Development of Nanosilicate–Hydrogel Composites for Sustained Delivery of Charged Biopharmaceutics

Abstract

Nanocomposite hydrogels containing two-dimensional nanosilicates (NS) have emerged as a new technology for the prolonged delivery of biopharmaceuticals. However, little is known about the physical–chemical properties governing the interaction between NS and proteins and the release profiles of NS–protein complexes in comparison to traditional poly(ethylene glycol) (PEG) hydrogel technologies. To fill this gap in knowledge, we fabricated a nanocomposite hydrogel composed of PEG and laponite and identified simple but effective experimental conditions to obtain sustained protein release, up to 23 times slower as compared to traditional PEG hydrogels, as determined by bulk release experiments and fluorescence correlation spectroscopy. Slowed protein release was attributed to the formation of NS–protein complexes, as NS–protein complex size was inversely correlated with protein diffusivity and release rates. While protein electrostatics, protein concentration, and incubation time were important variables to control protein–NS complex formation, we found that one of the most significant and less appreciated variable to obtain a sustained release of bioactive proteins was the buffer chosen for preparing the initial suspension of NS particles. The buffer was found to control the size of nanoparticles, the absorption potential, morphology, and stiffness of hydrogels. From these studies, we conclude that the PEG–laponite composite fabricated is a promising new platform for sustained delivery of positively charged protein therapeutics.

Graphical Abstract

INTRODUCTION

Biologics have emerged as the new frontier in biopharmaceuticals, but are hampered by unfavorable pharmacokinetics and low bioavailability.¹ Sustained local delivery of biologics is a promising approach to overcome these limitations and enhance the safety and efficacy of biologics.² Typically, polymeric hydrogels are sought for biologic delivery due to their biocompatibility, biodegradability, and tunable properties.³ However, polymeric hydrogels suffer from limitations including initial burst release and minimal control over release rates mostly by modulating network mesh size or degradability.⁴ Sustained release of smaller peptides is also challenging due to the rapid diffusion of these molecules from the hydrogel network.^5,6 Further, loss in protein bioactivity and long-term sustained release is difficult to achieve using polymeric hydrogels.⁷

The addition of nanoparticles to polymeric hydrogels has been shown to improve the retention and delivery of biologics.⁸ Specifically, nanoparticles with a high surface area such as two-dimensional (2D) nanosilicates (NS) have shown potential for sustained and prolonged delivery of biologics.^9-12 Two-dimensional nanosilicates are an emerging class of 2D layered biomaterials with high biocompatibility and surface-to-volume ratio, and are biodegradable.^13,14 NS have commonly been used as a rheological modifier for cosmetics and mining,¹⁵ and more recently in biomedical engineering and pharmaceuticals.^16-19 NS are disc-shaped and have negatively charged surfaces and positively charged edges.²⁰ The high surface area and charge characteristics of NS can be used to electrostatically adsorb a range of biologics including large proteins and small peptides, offering strong potential for a plug-and-play type of delivery device.

The strong electrostatic interactions between NS and proteins can result in sustained delivery of biologics for a prolonged duration (>4 weeks) without burst release.^9-12,21 For example, Cross et al. demonstrated that bone morphogenic protein 2 and transforming growth factor beta could be delivered via electrostatic adsorption onto nanoparticles while maintaining protein stability and activity.²² In addition, NS have also been incorporated into cryogels and hydrogels to form nanocomposite materials with sustained release profiles.^23-26 Further, the addition of NS to polymeric hydrogels results in the formation of injectable shear-thinning hydrogels, which is highly beneficial for minimally invasive delivery as well as three-dimensional (3D) bioprinting.^25,27-29 Furthermore, unlike hydrogels alone, NS–hydrogel composites offer the potential to bind smaller peptide therapeutics for sustained release.^5,30 Importantly, NS–hydrogel composites have been shown to be biocompatible, biodegradable, and noncytotoxic.²⁶

NS have tactoid structures under dry conditions and individual nanoparticles are held together by sodium ions, which must be exfoliated using a hydrophilic solvent. The dispersion buffer can have a significant impact on NS structure and mechanical properties.³¹ However, little is known about the way the dispersion buffer affects NS interactions with proteins and subsequently protein release from NS–hydrogel composites. It is thought that NS–biologic complexes are formed, as evidenced by an increase in hydrodynamic diameter.^31,32 Additionally, little is known about the effect of protein size, charge, and concentration on NS–protein complex formation and subsequent release from NS–hydrogel composites. In one study, Koshy et al. demonstrated that a positively charged chemokine was released from a nanocomposite cryogel more slowly than other negatively charged proteins and that protein release was slowed with an increase in NS concentration.²³ However, the authors focused only on release data and did not investigate the formation of NS–protein complexes or the nature of NS–protein interactions. Such information is critical in determining the applicability of NS–hydrogel composites for protein delivery for specific applications.

In this study, we developed NS–hydrogel composites capable of prolonged and sustained release of proteins. Specifically, laponite nanoparticles, which have a diameter of ~30 nm and a height of ~1 nm, were used. Laponite has a dual-charged nature, offering electrostatic interactions with proteins with negatively charged faces or the pH-dependent positively charged edge of nanoparticles.²⁰ Nanocomposite hydrogels were prepared from poly(ethylene glycol) (PEG) with embedded NS. To observe the influence of NS dispersion buffer on NS properties, four NS dispersion buffers were tested: a physiological buffer (phosphate-buffered saline; PBS), a buffer used to exfoliate laponite particles (sodium pyrophosphate; SPP), a buffer used for hydrogel formation (triethanolamine; TEA), and deionized water (H₂O). Characterization of NS dispersed in these buffers as well as NS–PEG hydrogels and NS–protein complexes was performed using rheology, swelling measurements, dynamic light scattering (DLS), fluorescence correlation spectroscopy (FCS), and release studies. Fluorescence correlation spectroscopy allowed for direct, in situ measurement of protein diffusivities rather than relying upon indirect measurements of diffusion. The influence of protein charge on NS–protein interactions was explored using three model proteins with varying size and net charge: lysozyme (Lys), bovine serum albumin (BSA), and ribonuclease A (RNase). The developed nanocomposite hydrogels show potential as protein delivery devices.

RESULTS AND DISCUSSION

Influence of Buffer on NS Dispersion.

Different buffers were tested with the goal of effectively dispersing NS particles to maximize the surface area available for electrostatic interactions with proteins. In powder form, NS particles aggregate into stacklike tactoid structures held together by shared sodium ions (Scheme 1).²⁰ Four buffers were tested for their ability to disperse these NS tactoids into individual particles: DI water (H₂O), SPP (1 mM), PBS (1×), and TEA (0.3 M in PBS, pH 7.4). Due to its dual ionic nature, SPP is commonly used to de-gel self-gelling grades of laponite when added at a concentration of 1–10% by mass of NS.^20,33 In particular, pyrophosphate anions in SPP shield the NS particle edges, leading to the entire particle having a negative charge.^20,21 Then, sodium cations in SPP loosely associate around the particles, causing mutual repulsion between them. PBS is a common physiological buffer, and TEA is a mild base needed for PEG hydrogel formation.^4,34

Scheme 1. NS Hydrogel Preparation^a.

^aNS was dispersed in a buffer to break up tactoid structures. For protein studies, NS was incubated with protein solution for 30 min to allow NS–protein complexes to form. To form the hydrogel matrix, 4-arm PEG-Ac and PEG-diSH were added and gelation completed within 1 h.

The buffer in which NS was dispersed had a significant effect on NS dispersion (Figure 1). PBS and TEA were ineffective at dispersing NS particles, with large aggregates remaining even after sonication. These aggregates made the NS solutions appear optically opaque (Figure 1A). H₂O- and SPP-buffered NS solutions were optically transparent, indicating good dispersion of NS particles. Visual observations were corroborated with particle size analysis using DLS (Figure 1B). NS dispersed in H₂O and SPP had average diameters of 35 and 55 nm, respectively, which were comparable to the manufacturer’s value of 25–35 nm for individual NS particles. Hence, DI water and SPP were effective in dispersing NS into single particles. On the other hand, PBS- and TEA-buffered NS particles were significantly larger, with average diameters of ~1.5 and 2.0 μm, respectively. Previous studies have also shown micro-sized NS particles when PBS was used as the dispersant.^31,35 This is because TEA and PBS contain relatively high osmotic concentrations compared to DI water and SPP and were not able to displace the sodium atoms within the NS stacks to disperse individual particles.^14,31

Figure 1.

Effect of buffer solution on the dispersion of NS. (A) Visual observation of dispersion of NS (10 mg/mL) in different buffer solutions. (B) NS (1 mg/mL) particle diameter in each buffer analyzed by dynamic light scattering. * indicates statistically significant difference between all samples (N = 4, p < 0.05).

Note that DLS measurements do not consider the shape of NS particles or NS–protein complexes, making the distinguishment between a fully exfoliated NS particle and a stack of a few NS particles unreliable. Currently, the exact geometry, structure, and stoichiometry of NS–protein complexes are unknown, further making it difficult to take shape into account. However, DLS could reliably show the increased NS–protein complex sizes compared to individual NS particles and has also been used successfully by others for NS dispersion measurements.^36,37

Influence of NS Dispersion Buffer on NS–PEG Hydrogel Properties.

Buffers used for NS dispersion also affected the properties of NS–PEG hydrogels (Figure 2). Optical microscope images showed large surface features in PBS- and TEA- buffered NS–PEG hydrogels, while H₂O- and SPP-buffered NS–PEG hydrogels showed smooth surfaces like PEG-only hydrogels (Figure 2A). These surface features were due to the presence of large NS aggregates.

Figure 2.

Influence of dispersion buffer on properties of nanocomposite hydrogels. (A) Optical microscope images of PEG-only (no NS) at 100 mg/mL and nanocomposite hydrogels (100 mg/mL PEG, 10 mg/mL NS dispersed in different buffers) at 10× magnification. Scale bar represents 50 μm. Arrows point to surface features. Shear-rate sweep reveals the difference in shear-thinning behavior between PEG-only (no NS) at 100 mg/mL and NS–PEG (100 mg/mL PEG, 10 mg/mL NS) solutions (B) and hydrogels (C) in each of the NS-dispersion buffers. (D) Rheological measurements of storage modulus (G′) of hydrogels. * indicates statistically significant difference from no NS control; # indicates statistically significant difference from H₂O (N = 3, p < 0.05). (E) Swelling ratio as a function of time as an indirect measurement of hydrogel degradation for PEG-only (no NS, 100 mg/mL) and NS–PEG (100 mg/mL PEG, 10 mg/mL NS) hydrogels for each of the NS-dispersion buffers. Note that while NS was dispersed in the indicated buffer, swelling experiments were performed in PBS.

Because NS is shear-thinning, a shear-rate sweep was used to demonstrate the effectiveness of the NS dispersion buffers first in a PEG solution (Figure 2B) and then in a PEG hydrogel (Figure 2C). According to eq 2, an n value of <1 indicates a shear-thinning behavior, while an n value of ~1 indicates Newtonian flow characteristics. Fully dispersed NS solutions were expected to have smaller n values (close to 0) than poorly dispersed NS solutions, as NS is known to exhibit high shear-thinning behavior, while PEG is known to act as a Newtonian fluid in solution.³⁸ As expected, the PEG-only solution had an n value of 0.892 (Figure 2B), indicating more Newtonian-like characteristics, which is comparable to literature values of 0.96.³⁸ In contrast, all NS–PEG solutions (not crosslinked) had n values of <0.5, indicating shear-thinning behavior. The buffer in which NS was dispersed also affected the n values. Well-dispersed NS in H₂O and SPP had substantially lower n values than solutions containing large NS aggregates in PBS and TEA.

To ensure NS–PEG hydrogels maintained their shear-thinning characteristics, a shear-rate sweep of fully crosslinked hydrogels was performed (Figure 2C). Both PEG-only and NS–PEG hydrogels exhibited shear-thinning behavior and all NS–PEG hydrogels had lower n values than PEG-only hydrogels. A shear-thinning behavior is desired for drug delivery applications with NS–PEG hydrogels, as it allows for hydrogels to be injectable.^39,40 The NS dispersion buffer also affected the n values of the NS–PEG gels, with well-dispersed NS in H₂O and SPP having lower values of n than NS dispersed in PBS and TEA, similar to the behavior in PEG solutions. The NS dispersion buffer also had a significant impact on G′ of the resulting hydrogels (Figure 2D). SPP-buffered NS–PEG hydrogels were 29% stiffer than PEG-only hydrogels, while PBS- and TEA-buffered NS–PEG hydrogels were significantly softer than PEG-only hydrogels (27 and 36% softer for PBS and TEA, respectively). Previous studies have also shown this decrease in G′ for PBS-dispersed NS.³¹ DI water-buffered NS–PEG hydrogels had approximately the same G′ as PEG-only hydrogels.

We suggest that in PBS- and TEA-buffered NS–PEG hydrogels, persisting large NS aggregates affected gel formation by decreasing crosslink density, which in turn led to a decrease in G′ compared to PEG-only gels. For SPP-buffered NS–PEG hydrogels, dispersed individual particles could have reinforced the hydrogel structure by filling in hydrogel pores and displacing water, resulting in a higher G′. We also tested whether the dispersed particles could have participated in secondary interactions with the PEG polymer to create additional crosslinks, which would also lead to a higher G′. However, we saw no significant increase in diameter for NS dispersed in SPP or any of the other buffers and independently incubated with 4-arm PEG-Ac and PEG-diSH for 30 min, indicating no secondary interaction between PEG and NS particles in this time frame (Figure S1). Unlike NS dispersed in SPP, NS dispersed in H₂O did not result in higher G′. We suggest that while H₂O-buffered NS particles were also well-dispersed, particle surface-to-edge interactions could have persisted.²⁰ Such interactions could be disrupting efficient crosslinking, while reinforcing the hydrogel network, thereby countering each other. It is also possible that well-dispersed and hydrated NS particles in H₂O did not significantly affect the hydration of PEG and hence had no effect on G′.

Note that others have observed secondary interactions between PEG and NS when both were dispersed in water and incubated for 24 h. For example, Nelson et al. described interactions between PEG-OH and NS particles, where PEG molecules adsorbed onto the NS surface, increasing the measured layer thickness.⁴¹ Such secondary interactions between acrylated PEG and NS particles have also been utilized by others for the formation of physical gels.⁴² The difference between our work and the studies cited here is that we only used water for NS dispersion, but hydrogels were formed and then swollen in a buffer solution, TEA and PBS, respectively. As discussed earlier, the high ion concentration in those buffers prevented NS dispersion and we suggest it should also hinder secondary interactions between NS and PEG.

Finally, we observed that neither the presence of NS nor the buffer in which NS was dispersed affected NS–PEG hydrogel degradation; PEG-only (no NS) and all NS–PEG hydrogels degraded in ~23 days (Figure 2E). Degradation was indirectly measured by following hydrogel swelling over time. This data implies that protein release was governed mainly by NS–protein interactions, rather than different hydrogel degradation profiles in the presence of NS.

Influence of NS Dispersion Buffer on NS–Protein Complex Formation and Protein Release from NS–PEG Hydrogels.

The ability of NS particles to interact with proteins was affected by the extent of NS dispersion due to the dispersion buffer used (Figure 3). Three model proteins with varying size and charge were for protein studies: bovine serum albumin (BSA), ribonuclease A (RNase A), and lysozyme (Lys) (Table 1). First, we confirmed that particle size measurements were not affected by larger particles falling out of solution; sedimentation of NS–protein complexes was not significant for the time scale of the DLS analysis (Figure S2). DLS particle size analysis revealed that NS formed complexes with all proteins in H₂O-buffered NS solutions; we observed a 3.3-fold, 4.4-fold, and 6.88-fold increase in particle size when BSA, RNase, or Lys were added compared to NS particles only, respectively. Particle diameter did not change following the addition of protein when NS was dispersed in PBS and TEA, indicating no complex formation between proteins and NS particles in those buffers (Figure 3A). This is because PBS- and TEA-buffered NS particles were never fully dispersed and remained in a stacked structure, leaving little surface area for electrostatic interactions with proteins (Figure 3B). When SPP was used as the NS dispersion buffer, charge shielding occurred from dissociated sodium and pyrophosphate ions forming an electrical double layer, leading to minimal RNase and Lys interaction (Scheme 2).²⁰ However, SPP NS–BSA complexes were significantly larger than SPP NS–RNase and SPP NS–Lys complexes, possibly due to contraction of an electrical double layer of Na⁺ ions at higher osmotic pressures, allowing BSA to interact with positively charged NS particle edges. In H₂O-buffered NS solutions, particles were well-dispersed and charge shielding did not occur, leading to strong NS interaction with all proteins, resulting in complex formation.

Figure 3.

Effect of NS buffer on NS–protein interactions. (A) NS and NS–protein complex sizes analyzed with dynamic light scattering. NS solutions contain 10 mg/mL NS, while protein solutions contain 10 mg/mL NS and 1 mg/mL protein. * indicates statistically significant difference (N = 5, p < 0.05). Release profiles of 1 mg/mL BSA (B), RNase (C), and Lys (D) in control (PEG-only) and NS–PEG (10 mg/mL NS) hydrogels, where NS were dispersed in different dispersion buffers. (E) Normalized diffusivity of proteins in PEG-only (no NS) and NS–PEG hydrogels, where proteins were preincubated with NS for 30 min in various dispersion buffers. * indicates statistically significant difference from no NS (N = 6, p < 0.05).

Table 1.

Summary of Model Protein Properties

protein	Rh (nm)	MW (kDa)	pI	De in water at 25 °C (10−6 cm2/s)
BSA	3.5	66.5	4.7	0.59
RNase	2.65	13.7	8.54	1.21
lysozyme	1.89	14.3	11.35	1.12

Scheme 2. Summary of the Effect of NS Dispersion Buffer on NS Dispersion and NS–Protein Complexation^a.

^aThe presented NS–protein structures are based on interpretations of results described here and available literature. (A) In H₂O, NS was completely dispersed into individual particles with negatively charged faces and positively charged edges. (B) In the presence of positively charged proteins such as RNase, proteins electrostatically adsorb to the NS particle faces, forming NS–protein complexes, containing multiple NS particles. (C) Negatively charged proteins such as BSA will preferentially interact with NS particle edges, forming complexes that are typically smaller than those formed with positively charged proteins due to the difference in NS surface area available for binding. (D) In SPP, NS was effectively dispersed due to a peptizing effect caused by the association of pyrophosphate (P₂O₇⁴⁻) and sodium (Na⁺) ions on NS particle edges and surface, respectively. (E) The SPP ion association forms an electrical double layer resulting in a charge shield that prevents adsorption of positively charged proteins. (F) However, osmotic pressure in the form of unassociated ions can compress the thickness of this double layer, thereby allowing positively charged NS particle edges to interact with negatively charged molecules, including negatively charged proteins, forming an NS–protein complex. (G) Neither PBS nor TEA was able to effectively disperse NS particles as tactoid structures persisted due to the relatively high osmotic pressures compared to H₂O and SPP. (H, I) Because of this ineffective exfoliation, NS–protein complexes were unable to form with negatively or positively charged proteins.

While all NS–PEG hydrogels slowed protein release compared to PEG-only hydrogels, H₂O-buffered NS–PEG gels exhibited the slowest release, with only 21% of Lys, 36% of RNase, and 75% of BSA released after 19 days (Figure 3B-D). At 12 days, all Lys were released from SPP-, PBS-, and TEA-buffered NS–PEG hydrogels, with all Lys being released within 7 days from PEG-only hydrogels. After 19 days, RNase was completely released from PEG-only and SPP-buffered NS–PEG hydrogels, and 83% and 77% of RNase was released from PBS- and TEA-buffered NS–PEG hydrogels, respectively. PEG-Only and PBS- and TEA-buffered NS–PEG hydrogels showed complete release of BSA after 19 days, while SPP-buffered NS–PEG hydrogels showed 85% release. The effective diffusivities were quantified using eq 3 (Figure 3E). Protein diffusivity was 1.75 times, 2.8 times, or 4.63 times lower for BSA, RNase, and Lys, respectively, in H₂O-buffered NS–PEG hydrogels compared to PEG-only hydrogels. SPP-buffered NS–PEG hydrogels showed 1.36-fold reduction in diffusivity of BSA compared to the PEG-only group, but there was an insignificant reduction for RNase or Lys. Diffusivities of BSA, RNase, and Lys had similar values in PBS- and TEA-buffered NS–PEG hydrogels as in PEG-only hydrogels.

Using DLS, we confirmed that NS–PEG interactions did not have a significant effect on NS–protein complexation (Figure S3). We further confirmed that the inclusion of the NS dispersion buffer itself (no NS) did not have a significant effect on release profiles (Figure S4). Earlier, we also showed no significant differences in PEG–NS hydrogel degradation as a function of buffer (Figure 2E). Cumulatively, our results indicate that NS–protein interactions were the main factor in modulating protein release.

Note that the buffers used for this study vary in osmotic pressure at the concentrations at which the buffers were used. SPP was used at a concentration of 1 mM (2 mOsm/L) according to the manufacturer’s protocol for use as a dispersing agent, stating that SPP should be added at 1–10% by mass of NS.^20,33 PBS was used at 1× concentration (151.1 mOsm/L) and TEA was used at a concentration of 0.3 M (300 mOsm/L) for fabrication of PEG hydrogels. We chose to use the buffers at their typical concentrations rather than modulating their concentrations for maximum efficiency. To explore whether NS dispersion efficiency and NS–protein interactions were affected by buffer ionic strength, buffer ionic strength was modulated for SPP, PBS, and TEA, and the change in NS–protein complex diameter was measured, where Lys was used as the model protein (Figure S5). For each buffer, as the buffer concentration decreased (via 10-fold and 100-fold dilutions), the diameter of the formed NS–Lys complex increased. We suggest that this was due to decreased osmotic pressure and ionic strength in the diluted buffers, leading to more similarities with the properties of NS dispersed in H₂O. The trend of increased complexation in decreasing osmotic pressure was apparent for all NS dispersion buffers, though the varying slopes between buffers could indicate an impact of the specific ions present, rather than osmotic pressure alone.

H₂O was chosen for NS dispersion for all further experiments since it efficiently dispersed NS particles and facilitated NS–protein complex formation for all proteins tested prolonging protein release (Scheme 2). Further, unlike NS dispersed in other buffers, NS dispersed in H₂O did not significantly affect hydrogel stiffness G′. NS–PEG hydrogels made with H₂O-dispersed NS also had the lowest n of 0.015 compared to all other hydrogels, indicative of a shear-thinning behavior.

Effect of Incubation Time on NS–Protein Complex Formation.

To determine when NS–protein complexes reach equilibrium size, DLS was used to measure NS–protein complex size as a function of time. NS particles were dispersed in H₂O as indicated above. Positively charged RNase and Lys or negatively charged BSA (Table 1) were incubated with NS for 0–60 min and the effective particle diameter was measured as a function of time (Figure 4A). The proteins formed complexes when incubated together with NS, while NS particles remained dispersed in the absence of a protein, with a consistent diameter of ~40 nm. NS–BSA complexes reached an equilibrium size of 105 nm after 10 min of incubation, while NS–RNase complexes were 1.5 times larger, reaching an equilibrium diameter of 160 nm after 12 min of incubation. NS–Lys complexes reached an equilibrium diameter of 285 nm after 20 min of incubation, a size that was 2.7 times greater than NS–BSA complexes and 1.8 times greater than NS–RNase complexes. To assure equilibrium in NS–protein complex formation, a 30 min incubation time was adopted for all experiments unless otherwise noted. We suggest that the absorbed proteins mediate the interaction between multiple NS, leading to large complex sizes.

Figure 4.

Effect of NS–protein incubation time on NS–protein complex size and release from PEG hydrogels. (A) Complexation of NS (10 mg/mL) and protein (1 mg/mL) as a function of time as measured by DLS. Release profiles (B) and normalized diffusivities (C) of Lys, BSA, and RNase (1 mg/mL) in NS–PEG hydrogels (10 mg/mL NS) with incubation (solid lines) and without incubation (dotted lines) of NS with the protein prior to hydrogel encapsulation. NS particles were dispersed in H₂O buffer. * indicates statistically significant difference (N = 3, p < 0.05).

Incubating NS and protein solutions prior to PEG gel encapsulation allowed NS–protein complexes to form, providing another lever to sustain protein release from the resultant NS–PEG hydrogels. Release profiles of NS–PEG hydrogels revealed that incubation had a significant impact on the release rate of Lys, BSA, and RNase (Figure 4B). Without incubation, Lys was completely released within 7 days, and 97% BSA and 69% RNase were released after 19 days. When protein and NS solutions were incubated for 30 min prior to PEG hydrogel encapsulation, protein release was substantially slowed, with 21, 84, and 36% of encapsulated protein released after 19 days for Lys, BSA, and RNase, respectively. The release profiles were then quantified based on eq 3 to obtain diffusion coefficients (Figure 4C). Lys experienced an 8.6-fold decrease, BSA a 0.9-fold decrease, and RNase a 2.8-fold decrease of effective diffusivity when NS and protein were incubated prior to PEG gel encapsulation compared to no incubation. Positively charged Lys and RNase were more affected than BSA because of their ability to electrostatically interact with the faces of NS particles, while negatively charged BSA would interact with the edge of NS particles, which have substantially lower surface area.

Influence of Protein Type on NS–Protein Complex Formation and Protein Release from NS–PEG Hydrogels.

Both size and charge of the protein played a role in NS–protein complex formation and protein release from PEG-only and NS–PEG hydrogels (Figure 5). When each protein was incubated with H₂O-buffered NS, solutions became more opaque compared to the clear NS dispersion, due to the formation of NS–protein complexes (Figure 5A). DLS size analysis showed that the more positively charged a protein was, the larger the NS–protein complex that formed, due to electrostatic interactions between NS and the protein (Figure 5B). Lys, RNase, and BSA showed 70-, 30-, and 12-fold increases in size, respectively, when incubated with NS compared to the protein only in solution.

Figure 5.

Effect of protein type on NS–protein complex size and release from PEG hydrogels. (A) Visual observation of complexation of NS (10 mg/mL) with Lys, BSA, and RNase (2 mg/mL). (B) Diameter of NS only (no protein) and NS–protein complexes measured via DLS. NS concentration was 1 mg/mL and protein concentration was 1 mg/mL. * indicates significant difference (N = 6, p < 0.05). (C) Release profiles of BSA, RNase, and Lys from PEG-only (dashed lines) and NS–PEG (10 mg/mL NS, solid lines) hydrogels. (D) Normalized diffusivity of proteins in PEG-only (no NS) compared to NS–PEG hydrogels. * indicates statistically significant difference (N = 6, p < 0.05). (E) Diameter of NS only (no protein) and NS–protein complexes as a function of pH. NS (100 μg/mL) and protein (50 μg/mL) were incubated for 30 min prior to DLS measurements. Vertical dashed lines represent isoelectric points of BSA (blue; pI = 4.7), RNase (red; pI = 8.54), and Lys (green; pI = 11.35).

As expected, for inert PEG-only hydrogels, protein release was inversely correlated with protein size and independent of protein charge.^43,44 The largest protein BSA was released the slowest, while the smallest protein Lys was released the fastest (Figure 5C). As anticipated for hydrogel matrix delivery devices, all proteins exhibited an initial burst release. Burst release, an initial large bolus release of an encapsulate protein, can be caused by protein adsorption to the hydrogel surface or by nonuniform pore sizes.⁴⁵ Such a high initial release is undesired for sustained release applications, as it is often unpredictable, lessens the biotherapeutic delivery device lifetime, and can lead to localized protein concentrations above the therapeutic window. In NS–PEG hydrogels, the burst release was decreased, and all proteins were released slower than from PEG-only hydrogels. Unlike PEG-only hydrogels, protein charge dominated the release rate, as positively charged Lys and RNase were released slower than negatively charged BSA. Similar release results were seen by Kim et al. in bisphosphonate-modified hyaluronic NS hydrogels and were attributed to the surface area and charge characteristics of NS particles.²¹ The normalized diffusivities for Lys, RNase, and BSA were 10.0, 3.2, and 1.6 times lower in NS–PEG hydrogels compared to PEG-only hydrogels, respectively (Figure 5D). As expected, the formation of larger NS–protein complexes led to decreased diffusivities.

While proteins varied in size and electrostatic charge, we suggest that electrostatic charge was the predominant factor in modulating NS–protein complexation. To further confirm this, we followed the change in NS–protein complex diameter as a function of buffer pH (Figure 5E). For each protein, a significant drop in NS–protein complex size was observed when the buffer pH exceeded the protein pI. This result was expected, as the net charge of the proteins became negative at buffer pH > protein pI. Due to the dual charge characteristics of NS particles, negatively charged proteins are more likely to interact with the smaller edges of NS particles than with the larger negatively charged faces, leading to smaller NS–protein complex sizes. NS diameter was slightly higher at pH 3 compared to all other pH and remained stable between pH 6 and 12.

Note that despite the mostly negative charge of NS particles, negative proteins were still able to be adsorbed. This can be attributed to binding to the edge of nanoparticles as well as surface patch binding, in which the anisotropic charged surface of protein allows protein to bind “across its pH”.⁴⁶ For example, BSA contains regions where the surface charge is more positive due to acidic residues (Figure S6), potentially enabling these regions to interact with the negatively charged surface of NS particles.

Influence of NS Concentration on NS–Protein Complex Formation and Diffusivity.

To further investigate our hypothesis that the slowed protein release was due to the electrostatic interaction between proteins and NS, the diffusivity of a fixed concentration of protein (1 μg/mL) was determined at different concentrations of NS using fluorescence correlation spectroscopy (FCS) (Figure 6). FCS is a sensitive technique that, by measuring the spontaneous intensity fluctuations of a fluorescent solute within a femtoliter volume, allows direct determination of diffusion coefficients. Diffusivity is dependent on conformational changes, chemical reactions, electrostatic interactions, and crowding.^47,48 Experiments were performed using BSA, RNase, and Lys labeled with Atto 655 and varying concentrations of NS. Consistent with our hypothesis, protein diffusivity was inversely correlated with the concentration of NS, as evidenced by the rightward shift of autocorrelation curves (Figure S7) documenting the interaction between proteins and NS. A two-component fit (eq 6) was used for all three proteins, indicating the presence of two diffusing species: NS–protein complexes (slow diffusivity) and free protein (fast diffusivity).

Figure 6.

Diffusivity of NS–protein complexes with varying NS concentrations. (A) Diffusivity of 1 μg/mL BSA, RNase, and Lys in varying NS concentrations compared to protein in water. (B) Change in fraction of protein bound to NS with varying NS concentrations. (C) Change in fluorescence brightness per molecule with varying NS concentrations. (D) Diameter of NS–protein complexes at varying concentrations of NS with 1 μg/mL protein. * indicates significant difference between all groups (N = 3, p < 0.05).

For bulk release experiments, an NS concentration of 10 mg/mL was used to allow for substantial interaction between NS and proteins while preventing NS from undergoing self-gelation at concentrations above 25–35 mg/mL.⁴⁹ However, FCS measurements were performed at lower concentrations of NS to prevent bias due to sedimentation of large NS–protein complexes. At high concentrations of NS, large NS–protein complexes and NS aggregates led to sharp changes in the measured spontaneous fluorescent intensity. These intensity spikes affected the autocorrelation function, as could be seen in the residuals at longer correlation times. Above 1000 μg/mL NS, these intensity spikes prevented useful data from being obtained, so lower NS concentrations were studied.

Equation 8 was used to calculate the diffusion coefficients of BSA, RNase, and Lys in NS solutions, which were then normalized to the diffusion coefficients of the free protein in water. Our results show that the diffusivity of negatively charged BSA was less hindered than that of the positively charged RNase and Lys, especially at lower NS concentrations (Figure 6A). At the highest concentration of 1000 μg/mL NS used, there was no significant difference between the diffusivities of different proteins. Overall, RNase diffusivity decreased by 76% (compared to free RNase in PBS) even at the lowest NS concentration of 0.1 μg/mL and changed only slightly with further increase in NS concentration, reaching almost 100% decrease at NS concentration of 1000 μg/mL. Lys diffusivity was significantly hindered at all NS concentrations, with an 86% decrease at NS concentration of 0.1 μg/mL and almost 100% decrease at 1000 μg/mL NS (compared to free Lys in PBS). On the other hand, BSA diffusivity decreased by only 10% for the lowest NS concentration (compared to free BSA in PBS), but also reached an almost 100% decrease at the highest NS concentration used. We suggest that the ~100% decrease in BSA diffusivity was attributable to the excess NS used, where NS was able to both adsorb ~75% BSA in solution (Figure 6B) and form larger NS–protein complexes (Figure 6D) as discussed below. We later show that ~100% diffusivity decrease for BSA was not achieved when BSA was in excess of NS (Figure 7), further demonstrating that the excess NS was needed due to the small surface area of the edges to which BSA electrostatically adsorbs. Overall, the decrease in diffusivity was attributed to the formation of NS–protein complexes, with larger complexes leading to slower protein diffusion.

Figure 7.

Diffusivity of NS–protein complexes with varying protein concentrations. (A) Diffusivity of varying concentrations of BSA, RNase, and Lys in 10 μg/mL NS compared to protein in water. * indicates BSA significantly different from RNase and Lys at the given protein concentration (N = 3, p < 0.05). (B) Change in fraction of protein bound to NS with varying protein concentration. (C) Change in fluorescence brightness per molecule with varying protein concentration. (D) Diameter of NS–protein complexes at varying protein concentrations with 10 μg/mL NS. * indicates significant difference between all proteins (N = 3, p < 0.05).

To quantify the fraction bound at each NS concentration, a two-component fit (eq 6) was used for data analysis, one component for NS–protein complexes (slow diffusivity) and the other component for free protein (fast diffusivity). Using this method, we found that RNase and Lys were able to bind more effectively to NS particles as compared to BSA (Figure 6B) due to the positive charge of RNase and Lys. Even at the lowest NS concentration of 0.1 μg/mL, a bound fraction of ~0.70 for RNase and Lys was achieved. RNase and Lys bound fraction leveled off at 0.90 for NS concentration of 10 μg/mL and higher. The bound fraction of BSA increased much more rapidly than RNase or Lys, starting from 0.25 at NS concentration of 0.1 μg/mL and reaching 0.75 at NS concentration of 1000 μg/mL. BSA binding to NS did not level off for the tested NS concentration range. These trends corroborated the results from Figure 6A, where RNase and Lys diffusivity showed minimal change above 10 μg/mL NS, while BSA diffusivity continued to drop.

Molecular brightness was also measured using FCS (Figure 6C). Molecular brightness refers to the brightness of each diffusing molecule—the more fluorescent proteins present per NS–protein complex, the higher the complex brightness would be. For all three proteins, as NS concentration increased, so too did the brightness per molecule. This was expected because an increase in NS concentration led to larger NS–protein complexes, which means that more fluorescent proteins were present in these larger diffusing complexes. NS–Lys complexes showed the highest brightness, while BSA showed the lowest brightness, which was expected due to the formation of larger complexes for positively charged proteins such as Lys than for negatively charged proteins such as BSA.

DLS size analysis supported FCS results, as NS–RNase and NS–Lys complex sizes were significantly larger than NS–BSA complexes for all NS concentrations (Figure 6D). NS–RNase, NS–Lys, and NS–BSA complexes all increased in size as NS concentration increased. The NS–RNase and Lys complex diameters increased by ~1.4-fold and ~1.5-fold, respectively, for the tested concentration range, while the NS–BSA complex size increased by ~3.7-fold. These results indicate that higher NS concentrations allowed for more protein binding, leading to larger NS–protein complex sizes. Based upon the NS dimensions from the manufacturer (30 nm diameter, 1 nm thick for a single NS particle), the surface area of the two faces of the NS particle should be 1413 nm², while the surface area of the edges should be only 94 nm². The faces of NS are negatively charged, making them more likely to electrostatically interact with positively charged proteins such as RNase and Lys. Conversely, the edges of NS particles are positively charged, leading to electrostatic interaction with negatively charged proteins such as BSA. BSA is also considerably larger than RNase and Lys (Table 1), so individual BSA molecules occupy more surface area per protein molecule than RNase or Lys proteins. These factors could explain why RNase and Lys molecules interacted with NS particles more so than BSA.

Influence of Protein Concentration on NS–Protein Complex Formation and Diffusivity.

To determine if there was a saturation point at which additional proteins could no longer electrostatically interact with NS particles, FCS experiments were performed with varying protein concentrations and a fixed NS concentration of 10 μg/mL (Figure 7). For negatively charged BSA, diffusivity decreased with an increase in protein concentration up to 100 μg/mL protein. However, above this protein concentration, diffusivity dramatically increased, leading to a leftward shift of the autocorrelation curve for 1000 μg/mL of protein (Figure S8), possibly indicating competition between BSA bound to NS and free BSA, which is known to oligomerize at high protein concentrations above ~600 μg/mL (Figure S9).^50,51 For positively charged RNase and Lys, as protein concentration increased, so did the diffusion time, indicating slower diffusion.

Normalized RNase diffusion coefficient rose significantly, but only slightly above 1000 μg/mL of protein (1.4-fold increase from 100 to 2000 μg/mL of protein) (Figure 7A). Normalized Lys diffusivity was greatly slowed even at the lowest protein concentrations and showed relatively little increase in normalized diffusivity above 100 μg/mL. Conversely, BSA diffusivity increased substantially above 100 μg/mL of protein (5.6-fold increase in normalized diffusion coefficient from 100 to 2000 μg/mL of protein). Once the saturation point was reached, the fraction of bound protein decreased, occurring at around 1000 μg/mL for RNase and Lys and 100 μg/mL for BSA (Figure 7B). The bound fraction of RNase and Lys remained above 0.80 for all tested protein concentrations, while the bound fraction of BSA dropped precipitously from 0.51 to 0.12 above protein concentration of 100 μg/mL.

For all three proteins, molecular brightness decreased as total protein concentration increased. This was due to the fraction of fluorescently labeled protein to unlabeled protein decreasing, which means it was less and less likely that a fluorescent protein would be adsorbed to the measured complex. Further, with an increase in complex size, it was expected that labeled protein would get “buried” inside the opaque complex with fewer fluorescent proteins visible on the surface of the complex and available for detection by fluorescence. Note that, as explained in the Materials and Methods section, we increased the protein concentration by adding nonfluorescent protein as not to saturate FCS detectors. NS–Lys and NS–RNase complexes saw a 16-fold and 12-fold decrease in fluorescence brightness per complex, while NS–BSA complexes showed an 8-fold decrease over the tested range of protein concentrations.

These results were corroborated using DLS, which showed NS–RNase and NS–Lys complexes continued to increase in size as protein concentration increased (3.6-fold increase from 1 to 2000 μg/mL for RNase and 3.2-fold increase for Lys) (Figure 7D). NS–BSA complexes increased in size at low protein concentrations but maintained a consistent diameter of 110 nm at protein concentrations above 100 μg/mL, indicating saturation of particle surfaces at that BSA concentration. For all tested protein concentrations, NS–RNase and NS–Lys complexes were larger than NS–BSA complexes, indicating that more RNase or Lys proteins were able to interact with a single NS particle compared to BSA. Saturation was expected to occur at lower protein concentrations for BSA than for RNase or Lys due to the two-dimensional geometry and surface charge of NS particles.

We used the data from Figures 6 and 7 to understand the relationship between NS–protein complex diameter and complex diffusivity as well as bound protein fraction (Figure S10). Overall, an increase in NS–protein complex diameter led to a decrease in complex diffusivity and an increase in bound protein fraction. NS–protein complex diffusivity was proportional to 1/r_s as expected according to the Stokes–Einstein equation (eq 4) and corroborating other studies.^52-54

Our results show great promise for the use of NS–PEG nanocomposite hydrogels for sustained release of protein therapeutics, minimizing burst release and sustaining protein release for a longer period compared to hydrogels without NS. Here, we provided detailed analysis on the effect of: (i) NS dispersion buffer, (ii) NS concentration, (iii) NS–protein incubation time, (iv) protein type and charge, and (v) protein concentration on NS–protein complex formation, size, and diffusivity and protein release from nanocomposite NS–PEG hydrogel. Such detailed knowledge presents a useful lever to control protein release kinetics. While here we focused on proteins, it is likely that NS particles will similarly form complexes with small molecules or peptide drugs, offering great potential for delivery of a wide range of therapeutics.

Further, understanding NS–protein interactions can inform the design for a variety of devices other than therapeutic delivery devices, including tissue engineering scaffolds, cellular delivery devices, and additive manufacturing technologies. For example, Shi et al. showed the presence of NS improved osteogenesis for bone tissue engineering.⁵⁵ Thakur et al. used nanocomposite hydrogels for stem cell delivery due to nanocomposite hydrogel’s shear-thinning and degradation behaviors.⁵⁶ Nanocomposite hydrogels have also been implemented in wound dressings and hemostats due to NS mechanical, degradation, and molecule-binding properties.^40,57,58 NS have been used in additive manufacturing to improve physical and mechanical material properties, including shear-thinning behavior, stiffness, and degradation.^42,59,60

It is of note that the work here was focused on the interaction between a single protein and NS particles in a solution or inside a hydrogel. Competitive or reversible adsorption in protein mixtures was not studied. In the in vivo environment, NS–protein complexes may increase in size due to the presence and adsorption of a variety of soluble proteins and small molecules.^61,62 Conversely, proteins in the physiological environment may also serve to dissociate bound proteins from NS particles, leading to more rapid release from nanocomposite hydrogels.⁶³ Future experiments will be performed to observe if such competitive adsorption is observed, or if proteins from the microenvironment otherwise alter the structure, size, and degradation rates of NS–protein complexes.^64,65 Other groups have shown that controlled release of proteins from NS particles in vivo is similar to release in vitro, including Kim et al., who demonstrated protein release from nanocomposite hydrogels in a murine model and Cross et al., who showed sustained release of osteogenic differentiation factors.^21,22 Therefore, PEG–NS nanocomposite hydrogels should remain effective devices for protein delivery in an in vivo environment.

CONCLUSIONS

Water (H₂O) was found to be the most effective buffer for dispersing NS nanoparticles, while PBS and TEA failed to disaggregate NS particles. SPP dispersed nanoparticles but prevented protein–NS interactions. The variance in dispersion effectiveness and particle characteristics affected both solution and hydrogel properties. Protein charge was implicated as the main determinant for NS–protein interactions via electrostatic adsorption. Positively charged proteins interacted more strongly with NS particles forming larger NS–protein complexes than did negatively charged proteins, due to the charge and surface area characteristics of NS particles. The larger NS–protein complexes formed with positively charged proteins led to hindered diffusion and slower protein release from NS–PEG hydrogels compared to negatively charged proteins. However, for all proteins, NS–PEG hydrogels showed significantly slower release profiles than PEG-only hydrogels, showing the utility of nanocomposite hydrogels for therapeutic delivery applications.

MATERIALS AND METHODS

Materials.

4-arm PEG-Acrylate (4-arm PEG-Ac; 10 kDa) and PEG-dithiol (PEG-diSH; 3.4 kDa) were acquired from Laysan Bio Inc. (Arab, Al). Laponite XLG particles (disc-shaped, ~30 nm × 1 nm) abbreviated here as NS for nanosilicates were purchased from BYK Additives (Wesel, Germany). Fluorescent dye removal columns and Atto 655 NHS ester were obtained from Thermo Scientific (Waltham, MA). CoverWell perfusion chamber gaskets and silicone spacers were purchased from Grace Bio-Labs (Bend, OR). Bradford protein assay reagent, bovine serum albumin (BSA), ribonuclease A (RNase), lysozyme (Lys), sodium pyrophosphate (SPP), phosphate-buffered saline (PBS), and triethanolamine (TEA) were procured from Millipore Sigma (Saint Louis, MO).

Fabrication of Hydrogels.

Hydrogels were formed by Michael-type addition of PEG-diSH onto 4-arm PEG-Ac. A 200 mg/mL stock solution of 4-arm PEG-Ac and a 200 mg/mL stock solution of PEG-diSH were prepared separately in 0.3 M TEA (pH 7.4) immediately prior to hydrogel fabrication. For NS–PEG nanocomposite hydrogels, NS powder was first dispersed in buffer (DI water, 1 mM SPP, 1× PBS, or 0.3 M TEA as noted) and probe sonicated for 30 s using a Sonic Dismembrator 100 (Fisher Scientific, Hampton, NH) in a 1 s on/1 s off pattern (Scheme 1). This NS solution was then added to the 4-arm PEG-Ac precursor solution at a final concentration of 10 mg/mL (15% v/v in buffer) NS. For PEG–NS hydrogels containing proteins, NS (15% v/v in buffer) and protein (15% v/v in buffer) were mixed gently for 30 min prior to addition of 4-arm PEG-Ac. Finally, a 100 mg/mL PEG hydrogel was made by combining the 4-arm PEG-Ac and PEG-diSH stock solutions in a 1:1 molar ratio of Ac/SH groups. The final hydrogel precursor solution was mixed well and pipetted between two glass slides covered with Parafilm and separated by 1 mm silicone spacers. The solutions were left at room temperature for 1 h to complete gelation.

Characterization of Hydrogel Swelling and Mesh Size.

PEG hydrogels with and without NS were formed as described above in a slab geometry with a final volume of 50 μL per gel. The initial mass, M₀, of each gel was measured using a Mettler Toledo XS104 Balance (Columbus, OH). Hydrogels were then incubated in PBS for 24 h at 37 °C. Gels were removed from the PBS and excess liquid was removed by gently patting with a Kimwipe. The swollen mass, M_S, of gels was then measured as a function of time and the swelling ratio was calculated as M_s/M₀. An increase in the swelling ratio over time was used as an indirect measure of hydrogel degradation.⁶⁶

Dynamic Light Scattering.

All dynamic light scattering (DLS; Zetasizer Nano ZS, Malvern, Westborough, MA) measurements were performed at a backscattering angle of 173° and a laser wavelength of 633 nm. The sizes of NS-only particles and NS–protein complexes as a function of buffer type were measured via DLS. For NS-only particles, NS powder was dispersed via sonication in each buffer (DI water, SPP, PBS, and TEA) at a concentration of 10 mg/mL and immediately analyzed to avoid bias from sedimentation. For NS–protein complexes, NS powder was first dispersed in each buffer (DI water, SPP, PBS, and TEA) at a concentration of 12.5 mg/mL. Solutions of BSA, RNase, and Lys were prepared at 5 mg/mL in PBS. Protein properties are summarized in Table 1. Protein and NS solutions were combined by gentle mixing to give a final concentration of 10 mg/mL NS, 1 mg/mL in protein, and 80% v/v in NS buffer and incubated at room temperature for 30 min. The solutions were again gently mixed prior to DLS measurements to avoid bias from sedimentation. Sedimentation was not noted for 30 min after mixing (Figure S2), indicating that sedimentation should not be influencing DLS measurements of diameter (DLS measurements take ~5 min). Corroborating our data, others have shown that NS suspensions were stable over a period 30 days.¹⁴ Absorbance measurements for sedimentation study were performed at a wavelength of 500 nm using SpectramMax i3 spectrophotometer (Molecular Devices, San Jose, CA). Solutions were mixed and incubated for 30 min, then gently shaken prior to analysis to mimic the protocol used for DLS analysis. Centrifugation was performed at 2000g for 30 s.

To investigate possible interactions between NS and PEG, NS (0.5 mg/mL) was mixed with either 4-arm PEG-Ac (0.1 mg/mL) or PEG-diSH (0.034 mg/mL) in H₂O or SPP. Solutions were incubated for 30 min before DLS measurements. To observe the effect of NS–PEG interactions on NS–protein complexation, NS and 4-arm PEG-Ac or PEG-diSH were incubated for 30 min, then Lys (0.1 mg/mL) was added and incubated for an additional 30 min before DLS measurements. To investigate the effect of buffer pH on NS–protein complex size, NS (100 μg/mL) was incubated with BSA, RNase, or Lys (50 μg/mL) for 30 min in DI water of pH 3, 6, 7.4, and 12 and measured via DLS. To determine the effect of buffer ionic strength on NS–protein complex size, 10-fold and 100-fold dilutions of SPP, PBS, and TEA were made by diluting with DI water. The osmolarity of undiluted SPP, PBS, and TEA were calculated as 2, 152, and 300 mOsm/L, respectively. NS (2 mg/mL) and Lys (0.05 mg/mL) were incubated for 30 min in each buffer prior to DLS measurements.

To determine the effect of NS concentration on NS–protein complex sizes, NS powder at varying concentrations, dispersed in DI water via sonication, was incubated for 30 min with BSA, RNase, or Lys (final concentration of 0.1–1000 μg/mL for NS and 1 μg/mL for BSA, RNase, and Lys). To determine the effect of protein concentration on NS–protein complex sizes, BSA, RNase, or Lys were incubated for 30 min with NS dispersed in DI water (final concentration of 10 μg/mL for NS and 1–2000 μg/mL for BSA, RNase, and Lys). DLS measurements were performed immediately following incubation.

Rheological Measurements.

To measure hydrogel mechanical properties, 350 μL of hydrogels (100 mg/mL PEG) without NS and with NS (10 mg/mL) dispersed in DI water, SPP, PBS, or TEA were prepared and swollen in PBS for 24 h at 37 °C. Following swelling, hydrogels were cut into 1 × 20 mm slabs using a custom cutter. Storage modulus, G′, and loss modulus, G″, were measured with an AR 2000ex Rheometer (TA Instruments, New Castle, DE) as a function of angular frequency from 1 to 10 rad/s and 2% strain at room temperature.

Viscosity measurements were conducted for both NS–PEG hydrogels and NS–PEG solutions at shear rates of 10⁻³–10² s⁻¹ at 25 °C. NS concentrations of 10 mg/mL and PEG concentrations of 100 mg/mL were used for both hydrogels and solutions. For hydrogels, both 4-arm PEG-Ac and PEG-diSH were added, while solutions contained only 4-arm PEG-Ac to avoid crosslinking. Using the Ostwald–de Waele relationship, values of the flow consistency index, K, and the flow behavior index, n, were obtained as⁶⁶

$$\eta =K{\stackrel{.}{\gamma }}^{n-1}$$ (1)

where η is the measured viscosity in units of Pa·s and ϒ is the shear rate in s⁻¹.

Protein Release Studies.

NS–PEG hydrogels loaded with proteins were prepared as shown in Scheme 1. Stock solutions (5 mg/mL) of BSA, RNase, and Lys were made by dissolving each protein in PBS. These solutions were then gently mixed with NS (25 mg/mL) dispersed in buffer (DI water, SPP, PBS, and TEA as noted) on a rocking platform at room temperature for 30 min unless otherwise indicated. The NS–protein solutions were added to PEG hydrogel precursor solutions to achieve a final concentration of 10 mg/mL in NS, 1 mg/mL in protein, and 100 mg/mL in PEG and mixed thoroughly. To form hydrogel slabs, 30 μL volumes of the prepared solutions were pipetted between two Parafilm-lined glass plates separated by 1 mm silicon spacers and left to gel for 1 h. Protein-loaded NS–PEG hydrogels and control protein-loaded PEG-only hydrogels (no NS) were then soaked in PBS (1 mL) and placed on a shaker platform at 37 °C for release studies. At specified time intervals, 300 μL of release buffer was collected and replaced with fresh PBS. The releasates were stored at 4 °C for up to 3 weeks prior to analysis. Protein concentration in the releasates was quantified using the colorimetric Bradford Assay following the manufacturer’s protocol. Absorbance was measured at 595 nm with a SpectraMax i3 plate reader (Molecular Devices, San Jose, CA). A mass balance was performed to calculate the total mass of released protein at each time point as

$$\frac{M_i}{M_{i\text{nf}}}=C_{i}V+\sum C_{i-1}{V}_{\mathrm{s}}$$ (2)

where M_i is the concentration of protein released at time i, M_inf is the concentration of protein at infinite time, M_i/M_inf is the fractional release, C_i is the concentration of protein in the releasate at time i, V is the total volume of the release solution, and V_s is the releasate sample volume.

Calculation of Effective Diffusion Coefficient.

Using a modified form of Fick’s Law, the effective diffusion coefficient was calculated for short release times from⁶⁷

$$\frac{M_i}{M_{\text{inf}}}=2{\left[\frac{D_{\mathrm{e}}t}{\pi {\delta }^2}\right]}^{1∕2}$$ (3)

where D_e is the effective diffusion coefficient, t is the time, and δ is half of the hydrogel thickness (0.5 mm). Hydrogel slabs were ~10 mm in diameter and 1 mm thick. The effective diffusion coefficient was then normalized by the protein diffusivity in water, D₀, at 37 °C, calculated using the Stokes–Einstein equation

$$r_{\mathrm{s}}=\frac{k_{\mathrm{B}}T}{6\pi \eta D_0}$$ (4)

where r_s is the hydrodynamic radius of the protein, k_B is the Boltzmann’s constant, T is temperature, and η is the viscosity of water.

Fluorescence Correlation Spectroscopy.

Fluorescence correlation spectroscopy (FCS; Microtime200, PicoQuant, Berlin, Germany) was used to measure protein diffusivity directly in solution as a function of NS presence and concentration. To do so, BSA, Lys and RNase were first fluorescently labeled with Atto 655 NHS ester, according to the manufacturer’s protocol with 56, 58 and 64% labeling efficiency, respectively. Unbound dye was removed using dye removal columns with >95% removal efficiency. Fluorescent solutes (Atto 655, BSA, Lys and RNase) were dissolved in PBS and incubated for 30 min with NS dispersed in DI water. For all FCS measurements, fluorescent solute concentrations were 0.2 nM for Atto 655 (fluorophore only control), 15 nM for Atto 655-labeled BSA, 70 nM for Atto 655-labeled Lys and 73 nM for Atto 655-labeled RNase. Protein–NS solutions (40 μL) were placed in perfusion chamber gaskets adhered to #1.5 coverslip and capped to avoid evaporation during FCS measurements. Fluorescent solutes without NS were used as control. To observe the effect of NS concentration on NS–protein complex diffusivity, NS concentration was varied between 0.1 and 1000 μg/mL, while fluorescent solute concentration was kept constant. To observe the effect of protein concentration on NS–protein complex diffusivity, protein concentration was varied between 1 and 2000 μg/mL, while NS concentration was kept constant at 10 μg/mL. To avoid fluorescence saturation, fluorescent protein concentration was kept constant at 15 nM for BSA, 70 nM for Lys and 73 nM for RNase, while nonfluorescently labeled protein was added to increase the total protein concentration.

For FCS measurements, first Atto 655 (in DI water) was used to calibrate the confocal volume (1.096–1.245 fL) of the FCS instrument. A 640 nm ps pulsed laser was used at an optical power of ~11.4 μW for at least six measurements of 180 s for each sample. An autocorrelation function G(τ) was obtained for each measurement⁴⁸

$$G(\tau )=\frac{1}{N}\frac{1}{\left[1+\left(\frac{\tau }{{\tau }_{\mathrm{D}}}\right)\right]}\frac{1}{{\left[1+p\left(\frac{\tau }{{\tau }_{\mathrm{D}}}\right)\right]}^{0.5}}$$ (5)

where N is the number of fluorescent particles, p = r_o/z_o is an instrumental constant, r_o is the radius and z_o is the axial length of the focused laser beam spot, and τ_D is the solute diffusion time. For two noninteracting, diffusing solutes, eq 6 can be rewritten as^68,69

$$\begin{gathered} G(\tau )=1+m_1\frac{1}{\left[1+\left(\frac{\tau }{{\tau }_1}\right)\right]}\frac{1}{{\left[1+p\left(\frac{\tau }{{\tau }_1}\right)\right]}^{0.5}} \\ +m_2\frac{1}{\left[1+\left(\frac{\tau }{{\tau }_2}\right)\right]}\frac{1}{{\left[1+p\left(\frac{\tau }{{\tau }_2}\right)\right]}^{0.5}} \end{gathered}$$ (6)

where m₁ and m₂ are related to the quantum yield and the average number of each diffusing species and τ₁ and τ₂ are their respective diffusion times.

Here, we used the two-component autocorrelation function for all NS–protein solutions to capture the different diffusion times of free protein and NS–protein complexes. To do so, τ_D for free protein in DI water was measured separately (and determined from the single component fit in eq 6) and used as a fitted parameter (τ₂) in eq 7 to determine the diffusion time of the NS–protein complex (τ₁). Additionally, the autocorrelation function was fit using a Triplet model to account for the possible excitation of molecular triplet states at higher laser intensities. Finally, the autocorrelation function was normalized as follows

$$\text{Normalized}G(\tau )=\frac{G({\tau }_{\mathrm{D}})}{G({\tau }_0)}$$ (7)

where G(τ_D) is the value of eq 7 at each time point and G(τ₀) is the value of eq 7 at the initial time point. The effective tracer diffusion coefficient for each protein in solution was calculated from τ_D as⁶⁹

$$D_{\text{FCS}}=\frac{(r_0{)}^2}{4{\tau }_D}$$ (8)

Statistics.

The results of experiments are the mean values (±standard deviation) of at least three individual experiments of three to eight samples each. Comparisons between multiple groups were performed using single-factor analysis of variance (ANOVA) followed by Tukey’s post hoc test. Comparisons between two groups were performed with two-tailed Student’s t-tests and were considered statistically significant when p < 0.05. Statistical tests were performed with GraphPad Prism software (San Diego, CA).

Visualization of Anisotropic Charges of Proteins.

To visualize the anisotropic surface charges of Lys, BSA, and RNase, computer models of the proteins were generated using RasMol Software.⁷⁰ Protein models were downloaded from the Research Collaboratory for Structural Bioinformatics Protein Data Bank (RCSB PDB). The built-in functions were used to select all acidic and all basic residues, highlighting them red and blue, respectively (Figure S6).