*p* > 0.05, \*\*\* *p* < 0.001, when dendrimer-only-based nanocomplexes are compared to gold–dendrimer nanocomplexes.

#### *3.3. The Band Shift Assay*

The binding of mRNA to the prepared NPs can be seen in Figure 4.

**Figure 4.** Band shift assay of the interaction between (**A**) G5D, (**B**) Au:G5D, (**C**) G5D:FA, (**D**) Au:G5D:FA, and mRNA. Incubation mixtures (20 μL) in HBS contained varying amounts of the nanoparticle preparation and 0.05 μg FLuc-mRNA corresponding to *w*/*w* ratios of 1:1, 2:1, 3:1, 4:1, 5:1, 6:1, 7:1, and 8:1 in lanes 2–8, respectively (**A**–**D**). Lane 1: naked mRNA control. Arrows indicate endpoint ratios.

All prepared NPs were able to bind and complex with the mRNA. This can be credited to the ability of G5D to become protonated at physiological pH [11]. G5D and Au:G5D NPs completely retarded the mRNA at ratios of 2:1 and 3:1 (*w*/*w*), respectively, while both G5D:FA and Au:G5D:FA NPs completely retarded mRNA at a ratio of 4:1 (*w*/*w*).

#### *3.4. Ethidium Bromide Dye Displacement Assay*

All NPs displaced ethidium bromide (ETB), indicating a significant degree of mRNA compaction, which bodes well for their stability and protection under physiological conditions. The degree of mRNA compaction by the G5D and Au:G5D NPs ranged from 50 to 80%, while that of G5D:FA and Au:G5D:FA NPs ranged from 40 to 70% (Figure 5).

**Figure 5.** Ethidium bromide displacement assay of (**A**) G5D, (**B**) Au:G5D, (**C**) G5D:FA, and (**D**) Au:G5D:FA NPs. Arrows indicate a point of complexation.

#### *3.5. RNase A Digestion Assay*

To assess the ability of the NPs to protect the mRNA cargo against nucleases, which would be encountered in circulation in an in vivo system, an RNase A digestion assay was conducted.

Figure 6 clearly shows the exceptional ability of all NPs to fully protect mRNA following treatment with 10% RNase A, as depicted by the presence of undigested bands in all tested ratios. By contrast, the treatment of naked mRNA with RNase A showed complete degradation (negative control), as illustrated in Lane 2.

**Figure 6.** RNase A digestion assay of nanocomplexes. (**A**) G5D, (**B**) Au:G5D, (**C**) G5D: FA, (**D**) Au:G5D:FA. Control: naked mRNA in the absence (+ = positive control) or presence (− = negative control) of RNase A. Lanes 1–3 contain nanocomplexes at sub-optimum, optimum, and supra-optimum nanoparticle: mRNA ratios. (**A**) 1:1, 2:1, 3:1; (**B**) 2:1, 3:1, 4:1; (**C**) 3:1, 4:1, 5:1; (**D**) 3:1, 4:1, 5:1 (*w*/*w*). Red-colored numbers indicate the optimum binding ratios.

#### *3.6. The MTT Assay*

To monitor cell viability after treatment with prepared nanocomplexes in selected cell lines, the MTT assay was conducted. This assay uses the MTT reagent, which enters the cells and passes into the mitochondria, where it is reduced to an insoluble, purple-colored formazan product that can be quantified spectroscopically and used as an indication of metabolically active cells. No significant (*p* > 0.05) change in cell viability was observed following treatment with all nanocomplexes. Higher cell viabilities (80–97%) were observed in all cell lines for the Au:G5D:mRNA and Au:G5D:FA:mRNA nanocomplexes, compared to the G5D:mRNA and G5D:FA:mRNA nanocomplexes (68–78%) (Figure 7A,B).

**Figure 7.** *Cont*.

**Figure 7.** Cell viability assay of nanocomplexes containing (**A**) Au:G5D and Au:G5D:FA; and (**B**) G5D and G5D:FA, in HEK293, HepG2, Caco-2, MCF-7, and KB cells. Cells were incubated with nanocomplexes containing 0.05 μg F*Luc*-mRNA at indicated ratios (*w*/*w*). Nanocomplexes were prepared at sub-optimum, optimum, and supra-optimum ratios. Data are presented as means ± SD (*n* = 3). Control = untreated cells. \* *p* > 0.05.

Noticeably, all FA-targeted nanocomplexes showed higher cell viability than their untargeted nanocomplex counterparts (average cell viability of 88% for Au:G5D:FA and 72% for G5D:FA).

#### *3.7. Apoptosis Assay*

Cell death was also studied by evaluating the ability of NPs to induce apoptosis in selected cell lines. All nanocomplexes induced little or no apoptosis in the cells, as evidenced by very few apoptotic (yellow-orange/red) cells visible and low apoptotic indices (AI) (Figure 8 and Table 2). Noticeably, the AI values of Au:G5D:mRNA and Au:G5D:FA:mRNA nanocomplexes were significantly (*p* < 0.0001) lower than those of the G5D:mRNA and G5D:FA:mRNA nanocomplexes particularly, in all cell lines (Table 2).


**Table 2.** Apoptotic indices of nanocomplexes in selected cell lines.

**Figure 8.** Fluorescence images of (**A**) HEK293, (**B**) HepG2, (C) Caco-2, (**D**) MCF-7, and (**E**) KB cells treated with test and control nanocomplexes prepared at sub-optimum ratios for 24 h, showing induction of apoptosis. Green = live (L), light orange = early apoptotic (EA), and dark orange = late apoptotic (LA) cells. Scale = 100 μm.

#### *3.8. Transfection and Competition Assays*

The ability of the NPs to deliver mRNA was evaluated in folate receptor-negative cell lines, HEK293, Caco-2, and folate receptor-positive cell lines HepG2, MCF-7, and KB (KB > MCF-7 > HepG2), with KB cells often being used as a model for folate receptors (FRs) [40]. The transfection efficacy of the nanocomplexes was assessed as a function of weight ratios (sub-optimum, optimum, and supra-optimum). The transfection activity of the Au:G5D:mRNA and Au:G5D:FA:mRNA nanocomplexes (Figure 9A,B) was much higher than that of the naked mRNA (control). Moreover, the transfection levels in HEK293 and Caco-2 cells were significantly (*p* < 0.001) lower than those elicited in the receptorpositive cells.

All nanocomplexes showed excellent transfection activity, with Au:G5D:mRNA and Au:G5D:FA:mRNA nanocomplexes (Figure 9A) showing higher transfection efficiencies ranging from 5 × 107–6 × 108 RLU/mg protein. On the other hand, G5D:mRNA and G5D:FA:mRNA nanocomplexes (4 × <sup>10</sup>7–3 × <sup>10</sup><sup>8</sup> RLU/mg protein) produced decreased transfection activity (Figure 9B). Noticeably, the Au:G5D:FA:mRNA nanocomplexes showed a four-fold increase in transfection activity (6 × 108 RLU/mg protein), compared to Au:G5D:mRNA nanocomplexes (2 × <sup>10</sup><sup>8</sup> RLU/mg protein) at the optimum ratios in the FR positive cell line, MCF-7.

To confirm the mechanism of the cellular uptake of the nanocomplexes, a competition assay was conducted. This involved flooding the cells with excess free FA (250 μg) before exposure to the FA-targeted nanocomplexes (Au:G5D:FA:mRNA and G5D:FA:mRNA). The assay was conducted in the cell lines with overall higher targeted transgene expression, viz. MCF-7 and KB cell. A significant (*p* < 0.01) drop of approximately 30% in transgene activity was observed as depicted in Figure 10, which suggests that a large portion of these nanocomplexes were taken up by receptor-mediated endocytosis [41], confirming that FA receptor mediation was a key player in the high transgene expression obtained.

**Figure 9.** Transgene expression for (**A**) Au:G5D and Au:G5D:FA nanocomplexes, and (**B**) G5D and G5D:FA nanocomplexes in HEK293, HepG2, Caco-2, MCF-7, and KB cells. Nanocomplexes contained 0.05 μg mRNA with varying amounts of nanoparticles to constitute the sub-optimum, optimum, and supra-optimum (*w*/*w*) ratios. Control 1 = untreated cells. Control 2 = cells treated with naked F*Luc*mRNA. The transgene expression is reported as RLU/mg protein. Data are presented as means ± SD (*n* = 3). \*\*\*\* *p* < 0.0001 for optimum ratios.

**Figure 10.** Competition studies of FA-targeted mRNA nanocomplexes in (**A**) MCF-7 and (**B**) KB cells. Cells were first exposed to excess folic acid (250 μg) then treated with FA-targeted nanocomplexes at selected ratios. Transgene expression is reported as RLU/mg protein. Data are presented as means ± SD (*n* = 3). \*\* *p* < 0.01.

#### **4. Discussion**

All NPs were successfully synthesized to produce spherical, monodispersed NPs. NP synthesis was confirmed by UV-vis and NMR spectroscopy, which also confirmed that the G5D polymer and FA moiety were successfully conjugated to the AuNPs. The Au:G5D NPs produced a redshift in the spectrum, whereas the Au:G5D:FA NPs produced a blue shift. The G5D generally has a very weak peak between 280 and 285 nm [14], especially at higher or at physiological pH due to the presence of the protonated amine groups of the G5D [14,42]. In this study, a small peak was noted at 283 nm. However, this peak often seems to disappear at lower pH. In NMR, the formation of Au:G5D NPs resulted in the downfield shift of protons 4, 5, and 6 of G5D, which indicated the interaction of the surface of the AuNPs with the internal amines of the dendrimers [43].

Favorably sized NPs (<200 nm) with the most zeta potentials, except for the AuNPs on their own being above 20 mV, were produced. All nanocomplexes, with the exception of the G5D:FA nanocomplexes (265.2 nm) fell within the ideal size range (100–200 nm) required for gene delivery via non-specific or receptor-specific uptake [44–46]. Zeta potential measurements greater than +25 mV or less than −25 mV are reported to be associated with good colloidal stability [47]. These AuNPs alone showed poor stability (−7.3 mV), but upon G5D and FA functionalization, the stability improved immensely to +20.9 mV for Au:G5D and +29 mV for Au:G5D:FA. This confirms that the functionalization of the NPs with the dendrimers improved their stability, as seen in a recent study where dendrimer was used to functionalize selenium NPs [14]. The improved stability achieved with the targeted NPs could be due to the partial shielding effect imparted by FA, in addition to the repulsive cationic amine groups on the dendrimers, which prevents particle aggregation [14,48]. From these findings, it can be predicted that these nanocomplexes may be efficient in delivering mRNA.

The differences observed in the binding efficiency between the FA-targeted and untargeted NPs could be due to the possible shielding of the cationic charges of the dendrimers on the targeted NPs by the FA moiety, which meant that more positive charges and more NPs were required to fully neutralize the negative charges on the mRNA [49]. Overall, the NP:mRNA nanocomplex formation occurred at very low ratios (*w*/*w*), which could be accredited to the single-stranded nature of the mRNA, which is quickly embedded by the highly cationic G5D. The G5D and Au:G5D showed greater quenching of the ethidium bromide fluorescence, which could be attributed to more amine groups being available to bind the mRNA [14]. The compaction was seen for the targeted nanocomplexes, further suggested a weaker binding of the mRNA, which could translate into easy dissociation of the mRNA from the nanocomplexes during transfection, hence avoiding degradation by the lysosomal compartment, and in turn, enhancing gene-transfection efficiency [50]. Overall, all NPs were able to efficiently bind and compact mRNA to varying degrees.

The integrity of the nanocomplexes may be compromised by degrading nuclease enzymes such as RNase A, leading to a reduced transgene expression [51]. The good nuclease protection afforded by the NPs in this study could be due to the highly organized globular structures that formed as a result of the electrostatic interaction between the negatively charged single-stranded mRNA and the highly cationic G5D-containing NPs [52]. The use of the RNase enzyme was a stringent test for these NPs due to its specificity for RNA molecules and was reported previously [53,54]. Various studies have used the less-specific fetal bovine serum containing nucleases to determine the integrity of RNA-based nanocomplexes [55,56] to achieve similar results. In the circulatory system, it is possible that the nanoparticles may encounter less-specific enzymes and possibly at lower concentrations as well. However, this assay confirmed that all NPs afforded exceptional protection to the mRNA cargo, boding well for future in vivo studies.

The first step towards understanding the biocompatibility of a delivery system often involves the use of cell-culture-based studies, commencing with the assessment of cytotoxicity. The gold-containing NPs achieved higher cell viability, which may be due to the presence of the gold in the NP and partly to the reduction of the cationic charges of the 1◦ amines of the G5D, some of which are responsible for stabilizing the entrapped AuNPs [21]. Furthermore, unmodified AuNPs have been shown to have little or no impact on cytotoxicity in non-cancer HEK293 and cervical cancer (HeLa) cells [24], which could be attributed to their inherent biocompatibility and favorable physicochemical properties that have been widely mentioned. Noticeably, all FA-targeted nanocomplexes showed higher cell viabilities compared to their untargeted nanocomplex counterparts, which could be as a result of the shielding effect of FA, which may have covered a portion of the positive charges on the surface of G5D, hence reducing the strong electrostatic interaction between the cells and the NPs [17]. Overall, more than 80% of cells were still viable after

being exposed to the gold-containing nanocomplexes at the selected ratios, suggesting that these nanocomplexes were superior and well-tolerated in all tested cell lines, and therefore relatively safe to use. Apoptosis studies corroborated these results, confirming that the Au:G5D:mRNA and Au:G5D:FA:mRNA nanocomplexes were safe and stable and did not induce any significant apoptotic effects.

The introduction of naked mRNA into cells is known to be associated with poor transgene expression, mainly due to enzymatic degradation [57], as evidenced in the RNase A digestion assay. All nanocomplexes displayed significant transfection in the cell lines tested. The Au:G5D:mRNA and Au:G5D:FA:mRNA nanocomplexes produced the highest luciferase activity, which could be due to three reasons. First, since the translation of mRNA occurred in the cytoplasm—and the major limiting step, which is the nuclear pore entry, was avoided—resulting in an increased transgene expression. Second, the transfection studies were conducted over a duration of 48 h, and more protein may have expressed, considering that mRNA may have a limited half-life [5]. Lastly, the efficient encapsulation of the mRNA by the dendrimer and its exceptional buffering capacity could have helped protect the mRNA from degradation and facilitated the endosomal escape of the nanocomplexes [58].

HepG2 cells exhibited lower luciferase expression, possibly due to fewer receptors on their cell surface compared to MCF-7 and KB cells. The low targeted expression is associated with a lack of specific transcription factors and cell-surface receptors [35]. The higher transfection efficiencies of Au:G5D:mRNA and Au:G5D:FA:mRNA nanocomplexes can be accredited to the entrapment of AuNPs within the 1◦ amines of the dendrimers, which helped preserve the structural integrity of the dendrimers, allowing for efficient interaction between the dendrimers and the mRNA [21]. This could lead to favorable cellular uptake and high gene expression. The decreased transfection activities of the G5D:mRNA and G5D:FA:mRNA nanocomplexes could be due to their higher cytotoxicity compared to their gold-containing counterparts and the poor dissociation between the mRNA and the G5D due to their strong binding affinity. The mRNA may have been entrapped by a network formed by the branches of the dendrimer. Earlier studies have demonstrated a direct correlation between the binding affinity of the single-stranded mRNA to cationic polymers and transgene expression [59].

The Au:G5D:FA:mRNA nanocomplexes showed a superior transfection activity to the Au:G5D:mRNA nanocomplexes, most likely due to ligand–receptor interaction that occurred between the FA and the FRs abundantly, decorating the surface of the MCF-7 and KB cells [60]. It is generally known that FA has a high affinity for FRs overexpressed by a majority of cancer cells [35], with KB cells generally regarded as models for the folate receptor, as previously mentioned [40]. The significant (*p* < 0.01) drop in transgene activity in the competition assay suggested that a large portion of these nanocomplexes were taken up by receptor-mediated endocytosis, confirming that FA receptor-mediation was a key player in the high transgene expression obtained.

#### **5. Conclusions**

Both Au:G5D and Au:G5D:FA NPs were highly efficient in F*Luc*-mRNA binding and delivery. They formed stable nanocomplexes and afforded excellent protection to the mRNA against RNases. Furthermore, more than 80% cell viability was observed, suggesting that these nanocomplexes were well tolerated by all cells. This was also demonstrated in their superior transfection efficiency, indicating the significant and synergistic roles played by both the dendrimer and the AuNPs in their formulation. This study further confirmed that folate-receptor-mediated delivery was the main route of entry into the receptor-positive cells, as evidenced by the transfection levels in the FA receptor negative cell lines, being significantly lower than that in FA receptor positive cell lines. Since this proof in principle study has shown potential, future studies would encompass the NP optimization for in vivo delivery using a therapeutic mRNA molecule.

**Author Contributions:** Conceptualization, L.S.M. and M.S.; methodology, L.S.M. and F.M., software, L.S.M., F.M., and A.D.; validation, M.S. and A.D.; data curation, L.S.M., F.M., and A.D.; resources, M.S.; writing—original draft preparation, L.S.M., F.M., and A.D.; writing—review and editing, M.S.; supervision, M.S.; project administration, M.S.; funding acquisition, M.S. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the National Research Foundation of South Africa, grant numbers 120455 and 129263.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data and contributions presented in the study are included in the article. Further inquiries can be directed to the corresponding author.

**Acknowledgments:** The authors acknowledge members of the Nano-Gene and Drug Delivery group for advice and technical support.

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

#### **References**


## *Article* **Engineered Human Heavy-Chain Ferritin with Half-Life Extension and Tumor Targeting by PAS and RGDK Peptide Functionalization**

**Shuang Yin 1, Yan Wang 2, Bingyang Zhang 1, Yiran Qu 1, Yongdong Liu 3, Sheng Dai 4, Yao Zhang 3, Yingli Wang 2,\* and Jingxiu Bi 1,\***


**Abstract:** Ferritin, one of the most investigated protein nanocages, is considered as a promising drug carrier because of its advantageous stability and safety. However, its short half-life and undesirable tumor targeting ability has limited its usage in tumor treatment. In this work, two types of functional peptides, half-life extension peptide PAS, and tumor targeting peptide RGDK (Arg-Gly-Asp-Lys), are inserted to human heavy-chain ferritin (HFn) at C-terminal through flexible linkers with two distinct enzyme cleavable sites. Structural characterizations show both HFn and engineered HFns can assemble into nanoparticles but with different apparent hydrodynamic volumes and molecular weights. RGDK peptide enhanced the internalization efficiency of HFn and showed a significant increase of growth inhibition against 4T1 cell line in vitro. Pharmacokinetic study in vivo demonstrates PAS peptides extended ferritin half-life about 4.9 times in Sprague Dawley rats. RGDK peptides greatly enhanced drug accumulation in the tumor site rather than in other organs in biodistribution analysis. Drug loaded PAS-RGDK functionalized HFns curbed tumor growth with significantly greater efficacies in comparison with drug loaded HFn.

**Keywords:** ferritin; drug delivery; tumor targeting; half-life extension

#### **1. Introduction**

Ferritin is one of the most attractive protein nanocages for drug delivery, due to its extraordinary thermal and chemical stability. In mammals, ferritin is a 12 nm sphere with an 8 nm cavity, made up of 24 subunits [1]. Two types of ferritin subunits exist in mammal tissues, called heavy-chain (H-chain) and light-chain (L-chain) (21 kDa and 19 kDa), respectively. Both two types of subunits consist of five α-helices (helices A-E), one long loop connecting helix B and C (BC loop) and three turns connecting helices. Exposed BC loop of Human H-chain ferritin (HFn) has a binding site of human transferrin receptor 1 (TfR1) and gives rise to an intrinsic tumor active targeting ability [2]. Researchers have loaded various chemotherapeutics into H-chain ferritin and explored its anti-tumor efficacy. For example, 5-fluorouracil attached Au nanoparticles inside ferritin decreased IC50 against HepG2 cells by 15 times [3]. A single dose of doxorubicin (DOX) loaded HFn (HFn/DOX) successfully inhibited TfR1 overexpressed HT-29 human colon cancer cells growth in mice [4]. Neuronal drugs carbachol and atropine loaded ferritin is proven to be able to regulate pancreatic cancer progression [5].

**Citation:** Yin, S.; Wang, Y.; Zhang, B.; Qu, Y.; Liu, Y.; Dai, S.; Zhang, Y.; Wang, Y.; Bi, J. Engineered Human Heavy-Chain Ferritin with Half-Life Extension and Tumor Targeting by PAS and RGDK Peptide Functionalization. *Pharmaceutics* **2021**, *13*, 521. https://doi.org/10.3390/ pharmaceutics13040521

Academic Editors: Francisco José Ostos, José Antonio Lebrón, Pilar López-Cornejo and Patrick J. Sinko

Received: 24 February 2021 Accepted: 6 April 2021 Published: 9 April 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 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 (https:// creativecommons.org/licenses/by/ 4.0/).

In spite of ferritin's multiple advantages, it is still facing challenges as a drug nanocarrier. It has a half-life in circulation of approximate 2 h in rats, shorter than the majority of other drug nanocarriers because of its relatively small particle size. Wang fused albumin binding domain (ABD) to increase ferritin half-life to 17.2 h [6]. In addition, the innate tumor targeting ability of HFn cannot be guaranteed in all tumors. The expression level of its receptor, human TfR1, varies in different tumor cell lines and in different stages of tumor progression [7,8]. Head and neck cancer, colorectal cancer and cervical cancer tissues have the highest expression level of human TfR1, whilst no human TfR1 was detected in carcinoid, prostate and testicular tumor tissues [9]. Human TfR1 is also ubiquitously expressed in healthy human tissues, such as bone marrow, lung, colon and liver, to import iron into cells, so the usage of HFn has the risk of undesired drug accumulation in healthy tissues.

To address (1) the short half-life and (2) the limited tumor targeting ability of HFn, two functional peptides were fused to HFn subunit C-terminal to construct three functionalized HFns (HFn-PAS, HFn-GFLG-PAS-RGDK and HFn-PLGLAG-PAS-RGDK). One peptide, PAS peptide, comprises repetitive P, A and S residues. It was designed by Schlapschy, and aimed to mimic poly ethylene glycol (PEG) [10]. In three previous studies, Falvo et al. have fused 40 aa and 75 aa PAS peptides to human ferritin subunit at N-terminal to increase halflife in circulation [11–13]. Another peptide is a tetrapeptide named as RGDK. It belongs to tumor penetration peptide (TPP) and possesses two functions. RGDK enhances drug tumor delivery, and drug distribution inside whole tumor tissue instead of only tumor cells alongside tumor vessels [14]. It specifically binds to two receptors, integrin αvβ3/5 and neuropilin-1, both overexpressed in a wide range of tumor cells [15]. Integrin αvβ5 is highly expressed in cancers such as gliomas and urothelial cancer, and neoropilin-1 expression is upregulated in ovarian cancer, colorectal cancer and stomach cancer [9]. Therefore, the addition of RGDK peptide can improve HFn tumor targeting ability and broaden HFn application. The GFLG (Gly-Phe-Leu-Gly) and PLGLAG (Pro-Leu-Gly-leu-Ala-Gly) in HFn-GFLG-PAS-RGDK and HFn-PLGLAG-PAS-RGDK are enzyme cleavable sites responding to cathepsin B and matrix metalloproteinase-2/9 (MMP-2, MMP-9), respectively [16,17]. Both enzymes are overexpressed in tumors but Cathepsin B is located inside cell lysosome and MMP-2 is secreted outside tumor cells [18,19]. The PAS-RGDK functional moiety in these two dually-functionalized HFns is theoretically to be cleaved from HFn before and after cell internalization, respectively. In total, four HFn-based proteins were compared with each other to investigate the impacts of PAS and RGDK on HFn performance as a drug nanocarrier.

A total of four HFn-based proteins were expressed in *Escherichia coli* (*E. coli*), and purified. High-Performance Size Exclusion Chromatography coupled with Multiple Angle Laser Light Scattering (HPSEC-MALLS) was used to characterize protein structures. In vitro and in vivo tests were designed to compare anti-tumor drug delivery performance of four HFn-based proteins in tumors lacking overexpressed human TfR1. Therefore, 4T1, a BALB/c mice breast tumor cell line was selected; 4T1 does not express human TfR1 and overexpresses integrin αvβ3/5 and neuropilin-1 [20,21]. Cellular uptake assay investigated RGDK functionalization impact on 4T1 cellular internalization efficiency. Intracellular distribution monitored if drug can be released from proteins and enter nucleus for killing tumor cells after internalization. Cytotoxicity assay compared IC50 values of drug carried by four HFn-based proteins. Pharmacokinetic study mainly assessed PAS impact on halflife in circulation. Biodistribution study assessed tumor targeting ability of four HFn-based proteins. In vivo anti-tumor test was conducted to compare the tumor growth inhibition efficacy of DOX carried HFn and functionalized HFns.

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

#### *2.1. Materials*

A total of four recombinant HFn-based proteins (HFn, HFn-PAS, HFn-GFLG-PAS-RGDK and HFn-PLGLAG-PAS-RGDK) were expressed in *Escherichia coli* (*E. coli*) BL21 (DE3). HFn-PAS was constructed by inserting a 15 aa flexible linker (GGGSGGGTGGGSGGG), an

enzyme-cleavable site GFLG, a 40 aa PAS peptide (ASPAAPAPAPAAPAPSAPAASPAA-PAPASPAAPAPSAPA) together with another 5 aa flexible liner (GGSGG) to HFn Subunit C-terminus. HFn-GFLG-PAS-RGDK was constructed by adding RGDK tretapeptide to HFn-PAS C-terminus. HFn-PLGLAG-PAS-RGDK, was designed by substitution of enzymecleavable site GFLG in HFn-GFLG-PAS-RGDK by a six residue MMP-2 cleavable site PLGLAG. Proteins were purified using a two-step pathway. Briefly, HFn was purified by heat-acidic precipitation at 60 ◦C, pH 4.5 5 min followed by butyl fast flow hydrophobic interaction chromatography (GE Healthcare, Waukesha, WI, USA). The other three functionalized HFns were purified by heat-acidic precipitation at 60 ◦C, pH 4.5 5 min followed by mono Q ion-exchange chromatography (GE Healthcare, Waukesha, WI, USA).

Doxorubicin hydrochloride (DOX) was purchased from Dalian Meilun Biotechnology (Dalian, China). 4T1 cells were purchased from Cellbank (Sydney, NSW, Australia). RPMI-1640 medium, penicillin-streptomycin solution (100 ×), fetal bovine serum (FBS), 0.25% trypsin-EDTA (1 ×) solution, Hoechst 33258 reagent and MTT reagent were purchased from Invitrogen (Thermo Scientific, Adelaide, SA, Australia). Propidium iodide and trypan blue solution were bought from Sigma-Aldrich (Sydney, NSW, Australia). All of the other reagents were of analytical reagent quality. Mili Q water was utilized throughout the whole procedure, produced by Merck Mili Q direct (Melbourne, VIC, Australia).

#### *2.2. HPSEC-MALLS Characterization of Purified Proteins and DOX Loading*

The four protein purities were analyzed by reducing 12% SDS-PAGE (Bio-Rad, Gladesville, NSW, Australia). Sizes and molecular weights (Mws) were measured by HPSEC-MALLS. In HPSEC-MALLS analysis, TSK G4000 SWxl column (Tosoh bioscience, Tokyo, Japan) was connected to HPLC (Shimadzu, Melbourne, VIC, Australia) coupled with DAWN MALLS and Optilab refractive index (RI) detector (Wyatt, Santa Barbara, CA, USA). Equilibration buffer was 20 mM phosphate buffer (PB), 0.1 M Na2SO4, pH 7.0. Flow rate was 0.8 mL min−1. Absorbance of fractions at 280 nm was monitored. Sample loading volume was 50 μL.

In DOX loading, briefly, 1 mg mL−<sup>1</sup> HFn-based protein in 20 mM phosphate buffer, 5 mM guanidinium chloride, pH 7.5 was heated at 50 ◦C for 6 h with 0.2 mg mL−<sup>1</sup> DOX. Excessive DOX was separated from DOX loaded protein (protein/DOX) by desalting on Hitrap G25 desalting column (GE Healthcare, Waukesha, WI, USA) using AKTA PURE (GE Healthcare, Waukesha, WI, USA). Collected protein/DOX underwent measurement of OD280 and OD480. DOX has absorbance at both 280 and 480 nm, and protein only has absorbance at 280 nm. Therefore, two (2) assumptions were made: (1) OD480protein/DOX = OD480DOX; (2) OD280protein/DOX = OD280DOX + OD280protein. Standard OD vs. C linear curves of DOX and HFn-based proteins were determined by serial concentrations of DOX (1–40 μg mL−1) and proteins (0.1–1.2 mg mL−1). Standard curves were used to calculate the concentration of DOX (CDOX) and the concentration of proteins (Cprotein) in protein/DOX. Consequently, the calculation of loading ratio in protein/DOX was as follows:

$$\text{Loading ratio} = \frac{\text{number of DO}}{\text{number of protein}} = \frac{\text{CDOX} \bullet \text{Mw}\_{\text{protein}}}{\text{C}\_{\text{protein}} \bullet \text{Mw}\_{\text{DOX}}} \tag{1}$$

#### *2.3. Cellular Uptake Test*

The 4T1 cell line was cultured in RPMI-1640 medium supplemented with 10% FBS at 37 ◦C in a 5% CO2 atmosphere. Cellular uptake test procedure was modified from a previous paper [15]. For each protein/DOX group, three different treatments were conducted to obtain three fluorescence intensities, total fluorescence, internalized fluorescence, fluorescence after RGDK peptide pre-incubation.

The procedure was as follows: (1) Cell seeding. 4T1 Cells in the exponential growth phase were seeded in 24-well plates at a density of 1 × 105 cells per well and cultured for 48 h for attachment. (2) RGDK peptide pre-incubation in wells for fluorescence after RGDK peptide pre-incubation determination. To investigate the impact of fused RGDK

on cellular uptake characteristics, 500 μM free RGDK peptide was pre-incubated with the cell for 1 h at 37 ◦C to saturate RGDK specific receptors. (3) Drug incubation. The media with or without RGDK peptide in all wells were discarded and cells were washed with phosphate buffered saline (PBS) three times, prior to adding 100 μL serum-free culture medium containing free DOX or protein/DOX (15 μg mL−<sup>1</sup> DOX-equivalent). Then the cells were incubated for 90 min at 37 ◦C and washed three times with PBS to remove drugs. (4) Trypan blue quenching. In wells for internalized fluorescence and fluorescence after RGDK peptide pre-incubation determination in all five groups, cells were incubated with trypan blue (0.25% in 0.85% NaCl) for 5 min at 25 ◦C, and then washed five times with PBS to remove trypan blue. (5) Detachment of cells for flow cytometry analysis. A total of 400 μL of 0.25% trypsin–0.05% EDTA solution was added to all wells for digestion for 5 min at 37 ◦C and 2 mL of complete medium was added to stop the digestion. Detached cells were spun at 112 rcf for 3 min at 4 ◦C and re-suspended in 1 mL PBS. In total, five microliters of propidium iodide (PI) was added to incubate with cells for 10 min at 25 ◦C for differentiation of alive and dead cells in flow cytometry detection. (6) Flow cytometry analysis. Csampler flow cytometry (Becton Dickinson, San Jose, CA, USA) was employed to determine the mean fluorescence of 5000 cells in each sample. A cell control underwent PI staining but without drug incubation, trypan blue and RGDK peptide treatment was used for gating and parameter setting prior to sample detection. PE channel (excitation laser light: 488 nm, emission: 578 nm) was utilized for DOX fluorescence detection. Mean fluorescence intensity of each sample was recorded.

#### *2.4. Intracellular Distribution Analysis*

Intracellular distribution analysis was designed to monitor if DOX carried by HFnbased proteins could reach tumor cell nucleus for disruption of cell division. Exponentially growing 4T1 cells were placed on a 6-well plate at a density of 4 × <sup>10</sup><sup>5</sup> cells per well and cultured for 24 h. One cover-glass slide was put in each well prior to seeding. The medium was then discarded and cells were treated with fresh media containing protein/DOX or free DOX (20 μM DOX-equivalent) in 2 mL per well for 3 h. Drugs in wells were then removed and cells were washed three times using PBS. Fresh complete medium was added to wells for another 36 h incubation. Subsequently, the cells were washed three times with PBS and fixed with 4% paraformaldehyde for 10 min at 25 ◦C. Following another three times wash with PBS, cell nucleus were stained with 0.5 μg mL−<sup>1</sup> Hoechst 33258 at 25 ◦C for 5 min. A ZOE fluorescence cell imager (Bio-Rad, Gladesville, NSW, Australia) was used to visualize cells. Images of cells under bright filed channel, green channel (Excitation: 480/17 nm, Emission: 517/23 nm) and blue channel (Excitation: 355/40 nm, Emission: 433/36 nm) were captured. Green channel and green channel monitored Hoechst 33258 and DOX signal, respectively.

#### *2.5. Cytotoxicity Study*

The cytotoxicity of four protein/DOX and free DOX against 4T1 cell was evaluated by MTT assays. Exponential growth-phase cells were digested by 0.25% trypsin-0.05% EDTA, and cell density was adjusted to 1 × 105 cells per mL by complete medium. 100 <sup>μ</sup>L of cells were seeded in wells of 96-well plates. Then, four wells without cells were adopted as blank control on each plate. After incubation for 24 h, medium was replaced with new complete medium separately containing either free DOX or protein/DOX, whose concentrations ranged from 0 to 30 μg mL−<sup>1</sup> (equivalent DOX). Four 0 μg mL−<sup>1</sup> DOX wells on each plate were cell control wells. After incubation for another 60 h, the media were removed and cells were washed three times by PBS. Then, 90 μL of new complete medium with 10 μL of MTT solution was added to each well for another 4 h. A total of 100 μL dimethyl sulfoxide (DMSO) was added to wells to ensure complete solubilization of the formed form-azan crystals. Finally, the absorbance of the solution was measured at 595 nm (background: 630 nm) by a Microplate Reader (Biotek, Winooski, VT, USA). Absorbance of each well (Awell) was defined as A595–A630. Cell viability (%) were calculated using Equation (2). Acell was the Awell of cell controls, and Ablank was the Awell of blank controls. IC50 value of each group was calculated using dose-response fitting in origin 9.0 software (Originlab, Northampton, MA, USA).

$$\text{Cell viability (\%)} = (\text{A}\_{\text{well}} - \text{A}\_{\text{blank}}) / (\text{A}\_{\text{cell}} - \text{A}\_{\text{blank}}) \times 100 \text{ (\%)}\tag{2}$$

#### *2.6. Pharmacokinetics Study*

All animal experiments were performed with the approval of the medical ethics committee of Shanxi University of Chinese Medicine (Approval Number 2019LL137, approval date: 13 June 2019). Specific-pathogen free Sprague Dawley rats (male, 230–250 g, SPF Biotechnology Co., Ltd. Beijing, China) were randomly assigned to six groups (three rats in each group), and administrated with PBS, free DOX and protein/DOX (3.0 mg kg−<sup>1</sup> DOX equivalent) separately via intravenous injection at tail vein. After injection, blood samples were collected from the retro orbital sinus at fixed time points (10, 30 min, 1, 2, 4, 8, 12, 24, 36, 48 h) and followed by clotting for at least 0.5 h at 37 ◦C. Serum was obtained by centrifugation at 4032 rcf for 30 min at 4 ◦C. Finally, 100 μL of serum of each sample was transferred to a 96-well microplate, and the DOX contents were determined using SpectraMax i3x microplate reader (Molecular devices, San Jose, CA USA). Excitation wavelength was set at 480 nm and emission at 580 nm. Meanwhile, the standard curve of the fluorescence intensity with varying concentrations of DOX in rat serum was also measured for quantitative analysis. Half-lives of DOX and protein/DOX were calculated using Drug Analysis System 2.0 software (Drug China, Shanghai, China) by fitting data in single-compartment mode.

#### *2.7. In Vivo Imaging*

The four HFn-based proteins were first labelled by Sulfo-cy5 NHS ester (Lumiprobe, Hunt Valley, MD, USA) with a molar ratio of 1:30 (Protein to Cy5) and the uncoupled Cy5 was removed by Hitrap G25 desalting chromatography. As 4T1 is a BALB/c breast tumor cell line, female BALB/c mice were chosen to establish tumor-bearing animal model. <sup>1</sup> × <sup>10</sup><sup>6</sup> 4T1 cells in 100 <sup>μ</sup>L of PBS were injected into right armpit of 8-week old female BALB/c mice (specific-pathogen free, SPF Biotechnology Co., Ltd. Beijing, China) to form mice tumor model. Each group had three mice. When tumor volume reached about 300 mm3, a 150 μL sample of Cy5 or protein-Cy5 conjugates (0.2 mg kg−<sup>1</sup> Cy5 equivalent) was intravenously injected into the tumor-bearing mice via tail vein. After treatment, the mice were anesthetized using isoflurane at 2, 4, 6.5, 12, 24 and 52 h and fluorescence images were taken under excitation wavelength of 646 nm and emission wavelength of 662 nm using FX Pro in vivo imaging system (Bruker BioSpin, Carteret, NJ, USA).

#### *2.8. Anti-Tumor Assay*

Then, 1 × 106 4T1 cells in 100 <sup>μ</sup>L of PBS were injected into right armpit of 8-week old female BALB/c mice. For in vivo inhibition of tumor progression assessment, female BALB/c mice bearing 4T1 tumors of approximate 250 mm<sup>3</sup> in size were randomly assigned to six groups (*n* = 6 in each group) and treated with protein/DOX (3 mg kg−<sup>1</sup> DOX equivalent), free DOX (3 mg kg−1), or PBS via 200 μL intravenous injection. The drug injection was carried out every 4 days for two doses. The volumes of tumors were measured every other day. Mice were monitored for up to 17 days post-implantation and then sacrificed. Primary tumors were harvested for ex vivo imaging.

#### *2.9. Statistical Analysis*

Data were presented in Mean ± Standard deviation (SD). T-test was applied to evaluate statistical significance of results. *p* value < 0.05 was considered significant.

#### **3. Results**

#### *3.1. Purified HFn-Based Protein Characterizations and Drug Loading*

The purity of each protein after purification reached above 90% based on the SDS-PAGE gel (Figure 1A), calculated by density scan using software Image J [22]. The apparent subunit molecular weights of HFn-PAS, HFn-GFLG-PAS-RGDK and HFn-PLGLAG-PAS-RGDK on gel were higher than their theoretical molecular weights (26 kDa, 26.5 kDa and 26.6 kDa), which are due to the hydration of PAS peptides. Bands of two PAS-RGDK functionalized HFns in SDS-PAGE gel were slightly higher than HFn-PAS probably due to the presence of extra residues.

**Figure 1.** Characterizations of purified HFn-based proteins. (**A**), 12% reducing SDS-PAGE analysis of purified proteins. Lane 1, HFn; 2, HFn-PAS; 3, HFn-GFLG-PAS-RGDK; 4, HFn-PLGLAG-PAS-RGDK. (**B**), HPSEC-MALLS chromatogram of HFn. (**C**), HPSEC-MALLS chromatogram of HFn-PAS. (**D**), HPSEC-MALLS chromatogram of HFn-GFLG-PAS-RGDK. (**E**), HPSEC-MALLS chromatogram of HFn-PLGLAG-PAS-RGDK.

The apparent hydrodynamic radius and molecular weight (Mw) of four HFn-based proteins were further characterized by HPSEC-MALLS analysis. In Figure 1B–E, the horizontal Mw lines of the main peaks show the uniform Mws of all four HFn-based proteins. Table 1 lists the hydrodynamic size and Mw of each protein. Due to PAS peptides, HFn-PAS possessed 1.4 nm higher apparent hydrodynamic radius in contrast to HFn. The adding of RGDK peptide and enzyme-cleavable site into HFn slightly further increased hydrodynamic radius. HFn-GFLG-PAS-RGDK and HFn-PLGLAG-PAS-RGDK were 1.75 nm and 1.91 nm larger than HFn, respectively. MALLS determined Mw order is

in accordance with theoretical order: HFn-PLGLAG-PAS-RGDK > HFn-GFLG-PAS-RGDK > HFn-PAS > HFn, and average Mw of all three proteins determined are similar to their theoretical Mw (Table 1).


**Table 1.** Hydrodynamic radius and molecular weight determined by HPSEC-MALLS.

After purification, the model drug DOX was loaded by thermally induced passive diffusion. On average, incubation with DOX at 50 ◦C loaded 33.5, 38.4, 36.9 and 42.1 DOX in one HFn, HFn-PAS, HFn-PLGLAG-PAS-RGDK and HFn-GFLG-PAS-RGDK nanocage, respectively. HFn DOX loading ratio in this study is comparable with previous pH-induced disassembly–reassembly method [23] and 8 M urea method adopting HFn [4].

#### *3.2. Cellular Uptake Efficiency*

In cellular uptake test, we investigated the RGDK functionalization impact on cellular uptake efficiency and the mechanism. Figure 2 presents DOX fluorescence intensities of all groups. Total fluorescence of DOX measured in flow cytometry came from two sources, DOX internalized by 4T1 cells and un-specifically bound to cell membranes. Trypan blue treatment quenched the signal from membrane-bound DOX, and, therefore, a lower internalized fluorescence intensity compared with total fluorescence intensity was observed in all groups (Figure 2).

**Figure 2.** Mean DOX fluorescence intensity in 4T1 cellular uptake test. Data were represented as mean ± standard deviation (*n* = 3), \* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001. Symbol '\*' on top of column represents the significance of *p* value between this column and the white column (total DOX fluorescence) in the same group.

Free DOX showed significantly greater internalized cellular uptake than others. HFn-GFLG-PAS-RGDK/DOX and HFn-PLGLAG-PAS-RGDK/DOX had the second highest efficiencies and were significantly different from the rest two. This means the insertion of RGDK peptide has significantly enhanced cellular uptake efficiency. HFn-PAS/DOX and HFn/DOX had similar internalized cellular uptake efficiencies. RGDK peptide preincubation treatment, with the use of excessive amount of free RGDK, is intended to mask RGDK-specific receptors, integrin αvβ3/5 and neuropilin-1, on cells to hamper RGDK-related cellular uptake. In Figure 2, the uptake of HFn-GFLG/PLGLAG-PAS-RGDK groups were significantly inhibited by RGDK peptide pre-incubation while in other groups no obvious difference occurred. After the pre-incubation of RGDK peptide, internalized fluorescence intensities of HFn-GFLG-PAS-RGDK/DOX and HFn-PLGLAG-PAS-RGDK/DOX were similar to that of HFn/DOX and HFn-PAS/DOX. This proves that RGDK facilitated 4T1 cells' internalization of HFn-GFLG-PAS-RGDK/DOX and HFn-PLGLAG-PAS-RGDK/DOX by providing RGDK-specific receptor-mediated pathway.

The difference of tumor cell uptake efficiencies lies in various uptake mechanisms. DOX is a small molecule and enters cells via passive diffusion. Passive diffusion is energyfree and concentration gradient-driven. It is quicker compared with all other internalization pathways when directly incubating drugs with cells. As 4T1 does not express human TFR1, HFn/DOX and HFn-PAS/DOX probably enter the cell through non-specific pinocytosis. In contrast with HFn/DOX and HFn-PAS/DOX, HFn-GFLG-PAS-RGDK/DOX and HFn-PLGLAG-PAS-RGDK/DOX have an extra internalization pathway by binding to RGDK recognized receptors, integrin αvβ3/5 and neuropilin-1.

#### *3.3. Intracellular Distribution*

DOX is an anthracycline topoisomerase inhibitor and exerts its function mainly inside the cell nucleus [24]. Free DOX directly diffuses into nucleus and disrupts cell division after internalization, whilst the protein/DOX are supposed to first be broken down by enzymes in lysosome to release loaded DOX and then reach nucleus. Intracellular distribution test was designed to check if the drugs loaded on HFn-based proteins could enter cell nucleus to kill tumor cells. In Figure 3, the blue color indicates where cell nucleus is and green color represents the fluorescence from DOX. Clearly, the majority of DOX has entered and accumulated inside nucleus of 4T1 cells in all groups, as the cyan color is the dominant color in merged images. This shows that the DOX in four protein/DOX groups could accumulate in 4T1 cell nucleus, the same as free DOX.

#### *3.4. Functionalization Effect on Cytotoxicity*

In order to test the inhibition of protein/DOX on tumor cell proliferation, we adopted an MTT assay. All DOX loaded HFn-based proteins demonstrate obvious anti-proliferation abilities (Figure 4). Free DOX group had the lowest IC50 (Table 2) and this is due to its relatively high cellular internalization efficiency. Inhibition impacts of HFn-GFLG-PAS-RGDK/DOX and HFn-PLGLAG-PAS-RGDK/DOX on tumor cell growth were similar and the second strongest, HFn-PAS/DOX ranked third, and HFn/DOX showed the worst anti-proliferation effect. T-test shows there was significant differences between IC50 values of free DOX and HFn-GFLG/PLGLAG-PAS-RGDK/DOX group (*p* < 0.05). Significant differences of IC50 values were also found between HFn-GFLG/PLGLAG-PAS-RGDK/DOX and the other two HFn-based protein/DOX groups (*p* < 0.05). That implies the RGDK in HFn-GFLG/PLGLAG-PAS-RGDK/DOX has enhanced HFn/DOX performance in terms of drug cytotoxicity towards tumor cells.

**Figure 3.** Intracellular distribution of DOX in 4T1 cells. Under the Bio-Rad Zoe cell imager, blue: nucleus after being stained with Hoechst 33258. Green: DOX because of its intrinsic fluorescence. Cyan: merged florescence signal.

**Figure 4.** Proliferation inhibition on 4T1 cells. Data were mean ± standard deviation (*n* = 4).



#### *3.5. Functionalization Effect on Pharmacokinetic Profile*

Pharmacokinetic profile of all four protein/DOX and free DOX were obtained through tail vein injection of healthy Sprague Dawley rats. Line chart of DOX concentrations in plasma over time (10 min–48 h) is shown in Figure 5 and half-lives in circulation of all protein/DOX are listed in Table 3. Standard curve of fluorescence intensity-doxorubicin concentration is shown in supplementary material Figure S1. Plasma drug concentrations of free DOX group rats reduced rapidly right after administration (Figure 5). At 10 min, average plasma drug concentration was 15 μg mL−1. Then, 8 h later, almost all free DOX was eliminated from circulation. In terms of HFn/DOX group rats, their plasma drug cleaning out speed ranked the second, with average drug concentration of 21.6 μg mL−<sup>1</sup> at 10 min and below 5 μg mL−<sup>1</sup> after 12 h. Average plasma drug concentrations in all three functionalized HFn/DOX group rats (approximate 35 μg mL−1) were more than double of those in free DOX group at 10 min. 48 h after administration, more than 5 μg mL−<sup>1</sup> drug still remained in plasma of all three functionalized HFn/DOX group rats. Single compartment fitting of the drug concentration-time curve was applied to evaluate drug half-life in circulation. The results show free doxorubicin only had about 25 min of half– life in circulation (Table 3). HFn/DOX half-life, approximate 3 h, was 7.3 times of free DOX. PAS peptide in HFn-PAS/DOX has increased half-life almost 4.9 times (14.96 h) compared with HFn/DOX. The extra RGDK residues in HFn-GFLG-PAS-RGDK/DOX further extended the half-life to 17.61 h. HFn-PLALGA-PAS-RGDK/DOX possessed the longest half-life, 18.93 h. Differences in half-life of all three functionalized HFns/DOX compared with HFn/DOX and free drug were statistically significant in *t*-test (*p* < 0.001).

**Figure 5.** Plasma concentrations of protein/DOX and free DOX in Sprague Dawley rats of different groups. Data were expressed as mean ± SD (*n* = 3).

**Table 3.** Half-life of each protein/DOX in Sprague Dawley rats (*n* = 3).


#### *3.6. Functionalization Effect on Protein Biodistribution*

To monitor distribution of all HFn-based proteins in tumor-bearing mice after tail vein administration over time, we used in vivo imaging. In this analysis, fluorescence label Cy5 was attached to all employed proteins and free cy5 worked as free drug control. The reagent in use reacts with primary amine group on protein outer surface. On the outer surface of HFn assembly, there are 24 exposed subunit N-terminals –NH2 groups and 144 Lys (K) residues. Due to the large number of accessible reaction sites, the possibility of cy5 blocking some or all K residues of RGDK in HFn-GFLG/PLGLAG-PAS-RGDK is low. The cy5 conjugation is unlikely to affect RGDK function. Real-time biodistribution of Cy5 attached proteins and free Cy5 were visualized in BALB/c mice with 4T1 tumor in right armpit, and fluorescence intensities of tumor areas were recorded. Two control groups, mice injected with free cy5 and HFn-cy5 were scanned at the same time, and mice in other three groups (HFn-PAS-cy5, HFn-PLALGA-PAS-RGDK-cy5 and HFn-GFLG-PAS-RGDK-cy5) were scanned together.

Figure 6A shows the top half of the mice where there was fluorescence signal captured by camera. 4–12 h after injection, fluorescence of free cy5 in liver was captured by camera (Figure 6A). At 24 and 52 h, fluorescence was barely visible. Free cy5 preferred to accumulate in the liver, perhaps due to the fact that liver is the main organ for metabolism. Theoretically, as a nanoparticle, HFn has passive tumor targeting ability. However, from the results in Figure 6A, signal of HFn-cy5 fluorescence was captured in liver rather than in tumor from 4 to 12 h. It seems that HFn-cy5 did not show desirable tumor targeting ability and it preferred liver. The particle size of HFn is probably still too small to achieve desirable passive tumor targeting ability. No obvious fluorescence was captured at all time points in HFn-PAS-cy5 group (Figure 6A). However, as is shown in Figure 6B, there actually was fluorescence detected in tumor area. Perhaps because of the sharp contrast between signal intensities of HFn-PAS-cy5 and HFn-GFLG/PLGLAG-PAS-RGDK-cy5, lower intensity of HFn-PAS-cy5 failed to be captured by the camera under the same exposure time. In HFn-GFLG/PLGLAG-PAS-RGDK-cy5 groups, fluorescence signal was captured from 4 h to 52 h after injection (Figure 6A). The armpit fluorescence areas at 6.5, 12 and 24 h were larger than the area of armpit lymph node, proving the protein accumulation in tumor tissues. However, it is uncertain that if lymph node accumulation co-existed or not. Figure 6A shows that the tumor area of HFn-GFLG/PLGLAG-PAS-RGDK-cy5 groups had stronger signals than liver at all time points. At 52 h after injection, whilst HFn-cy5 and free cy5 were almost completely eliminated, HFn-GFLG/PLGLAG-PAS-RGDK-cy5 were still detectable in region of tumor site, implying functionalized HFns were retained in tumor by longer and stronger accumulation.

As is presented in Figure 6B, free cy5 and HFn-cy5 had the lowest tumor florescence intensity at all time points. At 2 h, free cy5 had a greater intensity than HFn-cy5 but was surpassed by HFn-cy5 afterwards. Free cy5 tumor area fluorescence intensity peaked at 4 h and decreased rapidly after that, suggesting a fast clearance. HFn-cy5 achieved the highest concentration in tumor at around 4 h after injection (Figure 6B). The difference in free cy5 and HFn-cy5 is likely to be due to a quicker distribution and a shorter half-life of small molecule cy5 than HFn-cy5. HFn-PAS-cy5 demonstrated significantly stronger and longer lasting tumor intensities than HFn-cy5 at all detected time points (*p* < 0.001). As proven in the pharmacokinetic study, the insertion of the PAS peptide could lead to a longer half-life in circulation and probably result in the slower clearance of HFn-PAS-cy5 than HFn-cy5. The best drug targeting delivery results were from HFn-GFLG/PLGLAG-PAS-RGDK-cy5. They had significantly greater signal intensities in tumor area at all times than all the other groups (*p* < 0.001). This shows that the RGDK peptide can significantly improve HFn biodistribution. A previous study of RGDK fused Albumin binding domain has also proven the tumor targeting ability improvement of RGDK peptide *in vivo*. [24] Overall, both PAS and RGDK functionalization, and particularly RGDK functionalization, improved the tumor biodistribution of HFn.

**Figure 6.** Biodistribution of cy5 and cy5 conjugated with HFn-based proteins. (**A**), in vivo fluorescence imaging of tumor-bearing mice at different time points, from left to right: HFn-PAS-cy5, HFn-GFLG-PAS-RGDK-cy5, HFn-PLGLAG-PAS-RGDK-cy5, HFn-cy5 and free cy5. (**B**), the sum fluorescent intensity of region of interest (ROI, tumor area) at each time point.

#### *3.7. Functionalization Effect on Protein/DOX Anti-Tumor Efficacy*

To compare tumor treatment efficacies of all protein/DOX and free DOX, 4T1 tumor bearing BALB/c mice model was built and 36 mice with around 250 mm<sup>3</sup> tumor were randomly assigned into six groups. Intravenous injections of four HFn-based protein/DOX, free DOX and PBS were conducted at day 0 and day 5. As is shown in Figure 7A, the fastest mice tumor growth rate was observed in PBS control group rats which underwent no drug treatment. The average tumor volume reached 2030 mm3 after 17 days. The second fastest tumor growth rate was in free DOX group mice and their average group tumor volume were 1667 mm<sup>3</sup> at day 17. HFn/DOX showed a better tumor growth inhibition and at day 17, tumor volume grew to 1521 mm3. In HFn-PAS/DOX group, mice tumor volume reached 1432 mm3 in the end. Two PAS-RGDK functionalized HFn/DOX treated group had the strongest tumor-growth inhibition. In spite of just two administrations, average tumor volume of these two group mice at day 17 were just around 1100 mm3, close to half of the volume of the PBS group tumor.

**Figure 7.** In vivo tumor inhibitory effects on 4T1 tumor-bearing mice. (**A**), tumor volume change over time. (**B**), group tumor weight/ PBS group tumor weight (%) on day 17. \*\*\* *p* < 0.001. Symbol '\*' represents the significance of *p* value between groups. (**C**), the photo of excised tumor tissues on day 17. Arrows indicated the injection days; data are mean ± RSD (*n* = 6).

On day 17, the corresponding average tumor weights were measured and the percentages of tumor weights compared with tumor weights of control PBS group mice are presented in Figure 7B. The photo of excised tumor tissues is shown as Figure 7C. Tumor weights on day 17 of HFn-GFLG-PAS-RGDK/DOX and HFn-PLGLAG-PAS-RGDK group mice were 54.2 ± 9.7% and 54.0 ± 10.8% of tumor weights of the control PBS group. HFn-PAS/DOX group mice had 69.5 ± 9.4% and HFn/DOX group mice had 72.99 ± 6.2% of PBS control group mice tumor weights. Free DOX group mice tumor weight was 82.9 ± 8.7% of PBS control group mice. T-test results demonstrate that there were significant statistical differences between final tumor masses of protein/DOX and free DOX group. Both HFn and functionalized HFns had significantly increased DOX anti-tumor efficacy (*p* < 0.001). Compared with HFn/DOX group, HFn-PAS/DOX did not show statistical distinction (*p* = 0.468), showing that PAS functionalization alone was not enough to significantly improve anti-tumor efficacy. Masses of tumors from two PAS-RGDK protein/DOX groups, however, were significantly lower than those of both HFn/DOX group and HFn-PAS/DOX group. This indicates RGDK functionalization primarily accounts for the significant improvement of growth inhibition efficacy of 4T1 tumor. Difference between HFn-GFLG-PAS-RGDK/DOX and HFn-PLGLAG-PAS-RGDK/DOX (*p* = 0.977) was not significant in statistical analysis. Two different enzyme-cleavable sites did not make a statistical difference in anti-tumor efficacy.

#### **4. Discussion**

Based on all of the results above, PAS and RGDK functionalization have both improved HFn anti-tumor performance. PAS functionalization impacts HFn mainly by extension of half-life in circulation. In pharmacokinetic study, differences in half-lives in circulation of all protein/DOX mostly stemmed from the insertion of PAS peptide. In biodistribution assay, PAS provided HFn-PAS with a longer retaining time than HFn in tumor area. Constituted by repetitive P, A and S residues, PAS peptide is hydrophilic and uncharged in neutral solutions and plasma. Circular dichroism shows it is a flexible random coil [25]. Its properties are similar to polyethylene glycol (PEG), but it is advantageous in terms of biodegradability and biocompatibility. When it is attached to another molecule, it can

attract water molecules to increase molecule hydrodynamic volume, thereby extending half-life in circulation. PAS peptide usually has to be over 200 residues to achieve half-life extension, but because of the repeated and organized presentation manner on ferritin shell, 40 aa and 75 aa PAS peptides have been proven to be long enough when fused onto N-terminal of ferritin subunit [11]. In this study, the enlargement of HFn hydrodynamic volume after PAS insertion was detected in HPSEC-MALLS characterization.

RGDK peptide significantly improved anti-tumor performance of HFn/DOX. In cellular uptake assay, it has increased cellular internalization efficiency through binding to specific receptors. It has also led to the best tumor targeting abilities in biodistribution assay and the greatest anti-tumor efficacy in cytotoxicity and in vivo anti-tumor assay.

Comparing HFn-GFLG-PAS-RGDK and HFn-PLGLAG-PAS-RGDK, the sequence length difference caused the minor differences in hydrodynamic volume in HPSEC-MALLS and half-life in circulation in pharmacokinetic study. However, the two enzyme-cleavable sites, GFLG and PLGLAG, did not make a significant difference in cellular uptake efficiency, and in any other in vitro and in vivo tests. This suggest the PLGLAG enzyme-cleavable site probably was not digested by MMP-2/9 before cell internalization. It could be caused by the insufficient activity of MMP-2/9 in vitro and in vivo and/or the low accessibility of PLGLAG to enzymes. A further detailed investigation of the cleavage of PLGLAG is needed.

Figure 8 illustrates the assumed tumor cell internalization pathways of all groups. Free DOX enters tumor cells via unspecific passive diffusion due to its small size (Figure 8A). The short half-life in circulation and the lack of tumor targeting ability caused a great loss of DOX before it reached tumor cells. As a result, the in vivo anti-tumor efficacy was the lowest. Both HFn/DOX and HFn-PAS/DOX enter cells through non-specific pinocytosis [26] (Figure 8B), because there is no corresponding receptor, human TfR1, on 4T1 cells. Therefore, in in vitro assessments, cellular uptake assay and cytotoxicity assay, there were no statistical differences between HFn/DOX and HFn-PAS/DOX.

**Figure 8.** Schematic of different tumor cellular internalization mechanisms of DOX and protein/DOX. (**A**), free DOX passive diffusion pathway. (**B**), HFn/DOX and HFn-PAS/DOX pinocytosis internalization. (**C**), two possible receptor-mediated internalization pathways of HFn-GFLG-PAS-RGDK/DOX and HFn-PLGLAG-PAS-RGDK. (Created with Biorender).

HFn-GFLG-PAS-RGDK/DOX and HFn-PLGLAG-PAS-RGDK/DOX have extra drug internalization mechanisms compared with HFn/DOX and HFn-PAS/DOX (Figure 8C). Since RGDK has RGD motif and an exposed free C-terminal K residue, it can be directly recognized by both integrin αvβ3/5 and neuropilin-1 (NRP1), two kinds of receptors overexpressed on 4T1 cells [14]. The overexpression of these two receptors has led to the increase of internalization efficiency of HFn-GFLG/PLGLAG-PAS-RGDK/DOX. As shown in Figure 8C, when HFn-GFLG/PLGLAG-PAS-RGDK/DOX reaches tumor tissue, there are two possible internalization pathways. RGDK can either directly bind to NRP1 or firstly interact with integrin αvβ3/5 and then be transferred to NRP1, followed by endocytosis. After that, some of the HFn-GFLG/PLGLAG-PAS-RGDK/DOX inside tumor cells would be digested in lysosome while some will travel to other cells nearby via paracellular pathway or transcytosis [27].

#### **5. Conclusions**

All three functionalized HFns expressed in *E. coli* have self-assembled into nanoparticles such as HFn. RGDK peptide has enhanced HFn tumor cell uptake efficiency and improved biodistribution, resulting in a significant improvement in anti-tumor treatment outcome. PAS has expanded HFn hydrodynamic volume and helped ferritin stay longer in circulation, which also has improved anti-tumor efficacy of ferritin. In summary, we successfully prepared and evaluated three new functionalized HFn constructs (HFn-PAS, HFn-GFLG-PAS-RGDK, HFn-PLGLAG-PAS-RGDK), especially two PAS-RGDK fused ones, which hold greater potentials as anti-tumor drug delivery nanoparticles than HFn.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/10.3390/ pharmaceutics13040521/s1, Figure S1: standard curve of fluorescence intensity-doxorubicin concentration in SD rat plasma.

**Author Contributions:** Conceptualization, J.B., Y.W. (Yinli Wang) and Y.L.; methodology, S.Y., Y.W. (Yan Wang) and B.Z.; software, S.Y.; validation, S.Y., Y.Q. and Y.Z.; formal analysis, S.Y.; investigation, S.Y.; resources, Y.L., J.B., Y.W. (Yan Wang) and Y.W. (Yinli Wang); writing—original draft preparation, S.Y.; writing—review and editing, J.B.; supervision, Y.W. (Yinli Wang), S.D., Y.L. and J.B.; project administration, J.B. and Y.W. (Yinli Wang); funding acquisition, Y.L., J.B. and Y.W. (Yinli Wang). All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was funded by joint PhD Scholarship Scheme of the University of Adelaide and Institute of Process Engineering, Chinese Academy of Sciences, the National Natural Science Foundation of China [Grant No. 21576267], Beijing Natural Science Foundation [Grant Number 2162041], and Shanxi Education Science "1331 project" special research project (Research and Development of Traditional Chinese Medicine Micro-emulsion and New Biological Preparation).

**Institutional Review Board Statement:** The animal study in this work was conducted with the approval of the medical ethics committee of Shanxi University of Chinese Medicine (Approval Number 2019LL137).

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data presented in this study are available on request from the corresponding author.

**Acknowledgments:** Thanks to Iain Comerford from the University of Adelaide for his help with flow cytometry. Great appreciations to Anton Middelberg from the University of Adelaide for his helpful advice and support on this work.

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

#### **References**


## *Article* **Microfluidic Synthesis and Purification of Magnetoliposomes for Potential Applications in the Gastrointestinal Delivery of Difficult-to-Transport Drugs**

**Carlos E. Torres 1,†, Javier Cifuentes 1,†, Saúl C. Gómez 1, Valentina Quezada 1, Kevin A. Giraldo 1, Paola Ruiz Puentes 1, Laura Rueda-Gensini 1, Julian A. Serna 1, Carolina Muñoz-Camargo 1, Luis H. Reyes 2,\*, Johann F. Osma 3,\* and Juan C. Cruz 1,\***


**Abstract:** Magnetite nanoparticles (MNPs) have gained significant attention in several applications for drug delivery. However, there are some issues related to cell penetration, especially in the transport of cargoes that show limited membrane passing. A widely studied strategy to overcome this problem is the encapsulation of the MNPs into liposomes to form magnetoliposomes (MLPs), which are capable of fusing with membranes to achieve high delivery rates. This study presents a low-cost microfluidic approach for the synthesis and purification of MLPs and their biocompatibility and functional testing via hemolysis, platelet aggregation, cytocompatibility, internalization, and endosomal escape assays to determine their potential application in gastrointestinal delivery. The results show MLPs with average hydrodynamic diameters ranging from 137 ± 17 nm to 787 ± 45 nm with acceptable polydispersity index (PDI) values (below 0.5). In addition, we achieved encapsulation efficiencies between 20% and 90% by varying the total flow rates (TFRs), flow rate ratios (FRRs), and MNPs concentration. Moreover, remarkable biocompatibility was attained with the obtained MLPs in terms of hemocompatibility (hemolysis below 1%), platelet aggregation (less than 10% with respect to PBS 1×), and cytocompatibility (cell viability higher than 80% in AGS and Vero cells at concentrations below 0.1 mg/mL). Additionally, promising delivery results were obtained, as evidenced by high internalization, low endosomal entrapment (AGS cells: PCC of 0.28 and covered area of 60% at 0.5 h and PCC of 0.34 and covered area of 99% at 4 h), and negligible nuclear damage and DNA condensation. These results confirm that the developed microfluidic devices allow highthroughput production of MLPs for potential encapsulation and efficient delivery of nanostructured cell-penetrating agents. Nevertheless, further in vitro analysis must be carried out to evaluate the prevalent intracellular trafficking routes as well as to gain a detailed understanding of the existing interactions between nanovehicles and cells.

**Keywords:** magnetoliposomes; microfluidics; oral drug delivery; magnetite nanoparticles

#### **1. Introduction**

Oral drug administration is one of the most convenient routes of drug delivery due to patient preference, shelf life, sustained delivery, cost-effectiveness, and ease of large-scale

**Citation:** Torres, C.E.; Cifuentes, J.; Gómez, S.C.; Quezada, V.; Giraldo, K.A.; Puentes, P.R.; Rueda-Gensini, L.; Serna, J.A.; Muñoz-Camargo, C.; Reyes, L.H.; et al. Microfluidic Synthesis and Purification of Magnetoliposomes for Potential Applications in the Gastrointestinal Delivery of Difficult-to-Transport Drugs. *Pharmaceutics* **2022**, *14*, 315. https://doi.org/10.3390/ pharmaceutics14020315

Academic Editors: Francisco José Ostos, José Antonio Lebrón and Pilar López-Cornejo

Received: 29 November 2021 Accepted: 12 January 2022 Published: 28 January 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 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 (https:// creativecommons.org/licenses/by/ 4.0/).

manufacture [1,2]. Additionally, orally administered drugs can be directed through the gastrointestinal tract to allow localized treatment of different pathologies, such as cancer, infections, inflammations, and various digestive system diseases [1]. Nevertheless, the success of this delivery route depends on the physicochemical properties of such drugs and, particularly, their water solubility and cell-membrane permeability [3]. Different approaches have been proposed to control pharmacokinetics and improve release efficacy and safety [4]. Among these, encapsulation is one of the most attractive ones for preserving compounds with biological activity, especially when exposed to conditions that might be detrimental to their chemical stability [5,6]. In the pharmaceutical industry, the delivery of drugs has been significantly improved by encapsulation into polymeric capsules and liposomes [7,8]. These liposomal vehicles have been widely studied for pharmaceutical preparations with limited passing across biological barriers, such as the blood–brain barrier and the intestinal epithelium, due to attractive features such as the flexibility of changing their chemical composition, structure, and colloidal size [9–13].

Additionally, a significant challenge in drug delivery has been to achieve better internalization and high bioavailability [9,10]. This challenge has been addressed by an increasing number of delivery vehicles that include both viral and non-viral vectors [11]. Among these, magnetite nanoparticles (MNPs) functionalized with translocating proteins and peptides have been studied as potent vehicles for cell penetration and endosomal escape. Moreover, a possible enhancement of escape is expected if these vehicles are encapsulated into liposomes [12,14–16].

Liposomes with encapsulated MNPs, called magnetoliposomes (MLPs), have been extensively used as carriers in the pharmaceutical industry due to their ability to release various active molecules at a given site without the need for molecularly targeted agents [17]. This is in addition to the improvement in the biocompability, drug delivery rate for some compounds, and cellular uptake without a significant reduction in the activity of the functional compounds immobilized and delivered employing MNPs [18,19]. Moreover, these novel drug delivery vehicles might offer potential improvements in targeting, stabilization of antimicrobial agents, and gastroretention. This might help to reduce various possible side effects of oral administration, such as the uncontrolled destruction of both pathogenic and non-pathogenic microbiota, and, therefore, prevent the appearance of complications such as dysbiosis [20]. This correlation between the microbiota's metabolic activity and the improvement of the bioavailability is particularly relevant in determining the overall efficacy of these novel drug delivery vehicles. Therefore, it is important to determine if the carrier nanovehicle can increase the compound bioavailability and reduce the associated biotransformations, which in turn define the bioactivity expression in response to a particular microbiome [21–23].

Currently, MLPs have been explored as drug delivery carriers to treat conditions as diverse as cancer, Parkinson's, and Alzheimer's [24–26]. Over the past few years, several techniques have been proposed for preparing MLPs, in which nanoparticles can be encapsulated in the aqueous lumen, embedded in the lipid bilayer, or conjugated on the surface of the liposome [27–30]. By implementing these techniques, liposome solutions' parameters vary considerably, posing some challenges related to their particle size, dispersity, lamellarity, entrapment efficiency, and, most importantly, the difficulty in separating the non-encapsulated/unbound MNPs [31,32].

Currently, liposome preparation techniques using microfluidic systems have allowed greater control over physical properties, yielding massive and robust production of MLPs with uniform size distribution, high loading efficiencies, and reduced costs [33–35]. Nevertheless, there is still a challenge in the separation and sample purification methods due to the minor size differences between the MLPs and the non-encapsulated nanoparticles [32,36]. Due to several limitations of current separation techniques, microfluidic systems have been proposed as potential low-cost particle separation systems based on different active or passive methods to separate nanoscale objects such as DNA, viruses, proteins, exosomes, and nanoparticles [37]. Exploring the scope of microfluidics separation approaches, there has been a growing interest in using magnetic gradients to retain excess MNPs without compromising the integrity of the MLPs. The significant traction gained by this approach could be mainly attributed to its applicability, versatility, and ease of implementation in many areas of the biomedical field, including disease diagnostics, therapeutics, and cell sorting [38–40].

This study proposes the synthesis of MLPs using a microfluidic approach; FEM simulations implemented in COMSOL Multiphysics® to study the separation of MLPs from nanoconjugates aided by a magnetophoretic microfluidic system; the manufacture and experimental validation of two separation devices; and, finally, the in vitro testing of the synthesized MLPs to evaluate whether this delivery vehicle is biocompatible and improves the cell penetration of orally administered CefTRIAxone, a drug with exceedingly low intestinal absorption. The prototypes of the microfluidic devices for MLPs synthesis were manufactured by a low-cost method based on laser cutting techniques and led to MLPs with acceptable physical properties and encapsulation efficiency. Additionally, the separation devices showed different efficiencies depending on the implemented approach which varied considerably compared with those obtained during the experimental validation. Nevertheless, qualitatively, both methods led to similar results, confirming their suitability for the intended objective. Finally, the preliminary in vitro evaluation demonstrated that MLPs showed high biocompatibility, low endosomal entrapment, and high internalization rates, which are crucial factors in developing novel vehicles for delivering difficult-totransport drugs.

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

#### *2.1. Magnetite Nanoparticles Synthesis and Functionalization*

Magnetite nanoparticles (MNPs) were synthesized by the chemical co-precipitation method. For this, FeCl2 (0.34 g, J. T. Baker, Phillipsburg, NJ, USA) and FeCl3 (0.93 g, Merck, Kenilworth, NJ, USA) were solubilized in 60 mL of type I water. In addition, 0.69 g of NaOH (PanReac AppliChem, Darmstadt, Germany) was added to 17 mL of type I water, and then both solutions were heated at 80 ◦C. Next, a NaOH solution was added dropwise to the iron chloride solution at a 5 mL/min rate under constant stirring. A black precipitate was observed, corresponding to the formation of MNPs. The obtained MNPs were then washed four times with NaCl solution (1.5% *w*/*v*) and twice with type I water aided by a neodymium magnet. Then, 100 mg of MNPs were silanized by adding 50 μL of glacial acetic acid (PanReac AppliChem, Barcelona, Spain) followed by 400 μL of (3-aminopropyl) triethoxysilane (APTES, 98%, Sigma-Aldrich, St. Louis, MO, USA). The MNPs solution was left to react under constant stirring (250 rpm) at 60 ◦C for 1 h and then washed as mentioned previously. Later, 100 mg of the silanized MNPs (MNP-APTES) was mixed with 2 mL of glutaraldehyde solution (2% *v*/*v*, Sigma-Aldrich, St. Louis, MO, USA) (solution previously stirred (220 rpm) at room temperature for 1 h to allow glutaraldehyde activation (MNP-APTES-GA)). Then, 5 mL of a NH2-PEG-propionic acid (99%, Merck, Darmstadt, Germany) solution (2 mg/mL) was added dropwise to the MNP-APTES-GA conjugates under constant stirring to obtain MNP-APTES-PEG conjugates. The solution was left to react at 220 rpm and room temperature for 24 h and washed four times with NaCl solution (1.5% *w*/*v*) and twice with type I water. Similarly, (3-[(2-aminoethyl)dithio]propionic acid) (AEDP, ThermoFisher, Waltham, MA, USA) immobilization was carried out using 14 mg of *N*-[3-dimethylammino)-propyl]-*N* -ethyl carbodiimide hydrochloride (EDC, 98%, Sigma-Aldrich, St. Louis, MO, USA) and 7 mg of *N*-hydroxy succinimide (NHS, 98%, Sigma-Aldrich, St. Louis, MO, USA) solution in 5 mL of type I water added to 100 mg of MNP-PEG in 50 mL of type I water to activate the terminal carboxyl groups. Nanoparticles were ultrasonicated (ultrasonic bath, Branson, Danbury, CT, USA) for 10 min, and 5 mL of an AEDP solution (5 mg/mL) was added dropwise under constant stirring. The solution was left to react at 220 rpm and room temperature for 24 h. The MNP-PEG-AEDP conjugates were washed with NaCl (1.5% *w*/*v*) and type I water. Finally, the antibiotic CefTRIAxone (CTA, Vitalis, 1 g I.M/I.V) was immobilized by following the same protocol for AEDP

immobilization, using 5 mL of CefTRIAxone solution (2 mg/mL). Scheme 1 shows the complete methodology for the synthesis.

**Scheme 1.** Schematic of the developed workflow for the synthesis of MNP-PEG-AEDP-CTA nanoconjugate. (1) Magnetite nanoparticles (MNPs) are synthesized by co-precipitation. (2) MNPs are silanized with APTES and subsequently with (3) NH2-PEG-propionic acid. This was followed by the conjugation of (4) AEDP and, finally, (5) the immobilization of the drug CTA.

The resulting MNP-PEG-AEDP-CTA conjugates were labeled with rhodamine B (95%, Sigma-Aldrich, St. Louis, MO, USA) for fluorescence-based assays. For this, 14 mg of EDC, 7 mg of NHS, and 5 mg of rhodamine B (RdB) were dissolved in 5 mL of type I water containing 2 mL of dimethylformamide (DMF, Supelco/Sigma-Aldrich, Bellefonte, PA, USA). Rhodamine B solution was left under constant stirring for 15 min to allow the activation of carboxylic groups. Next, the previously activated rhodamine B solution was added to 50 mL of MNP-PEG-AEDP-CTA aqueous solution (2 mg/mL) and left to react at 220 rpm, room temperature, and in complete darkness for 24 h. The resulting MNP-PEG-AEDP-CTA-RdB was washed several times with NaCl (1.5% *w*/*v*) and type I water to remove the excess reagents. MNP-PEG-AEDP-CTA-RdB nanoconjugates were resuspended in type I water and stored in complete darkness at 4 ◦C until further use for

the MLPs preparation described below. A schematic of the synthesized nanoconjugate is shown in Figure 1A.

**Figure 1.** (**A**) Schematic of the synthesized nanoconjugate for the MLPs production. (**B**) Microfluidic experimental setup for the synthesis phase. Microfluidic separation designs proposed: (**C**) System 1 and (**D**) System 2. (**E**) Microfluidic experimental setup for the separation phase.

#### *2.2. Magnetoliposomes Synthesis Using the Microfluidic Approach*

#### 2.2.1. Lipidic-MNPs Phase Preparation

First, 100 mg of soy lecithin (1-α-lecithin, soybean-cas 8002-43-5-calbiochem) (Merck, Kenilworth, NJ, USA) was dissolved in 10 mL of chloroform c2432 (>99.5%, Merck, Kenilworth, NJ, USA), and 1 mL and 2 mL of MNPs (1.7 mg/mL) were added. The sample was rotary evaporated for 1 h at 45 ◦C under a vacuum (rotary evaporator, Hei-VAP Value Digital Vertical, Heidolph, Schwabach, Germany). Then, 10 mL of ethanol (96% *v*/*v*) was added to the rotary evaporator flask. The sample was vigorously agitated and rotated for 30 min at atmospheric temperature and pressure.

#### 2.2.2. Microfluidic System Manufacture and Experimental Setup

The manufacture and design of microfluidic devices for MLP synthesis was based on the study presented by Aranguren et al. [41]. A laser cutting machine (TROTEC ® Speedy 100, 60 w laser cutter, TROTEC, Marchtrenk, Austria) was used for engraving and cutting the microfluidic channels proposed on a PMMA substrate. The two or three layers of the microfluidic devices were manually aligned and sealed using 96% (*v*/*v*) ethanol with a mechanical press placed on a hot plate (110 ◦C). For the synthesis, the microfluidic channels were purged with a 10 mL syringe filled with 96% (*v*/*v*) ethanol for 15 min. Then, a syringe filled with the lipidic-nanoconjugates phase and a syringe with NaCl (anhydrous, Redi-Dri™, free-flowing, ACS reagent, >99%) solution (0.05 M) were mounted on an infusion pump (MedCaptain MP30), as is shown in Figure 1B. The syringes were connected to the microsystems using two probes (Nelaton, Probes, Medex caliber 8) (Medex, Smiths Medical

Inc., Minneapolis, MN, USA). The synthesis was carried out using a total flow ratio (TFR) set at 2.5 mL/min and 5 mL/min with a varying flow rate ratio (FRR) from 1:1 to 5:1 (aqueous:solvent ratio).

#### *2.3. Magnetoliposomes Characterization*

Magnetoliposome size and polydispersity index were measured using the Zetasizer Nano ZS (Malvern, Panalytical, Egham, UK). Additionally, a morphology and size characterization using a TEM Tecnai F20 Super Twin TMP (FEI, Hillsboro, OR, USA) was performed to determine the synthesis effectiveness. For the TEM analysis, a sample drop was deposited on a copper grid with a carbon coating that was dried for 1 h. Next, the prepared sample was stained with 2% uranyl acetate by depositing one drop on the grid for 8 min, washed with deionized water, and left to dry for imaging at a total magnification of 71, 97, and 145 kX.

#### *2.4. Magnetolipsomes Encapsulation Efficiency (EE%)*

The encapsulation efficiency of synthesized MLPs was analyzed using a Spectrofluorometer (0239D-2219 FluoroMax plus C, Horiba, Miyanohigashi, Japan) to track changes in the fluorescence intensity before and after treatment with Triton X-100 (Sigma-Aldrich, St. Louis, MO, USA). For this characterization, 100 μL of MLPs prepared with MNP-PEG-AEDP-CTA-RdB was pipetted into a 96-well microplate for the first fluorescence-based analysis where the fluorescence intensity was measured. Then, 10 mL of Triton X-100 was added to the microplates to break the magnetoliposome membranes and allow the labeled nanoconjugates to escape. Finally, a second measure of the sample fluorescence intensity was carried out to analyze the changes compared with the initial intensity. The fluorescence spectrum of rhodamine B allowed tracking intensity by setting up excitation and emission filters at 546 nm and 568 nm, respectively. The encapsulation efficiency was calculated using Equation (1):

$$\text{EE}\left(\%\right) = 100 \times \frac{\left(\text{Int\\_Final}\right) - \text{Int\\_Int\\_Inital}\left(\text{Initial}\right) - \text{Int\\_Triend X-100}\right), \tag{1}$$

where *Int (Final)* is the emission post-Triton *X-100, Int (Initial)* is the emission pre—*Triton X-100* treatment, and *Int (Triton X-100)* is the blank emission of *Triton X-100*.

#### *2.5. Magnetoliposomes Purification*

#### 2.5.1. Lipidic-Nanoconjugates Phase Preparation

The design of the two components of the microfluidic devices proposed (System 1 and System 2) was conducted on two rectangular PMMA layers (width: 7.48 cm, height: 2.60 cm) via AutoCAD v23.0 (AutoDesk Inc., San Rafael, CA, USA). The microfluidic channel was superimposed over one of the two layers, followed by locating holes in each piece. The larger holes were to accommodate permanent neodymium magnets 6 mm in diameter and 8 mm deep, while the smaller ones were for the inlets and outlets of the microchannel. In this case, the diameter was 2.4 mm. (Figure 1C,D).

In these designs, the magnetophoretic separation was analyzed in silico with two different methods. The first one involved implementing the particle tracing module, where magnetoliposomes and nanoconjugates were considered separate components within the microfluidic device. The simulation was conducted by implementing the laminar flow, particle tracing, and the magnetic field without current modules of COMSOL Multiphysics® (COMSOL Inc., Stockholm, Sweden). The second method was based on a mixture model approach to simulate a dispersed phase, considered a ferrofluid (i.e., a suspension of magnetic nanoconjugates in water) due to the high concentration of nanoparticles in the domain. Several recent reports support this assumption (rather than considering individual particles) for FEM simulations, since it leads to results that are closer to those obtained experimentally [42–46]. In this case, the implemented COMSOL modules were the "mixture model with laminar flow", the "magnetic field without current", and "diluted species' transport". 2.5.2. Multiphysics Simulations of Magnetophoretic Separation via the Particle Tracing Module

The magnetic field was incorporated into the simulations through the "magnetic fields with no currents" model of the AC/DC module of COMSOL. In this case, the magnetic field intensity is calculated by solving the Maxwell equations for permanent magnets, resulting in the governing equations shown in Equations (2)–(4) [47].

$$\mathbf{H} = -\nabla \text{ V}\_{\text{m}\_{\text{V}}} \tag{2}$$

$$
\nabla \cdot \mathbf{B} = 0,
$$

where:

$$\mathbf{B} = \mu\_0 \mu\_\mathbf{r} \overrightarrow{\mathbf{H}}^\prime + \mathbf{B}\_\mathbf{r} \tag{4}$$

where <sup>→</sup> H is the magnetic field distribution, Vm is the magnetic scalar potential, B is the magnetic flux density distribution, and μ<sup>0</sup> and μ<sup>r</sup> the vacuum permeability and relative permeability. Finally, Br represents the remanent flux density that, in this case, was set to 1 T. The second physic coupled to the model was the laminar flow governed by the Navier– Stokes conservation of momentum equation for incompressible fluids, Equation (5), which is accompanied by the conservation of mass by the continuity equation (Equation (6)).

$$\left[\nabla \mathbf{I} - \mathbf{P} \mathbf{I} + \mu \left(\nabla \mathbf{u} + (\nabla \mathbf{u})^{\mathrm{T}}\right)\right] + \mathbf{F} = \mathbf{0},\tag{5}$$

$$
\rho \nabla \cdot (\mathbf{u}) = 0,\tag{6}
$$

where P is the pressure, μ is the fluid's dynamic viscosity, F accounts for the volumetric forces, and ρ is the fluid density. Finally, the particle tracing for fluid flow was coupled to the model, where Newton's second law governs the movement of the particles transported within the device. This is described by Equation (7):

$$\frac{\mathrm{d}(\mathrm{m}\_{\mathrm{p}}\nu)}{\mathrm{d}\mathrm{t}} = \mathrm{F}\_{\mathrm{t}\nu} \tag{7}$$

where mp is the mass of the particles, υ the velocity, and Ft the sum of all forces acting on the particle. In this case, the involved forces were the drag force, defined by Equations (8) and (9) by the Stokes law, and the magnetic force, dependent on the magnetic flux density distribution described by Equation (10).

$$\mathbf{F}\_{\rm D} = \frac{1}{\tau\_{\rm p}} \mathbf{m}\_{\rm P} (\mathbf{u} - \boldsymbol{\nu}),\tag{8}$$

$$\pi\_{\mathbb{P}} = \frac{\rho\_p \, \mathrm{d}\_{\mathbb{P}}^2}{18\mu} \,\mathrm{}^{\prime}\tag{9}$$

where mp is the particle mass, u is the velocity field, υ the particle velocity, ρ<sup>p</sup> the particle density, dp the particle diameter, and μ the dynamic viscosity.

$$\mathbf{F\_M} = \frac{\mathbf{V\_m} \Delta \mathbf{X}}{\mu\_0} (\mathbf{B} \cdot \nabla) \mathbf{B} \tag{10}$$

where ΔX is the magnetic susceptibility difference between the particle and the fluid. Finally, the FEM simulations to solve the set of equations for laminar flow and magnetic field were conducted via a stationary study. Additionally, for the particle tracing module for 1200 particles per component (i.e., nanoconjugates and MLPs), a bi-directionally coupled particle tracing was used with a MUMPS solver. The computational domain was meshed with 176,141 domain elements and 6523 boundary elements for System 1 and 47,152 domain elements and 1838 boundary elements for System 2. This module's boundary conditions were the drag force in all the microfluidic channel domains and the system's inlets as the


main entrance for the particles into the system. The model parameters are summarized in

**Table 1.** Parameters employed for the particle tracing module.

Table 1:

2.5.3. Multiphysics Simulations of Magnetophoretic Separation via the Mixture Model

The magnetic field was established using a magnetic field, no currents physic, as described previously. The governing equations for this simulation are presented in Equations (1)–(3). The ferrofluid was simulated, aided by the mixture model, laminar flow physics. The interface solves a set of Navier–Stokes equations for the momentum of the mixture. The pressure distribution is calculated from a mixture-averaged continuity equation, and the velocity of the dispersed phase is described by a slip model [48]. The momentum conservation equation and the continuity equation are presented in Equations (11) and (12):

$$\rho \frac{\text{d}\mathbf{u}}{\text{d}\mathbf{t}} + \rho(\mathbf{u} \cdot \nabla)\mathbf{u} = \left[-\text{pl} + \mu(\nabla \mathbf{u} + (\nabla \mathbf{u})^{\mathsf{T}} - \frac{2}{3}(\nabla \cdot \mathbf{u})\right] - \nabla \cdot \left[\rho \mathbf{C}\_{\mathsf{d}}(1 - \mathbf{C}\_{\mathsf{d}})\mathbf{U}\_{\text{slip}}\mathbf{U}\_{\text{slip}}{}^{\mathsf{T}}\right] + \mathbf{F},\tag{11}$$

$$\left(\left(\rho\_{\rm c} - \rho\_{\rm d}\right) \left\{\nabla \cdot \left[\Phi\_{\rm d} \left(1 - \mathbf{C}\_{\rm d}\right) \mathbf{U}\_{\rm slip}\right] + \frac{\mathbf{m}\_{\rm dc}}{\rho\_{\rm d}}\right\} + \rho\_{\rm c} \left(\nabla \cdot \mathbf{u}\right) = 0,\tag{12}$$

where P is the pressure, μ is the dynamic viscosity of the fluid, ρ<sup>c</sup> and ρ<sup>d</sup> the continuous phase density and dispersed phase density, Φ<sup>d</sup> the volume fraction of the dispersed phase, mdc the turbulent dispersed phase diffusion, and F the body forces, which in this case are described by the Kelvin body force due to a spatially non-uniform magnetic field according to Equation (13) [46]:

$$\mathbf{F} = (\overrightarrow{\mathbf{M}} \cdot \nabla)\overrightarrow{\mathbf{B}}\_{\prime} \tag{13}$$

where <sup>→</sup> <sup>B</sup> is the magnetic flux density distribution and <sup>→</sup> M the magnetization. Finally, the diluted species' transport was used to determine the effective concentration of the nanoparticles inside the channel. This solution inside the microchannel is described by the convective–diffusive Equation (14).

$$\frac{d\mathbf{C\_P}}{dt} + \nabla \cdot (-\mathbf{D\_P}\mathbf{C\_P}) + \mathbf{u} \cdot \nabla \, \mathbf{C\_P} = 0,\tag{14}$$

where Cp is the concentration, u is the velocity field provided by the mixture model, and Dp is the effective diffusivity of the NPs as calculated by Equation (15).

$$\mathbf{D}\_{\rm P} = \frac{\mathbf{K}\_{\rm B} \mathbf{T}}{3 \Pi \boldsymbol{\eta}\_{\rm ff} \mathbf{d}\_{\rm P}},\tag{15}$$

Here, KB is the Boltzmann constant, T is temperature, ηff is the ferrofluid viscosity, and dp is the diameter of the particles.

Finally, time-dependent simulations were carried out using a MUMPS solver with a phase volume fraction of 0.2 for each particle component entering the upper inlet. For System 1, complete mesh consists of 59,862 domain elements and 1845 boundary elements, while for System 2, the computational domain mesh comprised 66,735 domain elements

and 1946 boundary elements. The final meshing is shown in Supplementary Figure S2, and the parameters used for this model are presented in Table 2.


**Table 2.** Parameters employed for the mixture model.

2.5.4. Microfluidic System Manufacture and Experimental Setup

The manufacture of the microfluidic device for the MLPs synthesis (Supplementary Figure S1) was based on the methodology presented above and reported previously by us in the studies of Aranguren et al. and Campaña et al. [41,49]. For the first design, seven neodymium magnets were in proximity to the microchannels, as is shown in Figure 1C. In parallel, one syringe of 10 mL was filled with the solution of LPs and MNP-PEG-AEDP-CTA nanoconjugates and connected to the system inlet. The solution was pumped into the device with syringe pumps (78-8110C Programmable Touch Screen, Cole-Parmer®, Vernon Hills, IL, USA, and B Braun Perfusor® compact, B. Braun, Melsungen, Germany) at total flow rates (TFRs) from 1 to 3 mL/min. The second design included six neodymium magnets, as is shown in Figure 1D. Two syringes of 10 mL were filled up and connected to the system's inlets. The first one was filled with the LPs and MNP-PEG-AEDP-CTA nanoconjugates, while the second one with a NaCl solution (0.05 M) (Figure 1E). The solutions were pumped into the device with the syringe pumps at total flow rates from 1 to 3 mL/min by maintaining a 1:1 FRR. The samples recovered from each design were analyzed using a spectrofluorometer (0239D-2219 FluoroMax plus C, Horiba, Miyanohigashi, Japan) to track changes in the fluorescence intensity compared with the control sample, which is the solution before injection into the system. As for the EE experiment, the fluorescence spectrum of rhodamine B allowed intensity tracking by setting up excitation and emission filters at 546 nm and 568 nm, respectively.

#### *2.6. In Vitro Testing of MLPs*

#### 2.6.1. Hemocompatibility

To determine the hemocompatibility of the liposomes, MNP-PEG-AEDP-CTA nanoconjugates, and magnetoliposomes, a blood sample was extracted from a healthy human donor in a vacutainer tube containing EDTA. Erythrocytes were obtained by centrifugation at 1800 rpm for 5 min. The supernatant was discarded, and erythrocytes were washed five times with NaCl solution (0.9% *w*/*v*) and twice with PBS 1×. To form a stock solution, 1 mL of the washed erythrocytes was suspended in 9 mL of PBS 1× and carefully homogenized. The liposomes and magnetoliposomes were evaluated at 0.1, 0.05, and 0.025 mg/mL, while MNP-PEG-AEDP-CTA nanoconjugates were at concentrations ranging from 200 μg/mL to 12.5 μg/mL. Triton 100-X (10% *v*/*v*) and PBS 1× were used as positive and negative controls, respectively. To evaluate the hemolytic activity, 100 μL of the erythrocyte stock solution was seeded with 100 μL of the different treatments in a 96-well microplate. The

microplate was then incubated under constant stirring at 37 ◦C for 1 h. The plate was centrifuged, and the supernatants were then transferred to another 96-well microplate. Finally, absorbance was read at 450 nm, and hemolysis percentage was calculated by following Equation (16):

$$\text{Hemolysis}\left(\%\right) = 100 \times \frac{\left(\text{Abs}\left(sample\right) - \text{Abs}\left(\mathbb{C} - \right)\right)}{\left(\text{Abs}\left(\mathbb{C} + \right) - \text{Abs}\left(\mathbb{C} - \right)\right)}\tag{16}$$

#### 2.6.2. Platelet Aggregation

The platelet aggregation capacities of the liposomes, MNP-PEG-AEDP-CTA, and magnetoliposomes were evaluated by exposing them to a blood sample extracted from a healthy human donor in a vacutainer tube containing sodium citrate. Platelet-rich plasma (PRP) was obtained by centrifuging the collected blood at 1000 rpm for 15 min. Erythrocytes were discarded, and the supernatant containing PRP was used to run the test. The liposomes and magnetoliposomes were evaluated at 0.1, 0.05, and 0.025 mg/mL and the MNP-PEG-AEDP-CTA at concentrations ranging from 200 μg/mL to 12.5 μg/mL. Thrombin and PBS 1× were used as positive and negative references, respectively. The aggregation capacity was evaluated by exposing 50 μL of PRP to 50 μL of the different treatments in a 96-well microplate. The microplate was incubated at 37 ◦C for 5 min, and then absorbance was read at 620 nm. Platelet aggregation percentage was calculated by following Equation (17):

$$\text{Platelet aggregation} \left( \% \right) = 100 \times \frac{Abs \text{ (sample)}}{Abs \text{ (C+)}}, \tag{17}$$

#### 2.6.3. Cytotoxicity

The cytocompatibility of the liposomes, MNP-PEG-AEDP-CTA, and magnetoliposomes was determined as a measure of the impact on the metabolic activity in two different cell lines, namely, Vero (ATCC® CCL-81) and gastric cancer (AGS, ATCC® CRL-1739) cells, with the aid of a colorimetric assay based on 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium (MTT, Sigma-Aldrich, St. Louis, MO, USA). The liposomes and magnetoliposomes were evaluated at 0.1, 0.05, and 0.025 mg/mL and the MNP-PEG-AEDP-CTA at serial dilutions from 200 μg/mL to 12.5 μg/mL. Non-supplemented DMEM medium was used as the negative control. To evaluate cell viability in the different cell lines, 100 μL of a cell stock solution in DMEM medium supplemented with FBS (10%) was seeded in a 96-well microplate at a cell density of 10 × <sup>10</sup><sup>4</sup> cells/well. Microplates were incubated at 37 ◦C, 5% CO2, and a humidified atmosphere for 24 h. After that, DMEM medium supplemented with FBS (10%) was extracted and replaced with a non-supplemented DMEM medium containing the different treatments. Viability was studied at 24 and 48 h after the exposure. To determine the viability percentage, 10 μL of MTT reagent (5 mg/mL) was added to each well, and the microplates were then incubated, under the same conditions described above, for 2 h. Finally, supernatants were discarded, and 100 μL of DMSO was added to each well to dissolve the formed formazan crystals. The absorbance was read at 595 nm with the aid of a microplate reader (Thermo Scientific Multiskan™ FC Microplate Photometer). Cell viability was calculated by following Equation (18):

$$\text{Cell viability} \left( \% \right) = 100 \times \frac{\text{Abs (sample)}}{\text{Abs (C-)}}, \tag{18}$$

2.6.4. Cell Internalization and Endosomal Escape Analysis

Cell internalization and endosomal escape abilities of MNP-PEG-AEDP-CTA-RdB nanoconjugates and magnetoliposomes were assessed by colocalization between the labeled nanoconjugates and Lysotracker Green® DND-26 (Thermo Fisher, Waltham, MA, USA) in Vero (ATCC® CCL-81) and gastric cancer cells (AGS, ATCC® CRL-1739). For this, cells were seeded on glass slides deposited into a 24-well microplate at a cell density of

<sup>5</sup> × <sup>10</sup><sup>4</sup> cells/well. Cells were then incubated with DMEM medium supplemented with FBS (10% *v*/*v*) at 37 ◦C and 5% CO2 for 24 h to allow cell adhesion. Once the incubation time was achieved, DMEM medium was extracted and replaced with supplemented DMEM medium containing the different treatments at 50 μg/mL and magnetoliposomes with an equivalent amount of MNP-PEG-AEDP-CTA-RdB of 25 μg/mL. Cells were incubated for 0.5 h and 4 h. Next, the medium was extracted, and cells were washed three times with PBS 1× to remove the excess of the treatments. After this, PBS 1× was drawn, and cells were exposed to a DMEM solution containing Hoechst 33,342 (Thermo Fisher, Waltham, MA, USA) (1:1000) and Lysotracker Green® DND-26 (1:10000) for 10 min before imaging via confocal microscopy. The images were acquired in an Olympus FV1000 confocal laser scanning microscope with a PlanApo 60× oil immersion objective. Imaging of nuclei, endosomes, and MNP-PEG-AEDP-CTA-RdB nanoconjugates was performed at the following excitation/emission wavelengths: 358 nm/461 nm, 488 nm/520 nm, and 546 nm/575 nm, respectively. Analysis was carried out by taking 10 images for each treatment with an average of 10 cells per image. The internalization and cytosol distribution were studied by calculating the surface area coverage. Image processing and analyses were performed on the software Fiji-ImageJ®. Statistical analyses and data processing were carried out on GraphPad Prism® V 6.01 software (GraphPad Software, La Jolla, CA, USA). Statistical comparisons were made using the unpaired t-test. Results of *p* ≤ 0.05 (\*) were considered significant.

#### *2.7. Statistical Analyses*

All data measurements are reported as mean ± standard deviation. Each experiment was carried out in triplicate. Data analysis was performed using the Graph Pad Prism V 6.01® software. Statistical comparisons were determined by running two-way ANOVA followed by post-treatment (Dunn's Multiple Comparison test). Results with *p*-value ≤ 0.05 (\*) were considered significant. (\*) corresponds to statistically significant difference with a *p*-value between 0.01 and 0.05; (\*\*) represents 0.001 ≤ *p*-value < 0.01; (\*\*\*) represents 0.0001 ≤ *p*-value ≤ 0.001; and (\*\*\*\*) represents *p*-value < 0.0001. In addition, "ns" represents no statistically significant differences between the treatments.

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

#### *3.1. Characterization of Magnetoliposomes Using the Microfluidic Approach*

Figure 2A,B shows the size and PDI of the synthesized MLPs using the two-layer device. In this case, no apparent differences in size were identified for the evaluated TFRs and the concentration of the nanoconjugates used in the experiment for different FRRs. Nevertheless, there is a slight decrease in size with the increase of the FRR in almost all cases except for the TFR of 5 mL/min at a concentration of nanoconjugates of 0.17 mg/mL for the 1:1 to the 2:1 FRR, where it is comparable with the results for the synthesis of liposomes for different FRRs [31,41]. Additionally, the size of the MLPs synthesized is smaller than 400 nm except for the case where the TFR, concentration, and FRR are the lowest. The PDI of the synthesized MLPs presented values under 0.5 in almost all cases, which indicates that the MLPs samples obtained had acceptable polydispersity.

Figure 2C,D shows the size and PDI of the synthesized MLPs using the three-layer device. In this case, there is a significant difference in size for the evaluated TFRs and the concentration of the nanoconjugates used in the experiment for different FRRs. The 5 mL/min TFR led to lower size values than the ones obtained using the 2.5 mL/min TFR, and there is a slight increase in the size for the MLPs at a higher concentration of nanoconjugates. These observations are contrary to those reported by Joshi et al., according to which the TFR has no impact on the liposome sizes [33]. The dimensions of the channels and the incorporation of nanoconjugates into the lipid phase before entering the system might be relevant factors to explain the identified differences. The obtained MLPs sizes are larger than those obtained using the two-layer device, but in this case such sizes are under 400 nm. Additionally, the PDI values of the MLPs were below 0.5 in nearly all cases, which

 **)55 6L]HQP** P/PLQPJP/ P/PLQPJP/ P/PLQPJP/ P/PLQPJP/ P/PLQPJP/ P/PLQPJP/ P/PLQPJP/ P/PLQPJP/  **)55 3',DX** P/PLQPJP/ P/PLQPJP/ P/PLQPJP/ P/PLQPJP/  **)55 6L]HQP** P/PLQPJP/ P/PLQPJP/ P/PLQPJP/ P/PLQPJP/  **)55 3',DX \$ & ' %**

indicates that the MLPs samples obtained have an acceptable polydispersity. Compared with the two-layer device, there is a slight increase in the PDI values. This strongly suggests the two-layer system provides superior control over MLP sizes and their PDI values.

**Figure 2.** MLPs size and PDI for the two-layer device and the three-layer device. Two-layer device: (**A**) MLPs size using TFRs of 2.5 mL/min and 5 mL/min for nanoconjugates concentrations of 0.17 mg/mL and 0.32 mg/mL and FRRs from 1:1 to 5:1. (**B**) MLPs PDI using TFRs of 2.5 mL/min and 5 mL/min for nanoconjugates concentrations of 0.17 mg/mL and 0.32 mg/mL and FRRs from 1:1 to 5:1. Three-layer device: (**C**) MLPs size using TFRs of 2.5 mL/min and 5 mL/min for nanoconjugates concentrations of 0.17 mg/mL and 0.32 mg/mL and FRRs from 1:1 to 5:1. (**D**) MLPs PDI using TFRs of 2.5 mL/min and 5 mL/min for nanoconjugates concentrations of 0.17 mg/mL and 0.32 mg/mL and FRRs from 1:1 to 5:1.

Figure 3 shows the TEM characterization of the MLPs obtained with both the two-layer and three-layer devices. The images show that the MLPs formed correctly and in agreement with previous reports of MLPs synthesized via microfluidics [34,35]. The images also show that the size of the MLPs obtained for the three-layer device is slightly larger than that obtained with the two-layer device, which agrees well with the hydrodynamic diameters measured via DLS.

#### *3.2. Magnetolipsomes Encapsulation Efficiency*

Figure 4 shows the encapsulation efficiencies (EE%) for the MLPs synthesized with both devices at different TFR values, nanoconjugates concentrations in the lipid phase, and FRRs. The EE% obtained at 0.17 mg/mL nanoconjugates concentration with the two-layer device was higher for almost all evaluated FRRs, while no identifiable trend was observable for the three-layer device. This agrees well with the notion that a superior

control of MLP assembly is achievable with the two-layer devices. Nevertheless, the results show efficiencies ranging from 20% to 90% for both devices for different FRRs, supporting the idea that the operating conditions strongly influence the performance of the devices. In addition, no correlation was found between FRR or TFR with the EE% values, which strongly suggests that the encapsulation process occurs randomly throughout the microfluidic device. However, the FRR is still an essential parameter in the size control of MLPs, which indicates that it is critical to define quality control strategies along with the nanoconjugates concentration [34]. Additionally, it is important to remark that the microfluidic synthesis of this type of drug delivery system provides a suitable route to enhance the therapeutic drug delivery efficiency compared with traditional methods. This provides further evidence for the relevance of the MLPs produced in this study, as they show consistent properties (e.g., size and morphology) without significant investments in infrastructure or instrumentation [34,50].

**Figure 3.** MLP characterization via TEM. (**A**) MLP synthesized with the three-layer device using TFR of 5 mL/min for nanoconjugates concentration of 0.32 mg/mL and a FRR set at 4:1 (**B**) Magnification by 145 k× of the MLP presented in A. (**C**) MLPs synthesized with the two-layer device using TFR of 5 mL/min for nanoconjugates concentration of 0.32 mg/mL and a FRR set at 4:1.

**Figure 4.** MLP encapsulation efficiency (EE%) for two-layer and three-layer devices. (**A**) Encapsulation efficiency for the MLPs synthesized using the two-layer device with TFRs of 2.5 mL/min and 5 mL/min at nanoconjugates concentrations of 0.17 mg/mL and 0.32 mg/mL and FRRs from 1:1 to 5:1. (**B**) Encapsulation efficiency for the MLPs synthesized using the three-layer device with TFRs of 2.5 mL/min and 5 mL/min at nanoconjugates concentrations of 0.17 mg/mL and 0.32 mg/mL and FRRs from 1:1 to 5:1.

#### *3.3. Magnetoliposomes Purification*

Supplementary Figure S3A shows the intensity of the magnetic field acting on the microfluidic separation channel for System 1. The magnetic field is higher in proximity to the permanent magnets, as was expected [51]. Figure 5A shows the experimental performance of System 1 and the identification of some of the particles' accumulation regions. Figure 5B shows the microfluidic system's separation performance where both nanoconjugates and MLPs are attracted to the channel wall near the higher magnetic flux density regions. However, calculations failed to show that the percentage of nanoconjugates trapped is higher compared to the MLPs. Figure 5C presents the obtained velocity fields for the dispersed phase within the channels for the mixture model approach. The results indicate that an increase in the dispersed phase's velocity matches each magnet's highest intensity locations. Finally, Figure 5D shows the concentration profile for the nanoconjugates within the channel as estimated by the transport of diluted species model during the first seconds of the study. This result indicates that the concentration tends to increase in regions where the magnetic flux density is higher, which directly results in streamlines targeting these regions along the entire channel.

**Figure 5.** Qualitative results of the separation System 1. (**A**) Zoom of nanoconjugates accumulation regions inside the magnetophoretic microfluidic channel. (**B**) Particle trajectories in the magnetophoretic separation channel (nanoconjugates are shown in yellow and MLPs in red). (**C**) Velocity profile of the mixture. (**D**) Concentration profile in the microchannel.

Supplementary Figure S3B shows the intensity of the magnetic field acting on the microfluidic separation channel for System 2. Figure 6A shows the experimental performance of System 2 and the identification of some of the particles' accumulation regions. As for System 1, these regions are located in the channel sections near the magnets. Figure 6B shows the microfluidic system's separation performance, where the behavior presented is almost identical to that of System 1. In the case of the results for the mixture model, Figure 6C illustrates the velocity field results for the dispersed phase within the channels. In contrast, Figure 6D shows that the concentration profiles for the nanoconjugates and their streamlines tend to increase in regions of high magnetic flux density.

Figures 5 and 6 show qualitative results that illustrate the general performance of the proposed magnetophoretic separation devices and the trajectories of the nanoconjugates along the devices' microchannels. Nevertheless, a quantitative analysis is critical to estimate separation efficiencies and further verify them experimentally. For the case of the mixture model analysis, we selected several locations (Supplementary Figure S4A,B) along the computational domain to determine the concentration of each type of particle component (i.e., MLPs and nanoconjugates) and to calculate their concentration difference close to the zones where the magnetic field is the highest. The final separation efficiency percentage (SE%) was calculated as the average percentage difference in concentration for all of the selected locations. For the particle tracing model, a particle counter for each type of particle was set at the system's outlet.

**Figure 6.** Qualitative results of the separation System 2. (**A**) Zoom of nanoconjugates accumulation regions inside the magnetophoretic microfluidic channel. (**B**) Particle trajectories in the magnetophoretic separation channel (nanoconjugates are shown in yellow and MLPs in red). (**C**) Velocity profile of the mixture. (**D**) Concentration profile inside the microchannel.

The separation efficiency is calculated as the percentage difference between each type of particle arriving at the outlet. Figure 7 shows the quantitative results for the separation efficiency calculated for the two simulation approaches implemented here and the corresponding comparison with the obtained experimental results for both systems. The results show that the best separation efficiency achieved was 31.55% for System 1 at a TFR of 1 mL/min, while for System 2 it was 51.22% at a TFR of 2 mL/min.

For System 1, the results show the dominance of magnetophoretic over hydrodynamic forces, as an increase in the TFR led to a decrease in the separation efficiency [52]. Despite the higher separation efficiencies obtained with System 2, such a correlation was not clear for this system. In addition, it is important to highlight that we found that separation with small TFRs might lead to a relatively large fraction of MLPs trapped along with nanoconjugates in the accumulation regions, thereby reducing the number of purified MLPs at the end of the process.

Compared with the mixture model, quantitative results for System 1 indicate that the particle tracing model led to results with a higher level of agreement with those observed experimentally. In contrast, the System 2 mixture model simulation approach showed better performance in predicting quantitative separation results than those obtained with the particle tracing model. Despite these results, it is important to highlight a significant difference in the choice of one approach over the other in terms of ease of implementation. In this regard, although the mixture model describes the suspended magnetic nanoconjugates as a ferrofluid, making the simulation more realistic, the computational cost of this modeling approach compared with particle tracing is much higher [42,43,52]. Our results recommend implementing both models for a more comprehensive understanding of the devices, as these two approaches complement each other and might provide much more robust insights for further manufacturing and experimental testing.

**Figure 7.** Microfluidic separation efficiency for Systems 1 and 2. (**A**) Comparison of both simulation approaches and the experimental results obtained for System 1 with a TFR ranging from 1 to 3 mL/min. (**B**) Comparison of both simulation approaches and the experimental results obtained for System 2 with a TFR ranging from 1 to 3 mL/min. Results with a *p*-value ≤ 0.05 (\*) were considered significant. (\*) corresponds to statistically significant difference with a *p*-value between 0.01 and 0.05; (\*\*\*): 0.0001 ≤ *p*-value ≤ 0.001; (\*\*\*\*): *p*-value < 0.0001; "ns" represents no statistically significant differences between the treatments.

Because the obtained separation efficiencies with System 2 were higher than System 1, it was selected to produce the MLPs for further experimentation.

#### *3.4. In Vitro Testing of MLPs*

#### 3.4.1. Biocompatibility

Figure 8 shows the cytocompatibility, hemocompatibility, and platelet aggregation results for the produced MLPs and the LPs. Figure 8A,B shows the viability of Vero cells after 24 and 48 h of exposure to the different treatments. MLPs and LPs show high biocompatibility at concentrations below 0.1 mg/mL. However, at concentrations above 0.1 mg/mL, the cell viability decreases to about 70%, showing a dose-dependent behavior. A similar tendency was observed for AGS cells (Figure 8C,D). In addition, AGS cells exhibited high tolerance to the treatments, reaching viability percentages above 80%, even at concentrations higher than 0.1 mg/mL. Figure 5A–D shows the cytocompatibility results for Vero and AGS cells exposed to nanoconjugates. Significant cytotoxicity levels were found at concentrations above 50 μg/mL in AGS cells, whereas in Vero cells, the viability remained above 80% even at higher concentrations. This result can be related to the significant sensitivity of AGS cells to CTA. Additionally, Figure 8E and Supplementary Figure S5E show the hemolysis percentage of the MLPs, LPs, and nanoconjugates compared with the positive and negative controls. The results show hemolysis percentages below 1% for MLPs and LPs and 1.16% for the nanoconjugates at the highest evaluated concentrations. Similar results were obtained previously for MLPs, where the hemolysis percentages were below 5% [12,18,53,54]. Figure 8F and Supplementary Figure S5F show the platelet aggregation percentages of the MLPs, LPs, and nanoconjugates compared with the positive control. The results indicate that the platelet aggregation of MNP, MLPs, and LPs remain below 55% even at high concentrations. Compared with the negative control, the observed aggregation is acceptable even at the highest evaluated concentrations but is slightly higher than those reported previously [12,54].

**Figure 8.** Biocompatibility assays for MLPs and LPs. Viability of Vero cells after 24 (**A**) and 48 h (**B**) of exposure. Viability of AGS cells after 24 (**C**) and 48 h (**D**) of exposure. (**E**) Hemolysis of MLPs and LPs with Triton X-100 as the positive control and PBS 1× as the negative control. (**F**) Platelet aggregation of MLPs and LPs with PBS 1× as the negative control and thrombin as the positive one.

#### 3.4.2. Cell Internalization and Endosomal Escape Analysis

Figures 9 and 10 show the confocal images corresponding to the delivery of MNP-PEG-AEDP-CTA-RdB nanoconjugates (labeled MNPs in the figures) and MLPs on Vero and AGS cells at different times (i.e., 0.5 h and 4 h). Staining with Hoechst 33,342 allowed determination of the impact of the treatments on the cell viability by analyzing the nucleus morphology and distribution. Images clearly show nuclei with regular spherical-shaped morphology with no visible fragmentation or DNA condensation [55]. These results demonstrate non-apoptotic cells, confirming high biocompatibility in both cell lines even after 4 h of exposure. These results provide further evidence of the high cell viability levels obtained via MTT (Figure 8A–D and Supplementary Figure S5A–D).

**Figure 9.** (**A**) Cell internalization and endosomal escape for magnetoliposomes (MLPs) and MNP-PEG-AEDP-CTA-RdB nanoconjugates (MNPs) in Vero cells with 40× magnification after 0.5 h and 4 h of exposure. The scale bar corresponds to 100 μm. Vero cells with 60× magnification and digital zoom to 120× after 0.5 h and 4 h of exposure to MLPs (**B**) and nanoconjugates (MNPs) (**C**). The scale bar for both (**B**,**C**) corresponds to 50 μm. The yellow arrows indicate colocalization between the green and the red channels, showing nanoparticles encapsulated into endosomes. The white arrows indicate non-colocalized zones, displaying nanoconjugates that escaped endosomes or reached the intracellular space by a different internalization mechanism.

**Figure 10.** (**A**) Cell internalization and endosomal escape for magnetoliposomes and MNP-PEG-AEDP-CTA-RdB nanoconjugates (MNPs) in AGS cells with 40× magnification after 0.5 h and 4 h of exposure. The scale bar corresponds to 100 μm. AGS cells with 60× magnification and digital zoom to 120× after 0.5 h and 4 h of exposure to MLPs (**B**) and nanoconjugates (MNPs). (**C**). The scale bar for both (**B**,**C**) corresponds to 50 μm. The yellow arrows indicate colocalization between the green and the red channels, showing nanoparticles encapsulated into endosomes. The white arrows indicate non-colocalized zones, displaying nanoparticles that escaped endosomes or reached the intracellular space by a different internalization mechanism.

In addition, internalization and endosomal escape abilities were studied as a measure of the colocalization of Lysotracker Green® with the rhodamine B labeled nanoconjugates and their distribution intracellularly. Figures 9 and 10 show cells with a strong red fluorescent signal in the intracellular space, confirming the internalization of both MLPs and nanoconjugates. High cell penetration rates are most likely a consequence of employing PEG for the functionalization of MNPs to obtain the tested nanoconjugates and the use of LPs as powerful vehicles which favor membrane fusion and, consequently, the effective delivery of cargoes. LPs have also been reported to improve nanoparticle transport and plasma half-life [56]. The versatility of the developed vehicle allows the transport of CTA into the intracellular space even in the absence of LPs. This promising result presents the PEGylated magnetite-based nanovehicles as a fascinating tool for designing more potent oral delivery platforms to transport molecules of difficult intestinal absorption, such as CTA. This approach has been studied and validated previously by Kawish and colleagues [57].

Figure 11 shows the quantitative results for the analysis of endosomal escape and the distribution of the nanoparticles into the cells. Pearson's correlation coefficient (PCC) was used as a statistic tool for quantifying colocalization and covered area percentage to measure MLPs and nanoconjugates internalization and distribution. In AGS cells, a non-statistically significant difference was observed between the MLPs and the nanoconjugates, indicating that encapsulation into LPs failed to increase the endosomal escape of nanoconjugates. This was confirmed by an increase in the PCC after 4 h of exposure. However, the covered area of nanoconjugates slightly increased after 4 h, whereas the covered area of MLPs almost doubled for the same time. We hypothesize that these results might be a consequence of the interplay of different internalization routes. In this regard, as opposed to endocytic routes, it is very likely that nanoconjugates prevalently enter cells by a rapid and direct translocation mechanism [56]. In contrast, due to their negatively charged surface, LP and MLP internalization occurs mainly by endocytic routes. This is in line with recent reports that indicate that internalization rates for nanostructures with anionic coatings are lower than those of cationic and neutral coatings [56].

**Figure 11.** Pearson correlation coefficient (PCC) and percentage of area covered by the nanoconjugates in Vero and AGS cells. Higher PCC values indicate a higher amount of nanoconjugates trapped in endosomes. (**A**) PCC of MLPs and nanoconjugates in Vero and AGS cells for 0.5 h and 4 h. (**B**) Area covered by the MLPs and nanoconjugates in Vero and AGS cells for 0.5 h and 4 h. Results with a *p*-value ≤ 0.05 (\*) were considered significant. (\*) corresponds to statistically significant difference with a *p*-value between 0.01 and 0.05; (\*\*\*): 0.0001 ≤ *p*-value ≤ 0.001; (\*\*\*\*): *p*-value < 0.0001; "ns" represents no statistically significant differences between the treatments.

In Vero cells, the vehicles led to entirely different results, as evidenced by a statistically significant decrease in the PCC for both treatments after 4 h, confirming, therefore, endosomal escape. Somewhat surprisingly, for the same time, the covered area showed a statistically significant decrease. This suggests that, after internalization, the nanoconjugates escape endosomes and likely accumulate in different organelles. Future work will be dedicated to confirming this hypothesis.

The different penetration levels achieved for the two evaluated cell lines can be attributed to their significantly different cell membrane compositions, which might substantially alter the cell–nanoconjugate interactions. For example, the overexpression of claudin proteins in AGS cell membranes is likely to interfere with the internalization routes, rates, and achieved intracellular distributions [58]. Based on this, the rational design and development of novel vehicles for specific therapeutic applications must include a comprehensive analysis of membrane composition for the targeted cells. This is critical to engineer nanovehicles capable of taking advantage of and avoiding the possible interactions leading to cell penetration.

The obtained results clearly show the great potential of the developed nanovehicles as versatile carriers for difficult-to-transport drugs with high biocompatibility and the possibility to selectively internalize in cells with specific characteristics. Despite these promising results, future studies should include detailed studies on the impact of MLPs and the delivered compounds on microbiota bioactivity and bioavailability. This with the main objective of ensuring the homeostasis of patients' microbiomes, which is controlled by molecular interactions and plays a central role in modulating metabolism, immunity, and response to infections. In addition, this understanding is critical to determining the bioavailability of the functional compounds, as it largely depends on the metabolism exerted by the microbiota during the delivery process [20,22]. We expect that our MLPs will show no impact on the microbiome balance and, therefore, contribute to novel oral delivery routes for patients suffering from complex diseases, such as eczema, inflammatory disorders, hypertension, and chronic kidney disease [20,59].

#### **4. Conclusions**

Over the past few years, MLPs have been studied to improve the cell penetration efficiency of nanovehicles for drug delivery after administration. This study presents a low-cost microfluidic device to produce MLPs with sizes ranging from 136.87 ± 3.97 nm to 787.47 ± 45.65 nm and polydispersity values ranging from 0.21 ± 0.02 to 0.58 ± 0.04. The device operates by putting into intimate contact lecithin liposomes (LPs) and functionalized magnetite nanoparticles (MNPs) within a serpentine microchannel. As reported elsewhere, the size of the MLPs appears to be strongly influenced by the flow rate ratio (FRR) between the components infused into the system, which supports the importance of considering this parameter for designing and optimizing the device's performance. We prepared MNPs functionalized with a polymeric spacer (PEG) and a molecule containing a reducible disulfide bond (AEDP) to evaluate encapsulation. Additionally, we selected the immobilization of antibiotic CefTRIAxone (CTA) for proof-of-concept, considering its limited passing of the intestinal lumen after oral delivery. For all evaluated FRRs (i.e., from 1:1 to 5:1), the obtained nanoconjugates (i.e., MNP-PEG-AEDP-CTA) were encapsulated into the LPs (to form the MLPs) with efficiency above about 80% at a concentration of 0.17 mg/mL and while operating the device at a TFR of 5 mL/min.

However, this approach is challenging, as purifying the obtained MLPs is not a simple task, due to the proximity in properties of the involved components. Here, we decided to take advantage of the magnetic properties of the nanoconjugates to develop a robust and high-throughput separation scheme enabled by microfluidics and permanent magnets. Accordingly, we designed two magnetophoretic microfluidic separation devices where permanent magnets can be located along the main separation channels to retain excess nanoconjugates. To investigate the feasibility of this approach, we conducted two different multiphysics simulations approaches that provided complementary qualitative and quantitative information regarding the purification efficiency of the proposed devices. The first one was based on a particle tracing model, while the second relied on a multiphase mixture model. After conducting parametric sweeps for the FRRs, the results of the two approaches allowed us to confirm that the designed systems were well-suited to retain the nanoconjugates at hot spots of high magnetic field intensity. Although the qualitative results of both simulations show adequate behavior for the nanoconjugates within the systems, quantitative results varied considerably between approaches, due to the differences in inlet parameters and conditions used for each approximation. Conversely, it is highly suggested to combine these approaches for the rational design of microfluidic separation devices. The experimental testing showed separation efficiencies ranging from 30% to 31% for System 1 and 47% to 51% for System 2, showing a weak correlation between the TFR and separation efficiency in System 1. According to these results, System 2 operating at a TFR of 2 mL/min was employed for purifying MLPs for further in vitro testing.

Finally, we validated and demonstrated the great potential of the developed nanovehicles as a versatile carrier for difficult-to-transport drugs by showing high hemocompatibility, low platelet aggregation, and high cytocompatibility in two relevant cell lines (i.e., Vero and AGS). Despite different cell internalization and endosomal escape results in these two cell lines, the achieved coverage shows a promising potency, which is attractive for applications in gastrointestinal delivery. Furthermore, the results suggest unique nanoconjugate–cell membrane interactions and, consequently, interplay of different internalization mechanisms, which need to be considered for further surface engineering experiments.

**Supplementary Materials:** The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/pharmaceutics14020315/s1, Figure S1: Manufactured separation Systems 1 and 2. (A) System 1 (B) System 2, Figure S2: Meshing used for multiphysics simulations. (A) System 1 for particle tracing; (B) System 2 for particle tracing; (C) System 1 for mixture model; (D) System 2 for mixture model, Figure S3: Magnetic flux density results for the separation Systems 1 and 2. (A) System 1 and (B) System 2, Figure S4: Evaluation points for mixture model simulations separation efficiency. (A) System 1 and (B) System 2, Figure S5: Biocompatibility assays for nanoconjugates (MNPs). Viability of Vero cells after 24 (A) and 48 h (B) of exposure. Viability of AGS cells after 24 (C) and 48 h (D) of exposure. (E) Hemolysis of MNPs with Triton X-100 as the positive control and PBS 1× as the negative control. (F) Platelet aggregation of nanoconjugates with PBS 1× as the negative control and thrombin as the positive one.

**Author Contributions:** Conceptualization, J.C.C. and J.F.O.; methodology, C.E.T., J.C., P.R.P., J.C.C. and J.F.O.; formal analysis, C.E.T., J.C. and J.C.C.; investigation, C.E.T., J.C., S.C.G., V.Q., K.A.G., P.R.P., J.A.S. and L.R.-G.; resources, C.M.-C., L.H.R., J.F.O. and J.C.C.; writing—original draft preparation, C.E.T. and J.C.; writing—review and editing, C.E.T., J.C., C.M.-C., L.H.R., J.F.O. and J.C.C.; visualization, J.C.; supervision, C.M.-C., L.H.R., J.C.C. and J.F.O.; project administration, J.C.C. and L.H.R.; funding acquisition, C.M.-C., J.C.C. and L.H.R. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Colombian Ministry of Science, Technology, and Innovation (Minciencias), Grant IDs 782-2019 and 845-2018. Additional funding was provided by the 2019 Fundación Santafé de Bogotá-Uniandes grant "Production of recombinant antimicrobial peptides to modify materials of biomedical interest" and 2018 Fundación Santafé de Bogotá-Uniandes grant "Development of multifunctional magnetoliposomes as vehicles for the delivery of combined therapies of low dosage and high bioavailability for the treatment of Parkinson's disease".

**Institutional Review Board Statement:** Human blood samples were collected under the permission granted by the ethics committee at Universidad de los Andes (minute number 928–2018).

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Acknowledgments:** We would like to thank the Department of Biomedical Engineering, the Department of Food and Chemical Engineering, and the Department of Electrical and Electronics Engineering at Universidad de los Andes for the financial and technical support. Additionally, we thank Juan Camilo Orozco for performing the confocal experiments and Johana Arboleda for TEM imaging.

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

#### **References**


## *Article* **Complementary Nucleobase Interactions Drive Co-Assembly of Drugs and Nanocarriers for Selective Cancer Chemotherapy**

**Fasih Bintang Ilhami 1, Enyew Alemayehu Bayle <sup>1</sup> and Chih-Chia Cheng 1,2,\***


**Abstract:** A new concept in cooperative adenine–uracil (A–U) hydrogen bonding interactions between anticancer drugs and nanocarrier complexes was successfully demonstrated by invoking the co-assembly of water soluble, uracil end-capped polyethylene glycol polymer (BU-PEG) upon association with the hydrophobic drug adenine-modified rhodamine (A-R6G). This concept holds promise as a smart and versatile drug delivery system for the achievement of targeted, more efficient cancer chemotherapy. Due to A–U base pairing between BU-PEG and A-R6G, BU-PEG has high tendency to interact with A-R6G, which leads to the formation of self-assembled A-R6G/BU-PEG nanogels in aqueous solution. The resulting nanogels exhibit a number of unique physical properties, including extremely high A-R6G-loading capacity, well-controlled, pH-triggered A-R6G release behavior, and excellent structural stability in biological media. Importantly, a series of in vitro cellular experiments clearly demonstrated that A-R6G/BU-PEG nanogels improved the selective uptake of A-R6G by cancer cells via endocytosis and promoted the intracellular release of A-R6G to subsequently induce apoptotic cell death, while control rhodamine/BU-PEG nanogels did not exert selective toxicity in cancer or normal cell lines. Overall, these results indicate that cooperative A–U base pairing within nanogels is a critical factor that improves selective drug uptake and effectively promotes apoptotic programmed cell death in cancer cells.

**Keywords:** adenine–uracil base pair; complementary hydrogen bonded drug carrier system; controlled drug delivery; supramolecular nanogels; selective cytotoxicity

#### **1. Introduction**

The specific sequences present in biopolymers such as DNA, RNA, and proteins are responsible for the survival of complex, adaptable living organisms [1]. Analogous synthetic polymers with well-controlled, designed sequences have been predicted to function as components in a wide range of applications [2–4]. Complementary, noncovalent multiple hydrogen bonding interactions—such as adenine-thymine (A–T), guanine-cytosine, and adenine–uracil (A–U) base pairing between nucleic acids—provide versatile tools to control and tune the structure and function of polymers containing complementary nucleobases [5–8]. In recent decades, numerous research groups have reported that "bioconstituted" hydrogen bonding interactions could facilitate the self-assembly of various structures with nanometer-scale features. Moreover, well-controlled, dynamic physical properties and varied stimuli-responsive properties in response to environmental changes could be achieved by varying the amounts and strength of the nucleobase pairs [9–11]. For instance, supramolecular amphiphilic block copolymers with complementary adenine and thymine hydrogen bonding interactions that spontaneously self-assemble into nano-sized micelles in aqueous solution were reported by Kuang et al., and a well-controlled drug release profile could be obtained by tuning the content of A–T in the polymer structure [12]. Wang et al. designed a unique supramolecular phospholipid with moderate hydrogen

Cheng, C.-C. Complementary Nucleobase Interactions Drive Co-Assembly of Drugs and Nanocarriers for Selective Cancer Chemotherapy. *Pharmaceutics* **2021**, *13*, 1929. https://doi.org/10.3390/ pharmaceutics13111929

**Citation:** Ilhami, F.B.; Bayle, E.A.;

Academic Editors: Francisco José Ostos, José Antonio Lebrón and Pilar López-Cornejo

Received: 19 October 2021 Accepted: 12 November 2021 Published: 15 November 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 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 (https:// creativecommons.org/licenses/by/ 4.0/).

bonding interaction recognition between adenosine and uridine; this phospholipid could self-assemble into spherical liposomes with highly pH-responsive ability, and had high potential for controlled drug release [13]. Fan and co-workers blended an adenine-containing amphiphilic block copolymer and uracil-functionalized crosslinking agent to self-assemble new micelles that exhibited rapid, pH-controlled drug release to significantly reduce cell viability [14]. Based on the examples above, synthetic supramolecular polymers containing nucleobase pairs are a promising concept in various areas of research, and the exploitation of non-covalent interactions to confer polymeric structures with dynamic characteristics may expand their applications.

The development of polymeric nanocarriers in order to improve chemotherapeutic efficacy and intracellular delivery for cancer therapy has attracted overwhelming enthusiasm in modern pharmaceutical technology. In terms of biocompatible nanocarriers, polyethylene glycol (PEG) is regarded as one of the most important components of nanocarriers for drug delivery, owing to its high level of water solubility, considerable chain mobility, and low toxicity, among other suitable physicochemical properties [15–17]. Moreover, the introduction of PEG segments into nanocarrier structures decreases their tendency to form aggregates, and thus results in enhanced structural stability and avoids the clearance of the nanocarriers by the reticuloendothelial system (RES) [18–20]. Koo et al. noted that shell cross-linked polymer micelles based on a combination of PEG and polyamino acids exhibited high stability in water, improved biocompatibility towards normal and cancer cells, and enhanced the intracellular release triggered by glutathione (GSH) [21]. Yokoyama et al. developed a new PEG-*b*-poly(α,β-aspartic acid) block copolymer; the resulting drug-loaded micelles were extremely stable in aqueous solution, and their size and distribution could be easily tuned [22]. Nevertheless, previous reports mentioned that not all hydrophilic PEG segments provide advantageous behavior in an aqueous environment [23–25]. A range of hydrophilic PEG segments have varied stabilization effects, including invoking side-effects via immunological responses, inhibiting the intracellular uptake of the nanocarrier, and non-biodegradability. Thus PEG segments with a molecular weight below 20 kDa are preferable for use in drug carriers [23–27]. Inspired by the specific features of nucleobases, we confidently speculated that the incorporation of nucleobase molecules into the terminal end-groups of PEG may confer unique self-assembly behavior and physical properties in aqueous solution, leading to potential candidates for drug delivery applications.

A recent series of studies in our laboratory demonstrated that the introduction of complementary A–U base pairs within polymer structures confers the ability to spontaneously self-assemble into stable, physically crosslinked networks [28], resulting in excellent filmforming capability, tailorable mechanical performance and intriguing self-healing capacity upon tuning of the content of the A–U complexes in the matrix [29]. To further extend the concept of complementary nucleobase interactions to drug carrier systems, we herein design and synthesize a new difunctional, uracil-terminated PEG macromer (BU-PEG) that associates with adenine-functionalized rhodamine (A-R6G) in an aqueous environment via complementary A–U interactions, and thus results in the formation of self-assembled spherical nanogels with high structural stability (Scheme 1) [30]. In addition, the resulting A-R6G/BU-PEG nanogels possess a number of unique physical properties, including extremely high A-R6G-loading capacity, wide-range tailorable A-R6G-loading content, distinct green fluorescence behavior, and well-controlled pH-triggered A-R6G release. Importantly, a series of in vitro experiments clearly confirmed that the A-R6G-loaded BU-PEG nanogels highly selectively targeted cancer cells, and were selectively internalized by them, and could thus promote high levels of rapid apoptotic death through an endocytotic pathway, but they were not internalized by and did not harm normal cells. As far as we are aware, this is the first study to develop a complementary drug\_carrier system based on hydrogen bonding interactions between stable A–U base pairs, with the goals of improving both the safety and effectiveness of cancer chemotherapy. Thus, this new concept for the fabrication of drug\_carrier systems via complementary A–U base pairing offers in-depth

insight into the benefits of manipulating the drug delivery behavior of nanocarriers with potential in various biological and biomedical applications.

## pairs under various pH conditions. **2. Materials and Methods**

#### *2.1. Chemicals and Materials*

Uracil (≥99% purity) and adenine (>99.5% purity) were purchased from Acros Organics (Geel, Belgium). Rhodamine 6G (R6G), polyethylene glycol (PEG, average molecular mass: 1900–2200 g/mol), dimethylformamide (DMF), potassium *tert*-butoxide (*t*-BuOK), triethylamine (TEA), deuterated chloroform (CDCl3), and deuterium oxide (D2O) were obtained from Sigma-Aldrich Chemical (Milwaukee, WI, USA) at the highest purity available. The HPLC-grade organic solvents were used as received from TEDIA (Fairfield, OH, USA). Phosphate-buffered saline (PBS), Dulbecco's modified Eagle's medium (DMEM), fetal bovine serum (FBS), penicillin-streptomycin, trypsin-EDTA, trypan blue, 4 ,6-diamidino-2 phenylindole (DAPI), the Dead Cell Apoptosis Kit with Brilliant Violet-421™ Annexin V (BV421-Annexin V), and Ghost Dye™ Red 780 (GDR780) were purchased from Thermo Fisher Scientific (Waltham, MA, USA). All chemicals and reagents were employed as received. Adenine-functionalized rhodamine derivative (A-R6G) and PEG diacrylate (PEGDA, number average molecular weight (Mn) = ~2000) were synthesized and characterized according to procedures that have been described previously [31–33].

**Scheme 1.** Schematic illustration of the co-assembly process, fluorescent properties, and cancer cell-selective cytotoxic behavior of hydrogen-bonded A-R6G/BU-PEG complexes. The upper-right green and red arrows represent the association and dissociation of the complementary A–U base

#### *2.2. Synthesis of BU-PEG*

PEGDA (4 g, 2 mmol) and uracil (0.5 g, 4.5 mmol) were dissolved in 300 mL of DMF, and agitated for 48 h at 60 ◦C with a small quantity of *t*-BuOK as a catalyst (0.04 g, 0.003 mmol). After removing DMF by vacuum distillation, the crude product was dissolved in chloroform (200 mL) and the insoluble impurities were removed by filtration through a Büchner funnel. Subsequently, the chloroform was evaporated by rotary evaporation, then the obtained product was washed three times with diethyl ether and dried in an oven at 30 ◦C for 1 day. The yield was 83% (3.8 g).

#### *2.3. Characterization*

#### 2.3.1. Proton and Carbon Nuclear Magnetic Resonance (1H NMR and 13C NMR)

The 1H NMR and 13C NMR spectra were obtained on a Bruker AVIII NMR spectrometer (Billerica, MA, USA) at 500 MHz; the 25 mg samples were dissolved in 1 mL of CDCl3 and D2O, respectively.

#### 2.3.2. Gel Permeation Chromatography (GPC)

Molecular weight information was obtained using a Waters Alliance 2690 HPLC Separation Module (Waters Corporation, Milford, MA, USA) with tetrahydrofuran (THF) as the mobile phase at a flow rate of 1.0 mL/min and 40 ◦C. The *M*n and the polydispersity index (PDI) were determined by comparison against a series of narrow distribution polystyrene standards.

#### 2.3.3. Critical Micelle Concentration (CMC)

Pyrene was utilized as a fluorescent probe to measure the CMC of the PEG and BU-PEG polymers. Varied concentrations of samples (0.00001 to 0.4 mg/mL) were prepared in water in advance. Next, 10 μL of pyrene solution was dropped into the tubes, then the mixtures were sonicated and incubated at 4 ◦C overnight to allow the polymers and aqueous phase to stabilize fully. All samples were examined using a fluorescence spectrometer (Jasco FP-8300 Spectrophotometer Hitachi, Tokyo, Japan) at an excitation wavelength of 335 nm. The emission intensities at 373 nm and 392 nm were recorded and plotted against the sample concentration in order to determine the CMC values.

#### 2.3.4. Particle Size and Surface Charge

Hydrodynamic particle size, PDI, and zeta (*ζ*) potential values were measured with a dynamic light scattering particle analyzer (DLS, Nano Brook 90Plus PALS, Brookhaven, Holtsville, NY, USA) connected to a 632-nm He-Ne laser beam (scattering angle: 90◦). DLS measurements for each sample were repeated ten times, and averaged.

#### 2.3.5. Photoluminescence (PL) and Ultraviolet-Visible (UV-Vis)

The PL and UV-Vis spectra of samples in aqueous solution were acquired at 25 ◦C using Hitachi F4500 luminescence and Jasco FP-8300 spectrophotometers (Hitachi, Tokyo, Japan), respectively.

#### 2.3.6. Atomic Force Microscopy (AFM) and Scanning Electron Microscopy (SEM)

The surface morphologies of the samples were assessed by AFM (NX10; AFM Park Systems, Suwon, Korea) and SEM (JSM-6500F system JEOL, Tokyo, Japan). Specimens were prepared by spin-coating diluted aqueous solutions onto silicon wafers at 1250 rpm for 15 s, and subsequently vacuum-dried at ambient temperature overnight.

#### *2.4. Preparation of A-R6G/BU-PEG and R6G/BU-PEG Complexes*

Different amounts of A-R6G or R6G (0.1 mg, 0.5 mg, and 1 mg) were added to BU-PEG (1 mg) in DMF (2 mL), stirred for 24 h, and then dialyzed against PBS (pH 7.4, 10 mM) or distilled water (DW) for 24 h (1000 Da molecular weight cut-off (MWCO)); the PBS (or DW) was replenished every 4 h. Finally, the absorption spectra of the A-R6G/BU-PEG solutions were obtained via UV/Vis spectrophotometry at λ = 525 nm to establish the concentration–absorbance standard curves of R6G and A-R6G. The absorption spectra of the A-R6G/BU-PEG and R6G/BU-PEG complexes (1 mg/mL) were compared to the standard curves for R6G and A-R6G. The following equation was used to calculate the drug loading content (DLC) and drug loading efficiency (DLE):

$$\text{DLC}\% = \frac{\text{Weight of drug loaded in polymeric nanogels}}{\text{Weight of drug loaded polymeric nanogels}} \times 100\%$$

DLE% <sup>=</sup> Weight of drug loaded in polymeric nanogels weight of drug input <sup>×</sup> <sup>100</sup>

#### *2.5. Evaluation of the Stability of Pristine BU-PEG and R6G/A-R6G-Loaded BU-PEG Nanogels*

The stability of blank and R6G/A-R6G-loaded BU-PEG nanogels in DMEM were investigated in the presence of FBS, which functions as a nanoparticle-destabilizing agent [34]. Pristine BU-PEG and A-R6G-loaded or R6G-loaded BU-PEG nanogels were mixed with DMEM containing 2:1 *v*/*v* serum, and the particle size and distributions were measured over 24 h by DLS.

#### *2.6. Study of the Drug Release Behavior of Drug-Loaded Nanogels*

The A-R6G-loaded or R6G-loaded BU-PEG nanogels (5 mL) were placed in dialysis tubing (MWCO = 1000 Da) and immersed in 50 mL of PBS with various pH values (7.4, 6.5, or 6.0) at 25 ◦C; the PBS was stirred (100 rpm) using a magnetic stirrer. At predefined intervals, 5 mL of external buffer solution was sampled and replaced with 5 mL of new PBS solution with the same pH. The amount of released R6G (or A-R6G) was measured by UV-Vis spectrometry at λ = 525 nm against a standard calibration curve for free R6G in PBS, and plotted vs. time.

#### *2.7. Hemolysis Assay*

Sheep red blood cells (SRBCs, Cosmo Bio, Tokyo, Japan) were used to assess the hemolytic activity of A-R6G-loaded and R6G-loaded BU-PEG nanogels. Briefly, 1 mL of the SRBCs and 0.5 mL of PBS were added into a microcentrifuge tube, centrifuged at 12,000 rpm for 15 min, and the supernatant was removed as plasma. Next, 1.5 mL of PBS was added, vortexed, and centrifuged; this wash step was repeated three times until the supernatant was clear. Subsequently, the resulting SRBC solutions (500 μL) were added to various concentrations of drug-loaded nanogels (10, 20, 40, 100, and 150 μg/mL). Triton X-100 solution (1%) was used as a positive control and PBS as a negative control. All samples were placed in a 5% CO2 incubator at 37 ◦C for 4 h, then centrifuged, and then 100 μL of the supernatants were transferred into a 96-well plate and the absorbance values were measured at a wavelength of 540 nm on an ELISA reader. The hemolysis index was calculated as follows:

$$\text{Hemolyis} \%= \frac{\text{A}\_{\text{sample}} - \text{A}\_{\text{negative}}}{\text{A}\_{\text{positive}} - \text{A}\_{\text{negative}}} \times 100\%$$

where a represents the optical density (OD) values of the test sample, positive control (1% Triton X-100), or negative control (PBS).

#### *2.8. Cell Lines and Culture Conditions*

HeLa cells (human cervical cancer cell lines), MG-63 (human bone cancer cell lines), and NIH/3T3 cells (mouse embryonic fibroblast cell lines) were obtained from the ATCC (American Type Culture Collection, Manassas, VA, USA) and routinely cultured in DMEM supplemented with 10% FBS containing 1% penicillin-streptomycin in T-75 culture flasks at 37 ◦C in a 5% CO2 incubator.

#### *2.9. In Vitro Cell Cytotoxicity Assay*

The cytotoxicity of the A-R6G-loaded and R6G-loaded BU-PEG nanogels were investigated against normal cell lines (NIH/3T3 cells) and cancer cell lines (HeLa and MG-63 cells) using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) method. Briefly, 1 × 105 cells/well were seeded into 96-well plates in 100 <sup>μ</sup>L of complete DMEM medium, allowed to adhere for 24 h, then the culture media were replaced with new media containing 0.1–100 μg/mL of blank BU-PEG, free R6G or A-R6G, or A-R6G-loaded or R6G-loaded BU-PEG nanogels. The plates were incubated for 24 h, the media were removed, and 100 μL of MTT assay solution (5 mg/mL) was added per well, incubated

for 4 h, then 100 μL of dimethyl sulfoxide was added to dissolve the formazan crystals in each well. The optical densities were measured using an ELx800 microplate reader (BioTek, Winooski, VT, USA) at 570 nm.

#### *2.10. Cellular Uptake of R6G/A-R6G-Loaded BU-PEG Nanogels*

Intracellular uptake of A-R6G-loaded and R6G-loaded BU-PEG nanogels was evaluated in NIH/3T3 and HeLa cells by confocal laser scanning microscopy (CLSM) and flow cytometry.

CLSM: NIH/3T3 and HeLa cells were seeded into 35-mm microscopy dishes at a density of 1 × 105, cultured for 24 h, then the medium was replaced with fresh serum-free medium containing A-R6G-loaded or R6G-loaded BU-PEG nanogels and incubated for various times (3, 12, or 24 h). Subsequently, the NIH/3T3 and HeLa cells were gently rinsed three times with cold PBS and fixed in 4% paraformaldehyde for 30 min at 25 ◦C. Thereafter, the nuclei of the cells were stained using 4 ,6-diamidino-2-phenylindole (DAPI) for 30 min. The fluorescence of the cells was visualized via confocal microscopy (iRiS™ Digital Cell Imaging System, Logos Biosystems, Anyang-si, Korea).

Flow cytometry: NIH/3T3 and HeLa cells were seeded into 6-well plates at a density of 2 × <sup>10</sup><sup>6</sup> cells/well, incubated for 24 h at 37 ◦C, then A-R6G-loaded or R6G-loaded BU-PEG nanogels dissolved in fresh free-serum medium were added to the wells and cultured for 1, 3, 6, 12, or 24 h. The cells were detached using trypsin, re-suspended with cold PBS (0.5 mL) and analyzed on a BD FACSAria III flow cytometer (BD Biosciences, San Jose, CA, USA). Data events were collected and determined by FlowJo software.

#### *2.11. Analysis of Apoptosis Induced by R6G/A-R6G-Loaded BU-PEG Nanogels*

The BV421-Annexin and Ghost Red Dye-780 Detection Kit was used to double stain the NIH/3T3 and HeLa cells. Briefly, the cells were seeded in 6-well plates at 1 × <sup>10</sup><sup>6</sup> cells/well, allowed to adhere for 24 h, then incubated with A-R6G-loaded or R6G-loaded BU-PEG nanogels for different periods of time (1, 3, 6, or 12 h). Untreated NIH/3T3 and HeLa cells were prepared as controls. Next, the cells were re-suspended in binding buffer and stained using BV421-Annexin (15 min), followed by Ghost Red Dye-780 (15 min) in the dark, in accordance with the manufacturer's protocol, then analyzed by flow cytometry (BD FACSAria III).

#### *2.12. Statistical Analysis*

All results are provided as the means and standard deviations of at least three independent experiments.

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

Herein, our research aims and objectives focused on the development of cooperative multiple hydrogen bonds in nucleobase-functionalized groups between a nanocarrier and drug molecules to enhance the safety and efficiency of cancer chemotherapy, as illustrated in Scheme 1. A new a water-soluble uracil-end-capped BU-PEG polymer was developed and was prepared via a one-step Michael addition reaction of PEGDA [32,33] to uracil using a *t*-BuOK catalyst, resulting in a high-yielding product (83%). The resulting BU-PEG polymer showed the desired chemical structures and molecular weight, as determined by 1H NMR, 13C NMR, and GPC (see Supplementary Materials, Figure S1). In addition, BU-PEG is a white, semi-crystalline powder and can easily dissolve in water at 25 ◦C, even at concentrations as high as 50 mg/mL. Due to the formation of A–U base-pairings between BU-PEG and adenine-functionalized molecules, a strong cooperative hydrogen bonding partner drug, A-R6G, was synthesized according to our previous report [31]. After the introduction of an adenine moiety into the R6G structure, the drug exhibited extremely poor solubility in water, buffer and biological media, demonstrated unique green-fluorescent behavior, and exerted strong cytotoxic effects in various normal and cancer cell lines [31,35–37]. Due to the significant differences in the water solubility between

the BU-PEG polymer and A-R6G drug, these intriguing findings motivated our interest in exploring the co-assembly behavior of BU-PEG/A-R6G complexes in water.

Before discussing the hydrogen-bonded complexes of BU-PEG with A-R6G in water, we first determined the basic physical characteristics of BU-PEG in water to help our understanding and control the self-assembly structures and dynamics of BU-PEG in water and to guide the construction of a drug\_carrier system based on the combination of BU-PEG and A-R6G. To investigate the effect of the hydrogen-bonded uracil groups on the amphiphilic features of the hydrophilic PEG backbone in aqueous environments, UV-Vis measurements using hydrophobic pyrene as a fluorescent probe were employed to determine the CMC values of pristine PEG (Mn = ~2000) and BU-PEG in water [38]. As shown in Figure 1a, BU-PEG showed a low CMC value of 2.6 × <sup>10</sup>−<sup>2</sup> mg/mL, whereas pristine PEG did not exhibit any CMC characteristics across a broad concentration range from 10−<sup>5</sup> to 1 mg/mL, suggesting that the introduction of a uracil moiety into the PEG end-groups significantly impacted the amphiphilicity of the PEG backbone in water. This observation was possibly due to the presence of the self-complementary, interpolymer uracil hydrogen bonding interactions between the chains, which thus prompted the formation of relatively longer linear polymer chains connected by hydrogen-bonded uracil dimers that subsequently altered the overall amphiphilicity of BU-PEG. Next, we studied the self-organized structure of BU-PEG in water via complementary DLS, AFM and SEM measurements. When the concentration of the BU-PEG solution was 0.5 mg/mL (above its CMC value), BU-PEG had a mean hydrodynamic diameter of 86 nm, a PDI value of 0.27 and a *ζ*-potential value of −41.12 ± 0.36 mV (Figure 1b, Table S1), suggesting that BU-PEG could self-construct uniform, mono-distributed nanogels in water; this observation was attributed to the presence of the self-complementary uracil hydrogen bonding interactions within the polymer structure. In validation of the DLS results, AFM and SEM images confirmed that BU-PEG formed spherical nanogels with a diameter ranging from 35 to 70 nm (Figure 1c,d). Thus, these results revealed that the uracil units within the polymer structure served as a key governing force in hydrogen bonding molecular recognition to induce the self-assembly of the polymer in an aqueous environment.

After exploring the self-assembled structures and the characteristics of BU-PEG in water, we examined the molecular recognition of the A–U base-pairing between BU-PEG and A-R6G in D2O using 1H NMR spectroscopy. When a binary mixture of A-R6G and BU-PEG was prepared at a 1:10 blending weight ratio in D2O, the hydrophobic A-R6G almost completely dissolved in D2O and the mixed solution had a light orange color, while pristine A-R6G entirely precipitated in D2O, implying that A-R6G has a high tendency to interact with BU-PEG via strong complementary A–U interactions, thus leading to a significant increase in the solubility of A-R6G in D2O. Further 1H NMR analysis led to clear observation of the characteristic A-R6G peaks for the A-R6G/BU-PEG blend in D2O, whereas no peaks were detected for pristine A-R6G due to its poor water solubility (Figure S2). These results clearly and unequivocally demonstrate that the adenine units of A-R6G underwent complementary hydrogen bonding interactions with the uracil end-groups of BU-PEG in the aqueous solution, which thus promoted the formation of the A-R6G/BU-PEG complex and drastically improved the water-solubility of A-R6G. In addition, these results also implied that the existence of the complementary A–U interactions within the A-R6G/BU-PEG complexes may significantly improve the encapsulation capacity and stability of A-R6G after blending/mixing with BU-PEG. Thus, we subsequently performed an A-R6G encapsulation experiment to evaluate the drug-loading performance of BU-PEG nanogels via a dialysis method (see further details in the Experimental Section) in order to validate whether the complementary A–U interactions within the drug–carrier system successfully conferred high drug entrapment efficiency and improved drug-entrapment stability. As expected, the A-R6G-loaded BU-PEG complexes with a weight ratio of 1:1 had the maximal DLC of 69.04 ± 2.89%, whereas the R6G-loaded BU-PEG complexes with the same mixing ratio only achieved an R6G-loading content of 22.67 ± 3.45% (Table S1). Additionally, the resulting A-R6G-loaded BU-PEG complexes exhibited a wide-range

tunable DLC, and the desired A-R6G-loading content could be achieved by controlling the A-R6G and BU-PEG blending ratio. In contrast to the A-R6G/BU-PEG system, the R6G and BU-PEG complexes prepared using different ratios of both materials exhibited similar DLC values after purification by dialysis (Table S1), indicating that R6G-loaded BU-PEG cannot stably encapsulate R6G due to the lack of complementary A–U interactions within the complexes, thus explaining the relatively low DLC and non-tailorable drug-loading content. In other words, the complementary hydrogen bonding interactions between uracil and the adenine moieties within the complexes increase the affinity and specificity of the nanocarriers for encapsulated hydrophobic drugs, and thus conferred extremely high and tunable DLC.

**Figure 1.** (**a**) Determination of the CMC values of PEG and BU-PEG in water in the presence of pyrene. (**b**) Hydrodynamic particle size of the BU-PEG polymer in water as determined by DLS at 25 ◦C. Surface morphologies of spin-coated BU-PEG polymers obtained by (**c**) AFM and (**d**) SEM at 25 ◦C.

The DLS analysis further demonstrated that A-R6G-loaded BU-PEG (containing 69% A-R6G) and R6G-loaded BU-PEG (containing 23% R6G) displayed mean hydrodynamic diameters of 181 ± 5.97 nm (PDI = 0.461) and 160 ± 5.53 nm (PDI = 0.237), respectively (Figure S3a, Table S1), suggesting that the complexes increased in size to offer a relatively large capacity to accommodate a large number of drug molecules compared with the pristine BU-PEG nanogels. In addition, the *ζ*-potential value of the A-R6G-loaded BU-PEG increased progressively from −38.29 ± 6.44 mV to 51.21 ± 3.66 mV with a gradual increase of DLC. All of the A-R6G-loaded complexes exhibited significantly larger *ζ*-potential values compared to the R6G-loaded BU-PEG and the pristine BU-PEG (Table S1), indicating the *ζ*-potential value of the A-R6G-loaded BU-PEG complexes gradually increased to improve the A-R6G encapsulation efficiency and stability of the complexes. In order to verify the results obtained by DLS, we examined the morphological structure of A-R6G-loaded and R6G-loaded BU-PEG complexes using AFM and SEM. As illustrated in Figure 2a and Figure S3b–d, both systems had nearly spherical shapes with sizes varying between 120–150 nm and 90–130 nm, respectively, consistent with the DLS results (Figure S3a). Over-

all, these findings further revealed that dynamic complexes of BU-PEG with hydrophobic A-R6G suspended in water were successfully constructed due to the complementary A–U interactions, and resulted in formation of a spherical-like A-R6G-encapsulated nanogel with tunable DLC capacity.

**Figure 2.** (**a**) AMF images of A-R6G-loaded BU-PEG nanogels measured at 25 ◦C. (**b**) PL spectra of A-R6G-loaded and R6G-loaded BU-PEG nanogels in water at 25 ◦C. Inset: photographs of A-R6G-loaded and R6G-loaded BU-PEG nanogels in water exposed to (1) natural lighting and (2) broadband UV lighting conditions. (**c**) Time-dependent DLS analysis of the structural stability of A-R6G-loaded and R6G-loaded BU-PEG nanogels in media containing 10% FBS at pH 7.4 for 24 h. (**d**) In vitro hemolytic assay of different concentrations of A-R6G-loaded and R6G-loaded BU-PEG nanogels on SRBCs. Inset: photographs showing SRBCs incubated with varying concentrations (1–100 μg/mL) of A-R6G-loaded or R6G-loaded BU-PEG nanogels at 37 ◦C in 5% CO2 for 4 h.

Due to the intrinsic fluorescence emission of R6G and A-R6G, we further investigated the effect of A–U base pairing on the fluorescence properties of the A-R6G-loaded and R6G-loaded BU-PEG nanogels in water with different DLC values using PL spectroscopy. As shown in the right-upper inset of Figure 2b, when A-R6G and the BU-PEG were blended at a 1:1 weight ratio, the resulting aqueous solution (containing 51% A-R6G) exhibited a strong, bright-green fluorescence under excitation with a broadband UV, while the R6Gloaded BU-PEG solution (weight ratio of 1:1; containing 19% R6G) exhibited relatively weak green fluorescence. These results indicated the complementary A–U interactions within the nanogels dramatically enhanced the fluorescence emission behavior of A-R6G in aqueous solution. A quantitative analysis of the fluorescence enhancement of A-R6Gloaded BU-PEG nanogels in water was conducted by PL measurements with excitation at 480 nm. As shown in Figure 2b, the PL spectra of R6G-loaded BU-PEG nanogels containing 13% and 19% R6G (weight ratios of 1:1 and 0.5:1, respectively) in water exhibited maximum

fluorescence peaks at 555 nm with the same intensity of around 1300, possibly due to the presence of the similar DLC of the nanogels. Interestingly, the maximum PL fluorescence peaks and intensities of the A-R6G-loaded BU-PEG nanogels with various DLC values were remarkably different to those of the R6G-loaded BU-PEG system: the A-R6G-loaded nanogels exhibited a significant blue-shift in their maximum fluorescence peaks from 555 nm to 549 nm, and a gradual increase in the maximum fluorescence intensity from 2740 to 3980 as the DLC increased, suggesting that the physical encapsulation of the A-R6G in the nanogels through complementary A–U interactions efficiently prevented the aggregation of the polycyclic aromatic rings of A-R6G within the nanogels [39,40], resulting in a significant blue-shift of the maximum fluorescence peak and a progressive enhancement in fluorescence intensity with the DLC, even with A-R6G-loading contents as high as 51%. Thus, these findings further demonstrate that the A–U interactions within this complementary drug-nanocarrier system manipulate the drug-encapsulation and fluorescence behavior of the nanogels in aqueous environments.

An ideal nanoparticulate drug-loaded delivery system must maintain high drugentrapment stability in the normal cellular environment to ensure safe, effective delivery of drugs. Therefore, we studied the structural stability of A-R6G-loaded and R6G-loaded BU-PEG nanogels in culture media (DMEM supplemented with 10% serum FBS) and the SRBC hemolysis assay. FBS and SRBCs were employed as nanoparticle-destructuring agents to induce the rapid disassembly of the self-assembled nanogels [34]. As shown in Figure 2c, pristine BU-PEG exhibited a gradual decrease in the mean hydrodynamic diameter from 84 nm to 57 nm after 24 h incubation with FBS-containing DMEM, implying that the self-complementary hydrogen bonding uracil moieties could not preserve the structural integrity of the BU-PEG nanogels in FBS/PBA-mixed medium. After the encapsulation process, A-R6G-loaded BU-PEG nanogels remained at an almost constant mean hydrodynamic diameter after 24 h monitoring, whereas a progressive decrease in the particle size of the R6G-loaded BU-PEG nanogels from 155 nm to 66 nm was clearly observed, indicating the stable complementary A–U interactions between the A-R6G and the BU-PEG complex substantially improved drug-retention stability and prevented initial drug leakage from the nanogels. Similar trends in the improvement of structural stability were obtained in the hemolysis assay. As presented in Figure 2d and the inset photographs, the SRBC hemolytic assay indicated that a broad range of concentrations (1–100 μg/mL) of A-R6G-loaded BU-PEG nanogels did not exert significant hemolytic activity. Even at high concentrations up to 100 μg/mL, the A-R6G-loaded nanogels showed a low hemolytic activity of 4.1%, indicating the excellent compatibility of the nanogels with blood, which is potentially favorable for in vivo applications [41,42]. In contrast to the A-R6G-loaded BU-PEG system, the percentage hemolysis gradually increased as the concentration of R6G-loaded BU-PEG nanogels increased. The percentage hemolysis was up to 21.2% at 100 μg/mL of R6G-loaded BU-PEG, indicating that BU-PEG nanogels cannot maintain their structural stability due to the lack of complementary interactions between BU-PEG and R6G, thus resulting in significant hemolysis. The combination of long-term drugentrapment stability and low-level hemolysis for A-R6G-loaded BU-PEG is an extremely attractive set of features that are rarely observed in traditional drug carrier systems. Therefore, A-R6G-loaded BU-PEG nanogels may represent a potential drug-delivery system that can provide a safe, reliable and efficient delivery of A-R6G within cellular environments.

The above findings prompted us to further evaluate the drug release behavior of A-R6G (or R6G) from BU-PEG nanogels in PBS solutions at different pH values (pH 7.4, 6.5, 6.0) at 25 ◦C using a dialysis method. At pH 7.4, the cumulative release of A-R6G from BU-PEG nanogels was only 30% after 48 h, whereas R6G release of up to 56% was observed from the BU-PEG nanogels, suggesting that the encapsulated A-R6G within the nanogels exhibited high structural stability under normal physiological conditions due to the complementary A–U interactions, thus leading to slow drug release and low cumulative drug release (Figure 3a,b). Interestingly, when the environmental pH was decreased to 6.5 or 6.0, the A-R6G or R6G was released much more quickly than at a

normal physiological pH of 7.4. Both systems exhibited a significant initial burst, with over 50% of the drug released in the first 12 h at either pH 6.5 or 6.0, followed by cumulative release of 94% and 88% after 48 h, respectively, implying that weakly acidic environments can trigger the structural disassembly of both drug-loaded nanogels, and thus induce a rapid drug release rate and high cumulative drug release. Overall, due to the transient A–U interactions within their structure, A-R6G-loaded BU-PEG nanogels exhibit high A-R6G-loading capacity and excellent structural stability that directly inhibits premature drug release and a confers a low drug release rate under normal physiological pH 7.4. However, the A-R6G-loaded BU-PEG nanogels disassembled and release the drug rapidly in mildly acidic conditions, and thereby possessed well-controlled pH-triggered drug release properties. Thus, this self-assembled drug–carrier system based on complementary A–U interactions may represent an attractive and potential route for safer, more efficient in vitro and in vivo delivery and release of medication.

**Figure 3.** In vitro drug release profiles of (**a**) A-R6G-loaded and (**b**) R6G-loaded BU-PEG nanogels. In vitro cytotoxicity of A-R6G-loaded and R6G-loaded BU-PEG nanogels against (**c**) NIH/3T3 cells and (**d**) HeLa cells after 24 h incubation.

Potential drug delivery nanogels must be highly biocompatible and exert low cytotoxicity against normal and cancer cells, and they must also only release the encapsulated drug under specific conditions in the cellular environment. Thus, an MTT-based chromogenic assay was used to quantify the cytotoxic activity of pristine BU-PEG, A-R6G-loaded-, and R6G-loaded BU-PEG nanogels toward normal NIH/3T3 cells and cancerous HeLa and MG-63 cells. As indicated in Figure 3c,d and Figure S4, a range of concentrations of BU-PEG nanogels exerted negligible cytotoxic effects in normal and cancer cells after 24 h, indicating the BU-PEG was highly biocompatible. In contrast, pristine the A-R6G and R6G showed highly potent cytotoxic activities against normal and cancer cells, with half-maximal inhibitory concentrations (IC50) ranging from 1 to 43 μg/mL. Extraordinarily, after culture with A-R6G-loaded BU-PEG nanogels at concentrations up to 100 μg/mL for 24 h, the viability of NIH/3T3 cells remained above 85%, while the R6G-loaded BU-PEG nanogels strikingly reduced the viability of NIH/3T3 cells, with an IC50 value of 40.5 μg/mL (Figure 3c). These observations clearly suggested the complementary A–U interactions within the A-R6G-loaded BU-PEG nanogels significantly improved the structural stability of the nanogels and minimized premature A-R6G leakage under physiological

conditions in NIH/3T3 cells. Conversely, the lack of complementary interactions within the R6G-loaded nanogels led to a significant reduction in NIH/3T3 cell viability. However, both the A-R6G-loaded and R6G-loaded BU-PEG nanogels exerted significant cytotoxic activities against HeLa and MG-63 cancer cells, with remarkable IC50 values of 39.7 μg/mL and 79.3 μg/mL in HeLa cells and 2.8 μg/mL and 9.3 μg/mL in MG-63 cells, respectively (Figure 3d and Figure S4). These results suggest that the pH-induced structural disassembly of both drug-loaded nanogels in the acidic extracellular environment of cancer cells led to rapid drug release, subsequently resulting in selective cytotoxic effects [43,44]. Thus, the weakly acidic extracellular cancer cell environment may facilitate the release of A-R6G within the interior of the cells and thus facilitate selective, potent cytotoxicity toward cancer cells, while reducing the adverse impacts of A-R6G-loaded nanogels in normal cells. Therefore, A-R6G-loaded BU-PEG nanogels could potentially significantly enhance the chemotherapeutic safety and effectiveness of anticancer drugs. However, the exact mechanisms of action of the selective cytotoxicity of A-R6G-loaded BU-PEG nanogels still remain uncertain. We are conducting research to more precisely define the structural features and in vivo cytotoxic effects of this drug–carrier system in order to confirm the selective and targeted cytotoxic effects of A-R6G towards cancer cells for chemotherapy applications.

In order to obtain further insight into the mechanisms of cellular uptake and intracellular drug release by A-R6G-loaded BU-PEG towards normal and cancer cells, NIH/3T3 and HeLa cells cultured with the drug-loaded nanogels for 3, 12, or 24 h, and then analyzed by CLSM to observe the cellular morphology and internalization of the nanogels [45]. Blue-fluorescent DAPI was used to stain the nuclei; A-R6G and R6G exhibit strong green fluorescence emission. As indicated in Figure 4a,b, the CLSM images clearly indicated no significant green fluorescent could be observed after 24 h incubation of NIH/3T3 cells with A-R6G-loaded BU-PEG nanogels. In contrast, remarkable green fluorescence was randomly distributed throughout the cytoplasm of the HeLa cells after 3 h incubation with A-R6Gloaded nanogels, and the green fluorescent signal gradually shifted into the nucleus after 24 h of incubation. In contrast to the A-R6G-loaded BU-PEG nanogels, a gradual increase in the green fluorescent signal and intensity was observed within the nuclei of NIH/3T3 or HeLa cells incubated with R6G-loaded BU-PEG nanogels for 3 to 24 h (Figure S5a,b). These findings are in good agreement with the MTT assay, and suggest the A-R6G-loaded BU-PEG nanogels were selectively internalized by the cancer cells and did not internalize in the normal cellular environment [46]. In contrast, the R6G-loaded BU-PEG nanogels underwent intensive, non-specific uptake by normal and cancer cells, possibly due to specific interactions between the A–U base-paring moieties of the nanogels and the surface of cancer cells. Overall, the above-mentioned findings indicate this supramolecular drug–carrier system containing complementary A–U pairs could promote the selective uptake of drugs by cancer cells and effectively induce cancer cell death while minimizing the cytotoxicity in normal cells.

In order to verify the CLSM images, we performed quantitative and qualitative flow cytometry analysis to further investigate the selective internalization of A-R6G-loaded BU-PEG nanogels by HeLa cells. As shown in Figure 5a,b, after incubation with A-R6Gloaded BU-PEG nanogels for 24 h, NIH/3T3 cells exhibited almost no change in A-R6G fluorescence intensity, whereas the A-R6G fluorescence intensity of HeLa cells gradually increased with the duration of incubation, indicating the complementary A–U base-pairs within the nanogels were endowed with a strong affinity for HeLa cells and thereby promoted selective, rapid internalization of the nanogels and subsequently led to the death of the cancer cells. In contrast, flow cytometry of NIH/3T3 and HeLa cells incubated with R6G-loaded BU-PEG nanogels for various periods of time revealed a gradual increase in R6G fluorescence intensity between 1 and 24 h, suggesting that the R6G/BU-PEG system was not able to selectively promote the internalization of nanogels by cancer cells. In addition, these results also revealed that R6G-loaded BU-PEG nanogels exhibited much higher and faster cellular uptake in HeLa cells than in NIH/3T3 cells (the lower left and right regions of Figure 5), possibly due to the differences in the surface charge and affinity

between the cells. Moreover, we further assessed how the complementary hydrogen bonding interactions between the drug and carrier affected cellular uptake ability by plotting the average fluorescence intensity of the flow cytometry data versus incubation time. As shown in Figure S6, the internalization rate of the A-R6G-loaded BU-PEG nanogels was approximately 74 times higher in HeLa cells than NIH/3T3 cells after 24 h, whereas the internalization rate of the R6G-loaded BU-PEG nanogels was only 5.7 times higher in HeLa cells than NIH/3T3 cells after 24 h, which is in good agreement with the CLSM results (Figure 4 and Figure S5). These results further demonstrate that the A-R6G-loaded BU-PEG nanogels were selectively internalized into the HeLa cells but only minimally taken up by NIH/3T3 cells, whereas the R6G-loaded BU-PEG nanogels were non-specifically internalized by cells. Collectively, these findings clearly prove the complementary A– U interactions within the nanogels not only controlled their drug delivery and release properties, but also promoted the selective uptake of drugs into cancer cells and accelerated cell death; they may thus potentially enhance the overall efficiency of chemotherapy.

**Figure 4.** CLSM images of (**a**) NIH/3T3 cells and (**b**) HeLa cells cultured with A-R6G-loaded BU-PEG nanogels at normal physiological conditions (pH 7.4 and 37 ◦C) for 3, 12, or 24 h. The scale bars in all CLSM images are 20 μm.

To further identify the cytotoxic pathways and assess the mechanisms of cell death for A-R6G-loaded and R6G-loaded BU-PEG nanogels, dual fluorescent staining and flow cytometry were used to quantify viable, dead, and total cells after exposure of NIH/3T3 and HeLa cells to A-R6G-loaded and R6G-loaded BU-PEG nanogels for various periods of time. BV421 Annexin-V was used to detect phosphatidylserine expression on early apoptotic cells, while GDR-780 was used to label intracellular DNA, which is released after the integrity of the plasma membrane has been compromised in late apoptotic cells [47,48]. As shown in Figure 6d,h, after the incubation of HeLa cells with A-R6G-loaded or R6G-loaded BU-PEG nanogels for 12 h, respectively, over 90% of the NIH/3T3 cells incubated with A-R6G-loaded nanogels survived, whereas the R6G-loaded nanogels increased the proportions of early and late apoptotic cells to 27.7% and 29.4%, respectively, indicating that the A-R6G-loaded BU-PEG nanogels had extremely stable drug entrapment stability in a normal cellular environment, which significantly reduced the leakage of hydrophobic A-R6G during the drug delivery process. In contrast to the general trend in NIH/3T3 cells, the overall proportion of apoptotic HeLa cells gradually increased between 1 and 12 h, suggesting that the A-R6G-loaded nanogels taken up by the HeLa cells moved progressively toward the nucleus, and that the A-R6G was gradually released from the nanogels (Figure 6i–l). After

12 h of incubation with A-R6G-loaded nanogels, the proportions of early and late apoptotic HeLa cells dramatically increased to 38.3% and 1.53%, respectively, while the proportion of viable cells still remained high, at 56% (Figure 6l). By comparison, the apoptotic trends in the HeLa cells cultured with R6G-loaded BU-PEG nanogels were similar to the results for the NIH/3T3 cells presented in Figure 6h,p. These observations clearly confirm that the mildly acidic cancer microenvironment promotes the rapid intracellular release of A-R6G from BU-PEG by inducing dissociation of the complementary A–U interactions within the nanogels, and that the A-R6G subsequently promotes programmed cell death. Thus, this newly developed system based on complementary hydrogen bonding A–U interactions between A-R6G and BU-PEG may potentially remarkably enhance the effects of chemotherapy in cancer cells while substantially reducing adverse effects in healthy cells.

**Figure 5.** Flow cytometry histograms of (**a**) NIH/3T3 cells and (**b**) HeLa cells cultured with A-R6Gloaded or R6G-loaded BU-PEG nanogels at normal physiological conditions (pH 7.4 and 37 ◦C) for 1, 3, 6, 12, or 24 h.

**Figure 6.** Flow cytometric dot plot quadrant charts of NIH/3T3 and HeLa cells cultured with A-R6G-loaded and R6Gloaded BU-PEG nanogels at normal physiological conditions (pH 7.4 and 37 ◦C) for 1, 3, 6, or 12 h before staining with BV421 Annexin V and GDR780. Figures (**a**)–(**p**) represent the results of flow cytometry for different time points during the co-culture period. The graph quadrants from the lower left to the upper left (turning anti-clockwise) represent viable cells (BV421 Annexin V- , GDR780−), early apoptotic cells (BV421 Annexin V+, GDR780−), late apoptotic cells (BV421 Annexin V+, GDR780+), and necrotic cells (BV421 Annexin V<sup>−</sup>, GDR780+). The numbers inside each quadrant refer to the proportions of cells.

#### **4. Conclusions**

We successfully developed a complementary drug delivery system based on cooperative hydrogen bonding A–U interactions between a drug and nanocarrier to achieve selective uptake by cancer cells, improve chemotherapeutic efficacy, and reduce adverse

effects in normal cells. A R6G-based anticancer agent containing a hydrogen-bonded adenine unit (A-R6G) was obtained via a simple three-step chemical reaction. A-R6G displays poor aqueous solubility and unique green fluorescent behavior, and exerts potent cytotoxic effects against a variety of cell lines. A complementary hydrogen bonding partner, water-soluble uracil-end-capped BU-PEG polymer, was prepared at high yield (83%) via a one-step Michael addition reaction. BU-PEG can spontaneously self-assemble into well-defined nanoparticles with a variety of unique physical properties in water, such as interesting amphiphilic and morphological characteristics. The formation of A–U base pairs between BU-PEG and A-R6G leads to the formation of well-dissolved A-R6G-loaded BU-PEG nanogels in water. Interestingly, the resulting A-R6G-loaded BU-PEG nanogels exhibited a number of unique physical properties, including extremely high A-R6G-loading capacity (69.4%), a widely tunable A-R6G-loading content, singular green-fluorescence characteristics, and well-controlled pH-responsive drug-release behavior. Moreover, A-R6G-loaded BU-PEG nanogels showed excellent structural stability in cell culture media and low hemolytic activity towards SRBCs. The combination of long-term drug-entrapment stability and low-level hemolytic activity offered by A-R6G-loaded BU-PEG is extremely attractive, but rare in traditional drug–carrier systems. In vitro cytotoxicity studies clearly confirmed that A-R6G-loaded BU-PEG nanogels exhibit potent cytotoxic activity against cancerous HeLa and MG-63 cells, and only minimal cytotoxic effects in normal NIH/3T3 cells, whereas R6G-loaded BU-PEG nanogels did not show selective cytotoxicity against these cell lines. Thus, the complementary A–U interactions critically improve the selective uptake of A-R6G into cancer cells and selectively induce cancer cell death. Importantly, analysis of intracellular cellular uptake using CLSM and flow cytometry clearly demonstrated that A-R6G-loaded BU-PEG nanogels enabled the selective uptake of A-R6G by HeLa cells via endocytosis and promoted controlled intracellular release of A-R6G in the weakly acidic microenvironment of the cancer cells, which subsequently induced programmed cell death in the cancer cells. The opposite trends were observed in NIH/3T3 cells, i.e., poor internalization and extremely low toxicity. Thus, this study clearly demonstrates that introduction of complementary A–U interactions within this drug–carrier system provides an effective approach to selectively delivering anticancer drugs into cancer cells, subsequently improving the safety and effectiveness of chemotherapy, without the need to incorporate targeting moieties onto the surface of drug carriers.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/10 .3390/pharmaceutics13111929/s1, Scheme S1: Process for the synthesis of BU-PEG, Figure S1: (**a**) 1H and 13C NMR spectra of BU-PEG in CDCl3 obtained at 25 ◦C. (**b**) GPC curves for PEG and BU-PEG in THF at 40 ◦C, Figure S2: 1H NMR spectra of BU-PEG, A-R6G, and A-R6G/BU-PEG complexes in D2O obtained at 25 ◦C, Figure S3: (**a**) DLS profiles for A-R6G-loaded and R6G-loaded BU-PEG nanogels in water at 25 ◦C. (**b**) AFM image of R6G-loaded BU-PEG. SEM images of (**c**) A-R6G-loaded BU-PEG and (**d**) R6G-loaded BU-PEG nanogels, Figure S4: In vitro MTT assay cytotoxicity of BU-PEG, A-R6G, R6G, and A-R6G-loaded and R6G-loaded BU-PEG nanogels towards MG-63 cells, Figure S5: CLSM images of (**a**) NIH/3T3 cells and (**b**) HeLa cells cultured with R6G-loaded BU-PEG nanogels at normal physiological conditions (pH 7.4 and 37 ◦C) for 3 h, 12 h, or 24 h. The scale bars in all CLSM images are 20 μm, Figure S6: Fluorescence intensity of NIH/3T3 and HeLa cells after incubation with A-R6G-loaded and R6G-loaded BU-PEG nanogels for different periods of time (1, 3, 6, 12, or 24 h), Table S1: Hydrodynamic particle size, zeta potential, drug-loading content (DLC) and drug-loading efficiency (DLE) of A-R6G-loaded and R6G-loaded BU-PEG nanogels.

**Author Contributions:** Conceptualization, C.-C.C.; investigation, F.B.I. and E.A.B.; writing—original draft preparation, F.B.I.; writing—review and editing, C.-C.C.; supervision, C.-C.C. All authors have read and agreed to the published version of the manuscript.

**Funding:** Ministry of Science and Technology, Taiwan (contract no. MOST 107-2221-E-011-041-MY3 and MOST 110-2221-E-011-003-MY3).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Acknowledgments:** The authors would like to acknowledge funding from the Ministry of Science and Technology in Taiwan for this study (contract no. MOST 107-2221-E-011-041-MY3 and MOST 110-2221-E-011-003-MY3). This study was also partially supported by the Yushan Scholar Program by the Ministry of Education in Taiwan.

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

#### **References**


## *Article* **Interaction of Supramolecular Congo Red and Congo Red-Doxorubicin Complexes with Proteins for Drug Carrier Design**

**Anna Jagusiak 1,\*, Katarzyna Chłopa´s 1, Grzegorz Zemanek 1, Izabela Ko´scik <sup>1</sup> and Irena Roterman <sup>2</sup>**


**Abstract:** Targeted immunotherapy has expanded to simultaneous delivery of drugs, including chemotherapeutics. The aim of the presented research is to design a new drug carrier system. Systems based on the use of proteins as natural components of the body offer the chance to boost safety and efficacy of targeted drug delivery and excess drug removal. Congo red (CR) type supramolecular, self-assembled ribbon-like structures (SRLS) were previously shown to interact with some proteins, including albumin and antibodies complexed with antigen. CR can intercalate some chemotherapeutics including doxorubicin (Dox). The goal of this work was to describe the CR-Dox complexes, to analyze their interaction with some proteins, and to explain the mechanism of this interaction. In the present experiments, a model system composed of heated immunoglobulin light chain Lλ capable of CR binding was used. Heat aggregated immunoglobulins (HAI) and albumin were chosen as another model system. The results of experiments employing methods such as gel filtration chromatography and dynamic light scattering confirmed the formation of the CR-Dox complex of large size and properties different from the free CR structures. Electrophoresis and chromatography experiments have shown the binding of free CR to heated Lλ while CR-Dox mixed structures were not capable of forming such complexes. HAI was able to bind both free CR and CR-Dox complexes. Albumin also bound both CR and its complex with Dox. Additionally, we observed that albuminbound CR-Dox complexes were transferred from albumin to HAI upon addition of HAI. DLS analyses showed that interaction of CR with Dox distinctly increased the hydrodynamic diameter of CR-Dox compared with a free CR supramolecular structure. To our knowledge, individual small proteins such as Lλ may bind upon heating a few molecules of Congo red tape penetrating protein body due to the relatively low cohesion of the dye micelle. If, however, the compactness is high (in the case of, e.g., CR-Dox) large ribbon-like, micellar structures appear. They do not divide easily into smaller portions and cannot attach to proteins where there is no room for binding large ligands. Such binding is, however, possible by albumin which is biologically adapted to form complexes with different large ligands and by tightly packed immune complexes and heat aggregated immunoglobulin-specific protein complex structures of even higher affinity for Congo red than albumin. The CR clouds formed around them also bind the CR-Dox complexes. The presented research is essential in the search for optimum solutions for SRLS application in immuno-targeting therapeutic strategies, especially with the use of chemotherapeutics.

**Keywords:** supramolecular self-assembled ribbon-like structures (SRLS); Congo red (CR); doxorubicin (Dox); bovine serum albumin (BSA); immunoglobulin light chain λ (Lλ); heat aggregated immunoglobulins (HAI); dynamic light scattering (DLS); elution volume (Ve)

**Citation:** Jagusiak, A.; Chłopa´s, K.; Zemanek, G.; Ko´scik, I.; Roterman, I. Interaction of Supramolecular Congo Red and Congo Red-Doxorubicin Complexes with Proteins for Drug Carrier Design. *Pharmaceutics* **2021**, *13*, 2027. https://doi.org/10.3390/ pharmaceutics13122027

Academic Editors: Francisco José Ostos, José Antonio Lebrón and Pilar López-Cornejo

Received: 19 October 2021 Accepted: 22 November 2021 Published: 28 November 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 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 (https:// creativecommons.org/licenses/by/ 4.0/).

#### **1. Introduction**

Drugs designed to reach molecular targets, among which monoclonal antibodies and kinase inhibitors are most frequently used, are the basis of modern therapy. Targeted immunotherapy is also expanded to simultaneous delivery of drugs, including chemotherapeutics. Immuno-targeting, defined as the use of immunological specificity directed to target connected with therapy, is still the subject of many investigations. Design and development of efficient carriers of anti-inflammatory and anticancer drugs are now extensively studied in order to increase the effectiveness and safety of the targeted therapies [1–5].

Self-assembled structures presented in this work are the group of compounds (polyaromatic molecules of an elongated shape with appropriately located polar groups) showing a tendency to self-associate via non-covalent interactions thus creating greater supramolecular systems. This phenomenon is also observed during the formation of microtubules or biological membrane structures. Some of these systems form elongated structures referred to as self-assembled ribbon-like structures (SRLS). These kinds of structure have the potential to be a part of systems delivering chemotherapeutics to cancerous tissue by immuno-targeting.

This is possible because of their ability to selectively interact with immune complexes. SRLS are examples of a novel type of protein ligand, as they bind to proteins via different interactions than the classic type [6]. SRLS systems bind to proteins at sites of local structural instability caused by unfolding conditions or function-derived structural changes in the protein molecule. The binding of SRLS to antigen-antibody complexes, with simultaneous lack of binding of free antibodies, can serve as an example. The described interaction is a foundation for using those compounds in immuno-targeting [7]. At the same time, SRLS systems can intercalate other molecules, including drugs, forming co-micellar systems [8].

Previous research has shown that SRLS can be applied in vivo as potential drug carriers. Such systems were easily bound to immune complexes formed in the body and then were gradually eliminated. Immune complexes are highly complex systems and their structural analysis is difficult. Conformational changes observed in Lλ under subdenaturing conditions, which mimic those in antigen-complexed antibodies, contribute to CR binding. This is the reason why immunoglobulin light chain (Lλ) heated to 45 ◦C was used for the research on interaction between SRLS and antigen-bound antibodies as a model system [9]. Heat aggregated immunoglobulins such as immunoglobulin G (HAI) were investigated as another immune complex model system. One more protein that can be used in targeted therapies, albumin, was also found to bind supramolecular ligands. Some therapeutic agents, especially of cationic nature (like the widely used chemotherapeutic doxorubicin) cannot bind to albumin directly. Thus, albumin can be used in targeted therapy other than in combination with negatively charged carriers. In particular, when SRLS bind drugs, co-micellar structures are formed that can interact with albumin which allows for effective drug delivery and also protects the body against uncontrolled drug action [6,10–16].

The standard example of SRLS is Congo red (CR), which forms assemblages of elongated shape, with a high level of plasticity [6]. CR is a molecule with polyaromatic rings of elongated planar symmetric structures, substituted by amino and sulfonic groups [17,18]. The presence of sulfonic groups makes the structure polyanionic at neutral pH. Symmetrical charge arrangement and hydrophobic interaction between groups in the central part of the molecules, which is noncovalent, make the ribbon-like structure stable but also guarantees its high plasticity [6]. In the presented work, only CR was used as a model because its properties as SRLS are well known. In the future, it can be replaced by other, more biocompatible compounds with similar properties (e.g., Evans blue) [19–21].

The special property of supramolecular CR is that it is able to interact with proteins especially those containing β structure fragments. The mechanism of interaction of supramolecular ribbon-like CR with protein is different from classic protein–ligand interaction in the protein active site. Such complexes can be formed on the condition that protein β-sheet part is at least partially destabilized, which allows for function-derived conformational rearrangement. Thus, susceptibility of CR structures to deformations allowing the best fitting to the protein binding site is the important condition for optimal CR-protein binding. Since there is no defined, specific amino acid sequence that binds CR and the supramolecular ligand is capable of changes due to its plasticity, a variety of proteins are able to form such complexes. Proteins binding to supramolecular CR are native proteins (normal and pathological) as well as proteins treated with denaturing factors [6,22,23].

Examples of native CR binding proteins include molecules in which supramolecular CR binding site arises from function-derived intramolecular strain. Antibodies belong to such group of proteins. In antibodies, antigen binding induces some structural changes [24,25]. Native, antigen-unbound immunoglobulins G do not form complexes with CR but gain this ability upon antigen binding. In the presence of CR, the effect of antibody–antigen interaction enhancement was observed [26]. The property of selective CR binding shown by antigen-bound antibodies gives opportunities for its application in targeted drug delivery [6]. The model system for such interaction was developed as an isolated immunoglobulin light chain (Lλ) heated to 45 ◦C or immunoglobulin G heated to 63 ◦C (HAI). Upon heating, the protein is destabilized. The N-terminus is locally unfolded, which opens up the V domain. The same phenomenon takes place when antibodies are bound to the antigen. At the Fab ends associated with the antigen, the beta structure rich polypeptide is revealed and those structures are penetrated by and bound to CR. CR-Dox co-micelles can bind at the same region. However, it should be underlined that it is just a model system, and the real mechanism of CR-Dox co-micelle binding to immunological complexes might be different than that described as binding to the cavity emptied by the N-terminal chain fragment [9]. To our knowledge, there is no other research on the binding of CR complexed with drugs to heated Lλ or HAI.

Another example of CR-binding native protein is serum albumin. This universal carrier of many hydrophobic compounds (especially anionic) can bind CR as single molecules as well as a supramolecular ligand thanks to its structure with the binding cavity [6,11–16]. Albumin is a protein adapted to the transport of a variety of anionic compounds that can bind CR without preliminary structure change. The molecule of serum albumin binds up to 16 CR molecules and at most nine molecules of Evans blue, which has a similar structure to CR but shows weaker self-association properties [10]. Until now there have been no investigations of albumin ability to bind CR-drug complexes. There have also been no studies on drug transfer from a carrier such as albumin to the immune complex (presented here as the HAI model).

SRLS shows the ability to form complexes with planar molecules including chemotherapeutics, such as doxorubicin [8]. Doxorubicin (Dox) is one of the most effective drugs used in the therapy for many types of cancer [27]. At the same time, it is a highly cardiotoxic drug, and its use can result in the inhibition of hematopoiesis and gastrointestinal disorders [28,29]. Despite this, it is still widely used because of its high effectiveness and wide spectrum of anticancer effects. That is why further investigations of doxorubicin carriers are very important as they can improve the efficiency of its delivery and reduce its toxic effects. Earlier experiments have shown that doxorubicin binds to supramolecular ribbon-like CR structure (Figure 1) [30–33].

Previously published research on the structure of free CR and CR-Dox complex showed large differences between these systems. Positively charged Dox binds to negatively charged CR and the complex migrates faster than free CR towards the anode during electrophoresis [34]. The absorption spectra of CR and its complex with Dox also differ (hypochromic effect). CR-Dox complexes have increased size compared to free CR, which was confirmed using DLS and molecular modeling methods [8]. The fuzzy oil drop model, applied in the previous study, aimed to detect binding sites for supramolecular ligands in albumin (particularly between its pseudo-symmetrical fragments) as well as in V domains, indicated that complexation of dye molecules led to the formation of a stable supramolecular structure, anchored between antibodies that participate in the immune complex [35].

**Figure 1.** Structure of Congo red and Doxorubicin and the supramolecular complex CR-Dox.

The main goal of the present analysis was to compare the capability of drug binding mediated by CR to proteins: albumin, heated light chain, which is a model system of antibody bound to antigen, and antibodies bound to surface-immobilized antigens. Using DLS and molecular sieve methods, sizes of the free supramolecular system (CR) and its complex with a drug (CR-Dox, molar ratio 2:1) were compared at various concentrations of components and ionic strength of buffers. Subsequently, the interaction of free CR and its drug-bound complexes (CR-Dox) with proteins was investigated. The properties of SRLS, particularly their ability to bind drugs, antibodies in immune complexes and albumin justify research into the application of the described systems in targeted therapy (immuno-targeting). This research with the use of carriers specifically interacting with some proteins is important for targeted transport of drugs to the desired parts of the body, where inflammation or neoplastic process continues.

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

#### *2.1. Materials*

Congo red (CR, 96% purity, Aldrich Chemical Company, Inc., Milwaukee WI 53233, USA), doxorubicin hydrochloride (Dox, 98% purity, Sigma-Aldrich, Co., 3050 Spruce Street, St. Louis, MO 63103, USA), bovine serum albumin (BSA, 96% purity, Sigma-Aldrich, Co., 3050 Spruce Street, St. Louis, MO 63103, USA), and immunoglobulin light chain λ dimer (Lλ) were obtained from the urine of a patient with multiple myeloma. After salting out and dialysis it was purified on Sephacryl S300 column (Pharmacia). Immunoglobulin G was obtained from Baxter Healthcare Corporation, Hyland Division Glendale, CA 91203, USA. All other reagents used were of analytical grade and were purchased from commercial sources.

#### *2.2. Methods of CR-Dox Preparation*

CR-Dox complexes of the 2:1 molar ratio were used because in the previously optimized CR:Dox ratio doxorubicin is completely bound to CR. CR-Dox complexes were created by adding 2 volumes of preheated (2 min. at 100 ◦C) Congo red (1.43 mM CR in 0.05 M Tris/HCl buffer pH 7.4, 0.154 M NaCl) to 1 volume of 1.43 mM Dox dissolved in the same buffer. The mixture was incubated for 15 min at room temperature. The cohesion of the CR molecules forming the ribbon-like structure is not high, but in the presence of alkaline doxorubicin complexes are formed and its cohesion increases significantly. The complexes were passed through a Sephadex G-200 column to remove unbound components.

For DLS analysis of the effect of concentration on the hydrodynamic diameters of the analyzed probes, different concentrations were used. The final concentrations of higher concentration probes were: CR (1.43 mM), Dox (0.715 mM), and CR:Dox (molar ratio = 2:1, CR = 1.43 mM, Dox = 0.715 mM). All probes were dissolved in 0.05 M Tris/HCl. pH 7.4 buffer with 0.154 M NaCl.

The final concentration of lower concentration probes were: CR (0.31 mM), Dox (0.15 mM) and CR:Dox (molar ratio = 2:1, CR = 0.31 mM, Dox = 0.15 mM). All probes were also dissolved in 0.05 M Tris/HCl. pH 7.4 buffer with 0.154 M NaCl.

For DLS analysis of the effect of buffer ionic strength on the hydrodynamic radius, different concentrations of NaCl in the buffer were used (0.05 M Tris/HCl, pH 7.4 buffer with 0.154 M NaCl or 0.3 M NaCl). The final concentrations of probes in this experiment were lower: CR (0.31 mM), Dox (0.15 mM) and CR:Dox (molar ratio = 2:1, CR = 0.31 mM, Dox = 0.15 mM).

#### *2.3. Methods of CR and CR-Dox Binding with Protein*

#### Complexes of CR and CR-Dox with Proteins

CR and CR-Dox complexes with three different types of acceptor proteins were analyzed: 1. partly unfolded immunoglobulin chain Lλ; 2. partly unfolded immunoglobulin G (HAI); 3. plasma albumin.

#### 1. The immunoglobulin light chain λ (Lλ)

Partly unfolded Lλ (dimer) was used as a model protein that binds CR or CR-Dox complexes according to the same mechanism as the one observed in antigen-complexed antibodies, where natural structural destabilization is caused by intramolecular constraints evoked by simultaneous interaction with two antigenic determinants [25]. Lλ was isolated from the urine of a patient with multiple myeloma. Complexes were formed by mixing partially unfolded Lλ (obtained by 20 min. incubation at 45 ◦C (sub-denaturing conditions) with a ten-fold molar excess of CR or CR-Dox complexes (2:1 molar ratio). Under such conditions the N-terminal polypeptide loop of the Lλ undergoes local structural destabilization creating the binding site for 4 CR molecules (per monomer) [9].

#### 2. Heat aggregated immunoglobulin G (HAI)

Human immunoglobulins G at a concentration of 10 mg/mL (0.05 M PBS buffer) were heated for 20 min at 63 ◦C. The aggregate dissolves upon the addition of 100-fold molar excess of CR. To remove free or weakly bound dye molecules the protein-dye complex was filtered through Sephadex G-200.

#### 3. Albumin

Albumin is a model system for studying the CR and CR-Dox interaction with a typical carrier protein. Bovine serum albumin (BSA) in 0.05 M Tris/HCl buffer, 0.145 M NaCl, pH 7.4 was mixed and incubated (15 min.) with CR or CR-Dox at 10-fold molar excess of CR to BSA, and 2:1 ratio of CR to Dox.

#### *2.4. Characterization of Free CR-Dox Complexes or CR-Dox Bound to Albumin, Lλ Light Chain, or HAI*

#### 2.4.1. Dynamic Light Scattering (DLS)

Hydrodynamic radii of CR, Dox, and CR-Dox complexes were measured by using the dynamic light scattering (DLS) method (detector Zetasizer Nano ZSP, Malvern, United Kingdom) with laser incident beam at λ = 633 nm and a fixed scattering angle of 173◦. For the measurements, dispersants with the following parameters of viscosity and refractive index were used: (1) Tris/HCl 0.05 M with NaCl 0.3 M; viscosity 0.9208 cP; Refractive Index = 1.334; and (2) Tris HCl 0.05 M with NaCl 0.154 M; viscosity 0.9068 cP; Refractive Index = 1.332. Each measuring probe was incubated inside the instrument (3 min/25 ◦C). A measurement comprised 5–9 repetitions each of which was an average of 15 records measured for 9 s. Outliers were rejected from analysis and the results were averaged.

#### 2.4.2. Gel-Filtration Chromatography (BioGel P-10 and BioGel P-300)

Elution volume (Ve) analysis for fractionation within a different size range was performed on 100 mm Bio®Spin columns (BioRad, Hercules, CA, USA). Columns were filled with 4 mL of BioGel P-10 in 0.01 M PBS buffer, pH 7.4 for supramolecular compound

solutions (CR, CR-Dox). For complexes of CR and CR-Dox with proteins (BSA and Lλ), the BioGel P-300 in 0.01 M PBS buffer, pH 7.4 was used. The flow rate reflects the size of the complex (or the supramolecular entity). 80 μL of the sample was loaded onto the column. Then, it was rinsed with 0.01 M PBS buffer, pH 7.4.

The presence of the protein in the eluate was detected by dot-staining with bromophenol blue, CR concentration was determined spectrophotometrically and Dox fluorometrically.

#### 2.4.3. Gel Electrophoresis

Electrophoresis was carried on 1% agarose plates (in 0.06 M sodium barbital buffer, pH 8.6) at 160 V for about 40 min. The position of CR-complexes was recorded, plates were then fixed with picric acid and the excess of CR was removed by reduction with sodium dithionate followed by staining for protein with bromophenol blue.

#### 2.4.4. TEM

Transmission Electron Microscopy (TEM; JEOL JSM-7500F) was used to evaluate the structure of the HAI-CR complexes. Samples for TEM were prepared by heating the aqueous solution of immunoglobulin G at 63 ◦C for 20 min. After gradual cooling, the CR solution was added (the CR was labeled with AgNO3). The resulting complexes of HAI-CR were purified from the excess of unbound CR using a thin layer chromatography on Sephadex G200. The procedure allows for obtaining well-dispersed HAI loaded with supramolecular CR "clouds". For TEM analysis, 1 μL of suspension was applied to the surface of the copper grid (300 mesh) and dried in vacuum.

#### *2.5. Methods of Analysis of Competition between BSA and HAI for Binding of the CR-RhoB Complexes (Congo Red-Rhodamine B) or CR-Dox (Congo Red-Doxorubicin) Complexes*

Additionally, BSA, BSA-CR complex (10:1), and BSA-CR-Dox (or BSA-CR-RhoB) complexes (CR-Dox or CR-RhoB ratios were 1:1, 2:1, or 5:1) were prepared. To remove excess CR unbound with the triple complex, gel-filtration chromatography on BioGel P-300 medium was used for all systems. HAI was added to a part of the sample to observe the transmission of CR and CR-Dox complex from the initial complexes with BSA.

#### **3. Results**

#### *3.1. Characterization of CR-Dox Co-Micelles*

3.1.1. DLS Analysis of CR-Dox Complexes—The Effect of Concentration on the Complex Size

DLS size analysis revealed well-defined peaks for CR-Dox complexes. The results indicate that the CR-Dox complex is bigger than free CR and free Dox. The size of the CR-Dox complex depends upon the initial concentrations of the components. The samples containing CR-Dox complexes at 2:1 CR:Dox molar ratio were compared with complexes formed at different component concentrations (lower: CR: 0.22 mg/mL, Dox: 0.09 mg/mL and higher: CR: 1 mg/mL, Dox: 0.83 mg/mL). In the case of free Dox and free CR, no effect of concentration on the hydrodynamic diameter was observed—for Dox it was 0.6 nm and for CR 2.3 nm for both concentrations. In the case of the CR-Dox complex, the size increased with increasing concentration, amounting to 3.6 nm for lower to 4.85 nm for the higher concentration (Figure 2).

The above results show that Dox alone does not form supramolecular structures and that the dimensions of the supramolecular CR ribbon are not influenced by the concentration while in the case of CR-Dox complex, the size increases with concentration (for samples with the same proportion of the components).

To explain the above results, it is necessary to clarify how the hydrodynamic diameter is read in the DLS analysis. In the case of spheroidal particles, it is just the diameter, but in the case of supramolecular CR, it is rather the width of the ribbon-like structure it creates in the solution. Both CR and Dox molecules contain charged groups, negative in the case of CR and positive in the case of Dox. As a result, supramolecular ribbons created by CR repel each other and the measured value of the hydrodynamic diameter is the same, even if the length of the ribbon differs, being higher at higher concentrations. A similar effect concerns Dox. In the case of CR-Dox complexes, the electrostatic interactions may stabilize the complex and the effect of electrostatic repulsion between separate ribbon-like structures disappears. This allows the individual ribbons to create more complex, tangled structures, registered in DLS analysis as the ones with higher hydrodynamic diameters.

**Figure 2.** DLS analysis of CR-Dox complexes. The effect of concentration on the hydrodynamic diameter of CR, Dox, and CR-Dox complex. (**A**). lower concentration probes (final concentration of CR: 0.31 mM, Dox: 0.15 mM); (**B**). higher concentration probes (final concentration of CR: 1.43 mM, Dox: 0.715 mM).

#### 3.1.2. DLS Analysis of CR-Dox Complexes—The Effect of Ionic Strength on Complex Size

The effect of ionic strength of the solution on hydrodynamic diameter of CR, Dox, and CR-Dox complex (2:1) was analyzed. Complexes were prepared in 0.05 M tris/HCl buffer with the addition of 0.154 M or 0.3 M NaCl. Higher ionic strength led to significantly increased size of CR-Dox complexes (3.6 vs. 78.82 nm) while it only slightly influenced the hydrodynamic diameter of CR ribbons (2.3 vs. 2.7 nm) and did not affect the results for Dox (0.6 nm for both NaCl concentrations, Figure 3).

To interpret the result, as in the case of the previous experiment, we need to note that in DLS analysis the hydrodynamic diameter value read by the DLS instrument reflects the diameter of the supramolecular ribbon formed by the analyzed molecules in the solution. As the ionic strength increases more charged groups become shielded due to the interaction with salt ions. For free CR the observed effect is rather small, but in the case of the CR-Dox complex, the neutralization of charges promotes the interaction between individual ribbon-like structures which produce bundles characterized by significantly higher hydrodynamic diameter.

**Figure 3.** DLS analysis of CR-Dox complexes. The effect of ionic strength on the hydrodynamic diameter of CR, Dox, and CR-Dox complex. Lower concentration probes (final concentration of CR: 0.31 mM, Dox: 0.15 mM) at two different 0.05 M Tris/HCl buffers supplemented with: 0.154 M or 0.3 M NaCl.

> 3.1.3. Gel-Filtration Chromatography (BioGel P-10): Elution Volume (Ve) of CR, Dox, and CR-Dox

Gel filtration chromatography on BioGel P-10 columns was used to estimate the sizes of the complexes. Elution volumes (Ve) for CR, Dox, and CR-Dox (2:1 molar ratio) complex were compared (Table 1).

**Table 1.** Elution volume for CR, Dox, and CR-Dox complex.


CR flows through the column very slowly and its elution volume is practically 100%, which may suggest adsorption to the BioGel bed. CR-Dox complex migrates faster than Dox alone indicating the formation of stable mixed SRLS created by CR and Dox. CR-Dox elution volume is 13% and Dox elution volume is 40%.

*3.2. Interaction of Free CR and CR-Dox Co-Micelles with Proteins*

3.2.1. Agarose Gel Electrophoresis and Chromatographic Analysis

1. CR-Dox co-micelles form complexes with BSA (BSA-CR-Dox)

Albumin, a universal transporter protein, has the binding site for CR [10]. To check its ability to bind CR-Dox supramolecular ligand, the agarose gel electrophoresis employed as binding for CR-Dox (2:1 molar ratio) changes the net charge of the protein. As seen in Figure 4, the BSA-CR complex migrates faster than BSA (but slower than free CR). The fastest migration was observed for the CR-Dox complex. This can be explained by a strong interaction and high cohesion between molecules, leading to the increased dissociation of CR sulfonic groups and thus more acidic properties and higher electrophoretic mobility. Migration of BSA-CR-Dox complex is similar to that of BSA-CR, but two-dimensional separation (electrophoresis in the direction pointed by the arrow 1, followed by chromatography

of the filter paper replica (arrow 2) reveals the presence of Dox in the complex. This confirms the presence of CR-mediated binding of Dox to BSA. Under the above experimental conditions, formation of stable complexes between BSA and Dox was not observed.

**Figure 4.** Complexes between BSA and CR-Dox (2:1)—replica on filter paper applied to bromophenol blue-stained gel (after agarose gel electrophoresis at pH 8.6); migration towards the anode (+): (**a**) CR, (**b**) BSA, (**c**) BSA-CR complex, (**d**) BSA and Dox (Dox can be seen as migrating towards the cathode (−), pointed by arrow), (**e**) CR-Dox complex, (**f**) BSA-CR-Dox (2:1) complex. The presence of Dox in the complex (BSA-CR-Dox) was confirmed chromatographically. The separation of CR-DOX mixtures was performed by Whatman 3 paper chromatography in butanol:acetic acid:water (5:1:4) solvent. Dox is seen as bright-orange fluorescence. For semiquantitative evaluation, DOX was eluted and the fluorescence was measured (emission signal at 550 nm upon excitation with a 470 nm laser beam) (the picture on the right; arrow no. 1 shows the direction of electrophoresis, and arrow 2 the direction of chromatography)—arrow (**A**) points to Dox released during chromatography from its complex with CR, while arrow (**B**) points to Dox released from the BSA-CR-Dox complex.

This experiment shows that this type of drug can be bound to albumin (BSA) via CR–a model SRLS.

#### 2. CR-Dox co-micelles do not form complexes with Lλ

Conformational changes in the N-terminal loop of the immunoglobulin light chain λ (Lλ) heated up to 45 ◦C create the binding site for free CR [24]. Complexes with a defined stoichiometry (4 CR molecules per L chain monomer) can be visualized by agarose gel electrophoresis.

The possibility of creating complexes between Lλ and CR-Dox co-micelles was analyzed. In agarose gel electrophoresis (pH 8.6), Dox migrates towards the cathode and free CR (a), Lλ chain (b), Lλ-CR complex (c), CR-Dox complexes (e) migrate towards the anode. CR forms a strong complex with Lλ, which migrates towards the anode twice as fast as free protein. For CR-Dox molar ratio 2:1, the formation of CR-Lλ complex is not observed, and all the CR is incorporated in CR-Dox faster-migrating complex, while Lλ migrates at the same speed as free unbound protein as in lane "b". It suggests a strong competition between Lλ and doxorubicin for CR binding. All in all, preferentially the CR-Dox complex (migrating at the front) is formed first. This complex binds all the CR and thus Lλ-CR complex do not assembled, which is confirmed by the same migration distance of the protein as that of CR-free protein (f) (Figure 5).

We conclude that Lλ binding site for SRLS created in the Lλ-chain at 45 ◦C can accommodate CR but not CR-Dox, suggesting that the binding site is specific and the incorporation of Dox changes the supramolecular properties of the CR.

**Figure 5.** The possibility of creating complexes between Lλ and CR-Dox (2:1 molar ratio) was analyzed—replica on filter paper applied to bromophenol blue stained gel (after agarose gel electrophoresis at pH 8.6); migration towards the anode (+): (**a**) CR, (**b**) Lλ, (**c**) Lλ-CR complex, (**d**) Lλ-Dox mixture (Dox can be seen as migrating towards the cathode (−), pointed by arrow), (**e**) CR-Dox (molar ratio 2:1), (**f**) Lλ-CR-Dox (CR-Dox molar ratio 2:1); The absence of Dox in the Lλ-CR-Dox mixture was confirmed chromatographically. Arrow (**A**) points to Dox eluted from CR-Dox complex (fastest migration), arrow (**B**) points to the position of Lλ−CR complex (where Dox is absent).

3.2.2. Gel-Filtration Chromatography (BioGel P-300): Elution Volumes of CR and CR-Dox Complexes with Proteins

Elution volumes (Ve) of protein (BSA, Lλ) complexes with CR and CR-Dox were analyzed using gel filtration chromatography on BioGel P-300 columns. Ve values of free ligands and proteins (CR, Dox, BSA, Lλ), two-component complexes (BSA-CR, Lλ-CR), and three-component (triple) complex of BSA with CR-Dox co-micelle (BSA-CR-Dox) or a mixture of Lλ with CR-Dox co-micelle were compared (Table 2).

A low value of the elution volume for some complexes comparable to that of protein indicates the formation of large and stable co-micelles by e.g., CR and DOX.


**Table 2.** Gel-filtration chromatography: elution volume (Ve) of BSA, Lλ, CR, and Dox in mixtures of BSA-CR-Dox, Lλ-CR-Dox, CR-Dox, BSA-CR, Lλ-CR, and in free BSA and Lλ (Biogel P-300). The molar ratio of CR:Dox in all mixtures was 2:1. Ve DOX free = 2.1 mL (not shown).

Free CR flows very slowly through the column, with tailing, which results from its adsorption to the BioGel. The elution volume of free BSA (Ve BSA free) was 0.7 mL, while for BSA-CR it was Ve BSA-CR = 0.6 mL (both for separately determined CR and BSA). The BSA-CR complex shows a higher flow rate than ligand-free BSA. Simultaneous elution points to the formation of the BSA-CR complex. Elution volume for CR-Dox was Ve CR-Dox = 0.3 mL

(both for separately determined CR and Dox). Simultaneous elution points to the formation of the CR-Dox co-micelle. So, we can conclude that CR-Dox at 2:1 molar ratio forms a stable, fast-migrating, and large complex which shows the migration rate twice as fast as that of BSA-CR and free BSA. The same elution volume for separately determined BSA, CR, and Dox was Ve BSA-CR-Dox = 0.3 mL. Simultaneous elution points to the binding of the CR-Dox co-micelle by BSA and formation of a stable, ternary complex with BSA.

The elution volume of free Lλ (Ve L<sup>λ</sup> free) was 1 mL. In the case of Lλ-CR, the Ve value differs for Lλ and CR (Ve <sup>L</sup><sup>λ</sup> = 0.9; Ve CR = 0.8). The separate elution from the column, but with increased flow rate for CR (as compared to the control, which is free CR), suggests that the complex is formed but dissociates during the flow through the column (which adsorbs CR).

In the sample containing the mixture of Lλ and CR-Dox, different Ve values for CR-Dox co-micelle and Lλ were determined (Ve <sup>L</sup><sup>λ</sup> = 0.8; Ve CR-Dox = 0.4). Separate elution points to the inability of Lλ to bind CR-Dox co-micelle.

Lλ migrates more slowly through the column than BSA due to its lower molecular weight. The differences in flow rates of Lλ-CR complex vs. free protein and free CR were observed. The mixture of Lλ and CR-Dox (2:1) did not produce a ternary complex (as in the case of BSA)–CR-Dox and Lλ were eluted separately from the column with Ve for CR-Dox equal to 0.4 and Ve for Lλ equal to 0.8.

#### 3.2.3. CR Binds to Heat Aggregated Immunoglobulins G

HAI was dissolved with CR contrasted by AgNO3. The formed complexes were isolated using gel filtration chromatography on Sephadex G-200 thin layer. HAI surrounded by clusters of supramolecular Congo red were visible using transmission electron microscopy. HAI are more diffused in the presence of CR and form smaller clusters than HAI aggregated without CR. (Figure 6).

**Figure 6.** Complexes observed using TEM. (**A**). Heat aggregated immunoglobulins (HAI) with CR; (**B**). HAI without CR. Clusters of supramolecular Congo red can be distinguished around antibodies (darker spots). The images of immunoglobulins (brighter spots) are indicated by arrows.

3.2.4. Competition of BSA and HAI for Binding of the CR-RhoB Complexes (Congo Red-Rhodamine B) or CR-Dox (Congo Red-Doxorubicin) Complexes

Albumin has been shown to compete with HAI for binding of the CR-RhoB (or CR-Dox) and that it may eventually donate it to the immune complex model.

BSA-CR-RhoB complexes are formed when CR-RhoB is added to albumin. In electrophoresis, there is a clear difference in the location of BSA band (at the front, running faster) and HAI band (runs much slower, close to the loading site). After the addition of HAI to BSA-CR-RhoB complexes, CR-RhoB dissociates from BSA-CR-RhoB complexes and simultaneous binding of CR-RhoB to HAI is visible (Figure 7). Similar results were obtained with the drug-doxorubicin. These results indicate that albumin can be used as a

carrier that binds other compounds, including CR-associated drugs. Such a drug can target the immune complex.

**Figure 7.** (**a**) HAI, (**b**) BSA, (**c**) BSA-CR, (**d**–**f**) BSA-CR-RhoB (1:1), (2:1), and (5:1) respectively, (**g**–**i**) BSA-CR-RhoB (1:1), (2:1), and (5:1) respectively after adding the same amount of HAI. The data in the graphs compare controls (blues bars) with results (red bars) and represent the mean ± SEM; test Student T. \*\* *p* < 0.01.

A loss of the CR-RhoB complex initially bound to albumin was observed after the addition of HAI in the amount of respectively: 50.6% in the case of the initially added CR:RhoB at molar ratio of 1:1 (compare "d" with "g" on the electrophoresis slab and with (1) on the histogram); 41.7% (for 2:1; compare "e" with "h" on the electrophoresis slab and (2) on the histogram) and 63.3% (for 5:1; compare "f" with "i" on the electrophoresis slab and (3) on the histogram). Complex binding to HAI is visible. The data in the graphs compare controls (blues bars) with results (red bars) and represent the mean ± SEM; test Student T. \* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* <0.001.

#### **4. Discussion**

The design of safe and reliable carriers for drugs is an important goal in the development of new therapies. The safe delivery of drugs using albumin, heat aggregated immunoglobulins, or antibodies as carriers in targeted immunotherapy, is already used as a solution which either supplements or substitutes the hitherto used therapies [36,37].

Earlier studies showed that stable mixed supramolecular assemblies can be formed by CR and other molecules characterized by planar, polyaromatic ring structure [30–33]. Supramolecular structure of Congo red dissociated by heating (80 ◦C) was found to form chaotic, frozen organization after rapid cooling, but it reorganizes upon gradual heating at about 25 ◦C deg forming standard ribbon-like structure and dissociates again above 60 ◦C [38]. In addition, the interaction of CR, as a model supramolecular ligand, with plasma albumin [10], partly unfolded immunoglobulin light chain [9], with HAI [6], and with antigen-complexed antibodies [25] was described.

In this study, the analysis of the CR-drug complex (using DOX as an example) was extended, based on the previously conducted research [8,34,35,39,40]. Attempts were also made to introduce co-micelles formed from Congo red bound to Dox into all analyzed proteins. Moreover, the latest research using oil drop modeling has shown that not only free CR but also its complex with the drug can be bound to albumin. However, only free CR binds to the light chain while the CR-Dox complex is not bound [35].

The presented results provide new information about the properties of the SRLSdrug complexes and the capability of their interaction with proteins, and thus about their possible role in the delivery of drugs. It was shown that the L chain λ forms complexes with supramolecular CR but not with CR-Dox complexes (created at the 2:1 molar ratio). In the case of plasma, albumin complexes can be formed with both CR and CR-Dox. CR-Dox co-micelles were also bound by HAI, and additionally, the transfer of a part of the co-micelle-bound drug from the BSA-CR-Dox complex to HAI was observed.

The interaction of positively charged Dox with negatively charged CR molecules probably changes the regular, ribbon-like architecture of the supramolecular CR assembly. The strong interaction between CR and Dox results in the increased electrophoretic mobility of CR-Dox as compared to CR. The explanation for this phenomenon is the face-to-face alignment of Congo red molecules in supramolecular ribbon-like structures. In the electric field, these systems become uniquely oriented dipoles (due to the delocalization of pi electrons from stacked aromatic rings). Electron relocation affects the polar groups of Congo red, changing their dissociation constants and consequently their charge. Hence, accelerated electrophoretic migration towards the anode upon increasing CR concentration is observed, which is indicative of the concentration-dependent rise in acidity of Congo red. CR-Dox complexes are formed by the intercalation of doxorubicin between the Congo red molecules forming the ribbon. The mechanism of the dipole formation is the same and hence the acceleration of the CR-Dox complex migration towards the anode [34,41].

The CR-Dox complex is probably too large, and perhaps too compact, to enter the binding site in a partly unfolded L chain. This binding site is capable of accommodating the supramolecular CR micelle composed of just 4 CR molecules [6] and requires certain plasticity of SRLS which is a characteristic of CR alone and is limited (or absent) in CR-Dox complex.

Therefore, the full analysis of protein complexes with a large CR-Dox complex could not be achieved using the partly unfolded L chain as a model. Supramolecular ligands, such as CR and CR-Dox can interact with proteins producing complexes in which some ligand molecules interact with the protein directly, while others are bound indirectly, as components of the SRLS. Such type of interaction would be possible in the case of HAI, where CR or CR-Dox could be localized also in-between immunoglobulins. It is thus possible that CR creates a kind of a scaffold structure capable of accommodating CR-Dox, and could work in the same way as in antigen-bound antibodies.

The potential of albumin to complex large-sized ligands (such as the CR-Dox complex) is much greater than that of the light chain model. Albumin, which has a ligand binding pocket, has greater ability to adapt this pocket to large ligands [35]. The structure of the ordered Congo red micelle changes significantly upon intercalation of other molecules (doxorubicin). The linear arrangement of the Congo red tape structure is likely to change, as indicated by the obtained results of the analysis using DLS and molecular filtration.

Albumin is one of the most often used drug carriers approved by the FDA, due to its high biocompatibility, availability, and accumulation in tissues that show high metabolic rate. It can also protect tissues against the harmful effects of the drugs it transports. Nab albumin nanoparticles (American Bioscience) have a diameter of 130–150 nm [42]. Albumin-based drug carriers are neither cytotoxic nor immunogenic, present optimum pharmacokinetics, are biodegradable, and easy to prepare [43–45]. Cancerous tissue presents increased capillary permeability and retention [EPR] and thus is more readily infiltrated by large particles, including albumin-based drug carriers. Here we show that native albumin can bind large ligands and can serve as a transporter for transport of SRLS-drug complexes.

Small molecules, like free Dox, can easily penetrate both healthy and diseased tissues while macromolecules (including albumin-drug conjugates or complexes) can easily infiltrate cancerous tissue but do not pass the endothelium of healthy ones. It increases the selective action of albumin-drug conjugates [46,47].

Albumin-CR-Dox complexes presented in this paper could serve as an alternative drug carrier. Due to the positive charge of Dox, its interaction with albumin is weak, and not observed in experimental systems presented here. Spectroscopy and docking results obtained by Agudelo et al. demonstrated that doxorubicin was able to bind to BSA and HSA via hydrophilic and hydrophobic interactions with more stable complexes created with human serum albumin than with bovine serum albumin. They also showed that drug-protein binding engaged several amino acid groups which were stabilized by a network of hydrogen bonds. The drug-protein interaction changed secondary structure of both bovine and human albumin causing partial destabilization of the protein. It can explain weak binding or no binding between BSA and Dox observed by us earlier [48].

Supramolecular systems, due to a wide range of their reactions with proteins and a variety of functional effects, become a new research tool in biology and pharmacology. The presented systems can be used as an alternative to the Nab technology (American Bioscience), consisting in mixing a hydrophobic drug suspended in an oil phase with an aqueous albumin solution and homogenization of the resulting mixture. The resulting albumin nanoparticles have a drug portion locked inside [42]. However, this technology does not work with hydrophilic, positively charged doxorubicin. The simplicity and speed of complex formation in the proposed system is also an interesting alternative to the previously tested combinations of albumin and doxorubicin [49]. On the other hand, drug delivery via CR that interacts with immune complexes is important due to the enhancing effect described in the literature [6].

Function-dependent Congo red binding to a protein like in the case when complexed with antigen-bound antibodies, consequently has a strong influence on this function. For instance, the affinity of antibodies forming immune complexes markedly increases in the presence of Congo red. Increasing the affinity of antibodies allows for the use of lowaffinity antibodies in immune complexes, and an additional advantage is that an increased amount of drugs can be delivered in this way. The presented system offers a wide range of biomedical applications including drugs delivery to cancer cells.

#### **5. Conclusions**

To conclude, our results confirms that Congo red type supramolecular, self-assembled ribbon-like structures form complexes with the chemotherapeutic agent doxorubicin. CR-Dox are large-sized structures with properties different from the free CR. A model system composed of heated immunoglobulin light chain Lλ capable of CR binding, did not bind CR-Dox complexes. Heat aggregated immunoglobulins (HAI) and albumin were able to bind both free CR and CR-Dox complexes. Additionally albumin-bound CR-Dox complexes were transferred from albumin to HAI upon addition of HAI. This kind of interaction between CR-Dox and the described proteins, may in future become an important therapeutic system with the possibility of targeted drug transport and delivery. Supramolecular ribbon-like CR complexed with doxorubicin is a promising system in the treatment of cancers and may open new avenues for novel treatment strategies.

**Author Contributions:** Conceptualization, A.J. and I.R.; methodology, K.C., G.Z., I.K. and A.J.; software, G.Z. and K.C.; validation, A.J. and I.R.; formal analysis, K.C., G.Z., I.K., A.J. and I.R.; investigation, A.J.; resources, A.J.; data curation, K.C., G.Z., I.K., A.J. and I.R.; writing—original draft preparation, A.J.; writing—review and editing, I.K. and I.R.; visualization, A.J. and R.I.; supervision, A.J. and I.R.; project administration, A.J.; funding acquisition, A.J. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the National Science Centre, Poland, grant number 2016/21/D/ NZ1/02763 and by the Polish Ministry of Science and Higher Education, grant number N41/DBS/000715.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data presented in this study are available on request from the corresponding author.

**Acknowledgments:** Many thanks to Leszek Konieczny for his valuable comments. Many thanks to Barbara Piekarska, Barbara Stopa and to Radoslawa Wrobel for helped edit the manuscript. Many thanks to Dorota Duraczynska from Jerzy Haber Institute of Catalysis and Surface Chemistry, Polish Academy of Sciences for transmission imaging of the samples.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Pharmaceutics* Editorial Office E-mail: pharmaceutics@mdpi.com www.mdpi.com/journal/pharmaceutics

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18

www.mdpi.com

ISBN 978-3-0365-3378-0