MAL/MRTF-A controls migration of non-invasive cells by upregulation of cytoskeleton-associated proteins

Cell migration is a key feature during tissue formation, wound healing and neoplastic processes. It is the result of a tightly coordinated interplay between cell–cell adhesion, cell–matrix adhesion and localised cytoskeletal reorganisation (Vicente-Manzanares et al., 2009; Friedl and Wolf, 2010). Extracellular signals and chemokines can directly and rapidly affect the directionality and speed of cellular movement by spatially controlling adhesion actin polymerisation (Vicente-Manzanares et al., 2009). However, migratory behaviour is predetermined by qualitative and quantitative differences in the expression of adhesive and cytoskeletal building blocks (Lauffenburger and Horwitz, 1996). Strengthening of cell matrix contacts through either increased integrin α5 expression or elevated fibronectin deposition slows cell motility of fibroblasts (Palecek et al., 1997). In epithelial cells, migration of individual cells is restricted by cell–cell contacts (Friedl and Wolf, 2010). The armadillo proteins plakophilin 1–plakophilin 3 localise to the cytoplasmic plaque of junctions and orchestrate desmosome formation (Bass-Zubek et al., 2009). Alterations in plakophilin expression results in impaired cell adhesion and migration (South et al., 2003; Grossmann et al., 2004; Kundu et al., 2008; Godsel et al., 2010).

The myocardin-related transcription factors MAL/MRTF-A (also known as MKL1) and MRTF-B (also known as MKL2) regulate through serum response factor (SRF) the expression of signalling molecules, transcription factors and numerous cytoskeletal components, including actin genes, non-muscle myosins and vinculin (Schratt et al., 2002; Cen et al., 2003; Miralles et al., 2003; Posern and Treisman, 2006; Sun, Q. et al., 2006; Descot et al., 2009; Medjkane et al., 2009). SRF depletion in mice results in embryonic lethality at gastrulation and impaired embryonic stem (ES) cell adhesion and motility (Arsenian et al., 1998; Schratt et al., 2002). MRTF double knockout in mouse embryonic fibroblasts (MEFs) impairs cell migration, consistent with an important role for SRF and MRTFs in cytoskeletal organisation and actin homeostasis (Mokalled et al., 2010). However, depletion of either MAL/MRTF-A or MRTF-B leads to relatively mild defects in differentiation of breast myoepithelial cells and cardiac neural crest cells, respectively, suggesting at least some functional redundancy (Li et al., 2005; Oh et al., 2005; Li et al., 2006; Sun, Y. et al., 2006). Recently, MRTFs have been implicated in experimental metastasis. Knockdown of MRTFs reduced adhesion, migration and invasion of highly invasive cancer cells, but also colonisation of distal organs (Medjkane et al., 2009). By contrast, the closely related SRF coactivator myocardin induces sarcoma cell differentiation, blocks malignant growth and acts as a tumour suppressor (Milyavsky et al., 2007). Activated MAL impairs cell proliferation, and actin–MAL signalling induces transcription of Mig6, a negative regulator of the epidermal growth factor receptor (EGFR) tyrosine kinase family (Descot et al., 2009).

Changes in actin dynamics, in turn, regulate transcription mediated by MAL–SRF (Sotiropoulos et al., 1999; Du et al., 2004; Kuwahara et al., 2005; Posern and Treisman, 2006). Serum-induced gene expression is directly controlled by monomeric G-actin, which forms a repressive complex with MRTFs (Posern et al., 2002; Miralles et al., 2003; Posern et al., 2004; Vartiainen et al., 2007; Mouilleron et al., 2008). Upon dissociation from G-actin, MAL/MRTF-A upregulates a Rho-family dependent, but MAPK-independent, subset of SRF target genes (Gineitis and Treisman, 2001; Vartiainen et al., 2007). These genes and their promoters respond differentially to actin binding drugs: treatment with cytochalasin D activates transcription by releasing MAL from G-actin, whereas latrunculin B stabilises the G-actin–MAL complex and inhibits gene expression (Posern et al., 2002; Posern et al., 2004; Vartiainen et al., 2007; Mouilleron et al., 2008).

Rearrangements of the actin cytoskeleton are thus communicated to the nucleus by MRTFs and establish an actin–MRTF–SRF circuit that can be assumed to play a pivotal role in motile cell functions (reviewed in Olson and Nordheim, 2010). It remains unclear, however, how MAL–SRF-mediated gene expression affects the motile behaviour of normal cells. Here, we analyse the migratory effect of MAL signalling in non-invasive highly adherent cells. Ectopic expression of active MAL caused impaired cell migration in wound healing and Boyden chamber assays, whereas a reduction of MAL/MRTF activity enhanced motility. A comparable antimigratory effect was observed in both mesenchymal fibroblasts and normal mammary epithelial cells. Invasive tumorigenic MDA-MB-231 cells, however, showed a reversed migratory response to changes in MAL activity, but in these less-adherent cells the basal transcriptional activity of MRTF–SRF was strongly reduced. To determine the responsible MAL targets, several newly identified G-actin-regulated genes (Descot et al., 2009) were characterised. Knockdown of three of these MAL targets, integrin α5, the four-and-a-half LIM domain protein FHL1 and plakophilin 2 (Pkp2), enhanced cell migration of either fibroblasts or epithelial cells and phenocopied the MRTF knockdown.

Many genes that are regulated by actin–MAL signalling to SRF encode cell adhesion and cytoskeletal components, and activated MAL increases cell flattening and spreading (Descot et al., 2009). To investigate whether MAL/MRTFs affect cell motility, we analysed the migration of non-invasive cell lines. NIH 3T3 fibroblasts were transiently infected with retroviral constructs harbouring MAL full length (fl.), the shorter version MAL (met), which starts at the first ATG, and the constitutively active MAL ΔN (Miralles et al., 2003). Conversely, the dominant negative constructs MAL ΔNΔB and MAL ΔNΔC were used, which either lack the SRF binding or the transactivation domain, respectively.

The directed motility of infected cells through the pores of uncoated transwell chambers was significantly impaired by active MAL (Fig. 1A,B). Similar results were obtained by analysing undirected migration in serum-containing medium, with the strongest inhibition observed for constitutively activated MAL ΔN lacking the negatively regulating RPEL motifs (Fig. 1C). During the 2–8 hours of these experiments, cell viability was unchanged, excluding that MAL-mediated antiproliferative effects were causative for the reduced motility (not shown). Moreover, dominant negative MAL enhanced cell motility considerably (Fig. 1A,B).

To test whether these effects exclusively occurred in cells squeezing through narrow pores of ~ 8 μM, two-dimensional wound healing assays were performed in the presence of mitomycin C to block cell cycle progression. The closure of preformed gaps in the cell layer was significantly reduced by MAL ΔN; MAL (met) also showed some inhibition, whereas wound closure was unaffected by MAL (fl.) (Fig. 1D,E). Vice versa, dominant negative constructs increased cell migration (Fig. 1F). The presence of the overexpressed MAL proteins was verified by western blot analysis (Fig. 1G). This suggests that MAL activity negatively correlates with fibroblast migration.

To assess whether the migratory effect of MAL is related to altered cell adhesion, transiently infected cells were plated on fibronectin and analysed for focal adhesion markers by immunofluorescence. Vinculin, a known MAL–SRF target, and integrin α5 colocalised to thickened and irregularly shaped patches of focal adhesions, which were often bundled at protrusion sites in MAL-infected cells after a short time of attachment (Fig. 2; supplementary material Fig. S1). Elongation of vinculin-containing focal adhesions was also observed after overnight adhesion of MAL-expressing cells. Under these conditions, integrin α5 was distinctly present in thickened patches that did not strictly colocalise with F-actin in fibrillar adhesions upon MAL expression (supplementary material Fig. S2). Phase contrast microscopy revealed that cells with activated MAL were more spread out, showed enhanced protrusions and covered a larger area, suggesting stabilised cell matrix adhesions (supplementary material Fig. S3).

To corroborate our findings from overexpression, knockdown by a short hairpin RNA (shRNA)-encoding vector targeting both MAL/MRTF-A and MRTF-B was performed (Medjkane et al., 2009). Knockdown by transient infection elevated the migration of NIH 3T3 cells in both transwell and wound closure assays (Fig. 3A,B). Analysis of the knockdown efficiency showed that some 40% of mRNAs encoding MAL and MRTF-B remained under the conditions used for migration assays (Fig. 3C), potentially explaining the relatively mild effects observed.

We therefore established clonal lines selected for improved knockdown efficiency. Using luciferase reporter assays, the basal transcriptional MRTF–SRF activity was significantly reduced at least threefold in four independent lines; inducibility by either serum or cytochalasin D was strongly diminished (Fig. 3D,E). When assayed for cell motility, a significant increase in directed and undirected transwell migration as well as in wound closure was observed for the MRTF double knockdown lines, in comparison with parental and control vector infected NIH 3T3 cells (Fig. 3F,G). Changes were more prominent for undirected transwell motility, either reflecting the lower basal migration or an enhanced dependency on MRTFs under these conditions. Together, the results demonstrate that MAL impairs migration of non-invasive, mesenchymal cells.

The motile behaviour of fibroblasts differs considerably from that of epithelial cells. Therefore, a potential cell-type specificity of the MAL effects was investigated in the mouse mammary epithelial cell line EpRas, which exhibits many features of untransformed epithelial cells (Oft et al., 1996). Stable infection of active MAL into these cells significantly impaired migration in transwell chambers, whereas dominant negative constructs enhanced undirected and directed motility (Fig. 4A,B; and data not shown). The migratory behaviour correlated with the activity of the MAL–SRF luciferase reporter (Fig. 4C); immunoblot analyses showed only mild overexpression in the stably infected pools (data not shown).

Moreover, knockdown of MRTFs significantly and sharply increased the number of migrated cells in transmigration assays (Fig. 4D). Interestingly, the effect of MRTF reduction was again more pronounced for undirected migration in the absence of serum. This suggests that MAL-controlled repression of EGFR–MAPK signalling through Mig6 (Descot et al., 2009) is unlikely to entirely explain the migratory effects. Together, the findings demonstrate antimigratory functions of MAL-mediated transcription in both untransformed mesenchymal and epithelial cells, irrespectively of their fundamental differences.

In metastatic human breast cancer MDA-MB-231 cells, MRTF knockdown has recently been shown to inhibit motility, invasion and colonisation of distal organs (Medjkane et al., 2009). We addressed this apparent discrepancy with our results obtained in highly adherent, untransformed fibroblasts and epithelial cells. First, a significant antimigratory effect of MRTF double knockdown using short interfering RNAs (siRNAs) in MDA-MB-231 cells could be reproduced (Fig. 4E,F). Similar results were obtained with the shRNA constructs (data not shown). Also, knockdown efficiencies were comparable with our previous experiments: mRNAs of MRTFs were partially reduced, and reporter inducibility diminished (supplementary material Fig. S4). This excluded underlying experimental differences, but strongly pointed towards opposing functions of MAL in tumorigenic and normal cells.

To substantiate this, active MAL was transiently overexpressed, resulting in enhanced reporter activity, as expected (supplementary material. Fig. S4B). In stark contrast to fibroblasts and EpRas epithelial cells, MAL enhanced the motility of MDA-MB-231 cells in both transwell and wound closure assays (Fig. 4E,F). Similarly, invasive EpRasXT cells, which alter their adhesion properties when undergoing TGF-β induced epithelial–mesenchymal transition, showed enhanced transwell migration upon expression of active MAL (supplementary material Fig. S5).

As a possible cause for effects on migration, cell adhesion was then studied. Attachment of MDA-MB-231 cells to coated surfaces was enhanced by active MAL, with most prominent effects obtained using low fibronectin concentrations (Fig. 4G). Conversely, MRTF knockdown reduced adhesion. This indicates that, in invasive tumorigenic cells, migration correlates with adhesion as the motility-limiting factor. However, noninvasive cells such as fibroblasts and epithelial cells are highly adherent and attached in less than10 minutes (data not shown). Further strengthening of adhesion by MAL, as indicated by the more spread-out morphology, might therefore repress the migratory capacity (supplementary material Fig. S3) (Descot et al., 2009; Medjkane et al., 2009).

Because MRTFs are implicated in cell adhesion, we compared the basal transcriptional activity of the MRTF–SRF module between the three cell lines, using a mutated reporter for normalisation (Busche et al., 2008). Whereas NIH 3T3 and EpRas showed comparable results and differed only when cultivated in serum, MRTF–SRF activity was strikingly reduced in MDA-MB-231 cells (Fig. 4H). Protein expression of MAL and MRTF-B was also diminished (Fig. 4I). This suggests that MRTFs determine cellular adhesion, and provides an explanation for the observed opposing effects of MAL in normal and tumorigenic cells (see Discussion).

To investigate which genes potentially mediate the antimigratory effect of MAL in fibroblasts, we re-analysed our previous whole genome microarrays for genes differentially expressed by signalling through G-actin (Descot et al., 2009). In addition to the many well-characterised cytoskeletal components, several novel G-actin-regulated targets with potential roles in migration were identified (Fig. 5A). These included two independent probe sets for plakophilin 2 (Pkp2) and integrin α5 (Itga5), as well as FHL1 (Fhl1). The statistical significance of these microarray results was high, as indicated by low q-values. Validation by quantitative reverse transcription PCR (qRT-PCR) showed that all mRNAs were upregulated by cytochalasin and repressed by latrunculin, comparable with the microarray data (Fig. 5B). Together, this suggests that the genes analysed responded directly to changes at the G-actin level, which is known to control MAL/MRTF activity.

To investigate the potential relevance for cell migration, transient transfection of siRNAs specific for these genes was performed. None of the siRNAs against Fam126b, Ereg, Lima1, Nexn and Plaur, which showed a specific knockdown, led to a reduction of motility (data not shown). Instead, the siRNAs targeting Itga5 and Pkp2 significantly enhanced the migration distance in NIH 3T3 wound closure assays (Fig. 5C). Similarly, Itga5 and Pkp2 knockdown also increased transwell motility (Fig. 5D). Knockdown efficiency was found to be >70%, based on mRNA levels (Fig. 5E), and the protein levels were strongly decreased (Fig. 5G, upper panels). A second and even more efficient siRNA for Pkp2 was also tested, but its migratory effects were not interpretable due to cytotoxicity in NIH 3T3 cells (data not shown).

In EpRas cells, knockdown of Itga5, Pkp2 and Fhl1 strongly and significantly enhanced transwell motility (Fig. 5F). Surprisingly the basal Itga5 mRNA level in EpRas cells was comparable with that in NIH 3T3 cells (data not shown), and integrin α5 protein was also readily detected (Fig. 5G). This was unexpected because epithelial cells normally do not come into contact with fibronectin. However, of the two Itga5-targeting siRNAs, the one with the stronger knockdown efficiency resulted in a weaker effect in both cell lines, suggesting that an integrin α5 threshold is required for basal migration. In the epithelial EpRas and MDCK cells, Pkp2 protein level was much higher than in fibroblasts, whereas basal FHL1 was only detectable by its mRNA (Fig. 5F,G). Nevertheless, the knockdown efficiency of both factors nicely correlated with the enhanced transmigration, indicating that Pkp2 and FHL1 limit the migratory potential of epithelial cells.

To show that Fhl1, Pkp2 and Itga5 are directly regulated by MAL, we transiently infected NIH 3T3 cells with active MAL and analysed their endogenous mRNA. Fhl1 was upregulated almost 100-fold by transient infection with MAL ΔN; significant upregulation of both mRNA and protein was also observable with MAL(met) and MAL (fl.) (Fig. 6A,D). Pkp2 mRNA was induced 12-fold by activated MAL (Fig. 6B). Again, protein levels were also increased (Fig. 6D). Itga5 mRNA was upregulated tenfold and sixfold by MAL ΔN and MAL (met), respectively (Fig. 6C). Conversely, shRNA-mediated knockdown of MRTFs resulted in significantly decreased Fhl1-, Pkp2- and Itga5-inducibility by serum (Fig. 6A-C, right panels). In addition, integrin α5 protein was upregulated by MAL ΔN and serum stimulation, but impaired by MRTF knockdown (Fig. 6D).

Protein induction of the three putative MAL targets was further analysed in cells treated with inducers of endogenous actin–MAL signalling. Epithelial MDCK cells activate MAL–SRF mediated transcription upon dissociation of E-cadherin, e.g. by reduction of extracellular calcium (Busche et al., 2008; Busche et al., 2010). Indeed, dissociating cell–cell contacts by calcium withdrawal increased Pkp2 protein level (Fig. 6E). Similarly, FHL1 and integrin α5 were induced in MEF cells treated with cytochalasin. Together, these data indicate that Fhl1, Pkp2 and Itga5 are commonly regulated by the actin–MAL pathway.

The Fhl1 gene contains a conserved consensus CArG box in its first intron and is activated by SRF-VP16 in reporter assays, strongly suggesting its direct transcriptional regulation by MAL–SRF recruitment (Sun, Q. et al., 2006). In the Itga5 promoter we identified in silico four potential SRF binding sites, which are CArG-like elements (Fig. 7A). To see whether any of these mediate activation by the actin–MAL pathway, we performed luciferase reporter assays with consecutive deletions of the Itga5 promoter. The 2.3 kb promoter and two deletions showed significant induction by MAL ΔN, cytochalasin or the MAL-activating mutant actin G15S, as well as repression by latrunculin pretreatment or co-transfection of the nonpolymerisable mutant actin R62D (Fig. 7B; supplementary material Fig. S6). However, removing the most distal CArG box decreased the induction, and further shortening had no effect, suggesting a role for CArG 1, but not for CArG 2. Mutation of either CArG 1 or CArG 4, which disrupt SRF binding (Hill and Treisman, 1995), reduced inducibility by MAL ΔN to ~30% and diminished the response to actin G15S, whereas mutation of both CArG boxes completely abolished reporter regulation (Fig. 7B; supplementary material Fig. S6). This revealed an important role of the most distal and proximal CArG-like elements.

Next, we analysed the recruitment of MAL and SRF to Itga5 promoter fragments comprising the identified CArG boxes using chromatin immunoprecipitation with specific antibodies. MAL and SRF were recruited to CArG 1 and CArG 4 of the Itga5 promoter after cytochalasin stimulation, comparable with the known MAL target gene Srf, which served as a positive control (Fig. 7C). Quantification by real time PCR revealed the strongest signals for CArG boxes 1 and 4, consistent with the hypothesis that these CArG boxes are responsible for inducible SRF and MAL binding in the native chromatin context of the Itga5 promoter (Fig. 7D). Together, these results demonstrate that two SRF binding sites act together to regulate the Itga5 promoter in response to actin–MAL–SRF signalling.

The Pkp2 gene contains a CArG-like element in the first intron at nucleotide 2894 after the transcription start site, which is conserved in human, rat and mouse. Upon treatment with cytochalasin, MAL was readily recruited to this chromatin region, similarly to the known MAL target gene Srf (Fig. 7E). The quantification showed a significant fivefold induction of MAL-binding to this cis-regulatory element (Fig. 7F). Recruitment of SRF to Pkp2 was also increased, in contrast toits more constitutive binding at the Srf promoter (Fig. 7C,E,F). This result suggests that Pkp2 expression by actin–MAL signalling is facilitated by direct binding of MAL and SRF to the Pkp2 gene.

To investigate whether the migratory effects of MAL observed in the stable EpRas cell lines could be mediated by regulation of the MAL targets Itga5, Pkp2 and Fhl1, their relative mRNA level was determined. Pkp2 and Fhl1 were elevated in EpRas stably expressing MAL ΔN and MAL (met), whereas Itga5 was essentially unaffected (Fig. 8A). Conversely, Fhl1 was considerably decreased by the dominant negative MAL constructs, whereas Itga5 and Pkp2 were little affected. Further analysis of the EpRas cells harbouring a stable but partial MRTF knockdown revealed a significant reduction of all three target genes, with the strongest effect on Fhl1 (Fig. 8B). Together, the MAL targets Pkp2 and Fhl1 could potentially contribute to the migratory effects observed in the EpRas cell lines, with Fhl1 being a promising candidate because its strongly regulated expression correlates well with the transmigration data shown in Fig. 4A-D.

We therefore tested whether knockdown of Fhl1, Pkp2, or both, restores part of the antimigratory effect of MAL. EpRas cells stably expressing MAL (met) showed a more than twofold increase in migrated cells upon transient knockdown of either Fhl1 or Pkp2, compared with the cells transfected with a control siRNA (Fig. 8C). The double knockdown of Fhl1 in combination with Pkp2 showed a significant and even stronger enhancement of transmigration. However, migration was only partially restored when compared with the cells not stably overexpressing MAL (met) (Fig. 8C). Analysis of the mRNA levels in this experiment revealed a Pkp2 knockdown efficiency of between 60% and 90% (Fig. 8D). The knockdown efficiency of Fhl1 was around 70%; thus the remaining Fhl1 mRNA amount in the MAL-expressing cells was still three times higher than in the control infected cells, which expressed around 10% (Fig. 8D). Therefore, the role of FHL1 in this rescue experiment is probably underestimated, due to limitations in the Fhl1 knockdown efficiency. Taken together, these results show that knockdown of Fhl1 and Pkp2 partially reverses the antimigratory MAL effect and suggest that their upregulation, probably in conjunction with other targets, is involved in MAL-regulated cell motility.

We have shown that activated MAL/MRTFs inhibit migration of non-invasive, highly adherent cells. Conversely, reduction of MRTF activity by dominant negative constructs or knockdown enhances motility. The cytoskeleton-associated factors integrin α5, Pkp2 and FHL1 were newly characterised as transcriptionally regulated actin–MAL–SRF target genes, and they were in part responsible for the impaired migration of cells harbouring activated MAL.

How does the induction of cytoskeleton-associated proteins inhibit motility of non-tumorigenic cells? Strengthening of cell adhesion and stabilisation of cytoskeletal structures by elevated levels of building blocks might underlie the antimigratory effects. A biphasic relationship between cytoskeletal components and motility might exist, where maximum migration occurs at intermediate adhesiveness. This has been nicely demonstrated for integrin α5: both too much and too little of the receptor (or its ligand fibronectin) reduces migration (Palecek et al., 1997). Indeed, the weaker knockdown of integrin α5 in our experiments had a stronger promigratory effect (Fig. 5), which is consistent with abrogated motility upon complete integrin α5 depletion. On the basis of our results, we propose that the biphasic model applies beyond integrins to other cytoskeleton-associated factors and probably to MAL activity itself (Fig. 8E).

MAL might generally influence adhesiveness and cytoskeletal stability through the regulation of target genes, thereby controlling motility. The highly adherent cells investigated here might harbour an adhesive or cytoskeletal repertoire that is beyond the migratory optimum. Consistent with this hypothesis, more MAL activity inhibited migration, whereas its partial knockdown lead to an enhancement. Although we were unable to achieve a complete knockdown of MRTFs in our experiment, our model predicts that total depletion of MRTF activity would nevertheless block migration. Indeed, MEFs from recently described MAL/MRTF-A and MRTF-B double knockout mice showed a severe impairment in wound closure (Mokalled et al., 2010).

In invasive cancer cell lines, however, the migratory effects of MAL are exactly the opposite, in line with a recent report (Medjkane et al., 2009). By confirming the knockdown results and extending them to active MAL, we investigated the reason for the apparent contradiction. We think that the biphasic model provides an explanation. Invasive cells such as MDA-MB-231 are less adherent to the ECM or to neighbouring cells and probably have a reduced cytoskeletal repertoire. Intriguingly, MDA-MB-231 cells have strongly diminished basal activity of the MRTF–SRF module (Fig. 4H). We therefore suspect that the MRTF activity and the repertoire of adhesive components is below a threshold required for optimal migration. Further knockdown of MRTFs leading to decreased cytoskeleton-associated targets therefore reduces motility and adhesion in these cells, whilst activated MAL has the opposite effect. An alternative explanation, however, includes the activation of oncogenes and multiple signalling pathways in transformed and invasive cells, which distinctly affect the migratory response to altered MAL target gene expression.

Although fibroblasts and normal epithelial cells are thought to differ considerably in their means of cell motility, cell–cell and cell–matrix adhesion, the antimigratory effects of MAL were strikingly comparable. In fact it was somewhat surprising to see that EpRas cells which form tight epithelial cell contacts readily migrated through transwell filters. It remains unclear whether this non-invasive motility represents individual cell migration of the amoeboid or mesenchymal type, or multicellular streaming (Friedl and Wolf, 2010). However, strengthening of cell–cell contacts by e.g. MAL-mediated Pkp2 upregulation potentially converts the mode of EpRas motility to collective migration, which is unlikely to be scored correctly in transwell assays.

Pkp2 is essential to coordinate actin-dependent maturation of desmosome precursors by interacting with the intermediate filament-desmoplakin complex, PKCα, and possibly F-actin (Grossmann et al., 2004; Bass-Zubek et al., 2009; Godsel et al., 2010). Amongst the plakophilins, Pkp2 has a relatively broad expression and is also detectable in fibroblasts (Fig. 5 and 6) and mesenchymal cells where it colocalises with classical cadherins at sites of cell–cell contacts (Rickelt et al., 2009). Thus it appears likely that the promigratory effect of Pkp2 knockdown reported here is mediated by weakening of cell–cell contacts, both in fibroblasts and epithelial cells. Intriguingly, the intermediate filament system has also been implicated in control of cell migration (Eckes et al., 1998; Beil et al., 2003). However, we cannot exclude a role for other functions of Pkp2 in the nucleus, at catenin-containing junctions, or on PKC regulation.

We identify and characterise Pkp2 as a novel actin–MAL–SRF target gene, as concluded from its transcriptional induction in the unbiased microarray screen, its MAL-controlled expression pattern, and its chromatin IP with MAL- and SRF-specific antibodies. Intriguingly, the recently described skin-specific SRF knockout resulted in a hyperproliferative skin disease, associated with impaired desmosome formation (Koegel et al., 2009). Whether expression of Pkp2 in these cell layers is affected by SRF depletion, or even causative for desmosome malformation, remains to be investigated.

We also characterised the molecular details of the integrin α5 regulation by MAL. Found in the microarray for G-actin-regulated targets, the Itga5 gene is directly controlled by RhoA and SRF (Descot et al., 2009), and by actin–MAL (Fig. 6). We further identify another conserved CArG-like element (CCTCATTAGG) in the proximal Itga5 promoter, which is regulated by SRF, in addition to the distal CArG box previously described following bioinformatics (Sun, Q. et al., 2006). In fact, MAL and SRF are much more strongly recruited to the proximal CArG box (Fig. 7). However, both CArG-like elements are required for proper regulation, and our data suggest a synergistic effect in cis on promoter activity, possibly by oligomerisation of MAL. We note that the CArG-likeelements in the Itga5 promoter, as well as the intronic CArG-like element in Pkp2 (CCTTGTAAGG), recruit MAL and also SRF mainly upon induction. This differs slightly from genes that contain a consensus CArG box in their proximal promoters (such as vinc, srf, cyr61 and eplin), where SRF is usually bound in a more constitutive manner (Miralles et al., 2003; Descot et al., 2009; Leitner et al., 2010). The discrepancy suggests a cooperative binding of MAL and SRF to these non-consensus sites, which might be stabilised further by MAL-mediated DNA contacts.

The LIM-only proteins FHL1–FHL3 (also known as SLIM1–SLIM3) are thought to act as scaffolds and interact with F-actin and focal adhesions (Robinson et al., 2003; Shen et al., 2006; Wixler et al., 2007). Knockdown of FHL1 clearly enhanced motility of epithelial EpRas cells and partially reversed the antimigratory effects of MAL. The Fhl1 mRNA was readily detectable and strongly and consistently affected by all MAL constructs in EpRas cells (Fig. 8A,B). Although FHL1 had little effect on fibroblast migration, its expression at both mRNA and protein levels was also tightly controlled by actin–MAL signalling in NIH cells, strongly suggesting that Fhl1 is a direct MAL–SRF target gene. In line with this, FHL1 has a conserved and consensus intronic CArG box, which facilitates strong reporter activation by SRF-VP16 (Sun, Q. et al., 2006). Moreover, the closely related Fhl2 gene is a known MAL–SRF target, and expression of FHL1 follows that of FHL2 (Philippar et al., 2004). We thus conclude that FHL1 plays an important role in mediating the antimigratory effects of MAL, at least in epithelial cells. In Src-transformed fibroblasts, FHL1 has also been shown to inhibit migration unless it is suppressed by Cas phosphorylation (Shen et al., 2006). FHL1 localises to focal adhesions and the nucleus in myoblasts, where it affects adhesion, spreading and migration (Robinson et al., 2003). However, its role probably goes beyond a simple cytoskeletal function because the LIM domains facilitate a wide range of protein–protein interactions. Interestingly, FHL1 negatively correlates with tumour progression (Shen et al., 2006; Ding et al., 2009) and is thus another addition to the increasing number of MAL–SRF-regulated tumour suppressor genes, such as those encoding Mig6 and Eplin-α (Descot et al., 2009; Leitner et al., 2010).

The functional analysis in this study deliberately focussed on previously unknown G-actin-regulated genes. Our results indicate that Itga5, Pkp2 and Fhl1 are good candidates to mediate part of the antimigratory MAL effects. A negative crosstalk through Mig6 into the EGFR-MAPK signalling axis might also contribute to the antimigratory MAL effects (Descot et al., 2009), but it is unlikely to fully explain the effects observed because they were recapitulated in serum-free conditions. In addition, other known cytoskeletal targets probably contribute to MAL-mediated restriction of cell motility. The MAL-regulated focal adhesion protein vinculin stabilises cell contacts, and its depletion results in accelerated motility (Xu et al., 1998; Cen et al., 2003). Enhanced actomyosin contractility has been shown to stabilise focal adhesions and results in stress fibre bundling, which is not compatible with rapid cycles of de-adhesion, re-adhesion and F-actin remodelling (Chrzanowska-Wodnicka and Burridge, 1996). Conversely, depletion of non-muscle myosin IIA impairs stress fibre formation and focal adhesion assembly and results in increased motility (Even-Ram et al., 2007). Importantly, non-muscle myosin IIA is encoded by the MAL–SRF target gene Myh9 and is associated with the myosin light chain Myl9, which is also a MAL target (Descot et al., 2009; Medjkane et al., 2009). Cells expressing activated MAL show enlarged focal and fibrillar adhesions (Fig. 2; supplementary material Figs S1, S2), a strongly spread-out morphology (supplementary material Fig. S3) (Descot et al., 2009; Medjkane et al., 2009) and increased adhesion (Fig. 4G), which might be indicative of altered motile properties. In line with this, expression of constitutively active MRTFs in Ras- or Src-transformed rat intestinal epithelial cells reversed the morphological abnormalities and impaired invasiveness and anchorage-independent growth (Yoshio et al., 2010).

Together, it is conceivable that the transcriptional upregulation of the cellular adhesive and contractile machinery by MAL results in a net reduced migration of non-invasive cells. Endogenous MAL and target gene expression, however, is in turn activated by increased actin dynamics and Rho family activation, which also occurs upon initiation of migration, e.g. after wounding. Thus, MAL-mediated transcription might be at the centre of a negative feedback loop in order to blunt inappropriate migration of non-transformed cells.

The Itga5 luciferase constructs were created by cloning promoter fragments of varying lengths from murine liver tissue into the luciferase reporter plasmid pGL3. CArG box 1 was mutated to ACCTATCGGG, and CArG box 2 to CCCCATCGGG, using the overlap extension method. pLPCX–MAL constructs were described previously (Descot et al., 2009). Deletions in murine MAL were as follows: met, 1–91; ΔN, 1–171; ΔB1, 316–341; ΔC, 563–1021; numbering based on MAL (fl.) (Miralles et al., 2003). pSuperRetro-Mrtf was cloned by introducing 5′-GATCCCCGCATGGAGCTGGTGGAGAAGAATTCAAGAGATTCTTCTCCACCAGCTCCATGTTTTTGGAAA-3′, which generates an shRNA perfectly matching the sequence 5′-ATGGAGCTGGTGGAGAAGAA-3′ of both murine MAL and MRTF-B (Medjkane et al., 2009). The SRF luciferase reporter p3D.A-Luc and the mutated p2M.A-Luc used for inter-cell line normalisation have been described (Busche et al., 2008). NIH 3T3 were cultivated in DMEM (Gibco) supplemented with 10% fetal calf serum (FCS; Gibco). MDA-MB-231 were grown in RPMI medium (Gibco) supplemented with 10% FCS. EpRas were cultivated in DMEM supplemented with 4% FCS. To generate cell lines stably expressing various MAL constructs and the shRNA targeting MAL and MRTF-B, NIH 3T3 and EpRas cells were retrovirally infected and selected for 10 days in 6 μg/ml puromycin and 3 μg/ml puromycin, respectively. Cytochalasin D, latrunculin B and puromycin were purchased from Calbiochem (Beeston, UK). Cycloheximide and mitomycin C were from Sigma (Taufkirchen, Germany).

Transfection of NIH 3T3 and MDA-MB-231 cells with siRNA was carried out using RNAiMAX (Invitrogen, Karlsruhe, Germany). Cells were seeded at low density of 8.5×10⁴ NIH 3T3 cells and 1.7×10⁵ MDA-MB-231 cells per 6-cm dish and transfected the next day with 60 and 48 pmol siRNA, respectively. For electroporation, 5×10⁶ EpRas cells in 200 μl Opti-MEM (Gibco) were mixed with 600 pmol siRNA in a 4-mm cuvette (Bio-Rad). Electroporation was performed with a GenePulser Xcell with CE and PC modules using the time constant protocol (voltage, 250 V; pulse length, 70 ms). siRNAs used are listed in supplementary material Table S1.

Transfection of MDA-MB-231 cells with plasmid DNA was carried out using Lipofectin (Invitrogen); 1.7×10⁵ cells per 6-cm dish were transfected with 6 μg of plasmid DNA. Transfection of EpRas cells was carried out with Lipofectamine 2000 (Invitrogen); 3.5×10⁴ cells per well of a 12-well plate were transfected with 1.6 μg plasmid DNA.

For luciferase assays, 3.5×10⁴ EpRas cells per 1-cm dish were transfected with 80 ng luciferase reporter plasmid and 160 ng pRL-TK in a total of 1600 ng DNA using Lipofectamine 2000 (Invitrogen). Some 3.5×10⁴ NIH 3T3 cells and 7×10⁴ MDA-MB-231 cells per 1-cm dish were transfected with 25 ng luciferase reporter plasmid, 50 ng pRL-TK and 10 ng of MAL-expressing plasmid in a total of 500 ng DNA using Lipofectamine (Invitrogen).

For retroviral infection, 10⁷ cells of the retroviral packaging line Phoenix E were transfected on a 15-cm dish with 50 μg of plasmid using calcium phosphate. Following virus production for 24 hours, the virus-containing medium was filtered through a 0.45 μm PVDF membrane (Millipore, Schwalbach, Germany), concentrated on a Vivaspin 20 column (30,000 Da molecular mass cut-off,polyethersulfone membrane; Sartorius, Goettingen, Germany), and used to infect 1.53×10⁵ NIH 3T3 cells seeded in 6-cm dishes in the presence of polybrene (8 μg/ml). The procedure was repeated 8 hours later. For retroviral infection of EpRas cells, the virus-containing supernatant of 3×10⁷ Phoenix A cells was used to infect 3.36×10⁵ cells per 6-cm dish.

Luciferase reporter assays were carried out using the Dual-Glo luciferase assay kit according to manufacturer's protocol. Optionally, cells were serum-starved upon transfection for 24 hours following treatment with 2 μM cytochalasin D or 15% FCS for 7 hours. Luciferase activity was determined and normalised as previously described.

Immunoblotting was done using anti-Pkp2 (1:100; Acris Antibodies, Hiddenhausen, Germany), anti-FHL1 (1.1000; ProteinTech Group, Chicago, IL), anti-integrin α5 (1:1000; Millipore), anti-tubulin (1:5000; Sigma), anti-HA peroxidase conjugate (1:700; Roche, Penzberg, Germany) and anti-MAL (1:1000; homemade rabbit antiserum #79).

RNA preparation (QIAGEN; Hilden, Germany) and first-strand cDNA synthesis (Thermo Scientific, Epsom, UK) were done according to the manufacturers' protocols. For cDNA synthesis, 1 μg of total RNA and 500 ng of anchored oligo-dT primer were used. For cDNA quantification, one-fortieth of the reverse transcriptase reaction was mixed with gene-specific primers (0.5 μM) and Fast SYBR Green Master mix (Applied Biosystems, Darmstadt, Germany) to a total volume of 15 μl. The PCR was carried out on a StepOnePlus instrument (Applied Biosystems), according to the manufacturer's instructions. Calculations were done using the ΔΔCt method (Winer et al., 1999). Gene-specific primers used are listed in supplementary material Table S2.

For each immunoprecipitation, 2×10⁷ NIH 3T3 cells were stimulated with 2 μM cytochalasin D for 40 minutes and fixed with 1% formaldehyde for 10 minutes at room temperature. Cross-linked cells were sonicated in RIPA buffer (0.1% SDS) using a Bandelin HD2200 sonicator with a MS72 tip. Then, 800 μl of sonicated chromatin was incubated overnight at 42°C with 5 μg of anti-SRF (G-20; Santa Cruz Biotechnology, Heidelberg, Germany) or home-made anti-MAL rabbit serum (#79) pre-coupled to 40 μl of protein G magnetic beads (Invitrogen). Following five washes with LiCl buffer and one wash with Tris–EDTA buffer, the DNA–protein complexes were eluted with 0.1 M NaHCO₃ containing 0.1% SDS. Cross-links were reversed overnight at 65°C, and DNA purified using a PCR purification kit (QIAGEN). Quantification was done by real time PCR and is shown as the percentage of input chromatin. Gene-specific primers for amplification of immunoprecipitated DNA are listed in supplementary material Table S3. Primers for Gapdh and Srf were published previously (Vartiainen et al., 2007).

For wound closure assay, 2×10⁴ NIH 3T3 cells and 6×10⁴ MDA-MB-231 cells in 70 μl medium were seeded into each of two reservoirs of migration inserts (Ibidi, Martinsried, Germany). After attachment overnight 10 μg/ml mitomycin C containing medium was added, the inserts were removed, and cells started to migrate into the separating zone of 500 μm between the two cell patches. Micrographs were taken at time point zero and after 12 or 24 hours. After precisely marking the migration front, the free area devoid of migrated cells was measured using the MetaVue software. The area difference of experiment starting and end point was normalised to the control.

Cell were seeded into polycarbonate transwells of 8 μm pore size at a density of 9×10⁴ (EpRas), 3.75×10⁴ (NIH 3T3), and 3×10⁴ (MDA-MB-231) cells in 300 μl medium. Migration conditions were as follows: EpRas, directed movement: 16 hours (13 hours in the case of Itga5 knockdown) from serum-free medium containing 0.2% BSA towards 10% FCS; EpRas, undirected movement: 23 hours in serum-free medium; NIH 3T3, directed movement: 2 hours from serum-free medium towards 1% FCS; NIH 3T3, undirected movement: 8 hours in 10% FCS; MDA-MB-231, directed movement: 16 hours from serum-free medium towards 10% FCS. After removing the non-migrating cells and staining the transmigrated cell with crystal violet, micrographs were taken at 5× magnification. Cell migration was deduced by measuring the membrane area covered with migrated cells using the Photoshop CS3 extended measurement feature.

For immunofluorescence microscopy, cells were plated 40 hours after transient infection on coverslips coated with fibronectin (2 μg/cm²). After attachment, cells were fixed with 4% paraformaldehyde in PBS, permeabilised in 0.2% Triton X-100, blocked with 3% BSA and 0.05% Triton X-100 in PBS, and washed in 0.05% Triton X-100 in PBS. Staining condition were as follows: mouse anti-vinculin (1:1000; clone hVIN-1, Sigma-Aldrich), secondary antibody was Alexa Fluor 488 anti-mouse (1:500; Molecular Probes); rabbit anti-ITGA5 (1:1000; AB1928, Millipore), secondary antibody was Alexa Fluor 488 anti-rabbit (1:500; Molecular Probes); Alexa Fluor 546–phalloidin (1:200; Molecular Probes). Micrographs were taken with a 63× NA 1.4 objective using a Zeiss Axioplan 2 with MetaVue software (Molecular Devices).

For adhesion assays, 96-well plates were coated with 5 μg/cm² fibronectin (high concentration), 0.625 μg/cm² fibronectin (low concentration), or 0.01% (w/v) poly-L-lysine (Sigma-Aldrich) in PBS at 4°C overnight. Coating with poly-L-lysine allowed receptor-independent cell attachment and served as an unspecific positive control. Wells were washed once with PBS and blocked with 1% BSA in PBS for 1 hour at 37°C. MDA-MB-231 cells were detached using PBS–EDTA (10 mM) and washed twice with PBS. Then, 3.5×10⁴ cells were plated in serum-free medium per well. After attachment for 30 minutes at 37°C, non-adherent cells were washed away. Adherent cells were fixed with 1% formaldehyde for 10 minutes at room temperature and stained with crystal violet. Micrographs were taken at 4× magnification. Cell migration was deduced by measuring the membrane area covered with migrated cells using the Photoshop CS3 extended measurement feature.

We thank Philip Vlaicu for providing the algorithm to quantify migration in Photoshop, and Monika Rex-Haffner for technical assistance.