Mesenchymal cells stimulate capillary morphogenesis via distinct proteolytic mechanisms

Abstract

During angiogenesis, endothelial cells (ECs) degrade their surrounding extracellular matrix (ECM) to facilitate invasion. How interactions between ECs and other cells within their microenvironment facilitate this process is only partially understood. We have utilized a tractable 3D co-culture model to investigate the proteolytic mechanisms by which pre-committed or more highly committed mesenchymal cells stimulate capillary formation. On their own, ECs invade their surrounding matrix, but do not form capillaries. However, in the presence of either mesenchymal stem cells (MSCs) or fibroblasts, ECs form polarized, tubular structures that are intimately associated with mesenchymal cells. Further, ECs upregulate gene expression of several extracellular proteases upon co-culture with either mesenchymal cell type. The administration of both broad spectrum and specific protease inhibitors demonstrated that MSC-stimulated capillary formation relied solely on membrane-type matrix metalloproteinases (MT-MMPs) while fibroblast-mediated sprouting proceeded independent of MMP inhibition unless the plasminogen activator/plasmin axis was inhibited in concert. While other studies have established a role for the ECM itself in dictating proteolysis and matrix degradation during capillary morphogenesis, the present study illustrates that heterotypic cellular interactions within the microenvironment can direct the proteolytic mechanisms required for capillary formation.

Introduction

Angiogenesis, the sprouting of new vasculature from a pre-existing vascular network, is a process crucial to the growth and maintenance of tissues ¹. Understanding the mechanisms that govern this complex process may facilitate approaches to engineer tissues for regenerative medicine purposes ² and treat pathologies (e.g., cancer and diabetic retinopathy) characterized by misregulated vessel formation and maintenance ³.

A cell’s microenvironment, composed of its surrounding extracellular matrix (ECM), other cells, and factors (e.g., soluble proteins) secreted as a result of interactions between these constituents⁴, profoundly influences cell behavior ^{5, 6}. In the context of capillary morphogenesis, the identity ⁷ and concentration ^8–13 of ECM molecules regulate endothelial cell (EC) invasion and tube formation. Homotypic (EC-EC) interactions are crucial to guiding and sustaining this morphogenic process ^{14, 15}, while heterotypic interactions with other cell types (e.g., tumor cells or interstitial cells) also promote capillary morphogenesis through the production of numerous angiogenic factors ¹⁶. Extracellular proteases, especially those comprising the matrix metalloproteinase (MMP) and serine protease families ^{17, 18}, are amongst these factors, as they facilitate ECM degradation (in addition to several other functions, reviewed in ¹⁹), which is necessary for EC invasion to occur.

Macromolecules within the ECM dictate the proteases utilized by ECs for the purposes of invasion. For instance, capillary invasion in type I collagen matrices depends critically on membrane-bound MMP14 (also known as MT1-MMP) ²⁰, while invasion of the provisional clot (composed primarily of the plasma protein fibrinogen in its cleaved form, fibrin)present in wounds, sites of inflammation, and tumors may depend less on MMPs and more on the serine proteases that comprise the plasminogen activator (PA)/plasmin axis ^{21, 22}. However, in more complex tissue explant cultures, capillary invasion in fibrin matrices proceeds independent of the PA/plasmin axis and relies instead on MT-MMP activity ²³. These results imply that in addition to the ECM, other cell types have the potential to bias the proteases utilized by ECs to undergo capillary morphogenesis.

The present study explores whether different stromal cell types modulate EC proteolysis via distinct mechanisms during capillary morphogenesis in a three-dimensional (3D) ECM. Using an established co-culture model ^{9, 10, 24} that generates true capillaries that are functional in vivo ²⁵, we show that two mesenchymal cell types (mesenchymal stem cells (MSCs) and fibroblasts) guide ECs to undergo a virtually complete angiogenic program via distinct proteolytic mechanisms. Specifically, ECs rely solely on MT-MMPs to undergo capillary morphogenesis in EC-MSC co-cultures, but utilize MMPs and the PA/plasmin axis to form capillary networks when co-cultured with fibroblasts. These data demonstrate that in addition to the ECM, interactions with other cell types determine the proteolytic repertoire utilized by ECs to undergo capillary morphogenesis, and point further to the importance of the integrated microenvironment in regulating cell behavior.

Materials and Methods

Construction of a 3D co-culture model to study capillary morphogenesis and quantify vessel networks

Construction of the 3D co-culture model was performed as described previously ¹⁰. In brief, mCherry-transduced human umbilical vein ECs (p3-p4) were cultured on Cytodex™ microcarrier beads (Sigma-Aldrich Co., St. Louis, MO) and mixed with either normal human lung fibroblasts (Lonza Biologics Inc., Portsmouth, NH; used prior to p10) or MSCs (Lonza, used at p4 or p5) within a 2.5 mg/mL fibrin matrix (the approximate physiologic clotting concentration of fibrin ²⁶). Medium (EGM-2, Lonza) was added after clotting and changed at days 1, 3, and 5. For studies involving inhibitors, the appropriate vehicle or inhibitor(s) was mixed withmedium prior to addition to the culture. For non-small molecule inhibitors (e.g., antibodies), the inhibitor was added at the indicated concentration in the fibrinogen solution prior to clotting and to the culture medium as well. Inhibitors were added at the following concentrations: GM6001, 10–40 µM (Calbiochem, San Diego, CA); BB2516, 3.3–10 µM (Tocris Bioscience, Ellisville, MO); aprotinin, 2.2 µM (Sigma); ε-aminocaproic acid, 50 mM (Sigma); human tissue inhibitor of metalloproteinase (TIMP)1, 5 µg/mL (R&D Systems, Minneapolis, MN); human TIMP2, 5 µg/mL (R&D Systems); a function-blocking antibody directed towards the urokinase plasminogen activator receptor (uPAR), 25 µg/mL (American Diagnostica, Inc., Stamford, CT); a function-blocking antibody targeting HGF, 1 µg/mL and 10 µg/mL (R&D Systems) ²⁷; and a neutralizing antibody directed at TNFα, 1 µg/mL and 10 µg/mL (Abcam, Cambridge, MA).

Low magnification (4x, to image as large a field as possible) fluorescent images of isolated beads were randomly captured at day 7. To maximize objectivity, an automated approach was adapted to quantify the total length of the vascular networks. Images were processed in a batch mode, using a script based on the Angiogenesis Tube Formation Application Module in Metamorph imaging software (Molecular Devices, Sunnyvale, CA). Briefly, each image was segmented based on a constant threshold of a background corrected image. The Adaptive Background Correction™ function of Metamorph was used to crrect uneven image backgrounds throughout the image by adapting to local content. This allowed for more robust segmentation and analysis Any tubule-like pattern of the image less than a specified minimum width was considered to be noise and was excluded from the segmentation. A maximum width was used to differentiate tubules from beads. A sample image of network tracings, visualized and reconstructed in Imaris (Bitplane AG, Zurich, Switzerland), is presented in Figure 1.

Figure 1. Depiction of network visualization and tracing.

A) Low magnification fluorescent images of day 7 capillary networks were segmented based on a constant threshold of a background corrected image. Minimum and maximum thresholds were set to delineate capillaries from noise or beads/nodes, respectively, before applying the Angiogenesis Tube Formation Application Module in Metamorph imaging software. B) A sample tracing generated by automated processing. The filament tracing function was used to visualize networks (blue) using Imaris.

For association studies, fibroblasts or MSCs were first retrovirally transduced with a green fluorescent protein (GFP) retrovirus as described previously ¹⁰. Tissue construction and maintenance then proceeded exactly as described. Imaging was conducted on day 7.

Immunofluorescent Staining

Cultures were constructed within 8 well LabTek™ chamberslides with a No.1 thickness borosilicate bottom (ThermoFisher Scientific; Rochester, NY) and maintained until the appropriate timepoint. Samples were fixed in formalin for 10 minutes, permeabilized for 10 min, and blocked in a 2% solution of bovine serum albumin in TBS-T for 1 hour. Primary antibody/staining agent was diluted in blocking solution at the appropriate concentration (collagen IV, 1:200 (Sigma); phalloidin, 1:50 (Molecular Probes, Eugene, OR); alpha-smooth muscle actin (αSMA), 1:200 (Sigma); NG2, 1:200 (Chemicon International, Inc., Temecula, CA)) and incubated with the gel at 4°C for 2.5 hours. Samples were washed in TBS-T overnight and incubated with the appropriate secondary antibody (1:200) the next day (2 hours at 4°C). After another overnight wash, samples were incubated with a 250 µM solution of DAPI (in PBS) for 10 minutes in order to visualize nuclei, washed, and finally imaged on a Zeiss LSM 510 Meta multiphoton confocal microscope (Carl Zeiss, Inc.; Jena, Germany). Images obtained were 12 bit images containing 1024× 1024 pixels. Four scans were averaged per pixel.

DQ Collagen Cultures

To visualize the localization of proteolytic activity in EC/mesenchymal cell co-cultures, 1% (v/v) of a 2.5 mg/ml solution of DQ collagen, type I (Molecular Probes) was added to the 2.5 mg/ml fibrinogen solution prior to polymerization. Gels were constructed as described above within 8 well No. 1 thickness Labtek chamberslides and imaged at day 7 using the Zeiss LSM 510 Meta. Image slices were acquired every 5 microns and reconstructed using Zeiss LSM Image Browser.

RNA Isolation

For each condition and timepoint, 24 tissues containing >100 HUVEC-coated beads +/− 25K overlying fibroblasts or MSCs each was constructed. To isolate RNA from fibroblast or MSC monolayers, gels were twice washed with PBS, then 100 µl of a concentrated trypsin solution (2.5%; Sigma) was added until the monolayer was completely removed (30–60 seconds). Trypsin was immediately neutralized with 200 µl of Media 199 (Invitrogen; Carlsbad, CA) containing 20% fetal bovine serum (FBS). Neutralized solutions were collected, spun down for 5 minutes at 200g, lysed using RNA lysis buffer (Qiagen; Valencia, CA), and stored at −80 °C.

To isolate EC, gels stripped of mesenchymal cell monolayers or gels containing only EC-coated microcarriers were washed with PBS before dislodging gels from the well siding using a small spatula. Gels were subsequently dissolved by adding 500 µl of trypsin (2.5%) and incubating at 37°C for ~20 minutes with periodic agitation. Upon dissolution, 1 ml of M199 containing 20% FBS and 1% P/S was added to quench trypsin activity. To ensure complete removal of EC from microcarriers, the contents of each well were triturated via repeated pipetting. The contents of each well were collected, microcarrier beads were removed by gravity, and the remaining solution was centrifuged to collect ECs. ECs were suspended in RNA lysis buffer before storage at −80°C.

Determination of mRNA transcript levels via RT-PCR

RT-PCR was conducted to analyze the expression of several genes in fibroblasts, MSCs, or ECs. Table I lists the genes analyzed.

Table I.

Description of primers used for RT-PCR

Gene	Primers (F = forward; R = reverse)	Anneal Temp. (°C)	Amplicon size (bp)
MT1- MMP	F: 5’-GCTTGCAAGTAACAGGCAAA-3’ R: 5’-AAATTCTCCGTGTCCATCCA-3’	57	589
MT2- MMP	F: 5’-TCGACGAAGAGACCAAGGAGT-3’ R: 5’-CTTGAAGTTGTCAACGTCCT-3’	55	578
MT3- MMP	F: 5’-ATGTGCTACAGTCTGCGGAAC-3’ R: 5’-TATCCACATCACGTTTGCCA-3’	59	461
MMP2	F: 5’-CAAGTGGTCCGTGTGAAGTA-3’ R: 5’-CTTGCGGTCATCATCGTAGT-3’	56	504
MMP9	F: 5’-GACCTCAAGTGGCACCACCA-3’ R: 5’-GTGGTACTGCACCAGGGCAA-3’	62	483
uPAR	F: 5’-ACAGGAGCTGCCCTCGCGAC-3’ R: 5’-GAGGGGGATTTCAGGTTTAGG -3’	58	414
uPA	F: 5’-GGCAGCAATGAACTTCATCAAGTTCC-3’ R: 5’-TATTTCACAGTGCTGCCCTCCG-3’	56	135
tPA	F: 5’-CCAGCAACATCAGTCATGGC-3’ R: 5’-GCACTTCCCAGCAAATCCTTC-3’	60	205
GAPDH	F: 5’-CCCTGTTGCTGTAGCCGTA-3’ R: 5’-CCGGTGCTGAGTATGTCG-3’	58	442

Total RNA was isolated from samples using the RNeasy kit (Qiagen) per manufacturer’s protocol. The concentration of purified RNA was quantified using the Quant-iT™ RiboGreen RNA assay kit (Molecular Probes) according to the manufacturer’s instructions. Equal amounts of total RNA from each sample were used to synthesize cDNA templates using the AMV reverse transcriptase kit (Promega; Madison, WI), also according to manufacturer’s protocol. The cDNA for each sample (2 µl) was amplified via PCR for 35 cycles in a total reaction volume of 50 µl. The reaction mixture was comprised of 200 nM of the appropriate forward and reverse primers, 0.2 mM deoxynucleoside triphosphate, 1.5 mM MgCl₂, and 1U of Taq polymerase (Promega). For each gene analyzed, forward and reverse primer sequences, annealing temperature, and amplicon size are listed in Table I. Amplified products were resolved via electrophoresis through 2% agarose gels containing 0.05% ethidium bromide, visualized using a BioRad ChemiDoc XRS imaging system (BioRad Laboratories; Hercules, CA), and verified via comparison of the known amplicon size with a DNA ladder. Images were inverted to facilitate visualization of bands.

Statistical Analyses

Data are presented as mean ± S.D. Statistical analyses were conducted using SigmaStat 3.5 (Systat Software, Inc., San Jose, CA), with p values < 0.05 deemed significant. The specific method of statistical analysis used for a given experiment is described in the appropriate figure legend.

Results

MSCs and fibroblasts stimulate ECs to form mature capillary networks

Capillary sprouting can be divided into four distinct steps, including: EC invasion and proliferation, primitive tube formation, stabilization of nascent tubules (which involves basement membrane deposition and pericytic association) and pruning of the vascular network ¹. EC sprouting in MSC- or fibroblast-laden ECMs was monitored over a 7 day period to determine whether both mesenchymal cell types stimulate ECs to undergo these steps. ECs were transduced to express the red fluorescent protein mCherry in order to differentiate capillary networks form surrounding interstitial cells.

ECs invaded surrounding ECM as early as day 1, formed primitive tubular structures by day 3, and elaborated into extensive vascular networks by day 7 (Fig. 2A and A’). In both culture conditions (EC-MSC and EC-fibroblast), vessels were surrounded by basement membrane sleeves, as gauged by IF staining cultures for collagen IV, which is a prominent component of the vascular basement membrane ⁷ (Fig. 2B and B’, see white arrows). Vascular mural cells, or pericytes stabilize nascent endothelium and are characterized by physical association with ECs as well as expression of molecular markers such as αSMA and NG2 ²⁸. MSCs and fibroblasts both exhibited pericytic behavior: these mesenchymal cell types intimately associated with capillaries, demonstrated by co-culture of GFP-MSCs or GFP-fibroblasts with mCherry-ECs (Fig. 2C and C’). IF staining revealed that MSCs ubiquitously expressed αSMA in addition to occasional NG2 expression (Fig. 2D), while EC-associated fibroblasts only stained positively for αSMA (Fig. 2D’). Collectively, these results demonstrate that both uncommitted MSCs and their differentiated progeny guide ECs to form mature vascular networks in 3D co-cultures.

Figure 2. MSCs and fibroblasts stimulate ECs to form mature capillary networks.

MSCs (A–D) or fibroblasts (A’–D’) were interspersed throughout fibrin ECMs in the presence of microcarrier beads coated with mCherry-transduced ECs. A, A’) Beads were monitored over a 7 day period and imaged at days 1, 3, and 7. Red channel fluorescence (i.e., ECs) is shown. Scale = 200 µm. B, B’) Cultures were fixed and IF stained at day 7 for F-actin (green) and collagen IV (red). Basement membrane deposition (white arrows) was observed in both conditions. Scale = 50 µm. C, C’) Mesenchymal cells expressing GFP were interspersed with mCherry-transduced EC. Physical association of both cell types was observed in both conditions (white arrows). Scale = 25 µm. D, D’) Cultures containing mCherry-transduced ECs and either MSCs (D) or fibroblasts (D’) were fixed and IF stained for pericyte markers αSMA (aqua) and NG2 (white). DAPI-stained nuclei are visible in the blue channel. Scale = 50µm.

Mesenchymal cell-mediated capillary morphogenesis is accompanied by enhanced periendothelial proteolysis

Localized proteolytic activity is necessary to facilitate key steps of the angiogenic cascade ^{23, 29–31}. We have previously observed that the addition of mesenchymal cells contributes to bulk increases of MMP levels in 3D cultures ⁹. However, whether the mesenchymal cells were directly responsible for the enhanced proteolytic activity observed or whether these cells were inducing increased proteolytic activity by ECs was unclear. To visualize localized proteolytic activity, 3D ECMs were doped with a small amount (1%) of a fluorescein-laden substratum (DQ Collagen, type I), which emits a green fluorescent signal upon cleavage by collagenases.

As shown previously ²⁴, ECs cultured in the absence of interstitial cells (or even in the presence of medium conditioned by interstitial cells ^{24, 32}) do not form capillary-like structures under these conditions, but do invade the surrounding fibrin ECM. ECs cultured alone elicited a focal signal from the DQ substratum (Fig. 3A). Conversely, a strong, primarily periendothelial DQ signal was observed in both EC-MSC (Fig. 3B) and EC-fibroblast (Fig. 3C) co-cultures.

Figure 3. Interstitial cell-mediated capillary morphogenesis is accompanied by periendothelial proteolysis.

mCherry-transduced ECs were cultured within fibrin ECMs containing 1% DQ Collagen either A) alone, B) with MSCs, or C) with fibroblasts. Red (mCherry-ECs), green (proteolyzed ECM), and merged channels are shown. Autofluorescence from the microcarrier bead is evident in A). Scales = 50 µm.

To confirm this apparent upregulation in periendothelial proteolysis, rather than spread mesenchymal cells throughout EC-containing matrices, MSCs or fibroblasts were seeded on top of fibrin matrices. This co-culture method allowed the effective isolation of ECs from mesenchymal cells in order to determine how cross-talk between these spatially distinct cell populations affected transcript expression of each cell type ²⁴. MSCs and fibroblasts both stimulated robust capillary network formation by 7 days with this altered configuration. Again, ECs cultured within the 3D ECM on their own invaded the surrounding ECM but did not form capillary-like structures (Fig. 4A). We examined mRNA expression levels of both mesenchymal and endothelial cell populations at day 1, day 3, and day 7 via RT-PCR. The mRNA expression of several fibrinolytic enzymes-- including MT1-, MT2-, and MT3-MMP, MMP2 and MMP9, and members of the PA/plasmin axis including tissue PA (tPA), urokinase-type PA (uPA), and uPAR-- was probed. The most notable differences in expression were observed at day 3 (Fig. 4B), which coincided with the time point that ECs were first observed to form lumen-containing structures in co-cultures (Fig. 4A, middle column, white arrows).

Figure 4. Onset of capillary morphogenesis coincides with upregulated expression of fibrinolytic enzyme transcripts by EC.

A) Brightfield images (10x) of EC-coated microcarrier beads at days 1, 3, and 7 cultured either alone (top row), with overlaid MSCs (middle row), or with overlaid fibroblasts (bottom row). Lumen formation is first evident in both co-cultures by day 3 (white arrows). Scale = 100 µm. B) Fibrinolytic enzyme and GAPDH (internal control) transcript expression for mesenchymal cells and ECs cultured separately (“alone”) or together (“co-cultured”) probed via RT-PCR (day 3). This experiment was repeated twice and a representative sample is shown.

Aside from MT1-MMP expression, which was increased by fibroblasts upon culture with ECs, neither mesenchymal cell type altered its expression of the fibrinolytic enzymes investigated in response to culture with ECs (Fig. 4B). However, it was apparent that fibroblasts expressed much higher levels of uPA transcript than MSCs (Fig. 4B). This result was confirmed at the protein level both by Western blot and plasminogen zymography (data not shown). On the other hand, ECs only expressed MT3-MMP, uPAR, and tPA when cultured alone. However, upon culture with MSCs, EC expression of MT1-MMP, MT3-MMP, MMP2, and MMP9 was strongly increased (Fig. 4B). In contrast, ECs exhibited a drastic increase in MT3-MMP expression along with decreased uPAR expression when cultured with fibroblasts (Fig. 4B). Taken together, the data presented in Figure 3 and Figure 4 suggest that: 1) MSCs and fibroblasts induce increased proteolytic activity in the periendothelial compartment, 2) MSCs and fibroblasts each upregulate EC expression of fibrinolytic enzymes in signature fashion, and 3) that this enhancement early on is crucial to the morphogenic process. We thus applied broad spectrum protease inhibitors to determine whether MSC- and fibroblast-mediated sprouting proceed through different proteolytic mechanisms.

Broad spectrum protease inhibition suggests that MSCs and fibroblasts stimulate capillary morphogenesis through different proteolytic mechanisms

To test the proteolytic dependence of MSC- and fibroblast-mediated capillary morphogenesis, co-cultures were treated with increasing concentrations of the broad spectrum MMP inhibitor GM6001 ³³, the serine protease inhibitor aprotinin, or a combination of the two (“dual,” Fig. 5–Fig. 6).

Figure 5. Broad spectrum inhibition reveals that MSC-mediated capillary morphogenesis is MMP dependent.

A) mCherry-transduced ECs were coated onto microcarrier beads and cultured within fibrin matrices interspersed with unlabelled MSCs. Shown are fluorescent images at day 7 of capillary network formation in vehicle (DMSO) treated cultures, those treated with 10 µM, 20 µM, or 40 µM of the broad spectrum MMP inhibitor GM6001, with 2.2 µM of the serine protease inhibitor aprotinin, or with a combination of GM6001 (10 µM) and aprotinin (2.2 µM) (“both”). Scale = 200 µm. B) Total network lengths were quantified from a minimum of 10 images over 2–3 independent experiments per condition as described in Materials and Methods. Applied concentrations of BB2516 were 3.3 and 10 µM. * P<0.05 when comparing the indicated conditions to vehicle control. One-way ANOVA followed by post-test analysis using Dunn’s method was used to compare treatment groups.

Figure 6. Broad spectrum inhibition reveals that fibroblast-mediated capillary morphogenesis is both MMP and serine protease dependent.

A) mCherry-transduced ECs were coated onto microcarrier beads and cultured within fibrin matrices interspersed with unlabelled fibroblasts. Shown are fluorescent images at day 7 of capillary network formation in vehicle (DMSO) treated cultures, those treated with 10 µM, 20 µM, or 40 µM of the broad spectrum MMP inhibitor GM6001, with 2.2 µM of the serine protease inhibitor aprotinin, or with a combination of GM6001 (10 µM) and aprotinin (2.2 µM) (“both”). Scale = 200 µm. B) Total network lengths were quantified from a minimum of 10 images over 2–3 independent experiments per condition as described in Materials and Methods. Applied concentrations of BB2516 were 3.3 and 10 µM. * P<0.05 when comparing the indicated conditions to vehicle control. One-way ANOVA followed by post-test analysis using Dunn’s method was used to compare treatment groups.

MSC-mediated sprouting was drastically inhibited by the application of broad-spectrum MMP inhibitors when compared to vehicle (DMSO)-treated conditions (Fig. 5). Doses of GM6001 (10, 20, 40 µM) nearly 1000x greater than IC₅₀ values reported for several MMPs (e.g., MMP2, MMP9, MT1-MMP ³⁴) abrogated network formation at every concentration tested (Fig. 5A–B). To lessen the probability of off-target effects, a second broad spectrum MMP inhibitor was applied to these cultures. BB2516, which has IC₅₀ values that are roughly one third of those measured for GM6001 (as reported by the manufacturer), also virtually eliminated capillary formation at both concentrations tested (3.3 and 10 µM) (Fig. 5B). On the other hand, application of an endogenous serine protease inhibitor (aprotinin, 2.2 µM ³⁵) had no effect on the formation of vascular networks, demonstrating that capillary network formation proceeds independent of plasmin-mediated proteolysis in EC-MSC cultures (Fig. 5A–B). As expected, inhibiting MMPs and serine proteases in concert eliminated network formation (Fig. 5A–B).

A strikingly different response to protease inhibition was noted in EC-fibroblast cultures (Fig. 6). GM6001 did not affect fibroblast-stimulated capillary network formation when compared to vehicle-treated conditions; only a 40 µM dose of the inhibitor significantly reduced, but did not eliminate, sprouting (Fig. 6A–B). Similar results were obtained with BB2516 (Fig. 6B). The serine protease inhibitor aprotinin also did not alter network formation. However, dual application of GM6001 (10 µM) and aprotinin (2.2 µM) eliminated network formation (Fig. 6A–B). The synergistic nature of combinatorial MMP and serine protease inhibition was confirmed using BB2516 in place of GM6001 and/or by replacing aprotinin with ε-aminocaproic acid (EACA, 50mM), which inhibits the association of plasmin and fibrin ³⁶ (data not shown).

In sum, these data demonstrate that MSC- and fibroblast-mediated capillary morphogenesis proceed via distinct proteolytic mechanisms and suggest specifically that MSC-mediated sprouting is driven solely by MMPs. More specific endogenous MMP inhibitors were applied to confirm this result.

Specific inhibition reveals that MSC-stimulated capillary morphogenesis depends on MT-MMP activity

To eliminate the potential off-target effects associated with synthetic small molecule inhibitors such as GM6001 and BB2516, endogenous MMP inhibitors were applied to EC-MSC and EC-fibroblast co-cultures. Compared to a vehicle control, TIMP1, which inhibits soluble MMP activity ³⁷, did not affect MSC- (Fig. 7A, B) or fibroblast-mediated network formation (Fig. 7A, C). Conversely, TIMP2, which inhibits the activity of membrane-bound and soluble MMPs ³⁷, severely compromised sprouting in EC-MSC co-cultures (Fig. 7A–B) but had no effect on network formation in EC-fibroblast co-cultures (Fig. 7A, C), suggesting that MT-MMPs play a critical role in MSC-mediated sprouting. A function blocking antibody targeting uPAR did not restrict network formation in either case, confirming that the activity of this key membrane-bound member of the PA/plasmin axis is not crucial for MSC- or fibroblast-mediated sprouting on its own.

Figure 7. Endogenous inhibitors confirm that mesenchymal cells stimulate capillary morphogenesis via distinct proteolytic mechanisms.

A) mCherry-transduced ECs were coated onto microcarrier beads and cultured within fibrin matrices interspersed with unlabelled MSCs (top row) or fibroblasts (bottom row). Shown are fluorescent images at day 7 of capillary network formation in vehicle (PBS) treated cultures, or those treated with 5 µg/mL TIMP1 or TIMP2, or with 25 µg/mL of a function blocking antibody directed towards uPAR (α-uPAR). Scale = 200 µm. B, C) Total network lengths were quantified from a minimum of 5 images over 2–3 independent experiments per condition as described in Materials and Methods for EC-MSC (B) and EC-fibroblast (C) cultures. * P < 0.001 when compared to vehicle control. One-way ANOVA followed by post-test analysis using the Holm-Sidak method was used to compare treatment groups.

Discussion

It is now well-established that MMPs and the PA/plasmin axis play important roles in EC invasion during capillary morphogenesis. How other cell types influence the proteolytic means by which ECs conduct this process is less clear. In the present work, we have utilized a co-culture model which recapitulates most phases of angiogenesis, including EC invasion, nascent tube formation, and stabilization of these tubes by both pericytic association and basement membrane deposition ¹, to demonstrate that both multi-potent MSCs and more committed fibroblasts stimulate capillary morphogenesis within 3D fibrin ECMs, but do so via distinct proteolytic mechanisms. Specifically, MSC-mediated capillary formation relied solely on MMPs, while fibroblast-mediated capillary formation proceeded in spite of MMP inhibition unless the PA/plasmin axis was inhibited in concert. Collectively, these data demonstrate that the cellular component of the microenvironment influences the proteolytic means by which ECs undergo capillary morphogenesis.

Mesenchymal cells stimulate an increase in the transcript expression of extracellular proteases by ECs early on, and this appears to be key to stimulating the morphogenic program. In the presence of several angiogenic factors present in the culture medium, ECs invaded the surrounding fibrin ECM but failed to form vessels when cultured on their own. Co-culture with mesenchymal cells stimulated the formation of capillary-like structures as early as day 3, and RT-PCR revealed that tube formation coincided with an induction of fibrinolytic enzymes at this time point. However, differences in the pattern of transcript upregulation induced by MSCs and fibroblasts were evident, and may underlie the functional differences observed when inhibitor studies were performed with both exogenous and endogenous protease inhibitors. For instance, co-culture with MSCs resulted in enhanced EC expression of MT1-MMP, MT3-MMP, MMP2, MMP9, and uPAR. This expression pattern suggested that MSC-mediated sprouting relies on MMP induction, and indeed sprouting in EC-MSC cultures was ablated by the broad spectrum MMP inhibitors GM6001 and BB2516. TIMP-based studies further specified that MSC-mediated sprouting depends on MT-MMPs.

In contrast, co-culture with fibroblasts only stimulated enhanced expression of MT3-MMP by ECs. But when compared to MSCs, fibroblasts expressed significantly higher levels of uPA at both the transcript and protein levels. Interestingly, functional assays revealed that sprouting persisted in EC-fibroblast cultures in spite of MMP inhibition. Furthermore, serine protease inhibition (via aprotinin) did not alter vascular network formation, either. Instead, dual application of MMP and serine protease/plasmin inhibitors was necessary to block capillary network formation. Thus, in fibrin gels, fibroblasts appear to stimulate ECs to expand their proteolytic repertoire and utilize both the MMP and PA/plasmin axes while undergoing capillary morphogenesis.

The mechanisms underlying this behavior are unclear. While plasmin activates a number of MMPs ³⁷, it is unlikely that an increase in the bulk concentration of plasmin (due to the increased levels of uPA activity observed in fibroblast cultures) results in levels of MMP activity that overwhelm either MMP inhibitor used, as each was applied at a dose that exceeded its reported IC₅₀ values by greater than 1000-fold. Instead, it is plausible that in fibroblast-EC cultures, a third type extracellular protease (e.g., cysteine cathepsins ¹⁸) is involved in the morphogenic process, and its activity is not quenched until both MMP and plasmin activity is inhibited.

Disparate growth factor secretion profiles may also underlie the differences observed in the proteolytic mechanisms utilized by ECs as they sprout in EC-MSC vs. EC-fibroblast cultures. We did not exhaustively explore this possibility, but did examine some likely candidates: VEGF, which signals through VEGF receptors to increase MT1-MMP expression ³⁸; HGF, shown to induce MT1-MMP, MMP2, uPA, and uPAR expression in cancer cells ^{39, 40}; and TNFα, which enhances uPAR expression by ECs ⁴¹. As shown in Supplemental Figure 1, sprouting in both EC-MSC and EC-fibroblast cultures was partially mediated by exogenous VEGF and endogenous HGF, but proceeded fully despite TNFαinhibition. Notably, differences were not observed between the two culture conditions.

We also considered that fibroblasts may directly or indirectly alter the deposition and arrangement of proteins in the periendothelial compartment ⁴², in turn affecting the dynamics of uPAR on the EC surface. We were unable to determine surface localization of uPAR in 3D culture using a fluorescently conjugated antibody targeting the membrane-bound receptor ⁴², and sprouting persisted in these cultures despite the addition of a function blocking antibody that specifically targets uPAR. Ablation of surface bound uPAR will be necessary to conclusively determine if this membrane-bound member of the PA/plasmin axis is a key node in the MMP-independent form of capillary morphogenesis observed in EC-fibroblast cultures.

A number of prior studies have demonstrated that another principle component of the microenvironment, the ECM, alters the proteolytic landscape utilized to facilitate morphogenic events. This is the case not only for capillary morphogenesis, but branching morphogenesis of mammary and salivary epithelia ^{43, 44}. Of course, substrate specificity comes into play, as some proteases are restricted to digesting a handful of ECM molecules, while others are less selective ³⁷. Though several MMPs are capable of cleaving type I collagen ³⁷, MT1-MMP is indispensable for invasion and capillary lumen formation within collagen I matrices ^{20, 45, 46}. In fibrin, the story is not as straightforward. Homotypic culture models have demonstrated that EC invasion of fibrin matrices relies solely on uPAR and not on MMPs ^{21, 22, 47}. However, other models have demonstrated EC tubulogenesis within fibrin matrices is MT-MMP dependent (though not necessarily dependent on MT1-MMP) when ECs are stimulated by either a broad array of growth factors ⁴⁸ or by other cell types (such as in a mouse aortic vessel explant assay ²⁰). Indeed, endowing MT-MMP naïve cells with either MT1-, MT2- or MT3-MMP is sufficient to confer them with the ability to invade fibrin ⁴⁹. Thus, the ability of MT1-null aortic explants to sprout in fibrin ²⁰ may be the result of heterotypic interactions which tune the cell surface proteases utilized by ECs as they undergo this morphogenic program. The data presented here suggest that changing the identity of the cell type on the other end of this heterotypic interaction can also modulate the proteases utilized by ECs to facilitate capillary morphogenesis.

A final implication of the present work is that the systems used in this study can be applied to model physiological and pathological angiogenesis. Recent work demonstrated that MSCs occupy a perivascular niche in a variety of organs and function as pericytes with multi-lineage potential ⁵⁰. Coupled with their ability to home to sites of injury ⁵¹ and foster a regenerative microenvironment via trophic activities ^{52, 53}, it is tempting to speculate that both bone marrow-derived and resident tissue MSCs function physiologically to aid in the repair of damaged mesenchyme, in part by stimulating and stabilizing neovasculature. We noticed a qualitative reduction in vessel coverage by fibroblasts compared to MSCs (Figs. 2C and C’). Pericytic investment stabilizes vessels and protects EC from apoptosis ⁵⁴, and reduced or sporadic vessel coverage is a hallmark of the reactive stroma present in chronic wounding disorders and surrounding carcinomas ⁵⁵, as is sustained fibroblast activation to the myofibroblast phenotype ^{56, 57} (marked by αSMA expression, which was observed in this and our prior study ¹⁰). Thus, the EC-MSC model of capillary morphogenesis may effectively simulate physiologic conditions while the EC-fibroblast model simulates pathological neovascularization. If true, this demonstrates the importance of using appropriate experimental models to uncover effective therapies for a given process. For instance, the failures of broad spectrum MMP inhibitors to effectively halt tumor dissemination and pathologic angiogenesis are well documented ⁵⁸. The present study illustrates how modeling complex interactions in appropriate experimental models could uncover novel approaches to address these shortcomings.

Supplementary Material

01. Supplementary Figure 1.

ECs exhibit VEGF- and HGF-dependent sprouting in both MSC and fibroblast laden matrices. ECs were co-cultured with either MSCs (top row) or fibroblasts (bottom row) within 3D ECMs under the following conditions: 1) with the addition of a mouse IgG control (10 µg/ml), 2) VEGF withdrawal from the culture medium, 3) with the addition of either 1 µg/ml or 10 µg/ml of a function blocking antibody targeting HGF (higher dose shown), or 4) with the addition of either 1 µg/ml or 10 µg/ml of a neutralizing antibody targeting TNFα (higher dose shown). Scale bar = 100 µm.

Abbreviations

3D

three-dimensional

αSMA

alpha-smooth muscle actin

EACA

ε-aminocaproic acid

EC

endothelial cell

ECM

extracellular matrix

GFP

green fluorescent protein

IF

immunofluorescent

MMP

matrix metalloproteinase

MSC

mesenchymal stem cell

MT-MMP

membrane-type matrix metalloproteinase

PA

plasminogen activator

TIMP

tissue inhibitor of metalloproteinases

uPA

urokinase-type plasminogen activator

uPAR

urokinase plasminogen activator receptor

tPA

tissue plasminogen activator