Angiogenic sprouting is regulated by endothelial cell expression of Slug

ABSTRACT

The Snail family of zinc-finger transcription factors are evolutionarily conserved proteins that control processes requiring cell movement. Specifically, they regulate epithelial-to-mesenchymal transitions (EMT) where an epithelial cell severs intercellular junctions, degrades basement membrane and becomes a migratory, mesenchymal-like cell. Interestingly, Slug expression has been observed in angiogenic endothelial cells (EC) in vivo, suggesting that angiogenic sprouting may share common attributes with EMT. Here, we demonstrate that sprouting EC in vitro express both Slug and Snail, and that siRNA-mediated knockdown of either inhibits sprouting and migration in multiple in vitro angiogenesis assays. We find that expression of MT1-MMP, but not of VE-Cadherin, is regulated by Slug and that loss of sprouting as a consequence of reduced Slug expression can be reversed by lentiviral-mediated re-expression of MT1-MMP. Activity of MMP2 and MMP9 are also affected by Slug expression, likely through MT1-MMP. Importantly, we find enhanced expression of Slug in EC in human colorectal cancer samples compared with normal colon tissue, suggesting a role for Slug in pathological angiogenesis. In summary, these data implicate Slug as an important regulator of sprouting angiogenesis, particularly in pathological settings.

INTRODUCTION

Angiogenesis is a multi-step, tightly regulated process that plays a crucial role during embryogenesis and wound healing, as well as in pathological conditions such as tumor growth (Conway et al., 2001; Folkman, 1985; Risau, 1997). During sprouting angiogenesis, endothelial cells (EC) are activated in response to angiogenic stimuli, the best characterized of which is vascular endothelial growth factor (VEGF) (Carmeliet, 2000; Conway et al., 2001). EC activation triggers a cascade of events, including degradation of the adjacent basement membrane, migration of nascent sprouts into the surrounding extracellular matrix (ECM), formation of lumens, branching, anastomosis and a return to quiescence once support cells have been recruited to the newly formed vessel (Carmeliet, 2000; Conway et al., 2001; Risau, 1997). Initiation of sprouting requires generation of at least two distinct EC phenotypes – tip cells and trunk cells. Each assumes a different morphology and performs unique functions. A tip cell leads the sprout; it is polarized along its anterior-posterior axis, rarely proliferates and is highly migratory (del Toro et al., 2010; Hellström et al., 2007; Jakobsson et al., 2010; Sainson et al., 2008). Trunk cells trail tip cells; they are proliferative, apically–basally polarized and form the vessel lumen (Ribatti and Crivellato, 2012). Gene expression profiles reveal tip cells to be highly enriched in VEGF receptor 2 (VEGFR2) (Gerhardt et al., 2003; Jakobsson et al., 2010; Ribatti and Crivellato, 2012; Sainson et al., 2008), platelet-derived growth factor B (PDGFB) (Ribatti and Crivellato, 2012; Sainson et al., 2008), neuropilin receptor 2 (NRP2) (Sainson et al., 2008), Jagged 1 (Jag1) (Johnston et al., 2009; Sainson et al., 2008), membrane type 1 matrix metalloproteinase (MT1-MMP) (van Hinsbergh and Koolwijk, 2008; Yana et al., 2007), and delta-like 4 (Dll4) (Hellström et al., 2007; Suchting et al., 2007). Expression of tip cell genes and induction of angiogenic sprouting are stimulated and regulated by pro-angiogenic cytokines including VEGF (Conway et al., 2001; Ribatti and Crivellato, 2012), tumor necrosis factor α (TNFα) (Otrock et al., 2007; Sainson et al., 2008), transforming growth factor β (TGFβ) (Otrock et al., 2007), fibroblast growth factor (FGF) (Conway et al., 2001; Otrock et al., 2007) and hepatocyte growth factor (HGF) (Sengupta et al., 2003). During pathological events such as inflammation and tumor growth, several of these growth factors induce expression of the transcription factor Slug (Snai2), and expression of this gene in tumor cells contributes to invasion and to metastasis (Barrallo-Gimeno and Nieto, 2005; Romano and Runyan, 2000; Thiery, 2002).

The Snail family of zinc-finger transcription factors are evolutionarily conserved and involved in processes that require cell movement. Expression of these genes is essential during embryonic development in events such as mesoderm, neural crest and heart cushion formation (Cobaleda et al., 2007; Niessen et al., 2008). During epithelial-to-mesenchymal transitions (EMTs), Slug acts as a transcriptional repressor by binding E-box elements in target promoters. Under certain conditions, Slug represses transcription of genes involved in formation of both adherens junctions (E-Cadherin), and tight junctions (claudins, occludins, ZO1), and promotes disassembly of desmosomes (Barrallo-Gimeno and Nieto, 2005; Cobaleda et al., 2007; Nieto, 2002). Slug also indirectly induces expression of genes that degrade ECM, such as matrix metalloproteinases (MMPs) (Barrallo-Gimeno and Nieto, 2005; Huang et al., 2009; Zhang et al., 2011). A specialized form of EMT is an endothelial-to-mesenchymal transition (EndMT). This event was first observed in developmental studies of heart formation (Armstrong and Bischoff, 2004), and studies in the heart continue to reveal mechanistic insights, including a role for Notch signaling and induction of Slug during EndMT (Niessen et al., 2008). Interestingly, Slug expression is upregulated in tumor-associated EC (Lu et al., 2007) and EndMT has been identified as an origin of cancer-associated fibroblasts (Zeisberg et al., 2007). Here, we provide evidence that Slug is expressed in angiogenic EC and is a crucial mediator of angiogenic sprouting. Interestingly, we find that Slug regulates expression of MT1-MMP, but not of VE-cadherin, and that, although it promotes EC migration, it does not lead to a loss of EC–EC junctions or to the separation of EC from their neighbors. Collectively, these studies suggest that Slug expression in EC promotes only a partial EndMT during angiogenesis.

RESULTS

Slug expression is temporally regulated during in vitro angiogenesis

In order to study the mechanisms regulating EC morphogenesis, we use an in vitro angiogenesis model (Nakatsu and Hughes, 2008) in which EC sprout into fibrin gels. The assay recapitulates several crucial steps of angiogenesis, including sprouting, lumen formation, branching and anastomosis (Fig. 1A). Using this assay we analyzed Slug expression in angiogenic EC at several time points up to 10 days, a point at which extensive, lumenized sprouts are present. Slug mRNA expression is strongly induced on day 3, when sprouts first begin to emerge from the beads, and remains highly expressed up to day 6, the time at which protein expression is highest (Fig. 1B,C). At this point, lumen formation begins to dominate the cultures, with fewer new sprouts emerging, and this correlates with a slow decline in Slug expression over the next 10 days (Fig. 1B,C). Thus, in an in vitro assay that mimics pathological and/or wound healing angiogenesis, Slug expression in EC correlates with neovessel sprouting. We also examined expression of the closely related transcription factor Snail. Like Slug, Snail was also induced during sprouting but with a slower time course, with expression peaking at day 6 (supplementary material Fig. S1A).

Fig. 1.

Angiogenic EC express Slug. (A) Representative images depicting EC morphogenesis during in vitro angiogenesis in fibrin gels. Nascent sprouts (arrowhead) are observed on day 3 and continue to proliferate, migrate, branch (arrow) and form lumens (asterisk) through days 6–10. Scale bars: 150 µm. (B) EC were harvested on the indicated days from fibrin gels and Slug mRNA levels were assessed by qRT-PCR. Results are conveyed as fold change over day 0±s.e.m. (n = 5; *P<0.01 and **P<0.0001; Student's t-test). (C) Western blot analysis of Slug protein levels in EC isolated from fibrin gels on the indicated days. (D) Formalin fixed, paraffin-embedded sections of de-identified (i) normal human colon tissue, (ii,iii) human colorectal cancer tissue and (iv) mouse colorectal cancer tissue stained for Slug (brown) and CD31 (blue), and counterstained with tri-methyl green. Red arrows depict Slug-positive EC. Scale bars: 20 µm. Two representative images of five human patient samples were analyzed.

Tumor-associated blood vessels in multiple cancers express Slug

To examine whether Slug is expressed in EC during pathologic angiogenesis in vivo we first surveyed cancer tissues stained for Slug in the Human Protein Atlas Database (www.proteinatlas.org). We observed Slug expression in vessels of gliomas (patient ID: 3120 and 3174), breast carcinomas (patient ID: 1882 and 2091), squamous cell lung carcinomas (patient ID: 1765, 1428 and 2231), liver carcinomas (patient ID: 2279, 2280 and 887) and colon adenocarcinomas (patient ID: 2060 and 2106), among others. Slug expression was not exclusive to vessels, however, as many of the tumor cells were also Slug positive. To confirm that Slug is expressed in the EC of pathological vessels, we obtained samples of normal human colon and colorectal cancer (CRC), and used double-labeling immunohistochemistry to look for Slug expression in CD31-positive EC. As shown in Fig. 1D, EC that line normal vessels only rarely express Slug. In sharp contrast, we found numerous Slug-positive EC in blood vessels in the reactive stroma, within and adjacent to colorectal tumor tissue. Some perivascular cells (possibly pericytes) were also positive in some vessels. Non-vascular cells expressing Slug, in both normal and tumor tissues are likely to be pericryptal myofibroblasts. We quantitated these findings and found fewer than 1% of vessels in normal tissues containing Slug-positive EC, whereas in two CRC tumors examined the proportions of Slug-positive vessels were 44% and 55%. We also examined vessels in an orthotopic, syngeneic (CT26) mouse colorectal cancer model, and here again we observed Slug staining in the vessels (Fig. 1D,iv). We also noted expression of Snail in the vasculature of human colorectal adenocarcinomas (supplementary material Fig. S1F). Thus, in the pathological setting of cancer, EC in angiogenic vessels express Slug and Snail, consistent with our in vitro model of pathological angiogenesis.

Loss of Slug inhibits EC sprouting

To determine whether Slug is required for vessel formation, we used small interfering RNA (siRNA) oligonucleotides to inhibit Slug expression in several in vitro angiogenesis assays. We first confirmed that targeting Slug with siRNA in EC resulted in robust inhibition of mRNA and protein expression (Fig. 2A,B). Next, we examined the effect of Slug knockdown on the ability of EC to sprout into fibrin gels, and consistently observed a dramatic loss of sprout formation (Fig. 2Ci,Cii,D). In addition, those sprouts that did form appeared to have a reduced ability to form lumens (Fig. 2Ci,Cii,E), a finding we confirmed in a second assay (Koh et al., 2008) that specifically models lumen formation (see below). Importantly, Slug knockdown was still over 60% at the mRNA level on day 5, the latest time at which phenotypes were quantified (Fig. 2C).

Fig. 2.

Loss of Slug inhibits EC sprouting in multiple in vitro angiogenesis assays. (A) EC were transfected with control or Slug siRNA and Slug mRNA levels were assessed by qRT-PCR 48 hours later. Results are shown as percent of control set to 100±s.e.m. (n = 3; ***P<0.0001; Student's t-test). (B) EC were transfected with control or Slug siRNA and harvested at 72 hours for analysis of Slug protein levels by western blot. (C) EC transfected with control or Slug siRNA were used in fibrin gel sprouting assays (i,ii), in 3D collagen I invasion assays (iii,iv) and in 3D vascularized tumor spheroids (v,vi). Graphs to the right of i and ii, and iii and iv indicate the respective percentage Slug expression on the day of quantification. Representative images from one of at least three similar experiments are shown. Scale bars: 150 µm in i,ii; 100 µm in iii–vi. (D,E) Sprouting, defined as a vessel with length greater than or equal to the diameter of the bead (150 µm), and lumen formation, defined as a vessel with a lumenal space throughout the entire vessel, were quantified on day 5 of the fibrin-gel sprouting assay. Results are expressed as mean±s.e.m. (n = 3; *P<0.05; Student's t-test). (F) Sprout invasion into collagen gels was analyzed 24 hours after seeding. Results are shown as percent of control set to 100±s.e.m. (n = 3; *P<0.05; Student's t-test). (G–I) Sprouting phenotypes from 3D vascularized tumor spheroids were quantified on day 7 (n = 3; *P<0.05, **P<0.001, ***P<0.0001; Student's t-test).

To confirm the loss of sprouting in a second assay we looked at the ability of control or Slug knockdown EC to invade collagen I gels in response to pro-angiogenic chemokines (Koh et al., 2008). Again, loss of Slug severely limited EC sprouting (Fig. 2Ciii,Civ,F). To rule out off-target effects of the siRNA, we obtained a second, independent sequence (Ambion) and repeated this assay. Once more, siRNA-mediated loss of Slug expression strongly inhibited EC sprouting (supplementary material Fig. S2D,E). Thus, Slug expression is necessary for sprouting in both fibrin and collagen gels. We also investigated the requirement for Snail expression in these assays. In both the fibrin gel sprouting assay and the collagen gel invasion assay, loss of Snail resulted in strong phenotypes, including loss of sprouting, invasion and lumen formation (supplementary material Fig. S1B-E). In these assays, the phenotypes were indistinguishable from those seen with loss of Slug expression. Clearly, the two transcription factors are not acting redundantly.

Our data showing a role for Slug during EC sprouting into fibrin gels suggest that it may be particularly important during pathological angiogenesis – indeed, it is already known from mouse knockout studies to be dispensable for developmental angiogenesis (Jiang et al., 1998). We therefore turned to an in vitro 3D vascularized tumor model to explore the role of Slug further. Co-cultures of EC transfected with either control or Slug siRNA, and colon cancer SW620 cells transduced to express GFP, were formulated into multicellular spheroids and embedded in fibrin gels distributed with fibroblasts. After 7 days, tissue constructs were fixed and tumor vessel networks were assessed. In the absence of Slug expression, we observed fewer sprouts compared with control cultures and, when EC did form sprouts, fewer than 20% of vascularized spheres had greater than five vessels, which is 70% less than the number of controls with more than five vessels (Fig. 2Cv,Cvi,G,H). The average total vessel length was also significantly decreased in the absence of EC Slug expression (Fig. 2I). Collectively, these data demonstrate that Slug is crucial during angiogenesis in the pathological setting of an in vitro 3D tumor.

Slug regulates lumen formation

Several mechanisms have been suggested for the formation of lumens during angiogenesis and the likelihood is that different mechanisms may pertain to large and small vessels, and developmental and pathological processes (Iruela-Arispe and Davis, 2009; Lubarsky and Krasnow, 2003). A widely accepted mechanism for lumen formation in small vessels involves formation of intracellular pinocytic vesicles, the fusion of these into larger intracellular vacuoles and, finally, the joining of these between neighboring EC to form a contiguous intercellular lumenal space (Iruela-Arispe and Davis, 2009). This is the process we see most often in vitro. To examine the role of Slug in EC undergoing lumen formation, we used an assay originally devised by the Davis lab in which EC are induced to form lumens in collagen gels (Koh et al., 2008). As shown in supplementary material Fig. S3, knockdown of Slug reduced both mean luminal area as well as the number of lumens per high-power field (supplementary material Fig. S3A-C). Again, we confirmed this finding using a second, independent siRNA (supplementary material Fig. S3A,D,E). We next assessed early stages of lumen formation by quantifying the number of intracellular vesicles in control and Slug knockdown EC in the presence of FITC-dextran – FITC-dextran is incorporated into the newly formed pinocytic vacuoles (Davis and Camarillo, 1996). We found no difference between control and Slug-knockdown EC, suggesting that the effects of Slug on lumen formation are downstream of the early, vesicle-forming stage, and likely at the stage of intercellular lumen formation (supplementary material Fig. S3F-H).

Inducers of Slug expression in EC

To gain insight into the induction of Slug expression, we tested several pro-angiogenic growth factors known to be present in our in vitro angiogenesis models. Some of these were added to the medium and the fibroblasts provide several more (Newman et al., 2011). We therefore tested the ability of these individually, or in combination, to induce Slug mRNA and protein in monolayer cultures (supplementary material Fig. S4). Several factors induced moderate Slug expression when tested independently, and more robust expression when used in combination. These data suggest that the expression of Slug depends on integration of multiple signals, potentially including those derived from the 3D microenvironment.

Slug misexpression promotes sprouting

To determine whether forced expression of Slug would promote sprouting and whether Slug-expressing EC sprout preferentially, EC were transduced with Slug lentivirus in which Slug was directly linked to copGFP via the self-cleaving peptide T2A, which permits visualization of Slug expression (these cells are referred to as EC^Slug/GFP). A second set of EC was transduced with copGFP lentivirus lacking Slug and these served as a control (referred to as EC^GFP). EC^Slug/GFP exhibited overexpression of Slug compared with EC^GFP and untransduced EC (EC^Control), as confirmed by western blot (Fig. 3A). We then tested these cells in the fibrin gel angiogenesis assay. Compared with EC^GFP, the EC^Slug/GFP cells showed a dramatic increase in their ability to form sprouts (Fig. 3B,D). Thus, Slug expression can drive angiogenic sprouting.

Fig. 3.

Slug misexpression in EC promotes angiogenic sprouting. (A) EC were transduced with pCDH-T2A-copGFP (EC^GFP) or pCDH-Slug-T2A-copGFP (EC^Slug/GFP) lentivirus, or were left untransduced (EC^Control), and then analyzed for Slug expression by western blot. (B) Transduced EC (EC^GFP or EC^Slug/GFP) were mixed with EC^Control and beads were then coated such that 10%, 25% or 100% of the cells were transduced and the remainder were untransduced. Fibrin-embedded beads were then examined for sprouting on day 6. Arrowheads indicate detached sprouts. (C) Confocal microscopy of sprouts from 25% transduced EC assays stained for nuclei (DAPI, blue) and F-actin (red). Arrowheads depict sprouts lacking GFP-expressing EC. Arrows indicate detached vessels; a detached vessel was defined as a sprout no longer attached to a Cytodex bead. Scale bars: 50 µm. (D) Quantification of sprouts/bead at the indicated ratios of transduced cells. (E) Quantification of detached vessels at the indicated ratios of transduced cells. (F) Quantification of sprouts that contain at least one EC^GFP or EC^Slug/GFP-positive EC. All results expressed as mean±s.e.m. (n = 3; *P<0.01, **P>0.001; GLMM).

To test whether this effect is cell-autonomous, we mixed EC^Slug/GFP with EC^Control at different ratios and again looked at sprouting in the fibrin gel angiogenesis assay, comparing this mixture with the same ratios of EC^GFP with EC^Control. As shown in Fig. 3D, at each ratio (10%, 25% and 100% EC^Slug/GFP) there was more sprouting compared with the cultures containing 10%, 25% or 100% EC^GFP. Interestingly, there was a disproportionate number of sprouts containing GFP-positive cells in 10% and 25% EC^Slug/GFP cultures compared with EC^GFP cultures of the same percentages (Fig. 3F). Indeed, almost all of the sprouts in 25% EC^Slug/GFP cultures contained Slug-positive cells and almost all of the cells within the sprout were Slug positive (Fig. 3F,C). Although the expression of Slug clearly pre-disposes EC to sprout, these data also suggest that Slug-expressing cells may suppress neighboring cells from sprouting (see Discussion). We also noted a secondary phenotype resulting from Slug expression – the detachment of sprouts from the beads, which became progressively more apparent at higher ratios of Slug-expressing cells (Fig. 3E,C).

Loss of Slug reduces MT1-MMP expression but does not affect VE-Cadherin

In epithelial cells, genes of the Snail family regulate expression of E-Cadherin, and thereby the ability of cells to release from each other (EMT). We therefore examined the expression of VE-Cadherin (the EC equivalent of E-Cadherin) in Slug knockdown EC during sprouting into fibrin gels. Interestingly, we saw no change in the mRNA expression of this gene using either of the siRNAs (Fig. 4A; supplementary material Fig. S2C). In addition, we evaluated VE-Cadherin protein localization in EC undergoing vessel formation in the absence of Slug expression and saw no differences compared with control (Fig. 4C). This is consistent with our finding that misexpression of Slug does not lead to a loss of EC junctional integrity (Fig. 3).

Fig. 4.

Loss of Slug reduces MT1-MMP expression, but does not affect VE-Cadherin. (A,B) EC were transfected with control or Slug siRNA, seeded into fibrin gels and harvested on day 5 for analysis of VE-Cadherin, or MT1-MMP expression by qRT-PCR. Results are expressed as mean±s.e.m. (n = 3; **P<0.001; Student's t-test). (C) EC were transfected with control or Slug siRNA, seeded into fibrin gels and examined by confocal microscopy on day 5 for expression of VE-Cadherin (green). Nuclei were visualized with DAPI (blue). Arrows indicate VE-Cadherin-positive adherens junctions. Scale bar: 10 µm. (D,E) EC transfected with control or Slug siRNA were seeded on top of collagen I gels and stimulated to invade for 24 hours. EC were harvested at the indicated time points and mRNA levels of Slug and MT1-MMP were determined by qRT-PCR. Results shown as fold change over time 0±s.e.m. (n = 3; **P <0.01 and ***P <0.001; ANOVA). (F) EC were transduced with the indicated lentiviral vectors and examined for expression of MT1-MMP by western blot. (G) Transduced EC were subsequently transfected with control or Slug siRNA, seeded onto collagen gels and stimulated to invade for 24 hours. Gels were fixed and stained, and invading cells were quantified (n = 3; **P <0.01; ANOVA). (H) Representative images from G captured at 24 hours. Arrows indicate invading cells. Scale bars: 100 µm.

An early, crucial stage of angiogenesis is the establishment of a tip cell that leads migration of the nascent sprout (Ribatti and Crivellato, 2012). In light of the sprouting defect observed in Slug knockdown cells, we hypothesized that Slug might regulate EMT-related genes and/or known tip cell genes. We therefore examined mRNA levels for the following genes in the presence or absence of Slug in the fibrin gel angiogenesis assay: VEGFR2, PDGFB, NRP2, Jag1, Dll4, integrin αv, integrin β3, vimentin, N-Cadherin and MT1-MMP. Of these, only levels of MT1-MMP (Fig. 4B) and Jag1 (not shown) were consistently decreased in Slug knockdown EC. We chose to pursue further studies with MT1-MMP and confirmed regulation by Slug using a second independent Slug siRNA (supplementary material Fig. S2B). MT1-MMP, a membrane-tethered MMP, is expressed in tip cells during angiogenesis (van Hinsbergh and Koolwijk, 2008; Yana et al., 2007) and is required to facilitate migration through both fibrin and collagen matrices (Genís et al., 2006; Hiraoka et al., 1998; Itoh and Seiki, 2006). We therefore examined Slug regulation of MT1-MMP in the collagen gel invasion assay. Slug was strongly induced at 24 hours and this induction was completely blocked by Slug siRNA (Fig. 4D). In the same cells, MT1-MMP mRNA was also strongly induced at 24 hours and this induction was blocked 50% by loss of Slug (Fig. 4E). Flow cytometry analysis confirmed upregulated surface expression of MT1-MMP protein and a concomitant decrease in cells treated with siRNA (data not shown). These data were also confirmed with a second independent siRNA to Slug (supplementary material Fig. S2F,G). As further confirmation that the decreased sprouting seen with Slug knockdown cells is due (at least in part) to loss of MT1-MMP expression, we performed a rescue experiment. EC were transduced with lentivirus expressing either GFP or MT1-MMP, and then transfected with control or Slug siRNA and tested for their ability to invade collagen gels. Expression of transduced MT1-MMP was confirmed by western blot (Fig. 4F). Knockdown of Slug reduced invasion by over 50% and this was not affected by expression of GFP (Fig. 4G,H). However, expression of MT1-MMP completely rescued the loss of sprouting due to Slug knockdown, confirming that MT1-MMP is a crucial downstream target of Slug during angiogenic sprouting.

Slug indirectly regulates activity of MMP2 and MMP9

During sprouting angiogenesis, the enzymatic activity of several MMPs is required to degrade and remodel the surrounding 3D ECM (Sang, 1998). MMP2 is a secreted protease that is inactive in its native form; however, in the presence of TIMP2 (tissue inhibitor of metalloproteinases 2), it is cleaved and activated by surface-expressed MT1-MMP (Visse and Nagase, 2003). Interestingly, several studies have reported that expression of Slug correlates with an increase in activity of several MMPs (Barrallo-Gimeno and Nieto, 2005; Huang et al., 2009; Zhang et al., 2011). We therefore reasoned that the decrease of MT1-MMP expression observed in the absence of Slug might result in decreased enzymatic activity of MMP2 and perhaps other MMPs such as MMP9. Indeed, this was the case. Using gelatin zymography we found that knockdown of Slug in EC reduced both MMP2 and MMP9 activity by 50% when compared with control (Fig. 5A,B,D). This result was confirmed using a second independent siRNA targeting Slug (supplementary material Fig. S2H–J). Interestingly, we saw no decrease in mRNA levels of either MMP2 or MMP9 at 24 hours, although we did see strong induction of MMP9 in this assay (Fig. 5C,E). These data are consistent with Slug regulating the activity of MMP2 through MT1-MMP; however, the mechanisms underlying the effects of Slug knockdown on MMP9 activity are as yet unclear, as MMP9 does not require activation by MT1-MMP. Interestingly, TIMP1, which blocks MMP2 and MMP9, but not MT1-MMP, blocked sprouting (data not shown), suggesting that MMP2 and MMP9 may have a role in this process. In aggregate, our data show that Slug regulates EC protease activity during angiogenic sprouting.

Fig. 5.

MMP2 and MMP9 activity is indirectly regulated by Slug. (A) EC were transfected with control or Slug siRNA, seeded on top of collagen gels, and stimulated to invade. After 24 hours, culture medium was collected and MMP activity assessed by gelatin zymography. (B,C) Quantitative analysis of MMP2 activity and mRNA expression after Slug knockdown. Results of the zymography are shown as percent of control set to 100±s.e.m. (n = 3; *P<0.05; Student's t-test). Results of the qRT-PCR analysis are shown as fold change over time±s.e.m. (n = 3; ***P <0.001; ANOVA). (D,E) Quantitative analysis of MMP9 activity and mRNA expression after Slug knockdown. Details as for B,C.

DISCUSSION

In recent years there has been a dramatic increase in our understanding of the growth factors and receptors that drive angiogenesis, and a growing appreciation of the signaling pathways downstream of these receptors. Our understanding of the transcription factors that form the link between these signals and new gene expression is, however, much less complete. Here, we define a role for the transcription factor Slug in sprouting angiogenesis. Slug expression drives sprouting through the induction of MT1-MMP and the regulation of MMP2 activity. In the absence of Slug, EC sprouting is disrupted and this can be overcome by re-expression of MT1-MMP. Importantly, we also find Slug expression in tumor-associated vessels in multiple cancers. Our data therefore suggest that Slug potentially regulates pathological angiogenesis in settings such as cancer. We also find a role for Snail in sprouting angiogenesis and are currently pursuing a deeper analysis of its mechanisms of action.

Slug is perhaps best characterized as a member of a family of transcription factors, including Snail, Twist, ZEB1 and ZEB2, that drive EMTs (Potenta et al., 2008). EndMT has been previously described during cardiac cushion morphogenesis (Niessen et al., 2008), and several studies have suggested that EndMT provides a source for cancer-associated myofibroblast cells (Potenta et al., 2008; Zeisberg et al., 2008). We therefore wondered whether Slug expression during angiogenesis was driving a partial EndMT, particularly affecting tip cells. Slug certainly drives migration and invasion, through MT1-MMP expression; however, we saw no change in VE-Cadherin expression, nor did we see regulation of several genes, other than MT1-MMP and Jag1, which are known to be upregulated in tip cells (van Hinsbergh and Koolwijk, 2008; Yana et al., 2007). Somewhat surprisingly, our hypothesis that Slug-expressing cells would localize preferentially to a tip location was not borne out. Instead, Slug-expressing cells were found throughout the sprout, suggesting that Slug expression in EC may be a more general marker for an activated, angiogenic phenotype rather than a specific marker for EndMT-like processes occurring in tip cells. Strikingly, when EC were forced to express Slug by lentiviral-mediated transduction, and these were mixed 1:3 with untransduced EC, the vast majority of cells locating to sprouts expressed Slug. In sharp contrast, when GFP-expressing EC were mixed 1:3 with untransduced EC, GFP-expressing cells were found both in and out of sprouts. The strong implication is that Slug-expressing cells not only preferentially localize to sprouts, but also actively suppress non-Slug-expressing cells from sprouting. Without further experimentation, we cannot be sure of the mechanism underlying this finding; however, data from our lab (Sainson et al., 2005) and others (Hellström et al., 2007; Suchting et al., 2007) may implicate Notch signaling. Notch ligand expression, especially Dll4, suppresses neighboring cells from sprouting both in vitro and in vivo (Hellström et al., 2007; Sainson et al., 2005; Suchting et al., 2007); however, our preliminary data did not show a loss of Dll4 expression in Slug-knockdown cells, although Jag-1 was suppressed. Further work will be required to determine the interactions between Slug and the Notch pathway in this process.

MMPs, including MT1-MMP, are crucial mediators of angiogenesis that are responsible for matrix degradation (Carmeliet, 2000; Carmeliet, 2003; Genís et al., 2006; Sang, 1998; Visse and Nagase, 2003) as well as release of matrix-bound pro-angiogenic factors, including bFGF and VEGF (Ziyad and Iruela-Arispe, 2011). MT1-MMP directly degrades both fibrin and collagen (Genís et al., 2006; Hiraoka et al., 1998; Itoh and Seiki, 2006) and acts in concert with TIMP2 to cleave pro-MMP2 into its active form (Visse and Nagase, 2003). Several studies have also shown that MT1-MMP is required for both sprouting (Fisher et al., 2009) and lumen formation in vitro (Stratman et al., 2009) – a finding we suggest is linked to expression of Slug (Fig. 4). These data are consistent with several previous reports that Slug regulates MMP expression and activity in cancer cells. For example, Slug regulates both MT1-MMP and MMP9 in pancreatic cancer (Shields et al., 2012; Zhang et al., 2011), and has also been shown to regulate MT4-MMP (Huang et al., 2009). Moreover, we find that Slug is upregulated in blood vessels adjacent to invasive tumors, but is largely absent in quiescent vessels (Fig. 1D). Finally, a previous report found Slug in invasive ovarian tumor-associated EC (Lu et al., 2007). In aggregate, these data support a role for Slug-regulated MMP expression in both tumor cells, and their associated angiogenic vasculature.

Interestingly, Slug knockout mice are viable with no major phenotype (Jiang et al., 1998), although loss of the closely related gene Snail causes early embryonic lethality due to problems with gastrulation (Carver et al., 2001). It is therefore possible that Snail compensates for the loss of Slug during early development, masking a potential role for Slug in this process. In our in vitro studies, by contrast, we find that Snail cannot compensate for Slug in the pathological setting of invasion into fibrin gels. We find that Snail is expressed under these conditions, although along a different time course than Slug, and that its expression is also required for proper sprouting (supplementary material Fig. S1). It is likely, therefore, that under these conditions Slug and Snail regulate a separate but potentially overlapping suite of genes. We are currently investigating this possibility. Importantly, there are a number of precedents for genes being crucial for pathological angiogenesis but dispensable for developmental angiogenesis, including tetraspanin CD151 (Takeda et al., 2007), aminopeptidase N (CD13) (Rangel et al., 2007) and TNFRI (CD120) (Kociok et al., 2006). In summary, our data suggest a crucial role for Slug expression in angiogenic EC upstream of MT1-MMP expression, and suggest that Slug may be a useful target for regulating angiogenic EC in multiple human tumor types.

MATERIALS AND METHODS

Cell culture and small-interfering RNA transfection

Primary human umbilical vein endothelial cells (HUVECs) were isolated from umbilical cords obtained from local hospitals under University of California Irvine Institutional Review Board approval. HUVECs were routinely cultured in 1×M199 (Life Technologies) supplemented with 10% fetal bovine serum (FBS) and endothelial cell growth supplement (ECGS; BD Biosciences) at 37°C and 5% CO₂. Normal human lung fibroblasts (NHLF) were purchased from Lonza, routinely grown in 1×M199 supplemented with 10% FBS at 37°C and 5% CO₂. HUVECs at 80% confluency were transfected with 50 nM siRNA purchased from Invitrogen or with 16 nM siRNA purchased from Ambion using Lipofectamine 2000 in Opti-MEM (Invitrogen) for 4 hours with transfection mixture and recovered in endothelial growth media 2 (EGM-2; Lonza) overnight. The non-targeting stealth RNAi-negative control high-GC duplex #2 (Invitrogen) or the silencer select negative control #1 siRNA (Ambion) was used as a control for sequence-independent effects of siRNA delivery. Transfection efficiencies were determined by qRT-PCR and western blot analysis. siRNA oligonucleotide sequences listed in supplementary material Table S1.

Lentiviral constructs and transductions

Full-length human HA-tagged MT1-MMP or full-length human Slug was cloned into the lentiviral vector pCDH (CD521A-1; System Biosciences). Lentivirus was made by transfection of pCDH constructs along with the packaging lines psPAX2 and pCMV-VSV-G into 293T cells using Lipofectamine 2000 in Opti-MEM, according to the manufacturer's protocol. Viral supernatants were collected and precipitated using 50% polyethylene glycol (PEG) and passage 0 HUVECs were transduced with virus using polybrene (8 µg/ml; Santa Cruz Biotechnology).

In vitro fibrin gel angiogenesis assay

Fibrin gel angiogenesis assays were performed as previously described (Nakatsu and Hughes, 2008). Briefly, HUVECs were coated onto Cytodex 3 microcarrier beads (Amersham) at a concentration of 150 cells/bead for 4 hours and allowed to adhere overnight. HUVEC-coated beads were then resuspended in a 2.5 mg/ml fibrinogen solution (MP Biomedicals) at a concentration of 250 beads/ml. Gels were formed by adding 500 µl of the fibrinogen/bead suspension to each well of a 24-well plate containing 0.5 U of thrombin (Sigma-Aldrich). Once gels clotted, 1 ml of EMG2 containing 20,000–50,000 NHLF was added to each well. Assays were quantified between days 5 and 6 by live-culture imaging using bright-field microscopy. Thirty beads per condition were quantified per experiment.

For RNA and protein isolation, HUVECs were isolated from the fibrin gels by removing fibroblasts with 3 mg/ml trypsin (Sigma-Aldrich) under gentle agitation. Residual fibroblasts were removed by washing the gels using 1×Hank's Balanced Salt Solution (HBSS; Cellgro). Fibrin gels were digested with 4 mg/ml trypsin and gels were dislodged from the wells of the 24-well plate. The entire contents of each well was transferred to a conical tube and placed under rotation at 37°C to achieve complete digestion. When harvesting cells for studies of Slug protein expression, the cells were pre-treated 10 µM MG-132 (Calbiochem) for 1 hour to retard proteasome-mediated degradation.

In vitro fibrin gel sandwich assay

HUVECs were transfected with control or Slug siRNA (Ambion) as described above. 500 µl of 2.5 mg/ml fibrinogen was mixed with 0.5 U thrombin (Sigma-Aldrich) in four wells of a 12-well plate and allowed to clot at 37°C. HUVECs were seeded on top of each gel at a concentration 2.5×10⁵ cells/ml in EGM2 and allowed to adhere at 37°C for 3 hours. EGM2 was aspirated and 500 µl of 2.5 mg/ml fibrinogen pre mixed with 0.5 U thrombin was added to create a fibrin-gel sandwich, gels were allowed to clot at 37°C and 1 ml of EMG2 containing 40,000 NHLF was added to each well. HUVECs were allowed to undergo morphogenesis for 3 days. HUVECs were isolated from the fibrin-gel sandwich by removing fibroblasts using 3 mg/ml trypsin (Sigma-Aldrich). Gels were washed with 1×HBSS to remove residual NHLF. Fibrin gels were digested with 4 mg/ml trypsin and HUVEC isolation was monitored under a microscope. The contents of four wells/condition were combined, and the digested product was centrifuged at 335 g. The resulting pellet containing HUVECs was resuspended in TRIzol for qRT-PCR analysis as described below.

Human tissue and immunohistochemistry

Formalin-fixed, paraffin-embedded sections of de-identified human colorectal cancer slides were obtained from the Experimental Tissue Resource in accordance with UCI Biorepository procedures. Deparaffinized human colorectal cancer tissue sections underwent citrate-based antigen retrieval, and blocking with 5% goat serum. Sections were incubated in rabbit anti-Slug (1∶200; Cell Signaling, 9585) and biotinylated goat anti-rabbit antibody, followed by development with a peroxidase-based Vectastain ABC kit. The stained slides were re-blocked with goat serum, and incubated in mouse anti-CD31 (1∶100; Dako, IR610) followed by ImmPRESS alkaline phosphatase-conjugated anti-mouse IgG (Vector Laboratories, MP-54020) and developed with Vector-Blue substrate. Counterstaining was performed with tri-methyl green.

In vitro invasion assays and in vitro lumenogenesis assays in 3D collagen matrices

Assays were performed as previous described (Koh et al., 2008). For invasion assays, collagen gels were made with 30 µl of rat-tail collagen I (3.75 mg/ml) supplemented with 200 ng/ml SDF1α (PeproTech) and 1 µM S1P (Biomol). Gels were added to each well of a 4.5 mm diameter 96-microwell plate (Corning) and incubated at 37°C until polymerized. HUVECs were then suspended in serum-free culture media of 1×M199 containing 1×ITS+3 (Sigma-Aldrich), 40 ng/ml VEGF (R&D Systems), 40 ng/ml FGF-2 (R&D Systems), 50 µg/ml ascorbic acid (Fisher Scientific) and 50 ng/ml PMA (Calbiochem) at a concentration of 1×10⁵ cells/ml and 100 µl of cell suspension was added to each well. HUVECs were allowed to invade for 24 hours at 37°C and 5% CO₂. Cultures were fixed in 3% glutaraldehyde for 30 minutes, washed with sterile water and stained using 1% Toluidine Blue in 30% methanol for 1 hour. Assays were destained with water and bright-field images (three gels/condition) were taken a few micrometers below the monolayer in order to quantify the number of invading HUVECs. To isolate HUVECs, 65 gels/condition were digested in 5 mg/ml collagenase (Worthington Biochemical) dissolved in dPBS (Gibco) and the cellular pellet was resuspended in 1 ml of Trizol (Invitrogen).

Alternatively, HUVECs used in lumenogenesis assays were suspended in 30 µl of rat-tail collagen I (3.75 mg/ml) gels at a final concentration of 6×10⁵ cells/ml, added to each well of a 4.5 mm diameter microwell plate (Corning) and incubated at 37°C until polymerized. 100 µl of serum-free culture media described above (omitting cells) was added to each well. HUVECs were allowed to undergo morphogenesis for 24–48 hours at 37°C and 5% CO₂, and were fixed, stained and destained as described for invasion assays. Four bright-field images were captured per well (three wells/condition) and intercellular lumens were manually traced using NIH ImageJ, converted from pixels to square micrometers and averaged for each condition. An EC lumen was defined as a multicellular lumenal space in addition to intracellular lumen compartments.

Early stage lumen formation was assessed using the assay described above with the addition of 5 mg/ml FITC-dextran (Molecular Probes) to the culture media. After 4 hours of morphogenesis, gels were digested with 5 mg/ml collagenase type I for 10 minutes at 37°C and the contents of three microwells were added to 500 µl Phenol Red-free 1×M199. Cells were seeded onto glass coverslips coated with 50 µg/ml type I collagen and allowed to adhere for 10 minutes at 37°C. Coverslips were mounted and the percentage of cells containing fluorescently labeled intracellular lumens/high-power field (HPF) and the number of fluorescently labeled intracellular lumens/cell were quantified for each condition (n = 400 cells).

Methylcellulose production and 3D vascularized colon cancer spheroid assay

Methylcellulose was generated by autoclaving 1.2 g of powder in a 250 ml beaker at 120°C for 20 minutes. Under sterile conditions, 50 ml of preheated (60°C) endothelial basal media (EBM; Lonza) was added to the autoclaved methylcellulose and dissolved by stirring at 60°C for 20 minutes. Once dissolved, an additional 50 ml of EBM was added for a final volume of 100 ml. The methylcellulose solution was covered with foil and mixed for 2 hours at 4°C. The solution was centrifuged at 2000 g for 2 hours. Supernatant was removed, ∼90% of the volume, and the resulting methylcellulose was stored at 4°C. The PDMS-retaining rings used to generate tissues had a diameter of 8 mm and a height of 0.8 mm. For quantification, a total of five tissues/condition were quantified for each independent experiment (n = 3) and one tissue contained approximately eight vascularized spheroids.

HUVECs and SW620 cells were seeded into EGM-2 containing 15% methylcellulose at 7.5×10⁴ cells/ml and 2.5×10⁴ cells/ml, respectively. Cellular suspensions were aliquoted (150 µl/well) into a 96-well U-bottom plate (Greiner Bio-one, CellStar) and allowed the formation of spheres overnight. Spheroids were resuspended in fibrinogen (2.5 mg/ml; Sigma) containing NHLF at 1×10⁶ cells/ml. 50 µl of spheroid/cell suspension was added onto a 12 mm circular glass coverslip with an affixed polydimethylsiloxane (PDMS)-retaining ring and mixed with 5×10⁻³ U thrombin (Sigma-Aldrich). Tissues were fed with EGM2 and maintained at 37°C and 5% CO₂. On day 7, tissues were fixed and immunofluorescent staining was performed.

Growth factor treatments

HUVECs were cultured as previously described. At 100% confluency, HUVECs were serum starved in 1×M199 containing 2% FBS overnight. The following day, HUVECs were treated with the following growth factors in 1×M199 containing 2% FBS for 24 hours, plus SDF1α (200 ng/ml; PeproTech), VEGF (40 ng/ml; PeproTech), bFGF (40 ng/ml; PeproTech), S1P (1 µM; Biomol), PMA (50 ng/ml; Calbiochem), TGFβ1 (5 ng/ml; PeproTech), TNFα (10 ng/ml; PeproTech), TGFα (50 ng/ml; PeproTech), HGF (100 ng/ml; PeproTech), ANG1 (250 ng/ml; PeproTech) and angiogenin (250 ng/ml; R&D Systems). HUVECs were then harvested and expression levels were detected via western blot or qRT-PCR analysis as described below.

Quantitative real-time PCR

Total RNA was isolated from HUVEC using Trizol reagent (Invitrogen) according to the manufacturer's protocol. Isolated RNA was treated with RQ1 DNase (Promega) for 1 hour. Total RNA was used for cDNA synthesis using an iScript cDNA Synthesis Kit (BioRad). A BioRad iCycler and HotStartTaq DNA Polymerase (Qiagen) was used to perform qRT-PCR with SYBR Green (Molecular Probes) as the readout. Average C_T values were normalized to GAPDH expression levels and all samples were measured in triplicate. Primers were synthesized by Integrated DNA Technologies and sequences can be found in supplementary material Table S1.

Western blot

HUVECs isolated from the fibrin gel angiogenesis assay as described above were lysed on ice in RIPA buffer [50 mM Tris-Cl (pH 7.4), 1% NP-40, 0.5% sodium deoxycholate, 150 mM NaCl] supplemented with 1 mM phenylmethylsulfonyl fluoride (PMSF), 1 mM EDTA, 5 mM DTT and 1×protease inhibitor cocktail. Lysates were sonicated twice at 10 W for 15 seconds, and cellular debris and beads were cleared by centrifugation: 16,000 g for 10 minutes at 4°C. Alternatively, protein lysates from monolayer HUVECs were extracted directly from culture dishes by adding supplemented RIPA buffer to culture dishes placed on ice for 10 minutes. Dishes were scraped and the cellular contents were added to a microfuge tube and allowed to lyse for an additional 10 minutes on ice. Lysates were sonicated and spun at 16,000 g for 10 minutes at 4°C. Protein concentrations were determined using bicinchoninic acid assay (Sigma-Aldrich) according to manufacturer's instructions. Samples were mixed 3∶1 with Laemmli 4×sample buffer (BioRad), boiled for 5 minutes at 95°C and equal amounts of protein (40–100 µg) were loaded and electrophoresed in 4–20% Mini-PROTEAN TGX polyacrylamide gels (BioRad) under denaturing and reducing conditions. Proteins were transferred to a polyvinylidene fluoride membrane (Millipore). Membranes were blocked for 2 hours in TBS/0.1% Tween 20 (0.1% TBST)-containing 5% non-fat dry milk. Membranes were then incubated overnight at 4°C in primary antibodies – primary rabbit monoclonal anti-Slug (1∶750; Cell Signaling, 9585) or primary rabbit monoclonal anti-MMP14 (1∶2000; Epitomics, 2010-1) were used. Anti-Slug antibody was diluted in 0.1% TBST containing 5% BSA and anti-MMP14 was diluted in 0.1% TBST containing 2% milk. The following day, membranes were washed with TBS/0.2% Tween 20 (0.2% TBST) before secondary antibody was added. HRP-conjugated goat anti-rabbit secondary antibody (1∶5000; Abcam) was diluted in 0.1% TBST containing 5% BSA and added to the blot for 2 hours at room temperature. Protein expression was detected using Amersham ECL Prime Western Blotting Detection Reagent (GE Healthcare) and membranes were imaged using a Nikon AF 50 mm f/1.4D camera (Nikon). To ensure equal loading, membranes were stripped using restore stripping buffer (Thermo Scientific), blocked for 1 hour in 0.1% TBST containing 5% BSA and re-probed for 2 hours with HRP-conjugated GAPDH (1∶5000; Abcam, ab9482) antibody.

Gelatin zymography

Supernatant/culture media from 3D collagen I invasion assays (see above) were collected from 20 wells/condition, combined and cellular debris was removed by centrifugation. Collected media was concentrated using ultra centrifugal devices with a 3000 nominal molecular weight limit (Amicon) according to the manufacturer's protocol. 25–100 µg of protein was resolved on 10% polyacrylamide gels containing 1% (w/v) gelatin (BioRad). Zymogram reagents were purchased from BioRad and the manufacturer's protocol was followed. Briefly, gels were washed four times for 15 minutes in 25 ml of 1×Renaturation Buffer (BioRad), incubated in Development Buffer (BioRad) for 20 minutes at 37°C, stained with 0.1% amido black (Sigma-Aldrich) in 30% methanol (v/v) and 10% acetic acid (v/v), and then destained in 30% methanol (v/v) and 10% acetic acid (v/v). Zymograms were imaged using a Gel Doc 2000 equipped with an 8-bit CCD camera and Quantity One software (BioRad) and densitometry quantification was completed using NIH ImageJ.

Immunofluorescence

Fibrin gel angiogenesis assays used for immunofluorescence were performed in Lab-Tek II four-well chambered borosilicate coverglass system (No. 1.0; Thermo Fisher Scientific). Prior to staining, the NHLF monolayer was removed as described above. Assays were fixed in 4% PFA for 15 minutes and extensively washed in 1×PBS containing 0.3 M glycine to remove fixative. Assays were permeabilized and blocked for 2 hours at room temperature using 1×PBS supplemented with 0.3 M glycine, 5% BSA, 5% goat serum, 0.2% sodium azide and 0.3% Triton X-100. Assays were treated with primary monoclonal rabbit anti-VE-Cadherin antibody (1∶75; Enzo, ALX-210-232) diluted in blocking/permeabilization solution and incubated at 4°C overnight. The following day cultures were treated with secondary goat anti-rabbit 488-conjugated antibody (1∶200; Invitrogen, A11008) overnight at 4°C. Cultures were extensively washed in 1×PBS. Nuclei were stained with 1 µg/ml DAPI (Sigma-Aldrich) and F-actin was stained with 0.2 µM Texas Red-X phalloidin (Invitrogen). All steps were completed under gentle agitation.

Vascularized 3D colon cancer spheroids were fixed in 10% formalin (Fisher Scientific). Tissues were permeabilized for 30 minutes at room temperature using 1×PBS supplemented with 0.5% Tween-20. Non-specific binding was blocked with 1×PBS containing 2% BSA and 0.1% Tween-20. Tissues were incubated overnight at 4°C using a mouse anti-CD31 antibody (1∶100; Dako, IR610) diluted in blocking buffer followed by a goat anti-mouse 568-conjugated (1∶500; Invitrogen, A11004) secondary antibody also diluted in blocking buffer. Tissues were extensively washed with 1×PBS containing 0.3 M glycine to remove background. All steps were completed under gentle agitation.

Microscopy

An inverted microscope (IX70; Olympus) was used for all conventional bright-field images. Images were captured using a SPOT Idea 3.0 megapixel color mosaic camera and Spot acquisition software (Sport Imagining Solutions). For confocal microscopy, a Nikon Eclipse Ti inverted confocal microscope (Nikon) equipped with a CoolSNAP ES2 CCD camera (Photometrics) and EZ-C1 acquisition software (version 3.91; Nikon) was used. Confocal images were 12-bit (containing 1024×1024 pixels) and four scans were averaged per pixel. Adjustments to image brightness and/or contrast were performed using Adobe Photoshop software – images between difference conditions were treated identically.

Statistical analysis

Researchers were blinded to experimental conditions prior to performing quantifications. All experiments were repeated at least three times. Data are reported as mean±s.e.m. Student's t-test was used to analyze differences between experimental groups of equal variance when only two groups were being compared. For comparisons involving three or more conditions and/or two independent time points, a two-way analysis of variance (ANOVA) with multiple comparisons was performed and the TukeyHSD probability value was used to determine significance. For analysis of Slug overexpression data (Fig. 3), a generalized linear mixed model (GLMM) was performed using SPSS software and an LSD pairwise contrast method was used to determine significance.