DLC3 suppresses MT1-MMP-dependent matrix degradation by controlling RhoB and actin remodeling at endosomal membranes Free

Loss of epithelial polarity is an early event in carcinogenesis, followed by the acquisition of invasive motility as the tumor progresses. Apart from cell-autonomous genetic alterations, the transition from an epithelial to a mesenchymal (EMT) phenotype is influenced by the tumor microenvironment, with paracrine factors such as transforming growth factor β (TGF-β) inducing signaling and gene expression that is supportive of cancer cell invasion (Derynck et al., 2014; Ye and Weinberg, 2015). Such cancer cells frequently form specialized actin-rich adhesion structures, termed invadopodia, that have the ability to degrade the surrounding extracellular matrix (ECM) to promote invasion and tumor cell dissemination during metastasis (Eddy et al., 2017; Murphy and Courtneidge, 2011). Although the precise morphology of these structures observed in vitro depends on the cell culture conditions, including ECM rigidity and composition, it is now accepted that tumor cells in vivo also rely on invadopodia-like structures for intravasation (Murphy and Courtneidge, 2011; Eddy et al., 2017).

One of the main proteases involved in matrix degradation at invadopodia is the surface-exposed membrane type 1-matrix metalloproteinase (MT1-MMP, also known as MMP14) whose correct delivery to invadopodia is critical for their formation and functionality. The upregulation of MT1-MMP surface levels is linked to malignancy in multiple cancer types including lung, colon and breast cancer (Castro-Castro et al., 2016). MT1-MMP is concentrated at invadopodia by polarized secretion upon activation of cell adhesion receptors, and by the recycling from endocytic compartments involving Rab7-positive late endosomes/lysosomes or Rab5A- and Rab4-dependent endocytic and exocytic cycles (herein, Rab7 refers to Rab7A, and Rab4 to Rab4A) (Frittoli et al., 2014; Monteiro et al., 2013; Macpherson et al., 2014; Williams and Coppolino, 2011).

The fission, movement and fusion of endosomal transport carriers during the targeted delivery of proteins to the plasma membrane rely on the remodeling of the actin and microtubule cytoskeleton. Cytoskeletal remodeling is controlled by members of the Rho GTPase family, comprising RhoA, RhoB and RhoC (Ridley, 2006; Croisé et al., 2014). RhoA and RhoC are mainly involved in the regulation of actin polymerization at the plasma membrane, whereas RhoB is known to control cytoskeletal remodeling at endosomal membranes through activation of the Dia family of formins and Src kinase (Fernandez-Borja et al., 2005; Sandilands et al., 2004; Wallar et al., 2007). How endosomal RhoB itself is regulated is still poorly understood.

Rho proteins cycle between an active GTP-bound state to interact with effector proteins, modulating their activity and localization, and an inactive GDP-bound state. This activation of Rho proteins is controlled by the guanine nucleotide exchange factors (GEFs), which promote the release of bound GDP and facilitate GTP binding, and the GTPase-activating proteins (GAPs), which increase the intrinsic GTPase activity of Rho GTPases to accelerate the return to the inactive state (Bos et al., 2007; Fritz and Pertz, 2016). Together, these proteins ensure the tight control of Rho activation in space and time. Apart from the dysregulation of the Rho proteins themselves, the inappropriate Rho signaling observed in transformed cells can also result from an imbalance of the respective GEFs and GAPs (Porter et al., 2016). Despite their importance, the nature and isoform specificity of the GEFs and GAPs that regulate Rho signaling in time and space is still largely unknown.

The three members of the deleted in liver cancer (DLC) family of RhoGAP proteins have attracted growing attention due to their frequent downregulation in various types of cancers, including those of the breast (Durkin et al., 2007b; Wang et al., 2016). DLC3, also known as STARD8, is the least characterized member of the DLC family. Similar to what is found for DLC1 and DLC2 (also known as STARD12 and STARD13, respectively), DLC3 contains an N-terminal sterile α motif (SAM), a C-terminal steroidogenic acute regulatory protein-related lipid transfer (START) domain and a GAP domain that regulates RhoA (Braun and Olayioye, 2015; Durkin et al., 2007a; Kawai et al., 2007). Although the DLC proteins are structurally related, our previous work has uncovered isoform-specific cellular functions for DLC3 that are associated with its distinct subcellular localization. In breast and colonic epithelial cells, DLC3 is recruited to the plasma membrane by interacting with the PDZ domain-containing scaffold protein Scribble where DLC3 is required for adherens junction stability and polarized 3D morphogenesis (Hendrick et al., 2016; Holeiter et al., 2012). Here, we investigated the role of DLC3 in breast epithelial cells that possess invasive properties. We show in TGF-β-treated MCF10A and in triple-negative MDA-MB-231 breast cancer cells that DLC3 dynamically associates with endosomal membranes to regulate RhoB activity and actin polymerization. Importantly, DLC3 depletion enhanced protease-dependent matrix degradation through the aberrant Rab4-dependent recycling of MT1-MMP. Our findings thus shed light onto the mechanisms by which the inactivation of DLC3 function in cancer cells may contribute to tumor progression.

MCF10A cells are known to undergo a reversible EMT-like program in response to TGF-β stimulation (Pignatelli et al., 2012; Zhang et al., 2014). The exposure of MCF10A cells to TGF-β for 1 week induced the moderate upregulation of the mesenchymal markers N-cadherin and vimentin, while E-cadherin expression was maintained (Fig. S1A). Depletion of DLC3 had no marked effect on total vimentin and E-cadherin levels; however, N-cadherin levels were elevated in the DLC3-knockdown cells (Fig. S1A). In HeLa cells, we observed redistribution of N-cadherin to the Golgi complex upon DLC3 depletion (Fig. S1B). DLC3 might thus regulate N-cadherin at the protein level, rather than driving gene expression changes associated with the EMT process. Interestingly, depletion of DLC3 using independent short-interfering siRNAs (spDLC3 and siDLC3) significantly increased TGF-β-induced cell adhesion and spreading (Fig. S1D,E) compared to the TGF-β-treated siRNA control cells. Western blot verified the effective knockdown of DLC3 in TGF-β-treated cells (Fig. S1C).

TGF-β stimulation of mammary epithelial cells has been shown to promote invadopodia formation and ECM degrading activity (Mandal et al., 2008). To analyze whether DLC3 controls matrix degradation, TGF-β-treated MCF10A cells were plated onto fluorescently labeled crosslinked gelatin. Cells lacking DLC3 showed a more-spread morphology, pronounced stress fiber formation and large actin-rich degradative structures, as visualized by phalloidin staining and staining of the actin-binding invadopodial marker protein cortactin (Fig. 1A). Most importantly, quantification of the degraded area relative to the fluorescence intensity of the non-degraded matrix revealed a significant increase in matrix degradation upon loss of DLC3 expression (Fig. 1B). The broad spectrum MMP inhibitor GM6001 efficiently blocked TGF-β-induced matrix degradation of the control and DLC3-knockdown cells, demonstrating the requirement for MMP activity in this assay (Fig. 1B). We then performed a rescue experiment to address whether the increased matrix degradation upon DLC3 depletion was dependent on Rho GTPases. The knockdown of RhoB resulted in a significant reduction of the gelatin degradation activity of DLC3-depleted cells, whereas the knockdown of RhoA had no effect (Fig. 1C). The F-actin structures observed in the DLC3-depleted cells were largely unaffected upon RhoB knockdown (Fig. S2A), pointing at a rescue of invadopodia functionality rather than general cytoskeletal organization of the cell. It should be noted that the upregulation of RhoB in RhoA-knockdown cells (Fig. 1D), which has been observed previously (Braun et al., 2015; Ho et al., 2008; Vega et al., 2011), most likely compensates for the loss of RhoA. Simultaneous RhoA and RhoB knockdown failed to suppress the increased matrix degradation in DLC3-depleted cells, which again is probably explained by the maintenance of RhoB expression in the absence of RhoA (Fig. 1D). Taken together, these data suggest that DLC3 suppresses TGF-β-induced matrix degradation, most likely by regulating RhoB and perhaps also RhoA.

To determine the Rho GTPase specificity of DLC3, we used established biosensors for RhoA, RhoB and RhoC (Pertz et al., 2006; Reinhard et al., 2016; Zawistowski et al., 2013) and measured, by determining the global fluorescence resonance energy transfer (FRET) efficiencies upon acceptor photobleaching, their regulation by DLC3 in TGF-β-treated MCF10A cells. Notably, upon DLC3 depletion, only the activity of the RhoB biosensor significantly increased in the entire cell (Fig. 2A), whereas the activities of the RhoA and RhoC biosensors were unchanged (Fig. S2B). The FRET efficiency of the RhoB biosensor was plotted as pseudocolor images representing high (red) and low (blue) Rho activity (Fig. 2D). Consistent with the original report (Reinhard et al., 2016), the RhoB biosensor was strongly enriched on endosomal membranes, as seen by the colocalization with the early endosomal marker EEA1 and the late endosomal marker Rab7 (Fig. S2C). Accordingly, in DLC3-depleted cells, the FRET efficiency of the RhoB biosensor was particularly increased on vesicular structures (Fig. 2B), but not at the plasma membrane (Fig. 2C). To prove that DLC3 can directly regulate RhoB, we co-expressed the RhoB biosensor along with wild-type (WT) and GAP-inactive DLC3 (DLC3-K725E), respectively, in HeLa cells. The GEF protein Dbl (also known as MCF2) was used as a positive control. Expression of DLC3-WT but not DLC3-K725E caused a significant decrease in FRET efficiency, demonstrating that DLC3 promotes the GTP-hydrolyzing activity of RhoB (Fig. S2D). The low RhoB levels in MCF10A cells precluded the detection of endogenous RhoB activity. However, in HeLa cells, endogenous RhoB activity was also increased in the absence of DLC3, as measured by Rho-binding domain (RBD) pulldowns, suggesting that in a cellular context, DLC3 functions as a GAP for RhoB (Fig. 2E,F). The increased Rho-GTP levels might additionally result from increased RhoB protein levels, which was especially obvious in the siDLC3 cells (Fig. 2E). We next aimed to visualize the dynamics of the DLC3 and RhoB association in living cells. For these localization studies, we employed GAP-inactive DLC3 because ectopic expression of wild-type DLC3 caused a strong reduction of actin stress fibers and alterations of cell morphology as shown previously (Braun et al., 2015; Holeiter et al., 2012). To image membrane proximal vesicles, TGF-β-treated MCF10A cells transiently expressing GFP–DLC3-K725E and Orange-tagged RhoB were analyzed by total internal reflection fluorescence (TIRF) live-cell microscopy. Indeed, we observed merging and co-trafficking of vesicles positive for GFP-DLC3 K725E and Orange–RhoB (Fig. 2G; Movie 1), proving the dynamic, vesicular colocalization of DLC3 and RhoB in living cells and supporting the conclusion that DLC3 directly regulates RhoB at endosomal membranes.

We next asked the question how DLC3 is recruited to RhoB-positive endosomes. By performing proteomic mass spectrometry analysis, we previously identified the early endosomal adaptor protein sorting nexin 27 (SNX27) as a candidate DLC3-interacting protein (Hendrick et al., 2016). Considering that DLC3 contains a PDZ ligand (PDZL) motif and SNX27 a PDZ domain, we aimed to validate the biochemical interaction of the proteins. To this end, Myc-tagged SNX27 was transiently expressed in HEK293T cells, immunoprecipitated from cell lysates and binding of co-expressed FLAG-tagged DLC3-K725E was analyzed by western blotting. Whereas co-immunoprecipitation of the full-length proteins was readily observed, no co-immunoprecipitation was seen with a SNX27 construct lacking the PDZ domain (SNX27-ΔPDZ) or a DLC3 construct lacking the C-terminal PDZL motif (FLAG–DLC3-ΔPDZL) (Fig. 3A). As expected, in TGF-β-treated MCF10A cells, SNX27 colocalized with the early endosomal marker EEA1 at vesicular structures, many of which were also decorated with ectopically expressed GFP–RhoB (Fig. 3B). Importantly, whereas GFP–DLC3-K725E localized at EEA1-positive vesicles in these cells, endosomal localization of GFP–DLC3-K725E was compromised upon SNX27 depletion with two independent siRNAs (Fig. 3C,D; Fig. S3B). Efficient knockdown of SNX27 was verified by western blotting (Fig. S3C). GFP-tagged wild-type DLC3 was also found at SNX27-positive vesicles, indicating that the functional DLC3 protein is also recruited to endosomes (Fig. S3A). These results suggest that RhoB regulation at endosomal membranes relies on the SNX27-dependent recruitment of DLC3.

Transport of metalloproteinases, such as MT1-MMP to invadopodia, is an important step in extracellular matrix remodeling underlying the mesenchymal mode of cell invasion. To investigate whether DLC3 affects MT1-MMP trafficking, we made use of the invasive triple-negative breast cancer cell line MDA-MB-231 stably expressing Cherry–MT1-MMP (Sakurai-Yageta et al., 2008). In accordance with the results obtained in TGF-β-treated MCF10A cells, DLC3 depletion also significantly increased the matrix degradation activity of MDA-MB-231 cells stably expressing MT1-MMP (Fig. S4A) and triple-negative BT549 breast cancer cells (Fig. S4B), demonstrating that DLC3 function is conserved in these cell lines. In addition, ectopically expressed GFP–DLC3-WT and -K725E were also found at SNX27-positive vesicular structures in MDA-MB-231 cells (Fig. S4C). We next investigated whether the localization of Cherry-tagged MT1-MMP in MDA-MB-231 cells depended on DLC3 expression. In cells lacking DLC3, EEA1 was strongly and significantly enriched at vesicular structures, many of which were positive for MT1-MMP (Fig. 4A,B). Quantification revealed a significantly increased overlap of the MT1-MMP and EEA1 signals in the DLC3-depleted cells, indicating an accumulation of MT1-MMP in early endosomes upon DLC3 knockdown (Fig. 4C). Moreover, F-actin was also significantly enriched at early endosomes, as assessed by the strong recruitment of the F-actin-binding protein LifeAct (Riedl et al., 2008) to EEA1-positive structures (Fig. 4D,E). By contrast, no accumulation of Cherry–MT1-MMP was observed in Rab7-positive late endosomes, suggesting that DLC3 loss leads to the specific accumulation of MT1-MMP at the early endosomal stage (Fig. S5A,B). In TGF-β-treated MCF10A cells lacking DLC3, EEA1-positive endosomal structures, which were also significantly increased compared to those in the control (Fig. S3D), contained significantly higher levels of ectopically expressed Cherry–MT1-MMP (Fig. S3E), agreeing with our observations in MDA-MB-231 cells. Importantly, compared to the DLC3-knockdown cells, the early endosomal accumulation of MT1-MMP was markedly attenuated by the co-depletion of RhoB (Fig. 4F). Finally, treatment of cells with low doses of the actin-disrupting agents latrunculin B or cytochalasin D (Duleh and Welch, 2010; Puthenveedu et al., 2010; Neel et al., 2007) fully abrogated the aberrant matrix-degrading activity of DLC3-depleted cells, whereas the activity of the control cells was not affected by the treatment (Fig. 4G). Taken together, we conclude that DLC3 regulates MT1-MMP trafficking and matrix degradation in an actin-dependent manner and through the local control of RhoB at early endosomal membranes.

We next investigated whether the subcellular distribution of MT1-MMP depended on DLC3 expression. Quantification of immunoblots revealed moderately enhanced total MT1-MMP levels upon DLC3 knockdown, which reached statistical significance in the case of the siDLC3 cells (Fig. 5A). By performing a biotinylation experiment for proteins on the cell surface, we could distinguish between the plasma membrane and intracellular MT1-MMP pool. The increase in MT1-MMP levels was reflected almost equally at both locations, indicating that MT1-MMP is not simply trapped intracellularly in the absence of DLC3 (Fig. 5B). To directly measure the arrival of MT1-MMP at the plasma membrane we employed MT1-MMPpHluorin, originally described by Lizarraga et al. (2009). Here, MT1-MMP is fused to a fluorophore that emits green fluorescence only at a neutral pH, whereas fluorescence is quenched in the acidic endosomal environment. Fusion of MT1-MMPpHluorin-containing vesicles with the plasma membrane can then be detected in real-time by TIRF live-cell microscopy as the appearance of bright green bursts of fluorescence. Remarkably, compared to control MDA-MB-231 cells ectopically expressing MT1-MMPpHluorin, the number of fluorescent bursts recorded at the plasma membrane was significantly enhanced in the absence of DLC3 (Fig. 5C; Movie 2), suggesting that DLC3 regulates MT1-MMP exocytosis.

At early endosomes, MT1-MMP is sorted to the Rab7-positive late endosomal compartment from where it is either recycled to the plasma membrane or targeted for lysosomal degradation. Alternatively, in response to motogenic stimuli, MT1-MMP can be directly recycled from early endosomes by a Rab4-dependent fast recycling route (Frittoli et al., 2014; Stenmark, 2009). Considering the lack of MT1-MMP accumulation in Rab7-positive endosomes and the enhanced frequency of MT1-MMP exocytic events in DLC3-depleted cells, we speculated that, in these cells, the MT1-MMP pool accumulated in the early endosomal compartment is redirected to the plasma membrane by a Rab4-dependent mechanism. Indeed, in DLC3-depleted MDA-MB-231 cells, we observed frequent and significantly enhanced colocalization of Cherry–MT1-MMP with GFP–Rab4-positive vesicles compared with the siRNA control cells (Fig. 6A,B). To demonstrate the involvement of Rab4 in the increased matrix degradation in cells lacking DLC3, we co-depleted DLC3 and Rab4 in MDA-MB-231 cells and analyzed the matrix-degrading ability of the cells. Whereas Rab4 knockdown alone using two independent siRNAs did not affect the degree of matrix degradation, the matrix-degrading activity of DLC3-depleted MDA-MB-231 cells was no longer elevated when Rab4 was co-depleted (Fig. 6C). Efficient knockdown of Rab4 was verified by quantitative real-time RT-PCR (qRT-PCR) (Fig. S5C). Rab4 depletion also prevented the enhanced matrix degradation seen in TGF-β-treated MCF10A cells lacking DLC3, proving that DLC3 signals through a common mechanism (Fig. 6D). These findings provide strong evidence for a requirement of DLC3 in endosomal sorting, whereby cell surface exposure of MT1-MMP and matrix degradation activity are suppressed (see Fig. 7).

Although it has been known for many years that RhoB associates with endolysosomal membranes, relatively little is known about the molecular mechanisms governing the local RhoB activation and inactivation cycles. By using a RhoB biosensor, biochemical RBD pulldowns and colocalization studies in fixed and living cells, we provide strong evidence for the direct stimulation of DLC3-mediated RhoB-GTP hydrolysis at endosomal membranes. The importance of RhoB regulation by DLC3 for endosomal trafficking is underscored by the excessive accumulation of F-actin at and trapping of MT1-MMP in early endosomal structures in cells depleted of DLC3, which was associated with aberrant Rab4-dependent MT1-MMP recycling and invadopodial matrix degradation.

Rho GTPases are critically involved in the assembly and maturation of invadopodia. Whereas RhoA regulates the exocyst-dependent delivery of MT1-MMP to invadopodia and invadopodia assembly (Sakurai-Yageta et al., 2008; Kuroiwa et al., 2011; Castro-Castro et al., 2016), spatially confined RhoC activity at the outer regions of invadopodia is important for their maturation (Bravo-Cordero et al., 2013; Spuul et al., 2014). To the best of our knowledge, this is the first report demonstrating the functional involvement of endosomal RhoB in invadopodial matrix degradation. Although a requirement for RhoC, but not RhoA, was previously reported for TGF-β-induced matrix degradation in MCF10A cells (Pignatelli et al., 2012), RhoC did not seem to play a role under conditions of DLC3 depletion. This can probably be explained by our TGF-β stimulation conditions, which primed the cells, but did not induce a robust EMT phenotype relying on RhoC. The prominent association of DLC3 with endomembranes in TGF-β-treated MCF10A cells provides support for endosomal RhoB as a primary DLC3 target. This agrees with the elevated RhoB biosensor activity and F-actin accumulation at vesicular structures in cells lacking DLC3. Because loss of RhoA can be compensated for by elevated RhoB expression, we cannot rule out RhoA contributing to MT1-MMP transport to invadopodia or local effects of DLC3 at the invadopodia themselves. Furthermore, in MDA-MB-231 cells, MT1-MMP followed a Rab8-dependent exocytic trafficking pathway (Bravo-Cordero et al., 2007) and we previously reported localization of DLC3 to Rab8-positive membrane tubules in HeLa cells (Braun et al., 2015). We thus cannot exclude that DLC3 also contributes to the regulation of Rab8-dependent polarized exocytosis of MT1-MMP. However, considering the strong suppression of matrix degradation upon Rab4 co-depletion, the primary function of DLC3 appears to be the regulation of MT-MMP1 recycling from early endosomes, rather than from late endosomes or in exocytosis of newly synthesized protein.

Actin assembly, triggered by Arp2/3 and the WASH complex, organizes early endosomal membranes into functional subdomains and is required for cargo sorting and the generation of transport intermediates (Puthenveedu et al., 2010; Derivery et al., 2009; Gomez and Billadeau, 2009). SNX27 is targeted to these endomembranes by lipid binding via its PX domain and cooperates with retromer and WASH in cargo recognition and recycling of proteins containing PDZL motifs (McGough et al., 2014; Steinberg et al., 2013; Temkin et al., 2011). Apart from the sorting of transmembrane proteins, SNX27 has been implicated in the transport of the GEF/GAP protein complex between βPIX (also known as ARHGEF7) and GIT1 to focal adhesions (Valdes et al., 2011). Although the SNX27–DLC3 interaction may similarly serve as a transport mechanism, our data favor more the idea of DLC3 being an integral component of the endosomal actin remodeling and sorting machinery. This is supported by siRNA-mediated knockdown of DLC3 mimicking the effects of constitutively active RhoB with regard to endosomal actin coat assembly (Fernandez-Borja et al., 2005), which results in delayed endosome maturation and transport of the epidermal growth factor to the lysosomal compartment (Gampel et al., 1999; Braun et al., 2015). A role for RhoB in the sorting decisions to either the recycling or degradation pathway was also previously proposed for the chemokine receptor CXCR2 (Neel et al., 2007). Very recently, the direct binding of endosomal actin was shown to be required for efficient constitutive recycling of the EGFR and MT1-MMP (MacDonald et al., 2018). Intriguingly, abrogation of actin-dependent recycling of MT1-MMP resulted in defective matrix degradation and invasion of triple-negative breast cancer cells (MacDonald et al., 2018). Enhanced sequestering of MT1-MMP into F-actin-enriched endosomal subdomains through direct actin interactions might thus provide a mechanistic explanation for the increased Rab4-dependent recycling of MT1-MMP in DLC3-depleted cells. A recent report investigating the trafficking of RhoB linked a block in endosomal maturation to the stabilization of the RhoB protein on early endosomal membranes (Gong et al., 2018). In the context of DLC3 depletion, this could provide an additional mechanism by which the elevated early endosomal RhoB-GTP pool is maintained.

In response to hepatocyte growth factor stimulation, a Rab5–Rab4 trafficking route has been shown to be required for MT1-MMP fast recycling and matrix degradation (Frittoli et al., 2014). Both Rab5 and Rab4 upregulation in invasive breast cancer has been further linked with poor prognosis and metastasis formation in vivo (Frittoli et al., 2014). Based on our data, we propose that the pro-invasive function of Rab5 and Rab4 is antagonized by the anti-invasive function of DLC3. The downregulation of DLC3 observed in breast cancer (Durkin et al., 2007a; Wang et al., 2016) might thus contribute to cancer progression through GAP-dependent mechanisms, although the precise role of DLC3 in cell invasion and metastasis remains to be defined. Whereas overexpression or hyperactivation of RhoA and RhoC is generally found to positively correlate with tumor progression in a number of tumor types, the role of RhoB in tumorigenesis appears to be context dependent and/or cell type specific (Haga and Ridley, 2016; Porter et al., 2016; Ju and Gilkes, 2018). Interestingly, RhoB has been identified as a transcriptional target of the TGF-β pathway and loss of RhoB impairs TGF-β-induced migration of keratinocytes and prostate cancer cells (Vasilaki et al., 2010). In a recent lung cancer study, however, RhoB appeared to act as a tumor suppressor and suppressed EMT (Dubois et al., 2016). Here, we show that in TGF-β-treated MCF10A cells, endosomal RhoB activity increases upon DLC3 depletion, leading to enhanced MT1-MMP-dependent matrix degradation, which was conserved in MDA-MB-231 and BT549 breast cancer cells. Thus, in these triple-negative breast cancer cells, downregulation of DLC3 appears to favor a pro-invasive phenotype.

RhoB has been implicated in the trafficking of different types of cargo, some of which are involved in invadopodia biogenesis and function. For example, the kinase c-Src, a key regulator of invadopodia formation, is trafficked by RhoB-dependent actin remodeling (Sandilands et al., 2004; Arnette et al., 2016). Similarly, in primary vascular smooth muscle and human umbilical vein endothelial cells (HUVECs), the trafficking of Cdc42 and Rac, which are involved in invadopodia formation and maturation, is regulated by RhoB (Huang et al., 2011; Marcos-Ramiro et al., 2016). Taking into account the enhanced adhesion and spreading of DLC3-depleted cells, it is possible that the trafficking of integrins is also regulated in a DLC3-dependent manner. In addition to the local Rho regulation at the plasma membrane, the altered membrane trafficking in cells depleted of DLC3 most likely contributes to the impairment of epithelial adherens junctions and polarized morphogenesis described previously (Holeiter et al., 2012; Hendrick et al., 2016). In future studies, it will thus be interesting to determine, through a global approach, how DLC3 modulates the spatial distribution of membrane-associated cellular proteins.

Antibodies used in this study were: rabbit anti-E-cadherin monoclonal antibody (mAb) (24E10), rabbit anti-EEA1 mAb, rabbit anti-RhoB polyclonal antibody (pAb) and rabbit anti-Rab7 mAb (D95F2) from Cell Signaling (Danvers, MA); mouse anti-N-cadherin mAb and mouse anti-vimentin mAb from BD Transduction Laboratories (Heidelberg, Germany); mouse anti-RhoA mAb (26C4) and mouse anti-DLC3 mAb (E-2) from Santa Cruz Biotechnology (Dallas, TX); mouse anti-Actin mAb (AC-40), mouse anti-Flag mAb (F1804), mouse anti-α-tubulin mAb and rabbit anti-GAPDH pAb (G9545) from Sigma-Aldrich (St Louis, MO); mouse anti-Cortactin mAb (clone 4F11) from Millipore (Burlington, MA); rabbit anti-mCherry pAb, mouse anti-SNX27 mAb (1C6) and rabbit anti-giantin pAb (ab24586) from Abcam (Cambridge, UK); mouse anti-c-Myc mAb (9E10) from Heiner Böttinger (Institute of Cell Biology and Immunology, University of Stuttgart, Germany). IR Dye 800-conjugated anti-GST antibody (600-132-200) was purchased from Rockland, HRP-labeled secondary anti-mouse-IgG and anti-rabbit-IgG antibodies from GE Healthcare and Dianova, IRDye 680- and IRDye 800-conjugated secondary anti-mouse-IgG and anti-rabbit-IgG polyclonal antibodies from Li-COR (Licor Biotechnology, Bad-Homburg, Germany), Alexa-Fluor®-labeled secondary IgG antibodies and Alexa-Fluor-labeled phalloidin were from Invitrogen (Carlsbad, CA). DAPI was obtained from Sigma-Aldrich, TGF-β from Peprotech (Rocky Hill, NJ), GM6001 from Tocris Bioscience (Bristol, UK) and Rho Inhibitor I from Cytoskeleton (Denver, CO). Oregon Green 488-conjugated gelatin was obtained from Thermo Fisher Scientific (Waltham, MA, USA) and GFP-Booster from Chromotek (Planegg-Martinsried, Germany). Latrunculin B and Cytochalasin D were obtained from Enzo Life Sciences (Lörrach, Germany). See Tables S1 and S2 for additional antibody details and dilutions used.

MCF10A cells (provided by Mohamed Bentires-Alj, Department of Biomedicine, University Hospital, Basel, Switzerland) were cultivated in DMEM/F12 medium (Invitrogen) supplemented with 5% horse serum (Invitrogen), 20 ng/ml EGF (R&D), 10 μg/ml insulin (Sigma), 0.5 μg/ml hydrocortisone (Sigma), 100 ng/ml cholera toxin (Sigma), and 50 mg/ml streptomycin (Invitrogen). MDA-MB-231 and BT549 cells (CLS, Eppelheim, Germany; authenticated in 2018) were cultured in Dulbecco's modified Eagle's medium (DMEM; Invitrogen), and HEK293T and HeLa cells (ATCC; authenticated in 2016) were cultured in RPMI 1640 (Invitrogen) supplemented with 10% fetal calf serum (FCS; PAA Laboratories, Cölbe, Germany). Cells were incubated in a humidified atmosphere of 5% CO₂ at 37°C. MDA-MB-231 cells stably expressing mCherry–MT1-MMP (kindly provided by Philippe Chavrier, Institute Curie, Paris, France) were cultivated in Leibovitz's L-15 medium (Thermo Fisher Scientific) supplemented with 15% FCS and 50 ng/ml G418 (Calbiochem) and incubated in a humidified atmosphere of 1% CO₂ at 37°C. MCF10A cells stably expressing GFP–DLC3-WT and -K725E were described previously (Holeiter et al., 2012). Cells were routinely tested for mycoplasma, frozen stocks were thawed and kept in culture for no longer than 3 months. For TGF-β stimulation, MCF10A cells were plated at 10⁶ cells per 6 cm dish and 5 ng/ml TGF-β was added to the medium the next day. Control cells were left untreated. Medium was replenished every 2–3 days for a total duration of 7 days before using cells for RNAi or overexpression experiments. For plasmid transfections, Lipofectamine LTX with Plus Reagent (Invitrogen) was used according to manufacturer's instructions. For RNAi, cells were transfected with siRNA using Lipofectamine RNAiMAX (Invitrogen) according to manufacturer's instructions. The siRNAs used were: negative control siRNA [siCon, ON-TARGETplus® non-targeting control pool D-001810-10 from Dharmacon (Lafayette, CO)], spDLC3 (siGENOME SMARTpool human STARD8 M-010254 from Dharmacon), siDLC3 (Silencer® Select STARD8 s18826), siSNX27#1 (Silencer® Select SNX27 s37695) and siSNX27#2 (Silencer® Select SNX27 s37696), siRab4#1 (Silencer® Select s11677) and siRab4#2 (Silencer® Select s11675, all from Ambion life technologies), spRhoA (ON-TARGETplus® SMARTpool RhoA J-003860 from Thermo Fisher Scientific) and spRhoB (ON-TARGETplus® SMARTpool RhoB J-008395 from Thermo Fisher Scientific). The biosensors pTriEx-RhoA FLARE.sc and RhoC FLARE.sc mCer were from Addgene [#12150 (Pertz et al., 2006); #65071 (deposited by Klaus Hahn; Zawistowski et al., 2013)]. The RhoB FRET Biosensor has been described previously (Reinhard et al., 2016). pEGFPN1-Lifeact was provided by Michael Sixt (Institute of Science and Technology, Klosterneuburg, Austria), pEGFP-Rab4 vector by Hesso Farhan (Department of Molecular Medicine, Institute of Basic Medical Sciences, University of Oslo, Norway) and pEGFC1-Rab7 by Lucas Pelkmans (Institute of Molecular Life Sciences, University of Zurich, Switzerland). Plasmids encoding MT1-MMP–mCherry and MT1-MMPpHluorin were kindly provided by Philippe Chavrier (Lizarraga et al., 2009; Sakurai-Yageta et al., 2008). pEGFP-C1-RhoB and pmOrange2-C1-RhoB were generated by PCR amplification using pECFP-endo (encoding RhoB; Clontech) as a template and the following forward and reverse primers: 5′-CCGGAATTCCCCCATGGAGCAGAAGCTGATC-3′ and 5′-CCGGGATCCTCATAGCACCTTGCAGCAGTTG-3′. The PCR product was cloned into the pEGFP-C1 and pmOrange-2-C1 vectors, respectively, through BamHI and EcoRI restriction enzyme sites. pCMV-myc-SNX27 and pCMV-myc-SNX27 ΔPDZ were described previously (Valdes et al., 2011) and were kindly provided by Martin Playford (National Heart, Lung and Blood Institute, National Institutes of Health, Bethesda, USA). pCRV62-Met-Flag-DLC3 WT and K725E were described previously (Braun et al., 2015; Holeiter et al., 2012). pCR3.V62.Met-Flag-DLC3 K725E lacking the five C-terminal amino acids PETKL (ΔPDZL) was generated by PCR amplification using pCRV62-Met-Flag-DLC3 K725E as a template and the following forward and reverse primers: 5′-CCGGAATTCCCTCTGCTGGACGTTTTCTG-3′ and 5′-CCGGAATTCTCAGCCCGCTGCCTGCAGGG-3′. The PCR product was cloned into the pcR3.V62.Met-Flag vector using the EcoRI restriction enzyme site.

Cells were lysed in RIPA buffer [50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% Triton-X-100, 0.5% sodium deoxycholate, 1 mM EDTA, 0.5 mM PMSF, 0.1% SDS, 1 mM sodium orthovanadate, 10 mM sodium fluoride, and 20 mM β-glycerophosphate plus Complete protease inhibitors without EDTA (Roche)] and lysates were clarified by centrifugation at 16,000 g for 10 min. Protein concentration was determined by performing a Bio-Rad DC protein assay. Proteins were separated by SDS-PAGE (NuPage® Novex® 4–12% Bis-Tris gels, Invitrogen) and transferred to nitrocellulose membranes using an iBlot® device (iBlot^®Gel Transfer Stacks; Invitrogen). The membrane was blocked with 0.5% blocking reagent (Roche) in PBS containing 0.1% Tween-20 and then incubated with primary antibodies, followed by IRDye 680- or 800-labeled secondary antibodies for visualization with the Odyssey® Imaging System (LI-COR Biosciences) or, alternatively, by HRP-labeled secondary antibodies for ECL-based (Pierce, Rockford, IL) visualization with the Fusion Solo system (Vilber Lourmat). For immunoprecipitation, cells were lysed in 0.5% NEB buffer [50 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.5% NP-40, 1 mM EDTA, 0.5 mM PMSF, 1 mM sodium orthovanadate, 10 mM sodium fluoride, and 20 mM β-glycerophosphate plus Complete protease inhibitors without EDTA (Roche)], equal amounts of protein were diluted with lysis buffer to a final concentration of 1 mg/ml and incubated with 1.5 μg of anti-c-Myc antibody overnight at 4°C. Immune complexes were collected using protein G agarose (KPL) for 1 h at 4°C and washed three times with 0.25% NEB buffer. Proteins were analyzed by SDS-PAGE, transferred to polyvinylidene difluoride membranes (Roth, Karlsruhe, Germany) and further processed as described above.

BL21 bacteria were transformed with a pGEX vector encoding the RBD of rhotekin and expression was induced with 0.1 mmol/l isopropyl-β-d-1-thiogalactopyranoside for 4 h at 37°C. Bacteria were harvested, resuspended in PBS containing Complete protease inhibitors without EDTA (Roche), and sonicated. Triton X-100 was added (1% final) and the lysate was centrifuged for 10 min at 8000 g. GST–RBD was purified with glutathione resin (GE Healthcare). For pulldowns, cells were stimulated with 100 ng/ml EGF for 5 min and lysed in RBD lysis buffer [150 mM Tris-HCl pH 7.5, 1 mM EDTA, 500 mM NaCl, 10 mM MgCl₂, 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS, 10% glycerol, 0.5% 2-mercaptoethanol plus Complete protease inhibitors without EDTA (Roche)]. Equal amounts of cleared lysates were incubated with GST–RBD beads for 1 h at 4°C. Beads were washed four times with RBD lysis buffer, bound proteins were separated by SDS-PAGE and RhoB was analyzed by western blotting.

For adhesion assays, cells were seeded at low density (20,000 cells) into collagen-coated (10 µg/ml) 96-well E-plates (Roche) in serum-free medium, and the impedance of cells was measured using an xCELLigence device (Roche). For Crystal Violet staining, cells (50,000 cells per well) were seeded in growth medium in 24-well plates, fixed after 2, 4, 6 and 20 h and stained with 0.2% Crystal Violet. Images were acquired with an EVOS Cell Imaging System (Thermo Fisher Scientific).

Cells grown on glass coverslips coated with 10 μg/ml collagen R (Serva; Heidelberg, Germany) or cross-linked gelatin, were fixed for 15 min with 4% paraformaldehyde. After washes in PBS, cells were incubated for 15 min with 150 mM glycine in PBS and permeabilized for 5 min with 0.2% Triton X-100 in PBS. Blocking was performed with 5% goat serum (Invitrogen) in PBS containing 0.1% Tween 20. Fixed cells were incubated with primary antibodies diluted in blocking buffer for 2 h at room temperature. Following three washing steps with PBS containing 0.1% Tween 20, cells were incubated with Alexa Fluor® (488, 546, 633)-labeled secondary antibodies and phalloidin in blocking buffer for 1 h at room temperature. Nuclei were counterstained with DAPI, and cells were washed twice with PBS containing 0.1% Tween 20. Coverslips were mounted in Fluoromount-G® (SouthernBiotech, Birmingham, AL) and analyzed at room temperature on a confocal laser scanning microscope (LSM 710, Carl Zeiss; Oberkochen, Germany) equipped with EC Plan-Neofluar 40×/1.30 DIC M27 and Plan-Apochromat 63×/1.40 DIC M27 (Carl Zeiss) oil immersion objectives using 405-, 488-, 561- and 633-nm excitation lasers. Linear adjustments to brightness and contrast, as well as maximum intensity projections were made using the ZEN software (Carl Zeiss). For quantification of mean fluorescence intensities, images were acquired with the same confocal settings and analyzed using the ImageJ software (NIH). The Manders' coefficient was determined using the JACoP plugin (Bolte and Cordelieres, 2006).

For matrix degradation assays, coverslips were incubated with fluorescently labelled gelatin (1 mg/ml) in an inverted manner on ethanol-sterilized parafilms protected from light for 30 min in a humidified chamber. Gelatin-coated coverslips were then cross-linked with 0.5% glutardialdehyde (Roth, Karlsruhe, Germany) in H₂O for 30 min, transferred to dishes, washed three times with PBS and stored at 4°C until further use. Before seeding the cells, coverslips were equilibrated with medium at 37°C for at least for 30 min. Cells (100,000 cell per coverslip) were replated on gelatin coated coverslips for 3 h (MDA-MB-231 cells) and 6 h (MCF10A and BT549 cells) and coverslips were processed as described in the immunofluorescence microscopy section. 40 confocal images (40× oil magnification) per sample were acquired using identical settings (488 and DAPI channel). Quantitative image analysis of gelatin degradation was performed with CellProfiler software version 3.0.0 (Carpenter et al., 2006; Kamentsky et al., 2011). In brief, the total area of gelatin degradation per image was measured and normalized to the number of nuclei.

At 2 days post siRNA transfection, control and DLC3-depleted cells were transiently transfected with the vectors encoding the Rho biosensors. The next day, cells were fixed and mounted in Mowiol® (Polysciences, Warrington, PA). FRET efficiencies were determined by using the acceptor photobleaching method. CFP was excited with a diode UV laser at 405 nm and emission was detected in the 454–515 nm spectral window. YFP was excited with the 514 nm laser line of an argon laser and emission was detected from 515–621 nm. Donor and acceptor images were acquired pre- and post-bleaching. Cells were bleached for YFP with the 514 nm argon laser line (90% intensity, 30 iterations, pixel dwell 50.4 µs). The FRET efficiency was calculated from the increase of the donor intensity (CFP) after acceptor bleaching using the FRET module of ZEN 2009 software (Carl Zeiss). FRET efficiency images were generated with a MATLAB script (developed by Dr Felix Neugart, University Stuttgart, Germany; available from the corresponding author upon request) that allows background suppression and visualization of the FRET efficiency at the same time by using a two dimensional look up table (total fluorescence intensity coded by pixel brightness, FRET efficiency coded by color).

Transfected cells were replated onto collagen-coated 35 mm high glass bottom µ-Dishes (ibidi, cat. no: 81158). TIRF studies were performed on a Zeiss Axio Observer, equipped with a motorized TIRF illuminator (Laser TIRF 3), an EMCCD camera (Photometrics Evolve 512) and an alpha Plan-Apochromat 100×/1.46 NA Oil objective. MT1-MMPpHluorin was visualized with a 488 nm diode laser in combination with a 525/50 nm emission filter. TIRF images were acquired at 37°C and 5% CO₂ every 10 s for a time interval of 30 min. Image processing was done with Zen 2.3 blue software.

Cells were incubated with 1 mg/ml nonpermeable Biotin (EZ-Link Sulfo-NHS-SS-Biotin, Sigma-Aldrich) for 30 min on ice under gentle agitation. Cells were then washed with PBS, quenched for 5 min and washed once more with PBS before lysis in RIPA buffer. Lysates were incubated with streptavidin beads for 2 h at 4°C to precipitate the biotinylated surface proteins, followed by three washes with 1% NEB buffer. Proteins were released by boiling in SDS sample buffer and analyzed by immunoblotting.

RNA was isolated with the NucleoSpin® RNA Kit (Macherey-Nagel, Hœrdt, France). qRT-PCR was performed with a Power SYBR® Green 1-Step kit (Thermo Fisher Scientific) using a Cfx96 device (BioRad) according to the manufacturer's protocol for one-step RT-PCR. Primers used were 5′- GTCCGTGACGAGAAGTTATTACC-3′ and 5′-TGAGCGCACTTGTTTCCAAAA-3′ for Rab4 (Eurofins, Luxembourg) and Hs_PPIA_4_SG QuantiTect Primer Assay (Qiagen). Changes in the relative expression level were determined using the 2^–ΔΔCt method (Biorad CFX manager software 3.1). PPIA was used as a control gene for normalization.

Bar graphs show the mean±s.e.m. For box plots, the box represents the 25–75th percentiles, and the median is indicated; whiskers extend 1.5 times the interquartile range from the 25th and 75th percentiles, and outliers are represented as determined by GraphPad Prism 7 software. ‘N’ refers to the number of sample points for each box blot and ‘n’ to the number of independent experiments. Statistical significance was analyzed by one-way ANOVA followed by Tukey's post-test if not otherwise stated (GraphPad Prism version 7; GraphPad Software Inc., La Jolla, CA). Results were considered significant when P-values were below 0.05 [*P<0.05; **P<0.01; ***P<0.001; ****P<0.0001; n.s. (not significant) P>0.05], all P-values are provided in Table S3.

We thank Angelika Hausser, Cristiana Lungu and Sebastian Lieb (University of Stuttgart, Germany) for helpful discussions, critical reading of the manuscript and figure design. We are grateful to Philippe Chavrier (Institute Curie, France) for providing stable MDA-MB-231 mCherry-MT1-MMP cells and MT1-MMP plasmids, and to Martin Playford (NIH, USA), Hesso Farhan (University of Oslo, Norway), Michael Sixt (Institute of Science and Technology, Austria) and Lucas Pelkmans (University of Zurich, Switzerland) for providing plasmids.