VANGL2 regulates membrane trafficking of MMP14 to control cell polarity and migration

It has been a decade since homologues of fly planar cell polarity (PCP) genes were shown to be required for convergence and extension movements during frog and fish gastrulation (Djiane et al., 2000; Wallingford et al., 2000; Jessen et al., 2002). Underlying these movements are polarized cell behaviors that contribute to mediolateral narrowing and anterior–posterior elongation of the developing embryo. Loss of PCP gene function in vivo disrupts polarity, as indicated by changes in cell elongation, orientation and protrusive activity (Wallingford et al., 2000; Topczewski et al., 2001; Jessen et al., 2002). As a result of altered cell polarity, zebrafish mutants including trilobite/vang-like 2 (vangl2) exhibit a distinct convergence and extension phenotype characterized by shortened and broadened body axes (Sepich et al., 2000; Jessen et al., 2002). Although it is thought that Wnt-mediated PCP signaling impacts the actin cytoskeleton to regulate polarity and protrusion dynamics in migrating cells, few downstream effector proteins have been identified. Previously, we showed that zebrafish membrane type-1 matrix metalloproteinase (Mmp14) is also required for cell polarity and convergence and extension movements exhibiting a genetic interaction with the Wnt co-receptor knypek/glypican4 (Coyle et al., 2008). The temporal requirement for Mmp14 function during late gastrulation coincides with both that of Vangl2 (Sepich et al., 2000) and the appearance of a fibrillar extracellular matrix (ECM) meshwork (Latimer and Jessen, 2010), suggesting a functional relationship between cell polarity and proteolysis. However, despite previous connections between PCP and ECM assembly (Goto et al., 2005; Dzamba et al., 2009), it is unknown whether PCP signaling proteins themselves regulate degradation and remodeling of the ECM. Our previous work used HT-1080 cells to probe the relationship between human VANGL2, MMP14 and cell–ECM interactions. Knockdown of VANGL2 using siRNA increased secreted levels of active MMP2 and promoted invasion through an ECM substrate (Cantrell and Jessen, 2010). Because MMP2 is activated by MMP14 at the cell surface (Sato et al., 1994), we hypothesized that Vangl2 might influence PCP in migrating cells by directly regulating Mmp14 activity. In this report, we identify VANGL2 as a regulator of MMP14 endocytosis and demonstrate that this membrane-tethered metalloproteinase acts downstream of Vangl2 to control zebrafish convergence and extension cell movements.

Both endocytosis and recycling of MMP14 provide possible mechanisms to control the level of proteolytic activity in polarized cells (Jiang et al., 2001; Remacle et al., 2003; Poincloux et al., 2009). To determine whether VANGL2 regulates MMP14 trafficking, we performed an in vitro biotinylation assay using a cleavable form of biotin to quantify cell-surface, internalized and recycled MMP14 levels in cells transfected with siRNA against VANGL2. As shown in Fig. 1A, VANGL2-knockdown cells had increased levels of cell-surface MMP14 but not total MMP14 protein (supplementary material Fig. S1). After we stimulated endocytosis, VANGL2-knockdown cells had reduced levels of internalized MMP14 (Fig. 1B,C). To test whether MMP14 recycling is impaired, after endocytosis cells were treated with MESNA and incubated at 37°C for an additional hour followed by a second MESNA treatment. Here, western blot of biotin-labeled proteins gives a quantitative measurement of the remaining non-recycled pool of MMP14. Our data show that a similar ratio of intracellular MMP14 was recycled to the cell surface in both control and VANGL2-knockdown cells (Fig. 1B,C). To confirm these findings, we performed an antibody-uptake assay to visualize MMP14-positive endocytic vesicles. Cells were labeled with antibody against MMP14 and either fixed or shifted to 37°C to permit endocytosis. VANGL2-knockdown cells were confirmed to have a decreased ability to internalize cell-surface MMP14 (Fig. 1D). To determine whether loss of VANGL2 globally disrupts endocytosis, we performed endocytosis assays using fluorescently labeled transferrin (to ascertain clathrin-mediated endocytosis) or EEA1 antibody (to detect early endosomes). In neither case did we observe a difference between control and VANGL2-knockdown cells (Fig. 1E and supplementary material Fig. S2). These data indicate that VANGL2 specifically regulates endocytosis of MMP14; however, it is possible that VANGL2 influences the trafficking of additional membrane proteins in polarized cells. For example, studies from fly and fish suggest that PCP proteins regulate endocytosis of cadherins to modulate cell adhesion (Classen et al., 2005; Ulrich et al., 2005).

MMP14 expression in vitro reflects its regulation by endocytosis and recycling in that little protein is present on the plasma membrane with the majority localized to intracellular vesicles (Steffen et al., 2008). Recruitment of fly Van Gogh to the plasma membrane is thought to be a crucial event for PCP signaling in epithelial cells (Strutt and Strutt, 2009). However, recent in vitro and in vivo studies have identified VANGL2 expression within vesicular compartments (Merte et al., 2010; Cha et al., 2011). VANGL2 and MMP14 colocalization in HT-1080 cells was detected at the plasma membrane (Fig. 2A and supplementary material Fig. S3A). VANGL2 was also expressed in vesicle populations that labeled positive for MMP14, EEA1, RAB11 (recycling endosomes), LAMP1 (late endosomes) and VAMP1 (post-Golgi vesicles) (Fig. 2B–F and supplementary material Fig. S3A,B). Here, co-labeling of VANGL2 with MMP14 and EEA1 or RAB11 represented the largest fraction of triple positive vesicles compared with LAMP1 or VAMP1. Neither VANGL2 nor MMP14 expression overlapped with the cis-Golgi marker GM130 (supplementary material Fig. S3C). Taken together, our data demonstrate that the cellular distribution of VANGL2 is dynamic and HT-1080 cells do not constitutively express this protein at the plasma membrane, but rather recruit VANGL2 (and MMP14) to specific membrane domains. These results are consistent with the notion that VANGL2 could influence MMP14 endocytosis by regulating protein interactions and/or actin cytoskeletal dynamics at either the plasma membrane or within endosomal vesicles.

Although regulation of MMP14 exocytosis facilitates targeting of protease activity to sites of focal degradation (Bravo-Cordero et al., 2007; Steffen et al., 2008), the concept of endocytic removal of plasma membrane MMP14 as a means of restricting proteolytic degradation is less well characterized. In vitro, MMP14-mediated cleavage of ECM at sites of cell–matrix adhesion was shown to influence cell polarity and migration (Takino et al., 2007), suggesting that regulation of cell-surface MMP14 activity is coordinated with integrin–ECM adhesion and signaling. This notion is supported by data showing that increased focal adhesion kinase (FAK) activity, including Y397 autophosphorylation suppresses MMP14 endocytosis through effects on endophilin A2 phosphorylation and inhibition of dynamin (Wu et al., 2005). Notably, injection of dominant-negative dynamin into frog embryos was shown to inhibit convergent extension gastrulation cell movements (Jarrett et al., 2002). Based on these data, we examined the effects of VANGL2 knockdown on focal adhesion formation and FAK phosphorylation. Using paxillin–GFP to label focal adhesions (Turner, 1998) in HT-1080 cells, we found that VANGL2-knockdown cells had fewer focal adhesions than controls, and the remaining adhesions were largely restricted to the cell periphery (Fig. 3A). Phosphorylation of FAK was analyzed by immunolabeling polarized cells in a wounded monolayer and by western blot. Both assays demonstrated increases in Y397-phosphorylated FAK in VANGL2-knockdown cells (Fig. 3B), suggesting that VANGL2 function antagonizes FAK activation. In addition to affecting MMP14 endocytosis, phosphorylation of FAK at Y397 induces disassembly of focal adhesions (Hamadi et al., 2005); thus our data also suggest that VANGL2 regulates focal adhesion turnover. To test whether the ability of VANGL2 to control MMP14 endocytosis is directly mediated by FAK activation, we used an inhibitor of FAK Y397 phosphorylation (Fig. 3C) in conjunction with siRNA knockdown. Our data show that treatment of VANGL2-knockdown HT-1080 cells with FAK inhibitor rescues both the VANGL2-dependent increase in cell-surface MMP14 levels and defective MMP14 endocytosis (Fig. 3D,E). FAK is a principle kinase implicated in integrin signalling, and phosphorylation of Y397 is a key determinant of focal adhesion stability and turnover (Hamadi et al., 2005). Based on our data, we propose a model whereby VANGL2 regulates MMP14 endocytosis at sites of cell–matrix adhesion downstream of effects on FAK activation. However, we do not rule out additional points of intersection between VANGL2 and MMP14, such as in EEA1-positive early endosomes.

Our data suggest that matrix metalloproteinases act downstream of VANGL2 in vitro. To corroborate these findings in zebrafish, we tested whether trilobite^m209/vangl2 mutant embryos (Jessen et al., 2002) exhibit increased metalloproteinase activity and have altered ECM organization. First, proteins were collected from wild-type and trilobite^m209/vangl2 homozygous mutants at various embryonic stages and incubated with fluorogenic protease substrates. Quantification of fluorescence showed a striking increase in enzyme activity in trilobite^m209/vangl2 mutants that was reduced to near background levels by treatment with the metalloproteinase inhibitor GM6001 (Fig. 4A). Interestingly, embryonic protease activity increases in wild-type embryos between mid-gastrulation and the end of gastrulation (80% epiboly-1 somite) correlating with the appearance of ECM (Latimer and Jessen, 2010) and the onset of the trilobite^m209/vangl2 mutant phenotype (Sepich et al., 2000). To determine whether increased protease activity in trilobite^m209/vangl2 mutant embryos is mediated by Mmp14, we performed a gelatin protease assay using embryos injected with mmp14 morpholino. Notably, knockdown of Mmp14 in trilobite^m209/vangl2 homozygous mutants reduced metalloproteinase activity to near wild-type levels (Fig. 4A), indicating that Mmp14 is aberrantly active when Vangl2 function is reduced. To test the effect of Vangl2 loss of function on ECM assembly, we performed fibronectin immunolabeling on whole-mount and sectioned embryos. Consistent with their increased metalloproteinase activity, trilobite^m209/vangl2 mutants had decreased levels of fibronectin (Fig. 4B). Fibronectin assembly into a fibrillar matrix is a cell-mediated process requiring engagement of integrin receptors and generation of cell tension (Schwarzbauer and Sechler, 1999). Because inhibition of PCP signaling in frog embryos interferes with cell adhesion, polarity and fibronectin fibrillogenesis (Dzamba et al., 2009), it is possible that disrupted polarity in trilobite^m209/vangl2 mutants also contributes to loss of ECM. However, despite having a severe cell polarity defect (Topczewski et al., 2001), fibronectin is not reduced in knypek^m119/glypican4 mutants (N.A.M., R.E.Q. and J.R.J., unpublished data). Although we do not exclude the contributions of cell polarity and adhesion to ECM assembly, our data suggest a significant role for matrix metalloproteinases in regulating ECM degradation and remodeling during zebrafish gastrulation.

The above data suggest that loss of Vangl2 function in zebrafish trilobite mutants disrupts Mmp14 trafficking, resulting in increased cell-surface expression and degradation of ECM. To confirm this hypothesis, we tested whether the Vangl2 loss-of-function convergence and extension phenotype is suppressed by inhibition of Mmp14 activity. Injection of a vangl2 antisense morpholino produced a moderate or severe convergence and extension phenotype in ~95% of wild-type embryos (Fig. 4C,D). By contrast, co-injection of mmp14 and vangl2morpholinos consistently rescued this defect in a small percentage of embryos, as indicated by normal anterior–posterior body length and mediolateral width (Fig. 4C,D). In these experiments, high doses (10 ng/embryo) of mmp14 morpholino were required to achieve consistent rescue, whereas lower doses (5 ng/embryo) had little effect. The Vangl2-knockdown phenotype was exacerbated in some mmp14 morpholino-injected embryos, indicating a requirement for tight regulation of Mmp14 activity during embryogenesis. Indeed, we previously showed that high doses of mmp14 morpholinos disrupt cell polarity and induce a convergence and extension defect (Coyle et al., 2008). Taken together, our data show that Mmp14 acts downstream of Vangl2 to mediate the establishment or maintenance of PCP.

The ability of Vangl2 to regulate the proteolytic activity of Mmp14 in migrating gastrula cells can partly explain how cell polarization is propagated across a field of cells. We propose that a proper balance of matrix metalloproteinase activity is required during gastrulation to regulate integrin–ECM interactions and the alignment or stabilization of membrane protrusions along fibronectin. Asymmetrical Vangl2 expression in polarized cells would promote or exclude Mmp14 and proteolytic activity from certain cell surfaces and thereby influence cell polarity autonomously and non-autonomously. One prediction from this model is that excess metalloproteinase activity and loss of ECM in trilobite/vangl2 mutant embryos would disrupt polarized membrane protrusive activity. In agreement, migrating cells in trilobite/vangl2 mutants are able to generate protrusions, but they cannot be polarized and/or stabilized (Jessen et al., 2002). Future studies will be needed to establish how Vangl2 regulates Mmp14 trafficking in vivo and to identify the specific roles of Mmp14 (and Mmp2) for cell polarity and directed migration underlying convergence and extension.

HT-1080 cell culture, siRNA-mediated knockdown of VANGL2, and internalization and recycling assays were performed as described (Remacle et al., 2003; Cantrell and Jessen, 2010). Biotinylated proteins were isolated using streptavidin Dynabeads (Invitrogen). After magnetic separation, samples were boiled in Laemmli buffer and analyzed by western blot. The blots were stripped and re-probed for GAPDH. Densitometry was performed using ImageJ (rsb.info.nih.gov/ij) to quantify MMP14 and GAPDH protein bands, and data were normalized to GAPDH. MMP14 and transferrin internalization were visualized fluorescently as described (Remacle et al., 2003).

Imaging was performed using a Leica DMI6000B epifluorescence microscope equipped with an ORCA monochromatic camera (Hamamatsu), a 40× oil objective, and Simple PCI software (Hamamatsu). Certain images (control and experimental) were manipulated using the brightness and contrast functions of the above software. Images of zebrafish embryos were collected using a Leica MZ16F stereomicroscope and DFC280 camera. Confocal images were captured using an Olympus FV1000 laser-scanning microscope using a 100× oil objective (2× optical zoom) and an optical thickness (z) of 0.67 μm (provided by the VUMC Cell Imaging Shared Resource).

Immunolabeling and cell monolayer wounding were performed as described (Cantrell and Jessen, 2010). Primary antibodies were used against the following proteins: MMP14 for localization studies and western blot (MAB3328, Millipore), MMP14 for endocytosis assay (Z-812, Santa Cruz Biotechnology), VANGL2 (AF4815, R&D Systems), LAMP1 (Developmental Studies Hybridoma Bank), EEA1 and RAB11 (C45B10 and D4F5, Cell Signaling Technologies), VAMP1 (ab3346, Abcam), GM130 (P-20, Santa Cruz Biotechnology), FAK and fibronectin (RB-477 and AB-10, Lab Vision Corporation), Y397-P FAK (KAP-TK131, Enzo Life Sciences) and GAPDH (clone 6C5, Ambion). The FAK inhibitor PF-573,228 was dissolved in DMSO at 10 mM and used as described (Slack-Davis et al., 2007) at a concentration of 35 μm (chosen because it provided inhibition of FAK Y397 phosphorylation without significant loss of cell viability).

To quantify colocalization of VANGL2 in vesicles, we generated an HT-1080 cell line stably transfected with a GFP–VANGL2 fusion protein. This construct exhibits an identical (but brighter) expression pattern to that detected by the VANGL2 antibody. Cells were fixed in 4% paraformaldehyde (PFA) and subjected to immunolabeling for MMP14, EEA1, RAB11, LAMP1 or VAMP1, as indicated. The z-stack from each field containing the highest number of GFP-expressing vesicles was digitally analyzed using ImageJ software to determine the total number of VANGL2, MMP14 or additional marker-positive vesicles in three independent experiments. Fields of ~400 μm² (indicated by boxes in Fig. 2B–D and supplementary material Fig. S3B) were quantified in >20 cells labeled with VANGL2 and MMP14 and 6–10 cells labeled with all other markers. The multi-point tool in ImageJ software was used to count focal adhesions in images of non-targeting control (NT) and VANGL2-knockdown cells. For comparison of the differences between means, we used a two-tailed Student's t-test.

Zebrafish were maintained under standard conditions and embryos staged according to morphology. Embryos were dechorionated with pronase, deyolked and placed in ice-cold lysis buffer without protease inhibitors (15 mM NaCl, 10 mM HEPES, 0.1% Triton X-100, pH 8). Proteins were quantified and assays performed using either DQgelatin or DQcollagen IV as directed (Invitrogen). In some experiments, 100 μM GM6001 metalloproteinase inhibitor (Calbiochem) or DMSO were added. Alternatively, embryos were injected at the one cell-stage with mmp14 morpholino, as described below.

The vangl2 morpholino (Gene Tools, LLC; 5′-AGTTCCACCTTACTCCTGAGAGAAT-3′) was injected into single-cell embryos at 3 ng/embryo. Morpholinos to both zebrafish Mmp14 isoforms (Coyle et al., 2008) were injected at doses of 5 ng of each isoform per embryo (10 ng total). We compared non-injected controls with embryos injected with vangl2 morpholino or a combination of vangl2 and mmp14a/mmp14b morpholinos. Embryos were fixed at the ~12-somite stage and analyzed by in situ hybridization. Probes used were zic1 (forebrain), egr2 (hindbrain rhombomeres) and myod1 (somites). Embryos were placed in different phenotypic classes: (1) mild or partially rescued, less severe than the moderate vangl2 phenotype; (2) moderate, similar to the moderate vangl2 phenotype; (3) severe, worse than the moderate vangl2 phenotype.

We thank Alissa Weaver, Department of Cancer Biology, Vanderbilt University Medical Center for paxillin–GFP HT-1080 cells and J. Dunlap for technical assistance.