Matrix Metalloproteinase–Mediated Degradation Governs Angioarchitecture within Poly(ethylene Glycol) Hydrogels

Abstract

Current tissue engineering strategies struggle to recapitulate the full diversity of tissue-specific vascular niches. Here, matrix metalloproteinase (MMP)-degradable poly(ethylene glycol) (PEG) hydrogels are used to define design principles governing vascular morphogenesis. Vasculogenesis was modeled by encapsulating human umbilical vein endothelial cells (HUVECs) in PEG hydrogels co-cultured with human mesenchymal stem cells (hMSCs), enabling the formation of 3D vessel networks over two weeks. Angiogenesis was modeled by entrapping HUVEC–hMSC spheroids within PEG hydrogels and assessing endothelial sprouting behavior. In parallel, comparative kinetic parameters for peptide degradation were calculated for vasculogenic (k_vasc) and angiogenic (k_ang) conditions; k_vasc values for IPES↓LRAG (8.8 × 10⁻⁷ hr⁻¹ cell⁻¹) and VPLS↓LYSG (1.4 × 10⁻⁶ hr⁻¹ cell⁻¹) were three-to-five-fold higher than GPQG↓IWGQ (2.9 × 10⁻⁷ hr⁻¹ cell⁻¹), indicating greater relative susceptibility to vasculogenic proteolysis. Consistent with these results, HUVECs encapsulated in hydrogels crosslinked with IPES↓LRAG or VPLS↓LYSG formed vessel networks with over two-fold greater diameter, length, and density compared to networks formed in GPQG↓IWGQ-crosslinked hydrogels under vasculogenic conditions, driven by greater overall degradability. In contrast, degradability influenced early sprouting dynamics under angiogenic conditions but did not affect overall network development. Additionally, vascular networks formed in non-degradable and GPQG↓IWGQ-crosslinked hydrogels exhibited compensatory upregulation of MMP and tissue inhibitor of metalloproteinase (TIMP) expression. Together, these findings elucidate how engineered extracellular matrix (eECM) degradability regulates vascular morphogenesis via cellular proteolytic feedback mechanisms, establishing a framework for designing eECM biomaterials that support tissue-specific microvascular modeling and improved vascular engraftment for regenerative medicine.

Graphical Abstract

1. Introduction

Microvasculature is necessary for supplying nutrients and oxygen to nearly all tissues in the body [1]. Tissue-specific vascular adaptations often manifest as unique angioarchitectural features essential for organ function. For example, the choriocapillaris, the densest portion of the retinal choroid, is a thin tissue (9.9 ± 1.7 μm) with a high vascular density (61.0 ± 0.7%), wide, flattened capillaries (19.1 ± 0.4 μm), and fenestrations [2–4]. These structural characteristics enable rapid nutrient and oxygen exchange, supporting the energy demands of the outer retina [5, 6]. In contrast, the capillaries of the blood-brain barrier form extremely tight barriers that exclude foreign substances, protect against inflammation, and maintain neural homeostasis [7]. The microvasculature of the cerebral cortex features small-diameter vessels (8.5 ± 11.5 μm), tight endothelial cell junctions, and lower overall volumetric density (about 3% throughout the brain) [8–12].

Despite the vast diversity of microvasculature, current tissue engineering approaches fail to capture these unique properties in models of vascular morphogenesis. Indeed, most methods for engineering microvasculature generate homogeneous vascular networks rather than tissue-specific vasculature. Consequently, tissue-engineered microvascular grafts often fail to support tissue function and/or regenerative outcomes due to poor host integration and structural mismatches from native tissue [13–15]. Hence, improved tissue engineering approaches that can recapitulate specific vascular niches are needed.

While natural biomaterials, such as collagen, fibrin, or hyaluronic acid, are commonly used to support vasculogenesis and angiogenesis, they are limited by the inflexibility of their chemical and physical properties, high batch-to-batch variability, and poor reproducibility [16, 17]. Alternatively, engineered extracellular matrices (eECMs) such as poly(ethylene glycol) (PEG) can be rationally designed to guide vascular growth by integrating adhesive, degradable, and mechanical cues. PEG is a biocompatible synthetic eECM biomaterial that enables precise control over biochemical cues via synthetic peptides (i.e., cell adhesion, degradability, growth factor integration) and biophysical cues (i.e., modulus, mesh size) to direct microvascular development through instructive eECM signaling [18, 19]. We have previously demonstrated that conjugating paracrine peptide factors to MMP-degradable peptide linkers enables their cell-mediated release from PEG hydrogels [20, 21]. Increasing the number of collagenase-sensitive peptide crosslinkers enhances the degradability of PEG hydrogels and promotes both fibroblast migration and endothelial cell angiogenesis [22–25]. Crosslinker peptide identity also affects vasculogenesis and angiogenesis in PEG hydrogels [26–28], but previous estimates of degradation rates have been poor predictors of vascular morphogenesis, suggesting that cell-specific differences in MMP responses to peptide crosslinkers could explain these inconsistencies [18]. Hence, there is a need to directly link effective cell-driven peptide degradation, rather than cell-type-agnostic catalytic efficiency, to predict vascular morphogenic outcomes.

This work investigates the use of a poly(ethylene glycol) (PEG)-based eECM to support vasculogenesis and angiogenesis in vitro [20, 29–31]. Thiol-ene chemistry was employed to crosslink hydrogels with MMP-degradable peptide crosslinkers and to tether the cell-adhesive peptide RGD [31–33] (Fig. 1A, B). Three different MMP-degradable crosslinkers with distinct predicted catalytic efficiencies were evaluated: GKKC-GPQG↓IWGQ-CKKG (GPQG) [34, 35], GKKC-IPES↓LRAG-CKKG (IPES) [35, 36], GKKC-VPLS↓LYSG-CKKG (VPLS) [35, 36] (↓ denotes the cleavage site) (Fig. 1C–E, Table 1) with a non-degradable PEG di-thiol crosslinker (ND) as a control. To model vasculogenesis, monodisperse human umbilical vein endothelial cells (HUVECs) were encapsulated in MMP-degradable eECM hydrogels [37] and co-cultured with human mesenchymal stem cells (hMSCs) to support vascular stability (Fig. 1F) [31, 38–43]. Angiogenesis was modeled by entrapping HUVEC-hMSC spheroids in MMP-degradable PEG hydrogels (Fig. 1G). An in vitro peptide degradation assay was used to quantify peptide degradation kinetics relevant to vascular morphogenesis, comparing soluble and membrane-MMP and TIMP expression in response to eECM MMP-degradability.

Fig. 1: Tunable PEG hydrogel engineered extracellular matrices can be used to emulate vasculogenesis and angiogenesis in vitro.

A) Hydrogel precursor mixtures containing 20 kDa 8-arm poly(ethylene glycol) norbornene, thiol-functionalized cell adhesive peptide GCGRGDSG, and di-thiol functionalized peptide crosslinkers are polymerized under UV with the photoinitiator lithium phenyl-2,4,6-trimethylbenzoylphosphinate (LAP). B) chemical structure for 8-arm poly(ethylene glycol norbornene, C) cell adhesive peptides, D) MMP-degradable peptide crosslinkers. E) Three MMP-cleavable peptides, GKKC-GPQG↓IWGQ-CKKG (GPQG), GKKC-IPES↓LRAG-CKKG (IPES), or GKKC-VPLS↓LYSG-CKKG (VPLS) were used to investigate if average catalytic efficiency correlated with cell-degradability and microvascular development. F) To emulate vasculogenesis, HUVECs were encapsulated in eECM PEG hydrogels and co-cultured with hMSC monolayers in 24 well plates. Light blue: PEG hydrogel. Pink: cell culture media. G) To model angiogenesis, HUVEC/hMSC spheroids were encapsulated in PEG hydrogels. Light blue: PEG hydrogel. Pink: cell culture media.

Table 1: MMP-specific kinetic catalytic parameters (k_cat/K_M, M⁻¹ s⁻¹) for GPQG, IPES, and VPLS.

MMP-specific catalytic parameters that have not been determined experimentally have been left blank. This table was adapted from Patterson et al. [35].

Peptide Sequence	MMP-1	MMP-2	MMP-3	MMP-7	MMP-8	MMP-9	MMP-10	MMP-14	Ref.
GPQG↓IWGQ (GPQG)	434	555	56.0	-	11,100	214	-	-	[34, 35]
IPES↓LRAG (IPES)	870	24,000	1,500	12,000	-	12,600	-	6,900	[35, 36]
VPLS↓LYSG (VPLS)	2,100	61,000	1,390	22,000	-	49,000	-	5,500	[35, 36]

2. Materials and Methods

2.1. Synthesis and characterization of 8-arm norbornene-functionalized poly(ethylene glycol)

20 kDa 8-arm poly(ethylene glycol)-amide-norbornene (8-arm PEG-NB) with ≥ 90% functionality was synthesized using carbodiimide chemistry, using a previously described four-step synthesis starting from 8-arm 20 kDa PEG-OH [31, 38, 44–48]. Briefly, PEG-OH was sequentially converted to PEG-mesylate (Fig. S1), then PEG-azide (PEG-N₃) (Fig. S2), and PEG-amine (PEG-NH₂) (Fig. S3) before norbornene functionalization. To synthesize norbornene-functionalized PEG (PEG-NB), N,N’-dicyclohexylcarbodiimide coupling was used as previously described [31, 38, 44, 45]. The resulting product was filtered, precipitated, and centrifuged to create a PEG-NB suspension. This suspension was dialyzed in deionized water (1000 MWCO dialysis tubing) then lyophilized to create solid 20 kDa 8-arm PEG-NB. Proton nuclear magnetic resonance (¹H NMR) was used to characterize 8-arm PEG-NB (Fig. S4).

2.2. Synthesis and characterization of peptides

The cell adhesive peptide ligand GCGRGDSG (RGD), the non-adhesive peptide control GCGRGESG (RGE), matrix metalloproteinase (MMP)-degradable crosslinkers GPQG: GKKC-GPQG↓IWGQ-CKKG, IPES: GKKC-IPES↓LRAG-CKKG, and VPLS: GKKC-VPLS↓LYSG-CKKG, and β-alanine modified peptides βA-GPQG: Ac-βA-K-GPQG↓IWGQ-K-βA-NH₂, βA-IPES: Ac-βA-IPES↓LRAG-βA-NH₂, and βA-VPLS: Ac-βA-K-VPLS↓LYSG-KK-βA-NH₂ (↓ denotes the cleavage site) were synthesized via solid-phase peptide synthesis using a Liberty Blue peptide synthesizer (CEM) on Fmoc-Gly-Wang resin (EMD), as previously described [21, 30, 31, 38, 49] (Table 2). Peptides on resin were collected after synthesis and cleaved using a cocktail made of trifluoroacetic acid (TFA, 92.5 % v/v, Alfa Aesar), triisopropylsilane (TIPS, 2.5 % v/v, Alfa Aesar), 3,6-dioxa-1,8-octanedithiol (DODT, 2.5% v/v, Alfa Aesar), and distilled, deionized water (ddH2O, 2.5 % v/v). For peptides containing arginine (R), the cleavage cocktail contained 90% TFA, 2.5% TIPS, 2.5% DODT, 2.5% ddH2O, and 2.5% thioanisole (v/v) (Alfa Aesar). Peptides were then vacuum filtered to remove resin, precipitated in cold diethyl ether, and centrifuged to collect precipitate.

Table 2: Synthesized peptide sequences and βA-modified peptide sequences for assessing degradation kinetics.

For βA-GPQG, and βA-VPLS, lysine residues were added so that their respective peaks could be more easily identified in LCMS readouts. ↓ denotes the MMP cleavage site [34–36].

Peptide Sequence	Function	Molecular Weight (Da)	Acronym
GCGRGDSG	Adhesive Peptide	708	RGD
GCGRGESG	Non-Adhesive Peptide	722	RGE
GKKC-GPQG↓IWGQ-CKKG	Degradable Peptide Crosslinker	1675	GPQG
GKKC-IPES↓LRAG-CKKG	Degradable Peptide Crosslinker	1675	IPES
GKKC-VPLS↓LYSG-CKKG	Degradable Peptide Crosslinker	1668	VPLS
Ac-βA-K-GPQG↓IWGQ-K-βA-NH2	Degradable Peptide (LCMS)	1025	βA-GPQG
Ac-βA-IPES↓LRAG-βA-NH2	Degradable Peptide (LCMS)	1281	βA-IPES
Ac-βA-K-VPLS↓LYSG-KK-βA-NH2	Degradable Peptide (LCMS)	1402	βA-VPLS

Matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) was used to verify peptide molecular weights (Fig. S5). Peptides were lyophilized and stored at −20 °C. For hydrogel formation, peptides were later dissolved in Dulbecco’s phosphate-buffered saline (DPBS) at 20 mg ml⁻¹ (for RGD and RGE) or 100 mg ml⁻¹ (GPQG, IPES, VPLS). Free thiol concentration was quantified using Ellman’s reagent (Thermo Scientific) dissolved in a buffer (0.1 mM sodium phosphate, 1 mM ethylenediaminetetraacetic (EDTA), pH 8.0) using L-Cysteine (Thermo Scientific) as a standard as previously described [38].

2.3. Synthesis of β-Alanine protected peptides for peptide degradation assays.

To isolate MMP-mediated degradation from non-specific exo/endopeptidase proteolytic activity, MMP-degradable peptides modified with β-Alanine (βA) were used (Table 2). In the absence of β-alanine, soluble peptides are degraded similarly due to non-specific peptide cleavage by exopeptidases and endopeptidases [50]. Hence, acetylated β-alanine (Ac-βA) was added to the C terminus, and β-alanine amine (βA-NH₂) was added to the N terminus to prevent non-specific degradation. Because crosslinking was unnecessary for the peptide degradation assays, thiol-containing cysteine residues were not included in the βA-modified peptides sequences to reduce disulfide formation [51]. To distinguish each βA-peptide in mass spectrometry, Lysine (K) amino acids were also introduced to the sequence to alter the molecular weights [50]. The respective modified peptide sequences for GPQG, IPES, and VPLS were as follows: βA-GPQG: Ac-βA-K-GPQG↓IWGQ-K-βA-NH₂, βA-IPES: Ac-βA-IPES↓LRAG-βA-NH₂, and βA-VPLS: Ac-βA-K-VPLS↓LYSG-KK-βA-NH₂ (↓ denotes the cleavage site) (Table 2).

2.4. Formation of PEG Hydrogels

To form PEG hydrogels, 4 wt% 8-arm PEG-NB, RGD and/or RGE, MMP-degradable crosslinkers, and 0.05 wt% of the photoinitiator lithium phenyl-2,4,6-trimethylbenzoylphosphinate (LAP), synthesized as previously described [52], were mixed in Dulbecco’s phosphate buffered saline then photopolymerized by exposing the gel precursor mixture to low-intensity long-wave UV light (5 mW cm⁻², 365 nm) for 3 minutes (Fig. 1A). Hydrogel physical properties including modulus, crosslinking density, and polymer density, were controlled by adjusting the proportions of crosslinkers. RGD concentration ranged from 0 mM to 5.0 mM (Fig. S6). For 0 mM, 1 mM, and 2.5 mM RGD hydrogels, the non-adhesive peptide RGE was included to maintain 5 mM total tethered peptide ratios. To mimic the differences in modulus and mesh size that exist in pro-vasculogenic versus pro-angiogenic ECMs [53–57], a lower modulus was used for the vasculogenesis-mimicking hydrogels, while a higher modulus and lower mesh size were used for the angiogenesis-mimicking hydrogels. MMP-degradable peptide concentration was 5.0 mM (65% crosslinking ratio) for vasculogenesis-emulating hydrogels, and 7.2 mM (80% crosslinking ratio) for angiogenesis-emulating hydrogels.

2.5. Mechanical characterization of PEG hydrogel properties

Unconfined compressive elastic modulus of acellular hydrogels was measured using an MTS Synergy 100 mechanical testing machine (MTS Systems, MN) equipped with a 2 N load cell, compressing between 5 and 10% of the initial hydrogel height at a rate of 0.1 mm s⁻¹. The average compressive elastic modulus was 6.1 ± 1.1 kPa for hydrogels used in the vasculogenesis assay (Fig. 2B) and 10.8 ± 1.0 kPa for hydrogels used in the angiogenesis assay (Fig. 3B). To calculate the mesh size ξ, the Canal-Peppas equation and the Flory-Rehner relation were used [58–65] (Equation S1). The swollen and dry mass and dimensions of PEG hydrogels were measured to calculate the polymer volume fraction φ_s and molecular weight between crosslinks M_c. The mesh size was 26 ± 3 nm for vasculogenic hydrogels and 18 ± 3 nm for angiogenic hydrogels.

Fig. 2: eECM degradability affects microvascular architecture in vasculogenesis-emulating hydrogel cultures.

A) HUVECs were encapsulated in PEG hydrogels then co-cultured with a feeder layer of hMSCs in the wells of a 24 well plate. B) Compressive modulus for PEG hydrogels crosslinked with GPQG, IPES, VPLS, or a non-degradable PEG crosslinker (ND). C) Representative Z-stack maximum projection confocal images depicting CD31+ PEG-encapsulated HUVECs forming microvessel networks in MMP-degradable PEG hydrogels. Scale bars: 200 μm. D) Average microvessel diameter, density, branching points, total network length, and average vessel length of PEG-encapsulated HUVEC networks on day 14. Statistics: one way ANOVA with Tukey’s post hoc comparisons test. Mean ± SD. N = 3-8. *: p<0.05. **: p<0.005.

Fig. 3: In angiogenesis-mimicking conditions, MMP-degradability is necessary for angiogenesis, and specific peptide sequences may regulate early HUVEC remodeling activity.

A) Schematics depicting spheroid sprouting assay. HUVEC-hMSC spheroids embedded in PEG are embedded in hydrogels and allowed to form vessel sprouts for 5 days. B) The compressive elastic modulus of hydrogels for the angiogenesis assay was about 11 ± 1 kPa across all four groups, with no significant differences. C) Confocal max projection images of HUVEC/hMSC spheroids, with cell trackers used to differentiate HUVECs (green) from hMSCs (magenta). In non-degradable (ND) hydrogels, HUVEC-hMSC spheroids grew very few sprouts. Vessel sprouts grew from HUVEC-hMSC spheroids encapsulated in hydrogels with degradable crosslinkers. Scale bar: 150 μm. D) Quantification of total network length, number of sprouts, and number of branches per sprout by HUVEC-hMSC spheroids with different crosslinkers. Mean ± SD. N = 9-12. Statistics: two-way ANOVA with multiple comparisons. *: ND vs. all degradable peptide groups. #: GPQG vs IPES. &: GPQG vs VPLS. $: IPES vs VPLS. *p < 0.05. **p < 0.01. ***p < 0.001. ****p < 0.0001.

2.6. Vasculogenesis analyses

To emulate vasculogenesis, HUVECs (Lonza) at passage 5 or less were cultured in endothelial growth media (EGM, M200500, ThermoFisher), 2% low serum growth supplement (LSGS, S00310, ThermoFisher), 2% fetal bovine serum (FBS, A5670701, ThermoFisher) and 1% antibiotic-antimycotic (15240062, ThermoFisher). Human mesenchymal stem cells (hMSCs) were isolated from bone marrow aspirates as previously described [66, 67]. hMSCs were cultured in low glucose Dulbecco’s Modified Eagle Medium (DMEM, 11885084, ThermoFisher) supplemented with 10% FBS and 1% antibiotic-antimycotic. hMSC feeder layers were seeded 1 day prior to gel formation at a cell density of 75,000 cells cm⁻² in the wells of 24-well tissue culture plates. Monodisperse HUVECs were included in the PEG hydrogel precursor mixture for all cell experiments at a cell density of 3×10⁶ cells ml⁻¹. 50 μL of the cell/hydrogel precursor solution was injected into truncated 1 mL syringe molds, which were placed under UV light to polymerize. Polymerized hydrogels were then cultured with hMSC feeder layers for 14 days in EGM, which was exchanged every other day.

2.7. Angiogenesis analyses

To model angiogenesis, HUVECs were cultured in EGM with 2% FBS, 2% LSGS, and 1% antimycotic-antibiotic. hMSCs were cultured in low glucose DMEM supplemented with 10% FBS and 1% antibiotic-antimycotic. 3D cell spheroids were formed using previously described methods [38]. Briefly, HUVECs and hMSCs were combined at a 1:1 ratio in EGM with 0.24 wt% methyl cellulose (M0512, Sigma). The cells were cultured in 96-well round bottom suspension plates overnight, such that there were about 3000 cells/spheroids/well. Prior to spheroid formation, HUVECs were incubated with 10 μM CellTracker Green (C2925, ThermoFisher), and hMSCs were incubated with 10 μM CellTracker Orange (C34551, ThermoFisher) to enable longitudinal monitoring of cell sprouting via confocal microscopy. Spheroids were then encapsulated in 30 μL hydrogels to evaluate for cell sprouting over 5 days. Confocal microscopy was used to image hydrogel cells on days 1, 2, 3, and 5.

2.8. Assessing peptide degradation via liquid chromatography mass spectrometry (LCMS)

To quantitatively assess differences in MMP degradation of the various MMP-cleavable crosslinkers, a previously developed liquid chromatography mass spectrometry method was used [50, 68]. For soluble peptide MMP-degradation studies, HUVECs, hMSCs, and HUVEC-hMSC were co-cultured in 48-well plates. HUVECs were plated at 68,000 cells cm⁻², hMSCs were plated at 32,700 cells cm⁻², and HUVEC/hMSC co-cultures were plated at 34,000 HUVECs cm⁻² and 16,300 hMSCs cm⁻². 24 hours after seeding, cell culture media containing 40 μM each of βA-GPQG, βA-IPES, βA-VPLS, and the non-enzymatically degradable internal standard NH₂-βF(βA)₆-NH₂ (βA is β-Alanine, βF is β-homophenylalanine). Notably, peptide degradation measured in these studies is an indirect measure of hydrogel remodeling, as peptides are free and not used as hydrogel crosslinks. 40 μL of media were collected and frozen at −80 °C at t = 0, 24, and 48 hours. βA-peptides were also added to HUVEC-conditioned media, hMSC-conditioned media, or HUVEC / hMSC co-culture-conditioned media to discriminate between soluble and membrane-bound MMPs. Immediately after extraction, samples were acidified using 4 μL of glacial acetic acid to halt peptide cleavage. 10 μL of crude solution was injected into a Thermo Scientific Liquid Chromatography System (ThermoFisher) equipped with a C18 column (ProntoSIL C18 AQ, 120 Å, 3 μm, 2.0 mm × 50 mm HPLC Column, PN 0502F184PS030, MAC-MOD Analytical Inc.) and analyzed via a Thermo Scientific LTQ XL Linear Ion Trap Mass Spectrometer (ThermoFisher) as previously described [68]. LCMS data analysis was performed using Xcalibur Software (Thermo Scientific). Mass/charge ratio (m/z) and expected retention time for each peptide were used to identify peptides and cleaved peptide fragments and isolate each peptide species from the total ion count (TIC) trace. All peptides were normalized to the internal standard NH₂-βF(βA)₆-NH₂. Area ratios, defined by the integrated peak area of a given peptide species divided by the internal standard, were used to calculate concentration of each peptide species [50, 68].

2.9. Assessing MMP and TIMP expression

Under vasculogenesis-emulating (Fig. 5) and angiogenesis-emulating (Fig. S13) conditions, cell-conditioned media were collected on days 1, 7, and 14 and frozen at −80 °C. Media samples were centrifuged at 3000 × g at 4 °C to remove any debris, and the supernatant was assessed for MMP and TIMP expression using the Human MMP and TIMP 12-Plex Discovery Assay^® Array for Cell Culture and non-blood samples (HMMP/TIMP-C,O) (Eve Technologies).

Fig. 5: Hydrogel degradability affected cellular MMP and TIMP expression by vasculogenic HUVEC / hMSC PEG hydrogel co-cultures.

A) Protocol for Luminex analysis of MMP and TIMP expression. B) Luminex results for MMP-1, MMP-2, MMP-3, and MMP-10. MMP-7, 8, 9, 12, and 13 were also tested but were below the detection limits of the Luminex assay. C) Luminex results for TIMP-1, TIMP-2, and TIMP-4. Statistics: Two-way ANOVA with Tukey’s post-hoc multiple comparisons within each time point. *: p<0.05. **: p<0.005. ***: p<0.0005. Mean ± SD. N=3. For multiple comparisons, *: VPLSLYSG vs. ND. %: VPLSLYSG vs. GPQGIWGQ. &: VPLSLYSG vs. IPESLRAG. $: ND vs. GPQGIWGQ. #: ND vs. IPESLRAG. @: GPQGIWGQ vs. IPESLRAG.

2.10. Analysis of vascular network development via immunohistochemistry

Vascularized hydrogels were fixed in 4% paraformaldehyde on day 14. Hydrogels were treated with a blocking and permeabilization buffer containing 0.25% Triton X-100 (X100-5ML, Sigma) / 1% bovine serum albumin (A9430-100G, Sigma) in DPBS overnight. Hydrogels were incubated with primary, secondary, and pre-conjugated antibodies in separate steps overnight at 4 °C. Between staining steps, hydrogels were rinsed three times with staining buffer. Detailed information about antibodies used in this work are listed in Table S1.

2.11. Quantification of vessel parameters based on confocal microscopy images

For vasculogenesis-emulating conditions, vessel positive areas of images were identified using endothelial cell markers CD31 or VE-Cadherin. Vessel parameters, including vessel diameter, density, total network length, number of branching points, and average branch length were measured using Fiji with the Vessel Analysis plug-in package (diameter) or AngioTool (density, network length, branching points, average branch length) for Z stack confocal images [69, 70]. For angiogenesis analyses, Z-stack images were quantified using the ImageJ “Sprout Morphology” plug-in to quantify cell sprouting in hydrogels. Because HUVECs and hMSCs were co-localized in angiogenesis-mimicking experiments (Fig. S9), both channels were used for angiogenesis sprout characterization. Cell trackers were used to distinguish HUVECs and hMSCs, respectively.

2.12. Statistical Analysis

All data were analyzed using GraphPad Prism, Microsoft Excel, and JMP Pro. One-way ANOVA or two-way ANOVA with Tukey’s post hoc test was used to identify significant differences between various groups. All data are expressed as mean ± SD unless otherwise noted. In the vasculogenesis-mimicking assay, a single hydrogel with encapsulated cells was considered a discrete experimental unit (technical replicate). In the angiogenesis-mimicking assay, a single spheroid was considered a discrete experimental unit (technical replicate). Only one image from each experimental replicate was treated as an independent replicate for quantification purposes. For the in-situ peptide degradation assay, a single well was treated as a discrete experimental unit (technical replicate).

3. Results

3.1: MMP degradability increases vasculogenic vessel diameter, density, and network length

To mimic vasculogenesis in vitro, HUVECs were encapsulated in PEG hydrogels and co-cultured with hMSCs in 24 well plates (Fig. 2A). During embryonic growth, when the ECM has less collagen, a lower overall modulus, and less interstitial crosslinking, vasculogenesis drives most vascular development [71]. To emulate pro-vasculogenic ECM, the modulus of the hydrogels was 6.1 ± 1.1 kPa, and the crosslinking ratio was 65% (Fig. 2B). HUVECs encapsulated in PEG hydrogels with 5 mM RGD generated vessels with higher average diameter and density compared to 1 mM RGD or 2.5 mM RGD hydrogels, so 5 mM RGD was used for all other vasculogenic hydrogels (Fig. S6).

Non-degradable (ND) hydrogels had the lowest levels of vessel-like structural development, with minimal average vessel diameter, network density, and network length; rounded CD31+ cells that were unable to form vascular networks were visible in non-degradable hydrogels (Fig. 2C, D). In contrast, HUVECs encapsulated in hydrogels formed with MMP-degradable peptide crosslinkers, GPQG, IPES, or VPLS, formed distinct 3D microvessel networks after 14 days (Fig. S7). The average vessel diameters for networks within PEG hydrogels crosslinked with either IPES or VPLS were 2-fold higher compared to hydrogels crosslinked with GPQG (IPES: 12 ± 3 μm, VPLS: 13 ± 1 μm, GPQG: 6 ± 6 μm) (Fig. 2D, Table S2). Total network length was 3-fold higher for VPLS and IPES crosslinked hydrogels compared to GPQG crosslinked hydrogels (IPES: 30,000 ± 10,000 μm, VPLS: 30,800 ± 9,700 μm, GPQG: 13,000 ± 11,000 μm). Similarly, the average length per vessel was about 3-fold higher for HUVEC networks in IPES- and VPLS-crosslinked hydrogels than in GPQG-crosslinked hydrogels (IPES: 1,300 ± 1,000 μm, VPLS: 1,020 ± 360 μm, GPQG: 310 ± 300 μm).

Average vessel density and number of branching points were used to assess vascular network complexity. HUVECs formed networks in IPES- and VPLS-crosslinked hydrogels developed over 2-fold as many vessel branches compared to networks in GPQG-crosslinked hydrogels (IPES: 280 ± 130 branches, VPLS: 260 ± 120 branches, GPQG: 113 ± 90 branches) (Fig. 2D). The average vessel density (total vascular area/tissue area) of vascular networks in IPES- and VPLS-crosslinked hydrogels was 2-fold higher compared to GPQG hydrogels and 5-fold higher compared to networks in non-degradable hydrogels (IPES: 33 ± 11%, VPLS: 33 ± 9%, GPQG: 15 ± 11%, ND: 6 ± 6%) (Fig. 2D, Table S2).

3.2: MMP-degradability is necessary for angiogenesis, and degradation regulates early remodeling

To assess the impact of MMP-degradable crosslinkers on angiogenic behavior, spheroids formed from HUVECs and hMSCs were encapsulated in PEG hydrogels, and angiogenic sprouting was assessed over time (Fig. 3A) [38]. Compared to the vasculogenic model, crosslinking was increased to 80% to increase the modulus to 11 ± 1 kPa (Fig. 3B), thereby recapitulating the properties of a mature ECM in which angiogenesis would occur [53–55]. Cell trackers were used to distinguish HUVECs (green) from hMSCs (magenta) (Fig. 3C).

Total network length was assessed to characterize the ability of new microvessels to sprout and migrate into the bulk of the eECM via degradation. The total number of sprouts and the average number of branches per sprout were used to characterize the effects of eECM degradability on microvascular density and complexity. Consistent with the results of the vasculogenesis assay, non-degradable hydrogels did not support vessel network formation, with minimal total network length (70 ± 80 μm), sprouting (4 ± 2 sprouts), and branches per sprout (0 ± 1 branches), averaged through day 5. Interestingly, total network length was not affected by peptide crosslinker identity. By day 2, spheroids encapsulated in GPQG-, IPES-, and VPLS-crosslinked hydrogels developed networks with similar average total network lengths of about 6,100 ± 2,600 μm, each significantly exceeding the minimal vessel networks of non-degradable hydrogels (GPQG: 7,100 ± 2,500 μm, IPES: 5,700 ± 3,200 μm, VPLS: 5,600 ± 2000 μm, ND: 70 ± 60 μm) (Fig. 3D, Table S3). By day 5, the average total network length for all MMP-degradable hydrogels increased to 22,000 ± 4,800 μm.

Early measurements of sprout number indicate that GPQG- and VPLS-crosslinked hydrogels supported early angiogenic sprouting, whereas the vascular density in IPES-crosslinked hydrogels was lower. On day 1, spheroids within GPQG- and VPLS-crosslinked hydrogels formed about 50% more sprouts compared to IPES-crosslinked hydrogels (GPQG: 32 ± 3 sprouts, VPLS: 31 ± 10 sprouts, IPES: 21 ± 10 sprouts) (Fig. 3D, Table S3). After the initial sprouting stage, the number of vessels decreased by 40% in GPQG-crosslinked hydrogels and by 20% in VPLS-crosslinked hydrogels by day 5. In contrast, the number of sprouts in IPES-crosslinked hydrogels remained stable. (GPQG: 20 ± 5 sprouts, IPES: 22 ± 4 sprouts, VPLS: 25 ± 8 sprouts). The number of branches per sprout was consistent among all MMP-degradable hydrogels at most time points, reaching about 13 ± 8 branches by day 3 and increasing to 29 ± 10 branches by day 5, indicating that peptide crosslinker identity had little impact on microvascular complexity. Altogether, these data confirm that ECM MMP-degradability is necessary for angiogenesis, but differences in peptide identity do not impact overall vessel architecture.

3.3: MMP-degradable peptides exhibit cell–type–specific differences in susceptibility to proteolytic cleavage

To quantify proteolytic susceptibility of the various MMP-cleavable crosslinkers to cell-type specific membrane-bound and soluble MMPs, a cell-based peptide degradation assay was used [50, 68]. Soluble βA-peptides were introduced to HUVECs, hMSCs, and HUVEC-hMSC co-cultures (Fig. 4A, Table 2). To isolate soluble MMP activity, βA-peptides were also incubated with HUVEC, hMSC, or HUVEC/hMSC co-culture conditioned media to exclude cleavage by membrane-bound MMPs, enabling differentiation between vasculogenesis- and angiogenesis-emulating degradation activity (Fig. 4B). To quantify peptide proteolysis, the effective degradation coefficient k_obs, which estimates catalytic activity, was calculated using the equation (k_obs = −ln(C_t / C_o) / Δt), where C_t represents the concentration at a given timepoint, C_o is the initial concentration, and t represents time (Fig. 4C, Table S4).

Fig. 4: Effective peptide degradation rate correlates with sequence specific catalytic efficiency k_cat/K_M. βA-IPES and βA-VPLS are more readily degraded in vasculogenesis emulating conditions.

A, B) βA-modified peptide sequences were incubated with HUVECs, hMSCs, or HUVEC/hMSC co-cultures or conditioned media conditions. LCMS was used to measure the remaining relative amount of full-sized βA-peptide in the media at t=0, 24, and 48 h. Made with Biorender. C) The effective degradation coefficient k_obs (hr⁻¹) was calculated for each peptide in cell culture or conditioned media conditions using LCMS-based peptide degradation data. Average values for k_obs ± SD are available in Table S4. D) k_Δ ± SD for hMSCs, HUVECs, and HUVEC/hMSC co-cultures, reflecting differences in proteolytic activity in cell culture conditions versus cell conditioned media conditions. k_{Δ, Cells} = k_{obs, Cells + Media} - k_{obs, CM}. E) k_vasc ± SD and k_ang ± SD for βA-GPQG, βA-IPES, and βA-VPLS were calculated using k_obs. k_vasc = k_HUVECs + k_{hMSC conditioned media}. k_ang = k_{HUVECs +media} and k_{hMSCs + media}. Mean ± SD. N = 11-14. Statistics: Two-way ANOVA with Tukey’s post-hoc multiple comparisons. *p < 0.05. **p < 0.01. ***p < 0.001. ****p < 0.0001.

βA-IPES and βA-VPLS were degraded more rapidly than βA-GPQG. For all cell culture conditions, βA-GPQG exhibited the lowest average degradation rate; the average k_{obs, Cells + Media} for βA-IPES and βA-VPLS were over 3-fold higher than βA-GPQG (βA-IPES: 0.03 ± 0.02 hr⁻¹, βA-VPLS: 0.04 ± 0.02 hr⁻¹, βA-GPQG: 0.02 ± 0.01 hr⁻¹) (Fig. 4B, Table 3, Table S5). Similarly, in conditioned media, in which degradation was limited to soluble MMPs only, βA-GPQG had the lowest overall degradation rate. Compared to βA-GPQG, average k_{obs, Conditioned Media (CM)} was about 7-fold higher for βA-IPES and nearly 6-fold higher for βA-VPLS (βA-GPQG: 0.01 ± 0.01 hr⁻¹, βA-IPES: 0.07 ± 0.05 hr⁻¹, βA-VPLS: 0.06 ± 0.02 hr⁻¹) (Fig. 4B, Table S4).

Table 3: Calculations for k_vasc and k_ang.

k_vasc represents vasculogenesis, including soluble and membrane-bound MMPs from HUVECs and soluble MMPs only from hMSCs (k_vasc = (k_{HUVECs + media} + k_hMSC-CM) / # of cells) (hr⁻¹ cell⁻¹) (Fig. 4E). k_ang represents angiogenesis, including the soluble and membrane-bound MMPs from both HUVECs and hMSCs (k_{ang, 1} = (k_HUVECs+Media + k_hMSCs+Media) / # of cells) (hr⁻¹ cell⁻¹).

Calculated kinetic parameter	βA-GPQG	βA-IPES	βA-VPLS
kvasc (hr−1 cell−1) × 10−7	2.9 ± 2	8.8 ± 3	13.6 ± 1
kang (hr−1 cell−1) × 10−7	2.7 ± 1	5.5 ± 2	8.1 ± 1

The degradation rate of different MMP-degradable peptides was affected by specific cell type and exposure to membrane-bound MMPs. The degradation of βA-IPES was similar in all cell cultures, but 2-fold faster in HUVEC-conditioned media and HUVEC + hMSC-conditioned media compared to hMSC-conditioned media, suggesting that secreted MMPs from HUVECs had a greater specificity for the degradation of βA-IPES (k_{obs, HUVEC-CM}: 0.08 ± 0.05 hr⁻¹, k_{obs, hMSC-CM}: 0.04 ± 0.02 hr⁻¹, k_{obs, Co-Culture-CM}: 0.08 ± 0.05 hr⁻¹) (Fig. 4C, Table S4). For βA-VPLS, k_obs was about 2-fold higher in hMSC cultures and hMSC-conditioned media compared to HUVEC cultures, HUVEC + hMSC co-cultures, HUVEC-conditioned media, or HUVEC + hMSC-conditioned media, indicating that MMPs specific to hMSCs have a higher specificity for βA-VPLS degradation (Fig. 4C, Table S4). The degradation of βA-GPQG was similar in cell cultures and conditioned media regardless of cell type (Fig. 4B, Table 3).

To determine differences in peptide susceptibility to soluble or membrane-bound MMP degradation, we calculated k_{Δ, Cells}, which represents cell membrane-bound MMP activity, using the formula (k_{Δ, Cells} = k_{obs, Cells + Media} - k_{obs, CM}). k_{Δ, Cells} was negative for both βA-IPES and βA-VPLS, indicating that membrane-bound MMPs had a low impact on degradation (βA-IPES: k_{Δ, Cells} = −0.04 ± 0.05 hr⁻¹, βA-VPLS: k_{Δ, Cells} = −0.01 ± 0.03 hr⁻¹) (Fig. 4D, Table S5). The higher degradation rates in conditioned media (without cells) suggest that HUVECs and hMSCs regulate MMP activity in cell cultures by upregulating TIMPs, leading to reduced peptide cleavage [72–74].

To correlate peptide degradation with the observed vascular activity in the vasculogenesis and angiogenesis-mimicking conditions, k_vasc and k_ang were calculated using k_obs. To capture vasculogenesis, k_vasc was calculated as the sum of k_{obs, HUVECs + media} and k_{obs, hMSC-CM} divided by the number of cells, representing the soluble MMP and membrane-bound MMP contributions of HUVECs and the soluble MMP contributions of hMSCs (k_vasc = (k_{obs, HUVECs + media} + k_{obs, hMSC-CM}) / number of cells). In the spheroid-based angiogenesis-mimicking conditions, both membrane-bound MMPs and soluble MMPs from HUVECs and hMSCs contributed to the degradation of the eECM. Hence, k_ang was calculated as the sum of k_{obs, HUVECs + media} and k_{obs, hMSCs + media} normalized to the number of cells (k_ang = (k_{obs, HUVECs + media} + k_{obs, hMSCs+media}) / number of cells). k_vasc was lowest for βA-GPQG, about 3-fold lower than βA-IPES and 5-fold lower than βA-VPLS (βA-GPQG: 2.9 × 10⁻⁷ hr⁻¹ cell⁻¹, βA-IPES: 8.8 × 10⁻⁷ hr⁻¹ cell⁻¹, βA-VPLS: 1.4 × 10⁻⁶ hr⁻¹ cell⁻¹) (Table 3). Similarly, k_ang was lowest for βA-GPQG, about 2-fold lower than βA-IPES and 3-fold lower than βA-VPLS (βA-GPQG: 2.7 × 10⁻⁷ hr⁻¹ cell⁻¹, βA-IPES: 5.5 × 10⁻⁷ hr⁻¹ cell⁻¹, βA-VPLS: 8.1 × 10⁻⁷ hr⁻¹ cell⁻¹) (Fig. 4E, Table 3). Both k_vasc and k_ang were lower for βA-GPQG than for βA-IPES and βA-VPLS, consistent with its slower overall effective degradation. For both βA-IPES and βA-VPLS, k_vasc was about 40% higher than k_ang, while k_ang and k_vasc were similar for βA-GPQG.

3.4: Cellular MMP/TIMP expression is affected by peptide crosslinker identity and varies with time.

Cell-driven differences in MMP and TIMP expression affect vascular tissue development. Specifically, MMP-1, also known as interstitial collagenase, is essential for angiogenesis, tube formation, and endothelial cell migration [75, 76]. MMP-2 (type IV collagenase or gelatinase A) is essential for vasculogenesis of endothelial progenitor cells and vessel stability, and it promotes angiogenic tube formation [73, 77–79]. MMP-3 and MMP-10 are both pro-angiogenic members of the stromelysin family [80–83]. TIMP-2 and TIMP-4 are similar anti-angiogenic proteins that regulate MMP-2 activity [73, 84–88]. To investigate how cellular MMP and TIMP expression change over time in response to eECM MMP-degradability, media samples from vasculogenic hydrogel cultures were collected and analyzed for MMP and TIMP production (Fig. 5A).

Overall, MMP expression was highest in non-degradable hydrogels and lowest in IPES-crosslinked hydrogels, with sequence-specific variability observed between VPLS- and GPQG-crosslinked hydrogels. On day 1, MMP-1 expression by cells entrapped within VPLS-crosslinked hydrogels was 6-fold higher compared to non-degradable hydrogels and over 1500-fold higher than IPES- and GPQG-crosslinked hydrogels before decreasing by day 14 (VPLS: 168,000 ± 21,000 pg ml⁻¹, ND: 31,000 ± 16,000 pg ml⁻¹, IPES: 22 ± 4 pg ml⁻¹, GPQG: 110 ± 20 pg ml⁻¹.) (Fig. 5B). MMP-2 production was highest by cells in GPQG-crosslinked and non-degradable hydrogels, about 2-fold higher compared to IPES- and VPLS-crosslinked hydrogels on day 1 and 3 to 4-fold higher by day 14 (GPQG: 37,000 ± 3,500 pg ml⁻¹, ND: 38,000 ± 8,800 pg ml⁻¹, IPES: 18,700 ± 1,600 pg ml⁻¹, VPLS: 15,000 ± 2,800 pg ml⁻¹). (Fig. 5B). On day 1, MMP-3 expression was highest in non-degradable hydrogels, about 2-fold higher compared to GPQG-crosslinked hydrogels, 13-fold higher compared to VPLS-crosslinked hydrogels, and nearly undetectable in IPES-crosslinked hydrogels (GPQG: 218 ± 11 pg ml⁻¹, VPLS: 39 ± 6 pg ml⁻¹, ND: 540 ± 60 pg ml⁻¹) (Fig. 5B). Similarly, MMP-10 expression by cells in non-degradable and VPLS-crosslinked hydrogels was elevated on day 1, but minimal in IPES- and GPQG-crosslinked hydrogels (ND: 890 ± 320 pg ml⁻¹, VPLS: 480 ± 81 pg ml⁻¹). By day 14, MMP-10 cellular expression was similar in all groups (Fig. 5B). MMP-7, MMP-8, MMP-9, MMP-12, and MMP-13 were also measured, but levels were below the detection limit for the Luminex assay.

Generally, TIMP expression was higher in non-degradable and GPQG-crosslinked hydrogels compared to IPES- or VPLS-crosslinked hydrogels. On day 1, TIMP-1 expression from cells in non-degradable and GPQG-crosslinked hydrogels was about 2 to 3-fold higher compared to VPLS- and IPES-crosslinked hydrogels (GPQG: 109,000 ± 6,800 pg ml⁻¹, ND: 113,000 ± 6,100 pg ml⁻¹, IPES: 40,300 ± 3,200 pg ml⁻¹, VPLS: 50,200 ± 7,600 pg ml⁻¹). Cellular TIMP-1 expression converged to similar levels across all groups by day 14 (Fig. 5C). TIMP-2 expression was similar in all conditions on day 1. By day 14, cellular TIMP-2 expression in GPQG-crosslinked hydrogels was 30% higher compared to non-degradable hydrogels and 3-fold higher compared to IPES- and VPLS-crosslinked hydrogels (GPQG: 44,000 ± 8,600 pg ml⁻¹, ND: 33,000 ± 10,000 pg ml⁻¹, IPES: 15,200 ± 1,100 pg ml⁻¹, VPLS: 14,900 ± 2,700 pg ml⁻¹) (Fig. 5C). On day 1, TIMP-4 expression was 2 to 3-fold higher for cells in GPQG-crosslinked and non-degradable hydrogels than in IPES- and VPLS-crosslinked hydrogels (GPQG: 101 ± 14 pg ml⁻¹, ND: 97 ± 9 pg ml⁻¹, IPES: 24 ± 9 pg ml⁻¹, VPLS: 31 ± 8 pg ml⁻¹). TIMP-4 expression increased over time in all groups; by day 14, cellular TIMP-4 expression in GPQG-crosslinked hydrogels was 3 to 4-fold higher compared to IPES- and VPLS-crosslinked hydrogels and about 40% higher than non-degradable hydrogels (GPQG: 255 ± 18 pg ml⁻¹, IPES: 72 ± 11 pg ml⁻¹, VPLS: 71 ± 19 pg ml⁻¹, ND: 180 ± 50 pg ml⁻¹).

Hydrogel MMP degradability had little effect on MMP and TIMP expression in HUVEC/hMSC spheroids in the angiogenesis-mimicking assay, consistent with the lack of observed differences between degradable peptide groups in angiogenic structural development. MMP-1, MMP-2, and MMP-13 expression in GPQG-crosslinked hydrogels was slightly higher compared to IPES- and VPLS-crosslinked hydrogels at certain time points; for example, on days 1 and 3, MMP-1 expression by cells in GPQG-crosslinked hydrogels was over 50% higher than in IPES- or VPLS- crosslinked hydrogels (GPQG: 280 ± 40 pg ml⁻¹, IPES: 190 ± 50 pg ml⁻¹, VPLS: 160 ± 40 pg ml⁻¹).

To compare MMP and TIMP expression in vasculogenesis-mimicking and angiogenesis-mimicking conditions, the measured concentration of each MMP or TIMP was normalized by the total number of cells in each respective culture condition. When normalized for total cell number, MMP-1 and MMP-2 expression in vasculogenesis-emulating and angiogenesis-emulating conditions were similar and normalized cellular expression of TIMP-1 and TIMP-2 was higher in angiogenesis-emulating conditions (Fig. S14). For example, in GPQG-crosslinked hydrogels, TIMP-1 expression was about 3-fold higher in angiogenesis-mimicking hydrogel cultures: (GPQG_Angiogenesis: 1.2 ± 0.6 pg ml⁻¹ cell⁻¹, GPQG_{Vasculogenesis}:0.40 ± 0.03 ml⁻¹ cell⁻¹). Similarly, TIMP-2 expression in angiogenesis-mimicking spheroid cultures was about 50% higher than in vasculogenesis-mimicking hydrogel cultures (GPQG_Angiogenesis:0.14 ± 0.05 pg ml⁻¹ cell⁻¹, GPQG_{Vasculogenesis}: 0.09 ± 0.05 ml⁻¹ cell⁻¹).

4. Discussion

This work explored how the susceptibility of engineered extracellular matrices (eECM) to cell-dictated degradation can be designed to control vascular morphogenesis. Leveraging PEG-based hydrogels crosslinked with MMP-degradable peptides in combination with an in vitro assessment of peptide degradation kinetics, vasculogenesis and angiogenesis were shown to be differentially regulated by the degradability of the crosslinkers used. Specifically, in vasculogenesis-mimicking conditions, HUVECs entrapped in IPES- and VPLS-crosslinked hydrogels formed vessel networks with over two-fold greater diameter, density, and total length compared to GPQG-crosslinked hydrogels, correlating with higher calculated vasculogenic kinetic parameters (k_vasc) for βA-VPLS and βA-IPES (βA-VPLS: 1.4 × 10⁻⁶ hr⁻¹ cell⁻¹, βA-IPES: 8.8 × 10⁻⁷ hr⁻¹ cell⁻¹, βA-GPQG: 2.9 × 10⁻⁷ hr⁻¹ cell⁻¹). Furthermore, the degradation of βA-VPLS and βA-IPES was primarily driven by soluble MMPs rather than membrane-bound MMPs (βA-IPES: k_{Δ, Cells} = −0.04 ± 0.01 hr⁻¹, βA-VPLS: k_{Δ, Cells} = −0.01 ± 0.03 hr⁻¹), indicating that bulk ECM remodeling is more impactful in vasculogenesis. In contrast, while relative crosslinker degradability influenced early sprouting in angiogenic conditions, it did not affect overall network development. Cells dynamically adapt their proteolytic activity to the physicochemical cues from the ECM; for example, MMP and TIMP expression were highest in non-degradable and GPQG-crosslinked hydrogels and lowest in IPES-crosslinked hydrogels, while MMP-1 expression was 1500-fold higher in VPLS-crosslinked hydrogels (VPLS: 168,000 ± 21,000 pg ml⁻¹). Altogether, these findings establish a paradigm for designing hydrogels that promote vasculogenic or angiogenic activity.

In the vasculogenesis model, IPES- and VPLS-crosslinked hydrogels supported vessel networks with higher average diameter, density, numbers of branch points, and network length, though the average vessel length was variable for IPES-crosslinked hydrogels. HUVECs encapsulated in IPES- and VPLS-crosslinked hydrogels formed comparable networks with similar structural properties, indicating that the hydrogels supported vasculogenesis. Cells in GPQG-crosslinked hydrogels exhibited a bimodal response: some formed sparse, disconnected endothelial structures or individual cells, whereas others generated networks similar to those in IPES- and VPLS-crosslinked hydrogels. This suggests that a threshold of eECM degradability must be overcome for successful network development. LIVE/DEAD staining showed substantial cell death in GPQG-crosslinked hydrogels, indicating failure to remodel the eECM (Fig. S15), similar to the non-degradable hydrogels. Collectively, these data indicate that vasculogenesis can be modulated by using different MMP-degradable peptides to alter the degradability of the eECM.

Similarly, the angiogenesis-emulating conditions showed that a threshold of eECM MMP degradability must be reached for vessel sprouting. MMP-degradability may have had a minor effect on early sprouting dynamics, but overall network length, sprout number, or branching under angiogenic conditions were not affected. Previous studies have similarly shown that angiogenesis and vasculogenesis are initiated only when certain mechanical, chemokine, and matrix degradability thresholds are met [18, 26, 89–91]. Interestingly, in IPES-crosslinked hydrogels, lower sprout densities were observed compared to GPQG- and VPLS-crosslinked hydrogels on day 1. In situ peptide degradation showed that membrane-bound MMPs had little impact on the cleavage of βA-IPES. It is possible that the early pro-angiogenic and pro-MMP/TIMP cell responses observed in GPQG-crosslinked, VPLS-crosslinked, and non-degradable hydrogels are not induced by IPES, thereby reducing early angiogenic vascular density. The reduced MMP and TIMP expression correlates with decreased early sprouting observed for IPES-crosslinked hydrogels. Hence, eECM degradability, mediated by feedback between cells and eECM, is critical for controlling the rapid development of vessels, even within angiogenic hydrogels.

To compare in situ peptide degradation, comparative kinetic parameters were calculated. Compared to βA-IPES and βA-VPLS, k_obs for βA-GPQG was lower in both cell culture and cell-conditioned media, reflecting the lower overall rate of degradation and correlating to the observed effects on angioarchitectural development in the vasculogenesis-emulating conditions. Similarly, k_vasc was 3- to 5-fold higher for βA-VPLS and βA-IPES compared to βA-GPQG. Based on differences between cell culture and conditioned media, βA-IPES appeared to be more susceptible to degradation by soluble MMPs from HUVECs, while βA-VPLS was degraded more rapidly in hMSC cultures or in hMSC conditioned media. Calculation of k_Δ, revealed that βA-IPES and βA-VPLS were more rapidly degraded in conditioned media than in cell cultures, suggesting that cells may have been upregulating TIMP expression in response to peptide recognition. These results suggest that differences in the effective proteolysis of peptides may be dependent on responsive changes in MMP and TIMP regulation.

Cells dynamically shift their proteolytic activity in response to crosslinker peptides and eECM degradability. In the vasculogenesis-mimicking conditions, MMP and TIMP expression were higher in non-degradable hydrogels and GPQG-crosslinked hydrogels compared to IPES- and VPLS-crosslinked hydrogels. It is possible that the slowly degradable GPQG-crosslinked and non-degradable hydrogels trigger compensatory pro-MMP and pro-TIMP responses by cells. Corroborating these data, in the vasculogenesis-emulating conditions, cells in non-degradable or GPQG-crosslinked hydrogels were unable to overcome a threshold of matrix degradability, resulting in rounded morphologies. HUVECs have been shown to upregulate MMP-2 and MMP-9 in hyaluronic acid hydrogels with minimal degradability [92]. MMP-1 and MMP-10 expression was higher for VPLS-crosslinked hydrogels from days 1-7. MMP-1 has relatively low specificity for VPLS compared to other MMPs [36]. Kinetic parameters for VPLS degradation by MMP-10 have not been reported, but it is possible that the early upregulation of MMP-1 and MMP-10 by cells in VPLS-crosslinked hydrogels represents a similar compensatory response due to low MMP-specificity for VPLS. These observations suggest that highly degradable eECMs require less MMP expression to support vascular remodeling and cellular migration, whereas cells entrapped in eECM crosslinked by peptides with lower enzymatic susceptibility upregulate MMP and TIMP expression in response. Cellular stress and mechanical cues from the ECM have been shown to induce upregulation of MMP and TIMP expression; accordingly, the elevated MMP expression in non-degradable hydrogels could be emblematic of an early stress response [84, 85, 93–97]. Further studies should assess cell viability, proliferative capacity, and cellular stress biomarkers (i.e., inflammatory cytokines, reactive oxygen species, mitochondrial stress markers) in response to different MMP-degradable and non-degradable crosslinkers. Together, these results indicate that eECM biomaterials with insufficient degradability may induce compensatory protease overexpression and/or cell stress. Hence, crosslinker choice is critical for preventing maladaptive cellular responses.

Shifts in MMP expression were nuanced and highly sensitive to peptide sequence. For example, MMP-3 expression was much higher in cells in GPQG-crosslinked hydrogels than in VPLS-crosslinked hydrogels, which instead exhibited higher MMP-10 expression. This was surprising, as MMP-3 and MMP-10 are members of the stromelysin family (stromelysin-1 and stromelysin-2, respectively) and share ~80% amino acid homology and degrade similar components of the ECM [81, 98–101]. MMP-3 stabilizes vessels by recruiting mural cells to the vasculature [80, 102]. In contrast, MMP-10 is a vessel-destabilizing factor, causing vessel regression and loss of endothelial cell-to-cell adhesion [83, 103–105]. These findings suggest that peptide identity can be altered to bias the expression of specific MMP subtypes, potentially enabling control over downstream processes such as vessel stability and regression.

Regulation of matrix degradation via cell-secreted TIMPs is also responsive to peptide crosslinkers, further demonstrating that physicochemical matrix cues can be engineered to influence proteolysis. In-situ peptide degradation showed that βA-IPES and βA-VPLS were degraded more rapidly in media-only conditions, which lack TIMP-mediated regulation of enzymatic activity. Indeed, the gradual decrease of cellular MMP-1, MMP-3, and MMP-10 expression in VPLS-crosslinked, GPQG-crosslinked, and non-degradable hydrogels correlated with increased TIMP-1, TIMP-2, and TIMP-4 expression [106–108]. Similarly, peptide crosslinks in regions farther from encapsulated cells show greater degradation than pericellular crosslinks [109]. Notably, cellular TIMP expression was lower in IPES- and VPLS-crosslinked hydrogels, which correlates with the higher observed k_vasc and degree of vasculogenesis compared to GPQG-crosslinked and non-degradable hydrogels. Hence, the interplay of peptide identity and TIMP-mediated regulation of proteolytic activity is another key design consideration for engineering eECM biomaterials for vascular morphogenesis.

In the angiogenesis-mimicking spheroid experiments, MMP and TIMP expression was similar in all MMP-degradable hydrogels, corroborating the lack of differences observed for the structural vessel parameters. When normalized for cell number, MMP expression was similar in both angiogenesis- and vasculogenesis-mimicking conditions. Normalized TIMP-1 and TIMP-2 expression were higher in angiogenesis-emulating conditions compared to vasculogenesis-emulating conditions, possibly suggesting that total encapsulation of the cells within the hydrogel eECM causes cellular upregulation of certain TIMPs.

Physical properties of the eECM, including modulus and mesh size, can be tuned to control vascular organogenesis towards pro-vasculogenic or pro-angiogenic developmental pathways. Vascular development comprises two distinct processes: vasculogenesis and angiogenesis. During embryonic growth, when the ECM has lower modulus and less interstitial crosslinking, vasculogenesis is the primary mechanism of vessel development [71]. As tissues mature and robust vasculature develops, this pro-vasculogenic environment shifts to a stiffer, collagen-rich ECM with higher modulus over time, signaling a shift from vasculogenic development to angiogenesis [56, 57, 110–114]. This developmental paradigm provides a framework for eECM design; to emulate these differences in vitro, hydrogel modulus was lower to promote vasculogenesis (6 ± 1 kPa) compared with angiogenesis (11 ± 1 kPa). While decreasing the crosslinking does increase mesh size (vasculogenic: ξ = 26 ± 3 nm, angiogenic: ξ = 18 ± 3 nm), these differences are unlikely to significantly affect the diffusivity of soluble MMPs through the eECM network due to their small hydrodynamic radii.

Some limitations of this work should be considered for further investigation. HUVECs were selected to model vasculogenesis and angiogenesis. However, HUVECs are prone to donor variability and may not fully recapitulate features necessary for specialized endothelial tissues [37, 115]. Future studies should confirm the trends observed herein using stem-cell-derived endothelial cells, which can adopt more robust endothelial phenotypes. To capture the differences between angiogenesis and vasculogenesis, cells were seeded in distinct arrangements, limiting direct structural comparisons. After normalization for cell number, MMP expression was comparable between both experimental modalities. Co-encapsulation of monodisperse HUVECs and hMSCs in PEG hydrogels for the vasculogenesis-mimicking conditions at a 3:1 ratio was also explored, but HUVECs were unable to consistently form microvascular network under these conditions (Fig. S10). While these findings demonstrate that crosslinking with different MMP-degradable peptides can affect both MMP/TIMP expression profiles and angioarchitectural development in vasculogenesis-mimicking conditions, a direct causal link between cellular proteolytic activity and vascular morphogenesis was not fully established. Further characterization of protein expression by cells through functional perturbation studies (i.e., genetic knockout) is warranted in future studies to determine if peptide-induced changes in MMP and TIMP expression directly lead to observed differences in structural development. While soluble and the aggregate of soluble and cell-membrane-bound MMPs were studied during in situ degradation experiments, the expression of membrane-bound MMPs was not measured directly. Additionally, steric hindrance in the crosslinked hydrogel environment could lead to differences in ECM degradation activity compared with soluble peptide degradation. The mesh sizes of both hydrogel formulations (vasculogenic: ξ = 26 ± 3 nm, angiogenic: ξ = 18 ± 3 nm) are an order of magnitude higher than the hydrodynamic radii of soluble MMPs; hence, the impact of diffusive constraints on peptide catalytic activity is expected to be minimal. In situ zymography or Förster resonance energy transfer (FRET) assays could also be used to specifically identify and assess membrane-bound MMP activity in future studies [116, 117].

5. Conclusions

Altogether, this study demonstrates that MMP-degradability of engineered ECMs can modulate angioarchitectural development in vasculogenesis. By tuning eECM degradability, vasculogenic development can be precisely controlled to mimic tissue-specific microvasculature. While MMP-degradable crosslinkers were essential for angiogenesis, only early sprouting activity was affected by eECM degradability. Furthermore, these data indicate that the proteolytic susceptibility of peptides varies with cell type and between soluble and membrane-bound MMPs. These results outline the dynamic interplay between ECM degradability and cellular MMP/TIMP expression, indicating that cellular proteolytic activity adapts to biochemical cues from the ECM. Collectively, this work establishes a framework for the rational design of eECM biomaterials, in which degradability, mechanics, and crosslinker peptide chemistry are tuned in tandem to engineer microvascular networks that recapitulate the structural and functional characteristics of specific tissue niches.

Supplementary Material

Supplementary material associated with this article can be found online.

Statement of Significance.

The microvasculature supplies nutrients and oxygen to nearly all tissues, yet current tissue engineering approaches struggle to recapitulate the diversity of tissue-specific vasculature. Poly(ethylene glycol) (PEG) hydrogels crosslinked with enzymatically degradable peptides were used to investigate how extracellular matrix degradability directs vessel development under vasculogenic and angiogenic conditions. Peptide crosslinkers with higher degradability promoted the formation of vessel networks with increased diameter, density, and length in vasculogenic conditions. Under angiogenic conditions, degradability regulated early sprouting without altering network properties. Additionally, the expression of matrix metalloproteinases and their inhibitors was elevated in slowly- or non-degradable hydrogels, suggesting a cellular attempt to overcome matrix constraints. These findings establish a biomaterial design strategy to engineer tissue-specific vasculature for disease modeling and regenerative medicine.