Missing-in-metastasis MIM/MTSS1 promotes actin assembly at intercellular junctions and is required for integrity of kidney epithelia

The actin cytoskeleton is a key regulator of cell morphology and migration. Consequently, it is crucial for embryonic development and many physiological functions in adult animals. Abnormalities in the function of the actin cytoskeleton are also central in cancer, and many actin-binding proteins have been directly linked to the metastatic behavior of cancer cells (Wang, W. et al., 2007a).

Processes mediated by the actin cytoskeleton, such as the formation of plasma membrane protrusions during cell migration, are often closely linked to direct membrane remodeling (Chhabra and Higgs, 2007). Recent studies demonstrated that the members of an extended protein family, characterized by the presence of a membrane binding and deforming BAR (Bin–Amphiphysin–Rvs) domain, function at the interface between the actin cytoskeleton and plasma membrane during the formation of membrane protrusions or invaginations (Doherty and McMahon, 2008; Frost et al., 2009; Itoh et al., 2005; Peter et al., 2004; Takano et al., 2008). These proteins can either generate positive membrane curvature to facilitate the formation of plasma membrane invaginations (e.g. BAR, N-BAR and F-BAR domain proteins) or induce negative membrane curvature to promote the formation of plasma membrane protrusions (I-BAR and IF-BAR domain proteins) (Guerrier et al., 2009; Mattila et al., 2007; Suetsugu et al., 2006).

Missing-in-metastasis (MIM/MTSS1), an actin and membrane binding protein, was first identified as a tumor suppressor, which is downregulated in bladder carcinoma cell lines with increased metastatic potential (Lee et al., 2002). Subsequent studies also demonstrated decreased expression of MIM in other cancer types (Ma et al., 2007; Wang, Y. et al., 2007a). In mice, MIM is expressed specifically in a few embryonic cell types, such as muscle and post-mitotic neurons, as well as in adult kidney, liver and Purkinje cells of the cerebellum (Machesky and Johnston, 2007; Mattila et al., 2003). MIM is composed of a C-terminal actin-monomer-binding WH2 (WASP homology 2) domain and an N-terminal I-BAR (inverse BAR) domain, also known as an IRSp53/MIM homology domain (IMD) (Mattila et al., 2003; Woodings et al., 2003; Yamagishi et al., 2004). The I-BAR domain was originally proposed to bundle actin filaments and thus induce the formation of filopodia-like plasma membrane protrusions when expressed in cells (Yamagishi et al., 2004). However, subsequent studies revealed that the I-BAR domain does not display significant actin filament bundling activity under physiological conditions, but rather binds to membranes rich in phosphatidylinositol (4,5)-bisphosphate [PtdIns(4,5)P₂], which it then deforms into tubular structures in vitro (Lee et al., 2007; Mattila et al., 2007; Saarikangas et al., 2009; Suetsugu et al., 2006). Interestingly, MIM was recently also implicated in the Sonic hedgehog (Shh) signaling pathway. Shh is a potent morphogen that controls many developmental processes, including left–right asymmetry and organ patterning. Ectopic or dysfunctional Shh signaling has been linked to many cancers (Varjosalo and Taipale, 2008). MIM was characterized as a Shh-responsive gene in both normal tissues and basal cell carcinomas, and was proposed to enhance Gli-mediated transcription through direct interactions with Gli1 and Gli2 transcription factors and by antagonizing Src-dependent phosphorylation of cortactin to regulate ciliogenesis (Callahan et al., 2004; Gonzalez-Quevedo et al., 2005; Bershteyn et al., 2010). This provided a plausible explanation for the correlation between MIM expression levels and metastatic potential of certain cancer types.

Despite the wealth of biochemical data, the physiological roles of MIM and its various activities (actin monomer binding, membrane deformation, Gli-mediated transcription) have not been reported. By inactivating the mouse Mim/Mtss1 gene, we demonstrate that full-length MIM protein is dispensable for embryonic development and Shh signaling. In adult mice, MIM deficiency leads to a progressive kidney disease characterized by abnormal tubular morphology, severe urine concentration defects, renal electrolyte wasting and bone abnormalities. These data, together with cell biological analysis of MIM localization and dynamics in MDCK cells, propose that MIM interacts with phosphoinositide-rich membranes and monomeric actin to modulate Arp2/3-mediated actin polymerization, and thus contributes to the integrity of cell–cell contacts in epithelia.

To analyze the physiological functions of MIM, we inactivated the mouse Mim gene by inserting a Neomycin (Neo) gene cassette, which contains several translation stop codons, into exon 1 (Fig. 1A). The successful gene targeting was verified by Southern blotting on genomic DNA from the targeted mouse embryonic stem (ES) cells (Fig. 1B). In-frame stop codons present in the targeting cassette are expected to destabilize the MIM transcript via the nonsense-mediated decay (NMD) pathway. Consistently, tissues from mice homozygous for MIM mutation (Mim^−/−) expressed only very low levels of Mim mRNA fragments as verified by qRT-PCR and RNA in situ hybridization (Fig. 1C,D). Furthermore, the lack of full-length MIM protein was confirmed by western blot analysis using an anti-MIM antibody (Fig. 1E,F).

Although our Mim^−/− mice do not express full-length MIM, it is important to note that another EST transcript is also associated with the Mim genomic locus. This shorter transcript contains an alternative 5′ exon, which is probably associated with an alternative promoter that is located in the intron 3 of the Mim gene. Thus, it can bypass our targeting construct and could possibly be translated to a protein lacking the first 81 residues of MIM, resulting in an I-BAR domain that lacks the membrane insertion motif, as well as a large part of the dimerization interfaces (supplementary material Fig. S1A–C). RT-PCR analysis revealed that this transcript is present in a subset of organs expressing MIM and is also found in the targeted MIM mutant mice (supplementary material Fig. S1D and data not shown). Importantly, our cell biological analysis revealed that the N-terminal I-BAR domain, which is crucial for the cellular role of MIM, is non-functional in the putative truncated protein and cannot deform membranes. Furthermore, the protein corresponding to the shorter transcript does not localize correctly to the plasma membrane when expressed in cells (supplementary material Fig. S1E,F). These data show that, if this transcript is indeed translated into a protein, it cannot promote membrane dynamics in cells. Thus, the MIM mutant mice generated and analyzed here do not express functional MIM protein that is capable of binding/deforming PtdIns(4,5)P₂-rich membranes.

The heterozygous MIM (Mim^+/−) mice were phenotypically normal and fertile. Their inter-crosses resulted in Mim^−/− mice at expected mendelian ratios. The MIM mutant mice survived into adulthood and had normal appearance and bodyweight (Fig. 1G and Fig. 2A). A detailed morphological inspection and histological analysis of all major organs was carried out on 2- to 3-month-old animals and did not reveal any gross abnormalities in Mim^−/− mice.

The mammalian IRSp53/MIM protein family, characterized by the N-terminal I-BAR domain, consists of five members and is divided in two branches (Scita et al., 2008). The first consists of MIM and ABBA, which show high sequence similarity within I-BAR and WH2 domains, but differ in the middle region, conceivably indicating divergence in the function. The tissue distributions of ABBA and MIM in embryonic and adult mice are mainly distinct (Saarikangas et al., 2008). Accordingly, the tissue distribution of ABBA was not changed in MIM mutant mice, suggesting that ABBA cannot compensate for the lack of MIM (supplementary material Fig. S2). The expression patterns of the three more distantly related I-BAR proteins, IRSp53, IRTKS and FLJ22582, are less well characterized, and it is thus possible that some of them might have partially redundant functions with MIM.

The Shh signaling pathway is indispensable for embryonic development, but the MIM-deficient mice showed no abnormalities in embryonic development or in the morphology of teeth, skull and hairs, which are ectodermal organs regulated by Shh during embryogenesis (Varjosalo and Taipale, 2008). Thus, we readdressed the suggested role of MIM in Shh signaling (Callahan et al., 2004; Bershteyn et al., 2010).

Cellular responses to Shh are transmitted through its receptor Patched (Ptc), which upon ligand binding, releases the repression of Smoothened, and activates signal transduction, ultimately leading to activation of Gli transcription factors (Varjosalo and Taipale, 2008). In Ptc-knockout (Ptch1^−/−) mice, Shh signaling is constitutively active causing embryonic lethality at E9–E10.5, probably because of abnormal development of the heart. Additionally, these embryos display a neural tube closure defect and are severely malformed (Goodrich et al., 1997). We crossed Mim^−/− mice with Ptch1^+/− mice to create double targeted mice. Loss of protein that activates Gli should reduce Shh-dependent transcription, which can be monitored by the expression of the marker gene β-galactosidase (β-gal) in the Ptch1^−/− mouse strain. However, similar high levels of β-gal expression were observed in both Ptch1^−/− and Mim^−/−Ptch1^−/− embryos indicating that loss of MIM did not decrease Gli activity (supplementary material Fig. S3A). Furthermore, if MIM was an activator of Gli1 or Gli2, loss of MIM should partially rescue the Ptch1^−/− phenotype by lowering the activity of Shh-dependent transcription in organs where both proteins are functioning. Because MIM is also highly expressed in the developing central nervous system and heart (Mattila et al., 2003), the phenotypes of Ptch1^−/− embryos should be at least partially rescued in double-knockout embryos (see Bai et al., 2002). However, comparative analysis between Ptch1^−/− and Mim^−/−Ptch1^−/− embryos revealed that the lack of MIM did not rescue the phenotype of Ptch1^−/− embryos to any detectable level, with no improvement in the morphology or survival of the embryos (supplementary material Fig. S3A). These data suggest that full-length MIM is not crucial for Shh signaling, at least during embryogenesis.

To exclude the possibility that putative shorter versions of MIM protein, which might be expressed in our MIM-deficient mice, could contribute to Shh signaling, we reassessed the role of MIM in Gli-mediated transcription by carrying out a Shh reporter assay in a NIH3T3 cell culture model. Here, we used corresponding constructs to the ones applied in the previous study concerning the role of MIM in Gli-dependent transcription (Callahan et al., 2004). In this Hedgehog-pathway-specific signaling assay, a firefly luciferase gene is under the control of a Gli-specific promoter (Taipale et al., 2000). This assay did not reveal any effect of MIM or ABBA in transcription by either endogenous Gli transcription factors induced by Shh or when MIM was co-expressed with Gli1 or Gli2 (supplementary material Fig. S3B and data not shown). Additionally, in contrast to the previous study (Callahan et al., 2004), Gli1 and Gli2 failed to interact with MIM–GFP in co-immunoprecipitation experiments (supplementary material Fig. S3C). Taken together, we found no evidence that full-length MIM has a crucial role in Shh-dependent transcription. However, based on this experimental system, we cannot definitely rule out a possible minor role of MIM in Shh-mediated transcription or its function in ciliogenesis in other cell-types, as suggested recently (Bershteyn et al., 2010).

In adult mice, MIM is strongly expressed in the kidney cortex (Mattila et al., 2003) (supplementary material Fig. S4C). To analyze the renal functions of MIM mice, we applied metabolic cages to collect 24-hour urine samples from 8-month-old wild-type and Mim^−/− littermates (n=10 per genotype). Urine analysis revealed alterations in several parameters. First, Mim-null mice displayed significantly higher water intake (P<0.01, Student's t-test) and urine production (P<0.002) compared with that in wild-type littermates (Fig. 2B,C). In addition, urinary wasting of electrolyte Mg²⁺ (P<0.02) over 24 hours was increased (Fig. 2D). Similarly, urinary wasting of Ca²⁺, K⁺ and Na⁺ was increased, although these changes were not statistically significant (data not shown). These results suggest that Mim^−/− mice suffer from reduced tubular water and electrolyte absorption in the kidney.

Alterations in bone metabolism are common in patients with chronic kidney disease (Magnusson et al., 2001) and were previously reported for mouse models with polyuria and renal electrolyte wasting (Hoenderop et al., 2003; Hou et al., 2007). A clinical chemistry screen revealed significantly increased plasma alkaline phosphatase (ALP) levels in Mim-null mice (P<0.05, one-way ANOVA), suggesting that they have a higher bone turnover rate compared with wild-type mice (Fig. 3A). Plasma ALP is considered a good marker for measuring bone metabolism in patients with renal disease because it is not influenced by variations in renal function (Magnusson et al., 2001).

Bone densitometry using dual-energy X-ray absorptiometry (DXA) revealed that both bone mineral content (BMC) (P<0.01) and bone content (P<0.005) were significantly decreased in MIM mutants (n=20 per genotype) when compared with wild-type littermates (Fig. 3B,C), and this phenotype was more pronounced in females. The observed changes cannot be explained by differences in body weight or bone length. We propose that the differences observed in the bone represent a secondary effect of the kidney defect (e.g. altered Ca²⁺ homeostasis).

Although no morphological or histological abnormalities were detected in the kidneys of 2- to 5-month-old MIM mice (Fig. 4A,B), many Mim^−/− mice became sick and died at around 8–18 months. Detailed histological analysis revealed dramatic changes in the kidney histology (Fig. 4C–F). Based on the observed pathological changes, the progression of the kidney disease in Mim^−/− mice was categorized into three stages. At stage I of the disease, dilated tubular structures were observed, and this was followed by infiltration of lymphocytes together with glomerulosclerosis (stage II). At the end stage of the disease, massive fibrosis together with glomerular degeneration was observed (stage III). The onset of the first histological signs of the renal disease varied widely between individuals. Some histologically abnormal kidneys were also detected in aged wild-type littermates, but with much milder anomalies compared with those in Mim^−/− mice.

To assess the detailed primary changes in Mim^−/− mice, we analyzed the kidney ultrastructures of three mice per genotype using transmission electron microscopy (TEM). The selected Mim^−/− mice displayed polyuria, but appeared otherwise normal (with normal body weight) and therefore were concluded to represent an early stage of the kidney disease. Interestingly, TEM analysis revealed apparent atrophic changes in the ultrastructure of the renal proximal tubules of the mutant mice, whereas such abnormalities were not observed in the control wild-type littermates. Dilated lateral intercellular spaces at the cell–cell contacts were observed in Mim^−/− mice (Fig. 4G–J). Furthermore, thickening of the basement membrane was detected in Mim-deficient mice but not in control mice (supplementary material Fig. S5). The lateral intercellular space represents a pathway for fluid and ion transport, and the observed changes in the ultrastructure of Mim^−/− kidneys are thus consistent with polyuria and renal electrolyte wasting in these mice. It is also important to note that EM analysis did not reveal alterations in the density or morphology of brush border in kidney proximal tubules (data not shown).

We also conducted immunohistochemical analysis using a tight junction marker (ZO-1), F-actin, Arp2/3 complex and glomerular markers (nephrin, synaptopodin, podocin and CD2AP) at the early stages of the disease. No apparent changes in the distribution or signal intensity between wild-type and MIM mutant mice were detected (supplementary material Figs S6 and S7). In addition, an RNA expression profiling analysis did not reveal differentially expressed genes in the kidneys of young Mim^−/− and wild-type mice (additional materials and methods; please contact the author for further information).

Similar ultrastructural changes in the kidney epithelia leading to polyuria and electrolyte wasting as described here for Mim^−/− mice were previously detected in mice lacking tight junction component claudin-16, as well as in humans suffering from cystic kidney disease nephronophthisis and kaliopenic nephropathy (Hou et al., 2007; Krishnan et al., 2008; Simon et al., 1999; Wang, W. et al., 2007b). This suggests that defects in cell–cell contacts compromising epithelial integrity can reduce osmotic gradient for effective water reabsorption or might result in backflow of water through the junctional space, and thus lead to polyuria. It is important to note that at early stages of kidney disease, no specific abnormalities were detected in the glomeruli of Mim^−/− mice, suggesting that the observed phenotype originates from kidney tubules. Moreover, no systematic proteinuria was detected in Mim-null mice, indicating that glomerular failure is probably not the primary cause of the symptoms.

Together, these data imply that MIM is not essential for the formation of cell–cell contacts in kidney epithelium, but instead contributes to their maintenance. It is important to note that the I-BAR protein IRSp53 is also expressed in kidney (supplementary material Fig. S4B,D). Interestingly, a recent study suggested that IRSp53 contributes to the formation or maintenance of tight junctions in cultured Madin-Darby canine kidney (MDCK) cells (Massari et al., 2009). Thus, I-BAR domain proteins appear to have an important, but probably at least partially redundant, role in the maintenance of specialized cell adhesion structures in the kidney epithelium.

Previous studies demonstrated that overexpression of MIM induces the formation of filopodia, microspikes and membrane ruffles in cultured mammalian cells (Bompard et al., 2005; Lin et al., 2005; Mattila et al., 2003; Woodings et al., 2003; Yamagishi et al., 2004). However, the subcellular localization, dynamics and role of MIM in confluent epithelial cells have not been reported. We used MDCK cell line to examine the localization of MIM in kidney epithelial cells. In the absence of a good antibody for immunofluorescence microscopy, MIM localization in polarized monolayer of MDCK cells was examined by using a MIM–GFP construct. Interestingly, in MDCK cells MIM localized to the plasma membrane and was strongly enriched at cell–cell adhesions, where it colocalized with F-actin and the adherens junction marker E-cadherin. Small amounts of MIM were also detected at tight junctions (Fig. 5A–C). Importantly, inactivation of the membrane binding and deforming I-BAR domain by previously characterized point mutations abolished the localization of MIM to cell–cell adhesions and plasma membrane, whereas inactivation of the actin-binding WH2 domain had no detectable effect on the subcellular localization of MIM (Fig. 5D,E). Fluorescence recovery after photobleaching (FRAP) analysis revealed that GFP–MIM localization to the cell–cell contacts was highly dynamic, with a t_1/2 of ~7 seconds (Fig. 6A,B). Together, these results demonstrate that in cultured kidney epithelial cells, MIM displays dynamic localization to cell–cell adhesions through its membrane binding and deforming I-BAR domain.

To elucidate the mechanism by which MIM is recruited to cell–cell contacts, we engineered a construct expressing a hybrid MIM protein. In this construct, the phosphoinositide binding of the I-BAR was inactivated by point mutations (see above), and the protein was fused to a PtdIns(4,5)P₂-specific PH domain from the N-terminus of PLCδ (Varnai and Balla, 1998). Thus, in the hybrid protein, the PtdIns(4,5)P₂-binding activity of the I-BAR domain was replaced by a PtdIns(4,5)P₂-binding PH domain without affecting the overall structure or dimerization properties of MIM or its I-BAR domain. Interestingly, whereas the wild-type GFP–MIM preferentially localized to intercellular junctions when expressed in MDCK cells, the chimeric MIM-PH protein displayed a more uniform plasma membrane localization to both lateral and apical regions of the cells (Fig. 6C,D).

To understand the possible role of MIM in actin and plasma membrane dynamics in MDCK cells, we compared cells expressing GFP–MIM with neighboring non-transfected cells. Interestingly, F-actin staining was brighter at the intercellular contacts of MIM-transfected cells compared with non-transfected cells. Quantification of the F-actin intensity at the intercellular contacts (additional materials and methods; please contact the author for further information) revealed that cells expressing moderate levels of wild-type MIM displayed an ~18% increase in the amount of F-actin at cell–cell contacts compared with non-transfected cells, indicating that MIM induces actin polymerization or recruits actin filaments to cell–cell contacts. Importantly, this activity was significantly reduced when the actin-monomer-binding WH2 domain of MIM was deleted (P<0.05) or when the lipid-binding site was inactivated (P<0.05) (one-way ANOVA) (Fig. 7A–C).

Previously, MIM was shown to promote Arp2/3-mediated actin polymerization through interaction with the actin-binding protein cortactin (Lin et al., 2005). More recently, a Drosophila I-BAR domain protein that is related to mouse MIM and ABBA was shown to interact with this actin-binding protein to promote directional cell migration (Lin et al., 2005; Quinones et al., 2010). We thus generated a MIM construct where the previously identified cortactin-binding site was deleted (Lin et al., 2005). Surprisingly, this mutant localized normally to the adherens junctions and was as efficient in promoting actin assembly at cell–cell contacts as wild-type MIM, indicating that cortactin-binding by MIM is not essential for its ability to promote actin filament assembly in cultured MDCK cells. MIM also undergoes Src-dependent phosphorylation at Tyr397 and Tyr398. Mutation of these residues was shown to lead to a reduced dorsal ruffling by MIM, indicating the importance of MIM phosphorylation in actin remodeling (Wang, Y. et al., 2007b). However, Tyr397Phe/Tyr398Phe mutation did not display significant effects on localization of MIM to cell–cell contacts or its ability to promote actin assembly at these sites in MDCK cells (Fig. 7B,C).

To examine whether the MIM-promoted actin filament assembly is dependent on the Arp2/3 complex, we inactivated the Arp2/3 complex by expressing a Scar WA construct in MDCK cells (Machesky and Insall, 1998). This inhibited the polarization of the cells and thus studies concerning the possible interplay between MIM and Arp2/3 were carried out in non-polarized MDCK cells. Importantly, inactivation of Arp2/3 diminished MIM-induced actin assembly (supplementary material Fig. S8). Collectively, these data propose that MIM mediates Arp2/3-dependent actin filament assembly at cell–cell contacts. This activity is dependent on the intact membrane-binding I-BAR domain and actin-monomer-binding WH2 domain, but does not require MIM phosphorylation at Tyr397 or Tyr398, or its interaction with cortactin.

To further examine the role of MIM in formation and maintenance of intercellular junctions, we disrupted cadherin-based adhesion of MDCK cells by depleting Ca²⁺ with 5 mM EGTA (supplementary material Fig. S9). After 15 and 30 minutes of Ca²⁺ removal, the adhesions in non-transfected cells started to dismantle, whereas in most cases, E-cadherin and phalloidin staining at the boundary between two adjacent cells expressing MIM–GFP was still prominent (supplementary material Fig. S9B,C). This suggests that MIM contributes to the stability of cadherin adhesions upon calcium removal. However, after 1 hour of Ca²⁺ depletion, the majority of both non-transfected cells and cells overexpressing MIM displayed a rounded morphology; the MIM–GFP signal was mostly cytoplasmic (supplementary material Fig. S9D). To analyze the localization of MIM–GFP during re-establishment of intercellular adhesions, we replaced the EGTA-containing medium with Ca²⁺-containing medium and monitored the reassembly of cell–cell adhesions. Interestingly, MIM did not accumulate at newly forming adhesions and was found only in more mature adhesions along with E-cadherin (supplementary material Fig. S9F,G). Together, these data suggest that MIM is not involved in the formation of intercellular junctions, but instead contributes to their maintenance.

MIM is a multifunctional protein that binds to actin and membranes and has been implicated in cancer metastasis (Machesky and Johnston, 2007). In addition, MIM was recently proposed to be a central component of the Shh signaling pathway (Callahan et al., 2004; Bershteyn et al., 2010). By generating and analyzing MIM-deficient mice, and by examining the localization and function of MIM in cultured cells, we show that MIM is necessary for the maintenance of cell adhesions in the kidney epithelia and provide evidence that I-BAR-domain-mediated membrane binding and WH2-domain-mediated actin-binding activities of MIM have crucial roles in this process. Surprisingly, analysis of MIM mutant mice and crossing them with Ptc-deficient mice did not provide evidence for the role of full-length MIM in Shh signaling in vivo. This conclusion is further supported by studies on an NIH3T3 cell culture model, which failed to detect any role of MIM in Shh-induced transcription in cells. Furthermore, MIM did not interact with Gli1 or Gli2 in a co-immunoprecipitation assay. Although these data demonstrate that full-length MIM is dispensable for Shh signaling, in the future it will be important to examine whether the N-terminally truncated version of MIM (AK030533) is expressed at the protein level and could have a role in Shh signaling. In this context, it is important to note that a recent study suggested that MIM additionally contributes to Shh signaling by promoting ciliogenesis, and that this activity would depend on the C-terminal region of MIM (Bershteyn et al., 2010).

Disruption of Mim resulted in a progressive kidney disease in adulthood and MIM-deficient mice displayed bone abnormalities. Interestingly, a recent study identified MIM as a transcriptional target of myocyte enhancer factor 2C (MEF2C) in polycystin-dependent sensing of fluid stress. Polycystin proteins are thought to act as mechanosensors in the kidney tubules, where they respond to changes in tubular fluid flow to induce downstream signaling and expression of target genes. Consistent with our findings, this study also showed that a lack of MIM leads to late-onset histological abnormalities in the kidneys (Xia et al., 2010). Thus, it is possible that upregulation of MIM in response to fluid flow sensing by polycystins promotes the strengthening of intercellular adhesions. The bone abnormalities of MIM-deficient mice identified in our study probably represent secondary effects resulting from the kidney phenotype, because MIM expression was undetectable in developing and adult bones, and similar bone abnormalities were previously reported for other mutant animals, and human diseases resulting in kidney defects (Hoenderop et al., 2003; Hou et al., 2007; Magnusson et al., 2001). In cultured MDCK cells, MIM concentrated to adherens junctions at the cell–cell contacts, where it induced an accumulation of F-actin. FRAP analysis revealed that association of MIM to intercellular contacts is highly dynamic and similar to that reported previously for actin and many actin-binding proteins, whereas E-cadherin, which colocalizes with MIM in MDCK cells, was reported to display slower dynamics (Yamada et al., 2005).

Two distinct F-actin populations are reported to exist at epithelial cell–cell contacts (Zhang et al., 2005; Yamazaki et al., 2006). The more stable population of thin actin bundles was proposed to be generated through re-organization of pre-existing actin filaments, whereas the more dynamic pool of junctional actin, is formed by de novo actin polymerization. Furthermore, two distinct actin filament nucleation pathways have been identified at epithelial cell–cell contacts: one mediated by Arp2/3 (Kovacs et al., 2002; Yamazaki et al., 2007) and the other by the formin mDia1 (Carramusa et al., 2007). Our data suggest that MIM induces F-actin assembly at the cell–cell contacts of MDCK cells in an Arp2/3-dependent manner, and that MIM displays rapid association–dissociation dynamics at these sites. Together, these results suggest that MIM is associated with the dynamic junctional actin. Owing to the relatively specific effects of MIM deficiency on the epithelial structures in mice, we suggest that depletion of MIM causes a local defect in the organization and/or dynamics of junctional actin rather than inducing more systematic defects in the actin cytoskeleton of epithelial cells. MIM might, for example, recruit or cluster F-actin assembly factors at correct foci of intercellular junctions by sensing or generating specific membrane curvature and/or by clustering phosphoinositides (Saarikangas et al., 2009). For example, PtdIns(4,5)P₂ clustering has a positive net influence on actin filament assembly (Saarikangas et al., 2010) and regulation of membrane curvature might be important for fine-tuning actin dynamics (Lee et al., 2010; Liu et al., 2008; Takano et al., 2008). Thus, MIM-mediated actin assembly might be important for the stabilization of cadherin adhesions or in generation of membrane protrusions for the re-establishment of cadherin-based adherens junctions (Fig. 7).

Importantly, mutant MIM, which was deficient in membrane binding as a result of inactivation of the I-BAR domain, did not localize to cell–cell contacts or induce an accumulation of F-actin to intercellular junctions. Furthermore, a construct in which phosphoinositide binding by the I-BAR domain was replaced with a PtdIns(4,5)P₂-binding PH domain failed to accumulate at the cell–cell junctions. These data suggest that the I-BAR domain of MIM does not localize the protein to cell–cell contacts by simply interacting with PtdIns(4,5)P₂, but might additionally sense membrane curvature to direct the protein to the correct site of action. Additionally, a MIM construct lacking the actin-monomer-binding WH2 domain displayed a significant defect in promoting F-actin assembly compared with the wild-type MIM, although this mutant still localized to cell–cell contacts. Thus, the WH2 domain of MIM is important for its role in assembly of actin filaments at intercellular contacts, possibly by increasing the local concentration of ATP–G-actin at these sites. Together, these data suggest that a dynamic interplay between the actin cytoskeleton and plasma membrane is important for the maintenance of cell–cell junctions in epithelia.

Previous studies demonstrated that the formation of intercellular adhesions in certain epithelial cells is preceded by a zippering stage driven by filopodia-like protrusions, which make contacts with the neighboring cells to facilitate epithelial sealing (Vasioukhin et al., 2000). However, no developmental abnormalities were detected in the kidneys of Mim^−/− mice. Instead, the abnormal kidney phenotype progressively developed during adulthood. It is possible that this might result from a functional redundancy between MIM and other I-BAR proteins such as IRSp53, which is also expressed in the kidney (supplementary material Fig. S4). Thus, future studies using IRSp53 (Kim et al., 2009; Sawallisch et al., 2009) and MIM double-knockout mice will be necessary to analyze whether these two proteins indeed have a redundant role in the development of the kidney epithelial structures. Alternatively, it is possible that MIM-induced actin and plasma membrane dynamics have an important role only in the regeneration of adhesive structures of kidney epithelium, or that the precise interplay between the actin cytoskeleton and plasma membrane is only required for the long-term maintenance of epithelial structures in animals.

The role of MIM in epithelial integrity might also provide an explanation for studies reporting that decreased MIM expression correlates with the metastatic potential of certain epithelial cancers (Lee et al., 2002; Ma et al., 2007; Wang, Y. et al., 2007a; Liu et al., 2010). Decreased expression of MIM in epithelial cancers could weaken the cell–cell contacts and subsequently promote epithelial–mesenchymal transition (EMT) of the cancer cells. In the future, it will be important to determine the exact mechanism by which MIM promotes actin and plasma membrane dynamics to maintain cell–cell contacts in various epithelia, and to elucidate how this activity is linked to the metastatic behavior of cancers.

The generation of the MIM targeting construct was carried out by Mu transposition (Turakainen et al., 2009). All animals were handled in strict accordance with good animal practice as defined by the relevant national and/or local animal welfare bodies, and all animal work was approved by the appropriate committee.

A primary systemic phenotype screen was performed at the German Mouse Clinic (GMC, http://www.mouseclinic.de) (Gailus-Durner et al., 2005).

The MIM–GFP expression plasmid and the WH2 deletion constructs were previously described (Mattila et al., 2003). The constructs for inactivation of the I-BAR domain (K146A,149A,150A) (Mattila et al., 2007), cortactin-binding region (by deleting residues 615–730) (Lin et al., 2005) and Src phosphorylation sites (Y397,398F) (Wang, Y. et al., 2007b) of MIM, and the fusion of MIM to PH domain of PLCδ domain (Varnai and Balla, 1998) were generated by PCR. The Scar W and WA constructs (Machesky and Insall, 1998) were a gift from Laura Machesky (Beatson Institute for Cancer Research, Glasgow, UK). MDCK, U2OS and NIH3T3 cells were cultured in D-MEM supplemented with 10% fetal calf serum. Shh reporter assays were performed as described (Taipale et al., 2000). Transfection of U2OS cells and quantification of filopodia was performed as described (Saarikangas et al., 2009). MDCK cells were polarized by culture for 3 days on Transwell 0.4 μm polycarbonate membrane permeable supports (Corning). Cells were transfected (day 1) using Lipofectamine 2000 (Invitrogen) following the manufacturer's guidelines and were fixed (day 3) in 4% paraformaldehyde. Cells were permeabilized with 0.3% Triton X-100 in PBS for 30 minutes, incubated with primary antibodies against E-cadherin (Transduction Laboratories) and ZO-1 (Zymed Laboratories) overnight at 4°C and with secondary antibodies for 2 hours at room temperature.

Further details of materials and methods, and any associated references can be obtained from the corresponding author.

We are grateful to R. Savolainen for the expertise in ES cell work, and E. Koivunen and A.-L. Nyfors for technical help. The Helsinki University Transgenic Animals Unit and the Viikki Laboratory Animal Centre are acknowledged for the expertise in generation and maintenance of the MIM-deficient mice. I. Thesleff, J. Pispa, J. Jernvall, and P. Munne are acknowledged for expertise in ectodermal organ analysis. J. Peränen is acknowledged for providing various antibodies and K. Tanhuanpää for help with image analysis. We thank Laura Machesky and Tamas Balla for sharing reagents. We thank the members of the German Mouse Clinic for comprehensive phenotyping of the MIM-deficient mice and fruitful discussions. We are grateful to Reinhard Seeliger, Elfi Holupirek, and Eleonore Samson and the GMC animal caretaker team for expert technical help. This study was supported by grants from the Academy of Finland (grants 121248 and 218021 to S.L., and 124137 to P.L.), Finnish National Technology Agency (TEKES) and the Finnish Cancer and Finnish Cultural Foundations. J.S. was supported by fellowships from Helsinki Graduate School in Biotechnology and Molecular Biology (GSBM), Emil Aaltonen, and Oscar Öflund Foundations. P.M. by fellowships from Viikki Graduate School in Biosciences and Alfred Kordelin, Maud Kuistila, Ida Montini, Farmos, and Biomedicum Foundations. M.V. was supported by GSBM, Paulo, Finnish Cultural, Emil Aaltonen, Wihuri, Orion-Farmos and Biomedicum Helsinki Foundations. The German Mouse Clinic received funding from the EU (EUMODIC grant LSHG-2006-037188) and the Federal Ministry of Education and Research (NGFNplus grants 01GS0851 to B.R. and E.W. and 01GS0850 to J.C.-W., M.T., W.H., M.H., H.F., V.G.-D., I.E. and M.H.A.).