MCAM contributes to the establishment of cell autonomous polarity in myogenic and chondrogenic differentiation

ABSTRACT Cell polarity has a fundamental role in shaping the morphology of cells and growing tissues. Polarity is commonly thought to be established in response to extracellular signals. Here we used a minimal in vitro assay that enabled us to monitor the determination of cell polarity in myogenic and chondrogenic differentiation in the absence of external signalling gradients. We demonstrate that the initiation of cell polarity is regulated by melanoma cell adhesion molecule (MCAM). We found highly polarized localization of MCAM, Moesin (MSN), Scribble (SCRIB) and Van-Gogh-like 2 (VANGL2) at the distal end of elongating myotubes. Knockout of MCAM or elimination of its endocytosis motif does not impair the initiation of myogenesis or myoblast fusion, but prevents myotube elongation. MSN, SCRIB and VANGL2 remain uniformly distributed in MCAM knockout cells. We show that MCAM is also required at early stages of chondrogenic differentiation. In both myogenic and chondrogenic differentiation MCAM knockout leads to transcriptional downregulation of Scrib and enhanced MAP kinase activity. Our data demonstrates the importance of cell autonomous polarity in differentiation.


INTRODUCTION
Determination of cell polarity is one of the core principles in developmental biology (Leung et al., 2016). Experiments in model organisms, ranging from the fruit fly to mouse, have demonstrated how cell polarity and differentiation are influenced by morphogen gradients and local tissue architecture (Heller and Fuchs, 2015;Lawrence and Casal, 2013;Lawrence et al., 2007). However, an important question has remained unansweredhow do differentiating cells achieve polarity in the absence of external signals? An example here is the skeletal muscle cell that even in vitro elongates in a highly polarized orientation.
Cell polarity is established by an intricate network of positive and negative protein interactions (Campanale et al., 2017;Devenport, 2014). SCRIB (Scribble) is a tumour suppressor and one of the regulators of cell polarity. It interacts with Rho guanine nucleotide exchange factor 7 (ARHGEF7; beta-PIX), thereby controlling cytoskeletal organization and diverse signalling pathways (Audebert et al., 2004). SCRIB binds directly to VANGL2 (Van-Gogh-like 2), another principal component of cell polarity establishment (Kallay et al., 2006). Additional members of the SCRIB complex include DLG1 (discs, large homolog 1) and LLGL1 (lethal giant larvae homolog 1). Knockout of these polarity regulators leads to severe embryonic malformations in mice (Caruana and Bernstein, 2001;Klezovitch et al., 2004;Murdoch et al., 2003Murdoch et al., , 2001Pearson et al., 2011;Yin et al., 2012). Disruption of basolateral SCRIB polarity complex causes expansion of apical PAR3-PAR6-aPKC complex, illustrating reciprocally repressive interactions (Bilder and Perrimon, 2000;Bilder et al., 2003). PAR3 and 6 are PDZ domain-containing scaffolding proteins that are essential in polarity establishment (Etemad-Moghadam et al., 1995;Hung and Kemphues, 1999). They interact with GTP-ase CDC42 (cell division cycle 42) to regulate downstream signalling pathways (Joberty et al., 2000). An ever-increasing number of additional proteins are involved in the establishment of cellular asymmetry. Much of our current knowledge about cell polarity arises from research in invertebrate models and only few studies have addressed its role in vertebrate mesoderm development. It has been shown how in embryonic development WNT11 can act as a directional cue for myotube elongation by activating the planar cell polarity pathway (Gros et al., 2009). Nevertheless, multiple concurrent signalling pathways may be induced by WNT11, which also regulates neuromuscular junction formation via β-catenin and VANGL2 (Messéant et al., 2017). Polarity pathway components are involved in the asymmetric division of satellite cells in the skeletal muscle (Le Grand et al., 2009;Ono et al., 2015). Cell polarity guides also chondrocyte proliferation in the elongation of long bones (Gao et al., 2011;Li and Dudley, 2009;Wang et al., 2011).
Cell migration in response to external signals relies on asymmetric distribution of surface receptors and cytoskeleton. Such transient polarization is required for providing the cell directionality and guiding the formation of protrusions. In migrating melanoma cells non-canonical WNT signalling leads to cytoskeletal rearrangement and asymmetric distribution of MCAM (melanoma cell adhesion molecule, CD146) through its controlled endocytosis (Witze et al., 2008). MCAM is targeted to the posterior end of the migrating cell, where it forms part of the WNT5Areceptor-actin-myosin-polarity (WRAMP) structure (Witze et al., 2008). WRAMP is stably maintained during sustained periods of directional cell migration, but disbanded in cells as they pause or change direction (Connacher et al., 2017). MCAM is highly expressed in embryonic development and is maintained in postnatal skeletal muscle satellite cells and osteogenic mesenchymal stromal cells (Alexander et al., 2016;Chan et al., 2005;Li et al., 2003;Pujades et al., 2002;Sacchetti et al., 2007;Shi and Gronthos, 2003;Shih and Kurman, 1996), and it is markedly upregulated in metastatic tumours (Johnson et al., 1996;Wang and Yan, 2013). Despite its high expression throughout embryonic development, still little is known on its function in cell differentiation. It can bind directly WNT1, 3 and 5 to regulate Dishevelled (DVL) and C-JUN ( jun proto-oncogene) phosphorylation (Ye et al., 2013). MCAM acts via NFAT (nuclear factor of activated T-cells) and JNK (MAPK8, mitogen-activated protein kinase 8) pathways to regulate asymmetry in zebrafish and Xenopus embryonic development (Gao et al., 2017). Here we aimed to investigate the role of MCAM in the establishment of cell autonomous polarity in differentiating cells. We show that MCAM is asymmetrically distributed at the tip of elongating myotube, where it colocalizes with actin binding protein Moesin (MSN) and cell polarity pathway regulators VANGL2 and SCRIB. CRISPR-Cas9 mediated knockout of MCAM or deletion of its endocytosis motif leads to loss of cell polarity and failure in myotube directional elongation. MCAM knockout has a detrimental impact on early chondrogenic differentiation. Our study reveals a novel role for MCAM in regulating cell polarity in mesoderm differentiation.

Generation and initial characterization of MCAM mutant cell lines
We sought a model that would enable us to follow cellular differentiation in vitro from the onset of fate commitment to terminal differentiation. We chose the multipotent embryonic 10T1/2 cells as these can be induced to chondrogenic fate by recombinant BMP2 (Denker et al., 1999;Shea et al., 2003) and to myogenic fate by exposing them to testosterone after initial brief culture with 5-azacytidine (Singh et al., 2003) (Fig. 1A)  CRISPR-Cas9-mediated genome editing, followed by FACS and single-cell cloning to generate three cell lines (Fig. 1B). Lines C149 and C164 were made by targeting the second exon of Mcam in order to achieve complete loss-of-function effect ( Fig. 1B; Fig. S1A  long) has a C-terminal endocytosis motif at amino acids 643-646 and since endocytosis has been proposed to regulate asymmetric MCAM trafficking in cell migration (Witze et al., 2008) we generated a third line, U125, that carries three mutations that all lead to the specific elimination of the endocytosis motif.
Mcam expression was significantly upregulated in wild-type 10T1/2 and U125 cells upon differentiation, when either testosterone or BMP2 was added to the culture medium ( Fig. 1C-E). Western blot analysis and immunohistochemistry proved the absence of MCAM in lines C149 and C164 (Fig. 1C,E). In their undifferentiated state the morphology of MCAM knockout cells did not differ from unedited 10T1/2 cells (Fig. 1F). Therefore MCAM is largely dispensable for fibroblasts, making our system suitable for studying its role in the context of differentiation.
MCAM is required for chondrogenic and myogenic differentiation in vitro BMP2 exposure of 10T1/2 cells leads to sequential activation of chondrogenic and osteogenic differentiation programs (Shea et al., 2003). BMP2 induced chondrogenic differentiation of 10T1/2 cells by 4 days in culture, which was evident in intense Alcian Blue staining ( Fig. 2A; Fig. S2A) and upregulation of Sox9 (Fig. 2C) (Bi et al., 1999). Knockout of MCAM abolished chondrogenic differentiation in all the established cell lines ( Fig. 2A,C). Wild-type BMP2-treated cells upregulated hypertrophic chondrocyte marker Col10a1 by the seventh day in culture, whereas its levels remained low in MCAM mutant cells (Fig. 2D). Alkaline phosphatase-positive osteogenic nodules formed in the culture of wild-type cells (Fig. 2B) and terminal osteogenic differentiation marker osteomodulin (Omd) was upregulated (Fig. 2E). Osteogenic differentiation was significantly impaired in the absence of MCAM (Fig. 2B,E).
In myogenic-promoting conditions early transcriptional regulators Myod1 and myogenin (Myog) were induced in both 10T1/2 and Mcam mutant cells after 4 days exposure to testosterone (Fig. 2F,H-I; Fig. S2B-C). Despite this seemingly unaffected initiation of the myogenic program, terminal differentiation into multinucleated myotubes was severely impaired in cells lacking MCAM or its endocytosis motif. MCAM mutant cells fused into multinucleated myogenic cells that failed to elongate ( Fig. 2G;  Fig. S2D). There was a correspondingly reduced expression of myosin heavy chain in Mcam mutant cell lines (Fig. 2J). These data show the importance of MCAM in regulating mesodermal differentiation. Our results highlight the differences between the role of MCAM in chondrogenic and myogenic commitment, whereby it is essential at the onset of the former, but not until the elongation stage in the latter.
We next analysed whether loss of MCAM may lead to enhanced expression of other CAMs. We found that neither neural cell adhesion molecule (Ncam1) isoforms 1 and 3 were upregulated in the knockout cell lines in comparison to wild-type cells (Fig. 3A). As expected, Ncam1 skeletal muscle-specific isoform was significantly upregulated in wild-type 10T1/2 cells in comparison to knockout cells after 7 day exposure to testosterone. Intercellular adhesion molecule 1 (Icam1) was only upregulated significantly in wild-type cells in myogenic growth conditions (Fig. 3B). Icam2 was expressed at low levels in all the cell lines. Vascular cell adhesion molecule 1 (Vcam1) was most highly expressed in undifferentiated cells, but its expression levels did not depend on the presence of MCAM (Fig. 3C). These data indicate that loss of MCAM does not cause a compensatory upregulation of other CAMs and rather the absence of MCAM leads to downregulation of CAMs that are characteristic of differentiating cells.

Asymmetric distribution of MCAM regulates cell autonomous polarity
We next addressed the mechanisms by which MCAM influences differentiation. First we approached the established signalling pathways downstream of MCAM: it is known to induce DVL2 and C-JUN phosphorylation in cancer cells and its knockdown leads to the induction of canonical WNT pathway (Ye et al., 2013). We noticed increased C-JUN phosphorylation under both osteochondrogenic and myogenic conditions, but this occurred even in the absence of MCAM (Fig. 4A). DVL2 activity decreases in myogenesis (Yamaguchi et al., 2012) and we confirmed this here, however there were no consistent differences between wild-type and MCAM mutant cells (Fig. 4B). We did not detect significant changes in the expression of canonical WNT transcription factors T-cell factor 4 (Tcf4) and lymphoid enhancer binding factor 1 (Lef1) (Fig. S3A). Neither did we observe nuclear translocation of β-catenin or change in WNT5A staining ( Fig. S3B-C). In cancer cells MCAM has been linked with Id1 expression (Zigler et al., 2011), which is a known inhibitor of myogenesis (Jen et al., 1992). However, we did not detect altered Id1 expression in MCAM mutant cell lines, indicating that this pathway is unlikely to explain defective myogenic differentiation (Fig. 4C). Several lines of evidence link MCAM with actin cytoskeleton. It interacts directly with actin binding proteins hShroom1 (Dye et al., 2009) and MSN . We tested the phosphorylation levels of MSN, but detected only a small decrease in its activity in MCAM mutant cells (Fig. 4D). We found significant morphological changes in MCAM mutant cells, which are clearly visible in actin cytoskeletal organization (Fig. 4E). Most notably, myoblasts fused together into multinucleated cells that expanded in size without elongating and aligning the actin cytoskeleton along the direction of the cell. This led us to hypothesize that myotube elongation relies on cell autonomous initiation of polarity and defects in this process may underlie the phenotype seen in MCAM mutant cells. Furthermore, MSN is a known regulator cytoskeletal organization in cell polarity (Fehon et al., 2010;Jankovics et al., 2002;Médina et al., 2002;Polesello et al., 2002;Speck et al., 2003). Abnormal localization of polarity complex rather than large changes in its activity may therefore lead to the observed differentiation defects. We analysed whether MCAM and MSN are asymmetrically distributed in differentiating cells. We found that in elongating myotubes MCAM colocalizes with MSN at the distal tip of the cell (Fig. 4F). We detected asymmetric distribution of MCAM also in in vivo E17.5 mouse forelimb developing myotubes (Fig. 4G,H;  Fig. S3D). Polarity in chondrogenic differentiation remains less obvious, although we could detect MCAM and MSN colocalization in 10T1/2 cells exposed to BMP2 (Fig. 4I). We analysed growing juvenile mouse ear cartilage and found that at the edges, where the cartilage is growing, MCAM is weakly expressed and cytoplasmic but acquires membrane localization in mature cartilage (Fig. 4J,K). Polarization may occur transiently in cartilage maturation and remains difficult to establish in vivo. We next analysed the localization of cell polarity-associated proteins in MCAM mutant cells in comparison to wild-type cells undergoing myogenic differentiation. Similarly to MSN and MCAM we found highly asymmetric distribution of VANGL2 at the distal tips of elongating myotubes (Fig. 5A,B; Fig. S4A). Remarkably, in MCAM knockout cells such polarity was not established and MSN labelled the whole plasma membrane, whereas VANGL2 was uniformly distributed across the cell ( Fig. 5C; Fig. S4A). We detected colocalization of SCRIB with MCAM at the tip of an elongating myotube ( Fig. 5D; Fig. S4B). Knockout of MCAM led to uniform distribution of SCRIB (Fig. 5E) and importantly, also its transcriptional downregulation (Fig. 5H). SCRIB gene regulatory elements are not known and the precise transcriptional link remains to be established. Intriguingly, elimination of MCAM endocytosis motif led to frequent localization of SCRIB on actin filaments (Fig. S4C-C″). SCRIB trafficking to cortical actin was recently established to rely on β spectrin (Boeda and Etienne-Manneville, 2015). Our data suggests that MCAM participates in SCRIB trafficking.
Since SCRIB complex antagonizes cell cortex targeting of PAR3, we analysed its localization and found that in wild-type cells PAR3 was largely cytoplasmic, whereas in MCAM mutant myotubes it was frequently enriched at the membrane (Fig. 5F,G). Also in MCAM endocytosis motif mutant cells, VANGL2 and PAR3 were mislocalized (Fig. 5I). We conclude that subcellular trafficking of MCAM is important in the establishment of cell autonomous polarity in myogenic differentiation.
We found significant downregulation of Scrib in MCAM mutant cell lines in osteochondrogenic differentiation (Fig. 5H). Intriguingly, while in wild-type BMP2-exposed cells VANGL2 showed only limited asymmetric localization, in the absence of MCAM it accumulated across the cytoplasm (Fig. 5J). We could also detect membrane targeting of PAR3 in MCAM mutant cells, although it remained less prominent than in myogenic differentiation (Fig. S4D). We propose that cell polarity has a more dynamic role in chondrogenic differentiation.
Several signalling pathways are regulated by cell polarity. Loss of SCRIB leads to enhanced MAP kinase pathway (ERK1/2) activity (Nagasaka et al., 2010) and since tight control over ERK1/2 signalling is required in chondrogenesis (Bobick and Kulyk, 2004) we tested its activity in MCAM mutant cell lines. While in both chondrogenic and myogenic differentiation wild-type cells downregulate ERK1/2 phosphorylation, in MCAM mutant cells their levels remained as high as in undifferentiated cells, providing a mechanistic insight into signalling changes in the cells (Fig. 5K).

DISCUSSION
We have established an in vitro assay that enabled us to monitor cell polarity determination throughout myogenic and chondrogenic differentiation. Our data enables us to propose a model on how myotubes achieve elongated tubular morphology. We propose that highly asymmetric localization of MCAM enables the adhesion and anchoring of the rear end of an elongating myotube (Fig. 6). MCAM is known to interact directly with MSN  and this interaction explains how MCAM-dependent polarity is translated to actin cytoskeleton arrangement in an elongating cell. We found that MCAM is required for the asymmetric distribution of cell polarity regulators SCRIB and VANGL2. In the absence of MCAM, asymmetric localization of MSN fails and actin cytoskeleton is dispersed in diverse directions across the cell, preventing myotube elongation. Thus the emerging model suggests that myotube elongation is reminiscent of cancer cell or leukocyte migration, with the exception that the rear of the cell is firmly adhered and does not follow the leading edge. The cell polarity-associated proteins have well-established roles in cell migration (Ebnet et al., 2017). Both MSN and MCAM are highly expressed in metastatic melanoma cells and known to regulate cell migration (Connacher et al., 2017;Estecha et al., 2009;Lei et al., 2015;Witze et al., 2008). Current evidence indicates that MCAM is a non-selective receptor for a wide number of extracellular signalling molecules that include WNT5A, WNT1, FGF4 and S100A8/A9 (Gao et al., 2017;Ruma et al., 2016;Ye et al., 2013). It is also a co-receptor for vascular endothelial growth factor receptor 2 (Jiang et al., 2012). Therefore diverse signals are capable of inducing or modulating MCAM activity in vivo. In our in vitro system cells polarized in the absence of extracellular signalling molecule gradients, suggesting that cells have also an autonomous polaritybased mechanism that regulates their shape. This intrinsic mechanism may explain how cell polarity and distinct morphology can arise in tissues such as skeletal muscle and cartilage where, due to high density and consequent poor penetrance, morphogen gradients are difficult to be established. Similarly to our study, merlin and ERMfamily proteins (ezrin, radixin, MSN) were found capable of inducing cell cortex cytoskeleton asymmetry in the absence of external cues in epithelial cells (Hebert et al., 2012). The importance of cell polarity has remained poorly characterized in cartilage. Non-canonical WNT signalling controls cell division plane in the development of long bones (Li and Dudley, 2009). Loss of VANGL2 leads to limb defects in mice . We show that MCAM knockout leads to failure to initiate chondrogenic differentiation program in vitro and causes mislocalization of actin cytoskeleton and VANGL2. We hypothesize that in the absence of MCAM tight intercellular contacts that are needed in chondrogenesis Mice deficient of MCAM or MSN do not show significant developmental defects, albeit in both models abnormalities have been observed in postnatal pathological conditions (Doi et al., 1999;Jouve et al., 2015;Okayama et al., 2008;Tu et al., 2013;Zeng et al., 2014). ERM proteins show functional redundancy in most cell types, explaining the limited phenotype of MSN knockout mice (Fehon et al., 2010). It is likely that in vivo other proteins may compensate for the lack of MCAM. The hereby used in vitro system has enabled us to establish its role in a minimal environment with limited ligands and adhesion substrates. Our data collectively provides evidence that cellular differentiation relies on an intrinsic, adhesion receptor dependent polarity mechanism.

Genome editing
CRISPR-Cas9 guide RNAs were designed in Benchling software (Benchling, Inc., https://benchling.com/academic). For N-terminal mutations in cell lines C149 and C164 guide RNA GCCAGTACCCACGCCCGACC/TGG, and for the C-terminal mutation in line U125 guide RNA GGGCAGCAACG-GTGACAAGA/GGG, was used (PAM sequence in bold). These guides were cloned into SpCas9(BB)-2A-GFP plasmid as previously described (Ran et al., 2013). pSpCas9(BB)-2A-GFP was a gift from Feng Zhang (Massachusetts Institute of Technology, Cambridge, MA, USA; Addgene plasmid # 48138). Authenticated 10T1/2 cells were purchased from Public Health England (European Collection of Authenticated Cell Cultures, Public Health England, Porton Down, UK) and cultured in MEM (GIBCO), supplemented with 1% L-Glutamine, 10% foetal bovine serum (FBS), 1% penicillin-streptomycin, 1× MEM Non-Essential Amino Acids Solution. Cells were transfected with guide RNA containing plasmid using Lipofectamine 3000 (Thermo Fisher Scientific) and after 24 h culture the GFP-expressing cells were sorted into 96well plates by BD FACS Aria. Confluent wells were split to isolate genomic DNA. The guide RNAs were designed so that indels would destroy a restriction enzyme binding site (XcmI for N-terminal and BpmI for C-termina edit). The PCR product was further verified by Sanger sequencing and analysis using CRISPR-ID (Dehairs et al., 2016). Selected lines were confirmed by cloning PCR product and sequencing individual bacterial colonies. For differentiation assays cells were plated at a density of 3200/cm 2 . Myogenic differentiation of 10T1/2 cells was carried out as described (Singh et al., 2003), by initial 3-day culture in 20 μM 5-Azacytidine (Sigma Aldrich, A2385), followed by 3-day recovery in growth medium and 4-7-day incubation with 100 nM testosterone. For chondrogenic differentiation cells were incubated with recombinant BMP2 (200 ng/ml, Cat. No. 355-BM-050 Systems Inc., Minneapolis, MN, USA) for 4-7 days.

Alcian Blue and alkaline phosphatase staining
Alcian Blue staining was carried out by washing the sample 3×5 min in PBS supplemented with 0.1% Tween (PTW) and 2×5 min in 3% acetic acid. Thereafter Alcian Blue (Sigma-Aldrich, 1% solution in 3% acetic acid) was added to the sample for 30 min, after which it was washed off with 3% acetic acid followed by 3× washing in PBS (5 min each). Alkaline phosphatase staining was carried out using Vector Laboratories Red AP kit.

Immunohistochemistry
Cells were grown on glass-bottom plates (24-well Sensoplate, Cat. No. 662892, Greiner Bio-One, Stonehouse, UK), washed with PBS and fixed in 4% paraformaldehyde for 10 min. The fixed cells were washed 3×5 min in PTW and incubated for 3 h in blocking buffer (10% donkey serum in PTW) and overnight with primary antibodies in blocking buffer. On the second day the samples were washed 3×5 min and 3×30 min in PTW, incubated for 30 min in 1% BSA in PTW and for 2 h with secondary antibodies in PTW (room temperature). Following the removal of the secondary antibody the samples were washed 3×15 min in PTW and 3×15 min in PBS. Inverted Zeiss Axio Observer Z1 was used together with Zeiss Zen software. Leica SP5 confocal was used together with the Leica Application Suite software. Images were acquired at 2× line average and 3-5× frame average, using sequential scanning mode. Single focal planes are shown.

Western blot
Cell samples were aspirated of media, washed 3× with cold PBS then placed onto ice. Cold lysis (RIPA) buffer was added and the cells were scraped into sterile, pre-cooled tubes. These were incubated (with rotation) for 30-60 min at 4°C. After incubation, the samples were centrifuged and supernatant removed to a new pre-cooled tube. Protein concentration of each sample was determined using a Bradford assay. 20 ng of each sample was reduced and run on a 4-12% NuPage Novex precast gel, following the NuPage Denaturing Electrophoresis protocol (Cat. No. NP0322BOX, Thermo Fisher Scientific Inc., Waltham, MA, USA). The gels were transferred onto nitrocellulose membranes following standard protocol. The membranes were blocked using either 5% non-fat milk or 5% BSA (for phosphorylation-specific antibodies) in TBS-Tween for 60-120 min at RT. Primary antibodies were added into the relevant blocking solution and membranes incubated overnight at 4°C. After incubation they were washed 3× with TBS-Tween. Secondary antibodies (in fresh blocking solution) were added to the membranes and incubated for 60-120 min at RT. They were washed 3× with TBS-Tween after incubation. HRP substrate was added to each membrane. Chemiluminescent film was exposed to the membrane and developed using an automatic processor, JP-33.

Antibodies
The antibodies were used at the following concentrations:

Statistical analysis
Statistical significance of RT-qPCR results was determined by two-tailed t-test in Microsoft Excel.