|Home | About | Journals | Submit | Contact Us | Français|
The rational design of immunoprotective hydrogel barriers for transplanting insulin-producing cells requires an understanding of protein diffusion within the hydrogel network and how alterations to the network structure affect protein diffusion. Hydrogels of varying crosslinking density were formed via the chain polymerization of dimethacrylated PEG macromers of varying molecular weight, and the diffusion of six model proteins with molecular weights ranging from 5,700 to 67,000 g/mol was observed in these hydrogel networks. Protein release profiles were used to estimate diffusion coefficients for each protein/gel system that exhibited Fickian diffusion. Diffusion coefficients were on the order of 10−6 to 10−7 cm2/s, such that protein diffusion time scales (td = L2/D) from 0.5 mm thick gels vary from 5 minutes to 24 hours. Adult murine islets were encapsulated in PEG hydrogels of varying crosslinking density, and islet survival and insulin release was maintained after two weeks of culture in each gel condition. While the total insulin released during a one hour glucose stimulation period was the same from islets in each sample, increasing hydrogel crosslinking density contributed to delays in insulin release from hydrogel samples within the one hour stimulation period.
Many cell and tissue replacement therapies are limited by the lack of appropriate delivery vehicles for cells and engineered tissues. Specifically, the regeneration or replacement of insulin-producing cells in patients with type 1 diabetes presents a unique challenge. Strategies for successfully transplanting adult islets or undifferentiated islet precursor cells are dependent upon the synthesis of a controlled transplantation environment that promotes cell survival, function, and, when appropriate, differentiation, while protecting transplanted cells from immune mediated destruction. Biocompatible, synthetic hydrogels present a promising platform for designing clinically relevant insulin-producing cell carriers with three-dimensional structures capable of supporting entrapped cell morphology and with tunable network structures for controlling the diffusion of molecules to and from entrapped cells (1–3). Because the permeability of many synthetic hydrogels is readily altered, these systems have been extensively studied for the delivery of cells with specific secretory functions (4, 5), including pancreatic islets (6–8). The necessity to understand and control diffusion in hydrogels designed for these applications is clear. For the delivery of adult islets and islet precursor cells alike, the importance of transport regulation extends beyond relatively low molecular weight insulin and high molecular weight immune cell secreted antibodies to many other proteins of intermediate size such as growth factors critical to cell survival or differentiation and remains independent of the micro- or macroencapsulation design.
Poly(ethylene glycol) (PEG) hydrogels formed via the photoinitiated polymerization of multifunctional macromolecules have shown promise for islet encapsulation, specifically with respect to in vitro biocompatibility (8). The ability to readily alter PEG network properties to influence the diffusion of a limited number of proteins has been demonstrated (6). However, a complete understanding of solute diffusion within PEG hydrogels requires an improved understanding of the gel network structure. Recent studies have demonstrated that PEG gels formed from the chain polymerization of functional macromers are not adequately described by classical hydrogel theories. The classic depiction of hydrogel network structure includes linear polymer chains connected by crosslinking points, occupying no volume, as shown in Figure 1A (9). The average mesh size of the network (ξ) is thus dependent on the length of the polymer chains between crosslinks and is used to calculate and predict solute diffusivity in the gel (10). In contrast, crosslinks within networks formed from the homopolymerization of high molecular weight divinyl macromers, such as dimethacrylated PEG, are not adequately described as single points. During polymerization, radicals propagate through the carbon-carbon double bonds of the methacrylate end groups to form polymethacrylate kinetic chains, and the crosslinks in this network are linear PEG molecules extending from each repeat unit of the kinetic chain to those of additional chains. In an ideal network, crosslinking density is low, and crosslinking molecules have negligible dimensions. However, the homopolymerization of divinyl macromers leads to networks with relatively high crosslinking densities, and crosslinking molecules substantially contribute to the structure and chemistry of the network. The addition of a crosslink dimension further complicates the concept of network mesh size. The idealized view of gels formed from dimethacrylated PEG macromers (Figure 1B) attempts to account for the crosslink dimension within the classical hydrogel depiction, representing these networks as homogeneous distributions of PEG crosslinks and polymethacrylate kinetic chains. However, the relative hydrophobicity of the polymethacrylate kinetic chains compared to the hydrophilicity of the PEG crosslink chains likely leads to the formation of complex structures in aqueous solution. Recent reports have proposed a PEG network structure composed of randomly coiled polymethacrylate chains with emanating, extended PEG chains (Figure 1C) (11, 12). This proposed network structure is supported by small angle neutron scattering (SANS) characterization of chain polymerized PEG hydrogels and their respective precursor solutions (13). In this description of gel structure, a gel mesh size defined by the distance between crosslinks is replaced by a characteristic length that represents the distance between polymethacrylate core chains (12).
Because of the inability of classical theories to accurately capture the structure of PEG gels and the paucity of data related to diffusion measurements in PEG gels, this work used an experimental approach to systematically investigate the diffusion of model proteins in gels formed from the chain polymerization of PEG macromers. The release of proteins of varying molecular weight (5,700 to 67,000 g/mol) from PEG gels formed via the photopolymerization of varying molecular weight PEG macromers (2,000 to 10,000 g/mol) was followed experimentally to provide insight into the diffusion of molecules with biological relevance in the application of these gels for insulin-producing cell delivery. These results also provide valuable experimental information for future efforts to develop theoretical relationships that more accurately describe solute diffusion in hydrogel structures formed via chain polymerization of macromolecular monomers. Finally, isolated murine islets were encapsulated in PEG networks formed from varying molecular weight macromers to investigate any direct effects of hydrogel formation or crosslinking density on encapsulated islet survival and the time scale and rate of insulin secretion in vitro.
Poly(ethylene glycol) dimethacrylate (PEGDM) was synthesized by reacting linear PEG (n = 2000, 4000, 6000, 8000, and 10000 g/mol) (Sigma, St. Louis, MO) with methacrylic anhydride (Sigma) at a molar ratio of 1:10 via microwave irradiation under solvent free conditions as previously described (11). Macromer product was collected by precipitation into chilled (4°C) ethyl ether (Sigma) and vacuum filtration, and macromer purification was achieved by dialysis in deionized water (diH2O) using cellulose ester dialysis tubing with a molecular weight cutoff of 1000 g/mol (Spectrum Laboratories, Rancho Dominguez, CA). Purified PEGDM was collected by lyophilization and stored at 4°C under nitrogen. Percent methacrylation was determined using 1H NMR by comparing the area under the integrals for the vinyl resonances (δ = 5.7 ppm, δ = 6.1 ppm) to that for the PEG backbone (methylene protons, δ = 4.4 ppm). Percent methacrylation for all macromers was approximately 95 ± 5%.
Hydrogels were formed from a precursor solution of 10 wt % PEGDM in phosphate buffered saline (PBS, pH 7.4, Gibco) and 0.025 wt % of the photoinitiator 2-hydroxy-1-[4-(hydroxyethoxy)phenyl]-2-methyl-1-propanone (Ciba-Geigy, Basel, Switzerland) exposed to 365 nm ultraviolet light at an intensity of ~7 mW cm−2 for 10 minutes. For protein release experiments, disc-shaped hydrogels were formed with a diameter of 10 mm and thickness of 0.4 mm to achieve a high diameter to thickness ratio (~25) and allow for the approximation of one-dimensional diffusion in the z-direction by assuming radial diffusion to be negligible. Hydrogel samples were swollen in PBS at 37°C for 24 hours and weighed to obtain the equilibrium swollen mass, Ms. Samples were then placed in deionized water to remove PBS salts, frozen, and lyophilized overnight, and the dry polymer mass, Md, determined. The volumetric swelling ratio, Q, was calculated from the mass swelling ratio, (Ms/Md), using density conversion factors (2), and the theoretical concentration of crosslinkable double bonds in the hydrogel precursor solution (mol/L) was used as a measure of hydrogel crosslinking density.
Hydrogels were incubated in 1 mg/ml protein solutions for 24 hours at 4°C for uniform loading of the following proteins: insulin, myoglobin, trypsin inhibitor, carbonic anhydrase, ovalbumin, and bovine serum albumin (Sigma). For protein diffusion coefficients on the order of 10−6 to 10−8 cm2/s, the time scale for loading into 0.4 mm thick gels ranges from 5 minutes to 12 hours. Therefore, 24 hour incubation times should be sufficient to achieve equilibrium protein concentrations within the gels. The diffusion time scales observed in similar studies of various model proteins through even denser polymer networks further support the rationale for a 24 hour loading time period (6, 14). For protein release, loaded gels were placed in PBS at 37°C and transferred to fresh solutions after 6, 12, 30, 60, and 90 minutes to maintain near sink conditions, and the amount of protein released into each solution was measured with the MicroBCA™ Protein Assay Kit (Pierce, Rockford, IL). Standard solutions of each protein were prepared immediately prior to gel loading and maintained in the same conditions as those used for gel loading and protein release until assay measurements were collected.
To determine the diffusion coefficients of each protein with respect to changes in the hydrogel crosslinking density, the measured protein release profiles were fitted to the following solution for diffusion through a sheet with uniform initial concentration and equal surface concentrations:
Here, Mt is the amount of solute that has diffused out of the sheet at some time t; M∞ is the amount after time equals infinity; l is the sheet thickness; and D is the diffusion coefficient of the given solute within the sheet (15). This solution assumes solute diffusion follows Fick’s second law, and comparisons between experimental release profiles and those predicted by this solution were made to comment on Fickian versus non-Fickian protein diffusion within the PEG hydrogels.
Islets from adult Balb/c mice were obtained from the Diabetes and Endocrinology Research Center at the Barbara Davis Center for Childhood Diabetes (Denver, CO). Briefly, after digestion of pancreatic tissue with collagenase, islets were isolated by density gradient purification and hand picked under a microscope. Isolated islets were cultured in RMPI 1640 (Gibco, Carlsbad, CA) supplemented with 10% fetal bovine serum (Gibco), 1% penicillin-streptomycin (Gibco), and 0.5 μg/mL fungizone (Gibco) at 37°C in humid conditions with 5% CO2. For each macroencapsulation sample, approximately 20 islets were suspended in 30 μl of hydrogel precursor solution prior to photopolymerization and entrapped throughout the resulting hydrogel disc upon gel network formation. Swollen, islet-containing hydrogels were approximately 4 mm in diameter and 1 mm in thickness with individual islets scattered throughout.
A membrane integrity assay, LIVE/DEAD®, from Molecular Probes, Inc. (Eugene, OR) was used to evaluate cell viability. Encapsulated islets were placed in the LIVE/DEAD® staining solution for 10 minutes at 37°C. After staining, live cells fluoresced green, and dead cells fluoresced red. Stained samples were visualized within the three-dimensional hydrogel environment using confocal laser scanning microscopy. Quantification of LIVE/DEAD® staining results is difficult in these samples, because clear delineations between individual live, stained cells are rare. However, when few dead cells are observed, stained islet viability can be estimated by comparing the number of dead cells to an approximate number of live cells based on the islet area within the fluorescent image (i.e., 100 cells per 100 μm2). Previous studies employed LIVE/DEAD® staining in observing encapsulated islet viability, because this technique does not require the destruction of the surrounding encapsulation barrier as common metabolic activity assays such as the MTT assay would (8, 16, 17).
Insulin secretion was evaluated by exposure of encapsulated islets to static glucose stimulation for one hour on specified days following hydrogel formation. Encapsulation samples were first placed in 1 ml of low glucose concentration solution (1.1 mM) for 45 minutes, followed by incubation in 1 ml of high glucose concentration buffer (16.7 mM) for one hour. The insulin concentration in the high glucose buffer solutions after one hour was measured by a mouse/rat insulin ELISA (Mercodia, Winston Salem, NC). The CellTiter-Glo® Luminescent Cell Viability Assay (Promega, Madison, WI) was used to measure the ATP content of each encapsulation sample. Islet-containing hydrogel samples were incubated in 0.5 ml culture media combined with 0.5 ml CellTiter-Glo reagent for 30 minutes on an orbital shaker (~200 rpm). The luminescence of each sample solution was measured using a microplate reader (Perkin Elmer Wallac Victor2, 1420 Multilabel Counter), and the insulin released from each sample was normalized by the ATP content of the respective sample to account for insulin secretion disparities between samples due to variations in the number of encapsulated cells.
For observation of insulin release within a one hour high glucose (16.7 mM) incubation period, samples were transferred to fresh high glucose solutions (250 μl each) at 2, 5, 10, 15, 20, 30, 45, and 60 minute time points within the one hour static stimulation period. The insulin content of each solution was measured by ELISA, and cumulative insulin secretion was plotted as a percentage of the total amount secreted from islets in hydrogels formed from each molecular weight macromer.
PEG hydrogels with varying crosslinking densities were synthesized using 10 wt % solutions of dimethacrylated PEG macromers formed from linear PEG polymers of different molecular weights (2,000 to 10,000 g/mol). The volumetric equilibrium swelling ratio and the concentration of crosslinkable double bonds, a measure of crosslinking density, for the resulting PEG gels are presented in Table 1. Both properties strongly impact solute diffusivity within the hydrogel network.
The release of proteins of varying molecular weight from PEG hydrogels was experimentally observed. Table 2 lists each protein studied with its corresponding molecular weight, literature reported hydrodynamic radius, and diffusion coefficient in aqueous solution at 37°C as calculated by the Stokes-Einstein equation:
Here, Do is the diffusion coefficient of the solute in a given solution; k is Boltzman’s constant (1.38 × 10−23 J/K); T is temperature (37°C = 310 K); η is the viscosity of the solvent (6.915 × 10-4 N•m/s2 for water at 37°C (18)); and Rs is the radius of the solute (20). Diffusion coefficients calculated using the Stokes-Einstein equation were used as rough approximations for comparison with experimentally derived diffusion coefficients for proteins in hydrogels of varying crosslinking density.
Individual proteins were loaded at 1 mg/ml into PEG hydrogels formed from varying molecular weight macromers, and protein release was experimentally measured as described. Diffusion coefficients for each protein in each PEG hydrogel system were estimated by fitting the release data to the one-dimensional Fickian diffusion model, given in Equation 1. Agreement between the experimentally measured release profiles and the theoretically predicted release profiles plotted in Figure 2 suggests that insulin, trypsin inhibitor, and carbonic anhydrase exhibit Fickian diffusion through hydrogels formed from the 10,000 g/mol PEG macromer. Experimental and theoretical release profiles for these proteins from all other PEG gels were also in agreement.
The release of ovalbumin from PEG gels formed from the 10,000 g/mol macromer followed Fickian diffusion (Figure 3A), but ovalbumin release from PEG gels with greater crosslinking densities did not agree with theoretically predicted profiles (3B). This inconsistency was also observed in the release of myoglobin, with Fickian diffusion apparent in PEG gels formed from macromers of molecular weight 6,000 g/mol and greater (Figure 3C), but poor agreement between experimental and theoretical release profiles in gels formed from 2,000 g/mol and 4,000 g/mol macromer (Figure 3D). Further analysis of protein diffusion in PEG gels was limited to protein/gel combinations that exhibited Fickian diffusion, allowing for the use of diffusion coefficients estimated by fitting release profiles to Equation 1.
Hydrogel samples incubated in a 1 mg/ml solution of bovine serum albumin did not release detectable concentration of protein suggesting that bovine serum albumin does not significantly penetrate these networks. Even if the diffusion coefficient of bovine serum albumin within the PEG gels was reduced by two orders of magnitude (0.8 × 10−9 cm2/s) relative to that in aqueous solution, protein loading of the gels should occur in approximately 12 hours. Although complete protein release would not be achieved in 2 hours, release of bovine serum albumin near the gel surface would be observed. For proteins that diffused from PEG gels in agreement with Fickian diffusion profiles, the calculated diffusion coefficients were plotted with respect to the concentration of crosslinkable double bonds in the hydrogel precursor solution (Figure 4). In comparison to the approximate diffusion coefficient in solution predicted by the Stokes-Einstein equation, insulin diffusivity is reduced by approximately 40% in the PEG gels with the lowest crosslinkable bond concentration and up to 60% in PEG gels with the highest concentration (Figure 4A). The diffusion coefficients of larger proteins, such as trypsin inhibitor and carbonic anhydrase, were decreased to approximately 10% of that in aqueous solution (Figure 4B and 4C). The following equation for the diffusion coefficient of a given solute through a gel network (Dg) relative to that of the solute in solution (Do) demonstrates how diffusion is dependent on the solute radius (rs) relative to a crosslinked network characteristic length (ξ) and the equilibrium water content of the hydrogel network, described here as the polymer volume fraction in the gel (ν2):
Y is the ratio of the critical volume required for a successful translational movement of the solute to the average free volume per liquid molecule and is usually approximated to be 1 (21). ν2 is simply the inverse of the equilibrium swelling ratio (Q) presented in Table 1.
Additionally, the diffusion coefficients of each protein in PEG gels formed from the 10,000 g/mol macromer were plotted relative to reported protein radii in solution (Figure 5). Insulin, with a radius of 1.47 nm, has the largest diffusion coefficient, and all other proteins, with radii greater than 2.0 nm, have significantly lower diffusion coefficients. The corresponding approximate time scale for insulin diffusion from the 0.4 mm thick PEG gels is 5 minutes, relative to 30 and 45 minutes for trypsin inhibitor and carbonic anhydrase, respectively.
Islet survival in PEG gels formed from varying macromer molecular weight was observed by LIVE/DEAD® staining after 14 days in culture. Encapsulated islets retained greater than 97% viability in all PEG gel conditions at this time point (Figure 6).
Total insulin release from islets encapsulated in PEG gels formed from each macromer molecular weight was measured following one hour static glucose challenge after one week in culture and normalized by sample ATP content to avoid sample variation due to different cell numbers between samples. Islet insulin release from within all PEG gel samples was similar to that released by unencapsulated islets (Figure 7).
To investigate any delays in islet response to high glucose stimulation introduced by the PEG gel structure, encapsulated islet samples were transferred to fresh glucose solutions repeatedly during the one hour glucose stimulation period. The insulin content of solutions collected after 2, 5, 10, 15, 20, 30, 45, and 60 minutes of glucose stimulation was measured and presented cumulatively over 60 minutes as a percentage of the total insulin released (Figure 8). Approximately 70% of the total insulin secreted is measured in the first five minutes of stimulation from unencapsulated islets. This value is reduced to 50% in islet samples encapsulated in PEG gels with the lowest crosslinking density, and further reduced with increasing crosslinking density. While islets in each PEG gel condition secreted similar total amounts of insulin (40–50 pg/ng ATP, Figure 7), insulin release to the solution surrounding encapsulated samples was delayed by the diffusion of insulin through the PEG gel. Considering the diffusion coefficient of insulin in the PEG gels with the lowest crosslinking density (~1.27 × 10−6 cm2/s), if an islet were positioned at the center of the hydrogel, the time scale for the diffusion of secreted insulin to the surrounding solution would be approximately 30 minutes. However, in these samples, islets were distributed throughout the gels, and only insulin secreted by islets positioned in the exact center of the gel were subject to the longest diffusion pathway of 0.5 mm. The observed insulin response delays were shorter than 30 minutes, because many encapsulated islets were located nearer to the hydrogel surface.
Early efforts to design hydrogels for adult islet encapsulation and transplantation largely focused on optimizing the immunoprotective barrier properties, and hydrogel systems that did not significantly hinder diffusion of small proteins, specifically insulin, to and from encapsulated cells but blocked the passage of large immune-cell secreted products such as antibodies. The concept of a molecular weight cutoff (MWCO), or a molecular size that cannot penetrate the network, was widely explored. However, relative to proteins which are monodisperse, the polymers used to synthesize encapsulation barriers are polydisperse, further complicating the concept of a gel MWCO and contributing to very broad MWCO ranges. Alginate-poly-L-Lysine capsules have been synthesized with a MWCO between 14,500 g/mol and 44,000 g/mol, as determined by measuring the diffusion of linear dextran molecules and model proteins (22). Other natural and synthetic polymer membrane studies have targeted a MWCO of 50,000 g/mol (23–28).
Even if the ideal encapsulation barrier were achieved, the in vivo environment for transplanted cells would still present harsh conditions for survival and function (29). In addition to an immediate foreign body inflammatory response to the transplanted graft and the islet-specific autoimmune rejection introduced by the type 1 diabetic immune system, the presence of the encapsulation barrier, no matter how thin, prevents revascularization within the islet mass, thus eliminating the original, physiological source for islet nutrient supply and the primary avenue for blood glucose regulation. Few studies have focused on the diffusion of proteins with sizes in between the extremes of insulin and IgG that may also be critical for successful insulin-producing cell transplantation. With a molecular weight of approximately 38,000 g/mol, vascular endothelial growth factor (VEGF) is one example of a protein that may influence encapsulated islet survival. Under hypoxic stress, islets secrete elevated levels of VEGF to promote localized vascularization (30, 31), and for encapsulated islets, the growth of blood vessels near the hydrogel surface may be of critical importance in establishing the proximity to the host circulation required for continued islet survival and function. Another growth factor, hepatocyte growth factor (HGF), has also been shown to promote transplanted islet viability and function (32). With an approximate molecular weight of 60,000 g/mol, transport of HGF to encapsulated islets will require a barrier with a MWCO greater than 50,000 g/mol and a PEG gel with a crosslinking density less than those synthesized in this work. The diffusion of various growth factors is equally important in the use of three-dimensional hydrogel environments for transplanting islet progenitor cells. Whether exogenously delivered or locally secreted by host tissues, factors necessary for guiding the differentiation and function of these cells must diffuse through the hydrogel barrier (33, 34).
In contrast, while the exclusion of high molecular weight proteins, such as antibodies (~150,000 g/mol) would provide protection from many cytotoxic immune molecules, other potentially toxic proteins such as IL-1β, with an approximate molecular weight of 17,500 g/mol, could easily penetrate the encapsulation barrier (35). The limitations of semipermeable membranes for immunoprotection have been reviewed (36), and an encapsulation barrier that could actively modulate the local immune response in addition to excluding many large toxic molecules may one day be beneficial to the preservation of encapsulated islet survival (37).
In the presented experiments, the diffusion of model proteins with molecular sizes between those typically studied was investigated in PEG hydrogels with network structures not accurately described by classical theories. Alterations in PEG gel crosslinking density resulted in changes in the amount of free volume available for diffusion within the gel, directly influencing protein diffusion within these gels. If the gel structure serves only to sterically interfere with protein diffusion, the gel can be regarded as a property of the solution, and the protein concentration gradient between the gel and the surrounding solution remains the primary driving force for diffusion, resulting in Fickian diffusion profiles. Not all of the proteins released in these experiments exhibited this type of diffusion in every PEG gel, indicating that other potential interactions are influencing diffusion. Complicating factors that may contribute to non-Fickian diffusion include changes in protein conformation, and therefore shape, during the release experiment, possible aggregation of protein molecules during the release experiment, and interactions, such as hydrogen bonding, between entrapped proteins and the PEG network. While significant adsorption of proteins to the hydrophilic PEG gels is unlikely, as the solute dimensions approach those that characterize the hydrogel network, diffusion is further complicated as solute contact with the gel structure is increased and even low specificity, low strength interactions such as van der Waals forces may affect solute diffusion.
The use of varying molecular weight PEG macromers not only resulted in hydrogels of varying crosslinking density, but also changes to the polymerization solution, specifically radical concentration profiles in time. Free radical concentration is influenced by not only the concentration of crosslinkable double bonds, but also factors such as solution viscosity and diffusion limitations introduced by gel network formation (2). For the PEG hydrogels tested, changes in the polymerization solution and the final crosslinking density did not affect encapsulated islet survival or alter islet insulin secretion over a one hour glucose stimulation period. At specific time points within the one hour high glucose incubation, however, insulin secretion was delayed from islets entrapped in 1 mm thick PEG gels relative to unencapsulated islet insulin release. The trend of increasing delays with increasing PEG crosslinking density suggests that diffusion of insulin through the gel network plays a role in the response time of encapsulated islets to changes in external glucose concentration. This diffusion-related delay could be offset by reducing the thickness of the gel barrier, and thus the distance between encapsulated islets and the surrounding environment. Given that total insulin release over the 1 hour incubation was not diminished with encapsulation, further tests of islet function are necessary to determine if the observed delays will be problematic for the ultimate goal of reversing hyperglycemia through implantation of islet-gel constructs.
The authors gratefully acknowledge the National Science Foundation and the Department of Education’s Graduate Assistantships in Areas of Nation Need Program for fellowships to LMW, the National Institute of Health/Howard Hughes Medical Institute Scholars Program for funding to CGL, and the National Science Foundation (Grant #EEC044771) and the Howard Hughes Medical Institute for research funding.