Supervillin couples myosin-dependent contractility to podosomes and enables their turnover

Protease-driven invasive cell migration depends on contact with the extracellular matrix (ECM) and degradation of matrix material (Friedl and Wolf, 2009). Both properties are combined in podosomes and invadopodia, which constitute a subset of cell-matrix contacts with an inherent ability to locally degrade the ECM. These structures, collectively called ‘invadosomes’, promote matrix degradation through release of ECM-lytic enzymes such as matrix metalloproteinases or ADAMs (Destaing et al., 2011; Linder, 2009; Linder et al., 2011; Saltel et al., 2011). Podosomes are formed constitutively in monocytic cells such as macrophages (Linder et al., 1999), dendritic cells (Burns et al., 2001) and osteoclasts (Destaing et al., 2003), and are also generated in other cell types such as endothelial cells (Moreau et al., 2003; Osiak et al., 2005; Tatin et al., 2006) or smooth muscle cells (Gimona et al., 2003). Current evidence is that these or similar structures exist in vivo, where they potentially regulate invasive cell migration in the context of transendothelial diapedesis (Carman et al., 2007) or vascular remodeling (Linder et al., 2011; Murphy and Courtneidge, 2011; Rottiers et al., 2009; Saltel et al., 2011).

Podosomes consist of an inner core of F-actin and actin-associated proteins, including the Arp2/3 complex (Linder et al., 2000a), cortactin (Bowden et al., 2006), gelsolin (Chellaiah et al., 2001) and Tks5 (Crimaldi et al., 2009; Thompson et al., 2008). Podosome cores are surrounded by a ring structure of adhesion plaque proteins such as talin or vinculin (Tarone et al., 1985; Zambonin-Zallone et al., 1989). Podosomes are highly dynamic structures with a lifetime of 2–12 minutes (Destaing et al., 2005), which suggests tight control of the turnover of both internal components and the podosome structure itself. Actomyosin-based contractility appears to be essential for podosome dynamics, although its exact influence is currently unclear (Burgstaller and Gimona, 2004; Clark et al., 2006; Collin et al., 2008; Kopp et al., 2006).

Podosomes have been seen as identical structures of similar size and molecular composition, especially in fibroblasts, in which podosome formation is induced by transformation with v-Src (Crowley et al., 2009; Mukhopadhyay et al., 2009; Nermut et al., 1991), and in smooth muscle cells treated with phorbol esters (Gimona et al., 2003; Hai et al., 2002). However, artificial induction of podosomes by activation of downstream regulators might bypass some regulatory steps involved in podosome function or diversification (Linder, 2007).

Especially in monocytic cells, podosomes are formed constitutively and exist in two subpopulations: (1) larger structures near the cell edge, called precursors, which undergo fission and fusion processes, and (2) smaller podosomes in the cell interior (Evans et al., 2003; Kopp et al., 2006). Following the established nomenclature of calling larger podosomes ‘precursors’ (Evans et al., 2003; Kopp et al., 2006), we propose to call the subpopulation of smaller podosomes ‘successors’. At the leading edge of a migrating cell, successors are found behind precursors, and many of the successors form after fission of precursors.

Current knowledge about the differential regulation of podosome subpopulations is limited. The kinesin KIF1C influences fission of precursors (Kopp et al., 2006) whereas the kinesin KIF9 contacts mostly successors (Cornfine et al., 2011). However, neither protein is a podosome component, and actual molecular differences between the two podosome subpopulations have not been described so far.

We show here that the membrane- and cytoskeleton-associated protein supervillin localizes preferentially to successor podosomes, but is also recruited to precursors upon their dissolution. Supervillin is a member of the villin and gelsolin family, and associates tightly with cholesterol-rich membrane signaling domains (Nebl et al., 2002; Pestonjamasp et al., 1997). A multi-domain protein, supervillin interacts with many cytoskeletal proteins, including F-actin, myosin II, the long form of myosin light chain kinase (L-MLCK), and as many as five microtubule-dependent motors (Chen et al., 2003; Smith et al., 2010; Takizawa et al., 2007; Takizawa et al., 2006). Supervillin induces L-MLCK-dependent myosin II contractility through its first N-terminal 171 amino acids, whereas residues 342–571 negatively regulate focal adhesions in COS7 cells (Takizawa et al., 2007; Takizawa et al., 2006). Interestingly, supervillin localization at invadopodia is associated with increased matrix degradation by MDB-MB-231 cells (Crowley et al., 2009), and it binds directly to both cortactin and Tks5 (Crowley et al., 2009; Smith et al., 2010). Our data now identify supervillin as the first podosome component to localize differentially to podosome subpopulations and show that supervillin couples myosin-dependent contractility to podosomes, thus enabling their turnover.

In a screen for differences in molecular composition between podosome subpopulations in primary human macrophages, we found that GFP-fused supervillin (GFP–SV) showed the most prominent differences. GFP–SV localized preferentially at successor podosomes in the inner regions of unpolarized cells (Fig. 1A–C) or at the rear of the podosome field of polarized macrophages (Fig. 1D–F). GFP–SV also was enriched in the cortex at the trailing edges of polarized cells (Fig. 1D–F).

A more detailed analysis showed that GFP–SV localizes to a distinct cap-like structure over the F-actin-rich cores of successor podosomes. Fluorescence intensity measurements of individual podosome cores (Fig. 1G–J) revealed that the intensities of GFP–SV and F-actin stained with Alexa-Fluor-568–phalloidin showed similar profiles (Fig. 1J). In 3D reconstructions of optical z-stacks (Fig. 1K–M; supplementary material Movie 1), GFP–SV (green) was found mostly at the top of the podosome core, although it partially colocalized with F-actin (red, overlap in yellow). This analysis was confirmed by co-staining of vinculin, a component of the podosome ring structure (supplementary material Fig. S1A–F). The highest intensities for GFP–SV-based fluorescence were found between the two maxima of vinculin-based fluorescence (supplementary material Fig. S1 and Movie 1). By contrast, the supervillin-related protein gelsolin showed no differential localization to podosome subpopulations and was present at both precursors and successors (supplementary material Fig. S2). GFP–SV expression significantly enhanced the number of polarized cells (Fig. 1N), suggesting a role for supervillin in the development of a migratory phenotype (and see below). Quantification showed significant enrichment of GFP–SV at successor podosomes compared with precursor podosomes (Fig. 1O). Collectively, these data show that GFP–SV is enriched at successor podosomes, where it forms a cap structure on top of the F-actin-rich core, with only minor localization at precursor podosomes and at the leading edge.

Supervillin has not previously been described in macrophages. Supervillin was detected on western blots of macrophage lysates as an appropriately sized ~205 kDa endogenous protein with the specific H340 antibody against the supervillin N-terminus (Nebl et al., 2002; Oh et al., 2003) (Fig. 1P). RT-PCR of macrophage mRNA using supervillin-specific primers also generated a band of the expected size (521 bp; Fig. 1Q). Mass spectrometric analyses of podosome-enriched macrophage fractions (not shown) identified 13 supervillin-specific peptides in the podosome-containing fraction and none in the corresponding fraction from macrophages in which podosomes had been disrupted by prior addition of the Src kinase inhibitor PP2 (Gringel et al., 2006; Linder et al., 2000b). The presence of endogenous supervillin at macrophage podosomes was confirmed by immunofluorescence staining with a newly developed antibody (HSV715) specific for amino acids 715–728 in human supervillin. Endogenous supervillin localized to F-actin-rich podosome cores and secondarily at subcortical actin filaments (Fig. 1R–T), confirming the results obtained with GFP–SV (Fig. 1A–M).

By measuring endogenous supervillin-based fluorescence intensity at randomly chosen successor and precursor podosomes, we found that successors had a significantly higher mean value of supervillin-based fluorescence (supplementary material Fig. S3). This difference was not as large as in cells expressing GFP–SV (compare supplementary material Fig. S3J with Fig. 1O), suggesting that supervillin overexpression accentuates the difference between precursor and successor podosomes.

To study the dynamics of supervillin and to clarify its localization in living cells, macrophages were transfected with constructs encoding GFP–SV and either mRFP–β-actin or mRFP–Lifeact to label F-actin and analyzed by videomicroscopy. In polarized (migratory) cells, GFP–SV was largely absent from precursor podosomes at the leading edge (Fig. 2A; supplementary material Movie 2), which was comparable to results with fixed cells (Fig. 1D–F). However, when cells retracted their leading edge and precursor podosomes were dissolved, the dissolving podosomes acquired GFP–SV; the gradual disappearance of the mRFP–β-actin (Fig. 2B–E,F–I) or mRFP–Lifeact signal (Fig. 2N–Q,R–U) was accompanied by a progressive enrichment of GFP–SV (Fig. 2B–E,J–Q,V–Y; supplementary material Movies 2,3).

Consistent with a role in podosome dissolution, overexpression of GFP–SV led to a pronounced decrease in podosome numbers in macrophages transfected for 48 hours or 72 hours (Fig. 3). This effect was only marginally evident at 24 hours (compare Fig. 3C with 3B), suggesting that podosome counts were relatively normal in our live-cell imaging experiments (Fig. 2), when expression levels of GFP–SV averaged only ~twofold that of endogenous supervillin (not shown).

On the basis of overexpression of GFP-tagged supervillin fragments, multiple regions of supervillin contribute to the regulation of macrophage podosome numbers (Fig. 3). Deletion of the myosin-IIA- and L-MLCK-binding N-terminal 170 amino acids in GFP–SV(171–1792) completely abrogated the reduction of podosome numbers observed with full-length GFP–SV (Fig. 3D). GFP-tagged supervillin residues 1–174 [GFP–SV(1–174)], by themselves, only modestly decreased podosome numbers (Fig. 3E), suggesting that this sequence is necessary but not sufficient for this effect.

Other deletion constructs suggested contributory regulation by the TRIP6-binding, focal-adhesion-targeting sequence in SV(343–571), by other supervillin N-terminal sequences and by the supervillin C-terminus. Although expression of GFP-tagged SV(343–571) alone had only a delayed effect on podosome numbers (supplementary material Fig. S4A), a supervillin construct lacking this region [GFP–SV(Δ343–570)] caused an even more pronounced reduction in podosome numbers than did full-length GFP–SV (Fig. 3F). GFP–SV(174–343) and GFP–SV(571–830), each of which also contains an F-actin binding site (Chen et al., 2003), also decreased podosome formation (Fig. 3G; supplementary material Fig. S4B). A potential co-regulatory role for both N- and C-terminal sequences is inferred from the lack of an effect of either the entire supervillin N-terminus [GFP–SV(1–830)] (supplementary material Fig. S4C) or the gelsolin and villin homology region [GFP–SV(1010–1792)] (supplementary material Fig. S4D) on podosome numbers.

Of the implicated domains, only the myosin-IIA- and L-MLCK-binding sequence, SV(1–174), appeared to regulate the targeting of supervillin to successor podosomes (Fig. 4). Successor podosome localization was preserved in SV(1–830) (Fig. 4A–C) and SV(Δ343–570) (Fig. 4D–F), indicating that the minimal sequence(s) responsible for this phenomenon were (1) present in the N-terminal half of the protein, but (2) were not present in the region 343–570, which is responsible for regulation of focal adhesions (Takizawa et al., 2006). Strikingly, GFP–SV(171–1792), which lacks the myosin-IIA- and L-MLCK-binding region, was not differentially localized, but was present at both precursor and successor podosomes (Fig. 4G–I). GFP–SV(1–174) alone did not show a pronounced localization to podosomes, but induced the formation of F-actin- and myosin-IIA-rich cables (Fig. 4J–L; supplementary material Fig. S5). Other GFP-tagged constructs containing isolated N-terminal domains [SV(174–343), SV(343–570), SV(571–830)] localized to all podosomes and other F-actin structures, including cortical cables (supplementary material Fig. S6A–I), probably as a result of their inherent F-actin binding activities.

To test whether overexpression of an isolated F-actin binding region could exert dominant-negative effects on the distribution of the full-length protein, we examined the effects of co-overexpressing GFP–SV(174–343) with mRFP–supervillin (mRFP–SV) (supplementary material Fig. S7). GFP–SV(174–343) localized to both precursor and successor podosomes (supplementary material Fig. S7C), which contained much-reduced levels of mRFP–SV (supplementary material Fig. S7B vs Fig. 7F). Competition for binding was less apparent at actin cables in the cell interior (supplementary material Fig. S7B), where the actin-bundling sites in full-length mRFP–SV (Chen et al., 2003) probably out-compete GFP–SV(174–343) for binding to the aligned actin filaments.

Taken together, these results suggest that supervillin-mediated loss of podosomes requires the myosin-IIA- and L-MLCK-binding sequence [SV(1–174)], and that this sequence is important for the association of supervillin with successor podosomes.

To explore possible roles for both supervillin and myosin IIA in podosome regulation and function, we performed siRNA-based knockdown experiments in macrophages and analyzed podosome lifetime and matrix degradation (Fig. 5). Knockdown of either protein (to 17% of endogenous supervillin levels or 5% of endogenous myosin IIA levels), using validated siRNAs (Materials and Methods; supplementary material Fig. S8), resulted in a pronounced ~twofold increase in lifetimes of both successors (Fig. 5A) and precursors (Fig. 5B). Strikingly, the combined knockdown of both proteins did not show an additive effect, with podosome lifetimes being statistically indistinguishable from the single knockdowns (Fig. 5A,B). (Note that because precursors can split and fuse, only precursors that formed and dissolved within the experimental period were evaluated.) Although there was no change in the percentage of cells with >100 podosomes (supplementary material Fig. S9A), average numbers did drop from ~400 podosomes/cell in luciferase-siRNA-treated cells to ~340 and ~270 podosomes/cell after treatment with siRNA against supervillin or myosin IIA, respectively (supplementary material Fig. S9B). The variability in the relative changes in podosome numbers over time also decreased somewhat (supplementary material Fig. S9C). Taken together, these results suggest that supervillin and myosin IIA function together in a pathway that stimulates podosome disassembly with a lesser effect on podosome formation.

By contrast, we found that supervillin and myosin IIA act synergistically to promote podosomal matrix degradation (Fig. 5C). Cells transfected with control siRNA against firefly luciferase or with specific validated siRNAs were scored into groups with low (0–25%) or high (26–100%) (Fig. 5C) levels of degradation of the underlying matrix. Unlike in previous studies of matrix degradation by tumor cells (Crowley et al., 2009), knockdown of supervillin alone did not significantly alter levels of matrix degradation (Fig. 5C–E). Reduction of myosin IIA or the supervillin-related protein gelsolin did reduce the percentage of highly degradative cells (Fig. 5C,F,G). Knockdown of supervillin in combination with either myosin IIA or gelsolin (Fig. 5C,H,I), or the triple knockdown of all three proteins (Fig. 5C,J), further reduced the extent of matrix degradation, with knockdowns of the single proteins all being significantly different from the double or triple knockdowns. The bar diagram (Fig. 5C) illustrates that the number of cells that degraded much less (≤25% of control) is highly and significantly increased from ~5% in control cells to ~45% in the double and triple knockdowns. These additive effects suggest that supervillin regulates a pathway that is functionally redundant with both myosin IIA and gelsolin and that all three proteins contribute to podosomal matrix degradation.

We next investigated the potential co-regulatory roles of supervillin and myosin IIA in podosome turnover in more detail. Consistent with a role for the myosin-IIA- and L-MLCK-binding region in supervillin-dependent regulation of podosomes, myosin IIA localization partially overlapped with F-actin and supervillin at successor podosomes and at cable-like structures between these podosomes (Fig. 6A–D). In the cell periphery, where precursors are located, myosin IIA was present in dash-like accumulations, which intercalated between precursors, but did not appear to be in direct contact with them (Fig. 6A–D). In polarized cells, GFP–SV colocalized with myosin IIA cables at the trailing cell edges (Fig. 6E–H). Supervillin-associated myosin IIA cables connecting successors were especially prominent in polarized cells (Fig. 6E–H). Such cable-like staining of myosin IIA was not found in control cells expressing GFP alone (not shown) (Kopp et al., 2006). Thus, modest overexpression of supervillin induces the formation or stabilization of myosin-IIA-containing cables between successor podosomes, suggesting the existence of contractile forces.

In live-cell imaging, small accumulations of GFP–myosin-IIA often appeared in proximity to mRFP-tagged supervillin at successor podosomes, especially at the rear of the podosome field (Fig. 6I–O; supplementary material Movie 4). Importantly, the appearance of these myosin IIA accumulations often preceded or coincided with podosome dissolution (Fig. 6I–M). During cell migration, successor podosomes enriched in supervillin and myosin IIA at the rear of the podosome field dissolved, and podosomes closer to the leading edge of the cell began to acquire mRFP–SV. Contractile myosin IIA cables were connected to successor podosomes, especially in cells expressing higher levels of myosin IIA (Fig. 6N,O; supplementary material Movies 5,6).

Interestingly, GFP-tagged L-MLCK localized to both successor and precursor podosomes in fixed and living cells (supplementary material Fig. S10 and Movie 7). No changes in GFP–L-MLCK signals were discernible before dissolution of podosomes from either subpopulation. These observations suggest that L-MLCK, which is present at all podosomes, and supervillin at successor podosomes recruit and/or activate myosin IIA, causing the dissolution of these structures.

To interrogate the order of these interactions, we co-immunoprecipitated cellular myosin IIA (Fig. 6P,Q) and L-MLCK (Fig. 6P) with GFP-fused supervillin constructs from macrophage lysates. As reported using recombinant polypeptides (Takizawa et al., 2007), GFP–SV(1–174) bound to both myosin IIA and to L-MLCK (Fig. 6P). Strikingly, myosin IIA contractility was required for binding to supervillin because inhibition of the myosin ATPase activity with blebbistatin (10 μM) before cell lysis abolished the co-precipitation of myosin IIA with both GFP–SV(1–174) and GFP–SV(1–830) (Fig. 6Q). These results confirm that SV(1–174) interacts with cellular L-MLCK and myosin IIA and suggest that supervillin binds selectively to contractile myosin IIA filaments.

Because supervillin increases myosin II contractility during cell spreading (Takizawa et al., 2007), we examined the level of phosphorylated myosin light chain (pMLC), a measure of myosin contractility (Matsumura, 2005; Vicente-Manzanares et al., 2009), in macrophages treated with siRNA to knock down supervillin (Fig. 7). Fluorescence-based measurements of pMLC associated with podosomes showed that supervillin knockdown significantly decreased pMLC levels, as compared with controls (Fig. 7A,B,G). Cells expressing GFP alone, GFP–SV or GFP–SV(1–174) showed no significant differences in pMLC levels at podosomes, whereas overexpression of GFP–SV(1–830) resulted in a pronounced increase in podosomal pMLC (Fig. 7C–G), despite lower levels of expression compared with GFP alone or GFP–SV(1–174) (not shown). Myosin IIA also was more compactly associated with individual podosomes in cells expressing GFP–SV(1–830) (Fig. 7H–J; supplementary material Movie 8), compared with podosomes with GFP–SV (Fig. 7K–M; supplementary material Movie 9). These data indicate that endogenous supervillin increases basal pMLC levels at podosomes and that SV(1–830) elevates the levels of pMLC at podosomes.

To test whether myosin contractility increases the localization of supervillin at successor podosomes, siRNA-mediated knockdown of myosin IIA, MLCK or a combination were performed, using luciferase siRNA as a control (Fig. 8A–L; supplementary material Fig. S8). MLCK knockdown was accompanied by a large reduction in pMLC (supplementary material Fig. S8E), indicating an important regulatory role for MLCK in macrophage myosin II activation. The preferential localization of GFP–SV to successor podosomes in control cells was significantly reduced by lowered levels of myosin IIA, MLCK, or both proteins (Fig. 8S). The difference between the single myosin IIA knockdown and double myosin IIA and MLCK knockdown conditions was not statistically significant. Differential supervillin recruitment to podosomes also was inhibited in cells treated with 2 μM blebbistatin (Fig. 8M–S), suggesting that MLCK-mediated activation of myosin IIA facilitates differential recruitment of supervillin to successor podosomes. This was further confirmed in live-cell experiments, where addition of 2 μM blebbistatin to cells resulted in prominent recruitment of GFP–SV to precursor podosomes (supplementary material Movie 10).

Similarly, both myosin IIA and MLCK were required for the increase in cell polarization observed after modest overexpression of GFP–SV in macrophages (Fig. 1N, Fig. 8T). Knockdown of supervillin, myosin IIA, MLCK, a combined knockdown of myosin IIA and MLCK, or blebbistatin treatment had no effect on control cells treated with either control luciferase siRNA or DMSO (Fig. 8T). By contrast, all of these treatments significantly reduced the extent of cell polarization induced by overexpression of GFP–SV (Fig. 8T). These data suggest that supervillin-induced cell polarization is largely dependent on MLCK-mediated myosin IIA contractility and is probably related to the selective recruitment of supervillin to successor podosomes.

We identify here supervillin as the first protein that shows a differential localization to podosome subpopulations in primary macrophages. Supervillin localizes primarily as a cap-like structure above the actin-rich podosome cores, a location that is consistent with the presence of three F-actin binding sites within the protein (Chen et al., 2003). This localization is also consistent with the colocalization of cortactin and supervillin at the cytoplasmic apexes of Src-induced invadosomes (Crowley et al., 2009) and apparently overlaps with the podosomal localization of the formin FMNL1 (Mersich et al., 2010). These results point to a more complex inner architecture of podosomes than was previously appreciated (Gimona, 2008; Linder and Aepfelbacher, 2003).

Supervillin promotes podosome dissolution by increasing the local activation of myosin II. GFP–SV reduces podosome numbers and its recruitment to podosomes coincides with the gradual disappearance of their actin-rich cores. Deletion of the myosin-II- and L-MLCK-binding N-terminal sequence [SV(1–174)] eliminates the reduction in podosome numbers and abolishes the differential localization of supervillin to successor podosomes. Conversely, the focal adhesion-targeting sequence in supervillin [SV(343–571)] (Takizawa et al., 2006) is not necessary for the reduction of podosome numbers. Any supervillin construct with a single F-actin binding site localizes to podosomes, suggesting that although F-actin binding can mediate podosome targeting, SV(1–174) is essential for promoting the differential localization to successors and for podosome dissolution. Moreover, siRNA-mediated knockdown of supervillin, myosin II, or both together led to a ~twofold increase in podosome lifetime, indicating that they function in a common pathway. Taken together, these data argue that supervillin-mediated myosin IIA activation is important for podosome turnover.

Our results confirm and extend earlier findings that supervillin induces myosin II hypercontractility (Takizawa et al., 2007; Takizawa et al., 2006), by showing that myosin II ATPase activity is required for the interaction with supervillin. [Note that myosin IIA is the predominant myosin II isoform in monocytic cells, such as macrophages (Maupin et al., 1994).] Myosin IIA and L-MLCK both co-immunoprecipitate with the supervillin N-terminus [SV(1–174)], and the association with myosin IIA is blocked by the myosin II ATPase inhibitor blebbistatin. Moreover, pMLC levels at podosomes are reduced by the knockdown of supervillin, myosin IIA or MLCK. These observations are in line with the identification of supervillin as part of the blebbistatin-sensitive myosin II interactome at focal adhesions (Kuo et al., 2011).

Our data clarify the role of myosin II in podosome regulation. Myosin II has been localized to the ring structure of podosomes in both osteoclasts (Krits et al., 2002) and dendritic cells (van Helden et al., 2008). Although the exact influence of myosin activity on podosome regulation is under debate (Burgstaller and Gimona, 2004; Clark et al., 2006; Collin et al., 2008; Kopp et al., 2006), a possibly unifying concept is that a low, basal myosin II activity supports the formation and maintenance of podosomes (Burgstaller and Gimona, 2004), whereas sudden increases in myosin II activity trigger podosome dissolution (van Helden et al., 2008). As reported previously (Burgstaller and Gimona, 2004), we found that myosin IIA localizes around and between macrophage podosomes. We also show that these myosin-IIA-positive structures co-stain for pMLC, confirming that these myosin cables are indeed contractile. GFP–myosin-IIA is enriched only at the most rearward, mRFP–SV-positive, successor podosomes, and myosin IIA enrichment immediately precedes podosome dissolution. By contrast, despite MLCK regulation of supervillin localization to successor podosomes, GFP-tagged L-MLCK itself is found at both successor and precursor podosomes, and no changes in GFP–L-MLCK signals were discernible before dissolution of podosomes from either subpopulation. Overexpression of GFP–SV increases the recruitment of myosin IIA and pMLC at and between podosomes: an effect that is even more prominent after expression of GFP–SV(1–830). These results are in agreement with the report that myosin II activation results in rear-localized disassembly of actin networks (Wilson et al., 2010) and suggest a sequence of events at podosomes during their dissolution.

We propose the following model for the coordinated regulation of podosome turnover by supervillin and L-MLCK-mediated activation of myosin II (Fig. 9). (1) Precursor podosomes form at the leading edge through supervillin-independent F-actin nucleation (Linder, 2007). (2) L-MLCK, which is a major MLC regulatory kinase in macrophages (supplementary material Fig. S8), localizes to all podosomes, probably because of its interactions with F-actin (Hatch et al., 2001; Yang et al., 2006). (3) Myosin IIA becomes activated at low levels around or at successor podosomes, probably, although not necessarily, by MLCK. (4) Supervillin binding to the activated myosin stabilizes the contractile filaments and helps connect them to L-MLCK and F-actin at successor podosomes. (5) The cross-bridging of L-MLCK and myosin IIA by supervillin (Takizawa et al., 2007) increases local MLCK-mediated phosphorylation of MLC, which promotes more supervillin binding to myosin IIA heavy chain in a positive-feedback loop that eventually triggers podosome dissolution.

Myosin II activation at the cell posterior also should be amplified by supervillin, which would explain the observed supervillin- and myosin II-dependent increase in cell polarization. Increased actomyosin contractility within the cell can break the symmetry of round, unoriented cells, leading to their spontaneous polarization and enhanced migration, even in the absence of a stimulatory gradient (Cramer, 2010). Thus, the effect of supervillin on cell polarity is not necessarily coupled to its effects on podosome dynamics, but would be consistent with a speculative role for podosome function in the regulation of directional migration of macrophages (Linder et al., 2011).

Our model is also consistent with observations that branched filament disassembly is triggered by increases in myosin II activation in both cellular and in vitro systems (Wilson et al., 2010) (L. Blanchoin, personal communication). Myosin contractility promotes the progressive alignment of actin filaments into arrays that facilitate myosin-based filament sliding. Incorrectly aligned filaments are lost from the coalescing structure and eventually disassemble. This process is further enhanced by the presence of F-actin bundling proteins. Therefore, supervillin-mediated actin bundling (Chen et al., 2003) might contribute to podosome dissolution. Indeed, myosin II itself is an important F-actin bundling and stabilizing protein (Choi et al., 2008; Xu et al., 2001).

Other mechanisms for supervillin-mediated podosome dissolution might also be involved. For instance, supervillin binding to cortactin (Crowley et al., 2009) could inhibit the conformational change in cortactin required for cortactin stimulation of Arp2/3-mediated actin polymerization (Evans et al., 2011). Simultaneous binding of the supervillin N-terminus to L-MLCK and cortactin might potentiate the association of L-MLCK with cortactin – an interaction that abolishes cortactin-mediated actin nucleation (Dudek et al., 2002). Supervillin also could indirectly inhibit cortactin-mediated actin polymerization at podosome cores by promoting the local formation of contractile assemblies of actin filaments containing tropomyosin, which inhibits Arp2/3-mediated actin nucleation from existing actin filaments (Blanchoin et al., 2001; Higgs et al., 1999; Machesky et al., 1999). Supervillin binding and bundling of actin filaments (Chen et al., 2003) might also directly inhibit their ability to bind activated Arp2/3.

The relationship between supervillin levels and podosome numbers is multi-faceted and dependent on cell-specific factors that might limit supervillin function at different points in a cyclical process. For instance, overexpression of EGFP–SV in cultured cells increases the number of invadosomes per cell (Crowley et al., 2009), in contrast to the decrease in podosome numbers observed here. Nevertheless, the decreased number of podosomes per cell observed after supervillin knockdown in macrophages is consistent with a role for supervillin in podosome formation or upkeep that is outweighed in macrophages by podosome dissolution as a result of localized myosin II activation. A role in podosome formation could include supervillin-mediated, actin-dependent rapid recycling (Fang et al., 2010) of podosome components from dissolving podosomes to newly forming ones. Rapid recycling of podosome components could also facilitate the dynamic re-assembly of cytoskeletal and membrane components of podosomes.

The absence of a major effect on podosome numbers by GFP–SV(1–830) hints further at the possibility of a cyclical process, in which the supervillin C-terminus plays a role in controlling podosome dynamics. Calponin is one podosome regulatory factor that interacts with the supervillin C-terminus (Fig. 4A) (Gangopadhyay et al., 2004; Gimona et al., 2003). Another likely player is Tks5 (Fig. 4A), which interacts with the supervillin C-terminus and promotes podosome formation and matrix degradation (Crimaldi et al., 2009; Seals et al., 2005; Smith et al., 2010).

Because knockdown of supervillin and myosin IIA showed additive effects on matrix degradation, we suggest that supervillin and myosin IIA function in complementary pathways to degrade matrix. Our results are consistent with reports showing that siRNA-induced knockdown of supervillin decreases the number of invadopodia-induced holes in matrix degradation by MDA-MB-231 cells (Crowley et al., 2009) and that biochemical inhibition of myosin II reduces matrix degradation by invadopodia (Alexander et al., 2008). For instance, supervillin-mediated recycling of podosome components might be independent of myosin II function, whereas multiple pathways, separately including supervillin and gelsolin, could provide input to myosin-II-mediated mechanosensing and/or matrix metalloproteinase secretion (Alexander et al., 2008; Arora et al., 2011; Myers et al., 2011).

Our data also show functional redundancy between supervillin and gelsolin in the regulation of podosomal matrix degradation, but not in the regulation of podosome numbers. Gelsolin and supervillin are homologs that share amino acids required for the formation of five gelsolin-like folding repeats and residues in surface loops that are required for polyphosphoinositide binding by gelsolin (Janmey et al., 1992; Pestonjamasp et al., 1997; Smith et al., 2010). Gelsolin is a podosome regulator (Chellaiah et al., 2001) that localizes to all podosomes and binds F-actin but not myosin II. Functional redundancy between supervillin and gelsolin is consistent with gelsolin-independent formation of podosomes in dendritic cells (Hammarfjord et al., 2011) and with the observation that supervillin and gelsolin synergize to promote matrix invasion by breast cancer cells (Crowley et al., 2009). These results suggest commonalities in downstream functions, although the large structural differences between supervillin and gelsolin imply different signaling mechanisms.

In conclusion, we identify supervillin as a crucial regulator of podosome turnover, podosomal matrix degradation and cell polarization in primary human macrophages. We further show that supervillin-mediated coupling of L-MLCK-dependent myosin IIA contractility to podosomes is required for the regulation of podosome turnover. Moreover, our results reveal that podosome subpopulations in macrophages differ in their molecular make-up, and that their composition alters during their life cycle. The identification of supervillin as the first differentially localized component, its localization to a cap structure on top of the podosome core, and the supervillin-dependent recruitment of contractile myosin to podosomes thus shed new light on the increasingly apparent intricacies of podosome composition, architecture and turnover.

Human peripheral blood monocytes were isolated from buffy coats (provided by Frank Bentzien, University Medical Center Hamburg-Eppendorf, Hamburg, Germany) and differentiated into macrophages as described previously (Linder et al., 1999).

Cells were transiently transfected using the Microporator (Peqlab, Erlangen, Germany). For transfection of primary human macrophages, the following parameters were used: 1000 V, 40 mseconds, 2 pulses, 0.5 μg DNA per 1×10⁵ cells.

GFP–SV, mRFP–SV, GFP–SV(171–1792), GFP–SV(Δ343–570), GFP–SV(1–830), GFP–SV(1–174), GFP–SV(1010–1792), GFP–SV(830–1792), GFP–SV(571–830), GFP–SV(174–343), GFP–SV(343–570) were generated as described (Chen et al., 2003; Fang et al., 2010; Takizawa et al., 2007; Takizawa et al., 2006; Wulfkuhle et al., 1999). Lifeact–GFP was a gift from Michael Sixt (Max Planck Institute for Biochemistry, Martinsried, Germany) (Riedl et al., 2008), GFP–myosin-IIA was a gift from Robert Adelstein (NIH, Bethesda, MD) (Wei and Adelstein, 2000). GFP–L-MLCK was a gift from Anne Bresnick (Albert Einstein College of Medicine, New York, NY) (Poperechnaya et al., 2000).

For siRNA-induced knockdown, specific targeting siRNA was generated as follows: 5′-GAUCUGAACUCGUUUGAGCTT-3′ (myosin IIA, Eurofins MWG, Ebersberg, Germany) (Vicente-Manzanares et al., 2009); 5′-GGCCAAACCUGCCGAAUAA-3′ (myosin IIA, Eurofins MWG) (Ivanov et al., 2008); 5′-CAGCCAUAAGGAAUCUAAAUAUGCU-3′ (supervillin Stealth siRNA, Invitrogen) (Crowley et al., 2009); 5′-UAUUAAGGUAGAAAGGUUGAUUCGC-3′ (supervillin, Invitrogen) (Smith et al., 2010); 5′-CAGUUCUAUGGAGGCGACAGCUACA-3′ (gelsolin, Stealth siRNA, Invitrogen) (Crowley et al., 2009); a pool of 5′-GGGAUGACGAUGCCAAGUA-3′ and 5′-GCAAUGAUCUCAGGGCUCA-3′ (L-MLCK and S-MLCK, Dharmacon, ThermoScientific, Lafayette, CO); 5′-AGGUAGUGUAACCGCCUUGUU-3′ (firefly luciferase negative control siRNA (cat. no. D-001210-02-20, Dharmacon) (Kopp et al., 2006). Primary human macrophages were transfected with siRNA (50 nM) twice at 0 hours and 72 hours, and evaluated after a further incubation period of 5 hours.

U2OS cells (ATCC, Manassas, VA) were treated with control or supervillin-specific siRNA for 48 hours and lysed with 150 mM NaCl, 1% NP-40, 0.5% deoxycholic acid, 0.2% SDS, 50 mM Tris-HCl, pH 8.0, with a protease inhibitor cocktail (8340, Sigma). Lysates were sonicated, clarified and examined for depletion of supervillin by immunoblotting with H340 and HSV715 anti-supervillin antibodies (supplementary material Fig. S8A).

Cells were fixed for 10 minutes in 3.7% formaldehyde and permeabilized for 5 minutes in ice-cold acetone. F-actin in podosome cores was stained with phalloidin coupled to Alexa Fluor 568, Alexa Fluor 488 or Alexa Fluor 647 (Invitrogen). Vinculin was stained using mouse monoclonal antibody (V9264, Sigma), myosin IIA with rabbit polyclonal antibody (M8064, Sigma) and gelsolin with mouse monoclonal antibody (G37820, BD Transduction Laboratories, Heidelberg, Germany). An affinity-purified rabbit polyclonal antibody against supervillin (H340) has been described previously (Nebl et al., 2002; Oh et al., 2003). The HSV715 antibody was generated against human supervillin aa 715–728 by GenScript (Piscataway, NJ) and affinity purified as described (Nebl et al., 2002). Cells were stained for supervillin by fixation in −20°C methanol for 30 seconds, post-fixed with 3.7% formaldehyde for 20 minutes, and permeabilized with 0.1% Triton X-100. Phosphorylated myosin light chain (pMLC) was detected using an antibody raised against a peptide corresponding to aa 12–27 in human MLC, KKRPQRAT(pS)NVFAMFD, after conjugation to Keyhole Limpet Hemocyanin (ab2480, Abcam, Cambridge, UK). Secondary antibodies were Alexa-Fluor-568- or Alexa-Fluor-488-labeled goat anti-mouse or goat anti-sheep (Invitrogen). Coverslips were mounted in Mowiol (Calbiochem, Darmstadt, Germany) containing p-phenylendiamine (Sigma) as anti-fading reagent and sealed with nail polish.

Microscopy was performed as described (Kopp et al., 2006). Images of fixed samples were acquired with a confocal laser-scanning microscope (Leica DM IRE2 with a Leica TCS SP2 AOBS confocal point scanner) equipped with an oil-immersion plan Apo 63× NA 1.4 objective. Acquisition and processing of images was performed with Leica Confocal Software (Leica, Wetzlar, Germany).

To detect pMLC at podosomes, a protocol was created in Volocity (Improvision, Coventry, UK) based on fluorescence intensity thresholds and object size. The ROI tool was used to restrict the selected objects to the podosome area. Levels of pMLC at podosomes are expressed as ratios of pMLC area and podosomal area.

Images were acquired with a spinning disc confocal system (Spinning disc CSU22, Yokogawa, Japan) fitted on a Zeiss Axiovert 200M microscope with a temperature and CO₂ controllable environmental chamber (Solent Scientific, Regensworth, UK), oil-immersion plan Apo 63× NA 1.4 objective and a CCD camera (EM-CCD C-9100-2, Hamamatsu, Japan). Acquisition and processing of images was performed with Volocity Software (Improvision). Cells were seeded on glass-bottom dishes (Ibidi, Martinsried, Germany) at a density of 2×10⁵ cells/dish and incubated for 5 hours or 72 hours before the start of the experiment.

To evaluate podosome lifetime, cells were transfected with respective siRNAs. After 2 days, cells were re-transfected with siRNAs, as well as Lifeact–GFP and seeded on glass-bottom dishes. After further incubation for 24 hours, cells were imaged using time-lapse microscopy. Podosome lifetime was evaluated using Volocity (Improvision). Individual podosomes were tracked manually. Only those podosomes were evaluated that formed during imaging. At least ten podosomes were evaluated per cell.

Immunoblotting was performed by standard procedures, using primary antibodies: rabbit polyclonal myosin IIA (M8064, Sigma), mouse monoclonal myosin light chain kinase (MLCK, M7905, Sigma), mouse monoclonal actin (MAB1501, Millipore, Billerica, MA), and mouse polyclonal GFP, a gift from Jan Faix (Medical University Hannover, Hannover, Germany). Secondary antibodies were horse-radish-peroxidase-coupled anti-mouse or anti-rabbit IgG (Dianova, Hamburg, Germany). Protein bands were visualized by using Super Signal Pico or Femto kit (Pierce, Rockford, IL) and X-Omat AR film (Kodak, Stuttgart, Germany).

6×10⁶ cells were cultured for 7 days, and total RNA was isolated using 1 ml TRIzol Reagent (Invitrogen). DNA was removed by DNase digestion (Novagen, Madison, WI). For cDNA synthesis, 1 μg of random primer (Promega) was annealed to 2 μg of RNA for 5 minutes at 70°C, and first strand synthesis was performed using Moloney murine leukaemia virus reverse transcriptase (Promega). Second strand synthesis was performed using an oligonucleotide primer pair corresponding to nucleotides 3517–3569 and 4071–4094 of the SVIL gene. As a control for quantitative removal of residual DNA, oligonucleotide primers specific for an exon in the human β-actin gene were used, corresponding to nucleotides 1161–1142 and 716–735.

Immunoprecipitations of GFP-fusion proteins were performed using the MACS GFP Tagged Protein Isolation Kit (Miltenyi Biotec, Bergisch-Gladbach, Germany), according to the manufacturer's instructions. For lysis, preparation of columns and washing, the following buffers were used: lysis buffer [150 mM NaCl, 1% Igepal CA-630, 0.5% sodium deoxycholate, 0.1% SDS, 50 mM Tris-HCl, pH 8.0, with Complete Mini protease inhibitor (Roche Diagnostics)]; buffer 1 (150 mM NaCl, 1% Igepal CA-630, 0.5% sodium deoxycholate, 0.1% SDS, 50 mM Tris-HCl, pH 8.0); and wash buffer 2 (20 mM Tris-HCl, pH 7.5).

Porcine gelatin (ROTH, Karlsruhe, Germany) was fluorescently labeled with NHS-Rhodamine (Thermo Scientific) (Chen, 1996). Coverslips were coated with labeled gelatin solution, fixed in 0.5% glutaraldehyde (ROTH) and washed with 70% ethanol and medium. Cells were seeded on coated coverslips with a density of 8×10⁵ cells/coverslip. Cells previously transfected with siRNA and incubated for the times indicated were seeded on coated coverslips with a density of 8×10⁵ cells/coverslip and incubated for a further 5 hours, followed by fixation and staining.

ImageJ software was used to analyze Rhodamine-labeled gelatin fluorescence intensity (Abramoff et al., 2004). Values of matrix degradation were determined by loss of fluorescence intensity, with the intensity of adjacent undegraded areas set to 100%. For comparability, laser intensity was not changed between measurements. For each value, 30 cells were evaluated in each of three experiments. Statistical analyses were performed with Graphpad Prism 5 software, using the Student's two-tailed t-test, or ANOVA for multi-sample comparisons. Results are presented as means ± s.e.m., P<0.05 was considered as statistically significant, P<0.01 as highly statistically significant.

We thank Robert Adelstein for GFP–myosin IIA, Anne Bresnick for GFP–L-MLCK, Michael Sixt and Roland Wedlich-Söldner for Lifeact–mRFP, Jan Faix for anti-GFP antibody, Frank Bentzien (UKE transfusion medicine) for buffy coats, Norio Takizawa for help with supervillin knockdown, Vanessa van Vliet for help with mass spectrometry, Bernd Zobiak and Virgilio Failla (UKE microscopy facility) for help with confocal microscopy, and Laurent Blanchoin and Mirko Himmel for helpful discussions. We also thank Barbara Böhlig, Jens Cornils and Andrea Mordhorst for expert technical assistance, and Martin Aepfelbacher and Peter C. Weber for continuous support.