Engineering PEG-based hydrogels to foster efficient endothelial network formation in free-swelling and confined microenvironments

Abstract

In vitro tissue engineered models are poised to have significant impact on disease modeling and preclinical drug development. Reliable methods to induce microvascular networks in such microphysiological systems are needed to improve the size and physiological function of these models. By systematically engineering several physical and biomolecular properties of the cellular microenvironment (including crosslinking density, polymer density, adhesion ligand concentration, and degradability), we establish design principles that describe how synthetic matrix properties influence vascular morphogenesis in modular and tunable hydrogels based on commercial 8-arm poly(ethylene glycol) (PEG8a) macromers. We apply these design principles to generate endothelial networks that exhibit consistent morphology throughout depths of hydrogel greater than 1 mm. These PEG8a-based hydrogels have relatively high volumetric swelling ratios (>1.5), which limits their utility in confined environments such as microfluidic devices. To overcome this limitation, we mitigated swelling by incorporating a highly functional PEG-grafted alpha-helical poly(propargyl-L-glutamate) (PPLGgPEG) macromer along with the canonical 8-arm PEG8a macromer in gel formation. This hydrogel platform supports enhanced endothelial morphogenesis in neutral-swelling environments. Finally, we incorporate PEG8a-PPLGgPEG gels into microfluidic devices and demonstrate improved diffusion kinetics and microvascular network formation in situ compared to PEG8a-based gels.

Introduction

In vitro tissue engineered models, often referred to as “Microphysiological Systems (MPSs)” or “organs-on-chips”, are poised to have significant impacts on drug development, both as platforms for basic discovery science in academia and industry, and translational models of human metabolism and toxicity in later stages of drug development^1,2. Although these tissue-engineered tools have already made inroads into providing important information in some applications, there remain significant gaps in how well these models recapitulate enough of the physiological architecture and function of native human tissues to be useful for other applications^3,4.

An essential functional feature of most native tissues is the microvasculature – a feature missing or only poorly represented in many existing in vitro model systems⁵. On the one hand, endothelial capillary networks deliver oxygen and nutrients necessary for growth and development of tissues and organs^6,7,8. On the other hand, endothelial cells also help to guide the development and organization of functional tissue structures in the developing embryo, and in regeneration and disease, based on paracrine signaling interactions with other types of cells^9,10. Such interactions are prominent in liver regeneration and neural stem cell differentiation, as well as many other normal physiological processes^11,12,13.

Several research groups have engineered microvasculature in vitro using cell-driven remodeling of natural collagen and fibrin extracellular matrix (ECM) protein gels^14,15,16, and extended these approaches to create perfusable capillary networks in accessible microfluidic device formats^17,18,19. These foundational studies have established the basis for producing perfusable tissues in vitro through cell-driven morphogenesis. Interest now extends to creating microvascularized organoids, which are typically cultured in Matrigel™, a murine tumor-derived ECM that contains hundreds of components including many growth factors with unknown influence on tissue development²⁰. Natural matrices like collagen, fibrin, and Matrigel™ face lot-to-lot variability issues, and the complex convolution of biochemical and biophysical cues hinder dissection of how individual variables contribute to biological processes studied in these systems^21,22. Most importantly, natural matrices have limited modularity (i.e., ability to deconvolute individual parameters) and tunability (i.e., ability to vary parameters systematically and independently across a wide range of values). These deficiencies make them difficult to adapt to the disparate needs of different cell types in a complex coculture environment, and can lead to practical problems like rapid degradation with concomitant tissue compaction, inhibiting their use in long-term tissue models²³.

To overcome these limitations, several research groups have explored the use of synthetic, fully defined ECMs to replace natural matrices as 3D substrates for tissue culture^24,25,26. Synthetic matrices, prominently poly(ethylene glycol) (PEG)- based hydrogels, are often highly modular and tunable, allowing them to be finely tailored for specific applications^27,28,29. The application of synthetic matrices to in vitro tissue models has seen extensive growth in recent years, with matrices designed to support intestinal organoid development and investigate cell-cell communication in complex, multicellular tissue models^30,31,32. Although a number of studies have reported the generation of some form of microvasculature in different synthetic or semi-synthetic hydrogel systems^{33,34,35,36,37,38,39,40,41,42} the successful integration of functional vasculature in a fully defined synthetic microphysiological model system remains a significant challenge.

Synthetic matrices need to be specifically engineered to capture the relevant bioactive cues present in native ECM. Previous efforts to engineer vascularized synthetic hydrogels take varied approaches in generating hydrogel matrices, and although they often include some degree of optimization, for example, evaluating different degrees of polymer density, detailed investigation of the matrix properties that influence endothelial morphogenesis in a given system usually obviated by the identification of a matrix formulation that supports some degree of vascularization^34,35,36,41. Notably, previously described synthetic matrices differ from one another in terms of their physical and biomolecular properties, but certain conserved relationships are evident across several systems, such as the diminishing of capillary network formation with increased material stiffness or polymer density^36,41. These differences between matrix systems may emerge based on the properties and quantities of the matrix components (polymers, peptides, full-length proteins) or the polymerization strategies used to form the gels (physical, chemical, enzymatic). The effective design of synthetic matrices requires a better understanding of the fundamental aspects of the matrix environment that influence cellular morphogenesis. A powerful feature of engineerable synthetic systems is the ability to isolate and independently alter the physical and biomolecular properties of the matrix environment to understand their impact on cell behavior. Here, we describe a selection of matrix design parameters (polymer density, crosslinking density, degradability, adhesive ligand density, structural heterogeneity) relevant to endothelial morphogenesis and systematically demonstrate their impact on hydrogel properties and endothelial network formation.

We use an engineerable synthetic hydrogel platform to probe ECM design principles that govern endothelial morphogenesis into microvascular networks. We first define categories of critical biomolecular and physical properties, then leverage the independently tunable parameters of this system to investigate the roles of these fundamental ECM properties in formation of microvascular networks. These design principles are then applied to engineer 3D vasculature that span tissue dimensions of at least 1 mm and exhibit certain key markers of vascular maturity. We address an important limitation of this and many other synthetic systems, hydrogel swelling, and control this property by incorporating an additional synthetic macromer based on PEG-grafted poly(propargyl-L-glutamate, “PPLGgPEG”) along with a commercially-available 8-arm PEG (“PEG8a”) macromer. Hybrid gels incorporating both PEG8a and PPLGgPEG macromers result in enhanced endothelial network formation compared to 8-arm PEG-based gels. Finally, we correlate this enhancement in morphogenesis with changes in hydrogel physical structure and fundamental properties, and demonstrate that we can use this novel material to improve endothelial network formation in microfluidic environments, a critical step towards the development of improved in vitro models of vascularized tissues.

Results and Discussion

Synthetic engineered microenvironments provide a platform for understanding how matrix properties impact morphogenesis

To describe design principles most relevant to endothelial morphogenesis, we systematically engineered synthetic microenvironments by tuning several material properties (Fig 1). According to the schematic in Figure 1, we used 8-arm reactive PEG8a macromers, cell-adhesive peptide sequences, as well as protease-sensitive peptide crosslinkers to generate synthetic hydrogel microenvironments. The modular nature of these synthetic materials permits precise control over the cellular microenvironment by independently modulating the matrix biophysical properties (Fig 1B). By systematically varying the concentration and identity of polymer, adhesion peptide, and crosslinker in the hydrogel compositions, the matrix properties that regulate endothelial morphogenesis can be precisely and independently controlled. To investigate endothelial morphogenesis in these environments, we encapsulated commercially available iPSC-derived endothelial cells (iPSC-EC). These cells have been previously characterized in detail and shown to behave similarly to human umbilical vein endothelial cells (HUVEC), with the added advantage of circumventing issues of variability that come with using primary cells^43,44. We use this system to investigate the design principles that govern endothelial morphogenesis in 3D hydrogel environments, which will provide a rational basis for further development of synthetic biomaterials for in vitro tissue engineering applications.

Figure 1. A modular and tunable synthetic hydrogel system.

A) Synthetic hydrogel fabrication scheme enables robust control over properties of defined hydrogel microenvironments. B) Synthetic microenvironments can be tuned independently along molecular and physical axes, guided by the needs of the cellular space.

Engineering the physical properties of synthetic matrices

We first investigated the physical properties axis (Fig 1) by varying polymer and crosslinking densities in synthetic matrices containing encapsulated iPSC-ECs. The fibronectin-based peptide PHSRN-K-RGD (see Methods for specific peptide sequences) was used to engender cell adhesion. The canonical RGD motif binds well to α_vβ₃, though with lower affinity for other RGD-recognizing integrins including α₅β₁⁴⁵. We have previously found that addition of the PHSRN sequence from the synergy site in the neighboring 9^th domain of fibronectin to fully synthetic biomaterials improves the attachment and function of many cell types, particularly certain primary cells that lack α_vβ₃ expression^{32,46,47,48,49}. Matrix remodeling was mediated by the addition of collagen-based protease-sensitive peptide crosslinker containing the GPQIWQG amino acid sequence (LW)²⁷. After 3 days of culture in these synthetic gel environments, samples were fixed and stained with phalloidin to visualize and quantify the extent of vascular network formation (Fig 2A).

Figure 2. Engineering the physical parameters of synthetic hydrogel to support endothelial network formation.

A) Elastic modulus of synthetic PEG8a gels measured by rheometry demonstrates tunability over a range of stiffness by modulating polymer density. Gels were formed at a fixed stoichiometric crosslinking ratio of 35% with LW and PHSRN-K-RGD concentration of 2 mM. B) Elastic modulus as a function of crosslinking ratio expressed here by the percentage of norbornene groups occupied by thiolated crosslinker (LW). Gels were formed at a fixed PEG8a concentration of 3 wt% and PHSRN-K-RGD concentration of 2 mM. C) Maximum z-projections over 100 m depth of F-actin-stained hydrogels after 3 days in culture illustrate the impact of polymer density and crosslinking density on 3D endothelial network formation. Gels were formed with the parameters described in A and B. D) Quantification of network length within a given image volume 400 m in depth reveals the impact of polymer density on endothelial network formation. E) Quantification of network length within a given image volume 400 m in depth reveals the dependence of endothelial morphogenesis on crosslinking density. (**** p < 0.0001, * p < 0.05, n = 3 hydrogel replicates, error bars represent SD).

In collagen and fibrin gels, angiogenesis and endothelial network formation are hindered as matrix density increases^50,51,52,53. Increases in matrix protein concentration simultaneously alter the mechanical properties, nanoscale architecture and permeability, adhesion ligand density, and number of protein linkages cells must degrade in order to remodel the microenvironment. In the synthetic ECMs used here, degradation rate and mechanical properties were varied independently of adhesion ligand concentration and total polymer density by varying the crosslink density and cleavage rate of crosslinking sequences. Systematic variation of both polymer density and crosslinking ratio, as the ratio of thiol groups on the crosslinkers to norbornene groups on the PEG8a macromers, produced hydrogels with a range of elastic moduli (Fig 2A–B). A decrease in endothelial network formation was observed as polymer density was increased (Fig 2C–D). Similarly, increasing the crosslinking ratio also produced stiffer environments and resulted in a decrease in endothelial network formation (Fig 2C–E). Other metrics related to endothelial network morphology including the length of the longest continuous network, the total number of branches, and endothelial cord diameter followed similar trends (Fig S1). The low crosslinking density regime over which endothelial network formation takes place most robustly corresponds with very soft hydrogels. Interestingly, although endothelial cell proliferation during early morphogenesis was found to decrease in high crosslinking regimes, no significant differences in proliferation rate were seen within the low to middle crosslinking regime. This suggests that the difference in endothelial network formation is mediated not by proliferation, but perhaps by matrix degradation, inhibited migration, or other critical events involved in morphogenesis (Fig S2).

These results suggest that the physical properties of synthetic matrices have a significant impact on endothelial morphogenesis, supporting extensive work from other hydrogel systems^35,44. However, it is necessary to consider that these bulk material property changes are reflected in multiple ways in the local cellular microenvironment. For example, increased crosslinking density not only increases the stiffness but also decreases the degradability of the pericellular environment by increasing the concentration of protease degradation sites that must be cleaved in order for cells to migrate. Crosslinking also changes gel swelling properties, and thus the effective concentration and accessibility of adhesion ligands.

Engineering the molecular properties of synthetic matrices

Independent of the physical properties investigated above, the bioactive properties of the matrices were modified to characterize their influence on endothelial morphogenesis. First, the nominal adhesion ligand concentration was varied between 0 and 3 mM, with the total peptide concentration kept constant by addition of a non-interactive RDG peptide. It should be noted that the actual concentration of ligand is dependent on hydrogel volumetric swelling. Increased network formation was observed as the global integrin ligand concentration was increased, with diminishing returns between 1 and 3 mM (Fig 3A–B). Interestingly although global adhesion ligand concentration had a significant impact, incorporating bioactive peptide sequences derived from either fibronectin or collagen I, designed to engage with different integrins (i.e. α₅β₁ versus α₂β₁) did not significantly alter endothelial morphogenesis (Fig S3).

Figure 3. Engineering the molecular properties of synthetic hydrogel to support endothelial network formation.

A) Maximum z-projections over 100 m depth of F-actin-stained hydrogels demonstrate the impact of adhesion peptide (PHSRN-K-RGD) density 3D endothelial network formation. Note that nominal concentration values are shown here, but actual concentration is dependent on hydrogel swelling and are therefore 50% of their nominal concentrations. Here gels were formed at 3 wt% PEG8a with 35% crosslinking (LW). B) Quantification of network length through 400 m depth reveals increasing network formation with adhesion ligand concentration. C) Maximum z-projections over 100 m depth of F-actin-stained hydrogels crosslinked with degradable (LW) or non-degradable (PEG-dithiol) linkers. D) Vascular network formation can be modulated using crosslinkers with different proteolytic cleavage properties. Gels are 3 wt% PEG8a with 35% crosslinking. (**** p < 0.0001, *** p < 0.001, ** p < 0.01, n = 3 hydrogel replicates, error bars represent SD).

Unlike natural matrices, where physical and molecular properties are highly intertwined, the synthetic matrices studied here allow for control over both adhesion ligand density and degradability independent of hydrogel physical properties. Notably, the adhesion ligand density regime that supported robust endothelial network formation (a nominal concentration of 1–3 mM, which represents an actual concentration of about 0.5–1.5 mM when adjusted for hydrogel swelling) is significantly below the adhesion site density (about 0.035 mM) in commonly used collagen and fibrin gels^54,55. This difference could derive from the presence of additional synergistic binding sites present in fibrin and collagen, which are conceptually similar but not identical to the PHSRN synergy site contained within the synthetic hydrogels used in this study^56,57. Furthermore, the nanoscale structure of natural matrices may permit higher ligand accessibility compared to the tight mesh structure of synthetic gels. Although not explicitly explored here, natural matrix proteins contain binding sites for growth factors and other matrix proteins that contribute to the complexity of the bioactive microenvironment by modulating growth factor bioavailability and matrix architecture^58,59. While other studies have demonstrated incorporation of heparin sulfate motifs for growth factor binding⁶⁰ and specific peptide sequences designed to sequester cell secreted ECM^32,61,62 we seek to engineer environments that support initial cellular morphogenesis events including the deposition of native ECM, whose complex properties enable downstream morphogenesis. As discussed below, we observed sufficient matrix deposition by endothelial cells in the absence of matrix binding peptides. Interestingly, although the concentration of PHSRN-K-RGD within the matrix influences the extent of network formation, the diameters of endothelial cords did not change in response to changes in adhesion ligand concentration, suggesting that cord diameter is regulated by the density of the matrix and ability of endothelial cells to degrade their surrounding environment, rather than the number of integrin binding sites initially present (Fig S1).

The degradability of the matrix can be modulated independently of its physical properties by altering the amino acid sequence of the crosslinker. The dependence of endothelial network formation on proteolytic matrix degradation can be demonstrated by comparing endothelial morphogenesis in matrices crosslinked with protease-sensitive peptide crosslinkers to those crosslinked with non-degradable PEG dithiol linkers (Fig 3C). Furthermore, the extent of endothelial network formation can be controlled by modulating the susceptibility of the crosslinker to proteolysis. Three peptide sequences were selected that were previously demonstrated to have varying catalytic efficiencies of cleavage by matrix metalloproteinases (MMPs) 1 and 2, two of the most prevalent MMPs detected in this system^63,64 (Fig S3). Incorporating these peptides into synthetic matrices revealed distinct differences in the extent of endothelial network formation after 3 days of culture, with the LA (GPQIAGQ) crosslinker supporting less extensive network formation than the collagen mimetic LW (GPQIWQG) linker, as predicted by the differences in cleavage efficiency (Fig 3D). Interestingly, VPMS (VPMSMRGG) exhibited less extensive endothelial network formation than LW despite its significantly higher k_cat reported for MMP-1 and MMP-2. Although we utilized iPSC-ECs in characterization of these matrices, we have also observed that other sources of endothelial cells were capable of forming endothelial networks in these materials without detectable secretion of MMP-1 or MMP-2, suggesting that other proteases may be involved in mediating endothelial morphogenesis in this hydrogel system (Fig S4). Because no significant differences were observed in the physical properties of gels formed with these linkers, these experiments demonstrate the ability to modulate cell-mediated matrix degradation independent of matrix physical properties (Fig S5).

The demonstration that different endothelial cell populations respond differently in the same hydrogel environment in relation to the specific proteases expressed by those endothelial cells suggests further possibilities in rational matrix design. By characterizing the activity of specific proteases and the expression of ECM receptors on the cell surface for a particular cell source of interest, one could rationally design a biomaterial environment to elicit the appropriate morphogenic responses via adhesion and degradation. Extended further, this approach derives a framework for understanding how to most appropriately modify or tune synthetic material environments to enable the generation of more physiologically relevant cellular organization and function in complex multicellular tissue models.

Engineered vasculature demonstrates markers of maturity

Using the design principles outlined above, microvascular networks were generated that exhibited consistent morphologies throughout the entirety of a 1 mm thick synthetic hydrogel, as assessed by histology and confocal imaging (Fig 4A). Although still in the early stages of development, these vascular networks maintained CD31 expression and exhibited several markers of functional maturity (Fig 4B). Namely, the endothelial cells deposited a laminin-rich basement membrane surrounding vessel structures (Fig 4B), and displayed contact inhibition marked by a decrease in proliferation to a basal level as the networks formed (Fig 4D). These phenomena have been previously demonstrated in endothelial populations during the formation of mature capillary networks^32,65,66,67. Short segments of the resulting vascular network structures exhibited defined lumens, suggestive of potential perfusability (Fig 4C); however, the proportion of lumenized vessels within the total structure was relatively small (<5%). This may be due, in part, to the lack of supporting mesenchymal cell populations, which have been used extensively in the generation of microvascular networks and shown to be essential in producing mature perfusable vasculature^{17,41,68,69,70}.

Figure 4. Design principles guide engineering of 3D endothelial networks in synthetic microenvironments.

A) Left: A representative vertical cross section of a 3wt% PEG8a, 35% crosslinked (LW) hydrogel with 2 mM PHSRN-K-RGD demonstrating consistent vascularization through 1 mm of gel. Boundaries indicated by dotted lines Right: A representative confocal image of engineered vasculature through 100 m of gel. B) Immunostaining shows expression of endothelial markers (left) and deposition of extracellular matrix (laminin, right) by endothelial cells in synthetic matrices. C) Confocal microscopy demonstrates the development of open lumen (white arrows) structures in short segments of the networks formed in synthetic gels. Scale bar represents 50 m D) Decrease in cell proliferation assessed by EdU incorporation during network formation suggests contact inhibition of endothelial cells. (n = 3 hydrogel replicates, error bars represent SD).

Extensive volumetric swelling of soft synthetic hydrogels limits their use in confined environments

Practically, the design principles outlined above suggest that endothelial network formation might be improved by further reducing crosslinking density; however, the hydrogels formed in the lowest crosslinking regimes examined here represent the functional lower limit of gel formation (Fig S6). Although the low crosslinking density gels described above were capable of supporting endothelial morphogenesis, one significant limitation towards practical applications was the high volumetric swelling ratio of these gels, particularly within the regime at which vascular network formation was most pronounced (Fig 5A). As hydrogels swell beyond their initial volumes, the nominal concentration of adhesion peptide, crosslinker, and cells at which gels were originally fabricated becomes diluted two-fold for gels that are 35% crosslinked.

Figure 5. Synthesis of soft synthetic hydrogels is limited by swelling in confined environments.

A) Volumetric swelling increases above neutral (dashed line) particularly at low crosslinking densities where endothelial morphogenesis takes place (shaded region). B) Comparison of expected crosslinking density in confined gels (restricted volume) and unconfined gels (freely swelling) demonstrates the impact of confinement on the local microenvironment, particularly in gels with low crosslinking densities. C) Schematic representation of geometrically confined gels that would otherwise swell leading to more restrictive polymer density. D) F-actin-stained endothelial cells encapsulated in identical synthetic hydrogels (3 wt% PEG8a, 35% crosslinking (LW), 2mM PHSRN-K-RGD) inside a microfluidic device (confined) or floating freely in media (unconfined). Scale bars represent 100 m.

The development of effective physiomimetic models depends on integration with advances in microfabrication and microfluidics to control the geometry and perfusion of tissues. Many existing synthetic hydrogel systems, including the hydrogel platform based on PEG8a macromers, suffer from extensive volumetric swelling at the low crosslinking densities required for endothelial network formation. Neutral swelling properties are critical for integration of synthetic hydrogels with emerging fabrication techniques and hardware, which require the maintenance of resolution and restricted or confined volumes, respectively. In circumstances where gels swell, but are confined to their original volume, we see a marked decrease in endothelial network formation compared to the same gel that is allowed to swell freely (Fig 5C–D). This may be due to an increase in effective polymer and crosslinking density compared to the unconstrained state (Fig 5B–C). Although the diffusion of biomolecules may also change in this confined environment, analysis of local growth factor concentration in the gels over time suggests that the differential diffusion of biomolecules is likely not the driver of the reduction in endothelial network formation that is seen in confined environments (Fig S7). A further limitation of these gels, in addition to the impairment of network formation, is that we observed gel swelling that significantly occluded abutting media channels in some devices.

PEG8a-PPLGgPEG hybrid gels promote enhanced endothelial network formation at neutral-swelling compositions

To overcome the limitations presented by hydrogel swelling, a second macromer, poly(propargyl-L-glutamate) grafted with short PEG chains (PPLGgPEG), was incorporated into the 8-arm PEG8a macromer solution at systematically varied ratios to create hybrid gels (Fig 6A). PPLGgPEG is a rigid, alpha-helical polymer that forms hydrogels across a range of compositions, resulting in gels that exhibit either neutral swelling or modest syneresis⁷¹. First, hydrogels were synthesized using PPLGgPEG as the sole macromer component. These hydrogels demonstrated the potential for 3D vascular network formation, but were limited in their tunability by the high thiol:norbornene ratio required to induce gel formation (Fig S8).

Figure 6. Incorporation of PPLGgPEG enables endothelial network formation in neutral-swelling synthetic hydrogels.

A) Schematic representation of PEG8a-PPLGgPEG dual macromer hydrogels illustrating formation of polymer-dense clusters contributing to heterogeneous gel structure. B) PEG8a-PPLGgPEG hydrogels can be engineered at neutral-swelling formulations. C) Elastic moduli of PEG8a-PPLGgPEG gels measured by rheometry. D) Maximum z-projections through 100 m of gel demonstrate PPLGgPEG-dependent increases in vascular network formation following 3 days of culture. E) Quantification of vascular network formation in PEG8a-PPLGgPEG hydrogels. All hydrogels are 3 wt% polymer (PEG8a + PPLGgPEG), 50% crosslinked (LW) with 2mM PHSRN-K-RGD (**** p < 0.0001, *** p < 0.001, n = 3 hydrogel replicates, error bars represent SD).

As an alternative approach to preserve modular properties, PEG8a-PPLGgPEG hybrid gels were formed by mixing PEG8a and PPLGgPEG macromers while keeping the total polymer content of the gels fixed at 3 wt%. Incorporating multi-arm PEG8a macromers allowed gels to form at lower crosslinking ratios, freeing up sufficient concentrations of free norbornenes for attachment of bioactive peptides. Interestingly, as the proportion of PPLGgPEG within a fixed total polymer density was increased, changes in hydrogel opacity were observed (Fig S8). Furthermore, clusters of polymer-rich regions on the order of hundreds of nanometers in scale were observed within the gel using fluorescently tagged macromers in PPLGgPEG gels, but not in PEG8a gels (Fig S8). This result is indicative of phase separation related to the difference in hydrophobicity of PPLGgPEG compared to PEG8a, and was previously shown to be controlled by grafting with ethylene oxide chains⁷². This hydrophobicity may also explain the influence of PPLGgPEG on swelling properties, in line with a recent report demonstrating that methacrylation level of dextran hydrogels could be used to control swelling³⁶. Other work from our groups has described this phase separation in greater detail⁷³. We suggest that addition of PPLGgPEG to PEG8a matrices generates structural heterogeneity in the gels, necessitating the presence of regions with lower polymer density. This may facilitate more efficient local degradation of the gels by encapsulated cells, leading them to form endothelial networks as if they were in a softer, lower polymer density environment as originally shown in Figure 2.

Volumetric swelling was measured over a range of polymer compositions and crosslinking densities to identify synthetic gel compositions that were near-neutral swelling (Fig 6B), and exhibited lower elastic moduli despite the decrease in swelling (Fig 6C). Strikingly, when endothelial cells were incorporated into PEG8a-PPLGgPEG matrices of 0 (PEG8a alone), 25, 50, or 75% PPLGgPEG, a robust increase in endothelial network formation was observed (Fig 6D–E). In Figure 2, we observed decreases in stiffness that correlated with increases in endothelial network formation, although different modifications (polymer vs. crosslinking density) may result in different extents of endothelial network formation at similar levels of stiffness, presumably due to the differences in hydrogel structure that result from those modifications. It is likely that the decrease in stiffness with increasing PPLGgPEG content reflects changes in structural properties that make the material more permissive to endothelial morphogenesis and also change the bulk stiffness of the material, but further work would be necessary to characterize this effect in more detail. Importantly in the context of this work, formulations of PEG8a gels with comparable elastic moduli undergo substantial volumetric swelling compared to neutral-swelling PEG8a-PPLGgPEG matrices (Fig 2B, 5A).

PEG8a-PPLGgPEG hydrogels support enhanced diffusion and vascularization in microfluidic environments

To demonstrate the utility of PEG8a-PPLGgPEG hybrid matrices in overcoming the limitations associated with swelling in confined environments, a neutral swelling hydrogel formulation consisting of an equal masses of PEG8a and PPLGgPEG (50% PPLGgPEG, 50% PEG8a) was incorporated into a microfluidic platform (Fig 7A) and compared to PEG8a alone. Although further increases in the PPLGgPEG content up to 75% showed improved endothelial network formation, these hydrogels also underwent significant syneresis, making them impractical for use in microfluidic devices requiring a gel barrier (Fig 6B–E). Diffusion experiments were performed in situ by flowing 70kDa FITC-dextran into the media channels and measuring fluorescence intensity over time in the gel channel. These experiments demonstrated that PEG8a-PPLGgPEG matrices supported enhanced diffusivity compared to PEG8a alone (Fig 7B). Furthermore, encapsulated endothelial cells exhibited enhanced network formation in PEG8a-PPLGgPEG gels compared to PEG8a gels in microfluidic devices (Fig 7C–E). These experiments taken together demonstrate that PEG8a-PPLGgPEG gels exhibit favorable bulk properties such as neutral swelling, while maintaining local properties that are capable of supporting enhanced endothelial morphogenesis and diffusion of biomolecules.

Figure 7. Neutral swelling PEG8a-PPLGgPEG gels improve endothelial network formation in confined microfluidic environments.

A) Above: Rendering of the microfluidic inverse mold design with media channels in red and green flanking a central gel channel (grey). Below: Cross sectional view of the device design. B) Fluorescence intensity tracking of 70kDa FITC dextran diffusion across gels in microfluidic devices reveals differences in diffusion kinetics in PEG8a and PEG8a-PPLGgPEG matrices with mean diffusivity reported for both hydrogels. C) Cross-channel micrograph (media channels above and below) and D) representative fields of PEG8a and PEG8a-PPLGgPEG gels in microfluidic devices show enhanced endothelial network formation in confined environments D) Quantification of network length reveals differences in endothelial network formation in microfluidic devices between PEG8a and PEG8a-PPLGgPEG matrices. PEG8a gels are 3 wt% PEG8a, 35% crosslinking (LW) and 2mM PHSRN-K-RGD. PEG8a-PPLGgPEG gels are 1.5 wt% PEG8a, 1.5 wt% PPLGgPEG, 50% crosslinking (LW), and 2mM PHSRN-K-RGD. (*** p < 0.001, n = 5 microfluidic devices, error bars represent SD).

Conclusions

Microvascularization of in vitro tissue models is desirable for both nutrient delivery to 3D tissue as well as proper physiological function via paracrine signaling and maintenance of blood-tissue barrier functions. The many reports of perfusable vessels achieved using fibrin or collagen have not yet been matched by demonstration of perfusable vessels achieved in synthetic materials. Synthetic materials are highly desirable as they potentially avoid the lot-to-lot variability that can impair reproducibility within and between labs. Further, they can be designed for complex co-culture experiments, as we previously described for co-culture of epithelial and stromal cells from the human endometrium^31,32. Through systematic investigation of the relationship between the physical and molecular properties of the microenvironment and the resulting impact on endothelial morphogenesis, we have established a set of critical design considerations for materials that support engineered microvasculature. Specifically, we demonstrate that lower polymer and crosslinking density support improved endothelial network formation. Similarly, increased degradability of matrices is a critical factor in driving endothelial morphogenesis, as well as the density of adhesive ligand. Furthermore, by incorporating a highly functionalizable PPLGgPEG macromer to overcome limitations associated with swelling, we suggest that heterogeneity in hydrogel structure may play a role in altering endothelial morphogenesis. This establishes feasibility for adapting this material approach to achieve improved vascularization within confined microfluidic environments. By building on the design principles outlined in this work, we expect to integrate endothelial networks with existing coculture models in 3D synthetic environments.

Materials and Methods

Functionalized polymers and peptides.

8-arm PEG norbornene 20kDa (PEG8a-NB) was purchased through JenKem Technology (Beijing). Peptide crosslinkers (Ac)GCRD-LPRTG-GPQIWGQ-DRCG(Am) (LW), (Ac)GCRD-LPRTG-GPQGIAGQ-DRCG(Am) (LA), (Ac)GCRD-VPMSMRGG-DCRG(Am) (VPMS) and adhesion ligands (Ac)GGYGGPG(GPP)₅GFOGER(GPP)₅GPC(Am) (GFOGER), (Ac)PHSRNGGGK-GGGERCG(Ac) -GGRGDSPY(Am) (PHSRN-K-RGD), (Ac)GCRE-RGDSP(Am) (RGD), and (Ac)GCRE-RDGS(Am) (RDG) were custom synthesized by Boston Open Labs (Cambridge, MA).

iPS-derived endothelial cell culture.

Induced pluripotent stem (iPS) cell-derived endothelial cells (iPS-EC) were purchased through Cellular Dynamics International (Madison, WI) and cultured according to manufacturer instructions. Briefly, cells were grown to 90% confluency on fibronectin coated flasks, with media changes on the first day and each subsequent day. VascuLife growth medium was purchased through Lifeline Cell Technology (Frederick, MD) and supplemented with iCell endothelial cell media supplement in place of recommended fetal bovine serum purchased through Cellular Dynamics International. Media was additionally supplemented with 1% penicillin/streptomycin (Gibco). Cells used in these experiments were passage 5 or lower. Cells were cultured in a humidified atmosphere at 37°C and 5% CO 2.

Fabrication of 8-arm PEG norbornene (PEG8a) hydrogels.

20 uL hydrogels were fabricated in a one-pot synthesis reaction within a 1 mL syringe from which the tip was removed. Between 3 and 5 wt% PEG8a-NB was mixed with appropriate concentration of adhesion peptide (0 – 3mM nominal) and crosslinked at stoichiometric ratios between 0.35 and 0.65, referred to as 35 or 65% crosslinking, respectively using LW. In experiments were adhesion peptide concentration was variable, a scrambled adhesion peptide (RDG) was added to keep the total concentration of adhesion peptide in the system consistent between conditions. Cells were added to the precursor solution in serum free medium at a final concentration of 12 million cells/mL. Gels were polymerized using 0.05 wt% IRGACURE 2959 (Ciba, Prod. No. 0298913AB) UV-irradiated for 45 seconds at ~100mW/cm².

Characterization of elastic modulus by rheometry.

25 uL hydrogels were polymerized as described above at the indicated polymer and crosslinking densities. All gels also contained 2 mM PHSRN-K-RGD as in encapsulation experiments. Gels were swollen overnight at 37°C and 5% CO₂ in 1X PBS prior to measurement. Using an MCR702 rheometer (AntonPaar) with an 8 mm sandblasted plate geometry, frequency sweeps between 0.1 and 10 Hz were obtained at a fixed strain rate of 5%. Storage modulus (G’) was averaged over the linear elastic regime and used to calculate the reported elastic moduli (E) where E = 2G’(1+ν) and ν is taken as 0.5. Reported measurements are averaged over at least 3 hydrogels.

Assessment of engineerable properties impact on vascular morphogenesis.

Culture:

iPS-EC encapsulated within hydrogels of defined formulation were cultured in iPS-EC media as described above, with a single 20uL hydrogel placed in one well of a 24-well plate containing 800uL of medium, such that gels were completely submerged but not touching the bottom of the plate. Plates were transferred to an incubator in a humidified atmosphere at 37°C and 5% CO 2. Media was changed after 24 hours in culture and every 48 hours after that if necessary.

Immunocytochemistry:

Gels were fixed in 4% paraformaldehyde solution in PHEM buffer for 30 minutes, followed by permeablization in 0.2% Triton X-100 in PBS for 20 minutes at room temperature on an orbital plate shaker set to 60 rpm. Gels were washed twice for 30 minutes at 60 rpm at room temperature in PBS, then blocked for 1 hour in 1% BSA at room temperature. Staining was performed overnight with DAPI (1:1000) and Phalloidin (1:100) at 4°C and 60 rpm in 1% BSA. Following staining, gels were washed three times for 30 minutes at 60 rpm at room temperature in PBS before imaging.

Imaging and analysis:

Z-stack images covering 400 m hydrogel depth were taken in 6–9 representative fields of view in each hydrogel for 3 hydrogels per condition with step sizes between 10–20 m using a Zeiss LSM 880 confocal microscope. Using FIJI, images were collapsed into maximum intensity z-projections with depths of 50–100 m. Images were modified by subtracting background, despeckling, increasing contrast, and performing a Gaussian blur to better distinguish vessel structures from background. A threshold was applied to produce a binary map of the vascular network, which was segmented and analyzed using the included “skeletonize” and “analyze skeleton” functions. The resulting data were analyzed in MATLAB to determine total network length within a representative image volume, including only segments beyond 200 m in the analysis in order to eliminate single or aggregated rounded cells from the analysis. Endothelial cord length is presented as the sum of measured lengths for each sub-projection within the total 400 m image depth.

In situ endothelial proliferation assay

Gels were formed as described above and cultured for up to 7 days with media changes ever 2–3 days. At 24h prior to the indicated time point, EdU was added to culture medium at 10uM according to the manufacturer protocol (Invitrogen C10337). At the indicated time point, gels were fixed in 4% PFA and samples were stained for proliferating and total nuclei. Briefly, permeabilization was performed with 0.5% TritonX-100 for 20 minutes. Gels were washed twice with 1X PBS + 1% BSA and stained with Click-iT reaction cocktail prepared according to manufacturer recommendations for 1h at room temperature, before being washed twice with 1X PBS + 1% BSA and stained with DAPI at 1:1000 for 20 minutes. Finally, gels were washed 3× 30min with 1X PBS prior to imaging with a Zeiss LSM 880 confocal microscope. Quantification of proliferating and total nuclei was performed using Fiji’s particle counter.

Volumetric swelling measurements

Volumetric swelling ratio (Q_V) was defined as the mass of the swollen gel over the mass of the formed gel at the time of polymerization. Gels were formed as described above with 4 wt% 8-arm 20kDa PEGNB (PEG8a), 2mM nominal concentration of PHSRN-K-RGD, crosslinked with LW peptide. Gels were weighed immediately after polymerization was complete and transferred to a 24 well plate containing 750uL 1X PBS and incubated for 24h at 37°C. Gels were removed from PBS and excess water was gently removed with gloved fingertips before weighing to obtain the swollen gel mass.

Synthesis of PEG-grafted poly(propargyl-L-glutamate) (PPLGgPEG)

PPLG was synthesized as previously reported^71,72,73. To generate macromers with the described functionality, O-(2-aminoethyl)-O’-(2-azidoethyl)nonaethylene glycol or N3-PEG10-NH2 (32 mg, 0.06 mmol), N,N,N’,N’,N”-pentamethyldiethylenetriamine (PMDETA) (12.7 L, 0.06 mmol) and 5-norbornene-2-carboxylic acid N-hydroxysuccinimide ester (15.7 mg, 0.067 mmol) were dissolved in dry DMF (1.94 mL) for 1 hr. Targeting 10% grafting of norbornene functionality, PPLG (100 mg, 0.6 mmol alkyne repeat units), crude N3-PEG10-norbornene solution (2.0 mL at 0.03 mmol/mL, 0.06 mmol azide), and PMDETA (12.5 L, 0.060 mmol) were dissolved in DMF (1.0 mL). The copper bromide catalyst (8.6 mg, 0.060 mmol) was added to the degassed solution, and the reaction solution was stirred for 1 hr at room temperature under nitrogen. 2-(2-azidoethoxy)ethanol (94 mg, 0.72 mmol) was added or 1 hr under a blanket of nitrogen. Finally, the solution was precipitated in diethyl ether, dissolved in distilled water, and dialyzed against water acidified by HCl (pH = 5) for 24 h and against distilled water for 12 h. After lyophilization, the polymer structure and degree of substitution were verified by 1H NMR in DMSO-d6. Stock solutions of PPLG macromers were dissolved at 10 wt% in ultrapure water and stored at −80 °C until use.

PEG8a-PPLGgPEG hybrid gel fabrication

PEG8a and PPLGgPEG precursor solutions were combined at the indicated weight ratio and a total polymer concentration of 3 wt% with 2 mM nominal concentration of PHSRN-K-RGD at a stoichiometric crosslinking ratio of 0.5. Polymerization with or without encapsulated cells, imaging, and image analysis was carried out as described above.

Microfluidic device fabrication and culture

Inverse molds for microfluidic devices were 3D printed (FormLabs Form2 printer) according to the specified dimensions (Fig S9). Polydimethylsiloxane (PDMS) (Sylgard 184) was combined with curing agent at a ratio of 15:1, degassed for 30 minutes, and poured into printed molds. Devices were cured overnight under vacuum at 60°C b efore removal from molds. The bottom surfaces of devices were cleaned with tape to remove debris, and placed under oxygen plasma for 30 seconds along with glass slides. Devices were then immediately bonded to slides to create channels for injection of hydrogel and media. At the time of cell seeding, 15 uL of gel precursor solution with or without encapsulated cells was injected into the central channel of the devices and polymerized by UV light as previously described, with inversion after 20 seconds to ensure cells remained suspended. Following polymerization, 200 uL culture media was added to the inlet and outlet ports of each of the flanking media channels. Media in devices was exchanged every 24 hours.

In situ diffusion measurements

Microfluidic devices with gels prepared as described above were swollen overnight in endothelial cell media. At the start of experiments, a 1 mg/mL solution of 70kDa FITC-dextran in endothelial media was injected into each of the flanking media channels. Images were taken every 2 minutes at 3 sites along 4 separate gel channels to track dextran diffusion from the media channels into the central gel. Quantification of fluorescence intensity in these areas over time was used to generate diffusivity values for the different gel conditions as described elsewhere.

hCMEC/D3 brain microvascular endothelial cell culture

Human cerebral microvascular endothelial cells (hCMEC/D3), human immortalized blood-brain barrier endothelial cells, were purchased from MilliporeSigma and cultured according to manufacturer instructions. Briefly, collagen type I (Millipore 08–115) was diluted 1:20 in 1X PBS and added to flasks for at least 1 hour at 37°C. Ce lls were maintained in EndoGRO-MV complete medium (Millipore SCME004) with FGF-3 (Millipore GF003) at 1 ng/mL, with media changes every 48 hours. Passaging was performed when cells achieved 80–90% confluence. For experiments, cells were used between passages 3 and 6.

Luminex assay for measurement of local growth factor concentrations and secreted MMPs

Local concentrations of VEGF-A, FGFb, and EGF were measured using a commercially available Luminex kit (R&D systems). To measure the concentration of growth factors in the local gel microenvironment, hydrogels of the indicated compositions were fabricated as described previously and incubated with media supplemented with growth factors for the indicated time periods. Immediately afterwards, gels were removed from the media and transferred to microcentrifuge tubes for sortase mediated dissolution as described previously³¹. As reported in the reference above, sortase is necessary as it enables rapid and bio-orthogonal breakdown of the hydrogels to release biomolecules contained within the gel without damaging them. Briefly, gels were suspended in a solution of 18mM GGG and 100uM sortase A for 30 minutes at 37°C shaking at 600 RPM to allow sortase to cleave and cap the LPRT motif in each crosslinking peptide with a terminal GGG, completely dissolving the gels. Growth factors in the resulting dissolved gel solution were quantified through Luminex controlled against dissolution of a blank gel, which had been incubated in PBS. Quantification was performed against a standard curve using a 4 parameter logistical fit. For measurement of secreted MMPs, a commercially available Luminex kit (R&D systems) was used to quantify supernatant levels of MMP1, MMP2, MMP3, MMP7, MMP8, MMP9, MMP10, MMP12, and MMP13 in culture medium obtained from 3D endothelial cell cultures. Note that a neat and 10-fold dilution was required to capture concentrations of all analytes.