The F-BAR domain protein PACSIN2 associates with Rac1 and regulates cell spreading and migration

Cell migration is an essential feature of physiological processes such as development, chemotaxis and wound healing. The capacity of cells to migrate is controlled by the actin cytoskeleton, which regulates cell polarity, the organization of adhesion structures and the generation of force. Such force is provided by actin polymerization at the front of polarized cells and actomyosin-based contraction at the rear (Ridley et al., 2003). In addition to localized actin polymerization, microtubule dynamics and vesicular transport are also important for cell migration. Intracellular vesicles regulate transport of signaling proteins to and from the leading edge and also control vesicular traffic, which is required for membrane protrusion at the leading edge. These events probably run in parallel, as many signaling molecules involved in cell motility can associate with the plasma membrane, as well as with intracellular vesicles. As a result, both endocytosis and exocytosis are increasingly recognized as important regulatory events, controlling cell motility in conjunction with the actin cytoskeleton (Fletcher and Rappoport, 2010; Scita and Di Fiore, 2010).

Cytoskeletal dynamics are controlled by small GTPases of the Rho family, in particular Rac1, Cdc42 and RhoA (Ridley et al., 2003). Rho GTPases cycle between an inactive GDP-bound state and an active GTP-bound state, and this transition is regulated by guanine-nucleotide-exchange factors (GEFs) that promote the exchange of GDP for GTP (Rossman et al., 2005) and by GTPase-activating proteins (GAPs) that stimulate the low intrinsic GTPase activity (Bernards and Settleman, 2004). Whereas most activated Rho GTPases are localized at the plasma membrane (del Pozo et al., 2000), inactive Rho GTPases associate with Rho guanine nucleotide dissociation inhibitor (RhoGDI), which is a cytosolic protein (del Pozo et al., 2002; Olofsson, 1999). Although there is increasing evidence that traffic to and from the plasma membrane constitutes an important aspect of Rho GTPase signaling, the underlying regulatory mechanisms are not well understood.

Among the family of Rho GTPases, Rac1 (Didsbury et al., 1989) is one of the most extensively studied members. Following activation, Rac1 can interact with a series of effector proteins, such as the p21-activated kinase (PAK) serine/threonine kinase, to trigger downstream signaling. Upon binding to activated Rac1, for example, following integrin activation, PAK in turn becomes activated, regulating cytoskeletal dynamics, adhesion and transcription (del Pozo et al., 2000; Price et al., 1998). In addition, Rac1 regulates formation of membrane ruffles and lamellipodia through members of the Scar and WAVE family of proteins (Tybulewicz and Henderson, 2009) that couple Rac1 to the Arp2/3 complex, which mediates actin polymerization (Chung et al., 2000).

In Rac1 proteins, the C-terminal hypervariable region is important for subcellular targeting and the regulation of signaling output (Michaelson et al., 2001; Nethe et al., 2010; Prieto-Sanchez and Bustelo, 2003; ten Klooster and Hordijk, 2007; van Hennik et al., 2003). In a proteomic screen for proteins associating with the C-terminus of Rac1, we previously identified the Rac1 and Cdc42 GEF β-Pix (also known as Rho guanine-nucleotide-exchange factor 7), which recruits Rac1 to leading edge focal adhesions (FAs) and to the peripheral membrane (ten Klooster et al., 2006). In addition, we found a series of proteins that bind to the Rac1 C-terminus and translocate upon Rac1 activation. These include: the PP2A inhibitor SET, which translocates from the nucleus to the plasma membrane in response to Rac1 activity and cooperates with Rac1 in cell migration (ten Klooster et al., 2007); the adapter protein CD2-associated protein (CD2AP; also known as CMS), which translocates to cell–cell junctions following Rac1 activation and supports Rac1-dependent cell adhesion (van Duijn et al., 2010); and caveolin-1, which is recruited to focal adhesions by Rac1 activity and regulates the polyubiquitylation and degradation of Rac1 in an adhesion-dependent fashion (Nethe et al., 2010).

Here, we describe the identification of PACSIN2 (for ‘protein kinase C and casein kinase 2 substrate in neurons 2’) as a Rac1-binding protein that, like CD2AP and caveolin-1, regulates endocytosis. PACSIN2, also known as syndapin 2, is a ubiquitously expressed 486-amino-acid membrane-associated adapter protein. Two additional PACSIN proteins exist in humans: PACSIN1, which is mainly expressed in brain tissues, and PACSIN3, which is expressed at high levels in skeletal muscle, heart and lung tissue (Modregger et al., 2000; Plomann et al., 1998; Ritter et al., 1999). PACSIN proteins form one branch of the Fer-CIP4 homology-BAR (F-BAR) domain protein family (Heath and Insall, 2008). They all share a highly conserved F-BAR domain at their N-terminus, which is a protein module that stabilizes and induces membrane curvature (Frost et al., 2009; Peter et al., 2004; Tsujita et al., 2006). The result of curvature induction by PACSIN proteins (and other F-BAR-domain-containing proteins) is the formation of invaginations and, subsequently, of vesicular-tubular structures that depend on the self-assembly of F-BAR modules into a helical coat (Frost et al., 2009).

In addition to the N-terminal F-BAR domain, PACSIN proteins harbor a central linker region, containing NPF motifs (in PACSIN1 and PACSIN2) or a proline-rich motif (PACSIN3), and a C-terminal Src homology 3 (SH3) domain. Many proteins are known to bind the PACSIN SH3 domain, including dynamin, N-WASP and synaptojanin (Chitu and Stanley, 2007; Kessels and Qualmann, 2004). The F-BAR domain also mediates homo- and hetero-oligomerization of PACSIN proteins (Kessels and Qualmann, 2006), and this oligomerization is important for their capacity to act as adaptor proteins, linking the actin-regulatory proteins with the endocytic machinery (Kessels and Qualmann, 2004).

In this study, we show that the F-BAR protein PACSIN2 interacts, through its SH3 domain, with the C-terminus of Rac1. We show that reciprocal regulation affects both Rac1 and PACSIN2. Rac1 activity controls the subcellular distribution of PACSIN2, whereas PACSIN2 negatively regulates Rac1 activity, cell spreading and migration by promoting Rac1 inactivation. To our knowledge, this is the first report describing the regulation of Rac1 signaling output by an F-BAR-domain-containing protein.

In the course of a proteomic screen for proteins that bind to the Rac1 hypervariable C-terminus, we identified by mass spectrometry the F-BAR-domain-containing protein PACSIN2 as a novel Rac1 interactor. The interaction between Rac1 and endogenous PACSIN2 was confirmed by streptavidin-based pull-down assays using a biotinylated peptide encoding the C-terminus of Rac1 in lysates of COS7, HeLa and mouse embryonic fibroblast (MEF) cells (Fig. 1A). To test whether PACSIN2 also interacts with other Rho family members, we performed a pull-down assay using biotinylated C-termini of related Rho GTPases. This experiment showed that PACSIN2 specifically interacts with the C-terminus of Rac1 (Fig. 1B). Using Myc-tagged versions of the three PACSIN isoforms expressed in human (Modregger et al., 2000; Plomann et al., 1998; Ritter et al., 1999) we showed that the C-terminus of Rac1 can bind, with comparable efficiency, to PACSIN1, 2 and 3 (supplementary material Fig. S1A). To confirm that endogenous PACSIN2 also binds full-length Rac1 we performed pull-down assays with bacterially purified GST, GST–Rac1 WT (wild-type protein), and GST–Rac1 ΔC (which lacks the hypervariable domain) using cell lysates of Jurkat, COS7 and HeLa cells. These experiments showed that PACSIN2 interacts with the full-length Rac1 protein and that PACSIN2 binding to GST–Rac1 ΔC is greatly reduced (Fig. 1C). This shows that the C-terminal domain of Rac1 is both necessary and sufficient for this interaction.

To study further the binding between PACSIN2 and Rac1 we performed pull-down experiments with bacterially purified GST and GST–PACSIN2 proteins using lysates of HeLa cells. These experiments showed that endogenous Rac1 interacts with GST–PACSIN2 (supplementary material Fig. S1B). Furthermore, using lysates of HeLa cells transfected with YFP–Rac1 WT, or the Rac1 Q61L, T17N and V12 mutants, we showed that all constructs can bind to GST–PACSIN2, although the active mutants of Rac1 (Q61L and V12) bound more efficiently (Fig. 1D). To investigate further the relevance of the Rac1–PACSIN2 interaction we studied whether the endogenous proteins can interact. To test this, we isolated endogenous Rac1-GTP with a biotinylated PAK1-CRIB domain (Price et al., 2003) and analyzed PACSIN2 binding to Rac1-GTP. This experiment showed that endogenous PACSIN2 was found in complex with endogenous Rac1-GTP (Fig. 1E). In addition, we studied whether Rac1 and PACSIN2 directly interact with each other. To study this, we used isolated GST and GST–PACSIN2 and performed a peptide pull-down assay using the biotinylated C-terminus of Rac1 and of several related Rho GTPases. This experiment showed that purified PACSIN2 interacts selectively with the Rac1 C-terminal peptide, indicating that this is a direct interaction (Fig. 1F).

The hypervariable C-terminus of Rac1 comprises two protein-binding motifs, a proline-stretch and a poly-basic region (Fig. 1G, upper panel) (van Hennik et al., 2003). Additional pull-down experiments using variant peptides of the Rac1 C-terminus (PPP→AAA; RKR→AAA) showed that binding of endogenous PACSIN2 is abolished when either the proline-stretch or the poly-basic region is mutated (Fig. 1G, lower panel). PACSIN2 does not interact with the 17–32 peptide (Fig. 1G, lower panel), which encodes part of the effector domain of Rac1 (Vastrik et al., 1999). SH3 domains are known to interact with proline-rich sequences (Li, 2005). To determine whether the PACSIN2 SH3 domain mediates the binding to Rac1 we performed pull-down assays with the Rac1 C-terminus, as well as with GST–Rac1, using HeLa cells transfected with Myc-tagged PACSIN2 WT or with an SH3 double-mutant, PACSIN2 Y435E/P478L (Modregger et al., 2000). These mutations in PACSIN2 abolished the binding to the Rac1 C-terminus, as well as to full-length Rac1 (Fig. 1H; supplementary material Fig. S1C). Together, these data identify PACSIN2 as an interactor of the small GTPase Rac1.

To study the intracellular localization of PACSIN2, we immunostained HeLa cells for endogenous PACSIN2, as well as for various markers and analyzed the distribution of these proteins by confocal laser-scanning microscopy. The majority of PACSIN2 localized to perinuclear vesicles and to vesicular-tubular structures at the leading edge of polarized cells (Fig. 2A). Co-staining for F-actin showed no obvious colocalization of PACSIN2 with actin structures (Fig. 2A, upper panels). Similarly, co-staining for paxillin showed that PACSIN2 does not localize to focal adhesions (Fig. 2A, middle panels). Finally, co-staining for α-actinin, which localizes predominantly to membrane ruffles and cortical actin structures, showed that PACSIN2 in the leading edge localizes distal to, rather than in, membrane ruffles (Fig. 2A, bottom panels). In primary human umbilical vein endothelial cells (HUVEC) and COS7 cells, endogenous PACSIN2 showed a similar subcellular distribution (supplementary material Fig. S2A,B). To confirm the specific detection of PACSIN2 in our immunostainings, we transfected HeLa cells with Myc-tagged constructs of the three different PACSIN proteins. As expected, the antibody recognizes PACSIN2 and not PACSIN3 on western blots (supplementary material Fig. S2C). The antibody cross-reacts to a limited extent with PACSIN1 (supplementary material Fig. S2C). However, PACSIN1 expression is restricted to neuronal cells (Plomann et al., 1998), which is why we are confident that we are only detecting PACSIN2 in the immunostainings. This was further confirmed by the loss of immunostaining in cells transfected with the PACSIN2 small interfering RNA (siRNA) (supplementary material Fig. S2D).

We next studied PACSIN2 distribution in live cells. We transfected HeLa cells with YFP- or mCherry-tagged PACSIN2 constructs and recorded their dynamics in real-time. Similar to the distribution in fixed cells, PACSIN2 was found in the periphery, at sites of membrane ruffling and in protrusions. Upon collapse of membrane ruffles, PACSIN2 was internalized and colocalized with newly formed elongated structures that move towards the cell center and appear similar to the structures seen in fixed cells (Fig. 2B; supplementary material Movie 1). In addition, PACSIN2 was found around perinuclear vesicles (supplementary material Fig. S2E). Together, these data are in good agreement with published findings on the tubulating activity of PACSIN proteins and the notion that PACSIN2 mediates endocytosis.

To determine whether these PACSIN2-containing tubular structures are indeed generated by endocytosis, we incubated HeLa cells, expressing YFP–PACSIN2, with various endocytic markers. In untransfected cells, internalized transferrin localized to PACSIN2-positive perinuclear vesicles but not to the peripheral tubular structures (Fig. 2C). Expression of PACSIN2 induced a reduction in internalized transferrin (data not shown), in line with previous observations (Modregger et al., 2000; Qualmann and Kelly, 2000). Furthermore, internalized Sulforhodamine101, which is a marker for clathrin-independent fluid-phase endocytosis (Wubbolts et al., 1996), localized to PACSIN2-negative tubular structures (supplementary material Fig. S2F). By contrast, internalized Alexa-Fluor-594-labelled Cholera toxin B (AF594-CtxB), which has been used previously as a marker to label tubular-endocytic structures (Verma et al., 2010), colocalized with PACSIN2 both in membrane patches and in peripheral tubular structures (Fig. 2D). This supports the notion that the PACSIN2-positive tubules are generated by endocytosis.

To identify the PACSIN2-positive perinuclear vesicles (Fig. 2), HeLa cells were stained for the early endosome marker EEA1 (Mu et al., 1995) and for Rab5, which also localizes to early endosomes (Christoforidis et al., 1999). These experiments showed that the perinuclear but not the peripheral pool of PACSIN2 resides on early endosomes (Fig. 3A,B). This was further confirmed by transfecting HeLa cells with GFP–FYVE. FYVE domains have been shown to recognize PtdIns(3)P and to localize to early endosomes (Gillooly et al., 2003). In line with the data in Fig. 3A,B, PACSIN2 localizes to GFP–FYVE-positive early endosomes (supplementary material Fig. S2G). Together, these results show that the perinuclear but not the peripheral pool of PACSIN2 localizes to early endosomes.

Given that PACSIN2 interacts with the small GTPase Rac1, it was important to establish whether and where PACSIN2 and Rac1 colocalize. Immunostaining of HeLa cells for endogenous PACSIN2 and endogenous Rac1 showed that both proteins are present at perinuclear vesicles, which we had identified as early endosomes, as well as in peripheral membrane protrusions (Fig. 3C). To confirm that Rac1 in fact localizes to early endosomes, we immunostained HeLa cells transfected with GFP–FYVE for endogenous Rac1. In good agreement with the data in Fig. 3A–C, endogenous Rac1 was found to localize to FYVE-positive early endosomes (supplementary material Fig. S2H). In addition to the colocalization of endogenous Rac1 and PACSIN2 on early endosomes, we found that endogenous Rac1 localized to PACSIN2 tubules (supplementary material Fig. S2I, arrow), although this localization was less clear owing to the high levels of Rac1 in the cytoplasm. In line with this result, we found, in live-cell imaging experiments, that in cells transfected with mCherry–Rac1 WT and YFP–PACSIN2, both proteins colocalized on endocytic structures (supplementary material Movie 2).

Previously, it has been shown that PACSIN proteins can co-immunoprecipitate α-tubulin and γ-tubulin, and PACSINs have been suggested to function in microtubule (MT) assembly (Grimm-Gunter et al., 2008). Moreover, because the MT network is important in the control of vesicle transport, we tested to what extent PACSIN2 localization was controlled by MTs. We found that a fraction of peripheral PACSIN2 aligned with microtubules (Fig. 4A; supplementary material Fig. S3). In cells treated with nocodazole to depolymerize MTs, PACSIN2 distribution was altered, in that the protein appeared restricted to vesicular structures that are evenly distributed throughout the cell. This was accompanied by a concomitant loss of the peripheral pool of PACSIN2 associated with vesicular-tubular structures (Fig. 4B; supplementary material Fig. S4A). Interestingly, whereas in control cells PACSIN2 only partially localized to early endosomes, after nocodazole treatment, all endogenous PACSIN2 was on early endosomes which were also dispersed (Fig. 4C; supplementary material Fig. S4B), indicating that PACSIN2 localization to the peripheral pool is dependent on MTs. This is further supported by the finding that 30–60 minutes after washout of nocodazole, PACSIN2 distribution into a perinuclear and peripheral pool was restored (Fig. 4B,C). Upon nocodazole treatment, the endogenous Rac1 distribution was dispersed, similar to that of PACSIN2 (supplementary material Fig. S4C). The effect of nocodazole on PACSIN2 distribution was mimicked by expression of constitutively active RhoA V14 (supplementary material Fig. S5A), which is in line with published data showing that MT depolymerization activates RhoA (Enomoto, 1996). These data show that the MT network regulates the intracellular distribution of PACSIN2.

Because PACSIN2 and Rac1 interact and colocalize, we next tested whether Rac1 controls PACSIN2 localization. Expression of a constitutively active Rac1 Q61L mutant drove endogenous PACSIN2 into the perinuclear pool, which was subsequently confirmed as comprising early endosomes, with a loss of peripheral PACSIN2 (Fig. 5A,B). To confirm this result, we treated HeLa cells with cytotoxic necrotizing factor 1 (CNF1) to activate endogenous Rac1 (Lerm et al., 1999; Nethe et al., 2010). CNF1-mediated constitutive activation of Rac1, as with expression of Rac1 Q61L, induced a loss of the peripheral pool of PACSIN2, concentrating PACSIN2 in the early endosomes (supplementary material Fig. S5B). Subsequent live-cell imaging studies showed that, whereas activated Rac1 induced a loss of peripheral PACSIN2 tubules, inhibiting Rac1 signaling by the expression of the inactive mutant Rac1 T17N induced an accumulation of PACSIN2-positive peripheral tubular structures (Fig. 6A and 6B, top and middle panels). Interestingly, Rac1 WT as well as Rac1 T17N clearly colocalized with PACSIN2 on the tubular structures (Fig. 6A), in line with the results for endogenous Rac1 and PACSIN2 (supplementary material Fig. S2I). It is important to note that levels of expressed mCherry–Rac1 were similar to those of endogenous Rac1. Moreover, expression of Rac1 mutants did not affect the levels of endogenous Rac1 (Fig. 6B, bottom panel). In addition to the Rac1-T17N-mediated accumulation of PACSIN2-positive tubular structures, we found that, following treatment of YFP–PACSIN2-expressing HeLa cells with a pharmacological inhibitor of Rac1 (EHT1864) (Shutes et al., 2007), the number of PACSIN2 vesicular-tubular structures increased (supplementary material Fig. S5C). A similar effect was seen after siRNA-mediated knockdown of Rac1 (Fig. 6C). Finally, we found that the SH3 domain mutant of PACSIN2, which cannot bind to Rac1, also promoted an accumulation of peripheral tubules, underscoring the notion that Rac1 activity, as well as Rac1 association with PACSIN2, is required to regulate the peripheral pool of PACSIN2 (supplementary material Fig. S5D). Together, these data show that Rac1 activity regulates the number of PACSIN2-positive tubular structures.

An important feature of cellular signaling pathways is the presence of feed-forward and feedback loops. The data in Figs 5 and 6 show that Rac1 signaling regulates PACSIN2 distribution. In a complementary set of experiments, we tested whether PACSIN2, in turn, regulates Rac1. Using siRNA transfection, we found that loss of PACSIN2 induces an increase in GTP loading of endogenous Rac1 of between twofold and fourfold (Fig. 7A), suggesting that PACSIN2 negatively regulates Rac1 signaling. In line with these data, we found that loss of PACSIN2 promotes cell spreading on fibronectin, a Rac1-dependent response (del Pozo et al., 2000; Price et al., 1998) (Fig. 7B, upper panel). The three different siRNAs to PACSIN2 all induced a similar increase in cell spreading as analyzed by electric cell–substrate impedance sensing (ECIS) (Fig. 7B, lower panel), which is in good agreement with the increased Rac1-GTP levels in these cells. Subsequent analysis of cell migration in a wound-healing assay (Lorenowicz et al., 2008) also showed that siRNA-mediated knockdown of PACSIN2 increased migration of HeLa cells (Fig. 7C).

We also tested the effect of PACSIN2 overexpression on GTP loading of Rac1. In line with the above data suggesting that PACSIN2 is a negative regulator of Rac1, we found an 80% reduction in the levels of Rac1-GTP in cells transfected with PACSIN2 (Fig. 7D). Expression of PACSIN2 Y435E/P478L, the SH3 domain mutant that is deficient in Rac1 binding, reduced Rac1-GTP levels by only ~35% compared with that by WT PACSIN2. This partial effect might result from dimerization with endogenous PACSIN2, which requires part of the F-BAR, rather than the SH3, domain (Fig. 7D). These data suggest that PACSIN2 association with Rac1 is important for efficient inactivation of Rac1. This effect was unrelated to the caveolin-1-dependent Rac1 ubiquitylation and degradation that we recently identified (Nethe et al., 2010), as loss or overexpression of PACSIN2 or its mutants did not affect the expression or the ubiquitylation of endogenous Rac1 (data not shown).

Because PACSIN proteins regulate endocytosis and vesicle transport, it is attractive to suggest that PACSIN2 negatively regulates Rac1 function by controlling internalization of activated Rac1 from the cell periphery. supplementary material Movie 1 shows that PACSIN2 is internalized in tubular structures. Expression of two separate F-BAR domain mutants of PACSIN2 (R50D and the double-mutant M124E/M125E), which can no longer induce membrane tubulation but still do associate with Rac1 (supplementary material Fig. S6A,B) (Shimada et al., 2010; Wang et al., 2009), induced a more than fourfold increase in GTP loading of Rac1 (Fig. 7E). This finding indicates a link between the formation of vesicular-tubular structures by PACSIN2 and Rac1 inactivation.

Dynamin has an important role in endocytosis mediated by its role in scission of newly formed vesicles from the membrane (Hinshaw, 2000). We therefore investigated whether inhibition of dynamin affects PACSIN2-associated regulation of Rac1. First, we studied the effects of dynamin inhibition on PACSIN2 dynamics by live-cell imaging studies. Fig. 8A and supplementary material Movie 3 show that upon addition of Dynasore, which inhibits dynamin by rapidly blocking coated vesicle formation (Macia et al., 2006), YFP–PACSIN2 accumulated on the plasma membrane and at peripheral vesicles. Furthermore, confocal studies showed that upon transfection of a dominant-negative mutant of dynamin (dynamin K44), endogenous PACSIN2 accumulated on tubules (Fig. 8B) with a concomitant loss of PACSIN2 localization to a distinct perinuclear pool (supplementary material Fig. S7A). Similar results were obtained for HeLa cells treated with Dynasore (supplementary material Fig. S7B). Taken together, these data indicate that inhibition of dynamin interferes with PACSIN2 localization and function. We therefore studied whether dynamin regulates PACSIN2-mediated downregulation of Rac1-GTP. Ectopic expression of dynamin K44 prevented the reduction in Rac1-GTP levels, as induced by Myc-tagged PACSIN2 (Fig. 8C). Furthermore, Dynasore transiently prevented PACSIN2-mediated downregulation of Rac1-GTP (Fig. 8D), suggesting that PACSIN2 regulates Rac1 inactivation in a dynamin-dependent fashion.

PACSIN2 could downregulate Rac1-GTP levels through two mechanisms. First, PACSIN2 could target Rac1 to specific sites to allow GAP-mediated inactivation. Alternatively, PACSIN2 could block Rac1 regulation by GEF proteins. To investigate this, we performed pull-down experiments with the biotinylated Pak1-CRIB domain using cell lysates of HeLa cells that had been transfected with either TrioD1 or Tiam1-C1199, GEF proteins known to activate Rac1 (Habets et al., 1994; van Buul et al., 2010), and studied whether endogenous Rac1 activation still occurred in the presence of PACSIN2. PACSIN2-mediated downregulation of Rac1-GTP was found in the control situation, as well as in the presence of GEF proteins. However, activation of Rac1 upon expression of the GEFs was detectable, even in the presence of PACSIN2 (Fig. 8E). These data show that PACSIN2 does not prevent Rac1 from being activated by GEF proteins and suggest that Rac1 inactivation occurs by its targeting to intracellular locations to allow GAP-stimulated GTP hydrolysis.

Collectively, these results suggest that PACSIN2 is an important regulator of the small GTPase Rac1 and that PACSIN2 limits Rac1-GTP signaling by promoting Rac1 inactivation. This effect of PACSIN2 requires its Rac1-binding capacity, mediated by the SH3 domain, as well as its membrane-tubulating activity, residing in the F-BAR domain.

Rho family GTPase signaling requires proper intracellular targeting of the activating GEF, the GTPase and effector proteins. For Rac1 GEFs, the relevant location is assumed to be at or near the plasma membrane, either in FAs or in peripheral membrane ruffles. A series of different signaling events upstream of exchange factor recruitment, such as inositol lipid turnover or kinase activation, have previously been identified and, as a result, this part of the Rac1 signaling pathway is relatively well established. By contrast, the mechanisms controlling the inactivation of Rac1, which needs to be equally well controlled in time and space, are less well understood. Here, we present data indicating that the F-BAR-protein PACSIN2, a regulator of endocytosis (Modregger et al., 2000), is part of a Rac1-inactivation pathway. Our data suggest that PACSIN2-regulated internalization of peripheral, Rac1-containing membrane protrusions downregulates Rac1 signaling, as reduction of PACSIN2 expression by siRNA promotes GTP loading of Rac1, cell spreading and migration. PACSIN2 does not localize to FAs, which suggests that inactivation of Rac1 at these sites requires other mechanisms. These could involve caveolin-1, which is recruited to FAs by activated Rac1 and regulates Rac1 ubiquitylation and degradation (Nethe et al., 2010).

The interaction with PACSIN2 is specific for Rac1 and is the first described association of PACSIN2 with a member of the Rho GTPase family. The PACSINs are not the only BAR domain proteins with Rac1-binding capacity. Previously, arfaptin, an effector of the Arf GTPase, was found to interact with GDP-bound, as well as GTP-bound, Rac1 through its BAR domain (Tarricone et al., 2001). The adapter protein IRSp53 (also known as brain-specific angiogenesis inhibitor 1-associated protein 2) interacts with activated Rac1 through its I (inverted)-BAR domain linking Rac1 to WAVE, which stimulates actin polymerization by the Arp2/3 complex (Miki et al., 2000). Thus, whereas the I-BAR protein IRSp53 might stimulate Rac1 signaling and protrusion, the F-BAR protein PACSIN2, which stimulates endocytosis, inhibits Rac1 signaling.

Through their F-BAR domain, PACSIN proteins bind membranes containing phosphatidylserine and phosphatidylinositol (4,5)-bisphosphate and induce membrane curvature, which results in the formation of invaginations and, subsequently, of vesicular-tubular structures which depend on the self-assembly of F-BAR modules into a helical coat (Frost et al., 2009). We show that upon collapse of membrane ruffles, PACSIN2 associates with these tubular structures, in line with the proposed role of PACSIN2 in endocytosis. In addition, PACSIN2 localizes to early endosomes. At tubules, as well as early endosomes, PACSIN2 colocalizes with Rac1. Our experiments using different markers for endocytosis indicate that PACSIN2 does not play a primary role downstream of clathrin-mediated endocytosis. The CtxB-labeling studies, however, indicate that PACSIN2 colocalizes with internalized lipid rafts in the vesicular-tubular structures. Given that Rac1 associates with PACSIN2-positive tubules, these data are in agreement with findings showing that Rac1 associates to cholesterol-rich membrane domains and that these domains are involved in internalization of active Rac1 (del Pozo et al., 2004).

We show that activation of endogenous Rac1 with CNF1 toxin or expression of a constitutively active mutant of Rac1 (Q61L) results in loss of PACSIN2-positive tubular structures, leaving PACSIN2 on early endosomes. Although the molecular basis for this is unknown, this finding suggests that fission of the vesicular-tubular structures, for example, mediated by dynamin (Hinshaw, 2000), is stimulated by Rac1 activity. Previously, constitutively active Rac1 has been found to inhibit transferrin-receptor-mediated endocytosis (Lamaze et al., 1996). Similarly, our findings show that activated Rac1 reduces the number of PACSIN2-positive tubules, which could reflect a reduction in PACSIN2-mediated endocytosis. In line with the observation that active Rac1 promotes PACSIN2 distribution to the perinuclear area and reduces the number of PACSIN2-positive tubules, inhibition of Rac1 results in an increased number of PACSIN2-tubular structures, as induced by a dominant-negative Rac1 mutant (T17N) and inhibition of endogenous Rac1 by a pharmacological inhibitor (EHT1864). Collectively, our data suggest that Rac1 activity regulates the localization of PACSIN2 and the abundance of PACSIN2-positive tubular structures.

We found that the differential PACSIN2 distribution into a peripheral and a perinuclear pool is regulated by MTs. In this respect, PACSIN2 behaves in a manner similar to amphiphysin-2, (Meunier et al., 2009), the distribution of which is also regulated by MTs. MT growth has been claimed to activate Rac1, but it is currently unclear to what extent this is directly linked to the regulation of Rac1 by PACSIN2. Our results indicate that, in epithelial cells, the MT network controls PACSIN2-mediated tubule formation and PACSIN2 distribution. PACSIN proteins are linked to the actin-regulating Arp2/3 complex through their binding to N-WASP (Qualmann and Kelly, 2000; Rohatgi et al., 1999). However, experiments using cytochalasin B to block actin polymerization were inconclusive, because cytochalasin B induced morphological changes in the cell, precluding proper analysis of PACSIN2 distribution (data not shown). Thus, the molecular mechanisms that underlie the Rac1-induced changes in the distribution of PACSIN2 remain to be established.

A key finding of the current study is that PACSIN2 is involved in the inactivation of Rac1 in a fashion that requires an intact F-BAR domain. As the F-BAR domain does not regulate the PACSIN2–Rac1 interaction, this suggests that the membrane-binding and/or tubulating activity of PACSIN2 is essential to inactivate Rac1. In addition, we show that inhibition of dynamin impairs PACSIN2 function and localization and prevents PACSIN2-mediated Rac1 inactivation. Together, these data suggest a model in which PACSIN2 regulates Rac1 by promoting its internalization. Interestingly, dynamin, which can associate with PACSIN2 (Kessels and Qualmann, 2004), was previously shown to regulate Rac1 (Schlunck et al., 2004). The dominant-negative dynamin K44 mutant was found to increase the formation of Rac1-positive tubular structures. Given that dynamin induces fission of tubules, generating intracellular vesicles, such as early endosomes, dynamin might act downstream of PACSIN2 in a Rac1-inactivation pathway.

In contrast to caveolin-1 (Nethe et al., 2010), we found no evidence for PACSIN2 targeting Rac1 to an ubiquitylation and/or degradation pathway. Our finding that GEF proteins are still able to activate Rac1 in the presence of PACSIN2, indicates that PACSIN2-associated Rac1 inactivation primarily relies on GAP proteins. Because the inactive Rac1 T17N mutant colocalizes to PACSIN2-positive tubules, it could well be that PACSIN2 contributes to targeting Rac1 to sites for GAP-mediated inactivation. One such GAP could be p50RhoGAP (also called Cdc42GAP and Rho GTPase-activating protein 1), which has GAP activity towards Rho, Cdc42 and Rac1 (Barfod et al., 1993; Lancaster et al., 1994) and has been found to localize to endosomal structures (Sirokmany et al., 2006). Furthermore, several BAR-domain-containing RhoGAPs exist, such as RICH1, which has GAP activity for Cdc42 and Rac1. RICH1 binds the F-BAR-domain protein CIP4, which is structurally similar to the PACSIN proteins (Richnau et al., 2004; Richnau and Aspenstrom, 2001). We are currently pursuing this issue to determine the localization and relevance of selected Rac GAPs, although these studies are complicated by the number of GAPs (Bernards, 2003) and the limited availability of sufficient well-characterized reagents.

We recently identified a series of other adapter proteins that interact with the C-terminal hypervariable domain of Rac1. These include CD2AP (CMS) and caveolin-1 (Nethe et al., 2010; van Duijn et al., 2010) that also, like PACSIN2, regulate endocytosis (Lynch et al., 2003; Parton and Richards, 2003). However, whereas CD2AP is recruited to cell–cell contacts upon Rac1 activation, caveolin-1 translocates to FAs. Thus, these proteins all reside at different intracellular locations and they all interact with Rac1 independently of each other (data not shown). These findings indicate that independent mechanisms regulate Rac1 signaling at different sites supporting the notion of parallel, but compartmentalized, Rac1 signaling. In these so-called ‘spatio-temporal signaling modules’ (Pertz, 2010), GTPases can interact specifically with different regulators and effectors, securing localized signaling.

In conclusion, our previous and current data suggest the following model (Fig. 9). Following GEF-mediated activation and interaction with effectors at the plasma membrane, Rac1 signaling can be terminated through at least two independent pathways. First, we recently showed that caveolin-1 can target Rac1-GTP for ubiquitylation and degradation in an adhesion-dependent fashion (Nethe et al., 2010). Second, we have shown that PACSIN2 promotes Rac1 inactivation, possibly by targeting Rac1 to sites where GAP proteins can stimulate GTP hydrolysis. This pathway is not associated with Rac1 degradation. Future experiments will be aimed at defining the subcellular locations to which PACSIN2 recruits Rac1 for its inactivation and the identification of the relevant RacGAPs.

Antibodies against the following proteins were used: PACSIN1 (Modregger et al., 2000); PACSIN2 (AP8088b, Abgent); GFP (632381, Clontech) and dsRed (8374-1, Clontech); HA (11867423001, Roche); Rac1 (610651, Transduction Laboratories), dynamin (610245, Transduction Laboratories), and paxillin (610620, Transduction Laboratories); Rac1 (05-389, Upstate Biotechnologies); β-Pix (AB3829, Millipore); Anti-SET/I2PP2A (SC-25564, Santa Cruz Biotechnology); α-actinin1 (A5044, Sigma), α-tubulin (T6199, Sigma); and actin (A3853, Sigma). F-Actin was detected using Texas-Red- or Alexa-Fluor-633-labeled phalloidin (Invitrogen). Secondary HRP-labeled antibodies for western blotting were from Pierce. Secondary Alexa-Fluor-labeled antibodies for immunofluorescence were from Invitrogen.

Dynasore (D7693, used at 80 μM at 37°C), nocodazole (M1404, used at 10 μM at 37°C) and EHT1864 (E1657; used at 50 μM at 37°C) were purchased from Sigma. GST–CNF1 was isolated as previously described (Pop et al., 2004) and used at 300 ng/ml. AF-594-CtxB (C22842; used at 0.5 μg/ml at 37°C), transferrin Texas Red (T2875; used at 20 μg/ml at 37°C), and Sulforhodamine 101 (S359; used at 25 μM at 37°C) were purchased from Molecular Probes.

To generate YFP–PACSIN2 and mCherry–PACSIN2, Myc–PACSIN2 WT was used as a template for PCR using, as a forward primer, 5′-GAGATCGGTACCTCTGTCACCTACGATGACTCT-3′ and, as a reverse primer, 5′-GAGATCGGATCCTCACTGGATAGCCTCGAC-3′. The product was cloned into pEYFP(C1) or pmCherry(C1) (Clontech Laboratories) using the KpnI and BamHI restriction sites. Myc- and YFP-tagged PACSIN2 R50D and M124E/M125E were generated by site-directed mutagenesis using Myc-PACSIN2 WT and pEYFP-PACSIN2 WT as templates for PCR. The R50D mutant was generated using, as a forward primer, 5′-GCATGAGCGGGCGGACATCGAGAAGGCG-3′ and, as a reverse primer, 5′-CGCCTTCTCGATGTCCGCCCGCTCATGC-3′. The M124E/M125E mutant was generated using, as a forward primer, 5′-CTTTCACAAGCAGGAGGAGGGCGGCTTCAAGGAGACCAAG-3′ and, as a reverse primer, 5′-CTTGGTCTCCTTGAAGCCGCCCTCCTCCTGCTTGTGAAAG-3′. To generate GST–PACSIN2, Myc-PACSIN2 WT was used as template for PCR using, as a forward primer, 5′-GAGATCGGATCCTCTGTCACCTACGATGACTCT-3′ and, as a reverse primer, 5′-GAGATCCTCGAGTCACTGGATAGCCTCGAC-3′. The product was cloned into pGex-6p-1 using BamHI and XhoI restrition sites.

To generate YFP and mCherry Rac1 fusions, Myc-Rac1 WT, Q61L, GV12, and T17N were used as templates for PCR using, as a forward primer, 5′-GATCCTCGAGTTCAGGCCATCAAGTGTGTG-3′ and, as a reverse primer, 5′-TTACAACAGCAGGCATTTTCTC-3′. The product was then cloned into pEYFP(C1) or pmCherry(C1) (Clontech Laboratories) using KpnI and EcoRI restriction sites. All fusion constructs and point mutations were confirmed by sequencing.

GST–Rac1 WT and GST–Rac1 ΔC were as described previously (ten Klooster et al., 2006). GFP–FYVE was a gift from Harald Stenmark (Department of Biochemistry, University of Oslo, Norway). Myc–Rab5WT and Myc–Rab5Q79L were a gift from Jacques J. Neefjes and pcDNA3+-DynaminK44 and pcDNA3-Tiam1-C1199 were a gift from John Collard (Netherlands Cancer Institute, Amsterdam, The Netherlands). pEGFP-TrioD1 was a gift from Jaap van Buul (Sanquin Research, Amsterdam, The Netherlands).

For siRNA-based knockdowns, the sequence for control siRNA was: 5′-CGUACGCGGAAUACUUCGAtt-3′ (Eurogentec). The sequence for PACSIN2 siRNA-1 was: 5′-GGAGAAGCUGGCUAUCUCACGAGAAtt-3′ (Eurogentec). PACSIN2 siRNA-2 (L-019666-02) was from Dharmacon. PACSIN2 siRNA-3 (SI02224292) was from Qiagen. Rac1 siRNA was as previously described (Nethe et al., 2010).

MEF, COS7, Jurkat and HeLa cells were maintained at 37°C under 5% CO₂ in Iscove's Modified Dulbecco's Medium (IMDM; BioWhittaker) containing 10% heat-inactivated fetal calf serum (FCS; Life Technologies, Breda, The Netherlands), 300 μg/ml glutamine, 100 units/ml penicillin and streptomycin. Cells were passaged by trypsinization. Primary HUVECs were purchased from Lonza and cultured in EGM2 medium, supplemented with singlequots (Lonza).

HeLa cells were transiently transfected with FuGENE (Roche) or TransIT (Mirus) according to the manufacturers' recommendations. Transfections of siRNA were performed with INTERFERin (Polyplus-transfection) according to the manufacturer's recommendations.

Peptide pull-downs were performed as described previously (ten Klooster et al., 2006). In short, cells were lysed in NP-40 lysis buffer (50 mM Tris-HCl pH 7.5, 100 mM NaCl, 10 mM MgCl₂, 10% glycerol and 1% NP-40) supplemented with protease inhibitors (Complete mini EDTA, Roche), centrifuged at 20,000 g for 10 minutes at 4°C. The supernatant was then incubated with the indicated Rho GTPase C-terminal peptides (5 μg) in the presence of streptavidin-coated beads (Sigma) at 4°C for 1 hour with rotation. Protein association was assayed by western blotting. Isolation of proteins and mass spectometry analysis was performed as described (Kanters et al., 2008). GST-fusion proteins were purified from BL21 bacteria as described previously (ten Klooster et al., 2006); 50 μg of the indicated GST-fusion constructs was used for each pull-down. Protein association was assayed by western blotting. For studying the direct interaction of Rac1 and PACSIN2, the GST–PACSIN2 fusion protein was eluted from the beads with a glutathione buffer (10 mM glutathione and 50 mM Tris-HCl pH 7.4) twice for 10 minutes (while rotating) at room temperature. Protein association was assayed by western blotting. Rac1 activation was assayed by a CRIB-peptide pull-down approach as described previously (Price et al., 2003); 30 μg of Pak1-CRIB peptide was used for each pull-down. Bound Rac1-GTP was detected by Western blot analysis.

Proteins were separated by SDS-PAGE (on 7.5, 10 or 12.5% polyacrylamide gels depending on the size of the proteins of interest) and transferred onto nitrocellulose transfer membrane (Whatman). Following blocking in 5% low-fat milk in TBST (tris-buffered saline Tween-20) the blots were incubated with the primary antibody overnight at 4°C. Next, the blots were washed three times for 10 minutes in TBST and subsequently incubated with HRP-coupled secondary antibodies (dilution 1:5000) in TBST for 1 hour at room temperature. Finally, blots were washed three times with TBST for 20 minutes each and subsequently developed by ECL (GE Healthcare)

Cells were seeded onto fibronectin-coated glass coverslips, transfected with the indicated plasmids and after 24 hours fixed with 3.7% formaldehyde (Merck) in PBS for 10 minutes and permeabilized with 0.5% Triton X-100 in PBS for 5 minutes. Coverslips were then incubated for 15 minutes with 2% BSA at 37°C. Immunostainings were performed at room temperature for 1 hour with the indicated antibodies. Fluorescent imaging was performed with a confocal laser-scanning microscope (LSM510/Meta; Carl Zeiss MicroImaging) using a 63× NA 1.40 or a 40× NA 1.30 oil lens (Carl Zeiss MicroImaging). Image acquisition was performed with Zen 2008 software (Carl Zeiss MicroImaging). For live-cell imaging, cells were seeded on fibronectin-coated glass coverslips, transfected with the indicated plasmids. After 24 hours, fluorescent imaging was performed.

For ECIS-based cell spreading experiments, gold ECIS electrodes (8W10E, Applied Biophysics) were treated with 10 μM L-cysteine for 15 minutes and subsequently coated with 10 μg/ml fibronectin in 0.9% NaCl for 1 hour at 37°C. Next, HeLa cells were treated as indicated, and seeded at a concentration of 100,000 cells per well in 400 μl IMDM with 10% FCS. Impedance was measured continuously at 45 kHz using ECIS model 9600. The increase in impedance, as a measure of cell spreading (Wegener et al., 2000), was recorded for 1 hour. When cells had formed a monolayer displaying stable transmonolayer resistance, electric wounding was performed (45 kHz and 4 V). The increase in resistance in the first hours after the wounding, as a measure of cell migration by wound healing, was subsequently recorded for up to 5 hours (Lorenowicz et al., 2008).

The authors thank Harald Stenmark for GFP-FYVE, J. J. Neefjes for Myc–Rab5WT and Q79L, John Collard for Tiam1-C1199, Jaap van Buul for Trio-D1 and Gudula Schmidt for CNF1 toxin and CNF1-encoding plasmid. This work was supported by an LSBR grant to P.L.H. (#0731).