The establishment of epithelial polarity is tightly linked to the dramatic reorganization of microtubules (MTs) from a radial array to a vertical alignment of non-centrosomal MT bundles along the lateral membrane, and a meshwork under the apical and basal membranes. However, little is known about the underlying molecular mechanism of this polarity-dependent MT remodeling. The evolutionarily conserved cell polarity-regulating kinase PAR-1 (known as MARK in mammals), whose activity is essential for maintaining the dynamic state of MTs, has indispensable roles in promoting this process. Here, we identify a novel PAR-1-binding protein, which we call microtubule crosslinking factor 1 (MTCL1), that crosslinks MTs through its N-terminal MT-binding region and subsequent coiled-coil motifs. MTCL1 colocalized with the apicobasal MT bundles in epithelial cells, and its knockdown impaired the development of these MT bundles and the epithelial-cell-specific columnar shape. Rescue experiments revealed that the N-terminal MT-binding region was indispensable for restoring these defects of the knockdown cells. MT regrowth assays indicated that MTCL1 was not required for the initial radial growth of MTs from the apical centrosome but was essential for the accumulation of non-centrosomal MTs to the sublateral regions. Interestingly, MTCL1 recruited a subpopulation of PAR-1b (known as MARK2 in mammals) to the apicobasal MT bundles, and its interaction with PAR-1b was required for MTCL1-dependent development of the apicobasal MT bundles. These results suggest that MTCL1 mediates the epithelial-cell-specific reorganization of non-centrosomal MTs through its MT-crosslinking activity, and cooperates with PAR-1b to maintain the correct temporal balance between dynamic and stable MTs within the apicobasal MT bundles.
Microtubules (MTs) are essential for many cellular functions, including vesicle transport and cell motility, polarity and division. These MT-dependent processes are crucially dependent on the spatial organization and dynamic nature of MTs, which are regulated by various MT-associated proteins (MAPs). In many animal cells, most interphase MTs are anchored by their minus ends to the centrosome located next to the nucleus and arranged in radial arrays (Vorobjev and Nadezhdina, 1987). By contrast, the majority of MTs in many differentiated cells, such as polarized epithelial cells and neurons, show cell-type-specific arrays that are not anchored to the centrosome (Bartolini and Gundersen, 2006; Keating and Borisy, 1999). These MTs are generally bundled and stabilized in order to sustain long-lived arrays during differentiation and morphogenesis (Baas and Lin, 2011; Bré et al., 1987). The correct assembly, positioning and maintenance of these non-centrosomal MTs are vital for many of the specialized functions of these cells, but the mechanisms underlying these MT organizations remain to be clarified.
Epithelial cell MTs are dramatically reorganized upon polarization, from a radial array to a vertical alignment of non-centrosomal MT bundles (apicobasal MT bundles) running along the lateral membrane with their minus ends towards the apical surface, and a meshwork under the apical and basal membranes (Bacallao et al., 1989; Mogensen et al., 1989; Müsch, 2004). Although the origin of the non-centrosomal MTs remains to be clarified (Oriolo et al., 2007), several studies have suggested that they are nucleated from the apical centrosome, released and transported. They are then captured by minus-end-binding proteins such as ninein at their final anchoring sites in the apical membrane or adherens junctions (Bellett et al., 2009; Keating et al., 1997; Mogensen et al., 2000). The interactions between APC, LL5, KIF17 and MT-plus-end-tracking proteins (+TIPs) were further demonstrated to be involved in membrane anchoring of the plus ends of these MTs and their stabilization (Hotta et al., 2010; Jaulin and Kreitzer, 2010; Mogensen et al., 2002). A recent study revealed that septin filaments navigate the growth and positioning of MTs along the apicobasal axis (Bowen et al., 2011). However, the mechanisms that promote MT assembly into higher-order structures (bundles) have not been clarified.
The serine/threonine kinase PAR-1 regulates cell polarity in various biological contexts as a component of an evolutionarily conserved protein machinery, the PAR–aPKC (atypical protein kinase C) system (Suzuki and Ohno, 2006). The mammalian PAR-1 homolog was originally identified as MARK (MAP and MT affinity-regulating kinase), the overexpression of which destabilizes MTs by phosphorylating classical MAPs and inducing their dissociation from MTs (Drewes et al., 1997). We recently demonstrated that small interfering RNA (siRNA)-mediated knockdown of PAR-1 suppressed continuous growth of MTs in vivo, indicating that PAR-1 is the kinase required for maintaining the dynamic state of MTs (Hayashi et al., 2011). Importantly, Cohen et al. (Cohen et al., 2004) demonstrated that one of the mammalian PAR-1 proteins, PAR-1b (also known as MARK2), is indispensable for establishment of the epithelia-specific MT assembly at the lateral membrane of MDCK cells. They speculated that PAR-1b mediates cell–cell adhesion-induced assembly or capture of MT minus ends at the lateral cortex (Müsch, 2004). However, the molecular basis for how PAR-1b mediates these processes remains to be clarified, especially in light of its activity in maintaining MT dynamics.
In this study, we identified a novel PAR-1-interacting protein, which we call microtubule crosslinking factor 1 (MTCL1), which colocalizes with the apicobasal MT bundles in polarized MDCK cells. MTCL1 is a novel MT-interacting protein that directly binds to MTs through its N-terminal and C-terminal portions independently. MTCL1 knockdown impaired the development of the apicobasal MT bundles, and the restoration of these defects required the N-terminal MT-binding region that crosslinks MTs with the aid of the subsequent coiled-coil motifs. Taken together with the results of MT regrowth assays, we conclude that MTCL1 mediates accumulation of MTs at the lateral cortex of polarizing epithelial cells. MTCL1 recruited a subpopulation of PAR-1b to the apicobasal MT bundles, and the interaction with PAR-1b was required for the MTCL1-dependent development of the apicobasal MT bundles. Taken together, we also suggest that MTCL1 functions as a scaffold protein for PAR-1b that incorporates PAR-1b activity to maintain the dynamics of the apicobasal MTs.
MTCL1 is a novel PAR-1-binding protein in MDCK cells
To identify molecules that cooperate with PAR-1 for epithelial polarity, we searched for new PAR-1-binding proteins in polarized MDCK epithelial cells. For this purpose, we established a stable MDCK cell line in which endogenous PAR-1b was replaced by streptavidin-binding peptide (SBP)-tagged PAR-1b expressed at an almost endogenous level. Among the proteins specifically copurified with SBP–PAR-1b, we reproducibly detected a large molecule with a molecular mass of ∼250 kDa (data not shown), which was subsequently revealed to be an uncharacterized protein, KIAA0802, by mass spectrometric analysis. On the basis of the results presented here, we named this protein microtubule crosslinking factor 1 (MTCL1) (see below). The longest cDNA encoding MTCL1 registered in public databases encodes a mouse protein of 1945 amino acids that is predicted to form long coiled-coil structures in the central region, but has no obvious homology with other proteins (Fig. 1A). The originally identified human cDNA encoding KIAA0802 (also annotated as SOGA2, see Discussion), which we first used in the following experiments, encodes a protein that lacks the N-terminal 352 amino acids and has an alternatively spliced C-terminal region (Fig. 1A). Given that the nucleotide sequence corresponding to the N-terminal region of full-length MTCL1 has an extremely high GC content (∼78%), and is thus highly refractory to typical PCR amplification, KIAA0802 might not represent a genuine MTCL1 isoform, but instead be an artificial transcript generated during high-throughput cloning processes. In fact, human cDNA databases have failed to register a cDNA for the full-length MTCL1 predicted from genome analyses.
An anti-MTCL1 polyclonal antibody (pAb) detected three bands in MDCK cells of around 250 kDa, which were commonly knocked down by three independent siRNAs (#1, #2 and #3) (Fig. 1B). The largest band exhibited a similar mobility in SDS-PAGE to V5-tagged full-length MTCL1 expressed in HEK293 cells. Furthermore, mass spectrometric analysis of the largest band identified several peptides within the N-terminal 352 amino acids of full-length MTCL1 (data not shown), indicating that MDCK cells do express the full-length MTCL1. Endogenous MTCL1 in MDCK cells was specifically coimmunoprecipitated with PAR-1b by an anti-PAR-1b antibody (Fig. 1C), whereas endogenous PAR-1b was coimmunoprecipitated by an anti-MTCL1 antibody (Fig. 1D). Thus, MTCL1 is a genuine endogenous partner of PAR-1b in MDCK cells.
Next, we narrowed down the PAR-1b-binding region by coimmunoprecipitation analyses in HEK293T cells using various deletion mutants of V5- or GFP-tagged MTCL1. As shown in Fig. 1E, full-length MTCL1 and KIAA0802 were each coimmunoprecipitated with HA-tagged PAR-1b. A subsequent analysis using mutants of KIAA0802 finally revealed that a central region (C1-2; amino acids 905–1022) immediately downstream of the coiled-coil domain was sufficient and essential for PAR-1 binding (Fig. 1E, bottom panels). In vitro binding assays using purified proteins further demonstrated that PAR-1 directly interacted with this region of MTCL1 (Fig. 1F).
MTCL1 colocalizes with the apicobasal MT bundles in polarized MDCK cells
To explore the physiological functions of MTCL1, we examined its subcellular localization in highly polarized filter-grown MDCK cells. To verify the antibody specificity, we heterogeneously knocked down MTCL1 in MDCK cells using pEB-super vectors simultaneously expressing EGFP and either of two kinds of shRNAs for MTCL1 (#1 and #2) (see Materials and Methods) (Masuda-Hirata et al., 2009). Immunofluorescence staining with the anti-MTCL1 antibody revealed submembranous signals that were specifically reduced in GFP-positive MTCL1 knockdown cells (Fig. 2A). Close inspection revealed that MTCL1 was localized inside the lateral membrane, which was identified by E-cadherin staining (data not shown), and showed colocalization with the apicobasal MT bundles running beneath the lateral membrane (Fig. 2B, bottom panels). SBP-tagged MTCL1 stably expressed at a low level exhibited a similar colocalization pattern with the apicobasal MT bundles (supplementary material Fig. S1). MTCL1 signals were also detected in the apical regions where MTs formed a dense network, and exhibited some colocalization with MTs, especially at the base of cilia (Fig. 2B, red arrows; see also Fig. 7A). However, in contrast to the MTs, which are condensed at the subapical region, MTCL1 signals did not show upregulation in the subapical region compared with the sublateral regions (see z-section view in Fig. 2B, bottom panels).
To assess the association of MTCL1 with MTs more clearly, we examined the subcellular localization of MTCL1 in sparsely-seeded MDCK cells in which individual MT filaments are clearly discernible (Fig. 2C). In these cells, the anti-MTCL1 antibody revealed granular staining that was concentrated on thick MT bundles. Careful examination of enlarged views further demonstrated that most of the granular signals of MTCL1 were localized on individual MT filaments (Fig. 2C, right panels), suggesting that MTCL1 stochastically associates with MT filaments along the lateral sides and is particularly condensed on MT bundles. Again, ectopically expressed MTCL1 exhibited a similar localization to MTs (supplementary material Fig. S1). Interestingly, in mitotic cells, endogenous as well as ectopically expressed MTCL1 showed specific accumulation in the spindle poles where MTs were tied in bundles, and in the midbodies that appeared at the end of cytokinesis (supplementary material Fig. S2). These results support the notion that MTCL1 is an MT-associated protein that tends to accumulate on MT bundles in various cellular contexts.
To examine the transition state of MTCL1 localization during the epithelial polarization process, we analyzed MDCK cells grown to confluence on coverslips and thereby displaying heterogeneous development of epithelial polarity. Under this condition, the cells in some regions were polarized and showed an accumulation of MTs at the sublateral region, while other cells were not fully polarized and displayed only the dense MT network (Fig. 2D). In these cells, the granular staining of MTCL1 decorated the developing MT bundles accumulated near the cell–cell boundary, but showed relatively less colocalization on the dense MT networks observed in unpolarized cells (asterisks in Fig. 2D). Taken together, these results suggest that MTCL1 gradually accumulates on the apicobasal MT bundles that develop during epithelial cell polarization.
MTCL1 is a novel MT-crosslinking protein
To investigate whether MTCL1 physically interacts with MTs, we next examined which regions of MTCL1 were responsible for the colocalization with MTs. For this purpose, we transiently expressed various deletion mutants of MTCL1 in HeLa-K cells (Fig. 3A). As shown in Fig. 3B, overexpressed full-length MTCL1 formed thick filamentous structures colocalizing with MT bundles. The same results were obtained for KIAA0802, indicating that the N-terminal 352 amino acids and the C-terminal alternatively spliced region are dispensable for MT binding. Subsequent analyses of KIAA0802 deletion mutants (Fig. 3B; supplementary material Fig. S3) indicated that the ∼100 amino acids in the C-terminal portion rich in conserved lysine and arginine residues (KR-rich region) was sufficient and essential for the colocalization with MTs, because a mutant protein lacking the KR-rich region (V5-KIAA0802 ΔKR) was distributed diffusely, while a short fragment corresponding to the KR-rich region (C6) formed bundles and colocalized with MTs (Fig. 3B). A direct interaction of the KR-rich region with MTs was confirmed by the result that purified GST–C6, but not GST alone, co-sedimented with taxol-stabilized MTs (Fig. 3C). Many MAPs, including +TIPs, induce MT bundle formation when overexpressed in cultured cells. Therefore, the above results do not necessarily indicate that the KR-rich region of MTCL1 has MT-bundling activity. However, the results demonstrate that MTCL1 is a novel MT-binding protein, and that one of the MT-binding regions is the KR-rich region.
To our surprise, in contrast to KIAA0802, full-length MTCL1 still showed colocalization with MTs even without the KR-rich region (Fig. 3D, ΔKR), suggesting that MTCL1 has another MT-binding region. Given that simultaneous deletion of the N-terminal 352 amino acids and the KR-rich region completely abolished the MT colocalization with MTCL1 (Fig. 3D, ΔN+ΔKR), the N-terminal region could exert the additional MT-binding activity. Consistent with this hypothesis, subsequent analyses revealed that the N-terminal fragment with 241 amino acids (N4) was sufficient for colocalization with MTs (Fig. 4B; supplementary material Fig. S4A) and co-sedimentation with taxol-stabilized MTs (Fig. 4C). Notably, overexpression of this N-terminal MT-binding fragment (N4) did not induce MT bundling, but it had intermittent colocalization with individual MT filaments (Fig. 4D). By contrast, a longer fragment, N1, which contained the subsequent two coiled-coil motifs in addition to the N-terminal MT-binding region, induced strong bundling of MTs (Fig. 4B). Taken together with the fact that the coiled-coil motifs of MTCL1 showed a significantly high probability of forming two-stranded structures (supplementary material Fig. S4B) (Wolf et al., 1997), these results suggest that the N1 fragment-induced MT bundling does not represent an artificial phenomenon, but indicates MTCL1 has MT-crosslinking activity through cooperation between the N-terminal MT-binding domain and the subsequent coiled-coil motifs. In accordance with this idea, the N1 fragment associated with itself, but not with other shorter mutants (N2 and N4) lacking the coiled-coil motifs (Fig. 4E). In addition, purified SBP-tagged N1, but not N4, induced significant bundling of MTs in vitro when mixed with taxol-stabilized Rhodamine-labeled MTs (Fig. 4F). We also confirmed that full-length MTCL1 showed self-association depending on the coiled-coil region (supplementary material Fig. S4C,D) and that endogenous MTCL1 in MDCK cell extracts was specifically co-sedimented with taxol-stabilized MTs (Fig. 4G). Taken together, these results demonstrate that MTCL1 is a novel MT-binding protein with two MT-binding regions, one in both its N- and C-termini, and that the former region functions to crosslink MTs with the aid of the coiled-coil region. This N-terminal MT-binding region was further confined to 190 amino acids that are also rich in basic residues (supplementary material Fig. S4A,E).
The N-terminal MT-binding region of MTCL1 is required for the development of the apicobasal MT bundles in MDCK cells
The MT-binding and MT-crosslinking activities of MTCL1 are consistent with its subcellular localization, and suggest a positive role for the development of the apicobasal MT bundles in polarized MDCK cells. In fact, when MTCL1 was heterogeneously knocked down in MDCK cells, MTCL1-depleted GFP-positive cells, but not control cells, had severely affected development of the apicobasal MT bundles along the lateral membrane. In control cells, MTs were accumulated at the sublateral region (Fig. 5A, arrowheads), and vertical staining of MTs was clearly observed along the lateral membrane in z-sectional views (arrowheads in Fig. 5B, top panels). By contrast, MTCL1 knockdown cells exhibited disorganized MT arrays (Fig. 5A, bottom panels) that did not show accumulation along the lateral membrane (see cells under the brackets in Fig. 5B). Even in cells showing a tendency to accumulate MTs at the sublateral region (asterisks in Fig. 5A, bottom panels), the development of the apicobasal MT bundles was only observed at the very basal side (Fig. 5B, arrowheads). Notably, these defects in MT organization were always associated with decreased cell height (59.6±3.5%) and increased cell area (137.9±6.0%), indicating impaired epithelial polarization (supplementary material Fig. S5). MTCL1 knockdown cells developed continuous tight junctions (supplementary material Fig. S5) with normal levels of transepithelial electrical resistance (data not shown). In addition, and consistent with a previous report that the apicobasal MT bundles regulate the efficiency, rather than the fidelity, of vesicle delivery to the polarized membrane domains in epithelial cells (Grindstaff et al., 1998), the cells showed normal distributions of apical (gp135) and lateral (NKA) polarity markers (supplementary material Fig. S5). Expression of siRNA-resistant MTCL1 rescued the defects of the knockdown cells (Fig. 5C), because MTCL1 knockdown cells (green) expressing V5-tagged MTCL1 (red) showed restoration of the strong accumulation of MTs along the lateral membrane (Fig. 5C,D) and regained their columnar shapes (Fig. 5E). These results demonstrate that MTCL1 is not required to establish the apicobasal axis and membrane polarity, but is essential for the late phase of epithelial polarization, including the development and/or maintenance of the apicobasal MT bundles and the epithelial-cell-specific columnar shape.
To examine whether the MT-binding and MT-crosslinking activities are involved in these MTCL1 functions for epithelial polarity, we assessed the rescue activities of the MTCL1 mutants lacking the MT-binding regions. In contrast to wild-type MTCL1, the ΔN mutant failed to restore the defects of MTCL1 knockdown MDCK cells (Fig. 5C–E). It should also be noted that, in contrast to wild-type MTCL1, the ΔN mutant exhibited a diffuse cytoplasmic localization in MTCL1 knockdown cells (Fig. 5C). This suggests that the sublateral accumulation of MTCL1 is also dependent on the MT-binding activity of the N-terminal region. In other words, MTCL1 and MTs synergistically promote the development of the apicobasal MT bundles in a mutually dependent manner. By contrast, and to our surprise, MTCL1 ΔKR exhibited accumulation to the lateral cortex (Fig. 6A) and exerted a slightly reduced, but substantial, rescue activity not only for MT organization (Fig. 6B) but also for the cell height and cell area (Fig. 6C). These results indicate that the MT-crosslinking activity of MTCL1 through the N-terminal MT-binding region is essential for organizing the apicobasal MT bundles and for the late phase of epithelial polarization. However, the C-terminal KR-rich region is almost dispensable for the development of the epithelial-cell-specific apicobasal MT bundles, despite the MT-binding activity observed in HeLa-K cells and in vitro.
MTCL1 is required for the accumulation of non-centrosomal MTs at the sublateral cortex
To further characterize the role of MTCL1 in the developmental process of the apicobasal MT bundles, we performed MT regrowth analyses in MDCK cells where the MTs had been completely depolymerized by a combination of cold treatment and treatment with 33 µM nocodazole (Fig. 7). As previously demonstrated (Bellett et al., 2009; Meads and Schroer, 1995), nocodazole removal first induced prominent regrowth of radial MTs from the apical centrosome (Fig. 7A, left panels, 20 minutes) that were positive for γ-tubulin (data not shown). Subsequently, these MT arrays gradually dissolved into less-focused non-centrosomal MT assemblies of the apical network (left panels) and the apicobasal bundles (right panels, Fig. 7A, 40 and 60 minutes). Note that, at the onset of the experiment, these assemblies of non-centrosomal MTs were predominantly positive for tyrosinated tubulin (a marker of dynamic MTs), but negative for acetylated tubulin (a marker of stable MTs). At the later stage, the content of acetylated tubulin was acutely increased, with a concomitant reduction of the tyrosinated tubulin signals. These results indicate that the non-centrosomal MT assembly is rather dynamic at the initial stage, and gradually becomes stabilized during the reconstruction of the epithelial-cell-specific MT organization (Bré et al., 1987).
In contrast to depolymerized tubulins, MTCL1 in nocodazole-treated cells demonstrated resistance to the detergent extraction required for specific detection of polymerized MTs and showed weak sublateral localization (Fig. 7A, right panels, 0 minutes). These results provided a sharp contrast to the above data that MTCL1 developed its sublateral localization in an MT-dependent manner during the epithelial polarization process, and suggest the possibility that MTCL1 became independent of MTs once the cells had established their epithelial polarity. Reconstructed z-sectional views confirmed that, at the initial stage of MT reconstruction, MTCL1 but not MTs showed a sublateral localization (Fig. 7B, 30 minutes, arrowheads). Subsequently, MTs appeared to elongate in this region, along the MTCL1 signals (Fig. 7B, 80 minutes, arrowheads). Taken together with the MT-crosslinking activity of MTCL1, these data suggest that MTCL1 mediates the recruitment of non-centrosomal MTs at the sublateral cortex.
To examine the above possibility, we next analyzed the effects of MTCL1 knockdown on the reconstruction process of MT organization after nocodazole removal. In these experiments, knockdown cells were labeled with nuclear-localizing EGFP that resists the detergent extraction (supplementary material Fig. S6). Again, MTCL1 knockdown (GFP-positive) cells showed reduced cell heights compared with the surrounding GFP-negative cells (Fig. 7C,D). Upon nocodazole removal, the GFP-positive cells showed regrowth of radial MTs from the apical centrosomes at the normal rate and positions compared with the surrounding GFP-negative cells (Fig. 7C) or cells expressing a control shRNA (data not shown). Furthermore, the development of the apical meshwork of non-centrosomal MTs was also normal (Fig. 7D, focus A and B). Nevertheless, the subsequent MT accumulation at the sublateral cortex was blocked in MTCL1 knockdown cells (Fig. 7D, focus C). These results support the notion that MTCL1 contributes to the lateral accumulation of MTs that are released from the apical centrosomes.
The interaction with PAR-1b is required for MTCL1 to fully exert its function to mediate the lateral MT bundle development
It has been demonstrated that PAR-1b knockdown in MDCK cells disrupts the epithelia-specific MT assembly at the lateral membrane and decreases the cell height (Cohen et al., 2004; Suzuki et al., 2004), which correspond well to the defects of MTCL1 knockdown cells observed in the present study. Considering the specific interaction between PAR-1b and MTCL1, these results suggest a functional link between MTCL1 and PAR-1b in epithelial polarization. In agreement with this notion, the ΔBR mutant of MTCL1, which lacks the minimum essential region for PAR-1 binding and thus loses the ability to bind to PAR-1, exhibited reduced rescue activities for the MTCL1 RNA interference (RNAi) phenotypes (see ΔBR in Fig. 6). Specifically, although many of the ΔBR-expressing cells showed MT accumulation to some extent, they could not organize the MTs into the well-developed apicobasal MT bundles (Fig. 6A,B). Correspondingly, ΔBR did not completely restore the RNAi-induced alterations in the cell height and cell area of MDCK cells (Fig. 6C). These results suggest that the interaction with PAR-1b is important for MTCL1 to fully function in the epithelial polarity development.
As far as we examined, the ΔBR mutant showed MT-bundling activity, dimerization activity and localization in MDCK cells that were comparable to wild-type MTCL1 (supplementary material Fig. S7), thereby meaning that it was unlikely that PAR-1 directly regulates the MT-regulating activity of MTCL1. Instead, the following results suggest that MTCL1 regulates PAR-1 localization in addition to its direct role in MT organization (Fig. 8). Specifically, although PAR-1 localizes to the lateral membrane of polarized epithelial cells (Böhm et al., 1997; Suzuki et al., 2004), detergent extraction before fixation removed the PAR-1b signals on the membrane and left weak, but substantial, signals that colocalized with F-actin and MTs at the sublateral regions (Fig. 8A). These signals were specific because they were sensitive to PAR-1b knockdown (Fig. 8A, asterisks), and ectopically expressed PAR-1b displayed a similar localization (Fig. 8B). MTCL1 knockdown did not affect this localization of the PAR-1b subpopulation (data not shown). However, the interaction with MTCL1 is likely to be, at least partially, responsible for these sublateral localizations of the detergent-insoluble PAR-1b signals, because ectopically expressed wild-type MTCL1 and ΔKR, but not ΔBR, specifically enhanced the localization (Fig. 8C). This was also supported by MT pull-down assays, which demonstrated that the MT-binding ability of PAR-1b in MDCK cell extracts was weakened by MTCL1 knockdown (supplementary material Fig. S8). Taken together, these results indicate the possibility that MTCL1 works as a scaffold protein for PAR-1b and assists this MT-regulating kinase in its interaction with MTs.
Thus, the question arises as to how PAR-1b cooperates with MTCL1 for the development of the apicobasal MT bundles. One insight into this question was provided by analyzing MDCK cells with high expression of MTCL1. These cells sometimes developed extensions of MT bundles from the basal side of the apicobasal MT bundles (Fig. 8D) that were strongly positive for PAR-1b (supplementary material Fig. S8). These protrusions were not observed in cells expressing the MTCL1 mutants ΔN and ΔBR (Fig. 8E), suggesting that these basal extensions of the apicobasal MTs reflected the ability of MTCL1 to promote the apicobasal MT bundle development. In this respect, it is interesting that the stems of the MT bundles in the protrusions were positive for MTCL1 and acetylated tubulin, whereas tyrosinated MTs were predominantly observed at the distal tips of the MT bundles from which they branched out (Fig. 8D, inset). This indicates that the dynamic states of MTs were altered along the protrusions. Taken together with the fact that PAR-1b is the kinase that maintains the dynamic state of MTs (Hayashi et al., 2011), these results indicate that MTCL1 exerts two opposite effects on MT dynamics, that is, not only crosslinking and stabilizing MTs, but also enhancing the plus-end dynamics of the MTs by recruiting PAR-1b, and thereby it promotes the development and/or maintenance of the apicobasal MT bundles (Fig. 8F, bottom).
Epithelial cells specifically develop highly elaborate arrays of non-centrosomal MTs, which are considered to play important roles in the development and maintenance of epithelial polarity (Bacallao et al., 1989; Grindstaff et al., 1998; Müsch, 2004). However, the mechanisms by which these higher-order assemblies of non-centrosomal MTs are constructed remain largely unknown (Bartolini and Gundersen, 2006; Bellett et al., 2009). In the present study, we identified a novel MT-binding protein, MTCL1, which plays essential roles in crosslinking non-centrosomal MTs in the apicobasal MT bundles running along the apicobasal axis of epithelial cells.
Many MAPs, including +TIPs, induce MT bundle formation when overexpressed in cultured cells. Therefore, caution is required when defining bona fide MT-crosslinking proteins (Bartolini and Gundersen, 2006; Chapin et al., 1991; MacRae, 1992). In such situations, several nonmotor MAPs, such as NuMA, PRC1 and TPX2, have been established as genuine MT-crosslinking proteins on the basis of their cellular localizations and loss-of-function phenotypes (Manning and Compton, 2008; Mollinari et al., 2002; Radulescu and Cleveland, 2010; Subramanian et al., 2010). The present results have established that MTCL1 is a new MT-crosslinking protein. First, immunofluorescence microscopy of endogenous MTCL1 revealed its intermittent association with the lateral sides of MT filaments and preferential accumulation to MT bundles in various cellular contexts. Second, siRNA-mediated depletion of MTCL1 impaired the development of the apicobasal MT bundles in polarized epithelial cells. Third, MTCL1 had a direct MT-binding region in its N-terminal region that exhibited intermittent colocalization with MTs. Significantly, this MT-binding region did not induce MT bundling unless it was ligated to the subsequent coiled-coil motifs, not only when it was overexpressed in cells but also in vitro. Because the MTCL1 coiled-coil region is likely to form two-stranded structures, and did indeed show oligomerization activity, these results indicate that, like NuMA, MTCL1 forms dimers that hold the two N-terminal MT-binding domains together, and thereby crosslinks MTs (Fig. 8F, top) (Radulescu and Cleveland, 2010). The proline-rich region between the N-terminal MT-binding region and the subsequent coiled-coil motifs (Fig. 3A) might function as a hinge that enables the N-terminal region to move flexibly. Here, we also identified another MT-binding region in the C-terminus of MTCL1. At the moment, it is not clear how this MT-binding region contributes to the MT-crosslinking activity of MTCL1.
Our rescue experiments demonstrate that the N-terminal MT-binding region is indispensable for the development of the apicobasal MT bundles in epithelial cells. Furthermore, the MT regrowth assays suggested that MTCL1 is involved in the accumulation of non-centrosomal MTs at the lateral cortex. Previous studies have revealed the molecular mechanisms involved in the initial step for the development of this epithelial-cell-specific non-centrosomal MT array, including nucleation, release, transport or guidance, and capture of either end of MTs (Bellett et al., 2009; Hotta et al., 2010; Jaulin and Kreitzer, 2010; Mogensen et al., 2000; Mogensen et al., 2002; Oriolo et al., 2007). In this context, our results supplement these previous studies by providing the molecular basis of the late steps regulating the higher-order array of the apicobasal MT bundles. During the development of epithelial polarity, MTCL1 gradually accumulates at the sublateral cortex (Fig. 2D) in a manner dependent on its MT-binding activity through the N-terminal region (Fig. 5C). However, once MTCL1 achieves this sublateral localization, it is maintained there even in the absence of MTs (Fig. 7A,B). This might indicate the presence of unknown structures in polarized epithelial cells that anchor MTCL1 to maintain its sublateral localization. Recently, filamentous GTPases (septins) were shown to guide the directionality of MT plus-end movement and support MT remodeling in polarizing epithelial cells (Bowen et al., 2011). It will be interesting to examine whether the septin scaffold provides the structural anchoring sites for MTCL1 at the sublateral cortex of polarized epithelial cells.
In the present study, we have demonstrated that MTCL1 is a novel binding partner of PAR-1b. MTCL1 knockdown cells mimic several defects of PAR-1b knockdown cells, and our rescue experiments demonstrated that the interaction with PAR-1 is essential for the complete activity of MTCL1 to facilitate epithelial polarization. Although we did not show the data here, PAR-1b specifically phosphorylates a certain serine residue of MTCL1 in the vicinity of its binding site. However, as far as we examined, the phosphorylation did not directly regulate the MT-regulatory activity of MTCL1. As we demonstrated here, even a deletion of the PAR-1b-binding region did not impair the MT-bundling activity, dimerization activity and subcellular localization of MTCL1. Although we cannot completely exclude the possible involvement of another site of phosphorylation by PAR-1, these results reduce the possibility that PAR-1b regulates MTCL1 activity. Instead, the present results favor the notion that MTCL1 assists the recruitment of the PAR-1b MT-destabilizing kinase to coordinate the assembly and dynamics of MTs in the apicobasal MT bundles (Fig. 8F, bottom). This idea is consistent with our observation that the MT stability in the apicobasal MT bundles changed dynamically during the reconstruction process of the structures (Fig. 7A). It is also supported by a recent study demonstrating that the coordination between stable and dynamic MTs is important for normal elongation of MT bundles (Bitan et al., 2012). Interestingly, a similar scaffold function for PAR-1 was suggested for septins that interact with PAR-1-related kinases in yeast (Spiliotis, 2010). Although future studies are required, the hypothesis that MT-bundling factors recruit PAR-1 to regulate MT dynamicity provides an important conceptual framework to clarify the molecular basis of non-centrosomal MT organization in various polarized cells. It should also be clarified in future studies why PAR-1b, as well as MTCL1 knockdown, commonly induces a decrease in cell height and increase in cell area. Although MT disruption in polarized MDCK cells did not significantly affect the cell height (Fig. 7), our present results demonstrate that the morphological defects in MTCL1 knockdown cells were dependent on the N-terminal MT-binding region of MTCL1. Therefore, correct MT organization could be tightly coupled with normal development of the epithelia-specific morphology specifically during the cell polarization processes.
Recently, public databases have been adopting the name ‘SOGA2’ for MTCL1, which is probably because a recent paper that named its paralog molecule suppressor of glucose by autophagy (SOGA) (Cowerd et al., 2010). The paper argued that the ‘MTCL1 paralog’ was secreted as 80 kDa and 25 kDa peptides after excision of an ‘internal peptide sequence’ and mediated adiponectin-dependent suppression of glucose production by inhibiting autophagy. However, the feature of this putative internal peptide sequence in the MTCL1 paralog is not at all conserved in MTCL1 and, as we demonstrated here, full-length MTCL1 is expressed intracellularly at the expected molecular mass and interacts with MTs. Furthermore, as far as we examined, the full-length cDNA encoding mouse SOGA (GenBank number NM_001164663) is also expressed intracellularly as the expected full-length protein with a molecular mass of 180 kDa, and shows a similar MT-bundling activity to MTCL1 through the conserved KR-rich region in the C-terminus (A.S., unpublished data). Therefore, we strongly claim that the name SOGA2 should not be adopted for the protein we identified in this study, and that it should be called MTCL1 based on the results presented here.
Materials and Methods
The human PAR-1b cDNA (NM_017490) was reported previously (Suzuki et al., 2004). The cDNA of human KIAA0802 (AB018345) was obtained from Kazusa DNA Research Institute (Chiba, Japan), and that of full-length mouse MTCL1 (AK147205) was purchased from Danaform (Kanagawa, Japan). To purify SBP-tagged N-terminal fragments of MTCL1 (N1 and N4) from human HEK293T cells, the nucleotides of the mouse cDNA were codon-optimized for human cells (Genescript, Piscataway, NJ). The cDNAs were subcloned into appropriate expression vectors: pEGFP (Takara Bio Inc., Shiga, Japan); pGEX (GE Healthcare Japan, Tokyo, Japan); pSRD-derivative vectors (pSRHA or pSRHis) (Masuda-Hirata et al., 2009); pCAGGS-V5 and pCAGGS-Flag-SBP (Yamashita et al., 2010). In some cases, the cDNAs were subcloned in pEB6CAG (Tanaka et al., 1999), pEB-Hyg (Wako Pure Chemical Industries Ltd, Osaka, Japan) or pOS-Tet14-SBP (Cong et al., 2010), all of which are Epstein–Barr virus (EBV)-based extrachromosomal vectors carrying a replication origin (oriP) and replication initiation factor (EBNA-1) sufficient for autonomous replication in human and canine cells. Site-directed mutations were introduced using a QuikChange site-directed mutagenesis kit (Agilent Technologies, Santa Clara, CA), followed by sequencing verification.
The anti-PAR-1b polyclonal antibody (pAb) was reported previously (Suzuki et al., 2004). The anti-MTCL1 pAb was raised by immunizing rabbits with the C-terminal fragment of KIAA0802 (C7: 1322–1420) fused with GST, followed by affinity-purification. Other antibodies were purchased from commercial sources as follows: anti-KIAA0802 (W19), anti-tyrosinated tubulin rat monoclonal antibody (mAb) (YL1/2), anti-β-catenin pAb (H102), anti-ZO-1 rat mAb, anti-GFP mAb (B2), anti-SBP mAb (SB19-C4) (Santa Cruz Biotechnology, Santa Cruz, CA); anti-α-tubulin mAb (DM1A), anti-acetylated tubulin mAb (Sigma-Aldrich, St. Louis, MO); anti-V5 mAb, anti-RFP pAb, anti-ZO-1 mouse mAb (Life Technologies Corporation, Carlsbad, CA); anti-E-cadherin mAb (BD Biosciences, San Jose, CA); anti-V5 pAb (Bethyl Laboratories, Montgomery, TX); anti-GFP pAb (MBL, Aichi, Japan); anti-Na+/K+ ATPase mAb (Millipore, Billerica, MA); anti-HA rat mAb (3F10) (Roche Applied Science, Indianapolis, IN).
Cell culture, transfection, and establishment of pEBV stable transformants
HeLa-K cells were a generous gift from Dr S. Tsukita (Osaka University, Osaka, Japan). MDCK II, HeLa-K and HEK293T cells were cultured in Dulbecco's modified Eagle medium (Life Technologies Corporation) containing 10% fetal bovine serum, penicillin (100 U/ml), streptomycin (0.1 µg/ml) and 1 mM glutamine. Plasmid transfections were performed using the Lipofectamine 2000 reagent (Life Technologies Corporation) or an electroporation system, NEPA21 (Nepa Gene Co. Ltd, Chiba, Japan) for MDCK II cells and Attracten (Qiagen, Valencia, CA) for HeLa-K cells, according to each manufacturer's instructions. Heterogeneous stable transformants of MDCK cells were established using pEB-based vectors as previously described (Masuda-Hirata et al., 2009).
The RNAi target sequences for canine MTCL1 were designed based on the cDNA sequence (Genbank number XM_547667) as follows: 5′-GGCACGGCUACAGAUCAAU-3′ (#1); 5′-CCUGUGACGAGUGCAAGUU-3′ (#2); 5′-GUUCAGCACGGGAAAACCU-3′ (#3). To establish MTCL1-depleted cells, MDCK cells were transfected with pEB6-Super-gfp encoding an shRNA sequence for canine MTCL1 (#1 or #2), and selected with neomycin (Masuda-Hirata et al., 2009). In some cases, we used pEB6-super-AcGFP1-Nuc, which was newly generated by replacing EGFP in pEB6-Super-gfp with AcGFP-Nuc in pAcGFP1-Nuc (Takara Bio Inc.). For rescue experiments, cells were co-transfected with pEB-Super-gfp expressing an shRNA for canine MTCL1 (#1) and pEB-Hyg expressing various mutants of human MTCL1, and subjected to selection using 800 µg/ml geneticin (Life Technologies Corporation) and 200 µg/ml hygromycin (Wako Pure Chemical Industries Ltd). As a nonsilencing RNAi oligonucleotide, Allstars negative control siRNA (Qiagen) was used. The nonsilencing control sequence used for pEB-super-gfp was described previously (Suzuki et al., 2004).
Cells were solubilized in lysis buffer (25 mM Tris-HCl pH 7.5, 150 mM NaCl, 2.5 mM MgCl2, 0.1% Triton X-100, 1 mM DTT) containing a cocktail of protease and phosphatase inhibitors (Roche Applied Science) for 30 minutes at 4°C, briefly sonicated, and centrifuged at 15,000 g for 30 minutes. The resulting supernatants were mixed with protein-A– or protein-G–Sepharose (GE Healthcare) conjugated with appropriate antibodies. To purify SBP-tagged proteins, streptavidin-conjugated Sepharose (GE Healthcare) was used, and the absorbed proteins were eluted with an appropriate buffer containing 10 mM biotin.
For pull-down assays, purified proteins dissolved in BRB buffer (80 mM PIPES-KOH pH 6.8, 1 mM MgCl2, 1 mM EGTA) were incubated with taxol-stabilized MTs (final concentration: 0.5 mg/ml) at 37°C for 30 minutes, and subjected to centrifugation (50,000 g) at 22°C for 30 minutes on a cushion of 40% glycerol in BRB buffer. When MDCK cell extracts were subjected to the assays, cells (1×106) were transfected with 60 pmol MTCL1-targeting siRNA (#1) using NEPA21, and homogenized in BRB buffer containing 1 mM DTT and 100 mM KCl at 4°C before mixing with taxol-stabilized MTs as described above. For MT-bundling assays, SBP-tagged proteins were purified from HEK293T cells, mixed with taxol-stabilized Rhodamine-labeled MTs at a molar ratio of 4∶1 at 37°C for 30 minutes, and gently spotted onto glass slides for immediate examination under fluorescence microscopy.
Unless otherwise noted, MDCK cells were seeded onto #3460 Transwell filters at 2.5×105 cells/cm2 and cultured for 3–4 days to polarize the cells sufficiently. HeLa-K cells were seeded at 3×104 cells/cm2 immediately before transfection, and subjected to fixation on the next day. The cells were fixed with cold methanol for 10 minutes at −20°C (for MTs and MTCL1) or 2% paraformaldehyde followed by permeabilization with 0.5% Triton X-100. To stain the detergent-resistant subpopulation of PAR-1b, cells were pretreated with PGEM buffer (100 mM PIPES-KOH pH 6.8, 1 mM EGTA, 1 mM MgCl2, 15% glycerol) containing 0.5% Triton X-100 for 1 minute before fixation. To observe the MT regrowth process, MDCK cells were treated with 33 µM nocodazole for 2 hours on ice followed by 1 hour at 37°C, and washed with cold medium three times before addition of warm medium and incubation at 37°C. At appropriate times after nocodazole removal, cells were subjected to pretreatment with PGEM buffer containing 0.5% Triton X-100 for 30 seconds followed by methanol fixation. The fixed cells were subjected to standard immunostaining procedures as described previously (Suzuki et al., 2004), using appropriate secondary antibodies conjugated with Alexa Fluor 488, Alexa Fluor 555 or Alexa Fluor 647 (Life Technologies Corporation). For visualization of F-actin, cells were labeled with Rhodamine–phalloidin (Life Technologies Corporation). Most of the samples were examined at room temperature using an AxioImager Z1 microscope (Carl Zeiss, Oberkochen, Germany) equipped with a CSU10 disc confocal system (Yokogawa Electric Corporation, Tokyo, Japan), an Orca II CCD camera (Hamamatsu Photonics, Shizuoka, Japan), and a 63× 1.4 NA Plan Apochromat or 100× 1.46 NA objective. Images were acquired using MetaMorph software (Molecular Devices, Sunnyvale, CA) and processed with ImageJ software to obtain appropriate brightness and contrast. To obtain images with higher resolutions, another confocal microscopy system (LSM710; Carl Zeiss) with a 100× 1.46 NA objective was used.
A.S. supervised and performed experiments, analyzed data and wrote the paper, Y.S., M.A. and Y.A. performed main parts of experiments. K.Y., M.I., K.S., A.Y. performed some experiments, H.H. and N.A. performed mass spectrometric analyses, T.M. and I.H. helped MT pull down assays, S.O. supported the experiments and provided helpful suggestions.
This work was supported by KAKENHI (to A.S. and S.O.); and the ‘Establishment of Research Center for Clinical Proteomics of Post-translational Modifications’ program as part of the Special Coordination Fund for Promoting Science and Technology ‘Creation and Innovation Centers for Advanced Interdisciplinary’ from the Ministry of Education, Culture, Sports, Science and Technology of Japan.