MMP2 sheds Glypican-4 to suppress Slit3-Robo4 signaling and angiogenesis

Abstract

Axon guidance molecules, initially identified for their roles in neural development, are now recognized as crucial regulators of angiogenesis and blood vessel patterning. Among these, Slit3 promotes endothelial cell migration, proliferation, and vascular network formation by signaling through the endothelial-specific receptor Robo4. This interaction depends on heparan sulfate (HS) as a co-receptor for assembly of the Slit3–Robo4 complex, although the specific HS proteoglycan involved has not yet been identified. Our recent work also shows that Robo4 ectodomain shedding by ADAM10 and ADAM17 suppresses this angiogenic pathway. Here, we identify matrix metalloproteinase-2 (MMP2) as an additional suppressor of Slit3–Robo4 signaling. MMP2 expression correlates with Slit3 in diaphragm mesenchymal and lung stromal cells and inhibits Slit3-induced endothelial cell proliferation, migration, and angiogenesis. Furthermore, we identify glypican-4 (GPC4) as an HS proteoglycan mediating Slit3–Robo4 co-receptor function and show that MMP2 cleaves GPC4, thereby disrupting Slit3–Robo4 signaling. These findings establish GPC4 as a vital co-receptor for Slit3–Robo4 signaling and reveal MMP2-mediated GPC4 shedding as a novel mechanism to regulate angiogenic responses. This study uncovers an additional level of proteolytic regulation in Slit3–Robo4 signaling-driven angiogenesis, broadening our understanding of how extracellular proteinases influence ligand–receptor/co-receptor interactions.

Introduction

Angiogenesis, the process of forming new blood vessels from existing vasculature, is a fundamental biological process essential for embryogenesis, tissue regeneration, and the progression of various diseases (1–3). During development, angiogenesis ensures the proper vascularization of growing tissues and organs, facilitating the delivery of oxygen and nutrients and the removal of metabolic waste. This process is coordinated by a complex network of signaling pathways, with vascular endothelial growth factor (VEGF) signaling playing a central role in promoting endothelial cell proliferation, migration, and survival (4). In addition to canonical growth factors, several signaling pathways originally described in neural development have emerged as key regulators of vascular morphogenesis (4, 5). These include axon guidance cues, including Slits, Netrins, Semaphorins, and Ephrins, which influence endothelial cell behavior through receptors Robos, UNC5, Plexins, and Ephs, respectively (6–8). The convergence of neural and vascular signaling modules highlights a conserved developmental strategy that coordinates the simultaneous patterning of both systems (6–8).

Slits are large, secreted glycoproteins conserved from C. elegans to mammals, with three isoforms—Slit1, Slit2, and Slit3—identified in vertebrates (9–12). These proteins share a modular structure that includes leucine-rich repeats, multiple epidermal growth factor (EGF)-like motifs, laminin G domains, and a cysteine-rich C-terminal region, which together facilitate receptor binding and downstream signaling (13, 14). Slit1 and Slit2 are primarily expressed in the central nervous system, whereas Slit3 has a broader distribution in peripheral tissues, including the placenta, mammary gland, skin, brain, and diaphragm, suggesting non-neuronal roles for Slit3 (15–17).

Slit proteins signal through the Roundabout (Robo) family of receptors, which includes four members (Robo1–4). Robo4 is uniquely expressed in endothelial cells (18). Our previous work demonstrated that Slit3 and Robo4 are highly co-expressed in the vasculature, identifying Slit3 as a strong angiogenic factor that acts through its interaction with Robo4 (19). Functional studies demonstrated that this signaling pathway promotes endothelial cell migration, proliferation, and vascular network formation, contributing to both physiological and pathological angiogenesis (19). We also found that Slit3-Robo4 signaling is modulated by heparan sulfate (HS), which binds Slit3, acting as a co-receptor to enhance the ligand-receptor binding and signal (20, 21). More recently, we reported that Robo4 undergoes proteolytic cleavage by ADAM10 and ADAM17, a process that suppresses Slit3-Robo4 signaling and limits Slit3`s angiogenic activity (22).

In this study, we identify matrix metalloproteinase-2 (MMP2) as a new suppressor of Slit3–Robo4 signaling. We also demonstrate that glypican-4 (GPC4) serves as a co-receptor for Slit3–Robo4 signaling and that MMP2 cleaves GPC4 to suppress Slit3-induced angiogenesis. These findings identify an HSPG that functions as a co-receptor for this angiogenic pathway and uncover co-receptor shedding as a new regulatory mechanism that modulates Slit3–Robo4 signaling, offering new insights into the complex regulation of Slit3-Robo4 signaling and angiogenesis.

RESULTS

Single-cell RNA sequencing shows that MMP2 expression is linked to Slit3 expression in murine diaphragm mesenchymal stem cells and lung stromal cells

Our previous studies showed that Slit3 promotes angiogenesis during diaphragm development through its interaction with the endothelial-specific receptor Robo4 (19, 20). This Slit3–Robo4 pathway is regulated by endothelial cell surface HS, which acts as a co-receptor by binding Slit3 (20, 21), and by Robo4 ectodomain shedding mediated by ADAM10 and ADAM17 (22). To investigate whether other molecules influence this pathway, we analyzed a publicly available single-cell RNA sequencing (scRNA-seq) dataset of over 100,000 cells from 20 murine organs and tissues (23). From this dataset, we focused on 870 cells from the diaphragm, including endothelial cells, lymphocytes, macrophages, mesenchymal stem cells (MSCs), and skeletal muscle satellite stem cells. Consistent with previous findings, Slit3 expression was limited to MSCs, while Robo4 was specifically expressed in endothelial cells (Fig. 1A).

Figure 1. Mmp2 is linked to Slit3 expression in diaphragmatic MSCs and lung stromal cells.

A, C. scRNA-seq profiling of Slit3 and Robo4 expression across various cell types in the mouse diaphragm (A) and lung (C). The thin lines indicate the range between the highest and lowest expression values, with boxes showing the first and third quartiles. A thick horizontal line within the boxes represents the median. B, D. KEGG enrichment analysis reveals that the Slit3-associated genes in diaphragm MSCs (Table S1) and lung stromal cells (Table S2) regulate multiple biological pathways involved in angiogenesis. E. Slit3-associated genes related to HS biosynthesis, HSPGs, MMPs, and TIMPs in mouse diaphragm MSCs or lung stromal cells. Adjusted p-values are calculated using the Benjamini-Hochberg procedure.

To identify transcriptional networks linked to Slit3 and Robo4, we performed Kendall’s τ correlation analysis on 241 MSCs and 79 endothelial cells. No genes showed significant correlation with Robo4 in endothelial cells, likely due to limited sequencing depth. In contrast, 45 genes were significantly associated with Slit3 expression in MSCs, with an additional 128 genes showing strong trends toward association (Table S1). Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis revealed that these Slit3-associated genes are involved in biological processes relevant to angiogenesis, such as focal adhesion, PI3K–Akt signaling, proteoglycans in cancer, cytoskeletal organization, regulation of actin dynamics, extracellular matrix (ECM)–receptor interaction, and adherent junctions (Fig. 1B).

To assess whether these transcriptional associations are conserved across different tissues, we conducted a parallel analysis using scRNA-seq data from murine lung tissue from the same study (23). Similar to the diaphragm, Slit3 and Robo4 showed cell-type-specific expression in lung stromal cells and endothelial cells, respectively (Fig. 1C). Kendall’s τ analysis identified over 200 genes associated with Slit3 in lung stromal cells (Table S2), and KEGG enrichment analysis revealed functional categories similar to those observed in diaphragm MSCs (Fig. 1D), indicating conserved regulatory roles for Slit3 across organ systems.

Among the Slit3-associated genes, we identified components involved in HS biosynthesis and remodeling, including Ext2 and Sulf2, as well as HS proteoglycans (HSPGs) GPC3 and GPC6 (Fig. 1E). These findings align with our previous work and others, which demonstrated that HS modulates Slit3–Robo4 signaling in angiogenesis (20, 21) and the critical roles of HS and HSPGs in angiogenesis in general (24–30). Considering our recent discovery that Robo4 shedding by ADAM10 and ADAM17 regulates Slit3–Robo4 signaling (22), we also examined genes involved in ectodomain shedding. Notably, Mmp2, Mmp23, and their endogenous inhibitor Timp2 were significantly associated with Slit3 expression. Mmp2 was consistently correlated with Slit3 in both diaphragm MSCs and lung stromal cells, and Timp2, a key inhibitor of MMP2, emerged as the second most Slit3-associated gene in the lung dataset. Given the established role of MMP2 in ECM remodeling and angiogenesis (31–33), these findings suggest that MMP2 may modulate Slit3-driven angiogenesis.

MMP2 reduces Slit3-induced angiogenesis, and Slit3 does not influence MMP2 expression or maturation in endothelial cells

Slit3 has been shown to strongly promote endothelial cell proliferation, migration, and tube formation across various endothelial cell types, including an immortalized mouse diaphragm endothelial cell (dEC) line that we previously generated, which expresses Robo4 (19–21). In vivo, Slit3 also stimulates neovascularization in mouse skin and corneal models (19–21). To determine whether MMP2 influences Slit3-induced angiogenesis, we first tested its effects on endothelial cell behavior. Exogenous MMP2 significantly reduced Slit3-induced proliferation and migration in both mouse dECs and human brain endothelial cells (Fig. 2A–D). Moreover, MMP2 suppressed Slit3-driven neovascularization in vivo in the Matrigel plug angiogenesis assay (Fig. 2E).

Figure 2. MMP2 inhibits Slit3-induced angiogenesis both in vitro and in vivo.

A–D. MMP2 blocks Slit3-induced proliferation (A, C) and migration (B, D) of mouse diaphragmatic endothelial cells (dECs; A–B) and human brain endothelial cells (bECs; C–D). Proliferation was measured after 36 hours in DMEM with 0.2% fetal bovine serum; migration was tested using transwell assays in serum-free DMEM for 12 hours. Cells were treated with Slit3 alone (1 μg/ml), Slit3 plus MMP2 (100 ng/ml), or Slit3 plus the MMP2 inhibitor ARP100 (ARP; 60 nM). Bovine serum albumin (BSA) served as a control. E. MMP2 reduces Slit3-induced new blood vessel formation in vivo, as shown by a Matrigel plug assay. Growth factor-reduced Matrigel with Slit3 or BSA (1 μg), with or without MMP2 (100 ng/ml), was injected subcutaneously into mice. Plugs were collected after two weeks; representative images are shown. Hemoglobin levels were measured (n = 5 mice per group). F-G. Knockdown (KD) of endogenous Mmp2 increases Slit3-induced proliferation (F) and migration (G) of dECs. Cells with stable expression of scrambled control (SC) or Mmp2 siRNA were used; knockdown efficiency is shown in Figure S1. The in vitro experiments were done in triplicate. Statistical significance was tested with Student’s t-test: ns, not significant; *, P < 0.05; **, P < 0.01; ***, P < 0.001.

We next examined endogenous MMP2 expression in dECs and found that dECs express MMP2, most of which is secreted into the culture medium (Fig. S1A). Inhibition of endogenous MMP2 activity with ARP100, a selective MMP2 inhibitor, increased dEC migration (Fig. 2A). Similarly, siRNA-mediated knockdown of Mmp2 enhanced dEC proliferation and migration (Fig. 2F, G; Fig. S1B, C). To determine whether Slit3 regulates MMP2 expression or activation, we performed gelatin zymography. Slit3 treatment did not change MMP2 protein levels or enzymatic activity, as indicated by the unchanged distribution of pro- and mature forms (Fig. S1D), suggesting that Slit3 does not influence MMP2 expression or maturation. Overall, these findings show that MMP2 acts as a negative regulator of Slit3-induced angiogenesis, and Slit3 does not influence its expression and maturation in endothelial cells.

MMP2 does not cleave Slit3 or Robo4 to inhibit Slit3-induced angiogenesis

Slit3 is a large secreted glycoprotein (~200 kDa) (12) that undergoes proteolytic processing to generate a ~140 kDa N-terminal fragment and a 50–60 kDa C-terminal fragment (19, 34). The full-length Slit3 and its N-terminal fragment are primarily membrane-associated and mediate Slit3’s biological functions, while the C-terminal fragment is more diffusely distributed, and its physiological role remains largely unknown (19, 34, 35). We previously reported that the Slit3 C-terminal fragment neutralizes heparin’s anticoagulant activity, thereby exerting a pro-coagulant effect (36). To determine whether MMP2 inhibits Slit3-induced angiogenesis by cleaving Slit3, we incubated recombinant Slit3 (amino acids 34–1116), which binds Robo4 and is the largest commercially available angiogenic Slit3 fragment (19), with MMP2 at 37°C for 6 hours. No degradation or appearance of lower molecular weight bands was observed (Fig. 3A). Similarly, treating A549 human lung carcinoma cells, which naturally express high levels of Slit3 protein (37), with MMP2 under the same conditions showed only full-length Slit3 in both conditioned medium and cell lysates (Fig. 3B). These results indicate that MMP2 does not cleave Slit3.

Figure 3. MMP2 does not cleave Slit3 or cause Robo4 shedding from the endothelial cell surface.

A. Recombinant Slit3 is resistant to cleavage by MMP2. His-tagged recombinant human Slit3 N-terminal fragment (amino acids 34–1116) was incubated with BSA or MMP2 in DMEM at 37°C for 6 hours. Proteins were separated by SDS-PAGE and detected with an anti-His antibody. The higher- and lower-molecular-weight bands likely represent glycosylated and non-glycosylated Slit3 isoforms. B. MMP2 does not cleave cell-secreted Slit3. A549 cells were treated with BSA or MMP2 in DMEM at 37°C for 6 hours. Slit3 in conditioned medium (CM) and cell lysates was detected by Western blot using an anti-human Slit3 antibody. Only full-length Slit3 (~200 kDa) was observed. Slit3 in cell lysates reflects its membrane-associated form. C, D. MMP2 does not promote Robo4 shedding from the endothelial cell surface. dECs were treated with BSA, MMP2, or ARP100 in DMEM at 37°C for 6 hours. In (C), Robo4 in CM and cell lysates was detected by Western blot using an anti-Robo4 N-terminal antibody and normalized to BSA-treated controls. In (D), dECs stably expressing Robo4-HA-FLAG were analyzed by Western blot with an anti-FLAG antibody. Protein levels were normalized to actin (where applicable) and expressed relative to control (mean ± SEM). Statistical analysis was performed using Student’s t-test: ns, not significant; *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Our recent studies showed that Robo4 is constantly shed from the endothelial surface by ADAM10 and ADAM17, and that the shed Robo4 fragment acts as a decoy to block Slit3-induced angiogenesis (22). To determine whether MMP2 similarly sheds Robo4, we treated dECs with MMP2 and measured Robo4 levels in both the conditioned medium and cell lysates. Neither the addition of MMP2 nor the knockdown of Mmp2 changed Robo4 shedding, as evidenced by unchanged levels of shed Robo4 and membrane-associated Robo4 (Fig. 3C). Additionally, MMP2 treatment did not influence the expression or shedding of Robo1 (Fig. S2), which binds Slit3 (38, 39) and forms Robo1–Robo4 heterodimers that regulate Robo4 function (40). Overall, these results indicate that MMP2 does not inhibit Slit3-induced angiogenesis by cleaving either Slit3 or Robo4.

MMP2 specifically cleaves Glypican4 from the surface of endothelial cells

Our previous studies showed that HS binds Slit proteins and acts as a co-receptor for Slit3, which is crucial for Slit3–Robo4-mediated angiogenesis (20). However, the specific HSPGs involved in this process remained unknown. Since MMPs are known to cleave HSPGs (24–26, 41, 42), we investigated whether MMP2 sheds any HSPGs from endothelial cells. Proteomic analysis of conditioned medium from serum-free dEC culture revealed the presence of syndecan-4 (SDC4), GPC1, and GPC4 (Fig. S3). To determine if MMP2 mediates shedding of these HSPGs from the endothelial cell surface, we treated dECs with MMP2 and observed increased levels of GPC4—but not GPC1 or SDC4—in the conditioned medium (Fig. 4A–C). Consistent with this, blocking MMP2 activity with ARP100 or using siRNA to knock down Mmp2 significantly reduced GPC4 shedding (Fig. 4D, E; Fig.S4). These results demonstrate that MMP2 specifically cleaves GPC4 from the surface of endothelial cells, but not GPC1 or SDC4.

Figure 4. MMP2 induces GPC4 shedding from the endothelial cell surface.

A, B. MMP2 does not promote shedding of GPC1 or SDC4. dECs were treated with MMP2, and the levels of GPC1 and SDC4 in conditioned media (CM) and cell lysates were analyzed by Western blot using antibodies targeting their N-terminal domains. C–E. MMP2 induces GPC4 shedding. GPC4 levels in CM and cell lysates were evaluated after treatment with MMP2 or ARP100 (C, D), or in dECs with stable Mmp2 knockdown (E). Western blot analysis using an N-terminal GPC4 antibody revealed increased GPC4 shedding following MMP2 treatment, which was blocked by ARP100 or Mmp2 knockdown. Protein quantification was normalized to actin (where applicable) and then compared to control levels, expressed as the mean ± SEM. Statistical analysis was performed using Student’s t-test: ns, not significant; *, P < 0.05; **, P < 0.01; ***, P < 0.001.

GPC4 mediates Slit3-induced angiogenesis, and MMP2 inhibition enhances this process depending on GPC4

To determine whether GPC4 influences Slit3-induced angiogenesis, we neutralized cell-surface GPC4 with a polyclonal rabbit antibody targeting its C-terminal domain, using naïve rabbit IgG as a control. The anti-GPC4 antibody significantly decreased Slit3-induced dEC migration (Fig. 5A). In vivo, the anti-GPC4 antibody also reduced Slit3-driven neovascularization in the Matrigel plug assay (Fig. 5B). Meanwhile, we silenced GPC4 expression in dECs using CRISPR interference (CRISPRi; Fig. S4). In transwell migration assays, CRISPRi-mediated GPC4 knockdown abolished Slit3-induced dEC migration, further supporting GPC4 as a HSPG co-receptor facilitating Slit3–Robo4 signaling (Fig. 5C). Notably, GPC4 knockdown led to a significant increase in Robo4 expression (Fig. S4A), suggesting a compensatory mechanism and reinforcing its role as a co-receptor in Slit3–Robo4 signaling. Additionally, GPC4 knockdown reduced the enhancement of Slit3-induced migration when MMP2 was inhibited with ARP100, indicating that the effect of MMP2 inhibition depends on preventing GPC4 shedding from the endothelial surface (Fig. 5C). Overall, these results demonstrate that GPC4 is required for Slit3–Robo4-mediated angiogenesis and that MMP2 suppresses this pathway by cleaving GPC4.

Figure 5. GPC4 is essential for Slit3-induced angiogenesis both in vitro and in vivo.

A. Anti-GPC4 antibody blocks Slit3-induced dEC migration. Transwell migration was conducted for twelve hours with either BSA or Slit3 (1 μg/ml), with naïve rabbit IgG or anti-GPC4 rabbit IgG. B. Knockdown of GPC4 reduces Slit3-induced angiogenesis, and MMP2 inhibition increases Slit3-induced dEC proliferation. dEC proliferation was measured over 36 hours with culture medium supplemented with BSA or Slit3, with or without ARP100. C. Anti-GPC4 antibody suppresses Slit3-induced angiogenesis in vivo. Matrigel plugs were supplemented with BSA or Slit3, with or without naïve rabbit IgG or anti-GPC4 rabbit IgG. Hemoglobin levels in the harvested plugs were measured and expressed as the hemoglobin index, normalized to the BSA + naïve IgG group (n = 10 plugs per group).

MMP2 disrupts GPC4-Robo4 and Slit3-Robo4 complexes; therefore, the functional Slit3- Robo4/GPC4 complex

Our previous studies showed that Slit3 binds to heparin, a highly sulfated form of HS, and to endothelial cell-surface HS, but we did not observe Robo4 ectodomain binding to heparin. This indicates that Robo4 likely does not interact with the HS moiety of GPC4 (20, 21). However, it remained unclear whether Robo4 binds to or is indirectly associated with the core protein of GPC4. To investigate this, we tested the GPC4–Robo4 interaction or association in the absence of Slit3. Human Robo4 with HA and FLAG tags (hRobo4-HA-FLAG) was transiently expressed in dECs, and cell lysates were subjected to anti-FLAG immunoprecipitation. GPC4 co-precipitated with hRobo4-HA-FLAG, and this interaction was significantly reduced after MMP2 treatment, suggesting GPC4 binds Robo4 likely via its co-protein, and MMP2 disrupts this binding (Fig. 6A). Next, we examined whether MMP2 affects the Slit3–Robo4 interaction. After incubation with Slit3-his to allow cell-surface binding, the cells were lysed, and the complex was pulled down using anti-His antibody-coated beads. Slit3 co-precipitated with hRobo4-HA-FLAG, but this complex formation was decreased when cell lysates were also treated with MMP2 (Fig. 6B). Overall, these findings suggest that Slit3 forms a functional complex with Robo4 and GPC4 on the endothelial surface, and MMP2 disrupts this complex, reducing Slit3–Robo4 signaling and angiogenesis (Fig. 6C).

Figure 6. GPC4 forms a ternary complex with Slit3 and Robo4 on the endothelial cell surface, and MMP2 disrupts these complexes to inhibit Slit3-induced angiogenesis.

A. dECs overexpressing human Robo4-HA-FLAG were treated with BSA or MMP2 for 6 hours. Cell lysates were immunoprecipitated using a mouse anti-FLAG antibody, and GPC4 and Robo4 were detected with rabbit anti-GPC4 C-terminal and anti-FLAG antibodies, respectively, in a Western blot. GPC4 co-precipitated with Robo4, and this interaction was reduced by MMP2 treatment. B. Cells were treated as in (A), with the addition of His-tagged Slit3. Lysates were immunoprecipitated with an anti-His antibody and probed for Slit3 and Robo4 with anti-His and anti-human Robo4 antibodies, respectively. Robo4 co-precipitated with Slit3, and this interaction decreased with MMP2. C. Proposed model: GPC4 interacts with Robo4 to form a receptor complex on the endothelial surface. Slit3 binds to HS chains of GPC4 and engages Robo4, forming a GPC4-Slit3-Robo4 ternary complex that promotes angiogenesis. MMP2 cleaves cell surface GPC4, disrupting complex formation and reducing Slit3-induced angiogenesis.

DISCUSSION

Axon guidance molecules, initially known for their roles in neural development, are now recognized as crucial regulators of angiogenesis. Growing evidence underscores their involvement in vascular biology, though the mechanisms underlying this regulation remain poorly understood. Our earlier work identified Slit3, a primarily peripheral axon guidance cue, as a potent angiogenic factor that signals through the endothelial-specific receptor Robo4 (19). This signaling depends on HS on the endothelial cell surface, which acts as a co-receptor by binding Slit3 and facilitating the formation of the Slit3–Robo4 complex (20, 21). We also demonstrated that Robo4 ectodomain shedding by ADAM10 and ADAM17 reduces the activity of this pathway (22). In this study, we identify GPC4 that acts as a HSPG co-receptor in Slit3-Robo4 signaling. Notably, we find that GPC4 can be shed by MMP2, which then disrupts Slit3–Robo4 signaling. These findings reveal an additional layer of regulation in angiogenesis, broadening our understanding of how extracellular proteinases influence ligand, receptor, and co-receptor interactions in Slit3-Robo4 signaling and angiogenesis.

Through transcriptional network analysis, we found that Slit3 expression in stromal cells of the diaphragm and lung correlates with HS biosynthesis and remodeling, supporting our previous finding that HS is crucial for Slit3–Robo4 signaling. Notably, MMP2 expression also correlates with Slit3 in these stromal populations, echoing our earlier observation that receptor shedding regulates this pathway (22). MMP2 is well known for its diverse roles in angiogenesis, including ECM remodeling, activation of pro-MMP2, release of angiogenic factors, and shedding of cell-surface receptors such as VEGFR2, thereby fine-tuning pro- and anti-angiogenic signals (32). Our functional studies show that MMP2 suppresses Slit3-induced angiogenesis by targeting endothelial GPC4, the co-receptor for Slit3–Robo4 signaling. Consistent with this, MMP2 is abundantly expressed and secreted by endothelial cells, and its inhibition or genetic deletion enhances Slit3-driven endothelial proliferation, migration, and new blood vessel formation in vivo. Overall, these findings identify MMP2 as a new suppressor of Slit3–Robo4 signaling and suggest it acts through both cell-autonomous and non-autonomous mechanisms to limit Slit3-Robo4 signaling-driven angiogenesis.

Although HS has been identified as a co-receptor for Slit3, the specific HSPGs involved remain unknown. Our study focused on the HSPGs that can be shed by MMP2. Among several candidates detected in the conditioned medium of dEC culture—including SDC4, GPC1, and GPC4—only GPC4 shedding increased upon treatment with MMP2. Co-immunoprecipitation confirmed that MMP2 reduces both GPC4–Robo4 and Slit3–Robo4 complexes, supporting the conclusion that MMP2 sheds GPC4 to suppress Slit3–Robo4 signaling. GPC4 was previously reported to undergo cleavage by furin-like proteases at R354, resulting in two subunits (43), and astrocyte-expressed GPC4 is robustly released by ADAM9 and, to a lesser extent, by GPI-anchor cleavage (44). Our findings identify MMP2 as a novel sheddase of GPC4. It is worth noting that our gelatin zymography analysis observed that dECs predominantly secrete MMP2, whereas human brain and mouse lung endothelial cells secrete abundant levels of both MMP2 and MMP9 (Fig. S1D). It will be interesting to determine whether MMP can shed GPC4 from brain and lung endothelial cells and, similarly, suppress Slit3–Robo4 signaling and angiogenesis.

GPC1 has been implicated as a co-receptor for Slit–Robo signaling in the rat brain (45), and syndecan has been shown to mediate Slit signaling during axon guidance and muscle patterning in Drosophila (46). Whether SDC4 and GPC1 function as HSPG coreceptors for Slit3–Robo4 signaling in endothelial cells remains unknown. Meanwhile, our unbiased proteomics analysis examined only HSPGs shed into the conditioned medium, leaving it unclear whether endothelial cells, such as dECs, express additional HSPGs beyond GPC1, GPC4, and SDC4 that may also serve as co-receptors. Addressing these questions will be an important direction for future studies.

Despite strong evidence that MMP2 sheds GPC4 to inhibit Slit3–Robo4 signaling, including increased GPC4 shedding in conditioned media, decreased GPC4–Robo4 and Slit3–Robo4 complexes, and the loss of enhanced Slit3-induced angiogenesis upon GPC4 knockdown, MMP2 treatment did not reduce total GPC4 levels on the endothelial surface (data not shown). This unexpected finding raises essential questions about GPC4 expression dynamics. The mechanism remains unclear, but one possibility is that MMP2 reduces GPC4 endocytosis by altering lipid raft composition and its interaction with the endocytic machinery (31, 33). These hypotheses warrant further investigation as well.

In summary, our findings uncover a complex regulation of Slit3–Robo4 signaling involving HSPGs and their proteolytic shedding. We identify GPC4 as an essential HSPG co-receptor required for Slit3–Robo4-mediated angiogenesis and demonstrate that MMP2 suppresses this pathway by cleaving GPC4 from the endothelial surface. This mechanism introduces an additional layer of regulation that fine-tunes angiogenic responses and likely intersects with other signaling pathways through shared co-receptors and sheddases. These insights provide a framework for understanding how extracellular proteinases integrate various cues to guide vascular growth and remodeling.

METHODS

Bioinformatics Analysis

scRNA-seq data for murine diaphragm and lung tissues were obtained from the Tabula Muris Consortium dataset (23). From these data, raw gene count matrices of the diaphragm and the lung were extracted for downstream analysis. Cells lacking detectable GAPDH expression were excluded during initial quality control. To account for differences in sequencing depth and library size, the expression values of all other genes were normalized to those of GAPDH within each cell. To explore genes whose expression patterns were associated with Slit3, we calculated pairwise correlations between Slit3 and all other genes within each cell population (diaphragm MSCs and lung stromal cells). Correlation strength and significance were assessed using Kendall’s rank correlation coefficient (τ) and the corresponding p-values. Genes showing significant correlation with Slit3 (raw p < 0.05) were selected for functional enrichment analysis. Over-representation analysis (ORA) was performed using the clusterProfiler R package (version 4.16.0) to test for enrichment against the KEGG and Gene Ontology (GO) Biological Process (BP) databases. The full murine genome served as the reference background. Enrichment results with Benjamini–Hochberg adjusted p-values below 0.05 were considered statistically significant. All analyses were conducted using R (version 4.5.1).

Reagents

Plasmids:

Plasmids were obtained from Addgene unless otherwise noted. Human Robo4 cDNA (NM_019055.5) and the pCDH-EF1-FHC vector (#64874) were used to generate the hRobo4-HA-FLAG construct. Lentiviral packaging was performed using the psPAX2 (#12260) and pMD2.G (#12259) vectors. For CRISPR-based gene modulation, the following lentiviral plasmids were used: LentiGuide-Puro (#52963), Lenti_dCas9-KRAB-MeCP2 (#122205), and pCDH-EF1-FHC (#64874). Mmp2 siRNA sets (io-43486) were purchased from Abmgood.

Antibodies:

Primary and secondary antibodies were sourced from various suppliers: LSBio (rabbit polyclonal anti-GPC4 IgG, LS-C808928); ABclonal (rabbit polyclonal anti-GPC4 IgG, A12805; anti-GPC1 IgG, A13019); Novus Biologicals (rabbit polyclonal anti-SDC4 IgG, NBP2–24630; anti-MMP2 [pro and active], NB200–193); Abcam (rabbit polyclonal anti-mouse Robo4 N-terminal IgG, #10547; naïve rabbit IgG, #172730); Invitrogen (mouse monoclonal anti-FLAG IgG2b, #14-6681-82; goat anti-mouse IgG-HRP, #626520); EMD Millipore (goat anti-rabbit IgG-HRP, AP187P); BioLegend (HRP-conjugated mouse monoclonal anti-His tag antibody, 652504); Thermo Scientific (goat anti-mouse IgM-Alexa Fluor 488, A-21042); GeneTex (mouse anti-heparan sulfate IgM, 10E4, GTX20073); and Santa Cruz Biotechnology (HRP-conjugated donkey anti-goat IgG, sc-2020).

Proteins and chemicals:

Recombinant human MMP2 (#230-00214-100) was obtained from RayBiotech. ARP100 was sourced from TOCRIS (#2621), Fisher Scientific (AAJ64151LB0), and AdooQ (A15292). His-tagged recombinant mouse Slit3 (3629-SL-050) was from R&D Systems. Additional reagents included a 30 kDa cut-off centrifugal filter unit (UFC803024, EMD Millipore); Lenti-X solution (#631232, Takara Bio); KwikQuant Ultra Digital-ECL substrate (R1002) and KwikQuant Imager (D1001) from Kindle Biosciences; Drapkin’s reagent (D5941–6VL, Sigma-Aldrich); iScript cDNA synthesis kit (#1708890); SYBR Green PCR kit (#1725270) from Bio-Rad; CIM plate (#05665817001); E-16 plate (#00300600890) from ACEA Biosciences; DTSSP (#21578); NeutrAvidin (#PI29200); serum-reduced Opti-MEM (#31985070) from Thermo Scientific; Sulfo-NHS-SS-Biotin (A8005, APExBIO); and Protein A-Agarose (sc-2001, Santa Cruz Biotechnology).

Cell lines

Mouse dECs were previously established by immortalizing primary cells from adult C57BL/6 mice with SV40 large T-antigen expression, followed by single-cell cloning (20, 47–50). These cell lines were maintained in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and penicillin-streptomycin (100 U/ml and 100 μg/ml, respectively) in a humidified incubator at 37°C with 5% CO₂. The human brain endothelial cell line hCMEC/D3 was obtained from Millipore Sigma (SCC066) and cultured in EBM-2 medium supplemented with vascular endothelial growth factor, insulin-like growth factor-1, epidermal growth factor, basic fibroblast growth factor, hydrocortisone, ascorbate, penicillin-streptomycin, and 2.5% FBS, according to the manufacturer’s instructions.

Preparation of concentrated conditioned media

To evaluate shedding of Robo4 and HSPGs and the presence of MMP2 in the conditioned medium (CM), 3 × 10⁶ endothelial cells were plated in 150-mm culture dishes and incubated for 12 hours. The cells were then starved in 10 ml of DMEM on an orbital shaker for 6 hours. CM was collected, centrifuged at 400 × g for 5 minutes to remove cellular debris, filtered through a 0.2-μm membrane, and concentrated to approximately 50 μl using a 30-kDa cutoff centrifugal filter unit.

To examine the role of MMP2 in shedding Robo4 and HSPGs, endothelial cells were starved for 6 hours in serum-free DMEM, then treated with DMEM containing BSA (100 ng/ml), MMP2 (100 ng/ml), or the MMP2 inhibitor ARP100 (60 nM). After 6 hours, conditioned medium was collected and analyzed for shed Robo4 and HSPGs.

Western blot

Concentrated CM and 2% of the corresponding cell lysates were separated by SDS-PAGE using either 7.5% or 12% polyacrylamide gels, then transferred to nitrocellulose membranes. Membranes were blocked for 30 minutes in 4% non-fat milk prepared in 1× TBST. Primary antibodies were used at concentrations of 0.5–1 μg/ml, and secondary antibodies at 0.1–0.5 μg/ml. Protein bands were visualized by chemiluminescence using the KwikQuant Imager system (Kindle Biosciences, D1001). Band intensities were quantified with ImageJ software.

Zymography

The conditioned medium concentrated 100-fold was loaded onto SDS-PAGE gels containing 1 mg/ml gelatin. After electrophoresis, gels were incubated twice for 20 minutes each in 2.5% Triton X-100 prepared in distilled water to remove SDS. Gels were then rinsed five times with distilled water and incubated in developing buffer (50 mM Tris-HCl, 200 mM NaCl, 5 mM CaCl₂, 0.02% Brij 35; pH 7.8) for 30 minutes at room temperature. Next, gels were transferred to fresh developing buffer and incubated for 16 hours at 37°C on an orbital shaker. Proteolytic activity was visualized by Coomassie Brilliant Blue staining.

Endothelial cell migration and proliferation assays

For migration assays, starved dECs were seeded at 20,000 cells per well in CIM-16 plates containing serum-free DMEM supplemented with BSA (1 μg/ml) or Slit3 (1 μg/ml), in the absence or presence of anti-GPC4 antibody (20 μg/ml), naïve rabbit IgG (20 μg/ml), or ARP100 (60 nM). Migration was monitored for 9 hours using the RTCA-DP system (ACEA Biosciences), and the migration index was calculated by normalizing to the BSA control.

For proliferation assays, dECs were seeded at 5,000 cells per well in E-16 plates and cultured in DMEM with 0.2% FBS, supplemented with BSA or Slit3 (1 μg/ml), with or without anti-GPC4 antibody (20 μg/ml), naïve IgG (20 μg/ml), or ARP100 (60 nM). Proliferation was monitored over 36 hours using the RTCA-DP system.

Matrigel plug angiogenesis assay

Female C57BL/6 mice (6–8 weeks old) were purchased from The Jackson Laboratory. All animal procedures received approval from the University of South Florida Institutional Animal Care and Use Committee and were performed in accordance with ARRIVE guidelines and standards established by the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC). For the in vivo angiogenesis assay, pre-chilled low-growth-factor Matrigel was supplemented with Slit3 or BSA (1 μg), with or without ARP100 (60 nM), naïve rabbit IgG (4 μg), or anti-GPC4 antibody (4 μg). A total of 0.5 ml was injected subcutaneously into both the left and right caudal ventral regions of each mouse. Matrigel plugs were collected two weeks after implantation, weighed, photographed, and homogenized in distilled water. These homogenates were centrifuged at 10,000 × g for 15 minutes, and hemoglobin levels in the supernatant were measured using Drabkin’s reagent (Sigma, D5941–6VL), following the manufacturer’s instructions.

Mass spectrometry

CM was collected from 20 confluent 150 mm petri dishes with dECs. The medium was initially centrifuged at 400 g for 5 minutes and then filtered through a 100 kDa cutoff protein filter (Millipore, #C7715) at 400 g centrifugation. The resulting flow-through was concentrated to 100 μL by centrifugation over spin columns with a 30 kDa molecular weight cutoff. The concentrated CM was separated on a 7. 5% PAGE gel, which was then cut into 10 bands. Each band was reduced with dithiothreitol (DTT), alkylated with iodoacetamide (IAA), and digested overnight with Trypsin/Lys-C at 37°C. After digestion, peptides were extracted from the gel using a solution of 50% water, 50% acetonitrile, and 0. 0.1% formic acid. The peptides were entirely dried in a speed-vacuum centrifuge and resuspended in 1% acetonitrile with 0. 1% formic acid. For LC- MS/MS analysis, peptides were separated on a 50 cm C 18 reversed-phase UHPLC column (ThermoFisher Scientific) on an Ultimate 3000 UHPLC system (ThermoFisher Scientific) with a 120-minute gradient from 2% to 32% acetonitrile/0—1% formic acid. Analysis was performed on a hybrid quadrupole-Orbitrap mass spectrometer (Q Exactive Plus, Thermo Fisher Scientific) using data-dependent acquisition (DDA), where the top 10 most abundant ions were selected for MS/MS fragmentation. Full MS scans ranged from 375–1500 m/z at a resolution of 70,000, with MS/MS scans at a resolution of 17,500. Raw data files were processed using MaxQuant (www.maxquant.org) and searched against 55,471 proteins from the UniProt Mouse Reference protein sequence database (Proteome ID: UP000000589, version May 4, 2020). Search parameters included a constant modification of cysteine by carbamidomethylation and variable modifications of methionine oxidation and protein N-terminal acetylation. The first search peptide tolerance was set at 20 ppm, and the main search peptide tolerance at 4.5 ppm. The fragment ion mass tolerance was 10 ppm. Protein identifications were accepted at a 1% false discovery rate for both proteins and peptides. The mass spectrometry analysis was carried out using a Q Exactive ^™ Plus Hybrid Quadrupole-Orbitrap ^™ Mass Spectrometer in the Proteomics Core at the University of South Florida.

CM was collected from twenty confluent 150-mm dishes of dECs. It was first centrifuged at 400 × g for 5 minutes to remove cellular debris, then filtered through a 100-kDa cut-off protein filter (Millipore, #C7715) at 400 × g. The flow-through was further concentrated to 100 μl using a 30-kDa cut-off centrifugal filter unit. The concentrated CM was separated on a 7.5% SDS-PAGE gel, and the gel lane was cut into ten bands for in-gel digestion. Each band underwent reduction with dithiothreitol (DTT), alkylation with iodoacetamide (IAA), and overnight digestion with Trypsin/Lys-C at 37°C. Peptides were extracted with a solution of 50% water, 50% acetonitrile, and 0.1% formic acid, dried in a SpeedVac centrifuge, and resuspended in 1% acetonitrile/0.1% formic acid. Peptides were separated on a 50-cm C18 reverse-phase UHPLC column (Thermo Fisher Scientific) using an Ultimate3000 UHPLC system with a 120-minute gradient (2–32% acetonitrile/0.1% formic acid). Mass spectrometry was performed on a Q Exactive Plus Hybrid Quadrupole-Orbitrap mass spectrometer (Thermo Fisher Scientific) using data-dependent acquisition (DDA), which selected the top 10 most abundant ions for MS/MS analysis. Full MS scans covered a mass range of 375–1500 m/z at a resolution of 70,000, while MS/MS scans were taken at a resolution of 17,500. Raw data were processed with MaxQuant (www.maxquant.org) and searched against the UniProt Mouse Reference Protein Sequence Database (Proteome ID: UP000000589, version May 4, 2020), which contains 55,471 protein entries. Search parameters included a fixed modification of cysteine carbamidomethylation and variable modifications of methionine oxidation and N-terminal acetylation. The first search peptide tolerance was set at 20 ppm, the main search tolerance at 4.5 ppm, and fragment ion tolerance at 10 ppm. Protein identification was filtered at a 1% false discovery rate (FDR) for both protein and peptide levels. Mass spectrometry analyses were performed at the Proteomics Core Facility, University of South Florida.

Knockout of mouse Gpc4

The Gpc4 in dEC cells was knocked out using the CRISPR-Cas9 method (49). Briefly, guide RNAs were designed to target protospacer adjacent motif (PAM) sites located 39 nucleotides (nt) and 209 nt upstream of the mouse Gpc4 transcription start site. The sequences for these guide RNAs are as follows: 1. Gpc4-cr-39F: CACCGTCCGCAAACTTGGCCAAGCG and Gpc4-cr-39R: AAACCGCTTGGCCAAGTTTGCGGAC; and 2. Gpc4-cr-209F: CACCGCAGCAGGGCGTGTCGCAAGA and Gpc4-cr-209R: AAACTCTTGCGACACGCCCTGCTGC. These oligos were cloned into the Lentiguide-Puro vector at BsmBI sites to generate the Gpc4-CRISPRi construct. Successful insertions were confirmed by PCR targeting the forward oligo and 3’ sequencing of the vector (5'-ATTGTGGATGAATACTGCC-3'). To produce lentivirus expressing mouse Gpc4 guide RNA or scramble, 10 μg of the Gpc4-CRISPRi plasmid or an empty vector (Lenti_dCas9-KRAB-MeCP2) was mixed with 5 μg of each lentiviral packaging vector psPAX2 and pMD2.g in 0.5 ml pre-warmed serum-reduced Opti-MEM after vortexing for 5 min. Additionally, 60 μl of 1 mg/ml linear Polyethylenimine hydrochloride (PEI) was incubated with 0.5 ml pre-warmed serum-reduced Opti-MEM for 5 min at room temperature after vortexing. The plasmid mixture and PEI were combined and incubated at room temperature for 20 min. This mixture was transfected into HEK293T cells at 80% confluence in a 10 cm culture plate with 5 ml of antibiotic-free DMEM containing 10% FBS at 37°C. After 24 h, conditioned media were collected and stored at 4°C. The cells were then incubated with an additional 5 mL of antibiotic-free DMEM containing 10% FBS at 37°C for 24 hours. The pooled conditioned media were centrifuged at 500g for 10 min to remove cell debris. The lentivirus was then mixed with one-third volume of 0.45-micron filtered PEG8000 overnight at 4°C and precipitated at 1500g for 60 minutes at 4°C. Lentivirus from each 10 cm plate of HEK293T cells was resuspended in 150 μl of pre-cooled PBS, and 50 μl was used to transfect one well in a 6-well plate containing 150,000 dECs. Gpc4 knockout efficiency was evaluated using quantitative PCR (qPCR).

Statistical Analysis.

Statistical analysis was performed using Prism 8 for Macintosh. All data are presented as mean ± SD or mean ± SEM and analyzed using a Student's t-test for two-group comparisons. P-values less than 0.05 were chosen as a threshold for statistical significance.

Abbreviations

ADAM

a disintegrin and metalloproteinase

bEC

brain endothelial cell

BSA

bovine serum albumin

CTF

C-terminal fragment

dEC

diaphragm endothelial cell

ECM

extracellular matrix

EGF

epidermal growth factor

FL-Robo4

full length Robo4

GPC4

Glypican 4

GPC1

Glypican 1

HS

heparan sulfate

HSPG

heparan sulfate proteoglycan

hRobo4-HA-FLAG

human Robo4 with C-terminal HA and FLAG tag

KEGG

Kyoto Encyclopedia of Genes and Genomes

MMP2

metalloproteinase-2

MMP9

metalloproteinase-9

MSC

mesenchymal stromal/stem cell

Robo

Roundabout

scRNA-seq

single-cell RNA sequencing

SDC4

Syndecan4

VEGF

vascular endothelial growth factor

Data Availability Statement:

The research data can be obtained from the corresponding author.