Rnd3-induced cell rounding requires interaction with Plexin-B2

ABSTRACT Rnd proteins are atypical members of the Rho GTPase family that induce actin cytoskeletal reorganization and cell rounding. Rnd proteins have been reported to bind to the intracellular domain of several plexin receptors, but whether plexins contribute to the Rnd-induced rounding response is not known. Here we show that Rnd3 interacts preferentially with plexin-B2 of the three plexin-B proteins, whereas Rnd2 interacts with all three B-type plexins, and Rnd1 shows only very weak interaction with plexin-B proteins in immunoprecipitations. Plexin-B1 has been reported to act as a GAP for R-Ras and/or Rap1 proteins. We show that all three plexin-B proteins interact with R-Ras and Rap1, but Rnd proteins do not alter this interaction or R-Ras or Rap1 activity. We demonstrate that plexin-B2 promotes Rnd3-induced cell rounding and loss of stress fibres, and enhances the inhibition of HeLa cell invasion by Rnd3. We identify the amino acids in Rnd3 that are required for plexin-B2 interaction, and show that mutation of these amino acids prevents Rnd3-induced morphological changes. These results indicate that plexin-B2 is a downstream target for Rnd3, which contributes to its cellular function. Summary: The Rho GTPase Rnd3 is known to alter cell shape and the actin cytoskeleton. Here, the semaphorin receptor plexin-B2 is identified as a new Rnd3 partner that mediates its effects on cell shape.


INTRODUCTION
The three Rnd proteins, Rnd1, Rnd2 and Rnd3, are members of the Rho family of small GTPases. They are atypical because they lack intrinsic GTPase activity and exist only in the GTP-bound form (Riou et al., 2010). They are instead regulated by reversible phosphorylation at their C-termini, which promotes binding of 14-3-3 proteins. The 14-3-3 proteins in turn extract Rnd proteins from membranes and inactivate them by translocating them to the cytosol .
In many mammalian cell types, Rnd1 and Rnd3 act antagonistically to RhoA by inducing loss of actin stress fibres and cell rounding (Guasch et al., 1998;Nobes et al., 1998). By contrast, Rnd2 does not induce cell rounding, probably because its localization is different to that of Rnd1 and Rnd3 (Oinuma et al., 2012). Rnd2 and Rnd3 play distinct roles in cortical neuronal migration during mouse development, in part because their expression is regulated by different transcription factors (Pacary et al., 2011). Interestingly, in endothelial cells, both Rnd2 and Rnd3 induce stress fibre assembly by upregulating RhoB expression (Gottesbuhren et al., 2013).
Several proteins that interact with Rnd proteins and could mediate their effects on cell shape and motility have been reported (Riou et al., 2010). The first to be identified was p190RhoGAP, which downregulates RhoA activity. Rnd3 interaction with p190RhoGAP stimulates its activity and induces loss of stress fibres (Wennerberg et al., 2003). Rnd3, but not Rnd1 or Rnd2, also binds to and is phosphorylated by the serine/threonine kinases ROCK1 and PKC (Madigan et al., 2009;Riento et al., 2005). Rnd1 interacts with the transmembrane receptor plexin-B1, and is reported to promote plexin-B1-induced RhoA activation (Oinuma et al., 2003;Rohm et al., 2000). Plexins are well known as neuronal guidance receptors that are activated by semaphorin ligands and frequently promote growth cone collapse (Püschel, 2007). They also contribute to multiple aspects of mammalian development, as well as angiogenesis, cancer progression and immune responses (Perälä et al., 2012;Worzfeld et al., 2014).
Plexins have a Rho-binding domain (RBD) in the middle of a split GAP domain, which is reported to interact with R-Ras and act as a GAP for Rap1 ( Fig. 1A) (Fansa et al., 2013;Pascoe et al., 2015). Several Rho GTPases have been reported to bind to the RBD of plexin-B1 in addition to Rnd1, including Rac1, RhoD, Rnd2 and Rnd3 (Fansa et al., 2013). Rnd proteins can also bind to other plexins, such as plexin-A1 and plexin-D1 (Rohm et al., 2000;Uesugi et al., 2009). Here, we report that Rnd3 interacts preferentially with plexin-B2 compared with plexin-B1 and plexin-B3, and map the amino acids in Rnd3 required for plexin-B2 interaction. Rnd3 interaction with plexin-B2 is required for it to induce cell rounding, indicating that it is a target for Rnd3 activity.
Plexin-B1 has been reported to interact with and act as a GAP for the small GTPases R-Ras or Rap1 (Oinuma et al., 2004;Wang et al., 2012). We therefore tested the effects of Rnd proteins on plexin-B interactions with R-Ras and the two closely related Rap1 proteins, Rap1A and Rap1B. Similar to plexin-B1, we found that plexin-B2 and plexin-B3 associated with R-Ras, Rap1A and Rap1B  (full-length and cytoplasmic region) showing the extracellular Sema domain, plexins-semaphorinsintegrins (PSI) domains and immunoglobulinsplexins-transcription factor (IPT) domains, and the intracellular split GAP domain, Rho-binding domain (RBD) and C-terminal PDZ-binding motif (PBM). (B,C) COS7 cells were transfected with expression vectors encoding VSV-epitope-tagged full-length plexin-B2 (B) or HA-epitope-plexin-B2-cyto (C). Cell lysates were incubated with GST or GST-Rnd3 on glutathione-Sepharose beads. Bound proteins were analysed by immunoblotting with the indicated antibodies. (D,E) Lysates of COS7 cells expressing full-length plexin-B2 (D) or HA-plexin-B2-cyto (E) and FLAG-Rnd3 were immunoprecipitated with anti-FLAG or anti-HA antibody, followed by immunoblotting. Full-length plexin-B2 migrates as a doublet at around 180 kDa. (F) Lysates of HeLa cells expressing FLAG-Rnd3 or transfected with empty vector (control) were immunoprecipitated with anti-FLAG antibody, followed by immunoblotting for endogenous plexin-B2. Results from three independent experiments are shown in the blot. All other blots are representative of three independent experiments.
( Fig. 3A-C). To determine whether Rnd proteins and/or plexin-B proteins affect R-Ras or Rap1 activity, active GTP-bound R-Ras and Rap1 (antibody recognizing Rap1A and Rap1B) were pulled down from cells expressing exogenous plexins and/or Rnd proteins. No significant changes in R-Ras activity or Rap1 activity were detected when Rnd1, Rnd3 and/or plexin-B1 or plexin-B2 were coexpressed (Fig. 3D,E). These results indicate that Rnd1 and Rnd3 proteins do not stimulate plexin-B1 or plexin-B2 GAP activity. This is consistent with the lack of change in the structure of the plexin-B1 GAP domain upon Rnd1 binding (Tong et al., 2009) or in its RapGAP activity (Wang et al., 2012).
The Rnd3 C-terminus and effector domain are both required for plexin-B2 association Because Rnd3 interacts preferentially with plexin-B2, we characterized the regions of Rnd3 required for plexin-B2 interaction, using deletion mutants and specifically designed point mutants (Fig. 4A), which were chosen on the basis of modelling predictions. The crystal structure of the Rho-binding domain of plexin-B1 with Rnd1 (PDB: 2REX) indicates that amino acids in the 'effector loops/switch 1 and switch 2 regions' of Rnd1 are required for plexin-B1 interaction (Tong et al., 2009;Wang et al., 2011). Modelling of Rnd3 into this crystal structure allowed us to identify potential amino acids required for Rnd3-plexin-B2 binding ( Fig. 4B; model kindly provided by David Komander).
Rnd3 amino acids T55, V56, F57, Y60, V87 and L90 in the effector loops were required for its interaction with plexin-B2, whereas N51 and N86 were not required (Fig. 4E). Based on our structural model (Fig. 4B), these results imply that Rnd3 interacts with the RBD of plexin-B2 in a similar way to Rnd1 interaction with the plexin-B1 RBD, but that additional C-terminal residues in Rnd3 are likely to be required to confer the specificity of Rnd3 for plexin-B2.
Interaction of Rnd3 with plexin-B2 stimulates Rnd3-induced rounding and enhances Rnd3 inhibition of invasion Previous studies have shown that stimulation of plexin-B1 by its ligand Sema4D and by Rnd1 can promote cell contraction in COS7 cells and induce growth cone collapse in neuronal cells (Oinuma et al., 2004(Oinuma et al., , 2003. Rnd3 is known to induce loss of stress fibres and reduce actomyosin contractility in a variety of cell types (Riou et al., 2010). We therefore investigated whether plexin-B2 interaction with Rnd3 affected cell morphology.
HeLa cells were transfected with constructs encoding wild-type Rnd3 and/or full-length plexin-B2. Rnd3 induced loss of actin stress fibres and thin protrusions, as expected (Fig. 6A). Plexin-B2 alone did not induce a phenotype. However, co-expression of plexin-B2 and Rnd3 strongly induced cell rounding, measured by a significant decrease in cell spread area (Fig. 6A,B). To test the physiological consequences of these morphological changes induced by Rnd3 and plexin-B2, we investigated their effects on HeLa cell invasion through Matrigel-coated Transwell filters. Rnd3 alone inhibited invasion of HeLa cells, and co-expression of plexin-B2 further reduced invasion (Fig. 6C).
We next tested the effects of Rnd3 mutations that prevent its interaction with plexin-B2 (see Fig. 3E). Rnd3ΔC, Rnd3-V56Y and Rnd3-V87R, which do not bind to plexin-B2, induced much less cell rounding than wild-type Rnd3 when co-transfected with plexin-B2 (Fig. 6D,E). These mutants also did not reduce cell area when co-expressed with plexin-B2, in contrast to wild-type Rnd3 (Fig. S1). Approximately 18% of cells expressing wild-type Rnd3 alone, without co-expressing plexin-B2, induced cell rounding, whereas the Rnd3 mutants induced significantly less cell rounding (Fig. 6E). This indicates that endogenous plexin-B2 is likely to contribute to the response to Rnd3.
These results indicate that plexin-B2 interaction with Rnd3 promotes cell rounding, which correlates with reduced cell invasion. We therefore depleted plexin-B2 to determine whether it is required for Rnd3 responses. Plexin-B2 knockdown by two different siRNAs strongly reduced Rnd3-induced cell rounding but did not affect control cells or alter endogenous Rnd3 expression (Fig. 6F,G;  Fig. S2). Taken together, these results indicate that Rnd3 interaction with plexin-B2 is important for it to induce actin remodelling and cell shape change, and define plexin-B2 as a Rnd3 effector.

DISCUSSION
Rnd proteins are known to bind to plexin-B receptors, but the physiological consequences of these interactions are still unclear.
Here, we show that plexin-B2 increases the ability of Rnd3 to induce cell rounding and reduce HeLa cell invasion, and that plexin-B2 depletion reduces Rnd3-induced rounding, consistent with it being an effector for Rnd3. We also compare for the first time the interaction of each of the three Rnd proteins with the three plexin-B proteins. Of the three plexin-B proteins, we find that Rnd3 preferentially interacts with plexin-B2, although it is also able to associate more weakly with plexin-B3 and plexin-B1 in cells.
Of the three plexin-B proteins, the interaction between plexin-B1 and Rnd1 has been the most characterized (Rohm et al., 2000). It was previously reported that the cytoplasmic domain of plexin-B1 bound to all three Rnd family members (Uesugi et al., 2009) and that Rnd1 interacted more weakly with plexin-B2 than plexin-B1 (Oinuma et al., 2003). In our hands, however, Rnd1interaction with all three plexin-B proteins was considerably weaker than that of Rnd2. Sequence alignment of plexin-B3 with that of plexin-B1 suggests that plexin-B3 should interact with Rnd1 (Wang et al., 2011). Our co-immunoprecipitation data indicate that there is a relatively weak association between plexin-B3 and Rnd1. Interestingly, plexin-B2 Rho-binding domain lacks several amino acids and has several amino acid differences compared with plexin-B1 and plexin-B3 (see Fig. 2A), which could explain its preferential interaction with Rnd3. Plexin-B1 dimerizes when it binds to its dimeric Sema4D ligand, and dimerization is required for Sema4Dinduced signalling (Janssen et al., 2010). However, plexin-B2 is predicted not to form stable dimers as a result of the presence of transmembrane prolines . Signalling by plexin-B2 must therefore be different to that of plexin-B1.
The GAP domain of plexin-B2 is central to its function in vivo, although its targets are not clear (Worzfeld et al., 2014). Plexin-B1 has been reported to interact with and act as a GAP for Rap1 as well as R-Ras and its close relative M-Ras (Oinuma et al., 2004;Saito et al., 2009;Wang et al., 2012). Plexin-B2 does not appear to act as a GAP for R-Ras in mice, although Sema4C acts through plexin-B2 to reduce R-Ras activity in HEK293 cells (Worzfeld et al., 2014). However, plexin-B2 and plexin-B3 have not been tested for interactions with R-Ras or Rap1. We found that all three plexin-B family members associate with R-Ras, Rap1A and Rap1B. However, neither plexin-B1 nor plexin-B2 decreased R-Ras or Rap1 activity, with or without Rnd1 or Rnd3 expression. Regulation of Rap1 or R-Ras activity by plexin-B1 normally requires stimulation with its ligand Sema4D (Oinuma et al., 2004;Wang et al., 2012), although this can be autocrine. For example, COS7 cells and cervical cancer cell lines express plexin-B1 and its ligand Sema4D (Damola et al., 2013;Liu et al., 2014), and thus Sema4D is likely to act in an autocrine manner to stimulate plexin-B1. In HeLa cells, we found that plexin-B2 induced some cell rounding in the absence of added ligand, suggesting that they secrete the plexin-B2 ligand Sema4C.
In addition to the GAP domain, plexin-B receptors share a common C-terminal PDZ-binding motif (Pascoe et al., 2015). In plexin-B1 this interacts with PDZ-RhoGEF and LARG to activate RhoA, leading to growth cone collapse (Swiercz et al., 2002). By contrast, Rnd proteins and plexin-B1 can antagonize RhoA signalling by activating p190RhoGAP (Sun et al., 2012;Swiercz et al., 2008;Wennerberg et al., 2003). Plexin-B2 could therefore act as a scaffold, bringing together Rnd3, p190RhoGAP and RhoA. However, in cortical neurons in vivo, plexin-B2 was reported to activate RhoA by competing with p190RhoGAP for Rnd3 binding, thereby blocking Rnd3-mediated inhibition of RhoA (Azzarelli et al., 2014). It is clear that the effect of plexins on RhoA activity is context-dependent, for example via its association with either ErbB2 or Met tyrosine kinase receptors (Swiercz et al., 2008). Although RhoA activity was not affected by plexin-B1 or plexin-B2 expression in our experiments, RhoA expression was reduced by Rnd1 or plexin-B2, which could reflect targeting of the inactive form of RhoA for ubiquitination and degradation by a family of RhoA-binding BTB domain adaptors (BACURD proteins) (Chen et al., 2009). A potential interaction of Rnd3 with BACURD3, identified in our yeast two-hybrid screen, might promote degradation of inactive RhoA. Intriguingly, BACURD1 and BACURD2 have recently been reported to interact with Rnd2 and Rnd3 (Gladwyn-Ng et al., 2016).
Our data indicate that both the Rnd3 effector domain and the Rnd3 C-terminal extension are required for interaction with plexin-B2. Additionally, mutation of the C-terminal CAAX box strongly reduced the interaction between Rnd3 and plexin-B2. Because the CAAX box is required for Rnd3 farnesylation and membrane association (Foster et al., 1996), this indicates that the Rnd3 C-terminal extension is most likely to interact with the membraneproximal region of plexin-B2. This interaction probably confers specificity to the interaction of Rnd3 with plexin-B2 as the Cterminal extensions of Rnd proteins are very different in sequence . Based on the structure of Rnd1 with the plexin-B1 RBD domain (Tong et al., 2009), the effector domain of Rnd3 interacts with the plexin-B2 RBD. The interaction between plexin-B2 and Rnd3 results in cytoskeletal remodelling to produce a rounded phenotype with decreased spread area. Furthermore, plexin-B2 knockdown strongly reduces Rnd3-induced cell rounding, indicating that plexin-B2 is essential for Rnd3 responses.  .m., n=3). Each dot represents a single cell; 19 cells were analysed for each condition. (C) HeLa cells were transfected with GFP (control), or GFP-Rnd3 with or without plexin-B2. Cell invasion was measured using Matrigel-coated Transwell filters. Invasion is shown relative to GFP-expressing control cells (mean±s.e.m., n=35). (D,E) HeLa cells (1×10 5 cells/ml) were transfected with pCMV5 (control), full-length plexin-B2 and wild-type (WT) FLAG-Rnd3 or FLAG-Rnd3 mutants that do not bind to plexin-B2 (Rnd3-V56Y, Rnd3-V87R and Rnd3-1-200). After 24 h, cells were fixed and stained for plexin-B2 (green) and FLAG-Rnd3 proteins (red). Actin filaments were detected with AlexaFluor-633-conjugated phalloidin (blue). Scale bar: 10 μm. (E) Mean percentage±s.e.m. of rounded cells from three independent experiments; n≥100 cells per experiment. (F) HeLa cells were transfected with two siRNAs targeting plexin-B2 or with control siRNA. After 48 h, cells were transfected with FLAG-tagged Rnd3. Cells were lysed or fixed, permeabilized and immunostained 24 h later. Plexin-B2 depletion was determined by immunoblotting with anti-plexin-B2 antibody. Endogenous Rnd3 and exogenous Rnd3 were detected using anti-Rnd3 or anti-FLAG antibody, respectively. Blots are representative of three independent experiments. (G) Mean percentage±s.e.m. of rounded cells from three independent experiments; n≥100 cells per experiment. For all graphs, *P<0.05, **P<0.01, ***P<0.001 ****P<0.0001; ns, not significant; one-way ANOVA with Tukey posthoc test.
We show that the plexin-B2-Rnd3 interaction does not require Rnd3 phosphorylation on any of the seven known phosphorylation sites. By contrast, Rnd3 phosphorylation is required for binding to 14-3-3 proteins, leading to translocation of Rnd3 from membranes to the cytosol and inhibition of Rnd3 cellular function . Our data indicate that plexin-B2 competes with 14-3-3 for Rnd3 binding, and that the binding of Rnd3 to plexin-B2 and that of Rnd3 to 14-3-3 are mutually exclusive, consistent with 14-3-3 being a negative regulator of Rnd3 signalling.
In summary we show here that, of the three plexin-B receptors, plexin-B2 binds preferentially to Rnd3, and that plexin-B2 is important for Rnd3-induced cell rounding. This specificity of Rnd3 for plexin-B2 is likely to be mediated by the requirement of the Cterminal domain of Rnd3 for plexin-B2 interaction. It is plausible that plexin-B2 brings Rnd3 and p190RhoGAP together to enhance Rnd3-induced rounding. The role of plexin-B2 as a target for Rnd3 signalling could explain the apparently opposing effects of plexin-B2 in cancer cell invasion and cancer progression: in some cancer types such as breast cancer, both Rnd3 and plexin-B2 appear to inhibit hallmarks of tumorigenesis (Malik et al., 2015;Zhu et al., 2014), whereas in gliomas they promote tumour progression (Clarke et al., 2015;Le et al., 2015).

Yeast two-hybrid screening
The yeast two-hybrid screen was carried out essentially as previously described (McKenzie et al., 2006), using mouse Rnd3 lacking the Cterminal CAAX box cloned into pGBT9 to encode the GAL4 DNA-binding domain fused to Rnd3 as bait (Riento et al., 2003). A Human Lung Matchmaker cDNA library in pACT2 (Clontech), which produces proteins fused to the GAL4-activation domain, was used to screen for proteins that interact with Rnd3. A β-galactosidase filter assay was used to identify yeast colonies expressing potential Rnd3 interacting partners.

Cell culture and transfection
COS7 cells (obtained from Michael Waterfield, Ludwig Institute for Cancer Research) and HeLa cells (obtained from Michael Way, Francis Crick Institute, London, UK) were grown in Dulbecco's modified Eagle's medium (DMEM) containing 10% foetal calf serum (FCS). Cell lines were not authenticated in our laboratory. For transfection, COS7 cells were washed with 5 ml of cold electroporation buffer (120 mM KCl, 10 mM K 2 PO 4 , 10 mM KH 2 PO 4 pH 7.6, 25 mM HEPES pH 7.6, 2 mM MgCl 2 , 0.5% Ficoll). The buffer was removed and cells were re-suspended in 250 µl of cold electroporation buffer and electroporated at 250 V and 960 µF with 5 µg of DNA. The cells were then plated onto 10-cm diameter dishes and incubated for 24 h prior to lysis. HeLa cells were transiently transfected with plasmids using Lipofectamine 2000 (ThermoFisher Scientific).
For immunoblotting, proteins were resolved by SDS-PAGE and transferred to nitrocellulose membranes (Schleicher and Schuell). Membranes were blocked with TBS (20 mM Tris-HCl pH 7.6, 137 mM NaCl) containing 5% non-fat dried milk and 0.05% Tween-20, and then incubated with primary antibodies. Bound antibodies were visualized with horseradish peroxidase-conjugated secondary antibodies and enhanced chemiluminescence (ECL; Amersham Pharmacia Biotech).

Invasion assay
Hela cells were transfected with plasmids encoding GFP (control), GFP-Rnd3 with or without VSV-tagged full-length plexin-B2 using Lipofectamine 2000 (ThermoFisher Scientific). The upper chambers of Biocoat Matrigel invasion chambers (Corning; 8-µm pore size) were rehydrated with 300 µl of serum-free medium for 2 h at 37°C. HeLa cells (2×10 5 for each condition) in 0.1% FCS were added to the upper chamber, and medium containing 10% FCS was used as a chemo-attractant in the lower chamber. After 21 h, cells in Transwell inserts were fixed with 3.7% paraformaldehyde for 15 min, and GFP-expressing cells on the top and bottom of the filter were detected using a Zeiss LSM510 confocal microscope and Zen software. Z-stacks (2.03 µm spacing) were acquired for 6-10 fields using a 20× objective (0.5 NA). Reflectance was used to identify the position of the Transwell filter. Invading cells were quantified from three independent experiments. Graphs were generated using Prism (GraphPad Software).

Statistical analysis
Cell area and cell rounding data, and western blot data, were analysed using one-way ANOVA with Tukey posthoc test for multiple comparisons.