**Toxicity of Carbon Nanomaterials and Their Potential Application as Drug Delivery Systems: In Vitro Studies in Caco-2 and MCF-7 Cell Lines**

**Rosa Garriga 1,\*, Tania Herrero-Continente 2, Miguel Palos 3, Vicente L. Cebolla 4, Jesús Osada 2,5, Edgar Muñoz <sup>4</sup> and María Jesús Rodríguez-Yoldi 3,5,\***


Received: 23 July 2020; Accepted: 16 August 2020; Published: 18 August 2020

**Abstract:** Carbon nanomaterials have attracted increasing attention in biomedicine recently to be used as drug nanocarriers suitable for medical treatments, due to their large surface area, high cellular internalization and preferential tumor accumulation, that enable these nanomaterials to transport chemotherapeutic agents preferentially to tumor sites, thereby reducing drug toxic side effects. However, there are widespread concerns on the inherent cytotoxicity of carbon nanomaterials, which remains controversial to this day, with studies demonstrating conflicting results. We investigated here in vitro toxicity of various carbon nanomaterials in human epithelial colorectal adenocarcinoma (Caco-2) cells and human breast adenocarcinoma (MCF-7) cells. Carbon nanohorns (CNH), carbon nanotubes (CNT), carbon nanoplatelets (CNP), graphene oxide (GO), reduced graphene oxide (GO) and nanodiamonds (ND) were systematically compared, using Pluronic F-127 dispersant. Cell viability after carbon nanomaterial treatment followed the order CNP < CNH < RGO < CNT < GO < ND, being the effect more pronounced on the more rapidly dividing Caco-2 cells. CNP produced remarkably high reactive oxygen species (ROS) levels. Furthermore, the potential of these materials as nanocarriers in the field of drug delivery of doxorubicin and camptothecin anticancer drugs was also compared. In all cases the carbon nanomaterial/drug complexes resulted in improved anticancer activity compared to that of the free drug, being the efficiency largely dependent of the carbon nanomaterial hydrophobicity and surface chemistry. These fundamental studies are of paramount importance as screening and risk-to-benefit assessment towards the development of smart carbon nanomaterial-based nanocarriers.

**Keywords:** cytotoxicity; carbon nanomaterials; drug delivery; doxorubicin; camptothecin; Caco-2; MCF-7

#### **1. Introduction**

Carbon nanomaterials are promising new materials to be used as drug nanocarriers suitable for medical treatments in biomedicine, due to their large surface area and chemical stability that allows efficient loading of drugs via both covalent and non-covalent interactions [1–3]. Although their interaction with lipid membranes and their subsequent intracellular transport is poorly understood,

it has been demonstrated that they can enter cells using various endocytic processes [4,5]. A combination of increased tumor vascular permeability and insufficient lymphatic drainage, resulting in what is termed as enhanced permeability and retention (EPR) effect, enables these nanoparticles to transport chemotherapeutic agents preferentially to tumor sites as compared to healthy tissues, thereby reducing toxic side effects [6]. Furthermore, these systems could be used for formulation of hydrophobic molecules which lack of suitable physicochemical characteristics required for development of stable pharmaceutical dosage form. Transition metal contamination from synthesis procedures can be avoided by purification [7,8]. Poor aqueous dispersibility and high aggregation tendency of pristine carbon nanomaterials can be sorted out by appropriate surface functionalization towards their applications as nanocarriers [9]. Functionalization of carbon nanomaterials can be achieved by either non-covalently coatings with amphiphilic macromolecules like lipid, polymers and surfactants, or covalently with hydrophilic functional groups. Specifically, in vivo studies have shown that functionalization with polyethylene glycol (PEG) allows to achieve prolonged circulation half-life, resulting in so-called 'stealth' behavior, and therefore improved accumulation in tumor by escaping opsonization-induced reticuloendothelial system (RES) clearance [10–13], making these nanocarriers good candidates for cancer diagnostics and treatment. Also PEGylated nanomaterials exhibited remarkably reduced in vivo toxicity, avoiding accumulation in liver or spleen.

To further enhance the therapeutic efficacy of drugs and simultaneously diminish their undesirable systemic side effects, different small targeting molecules such as folic acid (FA) [14,15], ligands with strong affinity against a given receptor overexpressed in a tumor [16], antibodies that recognize tumor-associated antigens [17–20] and also magnetic nanoparticles [3,15] can be further incorporated onto the drug-loaded carbon nanomaterials to confer either active targeting capabilities via receptor-mediated endocytosis or local nanocarrier accumulation induced by external magnetic field. However, addition of excess targeting ligands also increases clearance by the RES because more proteins are now 'visible' on the surface than with PEG. Also, adsorption of proteins and other biomolecules could shield the targeting ligands that have been grafted onto the surface of nanocarriers from binding to their receptors on tumors. 'Stimuli-responsive' drug delivery, that is, strategies to release drug cargo upon experiencing certain tumor-specific triggers (i.e., higher temperature and lower pH) can be extremely useful for selective and controllable drug release [15,21].

There is a trend to combine the therapeutics and early diagnostics (namely theranostics) together [3,18,22–26]. By combining imaging labels with therapeutics in the same platform, the location of the tumor can be precisely delineated, and the optimal drug doses as well as therapeutic time frame could be determined by acquiring the real-time drug distribution information in vivo. Imaging tags like radioactive nuclides [27–30] and fluorescence probes [30,31] can be conjugated in carbon multifunctional nanoplatforms to observe their intracellular trafficking and biodistribution in vitro and in vivo. Raman signals from nanocarbon materials can also provide a reliable method to monitor their distribution and metabolism in vivo [32,33].

In addition, carbon nanotubes, carbon nanohorns and reduced graphene demonstrate strong optical absorption in the near-infrared (NIR) region, making them promising materials for use in the photothermal ablation of tumors [25,34–36]. Carbon nanomaterials also offer promise in combination therapy, that generally refers to two or more therapeutic agents co-delivered simultaneously, and is becoming more popular because it generates synergistic anticancer effects, enabling a low dosage of each compound and overcoming multi-drug resistance (MDR) cancer [37,38]. While the delivery of cancer chemotherapeutic agents with carbon nanomaterials has been more widely attempted, carbon nanomaterials have also demonstrated promising potentials to be used to deliver many non-anticancer drugs, such as antimicrobial, anti-inflammatory, antihypertensive and anti-oxidant agents [1,26].

However, the biomedical applications of carbon nanomaterials arouse serious concerns, as more information on the pharmacokinetics, metabolism, long-term fate and toxicity is essential [27,29,31,39–41]. The issues of toxicity surrounding the biomedical applications of

carbon nanomaterials still remain controversial to this date, with studies demonstrating conflicting results [42–45]. Carbon nanomaterial-based drug delivery systems are still considered far from being accepted for use in actual clinical settings. Progress towards clinical trials will depend on the outcomes of efficacy and toxicology studies, which will provide the necessary risk-to-benefit assessments for carbon-based materials. Fundamental studies regarding the impact of size, shape, aggregation degree and functional groups of carbon nanomaterials are needed to provide the design criteria for successful nanomaterial-based strategies. In carbon nanomaterial safety assessment, in vitro cytotoxicity tests are an important research subject because they are fast, reproducible and easy to control the consistency of experimental conditions and should complement and/or supplement in vivo-animal tests. In vitro studies are able to provide information on the biological fate of nanomaterials at the cellular or multicellular levels. In the literature, there is a lack of comparative studies, as extensive variations in the nanomaterial source, functionalization, and experimental conditions do not allow direct comparison of the different results.

We here investigated in vitro toxicity of various Pluronic (F-127)-dispersed carbon nanomaterials in human epithelial colorectal adenocarcinoma (Caco-2) and human breast adenocarcinoma (MCF-7) cell lines. Representative examples, with different size, shape and functional groups on the surface, such as carbon nanohorns (CNH), carbon nanotubes (CNT), graphene nanoplatelets (CNP), graphene oxide (GO), reduced graphene oxide (RGO) and carbon nanodiamonds (ND) were systematically compared under the same experimental conditions.

Furthermore, their potential in the field of drug delivery of anticancer drugs, such as doxorubicin (DOX) and camptothecin (CPT), was also compared in this work. DOX and CPT have been chosen here as examples of hydrophilic and hydrophobic drugs, respectively. DOX belongs to anthracyclines, topoisomerase II inhibitors exhibiting multiple mechanisms of action and high clinical effectiveness against many types of cancer. Notably, cardiotoxicity is a major concern during therapy as it may be dose-limiting [46]. More effective and safer ways of delivering anthracyclines are hence of significant research interest. Also, resistance to anthracyclines and other chemotherapeutics due to P-glycoprotein (P-gp), a membrane transporter that actively pumps doxorubicin out of the cell, is a frequent problem in cancer treatment [47]. It has been reported that nanocarriers enhance doxorubicin uptake in drug-resistant cancer cells [48–50]. Thus, DOX enters the cells attached to nanocomposites bypassing the P-gp transporter, detaches from the nanocomposite surface following natural acidification of endosomes, and migrates reaching the cell nucleus. On the other hand, CPT is a potent anticancer agent with topoisomerase I-inhibiting activity. However, its practical use in viable cancer therapeutic systems is greatly hampered due to its low solubility in aqueous media [51]. The need to formulate water-soluble salts of CPT (that is, alkaline solutions for intravenous injections) led to chemical modifications of the molecule with loss of anti-tumor activity [52,53]. Thus, developing new drug delivery nanocarriers for CPT able to transport and deliver the drug inside the cancer cells has recently received considerable attention [54].

Finally, it has to be noted that the cancer cell lines targeted here, Caco-2 and MCF-7, correspond to cancers among those having the highest incidence in Western countries, hence the interest of anticancer therapeutic studies performed on them.

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

#### *2.1. Carbon Nanomaterials*

CNT used here are short (average length < 1 μm) and purified (95% wt. %) multi-walled carbon nanotubes from Nanocyl S.A. (Sambreville, Belgium), NANOCYL® NC3150™, produced via catalytic chemical vapor deposition (CCVD) process. CNH were single wall carbon nanohorns provided by Carbonium Srl (Padua, Italy), produced without any catalyst, by rapid condensation of small carbon clusters (C2 and C3) resulting from direct vaporization of graphite [55]). Single layer graphene oxide (GO, purity 99 wt. %) was supplied by Cheap Tubes Inc. (Grafton, VT, USA). RGO was from Sigma-Aldrich (777684, Darmstadt, Germany). CNP (purity 91 at.%.) and detonation nanodiamonds (ND, purified/grade G01) were purchased from PlasmaChem GmbH (Berlin, Germany).

#### *2.2. Characterization of Carbon Nanomaterials*

Transmission electron microscopes (TEM, Tecnai T20 and Tecnai F30, FEI, Hillsboro, OR, USA, operating at 200 and 300 KV, respectively) were used to characterize the structural features of carbon nanomaterials. During sample preparation, nanomaterials were dispersed in ethanol, and a drop was placed onto carbon coated copper grids, the sample excess was wicked away by means of a Kimwipe and allowed to dry under ambient conditions. Prior to TEM imaging, the samples on the grids were placed in a O2-Ar (20% O2) plasma cleaner (Model 1020 Fischione, Hanau, Germany) for 5–10 s to remove organic (hydrocarbon) contamination.

X-ray photoelectron spectroscopy (XPS) was performed on powder samples deposited onto double-sided carbon tape using an ESCA Plus spectrometer (Omicron, Taunusstein, Germany) provided with a Mg anode (1253.6 eV) working at 225 W (15 mA, 15 kV). CasaXPS software (version 2.3.15, accessed on 1 June 2020) was used for the peak deconvolution and Shirley type baseline correction was applied.

Nitrogen adsorption-desorption isotherms were measured at 77 K (Micromeritics ASAP 2020, Micromeritics Instrument Corp., Norcross, GA, USA) and surface area measurements of the powder samples were obtained using the Brunauer–Emmett–Teller (BET) method at values of relative pressure (p/p0) between 0.05 and 0.3.

#### *2.3. Dispersion of Carbon Nanomaterials*

Pluronic® F-127 (F-127), suitable for cell culture, average molecular weight 12.6 μDa, was purchased from Sigma.Aldrich (Darmstadt, Germany). F-127 solutions at 15 <sup>μ</sup>g·mL−<sup>1</sup> and 10 min bath sonication (100 W Branson 2510 bath sonicator, Branson Ultrasonics, Danbury, CT, USA) were used here to assist in carbon nanomaterials dispersion at 3.0 and 0.6 <sup>μ</sup>g·mL−<sup>1</sup> concentration in cell culture medium.

Dispersions of carbon nanomaterials were prepared in cell culture medium without fetal bovine serum (FBS), as it is known that bovine serum albumin (BSA) has different affinity towards carbon nanomaterials. Thus, it has been reported that BSA readily adsorbed on GO, resulting in a decrease in GO toxicity. In contrast, BSA loading capacity was ∼9-fold lower for MWCNT [56].

DOX and CPT loading on carbon nanomaterials was performed by simply mixing of solutions in cell culture media, agitated by using a vortex mixer and kept overnight in dark at room temperature. Due to its poor solubility in aqueous media, CPT was initially dissolved in dimethyl sulfoxide (DMSO, <sup>≥</sup>99.9%, from Sigma-Aldrich, Darmstadt, Germany) to a concentration of 1.6 mg·mL−1, and then diluted on cell culture medium to the required working concentrations.

#### *2.4. Cell Lines and Cell Culture*

Human Caco-2 cell line (TC7 clone) was kindly provided by Edith Brot-Laroche (Université Pierre et Marie Curie-Paris 6, UMR S 872, Les Cordeliers, France). Caco-2 cells (passages 40–60) were cultured in Dulbecco's Modified Eagles medium (DMEM) (Gibco Invitrogen, Paisley, UK) supplemented with 20% fetal bovine serum (FBS), 1% non-essential amino acids, 1% penicillin (1000 U/mL), 1% streptomycin (1000 μg/mL) and 1% amphotericin (250 U/mL), at 37 ◦C under a humidified atmosphere with 5% CO2. The cells were passaged enzymatically with 0.25% trypsin-1 mM EDTA and sub-cultured on 25 cm<sup>2</sup> plastic flasks at a density of 5 <sup>×</sup> 10<sup>5</sup> cells/cm2. Culture medium was replaced every 3 days. Experiments were performed in undifferentiated cells (24 h post-seeding to prevent cell differentiation).

Human breast adenocarcinoma MCF-7 cells were kindly provided by Carlos J. Ciudad and Dr. Verónica Noé (Departamento de Bioquímica y Fisiología, Facultad de Farmacia, Universidad de Barcelona, Spain). MCF-7 cells were maintained in the same conditions as described for Caco-2 cell line.

For comparison purposes, some experiments were performed with human dermal fibroblasts that were kindly provided by Dr. Julio Montoya (Departamento de Bioquímica y Biología Molecular, Facultad de Veterinaria, Universidad de Zaragoza, 50013 Zaragoza, Spain).

#### *2.5. Cell Viability Assay*

24 h after seeding in 96-well plates at a density of 4 <sup>×</sup> 10<sup>3</sup> cells/well, cells were treated for 24 and 72 h with carbon nanomaterial dispersions (including DOX and CPT anticancer drugs in some studies), and then 3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide (MTT) assay was performed for assessing cell metabolic activity. In short, 10 <sup>μ</sup>L of MTT (5 mg·mL−1) were added to each 100 μL sample well and incubated for 2 h. Mitochondrial dehydrogenases of viable cells reduce the yellowish water-soluble MTT to water-insoluble formazan crystals, which are later resolubilized by replacement of the medium with DMSO, obtaining a purple colored solution. Absorbance at 540/620 nm was measured using a SPECTROstar Nano microplate reader (BMG Labtech, Ortenberg, Germany). Control values (sample wells without treatment) were set at 100% viable and all values were expressed as a percentage of the control. All experiments were performed in triplicate. In each of the three independent experiments, each sample result corresponds to 16 wells, which sums up 48 wells per sample.

#### *2.6. Reactive Oxygen Species (ROS) Assay*

The reactive oxygen species (ROS) production was assayed by the 2- ,7- -dichlorofluorescein diacetate (H2DCFDA) molecular probe [57,58]. The cell-permeable H2DCFDA diffuses into cells and is deacetylated by cellular esterases to form 2- ,7- -dichlorodihydrofluorescein (H2DCF). In the presence of ROS, H2DCF is rapidly oxidized to 2- ,7- -dichlorofluorescein (DCF), which is highly fluorescent. Caco-2 and MCF-7 cells were seeded in 96-well plates at a density of 4 <sup>×</sup> 103 cells/well, incubated 24 h under standard cell culture conditions and then treated with nanomaterial dispersions (3 <sup>μ</sup>g·mL<sup>−</sup>1) for 24 h. Subsequently, cells were washed twice with PBS and incubated for 20 min with 100 μL of 20 μM H2DCFDA at 37 ◦C for in the dark. Fluorescence intensity (ex = 485/em = 535 nm) was measured with FLUOstar Omega microplate reader (BMG Labtech). % ROS production was compared to a negative control (untreated cells) and was normalized with MTT assays at 24 h incubation. All experiments were performed in triplicate. In each of the three independent experiments, each sample result corresponds to 16 wells, which sums up 48 wells per sample.

#### *2.7. Cell Death Study*

Caco-2 and MCF-7 cells were plated in 75 cm<sup>2</sup> flasks at a density of 5 <sup>×</sup> 105 cells per flask and incubated 24 h under standard cell culture conditions. They were then exposed to dispersions of the tested carbon nanomaterials (3 <sup>μ</sup>g·mL−1) for 72 h. Each sample result corresponds to a pool of two 75 cm<sup>2</sup> flasks. Quantitative flow cytometry (FCT) analysis was performed using propidium iodide (PI) intake and FITC annexin V staining according to manufacturer's instruction. Briefly, cells were washed twice with phosphate saline buffer (PBS) and 100 μL of annexing V-binding buffer (10 mM HEPES/NaOH pH 7.4, 140 mM NaCl, 2.5 mM CaCl2) were transferred to a 5 mL culture tube. Additions of 5 μL FITC annexin and 5 μL PI were made to each tube and then incubated for 15 min in the absence of light at room temperature. Cells were then resuspended in 400 μL of annexin V-binding buffer and analyzed with BD FACSAria flow cytometer (BD FACSDIVA version 7.0 software, accessed on 1 June 2020). Untreated cells were used as negative control and the positive control corresponds to cells treated with CPT (0.8 <sup>μ</sup>g·mL<sup>−</sup>1). Preliminary gating was used in flow cytometry analysis to identify the cells of interest based on the relative size and complexity of the cells, while removing debris and other events that are not of interest.

#### *2.8. Cell Cycle Assay*

Caco-2 and MCF-7 cells were plated in 75 cm<sup>2</sup> flasks at a density of 5 <sup>×</sup> 105 cells per flask and incubated 24 h under standard cell culture conditions. Each sample result corresponds to a pool of two 75 cm<sup>2</sup> flasks. They were exposed to carbon nanomaterial dispersions (3 <sup>μ</sup>g·mL−1) for 72 h and then washed with PBS, collected and fixed for 30 min at 4 ◦C and incubated with 70–80% ice-cold ethanol at −20 ◦C for 24 h. After washing with PBS and 5 min centrifugation at 2500× *g* rpm, cells were resuspended in PI/RNase staining buffer. PI-stained cells were analyzed for DNA content with a BD FACSArray bioanalyzer. PI fluorescence was measured in the orange range of the spectrum using a 562–588 nm band pass filter, and cell distribution was displayed on a linear scale. The percentage of cells on each cell cycle phase was determined by means of BD ModFit LT version 3.3 software (accessed on 1 June 2020).

#### *2.9. Statistical Analysis*

The experimental data were analyzed by one-way analysis of variance (ANOVA) followed by Bonferroni post-test using GraphPad Prism software (version 5.02, GraphPad Software, Inc., San Diego, CA, USA, accessed on 1 June 2020). Interval plots display 95% confidence intervals for the mean. Data were presented as means ± S.D. and differences were considered significant at *p* < 0.05.

#### **3. Results**

#### *3.1. Characterization of Carbon Nanomaterials*

Frequently, the most likely source of the apparent lack of uniformity in the results reported in the literature for in vitro and in vivo studies is the different structural and physicochemical properties of the diverse nanomaterials used. Thus, there are huge dissimilarities (i.e., length, diameter, surface defects, oxygen content, presence of impurities, etc.) among the batches employed by researchers. Therefore, thorough characterization studies of the carbon nanomaterials are required and must be taken into consideration to obtain meaningful results.

#### 3.1.1. Transmission Electron Microscopy (TEM)

The characterization of the structural features and textural properties of the carbon nanomaterials tested here provides useful information on their interaction with drugs and cells. CNH are conical-shaped single-walled tubules that arrange into 100 nm dahlia-like assemblies (Figure 1a). The CNT used here are relatively short MWCNT (up to 1 micron in length) and ~10 nm in diameter, comprising around six concentric nanotubes (Figure 1b). TEM micrographs of two-dimensional, graphene derivatives GO and RGO (Figure 1d,e, respectively) reveal that most flakes are up to 1 micron in length as well as their high exfoliation degree. On the contrary, CNP consist of aggregates of smaller, less exfoliated graphene sheets (Figure 1c). Finally, Figure 1f shows aggregates comprising ND of about 5 nm in diameter.

(**a**) (**b**) (**c**)

**Figure 1.** TEM micrographs of the tested carbon nanomaterials: (**a**) CNH, (**b**) CNT, (**c**) CNP, (**d**) RGO, (**e**) GO and (**f**) ND. Inset in (**b**) shows image analysis of a MWCNT with six concentric nanotubes.

(**d**) (**e**) (**f**)

#### 3.1.2. Photoelectron Spectroscopy (XPS)

X-ray photoelectron spectroscopy (XPS) provides important hints of the surface chemistry of the tested carbon nanomaterials (Figure S1). The ratio of oxygen and carbon atoms was calculated from the O1s and C1s peaks, and the results of the quantitative surface analysis are summarized in Table 1. XPS spectra of CNH, CNT, CNP and RGO are quite similar and correspond to C sp2-based nanomaterials materials, with a low O:C ratio and, therefore, are highly hydrophobic. In contrast, GO has a significant high oxygen content (49.2 at.%), as it contains abundant oxygen-containing

functional groups, which provide enhanced hydrophilicity. Although O content in ND is not as high as in GO, ND are known to disperse easily in polar solvents, as it will be commented later in the Discussion section. No significant transition metal contamination was observed in XPS spectra.

**Table 1.** Surface chemical analysis (at.%) of the carbon nanomaterials, obtained from XPS spectra.


<sup>1</sup> For ND, at.% N is 3.0, calculated from the N1s peak in XPS spectra.

#### 3.1.3. Specific Surface Area

Specific surface area for carbon nanomaterials in powder, determined using N2 adsorption and BET method, are shown in Table 2.

**Table 2.** Specific surface area of carbon nanomaterials determined by BET method.


The largest specific surface area value corresponds to CNP. While these values correspond to powder samples, sonication-assisted dispersion in solution significantly increases surface area, which is particularly relevant when it comes to GO exfoliation.

#### *3.2. Dispersion of Carbon Nanomaterials*

For efficient cellular uptake of carbon nanomaterials, it is necessary that that they remain dispersed and not aggregate in culture medium. Non-ionic polyether surfactants, such as poloxamer triblock copolymers (known also by the trade name Pluronic®), are frequently used as dispersants to prepare various nanoparticle suspensions, especially with hydrophobic nanoparticles, such as CNT and related materials. Pluronics are amphiphilic molecules that comprise two polyethylene glycol (PEG) blocks and one polypropylene glycol (PPG) block of various sizes and are frequently used for in vitro and in vivo nanotoxicity studies because they are considered non-toxic dispersants. Thus, the US Food and Drug Administration (FDA) has approved various Pluronic polymers for pharmaceutical usage and even intravenous administration [59,60]. However, it is known that Pluronics can be degraded during sonication, depending on sonication time, power, and frequency conditions, as the collapse of cavitation bubbles generated during sonication can create sufficient heat, pressures, and shear forces to degrade polymers containing PEG, PPG or both. It is therefore important to assess whether sonication of dispersants themselves contribute to the toxicity of sonicated nanomaterial suspensions so as not to misinterpret toxicity results [61]. Figure S2 shows that F-127 decreased MCF-7 and Caco-2 cell viability at high concentration. Thus, F-127 at low concentration (15 <sup>μ</sup>g·mL<sup>−</sup>1) and short bath sonication time (<10 min) was used here to assist in carbon nanomaterials dispersion in cell culture medium, while avoiding the generation of toxic degradation products. Moreover, it is well documented in the literature that above critical micelle concentration (CMC), Pluronics form nano-sized micellar structures which can act as drug nanocarriers, showing higher anticancer activity as compared to free drug [59,62]. It was also checked here that neither DOX nor CPT anticancer activity was enhanced due to drug encapsulation in F-127 micellar structures at the low F-127 concentration used here (Figure S3). Therefore, any improvement achieved in this study in cell killing ability over free drug against cancer cells can be attributed to the drug-nanocarrier complex.

#### *3.3. Carbon Nanomaterials Toxicity Assessment*

Cell viability assay, apoptosis detection, cell cycle analysis and ROS production assay are useful in vitro methods for the assessment of toxicity of nanomaterials.

#### 3.3.1. Cell Viability Assay

Figure 2 shows cell viability assays on carbon nanomaterial treatment at 3 <sup>μ</sup>g·mL−<sup>1</sup> after 24 and 72 h for Caco-2 and MCF-7 cell lines. Also results at 0.6 <sup>μ</sup>g·mL−<sup>1</sup> can be found in Figure S4. The MTT assays showed dose-dependence on both Caco-2 and MCF-7 cell lines. Cell viability followed the order of CNP < CNH < RGO < CNT < GO < ND. The decrease in cell viability was more pronounced for the Caco-2 cell line. No significant cell viability decrease was observed in Figure 2 for GO and ND at <sup>3</sup> <sup>μ</sup>g·mL−<sup>1</sup> (and also for CNT, at 0.6 <sup>μ</sup>g·mL<sup>−</sup>1, as shown in Figure S4).

**Figure 2.** Cell viability assays after 24 h (white) and 72 h (striped) of incubation with various carbon nanomaterials at 3 <sup>μ</sup>g·mL−<sup>1</sup> showing differential effects on (**a**) Caco-2 and (**b**) MCF-7 cells. Values that are significantly different from the control (*p* < 0.05) are denoted with asterisk (\*). Untreated cells were used as control.

Interestingly, MTT studies on human dermal fibroblast cells (Figure S5), as example of healthy cells, provided the same viability sequence after carbon nanomaterial treatment, showing less effects than on Caco-2 cells and very similar to those on MCF-7 cells.

#### 3.3.2. Reactive Oxygen Species (ROS) Assays

According to Figure 3, CNP produced the highest levels of ROS, more noticeably for Caco-2 cells.

**Figure 3.** ROS generation on (**a**) Caco-2 and (**b**) MCF-7 cells upon incubation with nanomaterial dispersions (3 <sup>μ</sup>g·mL<sup>−</sup>1) for 24 h. Significant results as compared to untreated control cells are marked by asterisk \* for *p*-value < 0.05.

#### 3.3.3. Cell Death Study

Figure 4 shows quantitative flow cytometry analyses for the Caco-2 and MCF-7 cell lines, treated 72 h with CNH at 3 <sup>μ</sup>g·mL−1, which showed the highest values of late apoptosis/necrosis among the carbon nanomaterials tested here. Experiments were performed as described in Section 2.7, and interpreted as follows: the percentage of viable cells is shown in the lower left quadrant (annexin V−/PI−), of early apoptotic cells in the lower right quadrant (annexin V+/PI−), and of late apoptotic and necrotic cells in the upper right quadrant (annexin V+/PI+). Additional results for CNT, CNP and RGO are compared in Figure S6.

**Figure 4.** Annexin V/Propidium iodide assay, providing quantitative information about living (lower left quadrant), early apoptotic (lower right quadrant) and late apoptotic and necrotic (upper right quadrant) cells for (**a**) Caco-2 and (**b**) MCF-7 cell lines, treated with CNH at 3 <sup>μ</sup>g·mL−<sup>1</sup> for 72 h. Data are presented as percentage of the cell population. Untreated cells, denoted as C, were used as negative control, and CPT (0.8 <sup>μ</sup>g·mL<sup>−</sup>1) treated cells were used as positive control.

#### 3.3.4. Cell Cycle Analysis

Figure 5 shows flow cytometric analysis of Caco-2 and MCF-7 cell cycle after treatment with CNH at 3 <sup>μ</sup>g·mL−<sup>1</sup> for 72 h. Additional results for CNT, CNP and RGO are compared in Figure S7.

**Figure 5.** Flow cytometric analysis of (**a**) Caco-2 and (**b**) MCF-7 cell cycle after treatment with CNH at <sup>3</sup> <sup>μ</sup>g·mL−<sup>1</sup> for 72 h. C denotes untreated control cells.

#### *3.4. Carbon Nanomaterials as Anticancer Drug Nanocarriers*

The potential in the field of drug delivery as nanocarriers of anticancer drugs of the four carbon nanomaterials studied here that showed the lowest effect on the cells (ND, GO, CNT and RGO) was compared in Figure 6. MTT assays were performed on Caco-2 cells at two drug concentrations, 0.2 and 0.8 <sup>μ</sup>g·mL−1. Carbon nanomaterial concentration was chosen as low as 0.6 <sup>μ</sup>g·mL−1, so that the observed decrease in cell viability could be attributable to the improved DOX or CPT efficacy when loaded on carbon nanomaterial nanocarriers rather than to any inherent toxicity of carbon nanomaterials. CPT showed more potent cytotoxic activity than DOX against both cancer cells (Figure 6). Carbon nanomaterial/drug complexes resulted in improved anticancer activity compared to that of the free drug. For CPT, the improvement follows the sequence ND < GO < CNT < RGO. For DOX, the sequence is the opposite (Figure 6). Thus, CNT and RGO showed significant enhanced anticancer activity compared to the free drug, but ND showed a significant improvement when it comes to DOX.

**Figure 6.** Cell viability assays after 24 h (white) and 72 h (striped) of incubation of Caco-2 cells with ND, GO, CNT and RGO at 0.6 <sup>μ</sup>g·mL−<sup>1</sup> concentration, free drug CPT (**a**) and DOX (**b**) at both 0.2 and 0.8 <sup>μ</sup>g·mL−<sup>1</sup> concentrations, and CPT- (**a**) and DOX- (**b**) loaded carbon nanomaterials. (\* and # represent significance at *p* < 0.05 when compared to untreated control cells and free drug-treated cells, respectively).

#### **4. Discussion**

CNT are a type of hollow one-dimensional (1D) carbon-based nanomaterial consisting of a graphene sheet rolled up to form a cylindrical structure with sp2 hybridized carbon atoms. CNT are classified into single-walled carbon nanotubes (SWCNT) and multi-walled carbon nanotubes (MWCNT), have high aspect ratios and needle-like shapes [63]. Comparing the two types, there has been a major debate over whether SWCNTs or MWCNTs generate more toxicity. Some research groups have reported that SWCNT cause more apoptosis than MWCNT, as they are more agglomerated [64–66]. Moreover, short CNT were found to be less toxic than longer CNT, which is comparable with the observed toxicity of asbestos [44,65,67,68]. CNT used here are MWCNT, with relatively short length (mean length 1 μm, Figure 1b), so that low toxicity is expected. The purity of this CNT material is high (>95.0%) so no significant toxicity should result from any traces of the transition metal nanoparticles used during CNT production.

Single-walled carbon nanohorns (SWCNH) are horn-shaped single-walled tubules with cone angles of approximately 20◦ that usually form aggregates with diameters of 80–100 nm [69,70] with a "dahlia-like" shape, as shown in Figure 1a. They are produced essentially metal-free and with high purity [71]. Their use in biomedical applications is still at a preliminary stage. SWCNH used here were produced without any catalyst by direct vaporization of graphite, as described in Section 2.1.

Another category of carbon nanomaterial is graphene, a two-dimensional (2D) sp2-bonded carbon sheet in a honeycomb structure and therefore, pristine graphene is hydrophobic in nature. On the contrary, GO contains abundant epoxy and hydroxyl functional groups attached to the basal plane and carboxylic groups attached to the edges, that disrupt the π conjugation, providing enhanced hydrophilicity which even enables the efficient dispersion in aqueous media. π conjugation and therefore hydrophobicity are partially restored upon reduction in RGO [72,73]. Size and morphological characteristics of graphene derivatives studied here, CNP, RGO and GO, are shown in Figure 1c–e.

As another important member of carbon nanomaterial family, ND consist of a highly ordered diamond core covered by a layer of functional groups on the surface, such as carboxyl, lactone, hydroxy and ketone, which stabilizes the particle by terminating the dangling bonds [74,75]. ND produced by detonation method are extremely tiny particles with average diameter between 4–6 nm (Figure 1f). ND are becoming increasingly useful in therapeutic and diagnostic applications due to their biocompatibility, scalability, and easy surface modification [76,77].

According to the XPS results summarized in Table 1, CNH, CNT, CNP and RGO nanomaterials have a low O:C ratio and can be considered as hydrophobic and difficult to disperse in polar solvents. On the contrary, GO has a remarkable high oxygen content and can be considered hydrophilic. Although oxygen content in ND is lower than in GO (Table 1), ND are known to disperse easily in polar solvents, which is due to the hydrophilic functional groups on the outer shell. No significant transition metal contamination for the tested nanomaterials was observed by TEM and XPS. Thus, we can claim that the toxicological effects of metal impurities in these nanomaterials are negligible.

Amphiphilic F-127 was used here to assist the dispersion of carbon nanomaterials in cell culture media through noncovalent functionalization, which involves the coating of the carbon nanomaterials with hydrophobic PPG motifs anchored onto the material surface, with the hydrophilic PEG ends extending to the aqueous solution and enabling the stability of the material in aqueous media.

Results of MTT assays upon treatment with carbon nanomaterials at 3 <sup>μ</sup>g·mL−<sup>1</sup> (Figure 2) and 0.6 <sup>μ</sup>g·mL−<sup>1</sup> (Figure S4) show that the cell viability was cellular type, time and dose-dependent. Viability decrease was more pronounced on the highly active metabolically Caco-2 cells. Cell viability follows the order CNP < CNH < RGO < CNT < GO < ND for both Caco-2 and MCF-7 line cells. The sequence in cell viability that resulted from the MTT assays for the different carbon nanomaterials tested here can be explained taking into account the surface chemistry of carbon nanomaterials. Thus, oxygen functional groups on the surface of carbon nanomaterials shield the hydrophobic domains. Two groups of carbon nanomaterials can be distinguished here, the hydrophilic ones, ND and GO, which present low effect on cells, and the hydrophobic ones, CNT, RGO, CNH and CNP, which inhibited cell viability in more extent. The highest viability values correspond to ND, whose surface is rich in functional groups, which make them ideal nanocarriers for building drug delivery systems. However, as it will later be discussed, ND efficiently load hydrophilic drugs, such as DOX, which readily attach to their functional groups on their surface, rather than hydrophobic drugs, such as CPT.

Figure 3 shows that, compared to the other carbon nanomaterials, CNP produced the highest ROS levels, more pronounced for Caco-2 cells. We also found enhancement of ROS levels for cells treated with CNH respect to those treated with CNT and RGO. No significant ROS level alterations were however observed for ND and GO.

As for the apoptosis study, the combination of annexin V and PI has been used to discriminate early apoptotic cells from late apoptotic and necrotic ones. Results collected in Figure 4 and Figure S6 show that the hydrophobic carbon nanomaterials induced late apoptosis/necrosis for both Caco-2 and MCF-7 cells, being more pronounced for CNH. Anticancer drug CPT at 0.8 <sup>μ</sup>g·mL<sup>−</sup>1, whose toxicity was much larger than that of all carbon nanomaterials studied here, was used as positive control.

The effect of carbon nanomaterials on cell cycle progression in Caco-2 and MCF-7 cells is shown in Figure 5 and Figure S7. Cytometric analysis showed no significant differences in the percentage of cells in the individual phases of the cell cycle for all the tested carbon nanomaterials and untreated cells, particularly for MCF-7 cells

Taking these results all together, we conclude that ND and GO show low toxicity, which is due to the oxygenated functional groups on their surface that shield the hydrophobic domains. On the other hand, CNH and CNP induce Caco-2 and MCF-7 late apoptosis/necrosis and enhanced ROS levels, which could be associated with the higher decrease in cell viability, compared to other hydrophobic carbon nanomaterials such as CNT and RGO. This could probably be due to the "dahlia-like" CNH morphology, consisting of small structures containing sharp conical ends, that may produce damage to cells, as well as the sharp edges of the highly fragmented CNP platelets [78,79]. Finally, CNP were found to induce the most elevated levels of ROS, which would contribute to the highest observed decrease in cell viability. The effect is noticeably more pronounced on the more rapidly dividing Caco-2 cells.

It is worth noting that CNH was reported to inhibit proliferation of human liver cell lines and promoted apoptosis [80]. In contrast, other authors reported low toxicity for CNH [69,81,82]. It has to be noted that low toxicity reports correspond to CNH synthesis methods leading to oxidized CNH, such as CO2 laser ablation or arc discharge. Thus, oxygen functional groups on the surface would shield hydrophobic carbon domains from interactions with cellular membranes. CNH studied here were produced by direct vaporization of graphite, with very low O content, which was confirmed by XPS (Table 1). This highlights the importance of carbon nanomaterial source when drawing meaningful conclusions from toxicity studies.

Finally, the potential of carbon nanomaterials materials in the field of drug delivery of anticancer drug was compared here. Drug delivery systems based on noncovalent interactions have several advantages compared with covalent conjugation. Thus, extra steps required in chemical conjugations are not necessary. Also, because the drug structure is not chemically altered, drug molecules released from such delivery systems are expected to exert their predicted pharmacological effects. Many clinically used chemical drugs possess aromatic rings, such as DOX, playing the π–π stacking interactions the major role in drug delivery systems [83]. It is known that the loading efficiency for DOX decreases when using CNTs with higher levels of PEGylation, due to the increased hydrophilicity of the surface. Furthermore, faster release rates of DOX were observed for these higher PEGylated CNTs owing to the lower affinity of DOX to the PEGylated CNT [84].

Due to their sp<sup>2</sup> carbon structure and inherent hydrophobic nature of carbon nanomaterials, all of them (except ND) are capable of establishing noncovalent π–π stacking interactions for the formation of anticancer DOX and CPT complexes. As for the hydrophobic drug CPT (Figure 7a), the more hydrophobic the carbon nanomaterial is, the more C sp2 domains has and the more efficient is the loading of CPT, through strong π–π interactions, which explains the results shown in Figure 6. Thus, the highest improvement in CPT anticancer activity compared to the free drug was observed for RGO and CNT nanocarriers.

On the other hand, because of its high surface free energy, ND rarely exist as a single particle, and usually form clusters of tens to hundreds of nanometers, even when they are dispersed in a solution by strong ultrasonication. Drug molecules can be assembled on the surface of ND clusters or in the nanoscale pores inside the ND clusters (Figure 7b) by noncovalent interactions [77,85]. The highest improvement in DOX activity compared to that of the free drug was observed for ND. However, ND were not efficient in loading the more hydrophobic drug, CPT. Results for higher ND concentration up to 20 <sup>μ</sup>g·mL−<sup>1</sup> shown in Figure S8 show that the efficiency was worse than those at 0.6 <sup>μ</sup>g·mL−1, probably due to ND aggregation forming higher size clusters, which offer less surface area for drug loading and more difficulty to enter the cells.

**Figure 7.** Schema of mechanism of interaction between drugs (in red) and carbon nanomaterials (in black): (**a**) CPT/CNT and (**b**) DOX/ND. (F-127 is depicted in blue).

Carbon nanomaterials display unique physicochemical properties making them potentially useful for bioapplications and competitive when compared to micelles, polymeric nanoparticles, dendrimers, and liposomes, to name a few. Thus, they offer high surface area for multiple drug adsorption through π–π stacking interactions and, as for ND, drugs bound to the abundant functional groups on their surface show enhanced chemotherapeutic efficacy. Much research activity has been devoted to perform in vivo experiments, either by systemic administration and localized drug delivery strategies [86]. Remarkably, carbon nanomaterials have also received much attention in imaging and diagnostics. Thus, due to their strong absorption in the IR or NIR regions, they can be used in cancer photothermal therapy (PTT). Also, they are useful in fluorescence [87,88] and photoacoustic imaging (PAI) [89–91]. Intrinsic carbon nanomaterial Raman vibrations allow monitoring their in vivo distribution and metabolism [32,33]. ND presenting nitrogen-vacancy centers have intrinsic fluorescence properties, and therefore are interesting tools for imaging and diagnostics [75]. Finally, carbon-based nanomaterials are emerging as potential candidates for the development of synthetic scaffolds in tissue engineering [92–95].

Long-term fate of carbon nanomaterials has been the subject of much concern and the origin of much skepticism surrounding their in vivo applications, as are presumed to be biopersistent. Despite discrepancies in findings on the clearance mechanism, majority of the studies have suggested that increasing the degree of functionalization enhanced renal clearance, while lower functionalization promoted RES accumulation (i.e., liver and spleen) [96]. Several groups have reported that carbon-based nanomaterials are susceptible to biodegradation as a result of the key role played by the immune system [97].

#### **5. Conclusions**

Cytotoxicity evaluation after 24 h and 72 h of incubation with various carbon nanomaterials shows differential effects on Caco-2 and MCF-7 cells. Cell viability followed the order CNP < CNH < RGO < CNT < GO < ND, being more pronounced in the more rapidly dividing Caco-2 cells. ND and GO showed the lowest toxicity, due to the presence of oxygen functional groups on carbon nanomaterial surface, that shield the hydrophobic carbon domains. High hydrophobicity, together with

the morphology containing sharp conical ends in CNH and sharp edges in CNP would account for the high cell viability decrease, enhanced ROS level and apoptosis/necrosis. Remarkable high ROS levels were obtained for CNP, more pronounced on Caco-2 cells.

There is a lack of ROS generation from both cell lines after incubation with ND, as well as the lowest apoptosis values, which further supports that ND provide the lowest toxicity among the carbon nanomaterials tested here, which make them an ideal carrier for designing drug delivery systems. ND form clusters of tens to hundreds of nanometers, wherein drugs can be loaded by interaction with their surface functional groups, and therefore ND will be much more efficient in loading hydrophilic drugs, such as DOX, which readily attach to their functional groups on their surface, rather than hydrophobic drugs. In contrast, CNT and RGO, which also have low toxicity among the hydrophobic carbon nanomaterials tested here, offer available surface area for π–π interactions with aromatic rings, leading to high CPT loading efficiency, due to the strong π–π stacking interactions formed with CPT. Remarkably, CPT is a more potent anticancer agent than DOX, so developing new drug delivery systems for CPT is of high interest.

Several obstacles must be overcome before carbon nanomaterials can be suitable for clinical use. The major challenge and current limitation in this area is still the potential long-term toxicity concerns of carbon nanomaterials. Comparative in vitro studies of cytotoxicity of carbon nanomaterial synthesized from different sources are needed as screening and risk-to-benefit assessment, together with drug loading efficiency studies, to further develop advanced multi-functional carbon nanomaterials for cancer theranostic applications.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2079-4991/10/8/1617/s1, Figure S1: High resolution XPS spectra of carbon nanomaterials; Figure S2: Cell viability assays showing the effect of F-127 at two concentrations; Figure S3: Cell viability assays after treatment with DOX and CPT, in the presence and absence of F-127; Figure S4: Cell viability assays after treatment with carbon nanomaterials at 0.6 <sup>μ</sup>g·mL<sup>−</sup>1; Figure S5: Cell viability assays after treatment with carbon nanomaterials at 3.0 <sup>μ</sup>g·mL−<sup>1</sup> on human dermal fibroblast cells; Figure S6: Cell death study comparing different hydrophobic carbon nanomaterials; Figure S7: Cell cycle analysis comparing different hydrophobic carbon nanomaterials; Figure S8: Cell viability assays after treatment with ND at several concentrations, DOX and CPT.

**Author Contributions:** R.G. and M.J.R.-Y. were responsible for the overall direction of the research. V.L.C. and E.M. contributed with nanomaterial characterization supervised by R.G., M.P. and T.H.-C. performed cell culture experiments supervised by R.G. and M.J.R.-Y., R.G., M.J.R.-Y. and J.O. analyzed the obtained data. R.G. drafted the manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was supported by the Diputación General de Aragón (project E25\_20R); the Spanish Ministry of Economy and Innovation under Grant (SAF 2016-75441-R); Aragón Regional Government (B16-17R, Fondos FEDER "Otra manera de hacer Europa"); CIBEROBN under Grant (CB06/03/1012) of the Instituto de Salud Carlos III and European Grant Interreg/SUDOE (Redvalue, SOE1/PI/E0123).

**Acknowledgments:** The TEM microscopy work has been conducted in the "Laboratorio de Microscopías Avanzadas" at "Instituto de Nanociencia de Aragón—Universidad de Zaragoza". The authors acknowledge the LMA-INA for offering access to their instruments and expertise. Quantitative flow cytometric studies were performed at "Centro de Investigación Biomédica de Aragón", Spain (http://www.iacs.aragon.es). The authors thank Alfonso Ibarra, Rodrigo Fernández-Pacheco, Ariel Ramírez-Labrada, Elvira Aylón and Isaías Fernández for valuable technical support and fruitful discussions.

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

#### **References**


© 2020 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/).

## **Delivery of siRNA to Ewing Sarcoma Tumor Xenografted on Mice, Using Hydrogenated Detonation Nanodiamonds: Treatment E**ffi**cacy and Tissue Distribution**

**Sandra Claveau 1,2, Émilie Nehlig 3, Sébastien Garcia-Argote 3, Sophie Feuillastre 3, Grégory Pieters 3, Hugues A. Girard 4, Jean-Charles Arnault 4, François Treussart 1,5,\*,**† **and Jean-Rémi Bertrand 2,**†


Received: 7 February 2020; Accepted: 15 March 2020; Published: 19 March 2020

**Abstract:** Nanodiamonds of detonation origin are promising delivery agents of anti-cancer therapeutic compounds in a whole organism like mouse, owing to their versatile surface chemistry and ultra-small 5 nm average primary size compatible with natural elimination routes. However, to date, little is known about tissue distribution, elimination pathways and efficacy of nanodiamonds-based therapy in mice. In this report, we studied the capacity of cationic hydrogenated detonation nanodiamonds to carry active small interfering RNA (siRNA) in a mice model of Ewing sarcoma, a bone cancer of young adults due in the vast majority to the *EWS-FLI1* junction oncogene. Replacing hydrogen gas by its radioactive analog tritium gas led to the formation of labeled nanodiamonds and allowed us to investigate their distribution throughout mouse organs and their excretion in urine and feces. We also demonstrated that siRNA directed against *EWS-FLI1* inhibited this oncogene expression in tumor xenografted on mice. This work is a significant step to establish cationic hydrogenated detonation nanodiamond as an effective agent for in vivo delivery of active siRNA.

**Keywords:** nanodiamond; tritium; biodistribution; Ewing sarcoma; drug delivery; siRNA; nanomedicine

#### **1. Introduction**

The use of nanoparticles as vectors for drug delivery has been largely described by the scientific community during these past decades, with numerous applications [1,2]. Nanoparticles facilitate intracellular delivery and protection of the cargo against degradation, therefore they present many advantages for the vectorization of small nucleic acids such as small interfering RNA (siRNA). The latter are used to control gene expression by silencing targeted genes. Considering their intrinsic poor cellular penetration and low stability in biologicals medium [3,4], siRNA must be associated to an effective carrier. Different types of siRNA transporting agents have been reported either organic

(e.g., liposomes, cationic polymers or dendrimers [5]), inorganic (e.g., metallic nanoparticles such as gold, iron, titanium [4]) or mineral like clay [6], silica nanoparticles [7] and nanodiamonds [8,9]. Most inorganic nanocarriers present a good-to-high chemical stability; they have a low toxicity at therapeutic dose of the drug:carrier conjugate and are able to deliver their cargo compounds into cells. Nevertheless, the high stability of inorganic nanoparticles goes with the fact that they are not biodegradable, and for pharmacological applications, this can be a limiting factor. Therefore, for the safe development of these particles, it is crucial to determine their possible elimination pathways after administration and study how they are distributed in the body.

Here, we report on the use of cationic detonation nanodiamonds (DND) for the delivery of siRNA directed against Ewing sarcoma (ES) junction oncogene *EWS-FLI1* to ES tumor xenografted on mice. Ewing sarcoma is a rare bone and soft tissue cancer [10] which is caused in the vast majority of cases by the formation of *EWS-FLI1* oncogene. To carry siRNA, DND surface needs to be modified to be able to bind negatively charged nucleic acids by electrostatic interactions. One strategy relies on cationic polymer coating of diamond surface, which is either done by electrostatic interaction [9,11] or by covalent grafting [12,13]. However, polymer coating may lead to the formation of large aggregates by crosslinking DND to other DND via polymer chains bridging. Another strategy, the one selected for this study, is based on the direct surface modification of DND using hydrogen gas combined with microwave plasma or with annealing method. Both hydrogenation methods were recently compared [14]. It was shown that after such reductive treatment, the hydrogenated DND (H-DND) can be dispersed in water, and acquire a positive surface charge characterized by a zeta potential of ≈+50 mV [15]. In a previous work, we described that H-DND can carry efficiently siRNA to ES cultured cells and promote specific targeted *EWS-FLI1* oncogene inhibition [16]. We want to extend this strategy to preclinical study now. Considering the very high chemical stability of diamond, these further investigations have to consider the risk of accumulation in the body after inoculation, as already described for larger nanodiamonds (around 50 nm diameter), produced by a different process than detonation [17]. However, owing to their small unitary size (3–8 nm) DND are good candidates for in vivo applications. Indeed, particles smaller than the filtration cutoff of kidney can be eliminated through urines, after glomerular passage, and since this limit for kidney is around 7–10 nm (depending of the molecule shape) [18,19], H-DND could be eliminated through this pathway. Indeed, Riojas et al. [20] showed that hydroxylated 7-nm sized DND further functionalized with radiolabeled amino groups are efficiently eliminated in urines if the solution is filtered before being intravenous injected.

In this work, we describe an original method that we developed to treat ES in vivo by injecting DND:siRNA complexes, and to determine the organ distribution of these DND. To this aim, we used DND labeled with tritium by annealing method [21]. We show that siRNA (i) can be loaded onto hydrogenated or tritiated DND (H-DND and T-DND, respectively), (ii) can efficiently inhibit *EWS-FLI1* in ES tumor model xenografted on mice and (iii) that the organ distribution and elimination of T-DND can be measured thanks to its radioactivity.

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

#### *2.1. Preparation of Hydrogenated and Tritiated DND and Associated Suspensions*

The procedures are detailed in reference [21]. DND (Reference: G02 grade, primary size 3–8 nm, from PlasmaChem GmbH, Berlin, Germany) were first manually milled in a mortar and then annealed using appropriate gas.

#### 2.1.1. H-DND Aqueous Suspension

A mass of 30 mg of as-received and manually milled DND were placed in a quartz tube (3.5 mL) with an isolation valve, an in/out gas connection and connected to a cold trap. Vacuum was created and H2 gas was loaded at 250 mbar pressure. The tube was placed in an oven, and connection was

made with a trapping set-up. The oven was turned on during 1 h at 550 ◦C. The powder was then pumped for 30 min before disconnection and air exposure. Particles were dispersed in ultrapure water (resistivity: 18.2 MΩ.cm) and sonicated (Model UP400S, 300 W and 24 kHz, from Hielscher GmbH, Teltow, Germany) for 1 h under a cooling system. Aggregates of particles were then removed by centrifuging the suspension for 40 min (acceleration: 2400× *g*, speed: 4754 rpm) and the supernatant was collected, forming the stock H-DND aqueous suspension. The final concentration was calculated by measuring the mass of particles after drying a calibrated volume of the supernatant.

#### 2.1.2. Mixture of T-DND and H-DND (Later Denoted T-DND to Simplify) Aqueous Suspension

As-received manually milled DND were first pre-oxidized to remove native C-H groups to achieve a more quantitative control of tritium added to the surface [21]. To this aim, 100 mg was placed in an alumina crucible under air for 1 h 30 min at 550 ◦C. Then, 30 mg of these pre-oxidized DND were annealed for 4 h at 550 ◦C, using the same process that for H-DND with 10% T2 and 90% H2 gas mixture (in order to obtain the desired activity). To remove all the labile tritium atoms, the treated powder was poured into methanol (4 mL) and the solvent was evaporated. The operation was repeated twice and the powder was then stored dried under a nitrogen atmosphere. Quantification of the tritium incorporation was assessed by measuring the activity (using liquid scintillation counting) in the combustion gas after the total combustion of T-DND in air (3 h at 600 ◦C). We measured a total specific activity of 13 mCi.mg<sup>−</sup>1.

The synthesized T-DND (4 mg) were dispersed in ultrapure water (3 mL) and sonicated (3 mm conical microprobe Vibracell 75043, 750 W, 28% amplitude, Sonics & Materials, Inc., Newton, CT, USA) for 1 h under a cooling system. To remove highly aggregated particles, the suspensions were centrifuged for 40 min (2400× *g*, 4754 rpm) followed by supernatant separation. A liquid scintillation counting was performed on the supernatant, yielding a specific activity of 18 mCi/mL.

To reduce the high specific activity of the suspension, T-DND from the supernatant (10 μL) were mixed with a solution of H-DND (2 mg/mL, 2.1 mL). The later was prepared following Section 2.1.1 method with a small adjustment: in order to be consistent with the tritium treatment, this H-DND suspension was also prepared from pre-oxidized DND treated for 4 h under H2 at 550 ◦C. The final DND solution was split in 3 sealed vials (0.7 mL each). The total activity measured by liquid scintillation counting of this final solution was 97 μCi/mL.

#### 2.1.3. Additional Pre-Injection Washing of T-DND Solution.

Before injection to mice, the T-DND solution was centrifugated (acceleration 10,600× *g* for 3 h at 10 ◦C) in order to eliminate residual labile tritium atoms by exchange with water. Washed T-DND could then be stably suspended in water, and only 2% of the total initial tritium radioactivity was lost in the supernatant (see Supplementary Table S1).

#### *2.2. Hydrodynamic Size and Electrophoretic Mobility Characterizations DND Solution*

Hydrodynamic size and zeta potential of H-DND and DND:siRNA complexes in solution were measured by dynamic light scattering (DLS) using a NanoBrook 90Plus PALS (Brookhaven Instruments, Holtsville, NY, USA) in 1 cm thick cuvette in deionized water. Hydrodynamic sizes are inferred from the scattered intensity autocorrelation function. The latter was then analyzed with the non-negative constrained least squared method [22], which is one of the methods of reference to infer the size from polydisperse suspensions. The size values reported correspond to the dominant population. For radioactive nanodiamonds, measurements were performed in a sealed cuvette.

#### *2.3. siRNA Sequences and Binding siRNA to Hydrogenated or Tritiated DND and siRNA Binding Capacity Assay*

#### 2.3.1. siRNA Sequences

siRNA was purchased from Kaneka Eurogentec S.A. (Seraing, Belgium). The sequence complementary to the *EWS*/*FLI1* fusion oncogene (siAS) is: sense strands 5- -GCAGCAGAACCCUUC UUAUd(GA)-3 and antisense strand 5- -AUAAGAAGGGUUCUGCUGCd(CC)-3- . The control irrelevant sequence (siCt) is: sense strand 5- -CGUUACCAUCGAGGAUCCAd(AA)-3 and the antisense strand 5- -UGGAUCCUCGAUGGUAACGd(CT)-3- .

#### 2.3.2. Binding siRNA to Hydrogenated or Tritiated DND and siRNA Binding Capacity Assay

siRNA complexation to cationic DND was performed by slowly dropping a siRNA solution to cationic DND solutions placed in a sonication bath (Ultrasonic cleaner, VWR International S.A.S., Fontenay-sous-Bois, France) at its maximum power during 10 min, maintaining room temperature in the bath. The measurement of hydrodynamic size and zeta potential were performed afterwards. The determination of the binding capacity of siRNA to nanodiamonds was performed by mixing to a fixed concentration of siRNA (0.3 μg/mL) an increasing concentration of H-DND from 0 to 600 μg/mL in 100 mM NaCl, 10 mM HEPES pH 7.2 buffer, in a fixed 60 μL final volume. After centrifugation (16,000× *g* at 10 ◦C for 15 min), non-complexed free siRNA concentrations were measured on 30 μL of the supernatants to which an equal volume of 1 μg/mL ethidium bromide (EtBr, Sigma-Aldrich S.a.r.l, Saint-Quentin Fallavier, France) was added. The mixtures were placed into a 96-wells plate, then analyzed with a fluorescence plate reader (Glomax Multi+, Promega, Charbonnières-les-Bains, France) set at 525 nm excitation and 580–640 nm bandpass detection wavelengths, to infer the free siRNA amount. The results are presented as the fraction of the fluorescence intensity relative to the one of the total amounts of siRNA before adding DND. Experiments were realized in triplicate.

#### *2.4. Measurement of EWS-FLI1 Inhibition in Cultured Cells by RT-qPCR*

One day before treatment, 3 <sup>×</sup> 105 human Ewing sarcoma cells A673 were seeded in 12 wells-plate in DMEM medium (Gibco, Life Technologies S.A.S., Courtaboeuf, France) containing 10% fetal calf serum (Gibco, Life Technologies S.A.S., Courtaboeuf, France) and 1% Penicillin, streptomycin solution (Gibco, Life Technologies S.A.S., Courtaboeuf, France) and then incubated at 37 ◦C, with 5% CO2 and 95% hygrometry. Then the medium was discarded and replaced by 500 μL of 75 nM siRNA bound to H-DND at a mass ratio of 5:1, 25:1 or 50:1 (H-DND:siRNA) in DMEM medium containing 10% fetal calf serum and 1% penicillin and streptomycin solution, for 24 h at 37 ◦C, 5% CO2 and 95% hygrometry. Comparatively, same conditions are used with 75 nM of siAS bounded to Lipofectamine 2000 (Life Technologies S.A.S., Courtaboeuf, France) added to cells in serum containing medium. Then, the medium was discarded, and the cells were lysed by 400 μL of Trizol solution (Invitrogen) and collected in Eppendorf tubes. Total RNA were extracted by adding 60 μL of chloroform:isoamyl alcohol (49:1). After centrifugation at 13,000 rpm for 15 min at 10 ◦C, 150 μL of the aqueous phase was precipitated by adding 150 μL isopropanol for 15 min at room temperature followed by centrifugation for 15 min at 13,000 rpm. Pellets were washed twice with 70% ethanol solution and dried. Plellets were then solubilized with water (10 μL) containing 0.5 U/μL of RNasin (Promega, Charbonnières-les-Bains, France) and RNA was quantified with a spectrophotometer (NanoDrop™, Life Technologies S.A.S., Courtaboeuf, France). The cDNA was prepared by heating 1.5 μg of RNA in 12.5 μL of water at 75 ◦C for 5 min and with 2 μL of 50 μg/mL random primer (Promega). Then, 4 μL of 5X M-MLV buffer (Promega), 0.5 μL dNTP 20 mM (Promega), 0.5 μL RNasin 40 U/μL (Promega) and 0.5 μL M-MLV Reverse Transcriptase 200 U/μL (Promega) were added and incubation was performed for 15 min at 25 ◦C followed by 1 h at 42 ◦C. *EWS-FLI1* mRNA expression is performed by qPCR on 5 μL of 1/20 diluted cDNA, 0.4 μL of each primer at 10 mM concentration, 4.2 μL deionized water and 10 μL of 2X KAPA SYBR® FAST Master Mix (Sigma-Aldrich S.a.r.l, Saint-Quentin Fallavier, France). PCR was performed for 40 cycles in fast mode on a StepOnePlus™ system (Applied Biosystems, Life Technologies S.A.S., Courtaboeuf, France). *EWS-FLI1* gene was amplified with *EWS*-forward primer: 5- -AGC AGT TAC TCT CAG CAG AAC ACC-3 and *FLI1*-reverse: 5- -CCA GGA TCT GAT ACG GAT CTG GCC-3- . As a control, we used GAPDH gene with forward primer: 5- -CAA GGT CAT CCATGA CAA CTT TG-3- , and reverse primer: 5- -GTC CAC CAC CCT GTT GCT GTA G-3- . We normalized the number PCR cycle threshold Ct for the target sequence to the one for GAPDH control gene. Experiments are performed in triplicate.

#### *2.5. In vivo Experiments*

#### 2.5.1. Animal Experimentation

Animal experiments were performed in accordance with the ethical project submitted and approved by the ethical committee N◦26 regulating the animal facility at Gustave Roussy Institute (Villejuif, France) and under national agreement N◦2013-062-01223.

#### 2.5.2. Biodistribution of Nanodiamonds in Mice

We injected 100 <sup>μ</sup>L of PBS buffer containing 3 <sup>×</sup> 106 A673 cells in the right flank of male nude mice. When the tumors appeared, we injected in the tail veins of the mice 100 μL of water for injectable preparation containing either siRNA (5 mg/kg) alone, T-DND (25 mg/kg) alone or siRNA complexed to T-DND (5 mg/kg siRNA mixed to 25 mg/kg T-DND, i.e., mass ratio: 5:1). We then placed the mice in metabolic cages for 4 h or 24 h. At these times, 3 mice of each condition were humanely sacrificed, the tissues were withdrawn and their weights are determined. About 0.1 g of each tissue is sampled and solubilized with 1 mL of Solvable™ (PerkinElmer, Courtaboeuf, France) heated during 3 h at 55 ◦C. After the transfer of the solubilized tissues in scintillation tube, solutions were clarified by adding 100 μL of 30% hydrogen peroxide (Sigma-Aldrich S.a.r.l, Saint-Quentin Fallavier, France). Then 10 mL of Ultima Gold (PerkinElmer, Courtaboeuf, France) were added before the radioactivity measurements were performed during 1 min with 1409 DSA scintillation counter (Wallac/PerkinElmer, Courtaboeuf, France). The results are expressed as deionized counts per minutes per unit mass of tissue (cpm/g) or as the percentage of the injected dose to the whole organs. All urines from cabinet containing animal group receiving a same treatment (in these conditions, measurement is a global statistical value) were collected. Experiments were carried out in triplicates for each condition. For a healthy mouse, the glomerular filtration rate of a compound intravenously injected is about 7 μL/min per g, leading to an almost full elimination after 3 h [23]. Therefore, at our observation times of 4 h and 24 h, we should be able to detect T-DND in urines, if they could be eliminated through this pathway.

#### 2.5.3. Measurement of *EWS-FLI1* Inhibition in Nude Mice

*EWS-FLI1* expression in tumor was measured after intravenous injection in the tail vein of nude mice of 100 μL of antisense or control siRNA 5 mg/kg bound to H-DND at mass ratio of 5:1 (H-DND:siRNA) or free H-DND for 24 h. Then a fragment of the tumor was sampled and placed in a tube containing 400 μL of Trizol solution (Invitrogen, Life Technologies S.A.S., Courtaboeuf, France). A ceramic ball was added and tissue homogenization was performed for 30 s in TissuLyser II (Quiagen Paris, Courtaboeuf, France). Finally, *EWS-FLI1* expression was quantified by RT-qPCR as previously described.

#### *2.6. Statistical Analysis*

Data are presented as means and either standard deviation or standard error on the mean. Statistical significance was evaluated using GraphPad Instat 3.10 software (GraphPad Sofware, San Diego, CA, USA). For all the comparisons between two conditions, we used the non-parametric Wilcoxon–Mann–Whitney two sample rank test (two-tailed *p* value provided), while for the comparison

between more than three conditions we applied a Kruskal–Wallis test, followed by Dunn's Multiple Comparisons Test. One star (\*) corresponds to *p* < 0.05, \*\* to *p* < 0.01 and \*\*\* to *p* < 10<sup>−</sup>3.

#### **3. Results**

#### *3.1. Characterization of the Hydrogenated and Tritiated Nanodiamond Suspensions*

After hydrogenation or tritiation by annealing using molecular gases (see methods and [21]), we suspended the treated DND powder in distilled water by sonication. We measured hydrodynamical diameters of DND clusters ranging from about 60 nm to about 90 nm for H-DND and T-DND colloidal suspensions, respectively (Supplementary Table S2). These results are in agreement with previous colloidal studies on hydrogenated DND [21,24] and also with the transmission electron microscopy observations we reported in Ref. [16] (see Figure 4a in this reference), showing H-DND aggregates of ≈50 nm in size at the cell membrane.

#### *3.2. Loading Capacity of Nanodiamonds for siRNA Binding*

We quantified the loading capacity of H-DND for siRNA binding. To this aim, we measured by fluorescence (see Materials and Methods), for a fixed amount of siRNA, the remaining free unbound siRNA for increasing amounts of hydrogenated nanodiamonds. The results are presented in Figure 1 for two conditions: one in which H-DND and siRNA were simply mixed, the other in which the mixing was done under sonication, both for the same duration. For these two conditions, the complete fixation of siRNA occurred at the same mass ratio *m*H-DND/*m*siRNA of 10:1. These data suggest the sonication assistance leads to a faster adsorption of siRNA onto H-DND, resulting from the larger thermal agitation brought by ultrasound waves.

**Figure 1.** Binding of siRNA to hydrogenated detonation nanodiamonds (H-DND). Free siRNA as function of increasing quantity of H-DND (H-DND-22 samples) added to a fixed initial quantity of siRNA. The results are presented as the percentage of initial siRNA. Orange/blue with or without (respectively) sonication during the siRNA to H-DND complexation. Experiments were performed in triplicates. Number over each point represent the mass ratio *m*H-DND/*m*siRNA. The statistical comparison tests between sonicated and non-sonicated conditions yield *p* = 0.021 (\*) for mass ratio 1:1 to 5:1 and non-significant differences starting from mass ratio 10:1 to higher ones.

#### *3.3. E*ff*ect of siRNA Binding to Nanodiamonds on the Hydrodynamic Size and Zeta Potential of Complexes*

Figure 2 displays the variation of the hydrodynamic size and zeta potential of H-DND:siRNA complexes upon increase of the DND:siRNA mass ratio. The experiment was carried out under two conditions: with or without sonication during complexation. We observed that for mass ratios lower than 10 the particles hydrodynamic size is around 80 nm (Figure 2a). Then, the size increased strongly for mass ratio between 10 and 40 to return to the initial size for mass ratio higher than 40. We conclude that strong aggregation occurred only for mass ratios between 10 and 40.

**Figure 2.** Variation in hydrodynamic size (in intensity) and zeta potential of siRNA bound to increasing mass ratio of H-DND (H-DND-22 sample). Complexes were prepared with or without sonication. (**a**) Size measurement by dynamic light scattering (DLS). (**b**) Zeta potential determination.

Looking at the zeta potentials of each solution (Figure 2b), we observed that for mass ratios lower than 10, the surface charge is negative (around −30 mV) corresponding to an excess of siRNA covering nanodiamonds. For mass ratios between 10 and 40, a surface charge inversion occurs with a zeta potential close to +40 mV for a mass ratio of 40. In between, the low zeta potential of the complexes promotes their aggregation, as observed on Figure 2a. For mass ratios higher than 40, the positive zeta potential reflects an excess of hydrogenated nanodiamonds. Sonication have no effect on surface charge but promote smaller aggregates, as observed on the DLS profiles (Figure 2a).

Moreover, considering that we inject the H-DND:siRNA conjugate in the blood, we could have some concern regarding its aggregation in such a complex environment. However, it is well established (see for example S. Hamelaar et al. [25]) that the serum favors the dispersion of electrostatically-charged nanoparticles when they are dispersed in a culture medium. In order to confirm these data with our specific conjugate, we formed a solution of H-DND:siRNA at the same mass ratio of 5:1 used in the in vivo experiment, and measured its hydrodynamic diameter in DMEM culture medium supplemented with 10% Fetal Bovine Serum. We found a diameter of 35 nm, a value even smaller than the diameter of 90 nm of the same conjugate in pure water (see Table S2), which is a strong indication that injecting intravenously an aqueous suspension of H-DND:siRNA (mass ratio 5:1), will rather induce a small deagglomeration than an aggregation in the blood circulation.

#### *3.4. Inhibition of EWS-FLI1 on Ewing Sarcoma Cultured Cells*

We first studied the efficacy of H-DND prepared by annealing under hydrogen gas, as vector for siRNA delivery to human Ewing sarcoma A673 cells. We used different mass ratios between H-DND and siRNA, from 5:1 to 50:1. We observed on Figure 3 that 35% inhibition is obtained with these cationic H-DND for mass ratios higher than 25. In comparison, commercially available cationic liposomes Lipofectamine 2000 used in similar conditions in the presence of serum during transfection show a lower efficacy with only 18% inhibition. It is to note that in recommended conditions (transfection in serum free medium) Lipofectamine 2000 yields 70% inhibition on this cell model [9]. The efficacy of siRNA transfection by H-DND in serum containing medium is crucial for further applications in animals.

**Figure 3.** Inhibition of *EWS-FLI1* expression in A673 Ewing sarcoma cultured cells by siRNA vectorized by H-DND (sample H-DND-24) or Lipofactamine 2000. Cells are treated during 24 h in DMEM medium containing 10% of fetal calf serum by 75 nM siRNA. The H-DND:siRNA mass ratio was 5:1, 25:1 or 50:1. siAS: antisense siRNA directed against *EWS-FLI1* oncogene; siCt: control irrelevant siRNA sequence (see methods). Experiments were performed in triplicate. Comparisons were all done relative to the "Ct" condition using the Wilcoxon–Mann–Whitney two sample rank test, that yielded *p*5:1 = 0.018, *p*25:1 = 0.009; *p*50:1 = 0.036, *p*lipo = 0.019; *ns*: non-significant.

#### *3.5. Inhibition of EWS-FLI1 Expression in Mice*

We then evaluated the efficacy of siRNA vectorized by T-DND to inhibit *EWS-FLI1* in tumor xenografted on mice. We produced *EWS-FLI1* tumor model by grafting A673 cells in the flank of nude mice. We used tritiated DND to conduct both the inhibition and biodistribution studies using the same mice. We selected the mass ratio of 5:1, first of all to ensure that all T-DND are covered by siRNA. Our hypothesis was that such a configuration would favor the detachment of siRNA molecules which interact only by a portion of their total length with the T-DND surface. Moreover, the mass ratio of 5:1 also ensures that the injected suspension is not too viscous, since at the given necessary dose of siRNA and volume of solution to be injected, the concentration of T-DND remains low enough to have a small impact on the viscosity of the aqueous suspension. Finally, high dose of cationic vector may be toxic as observed in cultured cells. For all these reasons, we favored the smallest H-DND concentration, corresponding to the 5:1 mass ratio.

We injected the T-DND radiolabels in the tail vein of mice, using either uncomplexed T-DND or with siRNA complexed to them. We did not include a free siRNA group because we already established that free siRNA is not able to inhibit *EWS-FLI1* oncogene [26]. We considered 6 to 8 nude mice per condition, which were placed in metabolic cage. After 4 h or 24 h, mice were humanly sacrificed. For the efficacy study, we only considered the 24 h group. Using RNA extracted from tumors, we then quantified *EWS-FLI1* expression by RT-qPCR. The normal level of *EWS-FLI1* expression is provided by mice treated by non-complexed T-DND. We also used an irrelevant siRNA complexed with T-DND as a specificity control. We observed on Figure 4 that the irrelevant T-DND:siRNA treatment has no effect. The *EWS-FLI1* antisense siRNA complexed to T-DND inhibits the *EWS-FLI1* expression by about 50%. These results confirm that H/T-DND:siRNA is able to inhibit *EWS-FLI1* in this Ewing sarcoma tumor model grafted on nude mice.

**Figure 4.** Inhibition of *EWS-FLI1* expression in tumor xenografted on mice treated for 24 h by siRNA vectorized by tritiated DND (T-DND) at a mass ratio of 5:1 (T-DND:siRNA). The tumor was formed from A673 Ewing sarcoma cells grafted in the right flank of the mice (*n* = 6 to 8 animals per condition). T-DND:siRNA was intravenously injected in the mouse tail vein. RNA was extracted from the tumor of mice sacrificed 24 h after treatment, and RT-qPCR was performed using GAPDH as housekeeping gene. Standard error on the mean are indicated. siAS:siRNA against *EWS-FLI1* oncogene; siCt: irrelevant. A significant difference between T-DND:siAS and T-DND-siCt is observed (\*) according to the *p*Dunn value <0.05, inferred from a Kruskal–Wallis test (*p* = 0.0094) followed by a Dunn multiple comparisons test.

#### *3.6. Biodistribution of Nanodiamonds in Nude Mice*

For an efficient use of H-DND produced by annealing for siRNA delivery to Ewing sarcoma tumor xenografted on mice, it is important to determine their tissue distribution and elimination, which is made possible by radioactivity measurement using their radioactive analogs T-DND. The different tissues of the mice of the efficacy study (Section 3.5) were collected, homogenized and then the radioactivity was determined as presented in Figure 5. We also collected urines and feces at the same time points. We observed that T-DND accumulated mainly in the liver, lung, spleen, kidney and also the heart for T-DND:siRNA. There are no significant changes between 4 h and 24 h in the quantity of T-DND found in the different tissues. For kidney, spleen, lung and heart, T-DND complexed with siRNA accumulated more than uncomplexed T-DND. The radioactivity is recovered in all tested tissues at a lower level. We did not observe high accumulation in the tumor. Nevertheless, the dose measured in the tumor is four time larger than the one in the blood.

One of the major questions for the use of mineral origin nanoparticles for biomedical applications is how they may be eliminated after injection in animals. The radioactive fraction recovered in urines was 0.15% of the injected dose per mouse. Unfortunately, after centrifugation of urines, we found that all radioactivity was recovered in the supernatant. This indicates that no T-DND was eliminated through the kidney pathway. The very small detected radioactivity is probably the one of water consecutive to exchange of H with labile T. Note that 4 h after injection in the kidney we detected about 0.25% of the radioactive T-DND:siRNA injected dose. This can be due either to the smallest T-DND being filtrated by the glomeruli but then reabsorbed by the tubules and/or to T-DND aggregates not being able to cross the glomeruli. This last hypothesis is in agreement with the fact that uncomplexed T-DND accumulated about twice less in kidney than the one associated to siRNA, which present a higher aggregation state. Another possible elimination pathway is by the feces. Thanks to the metabolism cage, we estimated that the total collected amount of feces after 24 h had a radioactivity representing about 0.19% for H-DND:siRNA and 0.13% for H-DND of the injected doses per mouse (Figure 5b), which is about 6 times larger than the one after 4 h (0.03% and 0.02% for H-DND:siRNA and H-DND, respectively), indicating a linear elimination with time, in perfect agreement with the constant value of radioactivity in feces sampled in the rectum at 4 h and 24 h (Figure 5a). Indeed, since

the majority of DND are captured by liver cells, their transfer to the intestine lumen by the bile may be the main way for T-DND elimination.

**Figure 5.** Mouse organ distribution of uncomplexed T-DND or complexed with siRNA at mass ratio 5:1 of T-DND:siRNA. The total radioactivity of each tissue is measured after 4 h and 24 h for 3 animals per series. Error bars: standard deviations. (**a**) Radioactivity presented as count per minutes (cpm) per gram of tissue or feces. The latter corresponds only to the fraction collected in the rectum of each animal. (**b**) Radioactivity from each full organ, presented as the percentage of the total injected dose. Feces were collected for all the animals simultaneously in the same metabolism cage, but the total amount, including non-expelled feces is not known which is why we could not include feces data in B. ND + siAS: DND complexed with siRNA sequence antisense for *EWS-FLI1* oncogene. The differences between error bars for the different conditions are due to (i) the variability between animal groups, and (ii) to the fact that the T-DND quantifications were done only on portions of ≈100 mg of each organ, and that we cannot exclude an inhomogeneous distribution of T-DND within the organs.

#### **4. Discussion**

In a previous work, we had demonstrated that cationic hydrogenated detonation nanodiamonds can deliver active siRNA to Ewing Sarcoma cell in culture [16]. Here, we wanted to establish if small hydrogenated DND are able to deliver active siRNA in vivo to mice model of Ewing sarcoma. In order to investigate simultaneously their biodistribution, we produced hydrogenated and tritiated detonation nanodiamonds via their exposure to a mixture of H2 and T2 molecular gases at 550 ◦C under controlled pressure. The characterization in size and surface charge was performed by DLS measurement and zeta potential determination (Table S2). We observed that H-DND are cationic with a zeta potential of ≈+45 mV and have a hydrodynamic size of ≈60 nm. H-DND positive charge favored electrostatic attachment of siRNA onto the nanoparticles and the measured DLS size larger than the 7-nm primary size indicated some aggregation. This size difference is partly due to the measurement methods, as unitary diamond size is the one of its hard core as observed by high-resolution transmission electron microscopy, whereas DLS yields the hydrodynamic size, that is larger because it takes into account the water dipole molecule attached to the core and moving with it under Brownian motion. Our data indicate that H-DND are present in the aqueous suspension in the form of aggregates of an average number of eight primary 7 nm sized particles.

The binding capacity of H-DND was studied by mixing to a constant quantity of siRNA an increasing quantity of H-DND and the full binding of siRNA was obtained for 10 times excess of DND (Figure 1). Hydrodynamic size and zeta potential variations during the titration process of siRNA with H-DND (Figure 2a) indicated that for mass ratio around this point of saturation (in the 10 to ≈40 mass ratio range) the size of H-DND:siRNA complexes increased to form large aggregates. The sonication during siRNA and H-DND association process decreased the size of the aggregates but did not prevent it. For mass ratio smaller than 10 or larger than ≈40 the nanoparticle hydrodynamic sizes were close to the ones of uncomplexed H-DND (<100 nm, Table S2). Furthermore, the slightly smaller

size of the H-DND:siRNA complex under sonication indicates its high colloidal stability. This strong binding capacity of H-DND presents advantages for animal applications to prevent desorption of the cargo before complexes reach their target cells. Finally, during the titration process, the zeta potential decreased from +50 mV for an excess of H-DND (mass ratios larger than 20) to a negatively charged complexes (zeta potential ≈-30 mV) for mass ratios smaller than 10 (Figure 2b). The sonication did not change the mass ratio transition threshold. This observation is consistent with the aggregation happening once all nanodiamonds have lost their charges due to the complexation with siRNA. It is worth to note that the global charge of the complexes could be tuned to negative (mass ratios smaller than 10) or to positive (mass ratios larger than 20), resulting in the formation of complexes that may interact differently with cell membrane and lead to different siRNA delivery efficacy [27].

In the prospect of using H-DND for anticancer gene delivery in mice, we first validated the delivery of these cationic DND by their capacity to transport a siRNA cargo directed against *EWS-FLI1* oncogene in cultured A673 human Ewing sarcoma cells and by detecting a gene inhibition efficacy. We observed that H-DND vectorized siRNA inhibited *EWS-FLI1* expression by 30–40% (Figure 3). In comparison with previous results [16], the inhibition of *EWS-FLI1* expression by siRNA transported by DND hydrogenated via annealing is lower than the one obtained with plasma hydrogenated DND vectors. This may be due to the slightly larger aggregation of DND hydrogenated via annealing, leading to a lower internalization efficiency [28].

One key question regarding the use of non-biodegradable nanoparticles for drug delivery in animal is to determine whether they are eliminated, through which pathways, and where in the body the non-eliminated fraction accumulates. It was shown that "large" ND of 50 nm primary size reside in animal tissues a few weeks after injection [17]. However, "small" ND of size compatible with kidney elimination may overcome this limitation. Indeed, nanoparticles smaller than 6–8 nm have been already reported to be filtered at the glomeruli level [18,19]. In this respect, detonation ND with primary size included between 3 and 8 nm and even smaller [29] are good candidates for such elimination. For this study, we prepared radioactive tritiated DND to be able to trace these ultra-small particles after injection to mice. We observed that the liver, spleen and lung contained much more radioactive amount than the other tissues (Figure 5). Moreover, T-DND carrying siRNA accumulated more in kidney, spleen, heart and lung, with a slow increase from 4 to 24 h. Unfortunately, we observed a limited accumulation of T-DND in the tumor compared to other tissues such as muscles, heart or kidney. This indicated that Enhanced Permeability and Retention (EPR) [30] effect did not occur for these complexes or/and that the tumor had no aberrant vascularization responsible for EPR effect. Nevertheless, T-DND:siRNA quantity was stable in the tumor between 4 h to 24 h which is the time range needed to induce the silencing effect. During the first 24 h, the blood radioactivity did almost not decrease, which indicated that DND were only very slowly eliminated from the circulation.

The main possible elimination pathways are urines and feces. Radioactivity from urines could come from both T-DND or tritiated water. The later could originate from a fraction of tritium desorbing from T-DND and exchanging with hydrogen in water. Using ultracentrifugation, we measured this labile tritium fraction to be 5% (see Table S1). In order to reduce the amount of free tritium in the injected solution, all T-DND:siRNA complexes were prepared from T-DND purified by this ultracentrifugation process. However, even after such purification, there is a remaining fraction of 2% free tritium. This value provides a lower bound of T-DND detection sensitivity in urines. Urines were collected globally for each group of mice and their radioactivity was measured in 1 mL of solution and in 1 mL of ultra-centrifugated supernatant. We observed the same values of radioactivity before and after centrifugation (Table S1) within a margin of error larger than the detection sensitivity lower bound. Hence, the radioactivity found in urines most likely comes from the residual free tritium present in the injected solution and not from T-DND. We therefore consider that T-DND were not eliminated in urines, or only very slowly. The possible reasons for the absence of T-DND in urine are (i) the formation of aggregates too large to be filtrated or/and (ii) ultrasmall DND reabsorbed by kidney tubules, the first one (i) being the most likely, considering previous observations of Rojas et al. [20]

who showed that membrane-filtrated DND are eliminated in urines, much more efficiently than the aggregates dominant in the original suspension.

We then explored the other possible elimination pathway, by collecting the feces from the rectum and we measured their radioactivity. We observed a high radioactive signal in the feces. The amount of radioactivity in the feces fraction is larger than the one of free tritium in the injected volume, according to the labile tritium release measurements (Table S1), therefore we shall conclude that DND are present in the feces which therefore constitute one of the elimination pathways. It is likely that DND are eliminated though the bile by liver filtration. Since the bile goes into the intestinal lumens, DND are then incorporated to the feces. This result is of high importance because it states that ultrasmall DND can be partly eliminated from mouse body after intravenous administration.

Furthermore, if we consider the radioactivity balance between injected doses per mouse and the total amount collected, only about 25% of the dose is recovered in the sampled fractions. The 75% of not-measured radioactive T-DND are localized in the carcass containing skin, not removed muscles, bones and so on, and probably a part has also been eliminated through other routes. Indeed, the elimination pathways are multiple in animals. Some authors propose expectoration as an alternative way of elimination. Considering that DND also accumulated into the lung (Figure 5), they may be engulfed by macrophages which further direct them into the alveoli where they may be finally expulsed through the pharynx by the mucociliary system [31]. Further investigations would be necessary to evaluate the expectorated fraction.

Finally, regarding the silencing efficacy of *EWS-FLI1* in Ewing sarcoma cells xenografted tumor by siRNA delivered by T-DND, we observed a 50% inhibition by the antisense siRNA compared to irrelevant siRNA. The latter yielded a similar effect than the T-DND alone, i.e., no inhibition. This result indicates that despite a low accumulation of T-DND in the tumor, the siRNA is delivered to the cancer cells where they silence the targeted oncogene. Note that although the in vitro inhibition efficacy is similar to the in vivo one, one cannot generally extrapolate in vitro results to in vivo ones which justifies our study. Finally, in order to treat the mouse and reduce the tumor size, we could combine a conventional anticancer treatment like vincristine to the H-DND-siRNA one, as we showed in our previous in vitro study [16].

Overall, our study represents a significant step towards the use of ultra-small solid nanoparticles which are able to deliver efficiently active siRNA to tumor cell in animals and are also eliminated from their body subsequently.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2079-4991/10/3/553/s1, Table S1: Determination of free tritium in nanodiamond suspension and in mice urine after injection; Table S2: Size measurement of hydrogenated and tritiated DND conjugated or not to siRNA, in water and culture medium supplemented with serum.

**Author Contributions:** J.-R.B. and F.T. conceived and designed the experiments; S.C. and J.-R.B.; performed the experiments; S.C., F.T. and J.-R.B. analyzed the data; É.N., S.G.-A., H.A.G., J.-C.A., S.F. and G.P. prepared the hydrogenated and tritiated nanodiamond solutions; J.-R.B. and F.T. wrote the article and did the figures. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by French National Research Agency ANR through the ERANET Euronanomed 2 program, grant number ANR-14-ENM2-0002.

**Acknowledgments:** We thank the animal experimentation platform from Gustave Roussy Institute for technical assistance and in particular Mélanie Polrot for her expertise in animal manipulation.

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

#### **References**


© 2020 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*

## **Carbon Dot Nanoparticles Exert Inhibitory E**ff**ects on Human Platelets and Reduce Mortality in Mice with Acute Pulmonary Thromboembolism**

**Tzu-Yin Lee 1, Thanasekaran Jayakumar 1, Pounraj Thanasekaran 2, King-Chuen Lin 3,4, Hui-Min Chen 5, Pitchaimani Veerakumar 3,4,6,\* and Joen-Rong Sheu 1,\***


Received: 26 May 2020; Accepted: 23 June 2020; Published: 28 June 2020

**Abstract:** The inhibition of platelet activation is considered a potential therapeutic strategy for the treatment of arterial thrombotic diseases; therefore, maintaining platelets in their inactive state has garnered much attention. In recent years, nanoparticles have emerged as important players in modern medicine, but potential interactions between them and platelets remain to be extensively investigated. Herein, we synthesized a new type of carbon dot (CDOT) nanoparticle and investigated its potential as a new antiplatelet agent. This nanoparticle exerted a potent inhibitory effect in collagen-stimulated human platelet aggregation. Further, it did not induce cytotoxic effects, as evidenced in a lactate dehydrogenase assay, and inhibited collagen-activated protein kinase C (PKC) activation and Akt (protein kinase B), c-Jun N-terminal kinase (JNK), and p38 mitogen-activated protein kinase (MAPK) phosphorylation. The bleeding time, a major side-effect of using antiplatelet agents, was unaffected in CDOT-treated mice. Moreover, our CDOT could reduce mortality in mice with ADP-induced acute pulmonary thromboembolism. Overall, CDOT is effective against platelet activation in vitro via reduction of the phospholipase C/PKC cascade, consequently suppressing the activation of MAPK. Accordingly, this study affords the validation that CDOT has the potential to serve as a therapeutic agent for the treatment of arterial thromboembolic disorders

**Keywords:** nanoparticles; carbon dots; platelet aggregation; arterial thrombosis; signaling molecules; bleeding disorder

#### **1. Introduction**

Platelet activation has been associated with several thrombotic diseases. While it plays a vital role in regulating hemorrhagic events, hyperactivity can lead to a range of complications. In general, patients with cardio- and cerebrovascular ailments are found to have more reactive platelets than normal, healthy individuals. Thrombotic diseases pose a severe threat to humans as they may elicit significant injury and even lead to death. Several studies have recommended that intravenous heparin and tissue plasminogen activators are effective for treatment [1,2]; nevertheless, these are unsafe and may lead to severe bleeding and problems associated with reocclusion and reinfarction [3]. The inhibitors of antiplatelet drugs, such as the P2Y12 receptor, integrin αIIbβ3, cyclooxygenase, and phosphodiesterase, are also widely used; however, they have serious limitations. Further, phosphatidylinositol 3-kinase inhibitors have been proposed as potential antithrombotic agents [4], but they also are associated with some major restrictions for use as drugs.

Nanoparticles can be defined as any particulate materials that range from 1 to 100 nm in size in at least one dimension [5], and they are ubiquitously distributed in the environment. In fact, humans are often exposed to airborne nanoparticles [6]. Their size can be manipulated to facilitate their passage across biological membranes and affect cell physiology [7]. Berry et al. [8] reported the substantial accumulation of nanoparticles in platelets in pulmonary capillaries and anticipated that there might be a predisposing factor for platelet aggregation and microthrombi formation. Nanoparticles are not proposed for systemic use as they can interfere with platelet function and increase the risk of cardiovascular diseases and vascular thrombosis [9]. However, some types of nanoparticles have been developed for therapeutic purposes that can target the injured vascular site to mimic platelet function [10] or enhance blood clotting [11]. Nevertheless, their undesirable, antiaggregating properties are of a significant concern in nanomedicine, impeding their widespread application in the clinical setting.

Carbon dots (CDOTs) have become the most important type of nanoparticles, considering their favorable biological properties. They are obtained from natural carbon sources and their average diameter is < 10 nm [12]. These nanoparticles have even been acquired from organic substances and are constant in water media, which is tremendously noteworthy from a biological point of view [13], especially in drug delivery, bioimaging, optical imaging, and biosensing due to their high biocompatibility [14]. A study reported that, in comparison to CDOTs, the application of noncarbon quantum dots has not received adequate attention, as they are associated with severe health and environmental concerns [15]. Yan et al. [16] reported the antihemorrhagic effects of novel water-soluble carbon quantum dots, and their results indicated the explicit hemostatic effect of these nanoparticles. CDOTs, isolated from egg yolk oil, demonstrated a hemostatic effect in mice via the stimulating intrinsic blood coagulation and fibrinogen systems [17]. Another relevant study showed that CDOTs from the Phellodendri Cortex carbonisatus considerably reduced bleeding time and coagulation parameters and significantly increased platelets without inducing toxicity when administered in mice [18]. Mariangela Fedel has recently reviewed the hemocompatibility of carbon nanostructures [19]. However, in general, the antiplatelet aggregating effects of CDOTs have not been extensively explored. Therefore, in this study, we investigated the antiplatelet and antithrombotic effects of a new type of CDOT in human platelets and mice, respectively.

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

#### *2.1. Reagents*

Collagen (type I), 9, 11-dideoxy-11α, 9α-epoxymethanoprostaglandin (U46619), and thrombin were purchased from Chrono-Log Corporation (Havertown, PA, USA). Anti-phospho-p38 mitogen-activated protein kinase (MAPK) (Thr180/Tyr182), anti-phospho-c-Jun N-terminal kinase (JNK) (Thr183/Tyr185), anti-phospho-(Ser) protein kinase C (PKC) substrate, anti-JNK polyclonal antibodies (pAb), and anti-p38 MAPK and anti-Akt monoclonal antibodies (mAb) were purchased from Cell Signaling Technology (Beverly, MA, USA). Anti-phospho-Akt (Ser473) pAb was purchased from Biorbyt (Cambridge, UK), and anti-pleckstrin (p47) pAb was purchased from Gene Tex (Irvine, CA, USA). Hybond-P polyvinylidene difluoride (PVDF) membranes, enhanced chemiluminescence (ECL) Western blotting detection reagent, and the analysis system were purchased from GE Healthcare Life Sciences (Buckinghamshire, UK). Horseradish peroxidase-conjugated goat anti-rabbit and anti-mouse immunoglobulin G antibodies were obtained from Jackson ImmunoResearch Laboratories (West Grove, PA, USA).

#### *2.2. Preparation of CDOTs*

Fresh garlic (*Allium sativum*) cloves were purchased from a local market in Taiwan, which were then peeled, crushed, and suspended in ultrapure water. This suspension was vigorously stirred for 1 h at 40 ◦C. The extract was filtered twice to remove insoluble materials and then freeze-dried. The obtained powder was stored at −20 ◦C until required. For CDOT synthesis, 100 mg of the garlic extract powder was dissolved in 3 mL water and poly (diallyldimethylammonium chloride) mixture (1/0.5, *v*/*v*). The clear transparent solution obtained was heated in a domestic microwave oven for 5 min at 600 W and then cooled to ambient temperature (25 ◦C). The obtained yellow-brown solution was diluted with ultrapure water and dialyzed against water for 2–3 h through a dialysis membrane.

#### *2.3. Characterization of the Synthesized CDOTs*

Crystallographic information pertaining to the CDOTs was collected using an analytical X-ray diffractometer (X'Pert PRO, Malvern, Worcestershire, UK) using Cu Kα radiation (λ = 0.1541 nm). A Fourier transform infrared (FT-IR) spectrometer (Bruker IFS28, Billerica, MA, USA) in the range of 4000–400 cm−<sup>1</sup> was used for the characterization of functional groups on the surface of the CDOTs, with an average of 21 scans. The sample was prepared as pellets using KBr (IR grade). Ultraviolet–visible (UV–vis) spectra were documented using a Thermo Scientific Evolution 220 spectrophotometer (Waltham, MA, USA), whereas fluorescence spectral measurements were taken using a PerkinElmer LS-45 spectrometer (Waltham, MA, USA). The morphological information on the prepared CDOTs was obtained through field-emission transmission electron microscopy (FE-TEM, JEOL JEM-2100F, Akishima, Tokyo, Japan).

#### *2.4. Preparation of Washed Human Platelets and Lactate Dehydrogenase (LDH) Release Assay*

This study was performed in accordance with the Declaration of Helsinki, and the Institutional Review Board of Taipei Medical University approved all protocols (IRB: N201612050). All volunteers provided informed consent before they participated in this study. Anticoagulated human blood with acid–citrate–dextrose (1:9) was collected from healthy human volunteers who had not eaten any drugs within a time of two weeks prior to the analysis. The method described by Sheu et al. [20] was used for preparing human platelet suspensions. The platelets were suspended in Tyrode's solution, and calcium chloride was then added, with the final concentration of Ca2<sup>+</sup> being 1 mM.

The cytotoxicity of the CDOTs was evaluated using an LDH release assay. Washed platelets (3.6 <sup>×</sup> 108 cells/mL) were pretreated with 50-500 <sup>μ</sup>M CDOTs or a solvent control (PBS; phosphate-buffered saline) for 20 min at 37 ◦C and then centrifuged at 5000 g for 5 min. The supernatant obtained was used for the assay. Briefly, 10 μL of the supernatant was placed on a Fuji Dri-Chem slide (LDH-PIII) (Tokyo, Japan), and the absorbance was measured at 540 nm using a UV–vis spectrophotometer (UV-160; Shimadzu, Japan). The LDH activity of 1% Triton X-100-treated washed platelets indicated 100% LDH release.

#### *2.5. Platelet Aggregation*

Platelet aggregation was monitored using a lumi-aggregometer (Helena Laboratories, Beaumont, TX, USA), as previously described [20]. The platelet suspension (3.6 <sup>×</sup> 108 cells/mL) was preincubated with various CDOT concentrations (25–120 μM) or an isovolumetric solvent control (PBS) for 3 min before the agonists were added. The reaction was permitted to continue for at least 10 min and the level of aggregation was calculated in light transmission units. The amplitude and slope of platelet aggregation were automatically calculated using the aggregometer.

#### *2.6. Western Blotting*

Washed platelets (1.2 <sup>×</sup> 109 cells/mL) were preincubated with the CDOTs (65 <sup>μ</sup>M and 90 <sup>μ</sup>M) for 3 min before treating them with collagen to induce platelet activation. A lysis buffer (200 μL) was used for platelet resuspension after the reaction was complete. Proteins (80 μg) from the supernatants were separated using 12% SDS-PAGE and electrophoretically transferred to PVDF membranes (Bio-Rad, Hercules, CA, USA). The membranes were blocked with 5% BSA in Tris-buffered saline (10 mM Tris-base, 100 mM NaCl, and 0.01% Tween 20) for 1 h and probed with various primary antibodies, followed by incubation with horseradish peroxidase-labeled anti-rabbit or anti-mouse immunoglobulin G antibodies for 1 h. Antibody-bound proteins on the membranes were detected using an ECL system and quantified using Bio-profil Biolight (version V2000.01; Vilber Lourmat, Marne-la-Vallée, France).

#### *2.7. Tail Bleeding Time in Mice*

ICR mice (20–25 g, 5–6 weeks old, male) were obtained from BioLasco (Taipei, Taiwan). All procedures and protocols were approved by the Affidavit of Approval of Animal Use Protocol, Taipei Medical University (LAC-2018-0360). The bleeding time was measured after 10 min of intravenous administration of 1 mg/kg CDOTs or PBS (control). The tail of anesthetized mice was cut 3 mm from the end and then directly immersed in normal saline at 37 ◦C. The bleeding time was recorded until no sign of bleeding was observed for at least 10 s.

#### *2.8. ADP-Induced Acute Pulmonary Thromboembolism in Mice*

According to our previous method, we used ADP to induce acute pulmonary thromboembolism in mice [21]. A fixed dose of the CDOTs (1 mg/kg) or PBS was intravenously injected, and after 10 min, ADP (0.7 mg/g) was injected into the tail vein. The lungs were then removed and placed in 4% formalin, and paraffin-embedded sections were stained with hematoxylin–eosin and then photographed using ScanScope CS (Leica Biosystems, Wetzlar, Germany). The mortality rate was recorded in all animal groups within 1 h of the injection.

#### *2.9. Statistical Analysis*

Data are expressed as mean ± standard error of the mean (SEM), and convoyed by the number of observations (*n*). n represents the number of experiments, and each experiment was performed using different blood donors. Statistical significances were analyzed for the in vivo experiments using unpaired Student's *t* test. One-way analysis of variance (ANOVA) was implemented to determine variations between the experimental groups and, if the analysis exhibited a significant difference, they were compared using the Student–Newman–Keuls test. *p* < 0.05 indicated statistical significance.

#### **3. Results**

#### *3.1. Characterization of the CDOTs*

#### 3.1.1. X-ray Diffraction Analysis

The X-ray diffraction (XRD) pattern revealed that one diffraction peak at 2θ of 23.6◦ corresponded to disordered carbon atoms and the (002) graphite lattice, as shown in Figure 1A, and this finding was consistent with that previously reported for CDOTs [22].

#### 3.1.2. FT–IR

The FT–IR spectrum observed for the garlic clove is rather similar to that of the CDOTs, indicating that the functional groups were, indeed, successfully provided the garlic clove, as illustrated in Figure 1B. The broad absorption band centering at 3427 cm−<sup>1</sup> should be associated with the O–H stretching vibration mode of the hydroxyl functional groups in the garlic clove. The weak bands at 2940 and 1413 cm−<sup>1</sup> confirm the presence of CH2 groups, whereas the bands at 921 and 1568 cm−<sup>1</sup> revealed the presence of oxygen-containing functional groups. The peaks at approximately 2944 and 1405 cm−<sup>1</sup> were assigned to the C–H and C–N stretching vibration modes, and the absorption at 680 cm−<sup>1</sup> could be ascribed to the C–S group [23]. Consequently, the as-prepared CDOTs were mainly composed of different functional groups on their surface, which is favorable for sustainable applications in biology.

#### 3.1.3. UV–vis Spectroscopy

The UV–vis absorption spectra of the CDOTs, as shown by the blue line in Figure 1C, showed a comparable absorption band ranging from 200 to 600 nm, concordant with an earlier study on N-doped CDOTs produced by Wu et al. [24]. The CDOTs water solution produced solid blue light under UV irradiation of 365 nm, as shown by the right inset in Figure 1C. The CDOTs exhibited very strong FL in the range of 380–600 nm, with the maximum peak at around 446 nm, as shown by the red line in Figure 1C.

**Figure 1.** Characterization of the synthesized CDOTs. (**A**) X-ray diffraction (XRD), (**B**) Fourier transform infrared (FT–IR) spectra, and (**C**) UV–vis absorption spectra, as described in the Materials and Methods section.

#### *3.2. LDH Assay and FE-TEM*

Herein, we explored the probable toxic effects of the synthesized CDOTs on platelets by observing the release of cytosolic LDH. The CDOTs (50 μM and 100 μM) did not provoke any substantial discharge of LDH from platelet cytosol, even at concentrations of up to 200 μM, as shown in Figure 2A. Thus, they evidently did not disturb platelet membrane integrity or induce cytotoxicity at concentrations as high as 200 μM. A slight increase was observed at a higher concentration of 500 μM. LDH activities measured from the 1% Triton X-100-treated platelets were regarded as 100% release.

The morphological features and average particle sizes of the CDOTs are shown in Figure 2B. The synthesized CDOTs had a crystalline structure and were well distributed in water without aggregation. Furthermore, they were round in shape with a normal diameter of 3 nm [25].

**Figure 2.** Cytotoxicity and morphology of the CDOTs. (**A**) Washed platelets (3.6 <sup>×</sup> <sup>10</sup><sup>8</sup> cells/mL) were preincubated with PBS (control) or the synthesized CDOTs (50, 100, 200 and 500 μM) for 20 min, and a 10 μL suspension of the supernatant was deposited on a Fuji Dri-Chem slide (LDH-PIII). (**Ba–f**) Field-emission transmission electron microscopic images. The arrows indicate sizes and morphologies of CDOTs. Values represent mean ± SEM (*n* = 6).

#### *3.3. Inhibition of Platelet Aggregation Stimulated by Collagen*

The CDOTs led to concentration-dependent (25–120 μM) inhibition of platelet aggregation induced by collagen (1 μg/mL), as shown in Figure 3A,B, but not by U46619 (1 μM), a prostaglandin endoperoxide (thromboxane A2 receptor agonist), or thrombin (0.01 U/mL), even at higher concentrations of 120 μM, as shown in Figure 3C,D. Almost full inhibition was observed at 90 μM in collagen stimulated aggregation, as shown in Figure 3B. As a result, the IC50 (65 μM) and maximal concentration (90 μM) of the CDOTs were chosen to observe the potential inhibitory mechanisms in collagen-activated platelets. The CDOTs suppressed maximal platelet aggregation, stimulated by collagen, but not by U46619 and thrombin, as shown in Figure 4A, whereas the slopes of platelet aggregation revealed that the CDOTs also significantly reduced the lag time induced by these agonists, respectively, as shown in Figure 4B.

**Figure 3.** Inhibitory effects of the CDOTs on human platelet aggregation. Washed platelets (3.6 <sup>×</sup> 108 cells/mL) were preincubated with PBS (control) or the synthesized CDOTs (25-120 <sup>μ</sup>M) and subsequently treated with (**A**,**B**) 1 μg/mL collagen, (**C**) 1 μM U46619, and (**D**) 0.01 U/mL thrombin to induce platelet aggregation. The IC50 and maximal inhibitory concentrations are shown in **B**. The inhibitory profiles (**A**–**D**) are representative examples of four similar experiments. The delayed lag phase of platelet aggregation noticed in CDOT-pretreated platelets stimulated by either U46619 (**C**) or thrombin (**D**).

**Figure 4.** Maximal aggregation and slope of aggregation curves. (**A**) Concentration-response bar diagrams of the synthesized CDOTs, demonstrating their inhibitory activity for maximal aggregation (%). (**B**) Slope of platelet aggregation, as calculated from the linear part of the aggregation trace. Values represent mean ± SEM (*n* = 4). \*\* *p* < 0.01 and \*\*\* *p* < 0.001, compared with the control (ctl; PBS-treated) group.

#### *3.4. Inhibition of PKC Activation (p47; Pleckstrin) and Akt, JNK1*/*2, and p38 MAPK Phosphorylation*

We additionally investigated the mechanisms by which the CDOTs inhibited platelet aggregation. Their effects on PKC activation (p-p47) and Akt (protein kinase B), JNK1/2, and p38 MAPK phosphorylation are shown in Figure 5. The CDOTs (65 and 90 μM) significantly and, in a concentration-dependent manner, suppressed PKC activation in collagen-activated platelets, as shown in Figure 5A. Akt is a serine/threonine-specific protein kinase, which acts a major element in numerous cellular events, such as platelet activation, cell proliferation, apoptosis, and cell migration [26]. The CDOTs markedly inhibited collagen-induced Akt phosphorylation, as shown in Figure 5B. Furthermore, the CDOTs inhibited JNK1/2 and p38 MAPK [27] phosphorylation, which were elevated in collagen-stimulated platelets, as shown in Figure 5C,D, respectively.

**Figure 5.** Effects of the CDOTs on PKC activation, and Akt, JNK1/2, and p38 MAPK phosphorylation in collagen-activated platelets. Washed platelets (1.2 <sup>×</sup> 109 cells/mL) were preincubated with PBS (control) or the synthesized CDOTs (65 and 90 μM), and subsequently, collagen (1 μg/mL) was added to trigger (**A**) PKC activation (p-p47) and (**B**) Akt, (**C**) JNK1/2, and (**D**) p38 MAPK phosphorylation. Values represent mean ± SEM (*n* = 4). \* *p* < 0.05 and \*\*\* *p* < 0.001, compared with the control (PBS-treated) group. # *p* < 0.05, ## *p* < 0.01, and ### *p* < 0.001, compared with the collagen-treated group.

*3.5. Effects of the CDOTs on Tail Bleeding and Mortality in Mice with ADP-Induced Pulmonary Thromboembolism*

Bleeding is a common side-effect of the antiplatelet drugs used in this study. We evaluated the effects of the CDOTs on bleeding time via a tail transection model. The bleeding time was 65.3 ± 4.2 s (*n* = 8) in the control group, as shown in Figure 6A. After 10 min of intravenous administration of the CDOTs (1 mg/kg), the bleeding time was 69.4 ± 5.7 s (*n* = 8). As is evident, the bleeding time was not significantly affected.

Further, we investigated mortality in mice with ADP-induced acute pulmonary thromboembolism treated with the CDOTs. The mortality rate of the animals with ADP-induced acute pulmonary thromboembolism (0.7 mg/g ADP) was 75% (i.e., deaths of 6 mice, *n* = 8); however, pretreatment with the CDOTs (1 mg/kg) considerably reduced the mortality rate to only around 25% (i.e., deaths of 2 mice, *n* = 8), as shown in Figure 6B. Hematoxylin–eosin was used to stain the lung tissues of the mice. As shown in Figure 6C, the control group exhibited severe pulmonary thrombosis (arrows), whereas the CDOTs (1 mg/kg) exerted substantial protective effects. Overall, these results showed that the synthesized CDOTs had an eminent antiplatelet effect in vivo without the side-effect of bleeding.

**Figure 6.** Effects of CDOTs on tail bleeding time and pulmonary thrombosis in experimental mice. (**A**) Bleeding time was measured through tail transection after 10 min of intravenous administration of PBS (control) or 1 mg/kg CDOTs. The bleeding time was continuously recorded until no sign of bleeding was observed for at least 10 s. (**B**) For the study of acute pulmonary thrombosis, PBS (control) or 1 mg/kg CDOTs was intravenously administered to mice, and ADP (0.7 mg/g) was then injected through the tail veins. (**C**) Pulmonary thrombosis (arrows) was observed by staining lung tissue sections with hematoxylin–eosin. Scale bar: 100 μm. Values represent mean ± SEM (*n* = 8). (**D**) Schematic illustration showing the inhibitory effect of CDOTs in human platelets. CDOTs potently inhibit human platelet activation by suppressing PKC activation and Akt, JNK1/2, and p38 MAPK phosphorylation without inducing cytotoxicity. CDOTs reduced the mortality in ADP-induced thromboembolic mice and did not affect bleeding tendency.

#### **4. Discussion**

CDOTs have extensively been applied in different fields for drug delivery. They have also been used in bioimaging and as effective biosensors for protein detection [14], considering their excellent biocompatibility, good water solubility, low toxicity, high photoluminescence, and high photostability. In this study, we synthesized a new type of CDOT from garlic. These nanoparticles were potent at hindering collagen-induced platelet aggregation and only reduced the slope of the aggregation curve (lag time) by U46619 and thrombin. Different physiological agonists (e.g., collagen, thrombin, and ADP) activated platelets. The primary activation of agonists may be enriched by the secondary activation induced by thromboxane A2 from arachidonic acid or by ADP from the granules in platelets. In the case of blood vessel injury, platelets adhere to the subendothelial matrix (collagen), causing granule secretion and platelet activation. Collagen, a matrix protein which exists in the vascular subendothelium and vessel wall, acts as substrate for platelet adhesion and potent platelet stimulator. In this manner, collagen exerts as a key player in platelet activation.

To exclude the possible cytotoxic effects of the synthesized CDOTs on human platelets, we estimated the leakage of cytosolic LDH. LDH, a soluble cytoplasmic enzyme which occurs in nearly all cells is released into the extracellular space when the plasma membrane is injured. We found that the alteration between the control and platelets subjected with 200 μM CDOTs was not substantial, suggesting potential hemocompatibility. This result is concordant with that reported by Shrivastava et al. [28], who established that silver nanoparticles did not disturb platelet membrane integrity, even at concentrations as high as 500 μM. In addition, LDH release was not noticed from platelets after exposure to 0.9–3.5 nM silver nanoparticles [29]. Moreover, in a recent study, Hajtuch et al. [30] reported that functionalized silver nanoparticles, such as AgNPs-GSH, AgNPs-PEG, and AgNPs-LA, ranging in size from 2 to 3.7 nm, inhibited platelet aggregation without releasing LDH. The results pertaining to the effects of nanoparticles on platelets are inconsistent. Studies have found that gold nanoparticles are inert [31] or activate [32] platelets. Silver nanoparticles have been reported to induce platelet aggregation both in human platelets and in an animal models [33]. Huang et al. [34] demonstrated that silver nanoparticles coated with polyvinyl pyrrolidone and citrate had no significant effects on human platelet aggregation. In this study, we found that the synthesized CDOTs effectively inhibited collagen-triggered platelet aggregation. Consistent with our findings, Ragaseema et al. [35] reported the inhibitory effects of silver nanoparticles on platelet aggregation. In addition, Shrivastava et al. [28] found that silver nanoparticles condensed ADP- and collagen-induced platelet activation with a reduction in the slope of aggregation. These inconsistencies could be attributed to differences in size, stabilization, and functionalization, as well as the method of nanoparticle synthesis.

In the present study, the CDOTs evidently inhibited collagen-stimulated platelet activation, implying that they were effective in inhibiting platelet activation via a prominent phospholipase C (PLC)-dependent mechanism. PLC, belonging to a family of kinases, hydrolyzes phosphatidylinositol 4,5-bisphosphate to yield two chief secondary messengers: diacylglycerol and inositol trisphosphate. Diacylglycerol activates PKC-inducing pleckstrin (p47) phosphorylation and ATP release in activated platelets, whereas inositol trisphosphate elevates calcium influx [36]. The observed antiplatelet effects of the CDOTs could be a result of the inhibition of the PLC–PKC cascade, leading to the suppression of Akt and MAPK activation. Akt (a downstream regulator of PI3K)-knockout mice have been found to demonstrate impaired platelet activation [22]. Hence, Akt inhibition may be considered as striking antithrombotic targets. Conservative MAPKs are classified into ERK1/2, p38 MAPK, JNK1/2, and big MAPK (ERK5) [37]. ERK1/2, JNK1/2, and p38 MAPK participate in platelet activation [37]. MAPK presents in platelets linked to the mechanistic role of several antiplatelet agents [38]. Adam et al. [39] reported that JNK1 knockdown reduced platelet aggregation, with JNK1−/<sup>−</sup> platelets displaying abnormal platelet granule secretion, and led to defective thrombus formation in mice. p38 MAPK is associated with thrombus formation, as evidenced in p38 MAPK−/<sup>−</sup> mice [39,40]. Therefore, PKC, Akt, JNK1/2, and p38 MAPK are regarded as major targets of antiplatelet agents. A study demonstrated that silica nanoparticles induced expressions of the phosphorylated JNK and p38 MAPK,

and suppressed ERK phosphorylation in human umbilical vein endothelial cells [41]. In the current study, the synthesized CDOTs markedly inhibited collagen-induced PKC, Akt, JNK1/2, and p38 MAPK phosphorylation in a concentration-dependent manner.

The GPVI receptor induces strong signaling through the protein tyrosine kinase pathways that results in the activation of PI3 and PLCγ and Ca2<sup>+</sup> release. Since, in this study, CDOT effectively inhibited collagen-induced platelet aggregation, GPVI receptor-mediated inhibitory signaling pathways maybe involved in this anti-aggregatory effect. Thus, we believe that the inhibition of these signaling molecules by the CDOTs may lead to inhibitory effects on platelet activation. Miller et al. proved the hypothesis that the biological activity of nanoparticles may be dictated by their composition, size, and charge [42]. They found that human- or bovine-derived nanoparticles inhibited platelet aggregation induced by two different agonists—one that activates the thrombin receptor and the other that activates the collagen receptor—and they suggested that the inhibitory effects may be nonspecific, possibly by reducing platelet–platelet interactions or by binding to these or other surface receptors. Consistent with these discoveries, the current in vitro observation of the potent inhibitory effect of CDOTs in collagen-induced human platelet aggregation may be due to its inhibition of platelet–platelet interactions or by preventing binding with the collagen receptor. However, the detailed mechanisms of these hypotheses remain to be explored.

The intravenous administration of nanoparticles has been previously reported to substantially inhibit platelet aggregation in mice, indicative of their in vivo antiplatelet effects [28]. Furthermore, Shrivastava et al. [28] conducted tail-bleeding assays to determine the presence of any opposing effect on bleeding time and found that mice continued to live normally after nanoparticle administration. Similarly, Kim et al. [43] found that silver nanoparticles were nontoxic to rodents, and in another more relevant study, gold nanoparticles were observed to inhibit both thrombosis and considerably improve the survival rates of mice, without increasing the bleeding risk [44]. These results are consistent with those of this study, where even we found that CDOT administration reduced mortality in thromboembolic mice without prolonging the bleeding tendency.

#### **5. Conclusions**

CDOTs have recently gained much attention worldwide. Herein we found that the synthesized CDOTs could actively inhibit human platelet activation by suppressing PKC activation and Akt, JNK1/2, and p38 MAPK phosphorylation. Furthermore, there was no cytotoxicity in vitro. The in vivo study revealed that the CDOTs had an antithrombotic effect on the ADP-induced pulmonary thromboembolic mice model. CDOTs attenuate ADP-induced severe pulmonary thrombosis via the potential recovering of lung histopathology, reducing mortality and maintaining the normal bleeding tendency in mice. Altogether, our results suggest that a direct application of CDOTs may contribute to the development of new antiplatelet drugs for the treatment of arterial thromboembolic diseases.

**Author Contributions:** Conceptualization and designing of the projects, P.V. and J.-R.S.; performing experiments, T.-Y.L., P.T., K.-C.L., and H.-M.C.; original manuscript preparation, T.J. and J.-R.S.; nanoparticle synthesis, P.V.; writing—review and editing, all authors. All authors have read and agree to the published version of the manuscript.

**Funding:** This work was supported by grants from the Ministry of Science and Technology of Taiwan (MOST-106-2113-M-001-032, NSC 102-2113-M-002-009-MY3, MOST 107-2320-B-038-035-MY2, and MOST 108-2320-B-038-031-MY3) and Taipei Medical University (DP2-109-21121-01-N-08-03).

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

#### **References**


*Nanomaterials* **2020**, *10*, 1254

© 2020 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* **Covalent Decoration of Cortical Membranes with Graphene Oxide as a Substrate for Dental Pulp Stem Cells**

**Roberta Di Carlo 1,**†**, Susi Zara 1,**†**, Alessia Ventrella 1,**†**, Gabriella Siani 1, Tatiana Da Ros 2, Giovanna Iezzi 3, Amelia Cataldi <sup>1</sup> and Antonella Fontana 1,\***


Received: 29 March 2019; Accepted: 8 April 2019; Published: 12 April 2019

**Abstract:** (1) Background: The aim of this study was to optimize, through a cheap and facile protocol, the covalent functionalization of graphene oxide (GO)-decorated cortical membrane (Lamina®) in order to promote the adhesion, the growth and the osteogenic differentiation of DPSCs (Dental Pulp Stem Cells); (2) Methods: GO-coated Laminas were fully characterized by Scannsion Electron Microscopy (SEM) and Atomic Force Microscopy (AFM) analyses. In vitro analyses of viability, membrane integrity and calcium phosphate deposition were performed; (3) Results: The GO-decorated Laminas demonstrated an increase in the roughness of Laminas, a reduction in toxicity and did not affect membrane integrity of DPSCs; and (4) Conclusions: The GO covalent functionalization of Laminas was effective and relatively easy to obtain. The homogeneous GO coating obtained favored the proliferation rate of DPSCs and the deposition of calcium phosphate.

**Keywords:** graphene oxide; covalent functionalization; cortical membranes; calcium phosphate deposition

#### **1. Introduction**

In this study, we focused our interest on cortical membranes, commonly used in oral surgical procedures, in order to improve their features thanks to a covalent enrichment with graphene oxide (GO). In particular, we used a type of cortical membrane, namely Osteobiol® Lamina provided by Tecnoss. Laminas, created by a registered trademark of Tecnoss, are made up of cortical bone of heterologous origin and are demonstrated to increase the rate of physiological resorption of the material [1]. Laminas have the compactness of bone tissue as well as a flexibility and adaptability that derives from the superficial decalcification process tuned for their preparation [1]. These materials are therefore used to improve bone tissue regeneration in cases in which it is important to reserve a space [2]. It is important to note that these tissue-derived materials are generally brittle and characterized by a low resistance to fracture. These drawbacks were overcome by enriching the original material, i.e., hydroxyapatite, with different additives such as alginate/chitosan [3], titania [4] or carbon nanotubes (CNT) [5].

Nowadays, graphene has emerged as a great alternative material for applications in biomedical and regenerative engineering. Graphene is a two-dimensional (2D) carbon-based material which has *sp<sup>2</sup>* bonded carbon atoms arranged in a honeycomb lattice structure, with extraordinary electrical, physical, and optical properties [6]. Since its discovery, graphene and its derivatives have been widely investigated for the development of electrical devices and for biomedical applications such as drug delivery systems, biosensors, and regenerative therapies [7]. Mechanically, graphene, despite its flexibility, appears to be one of the strongest materials ever tested. [6] It is transparent, able to conduct electricity and heat better than metals [8], chemically inert, and stable [9]. An increasing number of studies have recently focused on the expansion of new potential applications of graphene nanomaterials, with the aim to highlight the benefits of their use and to improve the application of these nanomaterials [10].

Despite these properties, graphene has a very low solubility in both organic and aqueous solvents. For this reason, hydrophilic graphene derivatives, namely Graphene Oxide (GO), have been widely used and tested for pharmaceutical and biomedical applications. GO is hydrophilic, does not tend to form aggregates, and is highly and homogeneously dispersible in water. GO has been demonstrated to be a biocompatible material whose limited cytotoxicity depends on final concentration, shape, sheet size, dispersibility, and degree of surface functionalization [10]. GO has been investigated for its ability to enhance the proliferation and differentiation of several types of stem cells [11].

The aim of this study was to achieve the covalent functionalization of Laminas, by exploiting, via a simple, cheap, and effective protocol, the capacity of oxygenated groups of GO to interact with cortical membrane surfaces. Indeed, previously investigated GO-coatings [12–14] were obtained by simply depositing GO on the elected substrates and therefore exploiting weak London, Van der Waals, or hydrogen-bonding interactions. The concentrations of GO chosen are those that, in preliminary biological assays and in previous studies [13,14], demonstrated not to be toxic for fibroblast cells and favor osteogenic differentiation in dental pulp stem cells (DPSCs) on collagen membranes. The idea is to demonstrate the ability of graphene oxide to improve Laminas biological properties as well as promote the adhesion, the growth and the osteogenic differentiation of DPSCs (Dental Pulp Stem Cells). DPSCs were chosen because of the easy access to the site collection. Besides, DPSCs have an extensive differentiation ability and their capacity to interact with biomaterials makes them ideal for tissue reconstruction [11].

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

#### *2.1. Materials*

Cortical membranes (0.5 <sup>×</sup> 0.5 <sup>×</sup> 0.2 cm) (Ostebiol® Lamina, Tecnoss) were a gift of Tecnoss dental s.r.l. Pianezza (TO), Italy. GO was purchased from Graphenea, San Sebastian, Spain as an aqueous solution of 4 mg/mL GO. This solution was diluted at the elected concentration and bath ultrasonicated for 10 min (Elmasonic P60H, 37 kHz, 180 W) before use.

All other reagents were product of analytical grade from Merck KGaA, Darmstadt, Germany and they were used as received.

#### *2.2. Enrichment with Graphene Oxide*

In order to prepare GO-coated Laminas, a protocol of covalent functionalization was optimized. Firstly, the cortical membranes was activated by using a UV/ozone lamp (PSD-UV4 Novascan UV Ozone System Base model, Novascan Technologies, Boone, NC, USA) for 15 min on each side. This permits the subsequent coating with the functional groups. Secondly, the Laminas were dipped in 1 M ethanolic solution of 3-aminopropyl triethoxysilane (APTES, commercial sample from Merck KGaA, Darmstadt, Germany) for 3 h to obtain a thin, stable aminosilane layer on the activated membranes. The so obtained aminosilane-functionalized membranes were rinsed with ethanol and deionized water. Thirdly, these aminosilane-functionalized cortical membranes were dipped in graphene oxide aqueous solution of two different concentrations, 5 or 10 μg/mL. In particular, 4 mL of homogenous dispersion

of GO in water were added to 10 cortical membranes (ca. 21 mg) in a baker. The GO solution was left into contact with samples overnight. Finally, membranes were left to dry at room temperature.

Samples were transferred in a 48 multi-well plate for the in vitro tests.

#### *2.3. Sterilization of Cortical Membranes*

Both pure and GO-coated Laminas were irradiated by using UV irradiation in the Herasafe KS 15, class II, type A2 biological safe cabinet (Thermo Fisher Scientific, North Logan, UT, USA) for 1 h on each side in order to sterilize the specimens.

#### *2.4. Apparatus for Chemico-Physical Characterization of Laminas*

Thermo-gravimetric analyses (TGA) were recorded on a TGA Q500 (TA Instruments, New Castle, DE, USA) on ca. 12 mg sample. The runs were performed under nitrogen atmosphere by equilibrating the samples at 100 ◦C for 20 min, following a ramp at 10 ◦C/min up to 800 ◦C.

The morphology of Lamina and GO-coated Laminas was evaluated by Atomic Force Microscopy (AFM), using a Multimode 8 Bruker AFM microscope (Bruker, Milan, Italy) coupled with a Nanoscope V controller and commercial silicon tips (RTESPA 300, resonance frequency of 300 kHz and nominal elastic constant of 40 N·m−1) with a typical apex radius of 8 nm in Peak Force and ScanAsyst™ in air mode.

By using this mode, it was possible, from the height panel, to calculate roughness and, from the force curves recorded at various points, to calculate the Young's modulus. In particular, NanoScope Analysis software 1.8 enables to select the force curves registered at each point of the scanned surface and, from each force curve, to calculate the Young's modulus by fitting the linear part of the retracting curve via a hertzian model. The deflection sensitivity and tip radius were calibrated, prior to use, against standard sapphire.

#### *2.5. Isolation and Culture of DPSCs*

The Local Ethical Committee of the University "G. d'Annunzio" Chieti-Pescara approved the project (approval number 1173, date of approval 31/03/2016), in agreement with the Declaration of Helsinki. Dental pulps were extracted from third molars derived from young male and female people (age range 18–28 years) which underwent surgical procedures. All patients signed informed consent. The study involved only impacted teeth without dental pathologies. After the extraction, the surrounding tissues were mechanically eliminated and processed as reported in our previous work [15].

Samples were rinsed with phosphate-buffered saline (PBS), maintained in Minimum Essential Medium (α-MEM) (Merck KGaA, Darmstadt, Germany) supplemented with 10% of Foetal Bovine Serum (FBS) and 1% antibiotics (penicillin/streptavidin mixture, EuroClone S.p.A, Milan, Italy) and sent to the laboratory for stem cells extraction [15]. When cells covered 80–90% of the flask area (subconfluence condition) they were subcultured. Antigen expression of CD29, CD45, CD105, CD73 CD90 and SSEA-4 was checked by flow cytometry [15].

#### *2.6. DPSCs Culture on Laminas*

Cells from the fourth passage (Figure S1 of the Supporting Information) were seeded on Laminas, 10,000 cells/cm<sup>2</sup> were used and cultured up to 28 days. Two hundred twenty Laminas for each experiment were used, fifty-five Laminas were used for each experimental point. Experiments were repeated for three times. At the established times cells were harvested and processed for the required analyses. The cells were cultured in α-MEM medium supplemented with differentiation factors such as 10 nM dexamethasone, 0.2 mM ascorbic acid and 10 mM β-glycerophosphate, as reported elsewhere [16,17].

#### *2.7. Scannsion Electron Microscopy (SEM) Analysis*

Samples were fixed with 1.25% glutaraldehyde in 0.1 M cacodylate buffer for 30 min, dehydrated through alcohol ascending series and then dried with hexamethyldisilazane followed by gold-coating. All micrographs were obtained at 15 kV on compact desktop Phenom XL SEM microscope.

#### *2.8. Alamar Blue Cell Viability Assay*

The Alamar blue test was performed in triplicate for each experimental sample at each experimental time. Cells viability was measured after 3, 7, 14 and 28 days of culture. The test is based on the reduction of Alamar blue reagent (Thermo Scientific, Rockford, IL, USA), performed only by viable cells, into a red product. At established experimental times the medium was replaced by a new one added with 10% of Alamar blue reagent, incubated for 4 h at 37 ◦C. A spectrophotometric reading at 570 and 600 nm wavelength was performed. The negative control was established as the value obtained without cells. The percentage reduction of Alamar blue reagent was calculated following the manufacturer instructions.

#### *2.9. Lactate Dehydrogenase (LDH) Cytotoxicity Assay*

To evaluate biomaterial cytotoxicity, LDH release into the medium was measured by means of CytoTox 96 non-radioactive cytotoxicity assay (Promega, Madison, WI, USA) at each time point (3, 7, 14 and 28 days). The LDH leakage in each well was normalized to the lysis value obtained in a lysis well of the same experimental point in which a lysis solution was added to the medium

#### *2.10. Alizarin Red S (ARS) Staining*

Alizarin red S is a calcium-sensing dye. DPSCs, differentiated towards the osteogenic phenotype, are able to deposit and to induce the mineralization of extracellular matrix rich in calcium phosphate, which can be identified by Alizarin red S. Calcium deposits are detectable for their bright orange-red color.

The DPSCs in each well were rinsed twice with PBS, PBS was discarded and DPSCs were fixed in paraformaldehyde 4% for 15 min at room temperature and then washed with deionized water. Alizarin red S staining solution 40 mM (Merck KGaA, Darmstadt, Germany) was added to each well and probed for 20 min at room temperature (RT) on a shaker. The wells were washed for five times in deionized water. Calcium deposits, stained in orange-red, were dissolved as follows: 10% acetic acid was added under shaking for 30 min. Laminas were scraped, the liquid containing deposits was collected and vortexed in a tube. Previously heated warm mineral oil (Merck KGaA, Darmstadt, Germany) was added, the tube maintained on ice for 5 min and eventually centrifuged at 20,000 *g* for 15 min. The supernatant was discarded and 10% ammonium hydroxide (Merck KGaA, Darmstadt, Germany) was added. The final solution was analyzed by a spectrophotometric reading performed at 405 nm wavelength.

#### *2.11. Statistical Analysis*

SPSS software version 16.0 (SPSS, Inc., Chicago, IL, USA) (Statistical Package for Social Science) and GraphPad Prism 5 were used to perform statistical analysis. Data were evaluated using one-way analysis of variance followed by the Tukey-Kramer post-hoc test. The results were expressed as the mean ± standard deviation (SD). *P* < 0.05 was considered to indicate a statistically significant difference.

#### **3. Results**

Laminas were enriched with GO at two different concentrations, 5 and 10 μg/mL. Photos of the obtained enriched cortical membranes are reported in Figure 1.

**Figure 1.** Photographs of (**A**) pure Lamina, (**B**) Lamina functionalized with 3-aminopropyl triethoxysilane (APTES) (see Experimental Section 2.2), (**C**) Lamina enriched with 5 μg/mL graphene oxide (GO) and (**D**) Lamina enriched with 10 μg/mL GO.

We tried to evaluate the amount of GO covalently attached to the functionalized Lamina by using TGA (Figure 2). While the GO sample presents the common behavior with an important weight loss at 200 ◦C, the Laminas profiles show a consistent weight loss at around 325 ◦C as for 5 μg/mL GO and 10 μg/mL GO (330 ◦C), even though for the last preparation a small implement of stability can be appreciated up to 270 ◦C with a difference in weight loss of 1.3% (9.1% vs. 10.4% in the case of control and 5 μg/mL GO).

**Figure 2.** Thermo-gravimetric analyses (TGA) of graphene oxide (black curve), bare Lamina (red curve), 5 μg/mL GO enriched Lamina (blue curve) and Lamina enriched with 10 μg/mL GO (green curve).

Bare and GO-enriched cortical membranes were analyzed by using AFM. In Figure 3, topographical and tridimensional micrographs as well as peak force error images of bare and GO-enriched (5 μg/mL and 10 μg/mL) cortical membranes are reported. From the images reported the changes of the topography of the surface on enrichment with GO are evident.

**Figure 3.** (**A**,**D**,**G**) Topographical, (**B**,**E**,**H**) Peak force error and (**C**,**F**,**I**) Three-dimensional Atomic Force Miscroscopy (AFM) images of bare Lamina (upper line), Lamina enriched with 5 μg/mL of GO (central line) and Lamina enriched with 10 μg/mL GO (bottom line).

From the height panel the Nanoscope analysis 1.8 software (Bruker, Milan, Itay) is able to recover the roughness indexes (i.e., the root-mean square roughness, Rq; the mean absolute value of the surface high deviations, Ra; i.e., the distance between the highest and lowest data points in the image, Rmax; the root-mean square of the surface slope, Sdq, and the ratio between the developed and the planar area, Sdr). We calculated these indexes for the bare membrane as the mean values of roughness recovered from two panels with a total surface area of 18 <sup>μ</sup>m2, were Rq <sup>=</sup> 53.0 <sup>±</sup> 10.2 nm, Ra <sup>=</sup> 43.3 <sup>±</sup> 9.3 nm, Rmax 255.5 ± 20.5 nm, Sdq 12.7 ± 1.6◦, and Sdr 2.5 ± 0.5%. The roughness indexes, calculated as the mean values recovered from three panels with a total surface area of 300 <sup>μ</sup>m2, were Rq <sup>=</sup> 216.0 <sup>±</sup> 21.2 nm, Ra = 175.0 ± 21.9 nm, Rmax 1303.3 ± 96.3 nm, Sdq 21.4 ± 1.8◦, and Sdr 7.16 ± 1.0% for the GO-coated sample with 5 μg/mL and Rq = 254.7 ± 56.12 nm, Ra = 205.7 ± 46.5 nm, Rmax 1311.0 ± 282.0 nm, Sdq 25.9 ± 10.9◦, and Sdr 11.5 ± 7.6% for the GO-coated sample with 10 μg/mL (See Supporting Information, Figures S8–S10).

Scansion electron microscopy (SEM) experiments (Figure 4) allowed to observe morphology differences in the investigated Laminas. As seen from SEM images (compare Figures 4B and 4A) the covalent functionalization with amino silane brought about a significant deformation of Lamina Surface. The subsequent coating with GO restored the typical layered structure of GO [13], with layered regions increasing on increasing GO concentration (compare Figures 4C and 4D).

#### *Nanomaterials* **2019**, *9*, 604

**Figure 4.** Scansion electron microscopy (SEM) images of (**A**) bare Lamina, (**B**) APTES-treated Lamina, (**C**) 5 μg/mL GO-coated Lamina and (**D**) 10 μg/mL GO-coated Lamina. Magnification 3000×. Scale bar: 200 μm.

DPSCs were cultured on Laminas with medium containing differentiation factors up to 28 days; 3, 7, 14, and 21 days were chosen as experimental times.

Before starting the evaluation of the biological parameters, an SEM analysis, after 7 and 14 days of culture, was performed in order to evaluate DPSCs morphology, spread and adhesion on Laminas. After 7 days of culture, cells are detectable on all the observed experimental points: DPSCs cultured on control Laminas are flat and spread throughout the surface, some granules of inorganic matrix are starting to be deposited. DPSCs cultured on APTES-treated Laminas appear isolated, with short cytoplasmic extensions, probably suffering and they do not cover all the surface of the Lamina. DPSCs grown on both 5 μg/mL GO- and 10 μg/mL GO-coated Laminas form a uniform cell layer on the biomaterial, they appear completely flat and in close contact with each other; an isolated cell is not recognizable. White granules of inorganic matrix can be identified especially on 5 μg/mL GO-coated Lamina. The same trend is revealed after 14 days of culture (Figure 5).

**Figure 5.** SEM images of Dental Pulp Stem Cells (DPSC)) cultured on bare Laminas (CTRL1), APTES-treated (CTRL2), 5 μg/mL GO-coated (GO5) and 10 μg/mL GO-coated (GO10) Laminas for 7 and 14 days. Magnification 3000×.

Cell viability was measured by means of Alamar Blue Assay after 3, 7, 14, and 28 days. After 3 days of culture the viability level does not show any significant difference among the tested experimental points, whereas after 7 days of culture an appreciable increase in viability level is recordable when DPSCs are cultured on GO-enriched Laminas. In particular, the cell viability is almost doubled for DPSCs cultured on 5 μg/mL GO-coated Laminas with respect to the control and it is comparably high for DPSCs cultured on 10 μg/mL GO-coated Laminas. Both the percentage of Alamar Blue reduction recorded on 5 μg/mL GO- and 10 μg/mL GO-coated membranes are statistically significant with respect to the control (*p* < 0.001). The metabolic activity of cell cultured on control and on GO-coated Laminas further augments (Figure 6) until 14 days of culture. By day 14, the number of viable cells reach a plateau, suggesting that those surfaces are advancing into confluence. On the other hand, it is worth noting that the proliferation rate of DPSCs cultured on APTES-treated Laminas is much lower, reaching the maximum percentage of Alamar Blue reduction at 28 days, when the difference with the other samples cancels out.

**Figure 6.** Alamar blue assay in DPSC cultured on bare Laminas (CTRL1), APTES-treated (CTRL2), 5 μg/mL GO-coated (GO5) and 10 μg/mL GO-coated (GO10) Laminas for 3, 7, 14, and 28 days. Forty Laminas were used for each experimental point, ten Laminas per experimental time. The histogram represents Alamar blue reduction percentage, data shown are the mean (±SD) of three separate experiments. Zero time % reduction Alamar Blue: 15.72%; \* Day 7: GO5 and GO10 Laminas vs. control (CTRL1) Laminas *p* < 0.001; Day 14 control Laminas, GO5-coated and GO-10 coated Laminas vs. APTES-treated Laminas (CTRL2) *p* < 0.001.

The cytotoxicity of the biomaterial was evaluated through LDH assay by measuring the percentage of released LDH within the culture medium after 3, 7, 14, and 28 days of culture. After 3 days of culture the cytotoxicity level is higher than 70% for all tested samples except for DPSCs cultured on 10 μg/mL GO-coated Laminas. In fact, this sample shows a released LDH percentage significantly lower than that measured for the three others samples. After 7 days of culture the cytotoxicity level starts to decrease with respect to that measured after 3 days of culture in all the investigated samples except for DPSCs cultured on APTES-treated Laminas. In fact, the released LDH percentage for this sample appears still higher than 70%, whereas the percentage decreases under 60% for cells grown on control Laminas and under 40% for DPSCs cultured on GO-coated Laminas. The cytotoxicity level does not change thereafter for GO-coated samples and a statistically significant reduction of released LDH percentage is detected for DPSCs grown on 10 μg/mL GO-coated Laminas with respect to the control (Figure 7). Again DPSCs cultured on APTES-treated Laminas show the highest released LDH percentage (>50%) among the four Laminas investigated.

Bone matrix deposition was measured through Alizarin Red staining, a calcium-sensing dye able to identify extracellular quantities of calcium phosphate. Alizarin Red staining was performed after 21 and 28 days of culture. After 21 days of culture, a marked decrease of synthetized calcium phosphate could be detected in DPSCs cultured on APTES-treated Laminas compared with the control and with GO-coated Laminas. Conversely, after 28 days of culture, a statistically significant increase in calcium phosphate deposition is detected in DPSCs cultured on 5 μg/mL GO-enriched Laminas with respect to all other tested samples (Figure 8).

**Figure 7.** Lactate Dehydrogenase (LDH) assay of DPSC cultured on bare Laminas (CTRL1), APTES-treated (CTRL2), 5 μg/mL GO-coated (GO5) and 10 μg/mL GO-coated (GO10) Laminas for 3, 7, 14, and 28 days. Forty Laminas were used for each experimental point, ten Laminas per experimental time. Released LDH is reported as percentage. Data shown are the mean (±SD) of three separate experiments. Zero time LDH release (%): 73.06 \* Day 3: 10 μg/mL GO-coated Laminas (GO10) vs. control (CTRL1) *p* < 0.05; \* Day 7: APTES-treated, 5 μg/mL GO-coated (GO5) and 10 μg/mL GO-coated Laminas (GO10) vs. control (CTRL1) *p* < 0.001; control Laminas, APTES-treated Laminas vs. 5 μg/mL GO-coated Laminas (GO5) *p* < 0.001; control Laminas, APTES-treated Laminas vs. 10 μg/mL GO-coated Laminas (GO10) *p* < 0.001; \* Day 14, day 28: APTES-treated, GO10 Laminas vs. control (CTRL1) Laminas *p* < 0.005.

**Figure 8.** The histogram represents optical density (OD) values of solubilized calcium deposits (orange-red stained) obtained after Alizarin Red staining on bare Laminas (CTRL1), APTES-treated (CTRL2), 5 μg/mL GO-coated (GO5) and 10 μg/mL GO-coated (GO10) Laminas. Twenty Laminas were used for each experimental point, ten Laminas per experimental time. Data shown are the mean (±SD) of three separate experiments. \* Day 21: APTES-treated Laminas vs. control (CTRL1) Laminas *p* < 0.005; \* Day 28: control Laminas, APTES-treated, 10 μg/mL GO-coated Laminas (GO10) vs. 5 μg/mL GO Laminas (GO5) *p* < 0.005.

#### **4. Discussion**

Laminas were covalently enriched with GO, by using APTES as the linker between the Lamina and the GO sheets. This type of functionalization was chosen in order to create on the biomaterial a layer of graphene oxide covalently bound to the scaffold. As a matter of fact, in a previous study [12], porcine bone granules, enriched with GO by exploiting simple physical deposition, were implanted in animals for three months and excess GO was detected in the form of GO aggregates in both hard and soft tissues.

TGA did not allow us to properly quantify the amount of GO functionalized onto Laminas, because the amount of GO was very low and the two materials, cortical membrane and GO, evidenced a weight loss at similar temperatures. An approximately 1.3% GO could be calculated at least for the more concentrated 10 μg/mL sample.

Nevertheless, the GO demonstrated good distribution, through AFM and SEM analyses, on the Lamina and changed completely the appearance of the surface of the cortical membrane. Despite SEM appearance, the enrichment with GO rendered the surface more rough, as already recently evidenced in the case of GO-enrichment of porcine bone granules and collagen membranes [12,13]. Indeed, SEM is not the best technique in order to discriminate GO coverage percentage, but SEM images highlight the formation of layered GO over the Lamina surface (Figure 4). AFM measurements instead evidenced that GO-enriched Laminas are rougher than bare membranes and present a much wrinkled structure (compare panels B, E, and H and C, F, and I in Figure 3).

In particular, the calculated roughness indexes, Rq and Ra, indicate that the non-coated membrane is characterized by lower peaks and therefore a lower roughness as compared to the GO-coated samples, as confirmed by Rmax values. The surface indexes, Sdr and Sdq, confirmed Ra, Rq, and Rmax data because they evidenced a surface enlargement induced by the presence of GO with more and steep peaks, respectively. Nevertheless no significant differences were highlighted between the samples enriched with 5 or 10 μg/mL GO, likely due to a saturation-like effect of the surface with the lowest concentration of GO investigated.

The measured Young's elastic modulus, obtained as an average value calculated from 20–25 force curves in samples of 10 μm × 10 μm dimensions, is 0.77 ± 0.46, 0.83 ± 0.62, and 1.00 ± 0.27 GPa in bare cortical membrane, 5 μg/mL and 10 μg/mL GO-coated membranes, respectively (See Supporting Information, Figures S2–S7). Therefore it does not change very much on GO enrichment, although a small increase could be monitored on increasing the concentration of GO, indicating that GO contributes to matrix stiffening. It is interesting to note that the SD is very high (60%) in the commercial cortical membrane due to the presence of pores and defects, keeps a high value for 5 μg/mL coated Laminas but reduces in Laminas enriched with 10 μg/mL GO, thus highlighting the presence of a homogeneous GO layer in the latter sample. These values are of the same order of those of polyethylene (1.5–2 GPa) and polystyrene (3–3.5 GPa), substrates which have been previously demonstrated to be ideal substrates for the growth of stem cells.

By considering that mesenchymal stem cells demonstrated [18] to sense matrix elasticity and preferentially differentiate depending of the stiffness of the substrate, such values, indicative of stiff matrices, appear proper to favor expression of an osteogenic lineage.

Laminas coated with 5 μg/mL and 10 μg/mL GO were then tested in an in vitro model, by seeding and culturing DPSCs on Laminas surface, in order to evaluate the biocompatibility of GO-enriched Laminas, in terms of cell viability, cytotoxicity, and mineralized matrix deposition.

During the DSPCs differentiation, GO enrichment positively modifies the biological parameters evaluated, thus indicating a good tolerability and an improved biocompatibility. Indeed, GO enrichment, both at 5 and 10 μg/mL concentration, improves the cell spread throughout the surface of the biomaterial thus allowing to hypothesize that GO enrichment is able to promote the adhesion process and favor the formation of a uniform cell layer (See Figure 5). Moreover, GO coating enhanced the growth rate of DPSCs. In fact, cells seeded on GO-coated Laminas after 7 days of culture show cell viability values two-fold that of bare Laminas evidencing confluence after 14 days of culture (see Figure 6). It is worth

noting that the above mentioned GO-induced cell viability value is not related to GO concentration as both samples evidence the same effect. Nevertheless it is important to stress that these similarities can be explained by the above mentioned saturation-like effect, with GO covering almost completely the Lamina surface already at the lowest investigated concentration. These results may be associated to the capacity of GO to favor protein adsorption [19]. Indeed, it has been demonstrated that serum proteins absorb quickly and spontaneously to graphene oxide surface to form a corona complex [20] and this adsorption, that demonstrate to be selective for different proteins, may affect adhesion, proliferation, and/or osteogenic differentiation of stem cells [19]. It was also demonstrated that induction of human mesenchymal stem cells (hMSC) differentiation towards different tissue lineages depended on the degree of π-π stacking with graphene and hydrogen bonding as well as electrostatic interactions with GO [19]. In the present case, a positive effect towards adhesion and viability of DPSCs may be induced also by the highly wrinkled surface associated to a small increase in stiffness obtained on GO enrichment. A similar evidence has been already demonstrated for highly convoluted methacrylate-functionalized GO membrane [21] that favored spontaneous stem cell differentiation towards bone lineage even in the absence of osteogenic growth factors.

The cytotoxicity level also appeared significantly reduced for GO-enriched Laminas during 28 days of culture and actually a statistical significant increase of membrane integrity was detected for Laminas enriched with 10 μg/mL GO at all the investigated times. It is worth noting that a slightly different behavior characterizes 5 μg/mL GO-coated Laminas, with cytotoxicity slightly increasing and reaching that of the bare Laminas at 14 and 28 days of culture. These results show that, although GO functionalization demonstrated to promote favorable biological effects on DSPCs, it is necessary to carefully tune the concentration of GO bound to Laminas in order to reach the best compromise of effectiveness and biocompatibility. Indeed, different studies evidenced that doses, as well as size, is a fundamental parameter to consider in order to fully characterize GO toxicity [22]. On the other hand, a relatively high toxicity was detected for APTES-treated Laminas. Despite an APTES-treated different material, such as nanoparticles, did not show any toxic effect on cell membrane integrity [23], polyamines demonstrated [24] to promote leakage of liposomal content from 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) liposomes due to interactions of primary ammonium groups with phospholipidic head groups. Similarly APTES-treated Laminas could promote analogous ammonium-phospholipidic headgroups interactions, thus explaining the sustained LDH leakage from cell membrane and the chronic cytotoxicity responsible for the lower proliferation rate as compared to cells seeded on the other investigated samples. These data are very interesting because they highlight that, after aminosilane-functionalization, the subsequent treatment with GO allows to override the negative effect evidenced in the presence of APTES-treated Laminas on cell cytotoxicity.

These results are further supported by mineralized bone matrix deposition which appears increased after 28 days of culture and therefore at the end of the osteogenic differentiation [25] but only for DPSCs cultured on 5 μg/mL GO-coated Laminas. They highlight the important role of GO as responsible for a faster and more intense promotion of bone matrix deposition. To conclude, this is a preliminary study on biocompatibility and lack of cytotoxicity of the GO-functionalized Laminas and further investigations are needed in order to fully characterize their biological properties. First of all a detailed investigation focused on the tuning of GO concentration needed for the covalent functionalization of Laminas will allow to optimize the risk-to-benefit balance and clarify all the factors affected by GO enrichment.

#### **5. Conclusions**

This study demonstrated that the relatively homogeneous coating of investigated commercial cortical membranes with GO was relatively easy to obtain. It favored the proliferation rate of DPSCs probably due to the capacity of GO to adsorb proteins present in the medium. Clear evidences of reduced toxicity were evidenced and Laminas enriched with GO 5 μg/mL demonstrated a statistical significant increase of calcium phosphate deposition. The present study is particularly promising and we believe that this material holds potential as useful substrate to facilitate in vivo bone regeneration. Nevertheless it highlights the need to further investigate GO-coated samples in order to tune the concentration of GO that demonstrates the best osteogenic activity and biocompatibility.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2079-4991/9/4/604/s1, Figure S1: DPSCs observed with a light microscope before detachment and seeding for osteoblastic differentiation, Figures S2–S7: Original AFM micrographs and representative force curves of pure Lamina, Lamina enriched with 5 μg/mL GO and 10 μg/mL GO used for the mechanical studies, Figures S8–S10: Original AFM micrographs used for the roughness index calculation of pure Lamina, Lamina enriched with 5 μg/mL GO and 10 μg/mL GO, Tables S1–S3: Roughness indexes for pure Lamina, Lamina enriched with 5 μg/mL GO and 10 μg/mL GO.

**Author Contributions:** R.D.C. enriched the Laminas with GO, S.Z. performed biological analyses, A.V. performed AFM measurements, G.S. and G.I. data curation and supervision, T.D.R. performed TGA analyses, A.C. and A.F. conceptualization, writing original draft, and supervision.

**Funding:** This work was carried out with the financial support from the University 'G. d'Annunzio' of Chieti-Pescara and MIUR.

**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*
