Antagonistic regulation of F-BAR protein assemblies controls actin polymerization during podosome formation

Dynamic remodeling of the cell membrane is mediated by the cooperative actions of membrane-molding proteins and the actin cytoskeleton (McMahon and Gallop, 2005). The BAR (Bin, amphiphysin, Rvs) domain superfamily has emerged as a crucial interface that links actin polymerization/depolymerization to membrane morphogenesis in a wide variety of fundamental cellular processes including membrane trafficking, cell migration, division and adhesion (Frost et al., 2009; Itoh and De Camilli, 2006). The BAR domain is typically divided into three subfamilies: classical BAR, FCH and BAR (F-BAR)/extended FCH (EFC) and Inverse-BAR (I-BAR), which share an overall structure composed of antiparallel coiled-coil dimers (Peter et al., 2004; Shimada et al., 2007; Suetsugu et al., 2006; Zhao et al., 2011). Biochemical and structural studies have indicated that the BAR domains probably form high-order oligomers upon membrane binding, which is believed to cause robust tubulation of the lipid bilayer (Frost et al., 2009; Itoh and De Camilli, 2006). Direct observations using cryo-electron microscopy have demonstrated that the F-BAR domain of FBP17 is capable of self-polymerizing into filaments, which adhere to the flat bilayer sheets and form a spiral protein coat around the tubulated membrane that they then induce (Frost et al., 2008; Shimada et al., 2007). This self-assembly of the F-BAR domain is proposed to be mediated by interaction with the membrane in a side-lying state followed by the formation of intermolecular associations, including tip-to-tip contacts and lateral interactions, to drive membrane tubulation (Frost et al., 2008).

Most of the BAR proteins contain SH3 domains that create a complex with two key components required for membrane dynamics that are regulated by the actin cytoskeleton: dynamin, the membrane fission protein and WASP family proteins, the actin nucleation promoting factor for Arp2/3 complex (Itoh et al., 2005; Takenawa and Suetsugu, 2007; Tsujita et al., 2006). These observations suggest that the BAR superfamily proteins not only generate membrane curvature, but also create a high-order scaffold on membrane surfaces that promotes membrane fission and/or localized actin polymerization. Indeed, Toca subfamily members of the F-BAR proteins (Toca-1, FBP17 and CIP4), which were originally identified as essential factors for N-WASP-mediated actin nucleation (Ho et al., 2004), have been shown to potently activate actin polymerization at the membrane (Takano et al., 2008). In addition, a recent study has reported that dimerization/oligomerization of WASP is important for its activation (Padrick et al., 2008), suggesting that F-BAR-mediated assembly at the plasma membrane is crucial for WASP-mediated actin polymerization. However, because the F-BAR domain of Toca proteins appears to possess the powerful membrane-deforming ability that would lead to the hyperactivation of WASP proteins, it is conceivable that their assembly needs to be tightly controlled and that the dysfunction of such regulation is obviously deleterious to cells. Recently, autoinhibitory mechanisms have been proposed for the regulation of BAR proteins (Eberth et al., 2009; Rao et al., 2010; Roberts-Galbraith et al., 2010). Such constraints are relieved by interaction with their binding partners or by dephosphorylation, promoting their oligomerization through BAR domains (Rao et al., 2010; Roberts-Galbraith et al., 2010). In light of the ability of the BAR domain to form oligomers for the generation and sensing of membrane curvature (Frost et al., 2009; Itoh and De Camilli, 2006), there is an intriguing possibility that each BAR protein functions in collaboration with other BAR proteins. Such collaborative action would promote or modulate their abilities to influence membrane morphology, ensuring a tight control of plasma membrane invagination and fission, driven by localized actin polymerization. Yet, the existence of a coordination between different BAR domain proteins also remains to be discovered.

Podosomes/invadopodia are highly dynamic adhesive actin-based structures with enrichment of matrix metalloproteases (MMPs) activity formed at the ventral surface of the cell body, which are seen in macrophages, osteoclasts, dendritic cells and some cancer cells (Albiges-Rizo et al., 2009; Murphy and Courtneidge, 2011). Therefore, they are thought to play crucial roles in extracellular matrix (ECM) degradation, the level of which must be strictly maintained; otherwise it may result in an increased inflammation or cancer metastasis (Kessenbrock et al., 2010). Podosomal actin structures are formed by the WASP/N-WASP-activated Arp2/3 complex at the plasma membrane (Albiges-Rizo et al., 2009; Murphy and Courtneidge, 2011). Recently, Toca proteins have been identified as key players in podosome/invadopodia formation (Chander et al., 2012; Hu et al., 2011a; Hu et al., 2011b; Linder et al., 2000; Pichot et al., 2010; Tsuboi et al., 2009), through their self-assembling property for the activation of WASP-mediated actin polymerization. However, it is unclear how the level of Toca protein assembly is determined for proper formation and dynamic turnover of podosomes. For this, a factor that counteracts the Toca protein assembly at the membrane should be involved.

In this study, we demonstrate that PSTPIP2, a macrophage-specific protein composed solely of an F-BAR domain suppresses the assembly of FBP17. Importantly, it has been known that the depletion of PSTPIP2 leads to auto-inflammatory disorder (Chitu et al., 2009; Grosse et al., 2006). Knockdown of PSTPIP2 in macrophages enhances the assembly of FBP17, leading to ‘hyperactivation’ of actin nucleation at podosomes and ECM degradation. Moreover, using in vitro reconstitutions as well as time-lapse imaging on acute pharmacological manipulations, we show that the assembly of FBP17 is antagonized by PSTPIP2 for the proper control of actin cytoskeleton during podosome formation.

Among F-BAR proteins, PSTPIP2 has a unique property owing to its primary structure composed solely of an F-BAR domain with membrane tubulation ability and no other functional domain modules (Tsujita et al., 2006). PSTPIP2 is mostly expressed in macrophages, but also in osteoclasts, mast cells and myeloid progenitor cells (Chitu et al., 2009; Chitu et al., 2012). It has been also reported that depletion of PSTPIP2 leads to a development of autoinflammatory disease in mouse models that resembles chronic recurrent multifocal osteomyelitis (CRMO) in humans (Chitu et al., 2009; Ferguson et al., 2006; Grosse et al., 2006). These features of PSTPIP2 led us to investigate its physiological function by RNA interference (RNAi) in Raw 264.7 cells, a macrophage cell line. In control cells, typical podosome structures were observed on fibronectin coated dish; intensive actin polymerization was limited to the F-actin core, which was surrounded by a podosome marker, vinculin (Fig. 1A, upper panel and supplementary material Fig. S1A). When PSTPIP2 was knocked down, abnormal podosome structures containing highly polymerized actin clusters were formed (approximately fourfold; Fig. 1A, bottom, B and supplementary material Fig. S1A). Similar result was obtained using siRNA with different target sequence (Fig. 1B). The aberrant structures also contained vinculin, but F-actin was broadly developed not only in the F-actin core but also in the surrounding area (Fig. 1A). We further performed a time-lapse imaging by TIRF microscopy to investigate a spatiotemporal dynamics of podosomes in control and PSTPIP2-depleted cells. Dynamics of podosomes in Raw 264.7 cells were monitored by coexpressing Lifeact–mCherry, an F-actin marker (Riedl et al., 2008). Consistent with previous observations (Albiges-Rizo et al., 2009), respective podosome in control siRNA-treated cells exhibited dynamic assembly and disassembly cycles (supplementary material Fig. S1B and Movie 1). Interestingly, the aberrant podosome induced by PSTPIP2 knockdown is also a highly dynamic structure with intense actin clusters that behave like an enlarged podosomal core as a whole (supplementary material Fig. S1C and Movie 2), suggesting that actin polymerization is hyperactivated at podosomes by the depletion of PSTPIP2. In addition, we also found that a robust membrane ruffling was induced in PSTPIP2-depleted cells (supplementary material Fig. S2A,B) as previously reported (Chitu et al., 2005), suggesting that PSTPIP2 is a negative regulator of actin polymerization.

Podosome formation is thought to be dependent on WASP-mediated actin nucleation (Albiges-Rizo et al., 2009; Murphy and Courtneidge, 2011). Consistent with this, abnormal podosome formation by PSTPIP2 knockdown was effectively suppressed by treatment with a WASP inhibitor, wiskostatin (Peterson et al., 2004) (supplementary material Fig. S2C), indicating that PSTPIP2 plays a negative role in WASP-dependent processes. Because Toca proteins are the upstream activators of WASP, we next questioned whether abnormal actin nucleation induced by the depletion of PSTPIP2 could be due to hyperactivation of Toca proteins. Based on the result of RT-PCR analysis showing FBP17 and CIP4, but not Toca-1, are expressed in Raw 264.7 cells (supplementary material Fig. S2D), we chose these two proteins in our experiments. As shown in Fig. 1C, TIRF microscopy revealed that FBP17 formed small patches and was accumulated around F-actin cores. These signals completely disappeared when FBP17 was knocked down (supplementary material Fig. S2E). Importantly, the accumulation of FBP17 was broadly enhanced around intense actin clusters in PSTPIP2-depleted cells (Fig. 1C bottom, D). Endogenous CIP4 protein also accumulated at the aberrant podosomes (supplementary material Fig. S2F), suggesting that high accumulation of Toca proteins in PSTPIP2-depleted cells contributes to abnormal activation of actin polymerization. To confirm this, we examined whether this aberrant podosomes phenotype can be rescued by further depletions of Toca proteins. Consistent with a previous study (Tsuboi et al., 2009), when FBP17 or CIP4 were knocked down, spontaneous formation of podosomes was reduced (data not shown). This effect was enhanced by simultaneous depletion of both proteins (∼66% decrease; Fig. 1B,E; supplementary material Fig. S1A), suggesting their essential roles in WASP activation. Interestingly, membrane ruffling was also inhibited (supplementary material Fig. S2A,B), indicating that FBP17 and CIP4 are also implicated in this process as reported for Toca-1 (Hu et al., 2011a). Notably, the aberrant podosomes and robust membrane ruffling induced by PSTPIP2 knockdown were significantly suppressed by the depletion of Toca proteins (Fig. 1B,E; supplementary material Fig. S1A; Fig. S2A,B), suggesting general antagonistic actions between PSTPIP2 and Toca family proteins on WASP-dependent actin polymerization at podosomes.

To characterize the importance of the F-BAR domain of PSTPIP2 on the regulation of podosome formation, we performed rescue experiments using an F-BAR domain mutant of PSTPIP2 (PSTPIP2 R168A), the analogous mutant of FBP17 K166A, with defective self-assembling ability (Shimada et al., 2007). We confirmed that this mutant is unable to induce tubular invaginations of the plasma membrane in COS-1 cells, a kidney epithelial cell line, which has been used to visualize membrane tubulation by overexpression of F-BAR proteins (Itoh et al., 2005; Tsujita et al., 2006) (supplementary material Fig. S2G). Expression of wild-type PSTPIP2 (PSTPIP2 WT) in PSTPIP2-knocked down Raw 264.7 cells completely suppressed aberrant podosome formation, whereas PSTPIP2 R168A failed to do so (Fig. 2A,B). In addition, overexpression of PSTPIP2 led to the inhibition of podosome formation (∼62.5% decrease; Fig. 2C,D). In contrast, no inhibitory effect was observed in cells overexpressing PSTPIP2 R168A (Fig. 2C,D), indicating that the self-assembling ability of the F-BAR domain is essential for the negative effect of PSTPIP2 on actin polymerization during podosome formation.

Given the functional relevance between podosome formation and matrix degradation, we examined whether aberrant podosomes enhance matrix degradation. Compared with control RNAi-treated cells, PSTPIP2-depleted cells exhibited a significant enhancement of the degradation of gelatin matrix overlaid with fibronectin (supplementary material Fig. S2H). These data suggest that PSTPIP2-depleted macrophages have a potent proteolytic activity because of the enhancement of podosome formation.

Given that FBP17 and PSTPIP2 play opposite roles in podosome formation, we examined their spatiotemporal behaviors at the podosomes by TIRF microscopy. GFP–FBP17 or GFP–PSTPIP2 expressed in Raw 264.7 cells were monitored with coexpressed Lifeact–mCherry. Consistent with endogenous staining of FBP17, time-lapse TIRF microscopy showed that fluorescent signals of GFP–FBP17 accumulated at podosomes as small clusters, and ∼72% of these patches (n = 90 from five cells) were associated with the F-actin structure (Fig. 3A top panels). Detailed image analysis revealed that the small clusters of GFP–FBP17 were formed at podosomes concomitant with nascent actin assembly (Fig. 3B arrowheads, C), which is reminiscent of the clustering patterns of integrins and vinculin (Fig. 1A) (Albiges-Rizo et al., 2009; Calle et al., 2006). The behavior of GFP–FBP17 on the plasma membrane was highly dynamic, apparently assembling around the F-actin core (supplementary material Movie 3). When overexpressed, GFP–FBP17 spots were numerous on the podosomes, leading to hyperactivation of actin nucleation (Fig. 3A bottom panels). Interestingly, overexpression of FBP17 tended to convert it from its ring-like structures to broad pattern that is similar to the situation induced by PSTPIP2 knockdown (Fig. 3A bottom; Fig. 1C), supporting opposite roles of the two F-BAR proteins. In addition, accumulation of GFP–FBP17 K166A at the podosomes was less (supplementary material Fig. S3), emphasizing the essential role of the self-assembling property of the F-BAR domain for clustering at the podosome. These data indicate that FBP17 assembles at the podosomal membrane, initiating actin nucleation for podosome formation.

In contrast, TIRF microscopy revealed that moderate expression of PSTPIP2 resulted in only a limited accumulation around the podosome core; only 6.8% of the GFP–PSTPIP2 patches (n = 55 from three cells) was associated with F-actin during podosome formation (Fig. 3D). When overexpressed, GFP–PSTPIP2 was clearly assembled around small actin dots (Fig. 3E arrowhead). Notably, time-lapse imaging showed that the assembly of PSTPIP2 correlated well with the decrease of actin polymerization (Fig. 3E,F). Together, these findings indicate that PSTPIP2 plays a negative role in actin polymerization during podosome formation.

The opposite effects of FBP17 and PSTPIP2 on actin polymerization at podosomes prompted us to investigate whether their assemblies compete with each other on the membrane surface. To directly address this possibility, we first attempted to visualize the behavior of FBP17 on liposomes. A GFP-tagged FBP17 recombinant protein was prepared, and its self-assembling ability was confirmed by native PAGE in the absence or presence of liposomes (supplementary material Fig. S4A).

Next, we examined a self-assembling ability of GFP–FBP17 by incubating it with fluorescently labeled liposomes with diameters of ∼1–5 µm by confocal microscopy. GFP alone (200 nM) did not result in any fluorescence signals on the membrane (supplementary material Fig. S4B), but GFP–FBP17 (200 nM) was assembled on liposome surfaces in a homogeneous manner (92%, n = 300 liposomes; supplementary material Fig. S4B middle panels). The clear localization of GFP–FBP17 at the surface of globular liposomes indicates that it can assemble on relatively flat membrane surfaces, probably in its side-lying state (Frost et al., 2008) (supplementary material Fig. S4B middle). At a lower concentration (50 nM), punctate or crescent-shaped fluorescence signals on rounded liposome surfaces were observed (supplementary material Fig. S4B,top, arrowheads). At a high concentration (600 nM), robust membrane tubules decorated with GFP–FBP17 were observed (supplementary material Fig. S4B bottom).

Next, we asked whether PSTPIP2 competes with FBP17 on liposome surfaces. When increasing concentrations of PSTPIP2 were added, the fluorescence intensity of GFP–FBP17 on liposomes was significantly reduced (Fig. 4A,B). Similar to those observed at a lower concentration (supplementary material Fig. 4B top), GFP–FBP17 at a physiological concentration (200 nM) distributed heterogeneously on spherical liposomes or membrane tubules induced by PSTPIP2 (83% in 600 nM PSTPIP2, n = 200 liposomes; Fig. 4A bottom panels, arrowheads, compared with those indicated by arrows). GFP–FBP17 was hardly detected on tubular liposomes probably induced by PSTPIP2 (Fig. 4A bottom). Our immunoprecipitation analysis showed that FBP17 and PSTPIP2 did not form a heterodimer, supporting their exclusive behavior (supplementary material Fig. S4D). It is also worth noting that the fluorescence signals of GFP–FBP17 did not decrease uniformly on liposomes, and fully or partially assembled patterns were also observed without changing their intensities. These observations further support the cooperative fashion of F-BAR assembly on membranes.

Actin polymerization on liposomes was also visualized using Rhodamine-labeled G-actin. As has been reported by many studies, N-WASP, a ubiquitous homolog of WASP, and Arp2/3 complex induce actin polymerization from certain areas on the liposome surface in vitro (supplementary material Fig. S4E top). When liposomes were precoated with FBP17 (200 nM) and then incubated with N-WASP (200 nM) and Arp2/3 complex (50 nM), intense actin assembly was observed from tubulated as well as spherical liposomes (Fig. 4C top, arrowheads; supplementary material Fig. S4E bottom), reflecting the high activity of FBP17 in promoting N-WASP-mediated actin polymerization on the membrane. In the absence of liposomes, FBP17 only had a low capacity to activate actin polymerization (data not shown). These data strongly suggest that self-assembly of FBP17 on the membrane surface promotes actin polymerization by WASP proteins. Strikingly, concomitant preincubation of FBP17 (200 nM) and PSTPIP2 (600 nM) with liposomes followed by the addition of N-WASP and the Arp2/3 complex clearly showed a potent inhibitory effect of PSTPIP2 on FBP17-dependent actin polymerization at the membrane (Fig. 4C,D). Importantly, PSTPIP2 itself did not inhibit N-WASP-dependent actin polymerization because branched actin filaments were diffusely present on the glass surface (data not shown). These data demonstrate that the assembly of FBP17 is antagonized by PSTPIP2 in a competitive manner, resulting in the inhibition of actin polymerization at the membrane surface.

The exclusive assemblies of FBP17 and PSTPIP2 on membranes in vitro suggested that both proteins could compete with each other in vivo. To examine this possibility, we first expressed GFP–FBP17 with or without PSTPIP2 in COS-1 cells. Overexpression of GFP–FBP17 in COS-1 cells induced a robust tubular invagination of the plasma membrane (Fig. 5A top panels, a cell on the right). When coexpressed with Myc–PSTPIP2, the FBP17-dependent tubule formation was significantly inhibited (Fig. 5A top, a cell on the left, and B). In control experiments, no competitive effect was observed when the full length FBP17 was coexpressed with the F-BAR-domain-only construct of FBP17 (1–340 aa); they were clearly colocalized on tubulated membranes (supplementary material Fig. S5). Interestingly, in 29% of cases (Fig. 5B), FBP17 and PSTPIP2 still showed tubular patterns decorated with one of the two F-BAR proteins in an exclusive manner (Fig. 5C arrow), or they were partially segregated (arrowhead), reflecting their heterogeneous assembly on the membrane observed in vitro (Fig. 4). Importantly, this effect was completely due to the polymerizing ability of F-BAR domains because the PSTPIP2 K168A mutant had no effect on FBP17-induced membrane tubulation in COS-1 cells (Fig. 5A bottom, B).

The competitive assembly of FBP17 and PSTPIP2 was also confirmed during podosome formation in Raw 264.7 cells. Using TIRF microscopy, we found that both GFP–PSTPIP2 and tagRFP–FBP17 were partially colocalized at the same clusters (Fig. 5D,E). However, the fluorescence intensities of the two proteins were inversely correlated (Fig. 5D,E), strongly supporting their competition. At higher magnifications, it was also observed that GFP–PSTPIP2 and tagRFP–FBP17 signals were partially segregated at patches or tubular structures in Raw 264.7 cells (Fig. 5F). These data indicate that the assemblies of FBP17 and PSTPIP2 are antagonized by each other at the plasma membrane, probably modulating the activity of FBP17 to promote WASP-dependent actin polymerization during podosome formation.

If the observed antagonism between FBP17 and PSTPIP2 is physiologically relevant, the acute dissociation of FBP17 could affect the self-organizing property of PSTPIP2 at podosomes. A recent study proposed that membrane binding/bending activity of the F-BAR domain of syndapin 1/Pacsin 1 is autoinhibited by an intramolecular interaction with its C-terminal SH3 domain, and the inhibition is relieved by a binding partner of the SH3 domain, dynamin (Rao et al., 2010). This knowledge prompted us to examine if dynamin and WASP, which bind to the SH3 domain of FBP17, are also involved in the regulation of membrane-binding ability of FBP17, and therefore its assembly at podosomes. When Raw 264.7 cells expressing GFP–FBP17 were treated with the dynamin inhibitor Dynasore (Macia et al., 2006), GFP–FBP17-positive podosomal clusters were apparently stabilized by losing their dynamics of formation and dissociation (supplementary material Fig. S6A,B). This is very similar to the observation that BAR domain proteins accumulate at sites of clathrin-mediated endocytosis in dynamin-knockout cells (Ferguson et al., 2009), and also suggests that invaginated membranes of podosomes are dynamically rearranged by dynamin. However, addition of the WASP inhibitor, wiskostatin, acutely induced a drastic dissociation of GFP–FBP17 patches from podosomes within a minute, followed by the disassembly of F-actin (Fig. 6A arrowheads; supplementary material Movie 4). Within 3 minutes after drug treatment, GFP–FBP17 changed its localization from an intense dot-like pattern to an entirely cytosolic distribution. Importantly, podosomal localization of GFP–FBP17 was largely unaffected by treatment with latrunculin B (LatB), an actin monomer-sequestering drug that inhibits polymerization (supplementary material Fig. S6C), indicating that the effect of WASP inhibition was not due to the perturbation of actin polymerization per se. Given the fact that wiskostatin can stabilize the WASP molecule in an auto-inhibited state (Peterson et al., 2004), we examined whether this treatment could decrease the association between FBP17 and WASP. As shown in supplementary material Fig. S6E, interaction of FBP17 with WASP was indeed diminished by wiskostatin treatment, indicating that WASP binding is required for the assembly of FBP17 at the plasma membrane in vivo.

We next analyzed the effect of wiskostatin on the behavior of GFP–PSTPIP2 in Raw 264.7 cells. Strikingly, a dramatic nucleation of GFP–PSTPIP2 at the plasma membrane was observed following the drug treatment (Fig. 6B; supplementary material Movie 5). Interestingly, this self-organization of GFP–PSTPIP2 closely originated from places where actin was polymerized. In particular, GFP–PSTPIP2-positive clusters seemed to emerge around an F-actin core of a podosome that was about to disassemble (Fig. 6B arrowheads; supplementary material Movie 5), opposite to the nucleation pattern of GFP–FBP17 during the formation of the F-actin core. In addition, this effect was also not due to the inhibition of actin polymerization, because LatB treatment did not induce any membrane clustering of PSTPIP2 (supplementary material Fig. S6D). Thus, the robust nucleation of PSTPIP2 is correlated with the loss of FBP17.

Next, we simultaneously observed the behaviors of the two F-BAR proteins at podosomes upon WASP inhibition. Treatment of Raw 264.7 cells coexpressing tagRFP-FBP17 and GFP–PSTPIP2 with wiskostatin caused a marked decrease of tagRFP-FBP17 patches (∼4.0-fold) followed by the assembly of GFP–PSTPIP2 (∼4.2-fold; Fig. 6C–E; supplementary material Movie 6). The time course of the generation of GFP–PSTPIP2 patches corresponded well with the disappearance of tagRFP–FBP17 assemblies from podosomes (Fig. 6D,E). Interestingly, tagRFP–FBP17 clusters were frequently taken over by GFP–PSTPIP2 patches (Fig. 6D), indicating that they both localize to the same podosomal membrane for their antagonistic action. Importantly, no accumulation of GFP–PSTPIP2 R168A was observed with the addition of wiskostatin even after the disappearance of tagRFP–FBP17 (Fig. 6F). Because wiskostatin was reported to have off-target effects (Guerriero and Weisz, 2007), we performed RNAi experiments to investigate the behavior of FBP17 and PSTPIP2 in WASP-knocked down cells. We found that the cluster formation of FBP17 was significantly inhibited in WASP-depleted cells in which podosome formation was significantly suppressed (Fig. 6G,I; supplementary material Fig. S6F), excluding the possibility of off-target effects of wiskostatin. In contrast, the number of PSTPIP2 patches was increased in the knockdown cells (Fig. 6H,I). Thus, we conclude that the self-assembly of PSTPIP2 is enhanced by the dissociation of FBP17, supporting an antagonistic action of their F-BAR domains for the assembly at podosomes.

In this study, we first provided evidence for a mutual regulatory mechanism between F-BAR protein assemblies in living cells. Our studies reveal a surprisingly dynamic behavior of F-BAR protein assembly in time and space. The results suggest that a competitive regulation along with a cooperative action of different BAR proteins might integrate the control of their assembly to exert various biological functions based on membrane morphogenesis driven by actin polymerization.

We directly visualized both in vitro and in living cells that the assembly of FBP17 corresponds to the initiation of actin nucleation. Our data strongly indicate that FBP17 could create a high-order membrane scaffold that triggers the oligomerization of WASP for localized actin polymerization at the ventral plasma membrane. Such protein scaffolds can be formed either at a flat membrane surface or membrane tubules that the F-BAR proteins induce, as observed in our in vitro experiments using giant liposomes (Fig. 4). Although our knowledge about the morphology of the podosomal plasma membrane is still limited, it is tempting to speculate that the F-BAR proteins contribute to the formation of invaginated membranes that has been observed by electron microscopy for podosomes in Rous-sarcoma-virus-transformed cells (Nitsch et al., 1989; Ochoa et al., 2000).

Although not required in vitro, WASP was necessary for the self-assembling ability of FBP17 in vivo (Fig. 6). This observation is somewhat surprising as the Toca protein itself has been proposed to recruit WASP to the plasma membrane (Takano et al., 2008; Tsuboi et al., 2009; Tsujita et al., 2006). We envisage that, in vivo, an initial assembly of FBP17 is triggered by WASP binding, which may induce a positive feedback loop that further propagates its assembly with WASP on the plasma membrane.

In contrast to FBP17, the self-assembly of PSTPIP2 clearly correlates with the inhibition of actin polymerization. Interestingly, besides its antagonistic role for FBP17, PSTPIP2 has been known to interact with PTP-PEST, a PEST-type tyrosine phosphatase that is thought to modulate WASP activation via its dephosphorylation (Veillette et al., 2009; Wu et al., 1998). Interestingly, PTP-PEST was localized to membrane tubules induced by PSTPIP2 in COS-1 cells (data not shown), suggesting that the self-assembly of PSTPIP2 may promote a clustering of PTP-PEST on the plasma membrane to synergistically modulate WASP activation.

We propose a plausible model of how the assembly of FBP17 is regulated by PSTPIP2 in podosome formation (Fig. 7). The clustering of PSTPIP2 competes with the assembly of FBP17 by disturbing its intermolecular interactions at the ventral plasma membrane. Their competition is clearly evident by our finding that the dissociation of FBP17 acutely promotes the nucleation of PSTPIP2 (Fig. 6). In contrast, the absence of PSTPIP2 appears to result in the robust assembly of FBP17 that leads to hyperactivation of actin nucleation (Fig. 1,7). Thus, the antagonistic action of different F-BAR proteins may ensure the proper control of actin polymerization at podosomes.

An important question is: what kind of molecule or information do the two F-BAR proteins compete for? One possibility might be the curvature of the plasma membrane in podosomes. It is conceivable that both F-BAR proteins prefer the similar extent of membrane curvature to which they wrap around, and the detachment of one from the invaginated membrane facilitates the recruitment of the other. This idea might explain the segregated localization of FBP17 and PSTPIP2 on a single membrane tubule in COS-1 cells or patches in Raw macrophages (Fig. 5). Another possibility is that the two F-BAR proteins compete for a common binding protein localized at the plasma membrane, which is an important subject of future studies.

A decreased expression or depletion of PSTPIP2 leads to a macrophage-associated autoinflammatory disease (Chitu et al., 2009; Grosse et al., 2006), suggesting that dysregulated assembly of Toca proteins and subsequent abnormal activation of WASP-mediated actin polymerization may contribute to the pathogenesis of this type of inflammation. In fact, abnormal activation of these cytoskeletal machineries by PSTPIP2 knockdown leads to aberrant podosome formation (Fig. 1), which leads to the enhancement of matrix degradation (supplementary material Fig. S2H). Consistent with this, absence of PSTPIP2 is reported to cause the erosion of interphalangeal articulations and phalangeal bones, probably promoting tissue damage, including osteolysis (Ferguson et al., 2006; Grosse et al., 2006). Interestingly, it has been reported that the absence of PSTPIP2 results in an increased population of osteoclasts, suggesting its suppressive role in the differentiation of macrophages into osteoclasts (Chitu et al., 2012). It was also found that the fusion of osteoclast precursors is inhibited by PSTPIP2 in a manner dependent on its F-BAR domain (Chitu et al., 2012). In this regard, podosome formation has been reported to play a crucial role for cell–cell fusion events in osteoclasts (Oikawa et al., 2012). Thus, it is possible that PSTPIP2 could negatively regulate the differentiation of macrophages by controlling their podosome formation. Given that matrix degradation activity of mature osteoclasts might be higher than that of monocytes/macrophages, the enhancement of MMP-mediated tissue damage caused by abnormally activated podosomes or the promoted differentiation of osteoclasts may be one of the most common mechanisms for the development of this inflammation in the PSTPIP2-depleted mice.

In conclusion, our study provides the first clue for a dynamic competition between F-BAR protein assemblies at the plasma membrane. The antagonistic actions between different membrane-deforming proteins may occur in a wider range of cellular events, including clathrin-mediated endocytosis which involves highly ordered, actin-dependent morphological changes in the plasma membrane.

Polyclonal anti-PSTPIP2 antibody was raised by immunizing rabbits with full-length PSTPIP2 recombinant protein, and was affinity purified. Polyclonal anti-FBP17 was generated as previously described (Itoh et al., 2005). The following antibodies were purchased: monoclonal anti-CIP4 (BD Biosciences), polyclonal anti-WASP (Cell Signaling), polyclonal anti-actin (MBL), monoclonal anti-Myc (Santa Cruz Biotechnologies), monoclonal anti-vinculin and anti-FLAG (Sigma-Aldrich) and secondary antibodies conjugated to Alexa Fluor 488 and 568 (Invitrogen). Fibronectin was obtained from Sigma-Aldrich. Dynasore, wiskostatin and latrunculin B were purchased from Calbiochem.

GFP-FBP17 (human) and GFP-PSTPIP2 (mouse) constructs were made as previously described (Tsujita et al., 2006). FBP17 was subcloned into pEF-BOS-FLAG, pEF-BOS-FLAG-GFP (with an N-terminal FLAG tag and GFP) and pTagRFP-C1 (Evrogen) vectors. PSTPIP2 was subcloned into pEF-BOS-Myc and pEF-BOS-FLAG vectors. For rescue experiments, PSTPIP2 siRNA-resistant mutants were constructed by substituting the nucleotides within the siRNA target no. 1 (GGACAGTCT) without changing the coding amino acids. Point mutations were made by PCR with mutated primers, and the sequences were confirmed. Lifeact-GFP and Lifeact-mCherry constructs were generated as previously described (Riedl et al., 2008).

Raw 264.7 cells were cultured in RPMI-1640 medium supplemented with 10% fetal bovine serum (FBS), and COS-1 cells was cultured in Dulbecco's modified Eagle's medium supplemented with 10% FBS. Transfection was performed using FuGENE HD (Roche) according to the manufacturer's protocol. Transfected cells were typically examined after 24 hours. For knockdown experiments, Stealth siRNAs were purchased from Invitrogen. The target sequences were as follows: PSTPIP2 no. 1, 5′-GCCGUGUGGACAGUCUGAGAUCAAU-3′; PSTPIP2 no. 2, 5′-UCGAGGCUCAGGAAUGUGAACGAAU-3′; FBP17, 5′-CAGGAGCAAUGGGAAUACUACCAUA-3′; CIP4, 5′-UGGAGCAGGCUUAUGCUAAGCAACU-3′, WASP, 5′-GAGCACCGAGUGGAUUCAAACAUGU-3′. Control siRNA (Stealth™ Negative Control Medium GC Duplex #3; Invitrogen) was used as control RNAi. siRNAs (40 nM) were transfected into Raw 264.7 cells with 7.5 µl Lipofectamine RNAi MAX (Invitrogen) in a six-well plate. After 24 hours, a second transfection was performed, and cells were cultured for 48 hours and subjected to various other experiments. For RT-PCR experiments, mRNA was extracted from confluent cells (Raw 264.7 and NIH Src) with Trizol reagent (Invitrogen). A first-strand cDNA was synthesized from the mRNA by SuperScript First-strand synthesis system for RT-PCR (Invitrogen). The PCR amplification was performed using the first-strand cDNA.

Cells were fixed with 3.7% formaldehyde in phosphate-buffered saline for 10 minutes, permeabilized with 0.2% Triton X-100 for 10 minutes, and immunostained with first and second antibodies for 1 hour each. Fluorochromes used include Alexa Flour 488 and 568. For visualization of F-actin, Alexa-Flour-488- or -568-conjugated phalloidin (Invitrogen) was incubated with fixed cells for 30 minutes. Fluorescence images were taken using a confocal microscope system (FluoView 1000-D; Olympus) equipped with 473- and 559-nm diode lasers through an objective lens [60× (cell analysis) or 100× (liposome analysis) oil immersion objective, 1.35 NA]. For live-cell imaging with total internal reflection fluorescence (TIRF) microscopy, Raw 264.7 cells were grown on fibronectin-coated glass-bottomed dishes (IWAKI). Before imaging, the culture medium was changed to an imaging medium [10 mM Hepes (pH 7.5), 150 mM NaCl, 5 mM KCl, 1 mM CaCl₂, 1 mM MgCl₂, 10 mM glucose] supplemented with 10% FBS. The images were acquired for 5–10 minutes with 500-ms exposures by using a TIRF microscopy system (Olympus) equipped with ion lasers (MELLES GRIOT) through a TIRF lens (100× oil immersion objective, 1.45 NA; Olympus) and with a Cascade II cooled charge-coupled devise camera (Photometrics), and then processed with ImageJ (NIH) and Photoshop (Adobe). The fluorescence intensity was calculated using ImageJ.

A normal podosome was defined as a dot structure surrounded by vinculin, whereas an aberrant podosome was determined by the morphology with F-actin cloud and vinculin staining. Total fluorescence intensities of F-actin on podosomes were calculated based on the intensity of Alexa-Fluor-488-conjugated phalloidin per area of normal and aberrant podosomes. A fluorescent-gelatin-coated dish was prepared as previously described (Berdeaux et al., 2004). Briefly, the glass-bottomed dish was coated with Oregon-Green-488-conjugated gelatin (0.2 mg/ml; Invitrogen) for 20 minutes, and the gelatin was then cross-linked with 0.5% glutaraldehyde for 15 minutes. After intensive washing with PBS, the dish was incubated with fibronectin for 20 minutes, and cells were plated on it in RPMI-1640 medium containing 10% FBS for 15 hours. Matrix degradation activity was indicated by loss of fluorescence of the gelatin. The area of these sites was calculated using ImageJ.

FLAG–GFP–FBP17, FLAG–FBP17 and FLAG–PSTPIP2 were expressed in FreeStyle 293-F cells (Invitrogen) using FreeStyle MAX reagents (Invitrogen); purified using anti-FLAG-affinity agarose (Sigma-Aldrich); and eluted with FLAG peptide (Sigma-Aldrich) in buffer A [20 mM Hepes (pH 7.5), 100 mM KCl, 1 mM MgCl₂ 1 mM EGTA and 1 mM DTT]. His-N-WASP was expressed in Sf9 cells and purified as previously described (Takano et al., 2008).

A relatively large liposome sample was prepared as follows. Samples containing brain lipids (95%; Folch fraction; Sigma Aldrich) and Rhodamine-conjugated phosphatidylethanolamine (5%; Avanti polar lipids) or DiOC18 (5%; Molecular probes) were dried under nitrogen gas and suspended in 0.3 M sucrose for 1 hour at 37°C followed by vortexing, to a concentration of 1 mg/ml. Then, liposomes were diluted with buffer A to allow the formation of large liposomes (∼1–5 µm in diameter), which was examined by confocal microscopy.

Purified GFP–FBP17 (50 or 200 nM) was incubated with liposomes (20 µg/ml) in buffer A for 10 minutes at room temperature and subjected to native electrophoresis. The gel was then subjected to sliver staining using Silver Quest™ (Invitrogen).

A glass-bottomed dish was coated with 0.2 mg/ml N-ethylmaleimide-inactivated myosin (Cytoskeleton). After washing with buffer A, the liposome solution was incubated on a myosin-coated glass-bottomed dish for 20 minutes. Then, the purified protein was added to a liposome (20 µg/ml)-coated glass-bottomed dish for 5 minutes at room temperature and observed by confocal microscopy. For actin polymerization assay on liposomes, purified GFP–FBP17, FLAG–FBP17 and/or FLAG–PSTPIP2 were preincubated with liposomes (20 µg/ml) on glass-bottomed dish for 5 minutes followed by incubation with 1 µM actin (40% Rhodamine-labeled actin; Cytoskeleton), 50 nM Arp2/3 complex (Cytoskeleton) and 200 nM N-WASP for 5 minutes in a polymerization buffer [20 mM Hepes (pH 7.5), 100 mM KCl, 2 mM MgCl₂, 500 µM ATP, 1 mM EGTA and 1 mM DTT], and observed by confocal microscopy. Note that FBP17-N-WASP-mediated actin nucleation on liposomes was very robust, as evidenced by the intensity of Rhodamine–actin saturation within a minute. The fluorescence intensity of GFP–FBP17 and Rhodamine-labeled actin on liposomes were calculated using ImageJ.

Statistical analysis was performed using two-tailed Student's t-tests. The differences were considered significant if P<0.05.

We thank Dr K. Takano for the gift of baculovirus for His-N-WASP expression.