectively) on the resulting EE TMZ values. (**C**) Response surface plot and (**D**) contour plot showing the effect of two factors (amount of TMZ and DCM, respectively) on the resulting EE TMZ values.

All five studied variables were part of an extensive interaction system. A significant two-factor interaction of O6BG amount with both PLGA and PVA amounts (Table 3) was verified (*p* < 0.05), as shown in Figure 5.

**Figure 5.** Interaction plots for EE of TMZ data means showing significant two-way interaction terms for the dependent variables. Solid blue lines display factors at a low level, whereas green dashed lines are the high level of the factors. (**A**) interaction term X2X3: amount of O6BG/amount of surfactant, (**B**) X2X5: amount of O6BG/amount of PLGA.

The described two-factor interaction between the PVA and O6BG amounts shows a significant PVA concentration effect and interaction term (Figure 5A). There is a difference among the means of the two PVA levels, but not a difference in the means among O6BG amount. This proves that although the amount of O6BG did not show a significant direct effect on the EE values, the interaction term exists. The amount of O6BG affected the nanoformulation at both low and high PVA amounts. At low PVA levels, the O6BG amount showed a positive effect in the TMZ encapsulation. The opposite was observed at high PVA amounts.

The same is verified in the two-factor interaction between the PLGA and O6BG amounts (Figure 5B).

3.1.3. Effect on the Encapsulation Efficiency of O6BG

The EE values of O6BG in the prepared NPs ranged from 70.9 (sample 3) to 99.8% (sample 11). The polynomial Equation (6) describes the relationship between the significant studied experimental variables (*p* < 0.05) and the EE of O6BG:

$$\text{P}\% \text{O6BG} = 104.4 - 4.12 \,amount \, PVA \, (\%) + 0.854 \, amount \, PLGA \, (mg) \tag{6}$$

PVA and PLGA amounts were the only studied independent variables that significantly influenced the EE values, as shown by the calculated *p*-values on Table 3. While PLGA concentration showed a positive effect on the EE of O6BG, PVA concentration exhibited a negative influence. In fact, increasing PVA concentration hampered the encapsulation of the drug, since it enhanced the partition of O6BG molecules into the aqueous phase during emulsification, decreasing the encapsulation efficiencies values [37]. Also, as already mentioned, higher amounts of PLGA polymer resulted in larger NPs and, consequently, higher encapsulation of the drug. Also, the increased viscosity caused by higher PLGA concentration mentioned above could complicate the diffusion of O6BG molecules into the aqueous phase, enhancing the drug's entrapment into the NPs' polymeric matrix [36].

Response surface analysis and contour graphs (Figure 6) show that predicted optimal nanoformulation with high O6BG molecules entrapment would fall into the high levels of PLGA and low levels of surfactant concentration. No significant two-factor interactions were observed.

**Figure 6.** (**A**) Response surface plot and (**B**) contour plot showing the effect of two factors (amount of PLGA and PVA, respectively) on the resulting EE values.

#### *3.2. Nanoformulation Optimization and Physicochemical Characterization*

The optimization process was performed by determining the optimal experimental values, which were obtained by solving the three determined polynomial regression equations and grid searching in the response surface graphs or contour plots applying the following criteria: to minimize the particle size (Y1), and to maximize the EE values for both drugs (Y2 and Y3). The optimum levels of the formulation factors are presented in Table 4.

**Table 4.** Optimal formulation parameters for the PLGA NPs determined by the experimental design.


A formulation checkpoint was prepared according to the predicted model to validate the reliability and the precision of the factorial design, using these optimal experimental conditions. The checkpoint

formulation was prepared in triplicates and the properties of the attained NPs were within the range of the predicted values, as shown in Table 5.

**Table 5.** Validation of the model by comparing the predicted values with the observed experimental values. The experimental data is given as the mean ± SD (*n* = 3).


Higher encapsulation verified for O6BG can be explained by its greater affinity for the organic phase, compared with TMZ. While TMZ exhibits a log P value of 0.36, for O6BG, log P is 1.66 (values obtained from Marvin Sketch Calculator software. ChemaxonTM,Version 16.4.25, Budapest, Hungary).

The developed NPs exhibit mean dimensions suitable for brain delivery. Although PdI and zeta potential values were not considered for the model, they were also determined for the checkpoint formulations. The developed NPs exhibited zeta potential values of −22 ± 1 mV and PdI values of 0.19 ± 0.01. The high negative zeta potential values are associated with low toxicity [38] and suggest that the NPs are stable [39]. In fact, all the nanoformulations prepared in the 18 runs of the FFD (data not shown) and the checkpoint formulations proved to be stable in storage conditions for at least six weeks, as shown in Figure 7.

**Figure 7.** Stability of the prepared checkpoint nanoformulations. Black line: Graphical representation of variations in zeta potential values over 6 weeks (data plotted on the left Y-axis); Red line: Graphical representation of variations in mean sizes over 6 weeks (data plotted on the first right Y-axis); Purple line: Graphical representation of variations in PdI values over 6 weeks (data plotted on the second right Y-axis). Results are given as mean ± SD (*n* = *3*).

The stability of the prepared colloidal suspension NPs is granted by a PVA layer on the NP's surface. During the NPs preparation, the PVA molecules adsorb on the surface of the resultant nanodroplets acting as a mechanic barrier that prevent coalescence and avoid NPs' aggregation [40]. TEM image (Figure 8) shows this stabilizer layer around the NPs' surface. Also, the attained PdI values suggest that the colloidal suspension is monodisperse. Therefore, it can be concluded that the method followed for the preparation of PLGA NPs produced well-stabilized monodisperse O6BG+TMZ-loaded PLGA NPs.

**Figure 8.** TEM photograph showing the morphology of the developed PLGA NPs. Scale bar is 200 nm.

#### *3.3. Drug Release from the PLGA NPs*

For the evaluation of the in vitro release profile of the developed NPs, the dialysis method was used. To simulate the physiological conditions (temperature, salt concentration, and pH), the release studies were conducted at 37 ◦C in PBS (10 mM, pH 7.4). The release curves were plotted and are shown in Figure 9.

**Figure 9.** In vitro release of TMZ and O6BG from PLGA NPs in PBS (10 mM, pH 7.4) at 37 ◦C. Free TMZ and O6BG were used as control. Results are given as mean ± SD (*n* = *3*). An enlargement of the first time points—6 and 24 h—is presented to highlight the burst effect.

PLGA NPs showed an initial burst release for both drugs, due to the presence of the drug molecules adsorbed at the NPs' surface. As Figure 9 depicts, at the first 24 h, 48 ± 3% of the total TMZ and 43 ± 3% of the total O6BG were released. Then, both drugs located at the polymeric matrix of the NPs were released in a slower and controlled manner that was prolonged for 20 days.

A faster release for TMZ was verified in comparison to O6BG. The PLGA NPs exhibited a release of TMZ of 81 ± 1% after 20 days, while for O6BG, only about 79 ± 1% of the entrapped drug was released at day 20 (Figure 9). This may occur due to the higher affinity of TMZ to the aqueous buffer (lower log P values). However, the release profile is similar for the two drugs, due to the erosion of the polymer matrix that allows the release of both molecules.

The obtained release profile proved that the developed PLGA NPs are able to maintain the sustained and controlled release of both TMZ and O6BG, and therefore could be used for the co-delivery of entrapped molecules. Further optimization of the release profile to reduce the initial burst release may improve the efficiency of the nanovehicle.

#### **4. Conclusions**

PLGA NPs as drug delivery systems have gained increasing interest in the last years, due to their unique physicochemical properties, such as stability, biocompatibility, and biodegradability. Encapsulating drugs in PLGA NPs allows them to maximize their therapeutic effects and to minimize their side effects. Since the biological fate and toxicity of the NPs depend on their physicochemical properties, the developed nanosystems must meet several criteria to be appropriate for brain delivery.

PLGA NPs for the entrapment of both TMZ and O6BG for further GBM treatment were developed in this work. TMZ is the used drug for GBM chemotherapy, and despite its therapeutics effects being well-known, the drug is not able to effectively cure GBM patients. Intrinsic resistance mechanisms are of one of the major obstacles for GBM treatment, due to the DNA repair by the MGMT protein. To overcome this issue, co-therapy with TMZ and an inhibitor of the MGMT protein, O6BG, was proposed. O6BG is a guanine analogue able to inhibit the activity of the MGMT protein, hampering the DNA repair. Thus, O6BG could decrease the resistance to TMZ's therapy.

Hence, an optimized nanoformulation was developed in this work, and for that, FFD was used. FFD proved to be a suitable approach for the design of co-loaded NPs allowing to optimize the single emulsion solvent evaporation production process with a low replica number in a reliable and precise manner. The regression analysis of the results proved to adequate to identify the main experimental variables affecting the physicochemical properties of the developed NPs. The optimal experimental conditions were chosen, and the optimal nanoformulation was prepared using 1 mg of both drugs, 0.5% (*w*/*v*) of PVA, 1 mL of dichloromethane, and 15 mg of PLGA.

The developed NPs exhibited high encapsulation efficiencies for both drugs and showed a sustained drug release. Thus, it is expected that these nanocarriers will provide an effective brain delivery, allowing the drug to reach the brain at desirable doses, leading to a significant improvement on the GBM treatment.

PLGA NPs provide a novel and potentially efficient approach for the co-administration of TMZ and O6BG, presenting a potential solution for the intrinsic resistance mechanisms to TMZ due to MGMT high expression. Though these NPs could potentially overcome the limitations of the currently available therapies, future in vitro and in vivo tests are necessary to assess the nanosystem efficacy.

**Author Contributions:** All authors contributed to the conceptualization of the experiments. M.J.R. conducted the experiments, data analyses, and the writing of the manuscript. M.d.C.P contributed to the review and editing of the manuscript.

**Funding:** This work was financially supported by: project UID/EQU/00511/2019—Laboratory for Process Engineering, Environment, Biotechnology and Energy—LEPABE, funded by national funds through FCT/MCTES (PIDDAC); Project POCI-01-0145-FEDER-006939, funded by FEDER funds through COMPETE2020—Programa Operacional Competitividade e Internacionalização (POCI) and by national funds (PIDDAC) through FCT/MCTES;—Project "LEPABE-2-ECO-INNOVATION"—NORTE-01-0145-FEDER-000005, funded by Norte Portugal Regional Operational Programme (NORTE 2020), under PORTUGAL 2020 Partnership Agreement, through the European Regional Development Fund (ERDF), and FCT doctoral grant- PD/BD/105984/2014.

**Acknowledgments:** The authors are grateful to Simone Morais for her comments and suggestions during the preparation of this manuscript.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

### **Nanoemulsion Structural Design in Co-Encapsulation of Hybrid Multifunctional Agents: Influence of the Smart PLGA Polymers on the Nanosystem-Enhanced Delivery and Electro-Photodynamic Treatment**

**Urszula Bazyli ´nska 1,\*, Julita Kulbacka <sup>2</sup> and Grzegorz Chodaczek <sup>3</sup>**


Received: 27 June 2019; Accepted: 8 August 2019; Published: 11 August 2019

**Abstract:** In the present study, we examined properties of poly(lactide-*co*-glycolide) (PLGA)-based nanocarriers (NCs) with various functional or "smart" properties, i.e., coated with PLGA, polyethylene glycolated PLGA (PEG-PLGA), or folic acid-functionalized PLGA (FA-PLGA). NCs were obtained by double emulsion (water-in-oil-in-water) evaporation process, which is one of the most suitable approaches in nanoemulsion structural design. Nanoemulsion surface engineering allowed us to co-encapsulate a hydrophobic porphyrin photosensitizing dye—verteporfin (VP) in combination with low-dose cisplatin (CisPt)—a hydrophilic cytostatic drug. The composition was tested as a multifunctional and synergistic hybrid agent for bioimaging and anticancer treatment assisted by electroporation on human ovarian cancer SKOV-3 and control hamster ovarian fibroblastoid CHO-K1 cell lines. The diameter of PLGA NCs with different coatings was on average 200 nm, as shown by dynamic light scattering, transmission electron microscopy, and atomic force microscopy. We analyzed the effect of the nanocarrier charge and the polymeric shield variation on the colloidal stability using microelectrophoretic and turbidimetric methods. The cellular internalization and anticancer activity following the electro-photodynamic treatment (EP-PDT) were assessed with confocal microscopy and flow cytometry. Our data show that functionalized PLGA NCs are biocompatible and enable efficient delivery of the hybrid cargo to cancer cells, followed by enhanced killing of cells when supported by EP-PDT.

**Keywords:** smart nanocarriers; folic acid; verteporfin; cisplatin; SKOV-3 cells; CHO-K1 cells; electroporation; theranostic cargo; double emulsion approach

#### **1. Introduction**

Effective nanocarriers (NCs) for cancer treatment need both passive and active targeting approaches to achieve highly specific drug delivery to cancer cells while avoiding rapid clearance by the mononuclear phagocyte system and cytotoxicity to normal cells [1]. Recently, the field of biomedical applications, including drug encapsulation, has raised much interest, in part due to the advancement of the biomaterials and "smart" polymers, which enable preparation of containers with novel functional properties (e.g., size, charge, interfacial functionalization) by means of nanoemulsion structural design. Nanoemulsion systems (so-called submicron emulsions, parenteral emulsions, or miniemulsions) are referred to in the literature as transparent or translucent (often bluish) isotropic dispersions of water and oil, with nano-domains coexisting in high kinetic equilibrium due to the occurrence of surfactant molecules at the oil/water interface [2,3]. Owing to their small size (usually in the range

20–200 nm), high kinetic stability, and much lower surfactant concentration (typically 3–10%) required for their formation in comparison to microemulsions (generally about 20% and higher, both oil-in-water (o/w) and water-in-oil (w/o), nanoemulsions were found to be very attractive for droplet engineering to obtain efficient nanocarriers with functional properties. The usefulness of these formulations as raw structures for container fabrication based on their surface templating approaches, by means of biocompatible polymers and polyelectrolytes, including very effective solvent evaporation and layer-by-layer techniques, have been proved by many scientific investigations [4–7]. Consequently, the structuration of the nanoemulsion droplets is typically carried out with three general approaches: layering, embedding, and clustering [8]. In the embedding process, the double emulsion method permits encapsulation of both hydrophilic and hydrophobic agents in the double compartment structure [7]. Such a hybrid cargo can be loaded simultaneously or separately in the NC liquid core and protected from the environmental conditions [9]. Additionally, the hybrid molecules may also migrate from the outer to inner phase and form a reservoir that enhances spectroscopic or functional properties of photodynamic therapy (PDT) and leads to synergistic anticancer activity. The biological effectiveness of such multifunctional nanosystems in the combined anticancer therapy may be also intensified by increasing their cellular uptake by the electroporation (EP) approach.

The electroporation method is widely used not only in chemotherapy for drug delivery but also in molecular biology and microbiology for gene transfection. The latest data show combination methods are used after electrochemotherapy, where plasmid electrotransfer is applied for stimulation of the immune response for better recovery [10]. It is also crucial that the electroporation technique significantly shortens the time of exposition to the "therapeutic substance". This approach is definitely safes then chemical methods because it does not introduce any additional components to the reaction environment. The EP approach was shown to efficiently increase both the uptake and the photodynamic activity of some second-generation photosensitive agents from the cyanine (e.g., AlPcS4) and chlorin (e.g., Ce6) families for PDT application [11]. The beneficial impact of EP was also demonstrated in the case of porphyrin-origin dyes (e.g., CoTPPS and MnTMPyPCl5), which were used against drug-resistant breast and colon cancer cells [12]. These and other studies performed with phthalocyanines (AlPc and Pc green) [13] showed the promising effect of EP in overcoming drug resistance. Moreover, electroporation can be used for supported delivery or fusion of nanosystems containing multifunctional cargo, which is not possible with chemical methods. Thus, the EP method was effective not only for drugs and photosensitizers in free form but also for encapsulated ones [14]. Our previous study shows that EP enhances delivery and photodynamic activity of encapsulated cyanine IR-780 combined with flavonoids [15]. Thus, the available data clearly indicate a new trend in the application of multifunctional methods in more specific drug delivery to cancerous cells.

Over the last years, different chemical compounds have been utilized in order to improve the stability of double emulsion. This has included various mixtures of surfactants, two types of emulsifiers, and co-surfactants or copolymers to support the final formulation [16,17]. Nevertheless, as has been recently shown, the formation of the nanoemulsion interface by the solvent evaporation approach using functionalized biopolymers, i.e., polyethylene glycol (PEG)-ylated polyesters, such as poly(d,l-lactic acid) (PDLLA), poly(glycolic acid) (PGA), polycaprolactone (PCL), and poly(lactic-*co*-glycolic acid) (PLGA), enhances the long-term colloidal stability of encapsulated bioactive compounds and extends the functional performance of nanoscopic emulsion formulations, rendering them non-toxic and stable in the bloodstream [7,18]. Furthermore, application of PLGA as one of the most promising Food and Drug Administration (FDA)-approved polymers for human use, for example functionalized by folic acid (FA) molecules with selective affinity to the folate receptor (FR), could be beneficial in treatment of human carcinomas with overexpression of FR, including ovarian, breast, and lung cancer cells [19,20].

Thus, in the present study, we prepared NCs coated with PLGA, PLGA-PEG, or PLGA-FA by means of the double emulsion (water-in-oil-in-water (w/o/w)) evaporation process, which enabled co-encapsulation of a hydrophobic, porphyrin-origin photosensitizing agent—verteporfin (VP) in combination with low-dose cisplatin (CisPt)—a hydrophilic chemotherapeutic drug. Obtained NCs

were characterized using a variety of analytical methods and functional assays. Finally, their utility in bioimaging as well as in anticancer applications upon human ovarian cancer SKOV-3 cells and control hamster ovarian fibroblastoid CHO-K1 cells was also analyzed when combined with EP and PDT.

#### **2. Materials and Methods**

#### *2.1. Chemicals and Reagents*

Poly(lactide-*co*-glycolide), ester endcap L:G 50:50 (PLGA, Mw ~ 25,000–35,000), poly(ethylene glycol) methyl ether-block-poly(lactide-*co*-glycolide) (PEG-PLGA, Mw ~ 5000:20,000), poly(lactide-*co*-glycolide)-folate, L:G 50:50 (FA-PLGA, Mw ~ 25,000–35,000) employed as biocompatible, "stealth" and functionalized "smart" polymers, respectively, for NC stabilization were obtained from PolySciTech® (West Lafayette, IN, USA). Cremophor A25 and didodecyldimethylammonium bromide, di-C12DMAB, applied as a hydrophilic non-ionic and a hydrophobic double chain cationic surfactant, were obtained from BASF Care Creations (Monheim am Rhein, Germany) and Sigma Aldrich (Poznan, Poland), respectively. Verteporfin (VP) and cisplatin (Cis-Pt), utilized as hybrid cargo, were from Sigma-Aldrich (Poznan, Poland). Supplementary chemical compounds were of commercial grade and were used as received. Doubly distilled water was purified using a Milli-Q purification system (Millipore, Bedford, MA, USA).

#### *2.2. Co-Encapsulation of Hybrid Agents in Polymeric Nanocarriers by Double Emulsion (w*/*o*/*w) Solvent Evaporation Method*

Polymeric NCs stabilized by PLGA, PEG-PLGA, FA-PLGA, and non-ionic and cationic surfactants for co-encapsulation of a therapeutic (CisPt), as well as a diagnostic and therapeutic agent—VP (both in the initial concentration of 130 μM)—were obtained using double emulsion (w/o/w) evaporation method [7,21]. Generally, we emulsified an aqueous internal phase (with CisPt) in dichloromethane (with VP, PLGA at a concentration of 5 mg/mL and di-C12DMAB) at 1:4 ratio using a homogenizer set to 25,000 rpm for 5 min. Next, the primary water-in-oil (w/o) nanoscopic emulsion was poured into 1% Cremophor A25 aqueous solution (a hydrophilic surfactant), stirred in a homogenizer for 10 min (25,000 rpm), and immersed in an ice-water bath to obtain the w/o/w emulsion. Then, we evaporated the organic solvent under reduced pressure in a rotary evaporator (Hei-VAP Value Digital, Heidolph Instruments, Schwabach, Germany) with a rotation speed of 150 rpm for 30 min at 25 ◦C and polymeric NCs with a PLGA, PEG-PLGA, or FA-PLGA shell. The hybrid cargo were collected the following day.

#### *2.3. Nanocarrier Size, Polydispersity, and Particle Charge*

The main physicochemical parameters of NCs, such as hydrodynamic diameter (DH), polydispersity index (PDI), and particle charge (ζ-potential), were analyzed by means of dynamic light scattering (DLS) and microelectrophoretic methods using Zetasizer Nano Series (Malvern Instruments, Worcestershire, UK) equipped with a He–Ne laser (632.8 nm). DLS measurements were conducted at 25 ◦C and the detection angle was 173◦, as previously described [7,21,22]. Each value was an average of three runs, with at least 10–20 measurements. We applied the DTS (Nano) program for data evaluation.

#### *2.4. Shape and Morphology*

The morphology of the obtained NCs was studied by atomic force microscopy (AFM) and transmission electron microscopy (TEM) according to our previous protocols [7,21]. The AFM observations were conducted using a NanoScope Dimension V instrument with an RT ESP tube scanner (Veeco Instruments, Plainview, NY, USA) Samples were analyzed at 0.5 Hz scanning speed using a low-resonance-frequency pyramidal silicon cantilever resonating at 250–331 kHz at a constant force of 20–80 N/m. The resonance amplitude was adjusted manually to the lowest possible amplitude enabling stable imaging within the contamination layer on the surface. We prepared the samples by adsorption of an NC droplet on mica that was freshly cleaved. After 18 h, the excess substrate was

removed by rinsing the mica plates in double distilled water for 1 min and drying for 2 h at room temperature. The TEM imaging of NCs was performed with an Field Electron and Ion Company (FEI) Tecnai G2 20 X-TWIN electron microscope (FEI, Brno, Czech Republic) by placing a few drops of diluted NCs on a Cu-Ni grid and leaving the specimens to dry for 20 h at room temperature.

#### *2.5. Encapsulation E*ffi*ciency*

The Ultraviolet - Visible (UV-VIS) absorbance of NCs with encapsulated VP and Cis-Pt was measured with a Metertech SP8001 spectrophotometer with a 1-cm length path thermostated quartz cell in order to evaluate the encapsulation efficiency (*EE*). The hybrid cargo concentration was calculated using calibration curves according to our previous protocol [21,22]. We determined *EE* as follows:

$$EE = \frac{\mathcal{W}\_{added} - \mathcal{W}\_{frc}}{\mathcal{W}\_{added}} \times 100 \text{ \%}$$

where *Wadded* is the amount of VP or CisPt added during the encapsulation procedure, and *Wfree* is the amount of free cargo in the supernatant quantified by UV-VIS spectroscopy after separation of NCs by centrifugation process (14,000 rpm for 30 min).

#### *2.6. Colloidal Stability*

The backscattering (BS) of pulsed near-infrared, IR light (l = 880 nm) was utilized to measure the long-term colloidal stability of NCs (Turbi-ScanLabExpert, Formulaction SA, Toulouse, France) [7]. In general, two synchronous optical sensors (transmission and backscattering detectors) recorded light transmitted through the sample (0◦ from the incident radiation) and light back-scattered by the sample (1358 from the incident radiation). The scanning of the sample was performed in a cylindrical glass cell at 25 ◦C by moving along the entire height of the cell. The BS profiles as a function of the sample height were then collected and analyzed using the instrument's software (Turbisoft version 2.0.0.33, Formulaction SA, Toulouse, France). We measured BS for freshly prepared NCs and after 30 days of the sample storage at 25 ◦C.

#### *2.7. Cell Lines*

The biological studies were performed on a human ovarian carcinoma cell line resistant to diphtheria toxin, cisplatin, and adriamycin (SKOV-3), and a hamster ovarian fibroblastoid cell line (CHO-K1) used as a model for transport studies in a pulsed electric field due to very low expression of endogenous ionic channels [23]. The SKOV-3 and CHO-K1 cells were purchased from ATCC® (American Type Culture Collection, distr. LGC Standards, Lomianki, Poland), cultured, and prepared according to the conditions described previously by our group [15].

#### *2.8. Uptake of Encapsulated Hybrid Cargo—Flow Cytometry Analysis*

The ability to internalize free and encapsulated VP/CisPt by CHO-K1 and SKOV-3 cells was analyzed by flow cytometry using fluorescence-activated cell sorter (FACS, Cube-6, SYSMEX EUROPE GmbH, Warsaw, Poland). Cells were harvested on 12-well plates and after obtaining 80% of confluence, appropriate nanosystems were added as follows. Free VP or NCs containing VP/CisPt were added with a final VP concentration equal to 2.0 <sup>×</sup> <sup>10</sup>−<sup>6</sup> M. Then, the cells were incubated for 24 h at 37 ◦C in a humidified atmosphere containing 5% CO2. In the next step, cells were detached with Trypsin-EDTA (Sigma-Aldrich Merck-Group, Poznan, Poland), washed in PBS, and resuspended in 0.5 mL of PBS. Flow cytometry analysis was performed using a Cube 6 flow cytometer (Sysmex, Warsaw, Poland). The fluorescence of VP was measured with a FL-4-H detector. Data were collected and analyzed by CyView software (Sysmex, Warsaw, Poland).

#### *2.9. Electroporation Protocol*

Electropermeabilization of cell membranes alone and with free or co-encapsulated VP and Cis-Pt was performed using Gene Pulser Xcell™ Electroporation System (BioRad Laboratories, Warsaw, Poland). When cells reached 80% of confluency they were trypsinized and centrifuged (5 min, 1000 rpm). Then, cells were counted and resuspended in 200 μL of electroporation buffer (EP buffer of low electrical conductivity of 0.14 S/m) at cell concentration of 3 <sup>×</sup> 106/mL [15]. Cells were maintained in suspension and pulsed in a cuvette (VWR) with two aluminum plate electrodes (4 mm gap). The following parameters of electroporation were applied: electrical field intensity *E*(appl) = 500 V/cm, 5 rectangular unipolar pulses of 1 ms duration. The EP conditions were established according to our previous study [14]. The EP experiments were performed using Gene Pulser Xcell™ Electroporation System 165-2660 (BioRad). After the pulse delivery, cells were incubated for 10 min at 37 ◦C, then gently centrifuged, resuspended in the cell culture medium (DMEM for SKOV-3 cells or HAM's F10 for CHO-K1 cells, Sigma-Aldrich Merck-Group, Poznan, Poland), and further analyzed by confocal microscopy and subjected to photocytotoxicity studies.

#### *2.10. Intracellular Internalization Studies by Confocal Microscopy*

The internalization of co-encapsulated VP by cancer and normal cells was studied with confocal microscopy. Briefly, SKOV-3 and CHO-K1 cells were seeded onto glass cover slips in Petri dishes (Sarstedt - distr. Equimed, Wroclaw, Poland) at a density of 1 <sup>×</sup> 104 cells per cover slip in a CO2 incubator for 24 h. Next, the cells were treated with NCs at a concentration corresponding to 2.0 <sup>×</sup> <sup>10</sup>−<sup>6</sup> <sup>M</sup> of VP and incubated at 37 ◦C for 24 h. For the EP-supported uptake, cells were first processed as described in the Section 2.9 and then seeded onto cover slips, thus the time of exposition to nanosystems was only 10 min. After 24 h, all samples were fixed in 4% formaldehyde (Polysciences Inc., Hirschberg an der Bergstrasse, Germany), washed, and placed onto basic glass slides (SuperFrost, Menzel, Braunschweig, Germany) upon mounting in an anti-fade medium (Roth – distr. Linegal Chemicals Sp. z o.o, Warsaw, Poland) with with 4 ,6-diamidino-2-phenylindole (DAPI) for nuclei staining. Microscopy was performed on a spinning disk confocal microscope (Cell Observer SD, Zeiss, Oberchochen, Germany). DAPI was visualized with a 405 nm laser and 450/50 emission filter, while VP fluorescence was excited with a 488 laser and collected with a 629/62 nm emission filter.

#### *2.11. Photodynamic Activity Protocol*

Photocytotoxicity of NCs in SKOV-3 and CHO-K1 cells was measured after standard photodynamic procedure and PDT combined with the EP protocol described above by cellular mitochondrial activity determined by the (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) colorimetric assay according to the manufacturer's procedure and our previous experiments [12–15]. The cells subject to PDT were irradiated after 24 h for 10 min with light in the range of 630–680 nm; the final energy delivered to the cell monolayer was 10 J/cm2. The MTT assay was performed after 24 h post irradiation for the NCs loaded with the theranostic cargo at a concentration equivalent to 1.0–5.0 × 10−<sup>6</sup> M of encapsulated VP. The measurements were performed on the GloMax® Discover multimode microplate reader (Promega, Madison, WI, USA). The cell viability in each group was expressed as a percentage of the value obtained for control (untreated) cells (average of three experiments).

#### *2.12. Statistical Analysis*

The results are presented as means ± standard deviation (SD) values for minimum *n* = 3 repeats. The results were analyzed by two-way ANOVA for multiple comparisons and α = 0.05 GraphPad Prism 7.05. The values where *p* ≤ 0.05 (marked with \*) were considered as statistically significant.

#### **3. Results and Discussion**

#### *3.1. Characteristic of "Smart" PLGA Nanocarriers Obtained by Nanoemulsion Structural Design*

The combination of biomaterials, "smart" polymers, and drug delivery systems for both therapeutic and diagnostic (theranostic) approaches enables the development of intelligent devices and brings enormous possibilities for biomedical applications [24]. Biodegradable polyesters, such as poly(glycolic acid) (PGA), poly(lactic acid) (PLA), and their copolymers are the favorite synthetic polymers for biomedical and pharmaceutical applications, since they were proved to be useful in the stabilization of different drug delivery systems and have excellent biocompatibility and bioresorbability [25].

With this in mind, we have tested highly biocompatible polymers made of PLGA (the FDA-approved copolymer of PGA and PLA, also functionalized with different moieties), namely PEG and FA. The "smart" co-polymers were used to design NCs co-loaded with VP, playing a dual role of a diagnostic and a therapeutic agent, and CisPt, a supporting chemotherapeutic drug. Engineered NCs with encapsulated hybrid cargo had an improved colloidal stability and therapeutic activity, as well as extended functional performance with unique attributes e.g., non-toxicity, stability in blood circulation, and cancer-targeting ability. Thus, according to the first phase of our general strategy presented in Scheme 1a, PLGA, PEG-PLGA, and FA-PLGA polymers were used for stabilization and structuration of nanoemulsion droplets involving the three-step w/o/w double emulsion evaporation approach, leading to co-encapsulation of the hybrid cargo, i.e., VP and CisPt in the NC's double compartment.

The second step (Scheme 1b) involved EP-supported PDT upon improved internalization of NCs by human ovarian cancer (SKOV-3) and normal ovary fibroblastoid (CHO-K1) cells. The SKOV-3 cells were selected as difficult-to-treat cells, which are extremely resistant to the wide spectrum of cytostatic drugs, but in particular to cisplatin. Consequently, in our study an attempt was made to design special nanosystems to overcome drug resistance phenomena in human ovarian cancer and also very probable secondary resistance to CisPt, and for rapid elimination of the drug from circulation [26]. The PLGA nanocarriers proposed here can significantly diminish this process, causing good bioavailability and the "willingness" of the cell to accept the natural carrier. Furthermore, as has been proved recently, different drug delivery systems with a negatively charged surface are generally less toxic compared to the positively charged ones, but their cellular uptake may be hindered due to the same negative charge present on the surface of target cells. Thus, the electropermeabilization may enhance the transport of theranostic cargo to target (malignant) cells in spite of their negative surface charge [15,27].

Accordingly, three types of PLGA-NCs loaded with theranostic cargo and three control NCs (systems V1-V6, Table 1) were successfully synthesized. The main physicochemical characteristics of the designed NCs are summarized in Table 1. We measured size (hydrodynamic diameter, D*H*), polydispersity index (PdI), zeta potential (ζ), and encapsulation efficiency of VP (EEVP) and CisPt (EECisPt). NCs with different PLGA shells displayed an average size between 187 and 200 nm, PDI of approximately 0.1–0.2, and ζ from −4 mV to −17 mV, proving efficient assembly of NCs with PEG-ylated and FA-functionalized shells [7,28]. Generally, loaded cargo did not significantly change the NC charge (ζ), as observed for control NCs with a FA-PLGA shell, loaded with VP, CisPt, or empty ones. Furthermore, the loaded NCs showed only a slightly larger size and less unimodal size distribution, which was probably caused by the incorporated cargo molecules. The EE was about 95% for VP and 90% for CisPt. The differences in the encapsulation of the hybrid cargo by PLGA, PEG-PLGA, and FA-PLGA shells are presented in Figure 1 as UV-VIS spectra of the co-encapsulated VP and CisPt compared to the control samples with only the cytostatic drug and with only the photosensitizer, as well as empty NCs. In all nanosystems a characteristic peak at 280 nm for CisPt as well as peaks at 340 nm, 415 nm, and 680 nm for VP can be observed, providing evidence of effective encapsulation of both ingredients.

**Scheme 1.** The general strategy of conducted studies: (**a**) double emulsion (w/o/w) design for co-encapsulation of Verteporfin (VP) and Cisplatin (CisPt) in various Poly(Lactide-*co*-Glycolide) (PLGA) nanocarriers (NCs) and (**b**) electroporation (EP)-supported photodynamic therapy (PDT) using obtained NCs against human ovarian cancer (SKOV-3) and hamster ovarian control (CHO-K1) cells.



**Figure 1.** Ultraviolet - Visible (UV-VIS) spectra and images of the VP and CisPt co-loaded nanocarriers with various PLGA shells (samples V1–V3) compared to the control samples loaded only by the cytostatic drug (V5), only by the photosensitizer (V4), and the empty NCs (V6). See descriptions for the systems (V1–V6) in Table 1.

Meanwhile, the correct imaging of NCs is a key parameter in the design of any drug delivery system dedicated to pharmaceutical applications. The obtained PLGA-stabilized NCs were characterized by transmission electron microscopy (TEM) and atomic force microscopy (AFM)—quick, efficient, and relatively non-invasive techniques that can provide evidence on shape and morphology and size distribution of these polymeric nanosystems. The TEM and AFM images of the loaded PLGA-NCs are shown in Figure 2. The TEM imaging demonstrated spherical particles with roughly uniform sizes related to AFM. Furthermore, we observed some differences in morphology as visualized by TEM for nanocarriers prepared with different PLGA shells. In the case of NCs covered by PLGA, the spherical nanoobjects with relatively smoother surface morphology were discovered by both TEM and AFM imaging, while the NCs stabilized by PEG-PLGA and FA-PLGA (Figure 2) had a typical core shell morphology, where the darkest part relates to the denser polymeric/PEG-ylated corona, demonstrating that these shells were successfully formed. The AFM tapping mode scanning presented as 2D and 3D images identified a semi-spherical shape of the NCs, being slightly less regular in the case of NCs stabilized only by PLGA (Figure 2). However, we did not see increased aggregation as it was found by TEM. The NC's size range was smaller than the distribution obtained by the DLS measurement presented in Table 1, as both TEM and AFM were carried out in dry conditions, and the obtained NCs have a tendency to shrink, resulting in losing their primary shape and size [29].

**Figure 2.** Imaging of VP and CisPt co-loaded PLGA (V1), PEG-PLGA (V2), and FA-PLGA (V3) NCs by means of transmission electron microscopy (TEM) and atomic force microscopy (AFM). See descriptions for the systems (V1–V3) in Table 1.

#### *3.2. Evaluation of Colloidal Stability*

The colloidal stability of NCs is one of the most critical factors for any potential biological application, since these nanostructures, when not stabilized electrostatically, are usually metastable due to short-range van der Waals attraction [30]. Consequently, to avoid NC aggregation due to their low colloidal stability, steric or electrostatic repulsion may be applied for stabilization. The literature data indicate the encapsulation of inorganic and organic molecules by surfactants and polymers as the best strategy for the enhancement of any nanostructure colloidal stability and functionality, leading to hybrid core/(polymer-)shell NCs [31]. However, the NC aggregation process is still hard to control, especially in biological environments, which is of essential significance for in vivo applications [6,30].

The detailed estimation of the colloidal stability of different PLGA co-loaded and empty NCs was conducted using turbidimetric method by means of time-dependent BS profiles. The BS levels expressed in % are indicated on the ordinate axis. The investigated formulation level in the measurement vial was expressed in mm and marked on the abscissa axis. By examining the BS profiles (Figure 3), we were able to determine the dynamics of any decomposition processes occurring within the sample. This was achieved through analysis of the distances between the curves in BS profiles of NCs at 0 days (freshly prepared) and after 30 days of their storage at room temperature (Figure 3).

Typically, rapid destabilization phenomena can be recognized by a large distance between the curves, while an overlap of the individual curves indicates the high stability of the analyzed sample and a slow rate of the destabilization process [32]. Based on the graphs shown in Figure 3, we conclude that the studied nanosystems have good colloidal stability, since no macroscopic changes in analyzed samples (aggregation, sedimentation, or creaming processes) were observed at the last day of the performed turbidimetric test.

**Figure 3.** Backscattering profiles of the various PLGA NCs co-loaded with VP and CisPt (**a**–**c**) (V1–V3) and an empty (**d**) (V6) FA-PLGA nanosystem as a function of sample height (mm) analyzed over 30 days (*T* = 30) of the NCs' storage. Solid lines represent measurement at 0 day (*T* = 0, freshly prepared NCs), while dashed lines indicate data obtained after 30 days of storage. See descriptions of the systems in Table 1.

#### *3.3. Cellular Internalization—Flow Cytometry and Confocal Microscopy Evaluation*

The evaluation of nanosystem uptake by cancer and normal cells is shown in Figure 4. The uptake efficiency of NCs was estimated by flow cytometry after 24 h of exposition. SKOV-3 cells revealed a higher uptake efficiency of VP-loaded NCs than CHO-K1 cells. The strongest fluorescent signal was observed for V3 (VP + CisPt) and V4 (VP) nanosystems, which were functionalized by FA moieties, proving the effective SKOV-3 targeting ability of the "smart" FA-PLGA-coated nanodevices [33]. The improved uptake in the case of nanosystem V3 can be explained by the content of isplaitin, which can slightly sensitize exposed cells and provoke better internalization of VP.

The confocal microscopy was conducted on cells treated with NCs and electroporated. The exposition time to treatment was much shorter and did not exceed 10 min. The acquired microphotographs are presented in Figure 5. Upon EP, the uptake of encapsulated drugs increased, in particular for FA-PLGA-coated NCs (V3 and V4) in both cell lines. Additionally, a stronger fluorescent signal was also found for the V2 (NCs-PLGA-PEG) nanosystem in electroporated ovarian cancer cells. A very low fluorescent signal was detected when cells were not electroporated. Thus, we conclude that EP is favorable when a short time of incubation is required for the therapeutic protocol. Moreover, FA significantly enhanced toxicity against ovarian cancer cells. Our results are in good correspondence with other studies that also indicate functionalization as a promising factor in anticancer protocols [32,34].

**Figure 4.** Influence of the nanocarrier surface on the encapsulated cargo uptake by SKOV-3 and CHO-K1 cells evaluated by flow cytometry analysis. The uptake of NCs was expressed as mean fluorescence intensity of VP in the gated cell population after 24 h incubation of cells with the loaded PLGA NCs (V1–V5), control cells (control), empty nanocarriers (V6). and free VP (Vp). See descriptions of the systems in Table 1.

**Figure 5.** Intracellular distribution of nanosystems after 24 h incubation with NCs (upper panel) or 10 min exposition to NCs after EP (lower panel) in (**a**) CHO-K1 cells and (**b**) SKOV-3 cells.

#### *3.4. Evaluation of PDT and EP-PDT*

The results of photocytotoxicity studies for SKOV-3 cells are presented in Figure 6a. The cellular viability was assessed after 24 h of incubation with the NCs followed by irradiation. We found that the photodynamic effect increased proportionally to the concentration of the applied nanosystems. Longer time of incubation induced a significant decrease of cellular viability (60–80% decrease), in particular for the mixed cargo, proving the supportive anticancer effect of CisPt [35]. In Figure 6b,c, both types of cells were first electroporated and then exposed to NCs for 10 min.

**Figure 6.** Photodynamic treatment (PDT) in SKOV-3 after 24 h of incubation with NCs (**a**) and photodynamic reaction facilitated by EP after 10 min exposition to nanosystems (*C*VP = 1 μM) in SKOV-3 (**b**) and CHO-K1 (**c**) cells. The missing red bars for V6 and free Cpt represent non irradiated negative controls. See description of the NCs in Table 1. Data represented as means ± SD for minimum *n* = 3, where \* *p* ≤ 0.05 was judged as statistically significant.

After irradiation that followed EP, we observed a significant decrease of cellular viability in ovarian cancer cells in a short time, even in the lowest concentration of the applied NCs (*C*VP = 1 μM).

This is in agreement with another study showing that EP enhances photodynamic reaction efficacy in the case of encapsulated drugs [12]. Thus, EP might be a promising and PDT-supporting tool for time-limited anticancer therapies. It is also worth noting that empty NCs (V6) with or without EP showed no toxic effect in SKOV-3 and CHO-K1 cultures, even at higher NC concentration (cell viability above 90%), proving the protective effect of the PLGA shell.

#### **4. Conclusions**

In this work, we demonstrate that a rationally designed double emulsion process leads to formulation of long-lasting, biocompatible, and "smart" NCs containing hybrid theranostic cargo with different hydrophobicity. We have designed, engineered, and characterized the effective polymeric nanocontainers (size ~200 nm) for efficient co-encapsulation of Cis-Pt and VP enabling drug delivery and synergistic anticancer activity via standard PDT and EP-enhanced PDT against human ovarian (SKOV-3) cancer cells. Loading hybrid cargo inside the oil core of NCs minimized its interaction with water environment, and thus, no drastic changes in physicochemical properties of the cytostatic drug and photosensitizer were observed. This feature is very favorable from the point of view of any bio-related application. Furthermore, EP with only 5 pulses and short time of loading exposition (10 min) enhanced delivery of encapsulated VP with all types of NCs. The highest photodynamic potential was noticed when VP was co-encapsulated with Cis-Pt and the "smart" NC functionalization with FA led to improved internalization by the SKOV-3 cells. Control CHO-K1 cells were significantly less sensitive to PDT, EP, and EP-PDT. In summary, the presented results reveal that the designed polymeric PLGA shells and colloidal cores in our NCs significantly improve PDT and EP-PDT and warrant future in vivo studies in animal models to fully prove their use for theranostic applications.

**Author Contributions:** Conceptualization, U.B. and J.K.; methodology, U.B., J.K. and G.C.; software, U.B. and J.K.; validation, U.B., J.K.; formal analysis, U.B., J.K. and G.C.; investigation, U.B., J.K. and G.C.; resources, U.B.; data curation, U.B., J.K. and G.C.; writing—original draft preparation U.B., J.K.; writing—review and editing, U.B., J.K., G.C.; visualization, U.B., J.K.; supervision, U.B.; project administration, U.B.; funding acquisition, U.B.

**Funding:** This research was funded by the National Science Centre (Poland) within the framework of the SONATA 8 program (No. 2014/15/D/ST4/00808), and by a statutory activity subsidy from the Polish Ministry of Science and Higher Education for the Faculty of Chemistry of Wroclaw University of Science and Technology.

**Acknowledgments:** We would like to thank Professor Katarzyna Bogunia-Kubik from the Institute of Immunology and Experimental Therapy (Wroclaw, Poland) for facilitating the electroporation tests on the Gene Pulser Xcell™ Electroporation System (BioRad, Warsaw, Poland).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### **Vitamin E-Loaded PLA- and PLGA-Based Core-Shell Nanoparticles: Synthesis, Structure Optimization and Controlled Drug Release**

### **Norbert Varga 1, Árpád Turcsányi 1, Viktória Hornok 1,2 and Edit Csapó 1,3,\***


Received: 27 June 2019; Accepted: 16 July 2019; Published: 22 July 2019

**Abstract:** The (±)-α-Tocopherol (TP) with vitamin E activity has been encapsulated into biocompatible poly(lactic acid) (PLA) and poly(lactide-*co*-glycolide) (PLGA) carriers, which results in the formation of well-defined nanosized (d ~200–220 nm) core-shell structured particles (NPs) with 15–19% of drug loading (DL%). The optimal ratios of the polymer carriers, the TP active drug as well as the applied Pluronic F127 (PLUR) non-ionic stabilizing surfactant, have been determined to obtain NPs with a TP core and a polymer shell with high encapsulation efficiency (EE%) (69%). The size and the structure of the prepared core-shell NPs as well as the interaction of the carriers and the PLUR with the TP molecules have been determined by transmission electron microscopy (TEM), dynamic light scattering (DLS), infrared spectroscopy (FT-IR) and turbidity studies, respectively. Moreover, the dissolution of the TP from the polymer NPs has been investigated by spectrophotometric measurements. It was clearly confirmed that increase in the EE% from ca. 70% (PLA/TP) to ca. 88% (PLGA65/TP) results in the controlled release of the hydrophobic TP molecules (7 h, PLA/TP: 34%; PLGA75/TP: 25%; PLGA65/TP: 18%). By replacing the PLA carrier to PLGA, ca. 15% more active substance can be encapsulated in the core (PLA/TP: 65%; PLGA65/TP: 80%).

**Keywords:** vitamin E; tocopherol; PLA; PLGA; core-shell nanoparticles; drug delivery; controlled drug release

#### **1. Introduction**

The encapsulation of the pharmaceutical ingredients in a macro- or a nanocarrier is a key factor in nanomedicine developments [1–3]. Utilization of drug delivery systems can increase and prolong the efficiency of the active drugs. Due to the good biocompatibility and structural properties, several materials such as proteins and mostly biodegradable polymers, e.g., chitosan, alginate, hyaluronic acid, poly(lactide) or poly(lactic acid) (PLA), polycaprolactone (PCL), poly(trimetilene-carbonate) (PTMC), etc., have been widely used as potential carriers of the nanosized drug delivery systems [4–12]. Many types of PLA copolymers, such as poly(lactide-*co*-glycolide) (PLGA), polylactide-poly(ethylene glycol) (PLA-PEG), poly(caprolactone-ethylene glycol-lactide) (PCELA), etc., are also well-known as drug carriers [13–17]. As a result of copolymerization, the hydrophilicity of the PLA can be remarkably tuned, which opens the possibility of encapsulating active ingredients that have hydrophilic or hydrophobic character into the polymer matrix.

The (±)-α-Tocopherol (TP) is one of the highest biological activity fat-soluble vitamins from the vitamin E family and is composed of eight tocopherols and tocotrienols (alpha (α), beta (β), gamma (γ) and delta (δ) isomers for both cases) [18]. Several studies focus on the encapsulation of the TP, because vitamin E can prevent and treat many chronic and age-related diseases, e.g., Alzheimer's disease [19]. Furthermore, the TP is described as functioning as an antioxidant, and the higher vitamin E content is associated with a lower risk of several cancer diseases (kidney, lung, bladder, etc.).

Alqahtani et al. synthesized TP-loaded PLGA50 NPs with an average size of ca. 130 nm using Polyvinyl alcohol (PVA) as a stabilizer, but only 4–4.5% of drug loading (in mg drug/100 mg PLGA) was achieved [20]. However, Zigoneanu and coworkers increased the TP loading to 8–16% using the PLGA50 carrier and PVA and sodium dodecyl sulphate (SDS) surfactant stabilizing agents, but it was confirmed that the 86% of the drug dissolved for NPs (d ~220–280 nm) with 8% TP loading, while 36% of the TP drug released from the NPs when the loading was 16% after1h[19]. However, Astete et al. synthesized Span80-stabilized TP-containing PLGA50-based nanosized particles in the range of 150–200 nm and the effect of salt concentration on the size and morphology of the NPs has been interpreted, but data for the drug loading and release were not presented [21]. Murugeshu and coworkers also synthesized chitosan/PLGA50-based TP-containing NPs with 8%, 16% and 24% initial loading, but the EE% was only 45–50% and the release studies were carried out in gastrointestinal conditions (pH ~1.50) [22]. Simon et al. successfully encapsulated the TP into PLGA50 using PVA, but only 2.5 mg drug/100 mg NPs can be obtained [23].

In our previous work, the TP as well as the water-soluble derivative of TP (D, α-Tocopherol polyethylene glycol 1000 succinate (TPGS)) and the non-steroidal anti-inflammatory ketoprofen (KP) have been successfully encapsulated into PLA and PLGA75 (lactide:glycolide ratio 75:25) and PLGA65 (lactide:glycolide ratio 65:35) nanocarriers [13]. Instead of the commonly used and above mentioned PLGA50, the PLGA65 and PLGA75 derivatives have been firstly used as carriers for the encapsulation of TP, and the hydrophilicity properties of the carriers and the model drugs on the EE% have been studied in detail. The release measurements of nanocomposites including hydrophobic drugs are very difficult to carry out. In order to facilitate the above-mentioned studies, stabilizing agents such as surfactants have been widely used [5,19]. Non-ionic poloxamer Pluronic F127 was applied previously, which stabilized the drug-containing polymeric NPs by increasing the biocompatibility. The optimal ratios of the carrier, the drug and the stabilizing agent have a dominant effect on the structure and the EE%, as well as the controlled drug release process of the nanosized drug delivery systems, which were not studied previously.

In the present work, the results of the TP-containing nanocomposites, using PLA, PLGA65 and PLGA75 carriers, have been completed and the determinative role of the concentration of the polymer, the TP and the PLUR stabilizing agent on the structure of the drug-containing nanocomposites has been investigated. Moreover, the drug release studies of the prepared TP-containing PLA, PLGA65 and PLGA75 core-shell NPs have also been interpreted.

#### **2. Materials and Methods**

#### *2.1. Materials*

Polylactide (PLA, Mw = 72,200 ± 15,000 Da) and two poly(lactide-*co*-glycolide) (PLGA) derivatives with a lactide to glycolide ratio at 75:25 (PLGA75, Mw = 69,900 ± 4000 Da) and at 65:35 (PLGA65, Mw = 93,000 ± 1000 Da) were synthetized according to the previously published procedure [13]. Pluronic F127 (PLUR), (±)-α-tocopherol (TP) and sodium phosphate monobasic monohydrate (NaH2PO4 · H2O, ≥99%) were obtained from Sigma Aldrich (Budapest, Hungary). Sodium phosphate dibasic anhydrous (Na2HPO4, ≥99%) and sodium chloride (NaCl, ≥99%) were purchased from Molar Chemicals (Halásztelek, Hungary). All other reagents and solvents were of analytical grade and used without further purification. The deionized water was obtained by Millipore purification apparatus (18.2 MΩ cm at 25 ◦C).

#### *2.2. Preparation of TP-Loaded PLA*/*PLGA NPs*

The TP-loaded PLA/PLGA NPs were prepared by nanoprecipitation method (Figure 1). The detailed experimental conditions were presented previously [13]. Briefly, the PLA/PLGA and TP with increasing concentrations (see Table 1) were dissolved in 1.5 mL of acetone, which was dropped slowly (10 μL/5 s) into the aqueous solution of the PLUR stabilizer (15 mL) under room temperature using magnetic stirring with 1000 rpm. In the interest of the evaporation of acetone, the prepared samples were further stirred (350 rpm) for two days. The dispersion was centrifugated with 40 mL MQ water at 12,000 rpm (*t* = 15 min, *T* = 25 ◦C). After the supernatant was removed, the NPs were redispersed in MQ water and the washing methods were repeated two times. The obtained NPs samples were freeze-dried by liquid nitrogen and lyophilized (by Christ Alpha 1-2 LDplus apparatus).

**Table 1.** The concentration of the components, the average particle diameter, the polydispersity index (PI), the encapsulation efficiency (EE%) and the drug loading (DL%) of the TP-loaded PLA NPs.


<sup>1</sup> The experimental error of the peak maximum is below 2.5%.

**Figure 1.** Schematic representation of the preparation of TP-loaded PLGA NPs stabilized by PLUR using nanoprecipitation technique.

#### *2.3. Characterization Methods*

The particle size was determined by dynamic light scattering (DLS) with Horiba Sz-100 (HORIBA Jobin Yvon, Longjumeau, France) equipped with a diode pumped frequency doubled (532 nm, 10 mW) laser. The measurements were carried out at 25 ± 0.1 ◦C with 90◦ of detection angle in every case. The transmission electron microscopy (TEM) images were obtained by Jeol JEM-1400plus equipment (JEOL Ltd., Tokyo, Japan) at 120 keV accelerating voltage. The Fourier transform infrared (FT-IR) spectra of the PLA/PLGA-TP NPs were registered by a Jasco FT/IR-4700 with ATR PRO ONE Single-reflection accessory (ABL&JASCO, Budapest, Hungary). The experiments were performed at room temperature from 600 cm−<sup>1</sup> to 3600 cm−1. The resolution of the spectra was 2 cm−1, which was determined by 128 interferograms.

The particles were dissolved in 1,4-dioxane to determine the EE% and DL% of the composites. The absorbance spectra of the prepared solutions were measured by a Shimadzu UV-1800 UV-Vis double beam spectrophotometer. The spectra were registered in the range of 200–500 nm using 1 cm quartz cuvette at room temperature. The characteristic absorbance band of the TP appeared at 294 nm. The concentration of the encapsulated drug was determined from the calibration curve (Figure S1). The data of the EE% in *w*/*w*% were calculated by the total drug mass used in the synthesis of the NPs (Equation (1)), while DL% was calculated by the mass of the NPs (Equation (2)).

$$EE\% = \frac{encapsulated\ mass\ of\ drug}{total\ mass\ of\ drug\ in\ synthesis} \times 100\tag{1}$$

$$\text{DL\%} = \frac{\text{encapusted mass of drug}}{\text{total mass of the nanoparticles}} \times 100 \tag{2}$$

#### *2.4. Determination of the Solubility Properties of TP Drug*

In order to determine the solubility of TP, 100 <sup>μ</sup>L of acetone solutions of TP (5 mg·mL−1) was added dropwise to 10 mL of PLUR solutions (0.1–1.2 mg·mL<sup>−</sup>1), and the turbidity was followed with a Precision Bench Turbidity Meter LP2000 (Hanna Ins. Service Kft., Szeged, Hungary). The experiments were performed in a pure aqueous medium and in a phosphate (PBS) buffer (pH = 7.4, 0.9% NaCl) solution at 25 ◦C and 37 ◦C.

#### *2.5. Critical Micelle Concentration (cmc) Studies*

The critical micelle concentration (cmc) of the applied PLUR was determined by inverse titration method in Krüss K100MK2 type surface tension equipment. The computer-controlled apparatus was supplied with a thermostat and an automatic burette. The Wilhelmy-plate method was applied. A volume of 50 mL of a 1.6 mM surfactant solution was titrated with MQ water or PBS in an aliquot of 10 mL in 40 steps. Each experimental point was the average of at least 5 measurements.

#### *2.6. In Vitro Release Study*

The in vitro studies of the different TP-loaded PLA/PLGA NPs were carried out by a UV-Vis spectrophotometer (500–200 nm). The release experiments were performed at 37 ◦C and a PBS buffer (pH <sup>=</sup> 7.4, NaCl 0.9%) containing 1 mg·mL−<sup>1</sup> of PLUR was used, which facilitated the easier feasibility of the release studies [5]. The TP-loaded samples were placed into the cellulose membrane (Sigma Aldrich) with 5 mL of PLUR/PBS medium inserted into 35 mL of a dissolution phase. During the measurements, 3 mL of the release media were taken at specified intervals to measure the released concentration of TP at 268 nm.

The release curves of the TP can be fitted by different kinetic models, such as the first order, Korsmeyer–Peppas, Peppas–Sahlin and Weibull models [24–27]. The measured points were fitted with a nonlinear regression by the QtiPlot 0.9.8.9 svn 2288 program. During the calculation session, the program finds the best fitting function for the measured points. The results consist of the fitted parameters, their standard deviation and the goodness of fitting (root mean squared error).

Depending of the release chemical condition (such as temperature, buffer solution, ionic strength, etc.), the shape of the polymers and the solubility of the drugs, the dissolution curves can be described with different kinetic models. For our calculation, the following kinetic models were used.

First order equation is a frequently used kinetic model. This formula is well applicable for drugs where the dissolution is continuously changing over time and depends only on the concentration.

$$c\_t = c e^{-kt} \tag{3}$$

where *ct* is the concentration of the solid drug in the matrix of the carried system at *t* time, *c*<sup>0</sup> is the initial concentration of the drug and *k* is the first order release constant.

The Korsmeyer–Peppas kinetic formula is a semi-empirical power law equation where the shape of the polymer matrix (such as film, cylinder or sphere) can be taken into account in the release curve.

$$\frac{c\_t}{c\_0} = k\_m t^n \tag{4}$$

where *ct* is the concentration of the dissolved drug at *t* time, *c*<sup>0</sup> is the initial concentration of drug, *km* is the kinetic constant and n is the diffusion dissolution index (for sphere shaped particles *n* = 0.42 for the diffusion-controlled mechanism, *n* = 1 for the Case II relaxation controlled mechanism and 0.42 ≤ *n* ≤ 1 for both of them).

The Weibull equation is a general empirical formula which can be used for all release profiles.

$$c\_l = 1 - \exp\left(\frac{-\left(t - T\_i\right)^b}{a}\right) \tag{5}$$

where *ct* is the concentration of the released component in *t* time, *Ti* is the lag time between the initial of measurement and the release of drug (in most cases *Ti* = 0), *a* is the time scale of the process and *b* is the shape parameter (shape of the release curve is exponential if *b* = 1, parabola if *b* < 1 or sigmoid if *b* > 1).

The kinetic formula reported by Peppas and Sahlin specifies the diffusion and the relaxation contribution in the drug dissolution process.

$$\frac{c\_t}{c\_0} = k\_1 t^m + k\_2 t^{2m} \tag{6}$$

where *ct* is the concentration of the dissolved drug in the *t* time, *c*<sup>0</sup> is the initial concentration of the drug and *k*1, *k*<sup>2</sup> and *m* are constants: *k*<sup>1</sup> is the Fick diffusion contribution, *k*<sup>2</sup> is the Case II relaxation contribution and *m* is the diffusion exponent (sphere shaped: *m* = 0.43, Fick diffusion mechanism; *m* = 0.85, Case II relaxation transport mechanism; 0.43 ≤ *m* ≤ 0.85, anomalous transport mechanism).

#### **3. Results**

#### *3.1. E*ff*ect of the Component Concentrations on the Core-Shell Structure*

In order to determine the role of the component quantities of the nanocomposites on the size and the structure, as well as on the EE% the concentration, only one building block (polymer carrier, drug or stabilizing agent) has been modified during the synthesis while the other parameters have been kept constant. The average particle diameters of the NPs were measured by DLS, and the results are summarized in Table 1. In case of PLA concentration dependence, regardless of the amount of the TP (c = 2.5 mg·mL−1), the diameter of the NPs permanently increases from 120 nm (cPLA <sup>=</sup> 1.25 mg·mL−<sup>1</sup> in aceton phase) to ca. 200 nm (cPLA = 10 mg·mL−<sup>1</sup> in the acetone phase). The morphology and the structure of the NPs have been investigated by TEM images as well (Figure 2). The images clearly represent that the structure of the TP/PLUR NPs in the absence of a polymer shows less amorphous structures and TP crystal-like objects are observed. Moreover, we established that an increase in the polymer concentration results in the formation of a well-defined core-shell structure at cPLA <sup>=</sup> 10.0 mg·mL<sup>−</sup>1. At lower polymer concentrations, this structure is not formed, and because of the low polymer concentration, the purification (centrifugation) of the NPs is impracticable; thus, the determination of the EE% was not possible. For TP that is concentration-dependent, the

diameters show a slightly increasing tendency to 2.5 mg·mL−1, but for a 5 mg·mL−<sup>1</sup> amount of TP, a higher size is obtained (dDLS = 252 nm) (Table 1). The TEM images also confirm this observation (Figure 2). The well-defined core-shell structure is formed at 2.5 mg·mL−<sup>1</sup> of TP quantity. With a further increase in the TP amount, the core, including the drug, shows a rather crystallized structure instead of the previously confirmed amorphous form. The EE% and the drug loading were determined for all composites. We obtained that the value of the EE% decreased from 91.28% (cTP <sup>=</sup> 0.5 mg·mL<sup>−</sup>1) to 66.15% (cTP = 5 mg·mL−1), while the DL% increased from 4.36% (cTP = 0.5 mg·mL−1) to 24.85% (cTP <sup>=</sup> 5 mg·mL<sup>−</sup>1). Considering the expected size of the NPs for optimal nanosized drug delivery systems (ca. 200 nm) as well as the crystallization of the core, the 2.5 mg·mL−<sup>1</sup> amount of TP will be used for further studies at 10.0 mg·mL−<sup>1</sup> of PLA polymer concentration. Besides optimizing the amount of carrier and active drug, the ratio of the stabilizing PLUR surfactant was also studied. It was clearly confirmed that in the absence of PLUR, the TP molecules were not capsulated into the polymer core, only the binding of the TP drugs onto the surface of the polymer shell is observed (Figure 2). Furthermore, it was established that the smallest particle diameter (d = 178 nm), as well as the highest DL% (ca. 20%), is obtained at 0.05 mg·mL−<sup>1</sup> of PLUR concentration. Increase in the PLUR amount resulted in a higher particle size (212 nm) as well as a lower DL% values (9%). Most probably, the increase in the PLUR concentration facilitated the solubility of the hydrophobic TP, thus hindering the encapsulation process. In order to confirm the above-mentioned phenomena, turbidity measurements were carried out (Figure S2). According to the TEM images (Figure 2) and the turbidity studies (Figure S2), we can conclude that a low quantity (0.1 mg·mL<sup>−</sup>1) of PLUR surfactant is advantageous for the formation of PLA-based core-shell NPs, but the presence of a higher amount of PLUR (>0.1 mg·mL<sup>−</sup>1) results in the decrease of the EE%. Moreover, the presence of a higher amount of stabilizer (≥1.0 mg·mL<sup>−</sup>1) causes aggregation (Figure 2).

**Figure 2.** Representative TEM images of the TP-containing PLA NPs using different component concentrations (PLA: 1.25–10.0 mg·mL<sup>−</sup>1; TP: 0–5.0 mg·mL−<sup>1</sup> and PLUR: 0–1.0 mg·mL<sup>−</sup>1).

Besides the determination of the optimal ratio of the composite building blocks, the TP was encapsulated in PLGA75 and PLGA65 copolymers that have increasing hydrophilicity. During the NPs synthesis, the previously optimized concentrations of the polymer carrier (10.0 mg·mL<sup>−</sup>1), the PLUR (0.1 mg·mL−1) and the TP (2.5 mg·mL−1) were used. Based on DLS studies, we established that the particle size increases from 203 nm to 226 nm with a decrease of the lactide part in the polymer (Figure 3A). Using the optimized component quantities, the core-shell structure was confirmed for PLGA75 and PLGA65-based TP-containing NPs (Figure 3B). The EE% and DL% has been determined for the PLGA75 and PLGA65-based system as well, and we observed that the replacement of the PLA carrier to PLGA75 and PLGA65 polymers resulted in the increase in the EE% to 75.72% and 87.69%, respectively. In addition, the drug loading also increased from 14.73% (PLA) to 15.92% (PLGA75) and to 17.98% (PLGA65) with decreasing lactide content. The higher EE% and DL% can be explained by the fact that the precipitation of the more hydrophilic PLGA carriers is slower than that for PLA and TP, which helps the formation of the well-defined core-shell structure [13].

**Figure 3.** The particle size distribution of the TP-loaded PLA and PLGA NPs (**a**), and representative TEM images of the PLGA75–(**b**) and PLGA65–(**c**) based composites.

#### *3.2. Structural Characterization of the TP-Loaded PLA and PLGA Core-Shell NPs*

To determine the interaction between the PLA and TP, the composites have been examined by infrared spectroscopy measurements. Figure 4 displays the spectra of the PLA-based NPs in the absence (Figure 4A) and in the presence of TP at different concentrations (Figure 4B,C). The TP-sensitive bands appear in the range of 3050–2800 cm−<sup>1</sup> and 1150–1000 cm−1. It is obvious that the intensity of all the determinative bands systematically increase by increasing TP content. At 2994, 2944 and 2968 cm<sup>−</sup>1, the asymmetric and symmetric CH stretching vibrations of the –CH2 and –CH3 groups of the drug appear. Due to the increasing TP concentration, these bands become more intense, indicating the presence of TP molecules in the polymer NPs. The carbonyl group of the PLA appears at 1750 cm−<sup>1</sup> [28,29]. The TP does not contain a C=O group, thus this band could originate only from the polymer. In the fingerprint region (<sup>ν</sup> <sup>≤</sup> 1500 cm<sup>−</sup>1), the deformation and bending vibration of the –CH3 and –CH2 groups appear at 1453, 1381 and 1267 cm<sup>−</sup>1, while the band at 1181 cm−<sup>1</sup> attributes to the stretching mode of the C–O–C (ester). At around 1086 cm<sup>−</sup>1, further bands of the C–O–C stretching vibration can be observed, which have shoulders (symmetric and asymmetric stretching vibrations). Because of the Ar–O–C group in the TP, the intensity of this C–O–C symmetric stretching vibration at around 1050 cm−<sup>1</sup> is increased. The band at 865 cm−<sup>1</sup> and 751 cm−<sup>1</sup> is characteristic of the polymer carrier, and no shift is observed. The IR measurements performed for PLGA75 and PLGA65 resulted in similar spectra. A strong irreversible interaction between the PLA (or PLGA75, PLGA65) and the TP cannot be discovered (Figures S3 and S4), which facilitates the spontaneous release of the TP active drug from the polymer NPs, but the presence of the TP in the different composites were definitely confirmed by IR.

**Figure 4.** IR spectra of the TP-loaded PLA NPs (A: cTP = 0 mg·mL<sup>−</sup>1; B: cTP = 2.5 mg·mL<sup>−</sup>1; C: cTP <sup>=</sup> 5 mg·mL<sup>−</sup>1, cPLA <sup>=</sup> 10 mg·mL−<sup>1</sup> and cPLUR <sup>=</sup> 0.1 mg·mL<sup>−</sup>1).

Turbidimetric measurements were performed to investigate the interaction between the TP and the PLUR stabilizing agent in a pure aqueous solution and in PBS at 25 ◦C and 37 ◦C (pH = 7.4, 0.9% NaCl); the results are presented in Figure 5. In MQ water (Figure 5A), the turbidity of TP is systematically decreased till ca. 0.9 mg·mL−<sup>1</sup> of PLUR concentration. Over time, more and more TP can dissolve in this medium. In contrast, in the PBS solution at 25 ◦C and at 37 ◦C, the titration curves exhibit steeper decreasing intensity, which may be due to the reduced critical micellization concentration (cmc) of PLUR in the presence of salt. In the case of the PBS solution, the turbidity remains constant from 0.7 mg·mL−<sup>1</sup> at 25 ◦C, while an increase in the temperature to 37 ◦C 0.6 mg·mL−<sup>1</sup> value is observed. It is important to mention that the solubility of the TP scarcely depends on time at 37 ◦C, which allows the possible use of the PLUR stabilizing agent for in vitro drug release measurements.

It is well known that the surfactant affects the solubility of the TP drug in the absence of a polymer carrier. Namely, above cmc, due to the micellization ability of the PLUR, the solubilization of the drug dominates forming TP-loaded individual micelles, while below cmc, only the solubility increases. Accordingly, the cmc of the PLUR was measured by surface tension measurements (Figure 6). In the literature, very different values can be found; the obtained range of cmc is 2–7 mg·mL−<sup>1</sup> [30,31]. Similar values were measured, but we determined that at 25 ◦C, the obtained cmc is decreased from 4.92 mg·mL−<sup>1</sup> to 1.99 mg·mL−<sup>1</sup> in the presence of the phosphate buffer. If the temperature rises from 25 ◦C to 37 ◦C, these values are further decreased. We can conclude that the concentration of the PLUR in the composites is significantly lower than the cmc (4.92 mg·mL<sup>−</sup>1) value, which excludes the presence of TP-loaded individual micelles.

**Figure 5.** The turbidity of the TP in the PLUR solution at 25 ◦C in an aqueous medium (**a**), in a PBS buffer at 25 ◦C (**b**) and at 37 ◦C (**c**) (pH <sup>=</sup> 7.4, 0.9 *<sup>w</sup>*/*w*% NaCl) (cTP <sup>=</sup> 0.05 mg·mL<sup>−</sup>1).

**Figure 6.** Determination of PLUR cmc at 25 ◦C in MQ water and in a PBS solution.

#### *3.3. In Vitro Drug Release Experiments*

The determination of the exact TP amount released from the different composites was carried out by the UV-Vis spectrophotometric method. The spectra of the TP and the calibration curve are presented in Figure 7. The absorption bands of the TP appear in the UV range.

**Figure 7.** UV spectra of the TP in the PBS solution (cPLUR <sup>=</sup> 1.0 mg·mL<sup>−</sup>1, 0.9 *<sup>w</sup>*/*w*% NaCl).

After the characterization of the TP-loaded PLA and PLGA NPs, the mechanism of the drug release was investigated. The drug dissolution profiles and the fitting of these curves by different kinetic models are demonstrated in Figure 8 (Figure S5). The suitability of several models like Korsmeyer–Peppas, Peppas–Sahlin, the first-order and the Weibull models were investigated. The release curves clearly show that a high amount of TP is retained in the polymers after 7 h. It was also observed that in the first half an hour, the dissolution of the drug occurs relatively quickly, but after that, measurable slow dissolution is observed. Moreover, we found that the active substance is released slowly with a decrease of the lactide part (PLA, 35.0%, PLGA75, 28.3%, PLGA65, 19.8%) in 7 h. The slowest dissolution occurred in the carrier-free TP (15%). Because of the higher hydrophobicity, a higher amount of the non-encapsulated drug can be attached to the surface of the particles, which confirms the above-mentioned dissolution order [13]. Thanks to the application of the nanosized drug carrier systems (polymer NPs), the TP molecules can bind to the enhanced specific surface area of the NPs, which facilitates the dissolution of more TP, in contrast to the bulk TP.

**Figure 8.** Release profiles and different kinetic models-predicted (Peppas-Sahlin model (**a**), Korsmeyer-Peppas model (**b**)) for the release curves of TP from PLA and PLGA NPs in the PBS buffer (pH = 7.4, 0.9 *w*/*w*% NaCl).

Taking into account the coefficient of the determination (*R*2), the Peppas–Sahlin model was the best kinetic formula for our systems (Table 2). The values of the Case II relaxation contribution (*k*2) were negative in all cases; therefore, this model does not provide complete information about the dissolution of the TP, but the low diffusion exponent (*m*) is referred to for the Fickian release.


**Table 2.** Determined parameters of the TP release by fitting several kinetic equations.

The second-best kinetic model was the Korsmeyer–Peppas model where the diffusion dissolution index (*n*) gives the information about the diffusion and the erosion of the matrix. The values of the *n* are increased from *n* = 0.182 (PLA) to *n* = 0.223 (PLGA65), which is referred to for the diffusion-controlled quasi Fickian drug release. Furthermore, it is important to note that the value of the diffusion dissolution index is lower than 0.42. This is caused by the high polydispersity of the particles (because they are lyophilized) and the very low degradation of the polymers [32]. Presumably, the Case II relaxation contribution from the Peppas–Sahlin model will be low by these effects.

During the slow degradation of the PLA/PLGA carrier, the drug diffuses with difficulty from the core of the particles. The measured and calculated results clearly stated that the released drug originates from the surface region of the particles. Based on this, we could calculate the approximate quantity of the TP in the core of the particles after 7 h: 65.0% (PLA), 71.7% (PLGA75) and 80.2% (PLGA65) from the encapsulated mass of the TP; therefore, a significant amount of the active ingredient can be encapsulated inside the particles.

#### **4. Conclusions**

In the work presented here, the successful encapsulation of the hydrophobic α-Tocopherol, one of the determinative natural forms of vitamin E, was carried out using the PLA, PLGA75 and PLGA65 biocompatible polymer carriers by increasing hydrophilicity. To the best of our knowledge, we first proved the formation of well-defined nanosized TP-core PLA/PLGA-shell structured nanocomposites. Optimization of the experimental conditions, such as optimal concentration of drug, polymer carrier and stabilizing PLUR non-ionic surfactant, resulted in the formation of core-shell NPs within the diameter range of ~200–220 nm. For the PLA-based system 14.7% of drug loading was achieved, where most of the TP molecules are encapsulated in the core (65%), while the remaining part of the active ingredient is located on the surface of the polymer shell. By replacing the hydrophobic PLA to PLGA copolymers, both the drug loading (PLGA75: 16%, PLGA65: 18%) as well the EE% can be increased (PLA/TP: 69.1%; PLGA75/TP: 75.7%; PLGA65/TP: 87.7%). Furthermore, the PLGA-based composites contain 71.7% and 80.2% of the encapsulated TP in the core. Considering the slower precipitation ability of the PLGA copolymers in contrast to the PLA and the active ingredient, the higher

encapsulation efficiency can be explored. Besides the preparation and the characterization of the composites, the drug release was also studied. The dissolution curves clearly show that depending on the polymer, more than 65–80% of the TP is contained in the composites after 7 h. In the first half an hour, the dissolution of the drug occurs relatively quickly, but after that, measurable slow dissolution was observed. Moreover, we found that the active substance is released slowly with decreasing lactide part (PLA, 35.0%; PLGA75, 28.3%; PLGA65, 19.8%) in 7 h. Thanks to the biocompatibility, cost-effectivity and tuneable hydrophilic properties, the PLA/PLGA polymers are potential candidates for controlled drug release and for the encapsulation of hydrophobic TP or similar sized and structured molecules in core-shell nanosized particles.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/1999-4923/11/7/357/s1. Figure S1: UV spectra of TP in 1,4-dioxane at different concentrations, Figure S2: The turbidity of TP (a) and PLA (b) in PLUR solution at 25◦C in aqueous medium (cTP <sup>=</sup> 0.25 mg·mL<sup>−</sup>1, cPLA <sup>=</sup> 0.25 mg·mL<sup>−</sup>1), Figure S3: IR spectra of the TP- free (a) and TP- loaded (b) PLGA75 NPs (cTP <sup>=</sup> 2.5 mg·mL<sup>−</sup>1, cPLA <sup>=</sup> 10 mg·mL<sup>−</sup>1, cPLUR <sup>=</sup> 0.1 mg·mL<sup>−</sup>1), Figure S4: IR spectra of the TP- free (a) and TP- loaded (b) PLGA65 NPs (cTP <sup>=</sup> 2.5 mg·mL<sup>−</sup>1, cPLA <sup>=</sup> 10 mg·mL<sup>−</sup>1, cPLUR <sup>=</sup> 0.1 mg·mL<sup>−</sup>1), Figure S5: Release profiles and different kinetic models-predicted (First Order model (a), Weibull model (b)) release curves of TP from PLA and PLGA NPs in PBS solution (pH = 7.4, 0.9 *w*/*w*% NaCl).

**Author Contributions:** Conceptualization, E.C.; methodology, N.V.; investigation, N.V. and Á.T. and V.H.; data curation, Á.T.; writing—original draft preparation, E.C.; writing—review and editing, V.H.; supervision, E.C.

**Funding:** University of Szeged Open Access Fund.

**Acknowledgments:** The research was supported by the National Research, Development and Innovation Office-NKFIH through the project GINOP-2.3.2-15-2016-00060. V.H. acknowledges the Premium Post Doctorate Research Program of the Hungarian Academy of Sciences for the financial support. This paper was supported by the János Bolyai Research Scholarship of the Hungarian Academy of Sciences (E. Csapó). The Ministry of Human Capacities, Hungary grant TUDFO/47138-1/2019-ITM is acknowledged.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Monitoring the Fate of Orally Administered PLGA Nanoformulation for Local Delivery of Therapeutic Drugs**

**Lucia Morelli 1, Sara Gimondi 2, Marta Sevieri 2, Lucia Salvioni 1, Maria Guizzetti 1, Barbara Colzani 1, Luca Palugan 3, Anastasia Foppoli 3, Laura Talamini 2, Lavinia Morosi 2, Massimo Zucchetti 2, Martina Bruna Violatto 2, Luca Russo 2, Mario Salmona 2, Davide Prosperi 1,4, Miriam Colombo 1,\*,**† **and Paolo Bigini 2,\*,**†


Received: 31 October 2019; Accepted: 4 December 2019; Published: 6 December 2019

**Abstract:** One of the goals of the pharmaceutical sciences is the amelioration of targeted drug delivery. In this context, nanocarrier-dependent transportation represents an ideal method for confronting a broad range of human disorders. In this study, we investigated the possibility of improving the selective release of the anti-cancer drug paclitaxel (PTX) in the gastro-intestinal tract by encapsulating it into the biodegradable nanoparticles made by FDA-approved poly(lactic-*co*-glycolic acid) (PLGA) and coated with polyethylene glycol to improve their stability (PLGA-PEG-NPs). Our study was performed by combining the synthesis and characterization of the nanodrug with in vivo studies of pharmacokinetics after oral administration in mice. Moreover, fluorescent PLGA-nanoparticles (NPs), were tested both in vitro and in vivo to observe their fate and biodistribution. Our study demonstrated that PLGA-NPs: (1) are stable in the gastric tract; (2) can easily penetrate inside carcinoma colon 2 (CaCo2) cells; (3) reduce the PTX absorption from the gastrointestinal tract, further limiting systemic exposure; (4) enable PTX local targeting. At present, the oral administration of biodegradable nanocarriers is limited because of stomach degradation and the sink effect played by the duodenum. Our findings, however, exhibit promising evidence towards our overcoming these limitations for a more specific and safer strategy against gastrointestinal disorders.

**Keywords:** PLGA-NPs; nanomedicine; gastrointestinal tract; paclitaxel; in vivo imaging

#### **1. Introduction**

The generation of therapeutic methods that can improve the organ specificity of many different types of therapeutic agents is one of the major challenges of 21st-century pharmacology [1]. The availability of transport systems that can successfully improve the targeted release of administered

substances would be, indeed, crucial in mitigating or even virtually eliminating the toxic side effects of already existing therapeutics. These improved transport systems can be used to limit the off-target spread and, consequently, reduce the potential side effects deriving from an unnecessary tropism [2]. Several strategies have been employed in the last years to improve the accuracy of drug targeting. Quite recently, our group and others have demonstrated that the systemic administration of anti-inflammatory and immunomodulatory agents, such as corticosteroids, can have a safer and more effective pharmacological profile when linked to a nanometric vector [3,4]. The potential prospects of nanocarriers for pharmacological purposes is extremely versatile complex. It is widely known that even the slightest modification of basic physicochemical strictures, such as the dimension, shape, and surface of the material itself, can indeed lead to a myriad of different results—causing a very vast range of reactions amongst the cells and organs of the host [5]. One of the primary and certainly most prominent, therapeutic strategies is the incorporation of the anti-neoplastic drug doxorubicin into liposomes for the treatment of tumors [6]. The first generation of nanoparticles (NPs) used for these purposes are mainly liposomes, polymeric micelles, and conjugates.

Since the turn of the century, the encapsulation of anti-neoplastic cytotoxic molecules inside NPs has been occurring with greater frequency. In 2004, an original formulation for the creation of nanocarriers to transport the anti-cancer drug paclitaxel (PTX) was developed [7,8]. In that case, different from other carriers, the biodegradable copolymer poly(lactic-*co*-glycolic acid) (PLGA) was utilized. In a more recent study, furthermore, it was shown that the systemic administration of PTX inside poly(lactic-*co*-glycolic acid)-polyethylene glycol-nanoparticles (PLGA-PEG-NPs), when functionalized with ANTI-EGFR, dramatically slows disease progression in triple-negative breast cancer-bearing mice, thus confirming the high efficacy, safety, and tenability of this material [9].

Nanoparticles utilizing FDA-approved PLGA, exhibit some important advantages. In particular, their biocompatibility causes low systemic toxicity in comparison to other polymers [10] and is highly versatile, allowing them to encapsulate various drugs, including small molecules and therapeutic biologics [11–13]. The importance of PLGA nanocarriers in biomedicine is further supported by evidence of its potential benefits in assisting drug transportation across biological barriers [14,15]. Most of their applications concern intravenous administration [16], but, in the last decade, many studies have been published regarding their potential efficacy in oral administration of drugs, even if the absorption efficiency needs to be more deeply investigated and optimized [17].

Although the majority of studies dealing with biodistribution, disease targeting, and therapeutic efficacy of NPs-based drug delivery systems has been conducted using systemic administration of nanoformulation, more recently the exploration of alternative routes, including topical [18,19], intranasal [20], intratracheal [21], and oral administration [22], are gaining increasing recognition.

Building on these advances, this work aimed to characterize the biodistribution of PTX alone or encapsulated in PEGylated PLGA-NPs after oral administration in healthy mice. This approach stems from the hope that, in so doing, the passage of the drug from the gastrointestinal (GI) tract to the bloodstream might be reduced, although not completely eliminating its release in the gut. While the active gastroduodenal absorption of PTX on oral ingestion is quite low, it is not possible to totally exclude systemic toxicity of both circulating cells and filter organs in the case of high doses or repeated cycles of chemotherapy. PLGA-NPs could, therefore, be useful to both protect the drug against gastric degradation and, at the same time, reduce the uptake by mucosae villi or Peyer's patches and the consequent release into the blood, while PEG grafting is expected to improve both the NPs stability and mucus-permeating properties in the GI tract [23,24]. In this study, PLGA-PEG-NPs were orally administered either loaded with PTX or labeled with a fluorophore (Rhodamine B), which allowed for the longitudinal tracking of NPs in both gastrointestinal and peripheral organs. This study could be of interest to understand the potentials of local delivery of anti-cancer agents in GI tumors by the oral administration of NPs, minimizing nonspecific absorption.

#### **2. Materials and Methods**

#### *2.1. Nanoparticles Synthesis*

#### 2.1.1. Materials

PLGA (Resomer® RG 504 H, 50:50 lactide:glycolide, acid terminal, MW 38,000–54,000 g/mol), paclitaxel (PTX, MW 853.91 g/mol), rhodamine B (RhB, MW 479.01 g/mol); polyvinlyl alcohol (PVA, MW 9000–10,000 g/mol, 80% hydrolyzed); *N,N* -dicyclohexylcarbodiimide (DCC, MW 206.33 g/mol, 99%); *N*-hydroxysuccinimide (NHS, MW 115.09 g/mol, 98%); dimethylamino pyridine (DMAP, MW 122.17 g/mol, 99%); ninhydrin (MW 178.14 g/mol); ethanolamine (MW 61.08 g/mol, ≥98%); dichloromethane (DCM, ≥99.8%); dimethyl sulfoxide (DMSO, ≥99.9%); acetonitrile anhydrous (ACN, 99.8%); methanol (MeOH, ≥99.9%) were purchased from Sigma–Aldrich Co (St Louis, MO, USA). Chloroform (≥99.8%) was purchased from Honeywell Riedel-de-Haën™ (Charlotte, NC, USA)). Poly (ethylene glycol) diamine (NH2-PEG-NH2, 6000 Da) was obtained from Rapp Polymer GmbH, (Tuebingen, Germany).

#### 2.1.2. PLGA-PEG-RhB Synthesis

For activation of the carboxylic groups of PLGA, the polymer was dissolved in DCM (15 mL, 40 mg/mL) and the reaction was performed in a round flask for 4 h (250 rpm, RT, under inert gas) after the addition of DCC (30 mg) and NHS (15 mg). Then, the solution was diluted with diethyl ether, and the polymer was collected after precipitation (20 min, 6200× *g*, 4 ◦C) and evaporation of the residual organic phase. Next, the conjugation with NH2-PEG-NH2 (molar ratio PEG/PLGA = 2.7) was performed in 10 mL of CHCl3 overnight (250 rpm, RT, under inert gas). To get rid of the unconjugated PEG, methanol (30 mL) was added, and the polymer was collected after precipitation twice [25]. The residual organic phase was then evaporated under reduced pressure, and the weight yield calculated (68.99 ± 14.94%). The degree of labeling was determined using the colorimetric ninhydrin assay (primary amine detection): Notably, 15 mg of the product (PLGA-PEG) was solubilized in DMSO (400 μL) then diluted with 100 μL of ninhydrin solution (3.5 mg/mL). Ethanolamine (18.75 mg/mL) and PLGA solutions were used as positive and negative controls, respectively. The samples were then incubated for 1 h at 65 ◦C (400 rpm), and the colored conjugate was detected by UV-Vis spectroscopy (Spectrophotometer Fluormax-Horiba Scientific, Rome, Italy) (λ = 600 nm), and the yield calculated using a calibration curve of ethanolamine standards. The obtained degree of labeling was 71.90 ± 12.70%.

Finally, RhB (molar ratio RhB/PLGA = 6.25) was added to the PLGA-PEG solution in CHCl3 (15 mL) with an excess of DCC and DMAP. The reaction was carried out overnight (250 rpm, RT, under inert gas). At the end of the reaction, the solution was evaporated, and the product was solubilized with DCM (10 mL) and purified thrice by precipitation (20 min, 4 ◦C, 6200× *g*) after diethyl ether addition (20 mL). The reaction yield was 38.16 ± 13.50%. RhB linked was detected by fluorescence spectrometry (λexitation 555 nm; λemission 574 nm) and the degree of labeling calculated by the comparison with a calibration curve (1.43 ± 0.46 μg/mg).

#### 2.1.3. Synthesis of PLGA-PEG-RhB-NPs and PTX-PLGA-PEG-RhB-NPs

NPs were synthesized according to the single emulsion method. PLGA-PEG-RhB solution (1 mL, 25 mg/mL in DCM) was emulsified with 8 mL of PVA 2% solution by sonicating twice (Sonifier Sound, Branson Ultrasonics, Shanghai, China; 30 s and 38% intensity) in an ice bath. The product was transferred immediately into a solution of PVA 2% (16 mL), allowing the organic solvent to evaporate (4 h, RT, 750 rpm). For PTX-PLGA-PEG-RhB-NPs synthesis, 5 mg of PTX was dissolved along with the polymer. After curing, NPs were collected by centrifugation at 19,500× *g*, 20 min, 4 ◦C (Heraeus Fresco 21; Thermo Fisher Scientific, Göteborg, Sweden), washed thrice with double distilled water and freeze-dried through an Alpha 1–2 LD freeze drier (Christ, Memmingen, Germany) at

0.500 mbar, −53 ◦C, 12 h, without the addition of any cryoprotectant. The process yield was calculated after the freeze-drying process, as the ratio between collected NPs and starting raw materials.

The amount of PTX loaded in the NPs was determined by HPLC analysis with UV detection (Waters Associates, Milford, MA, USA, model 2487 Variable Wavelength Detector, Wavelength: 230 nm). Briefly, 0.1 mL of the NPs solution was spiked with 5 μg of IS and extracted with 0.5 mL of CH3CN. After vortex for 10 s, samples were centrifuged at 13,000 rpm for 10 min. The organic phase was separated and dried under nitrogen, and the residues were dissolved with 250 μL of the mobile phase. Fifty microliters of the reconstituted samples was injected into the HPLC system. The apparatus was equipped with a Symmetry C18 column (5 μm, 4.6 × 150 mm), the mobile phase was composed of 50% ammonium acetate buffer (0.01 M, pH 5), 40% acetonitrile, and 10% methanol with a flux rate of 1.3 mL/min and 30 min run time. NPs solution without drug was used to prepare the calibration curve by the addition of PTX in the range 10–100 μg/mL.

#### *2.2. Nanoparticles Characterization*

#### 2.2.1. Dynamic Light Scattering (DLS) and Zeta Potential Measurements

All NPs were characterized in terms of size, size distribution, and Zeta potential through dynamic light scattering (DLS, Zetasizer Nano ZS; Malvern Instruments, Cambridge, UK) in ultrapure water at 1 mg/mL, at 25 ◦C. The scattered light from the NPs in suspension was used to calculate NPs' hydrodynamic diameter considering medium viscosity. NPs size distribution was described by the polydispersity index (PDI), where PDI ≤ 0.2 corresponds to the monodisperse NPs population.

#### 2.2.2. Scanning Transmission Electron Microscopy (STEM) Analysis

PLGA-PEG-RhB-NPs were observed by a Zeiss SEM-FEG Gemini 500, operating at 30 kV in scanning transmission electron microscopy (STEM) mode (Zeiss, Germany). The NPs suspension was deposited onto a formvar-coated 200-mesh copper grid (Ted Pella, CA, USA), negatively stained with 1% uranyl acetate and allowed to dry before examination.

#### 2.2.3. In Vitro Release Studies

PTX-PLGA-PEG-RhB-NPs equivalent to 8.75 μg of PTX were suspended in 1 mL of Tween/PBS and incubated at 37 ◦C with constant agitation. At predetermined time points (0.5, 1, 2, 4, 6, and 24 h), the suspension was centrifuged at 19,500× *g* for 15 min to separate NPs pellets and supernatants. PTX in the collected supernatants was analyzed with HPLC equipped with a UV detector (1260 Infinity II Series, Agilent Technologies, Palo Alto, CA, USA) and an Atlantis C18 column (25 cm × 4.6 mm, particle size 5 μm) (Supelco, St. Louis, MO, USA). The mobile phase was a mixture of acetonitrile and water (50:50) run in the isocratic mode at a flow rate of 1 mL/min. PTX was detected at 227 nm. PTX quantitation was performed using a calibration curve in a range of 1.25–40 μg/mL, and the results were expressed as a percentage of cumulative release (mean ± SD; *n* = 4).

#### 2.2.4. Nanoparticles Stability in Different Buffers

The hydrodynamic dimension of the nanocarrier was analyzed in different conditions: Artificial saliva (pH 7.6; KH2PO4 1.9 mM, NaHCO3 17 mM, Na2HPO4 1.8 mM, NaCl 8.5 mM, KSCN 3.4 mM, Urea 2.2 mM, alpha-amylase 0.007 mM), intestinal fluid (pH 6; KH2PO4 49.96 mM, NaOH 0.0154 M, pancreatin 1 g/L), and gastric juice (pH 1.2; NaCl 300 mM, pepsin 0.9 mM, HCl 840 mM) that simulate the media in GI tract. The NPs (15 mg/mL) were incubated for 0, 1, 2, 4, 6, and 24 h and analyzed by DLS. Upon establishing the size stability, the samples were further analyzed with fluorescence spectroscopy (λexitation 555 nm; λemission 574 nm) to monitor the stability of RhB conjugation. The release of RhB after 24 h was then calculated after PLGA-PEG-RhB-NPs removal by centrifugation (mean ± SD; *n* = 4).

#### *2.3. Cells*

Carcinoma colon cells (CaCo2) are a cell line derived from human colorectal adenocarcinoma and were purchased from ATCC® HTB-37™. CaCo2 cells were grown as a single layer in adhesion in DMEM-High Glucose (Dulbecco s Modified Eagle Medium High Glucose-Biowest, Nuaillé, France) with the addition of 10% of fetal bovine serum (FBS) and 1% l-glutamine (200 mM), 100 U/mL penicillin, 0.1 mg per mL streptomycin. Cells were maintained at 37 ◦C and in a 5% CO2 humidified atmosphere. The ability of PLGA-PEG-RhB-NPs and free RhB to be internalized by CaCo2 cells was evaluated. For each condition, the experiment was conducted in triplicate. Cells were seeded at a density of 40,000 cells/well on 13 mm diameter slides in 24-well plates. Forty-eight hours after sowing, CaCo2 cells were incubated with PLGA-PEG-RhB-NPs (100 μg/mL) and with RhB free (0.06 μg/mL) for 1, 4, and 24 h, while the wells destined for control did not receive any treatment. After incubation, the cells were washed with phosphate-buffer saline (PBS) three times and fixed with a 4% paraformaldehyde solution dissolved in PBS (0.1 M, pH 7.4) for 40 min and the vital nuclear dye Hoechst 33258 (2 μg/mL) was added to each well for 45 min. The slides thus obtained were assembled and analyzed with the Olympus BX51 epifluorescence microscope (Olympus, Tokyo, Japan). All acquisition parameters, including laser settings, were kept constant during all scans. To evaluate a possible cytotoxic effect of PLGA-PEG-RhB-NPs and free RhB, the metabolic activity of CaCo2 cells was evaluated by RealTime-Glo™ MT Cell Viability Assay (Promega, Madison, WI, USA). Cells were seeded at 16,000 cells/well in 96 opaque-walled tissue culture plates and maintained at 37 ◦C. The RealTime-Glo reagents were added at the same time as the test compounds, according to the manufacturer's protocol. At selected time points (4 and 24 h), the cell viability was monitored by a plate-reading luminometer (GloMax® Discover Microplate Reader, Promega, Madison, WI, USA). For each condition, 6 replicates were prepared. The viability was expressed as a percentage compared to non-treated cells.

#### *2.4. Animals*

All procedures involving animals and their health were conducted so as to minimize the number of mice used and their collateral suffering, in accordance with institutional guidelines, national laws (DL n. 24, 4 March 2014; Authorization No. 19/2008 A) and international laws and agreements (EEC Council Directive 2010/63, 6 August 2013; NIH Guide to the Care and Use of Laboratory Animals, US National Research Council, 2008). The research project was first reviewed by the Internal Ethics Committee of the Mario Negri Institute and was subsequently approved by the ministry and designated with the before mentioned code (42/2016-PR).

All animals were housed in SPF (specific pathogen-free) conditions. The housing rooms had a temperature of 22 ± 1 ◦C, relative humidity values ± of 50 ± 10% and a 12 h light/dark cycle. Furthermore, the animals were kept in cages with free access to water and food.

#### *2.5. Treatments*

As regards pharmacokinetics, six weeks old CD1 mice were treated by oral gavage with 20 mg/kg of Cremophor PTX (*n* = 16) or PTX-PLGA-PEG-RhB-NPs (*n* = 16), three animals were treated with saline solution as control. Blood was collected in heparinized tubes, 30 min, 1, 4, and 24 h after treatment (4 animals for each group) and centrifuged to obtain plasma. After the mice were sacrificed, the stomach, duodenum, colon, and liver were collected and stored at −20 ◦C until analysis.

For the biodistribution studies, we recruited 5 animals for each experimental group. To reduce background fluorescence due to the food, mice were fed an AIN-76A diet without alfalfa (Mucedola s.r.l., Settimo Milanese, Italy) for two weeks before the analyzes. The dose of the different formulations was standardized on the quantity of RhB present and fixed at 0.6 mg/kg mouse based on previous experimental studies. Vehicle treated mice received the same volume of saline solution. Mice were sacrificed at 1, 4, or 24 h to follow the fate of RhB. The sacrifice was performed by cervical dislocation, and the stomach, intestine, liver, and blood were collected to perform ex vivo analyses. Plasma for fluorometric analysis was obtained from the blood collected. Finally, a piece of liver and GI tracts of three animals for each experimental group were frozen at −80 ◦C for cryostatic sections and histological analysis.

#### *2.6. Molecular Imaging and Histology*

The in vivo biodistribution of the different formulations was monitored over time using an optical fluorescence imaging system (IVIS Lumina XRMS, PerkinElmer, Waltham, MA, USA). Ex vivo scans of organs from mice sacrificed at 1, 4, and 24 h after the treatment were performed by the same instrument. The following acquisition parameters were used: Excitation filter: 580 nm, emission filter: 620 nm, exposure time: Auto, binning factor: Medium, f/Stop: 2, Field of View: D (for the gastro line—intestinal), C (for peripheral organs). Very importantly, the Living Image Software 4.3.1 (Perkin Elmer, Waltham, MA, USA) conjugated with the spectral unmixing system was used to separate the RhB signal from tissue autofluorescence, image processing, and fluorescence signal quantification analysis.

Longitudinal sections of 20 μm of thickness were prepared and then, after adhesion in glass slides, were incubated with a PBS solution of Hoechst 33258 (2 μg/mL, Sigma–Aldrich) for 45 min and, after three washes in PBS, observed at the Microscopy Virtual Slide (Olympus, Tokyo, Japan), to obtain rapid organ scans of the whole section with high anatomical resolution.

#### *2.7. NPs Characterization from Homogenates and Biological Fluids*

The fluorescence of PLGA-PEG-RhB-NPs was analyzed in tissues explanted from previously treated animals. The analyzed tissues were liver, stomach, intestine, and plasma. Each tissue sample was weighed, homogenized in PBS 1X according to a 1:4 weight ratio, and centrifuged at 1200 rpm for 10 min at 4 ◦C. For all samples, the analysis was performed using the Infinite® M200 multimode plate reader exciting at the wavelength of 500 nm and recording the signal at a range of emission wavelength from 550 to 560 nm. For the stability studies in solutions mimicking gastrointestinal fluids, the NPs were incubated in stock solutions as previously described [26].

#### *2.8. Pharmacokinetics*

The total concentration of PTX in the different biological matrices was determined by HPLC-UV, as previously described [27]. For the determination of PTX in organs, tissues were previously homogenized in 0.2 M CH3COONH4 pH 4.5. Each study sample (0.3 mL for plasma and 0.5 mL for homogenate tissues) was assayed together with five points of a standard calibration curve prepared in the corresponding control biological matrix obtained from untreated mice at concentrations ranging from 0.05 to 5 μg/sample. The limits of quantification (LOQ) were 0.16 μg/mL and 0.6 μg/g for plasma and organs, respectively.

#### *2.9. Statistical Analysis*

All data were expressed as mean ± SD, Student's *t*-test and *p* values were done using the GraphPad Prism version 6.00 for Windows (Graph-Pad Software, San Diego, CA, USA).

#### **3. Results and Discussion**

#### *3.1. NPs Synthesis and Characterization*

The protocol for the polymer modification was set up, and PLGA-PEG-RhB was employed to synthesize the NPs to be used as carriers for PTX. The covalently bound RhB chromophore allowed the tracking of the PLGA fate in vivo. The single emulsion method allowed us to obtain both unloaded and PTX-loaded NPs, having controlled size and homogenous size distributions (Table 1). As measured by DLS analysis, the encapsulation of PTX did not significantly change the dimensional features of NPs. Moreover, a negative Z-potential was observed as opposed to PLGA-PEG-NPs, where the free amines of PEG determined the surface properties. However, a small difference in net negative charge values between PLGA-PEG-RhB-NPs and PTX-PLGA-PEG-RhB-NPs was attributable to the intramolecular reorganization of portions of the polymer chains due to hydrophobic interaction with PTX. PTX loaded NPs were further characterized by STEM, which showed pseudospherical shaped nanoparticles with the size around 200 nm (Figure S1). PTX encapsulation efficiency detected by HPLC analysis was 10.87 ± 1.13%, while the calculated loading efficiency was 4.77 ± 0.15%, in line with previous studies [28]. The drug release performance was evaluated, and the test was conducted in PBS containing 0.2 *v*/*v* % Tween 80, considering both its poor solubility and the analysis detection limits [29]. As previously reported with similar experimental settings, a fast PTX dissolution was observed (80% within 4 h) (Figure S2) [30].


**Table 1.** NPs characterization and process yield. Data represent mean ± SD (*n* = 3).

Stability in Mimicking Biological Fluids

The first study was carried out to verify the potential role of PEG on the fate of PLGA-NPs in solutions mimicking saliva, gastric, and proximal intestinal fluid, respectively (Figure 1A,B). Longitudinal measurement of the NPs size by DLS was possible because, diversely from the serum, they are very poor in macromolecules, and the interference on the recording is almost zero. Figure 1A shows that the presence of the PEG confers to PLGA-NPs a long-lasting stability for at least 24 h after incubation. In contrast, the lack of PEG rapidly induced aggregation in NPs incubated in gastric fluid and, to a lesser extent in intestinal fluid. Since we aimed to conserve considerable stability up until the large intestine, we decided to carry on our studies using pegylated PLGA-NPs exclusively.

**Figure 1.** Graphics showing the diameter, measured by dynamic light scattering (DLS), after incubation of (**A**) pegylated or (**B**) not pegylated poly(lactic-*co*-glycolic acid)-nanoparticles (PLGA-NPs) in artificial fluids mimicking the saliva (green lines), the gastric juice (blue lines), or proximal intestinal fluid (yellow line). For all conditions, the measurements were carried out at 1, 2, 4, 6, 24 h from the starting point.

Then, the stability of the RhB conjugation was performed by measuring the dye release in the same conditions (Figure 2A). After 24 h of incubation, the released RhB was below 15% in all cases, suggesting the system reliability for in vitro and in vivo NPs tracking. Additionally, as reported in Figure 2B, the fluorescence intensity of PLGA-PEG-RhB-NPs was not greatly affected by the incubation media showing that, once conjugated, the dye abolishes its pH-dependent emission properties [31].

**Figure 2.** Graphics showing (**A**) the amount of RhB released from PLGA-PEG-RhB-NPs after 24 h incubation in artificial fluids mimicking the saliva, the gastric juice, or proximal intestinal fluid (mean ± SD; *n* = 5); (**B**) emission spectra of poly(lactic-*co*-glycolic acid)-polyethylene glycol-RhB-nanoparticles (PLGA-PEG-RhB-NPs) incubated in different media.

#### *3.2. Pharmacokinetics*

To understand the possible influence of the nanoformulate on the biodistribution of drugs, we performed a pharmacokinetic study of the well-known anti-cancer agent, paclitaxel, administered orally as a free drug (PTX) or loaded into NPs (PTX-PLGA-PEG-RhB-NPs). PTX release Figure 3 shows the levels of PTX measured in the stomach (A), in the duodenum (B) and in the colon (C) of mice sacrificed 30 min, 1 and 4 h after treatment with PTX free (blue bars) or with PTX-PLGA-PEG-RhB-NPs (red bars). The drug concentration in the samples collected at 24 h after treatment resulted under the limit of detection. Overall, the drug concentration measured in all the gastrointestinal tissues was comparable between the two formulations and higher in the stomach and duodenum compared to the colon. Despite the comparable tissue drug level, interestingly, the nano-formulation reduced the systemic absorption of the drug. The measurement of the plasmatic levels, in fact, showed a clear reduction of systemic absorption of PTX when administered as PTX-PLGA-PEG-RhB-NPs, leading to a consequent reduction of liver accumulation, as shown in Figure 3D,E. Similarly to previous results achieved by our group [32,33], no evidence of acute toxicity was observed in mice receiving the single administration of PTX-PLGA-PEG-RhB-NPs.

**Figure 3.** Comparison of the paclitaxel (PTX) distribution in (**A**) stomach, (**B**) duodenum, (**C**) colon, (**D**) plasma, and (**E**) liver of mice after administration of a single treatment of PTX or PTX-PLGA-PEG-RhB-NPs (20 mg/kg p.o.). The bars are the mean value ± SD (*n* = 4).

#### *3.3. NPs Biodistribution and Nanosafety*

To better understand the mechanisms that govern the different behavior of the free and NPs-encapsulated drug, we decided to monitor the NPs transit at the digestive tract level, the interactions with the gastric and intestinal structures, the possible transition to the bloodstream and penetration into filter organs by marking PLGA with a fluorescent molecule, RhB. Since the monitoring of biodegradable NPs through an indirect method, such as fluorescence analysis, presents the risk of artifacts caused by the dye release or degradation [34], all studies were carried out by comparing animals that received PLGA-PEG-RhB-NPs with those animals treated with the same dose of free RhB. As the in vitro studies suggested, the stability and reliability of this indirect approach (see Figure 2), enable us to obtain further evidence from the biodistribution study that explains the different behavior between an orally administered small molecule and our nanocarrier.

Figure 4A shows the distribution of the signal associated with the GI tract in animals sacrificed 1, 4, and 24 h after treatment with the vehicle (left) or RhB (upper blue panel), and PLGA-PEG-RhB-NPs (lower red panel), respectively. The power of the excitation laser was set in vehicle-treated mice to avoid any possible overlapping between RhB and tissue autofluorescence. In each panel, it is possible to see an upper region shaped like a sack, which represents the stomach. The small intestine is the snake-like shape in the middle. The enlargement in the last part of the intestine includes cecum, colon, and rectum and is called "large intestine". The signal associated with RhB is clearly detectable in both groups receiving the same amount of dye. As expected, in both groups, it is also possible to see a progressive decay of signal and a shift from the distal part of the GI tract. However, in mice treated with NPs, the presence of the dye was more persistent, in particular in the distal part of the small intestine and in the large intestine. The quantification of the signal (Figure 4B) confirmed this observational study: In particular, the levels of the signal in the intestine at the 4th hour after the treatment was markedly higher in animals receiving the nanoformulation.

To better understand the interaction between PLGA-PEG-RhB-NPs and the GI tract structures, histological evaluation was carried out, also exploiting the presence of RhB to track them along the anatomic path and their development in time. Even in this case, the same doses of RhB free were injected into a further group of mice to exclude any possible misinterpretation of the results due to the possible loss of NPs stability and the release of the dye.

**Figure 4.** Signal distribution in the gastrointestinal (GI) tract of mice treated with RhB or PLGA-PEG-RhB-NPs. (**A**) Representative ex vivo scanning of the excised organs after washing with saline solutions to remove debris of feces. Animals were sacrificed at 1, 4, and 24 h after the treatment with the same dose of RhB. In the left column, a vehicle-treated mouse was shown to demonstrate the lack of the auto fluorescent component in this analysis. The interval of fluorescence signal intensity associated with the scale of colors is reported on the right. Five animals for each experimental group were used. (**B**) The quantification of the signal associated with the treatment was performed, dividing each sample into three tracts, as shown in the panel. The bars are the value of signal normalized to 100, considering the mean value measured in the stomach of mice treated with RhB alone and sacrificed 1 h after the oral administration. The bars are the mean value ± SD (*n* = 5). The Student's *t*-test was used to compare the levels between the two groups for each time point. \*\*\* *p* < 0.0001.

Figure 5A, upper panels, shows representative images taken from gastric sections of mice sacrificed 1, 4, and 24 h after PLGA-PEG-RhB-NPs administration. Although a progressive reduction of the signal can be seen, it is important to underline that the anatomical localization of the signal remains almost exclusively confined outside the gastric cells that are characterized by the intense blue staining due to the presence of the nuclear dye Hoechst 33258. Opposite, the RhB alone, Figure 5A lower panels, deeply penetrated inside the gastric parenchyma as clearly evidenced by the purple staining due to the merge between the red and the blue signal. This is more pronounced at the first hour and, interestingly, it does not involve the whole structure of the stomach, but it is almost confined to the superficial region of the gastric mucosa (left part of the picture at 1 h). At the 4th hour, the signal is lower but more penetrated and homogeneously spread in the parenchyma, whereas after one day, as already demonstrated by ex vivo quantification (Figure 4B), the fluorescent intensity strongly decreased. The different pattern of staining shown at the 4th and 24th hour strongly suggests that, in spite of the gastric activity, PLGA-PEG-RhB-NPs remain stable enough to avoid the release of the free dye. This is in line with the results reported in Figure 1A by DLS in solutions mimicking gastric juice.

The stability of the nanoparticles inside the stomach is essential to transport any encapsulated drug to the intestine. In Figure 4B, the signal measurement along the whole small intestine was evaluated, whereas histological analysis was focused on the more proximal part of the intestine, the duodenal region. The duodenum is one of the most critical portions of the GI for the absorption of metabolites and drugs. The active uptake by mucosae villi and Peyer's patches allows the absorption of many substances into the bloodstream and their consequent systemic distribution. Even if this process is required to provide energy and nutrients and to distribute therapeutic agents orally administered, it can be a hurdle for a localized gut delivery. In Figure 3, we have reported that the encapsulation in PLGA-PEG-RhB-NPs dramatically reduces the PTX absorption. This suggests that these kinds of NPs are able to pass through this first part of the intestine, maintaining their stability. Representative images from coronal sections of the duodenum from mice sacrificed 1 h after the treatment (lower panel on the left and higher magnification right in Figure 5B) shows that NPs are in the lumen and inside the intervilli space but are not absorbed by mucosa. A deeper interaction with villi can be seen at the 4th hour after treatment. However, even in this case, the red and blue signals are close but do not overlap. An overlapping was indeed clearly seen in distinction from mice treated with the same amount of RhB free at least up to the first 4 h after ingestion. A higher magnified picture furthermore confirms the restricted interaction between the red signal and the peripheral region of villi 1 h after treatment. As reported in the measurement of the intestinal levels by ex vivo scanning (Figure 4B), an almost complete disappearance of the red signal can be seen 24 h after the treatment in both experimental groups.

By histology, we found that RhB-free treated mice showed deep red staining in villi, whereas the red signal remained in the lumen of the intestinal tube in animals receiving RhB with NPs. Since enterocytes in villi are tightly connected to the vessels, it is, therefore, possible to hypothesize that RhB can easily penetrate the circulatory tree. To confirm that NPs can dramatically reduce the passage from the small intestine to the bloodstream, we compared the RhB levels both in plasma and in the liver of mice treated with PLGA-PEG-RhB-NPs or RhB free. Figure 6A, where plasmatic levels of RhB-related signals were normalized to the value measured in mice treated with RhB-free during the first hour of analysis, clearly reveals that the nanoformulation (red bars) leads to an almost complete abolishment of hematic absorption of RhB. This striking difference between the two groups supports the hypothesis that these NPs are stable and almost completely eradicate the absorption of themselves and of the relevant cargo by gastric and intestinal mucosae. The fast and quite elevated half-life of RhB in the blood led to an expected accumulation of the dye in the main filter organ, the liver, Figure 6B. Similar to the results obtained from the blood, the animals treated with RhB exclusively showed a well detectable red signal in liver sections (Figure 6C).

**Figure 5.** Representative images showing the distribution of RhB in (**A**) stomach and (**B**) duodenum of mice treated with either PLGA-PEG-RhB-NPs or RhB alone. (**A**) In the upper panels, the localization of signal (red) in gastric sections (blue) is shown 1, 4, and 24 h after the oral administration of PLGA-PEG-RhB-NPs. The same procedure has been used to track the presence of the dye in mice receiving the same amount of RhB (lower panels). Scale bar 100 μm. (**B**) Representative images of the duodenum are shown of PLGA-PEG-RhB-NPsand RhB-treated mice in upper and lower panels, respectively. Scale bar 200 μm. The thicker and more intense blue staining in the periphery of the sections represents the basal layer of the duodenum where the exchanges of tissue/blood occur. The red signal is more concentrated to the center, likely corresponding to the lumen close to the apical side of the villi. A higher magnified picture from a mouse sacrificed 1 h after the ingestion of NPs confirms the weak interaction between NPs and villi. In the duodenum of RhB-treated mice, a deep overlapping between the villi and the RhB was observed at both 1 and 24 h after treatment. A higher magnified image confirms the penetration of the dye into the external side of the villi. Scale bar 50 μm.

**Figure 6.** Signal measurement in (**A**) the plasma and (**B**) the liver of mice orally treated with RhB (blue bars) or PLGA-PEG-RhB-NPs (red bars). Animals were sacrificed at 1, 4, and 24 h after the treatment with the same dose of RhB. Five animals for each experimental group were used. In both graphics, quantification of the signal was normalized to the RhB level at 1 h and expressed as 100. The bars are the mean value ± SD (*n* = 5). The Student's *t*-test was used to compare the levels between the two groups for each time point. \*\*\* *p* < 0.0001. (**C**) Representative images of the signal related to the dye in a section of liver from mice treated with PLGA-PEG-RhB-NPs (upper panel) or RhB alone (lower panel), both sacrificed 1 h after the treatment. Scale bar 50 μm.

The last part of the study was carried out to investigate if these NPs were able to penetrate into CaCo2 cells and where they localize inside the cells. This experiment was aimed at exploring the possible application of our results to future local treatment of colorectal cancer. Figure 7A shows the progressive process of internalization of NPs (orange spots). The quantification of the occupied area of NPs inside the cell cytoplasm is reported in Figure 7B. Progressive penetration of NPs occurs and, at the 4th hour after incubation, they already occupied the 2% to 3% of the whole cell area. Although relatively low, this percentage in terms of a potential release of a therapeutic cargo cannot be considered negligible. The red arrow in Figure 7A and the higher magnified picture in Figure 7D confirm that NPs are deeply penetrated (orange spots) and the deep internalization of NPs inside the cell cytoplasm starting from the 4th hour of incubation. Moreover, the cell viability assay confirmed the safety of the materials we selected for this study. Indeed, neither RhB nor PLGA-PEG-RhB-NPs alone modify the healthiness of the cells for the whole duration of the treatment (Figure 7C).

**Figure 7.** Longitudinal study to evaluate PLGA-PEG-RhB-NPs internalization in CaCo2 cells. (**A**) Low magnified pictures showing the progressive penetration of NPs in cells. Representative images have been selected from non-treated cells (NT, left panel). For each image, the nuclei were stained with Hoechst 33258 (blue), whereas high excitation with the laser at the wavelength of 488 nm will allow unveiling the border of the cells by exploiting their auto fluorescent profile (green). Starting from the 1st hour, it is possible to see orange spots obtained by the merge between the red signal referred to the RhB and the green background. These spots became more evident 4 and 24 h after incubation. Scale bar 70 μm. (**B**) Quantification of the percentage of the area (pixels) occupied by NPs for every single cell at the different time-points. Data are expressed as mean value ± SD, *n* = 10. (**C**) Quantification of the viability of the cells of CaCo2 after exposure to RhB (orange line), PLGA-PEG-RhB-NPs (yellow line), and Benserazide as inner control (purple line) measured by RealTime-Glo™ MT Cell Viability Assay (Promega kit) 4 and 24 h after incubation The values obtained from non-treated cells were normalized to 100 for each time point. Values are expressed as mean ± SD (*n* = 6). The Student's *t*-test was used to compare the levels among the two groups for each time point. \*\*\* *p* < 0.0001 compared to NT. (**D**) Higher magnification pictures showing the same field of view achieved 24 h after PLGA-PEG-RhB-NPs incubation. In the left panel, the cell nucleus, in the middle panel, the red spots associated with NPs, and in the right panel, the merge between the three channels. The dotted line is the border of the cell. Scale bar 15 μm.

#### **4. Conclusions**

The current study sought to evaluate the effect of the nanocarrier on the transport of a drug and its behavior within the gastrointestinal tract and absorption to the bloodstream. It is important to note, however, that although RhB was originally used as a tracer to visualize NPs. The results obtained by comparing the biodistribution in mice of the free fluorophore and linked to the NPs is of further relevance for future developments. It is, in fact, interesting how PLGA-PEG-NPs manage to preserve the gastroduodenal absorption by using chemically different molecules along with different loading approaches to the nanocarrier. The low systemic exposure and, at the same time, equivalent drug concentration at the intestinal level, with even a trend to increase in the colon 4 h after the treatment, could have a significant positive outcome on the safety of a wide range of drugs targeting inflammatory and neoplastic diseases.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/1999-4923/11/12/658/s1, Figure S1: STEM image of PTX-PLGA-PEG-RhB-NPs, Figure S2: Release kinetics of PTX loaded NPs in PBS containing 0.2 *v*/*v*% Tween 80. PTX-PLGA-PEG-RhB-NPs equivalent to 8.75 μg of PTX were suspended in 1 mL of release medium and incubated at 37 ◦C with constant agitation. At predetermined time points (0.5, 1, 2, 4, 6 and 24 h), the NPs were separated by centrifugation and the supernatant analyzed by HPLC. The results are the mean value ± SD (*n* = 4).

**Author Contributions:** L.M. (Lucia Morelli), L.S., M.G., B.C. performed nanoparticles synthesis, functionalization, and characterization; S.G., M.S. (Marta Sevieri), L.T., L.R., M.B.V. performed in vivo experiments; L.M. (Lavinia Morosi) performed pharmacokinetic analysis; L.P., A.F. did the data curation; D.P., M.S. (Mario Salmona) gave critical comments on the manuscript; D.P., M.C., P.B., M.Z. designed the research and provided funding; P.B., M.C. wrote the manuscript; all authors revised the manuscript critically.

**Funding:** This research was funded by Academic Funding Unimib 2018 to M.C.; the Italian Ministry of University and Research (MIUR) through grant "Dipartimenti di Eccellenza-2017" to University of Milano Bicocca, Department of Biotechnology and Biosciences, Direzione Generale Ricerca, Innovazione, Università, Export e Internazionalizzazione of the Regione Lombardia.

**Acknowledgments:** We thank Tiziano Catalani for the STEM image.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### **Risperidone-Loaded PLGA–Lipid Particles with Improved Release Kinetics: Manufacturing and Detailed Characterization by Electron Microscopy and Nano-CT**

**Christopher Janich 1, Andrea Friedmann 2, Juliana Martins de Souza e Silva 3, Cristine Santos de Oliveira 3, Ligia E. de Souza 1, Dan Rujescu 4, Christian Hildebrandt 5, Moritz Beck-Broichsitter 5, Christian E. H. Schmelzer <sup>2</sup> and Karsten Mäder 1,\***


Received: 31 October 2019; Accepted: 2 December 2019; Published: 9 December 2019

**Abstract:** For parenteral controlled drug release, the desired zero order release profile with no lag time is often difficult to achieve. To overcome the undesired lag time of the current commercial risperidone controlled release formulation, we developed PLGA–lipid microcapsules (MCs) and PLGA–lipid microgels (MGs). The lipid phase was composed of middle chain triglycerides (MCT) or isopropylmyristate (IPM). Hydroxystearic acid was used as an oleogelator. The three-dimensional inner structure of Risperidone-loaded MCs and MGs was assessed by using the invasive method of electron microscopy with focused ion beam cutting (FIB-SEM) and the noninvasive method of high-resolution nanoscale X-ray computed tomography (nano-CT). FIB-SEM and nano-CT measurements revealed the presence of highly dispersed spherical structures around two micrometres in size. Drug release kinetics did strongly depend on the used lipid phase and the presence or absence of hydroxystearic acid. We achieved a nearly zero order release without a lag time over 60 days with the MC-MCT formulation. In conclusion, the developed lipid-PLGA microparticles are attractive alternatives to pure PLGA-based particles. The advantages include improved release profiles, which can be easily tuned by the lipid composition.

**Keywords:** controlled release; PLGA; risperidone; microparticles; microcapsules; oleogels; electron microscopy; three-dimensional X-ray imaging; nano-CT; biodegradable polymers; hydroxy-stearic acid

#### **1. Introduction**

The use of parenteral controlled release drug delivery systems (CR-DDS) is a highly attractive and rational way to improve drug therapy. Potential benefits include decrease of administration frequency, increase of bioavailability and less side effects. Polylactic acid (PLA) and poly(lactic-*co*-glycolic acid

(PLGA) products have existed for several decades and are, by far, the most widely used polymers for CR-DDS [1,2]. The major clinical use is the peptide-based treatment of GnrH-dependent diseases such as breast and prostate cancer. Other important and increasingly used applications include the local delivery of drugs to the eye [3], the brain [4], or the inner ear [5,6]. An important application of CR-DDS is also the controlled release of antipsychotic compounds to treat schizophrenia. Long-acting parenteral controlled release of antipsychotic drugs overcomes the poor compliance of patients with oral administration [7,8]. Risperidone is an important drug in the treatment of schizophrenia. Several oral and one parenteral controlled release formulations are available and a PLGA based CR-DDS formulation is marketed (Risperdal® Consta®). According to the treatment schedule [9], patients initially have to take tablets, although the CR-DDS has already been injected. This situation is not a favorable scenario of pharmacotherapy, because schizophrenia patients do frequently not comply and double medication should be avoided. The release mechanisms of Risperdal® Consta® have been studied in detail by the group of Burgess and good in vitro and in vivo correlations have been observed by means of scaling factors [10,11]. Initially, a lag phase with almost no release is observed, followed by a faster release. The group clearly showed that the release onset is related to the water penetration and polymer degradation-induced drop of the polymer glass transition temperature *T*<sup>g</sup> to 37 ◦C.

Several publications describe alternative PLGA formulations for the controlled release of risperidone. However, commonly, the presence of lag times, sigmoidal release curves, and rather short release times (<1 month) have been observed [10–15]. Souza et al. tested different formulations with single or repeated doses in vivo [16]. For single dose injections, peak maxima were observed after around one week. Unfortunately, no in vitro release data were published in the paper [16]. Therefore, it is clearly still desirable to develop biodegradable DDS with better risperidone release kinetics. The CR-DDS should provide a constant release over the 1–2 months without a lag time and the need for initial oral comedication. It should also be cost effective. Unfortunately, the selection possibility for biodegradable materials for parenteral controlled release is rather limited. In addition to PLA and PLGA, there are few other polymers and lipids. Liquid lipids are used clinically, but they provide only release control for a short period of few days. Solid lipids are also attractive as parenteral depot systems [17], but their use is often linked with problems of polymorphic transitions and limited enzymatic degradability due to the crystalline structure of the lipid. The use of polymer–lipid hybrid structures—compared to pure polymer or pure lipid-based DDS—offers additional opportunities. For example, increased drug loads may be achieved by the lipid component and the release rate might be controlled by the polymer shell. The incorporation of liquid oil into a microparticle might compromise the mechanical stability of the particle and could potentially lead to oily leakage and loss of release control. The incorporation of solid lipids will overcome this problem, but this advantage is at the cost of drug solubility in the crystalline lipid matrix and the problem of polymorphic transitions. It is desirable to combine the advantages of a liquid lipid phase with high mechanical stability. Therefore, we decided to investigate oleogels as possible alternatives. A suitable oleogelator for parenteral applications is 12-hydroxystearic acid (HS). Hydroxystearic acid induces gel formation in lipids due to its precipitation as a HS-fiber network [18,19]. Preclinical studies on injectable oleogels demonstrate the good biocompatibility and degradability of HS-oleogels in vivo [20,21].

We selected PLGA Expansorb® 75-7E from Merck KGaA (Darmstadt, Germany), which has the following properties: lactide/glycolide (L/G) ratio of 75/25, a molecular weight (MW) 80–115 kDa, and the end groups are esterified. The physical chemical properties of Expansorb®75-7E were similar to the PLGA in the market product Risperdal Consta® which contains a lactide/glycolide (L/G) ratio of 78/22, a molecular weight (MW) ~111 kDa and the end groups are esterified [22]. As the oil phase, we incorporated isopropyl myristate (IPM) or middle chain triglycerides (MCT) to form a microcapsule (MC) structure. Oily components with saturated fatty acids were used to avoid the danger of oxidation problems connected with unsaturated fatty acids.

The properties and the release kinetics of the micro-DDS will depend on the general structure and the physicochemical state of the drug and the excipients. Possible structures of (drug-free) micro-DDS are shown in Figure 1.

**Figure 1.** Possible structures of poly(lactic-*co*-glycolic acid (PLGA) and PLGA/lipid microparticles: I: PLGA microparticle; II: microcapsule with core–shell structure; III: microcapsule with multiple oily droplets; IV: microgel with core–shell structure; microgel with multiple gelled lipid domains.

PLGA microparticles are expected to form a homogeneous polymer matrix (structure I). Microcapsules composed of PLGA and oil could show structures with a single oily core and a PLGA shell (structure II) or multiple oily droplets encapsulated in PLGA (structure III). The addition of the oil gelator will increase the viscosity, but again, the oily phase could exist as a single core (structure IV) or as multiple cores (structure V). The drug could be dissolved in the polymer and the oil and/or exist as an own solid (crystalline or amorphous) phase. A rational development of the DDS requires an appropriate characterization. For particle sizing of micro-DDS, static light scattering is the standard method. We decided to use X-ray diffraction for the detection of crystalline structures in our DDS. In addition, we wanted to explore the microstructure inside the DDS. For this purpose, we used two complementary imaging techniques, which were focus ion beam scanning electron microscopy (FIB-SEM) and Zernike phase-contrast X-ray nano-computed tomography (nano-CT). By means of focused ion-beam preparation, we could cut the particles in situ and we were able to visualize the morphology of the microstructure at the location of the cut. In addition, we performed nano-CT experiments. Due to the high penetration depth of X-rays, nano-CT provided high-resolution three-dimensional images of the DDS and we were able to visualize their internal structures at any virtual slice without the need for cutting or slicing them.

In summary, the main goals of this work were:


#### **2. Materials and Methods**

#### *2.1. Materials*

Expansorb DLG 75-7E, batch number C100010962, (Merck, Darmstadt, Germany) PVA 4-88 (Merck, Darmstadt, Germany) Risperidone 95+% (Activate Scientific, Shanghai, China), Pioneer MCT (middle chain triglycerides) batch number: 1132229 (Hansen & Rosenthal, Hamburg, Germany); isopropylmyristate Ph-Eur. batch number: 1407015-01, (Euro OTC Pharma GmbH, Bönen, Germany); dichloromethane 99.9% (Carl-Roth, Karlsruhe, Germany); hydroxystearic acid was a gift from Alberdingk Boley GmbH, (Krefeld, Germany). Risperdal Consta® 50 mg batch number: GCSK002, (Janssen-Cilag GmbH, Neuss, Germany); 2.5 mL GASTIGHT® syringe (Hamilton Germany GmbH, Planegg, Martinsried, Germany), syringe pump (Pump 11 Elite, Harvard Apparatus, Holliston, MA, USA).

#### *2.2. Methods*

#### 2.2.1. Preparation of PLGA Microcapsules (MC) and PLGA Microgels (MG)

Risperidone-loaded microcapsules were prepared using an oil-in-water (o/w) emulsion solvent extraction/evaporation technique.

For this purpose, 100 mg PLGA 75-7E, 100 mg of the oil phase (MCT or IPM), and 50 mg of Risperidone were weighted into a 4 mL glass vial and dissolved into the final volume of 2 mL dichloromethane by vortexing. The organic solutions were drawn up into a 2.5 mL GASTIGHT® syringe (Hamilton Germany GmbH, Planegg, Martinsried, Germany) and injected with a flow rate 5.2 μL/min by a syringe pump (Pump 11 Elite, Harvard Apparatus, Holliston, MA, USA) directly into 300 mL of aqueous PVA 4-88 (Merck, Darmstadt, Germany) solution 0.25% (*w*/*v*) stirred with a Rotating Paddle, USP at 240 rpm at room temperature. The aqueous phase contains double distilled and filtrated water with 0.25% polyvinyl alcohol PVA 4-88 (Merck, Darmstadt, Germany). When particle formation was completed, the suspensions were transferred into a 500 mL round bottom flask and the remaining DCM were removed under vacuum 40 mbar for 20 min. Thereafter, sedimented microcapsules were transferred by a glass pipette into a 15 mL tube and the supernatants were removed carefully. Then, the microcapsules were washed thrice with sterile filtrated double distilled water, filled up to the final volume of 1 mL and lyophilized.

#### 2.2.2. Lyophilization

For freeze-drying, the samples were frozen rapidly with liquid nitrogen at −196 ◦C and lyophilised on a Christ Alpha 2–4 freeze dryer (Martin Christ Gefriertrocknungsanlagen GmbH, Osterode am Harz, Germany) in combination with a Vacuubrand RC 6 vacuum pump (Vacuubrand GmbH, Wertheim, Germany) for 24 h. The chamber was evacuated to 0.05 mbar, corresponding to −48 ◦C on the sublimation curve of ice. The prepared microcapsules were stored at −20 ◦C until further use.

#### 2.2.3. Static Light Scattering Measurements

Routine measurements of the size distribution were carried out by laser diffractometry on a Mastersizer 2000 (Malvern Instruments, Malvern, United Kingdom) combined with a Hydro 2000 S wet dispersion unit. The samples were measured at a laser obscuration of 5%, corresponding to 11%–15% obscuration of the blue laser, in purified water. A series of five runs was evaluated by the Mastersizer 2000 software version 5.60, using the Mie theory, assuming spherical particles with a refractive index of 1.44, absorption of 0.001, and a refractive index of 1.33 for the dispersant as optical properties.

#### 2.2.4. Powder XRD

XRD analysis was performed on a Bruker D8-Advance diffractometer, equipped with a one-dimensional silicon strip detector (LynxEye) operating with Cu Kα radiation. Diffraction was measured from 2θ=5◦ to 70◦ with a step size of 0.01◦ and a counting time of 1 s.

#### 2.2.5. High-Performance Liquid Chromatography (HPLC)

A HPLC Agilent 1100 Series (Agilent, Santa Clara, CA, USA) was used. Separations were carried out using a BDS-Hypersil® C18 5 <sup>μ</sup>m 250 <sup>×</sup> 4.6 column (Thermo Fisher Scientific, Waltham, MA, USA) and a mobile phase of 75% (*v*/*v*) methanol/25% (*v*/*v*) water, 1% (*v*/*v*) tetramethylammoniumhydroxid pentahydrate solution 25% *w*/*v*. Runs were carried out at 0.5 mL/min over 16 min and the absorption at 280 nm was recorded. The retention time of risperidone was 7.5 min. Injection volumes of 20 μL were used for determination of drug loading.

#### 2.2.6. Drug Loading and Encapsulation Efficiency (EE)

Risperidone-loaded microcapsules were weighed and diluted in methanol to get a final concentration of 0.5mg/mL. All the samples were sonicated for 10 min to accelerate the dissolving of MC. Thereafter all samples were diluted 0.05 mg/mL and filtered (Minisart SRP 4, 0.45 μm syringe filter, Sartorius Stedim Biotech GmbH, Göttingen, Deutschland). The concentration of risperidone was determined via HPLC. Drug loading and EE were calculated as following equations from literature: Drug loading = (weight of drug entrapped/weight of MC used) × 100%; EE = (experimental drug loading/theoretically drug) × 100% [23].

#### 2.2.7. In Vitro Drug Release Assay

The in vitro release of risperidone from MCs was carried out in a heated bath shaker (Memmert GmbH + Co. KG, Schwabach, Germany) at 37 ◦C and constant shaking at 50 rpm. A total of 6 mg of formulation I-IV and 3 mg of Risperdal Consta® were suspended in 1 mL of 10 mmol phosphate buffer saline (PBS) at pH 7.4 and transferred into a Spectra Por Float-A\_Lyzer ® G2 1000kD. The float-A Lyzer was placed into a 100 mL measuring cylinder and filled up with PBS to the final volume of 60 mL so that the float-A Lyzer is submerged in PBS. To maintain sink conditions, the buffer solution was drawn off half (half change) when the risperidone concentration reaches 7%–8% of the solubility in PBS. Drug quantity was determined by UV-1800 spectrophotometer (Shimadzu, Kyoto, Japan). All measurements were performed in triplicate and the measurement points were represented as the mean ± SD.

#### 2.2.8. Scanning Electron Microscopy and Focused Ion Beam Preparation

High resolution investigations of the microparticle morphology were done by scanning electron microscopy (SEM) using a Quanta 3D FEG (field emission gun) instrument from FEI Company, (Hillsboro, OH, USA). The microparticles were carefully placed on SEM sample holders using carbon tape and subsequently coated with a thin layer of platinum using a magnetron sputtering system (HVD, Dresden, Germany) to achieve a conducting surface. The SEM images were obtained under high vacuum conditions using an acceleration voltage of 5 keV, a working distance of 6–10 mm and electron beam currents of 12 pA determined with an Everhart–Thornley detector.

The microscope is designed as a dual beam (electron and ion beam) which gives the possibility to perform FIB preparations in the same device. Here, the gallium ion beam and the electron beam operate independently of each other. The point of coincidence of the two beams is located at a working distance of 10 mm. The angle between both beams is 52◦. To allow vertical cutting with the FIB, the sample was tilted by this angle. Thus observation of the cross-sections with the electron beam was also done at an angle of 52◦. After screening the samples by conventional SEM, the particle of interest was selected for cross-sectional preparation (target preparation). As the first step in cross-section preparation at predefined areas of interest coarse material was ablated with gallium ions accelerated at 30 kV with ion currents of 15 nA. After this coarse milling step lower beam currents of Afterwards, 1–0.5 nA were used to polish the cross sections. The samples were not cooled during FIB milling. The patterning conditions used conform to the standard patterning conditions for silicon materials. The SEM observation can be done during ion milling or after subsequent milling steps.

#### 2.2.9. Nano-CT

X-ray imaging experiments were performed at the Institute of Physics, Martin Luther University Halle-Wittenberg, in a Carl Zeiss Xradia 810 Ultra (Cr source, 5.4 keV, for instrument details see [24]). Phase-contrast imaging mode was used with a phase-ring positioned near the back focal plane of the zone plate. In each experiment, a total of 901 projections with a field-of-view (FOV) of 65 μm were obtained over 180◦, with an exposure time of 15 s per projection, a detector binning of 2, and voxel size of 128 nm. Image reconstruction was performed by filtered back-projection algorithm using the software integrated into the Xradia 810 Ultra. For each sample, two or three imaging

experiments were performed, moving the sample vertically to enable the visualization of one entire microsphere in one single direction (from now on named Z-direction). For each sample's datasets, the brightness and contrast were adjusted in ImageJ® and the datasets were then stitched together using the Merge module of the Thermo Fisher Scientific Avizo® software (version 9.2, Thermo Fisher Scientific, Hillsboro, OR, USA). Median denoising and nonlocal mean filter were applied to the datasets. The 3D renderings presented here were created using either the commercial software arivis Vision4D® (version 2.12.6, arivis AG, Rostock, Germany) or Avizo®. For the dispersed internal spherical structures (ISS) evaluation, a volume located in the center of the 3D image was extracted consisting of 325 × 325 × Z voxels. Z varies among the samples, as it is almost equal to the number of pixels of the microsphere diameter, but is smaller than it to prevent voxels that do not consist of sample to be added to the calculation. The results obtained here represent the ratio of ISS compared to the total extracted volume of the sample (PLGA+ISS). For binarization of the datasets into PLGA and ISS, automatic thresholding with IsoData criterion was used, followed by island removal for the removal of small clusters of voxels (<15). A surface was generated from the segmented dataset in order to smooth sharp voxel edges and the specific surface area (SSA) values obtained represent the ratio of the constructed surface area (SA) to the volume (V) for each sample.

#### **3. Results and Discussion**

#### *3.1. Preparation of MCs and MGs*

For the preparation of MCs and MGs, the conventional oil-in-water (o/w) emulsion solvent evaporation method was used [2,25]. All components were dissolved in dichloromethane. The components of the four different formulations are listed in the Table 1. In general, the composition followed a 2:2:1 mass ratio of PLGA/Lipid/Risperidone. Therefore, the theoretical drug load was 20% [m/m]. High encapsulation efficiencies (>92%) were achieved, as seen in Table 1.


**Table 1.** Composition and encapsulation efficiency of Risperidone-loaded PLGA microcapsules (MC) and microgels (MG).

MC = microcapsule, MG = microgel, PLGA = poly(lactide-*co*-glycolide), MCT = middle chain triglycerides, IPM = isopropylmyristate, HS = hydroxystearic acid.

To get reproducible particle sizes, it was important to keep all size-influencing parameters constant, e.g., the o-phase volume and solvent, the concentration and type of polymer, the volume of the continuous phase and the type and concentration of stabilizer, the temperature, and the stirring speed [26]. After lyophilization the MCs did not agglomerate and were resuspendable in water.

#### *3.2. Characterization*

#### 3.2.1. Particle Size Measurements

The particle size and morphology of MC are key factors which potentially affect EE, drug release rate, and the biodistribution [27,28]. The optimal size range for microparticles with controlled release characteristics is considered to be 10–200 μm. Particles in the lower micron range will show a faster drug release, and they are also more easily phagocytosed by immune cells [29]. Sizes larger than 200 μm are difficult to inject and will show more inhomogeneous degradation [30]. The particle size distribution is shown in Figure 2.

**Figure 2.** Volume-averaged particle sizes of microcapsules, microgels, and the reference product.

The particle size of the microcapsule formulations MC-IPM and MC-MCT are very similar (D(0.5) = 172.8 μm and 169.5 μm). The particle sizes for the corresponding microgels MG-IPM and MG-MCT were smaller compared to the microcapsules, but again very similar between IPM and MCT (D(0.5) = MG-IPM 157.2 μm; MG-MCT 155.9 μm), as seen Figure 2 and Table 2. It can be concluded, therefore, that the incorporation of hydroxystearic acid leads to slightly smaller particles, probably due to the impact on the surface tension during particle formation. The determined D(0.5) of the market product is 89.1 μm, which is in line with a size range between 25 and 150 μm reported in the literature [31]. The particle size distribution of the commercial product is considerably broader compared to our formulations (Span Risperdal Consta® 1.165 vs. span values between 0.612 and 0.637), as seen in Table 2.


**Table 2.** Particle size distributions determined by static light scattering.

3.2.2. Scanning Electron Microscopy and Focused Ion Beam Investigations

SEM images of Risperdal Consta® are shown in Figure 3. The market product Risperdal Consta® shows a high heterogeneity of particle size, shape, and surface structures. Surprisingly, structures showing partial engulfment, as seen in Figure 3d, and collapsed structures are visible, as seen in Figure 3c. Larger magnification of the surface indicated the presence of anisotropic material, which are most likely drug crystals, as seen in Figure 3e,f.

**Figure 3.** Electron microscopic pictures of Risperdal Consta®. The pictures at lower magnifications (**a**–**d**) show the presence of different particle sizes and shapes. Most particles show wrinkled surfaces; some have a smooth surface. High magnification shows the presence of small anisotropic material on the surface (**e**,**f**).

The microcapsules and microgels do not show the irregular structures of the commercial product. Most of MCT microcapsules are spherical, some show an elongated shape, which probably results from the coalescence of two particles, as seen in Figure 4a, white stars. The surface of the MC-MCT microcapsules shows a pattern similar to golf balls, as seen in Figure 4b,c. Very rarely, small anisotropic structures are found on the particle surface, as seen in Figure 4c, top middle, which most likely represent risperidone crystals. Cutting of the particles by focused ion beam (FIB) provides insights into the internal structure, as seen in Figure 4b,d. A multicore structure with dimensions in the lower micron range becomes visible. Within the internal structures, anisotropic shapes are visible which most likely present precipitated drug crystals, as seen in Figure 4d.

The electron microscopic pictures of the corresponding microgel MG-MCT shows spherical particles with a structured surface that appears smoother compared to the MC-MCT surface (Figure 5). In contrast to the MC-MCT microcapsules, we could not detect elongated shapes, as indicated by white stars in Figure 4a, which emerge from the coalescence of two particles. The pictures obtained by FIB cutting of the particles show internal multicore structures with dimensions in the lower micron range, as seen in Figure 5b,d.

IPM-loaded microcapsules (MC-IPM) have a spherical shape, as seen in Figure 6a. The surface appears similar to the MC-MCT capsules. Very rarely, small crystals can be detected on the particle surface, as seen in Figure 6c. After FIB cutting, multicore structures in the lower micron range become visible, as seen in Figure 6b,d.

The electron microscopy investigation on MG-IPM microgels shows the presence of spherical particles, as seen in Figure 7a. Higher magnifications of the particle surface show surface dips ("golf ball structure") similar to the MG-MCT microgels. In addition, elongated "hair-like" structures are visible, as seen in Figure 7c,d, which are not detectable in the other samples. Most likely, they are mainly composed of hydroxystearic acid, which is known to induce gel formation by elongated structures. Again, FIB cutting was used to explore the interior of the particles. Multicore microdomains became visible, as seen in Figure 7b, similar to the MG-MCT microgel particles with dimensions in the lower micron range.

**Figure 4.** Electron microscopic pictures of MC-MCT microcapsules, (for composition details see Table 1). Intact particles are presented at lower (**a**,**c**) higher magnification. Focused ion beam (FIB) cut particles are shown at lower magnification in (**b**) and after deeper FIB milling and higher magnification in (**d**). The white stars in (**a**) indicate anisotropic particles, which were most likely formed by coalescence of two particles.

**Figure 5.** Electron microscopic pictures of MG-MCT microgels (for composition see Table 1). Intact particles are presented at lower (**a**,**c**) higher magnification. FIB cut particles are shown at lower magnification in (**b**) and at higher magnification at (**d**).

**Figure 6.** Electron microscopic pictures of MC-IPM microcapsules (composition details are given in Table 1). Intact particles are presented at lower (**a**,**c**) higher magnification. FIB cut particles are shown at lower magnification in (**b**) and at higher magnification at (**d**).

**Figure 7.** Electron microscopic pictures of MG-IPM microgels (sample composition is given in Table 1). Intact particles are presented at lower (**a**,**b**) and higher (**c**,**d**) magnifications. A FIB cut particle is shown in (**b**). Elongated, "worm like" structures (most likely precipitated hydroxyl-stearic acid) are visible in (**c**,**d**).

The crystallinity of the formulations was evaluated by X-ray diffraction. The diffractograms of PLGA, the pure drug risperidone, the commercial product, and the developed microcapsules and microgels are shown in Figure 8. It is known that the drug can exist in different polymorphs [32–34]. As expected, the polymer PLGA does not show sharp peaks, but only broad lines, indicating an amorphous state. The pure drug shows multiple sharp peaks. The X-ray diffractograms of the commercial product Risperdal Consta® and the developed formulations show a similar pattern, but they are different from the pure drug. From the measurements, it can be concluded that risperidone exists in all formulations at least partially in the crystalline state. The crystalline state in the formulation is different from the crystal structure of the parent drug, but very similar to the crystal form of risperidone, which has been observed after micronization with supercritical fluids [33].

**Figure 8.** Wide angle X-ray diffractograms of microcapsules, microgels, Risperdal Consta®, and the pure drug risperidone.

#### **4. X-Ray Nano-Imaging (Nano-CT)**

X-ray imaging comprises a class of nondestructive techniques that are used to visualize the internal structure of different materials. In conventional X-ray imaging, differences in the X-ray absorption cross-section of different materials results in the image contrast [35,36]. Therefore, if the different parts composing one sample show only small differences in the X-ray attenuation, a poor image contrast will be obtained in conventional absorption-based X-ray imaging. This is specially the case for samples composed of elements with low electron density, making the characterization of the structures and visualization of the sample features a very hard task. Alternatively, phase-contrast imaging methods enable the visualization of features of weakly absorbing objects [35,37,38]. X-ray phase-contrast imaging is based on a different physical principle, and the contrast of the image is generated by the detection of the changes in the phase of the wavefront [36,39]. In this work, we imaged the different microspheres using Zernike phase-contrast imaging in a benchtop nano-CT scanner (Xradia Ultra 810). Phase-contrast nano-CT was used to characterize the inner structure of the commercial product Risperdal Consta® (Figure 9) and the developed microcapsules and microgels (Figure 10). Videos Supplementary Videos S1–S5 show the inner 3D structure of the microparticles. The diameter and volume distributions of the internal structures are presented in Supplementary Figures S1 and S2.

**Figure 9.** Volumetric representation the inner part of one reconstructed microsphere of Risperdal Consta® imaged with nano-CT. The left picture shows the X-ray signal in gray and the microsphere PLGA matrix pseudocolored in yellow. On the right, the small colored structures represent the dispersed internal spherical structures (ISS), which can be attributed to the dispersed drug particles. The volume of the ISS is about 13% of the total volume of the particle.

**Figure 10.** Volumetric representation of the inner part of the microspheres imaged with nano-CT. On the top from left to right, the MC-IPM, MG-IPM, MG-MCT, and MC-MCT microspheres pseudocolored in yellow are shown, with the dispersed internal spherical structures (ISS) represented in different colors. On the bottom, the volume used for the estimations done with nano-CT are shown for all samples.

All microgels and microcapsules show internal spherical structures (ISS) in the submicrometer range, with sphericity values close to unit. However, the volume percentage of the ISS in the microcapsules is between 35% and 46% and is, therefore, much higher than the volume percentage of the ISS phase in the Risperdal Consta® particles. The ISS phase can be attributed to the dispersed lipid phase for the microgels and microcapsules. In contrast, the ISS consists of risperidone only for the Risperdal Consta® system. A quantitative treatment of the 3D images indicates that the ISS have diameter around 2 μm for lipid-based systems and around 800 nm for the Risperdal Consta® product, as seen in Table 3. The ISS diameter and volume distributions are shown in Figures S1 and S2.


**Table 3.** Volume, diameter and sphericity of the internal dispersed phase (ISS), assessed by nano-CT.

Compared to other imaging techniques, Zernike phase-contrast nano-CT has the advantage of producing high-resolution three-dimensional images of low-absorbing samples, which enables the estimation of the values shown here. As with most high-resolution imaging techniques, nano-CT has the disadvantage of producing images limited to a small field of view, thus, to obtain the images of one single microparticle, it was necessary to image the sample two or three times and stitch the datasets. One of the benefits of phase-contrast nano-CT for imaging low-absorbing samples is that it does not require any special sample preparation procedure or induce any sample damage within the imaged area, in contrast with FIB-SEM. Overall, the results obtained by FIB-SEM correlate well with those obtained by nano-CT, and the same structures were observed. Both methods indicate the presence of a high percentage of internal structures (ISS) with sizes around two micrometers. Therefore, the microcapsules and microgels represent structures III and V of Figure 1, respectively.

#### **5. Drug Release**

The in vitro release of risperidone from the commercial product in the developed formulations is shown in Figure 11. For all formulations, a low percentage of the drug (<10%) is initially released. Most likely, drug molecules on the surface (which were observed by electron microscopy) are released very quickly. After the initial release, there is almost no release detectable for two weeks for Risperdal Consta®. Thereafter, the release accelerates. The developed formulations show different release profiles. The fastest release was observed for the MG-IPM microgel. The slowest release rate observed for the lipid formulations was for the MC-IPM microcapsules system. Therefore, it can be concluded that the oleogelator hydroxystearic acid had a strong impact on the release kinetics. The impact of the oleogelator hydroxystearic acid is also visible for the MCT systems, but to a smaller extent. Again, the microparticles with the incorporated oleogel (MG-MCT) release faster compared to the HS-free microcapsules (MC-MCT). The desired release profile was obtained with the MC-MCT microparticles: an almost linear release profile over two months with no lag time has been achieved.

**Figure 11.** In vitro release profiles of Risperidone microcapsules (MC), microgels (MG), and Risperdal Consta®.

#### **6. Conclusions**

The results of our study show that lipid–PLGA hybrid systems are interesting alternatives to pure PLGA microparticles. The saturated liquid lipids MCT and IPM were incorporated in the form of microcapsules and microgels. Hydroxystearic acid (HS) was used as an oleogelator. The particle structure was investigated by FIB-SEM and nano-CT imaging. Both methods support the existence of a highly dispersed internal phase with spherical particle sizes of around 2 micrometers. The noninvasive nano-CT imaging enabled the production of 3D representations of the microparticles' structures in high-resolution images. After image processing and segmentation, it was possible to estimate the amount, size distribution, and sphericity of the internal structures. The presence of hydroxystearic acid accelerated the drug release. The drug release kinetics can be tuned by the lipid composition. An almost ideal release profile for risperidone was achieved with PLGA–MCT microcapsules (MC-MCT). In contrast to Risperdal Consta®, no lag time and an almost linear release over 60 days was observed. Therefore, the developed systems are suitable candidates to overcome the need for double medication (injected microparticles and tablets), which is currently needed due to the release lag time of Risperdal Consta®.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/1999-4923/11/12/665/s1, Figure S1: Diameter distribution of the internal spherical structures (ISS) determined by nano-CT. Figure S2: Volume distribution of the internal structures (ISS) determined by nano-CT. Video S1: Video of the internal structure obtained by nano-CT for Risperdal Consta. Video S2: Video of the internal structure obtained by nano-CT for MC-IPM. Video S3: Video of the internal structure obtained by nano-CT for MG-IPM. Video S4: Video of the internal structure obtained by nano-CT for MG-MCT. Video S5: Video of the internal structure obtained by nano-CT for MC-MCT.

**Author Contributions:** Conceptualization, K.M., C.J., L.E.d.S., D.R., C.H., M.B.-B.; Particle manufacturing and drug release studies: C.J., L.E.d.S., K.M. Nano-CT measurements and data evaluation: J.M.d.S.eS. and C.S.d.O. F.I.B.-SEM: A.F., C.S.d.O. writing—original draft preparation, C.J., K.M.; writing—review and editing, C.J., A.F., J.M.d.S.eS., C.S.d.O., D.R., C.H., M.B.-B., C.H., C.E.H.S., supervision: K.M.

**Funding:** JMSS thanks DFG (WE 4051/21-1). This study was supported by funds from MilliporeSigma a Business of Merck KGaA, Darmstadt, Germany. CEHS acknowledges funding from the Fraunhofer Internal Program under Grant No. Attract 069-608203. The funding sponsors had no role in the design, execution, interpretation, or writing of the study.

**Acknowledgments:** The authors gratefully acknowledge Toni Buttler (Institute of Chemistry, Martin Luther University Halle-Wittenberg, Germany) for carrying out the powder-XRD measurements and Carl Zeiss for the loan of the phase-ring.

**Conflicts of Interest:** C.H. and M.B.-B. are employees of MilliporeSigma a Business of Merck KGaA, Darmstadt, Germany.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Dry Tablet Formulation of PLGA Nanoparticles with a Preocular Applicator for Topical Drug Delivery to the Eye**

### **Woo Mi Ryu 1,**†**, Se-Na Kim 2,**†**, Chang Hee Min <sup>1</sup> and Young Bin Choy 1,2,3,\***


Received: 25 October 2019; Accepted: 2 December 2019; Published: 4 December 2019

**Abstract:** To enhance ocular drug bioavailability, a rapidly dissolving dry tablet containing alginate and drug-loaded poly(lactic-*co*-glycolic acid) (PLGA) nanoparticles was proposed. For hygienic and easy administration of an accurate drug-dose with this tablet, the use of a preocular applicator was suggested. Herein, a dry tablet was prepared by embedding dexamethasone-loaded PLGA nanoparticles in alginate, which was deposited on the tip of the applicator. The nanoparticles were loaded with 85.45 μg/mg drug and exhibited sustained drug release for 10 h. To evaluate in vivo efficacy, dexamethasone concentration in the aqueous humor was measured after topical administration of the dry tablet, with the applicator, to rabbit eyes and was compared to that achieved with Maxidex®, a commercially-available dexamethasone eye drops. When applied with the preocular applicator, the dry tablet containing alginate could be fully detached and delivered to the eye surface. In fact, it showed up to 2 h of nanoparticle retention on the preocular surface due to tear viscosity enhancement, causing an estimated 2.6-fold increase in ocular drug bioavailability compared to Maxidex®. Therefore, the preocular applicator combined with a dry alginate tablet containing PLGA nanoparticles can be a promising system for aseptically delivering an accurate dose of ophthalmic drug with enhanced bioavailability.

**Keywords:** PLGA; nanoparticles; alginate; ophthalmic drug delivery; dexamethasone

#### **1. Introduction**

Topical drug administration is an easy route for ocular drug delivery. However, an ophthalmic drug, prepared as a solution and suspension, is known to clear very rapidly from the eye surface (<3 min), leading to a short preocular residence time, ultimately limiting drug bioavailability [1,2]. Less than 5% of a topically-applied drug can reach the interior target tissue of the eye [3,4]. Micro- or nano-particles have thus been suggested as ophthalmic drug carriers to achieve prolonged preocular residence time as well as sustained drug release [5]. Nevertheless, when formulated in the aqueous suspension, an additional fluid added to the eye would accelerate tear clearance, which would also expedite the drainage of the drug-loaded particles [6]. Therefore, an alternative strategy has been suggested and it includes a dry tablet formulation, where the drug-loaded particles are embedded in a tablet medium of a rapidly-dissolving polymer [7]. When applied to the eye surface, the tablet medium would dissolve in the tear fluid to free the drug-loaded particles on the preocular surface. The dissolution of the tablet medium would thus increase tear viscosity to delay tear clearance, which

in turn would improve the preocular retention of drug-loaded particles. However, applying a dry tablet to the sensitive eye may not be convenient for patients. In addition, this administration route might not be hygienic, as the tablet would be retrieved with the bare fingers [8,9].

Herein, a combined system of a preocular applicator and a rapidly-dissolving dry tablet containing drug-loaded nanoparticles was proposed. The applicator was designed to contain two compartments (i.e., a handle plus a tip to hold the dry tablet). Thus, for administration, a dry tablet on the tip could approach and touch the eye surface while the handle is held by one's fingers. The tablet would then be separated from the tip and its medium would be dissolved in the tear fluid to free the drug-loaded nanoparticles. This would provide an easy way to topically administer the dry tablet and importantly, allow tablets to remain aseptic during administration.

To test the system and strategy, a preocular applicator using the biocompatible polydimethylsiloxane (PDMS) was prepared in this study. The shape and dimension of the applicator herein was designed to be similar to that of the commercially-available, one-time use applicator of artificial tear fluid based on patients' familiarity [10]. The commercially-available applicator is basically a container of liquid; however, a dry tablet formulation was loaded at the tip of the applicator proposed in this work. For the formulation, poly(lactic-*co*-glycolic acid) (PLGA) nanoparticles were prepared to be loaded with dexamethasone, a widely used corticosteroid drug for the treatment of eye inflammation [11,12], to achieve sustained drug release. The nanoparticles were then suspended in a solution containing a mixture of water-soluble polymers, polyvinyl alcohol (PVA), and alginate, which was then freeze-dried on top of the applicator tip to produce a rapidly-dissolving dry tablet. PLGA has been widely used for sustained drug delivery due to its high biocompatibility and degradability to non-toxic by-products [13,14]. PVA has already been approved for clinical use in various ophthalmic applications [15,16]. In the present study, alginate was also employed in the tablet medium as a viscosity enhancer of tear fluids. Alginate is a highly-biocompatible polymer [17] that can form a gel in the presence of the multivalent cation, Ca2<sup>+</sup>, which is abundant in the tear fluid [18,19]. Due to this unique property, when dissolved in the tear fluid, alginate can further increase tear viscosity, thereby synergistically improving the preocular retention of the drug-loaded nanoparticles.

Dexamethasone-loaded PLGA nanoparticles (DX/NP) were prepared by the solid-in-oil-in-water (S/O/W) emulsion method and characterized by scanning electron microscopy (SEM) and dynamic light scattering (DLS), which were employed to assess nanoparticle size and morphology. The cytotoxicity of DX/NP was examined using human corneal epithelial cells (HCECs). To examine the effect of alginate on tear viscosity enhancement, two distinct tablet formulations containing DX/NP were prepared with media containing both PVA and alginate (DX/NP AL\_TAB) and PVA alone (i.e., without alginate) (DX/NP TAB), respectively. For in vivo experiments, these tablet formulations were applied to rabbit eyes, and drug concentration in the aqueous humor was assessed and compared to the commercially-available ophthalmic dexamethasone medication, Maxidex®, (i.e., an aqueous suspension of dexamethasone itself).

#### **2. Materials and Methods**

#### *2.1. Materials*

PLGA (lactic acid:glycolic acid = 50:50; i.v. = 48,000) was purchased from Evonik Industry (Essen, Germany). Dexamethasone and Nile red were purchased from Tokyo Chemical Industry (Tokyo, Japan). PVA (87–89% hydrolyzed), phosphate buffered saline (PBS) tablets, tween 80, trifluoroacetic acid (TFA; >99%), and calcium chloride dihydrate (>99%) were obtained from Merck (St. Louis, MO, USA). Dichloromethane (DCM; >99.5%), *N*,*N*-dimethylformamide (DMF; >99.5%), and acetone (>99.5%) were supplied by DaeJung (Siheung-si, Korea). Acetonitrile (ACN; >99.9%) was purchased from J.T. Bakers (Phillisburg, NJ, USA). PDMS (Sylgard 184) was obtained from Sewang Hitech Silicone (Bucheon-si, Korea). Alcaine® (0.5% proparacaine hydrochloride ophthalmic solution) and Maxidex® (0.1% dexamethasone ophthalmic suspension) were purchased from Alcon-Couvreur (Fort Worth, TX, USA).

#### *2.2. Preparation of Drug-Loaded Nanoparticles*

DX/NP was prepared via S/O/W emulsification [20]. Briefly, 300 mg PLGA and 100 mg dexamethasone were dissolved in 7 mL DCM. The resulting solution was then added to 8 mL of 1% *w*/*v* PVA solution, which was emulsified with a sonicator (Sonic Dismembrator Model 500, Fisher Scientific, Illkirch-Graffenstaden, France) at 160 W for 10 min. The emulsion was transferred into 100 mL of 1% *w*/*v* PVA solution at room temperature and stirred at 400 rpm under vacuum (−10 psi) for 60 min for solvent evaporation. The final suspension was then centrifuged at 200 rpm for 10 min, where the precipitates were eliminated and only the suspension at the top was collected. The collected suspension was then washed with deionized (DI) water three times via centrifugation at 13,000 rpm for 10 min. Thereafter, the suspension was lyophilized for 20 h to obtain dry DX/NP. PLGA nanoparticles loaded with Nile red (NR/NP) instead of dexamethasone were also prepared to evaluate its in vivo preocular retention properties. As such, 5 mg Nile Red and 500 mg PLGA were dissolved in 7 mL DCM for emulsification.

#### *2.3. Preparation of Tablet-Loaded Preocular Applicators*

A preocular applicator with a dry tablet on the tip was prepared, as depicted in Figure 1. To prepare a preocular applicator, two different molds were first made to fabricate a handle and tip (Figure 1a). In the molds, a mixture of a PDMS base and its curing agent (10:1, *v*/*v*) was poured and cured slightly at 80 ◦C for 2 h. Then, each of the constituent pieces was assembled and cured further at 80 ◦C for another 2 h to allow for their bonding. In this work, two different tablets, with media containing both PVA and alginate, and PVA alone, were produced to yield DX/NP AL\_TAB and DX/NP TAB, respectively. To prepare the DX/NP AL\_TAB, 20 mg/mL DX/NP were suspended in a solution containing 0.1% *w*/*v* PVA and 2% *w*/*v* alginate. Above this alginate concentration, the solution became too viscous to properly distribute the DX/NP. To prepare the DX/NP TAB, 20 mg/mL DX/NP were suspended in a solution of 0.1% *w*/*v* PVA only. Twenty μL of the resulting suspension was then poured in the reservoir that was made by tightly fitting a cylindrical connector with open ends to a tip of the applicator (Figure 1b). The whole piece was then rapidly frozen using liquid nitrogen, which was then lyophilized for 5 h at 0.01 mbar with the collector temperature set at −80 ◦C (FreeZone 6 Dryer system, Labconco, Kansas City, MO, USA) [21–23]. After that, a connector was carefully removed to produce the preocular applicator loaded with the tablet formulation. To evaluate the in vivo preocular retention properties, NR/NP was embedded instead of DX/NP to produce tablets of NR/NP AL\_TAB and NR/NP TAB, prepared under the same condition employed for DX/NP AL\_TAB and DX/NP TAB, respectively.

**Figure 1.** Schematic illustration of the fabrication of the DX/NP tablet-loaded applicator. (**a**) Fabrication of the PDMS applicator. (**b**) Tablet preparation on the tip of the applicator.

#### *2.4. Characterization*

The morphology of DX/NP was examined by SEM (JSM-7800F Prime, JEOL, Tokyo, Japan). Its size distribution was determined using DLS (ELS-2000ZS, Otsuka Electronics, Osaka, Japan) and a zetasizer (Nano ZS, Malvern, UK) with a particle suspension prepared in DI water [24]. To measure the drug loading amount, the solution of dexamethasone that was fully extracted from the DX/NP and tablet was respectively prepared. Briefly, 4 mg of DX/NP was completely dissolved in 4 mL DMF. The tablet at the applicator tip was fully immersed in 1 mL DMF, sonicated for 2 h, and centrifuged at 13,000 rpm. The supernatant was then diluted with ACN in a 1:1 ratio. Drug concentration in each resulting solution was measured using high-performance liquid chromatography (HPLC; Agilent 1260 series, Agilent Technologies, Santa Clara, CA, USA) with a Diamonsil C18 column (5 μm, 150 × 4.6 mm). Column temperature and absorbance wavelength were set at 37 ◦C and 240 nm, respectively. Injection volume and flow rate were 20 μL and 1.5 mL/min, respectively. The mobile phase was comprised of 0.1% TFA and ACN mixed in a 65:35 ratio. Given with the loading amount of the DX/NP, the encapsulation efficiency (EE) was calculated by the equation.

$$\text{EE}\left(\%\right) = \frac{\text{Amount of drug loaded in nanopparticles}}{\text{Initial amount of drug}} \times 100\tag{1}$$

The thicknesses of the tablets herein were measured using a caliper (ABSOLUTE Digimatic Caliper, Mitutoyo, Kanagawa, Japan). To assess the particle distribution in the tablet, the NR/NP TAB and NR/NP AL\_TAB were imaged with a fluorescence microscope (Leica DMI4000 B, Leica Microsystems, Wetzlar, Germany). To examine the effect of the tablet medium, DX/NP AL\_TAB and DX/NP TAB were each immersed in 1 mL of the medium mimicking tear fluid, i.e., PBS containing Ca2<sup>+</sup> (10 mM, pH 7.4, [Ca2+] = 39.4 μg/mL) [25], for 10 min at 37 ◦C. The viscosity of the resulting solution was then measured using a rheometer (Advanced Rheometric Expansion System, Rheometric Scientific, New Castle, DE, USA), where gap separation, temperature, and shear rate were set at 0.8 mm, 25 ◦C, and 100 s<sup>−</sup>1, respectively.

#### *2.5. In Vitro Drug Release Study*

To examine the in vitro drug release profiles, 400 μg DX/NP, two different tablets containing the DX/NP (i.e., DX/NP TAB and DX/NP AL\_TAB) and Maxidex®, all of which contained the same amount of about 35 μg DX were each placed in a dialysis membrane bag (SnakeSkinTM Dialysis Tubing, 10 kDa, Thermo Scientific, USA). The bag was then immersed in 5 mL pH 7.4 PBS containing 39.4 μg/mL Ca2<sup>+</sup> and 0.5% *w*/*v* Tween 80 to meet the sink condition of DX [26]. While being incubated at 37 ◦C, at scheduled times, 1 mL of the release medium was collected and the same volume of fresh PBS was added back. The amount of released dexamethasone was measured using HPLC as described above.

#### *2.6. Cytotoxicity Evaluation*

The in vitro cytotoxicity of DX/NP was evaluated using HCECs (PCS-700-010, ATCC, Manassas, VA, USA). HCECs were grown in a corneal epithelial cell basal medium (PCS-700-030, ATCC, USA) with supplements (PCS-700-040, ATCC) at 37 ◦C in a humidified environment with 5% CO2. Prior to the assay, HCECs were seeded in a 96-well plate at 1 <sup>×</sup> <sup>10</sup><sup>5</sup> cells/well and grown for 24 h. Subsequently, 100 μL of the DX/NP suspension, which was prepared in the cell growth medium at concentrations of 5, 10, 25, 50, 100, 250, 500, and 1000 μg/mL, was added to each well and incubated at 37 ◦C for 24 h. The medium was then completely removed and replaced with 100 μL of fresh medium. Thereafter, 10 μL of an EZ-Cytox solution was added to each well and incubated at 37 ◦C for 2 h under dark conditions. Cell viability was measured using a microplate reader, with absorbance and reference wavelengths of 450 nm and 600 nm, respectively (VersaMax ELISA Microplate Reader; Molecular Devices, San Jose, CA, USA).

#### *2.7. Animal Experiments*

In vivo experiments were conducted with the healthy eyes of male New Zealand White rabbits (weight 2.1–2.5 kg). Rabbits were granted free access to food and water and were housed in a controlled environment: temperature; 21 ± 1 ◦C, humidity; 55 ± 1%, and light/dark cycle; 12 h/12 h. The in vivo experimental protocols were approved by the Institutional Animal Care and Use Committee at Seoul National University Hospital (IACUC No. 19–0133, date: 30 July 2019).

First, the in vivo preocular retention properties of the nanoparticles were evaluated after topical administration to the eye. The preocular retention properties were assessed as reported in our previous studies, with slight modifications [27,28]. Briefly, the NR/NP AL\_TAB or NR/NP TAB on the applicator tip was applied on the lower cul-de-sac of the left eye of rabbits. At scheduled times, the rabbit eye was anesthetized by topical administration of a drop of Alcaine® and the entire preocular surface was thoroughly wiped with a surgical sponge (PVA spear; Sidapharm, Thessaloniki, Greece) to collect the NR/NP. The sponge was then fully immersed in 5 mL DMF and sonicated for 2 h to fully dissolve the NR/NP. The amount of Nile red in the sample was measured using HPLC-mass spectroscopy (LC-MS) with a Polaris 5 C18-A (2.7 μm pore size, 4.6 × 150 mm) and the following conditions: column temperature, 30 ◦C; absorbance wavelength, 243 nm; injection volume, 20 μL; and flow rate, 0.45 mL/min. The mobile phase was prepared by mixing 0.1% formic acid and ACN in the ratio, 55:45. For statistics, four animals (i.e., one left eye for each rabbit) were assigned per time point for each formulation. With the NR/NP AL\_TAB and NR/NP TAB, the in vivo profile of tablet disintegration was also assessed. For this, the eye was imaged using a digital camera (Galaxy S10, Samsung, Seoul, Korea) at 0 and 30 s after topical administration of the tablet to rabbit eyes.

To assess the in vivo ocular drug bioavailability, each of the three formulations (i.e., the DX/NP AL\_TAB and DX/NP TAB in the applicator, and 35 μL Maxidex®) with the same dose of dexamethasone (ca. 35 μg dexamethasone) was administered onto the lower cul-de-sac of rabbit eyes. At scheduled times, the rabbit was anesthetized with a subcutaneous injection of a cocktail containing 20 mg/kg ketamine and 10 mg/kg xylazine. Thereafter, approximately 100 μL of aqueous humor (AH) was collected using a 31 G needle (BD Ultra-Fine II, Becton Dickinson and Company, Franklin Lakes, NJ, USA). Drug concentration in AH was analyzed by LC-MS as described above. For statistics, three animals (i.e., the left eye of each rabbit) were assigned per time point for each formulation.

#### *2.8. Statistical Analysis*

Statistical analysis was performed using the amount of particles remaining on the preocular surface and drug concentration in AH by the Mann–Whitney U-test. A *p*-value < 0.05 was considered to indicate statistical significance (SPSS version 22, IBM, Armonk, NY, USA).

#### **3. Results**

#### *3.1. Characterization of the Formulations*

DX/NP was prepared using the emulsion method. As a result, the drug-loaded nanoparticles were found to display a spherical shape, as shown in Figure 2a. Particle diameter was 336.92 ± 5.56 nm (Figure 2b) while drug-loading amount was 85.45 ± 5.44 μg/mg (i.e., EE = c.a. 25.6%). The size and shape of NR/NP were similar to those of DX/NP (Figure S1 in the Supplementary Information). As shown in Figure 2c, the zeta potential of the blank PLGA nanoparticles was measured to be −3.9 mV, which was shifted to −27.2 mV with the DX/NP due to a negative charge of the encapsulated drug, DX [29]. Figure 3a shows the tablets prepared in the present experiment (i.e., the DX/NP TAB and DX/NP AL\_TAB). They displayed a cylindrical shape and were well attached at the tip applicator. As shown in Figure 3b, the tablets loaded with the NR/NP exhibited a evenly-distributed fluorescent signal throughout the medium, implying a homogenous distribution of the particles in the tablet. As the same amount of DX/NP was embedded during tablet preparation, the drug loading amount per tablet was similar between DX/NP TAB and DX/NP AL\_TAB (34.11 ± 0.48 and 34.89 ± 0.28 μg, respectively). For the same reason, the thicknesses of the DX/NP TAB and DX/NP AL\_TAB were also similar, which were measured to be 1.02 ± 0.03 and 1.02 ± 0.06 mm, respectively. When tablets were dissolved in Ca2+-containing PBS, the solution with DX/NP AL\_TAB (i.e., the tablet containing 400 μg alginate) had a higher viscosity of 0.93 Pa s than that with DX/NP TAB (i.e., the tablet without alginate) of 0.01 Pa s. However, both tablets herein were observed to be disintegrated in 30 s by dissolution in tear fluid when topically administered to rabbit eyes (Figure 3c).

**Figure 2.** Characterization of DX/NP. (**a**) Representative scanning electron micrographs of DX/NP. (**b**) Particle size distribution of DX/NP measured by the dynamic light scattering (DLS) method; Polydispersity index = 0.056. (**c**) Zeta potentials of the blank PLGA nanoparticles (-) and DX/NP ().

**Figure 3.** Characterizations of tablets. (**a**) Representative optical images of the applicator with the DX/NP TAB and DX/NP AL\_TAB. White arrows indicate the dry tablets prepared on the applicator tip. (**b**) Representative fluorescence images of the NR/NP TAB and NR/NP AL\_TAB. (**c**) Disintegration profiles of the NR/NP TAB and NR/NP AL\_TAB administered topically to rabbit eyes.

#### *3.2. In Vitro Evaluation Results*

Figure 4 shows the in vitro drug release profiles of DX/NP and the tablet formulations, which displayed sustained drug release for up to 10 h. The release profiles did not differ among the formulations as the tablet medium dissolved rapidly to achieve almost instantaneous freeing of DX/NP. Drug release was slightly more suppressed with DX/NP AL\_TAB, which could be due to a slight increase in the viscosity of the release medium in a dialysis membrane bag owing to the interaction between algnate and Ca2+. On the other hand, Maxidex®, which is basically a suspension of DX, exhibited almost complete dissolution of the drug in 2 h. When tested with HCECs, the DX/NP exhibited a cell viability greater than 90% at all testing concentrations (Figure 5), suggesting that the DX/NP was non-cytotoxic.

**Figure 4.** In vitro drug release profiles of Maxidex® (), DX/NP (), DX/NP TAB (), and DX/NP AL\_TAB (-).

**Figure 5.** In vitro cytotoxicity of DX/NP on human primary corneal epithelial cells.

#### *3.3. In Vivo Experimental Results*

To examine the effect of alginate on preocular retention of the nanoparticles, two different tablets embedded with the Nile red-loaded nanoparticles (i.e., NR/NP TAB and NR/NP AL\_TAB) were employed for in vivo evaluation. As shown in Figure 6, by incorporating alginate in the tablet, the amount of nanoparticles remaining at the preocular surface increased. For the tablet without alginate, the average percent of remaining particles was relatively low (i.e., 4.8 and 0.2% at 15 min and 1 h, respectively). However, owing to the presence of alginate in the tablet, 33 and 12% of particles remained at 15 min and 1 h, respectively, and could still be detected until 2 h. As conventional eye drops are known to completely disappear in 5 min [1], such findings suggest a greater improvement in drug retention at the eye surface. For the tablet with alginate, there was a high variability in particle retention observed at 1 h. Right after the tablet medium dissolved in the tear fluid to interact with Ca2<sup>+</sup> ions, the tear viscosity increased to allow for an evident retention of the NP at the preocular surface. However, as the time progressed, the tear fluid with alginate would be continuously diluted with a newly-generated fresh tear, which would vary greatly depending on the subject. After 2 h, almost all particles would be cleared from the preocular surface.

**Figure 6.** In vivo preocular retention profiles of NR/NP TAB (-) and NR/NP AL\_TAB (-) on rabbit eyes. \* indicated statistical significance between the formulations at each time point (*p* < 0.05).

The pharmacokinetic profiles of the tablet formulations were assessed and compared with that of the commercially-available dexamethasone eye drops, Maxidex®. Figure 7 and Table 1 display the drug concentration profiles in the AH and their pharmacokinetic parameters, respectively. At 30 min, Maxidex® had a Cmax of 285.93 ng/mL; however, this value rapidly decreased to an undetectable level at 6 h after the administration. Interestingly, albeit formulated in a dry tablet, the DX/NP TAB without alginate exhibited a much lower drug bioavailability. In addition, its area under the drug concentration-time curve (AUC) and Cmax were less than a half of those with Maxidex®. Such finding could be due to a lower amount of drug exposure at the eye surface, where the DX/NP disappeared relatively rapidly and only a small portion of drug within the particles was actually released during the early period post-administration. On the other hand, 100% drug would be exposed right after a bolus administration with Maxidex®.

**Figure 7.** Dexamethasone concentration in the AH of rabbit eyes after the administration of Maxidex® (), DX/NP TAB (), and DX/NP AL\_TAB (). \* DX/NP AL\_TAB was significantly different from Maxidex®.

**Table 1.** Pharmacokinetic parameters of dexamethasone in the AH.


**<sup>a</sup>** calculated using the trapezoidal rule.

The highest drug bioavailability was observed with the DX/NP AL\_TAB. In fact, its AUC was 2.5-fold greater than that of Maxidex®. By incorporating alginate in the tablet, it appeared to increase the viscosity of the tear fluid, enabling a larger amount of DX/NP to be retained in the preocular space for a longer period. During this period, the DX/NP would release and expose a significant amount of dexamethasone in a sustained manner. As a result, the drug concentration in tear would gradually increase before being absorbed into the AH to shift the time to reach Cmax (Tmax) to a later point (i.e., 2 h) and further increase the value of Cmax itself, compared to that of Maxidex®. Due to prolonged retention and sustained drug release, drug clearance from the AH was delayed, demonstrating that drug concentration in the AH was statistically significantly higher than that of Maxidex® at 4–10 h (*p* < 0.05). All tablets tested in this experiment were almost instantaneously fully detached and released from the applicator tip to the preocular surface when they were in contact with the lower cul-de-sac of rabbit eyes (Figure S2 in Supplementary Information).

#### **4. Discussion**

Topical drug administration is considered to be an easy route for ophthalmic drug delivery. However, for conventional formulations of drops and suspensions, the additional liquid accelerates tear clearance to further lower drug bioavailability in the eye [30,31]. To overcome this, a dry tablet loaded with microparticles had been proposed as topical drug-delivery formulations [7]. The tablet was prepared by compressing the medium of mannitol together with drug-loaded microparticles composed of PLGA and a mucoadhesion promoter, PEG. Thus, dissolution of mannitol increased the tear viscosity to allow for a prolonged time of interaction between PEG in particles and mucin in the preocular surface. However, the tablet with a high density appeared to dissolve in tear fluid in

minutes, during which a relatively large tablet could cause eye irritation. Therefore, the formulation was upgraded to be able to dissolve rapidly in tear fluid and for this, the dry tablet was prepared by freeze-drying [7]. In this way, the tablet medium could possess a porous structure and thus, after administration, it dissolved almost instantaneously in tear fluid to free the drug-loaded particles in the preocular surface. With this strategy, the particles still maintained a preocular retention property, as well as showing sustained drug release, hence improved drug bioavailability in the eye.

Herein, a rapidly-dissolving dry tablet formulation embedded with drug-loaded PLGA nanoparticles (i.e., the DX/NP) was proposed again for topical drug delivery to the eye. Owing to the presence of alginate in the tablet medium, tear viscosity increased because of the interaction between alginate and Ca2+. This hampered tear clearance, leading to a higher preocular retention of DX/NP (Figure 6). Therefore, in addition to residing longer in the preocular space, DX/NP could continuously release the drug to tear fluids, providing more time for drug diffusion into the eye and eventually improving ocular drug bioavailability (Figure 7).

In the present study, the tablets were prepared by freeze-drying the DX/NP suspension; this allowed the tablet medium of the water-soluble polymers to be highly porous for rapid dissolution of the tablet in the tear fluid. Considering commercialization, the moisture content in the tablet may need to be minimized further to provide with an appropriate shelf-life of the product. Owing to the nano-sizes of DX/NP, the sensitive eye surface would not be irritated [32]. Therefore, with the alginate content employed for the tablet herein, any sign of eye irritation or discomfort was observed during in vivo experiments. By using the applicator, the tablet could be delivered without being touched, thereby achieving a hygienic application. PDMS was employed as the constituent material for the applicator because of its biocompatibility, should the applicator touch the eye surface. More importantly, PDMS is inherently hydrophobic and it is known to possess a low surface release energy [33]. Therefore, the dry tablet composed of hydrophilic polymers could be separated almost instantaneously without loss when wet with tear fluids (Figure S2), thereby enabling the delivery of an accurate drug dose to the eye using the tablet formulation.

#### **5. Conclusions**

Herein, a formulation consisting of a dry tablet containing alginate and drug-loaded PLGA nanoparticles is derived to improve the bioavailability of drugs that are topically delivered to the eye. Owing to the ease and hygienic administration of the developed tablet formulation, the use of a preocular applicator is suggested. Based on the hydrophobicity and low surface release energy of the constituent material, PDMS, the applicator could almost instantaneously cause full separation of the dry tablet composed of hydrophilic polymers when applied to the eye surface. After topical delivery, the tablet medium dissolved rapidly to free the drug-loaded nanoparticles in the tear fluid, where alginate could react with Ca2<sup>+</sup> to increase its viscosity. Such findings suggest that the drug-loaded PLGA nanoparticles can be better retained in the preocular space and the drug can be subsequently released in a sustained manner, ultimately enhancing ocular drug bioavailability. Therefore, it is concluded that the combination of a dry tablet formulation with an alginate medium, drug-loaded PLGA nanoparticles, and a preocular applicator is a promising strategy to achieve patient-friendly, topical drug delivery to the eye, with enhanced drug availability.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/1999-4923/11/12/651/s1; Figure S1: Characterizations of NR/NP; Figure S2: Images of rabbit eyes.

**Author Contributions:** Conceptualization, Y.B.C.; Methodology, Y.B.C., W.M.R., and S.-N.K.; Validation, W.M.R. and S.-N.K.; Animal experiments, W.M.R, S.-N.K., and C.H.M.; Formal analysis, W.M.R. and S.-N.K.; Investigation, W.M.R. and S.-N.K.; Writing—original draft preparation, W.M.R. and S.-N.K.; Writing—review and editing, W.M.R., S.-N.K., and Y.B.C.; Visualization, W.M.R. and S.-N.K.; Supervision, Y.B.C.; Project administration, Y.B.C.; Funding acquisition, Y.B.C.

**Funding:** This research was supported by grants from the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI) funded by the Ministry of Health & Welfare, Republic of Korea (grant no. HI15C1744), the Technology Innovation Program (10060067, Technical development of nanosystem-based technology for drug therapy preventing side effects after cataract surgery) funded by the Ministry of Trade, Industry, and Energy (MI, Korea), and the Bio & Medical Technology Development Program of the National Research Foundation (NRF) funded by the Ministry of Science, ICT & Future Planning (2015M3A9E2030129).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

### **A Combinatorial Cell and Drug Delivery Strategy for Huntington's Disease Using Pharmacologically Active Microcarriers and RNAi Neuronally-Committed Mesenchymal Stromal Cells**

**Emilie M. André 1, Gaëtan J. Delcroix 2,3, Saikrishna Kandalam 1, Laurence Sindji <sup>1</sup> and Claudia N. Montero-Menei 1,\***


Received: 18 June 2019; Accepted: 2 October 2019; Published: 12 October 2019

**Abstract:** For Huntington's disease (HD) cell-based therapy, the transplanted cells are required to be committed to a neuronal cell lineage, survive and maintain this phenotype to ensure their safe transplantation in the brain. We first investigated the role of RE-1 silencing transcription factor (REST) inhibition using siRNA in the GABAergic differentiation of marrow-isolated adult multilineage inducible (MIAMI) cells, a subpopulation of MSCs. We further combined these cells to laminin-coated poly(lactic-*co*-glycolic acid) PLGA pharmacologically active microcarriers (PAMs) delivering BDNF in a controlled fashion to stimulate the survival and maintain the differentiation of the cells. The PAMs/cells complexes were then transplanted in an ex vivo model of HD. Using Sonic Hedgehog (SHH) and siREST, we obtained GABAergic progenitors/neuronal-like cells, which were able to secrete HGF, SDF1 VEGFa and BDNF, of importance for HD. GABA-like progenitors adhered to PAMs increased their mRNA expression of NGF/VEGFa as well as their secretion of PIGF-1, which can enhance reparative angiogenesis. In our ex vivo model of HD, they were successfully transplanted while attached to PAMs and were able to survive and maintain this GABAergic neuronal phenotype. Together, our results may pave the way for future research that could improve the success of cell-based therapy for HDs.

**Keywords:** tissue engineering; Huntington's disease; siRNA; nanoparticles; microcarriers; mesenchymal stromal cells

#### **1. Introduction**

Huntington's disease (HD) is a genetic neurodegenerative disorder caused by the abnormal repetition of CAG nucleotides in the Huntingtin (HTT) gene. This leads to a pathological expansion of polyglutamine (polyQ) and aggregation of the mutated HTT protein in the brain, more specifically in the striatum [1,2]. HD is characterized by a progressive degeneration of striatal GABAergic medium spiny projection neurons, followed by a progressive degeneration extending throughout the brain [3]. Clinically, this results in involuntary movements, cognitive impairment and psychiatric manifestations, culminating in death around 15–20 years after the onset of motor symptoms [4,5]. Currently, there is no proven medical therapy to alleviate the onset or progression of HD [6].

Mesenchymal stromal cells (MSCs), have emerged for clinical transplantation studies due to their easy availability, their immune-modulatory properties [7] and their capacity to release neurotrophic factors and to create a neuroprotective microenvironment [8]. Clinical trials using MSCs in the central nervous system are now underway for many neurological disorders and have shown the feasibility of this approach [9]. Pre-clinical studies with HD models have shown improvement in behavior and a reduced lesion volume after MSC implantation. These beneficial effects could be explained by the secretion of neurotrophic factors including brain-derived neurotrophic factor (BDNF), ciliary neutrotrophic factor, nerve growth factor, insulin-like growth factor 1 and epidermal growth factor (EGF) [10,11]. The trans-differentiation of MSCs into a neural/neuronal lineage, even if possible in vitro, is far less efficient in vivo and their functional maturity remains too scarce [12]. The rationale for their transplantation does not lie on their capacity to replace the damaged neuronal cells, but on their ability to neuroprotect and repair by their paracrine effects on the surrounding environment [8]. Even though they are not used in a cell replacement strategy demanding electrophysiolocally functional connections, if the cells are to be safely used in clinical trials, it is important that they blend into the microenvironment of the brain and present the same neuronal phenotype. Therefore, MSCs are required to be committed to a neuronal cell lineage and maintain this phenotype to ensure their safe transplantation in the brain. In this regard, REST/NRSF a repressor transcription factor functioning as a master negative regulator of neurogenesis by binding to a specific DNA domain named RE1 motif is an interesting target to inhibit and thus induce a neuronal specification [13,14]. It was thus recently shown that the silencing of REST, obtained by a recombinant lentivirus carrying a small interfering RNA (siRNA) for REST, induced a neural/neuronal differentiation of MSCs [15,16].

Nanoparticles have been developed to efficiently and safely deliver siRNA both in vitro and in vivo and to avoid insertional gene mutations and viral toxicity issues. We recently designed lipid nanocapsules (LNC) able to encapsulate siRNAs complexed to lipids, thus protecting the siRNA from degradation. LNCs consisting of a lipid liquid core of triglycerides and a rigid shell of lecithin and polyethylene glycol (PEG) can be formulated by a simple and easily industrialized solvent-free process based on the phase inversion of an emulsion [17,18]. They have a high stability and can destabilize lysosome's membranes by a proton sponge effect [19]. We also demonstrated that LNCs associated with siREST in MSCs were able to induce their neuronal commitment with a better efficiency than the commercial reagent Oligofectamine® [9]. However, MSCs are a heterogeneous population presenting different differentiation properties. A homogeneous subpopulation of MSCs, termed marrow-isolated adult multilineage inducible (MIAMI) cells, which present a unique genetic profile expressing several pluripotency markers (Oct4, Sox2, Nanog, SSEA4), share many proteins with embryonic stem cells, secrete more tissue repair factors than MSCs, protect the neurovasculature and can be induced to differentiate into cells from all three germ layers [20–22], emerge as a good alternative for cell therapy studies. MIAMI cells can be specified into the neuronal lineage with epidermal growth factor (EGF) and fibroblast growth factor-2 (FGF-2) [23] and induced to an immature neuronal dopaminergic phenotype presenting appropriate electrophysiological properties, regardless of the donor age, with a three-step protocol in vitro requiring neurotrophins [24,25]. In addition, laminin (LM) was shown to enhance the neuronal differentiation of these cells [8]. However, further efforts are needed to maintain a differentiated phenotype and increase engraftment/survival after transplantation in the brain.

Pharmacologically active microcarriers (PAMs) are biodegradable and biocompatible poly(lactic-*co*-glycolic acid) PLGA-based microspheres covered with extracellular matrix molecules (ECM) such as fibronectin or laminin providing an adequate three-dimensional (3D) biomimetic surface for the transplanted cells and thus enhancing their survival [22,24]. Moreover, PAMs can also release in a controlled manner an encapsulated growth factor allowing a better cell engraftment by stimulating the transplanted cells and/or the microenvironment [22,26]. In this context, PAMs transporting human stem cells and delivering different growth factors have been shown to be beneficial in several animal models of neurological disorders, cartilage and cardiac pathologies [27–30]. In this regard, BDNF, which is involved in neuronal GABAergic differentiation and neuronal survival [30–32] may maintain the differentiated phenotype and increase the survival of the transported MIAMI cells. Moreover, in the case of HD, several studies demonstrated that the expression of BDNF is reduced in

the patient's brains [33]. Promising results show that BDNF supplementation increases the survival of enkephalin-immunoreactive striatal neurons, reduces striatal interneuronal loss and improves motor function in HD animal models [31,34,35]. PAMs would have the benefit to prevent BDNF degradation while providing for its controlled release over time, hence maximizing the possible benefits of BDNF for neuroprotection in HD.

In this study, we aim to first improve the neuronal differentiation of a subpopulation of MSCs, the MIAMI cells, as well as to study their behavior after combining them with PAMs releasing BDNF in vitro and on organotypic brain cultures. We first evaluated the impact of siREST LNC transfection on the neuronal commitment of EGF/FGF-2 pre-treated MIAMI cells. We set up a differentiation protocol based mainly on sonic hedgehog (SHH) treatment followed by BDNF as previously described [36,37], to evaluate the GABAergic differentiation potential of these cells in vitro; GABAergic cells being the predominant cell type present in the striatum. We thus also studied the effects of the REST nanocapsule-silencing on MIAMI cells induction towards the GABAergic phenotype. The secretome and neuronal marker expression of MIAMI cells adhered onto LM-coated PAMs releasing BDNF was also studied in order to assess their potential therapeutic effect. We finally evaluated their behavior in an ex vivo organotypic model of HD. Therefore, the novelty of our paper lies in the combinatorial strategy using siREST silencing nanoparticles and cell transporting biodegradable polymeric PAMs releasing BDNF to achieve neuronally-committed MSCs releasing growth factors for a safe neuroprotective/neurorepair strategy for HD.

#### **2. Materials and Methods**

#### *2.1. siRNA-LNCs*

LNCs were formulated, as previously described [38] by mixing 20% *w*/*w* Labrafac® WL 1349 (caprylic-capric acid triglycerides, Gatefossé S.A. Saint-Priest, France), 1.5% *w*/*w* Lipoid S75-3® (Lecithin, Ludwigshafen, Germany), 17% *w*/*w* Kolliphor® HS 15 (Polyethylene glycol-15-Hydroxystearate PEGHS BASF, Ludwigshafen, Germany), 1.8% *w*/*w* NaCl (Prolabo, Fontenay-sous-Bois, France) and 59.8% *w*/*w* water (obtained from a Milli-Q system, Millipore, Paris, France) together under magnetic stirring. Briefly, three temperature cycles between 60 and 95 ◦C were performed to obtain phase inversions of the emulsion. A subsequent rapid cooling and dilution with ice cold water (1:1.4) at the last phase inversion temperature led to blank LNC formation. For liposome preparation, a cationic lipid DOTAP (1,2-dioleyl-3-trimethylammoniumpropane) (Avanti® Polar Lipids Inc., Alabaster, AL, USA), solubilized in chloroform, was mixed at a 1/1 molar ratio with the neutral lipid DOPE (1,2-dioleyl-sn-glycero-3-phosphoethanolamine) (Avanti® Polar Lipids Inc.) to obtain a final concentration of 30 mM of cationic lipid. After chloroform vacuum evaporation, the lipid film was rehydrated and liposomes sonicated. A simple equivolume mix of liposomes and siRNA resulted in lipoplexes characterized by a charge ratio of 5 between the positive charge of lipids and the negative charge of nucleic acids. To obtain siRNA-LNCs, the water introduced at the last phase inversion temperature was replaced by lipoplexes, i.e., REST siRNA: (sense sequence: 5 -CAG-AGU-UCA-CAG-UGC-UAA-GAA -3 ; Eurogentec, Seraing, Belgium) and control (scrambled) siRNA (sense sequence: 5 -UCUACGAGGCACGAGACUU-3 ; Eurogentec) complexed with cationic liposomes in a defined charge ratio as described above. To avoid the possible denaturation of siRNA the addition of lipoplexes was performed at 40 ◦C.

#### *2.2. Fluorescent siRNA-LNCs-DiD*

To formulate fluorescent siRNA-LNCs, a solution of DiD (1,1 -dioctadecyl-3,3,3 ,3 tetramethylindodicarbocyanine perchlorate; em. = 644 nm; exc. = 665 nm) (Invitrogen, Cergy-Pontoise, France) solubilized in acetone at 25 mg/mL was prepared.

For in vitro experiments, the DiD concentration was fixed at 200 μg/mL of LNC suspension or corresponding to 1.36 mg of DiD per grams of Labrafac®. The adequate volume of DiD solubilized in acetone was incorporated in Labrafac® and acetone was evaporated at room temperature. The formulation process was unchanged, and formulation was stored at 4 ◦C, protected from light. For siRNA fluorescent LNCs, a fluorescent Alexa 488 siRNA (Eurogentec) was used.

#### *2.3. BDNF-Releasing, Laminin (LM)-Coated PAMs*

Synthesis and characterizations of PLGA-P188-PLGA polymer were performed using Synbio3 platform supported by GIS IBISA and ITMO Cancer. BDNF-releasing PAMs were prepared as previously described using a solid/oil/water emulsion solvent extraction-evaporation method [30]. Briefly, BDNF and human serum albumin were first nanoprecipitated separately and nanoprecipitated proteins were dispersed in the organic phase containing the polymer at a protein loading of 1 μg of protein and 5 μg of human serum albumin/mg of PAMs. The suspension was emulsified in a poly(vinyl alcohol) aqueous phase and after solvent extraction in an aqueous phase, the microspheres were filtered and freeze-dried. Blank microspheres, without protein, were prepared following a similar process. To obtain LM-covered PAMS (LM-PAMs), PLGA-P188-PLGA microspheres were coated with LM and poly-d-Lysine (PDL) as previously described [29]. Briefly, the coating solutions prepared in Dulbecco's Phosphate-Buffered Saline (DPBS) were mixed under rotation with the microspheres at a final concentration of the coating molecules of 16 μg/mL of LM and 24 μg/mL of PDL (corresponding to a 40:60 ratio of LM:PDL). In vitro BDNF release from PAMs was performed as previously described by incubation of 5mg PAMs in citrate buffer and dosage by ELISA of collected fractions of the supernatant over time [30].

#### *2.4. LNC and PAM Characterization*

The size and Zeta potential of LNCs (*N* = 3) were measured by using the Dynamic Light Scattering (DLS) method using a Malvern Zetasizer® apparatus (Nano Series ZS, Malvern Instruments S.A., Worcestershire, UK) after dilution at a ratio of 1:200 with deionized water. PAM's size was measured with a Multisizer® coulter counter (Beckman Coulter, Roissy France), zeta potential was measured by DLS [30]. The laminin surface was characterized by confocal microscopy (Leica TCS SP8, France) after LM immunostaining as previously described [30]. Lyophilized PAMs were incubated for 30 min at room temperature (RT) under 15 rpm stirring in DPBS containing 4% bovine serum albumin (BSA), 0.2% Tween 20 (DPBS BT). After washing, anti-LM mouse monoclonal antibody (Sigma-Aldrich, St-Louis, MO, USA, 100 μg/mL in DPBS) was added for 1.5 h under rotation at 37 ◦C. After washing, biotinylated anti-mouse IgG antibody (2.5 μg/mL in DPBS BT) was added for 1 h, at RT, washed and incubated with streptavidin–fluoroprobe 547 (1:1000 in DPBS) at RT, for 40 min. (*N* = 3, *n* = 3)

#### *2.5. MIAMI E*/*F Cells*

MIAMI cells were isolated from human bone marrow (Lonza, donor #3515) and expanded on fibronectin (Sigma-Aldrich) coated flasks at 120 cells/cm2 in low oxygen tension (3% of O2 and 5% of CO2) in Dulbecco's Modified Eagle Medium-low glucose (DMEM, Gibco, Life Technologies, Paisley, UK), supplemented with 3% of Foetal Bovine serum (FBS), 100 μM of ascorbic acid and a mixture of lipids, as previously described [20]. Using the same density of cells and culture condition, a 10 days treatment with an addition of 20 ng/mL of EGF and 20 ng/mL of FGF-2 (both from R&D systems, Lille, France) and 5 μg/mL of Heparin (Sigma-Aldrich) was used to enhance neuronal specification [23] and these cells were named MIAMI E/F cells. Every three days, half of the culture medium was replaced.

#### *2.6. MIAMI E*/*F Cell Transfection*

MIAMI E/F cells were seeded at 2000 to 3000 cells per cm<sup>2</sup> in wells coated with LM (2 μg/cm2, Sigma-Aldrich). Experiments were performed in Opti-MEM® media (Life technologies, France). SiRNA-LNCs were incubated with cells at 37 ◦C in a humidified atmosphere with 3% O2 and 5% CO2 for 4 h before serum addition. Cells were harvested after appropriate time and assayed for mRNA expression levels by RT-qPCR or protein expression level by immunofluorescence.

#### *2.7. LNC Cell Time Retention*

MIAMI cells (2000 to 3000 cells per cm2) were seeded on glass slides coated with LM (2 μg/cm2, Sigma-Aldrich). SiRNA fluorescent LNCs were incubated with cells at 37 ◦C in a humidified atmosphere with 3% O2 and 5% CO2 for 4 h in Opti-MEM® media before DPBS washing and paraformaldehyde fixation (4% of paraformaldehyde during 15 min at 4 ◦C) or FBS addition. At Day 0 and before fixation, 100nM LysoTracker Red (Molecular Probes, Eugene, OR, USA.) was added to the media. After washing, cells were visualized from Day 0 to Day 6 post-transfection using a fluorescence confocal multispectral imaging, FCSI (Leica TCS SP8).

#### *2.8. MIAMI E*/*F Cell Neuronal Di*ff*erentiation*

MIAMI E/F transfected with siRNA-LNC cells were seeded (2000–3000 cells per cm2) on 175 cm2 cell culture flasks for Step 1 and on glass slides on 6 well plates for Step 2 coated with LM (2 μg/cm2, Sigma-Aldrich, St-Louis, MO, USA) and different conditions tested to obtain the best two-step GABAergic differentiation protocol (Figure 1). The first step was performed with DMEM/F12 (Glutamax, Gibco, Life Technologies, Paisley, UK) supplemented with 5% of N2 (1X) (both from Gibco, Life Technologies), and 200 ng/mL of sonic hedgehog (SHH, Peprotech, Rocky Hill, USA) for 14 days and 35 mL of media was used per flask. These cells are named MIAMI-SHH. Second step: Neurobasal media (Neurobasal, Gibco, Life Technologies) supplemented with or without 1 mM of Valproic acid (Sigma-Aldrich) and with or without 30 ng/mL of BDNF (Peprotech) for 14 days and 3 mL of media was used per well. Length and surface area were quantified using MetaVue software®. 6 pictures from each condition (24 in total) were taken with a 10× objective and used to determine total area and length. Only cells responding to the treatment (with neurite-like structures) were analyzed in this experiment.

**Figure 1.** Media tested for the GABAergic differentiation of marrow-isolated adult multilineage inducible (MIAMI) cells.

#### *2.9. Formation of PAMs-Cell Constructs*

Cell adhesion studies were performed based on previous published protocols [22,24]. At the end of Step 1, MIAMI-SHH-siREST cells were detached and resuspended in DMEM-F12 supplemented with either 3% FBS (Lonza, Verviers, Belgium), 5% of N2 (1X) (Gibco, Life Technologies, Paisley, UK) or 2% of B27 (1X) (Gibco, Life Technologies). Lyophilized PAMs (0.5 mg) were resuspended in coated microcentrifuge tubes (Sigmacote, Sigma) containing DMEM-F12 (Gibco, Life Technologies), and mixed with 0.5 mL of cell suspension (2.5 <sup>×</sup> 10<sup>5</sup> cells/0.50 mg PAMs). The mixture was then gently flushed and plated in 1.9 cm<sup>2</sup> Costar ultra low adherence plate (#3473, Corning, Avon, France). Plates were incubated at 37 ◦C during 4 h for MIAMI E/F, to allow cell attachment on PAM surface (covered with FN). PAMs/cell aggregates were pelleted by centrifugation at 200 g for 2 min. Cell adhesion to PAM surface was assessed by microscopic observation and cells adhered to PAMs were quantified using the Cyquant cell proliferation assay (CyQuant Cell proliferation Assay kit, Invitrogen). Complexes were further studied using light and confocal microscopy and scanning electron microscopy. Samples were prepared for scanning electron microscopy analysis as previously described [24].

#### *2.10. Preparation of Brain Organotypic Slices*

Organotypic cultures were prepared as previously described by our team [39]. Briefly, Albinos wild-type Sprague–Dawley rats from the SCAHU (Service Commun d'Animalerie Hospitalo-Universitaire, N◦49007002, Angers University, France) were used. Postnatal 9–11 (P9–11) days old Sprague–Dawley pups were used to generate the ex vivo model of HD. Animals were rapidly euthanized after anesthesia, by intraperitoneal injection of 80 mg/kg of ketamine (Clorketam 1000, Vetoquirol, Lure, France) and 10 mg/kg of xylazine (Rompum 2%, Bayer Health Care, Kiel, Germany). Brains were removed and rapidly dissected before being glued onto the chuck of a vibratome cooled with a bath of Gey's balanced salt solution supplemented with 6.5 mg/L of glucose and antibiotics. 400 μm thick slices were cut using a vibratome (Motorized Advance Vibroslice MA752, Campdem instruments) in different configurations to obtain a progressive degeneration of the GABAergic medium spiny neurons (MSNs). Each hemisphere was mechanically separated to culture 8 organotypic brain slices total (4 from each hemispheres). Obtained slices were next transferred to 30 mm diameter semi-porous membrane inserts (Millicell-CM, Millipore, Guyancourt, France) within a 6-well plate, containing Neurobasal medium (Gibco, Life Technologies) supplemented with 6.5 mg/L of glucose, 1 mM of l-glutamine, 1x B27 supplements (Gibco, Life Technologies) and antibiotics. Slices were incubated at 37 ◦C and 5% CO2 up to 30 days and half of the medium was removed every 2–3 days. Each slice was cultured on a single membrane to increase their survival over time.

#### *2.11. Injection of the Cells-PAMs Constructs into Organotypic Slides*

Three days after organotypic slice preparation, the cells-PAMs were injected into the striatum using a 22-gauge Hamilton needle (Hamilton, Bonaduz, Switzerland) connected to a micromanipulator. The total injection volume consisted of 2 μL of culture medium containing approximately 75,000 cells alone or adhering to 0.1 mg of PAMs. The injections were done at an infusion rate of 0.5 μL/minute. The needle was left in place for 5 min before removal to avoid the cells being expelled from the organotypic slices.

#### *2.12. Reverse Transcription and Real Time Quantitative PCR*

Experiments were performed following the guidelines of the PACeM core facility ("Plate-forme d'Analyse Cellulaire et Moléculaire", Angers, France). Total RNA of cells was extracted, purified using RNeasy Microkit (Qiagen, Courtaboeuf, France), treated with DNase (10 U DNase I/μg total RNA) and the concentration determined with a ND-2000 NanoDrop (Thermo Fisher Scientific, Wilmington, DE, USA). RNA integrity was verified on Experion RNA StdSens chip (Bio-Rad). First strand complementary DNA (cDNA) synthesis was performed with a SuperScriptTM II Reverse Transcriptase (Invitrogen), in combination with random hexamers, according to the manufacturer's instructions. cDNAs were purified (Qiaquick PCR purification kit, Qiagen) and 3 ng of cDNA mixed with MaximaTM SYBR Green qPCR Master Mix (Fermentas) and primer mix (sense and antisense at 0.3 μM, Table 1 (Eurofins MWG Operon, Ebersberg, Germany) in a final volume of 10 μL. Amplification was carried out on a LightCycler 480 (Roche): denaturation step at 95 ◦C for 10 min and 40 cycles of 95 ◦C for 15 s, 60 ◦C for 30 s. Specificity of the primers was controlled. The GeNormTM freeware (http://medgen.ugent.be/-jvdesomp/genorm/) was used to choose GAPDH and ACTB, as the most stable housekeeping genes. The relative transcript quantity (Q) was determined by the delta Cq method: Q = E(Cq min in all the samples tested − Cq of the sample), where E is related to the primer efficiency (E = 2 if the primer efficiency = 100%). Relative quantities (Q) were normalized using the multiple normalization method (Vandesompele et al., 2002). Q normalized = Q/(geometric mean of the stable housekeeping genes Q). The 2(−ΔΔCt) method was retained, using a housekeeping gene and gene of interest (Livak and Schmittgen, 2001) tested on control sample and treated sample.


**Table 1.** Primer sequences used for RT-qPCR.

#### *2.13. Immunocytofluorescence*

After treatment, cells were fixed with 4% paraformaldehyde (PFA, Sigma, St Louis, MO, USA) in DPBS (Lonza, Verviers, Belgium) pH 7.4 during 15 min at 4 ◦C, washed and non-specific sites were blocked with DPBS, Triton 0.1% (DPBS-T, Triton X-100, Sigma), bovine serum albumin 4% (BSA, Fraction V, PAA Lab, Austria), normal goat serum 10% (NGS, Sigma) during 45 min at RT. A mouse anti-human β3-tubulin (2 μg/mL, clone SDL.3D10, Sigma), a mouse anti-human neurofilament medium (NFM, 1:50, clone NN18, Sigma-Aldrich), a monoclonal rabbit anti-human dopamine- and cAMP-regulated neuronal phosphoprotein (DARPP32, 0,6 μg/mL, clone EP721Y, Abcam, Paris, France), a mouse anti-glutamic acid decarboxylase-67 antibody (GAD67, 2 μg/mL, clone 1G10.2, Millipore SA), and a rabbit anti-GABA transporter 1 (GAT1 500 ng/mL Millipore SA) were used to characterize cell differentiation. Cells were incubated overnight with the primary antibody diluted into DPBS-T, BSA 4% at 4 ◦C. After washes, slices were incubated with the biotinylated mouse (2.5 μg/mL, Vector Laboratories, Burlingame, CA, USA) or rabbit secondary antibody (7.5 μg/mL, Vector Laboratories) for 1 hour at RT. Then slices were washed and incubated with Streptavidin Fluoroprobes 488 or 547H (Interchim, Montluçon, France) diluted 1:1000 or 1:500 respectively in DPBS for 1 h before mounting and observation using a fluorescence microscope.

#### *2.14. Immunofluorescence*

Immunofluorescence of brain slices was performed as previously described [39]. Immunofluorescence was performed using antibodies against human mitochondria (hMito) (10 ng/mL, mitochondrial cytochrome C oxidase subunit II, Abcam), rabbit anti-human dopamine- and cAMP-regulated neuronal phosphoprotein DARPP32 (DARPP32, 0.6 μg/mL, clone EP721Y, Abcam) and a mouse anti-GAD67 (5 μg/mL, clone 1G10.2, Millipore SA). Isotypic controls were performed for each antibody. Free-floating slices were incubated in 1% DPBS-T (Sigma-Aldrich). After pre-blocking for 4 h with 4% BSA (fraction V, PAA Laboratories, Piscataway, NJ, USA), 10% NGS (Sigma-Aldrich) in DPBS-T, slices were incubated for 48 h at 4 ◦C with monoclonal antibodies against DARPP32, GAD67 or hMito with diluted in 4% BSA DPBS-T. After washing, the sections were incubated for 2 h slices with the biotinylated mouse (2.5 μg/mL, Vector Laboratorie,) or rabbit secondary antibody (7.5 μg/mL, Vector Laboratories) at room temperature. Then, slices were washed and incubated for 2 h with Streptavidin Fluoroprobes 488 or 547H (Interchim) diluted 1:500 or 1:1000 respectively in DPBS. Finally, the sections were washed mounted using fluorescent mounting medium (Dako, Carpinteria) and observed in a confocal microscopy.

#### *2.15. MIAMI Cell Secretome Analysis*

A Luminex® Multiplex secretome assay was used to quantify cytokines and growth factors secreted by MIAMI cells under different conditions. 72 h before the experiment, media was completely removed and replaced by media without serum for MIAMI E/F or appropriate media for the other conditions. The media was collected and frozen at 72 h. The secretome of MIAMI cells was compared to the secretome of E/F pre-treated MIAMI cells cultured for 7–10 days. The secretome of E/F MIAMI cells was then analyzed after exposure to different conditions: after commitment into GABA-like progenitors (transfected by the si-REST and exposure during 14 days to SHH, and adhered on PAMS during 72H with Neurobasal media and 1 mM of Valproic acid. Eight human growth factors (Brain-derived neurotrophic factor (BDNF), beta polypeptide-Nerve growth factor (b-NGF), stem cell factor (SCF), Leukemia inhibitory factor (LIF), Hepatocyte growth factor (HGF), Placenta growth factor-1 (PlGF-1), Stromal cell-derived factor 1 alpha (SDF-1alpha), Vascular endothelial growth factor-A (VEGFa)) were quantified using 2 Luminex® Multiplex assay panels: ProcartaPlex® Human Chemokine Panel I 9 Plex (#EPX090-12187-901, ThermoFisher), ProcartaPlex® Human Chemokine Panel 11Plex (#EPX110-12170-901, ThermoFisher). Samples were centrifuged at 4 ◦C for 10 min at 10,000 g and prepared as per the manufacturer's recommendations using a Bio-Plex Pro wash station (Bio-Rad, Hercules, CA). No sample dilution was performed. Quantification of growth factors was performed on a Magpix apparatus (Bio-Rad) and analyzed with the Bio-Plex Manager Version 3.0 software (Bio-Rad). Appropriate media were used as control and to determine the background. Background subtraction was performed with appropriate media depending on the sample, *N* = 2, *n* = 2. N represents one independent experiment done at one moment using 375,000 MIAMI cells adhered to 0.5mg of PAMs for each condition tested, n is the number of samples for each condition.

#### *2.16. Data Analysis*

Data are presented as the mean value of three independent experiments +/− standard deviation (SD) unless otherwise stated. Significant differences between samples were determined using an ANOVA test, followed by Dunnett's multiple comparison tests, unless otherwise stated. Luminex data was analyzed with Kruskal–Wallis test followed by pair-wise comparison. Threshold *p*-value was set to 0.05 and significant differences was depicted with a "\*".

#### **3. Results**

#### *3.1. MIAMI E*/*F Cell Transfection*

LNCs carrying REST siRNA of around 85 nm size and a positive surface charge of +7 mV were used for the incorporation of siREST into MIAMI E/F cells. To better understand the increase of neuronal markers in MSCs with siREST-LNCs [38], fluorescence confocal multispectral imaging was used to characterize LNC uptake and siRNA delivery on MIAMI E/F cells (Figure 2A). Triple-labeled LNCs, siRNA and lysosomes were generated, with respectively DiD, Alexa-488 and lysotracker to follow the distribution and localization of siRNA-LNCs over time. Four hours after the contact between LNCs and MIAMI E/F cells, referred to as Day 0, a heterogeneous distribution of LNCs within MIAMI E/F cells was observed (Figure 2A). LNCs (DiD-positive) co-localized with lysosome staining from Day 2 to Day 6 but the same co-localization of the siRNA (Alexa 488) was not observed confirming the results previously published by our team [40]. The number of fluorescent cells for the LNCs and siRNA decreased progressively from Day 0 until Day 9 (Figure 2B). Twelve percent of cells were still DiD-positive at Day 6 (Figure 2B). REST mRNA was quantified by RT-qPCR (Figure 2C) and down-regulated to 70.7% with the siREST at Day 5 when compared to the control cells (siCTRL), receiving a scrambled siRNA.

**Figure 2.** Cellular uptake and cell time retention of siRNA-LNC in MIAMI E/F cells. (**A**) After 4 h incubation, confocal microscopy was performed on MIAMI cells with siRNA-LNCs. Cells were fixed on glass slide and nucleus and lysosome staining were performed with DAPI (blue) and lysotracker (Red). Double fluorescent probes were used to follow siRNA and LNCs: lipophilic DiD (yellow) and Alexa488 siRNA (green). Scale bar represents 10μm. Analysis confirmed the internalization of siRNA-LNCs and its presence until Day 6. (**B**) DiD-positive cells representing the LNC-positive cells were counted using imageJ. 6 images per conditions in 10× objective were selected. (**C**) The expression of REST, measured by RT-qPCR, was decreased significatively (Fold decrease: 4.80 ± 1.80) at Day 5 after transfection compared to the control. *N* = 3. \* Significantly different means at *p* < 0.05.

#### *3.2. MIAMI E*/*F Cells Commitment into GABA-Like Progenitors*

To ensure that the transplanted cells have a similar phenotype to the neurons in the striatum and thus warrant the safety of their transplantation in that area, the MIAMI E/F cells have been differentiated into GABA-like neurons. Cell density during the first steps is important and the confluence has to be maintained under 30% to induce cell differentiation. The MIAMI E/F cells were transfected with siREST or with a negative control, scrambled siRNA, followed by sonic hedgehog (SHH) treatment (Figure 3A). Some of the MIAMI E/F cells exhibited small neurite-like structures. Fourteen days after transfection, most of the MIAMI-SHH-siREST cells exhibited very long neurite-like structures. The majority of the MIAMI-SHH-siREST cells presented a total length of around 600 μm. In comparison MIAMI-SHH siCTRL cells showed a flat morphology (Figure 3B). REST expression, detected by RT-qPCR, was diminished over time during the treatment when compared to the MIAMI E/F cells. As expected at this time-point no difference in REST expression was observed between the cells transfected with siREST and siCtrl. Pax 6 and Dlx2, transcription factors detected in the ventral telencephalon, were expressed by MIAMI E/F cells and decreased slightly 14 days after the treatment (Figure 3C), The

GABAergic marker, GAD67, was slightly increased following the treatment, at the mRNA and protein level (Figure 3C) and not detected in MIAMI E/F cells, as for DARPP32. No major differences were detected between the siREST and siCtrl conditions at this long-term time point, although there was a tendency in the MIAMI-SHH-siREST cells to express more GAD67 at the mRNA and protein level (Figure 3C,D). The early and late neuronal markers, β3-tubulin and NFM, respectively were expressed by the MIAMI-SHH-siREST cells and were only very slightly detected in MIAMI-SHH-siCtrl as shown by the immunofluorescence staining (Figure 3D).

**Figure 3.** MIAMI E/F cell commitment into GABA-like progenitors. (**A**) For GABA-like progenitors commitment, a simple protocol has been designed using siREST and SHH. Transfection of MIAMI E/F was performed with 250 ng/mL of siCtrl and siREST-LNCs. (**B**) After the culture period of 14 days, MIAMI-SHH-siREST emitted long neuritis exhibiting a neuron-like morphology compared to the flat morphology of MIAMI-SHH-siCtrl. (**C**) The expression of neuronal commitment genes (REST, PAX-6, and Dlx2) and of genes of GABAergic-like neurons (DARPP32 and GAD67) was quantified at 14 days by RT-qPCR in MIAMI E/F, MIAMI-SHH-siREST and MIAMI-SHH-siCTRL cells (*N* = 3). (**D**) After immunofluorescence staining, the neuronal commitment of MIAMI-SHH-siREST cells was confirmed by the expression of β3-tubulin and NFM and a very slight expression of GAD67, all of which are not detected in MIAMI-SHH-siCTRL cells. ND: not detected. \* Significantly different means at *P* < 0.05.

#### *3.3. MIAMI E*/*F Cell Commitment into GABAergic-Like Neurons*

The GABA-like progenitors were exposed to Valproic acid (VPA) from Day 14 for seven days and BDNF from Day 21 for seven more days or during 15 days to VPA and BDNF in the same time, without SHH (Figure 4A) in order to obtain GABAergic-like neurons. At this step, no morphological change was observed between cells transfected with siCtrl or siREST, but the cells differentiated using VPA and BDNF in the same time exhibited longer neurite-like structure than the cells receiving the protocol using VPA and then BDNF (data not shown). In the same way, the expression of GABA markers detected by RT-qPCR (Figure 4B) is higher with the protocol using VPA together with BDNF compared to the sequential protocol (VPA then BDNF). At the end of the differentiation period, a low expression of Dlx2 and REST was detected by RT-qPCR (data not shown). Moreover, the inhibition of REST had no effect on the expression of GABAergic neuron markers with a tendency to be diminished for GAD67. The majority of the GABA-like neurons were positive β3-tubulin and some for DARPP32, GAD 67 and GAT1 after immunofluorescence, a feature of medium spiny neurons (Figure 4C).

**Figure 4.** Differentiation and Characterization of GABA-like neurons. (**A**) Schematic procedure of GABAergic differentiation. For both protocols, the concentration of VPA was 10 mM and 30 ng/mL of BDNF). (**B**) The characterization of the differentiation was realized by RT-qPCR. The expression of genes DARPP32 and GAD67 was quantified at the end of the differentiation (*N* = 2, *n* = 1). (**C**) In vitro immunofluorescence against β3-tubulin, DARPP32, GAD67, GAT1 on GABA-like neurons was performed.

#### *3.4. Adherence of MIAMI E*/*F SHH-siREST on PAMs*

The particle size of PAMs measured using a Multisizer Coulter Counter was around 30 μm, as also observed using SEM (Figure 5A). LM completely covered the whole surface of PAMs in small patches (Figure 5B) and a continuous release of BDNF was observed over time. Up to 3000 ng/mL of BDNF was progressively released during 40 days from 5 mg PAMs (Figure 5C). During the commitment and the differentiation, no serum and no antibiotics were used. In order to respect these conditions, different media for cell adherence were tested (Figure 5D). Unfortunately, a high proportion of cells were observed floating in the media or adhered to the plastic, particularly with N2 and also with B27 media (Figure 5D). On the contrary, with the 3% serum, the cells adhered nicely onto the PAMs (Figure 5D). Different times of adherence were tested, and it was observed that the adherence of the GABA-like progenitors increased over time; it was better after 6 h and 72 h in contact with PAMs compared to only 4 h (Figure 5E). At 4 h, a large quantity of cells was not combined to the PAMs, but cell/PAM aggregates could be observed from 6 h onwards. (Figure 5E). Since the cells completely surrounded the PAMs in Figure 5E, the PAMs could not directly be seen anymore. Using SEM, MIAMI

E/F SHH-siREST cells were observed to adhere onto the PAMs with their lamellipodia surrounding the PAMs to form three-dimensional complexes (Figure 5F). The percentage of cells adhered onto PAMs' surface at the end of the cell attachment protocol (72 h) was about 95.2 ± 2.7%, as measured using the CyQUANT DNA quantification method.

**Figure 5.** PAM characterization and adherence of GABA-like progenitors on PAMs. (**A**) PAMs observed using SEM and (**B**) laminin overlay onto PAMs observed by confocal microscopy after immunofluorescence. (**C**) Controlled release of BDNF from PAMs, measured using ELISA. (**D**) Different media were tested for the adher ence: 2% of B27, 15% of N2 and 3% of FBS. The last condition has been chosen for the rest of the experiments. (**E**) The cells were incubated with PAMs and pictures taken by brightfield microscopy at 4 h, 6 h and 72 h. (**F**) Observation of microspheres and cells-PAMs complexes by scanning electronic microscopy showing the cells adhering onto the PAMs. White arrows point at PAMs associated with cells.

#### *3.5. Characterization of MIAMI-SHH-siREST on PAMs*

MIAMI E/F cells secreted a few growth factors, but interestingly SHH/siREST exposure affected the secretory profile of MIAMI E/F cells, which secreted more hepatocyte growth factor (HGF) and had a tendency to secrete more leukemia inhibitory factor (LIF) and vascular endothelial growth factor-a (VEGFa). They also secreted SDF-1α (Figure 6A, Table A1). As observed using RT-qPCR, the GABA-like progenitors expressed less BDNF and nerve growth factor (NGF), but after 72 h of adherence of these cells on the PAMs, the expression of BDNF, NGF and VEGFa increased (Figure 6B). The expression of GAD67 also tended to increase, while DARPP32 remained very low (23.00 ± 12.82 mRNA arbitrary units) At the protein level, the secretome analysis confirmed the release by the cells adhered onto PAMs of the above mentioned growth factors in the media as well as FGF2, EGF and stem cell factor (SCF). There was more placental growth factor (PIGF-1) and of course BDNF when the cells adhered onto PAMs releasing BDNF compared to blank PAMs (Figure 6C, Table A1).

**Figure 6.** Characterization of the secretome profile of the GABA-like progenitors combined with PAMs. (**A**) Media collected from MIAMI E/F cells or MIAMI-SHH-SiREST cells were analyzed by the Luminex apparatus for the quantification of growth factors and cytokines (*N* = 2, *n* = 2). (**B**) The expression of GABAergic genes and GAD67 and growth factors (BDNF, NGF and VEGFa) was quantified by RT-qPCR 72H after adherence to the PAMs (*N* = 3). (**C**) Media collected 72 h after adherence between cells and PAMs were analyzed by the Luminex apparatus for the quantification of growth factors and cytokines (*N* = 2, *n* = 2) N represents one independent experiment with all the conditions, n the number of samples/condition. \* Significantly different means at *P* < 0.05.

#### *3.6. Behavior of PAMs and MIAMI-SHH-siREST in Huntington's Disease Model*

MIAMI-SHH-siREST alone, pre-treated with BDNF, or complexed to PAMs or BDNF-PAMS were grafted in the recently reported ex vivo HD model [39] at Day 5 when 30% of GABAergic striatal cell degeneration has occurred. Immunofluorescence against human mitochondria was used to visualize MIAMI-SHH-siREST cells in the rat brain (Figure 7A) and antibody against human DARPP32 was used to localize GABAergic-like neurons 15 days after grafting in the HD organotypic slices (Figure 7B). Immunofluorescence staining was faint with cells alone, suggesting that some cells died. PAMs delivering or not BDNF improved cell survival and BDNF administered to the cells before transplantation also seemed to slightly improve survival. Moreover, the cells receiving BDNF showed an elongated morphology with neurite-like structures (Figure 7A). Some of the transplanted cells expressed DARPP32 in all the conditions studied except for the cells transplanted alone (Figure 7B). GAD67 and DARPP32 immunostaining of the GABAergic rat striatal cells was more intense after transplantation of MIAMI-SHH-siREST cells on PAMs particularly when delivering BDNF (Appendix A Figure A1).

**Figure 7.** Behavior of MIAMI-SHH-siREST grafted into an ex vivo model of HD. (**A**) Black and white picture at lower magnification showing the PAMs with the transplanted cells within the tissue. All the cells are visualized by DAPI staining and the transplanted cells by immunohistofluorescence against human mitochondria. (**B**) Immunohistofluorescence against human mitochondria (in red) and (**C**) human DARPP32 (in red) for MIAMI cells 14 days after grafting, either alone, or exposed to BDNF for three days before grafting, or on LM-blank PAMs for three days before grafting or on LM-BDNF PAMs for three days before grafting. Red circles were suggested on the images to emphasize the localization of PAMs. DAPI staining (in blue) was used to visualize all the cells. Scale bar is 10 μm.

#### **4. Discussion**

Mesenchymal stromal cell-based neuronal therapies can provide a limitless, easily accessible source of cells but survival and differentiation remain a drawback. Combinatorial strategies with nanoand microvectors designed to improve stem cell differentiation and engraftment are necessary. In this study, we used a combinatorial cell and drug delivery approach. We showed that a subpopulation of MSCs, the MIAMI cells, can differentiate towards the neuronal GABAergic lineage by an epigenetic RNA interfering approach inhibiting REST expression combined to GABAergic inducers. Moreover, GABA-committed MIAMI cells adhered onto PAMs maintained the GABAergic phenotype and secreted neural tissue repair factors. They could furthermore be transplanted while attached to PAMs delivering BDNF in an ex vivo model of HD, and were able to survive and maintain this GABAergic neuronal phenotype, particularly in response to BDNF.

Our strategy to differentiate MIAMI cells into striatal-like neurons was developed based on the simplifying scheme that the normal course of neuronal differentiation may be separated into three successive steps: namely, (*i*) neural induction, (*ii*) regional commitment, and (*iii*) neuronal maturation. To obtain neural induction, we pre-treated MIAMI cells with EGF and FGF-2 as previously published by our group [23]. As expected, MIAMI E/F expressed Pax6 and slightly Dlx2, showing their progression towards a neural/neuronal phenotype. To induce a strong GABAergic commitment and further neuronal maturation, we chose to combine REST inhibition with adapted published protocols for pluripotent stem cells [41,42]. REST expression is progressively reduced during neuronal differentiation in neural stem cells [14] and its inhibition engaged MSCs into a neuronal pathway [15,38]. We chose to transiently inhibit REST with siREST carried by nanoparticles as previously performed [38]. Permanent down-regulation of REST with strategies such as ShREST would likely not be satisfactory since REST inhibition is only transient during normal brain development [43]. The siREST-LNCs inhibited REST expression for at least 5 days, probably due to their long retention time within the cells. After this

transient REST inhibition together with SHH treatment followed by the combination of VPA and BDNF, we first obtained GABA-like progenitors and then immature neurons expressing striatal markers such as GAD67, DARPP32 and GAT1. The main difference observed after REST inhibition was a clear neuronal commitment of the GABAergic-like progenitors, which expressed the neuronal markers B3-tubulin and NFM. To our knowledge, our strategy is the first to describe a direct non-epigenetic transient modulation of REST with nanoparticles to direct MSCs toward a GABAergic lineage. Other approaches used to inhibit REST are burdened by intrinsic limitations such as the HIV origin of the product and host genome integrations of lentivirus or oligodeoxynucleotides, which may lead to unwanted side effects.

MIAMI cells committed or not towards a neuronal phenotype easily adhere onto PLGA PAMs with a biomimetic surface composed of extracellular matrix molecules in the presence of low serum concentrations [26,29,30]. Different serum free media were evaluated for the formation of PAM/cell complexes to avoid washing steps before implantation to dispose of the undesired serum. Moreover, in this way, the implantation procedure is simplified, and the cells implanted in their own conditioned media already containing some BDNF secreted by the cells and delivered from the PAMS. However, the GABA-like progenitor cells only adhered to PLGA-P188-PLGA PAMs with a LM biomimetic surface in the presence of 3% serum. In this study, we observed that MIAMI E/F derived GABA-like progenitors express more GAD67, a constitutive striatal marker, when adhered onto PAMs compared to the cells in a 2D culture. The laminin-covered PAMs stimulate this differentiation as it has been shown that laminin stimulates neuronal differentiation of MSCs [44] and of MIAMI cells [8]. Nevertheless, the 3D condition should also contribute as our previous results showed that PAMs offering a 3D biomimetic surface stimulate MIAMI cells neuronal differentiation [30]. In a similar manner neural stem/progenitor cells also more efficiently differentiated to neurons in a collagen/hyaluronan 3D matrix compared to a 2D culture condition [45].

MSCs have recently emerged as a promising cell population to protect degenerating neuronal cells and increase the function of the remaining cells in various neuronal disorders. In HD preclinical models, the functional recovery observed after MSC transplantation could be explained by the secretion of neurotrophic factors including BDNF, CNTF, NGF, insulin-like growth factor 1 and EGF [10,11]. Within this line, previous studies have shown that MIAMI cells, which are a homogeneous population of cells with an unique genetic profile, express several pluripotency markers (Oct4, Sox2, Nanog, SSEA4) and other proteins also expressed by embryonic stem cells, release more tissue repair factors than MSCs or embryonic stem cells and protect the neurovasculature [45]. Compared to naïve MIAMI cells, MIAMI cells pre-treated with E/F secreted more tissue repair factors such as VEGFA, NGF, LIF and HGF [30], and we observed that SHH/siREST treatment further increased the expression of HGF, but also of BDNF and SDF-1α. SDF-1α is an indispensable chemoattractant for neuron migration in different brain regions (CXCR4 regulates interneuron migration in the developing neocortex.). Moreover, it has been demonstrated that SDF-1 coexists in GABA-containing vesicles in the terminals of basket cells, regulates the strength of GABAergic input to nestin+-type 2 neural progenitors in hippocampal dentate gyrus, and plays a crucial role in adult neurogenesis [46], while laminin also enhances the chemotactic activity of SDF-1 in the thymus [47].

We previously demonstrated that PAMs delivering neurotrophin-3 can improve MIAMI cells survival, thereby preserving neural function in a Parkinson's disease animal model [8,22]. PAMs encapsulating BDNF and composed of a triblock copolymer of PLGA-P188-PLGA were used in this study as we previously showed that it allowed for a complete release of functionally active growth factors from the microspheres [28,30]. Exposure of MIAMI E/F cells to SHH/siREST and adhesion to PAMs had a drastic effect on the expression of NGF and VEGF at the mRNA level, and we also observed an augmentation at the protein level of BDNF and PIGF-1 in the media when cells were attached to BDNF-loaded PAMs compared to blank PAMs. The increase in BDNF in the media was likely due to the release of BDNF from PAMs since no difference was observed at the mRNA level in MIAMI SHH/siREST attached to PAMs with or without encapsulated BDNF. VEGF and NGF are known to increase neurogenesis [48], while VEGF, NGF, and PIGF-1 also all share in common the possibility of being of benefit to enhance reparative angiogenesis [49,50]. In this regard, we demonstrated in another study that VEGF secreted by MIAMI cells can improve vascularization maintenance after grafting in an ex vivo model of Parkinson's disease [22]. These findings thus support the use of PAMs to enhance the benefit of MIAMI SHH/siREST in the context of HD. In this study, we tested this innovative strategy combining mesenchymal stromal cells and biomaterials in an ex vivo model of HD [39]. The benefit of this model is that it preserves cell graft-host tissue crosstalk, while maintaining the cytoarchitecture of the original tissue. We here demonstrated the screening capacity of this HD model for new therapeutic strategies. After grafting, GABA-like progenitors alone did not survive while if associated to PAMs the 3D support seemed to stimulate the survival of the cells after transplantation as already demonstrated by our group [22–24]. BDNF pre-treatment also stimulated cell survival while the combination of the BNDF delivery and the 3D polymeric support, provided by the PAMs, seemed to further enhance this survival. Moreover, these GABA-like progenitors submitted to BDNF, known to induce neuronal survival and differentiation [30–32] showed an elongated morphology with neurite-like structures suggesting the maintenance of their neuronal phenotype. In this regard, they also expressed DARPP32, a GABAergic marker. Finally, the combination of cells and PAMs delivering BDNF seemed to delay the degeneration of GABAergic neurons in the HD model, thereby demonstrating the potential benefits of this strategy for the treatment of HD. Further studies in a transgenic mouse model of HD are however needed to obtain confirmation that this combinatorial therapeutic strategy leads to neurorepair and provides a functional benefit.

#### **5. Conclusions**

We here demonstrated the therapeutic potential of this novel combinatorial strategy using LNC-delivered siREST and neuronally-committed MSCs transported by laminin-covered microcarriers delivering BDNF. We showed the capacity of MIAMI E/F cells to differentiate into GABAergic-like neurons by using REST inhibition and the appropriate media cues in vitro. GABA-like progenitors adhered to PAMs increased their mRNA expression of NGF/VEGF-a as well as their secretion of PIGF-1, three chemokines known for their potential to enhance reparative angiogenesis. Together, our results indicate that MIAMI cell/PAMs complexes survive after transplantation in an ex vivo HD model, maintain a GABAergic-like phenotype and may alleviate cell damage in the context of HD through chemokine secretion.

**Author Contributions:** Conceptualization, E.M.A. and C.N.M.-M.; Formal analysis, E.M.A. and G.J.D.; Funding acquisition, C.N.M.-M.; Investigation, E.M.A., G.J.D., S.K., L.S. and C.N.M.-M.; Methodology, E.M.A., G.J.D. and S.K.; Project administration, L.S. and C.N.M.-M.; Supervision, C.N.M.-M.; Validation, C.N.M.-M.; Writing—original draft, E.M.A.; Writing—review & editing, G.J.D. and C.N.M.-M.

**Funding:** This research was funded by La Région Pays de la Loire, grant Name: "MECASTEM".

**Acknowledgments:** The authors thank the SCAHU (Service commune d'animalerie hospital-universitaire), particularly Pierre Legras and Jérôme Roux for animal care, the SCIAM ("Service Commun d'Imageries et d'Analyses Microscopiques") of Angers for the confocal microscopy as well as the PACeM (Plateforme d'Analyse Cellulaire et Moléculaire) of Angers for the use of PCR facilities. We also wanted to thank Professor Paul C. Schiller for helpful critiques in experiments. This project is supported by the "Education Audiovisual" and the executive cultural agency of the European Union through the NanoFar Erasmus Mundus joint Doctoral program, by "Angers Loire Métropole" and by "la Région Pays de La Loire grant name MECASTEM" & "Inserm", France.

**Conflicts of Interest:** The authors declare no conflicts of interest.

#### **Appendix A**

*Immunohistochemistry.* A mouse anti-GAD67 (5 μg/mL, clone 1G10.2, Millipore SA, Guyancourt, France) and a mouse anti-DARPP32 (0.25 μg/mL, clone 15, BD Bioscience, Le Pont de Claix, France) antibodies were used to observe striatal-GP GABAergic neurons. Isotypic antibodies were used to control background staining. After 2 h of saturation in DPBS, 4% BSA, 1% triton and 10% NGS, slices were incubated 48 h with the primary antibody diluted in DPBS, 4% BSA and 1% triton at 4 ◦C. After washes, slices were incubated with the biotinylated anti-mouse secondary antibody (2.5 μg/mL Vector Laboratories, Burlingame, CA, USA). Then slices were washed, and quenching of peroxidase was performed with 0.3% H2O2 (Sigma, Saint Quentin Fallavier, France) in DPBS, at room temperature for 20 min. After DPBS washes, slices were incubated with Vectastain ABC reagent (Vector Laboratories, Eurobio, Les Ulis, France) in DPBS at room temperature for 2 h. Sections were then washed and revealed with 0.03% H2O2, 0.4 mg/mL diaminobenzidine (DAB, Sigma, Saint Quentin Fallavier, France) in DPBS, 2.5% nickel chloride (Sigma, Saint Quentin Fallavier, France) and dehydrated before mounting.

**Figure A1.** Neuroprotective effect of cells injection on the GABAergic degeneration. Immunohistochemistry against striatal GAD67-positive neurons and DARPP32-positive neurons at Day 25 after injection of media, cells alone, cells on blank PAMs and cells on BDNF PAMs in the ex vivo model of HD. Scale bar is 100μm (*N* = 2).

**Table A1.** Raw data of the secrotome analysis. Media collected from MIAMI E/F cells, MIAMI-SHH-SiREST cells or 72 h after adherence between cells and PAMs were analyzed by the Luminex apparatus for the quantification of growth factors and cytokines (*N* = 2, *n* = 2).


#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **MRI**/**Photoluminescence Dual-Modal Imaging Magnetic PLGA Nanocapsules for Theranostics**

### **Yajie Zhang 1, Miguel García-Gabilondo 2, Anna Rosell 2,\* and Anna Roig 1,\***


Received: 30 October 2019; Accepted: 17 December 2019; Published: 21 December 2019

**Abstract:** Developing multifunctional and biocompatible drug delivery nanoplatforms that integrate high drug loads and multiple imaging modalities avoiding cross-interferences is extremely challenging. Here we report on the successful chemical reaction of the high quantum yield biodegradable and photoluminescent polyester (BPLP) with the poly(lactic-co-glycolic acid) (PLGA) polymer to fabricate biocompatible photoluminescent nanocapsules (NCs). Furthermore, we transform the PLGA-BPLP NCs into a magnetic resonance (MR)/photoluminescence dual-modal imaging theranostic platform by incorporating superparamagnetic iron oxide nanoparticles (SPIONs) into the polymeric shell. In vitro phantoms confirmed the excellent MRI-r2 relaxivity values of the NCs whilst the cellular uptake of these NCs was clearly observed by fluorescence optical imaging. Besides, the NCs (mean size ~270 nm) were loaded with ~1 wt% of a model protein (BSA) and their PEGylation provided a more hydrophilic surface. The NCs show biocompatibility in vitro, as hCMEC/D3 endothelial cells viability was not affected for particle concentration up to 500 μg/mL. Interestingly, NCs decorated with SPIONs can be exploited for magnetic guiding and retention.

**Keywords:** PLGA nanocapsules; magnetic resonance imaging; photoluminescence; drug delivery systems; magnetic targeting; multimodal imaging; theranostics

#### **1. Introduction**

Nanomedicine, which refers to the application of nanotechnology in medicine, offers valuable new tools for the diagnosis and treatment of many diseases. Nanoparticles are increasingly important by assisting to expedite the development of contrast agents, therapeutics, drug delivery vehicles, and theranostics in the context of nanomedicine [1]. Polymer-based nanoparticles are frequently proposed as drug carriers due to their biocompatibility and biodegradability, as well as the possibility of customizing their physicochemical properties for a specific drug or delivery route [2–4]. Among them, poly(lactic-co-glycolic acid) PLGA nanoparticles have gathered particular attention, since they encompass a number of interesting features: (i) FDA and European Medicine Agency approval in drug delivery systems for parenteral administration; (ii) well described formulations and methods of production adapted to various drugs i.e., hydrophilic or hydrophobic, small molecules, or macromolecules; (iii) protects the loaded drugs from degradation and possibility of sustained release; and, (iv) easy to modify to include targeted delivery or to provide tuned performance in a specific biological environment [5,6].

Theranostics is an emerging field that combines diagnostics and therapeutics into multifunction nanoparticle systems [7]. Drugs and nanocarriers in vitro/in vivo fate can be monitored while using noninvasive imaging techniques, such as magnetic resonance imaging (MRI) and fluorescent imaging, in order to optimize the route of delivery, biodistribution, and drug accumulation, among other

factors [4,8,9]. PLGA nanoparticles can be used in diagnostic and therapeutic imaging by the addition of the imaging moieties during the particle synthesis. To date, large efforts have been devoted to conjugating PLGA with semiconducting quantum dots and organic dyes to create photoluminescent PLGA nanocarriers [9–11]. However, conventional approaches for physically blending imaging probes within the nanoparticles can lead to incomplete conclusions or misinterpretations on the nanoparticles' biodistribution and fate [12,13]. Moreover, the inevitable photobleaching and low dye-to-polymer labeling ratios of organic dyes and the innate toxicity of quantum dots prevent their practical use in vivo [14,15]. Recently, a series of biodegradable photoluminescent (PL) polyesters (BPLPs) could derived inherently photoluminescent PLGA-BPLP copolymers with excellent biocompatibility, tunable luminescence and degradation rates, and good thermal and mechanical properties, thus expanding the biomedical applications of PLGA to highly desired optical imaging [16,17].

Here, we present PLGA nanocapsules (NCs) as a dual-modal imaging theranostic platform for magnetic targeting protein delivery (Figure 1). We report on the protocol for the fabrication of magnetic PLGA-BPLP NCs with intrinsic photoluminescence and MRI capacity being endowed by the incorporation of superparamagnetic iron oxide nanoparticles (SPIONs). Moreover, the NCs are functionalized with poly(ethylene glycol) (PEG), providing a hydrophilic surface that could result in an enhanced stealth effect. Unlike solid PLGA nanoparticles [4,18,19], in our NCs, the functional moieties are incorporated into the polymeric shell matrix to minimize interferences with the cargo being placed inside the NC; this is especially important for delicate payloads, such as proteins, enzymes, or microRNAs. In this study, bovine serum albumin (BSA) has been used as a model protein for evaluating the protein loading capability and release kinetics of the proposed nanocarrier.

**Figure 1.** Schematic illustration of poly(lactic-co-glycolic acid)-biocompatible photoluminescent polymer nanocapsule (PLGA-BPLP NC) as a dual-modal imaging theranostic platform.

#### **2. Experimental Section**

All of the reagents were purchased from Sigma-Aldrich (Merck KGaA, Darmstadt, Germany) unless specified otherwise.

#### *2.1. Synthesis of PLGA-BPLP Copolymer*

The PLGA-BPLP copolymer was synthesized by a slight modification of Yang's method [16,17]. Step one involves the synthesis of a hydroxyl terminated BPLP pre-polymer. Briefly, citric acid, 1,8-octanediol, and l-cysteine with the molar ratio of 1:1:0.2 (5.76 g, 4.38 g, 0.72 g) were added into a flask of 50 mL with a stirring bar. Under a constant flow of argon, the reactants were melted by heating to 140 ◦C until a clear solution formed. The solution was allowed to react at this temperature for 80 min. and then stopped before the stirring bar stopped stirring completely in the increasingly viscous solution by adding 25 mL of 1,4-dioxane to dilute the produced prepolymer. The product was purified by dropwise precipitation from the 1,4-dioxane solution in water to remove the unreacted monomers. The final BPLP product was collected by centrifugation and then lyophilized.

In the second step, the PLGA-BPLP copolymer was synthesized while using the BPLP as a macroinitiator to react with l-lactide and glycolide via a ring-opening polymerization that was catalyzed by Tin (II) 2-ethylhexanoate (Sn(OCt)2). Briefly, l-lactide, glycolide and BPLP with the molar ratios of (75:25):1 or (50:50):1 were added into a reaction tube. Subsequently, Sn (OCt)2 (0.1 wt% of l-lactide and glycolide mixture) was added as a solution in dichloromethane which was then evaporated in vacuum. The tube was flushed with argon and capped and subsequently immersed in a 160 ◦C oil bath for 48 h. The obtained product was dissolved in chloroform and then purified by dropwise precipitation into pure ethanol to remove unreacted raw materials. Finally, the PLGA-BPLP copolymer was recovered by centrifugation and then dried in vacuum at room temperature. All of the PLGA-BPLP mentioned in the text were with molar ratios of (75:25):1, unless specified otherwise.

#### *2.2. Synthesis of Oleic Acid Coated SPIONs*

The SPIONs were synthesized by microwave assisted thermal decomposition in a microwave synthesizer (Discover SP, CEM Corporation, Matthews, NC, USA), and then coated by oleic acid (OA) [20]. The process to obtain homogeneous OA-SPIONs with an average diameter of 9 nm was as follows: 3.5 mmol of iron precursor Fe (acac)3 was dissolved in 4.5 mL of benzyl alcohol in a microwave reaction glass tube. Microwave irradiation was initiated at 60 ◦C for 5 min. to fully dissolve the precursor and, subsequently, the irradiation was kept at 210 ◦C for 30 min., and the reaction was then stopped and cooled down to room temperature. 4 mL of oleic acid in toluene (0.8 mmol/mL) was added immediately into the as-synthesized SPIONs dispersion followed by incubation under ultrasound for 1 h. Subsequently, the obtained OA-SPIONs were separated by centrifugation in five-fold of acetone. The pellet was redispersed in 4 mL of toluene in a glass vial and a magnet was attached on the wall for 5 s, the un-adsorbed suspension containing SPIONs with smaller size was discarded and the adsorbed pellet was redispersed in 6 mL of dichloromethane (DCM), followed by centrifugation at 4000 rpm for 5 min. to sediment the unstable big particles. Finally, the stable OA-SPIONs (9 nm) dispersion in DCM was centrifugated in five-fold of acetone and the pellet product was dried under vacuum and redispersed in DCM at the concentration required for use.

#### *2.3. Fabrication of Functional PLGA-BPLP NCs and Encapsulation of BSA*

Functional PLGA NCs encapsulating BSA were prepared by a double emulsion solvent evaporation method. Briefly, 50 μL of inner aqueous phase (W1) containing BSA (30 mg/mL) was emulsified in 500 μL of DCM organic phase (O) that was composed of 50 mg of different proportions of PLGA (RG502H, Mn 12,000)/PLGA-BPLP/PLGA-PEG (PLGA Mn 7000, PEG Mn 5000) and a certain amount of OA-SPIONs by sonication (VC505, Sonics & Materials Inc., Newtown, CT, USA) at 200 W for 28 s to form the first emulsion (W1/O). Afterwards, 2 mL of external aqueous phase (W2) with polyvinyl alcohol (PVA) (20 mg/mL) was added and the second emulsion (W1/O/W2) was formed by sonication for another 28 s. The temperature during the whole emulsion process was kept at 4 ◦C by using an ice bath. The resulting double emulsion was poured into 50 mL of MilliQ water and mechanically stirred at RT for 2 h to allow for complete evaporation of the organic solvent DCM and the formation of NCs. Finally, the NCs were washed three times with MilliQ water and lyophilized in 6 mL of trehalose aqueous solution (2 mg/mL). The as-obtained powder was stored at 4 ◦C with desiccant silica gel.

The non-PEGylated and PEGylated NCs have been labelled, as follows: NC1.non PEGylated (90 wt% PLGA-BPLP + 10 wt% PLGA) and NC2.PEGylated (90 wt% PLGA-BPLP + 3 wt% PLGA + 7 wt% PLGA-PEG), both containing ~6 wt% of SPIONs, as gathered in Table 1.


**Table 1.** Summary of the two main formulations for the NCs. Oleic acid coated-SPIONs and PLGA-BPLP and PLGA-PEG were mixed in the organic phase during the double miniemulsion process.

### *2.4. Physicochemical Characterization of PLGA-BPLP, SPIONs, and NCs*

#### 2.4.1. Absorption Spectroscopies of the Polymers

Attenuated total reflectance-Fourier transform infrared (ATR-FTIR) characterization of the polymers was performed on a Bruker Vertex 70 FTIR spectrometer with a Pike Miracle Single-Bounce diamond crystal plate accessory at room temperature. The FTIR spectra were recorded over a wavelength range of 4000–500 cm−<sup>1</sup> with a resolution of 4 cm<sup>−</sup>1. UV-Vis absorbance of the fluorescent polymers was recorded on a Varian Cary-5000 UV-Vis spectrophotometer while using a quartz cuvette with an optical path of 1 cm.

#### 2.4.2. Size Distribution and Disperse Stability of the NCs

Dynamic light scattering (DLS) (Zetasizer Nano ZS, Malvern Panalytical, Madrid, Spain) measurement of the hydrodynamic diameter and size distribution of NCs by intensity was performed by redispersing 0.5 mg of lyophilized NCs powder into 1 mL of MilliQ water.

Turbiscan (Turbiscan Lab, Formulaction, Toulouse, France) is used to detect the destabilization of the NCs suspension. 15 mL of 2 mg/mL NCs in the Turbiscan cell were scanned at all of the heights of the suspension with a time interval of 2 min. during 24 h. The back scattering signals at different heights of the cell were recorded and the delta of back scattering intensity was calculated by subtracting the reference time 0 s. The bottom part was defined as 1/5 of the liquid level.

Nanosight (NS300, Malvern Panalytical, Madrid, Spain) is used to measure the averaged size and concentration of the NCs water suspension. 0.2 mg/mL of NCs with 50 times dilution was pumped into the cell and the data were acquired and analyzed through the Nanoparticle Tracking Analysis (NTA) software along with the instrument.

#### 2.4.3. Electron Microscopies

Field emitting scanning electron microscope (SEM, FEI Quanta 200 FEG, Thermo Fisher Scientific, OR, USA) and transmission electron microscope (TEM, JEM-1210, JEOL Ltd., Tokyo, Japan) were used to study the morphologies of SPIONs and NCs. 0.5 mg of lyophilized powder was redispersed into 1 mL of MilliQ water and centrifuged at 4000 rpm for 10 min. for the SEM sample preparation of NCs. Subsequently, the supernatant was discarded to remove the trehalose (used for cryopreserving during lyophilization) and 1 mL of fresh water was added, the pellet of NCs was redispersed in water with ultrasound. Finally, 6 L of the slightly turbid suspension was deposited onto a small slice of silicon wafer stuck on the top of a carbon layer and dried at room temperature overnight. The sample was sputtered with Au-Pd (20 mA 2 min, Emitech K550, Quorum Technologies Ltd., East Sussex, UK). The TEM samples were prepared by placing and drying one drop of the corresponding NCs or SPIONs dispersion on a copper grid at room temperature.

#### 2.4.4. Magnetometry

Superconductive quantum interference device (SQUID, MPMS5XL, Quantum Design, San Diego, CA, USA) was used to measure the magnetization of NCs and SPIONs and calculate the SPIONs loading (wt%-SPIONs) of the magnetic NCs. Zero-field cooling and field cooling (ZFC-FC) measurement were used to determine the blocking temperature (TB) of the SPIONs. A gelatin capsule filled with about

7 mg of samples, together with some cotton wool, was inserted into the SQUID magnetometer sample holder and the hysteresis loop was measured from −50 kOe to 50 kOe. The saturation magnetization of the NCs (MS-NCs, emu/g) and of SPIONs (MS-SPIONs, emu/g) was used to calculate wt%-SPIONs, as follows:

wt%-SPIONs = MS-NCs/MS-SPIONs × 100%

#### 2.4.5. Fluorescence Properties

Fluorescence spectra of the polymers and NCs were acquired on a spectrofluorometer (LS45, PerkinElmer Inc., Waltham, MA, USA). The excitation and emission slit widths were both set at 10 nm. The quantum yield of the polymers was measured by the Williams' method [16]. Briefly, a series of BPLP/PLGA-BPLP solutions in the corresponding solvents were prepared with gradient concentrations. Maximal excitation wavelength was determined, which generated the highest emission intensity. The fluorescence spectra were collected for the series of solutions in the 10 mm fluorescence cuvette (Figure S1). The integrated fluorescence intensity, which is the area of the fluorescence spectrum, was calculated and then noted. Afterwards, the UV-Vis absorbance spectra were collected with the same solutions and the absorbance at the maximal excitation wavelength within the range of 0.01–0.1 Abs units was noted (Figure S1). The graphs of integrated fluorescence intensity vs. absorbance were plotted. The quantum yield was calculated according to the equation:

$$\Phi \mathbf{s} = \Phi\_r(\frac{\text{Slope}\_s}{\text{Slope}\_r})(\frac{n\_\mathbf{s}}{n\_r})^2$$

where, Φ = quantum yield; Slope = slope of the straight line obtained from the plot of intensity vs. absorbance; *n* = refractive index of the solvent; *s* = subscript denotes the sample; and, *r* = subscript denotes the reference used. Here, anthracene (Φ = 27% in ethanol when excited at 366 nm) was used as the reference.

The fluorescence intensities of different concentrations of NCs in water were quantified in a 96-well by a microplate reader (Spark, Tecan Group Ltd., Männedorf, Switzerland). The NCs were excited at the maximal excitation wavelength and the fluorescence signal was collected by area scan at the maximal emission wavelength.

Photostability was measured by continuously illuminating the polymer or NCs in a fluorescent confocal microscope (Leica SP5, Leica Microsistemas S.L.U., Barcelona, Spain) while using UV diode 405 nm excitation at different laser power (10%, 20%, 100%). The fluorescence images were acquired at a time interval of 1 s for 10 min., changes of mean fluorescence intensity of six region of interests (ROIs) in 10 min. were calculated while using software Las AF. The real laser power (W) at different percentage during 10 min. was monitored using a laser power meter.

#### *2.5. MRI Phantoms of the NCs*

In vitro agarose phantoms of NCs were prepared in Eppendorfs where a series of concentrations of NCs were vortexed and sonicated in agarose water solutions before the gel formed. The volume was kept at 1 mL with 0.63 wt% of agarose (Conda, Madrid, Spain). The corresponding iron doses (mmol/L) were calculated according to the wt% of SPIONs in each sample. T2 maps of the phantoms were acquired at 7 T in a 70/30 Bruker USR Biospec system (Bruker GmbH, Ettlingen, Germany), as follows: multi-slice multi-echo (MSME) sequence with echo time (TE) = 13 ms, repetition time (TR) = 4000 ms, field of view (FOV) = 5.5 × 11 mm, and three slices of 1 mm thickness. The quantitative T2 values were obtained from hand-drawn ROIs by using curve fitting in the Image Sequence Analysis (ISA) software along with the instrument.

#### *2.6. In Vitro Toxicity Evaluation of the NCs*

Two parallel methods assessed the toxicity of the NCs on human brain endothelial cells (hCMEC/D3): a viability assay based on WST-8 tetrazolium salt reduction (cell counting Kit-8, Dojindo) and direct cell counting. First, 10<sup>4</sup> viable cells per well were seeded on a 24-well plate pre-treated with collagen (rat tail type I, Corning) in 400 μL of endothelial growth medium (EGM2 from Lonza with 2% fetal bovine serum and half the amount of the growth factors that were included in the kit). After incubation for 72 h at 37 ◦C with 5% of CO2, the cells were at 80–90% confluence and medium was changed to endothelial basal medium (EBM2, Lonza) containing NCs at 25, 50, 100, 500, and 1000 μg/mL. After 48 h, the cells were washed and incubated with 10% WST-8 solution for two hours. Culture supernatants were centrifuged at 20,000 rpm for 5 min. to remove NCs detritus that could interfere with the dye absorption before determining the absorbance at 450 nm. Cells were trypsinized and resuspended in growth medium and diluted 1:1 in Trypan Blue in order to perform cell counting in a Neubauer chamber. Cell viability and count are expressed as the percentage of absorbance or number of cells as compared with the control (vehicle without NCs). ANOVA test and Dunnett's multiple comparisons post-hoc test was performed vs. the control.

#### *2.7. In Vitro Observation of NCs Cellular Uptake*

The hCMEC/D3 endothelial cells were seeded in cover-slips that were pre-treated with collagen (2 <sup>×</sup> <sup>10</sup><sup>4</sup> cells/well in 24 well plates) and 24 h later were exposed to 50 <sup>μ</sup>g/mL of empty fluorescent NCs. After an additional 24 h of culture, the wells were washed with PBS, fixated with 4% paraformaldehyde, and mounted with Vectashield antifade mounting medium (Vector Laboratories) with propidium iodide for nucleic acid counterstaining. Additionally, some of the cells were stained with PKH26 lipophilic fluorescent dye for cell membrane labeling according to the manufacturer's protocol. Images at 63× were obtained on a fluorescent confocal microscope (LSM 980 with Airyscan 2 detector, Zeiss, Oberkochen, Germany).

#### *2.8. BSA Loading in NCs and Release Kinetics*

The albumin content of the NC was directly determined while using the CBQCA protein assay kit (Invitrogen™ ref. C6667), which determines the protein concentration based on the production of fluorescent products measurable at λex/λem = 450 nm/550 nm via non-covalent interaction between CBQCA and primary aliphatic amines of proteins. This highly sensitive fluorescence-based method showed compatibility with DMSO, SPIONs, detergents, and other substances that interfere with other commonly used protein determination methods. Lyophilized NCs encapsulating albumin as well as empty NCs as control were fully dissolved in DMSO at 100 mg/mL. The protein contents in the NCs lysates were measured and calculated based on the difference in fluorescence with the control and a calibration curve drawn with standard albumin solutions, the protein contents in the BSA solutions used for encapsulation were also measured. As listed in Table 1 for the two types of NCs, at least two replicate NCs batches of each were measured for the BSA loading and encapsulation efficiency (EE%) calculation. All of the measurements were performed in duplicate for each NCs batch. The experimental BSA loading in the NCs is expressed as μg of BSA per mg of NCs (μg/mg) or the wt% of the NCs, and the BSA EE% is calculated, as follows:

$$\text{EE\% BSA} = \frac{\text{Experimental BSA loading}}{\text{Nominal BSA loading}} \times 100\% \text{ } \%$$

For the release studies, lyophilized NCs were resuspended in phosphate buffered saline (PBS) (pH 7.4, Sigma ref. D1408) at 10 mg/mL in low protein binding microcentrifuge tubes (Thermo Scientific© ref. 90410). NCs solutions were incubated at 37 ◦C in a vertical rotator for different time measures to simulate the in vivo environment: right after the resuspension (time 0), and after three hours, six h, one day, and seven days of incubation. In all cases, an aliquot of 200 μL was frozen

at −80 ◦C until protein determination. Before the CBQCA assay, the aliquots were centrifuged at 15,000× *g* rcf to separate supernatant and pellet. The amount of protein release was directly calculated as total released protein and indirectly from the remaining protein in the pellet as indirect measure. For this purpose, the pellets were fully dissolved in DMSO and then compared with the intact NCs fully dissolved in DMSO (100 mg/mL) used as the 100% release set up. Four BSA-loaded NCs batches and one of H2O-NCs batch as control were used and measures done in duplicate. The release profiles were expressed in terms of cumulative release and plotted vs. time.

#### **3. Results and Discussion**

#### *3.1. Photoluminescent PLGA-BPLP Copolymer*

BPLPs are degradable oligomers synthesized from biocompatible monomers, including citric acid, aliphatic diols, and various amino acids via a convenient and cost-effective polycondensation reaction. BPLPs present some advantages over the traditional fluorescent organic dyes and quantum dots due to their cytocompatibility, minimal chronic inflammatory responses, controlled degradability, and excellent fluorescence properties [16]. Here, l-cysteine was selected and introduced into the polyester structure that was made of biocompatible monomers of citric acid and aliphatic 1,8-octanediol, since previously reported BPLP from this starting amino acid exhibited the highest quantum yield (62.3%) [16]. The fluorophore structure of BPLP was verified as a fused ring structure ((5-oxo-3,5-dihydro-thiazolopyridine-3,7-dicarboxylic acid, TPA) [21]. ATR-FTIR was also used to confirm the chemical structure of the as-synthesized BPLP (Figure 2A). Strong absorptions from the molecular backbone of the polyester were observed i.e., peaks at 1044 cm<sup>−</sup>1, 1176 cm−1, and 1716 cm−<sup>1</sup> are attributed to the C=O stretch, C–O asymmetrical, and symmetrical stretches of the ester bond, respectively, the peaks at 2930 cm−<sup>1</sup> and 2856 cm−<sup>1</sup> are attributed to the C-H stretches of alkane from 1,8-octanediol and the band near 3467 cm−<sup>1</sup> is from the <sup>−</sup>OH. NH bending of the secondary amide at 1527 cm−<sup>1</sup> and <sup>−</sup>SH at 2575 cm−<sup>1</sup> confirm that <sup>l</sup>-cysteine is chemically bound to the poly(diol citrate) chain. The shoulder band near 1635 cm−<sup>1</sup> is attributed to the C=O stretching of the tertiary amide from the TPA ring. The average molecular weight (Mw) of BPLP measured by matrix-assisted laser desorption/ionization time of flight mass spectroscopy (MALDI-TOF-MS) was 1044 g/mol (Figure S2). The BPLP oligomer served as a macroinitiator to react with l-lactide and glycolide via a ring-opening polymerization to produce PLGA-BPLP [17]. The as-synthesized PLGA-BPLP (75:25):1 with molar ratios equal to 75:25 for l-lactide to glycolide and equal to 1:100 for BPLP to total l-lactide and glycolide was reported to have desirable glass transition temperature (Tg, 32.5 ◦C), mechanical properties, fluorescence properties, and degradation rate of the resulting product [17]. Moreover, we have proved here that the 75:25 formulations are more suitable for the fabrication of NCs by a double mini-emulsion method than the PLGA-BPLP (50:50):1 one, as shown in the following section.

The obtained PLGA-BPLP copolymer (Figure 2B inset) exhibits the inherent photoluminescence from the BPLP. The similarities of bands and shapes of the IR spectra of the as-synthesized PLGA-BPLP and commercial PLGA (Figure 2B), for instance, bands at 1084 cm<sup>−</sup>1, 1165cm−1, and 1747 cm−<sup>1</sup> from the ester bonds in PLGA indicate their similar chemical structure, given the fact that BPLP is a very small portion of the PLGA-BPLP copolymer.

The fluorescence of the as-synthesized BPLP and PLGA-BPLP was evaluated and Figure 3A,B depict the excitation and emission spectra. The similar spectra further confirm the inherent photoluminescence of PLGA-BPLP from BPLP. Importantly, the fluorescence intensity of PLGA-BPLP only slightly decreased (10%) after 10 min. of continuous illumination under confocal microscope at 10% of laser power (0.40 ± 0.01 μW) (Figure 3C). When considering the laser power applied to observe stained cells is generally less than 10%, our photoluminescent polymer would exhibit good photostability at in vitro conditions. The calculated high quantum yields of BPLP (64%) and PLGA-BPLP (33%) from Figure 3D are consistent with the previously reported values [16,17]. The remarkable fluorescence properties that are shown here endow the PLGA-BPLP copolymer with high potential for the fabrication of functional photoluminescent NCs.

**Figure 2.** Attenuated total reflectance-Fourier transform infrared (ATR-FTIR) spectra of the as-synthesized biocompatible photoluminescent polymer (BPLP) and PLGA-BPLP copolymer confirming their chemical structures and the successful synthesis. (**A**) BPLP; (**B**) PLGA-BPLP and commercial PLGA as reference. Inset: fluorescence of the PLGA-BPLP under a UV lamp.

**Figure 3.** (**A**,**B**) excitation and emission spectra of the as-synthesized BPLP and PLGA-BPLP. Insets: fluorescence of BPLP and PLGA-BPLP dispersed in solutions under a UV lamp; (**C**) photostability evaluation of PLGA-BPLP powder under confocal microscope at different laser power, fluorescence intensity expressed as the percentage vs. the value at the initial time; and, (**D**) fluorescence intensity-absorbance curves of BPLP, PLGA-BPLP, and anthracene used as a reference used to calculate quantum yields.

#### *3.2. Fabrication of PLGA-BPLP NCs Combining Other Functional Moieties*

PLGA-based NCs with BSA encapsulated were prepared by a double emulsion solvent evaporation method with slight modifications from our previously reported method [22]. Functional moieties, such as PLGA-BPLP for the fluorescence imaging, PLGA-PEG to increase hydrophilicity and sealthness, and SPIONs for magnetic targeting and MRI (Figure 4A,B) can be incorporated to the organic phase during the NCs fabrication process (see Table 1). The first important remark is that NC morphology was not affected, even when using 100 wt% of PLGA-BPLP (75:25):1 (Figure S3A). A formulation with 90 wt% of PLGA-BPLP (75:25):1 was selected to allow for NC PEGylation by the addition of 7 wt% of PLGA-PEG and for the benefit of higher fluorescence intensity (Figure S4). Note that 7 wt% of PLGA-PEG (3 wt% PEG, Figure S3B) (NC2) was found to be the maximum amount that can be mixed in the organic phase during the NCs fabrication process, due to the amphiphilic property of PLGA-PEG. For a larger wt% of PLGA-PEG, the morphology and size of NCs were not maintained. As shown in the SEM image (Figure 4C), non-PEGylated and PEGylated NCs, as listed in Table 1, both contain ~6 wt% of SPIONs and depict homogeneous spherical morphologies and sizes (d.nm ~270) being similar to other reported PLGA systems that are suitable for intravenous administration [23]. The upper inset in Figure 4C shows a representative broken NC, exposing the hollow core where the protein drug is loaded. We have also found that the PLGA-BPLP (50:50):1 polymer from using initial molar ratio of LA:GA = 50:50 and BPLP:(LA + GA) = 1:100 was not as suitable for the fabrication of NCs. NCs with homogeneous morphology and narrow size distribution were attained only up to a maximum of 30 wt% of modified PLGA (Figure S3C) and without the possibility of further adding PLGA-PEG when using PLGA-BPLP (50:50):1 (Figure S3D). These results are in accordance with the higher glass transition temperature and better mechanical properties of PLGA-BPLP (75:25):1 over those of PLGA-BPLP (50:50):1 [17]. As expected, the NCs with a higher fraction of PLGA-BPLP (75:25):1 (90 wt%) show higher fluorescence intensity than the ones that were obtained with PLGA-BPLP (50:50):1 (30 wt%) at the same concentration (Figure S4).

Regarding magnetic loading, up to 6 wt% of SPIONs could be loaded without affecting the NCs morphology and yielding a saturation magnetization (MS) value of around 4 emu/g NCs (Figure 4D). In addition, the lower blocking temperature (TB, 33 K) of SPIONs in the NCs (Figure 4D inset) than the SPIONs of dry powder (55 K, Figure 4B inset) further demonstrates that the SPIONs are well dispersed in the polymer matrix. Note that a high magnetic loading is desirable for the magnetic retention of NCs. This is illustrated in the inset images of Figure 4D, where the darker-coloured water suspension of NCs with 6 wt% of SPIONs were adsorbed faster to the tube wall on the magnet side than the NCs with 1 wt% SPIONs loading at the same concentration, promisingly benefiting the magnetically targeted drug delivery as compared to the previously reported results [4,24]. Note that the superparamagnetic behaviour of NCs at room temperature (lack of coercivity, Figure S5) ensures no magnetic interactions among NCs in the absence of an external magnetic field, minimizing the risk of embolization during i.v. administration.

**Figure 4.** (**A**) Representative transmission electron microscope (TEM) image of the oleic acid (OA) coated SPIONs with the upper inset showing the selected area electron diffraction (SAED) pattern of maghemite and lower inset a size distribution histogram of the particles in the image; (**B**) hysteresis loop (5 K) and ZFC-FC (inset) for the OA-superparamagnetic iron oxide nanoparticles (OA-SPIONs), inserted picture shows the stable OA-SPIONs in dichloromethane attracted by an external magnet; (**C**) representative SEM image of lyophilized NCs with the upper inset showing the hollow core of a nanocapsule and lower inset a TEM image of three NCs with the SPIONs visible as black spots well distributed in the polymer matrix; (**D**) hysteresis loop (5 K) and ZFC-FC (inset) for the lyophilized nanocapsule batch NC1, inserted pictures show the water suspension of these NCs (2 mg/mL) where the 6 wt% loading of SPIONs were adsorbed faster to the tube wall on the magnet (diameter 8mm, surface field ~0.4 T) side than that of 1 wt% loading.

It is well reported that the surface PEGylation of engineered nanoparticles provides them with stealth character increasing blood circulation time since nanoparticles are less visible to the reticulo-endothelial system [6]. We evaluated the flocculation regime of the NCs with a Turbiscan to confirm successful surface modification with PEG, since the long hydrophilic PEG chains (Mn 5000) on the surface of the NCs are expected to increase the stability of the NCs in water suspension and decrease the sedimentation rate. Sedimentation of the NCs suspension was monitored for 24 h. Figure 5A shows that, as NCs sedimentation progresses, the back scattering signal of the bottom part of the suspension increases from an increasingly higher concentration of NCs, while the signal of the top part decreases. The sedimentation rates of non-PEGylated and PEGylated NCs were compared with or without a physiological concentration of BSA (0.5 mM) in Figure 5B, as expected the bottom back scattering signal of the PEGylated NCs media increases at a slower rate than the non-PEGylated ones

both with and without BSA, which demonstrates a better dispersibility of the NCs due to the surface hydrophilic PEG chains. Note that the sedimentation rates of non-PEGylated and PEGylated NCs both slow down with the physiological concentration of BSA probably due to the interaction of NCs with the dense BSA solution. Additionally, in Figure 5A, the back scattering signal of the middle part did not vary with time, which means that the NCs were monodispersed at the physiological concentration of BSA and flocculation or coalescence did not occur during the 24 h period. This is in consistent with the DLS size distribution results that are shown in Figure 5C, both non-PEGylated and PEGylated NCs remained monodisperse at the physiological concentration of BSA, which is of great advantage for the i.v. administration and in vivo blood circulation. Nanosight was also used as an additional technique for the determination of size and the concentration of the NCs (Figure 5D). The results show a similar size distribution as obtained by DLS. From the number concentration of the NCs, we can determine a mean mass of 1.06 <sup>×</sup> <sup>10</sup>−<sup>11</sup> mg/NC.

**Figure 5.** (**A**) Back scattering intensity change of the NC2 PBS suspension (shown inset) containing 0.5 mM of BSA at different height of the vial along 24 h measured by Turbiscan; (**B**) quantified back scattering intensity change of the bottom part of the vial along 24 h for NCs suspension in different media measured by Turbiscan; (**C**) dynamic light scattering (DLS) size distributions of NC1 and NC2 PBS suspensions with 0.5 mM of BSA along 24 h; (**D**) quantitative number concentration and size distribution of the nanocapsules measured by Nanosight (n = 3, mean ± SD with error bar).

#### *3.3. Imaging Performance of the Magnetic Photoluminescent NCs*

The fluorescence of the NCs was evaluated, and Figure 6A depicts the excitation and emission spectra. The spectra are similar to those of the PLGA-BPLP polymer shown in Figure 3B. Importantly, the incorporation of SPIONs in the polymer matrix does not quench the fluorescence of NCs. Note that a small displacement of the emission peak wavelength was observed for aqueous dispersed fluorescent NCs when compared to the emission peak of the polymer in a chloroform solution (Figure 3B), which we ascribe to the different interaction of the fluorescent probe with the two solvents. The fluorescence intensities of NCs show a linear dependency on the NCs concentration within a range of 0.1 to 1.0 mg/mL; at higher concentrations the fluorescence shows a trend towards saturation (Figure 6B). NCs can be clearly imaged with a fluorescence confocal microscope and they show a very good photostability while using 10% of laser power, which is ideal for the observation of in vitro cellular uptake (Figure 6C). The strategy used here confers intrinsic photoluminescence to the PLGA NCs without introducing any cytotoxic quantum dots or photo-bleaching organic dyes when compared to other more conventional approaches that physically blend imaging probes within the carrier that can lead to misinterpretations on the tracing of the carrier [12,13], which may greatly expand the applications of this drug carrier.

**Figure 6.** (**A**) Excitation and emission spectra of the nanocapsules (NCs) water suspension; (**B**) fluorescence intensity of different concentrations of NCs measured by microplate reader (n = 2, values represent mean ± sd and subtract values of control non-fluorescence NCs); and, (**C**) photostability evaluation of NCs under confocal microscope at 10% laser power, fluorescence intensity expressed as the percentage vs. the value at the initial time, inset: NC1 water suspension observed at 60× lens.

Phantom studies were conducted to confirm the MRI performance of the capsules. Phantoms of NCs that were dispersed in agarose gel at various concentrations were prepared (Figure 7A). Spin-spin relaxation time (T2) maps clearly exhibit signal decay in a concentration dependent manner. The calculated transverse relaxivity (r2) values at 7 Tesla of both non-PEGylated NC1 (263 mM<sup>−</sup>1s−1) and PEGylated NC2 (237 mM−1s−1) are similar as those seen in Figure 7B, as expected, further demonstrating the similar loading and distribution of SPIONs in the polymer shell matrix for both systems. When compared with other clinically used SPIONs systems, such as Feridex (98 mM<sup>−</sup>1s−1) and Resovist (151 mM<sup>−</sup>1s−1) [25], the much higher r2 value of our NCs formula is expected to be useful for in vivo MRI tracking of the NCs.

**Figure 7.** (**A**) T2 maps of a series of concentrations of NC1 in agarose phantoms; (**B**) r2 relaxivity evaluation for the NCs.

#### *3.4. Cell Viability after NCs Uptake*

Photoluminescent NCs were incorporated by brain endothelial cells after several hours in culture, as seen in Figure 8A, with cytoplasmic localization of the NCs in perinuclear structures compatible with Golgi bodies and endosomes. This subcellular localization was confirmed by Z-stack images (Figure 8B). Importantly, this cellular uptake was biocompatible for endothelial cells, as the main exposed cells during NCs circulation in blood vessels, since viability tests did not show signs of cell toxicity at a wide range of NCs concentrations up to 500 μg/mL (Figure 8C,D) and 48 h exposure. Only extremely high doses (1000 μg/mL) with noticeable occupying space difficulties for cell culturing showed a significant reduction in cell viability and number.

**Figure 8.** NCs uptake and cytotoxicity in human brain endothelial cells. (**A**) Representative images of hCEMC/d3 cells stained with membrane dye PHK26 and exposed to 50 μg/mL of NC1 for 24 h (63× magnification); (**B**) Orthogonal view of a Z-stack of Propidium Iodide (PI) stained cells (showing the cell nuclei) and the fluorescent NCs; (**C**,**D**) hCEMC/d3 cells were treated for 48 h with different concentrations of NC1 and cell viability was determined with WST-8 reduction assay or tripsinized and counted in a Neubauer chamber, (n = 3–4, values represent mean ± SEM, \*\* *p* < 0.01).

#### *3.5. Protein Loading and In Vitro Release*

Protein loading and encapsulation efficiency were determined by lysing NCs with DMSO and measuring total protein content. The BSA loading content was determined as ~10 μg BSA/mg NCs (1 wt%) with an EE% of around 40% for both NC2.PEGylated and NC1.non-PEGylated systems, as listed in Table 1, which indicated that the incorporation of PEG does not affect the protein encapsulation process. Note that a protein loading of 1 wt% is much higher than other reported values (0.03 wt% of vessel endothelial growth factor (VEGF) loaded PLGA NCs) [22] and the EE% of 40% is comparable to the PLGA NCs loaded with neurotrophin-3 or brain-derived neurotrophic factor (47%) [26].

BSA-loaded NCs were able to release protein cargo over time at physiological temperature in PBS media (32% protein release in one week). Figure 9A shows a fast BSA release within the first hours, but not after one day, which could be related to the protein degradation in ex vivo conditions of our assays. The amount of released protein in one week was similar when indirectly measured from the pellet retained protein (Figure 9B,C), although the release profile showed a more sustained pattern over time, which could be associated to the protein that was trapped within the PLGA polymer.

**Figure 9.** Protein release temporal profile of the NCs. (**A**–**C**) NC1 in PBS (10 mg/mL) were incubated at 37 ◦C in rotation and the released BSA quantified as total mass of protein released (**A**) together with BSA content in the remaining NCs pellet by DMSO lysis (**B**) and the released BSA calculated indirectly from the pellet values (**C**). The percentage was calculated versus intact unreleased NCs lysated also in DMSO (n = 8, values represent mean ± SEM).

By using the model protein BSA, here we were able to prove the loading capacity and release kinetics of the drug carrier, yet the preservation of protein functionality after the encapsulation process still needs to be investigated. Nevertheless, in a previous work, we have shown that VEGF could be encapsulated following a similar route and VEGF effect on cell proliferation could be determined [22], which implied the preserved protein functionality.

#### **4. Conclusions**

We here transform PLGA nanocapsules (NCs) into a highly sensitive, MRI/photoluminescence dual-modal imaging theranostic platform for drug delivery by integrating the biocompatible and photoluminescent polyester BPLP into the PLGA molecular structure, as well as by incorporating superparamagnetic iron oxide nanoparticles (SPIONs). Furthermore, we have shown that PEGylation provides a hydrophilic surface to the NCs slowing down their flocculation rated and without modiying the size, SPIONs content or protein loading capacity of the NCs. In all cases, the functional moieties are embedded in the PLGA shell with minimal interferences between them or with the therapeutic protein. The developed magnetic PLGA-BPLP NCs show biocompatibility in vitro. In this regard, the NCs did not affect the viability of endothelial cells in culture for concentration up to 500 μg/mL and 48 h incubation. Finally, we have shown that the NCs can contain 1 wt% of protein in their core achieved at fabrication level and that one third of the encapsulated protein is released in the first week. Interestingly, the NCs decorated with SPIONs can be exploited for magnetic retention and magnetic guiding.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/1999-4923/12/1/16/s1. Figure S1: UV-Vis absorbance spectra and fluorescence spectra of the as-synthesized A/B) BPLP, C/D) PLGA-BPLP and E/F) reference anthracene at a series concentrations for fluorescence quantum yield calculation, Figure S2: Matrix-assisted laser desorption/ionization time of flight mass spectroscopy (MALDI-TOF-MS) was used to determine the average molecular weight (Mw) of BPLP. Average Mw calculated from the mass spectrum of BPLP

*Pharmaceutics* **2020**, *12*, 16

is 1044 g/mol. For the analysis, the sample was spotted in a MALDI plate with 1,8,9-Antracenotriol (Ditranol) and analyzed using a mass spectrometer (Bruker Daltonics Ultraflex TOF/TOF) in reflectron mode, Figure S3: SEM images and DLS size distributions of NCs fabricated by different types of PLGA-BPLP, Figure S4: Fluorescence intensity of different concentrations of NCs measured by microplate reader at the maximal emission wavelengths (values have subtracted values of control non-fluorescence NCs), Figure S5: Hysteresis loop (300 K) and ZFC-FC (inset) measurements of the lyophilized nanocapsule batch NC2.

**Author Contributions:** Conceptualization, Y.Z. and A.R. (Anna Roig); Data curation, Y.Z. and M.G.-G.; Formal analysis, all authors; Investigation, Y.Z. and M.G.-G.; Methodology, Y.Z. and M.G.-G.; Resources, A.R. (Anna Rosell) and A.R. (Anna Roig); Supervision, A.R. (Anna Rosell) and A.R. (Anna Roig); Writing—original draft, Y.Z.; Writing—review & editing all authors. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was partially supported by the Spanish Ministry of Science, Innovation and Universities through the grants PCIN-2017-090, RTI2018-096273-B-I00, CTQ2017-87637-R, SAF2017-87670-R, SEV-2015-0496 and MDM-2017-0720 and by the Generalitat de Catalunya grants (2017SGR765 & 2017SGR1427). Yajie Zhang was supported by the China Scholarship Council (CSC). The Acciones Complementarias program from the Instituto de Salud Carlos III, Spain, co-financed by the European Regional Development Fund (AC17/00004 grant) as part of the MAGBBRIS project (Euronanomed III 8th joint call), and the RETICS INVICTUS RD16/0019/0021 were acknowledged. Miguel Servet program (CPII15/00003) from Fondo de Investigaciones Sanitarias-Instituto de Salud Carlos III and ERDF, supported A. Roig research contract.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### **Theranostic Sorafenib-Loaded Polymeric Nanocarriers Manufactured by Enhanced Gadolinium Conjugation Techniques**

**Tivadar Feczkó 1,2,3,\*, Albrecht Piiper 3, Thomas Pleli 3, Christian Schmithals 3, Dominic Denk 3, Stephanie Hehlgans 4, Franz Rödel 4, Thomas J. Vogl <sup>5</sup> and Matthias G. Wacker <sup>6</sup>**


Received: 15 August 2019; Accepted: 18 September 2019; Published: 23 September 2019

**Abstract:** Today, efficient delivery of sorafenib to hepatocellular carcinoma remains a challenge for current drug formulation strategies. Incorporating the lipophilic molecule into biocompatible and biodegradable theranostic nanocarriers has great potential for improving the efficacy and safety of cancer therapy. In the present study, three different technologies for the encapsulation of sorafenib into poly(d,l-lactide-*co*-glycolide) and polyethylene glycol-poly(d,l-lactide-*co*-glycolide) copolymers were compared. The particles ranged in size between 220 and 240 nm, with encapsulation efficiencies from 76.1 ± 1.7% to 69.1 ± 10.1%. A remarkable maximum drug load of approximately 9.0% was achieved. Finally, a gadolinium complex was covalently attached to the nanoparticle surface, transforming the nanospheres into theranostic devices, allowing their localization using magnetic resonance imaging. The manufacture of sorafenib-loaded nanoparticles alongside the functionalization of the particle surface with gadolinium complexes resulted in a highly efficacious nanodelivery system which exhibited a strong magnetic resonance imaging signal, optimal stability features, and a sustained release profile.

**Keywords:** gadolinium; drug release; polymeric nanocarrier; sorafenib; theranostic nanoparticles

#### **1. Introduction**

Liver cancer, of which the majority of cases are hepatocellular carcinoma (HCC), is a life-threatening disease and, according to global cancer statistics, the third leading cause of cancer-related mortality worldwide [1]. To date, sorafenib is the only drug able to prolong the lives of patients with HCC [2], although at the expense of severe side effects due to uptake of the drug into healthy tissues [3]. Sorafenib is a multi-kinase inhibitor targeting various receptor tyrosine kinases and rapidly accelerated fibrosarcoma (RAF) kinases. The pronounced lipophilicity of the molecule is responsible for its poor

bioavailability and the distribution of the compound into healthy tissues [3]. As a consequence, patients are treated using high doses of the drug and suffer from a number of side effects.

Nanocarrier-based delivery of sorafenib has the potential to improve drug therapy significantly. Theranostic nanodelivery systems offer a versatile combination of therapeutic and diagnostic features, and have previously been applied to this task [4]. A variety of contrast agents, such as gadolinium diethylenetriamine pentaacetic acid (Gd-DTPA), shorten the longitudinal relaxation time, and have been widely used for both vascular and tumor magnetic resonance imaging (MRI) [4,5]. One major limitation of this technique lies in the short half-life of the contrast agent, as well as its poor specificity for the target site.

In this context, Gd-DTPA-conjugated human serum albumin (HSA) nanoparticles improved the contrast of MRI compared to free Gd-DTPA aqueous solution in vivo, due to negative contrasting of the tumors [5]. Even after conjugation of polyethylene glycol (PEG) to the particle surface, HSA nanoparticles exhibited only a short circulation time [6] and the scale-up potential of this technology is rather limited [7].

Block copolymers comprising polylactic-*co*-glycolic acid (PLGA) and PEG have been processed to nanoparticles in pilot scale using either microfluidic technologies [8] or emulsion techniques [9]. Recently, targeted nanotherapeutic formulations successfully passed phase 1 clinical trials [10]. A major shortcoming of nanocarrier delivery is limited drug loading, resulting in high excipient concentrations and administration volume [11].

Several preparation techniques of PLGA nanoparticle preparation have been described in the literature. Previous investigations reported the loading of sorafenib into PLGA nanoparticles, achieving a drug load of 1.4% oil-in-water single emulsion–solvent evaporation method [12]. Preparations made with a nanoprecipitation–dialysis technique using a block copolymer comprising dextran and PLGA resulted in a drug load of 5.3% [13]. To attain synergistic effects of cytostatic agents, co-delivery of drug molecules has been considered. By employing a sequential freeze–thaw method followed by ethanol coacervation, a core–shell construct was manufactured [14]. After preparation of a polyvinyl alcohol (PVA)-doxorubicin nanocore, a thin shell of HSA covering sorafenib was used as a second drug molecule. A drug load of 2.4% was reached. Lipid–polymer hybrid nanoparticles were synthesized for the co-delivery of doxorubicin and sorafenib to enhance efficacy in HCC therapy [15].

Other approaches have focused on theranostic drug delivery systems containing gadolinium (Gd) and sorafenib. Theranostic liposomal carriers with a drug content of 4.3% (*m*/*m*) have been produced [16]. Another system used a multiblock polymer comprising (poly(lactic acid)-poly(ethylene glycol)-poly(l-lysine)-diethylenetriamine pentaacetic acid and the pH-sensitive material poly(l-histidine)-poly(ethylene glycol)-biotin. A drug content of 2.4% (*m*/*m*) of sorafenib was reached, and the MRI signal intensity was more beneficial than that of Magnevist® in vivo [12].

In the present study, the impact of three different preparation techniques on the physicochemical features of poly(d,l-lactide-*co*-glycolide) (PLGA) and polyethylene glycol-copolymer (PEG-PLGA) nanocarriers were compared. Nanocarriers loaded with sorafenib were manufactured using the nanoprecipitation, single emulsion, and double emulsion–solvent evaporation methods. Nanoparticles manufactured using the single emulsion–solvent evaporation method were further optimized with regards to the intended target product profile. In vitro cytotoxicity and cellular uptake were investigated in HepG2 cells. Finally, the nanoparticles were modified on their surface using a gadolinium complex, resulting in a theranostic nanocarrier system. For this purpose, the encapsulation of HSA into nanoparticles and the covalent modification of the surface using HSA or polyethylene imine were considered.

#### **2. Materials and Methods**

#### *2.1. Materials*

Resomer® RG 502H (PLGA, lactide: glycolide: 50:50, inherent viscosity: 0.16–0.24 dL/g), Resomer® RG 752H (PLGA, lactide:glycolide: 75:25 inherent viscosity 0.14–0.22 dL/g), and block copolymer Resomer® RGP d5055 (PEG-PLGA, PEG content: 3–7% (*m*/*m*), inherent viscosity: 0.93 dL/g) were obtained from Evonik Industries AG (Essen, Germany). PVA (*M*<sup>w</sup> = 30,000–70,000, 87–90% hydrolysed), polysorbate 80, Triton X-100, Pluronic F127, poly(methacrylic acid sodium salt) emulsifiers, dichloromethane (DCM), acetone, dimethyl sulfoxide (DMSO), sodium azide, D-trehalose dehydrate, mannitol, polyethyleneimine (PEI) (MW 25 kDa), 1-Ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDC), *N*-hydroxysuccinimide (NHS), Gd-DTPA, and HSA were obtained from Sigma Aldrich (St. Louis, MO, USA). Sorafenib (free base) was purchased from LC Laboratories (Woburn, MA, USA). Magnevist® was purchased from Bayer AG (Leverkusen, Germany). The micro bicinchoninic acid (μBCA) protein assay kit was bought from Pierce Biotechnology, Inc. (Waltham, MA, USA).

#### *2.2. Cell Culture Experiments in HepG2 Cells*

The human hepatoma cell line HepG2 was grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum (FCS), 100 U/mL penicillin, and 100 μg/mL streptomycin. The cells were cultured at 37 ◦C in a humidified atmosphere containing 5% carbon dioxide. The cells were trypsinised, resuspended, and precultured before use.

#### *2.3. Preparation of Nanoparticles Using Nanoprecipitation*

In brief, 5 to 10 mg of Resomer® RG 502 H, Resomer® RG 752H or Resomer® RGP d5055 and 0.5 to 4 mg of sorafenib were dissolved in between 0.5 and 1.0 mL of acetone under magnetic stirring. A water phase with a volume of 2.0 to 4.0 mL was composed of an aqueous solution (0.5 to 2.0% *w*/*v*) of emulsifying agent (polysorbate 80, poly(methacrylic acid sodium salt), Triton X-100, or Pluronic® F127), and added as a one-shot to the organic phase. Afterwards, the organic solvent was evaporated over a time period of 12 h at room temperature and 1 bar under constant stirring. The nanoparticles were centrifuged (Eppendorf 5424 R, Hamburg, Germany) at 37,565 *g* for 25 min, washed thrice and redispersed in an equal volume of purified water.

#### *2.4. Preparation of Nanoparticles Using the Single Emulsion Technique*

For the preparation of nanoparticles with the single emulsion–solvent evaporation method, the organic phase was prepared in two steps. An measure of 1 to 2 mg sorafenib was dissolved in 0.1 to 0.2 mL of acetone, and this solution was poured into a solution comprising 5 to 20 mg of Resomer® RG 502 H, Resomer® RG 752H, or Resomer® RGP d5055 in 1 to 2 mL DCM.

A volume of 4 to 8 mL of an aqueous solution of PVA (1 to 2% *w*/*v*) was added and sonicated using a Sonoplus HD2070, MS73 probe (Bandelin, Berlin, Germany) at an amplitude of 10% for 60 s. The organic solvent was evaporated over 2 h at atmospheric pressure and room temperature. The nanoparticles were purified as described above.

#### *2.5. Preparation of Nanoparticles Using Double Emulsion–Solvent Evaporation Technique*

The double emulsion–solvent evaporation technique was tested for the co-encapsulation of HSA into the sorafenib nanoparticles. The entrapped protein was further used for the covalent modification of the particle surface using the Gd-DTPA complex. In principle, the inner water phase was formed of 2.5 mg HSA in a volume of 0.1 mL of purified water. This solution was added to the organic phase, composed of 15 mg of Resomer® RG 752H or Resomer® RGP d5055 dissolved in 1.5 mL DCM, combined with 1.5 mg of sorafenib dissolved in 0.15 mL acetone. The first emulsification was performed by sonication using a Sonoplus HD2070, MS73 probe (Bandelin, Berlin, Germany) at an amplitude of 10% for 30 s. Afterwards, the water-in-oil emulsion was pipetted into 6.0 mL of an aqueous solution of PVA (1% (*w*/*v*)). A water-in-oil-in-water emulsion was formed by a second sonication step at an amplitude 15% for 45 s. The organic solvents were evaporated over a time period of 2 h under magnetic stirring at atmospheric pressure and room temperature. The nanoparticles were centrifuged at 37,565 *g* for 25 min (Eppendorf Centrifuge 5424 R), washed thrice and redispersed in 0.5 mL of phosphate buffer (pH 8), and, after gravimetric analysis, the suspensions were diluted to 20 mg·mL−<sup>1</sup> nanoparticle concentration.

#### *2.6. Particle Morphology and Particle Size Analysis*

The morphology of the nanospheres was investigated after centrifugation and redispersion in the described medium. The samples were examined using a FEI Talos F200XG2 high-resolution analytical microscope operated at 200 keV (Thermo Fischer Scientific, Waltham, MA, USA).

Additionally, the particle size and size distribution were determined using a Zetasizer Nano ZS (Malvern Instruments, Malvern, UK) equipped with a backscatter detector at an angle of 173◦. The particles were characterized for their intensity mean diameter and polydispersity index (PDI). Dynamic light scattering (also known as photon correlation spectroscopy) is based on the Brownian motion of particles dispersed in a liquid. The particle diameter is calculated from the intensity fluctuations of light scattered at the particle surface. The Zetasizer Nano ZS uses a HeNe gas laser to generate a signal of these intensity fluctuations, from which the size is calculated by applying the Stokes–Einstein equation. Consequently, the resulting size distribution is an intensity distribution.

#### *2.7. Storage Stability of Nanoparticle Formulations*

To improve the physical stability of the colloidal dispersion during storage, the nanoparticles were freeze-dried, and their storage stability was investigated after 6 months of storage. A concentration of 3% (*w*/*v*) of two lyoprotectors (trehalose dihydrate or mannitol) was evaluated. The solid concentration was adjusted to 7 mg·mL−<sup>1</sup> of nanoparticles and the samples were put into a Christ Epsilon 2–7 freeze dryer (Martin Christ GmbH, Osterode am Harz, Germany). The lyophilisation was conducted in two drying steps.

Initially, the temperature was decreased to −60 ◦C for 1 h to freeze the samples. Afterwards, primary drying was initiated by evacuating the chamber to 0.94 mbar. In parallel, temperature was raised to −30 ◦C over a time period of 150 min. The pressure was then reduced to 0.006 mbar and the temperature was increased to −10 ◦C over a time period of 60 min. These conditions were maintained for 35 h. The second drying was accomplished at a temperature of 10 ◦C for 60 min and at 20 ◦C for 10 h. The freeze dried samples were stored at 4 ◦C for a total duration of 6 months, and reconstituted in the same volume of purified water. The nanocomposite size and size distribution after redispersion were characterized by triplicated measurement.

#### *2.8. Nanoparticle Yield and Encapsulation E*ffi*ciency*

The nanoparticle yield was determined by microgravimetry. The drug loading and encapsulation efficiency were investigated dissolving 10 mg nanoparticles in 1 mL of DMSO. The solution was diluted using DMSO to be within the detectable linear calibration range (1–20 μg/L). The absorbance of the solutions was measured spectrophotometrically (Hitachi U-3000, Tokyo, Japan) at 285 nm. A concentration range of HSA solution in DMSO was prepared in order to correct the absorbance of HSA-loaded nanoparticles, also measured at 285 nm.

#### *2.9. Biorelevant In Vitro Drug Release Test Using the Centrifugation Method*

The biorelevant in vitro drug release test was conducted using the centrifugation method. In brief, 1.5 mg of the sorafenib-loaded nanocomposites was re-suspended in 1 mL of human blood plasma containing 0.03% sodium azide as a preservative. The nanoparticle formulations in release medium were filled into 2 mL tubes, and incubated at a temperature of 37 ◦C in a Thermomixer (Eppendorf, Hamburg, Germany) for 12 days at 700 rpm. At predetermined time points, 0.2 mL samples were collected. After each sampling time point the medium was replenished. The nanoparticles were separated from the plasma by centrifugation (Eppendorf Centrifuge 5424 R, 20 min at 37,565 *g*). The concentration of free sorafenib was determined using direct quantification of the drug remaining in the particle system. For this purpose, the nanoparticle pellets were dissolved in DMSO and the drug amount was detected spectrophotometrically.

#### *2.10. Surface Modification Using Human Serum Albumin and Polyethyleneimine*

The nanoparticle dispersions prepared by single emulsion technique were centrifuged and redispersed in phosphate buffer (pH 8), resulting in a nanoparticle concentration of 20 mg·mL<sup>−</sup>1. To increase the number of free amino groups on the particle surface, HSA or PEI were covalently bound to the nanoparticle surface. A 50-fold molar excess of EDC and the same excess of NHS, both calculated to the molar polymer concentration, were dissolved in 0.5 mL phosphate buffer (pH 8) and incubated for 60 min, centrifuged and washed three times, and redispersed in 1.5 mL phosphate buffer (pH 8).

The obtained carbodiimide-activated nanoparticle dispersion was added into 0.5 mL phosphate buffer (pH 8) solution containing equimolar amounts of HSA or PEI, and shaken overnight at 20 ◦C in an Eppendorf Thermomixer (Hamburg, Germany) at 700 rpm. Afterwards, the nanoparticles were dialysed for 2 h against 400 mL of purified water, using a 100 kDa membrane to remove residues of the dissolved polymers HSA or PEI, respectively. The dialysis step was repeated with purified water.

#### *2.11. Conjugation of Nanoparticles Using the Gd-DTPA Complex*

The Gd-DTPA complex was covalently attached to either HSA or PEI after activation using the carbodiimide. In order to activate the carboxyl groups of the gadolinium complex, a 10-fold molar excess of Gd-DTPA (calculated on the amount of the Resomer®) was dissolved in 1 mL of buffer (pH 8) and combined with a 5-fold molar excess of EDC and a 5-fold molar excess of NHS (calculated on the amount of Gd-DTPA). The resulting solution was incubated for 50 min and added to the purified nanoparticle suspension overnight. The preparation was purified by dialysis twice, using a membrane with a MWCO of 3.5 kDa and 500 mL of purified water for 2 h. The dialysed nanoparticle suspension was centrifuged and redispersed in 1.5 mL phosphate buffer.

#### *2.12. Quantification of Human Serum Albumin Using the Micro Bicinchoninic Acid Method*

The crosslinked or co-encapsulated HSA content of the nanocomposites was determined by the μBCA method after centrifuging 0.1 mL nanoparticle dispersion and removing the supernatant, while the pellet was dissolved in 0.5 mL DMSO. This solution was diluted 10-fold with purified water, and incubated for 1 h at 60 ◦C with the same volume of freshly prepared μBCA reagents mixture. DMSO was added to the calibrating HSA solution with the same ratio. After cooling the colored mixtures to room temperature for 20 min, their absorbance was evaluated by spectrophotometry at 562 nm (Hitachi U-3000, Tokyo, Japan).

#### *2.13. Quantification of Gadolinium Using Inductively Coupled Plasma Optical Emission Spectroscopy*

The Gd concentration was measured using a Spectro Genesis ICP-OES (Kleve, Germany) simultaneous spectrometer with axial plasma observation. Multielemental standards (Merck, standard solutions for ICP, Darmstadt, Germany) were used for calibration. The limits of detection of the element were calculated according to Equation (1):

$$\text{Limit of detection} = \text{Background signal} + \text{3} \times \text{SD}\_{\text{Background}} \times f\_{\text{dilution}} \tag{1}$$

The purified nanoparticles were dissolved in 5 M hydrochloric acid and diluted to the desired calibration range.

#### *2.14. In Vitro Investigation of Diagnostic Features by Magnetic Resonance Imaging*

In vitro MRI was carried out using box analysis to compare the contrast achieved with the nanoparticle suspension with that of Magnevist® (Gd-DTPA complex with meglumine) solution. A calibration was achieved by diluting Magnevist® to 0.01–2.5 mg·mL−<sup>1</sup> concentration. MR imaging was performed on a 3.0-T scanner (Siemens Magnetom Trio, Siemens Medical Solutions, Erlangen, Germany). The measurement conditions were T1-weighted 3D gradient echo sequences (fast low-angle shot) with the following parameters: TE (echo time) = 3.31 ms, TR (repetition time) = 8.67 ms, field of view = 100 × 78 mm, matrix acquisition = 640 × 480, slice thickness = 0.3 mm, flip angle = 16◦, fat suppression = fat saturated, and bandwidth = 180 Hz/Px.

#### *2.15. Labeling of Nanoparticles with a Fluorescent Dye*

A volume of 1.0 mL of the nanoparticle suspension (12 mg·mL<sup>−</sup>1) in phosphate buffer (pH 8) was added to 0.1 mL phosphate buffer (pH 8) containing a 25-fold molar excess of EDC and NHS (calculated on the amount of Resomer®) and incubated for 60 min, centrifuged, and washed, and redispersed in 1.0 mL phosphate buffer (pH 8). The obtained carbodiimide activated nanoparticle dispersion was given to 0.1 mL phosphate buffer (pH 8) solution containing 1 mg·mL−<sup>1</sup> Cyanine5 amine fluorescent (Cy5) dye, and shaken for 1 h at 20 ◦C in an Eppendorf Thermomixer (Hamburg, Germany) at 700 rpm. The nanoparticle dispersion was then centrifuged in an Eppendorf Centrifuge 5424 R (Hamburg, Germany) at 37,565 *g* for 25 min, washed three times, and redispersed in phosphate-buffered saline to a nanoparticle concentration of 10 mg·mL<sup>−</sup>1.

#### *2.16. In Vitro Cellular Uptake and Cytotoxicity*

Cellular uptake of the nanoparticles into the HepG2 cells was evaluated using flow cytometry. The cells were cultured in 24-well plates at a cell density of 2 <sup>×</sup> 105 cells per well at 37 ◦C and 5% CO2 for 24 h. After cultivation, 100 μg nanoparticles/well (10 mg/mL nanoparticle suspension was diluted to 100 μg/mL) was pipetted to the cells and incubated for 24 h. The cells grown without nanoparticles were used as control. The cells were washed in phosphate-buffered saline (PBS), trypsinised, and redispersed in PBS containing 2% (*m*/*v*) of bovine serum albumin. Flow cytometry was performed on a Cytoflex S cytometer (Beckman Coulter, Brea, CA, USA).

To further analyze cellular uptake and intracellular localization, fluorescence microscopy of HepG2 cells plated on cover slides and incubated with either Cy5-labeled PLGA or PEG-PLGA NP for 4 h and 24 h was employed. Membrane staining was performed by using Alexa488 concanavalin A (2.5 μg/mL; Thermo Fisher Scientific, Schwerte, Germany) and nuclei were counterstained with 4',6-diamidino-2-phenyl-indole (DAPI) solution (Merck, Darmstadt, Germany). Finally, slides were mounted with Vectashield mounting medium (Biozol, Eching, Germany) and images were obtained using an AxioImager Z1 microscope and Axiovision 4.6 software (Carl Zeiss, Jena, Germany).

The in vitro cytotoxicity in HepG2 cells was determined using 3-(4,5-dimethylthiazol-2-yl)-2,5 diphenyltetrazolium bromide (MTT) assay. Cells were seeded (50,000 cells/well) in 96-well plates. At 24 h pre-incubation, the media were replaced with 100 μL of fresh Dulbecco's Modified Eagle's Medium (DMEM) containing 10% FBS and sorafenib-loaded nanoparticles. Three different sorafenib concentrations (6.25 μg/mL, 12.5 μg/mL, and 25 μg/mL) were used, while the control samples contained the same amounts of free sorafenib in DMSO solution. After 24 h of incubation, a volume of 10 μL per well of MTT solution (5 mg MTT/mL) was added, followed by further incubation for 2 h. The supernatant was removed, and 0.2 mL MTT lysis solution was added into each well. The absorbance of cell suspension was determined at 595 nm using a spectrophotometer (EnVision 2104 Multilabel Reader, Perkin Elmer, Waltham, MA, USA). The percentage of viable cells was calculated by comparing the absorbance of treated cells against the untreated cells (negative control). The DMSO solution and the blank nanoparticle suspensions served as positive controls. The data are presented as mean and standard deviation with five replicates.

#### *2.17. Statistics*

All data are expressed as the mean value ± standard deviation (SD), which were calculated and plotted using Microsoft Excel (Microsoft, Redmond, WA, USA) and SigmaPlot 11.0 (Systat Software GmbH, Erkrath, Germany), respectively. All nanoparticle formulations were produced as three batches (*n* = 3).

#### **3. Results and Discussion**

In recent years, a variety of preparation methods have been evaluated for the synthesis of nanocarrier devices. The current study produced advanced theranostic drug carriers and compared the impact of manufacturing technology and surface modification on their physicochemical properties and in vitro features.

Initially, a particle size between 100 and 300 nm [17], a zeta potential of more than −15 mV [18,19], and a high drug load and particle yield were identified as key criteria for formulation development. Due to the enhanced permeability and retention (EPR) effect, these nanoparticles may be capable of targeting tumor tissues [17]. While smaller nanoparticles can be rapidly excreted by the kidneys, larger colloids with a size of more than 300 nm are quickly recognized by the macrophages of the reticuloendothelial system [11]. Among other aspects, the encapsulation efficiency and particle size play an important role in nanocarrier delivery. Selecting biodegradable polymers of the Resomer® type, three different techniques for the encapsulation of sorafenib were compared (Figure 1).

**Figure 1.** Scheme of the nanoparticle preparations with the preparation technique, materials used, and surface functionalization (from the left to the right). Each preparation technique was evaluated with the presented materials. Surface coating was undertaken for preparations manufactured by single emulsion technique only.

#### *3.1. Manufacture of Sorafenib-Loaded Core Particles Using Nanoprecipitation Technique*

The manufacture of sorafenib-loaded nanoparticles by nanoprecipitation using Triton X-100 or Pluronic® F127 stabilizers in aqueous solution resulted in a pronounced aggregation for all three polymers. Similar observations have been made at the medium scale, suggesting poor 'scalability' during the later stages of production [8]. Changing the stabilizer to polysorbate 80, similar aggregation occurred with Resomer® RG 502H and Resomer® RG 752H, respectively.

The use of Resomer® RGP d5055 (and polysorbate 80) resulted in nanoparticles within the desirable size range (153 ± 14 nm, Figure 2, upper micrographs). However, the PDI of more than 0.27, an encapsulation efficiency below 20%, and a particle yield between 20 and 40% were the major disadvantages of this formulation design. A lower density of the particle system due to the hydrophilic side chains of the polymer is the most likely explanation.

**Figure 2.** Scanning/transmission electron micrographs of nanoparticles comprising sorafenib in a Resomer® RG 752H (**upper left**) or Resomer® RGP d5055 (**upper right**) matrix prepared by nanoprecipitation, in a Resomer® RG 752H (**middle left**) and Resomer®, RGP d5055 (**middle right**) matrix prepared by single emulsion–solvent evaporation technique using polyvinyl alcohol (PVA) as a stabilizer, or in a Resomer® RG 752H-HAS (**lower left**) and Resomer®, RGP d5055-HSA (**lower right**) matrix by double emulsion–solvent evaporation technique using PVA as a stabilizer.

In comparison, the nanoprecipitation technique using PVA in combination with Resomer® RG 502H or Resomer® RGP d5055 led to particle systems broadly distributed in size, as indicated by elevated polydispersity indices ranging between 0.27 and 0.46. This was also confirmed by the intensity distributions, exhibiting a second and third fraction of larger particles in the micrometer range (Figure 3). The zeta potential of sorafenib-loaded Resomer® RG 502H or Resomer® RGP d5055 nanocomposites by nanoprecipitation and PVA emulsifier was found to be 13.2 ± 0.7 mV and 12.1 ± 1.2 mV, which also explains the poor stability resulting in a pronounced aggregation. Under similar conditions, crystallization of the drug accompanied by an increased PDI between 0.21 and 0.35 was reported by Lin et al. [18]. When applying the nanoprecipitation technique, a particle size in the desired range (196 ± 10 nm) as well as a reduced PDI of 0.21 ± 0.03, was achieved when using a polymer concentration of 10 mg·mL−<sup>1</sup> of Resomer® RG 752H and 1 mg·mL−<sup>1</sup> of sorafenib in the acetone phase.

Considering the difficulties in nanoparticle preparation when using the nanoprecipitation method, later efforts were focused on nanoparticle manufacture by single and double emulsion techniques.

**Figure 3.** Size distribution by intensity of nanocomposites containing sorafenib in Resomer® RG 502H (green line) and Resomer® RGP d5055 (red line) matrix nanoparticles prepared by nanoprecipitation, using PVA as an emulsifier.

#### *3.2. Manufacture of Sorafenib-Loaded Core Particles by Single Emulsion Technique*

For the preparation of nanoparticles utilizing the single emulsion method, the hydrophilic phase was composed of a 1% or 2% (*w*/*v*) aqueous solution of PVA in purified water (Table 1). A polymer concentration of 10 mg·mL<sup>−</sup>1, an initial drug amount of 10% (*m*/*m*), and a water-to-dichloromethane (DCM) ratio of 4:1 led to acceptable properties for each of the three polymers (Table 1). The morphology of the nanoparticles was tested by scanning/transmission electron microscopy. They were of spherical shape and within the expected size range (Figure 4).

**Table 1.** Properties of Resomer–sorafenib nanoparticles as a function of the encapsulating polymer and emulsifier (PVA) concentration.


**Figure 4.** Size distribution by intensity of sorafenib-containing nanoparticles prepared from Resomer® RG 502H (black line) and Resomer® RG 752H (red line) polylactic-*co*-glycolic acid (PLGA) polymers as well as Resomer® RGP d5055 (green line) by single emulsion method.

With an encapsulation efficiency ranging between 70.4 and 78.8% (Table 1) and a monomodal size distribution (Figure 4), the emulsion method was superior compared to the nanoprecipitation technique. The drug loading (8.9–12.0% *m*/*m*) almost reached the initial drug ratio (10% *m*/*m*) (Table 1). Further increase of the drug amount from 10% (*m*/*m*) to 20% (*m*/*m*) resulted in elevated PDI values due to the formation of larger aggregates.

The best outcomes were achieved by using Resomer® RG 752H or Resomer® RGP d5055 as matrices (Figure 2, middle micrographs). Resomer® RG 752H resulted in a particle yield of 73.7%, an encapsulation efficiency of 76.6%, and a drug load of 11.2% (Figure 5). For Resomer® RGP d5055, similar promising results were achieved. A particle yield of 76.1%, an encapsulation efficiency of 75.2%, and a drug load of 8.9% were obtained using a 2% *m*/*v* emulsifier concentration. These nanocomposites were selected for further surface modification and cellular uptake and cytotoxicity studies (Figure 5). The zeta potential indicated high colloidal stability, which was also confirmed during further processing. It is likely that the surface charge resulted from the high number of carboxyl groups present in the PLGA polymers. However, there were no major differences between the zeta potential values of the employed polymers even when the nanocarriers were manufactured from the PEGylated derivative. This indicates that PEGylation did not significantly reduce the number of the carboxyl groups. These nanoparticle preparations were further processed and evaluated for their release behavior.

**Figure 5.** Selected formulations for surface modification and further tests.

#### *3.3. Co-Encapsulation of HSA Using Double Emulsion Method or Surface Modification*

To covalently bind the Gd-DTPA complex to the particle surface, amino groups were introduced into the polymeric matrices. In a first approach, HSA was co-encapsulated into the nanoparticles using the double emulsion–solvent evaporation method. Although there was no significant difference between the diameters of nanoparticles manufactured from Resomer RG 752H (mean diameter 210.6 nm, PDI 0.113) and Resomer®, RGP d5055 polymers (mean particle diameter 210.3 nm, PDI 0.099), incorporation of the protein resulted in smaller particle sizes than the single emulsion method. The detected size range was confirmed by electron microscopy (Figure 2, lower micrographs). All nanocomposites were of spherical shape (Figure 3). The particle yield was much lower compared to the methods described earlier. A particle yield of 50.4 <sup>±</sup> 1.0% for Resomer® RG 752H and 49.0 <sup>±</sup> 5.3% for Resomer® RGP d5055 was achieved. A similar amount of HSA was incorporated into both polymers (Resomer® RG 752H 10.6 <sup>±</sup> 1.1%, Resomer® RGP d5055, 9.4 <sup>±</sup> 1.2%). The formation of smaller particles with reduced density could be responsible for the decreased particle yield.

However, the incorporation method resulted in a high protein loading and an increased number of functional groups available for the EDC reaction. In comparison, an amount of 1.0 ± 0.3% HSA for Resomer® RG 752H and 1.9 <sup>±</sup> 0.5% HSA for Resomer® RGP d5055 was bound to the particles using the surface coating technique.

#### *3.4. Surface Modification with Gadolinium*

The Gd-DTPA complex was conjugated to the amino groups of the HSA molecules after incorporation or covalent binding of the protein to Resomer® RG 752H and Resomer® RGP d5055 nanoparticles, respectively. As expected, the modification of nanocomposites with Gd-DTPA was limited by the availability of functional groups on the particle surface.

The nanoparticles modified on their surface with HSA were characterized by poor protein binding, with an amount of approximately 1.5 mg gadolinium per g Resomer® RG 752H, and 1.4 mg gadolinium per g Resomer® RGP d5055. Nanoparticles comprising HSA as part of their matrix structure exhibited a much higher binding of 2.3 mg gadolinium per g Resomer® RG 752H and 3.2 mg gadolinium per g Resomer® RGP d5055.

Nevertheless, further increase of the gadolinium content was achieved using a method described previously [20] with some modification. The binding of PEI to the surface resulted in a significant increase of the gadolinium content, with 15.7 mg gadolinium per g Resomer® RG 752H and 10.7 mg gadolinium per g Resomer® RGP d5055.

#### *3.5. Evaluation of Contrast Signal Using In Vitro MRI*

The MRI properties of theranostic nanocarriers were investigated in vitro. Magnevist® aqueous solutions were used as a reference. There was a linear correlation (*R*<sup>2</sup> = 0.9724) between the MRI signal and the gadolinium content in the range between 0.02 and 0.3 mg·mL−<sup>1</sup> (with MRI intensities of 470–670 at 0.02 mg·mL−1, and 1520–1750 at 0.3 mg·mL−1). For nanoparticles exhibiting the highest content of gadolinium, the theoretical particle concentration (calculated from a maximum injectable particle concentration of 10 mg·mL−<sup>1</sup> and the gadolinium load) fell into this calibration range. The signal intensities measured for the T1-weighed MRI were in good accordance with the inductively coupled plasma optical emission spectroscopy (ICP-OES). Earlier studies reported a loading of 4.7 to 8.5 mg gadolinium per g HSA nanoparticles after covalent binding of DTPA. This amount was found to be considerably high for in vivo MRI studies [5].

#### *3.6. Storage Stability of Theranostic Nanocomposite Formulations*

To freeze dry Resomer® RG 752H and Resomer® RGP d5055 nanoparticles manufactured by the single emulsion method, a concentration of 3% (*w*/*v*) of the lyoprotectors sucrose, trehalose, and mannitol has been previously applied [21]. The sorafenib-loaded nanoparticles with a drug load of approximately 10% were freeze dried in the presence of 3% (*w*/*v*) of trehalose or mannitol. The nanospheres remained stable over the time of storage. An increase in the PDI values was observed for formulations freeze-dried with mannitol. On this basis, trehalose at a concentration of 3% was identified to be the optimal lyoprotector for the developed particle system (Figure 6).

**Figure 6.** Size distribution by intensity of Resomer® RGP d5055 nanoparticles after preparation (red line), and following reconstitution after freeze-drying in the presence of 3% of trehalose (blue line) and 3% of mannitol (yellow line).

#### *3.7. Evaluation of In Vitro Drug Release of Sorafenib from Theranostic Nanocomposites*

For nanoparticles comprising polymers from the Resomer® family, a biphasic release pattern has been previously reported [22]. Preliminary investigations indicated a less pronounced burst release for Resomer® RG 752H compared to Resomer® RG 502H. After the initial release phase, a sustained release behavior was observed.

Consequently, the drug release of sorafenib-loaded Resomer®RGP d5055 and Resomer® RG 752H particles was investigated over a time period of 12 days in human blood plasma. Both of the nanocomposites were prepared under optimized conditions, that is, using the single emulsion method with an encapsulating polymer concentration of 10 mg·mL−1, an initial drug amount of 10% (*m*/*m*), a water-to-DCM ratio of 4:1, and 2% (*w*/*v*) of PVA. The initial burst release was higher with nanoparticles manufactured from Resomer® RG 752H polymer (18.2 <sup>±</sup> 2.9%) compared to Resomer®RGP d5055 copolymer, from which 8.8 <sup>±</sup> 2.1% of the drug was released during the first hour. Afterwards, a continuous release was observed for both composites (Figure 7). Surprisingly, the release from Resomer®RGP d5055 was substantially slower, reaching a plateau at 50.6 <sup>±</sup> 9.2%. A specific interaction of sorafenib with the hydrophilic side chain of the block copolymer could be responsible for this behavior.

**Figure 7.** Sorafenib release from Resomer® RG 752H (green line) and Resomer® RGP d5055 -sorafenib (PEG-PLGA-SFB) nanocomposites (blue line) in human blood plasma. Data are presented as mean ± SD from three independent samples for each concentration.

#### *3.8. Cellular Uptake and Cytotoxicity*

For the cellular uptake and cytotoxicity studies, sorafenib-loaded Resomer®RGP d5055 and Resomer® RG 752H nanocomposites were prepared by the single emulsion method at a polymer concentration of 10 mg·mL<sup>−</sup>1, an initial drug amount of 10% (*m*/*m*), a water-to-DCM ratio of 4:1, and 2% (*w*/*v*) of PVA. The internalization rates of sorafenib-containing nanoparticles in HepG2 cells were quantified by flow cytometry. Additionally, fluorescence microscopy was conducted to confirm the localization of the particles inside the cells (Figure 8).

**Figure 8.** Cellular uptake of either Cy5-conjugated Resomer® 752H or Resomer® RGP d5055 nanoparticles (red). Membrane staining was performed by using Alexa488 concanavalin A (green). Nuclei were stained with 4 ,6-diamidino-2-phenylindole (DAPIblue).

As expected, the uptake of drug-loaded nanoparticles prepared by using the PEGylated polymer was significantly lower (3.9 <sup>±</sup> 3.3%, Figure 9B) than for Resomer® RG 752H nanoparticles (49.9 <sup>±</sup> 4.9%, Figure 9C). The biocompatibility of nanomedicines is an important feature with regards to clinical success and commercialization.

**Figure 9.** Flow cytometry diagrams of untreated cell control (**A**), sorafenib-loaded Resomer®RGP d5055 (**B**), and Resomer® RG 752H nanocomposites (**C**).

The cytotoxicity of the unloaded and the drug-loaded nanoparticles was investigated in HepG2 cells. Solutions of the drug in dimethyl sulfoxide (DMSO), the organic solvent alone, and an untreated negative control were used as references. As shown in Figure 10, the viability of the cells remained at approximately 90% when exposed to 100 μg/well of both types of unloaded nanoparticles (50,000 cells/well). The non-toxic features of Resomer® polymers have been verified previously [4].

**Figure 10.** Viability of HepG2 cells treated with different concentrations of sorafenib (SFB) or Resomer® RG 752H or Resomer® RGP d5055 nanoparticles, and that of untreated cells (negative control).

As expected, sorafenib exhibited a concentration-dependent cytotoxicity in the HepG2 cells. Similar concentrations of sorafenib resulted in higher cytotoxicity for the free drug compared to the particle formulations (Figure 10), which could have been due to the slow release of the drug from the particle matrix.

#### **4. Conclusions**

Theranostic nanocomposites comprising PLGA or PEG-PLGA were loaded with the anti-tumor drug sorafenib and modified on their surface with the contrast agent Gd-DTPA. The single emulsion technique was found to be the most appropriate method for the effective preparation of monodisperse nanocomposites. These nanospheres exhibited superior properties compared to the particle systems described in the literature. Finally, the top-down manufacture combined with a modification of the particle surface using PEI and Gd-DTPA resulted in a strong MRI signal, optimal stability features, and a sustained release profile. On this basis, further investigations will focus on the in vivo performance of these nanocarriers.

**Author Contributions:** T.F. designed and achieved most of the experiments except cellular studies, and wrote the paper. A.P. designed the cellular studies and took part in the manuscript preparation. T.P., C.S., D.D. and T.J.V. performed the cytotoxicity and cellular uptake studies by MTT and flow cytometry, respectively. S.H. and F.R. designed and did the cellular uptake investigations by fluorescence microscopy. M.G.W. participated in designing the experiments and writing the manuscript.

**Funding:** This research was funded by Alexander von Humboldt Foundation (Ref. No.: 3.3-UNG/1161203 STP and 3.3-1161203-HUN-HFST-E) and the Ministry of National Economy (Hungary) via the Economic Development and Innovation Operation Programme (project No. BIONANO\_GINOP-2.3.2-15-2016-00017).

**Acknowledgments:** The authors are grateful to the LOEWE initiative of the state of Hessen for financial support to the research center Translational Medicine and Pharmacology. Tivadar Feczkó acknowledges the funding of Alexander von Humboldt Foundation (Ref. No.: 3.3-UNG/1161203 STP and 3.3-1161203-HUN-HFST-E) and that of BIONANO\_GINOP-2.3.2-15-2016-00017 project. S/TEM studies were performed at the electron microscopy laboratory of the University of Pannonia, established using grant no. GINOP-2.3.3-15-2016-0009 from the European Structural and Investments Funds and the Hungarian Government.

**Conflicts of Interest:** The authors declare no conflicts of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **E**ff**ect of Size and Concentration of PLGA-PEG Nanoparticles on Activation and Aggregation of Washed Human Platelets**

**Rana Bakhaidar 1, Joshua Green 2, Khaled Alfahad 2, Shazia Samanani 2, Nabeehah Moollan 2, Sarah O'Neill <sup>1</sup> and Zebunnissa Ramtoola 1,\***


Received: 23 August 2019; Accepted: 26 September 2019; Published: 4 October 2019

**Abstract:** Nanotechnology is being increasingly utilised in medicine as diagnostics and for drug delivery and targeting. The small size and high surface area of nanoparticles (NPs), desirable properties that allow them to cross biological barriers, also offer potential for interaction with other cells and blood constituents, presenting possible safety risks. While NPs investigated are predominantly based on the biodegradable, biocompatible, and FDA approved poly-lactide-*co*-glycolide (PLGA) polymers, pro-aggregatory and antiplatelet effects have been reported for certain NPs. The potential for toxicity of PLGA based NPs remains to be examined. The aims of this study were to determine the impact of size-selected PLGA-PEG (PLGA-polyethylene glycol) NPs on platelet activation and aggregation. PLGA-PEG NPs of three average sizes of 112, 348, and 576 nm were formulated and their effect at concentrations of 0.0–2.2 mg/mL on the activation and aggregation of washed human platelets (WP) was examined. The results of this study show, for the first time, NPs of all sizes associated with the surface of platelets, with >50% binding, leading to possible internalisation. The NP-platelet interaction, however, did not lead to platelet aggregation nor inhibited aggregation of platelets induced by thrombin. The outcome of this study is promising, suggesting that these NPs could be potential carriers for targeted drug delivery to platelets.

**Keywords:** PLGA-PEG; nanoparticles; platelet; activation; aggregation; binding; uptake

#### **1. Introduction**

Biocompatible and biodegradable nanoparticles (NPs) continue to be extensively investigated for the diagnosis, prevention, and therapy of various diseases, for cellular and molecular imaging, and for tissue engineering applications [1–5]. NPs have been reported to target various therapeutic agents to the central nervous system (CNS), tumour sites, and the vascular compartment [6–10]. The small size and high surface area of NPs are desirable properties that allow them to cross biological barriers, but also offer potential for interaction and interference with other cells and blood constituents, presenting possible safety risks. While the NPs investigated are predominantly based on the biodegradable, biocompatible and FDA approved poly-lactide-*co*-glycolide (PLGA) polymers, irrespective of their intended destination, NPs introduced into the bloodstream will interact with endothelial cells as well as blood constituents. It is therefore important to understand the bio-distribution of such NPs in human blood and their interaction with various blood components such as platelets [6,11,12]. Previous studies report that grafting a polyethylene glycol (PEG) chain to a PLGA polymer backbone significantly increases the circulation half-life of the modified copolymer, as it remains camouflaged from the macrophage [13,14]. The longer duration of exposure to blood components can result in enhanced interaction potential with blood cells, and in particular with platelets, maximising the possibility of modulating platelet, activation, secretion, and aggregation [15,16].

NP physicochemical characteristics, such as size and net charge (zeta potential (ZP)) that are reported to be key factors enhancing their efficacy, can substantially affect their bio-distribution [17,18]. Variations in NP characteristics may result in a diverse range of platelet reactivity but unfortunately, only a few studies regarding the interaction of NPs with washed platelets (WPs) have been published [18–20]. We previously reported that biodegradable micro and NPs of PLGA, PLGA-PEG, and chitosan, of median diameter (D50%) of 2–9 μm and 100–500 nm, and with various surface morphology did not induce or inhibit platelet aggregation at particle concentrations of 0.1–500 μg/mL [20]. In contrast, silica NPs of 10 nm, 50 nm, and 150 nm, were shown to cause platelet hyper-aggregability, promoting the risk of thrombus deposition, while particles of 500 nm did not pose such a risk [21]. Mixed carbon NPs, carbon nanotubes, single-walled nanotubes, and multi-walled nanotubes were reported to stimulate platelet aggregation by the upregulation of GPIIb/IIIa receptors on platelets, inducing subsequent vascular thrombosis [19]. In a separate study, the contribution of silver NPs to the effective inhibition of integrin-mediated WPs responses, such as aggregation, secretion, and adhesion to immobilized fibrinogen or collagen, was established [22].

The pro-aggregatory effects of some NPs and the antiplatelet effects of others raise questions regarding the safety of biodegradable nanomaterials in terms of interference with the haemostatic equilibrium [20,23]. To date, the potential interactions of PLGA-PEG NPs with blood elements remains scant. In particular, the effect of their size on blood constituents, as well as any potential hazard they may pose, have not been studied. Platelet activation is a precisely regulated event, critical for maintaining normal blood flow. When designing drug-loaded NPs intended to be delivered to the systemic circulation, a major consideration is to maintain platelets in an inactive state [22].

The aims of this study were to determine the impact of size-selected PLGA-PEG NPs on platelet activation and aggregation. Coumarin-6-labelled PLGA-PEG NPs of three average sizes of 112, 348, and 576 nm were formulated using the solvent dispersion technique [20]. The effect of NP size and NP concentration of 0.01–2.2 mg/mL on the activation and aggregation profiles of washed platelets (WP) were examined using platelet aggregation assays and flow cytometry. Confocal imaging was carried out on NPs incubated with WP to characterise the interaction of NPs with platelets in the resting and activated states.

#### **2. Materials and Methods**

#### *2.1. Materials*

Albumin, from human serum, Coumarin-6 (98%), Polysorbate 80 (Tween® 80), Dextrose, Sodium Chloride (NaCl), Sodium bicarbonate (NaHCO3), Potassium Chloride (KCl), Sodium Citrate (Tribasic, 10 dehydrate (HOC(COONa)(CH2COONa)2, KH2PO4 (monobasic potassium phosphate anhydrate or potassium dihydrogen phosphate), Calcium Chloride (CaCl2), Magnesium Chloride hexahydrate MgCl2·6H2O, and HEPES (N-2-Hydroxyethylpiperazine-N'-2-Ethanesulfonic Acid) were purchased from Sigma-Aldrich Dublin, Ireland. PLGA-PEG containing PEG at 10% *w*/*w* (poly-d, l-lactic-*co*-glycolic acid-polyethylene glycol diblock copolymer 50:50 mPEG; 33 kDa with inherent viscosity: 0.05–0.15 dL/g, Lakeshore Biomaterials™, was purchased from Evonik Industries AG, Essen, Germany. Deionised water was used throughout the experiments.

#### *2.2. Methods*

#### 2.2.1. Preparation and Characterisation of PLGA-PEG NPs

PLGA-PEG NPs containing a fluorescent marker, coumarin-6, were prepared using the solvent dispersion method [24]. Briefly, PLGA-PEG polymer was dissolved in acetone to form 10, 55, and 100 mg/mL solutions, and coumarin-6 at 0.05% *w*/*w* of the polymer was dissolved in the PLGA-PEG solutions. The polymer solutions were added dropwise to an external aqueous phase containing Tween 80® at 2% *w*/*v*, under constant stirring. The NPs formed were recovered by centrifugation at 11,000 rpm for 20 min (Rotina 35 R centrifuge, Hettich Zentrifugen, Tuttilingen, Germany). NPs were washed with deionised water, centrifuged, and the pellet resuspended in deionised water and stored at 4 ◦C. The average particle size (PS), polydispersity index (PDI), and zeta potential (ZP) of the NPs were determined by dynamic light scattering technique, using a Malvern Zetasizer Nano ZS 90 (Malvern Instruments, Worcestershire, UK). After centrifugation, NPs were suspended in deionised water, placed in disposable voltable zeta cells (DTS 1060) for analysis using DLS. Viscosity and refractive index of the dispersants were taken into account. For each sample, the average of five measurements was calculated. Results were expressed as an average of five measurements ± standard deviation.

The morphologies of NPs were examined under a MIRA3 variable pressure field emission scanning electron microscope (Tescan, Brno-Kohoutovice, Czech Republic) at an accelerating voltage of 5.0 kV and magnification of 30,000 to 50,000×. Samples of lyophilised NPs were applied onto aluminium stubs using a double-sided conductive tape and were sputter-coated with gold.

#### 2.2.2. Preparation of Washed Platelets

Venous whole blood was obtained from healthy human volunteers, free from aspirin and other non-steroidal anti-inflammatory agents for the previous 7–10 days. Ethical approval was obtained from the Research Ethics Committee, Royal College of Surgeons in Ireland. The blood obtained was centrifuged at 180 g for 12 min, to separate the platelet-rich plasma (PRP) that was acidified to pH 6.5 and the plasma pelleted by centrifugation at 720 g for 12 min. Washed platelets were then adjusted to approximately 250 <sup>×</sup> 10<sup>9</sup> platelets per millilitre by resuspending the pellets in HEPES (N-2-Hydroxyethylpiperazine-N'-2-Ethanesulfonic Acid) platelet buffer and leaving to stand for 30 min before adding calcium chloride (CaCl2) at a final concentration of 1.8 mM.

#### 2.2.3. Effect of Size and Concentration of NPs on Platelet Aggregation

Platelet aggregation was determined by measuring the change in the optical density of stirred WPs in the absence or presence of 0–2.2 mg/mL PLGA-PEG NPs of different sizes, following the addition of the platelet agonist, thrombin. Platelets were incubated with NPs for 4 min prior to the addition of 0.1 U/mL thrombin. Platelet aggregation data was analysed using an eight channel platelet aggregometer,-PAP-8 (Bio/Data Corporation, Horsham, PA, USA). Aggregation results were expressed as final percent aggregation (% PA) at the end of the reaction time of 12 min. Data are presented as mean of *n* = 4–6 ± SEM.

#### 2.2.4. Effect of Size and Concentration of NPs on Platelet Activation

WPs (250 <sup>×</sup> 103/μL) were incubated with 20 <sup>μ</sup>L phycoerythrin (PE)-labelled CD62P antibody and mixed with PLGA-PEG NPs of different sizes, in the absence and presence of 0.1 U/mL thrombin and incubated for 4 min at 37 ◦C. At predetermined time points, samples were fixed using 1% *v*/*v* formaldehyde (FA) and platelet activation was measured by flow cytometry. Platelet activation was determined by measuring the level and extent of CD62P antibody binding to the platelet surface, expressed as percent positive cells (% PP) and mean fluorescence intensity (MFI). Platelets with and without CD62P were used as negative controls and platelets activated with thrombin at 0.1 U/mL for 4 min was the positive control. Data are average of *n* > 3 ± SEM.

2.2.5. Confocal Microscopy of the Effect of Incubation Time on The Interaction of PLGA-PEG NPs with Washed Platelets

Coumarin-labelled NPs of 112, 348, and 576 nm, at 0.1 and 2.2 mg/mL, were incubated with resting WPs in suspension for 1, 5, 15, and 30 min. At each time point, samples were fixed with 1% formaldehyde, stained red with phalloidin-tetramethylrhodamine isothiocyanate (phalloidin-TRIT-C) and examined using confocal laser scanning microscopy (CLSM), (Carl Zeiss, Jena, Germany). WPs were also incubated with coumarin-labelled NPs of the three different sizes at 2.2 mg/mL for 1, 5, and 30 min, and allowed to attach to fibrinogen-coated glass slides (20 μg/mL). At each time point, samples were fixed with 1% formaldehyde, platelets stained red with phalloidin-TRITC, and examined by CLSM.

#### 2.2.6. Statistical Analysis

Statistical analysis was carried out using GraphPad Prism (version 7.00 for Windows; GraphPad Software, San Diego, CA, USA). One-way ANOVA and unpaired student T tests and One-way ANOVA followed by post hoc analysis using Dunnett's Tests were used to determine differences between samples and groups. A *p* value <0.05 was considered to be statistically significant.

#### **3. Results**

#### *3.1. Characterisation of NPs Formulated*

The size, PDI, and zeta potential of the NPs increased with increasing PLGA-PEG concentration (Table 1, Figure 1A–C). A PDI value of 0.10 indicating a nearly monosized distribution was observed for the smallest NPs of 112 nm, while larger PDI values of 0.54 and 0.70 were observed for the 348 and 576 nm NPs, respectively. A larger size distribution is expected for larger size NPs formulated using high polymer concentration. In Figure 1A, the SEM shows the ~100 nm NPs with a distinct spherical shapes and smooth surfaces, exhibiting a narrow range of sizes at different fields of view. In Figure 1B, the NPs of ~350 nm display an interparticular bridging/fusion. In Figure 1C, the NPs of ~600 nm average size maintain a spherical appearance similar to that for the NPs in Figure 1A. The SEMs did not show any internal or external porosity at different magnifications. Calculations by Image J software from SEM analysis showed an average size, which were lower than the sizes measured by DLS. SEM is a tool primarily utilised to analyse particle morphology. The obtained PS determined by microscopy techniques is usually considered the lower limits of PS [25].



PLGA: poly-lactide-*co*-glycolide; PEG: polyethylene glycol; PS: Particle size; PDI: Polydispersity Index, ZP: Zeta Potential.

**Figure 1.** Scanning Electron Microscope Images of PLGA-PEG NPs of average size of (**A**) 112nm, (**B**) 348 nm, and (**C**) 576 nm. Each SEM image is representative of all images taken.

The 112 nm NPs had a net negative charge of −22.20 ± 5.71 mV indicating good colloidal stability, whereas the two larger NPs displayed moderately stable colloidal suspensions, with surface charges from −12.40 ± 5.47 to −9.50 ± 14.56 mV. The ZP values of the larger NPs decreased to within the range of −16 to −23 mV on dilution to the concentrations used in the study. The stability of the NPs generated was examined over a 28-day time period to ensure stability over the time period of use. It was noted that storage-induced instability was minimal, particularly in the batches formed of NPs of small and mid-level particle diameters (~100 nm and ~350 nm), with no significant changes in particle properties. Particle PS and PDI properties were retained over this period of 28 days and for the largest size NPs over at least 14 days, revealing a reasonable stability profile for the NPs. This approach in examining the stability of NPs over time is supported by recent studies by Fornaguera and Solans [26].

#### *3.2. E*ff*ect of Size and Concentration of NPs on Platelet Aggregation*

Incubation of NPs with platelets for 4 min in absence of the agonist, thrombin, showed that, irrespective of the size and concentration of NPs (0.05–2.2 mg/mL), spontaneous platelet aggregation was not detected (0 ± 0 % PA, *p* > 0.05, *n* = 3–6). The addition of 0.1 U/mL thrombin, at 4 min in absence of NPs, resulted in a % PA of 72.25–74.75 (Table 2). In presence of 112 nm NPs, no significant difference in % PA was observed after 12 min incubation at all NP concentrations tested (Table 2). Similarly, larger NPs of 348 and 576 nm showed no significant effect on the % PA over the 12 min incubation at the lower NP concentrations of 0.01–1.0 mg/mL and 0.01–1.5 mg/mL for the 348 and 576 nm NPs, respectively (Table 2). At the higher NP concentrations of 1.5 and 2.2 mg/mL, the % PA was significantly lower at 28.40 ± 3.47 and 27 ± 3.10, respectively, when incubated with 348 nm NPs (*p* value < 0.001). Similarly, at the highest concentration of 2.2 mg/mL, NPs of 576 nm resulted in a significantly reduced % PA of 49.00 ± 5.18 (*p* < 0.05) (Table 2).

**Table 2.** Effect of size and concentration of PLGA-PEG NPs on percent platelet aggregation (% PA). Data are the average of *n* = 4–6 experiments ± SEM. \* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001 compared with 0.1 U/mL thrombin (control), one-way ANOVA, and unpaired student *t*-test, at each time point.


Examination of the platelet aggregation time profile showed no delay in platelet aggregation for the smallest NPs of 112 nm at any of the concentrations tested (Figure 2A). However, the reduced % PA observed at the higher NP concentrations of 348 and 576 nm NPs was associated with a significant delay in platelet aggregation at the earlier time points following addition of thrombin to the sample. Platelet aggregation proceeded significantly more slowly for the 348 and 576 nm NPs at the higher NP concentrations of 1.5 and 2.2 mg/mL (Figure 2B,C), respectively.

**Figure 2.** Effect of PLGA-PEG NPs of (**A**) 112 nm, (**B**) 348 nm, and (**C**) 576 nm at 0.05, 1.5, and 2.2 mg/mL on platelet aggregation profile for 4 min following the addition of thrombin. Data are the average of *n* = 4–6 experiments ± SEM. \* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001 compared with 0.1 U/mL thrombin (control), one-way ANOVA, and unpaired student *t*-test, at each time point.

#### *3.3. E*ff*ect of Size of NPs on Platelet Activation Profile*

The percent positive platelets (% PP) and mean fluorescence intensity (MFI) associated with resting WPs in the absence of NPs were not significantly different when incubated with or without CD62P antibody (Figure 3A,B). On addition of the agonist, thrombin, the % PP and MFI significantly increased as expected, indicating activation of platelets with secretion of CD62P, which binds to the CD62P antibody (*p* < 0.0001 for % PP; *p* < 0.05 for MFI) (Figure 3A,B).

**Figure 3.** Effect of size of PLGA-PEG NPs at 2.2 mg/mL on platelet activation measured by flow cytometry (**A**) % positive platelets (% PP) (**B**) Mean fluorescence Intensity (MFI). Data are the average of *n* = 3–5 experiments ± SEM. \* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001 compared with washed platelets (WPs) + Ab (control), one-way ANOVA followed by post hoc Dunnett's test.

In the presence of PLGA-PEG NPs, a significant increase in % PP for the smallest sized NPs of 112 nm was observed over the 30 min incubation period. However, for the larger NPs of 348 and 576 nm, a significant increase in % PP was only observed at the incubation time of 30 min, (*p* < 0.05, *n* = 3), (Figure 3A). No significant difference in MFI values for platelets incubated with the NPs was observed over the 30 min except for the 348 nm NPs at 30 min incubation (Figure 3B).

#### *3.4. E*ff*ect of size of NPs on Thrombin Activation of Washed Platelets*

In absence of NPs, an increase in the percentage of CD6P-positive platelets (% PP) and in CD62P binding level (MFI) were observed following the exposure of WPs to thrombin, as expected, confirming activation of WPs in response to the agonist. The presence of NPs at any of the sizes tested did not result in a significant change in the % PP or MFI following exposure of the WPs with thrombin. (Figure 4A,B).

**Figure 4.** Effect of size of PLGA-PEG NPs at 2.2 mg/mL on thrombin activation of washed platelets measured by flow cytometry (**A**) % positive platelets (% PP) (**B**) Mean fluorescence Intensity (MFI). Data are the average of *n* = 3–5 experiments ± SEM. \* *p* < 0.05, compared with WPs + Ab (control), one-way ANOVA followed by post hoc Dunnett's test.

#### *3.5. Confocal Microscopy of the E*ff*ect of Incubation Time on the Interaction of NPs and Washed Platelets*

A rapid interaction of NPs and WPs, within the first minute, was observed at all NP sizes tested and at both the low (0.01 mg/L) and high (2.2 mg/mL) concentrations of NPs (Figure 5). The percent of platelets bound to NPs of 112 nm was in the order of 44 ± 5.7% and 53.7 ± 7.1% at *t* = 1 min for low and high concentrations, respectively. Increasing the incubation time or concentration of NPs at any of the sizes tested, did not result in significant change in the percent of NPs-bound platelets. The observed interaction between resting platelets and PLGA-PEG NPs of the three sizes tested did not lead to any apparent morphological alteration of the shape of the platelets or development of

pseudopods, indicating no effect on platelet activation. This observation was relevant to both low and high concentrations of NPs tested.

**Figure 5.** Confocal microscopy images of WPs (red) in suspension incubated with coumarin-PLGA-PEG NPs at 2.2 mg/mL) (**A**) 1 min, (**B**) 5 min, (**C**) 15 min, and (**D**) 30 min, at RT; Magnification 100×. Scale bar represents 10 μm.

Confocal microscopy of platelets layered onto fibrinogen at 20 μg/mL showed the typically formed filopodial projections and heterogeneity expected for adhered platelets. In the presence of NPs, of all of the three sizes and over the incubation times tested, similar attachment of platelets to fibrinogen with alteration in the shape of the platelets and development of extended filopodia projections were observed. The majority of NPs, at all sizes tested, were observed to be bound to platelets (Figure 6). This association of NPs to platelets was of the order of 53.7 ± 1.5% to 59 ± 14.1%, respectively, for 1 and 30 min incubation for 112 nm NPs, at 51 ± 6.1%, and 43 ± 11.3% for 1 and 30 min incubation times, respectively, for the 348 nm NPs. For the 576 nm NP, NP binding was 35.3 ± 5.7% at 1 min and increased to 52.7 ± 2.1% after 30 min incubation. The NP interaction with platelets was maintained upon activation through to adhesion (Figure 6).

**Figure 6.** Confocal microscopy images of WPs (red) incubated with coumarin-PLGA-PEG NPs at 2.2 mg/mL and adhered to fibrinogen coated slides: (**A**) 1 min, (**B**) 5 min, and (**C**) 30 min. Magnification 100×. Scale bar represents 10 μm.

#### **4. Discussion**

The average NP sizes formulated were 112, 348, and 576 nm and were significantly different (*p* < 0.05). This ensured a reasonable comparison of the effects of size of NPs on how platelets may respond when in the presence of NPs with varying sizes. Addition of thrombin at 0.1 U/mL to washed human platelets produced a platelet aggregation of 72–74, which is similar to the percent platelet aggregation of 74 ± 7.8 reported [27]. Incubation of PLGA-PEG NPs, of any size or concentration tested, with WPs for 4 min did not cause any platelet aggregation. This is similar to the results reported for 50 μM silver NPs incubated with WPs for 2 min at 37 ◦C [22]. In the current study, the 112 nm NPs did not influence the thrombin-induced platelet aggregation profile. Similarly, the larger NPs showed no significant effect on the thrombin-induced platelet aggregation profiles at the lower NP concentrations of 0.01–1.0 mg/mL and 0.01–1.5 mg/mL for the 348 and 576 nm NPs, respectively. Interestingly, a tendency of inhibiting or slowing platelet aggregation was observed for the 348 nm and 576 nm PLGA-PEG NPs at their highest concentrations. This may be related to the presence of NPs in the medium, acting as a physical barrier preventing platelet-platelet contact, hence delaying and reducing platelet activation and aggregation. As the NPs have a negatively charged surface, they may interact with the positively charged thrombin, lowering the free thrombin concentration in the suspension medium, hence reducing the exposure of platelets to thrombin. The surface charge, rather than size, was reported to be the key factor governing this subcategory of NPs' interaction with platelets [28]. Physiological or non-physiological agents can initiate various signalling pathways in platelets. Platelet activation can also be induced by biomaterials interacting with the specific intrinsic coagulation pathway, which leads to thrombin generation and ultimately platelet activation. However, such mechanisms still remain unexplored [29].

The influence of size-selected NPs on the expression of CD62P on the surface of platelets was analysed by flow cytometry. CD62P is a 140 kD glycoprotein stored in the secretory α-granules of the platelets, mediating leukocyte rolling and platelet–leukocyte aggregation.

The percentage of positive platelets (% PP) for the antibody marker reflects the proportion of activated cells in the total platelet population; this was used to quantify platelet activation in the presence of the NPs. The incubation of the smallest NPs (~113 nm) with WPs resulted in a significant increase in platelet activation, almost similar to the activation observed for thrombin-induced platelet activation. Interestingly, no significant effect on platelet activation was observed in the presence of the larger NPs of 348 nm and the 576 nm. Smaller sized NPs have a higher surface area per unit mass, which result in enhanced interaction potential with platelets. Silica NPs at the smallest PS of 10 nm, at a concentration of 200 μg/mL, were reported to increase CD62P abundance on the platelet surface, while 50 nm-sized NPs did not show any influence [21]. Similarly, of the 10, 50, 113, and 150 nm-sized NPs of silica exposed to platelets, the smallest NPs (10 nm) induced the upregulation of CD62P when used at a low concentration of 10 μg/mL [15,21]. The tendency of 26 nm carboxylated-polystyrene NPs to activate platelets was reported, while particles of larger sizes showed no influence on platelet activation [30].

The presence of NPs of any of the three sizes tested did not influence the % PP or the CD62P abundance on the surface of thrombin-activated platelets. This suggests the investigated NPs did not limit the access of the PE-labelled marker to the WPs surface, irrespective of their sizes.

CLSM performed in this study showed a rapid physical interaction of NPs, of all three sizes examined, with platelets in resting state. This interaction appeared to progress to possible NP uptake by platelets over time. This NP-platelet interaction, however, did not lead to platelet aggregation nor inhibit aggregation of platelets induced by the agonist, thrombin. In contrast, a previous study regarding latex microspheres interaction with platelets suggested platelet activation to be a prerequisite in order to interact with and internalise those inert microparticles (MPs) [31].

The interaction and binding of the investigated NPs to the surface of platelets as well as possible internalisation of NPs into platelets is interesting. In view of their unique small sizes and their curvature, NPs have been shown to enter various human cells and organs, a desirable characteristic for drug targeting [9,24,32].

In the present study, in the presence of PLGA-PEG NPs, variable morphologies of platelets from dendritic spread to the fully spread shape were displayed when platelets were layered onto fibrinogen. This indicates that the NPs did not affect platelet adhesion to fibrinogen. Similarly, silver NPs at different concentrations were reported to have no impact on platelet spreading on fibrinogen [22].

#### **5. Conclusions**

The results of this study show that while the smallest NPs demonstrated a tendency to result in platelet activation and the larger NPs, at the highest concentrations, to delay platelet aggregation, the three sizes of PLGA-PEG NPs examined had little or no effect on platelet aggregation. The NPs did not interfere with the activation status of platelets and did not inhibit platelet aggregation by the agonist thrombin. Importantly, the rapid binding of the studied PLGA-PEG NPs to platelets is promising, suggesting that these NPs could be potential carriers for targeted drug delivery to platelets in certain disease states. The data presented here supports the use of such NPs as targeting carriers of anti-platelet agents or thrombolytics. It is clear from this study that the question of whether these NPs would be consumed by resting and/or activated platelets remains unanswered. Future strategies may include a more sensitive analysis of the potential for these NPs, pre-loaded with therapeutic drug-candidates, to be internalised by platelets. A number of imaging modalities may be considered such as z stacking, the use of transmission electron microscopy (TEM), or using a combination of high resolution confocal microscopy with flow cytometric analysis that has the potential to provide a precise evaluation of whether NPs are internalised or adsorbed onto the platelets surface, as described by Vranic et al. [33]. With respect to the effects of the interactions of PLGA-PEG NPs at the studied sizes on normal platelet functions, the use of a more sensitive platelet aggregation assay, such as flow-induced platelet aggregation in the presence of PLGA-PEG NPs, should be considered in future studies. An example would include a flow-mimicking technique such as quartz crystal microbalance with dissipation (QCM-D), which operates under similar flow conditions encountered in microvasculature [34]. Further understanding of the interaction between platelets and NPs is required to aid in the development of a safe platelet-targeted drug delivery system in the future.

**Author Contributions:** Conceptualisation and Design: Z.R., S.O.; Methodology, Investigations, and Analysis: R.B., J.G., K.A., S.S., N.M.; Writing—original draft preparation: R.B., S.O., Z.R.; Review and editing: Z.R., S.O.; Supervision: S.O., Z.R.

**Funding:** This research was funded by King Abdulaziz University in Jeddah, Saudi Arabia and the Ministry of Higher Education in Saudi Arabia.

**Acknowledgments:** The authors gratefully acknowledge research funding by the King Abdulaziz University in Jeddah, Saudi Arabia and the Ministry of Higher Education in Saudi Arabia.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Microfluidics-Assisted Size Tuning and Biological Evaluation of PLGA Particles**

**Maria Camilla Operti 1, Yusuf Dölen 1,2, Jibbe Keulen 1, Eric A. W. van Dinther 1,2, Carl G. Figdor 1,2 and Oya Tagit 1,2,\***


Received: 30 September 2019; Accepted: 6 November 2019; Published: 8 November 2019

**Abstract:** Polymeric particles made up of biodegradable and biocompatible polymers such as poly(lactic-co-glycolic acid) (PLGA) are promising tools for several biomedical applications including drug delivery. Particular emphasis is placed on the size and surface functionality of these systems as they are regarded as the main protagonists in dictating the particle behavior in vitro and in vivo. Current methods of manufacturing polymeric drug carriers offer a wide range of achievable particle sizes, however, they are unlikely to accurately control the size while maintaining the same production method and particle uniformity, as well as final production yield. Microfluidics technology has emerged as an efficient tool to manufacture particles in a highly controllable manner. Here, we report on tuning the size of PLGA particles at diameters ranging from sub-micron to microns using a single microfluidics device, and demonstrate how particle size influences the release characteristics, cellular uptake and in vivo clearance of these particles. Highly controlled production of PLGA particles with ~100 nm, ~200 nm, and >1000 nm diameter is achieved through modification of flow and formulation parameters. Efficiency of particle uptake by dendritic cells and myeloid-derived suppressor cells isolated from mice is strongly correlated with particle size and is most efficient for ~100 nm particles. Particles systemically administered to mice mainly accumulate in liver and ~100 nm particles are cleared slower. Our study shows the direct relation between particle size varied through microfluidics and the pharmacokinetics behavior of particles, which provides a further step towards the establishment of a customizable production process to generate tailor-made nanomedicines.

**Keywords:** PLGA; drug delivery systems; microfluidics; nanoparticles; microparticles

#### **1. Introduction**

As evidenced by the dramatic increase in the number of studies and clinical applications over time, polymeric particles play a crucial role in drug delivery, providing improved stability, targeted delivery, and sustained release of loaded therapeutic agents without causing off-target toxicities in vivo [1–5]. Drug delivery particles based on poly(lactic-co-glycolic acid) (PLGA) are among the most commonly studied vehicles due to excellent biocompatibility, tuneable degradation characteristics, and long clinical history of PLGA [6–8]. The remarkable physicochemical properties and high versatility of PLGA have proven to address several challenges in drug delivery. PLGA can be processed into almost any size and shape [7] and can encapsulate molecules of virtually any size such as small drugs [9–13], proteins [14–18], nucleic acids [19–23], and vaccines [24–29]. The colloidal features such as size and surface functionality have a direct influence on the cellular uptake, biodistribution, and thus therapeutic efficacy of the particles, dictating the in vivo fate of therapeutic cargo [30,31]. These colloidal features

are particularly important in cancer immunotherapy as the efficacy of any given therapeutic agent is highly dependent on its ability to reach either the tumor microenvironment (TME) or the lymph nodes. The incomplete endothelial lining due to rapid angiogenesis at the tumor site results in the formation of large and irregular pores (0.1–3 μm) [32], through which nanoparticles can escape from the circulatory stream and accumulate at the TME via a so-called enhanced permeability and retention (EPR) effect [32]. Toy et al. demonstrated that, within the size range of 60–130 nm, smaller particles exhibited greater lateral drift towards the blood vessel walls, which is a prerequisite for interaction with the tumor vascular bed and essential for escape into the TME [33]. Furthermore, both blood clearance and uptake by the mononuclear phagocyte system (MPS) at the liver and spleen depend on the size and composition of PLGA carrier systems. The slit size in the inter-endothelial cells of the spleen is about 200 nm, which facilitates the leakage and circulation of smaller particles for longer time periods [34] or enables their entry to the lymphatic system through direct drainage to the lymph nodes [35]. On the other hand, larger particles, which are not eligible for intravenous administration due to their excessive size, are more likely to be taken up by immune cells at the injection site, such as in the case of subcutaneous or intramuscular pathways [36].

Surface functionality also influences drug pharmacokinetics in vivo. The use of polyethylene glycol (PEG) has been shown to significantly increase circulation time in several studies [36–38]. When attached to nanoparticles' surface (a process so-called PEGylation), the hydrophilicity of PEG chains recruits specific proteins from plasma, that cloak and limit the interactions of particles with MPS cells, hence prolonging blood circulation ('stealth effect') [39,40]. The decrease in the aggregation, opsonization, and phagocytosis of particles entails the extension of their circulation time. Morikawa et al. reported additional benefits of PEGylation such as reduced particle size and improved encapsulation efficiency for curcumin-loaded nanoparticles [41].

In addition to particle size and surface functionality that have a large influence on the therapeutic efficacy in vivo, particle uniformity in terms of size and encapsulated cargo is also an important aspect particularly for the clinical translation of polymeric drug delivery formulations [42,43]. The uniformity of particles depends largely on the utilized manufacturing approach. The conventional production techniques based on emulsion solvent diffusion, emulsion solvent evaporation, and nanoprecipitation are suitable for the production of sub-micron and micron-size PLGA particles [7,8]; however, these methods lack the precision and full control over the particle size and uniformity particularly for larger scale processes [8]. In recent years, microfluidics technology has emerged as an effective tool to produce particles in a highly controllable manner [44,45]. Major advantages of microfluidics-based particle manufacturing include the requirement of low sample volumes, high surface area, and reduced system footprint [46]. This technology allows for rapid fluid mixing at the nanoliter scale and production of PLGA particles with highly specific sizes as well as surface functionalities only by altering specific parameters such as concentration of the starting materials or fluid flow rates through micron-size channels [41,47–52]. Additionally, traditional small-scale laboratory synthesis techniques suffer from batch-to-batch variations while microfluidics technology provides a precise size control, a high degree of particle uniformity, and reproducibility. Furthermore, formulations can be scaled up by increasing the quantities of fluids pumped through the system or by parallelizing multiple microfluidic mixers [53].

In this study, we report on the use of microfluidics technology as a platform for the production of PEGylated PLGA particles. We explored the feasibility to generate different sizes of particles with low polydispersity index (PDI) by varying process and formulation parameters such as total flow rate and flow rate ratio of organic and aqueous phases as well as surfactant and polymer concentration. We tuned the particle size at sub-micron (~100 nm and ~200 nm) and micrometer (>1000 nm) length scales, which represent biologically-relevant cut-off values that influence the particle biodistribution and clearance in vivo. A fluorescent dye was used as a model drug to assess encapsulation efficiency, release profile, and uptake by mouse-derived immune cells. Clearance of systemically administered, fluorescently labeled particles was also studied on a mouse model through in vivo imaging.

#### **2. Experimental**

#### *2.1. Materials*

PLGA (Resomer RG 502H), with a 50:50 ratio of lactic acid:glycolic acid and Mw 7000–17,000 Da was obtained from Evonik Nutrition and Care GmbH (Darmstadt, Germany). PEG-PLGA copolymer (PEG Mn 5000, PLGA Mn 7000), polyvinyl alcohol (PVA, 9000–10,000 Mw, 80%, hydrolyzed) and cholesteryl BODIPY™ FL C12 were obtained from Thermo Fisher Scientific (Waltham, MA, USA). Near-infrared emitting fluorescent dye VivoTag-S 750 was purchased from Perkin Elmer Inc. (Waltham, MA, USA) and acetonitrile (ACN, 99.95%) was from VWR International (Radnor, PA, USA). Ultrapure Milli-Q® water (18.2 MΩ.cm) was used where necessary (Merck KGaA, Darmstadt, Germany). Roswell Park Memorial Institute (RPMI) 1640 medium, Anti-Anti (AA), and ß-mercaptoethanol were obtained from Gibco (Thermo Fisher Scientific, Waltham, MA, USA). Granulocyte-macrophage colony-stimulating factor (GM-CSF) was obtained from Peprotech Inc. (Rocky Hill, NJ, USA). X-Vivo medium and ultraglutamine were from Lonza Group (Basel, Switzerland). Fetal bovine serum (FBS) was purchased from Hyclone Laboratories Inc. (GE Healthcare, Chicago, IL, USA).

#### *2.2. Equipment*

The microfluidics system was set up by connecting syringe pumps (Harvard PHD-2000 infusion 70–200) to a Y-junction mixer with staggered herringbone ridges (NanoAssemblr™, Precision Nanosystems Inc., Vancouver, Canada) through 0.8 mm polytetrafluoroethylene (PFTE) tubing (ID 0.8 mm, OD 1.58 mm) obtained from Sigma-Aldrich (St. Louis, MO, USA). A detailed characterization of the mixing geometry was reported in [44]. Two 20 mL NORM-JECT Luer Lock syringes were connected to 0.8 mm diameter needles (Braun Sterican 0.8 × 120 mm), which were inserted in the fittings connected to the inlets of the mixing cartridge. A PFTE tubing connected to the chip outlet was used for sample collection.

#### *2.3. Preparation of PLGA Particles*

Prior to particle production, the pipes and the mixing chip were primed first with the solvents (ACN for the organic phase inlet and MilliQ for the aqueous phase inlet) and then with the appropriate phases for 1 min at organic: aqueous flow rates of 2:2. After the priming step, pumps were operated at the desired flow rates. The product obtained within the first 2 min was discarded. The collected particles were left stirring overnight (350 rpm) at room temperature for organic solvent evaporation. A schematic representation of the process is shown in Scheme 1.

**Scheme 1.** Schematic illustration of the particle generation process via nanoprecipitation method using a Y-junction mixing cartridge. PLGA: poly(lactic-co-glycolic acid); ACN: acetonitrile; PVA: polyvinyl alcohol.

The particle size was tuned via varying the flow rates of organic and aqueous phases as well as PLGA and PVA concentrations. Upon determination of the optimized parameters for each target size, PEGylated PLGA particles encapsulating fluorescent dyes emitting either at visible or near-infrared regions were prepared by the following methods:

>1000 nm particles: the organic phase (5 mL ACN) contained 233.1 mg PLGA, 99.9 mg PEG-PLGA, and 50 μL of 1 mg/mL fluorescent dye in ethanol. A 3% PVA solution was used as the aqueous phase. Particles were produced at 6:2 organic:aqueous flow rates in triplicate.

~200 nm particles: the organic phase (5 mL ACN) contained 116.6 mg PLGA, 50 mg PEG-PLGA, and 50 μL of 1 mg/mL fluorescent dye in ethanol. A 1% PVA solution was used as the aqueous phase. Particles were produced at 2:6 organic:aqueous flow rates in triplicate.

~100 nm particles: the organic phase (8 mL ACN) contained 93.5 mg PLGA, 40 mg PEG-PLGA, and 80 μL of 1 mg/mL fluorescent dye in ethanol. A 1% PVA solution was used as the aqueous phase. Particles were produced at 4:6 organic:aqueous flow rates in triplicate.

After evaporation of the organic solvent, particles were washed three times with MilliQ water by centrifugation at 15,000 rpm for 35 min (for >1000 nm and ~200 nm particles) and by spin filtration using a 100,000 kDa MW cut-off centrifugal filter device (Millipore, Merck KGaA, Darmstadt, Germany) for ~100 nm particles. Particles were then lyophilized.

#### *2.4. Colloidal Characterization of PLGA Particles*

In total, 1.25 mg of lyophilized particles were dispersed in 1 mL of MilliQ water to determine the size distribution and polydispersity index (PDI) with dynamic light scattering using a Nanotrac Flex (Microtrac GmbH, Krefeld, Germany). Zeta potential measurements were obtained with a Zetasizer Nano ZS (Malvern Instruments, Malvern, United Kingdom). Prior to measurements, particles were suspended in a 50 mM NaCl solution. The average of three measurements were used to report the size, PDI, and Zeta potential for each sample.

Atomic force microscopy (AFM) images of particles were obtained with a Catalyst BioScope (Bruker, Billerica, MA, USA) coupled to a confocal microscope (TCS SP5II, Leica Mycrosystems, Wetzlar, Germany). Then, 50 μL of 10 mg/mL particle suspension was dried on clean glass substrates and particles were imaged in peak-force tapping mode using a silicon nitride cantilever with a nominal spring constant of 40 N/m (Bruker). AFM images were analyzed using NanoScope analysis software (Bruker, Billerica, MA, USA).

#### *2.5. Determination of Encapsulation E*ffi*ciency*

Dye loading was quantified by construction of a calibration curve using a BODIPY-C12 standard. A series of dye concentrations (in 0–0.25 mg/mL range) were prepared in a solvent composed of equal parts of ACN and MilliQ and fluorescence intensities were measured for each standard. The fit function applied to the linear portion of the curve was used for the calculation of the dye content of particles. The encapsulation efficiency was determined by comparing the total amount of dye in the lyophilized particles to the initial amount of dye supplied for the collected volume.

#### *2.6. In Situ Release Profile*

PLGA particles containing BODIPY-C12 were suspended in PBS in 5 mg/mL concentrations and were dialyzed at 37 ◦C for 14 days using dialysis tubes with a 1000 Da molecular weight cut-off membrane (GE Healthcare, Chicago, IL, USA). At different incubation times, the dialysis medium was collected for fluorescence measurements using an LS 55 Perkin Elmer fluorescence spectrometer (Waltham, MA, USA). Samples were excited at 488 nm and emission was recorded between 500 nm and 700 nm. After each measurement the dialysis medium was refreshed. Samples were studied in triplicate.

#### **3. In Vitro Cellular Uptake Experiments**

#### *3.1. Generation of Bone Marrow-Derived Dendritic Cells (BMDCs)*

Dendritic cells were generated from mouse bone marrow cells by culturing them in full RPMI medium containing 5 mL 200 mM l-glutamine (Gibco, Thermo Fisher Scientific, Waltham, MA, USA), 5 mL 100× antibiotics and antimycotics (Gibco, Thermo Fisher Scientific, Waltham, MA, USA), 10% fetal bovine serum (Gibco, Thermo Fisher Scientific, Waltham, MA, USA), and 500 μL β-mercapto ethanol (Sigma-Aldrich, St. Louis, MO, USA). Then, 5.0 <sup>×</sup> 10<sup>6</sup> cells in 13 mL of full medium were cultured in the presence of 20 ng/mL GM-CSF for 7 days. At day 3, 4 mL of complete medium containing 37.9 ng/mL GM-CSF was added. At day 6, 1 mL complete medium containing 158 ng/mL GM-CSF was added. Cells were used for uptake experiments at day 7.

#### *3.2. Isolation of Myeloid-Derived Suppressor Cells (mMDSCs) and Polymorphonuclear Myeloid-Derived Suppressor Cells (pmnMDSCs)*

Gr-1dimLy-6G<sup>−</sup> monocytic and Gr-1highLy-6G<sup>+</sup> polymorphonuclear myeloid-derived suppressor cells (mMDSCs and pmnMDSCs, respectively) were isolated from the spleen of tumor-bearing mice using an isolation kit (Miltenyi Biotec, Bergisch Gladbach, Germany) according to manufacturer's instructions. Briefly, the spleen was isolated under sterile conditions and meshed through a 100 μm cell strainer with a syringe plunger. The cell suspension was spun at 400× *g* for 5 min and resuspended in 3 mL of 1× ammonium chloride solution for the lysis of erythrocytes. After 5 min of incubation at room temperature, cells were washed with 10 mL of PBS. The cells were incubated with an Anti-Ly-6G-Biotin antibody and Anti-Biotin MicroBeads and were subsequently applied to a magnetic-activated cell sorting (MACS) column, which retained the pmnMDSCs. The flow-through containing mMDSCs were eluted as the positively selected cell fraction and were further purified by applying them to a second MACS column.

#### *3.3. In Vitro Cellular Uptake*

Firstly, 1.0 <sup>×</sup> 105 cells in 500 <sup>μ</sup>L complete medium were transferred to 5 mL propylene round bottom tubes (Falcon). Then, 10 μg of particles containing BODIPY-C12 water were added to the round bottom tubes and were incubated for time periods of 1, 2, 4, 6, 24, and 48 h. After incubation, particle uptake was determined by flow cytometry analysis on a FACSVerse (BD Biosciences, Franklin Lakes, NJ, United Sates).

#### **4. In Vivo Clearance Studies**

All animal experiments were performed according to guidelines of Radboud University's Animal Experiment Committee and Central Authority for Scientific Procedures on Animals (project number 2015-019TIL, date September 2015) in accordance with the ethical standards described in the Declaration of Helsinki. Wild-type BALB/cAnNCrl mice, aged 8–12 weeks, were obtained from Charles River, Germany and maintained under specific pathogen-free conditions at the Central Animal Laboratory (Nijmegen, The Netherlands). Drinking water and food were provided ad libitum. Mice were warmed either in a heating chamber or under a heating lamp and 1 mg PLGA nanoparticles (~200 nm and ~100 nm) containing VivoTag-S 750 were injected in 200 μL of phosphate-buffered saline (PBS) solution through a lateral tail vein using a 1 mL syringe with a 29 G needle. Then, 0.5, 3, 24, and 48 h after injections mice were shaved and imaged in an IVIS Lumina II (Perkin Elmer) system. Mice were euthanized, and organs were dissected and imaged separately at 24 and 48 h. Imaging settings were: exposure time: 3 s; binning: medium; F/stop: 2; fluorescent excitation filter: 745 nm; fluorescent emission filter: 810–885 nm. A fluorescent background acquisition was performed for each time point. Living Image software (Caliper Life Sciences, Hopkinton, MA, USA) was used for data analysis. Background values were subtracted from measurement values. Same sized regions of interest (ROI)

were applied on the liver and bladder for full body image analysis; also, same sized ROIs were applied on the isolated liver and spleen. Total flux (photon/s) per each ROI was calculated.

#### *Statistical Analysis*

An unpaired *t*-test was used to determine the significance of the difference in mean particle size values among the compared groups using GraphPad Prism.

#### **5. Results and Discussion**

#### *5.1. Preparation of Particles*

A commercially available NanoAssemblr™ cartridge [8] connected to syringe pumps was used for the production of PLGA particles through nanoprecipitation. In this method, water-miscible organic solvents are used to dissolve the polymer which, once at the interface between the aqueous medium and the organic solvent, will start to precipitate. Polymer deposition caused by fast diffusion of the solvent leads to instantaneous formation of a colloidal suspension and particle size is mainly determined by solvent diffusion coefficient (*D*) and mixing time (*t*mix) [54]. The architecture of the mixing chip also plays a crucial for the generation of a hydrodynamic-focusing flow pattern and the mixing process can be controlled by varying the width of the focused stream (*wf*) as shown in Equation (1) [44]:

$$t\_{\rm mix} \approx \frac{wf^2}{4D} \tag{1}$$

By varying the ratio of flow rates, the lateral width of the focused stream can be adjusted very accurately, which leads to a highly controlled diffusion and mixing at the interface of organic and aqueous phases. In our study, acetonitrile (ACN) was used as the organic phase and particles were formed upon rapid mixing of organic and aqueous phases in micron-size channels. Due to rapid diffusion of ACN to the aqueous phase, PLGA precipitated and was subsequently stabilized by PVA present in the aqueous phase. Process parameters such as (total) flow rates of organic and aqueous phases as well as formulation parameters such as PLGA and PVA concentration were varied to tune the size of PLGA particles from sub-micron to micron-scales.

The first process parameter tested was the flow rate of the organic phase composed of a 33.3 mg/mL PLGA solution in ACN. The organic phase flow rate was varied between 2 mL/min and 6 mL/min while the aqueous flow rate was kept constant at 2 mL/min for a 1% PVA solution (Figure 1A). A gradual increase in the particle size from ~400 nm to ~900 nm was observed with an increase of the organic phase flow rate. This was most likely caused by precipitation of higher amounts of PLGA at a given mixing time when relatively higher organic phase flow rates were used. Similarly, increasing the flow rate of the aqueous phase for a given set of formulation parameters and organic phase flow rate resulted in the formation of smaller particles (Figure 1B). Doubling the aqueous flow rate to 4 mL/min resulted in particles almost half the size of those produced at 2 mL/min. However, increasing the aqueous flow rate up to 6 mL/min did not result in a notable change in particle size.

Another process parameter affecting particle size was the total flow rate of the organic and aqueous phases. With the equal flow rate of organic and aqueous phases, the increase of the total flow rates from 4 mL/min (2:2) to 8 mL/min (4:4) led to a decrease in the particle size (Figure 1C). Increasing the total flow rate further to 12 mL/min (6:6), however, did not cause any notable difference in the particle size other than a slight improvement in the PDI. Increasing the total flow rates from 8 mL/min to 12 mL/min apparently did not cause a relevant variation of solvent diffusion time to induce a significant change in particle size. The total flow rate determines both the mixing time of the two phases in the cartridge and the collection rate of the particles at the outlet channel. Faster mixing by increasing the total flow rates has also been reported in other studies to result in the formation of smaller particles [41,47].

**Figure 1.** (**A**–**C**) Influence of process and formulation parameters on the particle size. The variation of particle size at different flow rates for 33.3 mg/mL PLGA as organic phase and 1% PVA as aqueous phase. (**D**) The influence of PLGA concentration on the particle size (organic:aqueous flow rate of 4:4 mL/min, 3% PVA as aqueous phase). Bars: intensity-averaged particle size; green dots: polydispersity index (PDI). QO: flow rate of organic phase, QA: flow rate of aqueous phase. Significant: \*: *p* < 0.05; \*\*: *p* < 0.005; \*\*\*: *p* < 0.001; \*\*\*\*: *p* < 0.0001. Non-significant (n.s.): *p* > 0.05.

Overall, particles in the sub-micron size range were obtained by varying only the flow parameters. Furthermore, we varied the PLGA and PVA concentrations in order to increase the particle size to micron scale. A representative data set in Figure 1D shows that particle size almost tripled when the PLGA concentration was doubled. In these batches, the PVA concentration was increased as well and organic:aqueous flow rates were kept equal (4:4 mL/min). The larger particle size obtained with a higher PLGA concentration can be explained by the higher viscosity of the organic phase, which resulted in a slower diffusion of ACN to the aqueous phase and an increased mixing time [55].

For all tested conditions, a good batch-to-batch reproducibility and a small PDI (≤0.2) was observed. After several trials, the optimal conditions and parameters were determined to obtain particles of sizes >1000 nm, ~200 nm, and ~100 nm. These parameters were then used to prepare particles with a surface functionalized with polyethylene glycol (PEG) and encapsulating a fluorescent dye (Table 1).


**Table 1.** Optimized formulation and process parameters to obtain PLGA particles with indicated size. PEG: polyethylene glycol.

For fluorescent labeling, two different fluorescent dyes with visible (BODIPY™ FL C12, green, 510 nm) and near-infrared (VivoTag-S 750, red, 750 nm) emission wavelengths were used. BODIPYs are low-polarity dyes with stable fluorescence emission properties [56] and are commonly used in fluorescence detection and photodynamic therapy applications [57]. VivoTag-S 750 is a near-infrared emitting (NIR) fluorochrome extensively used for in vivo imaging applications [58,59]. PLGA particles labeled with the green fluorescent dye were used for in situ release and in vitro uptake experiments, whereas those labeled with the red dye were used for in vivo imaging studies.

#### *5.2. Colloidal Characterization*

A detailed colloidal and functional characterization of PEGylated PLGA particles encapsulating BODIPY-C12 is shown in Figure 2. Both atomic force microscopy (AFM) images (Figure 2A–C) and dynamic light scattering (DLS) measurements (Figure 2D) revealed the formation of monodisperse particles (PDI ≤0.2) within the desired size range (i.e.; >1000 nm, ~200 nm, ~100 nm) also upon PEGylation and fluorescent dye encapsulation. It should be noted that although PDI values smaller than 0.3 are considered acceptable for drug delivery applications, more specific standards and guidelines have yet to be established by regulatory authorities.

**Figure 2.** Atomic force microscopy height images of (**A**) >1000 nm, (**B**), ~200 nm, and (**C**) ~100 nm PLGA particles. The scale bars display the Z-range and the scan sizes are shown on the *x*-axis. (**D**) Size distribution and (**E**) ζ potential values of PLGA particles measured using dynamic light scattering. (**F**) Release profile of BODIPY-C12 from >1000 nm, ~200 nm, and ~100 nm particles. Data obtained for >1000 nm, ~200 nm, and ~100 nm particles are represented in orange, blue, and red, respectively.

The micron-sized particles showed an average size of 1690 nm (±60 nm), while the sub-micron particles were 250 nm (±48 nm) and 106 nm (±5 nm) in diameter. For ease of reporting, we will continue referring to them as >1000 nm, ~200 nm, and ~100 nm particles. Slightly negative ζ potential values that spanned a range between −5 mV and −15 mV were observed for all particle types (Figure 2E). While the variation of ζ potential was negligible among sub-micron particles (i.e., ~200 nm, ~100 nm), the micron-size particles displayed the most negative value. PLGA nanoparticles stabilized with PVA have also been reported to bear a slightly negative ζ potential (−5 mV) [60] due to the presence of residual PVA on the particle surface, which affects the number of carboxylic acid end groups [61]. The hydrophobic acetate moieties in partially hydrolyzed PVA can lead to its entrapment within the PLGA matrix on the particle surface, thereby masking the surface charges to almost neutral ζ potential values [62]. Therefore, the more negative ζ potential obtained for micron-size particles can be related

to differences in the PVA content and the flow rate of the aqueous phase for micron-size particles compared to sub-micron particles, as well as higher PLGA content. Apparently, these variations in formulation and process parameters influence not only the particle size but also the particles' surface properties, which collectively determine the pharmacokinetic properties, cellular uptake, and particle biodistribution [63].

Encapsulation efficiency of BODIPY-C12 was determined using fluorescence spectroscopy by comparing the initial amount of dye supplied during particle production and the total amount of dye detected in the entire yield. The efficiency of dye encapsulation was significantly higher for smaller particles with the highest efficiency at 44% for ~100 nm particles, which decreased gradually to 21% and 13% for ~200 nm and >1000 nm particles, respectively. It should be noted that each particle type was produced using different formulation and process parameters, which equally play a role in determining the encapsulation efficiency [64]. An efficient encapsulation requires the rapid precipitation of polymer with the fluorescent dye to restrict the dye within the polymer matrix and prevent its diffusion to the aqueous phase. The micron-size particles with the lowest encapsulation efficiency were prepared using the highest polymer concentration and the highest organic:aqueous phase flow rate ratios. The high viscosity of the organic phase as well as the relatively low fraction of the aqueous phase probably resulted in a slower polymer precipitation, and therefore a lower encapsulation efficiency for micron-size particles.

The functional characterization of particles was achieved by monitoring the dye release. For this study, particles dispersed in PBS were dialyzed at 37 ◦C for a period of 14 days. At certain time points, fluorescence intensities of the dialysis media were measured using fluorescence spectroscopy. The dialysis medium was replaced with a fresh medium after each measurement. Each formulation type showed a distinct release profile that was strongly correlated with the particle size. After an initial burst, the release was steady for all particles (Figure 2F). The magnitude of burst release was higher for ~100 nm particles (approx. 30%), which was found as ~15% and ~10% for ~200 nm and >1000 nm particles, respectively.

In addition to the highest burst release, the overall release rate was also faster for ~100 nm particles such that they already released the majority of their content (~80%) by day 4. On the other hand, the released content of ~200 nm and >1000 nm particles barely reached 50% and 20%, respectively, by the end of the entire release period (two weeks). Indeed, several different release patterns such as mono-, bi-, and tri-phasic release are reported for PLGA particles, which are mainly regulated by the physicochemical properties of the cargo, PLGA type (molecular weight, lactide/glycolide ratio, etc.), and particle morphology (size, porosity, etc.) [65]. Since the same cargo and polymer type were used for the preparation of particles, the main reason for different release rates observed was the particle size. Larger surface area/volume ratio of smaller particles facilitated the faster diffusion of dye molecules located on or close to the surface. Such inverse relation between the release rate and particle size is in good agreement with the general trends reported in other studies [66–68].

#### *5.3. In Vitro Uptake Experiments*

Particle size and surface functionality are among the key parameters that influence their interactions with cells [69]. In this work, we used mouse-derived cells, namely bone marrow-derived dendritic cells (BMDCs), CD103<sup>+</sup> dendritic cells (CD103+), and myeloid-derived suppressor cells (MDSCs) of monocytic (mMDSCs) and polymorphonuclear (pmnMDSCs) sub-types to study the uptake of fluorescently labeled, PEGylated PLGA particles of varying sizes (Figure 3).

Particles incubated with BMDCs for different time periods were analyzed using flow cytometry (Figure 3A). In our previous study we reported that, once taken up by the cells, the integrity of PLGA particles is compromised before 72 h of incubation [70]. Hence, in the present work, we monitored particle uptake up to 48 h in order to avoid possible variations in the mean fluorescence intensities (MFIs) due to particle degradation at later time points. For ease of comparison, MFI data were normalized to 1 for the values obtained at 48 h of incubation time. A clear correlation between the

particle size and the trend in particle uptake was observed. For sub-micron particles (~100 nm and ~200 nm) a stable MFI was observed within the first 6 h, which increased at later time points (Figure 3A, red and blue curves). On the other hand, micron-size particles (>1000 nm) showed a time-dependent uptake behavior. For these particles, the intracellular MFI increased gradually within the first 6 h, reaching a plateau until the end of the incubation period (Figure 3A, orange curve).

**Figure 3.** (**A**) Intracellular mean fluorescence intensity (MFI) values obtained for >1000 nm (orange), ~200 nm (blue), and ~100 nm (red) PLGA particles incubated with bone marrow-derived dendritic cells (BMDCs) for different time periods, and (**B**) correlation between the particle size and intracellular MFI after 2 h of incubation with CD103<sup>+</sup> dendritic cells, (**C**) monocytic myeloid-derived suppressor cells (MDSCs), and (**D**) polymorphonuclear MDSCs isolated from mouse. Data were obtained using flow cytometry.

Overall, the uptake of sub-micron particles was more efficient compared to micron-size particles. The MFI obtained at 1 h of incubation already corresponded to ~75% and ~50% of the total MFI for 100 nm particles (Figure 3A, red) and 200 nm particles (Figure 3A, blue), respectively. It should be noted that each type of PLGA particles had different dye loading ratios. In this uptake study, equal amounts of PLGA particles with different total fluorescence intensities were used. Although normalization of MFI enabled a fair comparison of the particle uptake trend for different sizes, we further investigated the particle uptake on other types of mouse-derived immune cells using particles with equal fluorescence intensities instead of equal particle mass in order to compensate for the variations in dye encapsulation (Figure 3B–D). Flow cytometry histograms of the data presented in Figure 3B–D are shown in Figure S1. A clear correlation between the particle size and intracellular MFI was observed for the cells that were either generated in vitro (Figure 3B) or were isolated from the spleen of tumor-bearing mice (Figure 3C,D). After 2 h incubation period, the uptake of ~100 nm particles was ~1.5 fold higher than the >1000 nm particles for CD103<sup>+</sup> dendritic cells (Figure 3B) and mMDSCs (Figure 3C), and the difference was even higher (almost three-fold) for pmnMDSCs (Figure 3D). Similar studies also reported the lower uptake efficiencies of micron-size particles compared to sub-micron particles by dendritic cells [71],

which could be due to different uptake mechanisms associated with different particle sizes. The uptake of small particles (<100 nm) has been reported to be clathrin- and/or caveolin-mediated [72], whereas micropinocytosis and phagocytosis are the main mechanisms via which sub-micron particles (~200 nm) and micron-sized particles (>1000 nm) are taken up [73]. In addition to particle size, surface charge is an important parameter that influences the efficiency of particle uptake by cells. Prior to their internalization, particles need to attach on the cell surfaces that are decorated with negatively-charged proteoglycans [63]. Consequently, the uptake of positively-charged particles can be more efficient due to electrostatic interactions between the particles and cell surface. In our work, all particles had negative ζ potential values, which was highest for >1000 nm particles. Therefore, the poorer uptake efficiency observed for micron-size particles can be due to less-favorable surface charge in addition to larger particle size.

All the cell types used in our work are important regulators of immune response. BMDCs and CD103<sup>+</sup> are antigen presenting cells that can prime T cells to induce antigen-specific immune responses [74]. On the other hand, MDSCs play an important role in immune suppression in cancer as well as in tumor angiogenesis, drug resistance, and promotion of tumor metastases [75], representing an attractive potential therapeutic target, for things such as cancer immunotherapy. Among the studied PLGA formulations, ~100 nm particles with almost neutral surface charge would be the most efficient vehicle to deliver such things as antigens to dendritic cells (DCs) and immunomodulatory drugs to MDSCs for cancer immunotherapy.

#### *5.4. In Vivo Clearance of Particles*

Rapid clearance of particles from the bloodstream through the mononuclear phagocyte system (MPS) and reticuloendothelial system (RES) represents a major limitation to achieve preferential accumulation of particles in the target organs [76]. Tuning the size and surface properties of particles has proven useful in preventing their rapid removal from the bloodstream [77,78]. In order to evaluate particle clearance in vivo, we administered PLGA particles labeled with a near-infrared (NIR)-emitting dye (VivoTag-S 750) intravenously (i.v.) to mice. The acceptable particle size range for the i.v. injections have been reported as 10–1000 nm to prevent possible accumulation of larger particles in the lung capillaries [61,79,80]. Therefore, we used only sub-micron size particles for in vivo studies. The colloidal properties of ~100 nm and ~200 nm particles encapsulating the NIR-emitting dye were similar to those labeled with the green fluorescent dye (Figure S2). Whole-body and ex-vivo organ imaging were performed at different time points up to 48 h after i.v. administration of sub-micron PLGA particles (Figure 4).

Whole-body imaging revealed the presence of both particle types mainly in the liver and bladder already after 30 min following the injection (Figure 4A). The liver signal decreased gradually at later time points and was still above background levels at 48 h for both ~100 nm and ~200 nm particles. The presence of a high fluorescence signal in the bladder at 0.5 h was an interesting observation, which could be related to the excretion of free dye molecules that were burst-released. In fact, intact particles cannot pass through the renal filtration barrier since it has an effective size cut-off of ~10 nm [81]. The higher percentage of burst release displayed by ~100 nm particles in situ aligned well with the in vivo observations, in which the bladder signal resulting from the excretion of the free dye molecules was relatively lower for ~200 nm particles. Overall, the bladder signal did not provide a reliable measure of the systemic clearance of the particles due to the interference of the burst-released dye. Thus, we monitored the clearance of particles from the liver as representative of their systemic clearance. The influence of burst-released dye can be avoided via covalent attachment of the fluorescent dye.

**Figure 4.** (**A**) Fluorescent whole-body images of mice obtained at different time points up to 48 h after intravenous (i.v.) administration of ~200 nm and ~100 nm PLGA particles. The liver and bladder are encircled. (**B**) Variations of liver fluorescence in whole-body images. The fluorescence intensities are normalized at maximum values observed at 3 h after administration of ~100 nm (blue) and ~200 nm (green) PLGA particles. (**C**) Variations of liver fluorescence intensities of ~100 nm (blue) and ~200 nm (green) PLGA particles obtained on an isolated liver at 24 h and 48 h after particle administration. Data obtained for untreated mice (negative control) are shown in black.

The variation of liver signal was monitored using the whole-body images obtained at different time points (Figure 4B). For a better comparison, the values were normalized to maximum intensities measured at 3 h. From this point on, the decay of liver signal was almost linear for ~100 nm particles with a slope of <sup>−</sup>0.014 (R2 = 0.9926) (Figure 4B, blue curve), whereas the decay of ~200 nm particles better suited an exponential function (Figure 4B, green curve). Since the number of fit points was not sufficient for an accurate exponential fit, we used a linear fit for ~200 nm particles as well. These particles displayed a relatively faster decay with a slope of <sup>−</sup>0.018 (R<sup>2</sup> = 0.9537). Ex-vivo imaging of an isolated liver also showed similar variations in signal intensities, such that ~200 nm particles (Figure 4C, green curve) displayed a more pronounced decrease at 48 h compared to ~100 nm particles (Figure 4C, blue curve). Faster clearance of larger particles was also shown in other studies. When administered intravenously, particles larger than 200 nm activate human complement systems, are rapidly eliminated from the bloodstream, and gather mainly in the liver and spleen where the rate of accumulation is proportional to the size of the particles [30]. While particles with a diameter greater than 200 nm were most likely cleared by Kupffer cells, smaller particles displayed a decreased rate of clearance and an extended circulation time [82]. Of note, we did not observe a signal originating from the spleen in whole-body imaging. However, ex-vivo imaging of isolated spleen at 24 h and 48 h revealed signal intensities that were only slightly above the background noise, indicating that the spleen was not the preferential accumulation site for sub-micron PLGA particles (Figure S3). Indeed, it has been reported that the spleen receives barely 15% of the i.v. injected dose of nanomedicines [83]. Therefore, additional surface modification strategies may be needed for applications that require splenic accumulation of PLGA particles.

#### **6. Conclusions**

In this study, we demonstrated how PLGA particle size can be specifically tuned using a microfluidics system via modulating the formulation and process parameters. Through a series of optimization experiments, we obtained PEGylated PLGA particles in different sizes, which remarkably affected the characteristics of the particles in vitro and in vivo confirming the direct relation between the size and the pharmacokinetics behavior. This work can be considered as a further step towards the establishment of a production process that is able to generate tailor-made medicine for each individual clinical need.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/1999-4923/11/11/590/s1, Figure S1: Flow cytometry histograms of the uptake experiments, Figure S2: Size distribution and ζ potential values of PLGA particles encapsulating a NIR-emitting fluorescent dye, Figure S3: Variations of spleen fluorescence intensities of ~100 nm and ~200 nm PLGA particles obtained on an isolated spleen at 24 h and 48 h after particle administration.

**Author Contributions:** M.C.O., Y.D., J.K., E.A.W.v.D.: Performed experiments, data curation, analysis, and writing. C.G.F.: Funding acquisition, supervision, review and editing. O.T.: Conceptualization, supervision, data interpretation, writing, review and editing.

**Funding:** This research was funded by EU grant PRECIOUS (686089).

**Acknowledgments:** C.F. received the NWO Spinoza grant, ERC Advanced Grant Pathfinder (269019) and Dutch cancer society award 2009-4402.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Terahertz Spectroscopy: An Investigation of the Structural Dynamics of Freeze-Dried Poly Lactic-co-glycolic Acid Microspheres**

**Talia A. Shmool 1, Philippa J. Hooper 1, Gabriele S. Kaminski Schierle 1, Christopher F. van der Walle <sup>2</sup> and J. Axel Zeitler 1,\***


Received: 23 April 2019; Accepted: 11 June 2019; Published: 20 June 2019

**Abstract:** Biodegradable poly lactic-co-glycolic acid (PLGA) microspheres can be used to encapsulate peptide and offer a promising drug-delivery vehicle. In this work we investigate the dynamics of PLGA microspheres prepared by freeze-drying and the molecular mobility at lower temperatures leading to the glass transition temperature, using temperature-variable terahertz time-domain spectroscopy (THz-TDS) experiments. The microspheres were prepared using a water-in-oil-in-water (w/o/w) double-emulsion technique and subsequent freeze-drying of the samples. Physical characterization was performed by morphology measurements, scanning electron microscopy, and helium pycnometry. The THz-TDS data show two distinct transition processes, *T*g,*<sup>β</sup>* in the range of 167–219 K, associated with local motions, and *T*g,*α* in the range of 313–330 K, associated with large-scale motions, for the microspheres examined. Using Fourier transform infrared spectroscopy measurements in the mid-infrared, we were able to characterize the interactions between a model polypeptide, exendin-4, and the PLGA copolymer. We observe a relationship between the experimentally determined *T*g,*<sup>β</sup>* and *T*g,*<sup>α</sup>* and free volume and microsphere dynamics.

**Keywords:** terahertz spectroscopy; microspheres; drug delivery; formulation development; PLGA; molecular mobility

#### **1. Introduction**

Microencapsulation has been explored as a promising method for controlled drug release [1–3]. Polymeric microspheres, for example polylactic acid (PLA), polyglycolic acid (PGA), and poly(D,L-lactide-co-glycolide) (PLGA), can be used as effective drug-delivery systems protecting the encapsulated active agent and controlling the release rate over periods of hours to months [1,4]. By applying the correct methodology when preparing microspheres, challenges such as targeting the drug to a specific organ or tissue, and precisely controlling the rate of drug delivery to the target site can be addressed [3,5]. The technique of double water-in-oil-in-water (w/o/w) emulsion has been widely used for the encapsulation of hydrophilic drugs as this preparative method minimizes the loss of drug activity via contact with the organic solvent [2,6]. The basis of this methodology is to emulsify an aqueous solution of the active compound in an organic solution of the hydrophobic coating polymer. Then, this primary

water-in-oil (w/o) emulsion is dispersed in a second aqueous phase, forming a double water-in-oil-in-water (w/o/w) emulsion. The solid microsphere is produced as the organic solvent evaporates. In this work, we prepared blank PLGA microspheres (containing no polypeptide) of grades 50:50 and 75:25 (lactide:glycolide ratio), using the double-emulsion process [2]. Using the same technique, we also loaded PLGA microspheres at low and high concentrations of exendin-4, a glucose-dependent insulinotropic polypeptide used to treat type 2 diabetes [7]. The microspheres were subsequently freeze-dried. The objective of this work was to prepare, characterize, and understand the structural dynamics of these peptide-loaded biodegradable polymeric microspheres.

To optimize the formulation of microspheres and achieve release delivery of the active drug, the microsphere product ought to be stable [1,8]. When determining the chemical stability of a material, molecular mobility is a key factor [9–11] and it has been shown that an increase in molecular mobility is directly linked to an increase in the chemical degradation of a material [9] and aggregation [12], and therefore its storage stability. Thus, it is critical to understand the molecular mobility behavior of a material and its dependence on temperature, and specifically the molecular dynamics of the material leading up to the glass transition temperature (*T*g). It has been shown that a great number of polymers and amorphous pharmaceuticals, ranging from small molecules to high molecular weight peptides and proteins exhibit at least two dielectric relaxation processes: the primary, or *α*-relaxation process and a secondary, or *β*-relaxation process [13–15]. The former can be observed at temperatures above *T*<sup>g</sup> and can be designated as *T*g,*α*, and the latter occurs at temperatures below *T*<sup>g</sup> and is associated with a secondary glass transition process, which can be indicated as *T*g,*<sup>β</sup>* [16–20].

In previous work we investigated these processes in neat PLGA 50:50 and 75:25 of low, medium, and high molecular weights [21]. For all the copolymer PLGA film samples we observed three regions of relaxation behavior with distinct *T*g,*<sup>β</sup>* and *T*g,*<sup>α</sup>* transition temperatures. We showed that with an increase in temperature, the copolymer chains can transition through different conformational environments, each constrained by its characteristic potential energy landscape, and that the movement of each PLGA chain is restricted by adjacent entangled chains. At *T*g,*β*, the system has sufficient thermal energy and free volume to overcome the energy barrier arising from chain entanglement, allowing for local mobility to occur. Furthermore, we show that there is a relationship between free volume and the value of *T*g,*<sup>β</sup>* for the different film samples.

Fourier transform infrared spectroscopy (FTIR) is a vibrational spectroscopy used to gain structural information about a sample. The advantage of FTIR for structural analysis is it is a non-contact technique; only a small quantity of sample is needed (∼1 mg mL−1); the sample can be probed in different physical environments such as the solid state, liquid state, and when adsorbed to a surface; the sample preparation is minimal and a spectrum can be obtained in a few minutes [22]. For systems with multiple components such as polymer-peptide microspheres, FTIR can be used to identify the individual components, and interactions between the components which shift the peaks [23].

Terahertz time-domain spectroscopy (THz-TDS) is a valuable technique which can be used to detect *T*g,*<sup>α</sup>* and *T*g,*β*, associated with the *α*- and *β*-relaxation processes respectively [10,13]. It is a relatively recent technique which is used to investigate the molecular dynamics of relaxation processes at high frequencies [24]. The advantage of this technique is that it is a non-contact technique and can measure molecular mobility of a material over a broad temperature range over the spectral region of 0.1–3 THz. It can be used to investigate the microscopic mechanisms of amorphous polymer dynamics [25]. In the present study we examine samples of freeze-dried microspheres with different concentrations of encapsulated peptide. We investigate the molecular mobility of these materials and the behavior and trends these exhibit with respect to the temperature dependence using THz-TDS (Table 1), and physically characterise each system (Tables 2 and 3). Thus, the purpose of this work is to provide a comprehensive understanding of the relationship between the relaxation dynamics and the molecular structure of PLGA

microspheres with varying exendin-4 peptide concentrations, and to rationalize the behavior of these materials in relation to *T*g*<sup>α</sup>* and *T*g*β*.

#### **2. Materials and Methods**

#### *2.1. Materials*

Throughout this work the polymers are referred to by their monomer ratio used. For instance, PLGA 75:25 refers to a copolymer consisting of 75% lactic acid and 25% glycolic acid. Medium-MW (10–25 kDa) PLGA 50:50 and medium-MW (20–30 kDa) PLGA 75:25 were purchased from Evonik Corporation (Birmingham, AL, USA). The exendin-4 peptide (4.2 kDa) was provided by MedImmune Limited (Cambridge, UK).

#### *2.2. Microencapsulation Preparation*

Blank PLGA microspheres, were prepared for the study using the water-in-oil-in-water (w/o/w) double-emulsion technique: First, 500 mL of a 0.5% aqueous solution of polyvinyl alcohol (PVA) (87–90% hydrolyzed, 13–23 kDa) were emulsified in 12.5 mL of a 5% (w/v) PLGA dissolved in dichloromethane (DCM) by stirring at a rotation speed of 22,000 pm for 15 seconds using an ultra-turrax (IKA T-25, Cole-Parmer, UK). This primary emulsion was emulsified into 0.5% aqueous PVA under stirring at 200 rpm, on a stirring plate, for four hours. For the preparation of the peptide-loaded microspheres, exendin-4 was dissolved into a buffer consisting of citrate, citric acid, (pH = 4.5) and 0.5% PVA. This peptide solution was emulsified into the 5% (w/v) dispersion of PLGA in DCM, which was stirred at rotation speed 22,000 rpm for 15 s using the ultra-turrax. Different peptide loadings in the microspheres were achieved using exendin-4 concentrations of 1% and 10% (w/v) in the primary emulsion. Conceptually, the same method was followed to produce the peptide-loaded microspheres, as that for the blank microspheres. The solid microspheres were collected by centrifugation at 4000 *g* for 5 min, then washed with distilled water three times, and centrifuged once more at 4000 *g* for 5 min. Finally, after removal of the aqueous supernatant and dispersion, the microspheres were freeze-dried using a lyophilizer. First, prior to lyophilization, an annealing step was performed by cooling the shelf to 233 K for 240 min, raising the temperatures to 257 K for 200 min and cooling the shelf again to 233 K for 170 min at a pressure of 160 mbar. Then, lyophilization was performed using the following steps: primary drying was completed at 233 K for 30 min, and then the temperature was raised to 253 K for 2440 min at a pressure of 133 mbar; secondary drying was subsequently performed at 313 K for 960 min, also at 133 mbar. The vials were closed under a pressure of 266 mbar at 298 K using a rubber stopper, removed from the lyophilizer, and crimped with aluminum seals. Vials were stored at 278 K until all further measurements and analysis. Exendin-4 concentration was determined using the bicinchoninic acid protein assay kit (MilliporeSigma, St. Louis, MO, USA) following the manufacturer's instructions, with bovine serum albumin (BSA) used as the standard. The exendin-4 concentration was analyzed using a UV-VIS spectrophotometer (Agilent Cary 60 UV-Vis spectrophotometer, Agilent Technologies, Santa Clara, CA, USA) at 562 nm. The encapsulation efficiency-based BCA assay values were determined on the liquid samples. Based on these measurements, we calculated the encapsulation efficiency to be: 15.4% and 37.12% for the low and high peptide loading of PLGA 75:25 microspheres, respectively, and 42.88% and 36.72% for the low and high peptide loading of PLGA 50:50 microspheres, respectively. This experiment was completed one time. The water content for each lyophilized microsphere was determined using Karl Fischer (Mettler Toledo, Leicester, UK) coulometric titration.

#### *2.3. Helium Pycnometry Measurement*

The lyophilized microsphere samples were analyzed as powders using a helium pycnometer (Micromeritics Accupyc II 1340, Norcross, GA, USA) to determine the density and specific volume of the microparticles. Approximately 20 mg of each sample were placed in the instrument compartment and measurements were performed with helium gas at 298 K at a pressure of 10 mbar. No pretreatment conditions were required. The measurements were repeated for each sample five times. The average volume of the five repeat measurements was used to determine the density of each material. For comparison, polymer films of PLGA 50:50 and PLGA 75:25 were prepared using the vacuum compression molding (VCM) tool (MeltPrep, Graz, Austria), and these were also analyzed using helium pycnometry.

#### *2.4. Morphology Measurement*

The particle size and shape of the microsphere particles were characterized, for unlyophilized liquid samples and for lyophilized powder samples, using a Morphologi G3 instrument (Malvern Panalytical Ltd., Malvern, UK). Each unlyophilized liquid sample was dispersed in 1 mL of 5% aqueous PVA solution to allow for spatial separation of the particles and reduction of agglomerates. The lyophilized powder samples were prepared for measurement using a dry powder disperser. The instrument captured images of the individual particles and the morphological properties for each particle were determined from the images by image analysis using the Morphologi G3 software (Malvern Panalytical Ltd., Malvern, UK).

#### *2.5. Scanning Electron Microscopy*

The range of morphologies of the microspheres were characterized using scanning electron microscopy (SEM), with a Zeiss CrossBeam 540 instrument, equipped with a Gemini 2 column (Carl Zeiss Microscopy GmbH, Jena, Germany). To qualitatively analyze the shape and surface of the samples, the lyophilized microspheres were sputter coated with gold using an Emitech 550 (Emitech Ltd., Ashford, UK). The samples were then examined under vacuum at an acceleration voltage of 1 kV, and imaged using secondary electrons via an Everhart-Thornley detector (Carl Zeiss SMT GmbH, Oberkochen, Germany).

#### *2.6. Differential Scanning Calorimetry (DSC)*

A Q2000 Differential Scanning Calorimeter (TA Instruments, New Castle, DE, USA) was used to determine the calorimetric glass transition temperature (*T*g,DSC, defined by the onset temperature) for each material. 2–3 mg of sample material were placed in hermetically sealed aluminum pans under a constant flow of nitrogen atmosphere (flow rate of 50 mL min<sup>−</sup>1) and heated at a rate of 10 K min−<sup>1</sup> to 358 K, and subsequently cooled down to 293 K at 40 K min<sup>−</sup>1. Finally, the samples were heated from 293 K through *T*<sup>g</sup> to 358 K again at a rate of 10 K min−1. The temperature and heat flow of the instrument were calibrated using indium (*T*<sup>m</sup> = 430 K, Δ*H*fus = 29 J g<sup>−</sup>1).

#### *2.7. Fourier Transform Infrared Spectroscopy*

FTIR was used to examine the change in the secondary structure of the PLGA 75:25 microspheres at 278 K. For FTIR analysis, each microsphere material (300 μg) was mixed with 100 mg potassium bromide (KBr) using an agate mortar and pressed into 7 mm self-supporting disks using a load of 10 Tons. FTIR spectra were acquired using a Cary 680 FTIR spectrometer (Agilent Technologies, Santa Clara, CA, USA) with 60 scans and a resolution of 1 cm−1. At least four spectra were measured for each material. The recorded spectra were normalized based on the total area under the curve [26].

#### *2.8. Terahertz Time-Domain Spectroscopy (THz-TDS)*

#### 2.8.1. Sample Preparation and Experimental Methodology

For each sample 70 mg were weighed in under atmospheric protection in a glove bag (AtmosBag, Merck UK, Gillingham, Dorset, UK) which was purged with dry nitrogen gas (relative humidity < 1%) to avoid moisture sorption from atmospheric water vapor during preparation. The lyophilized powder samples were pressed into 13 mm diameter disks using a load of 1.5 tons. The tablets were between 300–650 μm in thickness each, and were placed between two z-cut quartz windows of 2.05 mm thickness. This sandwich structure was sealed in the sample holder, and used immediately following preparation for THz-TDS measurements.

The THz-TDS spectra were acquired using a commercial TeraPulse 4000 instrument across the spectral range of 0.2–2.2 THz (TeraView, Cambridge, UK). The sample temperature (90–360 K) was controlled using a continuous flow cryostat with liquid nitrogen as the cryogen (Janis ST-100, Wilmington, MA, USA) as outlined previously [27]. The cryostat cold finger accommodated both the reference (two z-cut quartz windows) as well as the sample (quartz/sample/quartz sandwich structure as described above). The two z-cut quartz windows that were used for the reference (same thickness and diameter dimensions as sample) were directly pressed to one another without any spacer in between the two windows to avoid internal reflections in the time-domain signal. The cryostat cold finger was moved vertically using a motorized linear stage to switch between sample and reference at each measurement temperature. For each temperature, the sample and the reference were measured at the center position, with 1000 waveforms co-averaged for each acquisition, resulting in a measurement time of approximately 1 min for each sample.

The temperature of the sample was measured using a silicon diode mounted to the copper cold finger of the cryostat. The temperature controller used was a Lake Shore model 331 (Westerville, OH, USA). For each series of measurements, a sample and a reference were loaded into the cryostat, the cryostat chamber was evacuated to 10 mbar and the cold finger was cooled to a temperature of 90 K. The cryostat was allowed to equilibrate for 10–15 min at 90 K and the first set of sample and reference measurements was acquired. Subsequently the cold finger was heated using temperature intervals of 10 K (at a rate of 2 K min<sup>−</sup>1). At each desired temperature point the system was allowed to equilibrate for 3 min before a set of sample and reference measurements were acquired.

#### 2.8.2. Data Analysis

To calculate the absorption coefficient and the refractive index of the sample a modified method for extracting the optical constants from terahertz measurements based on the concept introduced by Duvillaret et al. was used [27,28]. The changes in dynamics of the polymer sample were analyzed by investigating the change in the absorption coefficient at a frequency of 1 THz as a function of temperature using the methodology introduced in [21].

#### **3. Results**

#### *3.1. Differential Scanning Calorimetry Data*

The calorimetric *T*g,DSC was determined for each sample and the resulting values are listed in Table 1. We observed no significant difference in *T*g between samples. Additionally, we observed one *T*g for each material, indicating that no phase separation occurred for these samples.

**Table 1.** Gradient, *m*, of the linear fit (*y* = *mx* + *c*) for the respective temperature regions as outlined in Section 2.8.2 as well as the respective glass transition temperatures determined based on the terahertz analysis and by DSC.


#### *3.2. Morphology Measurement Analysis*

For each material we report the number of particles counted and the circular equivalent (CE) diameter *D*[*n*, 0.1], *D*[*n*, 0.5] and *D*[*n*, 0.9] percentiles. As shown from Table 2, preparing the microspheres using an oil-in-water emulsification yields products with a relatively broad size distribution for both the blank microspheres and the exendin-4 loaded microspheres before lyophilization. Following lyophilization, the particle size distribution is significantly narrower than before. This could be due to the removal water and moisture [5,9]. The Karl Fischer measurements we conducted revealed that the residual moisture for each vial was less than 1%. Notably, upon lyophilization, as water molecules are removed peptides and copolymers, for example, exendin-4, and PLGA microspheres, can form an extensive hydrogen bonding network [11,29,30] leaving few sites for bonding with water molecules. Additionally, the system is exposed to various stresses of temperature and pressure, and these could cause two or more adjacent pores to merge, as common pore walls rupture, and adjacent particles can combine. Thus, the removal of water in the drying step of lyophilization would affect the size distribution of the dry particles produced, and we can obtain a more narrow particle size distribution compared to a sample in solution [31]. It is worth noting that exendin-4 is inherently flexible and thus, chemical stability is of primary concern; however, given that exendin-4 is a peptide, it lacks a defined protein domain with a characteristic architecture that can undergo unfolding upon exposure to acute freezing and dehydration stresses of lyophilization [32,33]. Finally, when comparing the CE values of the lyophilized samples, the blank PLGA 50:50 samples have a higher CE compared to the 50:50 loaded samples, while the blank 75:25 samples have a lower CE compared to the 75:25 loaded samples. This can be explained by considering emulsion stability which dictates microsphere size. For example, exendin-4 would change the emulsion stability through surfactant-like activity, and the viscosity of the emulsion can increase with higher polymer concentration, polymer MW, and hydrophobicity [34]. Specifically, as the polymer is more hydrophobic, as is the case for PLGA 75:25 due to its a higher lactide fraction, more energy would be required to generate smaller droplets; while for PLGA 50:50 it is expected that smaller droplets could be generated with less energy input; however these parameters were not optimized for and beyond the scope of this work [33,34].


**Table 2.** Morphology data for unlyophilized samples and lyophilized samples. CE is the circular equivalent diameter: the diameter of a circle with the same area as the 2D projection image of the particle.

#### *3.3. Helium Pycnometry Analysis*

The pycnometry data shows that increasing exendin-4 loading increases the density of the material (Table 3). Our helium pycnometry measurements agree with values reported in the literature [4].



#### *3.4. Scanning Electron Microscopy (SEM) Characterization*

Using SEM, qualitative analysis of a representative set of images indicated that the microspheres examined are predominantly spherical in shape and exhibit a smooth and porous surface (Figure 1, see Supplementary Materials). The internal structure of fractured spheres revealed a porous interior. The qualitative analysis of the data showed that for the blank PLGA microspheres, there is no significant difference visually in the porosity of the blank 50:50 and 75:25 systems, respectively. In contrast, for both the 50:50 and 75:25 systems, the microspheres with high loading of peptide appeared more porous compared to the microspheres which contained a low loading of peptide. See Supplementary Materials for SEM data for blank PLGA 50:50 and 75:25 microspheres, and low polypeptide loaded PLGA 50:50 and 75:25 microspheres.

**Figure 1.** Representative SEM micrographs for low exendin-4 loaded PLGA 75:25 microspheres shown in (**a**–**c**), and high exendin-4 loaded PLGA 75:25 microspheres shown in (**d**–**f**).

#### *3.5. Fourier Transform Infrared Spectroscopy*

FTIR was performed for the PLGA 75:25 microsphere samples (Figure 2). The absorbance of the peaks at around 3300 cm−<sup>1</sup> and 2950 cm−<sup>1</sup> increased in intensity with an increase in polypeptide loading. The peak at 3300 cm−<sup>1</sup> originates from the -OH stretch in exendin-4, and the peak at 2950 cm−<sup>1</sup> is assigned to the C-H stretching modes in exendin-4 [26,35]. The amide I band at 1600–1700 cm−<sup>1</sup> can be attributed to the –C=O stretch, with contributions from the out-of-phase –CN stretch, –CCN deformation, and –NH in-plane bend and is sensitive to the structure of the protein backbone. Additionally, there is an increase in absorption in the amide I region around 1600–1700 cm−1, and a shift in the frequency of the mode for the sample with a high polypeptide loading, which is indicative of changes in the hydrogen bonding network [22]. Thus, the FTIR spectra confirm the increase in exendin-4 loading between the blank, low and high polypeptide formulations, and, with this, a change in the hydrogen bonding network. Notably, it was not possible to investigate an only exendin-4 control, as a stable product could not be freeze-dried.

**Figure 2.** FTIR spectra of blank microspheres (solid black line), low polypeptide loaded (solid red line) and high polypeptide loaded (solid blue line) PLGA 75:25 microspheres. (**a**) shows the wavenumber range of <sup>1300</sup> <sup>−</sup> 1700 cm−<sup>1</sup> and (**b**) shows the wavenumber range of 2500–4000 cm<sup>−</sup>1.

#### *3.6. Analysis of THz-TDS Data*

The terahertz spectra of all the microspheres showed an increase in absorption with frequency and temperature over the entire investigated range in line with previous measurements of amorphous molecular solids (Figure 3). As expected for non-crystalline materials, no discrete spectral features were present and the spectra were dominated by the monotonous increase with frequency that is characteristic for the rising flank of the peak due to the vibrational density of states (VDOS) [24]. In contrast, the refractive index subtly decreases with increasing frequency. To further investigate the relationship between the increase of absorption coefficient and temperature we examined the temperature-dependent changes in absorption losses at a frequency 1 THz in more detail. Given the lack of distinct spectral features we chose the frequency of 1 THz. The rationale for this choice is based on the fact that the signal-to-noise ratio of the measurement at this frequency is high and that we know from our previous work that at a frequency of 1 THz the minimum in losses in the spectral response is exceeded for all samples studied and hence the absorption is clearly dominated by the VDOS [27].

**Figure 3.** Absorption coefficient spectra of high MW PLGA 75:25 in the temperature range of 100–350 K, with 10 K temperature increments between spectra. Solid black line indicates the maximum absorption coefficient.

The changes in absorption at a frequency of 1 THz with temperature for the microsphere samples of PLGA 50:50 and PLGA 75:25, are plotted in Figures 4 and 5 respectively. For all the materials the absorption coefficient was found to increase in a linear fashion with increasing temperature and several distinct temperature regions can be identified for each material. *T*g,*<sup>β</sup>* was defined as the intersection point of the two best-fit linear fits at low temperatures for all cases, and *T*g,*α* was defined as the intersection point of the two best-fit linear lines at high temperatures, as outlined in [21].

**Figure 4.** Mean terahertz absorption coefficient as a function of temperature at 1 THz for PLGA 50:50 microspheres. Error bars represent the standard deviation for *n* samples. (**a**) blank (*n* = 3), (**b**) low (*n* = 4), and (**c**) high peptide loading (*n* = 3). Lines show the different linear fits of the respective regions.

**Figure 5.** Mean terahertz absorption coefficient as a function of temperature at 1 THz for PLGA 75:25 microspheres. Error bars represent the standard deviation for *n* samples (**a**) blank (*n* = 3), (**b**) low (*n* = 3), and (**c**) high peptide loading (*n* = 3). Lines show the different linear fits for the different regions.

For all of the microsphere samples the change in absorption with temperature can be observed to take place over three distinct regions and two transition temperatures, *T*g*<sup>β</sup>* and *T*g*α*, as determined using the methodology outlined above (see Figure 4 and 5 and Table 1). In this work we attempt to explain the origin of the transition temperatures by proposing a physical picture of the change in the microsphere dynamics with temperature and relate this to the free volume approach, and to peptide and polymer interactions. It is worth noting that the values of *T*g,*α*, as determined from the THz-TDS experiments, are in good agreement with our own calorimetric measurements, *T*g,DSC, as well as the values reported in the literature for these materials [36].

#### **4. Discussion**

#### *4.1. Understanding Peptide and Copolymer Interactions*

It has been shown that the *α*-relaxation process is associated with large-scale mobility, whereas the secondary or *β*-relaxation process is thought to be associated with local mobility, or small-scale mobility [24, 37]. To date, there remains a discussion in the literature as to the origin and molecular mechanisms associated with the *α*- and *β*-relaxation processes. One secondary relaxation, the Johari-Goldstein (JG) *β*-relaxation, also referred to as the slow *β*-relaxation, is considered a universal feature of all amorphous materials [38,39]. This process is observed at higher frequencies than the *α*-relaxation, and has been associated predominantly with the intermolecular degrees of freedom of a material [16,38,40]. Notably, recent experimental and theoretical work clearly highlights that the potential energy surface (PES) model proposed by Goldstein almost half a century ago is the most intuitive and comprehensive model to understand the molecular dynamics in amorphous systems and that intra- and intermolecular processes are always fundamentally coupled by means of the PES [41,42]. Goldstein explained that as a liquid flows it can move on the PES from one minimum to another minimum, and each minimum is associated with an energy barrier, yet as the liquid moves the volume nor energy of the liquid changes. When the liquid is cooled down to a glass, the liquid structure is trapped in a deep minimum, with some level of mobility remaining as a distribution of relaxation times [42]. Thus, the work of Goldstein provides an intuitive illustration that a liquid can exist in numerous transient structures and it is the potential energy barriers which determine the molecular motions of the viscous liquid which forms a glass: a picture which can be applied to bulk amorphous systems.

The different molecular motions of a copolymer chain can be tracked with changes in temperature. At low temperatures, the copolymer chains are completely disordered and are densely packed, and the motions of the copolymer are restricted. Upon heating to *T*g,*β*, the activation energy and free volume is sufficient for the copolymer to undergo local motion, giving rise to the *β*-relaxation processes [40]. Previously, we have shown that the *T*g,*<sup>β</sup>* is fundamentally linked to the onset of motions in an organic molecule that results in changes of the dihedral angle of one or multiple bonds in the system. Specifically, for PLGA, this could involve local motions of small segments of the copolymer backbone and side chain groups [37,43]. With a further increase in temperature, the copolymer chains are more loosely packed and the activation energy and free volume of the system increases further, allowing for intermolecular large-scale copolymer motions to occur at the temperature indicated by *T*g,*<sup>α</sup>* associated with the *α*-relaxation process.

Exendin-4 is a polypeptide composed of 39 amino acids, and its structure is thought to resemble a random coil chain [7,44]. The short-range local interactions of a polypeptide influence the conformational preferences of its amino acid chain [45–47]. In general terms, it is well established that the partial double bond character of a peptide bond gives rise to its planar structure, and free rotation is restricted about this bond [47]. We hypothesize that with sufficient thermal energy and free volume present in the system, two local motions could take place for exendin-4, which would give rise to the *β*-relaxation process: (1) local side chain rotations, and (2) local rotation about two single bonds [45,48]. Given that the amino acids of the exendin-4 polypeptide chain are linked by peptide bonds, rotational freedom arises from the single bonds between an amino group and the *α*-carbon atom and the carbonyl group of the peptide backbone [47].

However, the rotation of these bonds is limited by steric hindrance [47,48]. Based on the chemical structure of the exendin-4 this polypeptide can act as a hydrogen bond acceptor via its carbonyl groups (–C=O) or as a hydrogen bond donor via its amine group (–NH), and its hydroxyl groups (–OH) can act as both the hydrogen bond donor or acceptor. The linear PLGA—[C3H4O2]*x*[C2H2O2]*y*—chains include the methyl side groups of poly(lactic acid) (PLA) and oxygen atoms at every third position of the copolymer backbone, and C=O bonds which introduce significant structural rigidity to the copolymer backbone. Specifically, hydrogen bonds can form: (1) between the carbonyl groups of PLGA and the amine groups of exendin-4, (2) between the amine groups of exendin-4 and the hydroxyl group of PLGA, and (3) between the hydroxyl groups and the carbonyl groups of exendin-4 and PLGA. Additionally, van der Waals as well as dipole-dipole interactions between the peptide groups and the hydroxyl groups can serve to stabilize the system [49]. These strong interactions, the dynamics of which are infrared active, can reduce molecular mobility and improve the physical stability of these systems [3,29,30]. Notably, it is conceivable that with additional free volume and thermal energy input large-scale rotational motions of the backbone dihedral angles in the polypeptide chain could occur, which could lead to large-scale changes in chain conformation, contributing to the *α*-relaxation.

#### *4.2. Tracking the Dynamics of Microspheres Using THz-TDS*

We observe three regions with two distinct transition points, *T*g,*<sup>β</sup>* and *T*g,*<sup>α</sup>* for both the PLGA 50:50 and PLGA 75:25 microspheres (Figure 4). For the PLGA 50:50 formulation, the high exendin-4 loaded microspheres have a significantly higher *T*g,*<sup>β</sup>* = 219 K, than the blank and low polypeptide loaded microspheres (*T*g,*<sup>β</sup>* = 167 K, and *T*g,*<sup>β</sup>* = 168 K, respectively). In the polymer dispersion, exendin-4 can form hydrogen bonds with PLGA. Intuitively, an increase in exendin-4 loading would correspondingly increase the hydrogen bonding interactions between the polypeptide and PLGA [48], reduce the molecular mobility of the system, which would raise the value of *T*g,*<sup>β</sup>* [9,50]. Indeed, we observe a significant increase in *T*g,*<sup>β</sup>* for the high polypeptide loaded microspheres, compared to the *T*g,*<sup>β</sup>* value observed for the blank and low polypeptide loaded microspheres. Notably, for the blank and low exendin-4 loaded 50:50 microspheres, the onset for local mobility occurs at approximately the same value of *T*g,*β*. This could suggest that the low polypeptide loaded microsphere behave similarly to the blank microspheres, due to the limited interaction between peptide and polymer [46,51]. For the blank microspheres and low exendin-4 loaded microspheres, the onset for local mobility occurs at approximately the same value of *T*g,*β*, (*T*g,*<sup>β</sup>* = 167 K). For the high exendin-4 loaded microspheres, *T*g,*<sup>β</sup>* increases significantly (*T*g,*<sup>β</sup>* = 219 K). As more polypeptide is added to the PLGA matrix it forms an extensive hydrogen bonding network, which in turn reduces the configurational entropy of the system [44,45,51]. The resultant interactions between exendin-4 and PLGA appear to be stronger compared with the interactions between adjacent PLGA chains.

For the samples of the PLGA 75:25 microspheres, we observe a similar trend to the PLGA 50:50 microspheres. The value of *T*g,*<sup>β</sup>* increases from 179 K for the blank microspheres, increases further to 192 K for the low exendin-4 loaded microspheres, and rises to *T*g,*<sup>β</sup>* = 215 K, for the high exendin-4 loaded microspheres (Figure 5). Notably, we have previously shown that the methyl side group of PLGA 75:25 inhibits polymer mobility and introduces steric hinderance [43]. Thus, for PLGA 75:25, the steric hinderance caused by the lactide monomer can restrict the intermolecular interactions between exendin-4 and PLGA, limiting the sites that are able to participate in hydrogen bonding, most likely making the carbonyl group of PLGA more favorable for hydrogen bonding. With limited sites for hydrogen bonding, less hydrogen bonds can form between exendin-4 and PLGA, resulting in PLGA chain entanglement, lower mobility, and reduced free volume. Our measurement that the PLGA 75:25 microspheres exhibit higher *T*g,*<sup>β</sup>* values compared to the 50:50 microspheres could therefore be explained by steric effects.

Finally, we observe that for both PLGA 50:50 and PLGA 75:25 the value of *T*g,*α* increases in the order of blank < low polypeptide loaded < high polypeptide loaded microspheres (Figure 5). This suggests that at high temperature, sufficient activation energy and free volume must be available to facilitate mobility of the polymer and polypeptide. Thus, with increase in exendin-4 loading, due to steric hinderance the threshold for mobility increases, as reflected in the raised value of *T*g,*α* (Figure 6). Ideally if it were possible to produce a stable freeze-dried exendin-4 product, it may be feasible to determine whether the exendin alone could aggregate and lead to the changes observed.

**Figure 6.** Behavior of the different materials with increasing steric hinderance and free volume. Blue lines represent the PLGA copolymer (20–30 kDa MW), and solid red circles represent the exendin-4 molecules (4.2 kDa MW). Low loading and high loading refers to 1 mg/mL and 10 mg/mL of polypeptide loaded in PLGA microspheres, respectively. With an increase in free volume and decrease in steric hinderance, the values of *T*g,*<sup>β</sup>* is decreased.

#### **5. Conclusions**

We have produced and characterized lyophilized blank PLGA microspheres and lyophilized PLGA microspheres containing two different loadings of the polypeptide exendin-4. We studied the dynamics and relaxations and the glass transition behavior of these PLGA microspheres by performing variable temperature THz-TDS measurements. A monotonous increase of absorption coefficient with temperature was observed for all the materials examined, and all of the microspheres exhibit three temperature regimes, with a distinct *T*g,*<sup>β</sup>* and *T*g,*α*. We explain our experimental results using the concepts of free volume and discuss the interactions of the polypeptide and copolymer matrix and steric effects. We define *T*g,*<sup>β</sup>* as the point at which the material has sufficient amount of activation energy and free volume to allow for local motions to occur, and *T*g,*α* as the point at which large-scale movement can take place, and relate the onset of the transition temperatures to the interaction strength between the polymer and the peptide. Our work provides a physical explanation for the behavior of these microspheres leading to *T*g, and agrees with the PES concept outlined by Goldstein [42]. This work provides a framework for understanding the dynamics of complex systems, such as lyophilized microspheres, and considers the parameter of *T*g,*<sup>β</sup>* as a valuable criterion for preparing stable formulations. Finally, this work demonstrates that THz-TDS is an effective method to measure the molecular dynamics and temperature-dependent behavior of a polymer-polypeptide microsphere system.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/1999-4923/11/6/291/ s1: Figure S1: Representative SEM micrographs for blank PLGA 50:50 microspheres shown in (a–c) and blank PLGA 75:25 microspheres shown in (d–f), Figure S2: Representative SEM micrographs for low exendin-4 loaded PLGA 50:50 microspheres shown in (a–c) and high exendin-4 loaded PLGA 50:50 microspheres shown in (d–f), Figure S3: FTIR spectra of blank microspheres (solid black line), low polypeptide loaded (solid red line) and high polypeptide loaded (solid blue line) PLGA 75:25 microspheres, Figure S4: MDSC thermogram of PLGA 50:50 blank microsphere, Figure S5: MDSC thermogram of PLGA 50:50 low peptide loading microsphere, Figure S6: MDSC thermogram of PLGA 50:50 high peptide loading microsphere, Figure S7: MDSC thermogram of PLGA 75:25 blank microsphere, Figure S8: MDSC thermogram of PLGA 75:25 low peptide loading microsphere, Figure S9: MDSC thermogram of PLGA 75:25 high peptide loading microsphere, Figure S10: Terahertz absorption spectra of a blank PLGA 75:25 microsphere sample over 0.3–2.8 THz in the temperature range of 100–360 K, Figure S11: Refractive index spectra of a blank PLGA 75:25 microsphere sample over 0.1–2.8 THz in the temperature range of 100–360 K, Figure S12: Terahertz absorption spectra of a blank PLGA 50:50 microsphere sample over 0.3–2.8 THz in the temperature range of 100–350 K, Figure S13: Refractive index spectra of a blank PLGA 50:50 microsphere sample over 0.1–2.8 THz in the temperature range of 100–350 K, Figure S14: Terahertz absorption spectra of a low polypeptide loaded PLGA 75:25 microsphere sample over 0.3–2.8 THz in the temperature range of 100–350 K, Figure S15: Refractive index spectra of a low polypeptide loaded PLGA 75:25 microsphere sample over 0.1–2.8 THz in the temperature range of 100–350 K, Figure S16: Terahertz absorption spectra of a low polypeptide loaded PLGA 50:50 microsphere sample over 0.3–2.8 THz in the temperature range of 100–370 K, Figure S17: Refractive index spectra of a low polypeptide loaded PLGA 50:50 microsphere sample over 0.1–2.8 THz in the temperature range of 100–350 K, Figure S18: Terahertz absorption spectra of a low polypeptide loaded PLGA 75:25 microsphere sample over 0.3–370 THz in the temperature range of 100–350 K, Figure S19: Refractive index spectra of a low polypeptide loaded PLGA 75:25 microsphere sample over 0.1–3 THz in the temperature range of 100–370 K, Figure S20: Terahertz absorption spectra of a high polypeptide loaded PLGA 50:50 microsphere sample over 0.3–2.8 THz in the temperature range of 100–370 K, Figure S21: Refractive index spectra of a high polypeptide loaded PLGA 50:50 microsphere sample over 0.1–2.8 THz in the temperature range of 100–370 K, Figure S22: Terahertz absorption spectra of a medium MW PLGA 50:50 over 0.3–2.2 THz in the temperature range of 90–360 K, Figure S23: Refractive index spectra of a medium MW PLGA 50:50 over 0.3–2.2 THz in the temperature range of 90–360 K, Figure S24: Terahertz absorption spectra of a medium MW PLGA 75:25 sample over 0.3–1.9 THz in the temperature range of 90–350 K, Figure S25: Refractive index spectra of a medium MW PLGA 75:25 sample over 0.3–1.9 THz in the temperature range of 90–350 K. All raw data are available for download at https://doi.org/10.17863/CAM.40737.

**Author Contributions:** Conceptualization and methodology, T.A.S., C.F.v.d.W. and J.A.Z.; infrared spectroscopy, P.J.H.; formal analysis, T.A.S. and P.J.H.; data curation, T.A.S.; writing—original draft preparation, T.A.S.; writing—review and editing, all authors; supervision, J.A.Z., C.F.v.d.W. and G.S.K.S.; project administration, J.A.Z.; funding acquisition, J.A.Z.

**Funding:** The authors acknowledge funding from AstraZeneca UK Limited (Medlmmune Limited) and the UK Engineering and Physical Sciences Research Council (EP/N022769/1). T.A.S. would like to thank the AJA-Karten Trust and the AIA-Kenneth Lindsay Trust for their financial support.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **Abbreviations**

The following abbreviations are used in this manuscript:


#### **References**




© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Pharmaceutics* Editorial Office E-mail: pharmaceutics@mdpi.com www.mdpi.com/journal/pharmaceutics

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18