Ultrafast-relaxing and photopolymerizable PEG hydrogels enable viscoelasticity-mediated cell remodeling in synthetic matrices

Summary

Synthetic hydrogels provide powerful material platforms to engineer cellular microenvironments with control over stiffness, viscoelasticity, porosity, degradability, and biochemical signals. Here, we demonstrate how orthogonal crosslinking reactions allow fabrication of covalent adaptable networks to tailor photopolymerizable bioresin formulations relevant for tissue engineering. Specifically, we synthesize multifunctional poly(ethylene glycol) (PEG) macromers containing dynamic boronate ester bonds and dithiolane and norbornene moieties that allow for photopolymerization and projection-based biofabrication. These materials are used to print human mesenchymal stromal cells (MSCs) in formulations where the ratio of elastic versus adaptable crosslinks is engineered to study and manipulate MSC spreading, actin structure, and macroscopic material-level deformation. We demonstrate how material and print parameters, peptide ligands, actomyosin-modulating drug treatments, and cell types influence cell-material interactions and emergence of morphogenesis that is uniquely enabled by viscoelasticity. The presented materials introduce a versatile strategy for spatiotemporal control over dynamic mechanical properties in cell-laden matrices.

Introduction

Variations in viscoelastic properties of tissues define distinct niches in the body and influence mechanosensing processes relevant to development, regeneration and disease progression.¹ For comparison, cells residing in fat, liver, and brain experience matrix environments that relax in seconds, whereas muscle and skin can take hundreds to thousands of seconds to dissipate equivalent stresses. Matrix signals presented over different timescales provide cells with fundamentally different cues that can regulate function and fate.^2,3 With this in mind, biomaterial scientists have explored new types of viscoelastic matrices for 3D culture and to probe matricellular signaling. However, few matrices provide control over the relative content of permanent (elastic) and reversible (dynamic) crosslinks, which converge to shape both construct fidelity and cell-mediated remodeling. This requires material platforms in which these two crosslink types can be incorporated and tuned independently, providing control of the balance between stability and adaptability.

Protein-based hydrogels derived from the extracellular matrix (ECM) (e.g., collagen, Matrigel) are viscoelastic and readily promote cell-matrix and cell-cell interactions reminiscent of those observed in vivo. However, their application to the systematic study of cell responses to specific material properties is challenging, in part due to their ill-defined composition, interdependence of mechanical and biochemical properties, and topographical cues conferred by intrinsic fibrillar structures.^4,5 As a complement to naturally derived hydrogels, synthetic ECM mimics have facilitated new chemical approaches^6,7 to study mechanosensing,⁸ develop innovative biofabrication methods,⁹ and design controlled release profiles for diverse biomolecular cargo.^10,11 Synthetic hydrogels allow for well-defined molecular architectures by dictating the crosslinking chemistry, enabling precise structure-property relationships to be engineered.^12,13 Moreover, photoinitiated polymerizations confer spatiotemporal control over gelation and subsequent mechanical properties.¹⁴ By integrating a reversible, dynamic motif with an orthogonal photocrosslinking scheme, networks can be designed so that elasticity and viscoelastic stress dissipation are functionally decoupled.

For biomanufacturing methods, synthetic hydrogel formulations that allow integrated control of structural and spatiotemporal viscoelastic properties remain scarce. New formulations are needed that can allow experimenters to systematically probe how rapid, reversible mechanics influence cell behavior and to integrate such properties into light-based biofabrication workflows. Controlled hydrogel viscoelasticity has enabled the exploration of time-dependent cellular responses, where the timescale of relaxation can be matched to cell-generated forces to modulate mechanotransduction and morphological responses like spreading and polarization.^15–17 Numerous studies using 3D cell culture models have shown that viscoelastic hydrogels facilitate active cellular processes in a degradation-independent manner through stress dissipation and matrix rearrangement, allowing for spreading, migration, and morphogenesis of encapsulated cells and organoids without requiring irreversible changes in crosslinking density.^18–20 Despite evidence that faster dynamics promote more robust cellular responses,^21–23 most synthetic ECMs have been largely restricted to relatively slow relaxation profiles (typically half-times of hundreds of seconds or longer).²⁴ This limitation stems, in part, from processing challenges associated with highly adaptable networks, where rapid bond association and exchange complicates homogeneous mixing of multivalent network-forming precursors.²⁵ It also reflects the lack of systems where dynamic and permanent crosslinks can be deliberately combined in controlled ratios to interrogate how the balance between stability and adaptability influences both network mechanics and cell behavior in a spatially resolved manner.

Recent advances by Caliari and others have demonstrated that supramolecular photopolymerizable hydrogels (based on host-guest interactions) can unify ultrafast relaxation (i.e., stress dissipation on the order of five seconds or less) with cytocompatible, light-based cell encapsulation.^26–30 However, no analogous hydrogel platform utilizes dynamic covalent chemistries such as boronate ester exchange, which has similarly rapid timescales of relaxation but distinct frequency-dependent behavior.^31–34 Boronate ester-based networks have been explored in orthogonal photopolymerization schemes, but these efforts have typically relied on small molecule monomers that are incompatible with direct cell encapsulation.^35,36 As a result, there remains no boronate ester hydrogel system that simultaneously supports ultrafast relaxation, uniform and cytocompatible encapsulation, and user-directed photopolymerization. In contrast to reported host-guest systems, which may face solubility limitations when using small molecule crosslinkers, integrating complementary photocrosslinked and covalent adaptable networks using multifunctional macromers would enable modular tuning of viscous and elastic components, providing a versatile platform in which users can tune stress dissipation and mechanics while constructing patterned microenvironments that actively promote cellular remodeling and colonization.

To address these challenges, we introduce a PEG hydrogel platform that combines the dynamic covalent boronate ester chemistry with a photoinitiated dithiolane-ene polymerization, allowing for spatiotemporal viscoelastic network formation and light projection-based cell encapsulation. Popular step-growth photopolymerizations (e.g., thiol-ene) would require more challenging macromer design and photomediated chain-growth homopolymerizations (e.g., using acrylates) would lack selectivity over how each macromer is incorporated into the bulk network architecture. In contrast, the dithiolane-ene reaction confers both molecular control of network topology and rapid photocrosslinking through the formation of an elastic kinetic chain. By varying parameters such as resin formulation and light exposure, we demonstrate a balance of covalent and dynamic crosslinks that enables dramatic cell spreading and macroscale hydrogel compaction, while maintaining structural fidelity. Networks with insufficient elasticity (e.g., crossover between storage and loss moduli at higher frequencies) exhibit uncontrolled cell-driven remodeling and poor shape fidelity during culture, while networks with limited viscoelasticity retain their initial structure but restrict cell spreading. Moreover, fibronectin and N-cadherin mimicking peptides and actomyosin-modulating drug treatments are leveraged to show that these emergent cell-matrix interactions require both integrin binding and F-actin polymerization. Our approach establishes a mechanical design space where hydrogel stiffness and viscoelasticity can be tuned in concert to support both robust construct fidelity and rapid cellular remodeling, which are generalized in these matrices across multiple cell types, including myoblasts, fibroblasts, endothelial cells, and intestinal organoids. Furthering the utility of this system, these hydrogels are compatible with 3D printing methods such as digital light processing (DLP). This strategy enables a broad class of viscoelastic and cell-instructive materials, expanding the utility of synthetic hydrogels in a variety of bioengineering and biofabrication applications.

Results

Heterofunctional PEG macromers integrate dynamic and covalent crosslinking modes

To synthesize a photopolymerizable and ultrafast-relaxing hydrogel, we prepared a collection of multifunctional PEG macromers (Figure 1A). Briefly, eight-arm PEG-amine (20 kDa) was functionalized with approximately two lipoic acid groups per macromer via hexafluorophosphate azabenzotriazole tetramethyl uronium (HATU) coupling. This precursor macromer was divided into two equal portions and further modified on the remaining arms to yield complementary heterofunctional molecules with identical lipoic acid functionality. For the first macromer, residual amines were terminated with 4-carboxy-2-fluorophenylboronic acid (FPBA) via a second HATU coupling, producing PEG-boronic acid-lipoic acid (BLA). In parallel, a second macromer was functionalized with gluconolactone via base-catalyzed aminolysis, yielding PEG-gluconamide-lipoic acid (GLA). Next, PEG diamine (3.5 kDa) bearing a boc-protecting terminus was HATU-coupled to norbornene carboxylic acid, deprotected with trifluoracetic acid, and subsequently functionalized with gluconolactone in the same manner as the PEG-GLA, affording a heterobifunctional macromer (GNX). Additionally, a monofunctional PEG (2 kDa mPEG-amine) was prepared by aminolysis of gluconolactone for use as an inert competitor (GX).

Figure 1.

BELA hydrogel design. (A) BELA hydrogels consist of mixed heterofunctional PEG macromers. Eight-arm macromers are functionalized with approximately two lipoic acid groups and the remaining arms are converted to either boronic acid (BLA) or gluconolactone (GLA) residues. Heterofunctional linear PEGs are end-functionalized with gluconolactone on one side and either reactive norbornene (GNX) or unreactive methyl (GX) groups on the other. (B) Mixing BLA, GLA, and GNX (or GX) produces a soluble blend of oligomers linked by highly dynamic boronate ester crosslinks. Upon exposure to light in the presence of radical photoinitiator, rapid gelation occurs via dithiolane-ene polymerization.

Together, these molecules lay the framework for photopolymerizable and highly viscoelastic synthetic networks (Figure 1B) that we refer to as BELA (boronate ester/lipoic acid) hydrogels. While BLA and GLA macromer solutions spontaneously form viscoelastic gels, the addition of approximately stoichiometric GNX prevents gelation by competing with GLA for FPBA crosslinking sites on BLA. Importantly, GNX also participates in a photoinitiated free-radical dithiolane-ene polymerization between norbornene and lipoic acid (i.e., the lipoamide derivative of lipoic acid and PEG amine),³⁷ immobilizing all three macromolecular hydrogel precursors in a monolithic network upon exposure to light.

To image the effect of the competitor aspect of GNX, hydrogel precursor solutions were prepared at 10 wt% overall macromer containing a small quantity of a fluorescent PEG-norbornene, 0.1 wt% photoinitiator (lithium phenyl-2,4,6-trimethylbenzoylphosphinate; LAP), and either 1:1:1 BLA:GLA:GNX or BLA:GLA:mPEG-norbornene (i.e., no competitor). After curing these gels under 365 nm light (10 mW/cm²) for 5 minutes, the distribution of the fluorescent signal was assessed via widefield microscopy (Figure S1). As anticipated, hydrogels with only network-forming gluconolactone residues resulted in heterogeneous fluorescence, indicating an uneven distribution of the various macromer components throughout the matrix. In contrast, hydrogels containing GNX had a more homogeneous fluorescence intensity, denoting improved mixing prior to gelation and a more uniform network structure. These results illustrate the benefits of GNX as a competitive binder that maintains a soluble, well-mixed precursor solution, while also enabling subsequent photopolymerization to produce structurally and chemically uniform matrices. Previously reported dynamic hydrogels have relied on similar competitors for homogenization of crosslinking hydrazone bonds or complementary DNA strands,^38–40 but to our knowledge, this is the first example of a heterofunctional competitor that also acts as a photochemical crosslinker.

In situ mechanical properties of BELA hydrogels

Next, we explored the network evolution and resulting mechanical properties of BELA hydrogels prepared from various macromer concentrations to evaluate the accessible modulus range and to characterize the viscoelastic behavior under conditions of high conversion of the dithiolane-ene photopolymerization (effectively maximizing the ratio of covalent to dynamic bonds). Hydrogels were formulated at total macromer concentrations of 5, 10, or 20 wt% at a fixed stoichiometry (1:1:1 BLA:GLA:GNX) and containing 0.1 wt% LAP. All gels were cured under constant light intensity and duration (10 mW/cm² for 60 seconds). Oscillatory shear rheology was used to measure the storage (G’) and loss (G’’) moduli across this concentration range, providing insight into how network density influences both elastic and time-dependent responses (Figure 2A).

Figure 2.

Mechanical properties of BELA hydrogels with varied macromer concentration during network formation. (A) Photo-rheology of 5, 10, and 20 wt% BELA hydrogels containing 0.1 wt% LAP, spanning in situ shear storage moduli (G’) of hundreds of Pa to tens of kPa. Light on (365 nm, 10 mW/cm²) from 20 to 80s; measurements taken with 1% strain and a frequency of 1 rad/s. Inset shows the ratio of G”/G’ (tan d) which is consistently between 0.4-0.6 across gels of varied macromer concentration. ns = not significant, Welch’s one-way ANOVA test. (B) Normalized stress relaxation of cured BELA hydrogels (strain held at 10%). (C) The storage (G’) and loss (G”) modulus as a function of frequency for BELA hydrogels. n = 3 gels per condition.

First, monitoring the shear moduli during in situ photopolymerization confirmed that all formulations began as liquid precursors (G’ < 1 Pa, G’’ > G’), which then rapidly gelled upon irradiation with 365 nm light (10 mW/cm²), reaching stable plateau moduli (i.e., complete network formation) within 60 seconds of exposure. After curing, samples exhibited macromer concentration-dependent plateau moduli (mean ± SD, n = 3 gels): at 5 wt%, G’ = 620 ± 410 Pa and G’’ = 312 ± 172 Pa; at 10 wt%, G’= 7700 ± 430 Pa and G’’ = 3270 ± 140 Pa; and at 20 wt%, G’ = 11670 ± 1300 Pa and G’’ = 5970 ± 310 Pa. Interestingly, no significant differences were observed in the tan delta of these three formulations (approximately 0.5, measured at 1% strain and 1 rad/s), which may be a result of matched proportions of elastic (i.e., non-relaxing) and dynamic crosslinks in each condition. However, dramatic differences were observed in the stress relaxation behavior at higher deformations (10% strain), where the peak stress and immobile fraction of bonds both strongly correlated with the macromer concentration (Figure 2B and Figure S2). The 5 wt% hydrogels had a relaxation half-time of ~2 seconds and effectively dissipated the applied stress within 20 seconds, but the peak stress was only ~20 Pa. The 10 wt% hydrogels reached a peak stress of ~300 Pa, while maintaining a relaxation half-time of < 5 seconds, whereas the 20 wt% gels reached peak stresses of ~600 Pa but relaxed only one-third of the applied stress within 20 seconds. The overlap concentration of 8-arm 20 kDa PEG is approximately 13 wt%,⁴¹ suggesting that physical entanglements may contribute to these observations: similar dynamics at low strain (indicated by the measured tan delta) but significant differences in stress relaxation behavior at high strain.

To better characterize the time-dependent properties of these matrices, their frequency-dependent behavior was evaluated between 0.1-10 rad/s at 1% strain (Figure 2C). These frequencies were selected based on prior literature with boronate ester hydrogels containing network-stabilizing elastic crosslinks to observe the crossover between G’ and G’’.^31–33,35 Indeed, the frequency-dependent crossover for the BELA gels was observed for both 5 and 10 wt% gels at approximately 0.5 and 0.2 rad/s, respectively. Similar to their restricted relaxation behavior, 20 wt% BELA gels did not show complete crossover at any measured frequency.

In the interest of controlling viscoelasticity independent of concentration, we next explored how the formulation could be modified to shift the ratio of covalent to dynamic crosslinks. One simple approach was the omission of GLA entirely and relying on exclusively BLA:GNX interactions for both reversible and covalent bonding, effectively reducing the dithiolane content by a factor of two and limiting the extent of covalent crosslinking. This strategy yielded photopolymerizable and highly viscoelastic gels with markedly reduced storage modulus compared to the three-component system at matched polymer content, thus demonstrating how network topology controls matrix mechanics (Figure S3). Further exploiting this concept, we next varied the ratio of GNX to GX (the non-crosslinkable gluconolactone-terminated competitor), which also enabled predictable tuning of viscoelasticity at constant total macromer concentration and crosslinking conditions (Figure S4). Finally, we reinforced the elastic component of these networks by supplementing an intermediate formulation (1:1 GNX:GX) with an eight-arm norbornene-functionalized PEG macromer, thereby increasing the connectivity of covalent bonds. This modification enhanced elasticity (i.e., decreased the extent of stress relaxation), suppressed creep, and enabled the material to retain applied stress over repeated strain cycles, which was not observed in the analogous BELA hydrogel lacking the multifunctional norbornene macromer (Figure S5). These findings collectively demonstrate the flexibility of this macromer design approach for tuning hydrogel viscoelasticity by rationally controlling network topology.

BELA hydrogels support DLP-based cell encapsulation and undergo cell-mediated remodeling

Three-dimensional printing techniques allow for the crosslinking of patterned hydrogels through controlled light exposure of resins containing photoreactive macromers.⁹ Here, we used DLP as a biofabrication modality that relies on short light exposures and photoabsorbers to yield printed hydrogels that do not reach full conversion, but can be leveraged to tune processes such as cell-mediated hydrogel compaction. In particular, we identified several BELA formulations that span a range of final material properties and can serve as bioresins for DLP-based encapsulation of human mesenchymal stromal cells (hMSCs). Based on rheological characterization of fully cured BELA gels (Figure 2), we chose 10 wt% formulations for projection-based cell encapsulation and material printing, as this condition yielded matrices with both fast and extensive relaxation and storage moduli in the kilopascal range. Specifically, we prepared three BELA formulations with 1:1 GNX:GX and increasing amounts of eight-arm PEG-norbornene, which controlled the ratio of covalent to dynamic crosslinkers (C:D ratio; detailed formulations and molar ratios of functional groups are reported in Table S1). These polymer solutions were further modified for DLP-based printing and to support cell-matrix interactions by adding a photoabsorber (1 mM tartrazine) and thiolated RGD peptide (2.4 mM). LAP concentration was also increased to 0.5 wt% to maintain a similar photoinitiation rate, as LAP has a ~4-fold lower absorbance at 405 nm (used for DLP) compared to 365 nm used for our initial bulk hydrogel characterization.

Consistent with previous results, real-time tracking of the evolution of shear moduli for all three conditions showed liquid resins that quickly gelled upon exposure to visible light (Figure 3A). High and medium C:D gels were irradiated for 16 seconds and low C:D gels were exposed for a slightly longer time (24 seconds) to yield matrices with storage moduli ranging from approximately 600 to 2000 Pa. Of note, the low C:D hydrogel formulations led to a tan d >1 at 1 rad/s and 1% strain, indicating access to a unique high loss modulus regime in these materials compared to most previously reported dynamic hydrogel systems.²⁴ The frequency-dependent behavior (Figure 3B) showed decreasing crossover frequencies with increasing C:D ratios, illustrating how added elastic crosslinkers can shift the viscoelastic spectrum across a range of timescales.

Figure 3.

Encapsulated hMSCs actively compact DLP-printed BELA hydrogels. (A) Representative photo-rheology of 10 wt% BELA hydrogels with varied covalent to dynamic (C:D) bond ratio; gels contain 1 mM tartrazine and 0.5 wt% LAP. Light on (405 nm, 20 mW/cm²) for 16 s for medium and high C:D formulations and 24 s for low C:D formulation; measurements taken with 1% strain and a frequency of 1 rad/s. (B) Frequency response of BELA hydrogels of varied formulation, showing decreasing crossover frequency with increasing C:D ratio. (C) DLP-printed acellular BELA hydrogel after 48 hours in culture media, demonstrating gel homogeneity and retention of initial dimensions in the absence of cells. (D) 48 hour timelapse of collective hMSC remodeling of a low C:D BELA hydrogel, resulting in dramatic compaction. (E-F) 48 hour timepoints of medium (E) and high (F) C:D BELA hydrogels, showing increased cell spreading with increasing elasticity. Scale bars = 500 μm.

With these formulations in hand, we next assessed whether DLP-patterned hydrogels would retain sufficient cohesion and mechanical integrity in cell culture media, focusing on the softest and most viscoelastic condition (low C:D). We posited that the stiffer and less viscoelastic formulations would only show improvements in this regard (i.e., as more densely crosslinked networks, these would be slower to degrade). We projected patterns of various shapes (i.e., triangles, squares, ellipses, spatially varying arrangements of circles) to print well-defined, millimeter-scale features bound to acrylate-functionalized coverslips of uniform hydrogel thickness (250 μm, as used in our rheological characterization). In the absence of cells, the low C:D hydrogels were structurally stable, as defined by no shape changes over the course of 48 hours in cell culture medium (Figure 3C and Figure S6). This control was used as a baseline for comparison to cellularized samples. Beyond 48 hours, we expect these networks to remain structurally stable under culture conditions, as the PEG backbones are amide-linked and the boronate ester bonds are dynamic but not degradable. Potential interactions of boronic acids with soluble diols (e.g., glucose in culture medium) may alter the equilibrium crosslinking density over extended times, but no such effects were observed over the course of our 48 hour experiments.

The low C:D formulation was used for printing hMSC-laden constructs (5M cells/mL), where we observed dramatic cell-mediated matrix compaction of the hydrogel matrix in just 48 hours (Figure 3D). Quantifying these changes, the hydrogel surface area decreased by a factor of ~3 over this timespan, while the intensity of a network-bound fluorophore increased by ~50%, suggesting that compaction results in both cell and matrix densification (Figure S7). During the compaction process, cells extended and retracted protrusions that persisted on the order of hours and exerted a collective inward radial traction force detected by particle image velocimetry (PIV; Figure S8). Extended PIV analysis further revealed that displacement and velocity fields were most pronounced during the first 24 hours of culture and dropped to near zero thereafter, consistent with traction forces that plateau once compaction is essentially complete. These observations are similar to cell-driven condensation that occurs in fibrillar extracellular matrix proteins, such as collagen and Matrigel.⁴²

As the C:D ratio in the hydrogel network was increased, hMSCs showed greater spreading over 48 hours (Figure 3E–F), consistent with stable cell-matrix anchoring and force transmission in elastic matrices that are more resistant to cell-mediated deformation. Importantly, at the final timepoint, cells were viable (i.e., via a live/dead stain) and proliferative (i.e., via an EdU incorporation assay). For the latter, newly synthesized DNA was detected in 17% of encapsulated hMSCs in high C:D matrices over a 3-hour timespan and nearly 40% over 24 hours (Figure S9). Taken together, these results validate that BELA hydrogels support cell spreading, contraction, matrix remodeling, and proliferation under DLP-compatible fabrication conditions.

In addition to these proliferation assays, we examined whether non-uniformities inherent to DLP printing might confound the observed remodeling. Bulk rheology cannot capture depth-dependent crosslinking or gradients generated by light attenuation and surface immobilization, so we assessed this indirectly. An estimate of the axial light gradient confirmed a modest decay in dose through the z-axis, reaching a maximum of 30-35% attenuation at the upper gel surface (Figure S10). Consistent with this, cells appeared to polarize preferentially toward the upper gel surface, and compaction was more pronounced away from the coverslip, where immobilization reduces deformability. To further verify that the biological outcomes were not simply artifacts of printing heterogeneity, we encapsulated C2C12s in cast gels formulated without absorber and at reduced LAP concentration. These homogeneous gels underwent similar collective compaction events, albeit without the spatially programmed structures achievable by DLP. Together, these findings support that the observed cell-mediated remodeling behaviors arise from the intrinsic viscoelastic properties of the BELA network, rather than hidden heterogeneity introduced during printing.

Cell spreading and construct shape stability depend on viscoelasticity of BELA hydrogels

To more precisely evaluate how matrix mechanics regulate cellular and structural outcomes, we next quantified differences in cell spreading and gel remodeling across multiple printed hydrogels. Arrays of hMSC-laden and geometrically defined constructs (squares, triangles, and circles) were printed from each formulation and cultured for 48 hours (Figure 4A). Results demonstrated reproducible responses in both cell and matrix-level behavior in all geometrical shapes with a dependency on the C:D content. To elucidate the contribution of viscoelasticity apart from the network modulus/stiffness, we also included an elastic hydrogel control. The elastic control relaxed a much smaller fraction of the applied stress but had a matched polymer content and storage modulus to the high C:D gels (Figure S11). This control gel was generated by substituting GLA with 8-arm PEG-lipoic acid, which increases the density of covalent crosslinks while eliminating some moieties that participate in reversible boronate ester exchange. While this approach necessarily changes network topology, we matched peptide ligand density, polymer content, photopolymerization conditions, and approximate modulus at 1 rad/s to minimize compositional differences beyond crosslink type (Table S1). Thus, this elastic formulation provides a practical point of comparison for viscoelasticity-dependent cell responses. Minimal changes in matrix area or shape occurred when hMSCs were encapsulated in the elastic hydrogel formulation, as cells remained rounded even after 48 hours.

Figure 4.

Analyzing the influence of C:D ratio on hydrogel and cell-level spatial metrics. (A) Representative fluorescent images of acellular and hMSC-laden hydrogels, DLP-printed as various shapes in each BELA formulation and the elastic control after 48 hours in culture. Scale bars = 500 μm. (B) Quantification of hydrogel compaction based on change in projected area from acellular control. All viscoelastic conditions showed substantial area loss, while the elastic control remained essentially unchanged. (C) Shape similarity to original printed geometry (1 = perfect similarity); high C:D and elastic hydrogels had improved fidelity compared to the low and medium C:D formulations. (D) Actin:gel projected area ratio, showing a progressive increase in actin coverage with increasing elasticity among viscoelastic gels, but low spreading in the elastic control. (E) Mean distance to each nucleus’s three nearest neighbors, with no significant differences detected across all conditions. (F) Schematic of persistent homology analysis showing Betti-0 (H₀) and Betti-1 (H₁) feature tracking across increasing filtration sizes. (G) Heatmap of H₁ features binned by filtration size, indicating that characteristic loop size increases with elasticity. Data represent mean ± SD; statistical comparisons by Welch’s one-way ANOVA. ns = not significant, * p<0.05, ** p<0.01, *** p<0.001, **** p<0.0001. n ≥ 3 gels per condition.

For further quantitative analysis, we focused on the square hydrogel islands, as these provided a convenient balance of distinct angular features and maximum initial area for quantifying the extent of compaction. Measurements of projected hydrogel area revealed no statistically significant differences among the low, medium, and high C:D formulations, all of which compacted by at least 25% and in some cases up to 75% over 48 hours (Figure 4B). In contrast, the elastic control retained its original area, confirming that substantial cell-driven compaction occurs only in the presence of viscoelastic network dynamics.

To assess shape fidelity, we fit the perimeter of the largest object in each image using a Fourier descriptor and computed a rotational- and scale-invariant shape similarity metric relative to the acellular printed geometry (Figure S12). As expected, shape stability improved with increased elasticity, with both high C:D and elastic hydrogels showing significantly better preservation of the original printed geometry compared to the softer and more viscoelastic formulations (Figure 4C). Quantification of the projected actin:gel area ratio also revealed significant differences between the formulations (Figure 4D), with the low C:D and elastic gels displaying the lowest coverage (~10%), medium C:D increasing to ~45%, and high C:D reaching ~65%. This trend confirmed that greater matrix elasticity promotes cell spreading, whereas unbalanced stress dissipation or cell-confining elasticity restrict the relative actin area. Calculating the average distance to each cell’s three nearest neighbors did not identify any detectable variation across conditions (Figure 4E). While these metrics revealed condition-dependent trends, none could independently resolve all four formulations despite visually distinct features, prompting interest in topological analysis of actin distribution to better capture signatures of matrix-directed multicellular organization.

Complementing our other quantitative approaches, we explored topological data analysis (TDA). TDA has proven useful for distinguishing between spatiotemporal clustering modes of biological agents including epithelial cells, bacteria, honeybees, and birds.^43–46 Inspired by these studies, we developed a similar analytical approach for our hMSC-laden hydrogel structures and generated 2D point clouds from binarized versions of the actin channel. This data was randomly downsampled to 100-1000 points per image to facilitate computational tractability and filter out noise and redundancies. The original images typically contained 10-100 times more signal-positive pixels, but random down-sampling in this manner preserved the spatial distribution of the data, which is widely used in TDA. These reduced datasets are sometimes referred to as witness complexes. Notably, our chosen sampling ratio resulted in a signal density on the order of one point per cell. We viewed this as relatively ideal for our system, as individual cell boundaries cannot be reliably segmented for direct morphometric characterization because of high confluency and limited imaging resolution.

Point clouds were used to compute persistent homology by incrementally expanding the connectivity radius and tracking changes in topological features. This yielded Betti numbers, where Betti-0 (H₀) counts the number of connected components and Betti-1 (H₁) counts the number of enclosed loops, capturing both local clustering and more global architectural features of the cell population (Figure 4F). To visualize the distribution of these parameters, we processed three representative images of square hydrogel islands for each condition and binned filtration sizes (i.e., the diameter of the circular area around each point that defines the length scale of connectivity) in 5 μm intervals. By representing the number of features in each bin for both H₀ (Figure S13) and H₁ (Figure 4G) in a heatmap format, a clear trend emerged: the characteristic length scale of both metrics increased with matrix elasticity. Low C:D gels exhibited peak H₁ feature emergence (loops) around 25 μm, medium C:D around 40 μm, high C:D around 50 μm, and the elastic control showed a large shift to ~100 μm, consistent with more widely spaced actin structures in the absence of viscoelastic remodeling. For statistical comparison, we computed the mean filtration diameter associated with each Betti number for every image (bootstrapped across three rounds of independent random downsampling to ensure that the data reduction process did not influence statistical comparisons) and compared values using one-way ANOVA (Figure S14). When plotted against each other, H₀ and H₁ means displayed an approximately linear relationship across all conditions. We also evaluated the impact of sampling density, finding that 100–500 points per image yielded optimal discriminatory power among the four hydrogel formulations. While sampling with fewer (50) or more (1000) points still allowed for reasonable differentiation, these densities likely under- or over-represented the true spatial arrangement of cells reflected by actin signal distribution. These results bolster growing support for TDA as a useful tool for extracting meaningful organizational features from complex biological images without requiring sophisticated segmentation, and this analysis method was carried forward through the remainder of our studies.

BELA hydrogels afford exposure-based control of hMSC response in multi-property constructs

Given the strong dependence of cell behavior and matrix interactions on the viscoelastic properties of BELA hydrogels, we next manipulated the photopolymerization time to study and direct cellular response using a single BELA formulation. In situ photo-rheology of the high C:D formulation confirmed how extending light exposure time from 16 to 24 seconds, under otherwise identical conditions, can increase the final storage modulus, suppress frequency-dependent crossover behavior, and limit the extent of stress relaxation (Figure S15). Light dose similarly offers a simple and effective means for shifting the viscoelastic character of BELA hydrogels, enabling direct spatial patterning of mechanics without modifying the hydrogel composition.

As anticipated, encapsulation of hMSCs in the high C:D formulation had exposure-dependent differences in matrix remodeling, cell densification, and cytoskeletal organization (Figure 5A). Gels irradiated for 16 seconds compacted by more than 50% over 48 hours, whereas those irradiated for 24 seconds essentially retained their initial dimensions (Figure 5B). Actin coverage was similarly modulated, with the short exposure yielding more than double the projected actin:gel area ratio compared to the longer exposure (Figure 5C). Topological analysis reflected these differences, with the longer-exposure condition exhibiting nearly twice the mean filtration diameter for both H₀ and H₁ (Figure 5D), consistent with reduced local connectivity.

Figure 5.

Photopolymerization time modulates matrix mechanics and directs hMSC response within a single BELA formulation. (A) Representative images of high C:D BELA hydrogels encapsulating hMSCs and cured with either 16 or 24 seconds of visible light exposure. Scale bars = 500 μm (left, middle) and 100 μm (right). (B) Shorter exposure results in pronounced compaction over 48 hours, while longer exposure maintains original gel dimensions. (C) Actin:gel area ratios decrease with increased exposure. (D) Topological analysis shows substantially increased H₀ and H₁ mean filtration diameters with longer exposure, indicating reduced cellular connectivity. (E) hMSC-laden 4x4 mm hydrogel grid printed with exposure times from 10 to 25 seconds (1 second intervals) demonstrates a gradient of remodeling and cytoskeletal organization. Scale bars = 1 mm (top), 500 μm (bottom). (F) Macroscale construct patterned with alternating short and long exposures yields spatially distinct domains of compaction and actin architecture. Scale bars = 1 mm (top), 500 μm (bottom). Data represent mean ± SD; statistical comparisons by Welch’s t-test. ** p<0.01. n = 3 gels per condition.

To further characterize the range of cellular responses accessible with this approach, we patterned a 4x4 millimeter cellularized hydrogel grid using discrete exposure times from 10 to 25 seconds. Thus, using a single BELA formulation and a simple fabrication adjustment, we produced a continuum of hMSC behaviors with increasing exposure driving progressive reductions in compaction and actin coverage (Figure 5E). Projecting alternating short and long exposures in a greyscale pattern yielded a single construct with spatially distinct mechanical properties with well-demarcated zones of differential matrix remodeling and cell organization (Figure 5F). We also studied exposure modulation in the medium C:D formulation (Figure S16). In contrast to the high C:D condition, longer exposure in this intermediate formulation enhanced actin coverage while improving shape stability. Together, these results highlight that the regimes of cell spreading and matrix remodeling (ranging from uncontrolled compaction, to balanced spreading with optimized shape fidelity, to cellular confinement) are ultimately dictated by the balance of elastic and viscoelastic components, rather than by a single formulation. These results suggest that multi-exposure patterning of BELA hydrogels is a versatile strategy for fabricating multi-material constructs that provide simultaneously encapsulated cells with diverse physical cues.⁴⁷

hMSC-mediated remodeling of BELA hydrogels requires integrin binding and F-actin polymerization

Considering the robust cell-driven remodeling observed in high C:D BELA hydrogels, we next investigated the relative roles of integrin binding peptides, F-actin polymerization, and actomyosin-mediated tension in driving compaction (Figure 6A). To test the role of integrin binding, we compared hMSC responses in DLP-patterned high C:D hydrogels (16-second exposure) functionalized with 2.4 mM total thiolated peptide, using either RGD (integrin-binding), HAVDI (N-cadherin-mimetic), or a 1:1 mixture of the two. Previous studies have shown that HAVDI engagement with N-cadherin reduces hMSC mechanosensitivity by inhibiting cofilin phosphorylation and subsequent F-actin polymerization, while RGD promotes integrin activation and cytoskeletal assembly.^48–51 By varying the peptide ligand that facilitates cell-matrix binding, downstream functional outputs such as protein secretion and multicellular clustering can be modulated.^52,53 Consistent with these distinct signaling pathways, hMSCs in RGD-functionalized BELA gels adopted a spread morphology, whereas HAVDI-functionalized matrices supported limited spreading and minimal scaffold deformation (Figure 6B), resembling cell behavior observed in more elastic controls. Quantification of topological features revealed that RGD promoted denser actin connectivity, whereas HAVDI yielded sparse architectures, and the 1:1 ratio exhibited an intermediate responses (Figure 6C). These results complement prior studies, as BELA hydrogels allow one to probe cell responses to biochemical cues within microenvironments with controllable viscoelasticity, both temporally and spatially.

Figure 6.

Integrin binding and F-actin govern hMSC remodeling of BELA hydrogels. (A) General schematic illustrating molecular components involved in cell-matrix adhesion and cytoskeletal dynamics in these studies. (B) Representative images of hMSCs in high C:D BELA hydrogels (16 s exposure) functionalized with 2.4 mM total thiolated peptide as RGD, HAVDI, or a 1:1 mixture. RGD supported robust compaction and actin spreading, while increasing HAVDI concentration reduced these responses. Scale bars = 500 μm. (C) Topological quantification of actin organization across ligand conditions. (D) Representative images showing the effect of cytoskeletal inhibitors (blebbistatin, ROCK inhibitor (Y-27632), and cytochalasin D) on hMSCs in RGD-functionalized hydrogels. Scale bars = 500 μm. (E) Topological metrics detect distinct remodeling phenotypes: blebbistatin had minimal effect, ROCKi produced partial disruption, and cytochalasin D hindered actin connectivity in the same manner as eliminating integrin-binding (i.e., no RGD). n = 3 gels per condition.

To further probe outside-in signaling in BELA gels, small molecule inhibitors were introduced to cell-laden BELA gels functionalized with RGD to disrupt cytoskeletal processes and evaluate their contribution to matrix remodeling. Blebbistatin and ROCKi (Y-27632), which respectively inhibit non-muscle myosin II and ROCK,¹⁶ reduced cell spreading and altered matrix remodeling, yet compaction still occurred. In contrast, cytochalasin D, which blocks actin polymerization, completely prevented deformation and disrupted cell morphogenesis (Figure 6D). Topological analysis captured these differences: blebbistatin-treated samples resembled untreated controls, ROCK inhibition produced a mixed phenotype, and cytochalasin D resulted in sparse and unconnected actin structures (Figure 6E). The absence of RGD similarly abrogated compaction and significantly reduced actin coverage, consistent with integrin engagement being essential for matrix remodeling.

Collectively, these results suggest that BELA hydrogels elicit matrix remodeling through canonical mechanotransduction pathways, requiring integrin engagement and actin polymerization. However, the persistence of compaction under myosin or ROCK inhibition, combined with its loss under actin disruption, points to a primary dependence on actin-driven protrusion rather than contractility. Integrin binding-mediated mechanisms downstream of ROCK may facilitate compensatory F-actin polymerization (potentially via cofilin phosphorylation), but polymerized actin itself remains indispensable. These results are consistent with prior studies linking viscoelasticity sensing to canonical mechanotransduction pathways, including dynamic regulation of YAP/TAZ activity and actin turnover via cofilin phosphorylation.^19,48 In line with our observations of rapid protrusive activity and collective remodeling, it is very likely that these downstream signaling events are dynamically engaged and participate in biochemical feedback loops during compaction, though direct assessment was beyond the scope of this study.

Diverse applications of BELA hydrogels across multiple cell types

We next explored the generalizability of high C:D BELA hydrogels across multiple cell types and applications. Encapsulation of two widely used murine cell lines, C2C12 myoblasts and 3T3 fibroblasts, revealed distinct remodeling phenotypes. C2C12s produced more pronounced compaction over 48 hours, whereas 3T3s spread more sparsely over the same time period (Figure 7A). Both cell types formed highly spread multicellular networks, although there were apparent morphological differences between cell-laden samples (Figure S17), illustrating how the intrinsic balance between contractile and extensile properties modulates cellular responses to these viscoelastic matrices.

Figure 7.

Dynamic cellular responses in BELA hydrogels are generalizable across multiple tissue models. (A) C2C12 myoblasts and 3T3 fibroblasts exhibited distinct remodeling patterns in BELA matrices compared to hMSCs, reflecting intrinsic differences in phenotype with respect to contractility, proliferation, and spreading. Scale bars = 500 μm. (B) Co-encapsulation of HUVECs and hMSCs led to confluent and functional endothelial monolayers (i), marked by VE-cadherin expression (ii). Scale bars = 500 μm (left) and 100 μm (middle, right). (C) Centimeter-scale hMSC-laden BELA constructs patterned with anisotropic features directed multicellular organization (i) and local cytoskeletal alignment (ii). Scale bars = 3 mm (left), 500 μm (middle, right). (D) BELA hydrogels supported crypt-like morphogenesis (white arrowheads; organoids stained for nuclei in blue and F-actin in magenta) and Paneth cell differentiation of intestinal organoids, visualized via DEFA4-tdTomato (upper right). Scale bars = 50 μm.

To investigate whether BELA hydrogels could support self-organization of more complex multicellular structures, we co-cultured human umbilical vein endothelial cells (HUVECs) with hMSCs at a 2:1 ratio. After 48 hours, the encapsulated endothelial cells polarized to the surface of the hydrogels and formed a continuous VE-cadherin-positive monolayer supported by underlying mesenchymal cells (Figure 7B). This self-assembly is consistent with the emergence of functional endothelial barriers and lifelike interactions between vascular and stromal compartments. Then, as a demonstration of scalability, we used DLP to prepare centimeter-scale BELA construct containing embedded hMSCs. This macro-scale hydrogel with anisotropic features directed cytoskeletal alignment and cell spreading along patterned axes, further highlighting how construct geometry can influence supracellular organization, even over long length scales (Figure 7C).

As a final demonstration, we examined the potential for BELA matrices to support epithelial morphogenesis from intestinal stem cells, as a recent report from our group demonstrated that boronate ester-based PEG hydrogels within a similar modulus range facilitated intestinal organoid differentiation.⁵⁴ Here, undifferentiated intestinal stem cell colonies formed in Matrigel were transferred into BELA gels, where morphogenesis and differentiation were observed over the following 48 hours (Figure 7D). During this time, crypt-like structures emerged as marked by epithelial buckling and budding. Resident Paneth cells were used to define differentiated intestinal crypts and observed via a live fluorescent reporter of defensin alpha 4 (DEFA4-tdTomato) expression.⁵⁵ Together with results from myogenic, fibroblastic, stromal, and endothelial populations, these data highlight the versatility of BELA hydrogels for not only studying dynamic cellular behaviors but processing the cell-laden matrices into complex 3D scaffolds using light-based fabrication for a range of applications.

Discussion

In this work, we synthesized photopolymerizable PEG hydrogels that integrate dynamic boronate ester linkages with dithiolane-ene photopolymerization, yielding viscoelastic and modular synthetic biomaterials. Hydrogels with adaptive linkers and viscoelastic properties are of increasing interest as matrices for the culture of cells and regenerating tissues, and the presented boronate ester-based hydrogels provide a unique material chemistry with chemical versatility, advanced processing capabilities, and mechanical tunability. In studies complementary to the present work, Lin and coworkers developed boronate ester hydrogels that incorporated thiol-ene or dithiolane crosslinkers, allowing for stiffening via either elastic thioether formation or dynamic disulfide metathesis.^56,57 With respect to photocontrolled material properties, Kalow and coworkers demonstrated that azobenzene-functionalized boronic acids can reversibly modulate boronate ester equilibrium upon photoisomerization, enabling hydrogels with tunable and photo-switchable mechanics.^58–60 This advance represents a broader push toward developing hydrogels whose mechanical properties can be actuated through optical inputs.⁶¹ Additional approaches include azobenzene-based host-guest systems,^62,63 thiuram disulfide and allyl sulfide exchange,^64–66 as well as thiol-ene-based PEGylation of norbornene-functionalized alginate to enhance stress relaxation.^67,68 While these chemistries provide powerful routes for spatiotemporal modulation of hydrogel viscoelasticity, they have not been integrated into and likely require modifications for light-based additive manufacturing workflows such as DLP printing.

Although several reactions have been implemented for fabricating photopolymerizable viscoelastic hydrogels, such as thiol-ene reactions forming networks containing thioesters or dithioalkylidenes, these systems are much slower-relaxing than boronate esters and/or require processing conditions (e.g., organic solvents, small-molecule monomers) that limit their applicability to direct cell encapsulation and subsequent cell-matrix functional interactions (e.g., contractility).^69,70 In contrast, BELA hydrogels allow cell encapsulation, are inherently compatible with projection-based photopolymerization, and could be useful for other advanced biofabrication strategies like volumetric additive manufacturing or two-photon lithography.^9,71–74

From a biochemical perspective, unreacted norbornene or lipoic acid groups in BELA gels could be exploited for post-polymerization modifications using photoinitiated coupling or cleavage reactions, enabling staged mechanical reprogramming over user-defined timescales in biofabricated microenvironments. Collectively, these capabilities establish BELA hydrogels as an early example of a cytocompatible synthetic material that unites ultrafast relaxation and photocrosslinking in a platform that is easily tuned through multiple mechanisms. We expect that this type of integrated system will serve as a valuable tool for soft matter and tissue engineering researchers seeking to investigate the role of highly dynamic matrix mechanics in a variety of contexts where spatially controlled fabrication is useful, such as templating gels to optimize nutrient transport, an approach that has been shown to allow long-term culture and enhanced maturation of organoids.⁷⁵

In addition to morphological outcomes, patterned viscoelasticity has the potential to bias functional biological responses such as collective migration, spatially restricted differentiation, or juxtacrine signaling between cell populations residing in distinct mechanical domains. As has been established using a variety of hydrogel systems, mechanics can direct lineage specification and tissue organization, and we anticipate that BELA hydrogels will provide a versatile platform to probe many of these questions in the setting of controlled viscoelasticity. Here, we highlight representative patterns of differential spreading and compaction (Figure 5), and such observations suggest that BELA patterning could be extended to spatial control over more functional cell-cell communication and fate decisions in future work.

In this contribution, cellular behavior observed in BELA hydrogels was consistent with other reports of cells encapsulated in matrices with similar viscoelasticity and also enabled the discovery of new behaviors based on dynamic tailoring of the viscoelasticity (Figures 4 and 5). First, the fast-relaxing and photopolymerizable host-guest networks described earlier supported extensive spreading of encapsulated hMSCs.^26–29 Related work, using similar (but spontaneously crosslinked or slower-relaxing) materials, showed that hMSCs deposited more nascent ECM with increased matrix viscosity.^23,76 While not directly assayed here, such processes are very likely engaged in BELA constructs and represent an important direction for future studies. hMSCs have also been shown to maximally colonize and compact viscoelastic hydrogels when the residual elasticity (i.e., the immobile fraction of applied stress in relaxation studies) is on the order of 10-20%,⁷⁷ which is similar to the high C:D BELA hydrogels that best facilitated both cell spreading and construct shape stability.

When hMSCs are cultured in slower-relaxing hydrogels like those crosslinked by hydrazone bonds, spreading and collective organization are observed, albeit on longer timescales (days instead of hours).^21,23,78,79 In a complementary approach, soft contractile assemblies of photo-responsive fibrous hydrogels have also been engineered to be compatible with DLP-based cell encapsulation, affording qualitatively similar active densification of cells and matrix elements as what was observed in the BELA hydrogels (Figures 3 and 4).⁸⁰ However, a soft storage modulus alone (i.e., 1-1000 Pa) does not yield the type of cell-mediated remodeling seen with the BELA hydrogels in this study,⁸¹ demonstrating the critical role of viscoelastic properties to impart dramatic, collective cellular responses. We find that both the rapid stress relaxation and moderate stress retention of BELA hydrogels (Figure 3) fall within a biomechanically relevant regime that allows for active cellular remodeling and is consistent with emerging design principles for biomimetic synthetic matrices.^82–84

Beyond matrix mechanics, we used BELA hydrogels to modulate cell interactions through integrin- and cadherin-binding peptide ligands and pharmacological perturbation of the cytoskeleton. For example, HAVDI limited hMSC-matrix interactions as observed through lower spreading (Figure 6), similar to observations of hMSC morphologies in other peptide-functionalized viscoelastic gels.^85,86 Drug treatments that altered cytoskeletal structure and function provided further insight into how cells interact with these dynamic networks. Blebbistatin had minimal effects on cell spreading and gel compaction, while ROCK inhibition moderately reduced hydrogel colonization. In contrast, cytochalasin D completely blocked cell-mediated matrix remodeling, consistent with the essential role of actin polymerization underpinning supracellular organization.⁸⁷ These findings were somewhat expected, given the active dynamics of compaction and patterning processes.^88–90 However, it is notable that blebbistatin had little effect with dosing similar to or exceeding concentrations used to disrupt self-organization in muscle and kidney models,^91,92 suggesting either compensation through alternative contractile pathways or a greater dependency on F-actin than myosin activity in the BELA system. Altogether, the varied responses to peptide presentation and cytoskeletal inhibition highlight the complex interplay between cell-intrinsic programs and matrix adaptability in directing collective behavior and the usefulness of BELA hydrogels to study these interactions.

Morphogenetic processes exhibited by hMSCs in BELA hydrogels echo broader developmental trajectories mediated by mechanobiological feedback. The spontaneous emergence of polarization, compaction, and matrix condensation reflects a conserved capacity for cells to organize in response to viscoelastic environments, and explorations of tissue fluidity, adhesion-driven stiffening, and matrix remodeling in biological contexts signal general principles of force-guided patterning.^93–95 These behaviors suggest that BELA hydrogels may offer a useful model system for in-depth study of how simple physical cues guide cell and tissue organization and function, with potential relevance across co-culture systems, organoids, and disease models.^96–99 Continued refinement of these materials, alongside integration with engineered living components and synthetic biology strategies, lays the groundwork for constructing systems that blur the line between designer matrices and living tissues.^100–103

With respect to our analytical approaches (Figures 4–6), single-cell morphometry offers a compelling means to quantify physical features such as shape and orientation, enabling clear phenotypic distinctions across experimental conditions.^104–106 However, such measurements are typically used descriptively (i.e., to distinguish between experimental conditions), rather than as generative inputs into further analytical or predictive frameworks. In contrast, TDA (Figure 4) requires no segmentation or cell-level resolution and instead captures structural patterns inherent to the images themselves, reflecting collective cellular behaviors (in this study) without reliance on assumptions about individual object identity. Since cell morphology is a downstream reflection of intracellular programs,¹⁰⁷ including gene expression, metabolic activity, and cytoskeletal dynamics, future studies will focus on exploiting BELA hydrogels to investigate how these fundamental processes interact with material properties to govern multicellular behaviors. We hope that our TDA method for biological image analysis proves valuable for researchers with datasets that resist conventional quantification.

We also recognize there are limitations to our study that could be improved upon in the future. First, variability in the heterofunctional macromer synthesis and the complexity of this system (i.e., the number of components) can lead to variations in material properties with specific formulations. Moreover, the dithiolane-ene crosslinking reaction likely includes additional chemical pathways including dithiolane homopolymerization and chain-terminating thiol-ene crosslinking (e.g., single additions of dithiolanes to norbornenes). However, the underlying material properties remain easily accessible through modest re-optimization of the formulation, and the concepts illustrated throughout (tuning formulation elasticity by increasing covalent crosslinker or exposure time) can be generalized. Likewise, as with any hydrogel, numerous properties are intrinsically coupled through the crosslinking density and network topology, and these constraints make it effectively impossible to entirely decouple viscoelasticity from modulus (particularly in single-network gels).

In practice, changes to the ratio or connectivity of dynamic and covalent bonds inevitably alter multiple parameters simultaneously, including stiffness, relaxation timescale, frequency response, and network heterogeneity. For example, the elastic control formulation (generated by substituting GLA with 8-arm PEG-lipoic acid; Figures 4 and S11) reflects this coupling: although polymer content, ligand density, and modulus were carefully matched, differences in topology and crosslink distribution cannot be fully eliminated. Such interdependence is a general limitation of covalent hydrogels. Nonetheless, we emphasize that the same viscoelasticity-dependent trends can be reproduced within BELA hydrogels by adjusting the covalent:dynamic ratio and exposure time (e.g., Figure S16), supporting that the observed cellular behaviors are governed by matrix mechanics rather than the specific molecular architecture or chemical composition. Accordingly, we suggest that the rational molecular design of BELA hydrogels offers opportunities for systematic modulation of many of these material parameters, enabling one to investigate and harness multiple structure-property relationships in soft, dynamically crosslinked matrices.^108–110

It also remains difficult to precisely characterize local mechanical properties in the DLP-projected system. Bulk rheological measurements do not capture depth-dependent effects, such as gradients in crosslinking that can occur with light attenuation, diffusion of monomers, or surface immobilization from covalent bonding to the coverslip, all of which may contribute to heterogeneities in the cellular microenvironment throughout the matrix. Moreover, cell-mediated remodeling and densification of the matrix is a reciprocal interaction, which can alter the local modulus and viscoelastic properties over time. In this regard, microrheological approaches (e.g., optical tweezers, traction force microscopy) could provide spatiotemporally resolved insights into the initial distribution and evolution of local variations in mechanics. Finally, while this work focused on developing the BELA platform and demonstrating emergent cellular responses in cell-laden gels, there remain abundant opportunities to probe both subcellular and supracellular mechanisms in greater depth. Future studies will elucidate how specific molecular pathways sense and respond to dynamic material properties, as well as how multicellular systems biophysically reorganize these adaptable matrices over time.

A final consideration is long-term stability. Our study focused on 48 hour cultures, as the chemical components of BELA gels are not expected to undergo significant degradation on this timescale. Of further note, these macromers are unlikely to undergo irreversible degradation in culture media, as amides are hydrolytically stable and the boronic acid and gluconolactone motifs can readily re-associate if the boronate ester is hydrolyzed. While the thioether linkages from the dithiolane-ene reaction are stable, they could slowly oxidize to sulfones over long timescales. Thus, gradual softening or erosion of the gels is unlikely to influence the remodeling behaviors observed here. Rather, at extended culture times in the dense contracted networks, cell viability and nutrient transport might be expected to be limiting factors.

Conclusions

In summary, we developed BELA hydrogels as a photopolymerizable and viscoelastic material platform that features rapid stress relaxation on the timescale of seconds, modular mechanical tuning, and direct cell encapsulation that can be spatiotemporally templated. By exploring a range of formulations and processing conditions, we identified combinations of covalent and dynamic crosslinking that support dramatic hMSC remodeling and construct compaction, resulting in lifelike cellular dynamics and colonization of the encapsulating synthetic matrix. We also appraised the role of critical mechanosensing pathways at the cell-matrix interface and the cytoskeleton, finding that cells require both integrin binding and F-actin polymerization to remodel the surrounding hydrogel. Our results demonstrate that BELA hydrogels provide a highly configurable and cytocompatible microenvironment capable of eliciting complex cell behaviors relevant to a variety of biological applications.

Methods

Materials

Unless otherwise noted, all materials were purchased from Sigma-Aldrich. All PEGs were purchased from JenKem Technology USA with the exception of the 2 kDa monofunctional PEG amine (Biopharma PEG Scientific Inc.).

Macromer synthesis and characterization

Macromer syntheses generally followed a previously described protocol with some modifications.³⁴ In a typical synthesis of BLA and GLA, 1 g of 8-arm 20 kDa PEG amine was dissolved in DMF (5 mL per gram of PEG) and combined with lipoic acid (2.2 eq per macromer, i.e., approximately one of four arms functionalized), HATU (2.2 eq per macromer), and N-methylmorpholine (NMM, 4.4 eq per macromer). This reaction was protected from light and stirred for 48 hours before the PEG was collected by precipitation into excess ice-cold diethyl ether, centrifugation, and drying under ambient conditions. The partially modified crude product was split into equal portions for separate functionalization of the remaining endgroups. In the first portion, the remaining amines were capped with 4-carboxy-2-fluorophenylboronic acid (FPBA) by repeat HATU coupling in DMF (20 eq per macromer of both FPBA and HATU, 40 eq of NMM). The second portion was dissolved in methanol with gluconolactone and triethylamine (25 eq per macromer of each) and stirred for one week. The portion in methanol was concentrated by rotary evaporation, after which both portions were collected by precipitation in ether and centrifugation as above. Next, both macromers were dissolved in deionized water and purified by dialysis (MWCO = 8 kDa) against water for one week with water changes at least once per day, followed by flash freezing and lyophilization to yield dry off-white powder. GNX was prepared by a three-step synthesis. First, boc-NH₂-PEG-NH₂ (3.5 kDa) was coupled to norbornene carboxylic acid (2.2 eq per macromer) using HATU and NMM in DMF as before, then deprotected in 1:1 DCM:TFA to yield NH₂-PEG-norbornene, and finally functionalized with gluconolactone under the same aminolysis conditions used for GLA. The monofunctional competitor (GX) was prepared by reacting mPEG-amine (2 kDa) with gluconolactone and triethylamine in methanol in the same manner. Competitors were precipitated, dialyzed (MWCO = 1 kDa), and lyophilized prior to use. 8-arm PEG norbornene and 8-arm PEG lipoic acid (both 20 kDa) were prepared by HATU coupling in DMF (20 eq carboxylic acid and HATU, 40 eq of NMM) and collected and purified according to the same protocol as BLA and GLA. Macromer syntheses all had yields of ~80%, with losses due to incomplete precipitation. Macromer functionality was confirmed by ¹H NMR (Figure S18), with the exception of the gluconolactone endgroups, which are convoluted with protons associated with PEG; however, the aminolysis reaction is anticipated to be quantitative.

Hydrogel preparation

Macromers were dissolved at 20 wt% in sterile PBS and mixed to the desired final concentrations. For typical gel formulations, BLA, GNX, and GLA were combined in a 1:1:1 ratio by weight. To promote boronate ester complexation and prevent premature gelation, BLA was first mixed with GNX and PBS diluent, followed by addition of GLA and LAP (and cells, when applicable) to reach the final formulation. LAP was prepared as a 2 wt% stock in sterile PBS and added at the desired final concentration. For formulations with modified viscoelasticity or competition, GNX and GX were mixed at defined molar ratios as described. Increased elasticity was achieved by incorporating 8-arm PEG-norbornene at specified concentrations. The final formulations used for the DLP studies (low C:D, medium C:D, and high C:D) are detailed in Table S1.

Rheological characterization

Bulk hydrogel characterization was performed using a TA Instruments DHR-3 rheometer equipped with a parallel plate geometry consisting of an 8 mm sandblasted tool and a quartz curing stage equipped with a UV light guide accessory attached to an Omnicure 1000 light source (365 nm and 400-500 nm filter sets). 12.5 μL of macromer solution was placed onto the quartz plate and the probe was lowered to a height of 250 μm. A baseline oscillatory shear measurement (1% strain, 1 rad/s) was taken for 30s, then the light was turned on for variable time intervals depending on the sample (i.e., 60s for samples cured with 365 nm light, formulation-dependent exposures of either 16s or 24s using visible light, as reported throughout the main text). Following oscillatory shear measurements to monitor in situ gelation, other mechanical measurements were performed. Stress relaxation was measured by ramping to 10% strain over 2 seconds and holding for at least 2 minutes while recording stress decay. Frequency sweeps were performed at 1% strain with five points per decade from 0.1 to 10 rad/s. Creep tests were conducted by applying a constant stress of 100 Pa and monitoring strain over time.

Cell and organoid culture

Human mesenchymal stromal cells (hMSCs, PromoCell) were used for all in vitro experiments unless noted otherwise. hMSCs were expanded on tissue culture plastic in low-glucose Dulbecco’s modified Eagle’s medium (Gibco) supplemented with 10% v/v fetal bovine serum (Gibco), 1% v/v penicillin–streptomycin (Gibco), 0.5 μg/mL of Amphotericin B (Gibco), and 1.25 ng/ml basic fibroblast growth factor (R&D Systems). After expansion, hMSCs were passaged, using standard methods, into experiments at passage 3-8. The culture medium after hydrogel encapsulation consisted of high-glucose Dulbecco’s modified Eagle’s medium (DMEM, Gibco) supplemented with 10% v/v fetal bovine serum (Gibco), and 1% v/v penicillin–streptomycin (Gibco). C2C12s (CRL-1772, ATCC) and NIH 3T3s were cultured in growth medium containing high-glucose DMEM supplemented with 10% v/v fetal bovine serum, and 1% v/v penicillin–streptomycin. HUVECs were cultured in endothelial cell growth medium (EBM-2 basal medium containing EGM-2 SingleQuots supplement, Lonza). Murine small intestinal organoids were generated from crypts isolated from Lgr5-eGFP-IRES-Cre-ERT2 and Defa4-Cre-Rosa26-tdTom mice (Dempsey lab, CU Anschutz). Crypts were embedded in reduced growth factor Matrigel (Corning) and cultured in a defined stem cell expansion medium based on Advanced DMEM/F-12 (Thermo Fisher Scientific) supplemented with N2, B27, GlutaMax, HEPES, and penicillin–streptomycin. This basal medium was further enriched with EGF (50 ng/mL, R&D Systems), Noggin (100 ng/mL, PeproTech), R-spondin-conditioned medium (5% v/v, Organoid & Tissue Modeling Shared Resource, CU Anschutz), CHIR99021 (3 μM, Sell-eckchem), and valproic acid (1 mM, Sigma-Aldrich). Following encapsulation in BELA hydrogels, CHIR99021 and valproic acid were withdrawn from organoid medium to induce differentiation.

Bioink formulations

BELA hydrogels used for cell encapsulation were comprised of BLA, GLA, GNX (and other macromers as described in Table S1), in addition to lithium phenyl-2,4,6-trimethylbenzoylphosphinate (LAP, 0.5 wt%), tartrazine (1 mM), and Dulbecco’s Phosphate Buffered Saline (PBS). To visualize gel morphology, hydrogels were functionalized with thiolated fluorescein isothiocyanate (FITC) (GCDDD-fluorophore, 50 μM, Genscript). Thiolated RGD peptide (GCGYGRGDSPG, Genscript, 2.4 mM) was added to all formulations unless otherwise noted (i.e., where HAVDIGGGC, AAPPTec, 2.4 mM) or mixtures of the two peptides were used) and added to the resin from stock solutions. Cells were added to the resin prior to hydrogel fabrication to enable cell encapsulation for all experiments (final density: 5 million cells mL⁻¹). Colonies derived from single intestinal stem cells were transferred from 10 μL Matrigel droplets after three days of culture by washing with cold culture medium, centrifugation, and resuspension in 30-50 μL of BELA hydrogel precursor solution.

DLP biofabrication and organoid encapsulation

CAD models for solid shapes and complex geometries used in the study were designed in Fusion 360 (Autodesk, USA). Acellular and cell-encapsulating hydrogels were 3D printed on acrylated glass coverslips (22 mm, functionalized with (3-acryloxypropyl)trimethoxysilane, Millipore Sigma) using maskless projection stereolithography. Patterned illumination (exposure time = 16 s or 24 s, intensity = 20 mW cm⁻², wavelength = 405 nm) was projected using a Lumen Alpha DLP printer (Volumetric Inc.) into BELA bioink confined between the acrylate-functional coverslip and a hydrophobic coverslip with 250 μm-thick spacers. After printing, the hydrophobic coverslip was removed and hydrogels were washed with GX solution in sterile PBS (5 wt%) to remove excess resin and prevent viscoelastic gelation of unreacted macromers. Following this, coverslips with printed arrays of BELA hydrogels were transferred to well plates containing fresh cell culture medium. Organoid colonies were transferred to BELA precursor solution as previously described⁵⁵ and irradiated in a well plate in 10 μL hydrogel droplets before addition of differentiation medium. All samples were cultured for 48 hours before fixation, staining, and imaging.

Inhibitor study

Inhibitors were purchased from Cayman Chemicals. hMSC medium was supplemented with 50 μM blebbistatin, 5 μM ROCKi (Y-27632), or 5 μM cytochalasin D. Fresh medium containing inhibitor was replaced every 24 hours (i.e., immediately after encapsulation and again on the following day).

Proliferation and viability assays

To assess viability, cells were incubated with calcein AM (2 mM), ethidium homodimer-1 (4 mM), and Hoechst 33342 (1:1000) at 37 °C for 30 min following manufacturer’s protocol (Live/dead Cytotoxicity Kit, Thermo Fisher Scientific). EdU staining was similarly performed with Hoechst 33342 and a Click-iT Plus EdU Alexa Fluor 647 kit (Thermo Fisher Scientific) according to the manufacturer’s protocol. EdU staining was quantified by counting the number of EdU-positive cells relative to the total number of nuclei in FIJI.

Live imaging

For live imaging, hMSCs were incubated in CellMask Deep Red (Thermo Fisher Scientific) according to the manufacturer’s instructions prior to encapsulation. Cells were then encapsulated in the low C:D formulation and monitored using a Nikon Ti2-E widefield/spinning disk confocal microscope equipped with an OkoLab environmental chamber (37 °C, 5% CO₂). Images were acquired using a 10X/0.30 NA objective every 30 minutes for approximately 43 hours, capturing both the green fluorescence of the gel (tethered FITC fluorophore) and the cell membrane signal from CellMask, continuing until the compaction process was complete.

Cell fixation, staining, and imaging

To visualize cell morphology, samples were fixed 48 h post-printing with 10% v/v neutral buffered formalin, permeabilized with 0.1% v/v TritonX-100 for 20 min, and blocked with 5% w/v bovine serum albumin (BSA) for 1 h on a shaker. Organoids were fixed with 4% paraformaldehyde. After blocking, samples were stained with Alexa FluorTM 647 or rhodamine-conjugated phalloidin (1:500; Thermo Fisher Scientific) and Hoechst 33342 (1:500; Thermo Fisher Scientific) for 1 h on a shaker to visualize F-actin and the nucleus, respectively. To visualize endothelial cell junctions, fixed hydrogels encapsulating the HUVEC-hMSC co-cultures were stained with primary antibodies for VE-cadherin (Rabbit mAb, 1:200, D87F2, Cell Signaling Technology) overnight at 4°C on an orbital shaker, followed by multiple washes with DPBS and staining with an anti-rabbit fluorescent secondary antibody (1:500). Z-stack fluorescence images were acquired on a confocal microscope (Nikon Eclipse Ti2 with AXR scanner) using a 10X objective.

Image and statistical analyses

Multi-channel fluorescence images were analyzed in Fiji using custom macros to extract both local and global image features. Image stacks were processed by maximum intensity projection, and each channel was thresholded using Otsu’s method to segment relevant structures. Particle analysis was performed using defined size and circularity constraints, and the resulting ROIs were applied to the original (non-thresholded) images to measure area, mean intensity, centroid position, and shape descriptors. Additional macros were used to calculate total segmented area and area-weighted mean intensity for each image, allowing for comparison of overall signal distribution. Thresholded masks were further processed to generate outputs suitable for downstream shape and spatial organization analysis. Shape comparison was performed in Python using Fourier-based contour reconstruction followed by Procrustes alignment. Binary edge maps generated from thresholded gel fluorescence images were used as input. Contours were extracted using skimage.measure.find_contours, and the largest contour in each image was centered and rescaled. Each shape was reparametrized via arclength interpolation and smoothed using a truncated Fourier series (50 terms, 1000 interpolated points), then resampled to 200 uniformly spaced coordinates. Procrustes alignment was applied using scipy.spatial.procrustes to compare each sample to a reference shape, and similarity was defined as (1 – Procrustes disparity). Results were compiled across a batch of input files and exported to a .csv file for further analysis.

Approximate nuclear centroids were extracted from blue-channel segmentations and analyzed in R to quantify local cell density using 3-nearest neighbor (kNN) analysis. For each image, pairwise distances between each nucleus and its three nearest neighbors were computed using the FNN package. Edge lengths were calculated and summarized per image as the mean and standard deviation of kNN edge lengths. Summary statistics were calculated in R by grouping segmented ROIs by image and fluorescence channel. For each image, total area, average intensity, and area-weighted intensity were computed per channel. Area ratios between blue, green, and red channels were then calculated to assess relative cell content and gel organization. All metrics were compiled into a single table for comparison across samples.

Persistent homology was computed in R using the TDAstats package to quantify topological features (connected components and loops) in the actin channel of thresholded fluorescence images. Each binary image was converted to a point cloud by extracting (x, y) coordinates of pixels with intensity 255, then randomly downsampled to 500 points per sample (unless otherwise noted). Persistent homology was computed up to dimension 1 (H₀ and H₁). Resulting barcode diagrams were binned by filtration diameter (5 μm increments), with H₀ and H₁ counts accumulated per bin and visualized as CROCKER-style plots. Spatial persistence patterns were compared across conditions by analyzing the distribution of topological features over the filtration range. Mean H₀ and H₁ filtration values were computed as weighted averages, where each filtration bin’s contribution was scaled by the number of topological features present in that bin.

Particle Image Velocimetry (PIV) was performed using PIVlab in MATLAB on the time-lapse video of live cell compaction to visualize inward radial displacements. Actin fiber orientation (Figure 7C) was visualized using the OrientationJ plugin in FIJI to assess cytoskeletal alignment. All statistical analyses were performed in GraphPad Prism, with specific statistical tests denoted wherever relevant in figure captions.

Materials availability

There are restrictions to the availability of BELA macromers because of the lack of an external centralized repository for their distribution and our need to maintain the stock. We are glad to share BELA macromers, but we may require a payment and/or a completed materials transfer agreement if there is potential for commercial application.

Data and code availability

All data needed to evaluate the conclusions in the paper are present in the paper and/or the supplemental information. Additional datasets and code are available from the lead contact upon reasonable request. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.