Tmem2 regulates cell-matrix interactions that are essential for muscle fiber attachment

In vertebrates, most skeletal muscles derive from precursors found within the somites, repetitive segments of paraxial mesoderm that flank the embryonic notochord (Bryson-Richardson and Currie, 2008; Buckingham and Vincent, 2009). As muscle precursors mature, they elongate to form fibers that span each segment and attach to the somite boundaries (Goody et al., 2015). Attachments are created through direct interactions of muscle fibers with the extracellular matrix (ECM), and sites of attachment develop into the myotendinous junction (MTJ), which transmits muscular forces to the skeletal system (Charvet et al., 2012). Thus, cell-matrix connections facilitate the morphology, integrity and function of developing muscles. In contrast, failure to maintain fiber attachments can lead to the progressive tissue degeneration that underlies muscular dystrophy. Although numerous causative mutations have been associated with congenital muscular dystrophies (Bertini et al., 2011; Kirschner, 2013), our understanding of the molecular mechanisms that regulate muscle fiber attachment remains incomplete.

Several protein complexes are known to play primary roles in connecting muscle cells to the MTJ (Charvet et al., 2012; Goody et al., 2015; Thorsteinsdóttir et al., 2011). Within the ECM, deposition of both fibrillar fibronectin and polymerized laminin is crucial for successful anchoring of muscle fibers. These ECM molecules are engaged by a variety of transmembrane receptors at fiber termini, including integrin heterodimers and the dystrophin-associated glycoprotein complex (DGC). In collaboration with cytoplasmic proteins such as focal adhesion kinase (FAK) and paxillin, these receptors form cell-matrix adhesion complexes (CMACs) that link the extracellular environment to the cytoskeleton and thereby facilitate both force transmission and signaling. Whereas the importance of ECM and CMAC components is well documented, it is less clear how deposition of the ECM is controlled or how CMAC assembly is regulated in order to insure appropriate cell-matrix interactions.

The use of the zebrafish as a model organism provides valuable opportunities for interrogating the functions of genes involved in muscle fiber attachment (Berger and Currie, 2012; Gibbs et al., 2013). Here, we show that the zebrafish gene transmembrane protein 2 (tmem2) plays an important and previously unappreciated role in regulating cell-matrix interactions at the MTJ. Tmem2 is a type II transmembrane protein with a small cytoplasmic domain, a single-pass transmembrane domain and a large ectodomain (Smith et al., 2011; Totong et al., 2011). Prior studies have demonstrated that tmem2 regulates the regional restriction of the cardiac atrioventricular canal (Smith et al., 2011; Totong et al., 2011). In addition, embryos lacking both maternal and zygotic supplies of tmem2 (MZtmem2) exhibit earlier defects in multiple tissues, including aberrantly shaped somites (Totong et al., 2011). Through analysis of the somite defects in MZtmem2 mutants, we find that loss of tmem2 function leads to muscle fiber detachment. Our results indicate that tmem2 is required for appropriate ECM deposition during skeletal muscle morphogenesis, as well as for deposition of the ECM that surrounds cardiomyocytes during heart tube formation. In addition, tmem2 promotes the glycosylation of α-dystroglycan within the DGC at the MTJ. Thus, our studies suggest that Tmem2 impacts cell-matrix interactions by influencing both the organization of the ECM and the post-translational modification of the CMAC.

Our previous studies indicated that zebrafish embryos lacking both maternal and zygotic supplies of tmem2 (MZtmem2) exhibit abnormal somite morphology, whereas embryos lacking only maternal supplies of tmem2 (Mtmem2) are indistinguishable from wild-type (Fig. 1A,D) (Totong et al., 2011). Notably, instead of the chevron-shaped somites seen in Mtmem2 embryos, MZtmem2 mutants display U-shaped somites (Fig. 1B,E). Formation of chevron-shaped somites requires Hedgehog signaling from the notochord (Barresi et al., 2000; Blagden et al., 1997); however, the morphology, integrity and differentiation of the MZtmem2 notochord appear relatively normal (Fig. 1C,F; Fig. S1B,D). Moreover, the MZtmem2 somite shape does not seem to result from defective Hedgehog signaling, since ptc1 expression appears to be intact in MZtmem2 mutants (Fig. S1A,C).

We investigated whether defects in muscle fiber morphogenesis could underlie the aberrant somite shape in MZtmem2 mutants. MZtmem2 embryos exhibit a normal number of somites (Fig. 1A,D) and have no apparent defects in initial somite boundary formation (Fig. S2A,B). However, muscle fiber attachment defects are prevalent in MZtmem2 mutants (Fig. 1G,H). Both fast and slow fibers show detachment from the MTJ (Fig. 1H; Fig. S3); in addition, some muscle fibers aberrantly cross the MTJ (Fig. 1H). Fiber detachment becomes more widespread as development proceeds (Fig. 1J,K; Fig. S4A,B), indicating failure to properly maintain attachments. Consistent with this, although zygotic tmem2 (Ztmem2) mutants exhibit normal somite morphology at early stages, defects in somite shape and muscle fiber integrity emerge in some Ztmem2 mutants over time (Fig. S4C-G), presumably as maternal supplies of tmem2 are depleted. Together, these results provide the first demonstration that tmem2 plays an important role in preserving muscle fiber attachment to the MTJ.

Establishment and maintenance of muscle fiber attachment at the MTJ require successful interactions with the ECM molecules that compose the basement membrane (Goody et al., 2015; Snow and Henry, 2009). Moreover, the MZtmem2 phenotype shares some characteristics with the phenotypes of laminin-deficient and fibronectin-deficient embryos, including the presence of fibers that cross the MTJ (Snow et al., 2008a,,b), prompting us to investigate the ECM in MZtmem2 mutants. Instead of the normally concentrated deposition of laminin at the MTJ in Mtmem2 embryos (Fig. 2A,C), we observed diminished and poorly organized laminin in MZtmem2 mutants (Fig. 2B,D), particularly in locations where fibers were detached (Fig. 2D). In contrast, fibronectin deposition appears relatively robust, albeit somewhat disorganized, in MZtmem2 mutants (Fig. S2A-F; Fig. 2E,F). During the usual progression of muscle morphogenesis (Snow and Henry, 2009; Jenkins et al., 2016), fibronectin levels degrade at the MTJ over time (Fig. 2E,G), in conjunction with accumulation of laminin (Fig. 2A,C). However, in MZtmem2 mutants, fiber attachment defects are accompanied by aberrantly increased fibronectin localization (Fig. 2F,H). This may represent a secondary consequence of laminin deficiency, since organized laminin has been shown to play an indirect role in facilitating fibronectin degradation at the MTJ (Jenkins et al., 2016); alternatively, increased fibronectin could be a secondary response to muscle fiber detachment, akin to the increased fibronectin fibrillogenesis seen in association with some myopathies (Hori et al., 2011; Rampoldi et al., 1986; Zacharias et al., 2011).

The deficient and disorganized ECM at the MZtmem2 MTJ made us wonder whether ECM defects could account for other aspects of the MZtmem2 mutant phenotype. MZtmem2 mutants exhibit cardia bifida, reflecting an early failure of cardiac morphogenesis (Totong et al., 2011). In wild-type embryos, bilateral populations of cardiomyocytes move toward the midline, where they meet and merge to assemble the heart tube through a process called cardiac fusion. MZtmem2 mutants fail to execute cardiac fusion and instead display two separated groups of cardiomyocytes in bilateral positions (Fig. 3F,G) (Totong et al., 2011). The composition of the basement membrane has a potent influence on cardiac fusion: either diminished or excessive ECM deposition can inhibit cardiomyocyte movement (Arrington and Yost, 2009; Garavito-Aguilar et al., 2010; Trinh and Stainier, 2004). Interestingly, the ECM adjacent to the MZtmem2 myocardium exhibits irregular and disorganized deposition of both laminin and fibronectin (Fig. 3A-E), which could account for the failure of cardiac fusion in MZtmem2 mutants. Thus, our data suggest that Tmem2 regulates both cardiac and skeletal muscle morphogenesis via modulation of the ECM.

Since the biochemical function of Tmem2 is currently unknown, it remains unclear whether this protein could exert its influence on the basement membrane through direct interaction with ECM components. To evaluate whether the Tmem2 ectodomain is sufficient to execute its functions, we replaced the transmembrane and cytoplasmic domains of Tmem2 with a signal peptide and tested whether this modified version of Tmem2 can rescue the MZtmem2 mutant phenotype. Injection of wild-type tmem2 mRNA into MZtmem2 mutants can rescue both muscle fiber attachment (Fig. 1I,L; Table S1) and cardiac fusion (Fig. 3F-K; Table S2). Similarly, we found that the Tmem2 ectodomain can also ameliorate both of these features of the MZtmem2 phenotype, although less efficiently than full-length Tmem2 (Fig. 1L; Tables S1 and S2). Therefore, the Tmem2 ectodomain can mediate at least some of the molecular functions of Tmem2, consistent with a model in which Tmem2 functions within the extracellular environment.

Our results suggest that the muscle fiber detachments in MZtmem2 mutants could be a direct consequence of faulty ECM organization. Since fiber attachment also relies upon effective CMAC assembly (Goody et al., 2010, 2015; Jackson and Ingham, 2013), we investigated whether the MTJ defects in MZtmem2 mutants are restricted to the basement membrane or are also reflected in the localization of CMAC components. Examination of three components of the DGC – the scaffolding protein paxillin, a phosphorylated form of focal adhesion kinase (pFAK), and the core complex component β-dystroglycan (βDG) – demonstrated that each was localized to the MTJ in MZtmem2 mutants (Fig. 4A-F). However, the distribution of each component was affected: paxillin was not properly concentrated (Fig. 4A,B), pFAK levels appeared to be reduced (Fig. 4C,D) and some gaps in βDG localization were observed (Fig. 4E,F). These aberrations could reflect ineffective CMAC assembly as a result of poor ECM engagement, or they could represent CMAC displacements that are secondary to fiber detachment (Bassett et al., 2003; Jacoby et al., 2009). Together, these observations suggest that recruitment of CMAC components to the MTJ does not require Tmem2, but that Tmem2 influences CMAC organization and integrity.

Our analysis of dystroglycan localization at the MZtmem2 MTJ also revealed a significant defect in the glycosylation of α-dystroglycan (αDG) (Fig. 4E,F). Dystroglycan is post-translationally cleaved into two subunits, αDG and βDG (Moore and Winder, 2012). αDG functions as a laminin receptor and its affinity for laminin depends upon its proper glycosylation (Sciandra et al., 2013). Strikingly, the glycosylated form of αDG is barely detectable at the MZtmem2 MTJ, even though βDG localization is robust (Fig. 4F,H, Fig. S5C, Table S3). The influence of Tmem2 on αDG glycosylation may require its transmembrane and/or cytoplasmic domains: whereas full-length Tmem2 can rescue glycosylation in MZtmem2 mutants, the Tmem2 ectodomain cannot (Fig. 4G,H, Fig. S5C, Table S3). Thus, in addition to its effects on ECM organization, Tmem2 promotes αDG glycosylation and, presumably, DGC activity, and this function of Tmem2 may employ a mechanism that is distinct from its other roles.

Together, our data establish Tmem2 as a previously unappreciated player in the cell-matrix interactions that control muscle morphogenesis. Tmem2 influences two distinct elements that enforce muscle fiber attachment: ECM deposition and CMAC composition. Since reduced laminin deposition interferes with fiber attachment (Goody et al., 2010; Hall et al., 2007; Jacoby et al., 2009; Snow et al., 2008b), it is likely that the ECM disruption in MZtmem2 mutants contributes to their muscle defects. The onset of fiber detachment in MZtmem2 mutants corresponds to the timeframe when laminin enrichment normally begins at the somite boundary (Crawford et al., 2003). Furthermore, ECM disorganization could explain the cardia bifida in MZtmem2 mutants (Arrington and Yost, 2009; Garavito-Aguilar et al., 2010; Trinh and Stainier, 2004). In addition, since DGC glycosylation promotes its engagement of the ECM (Sciandra et al., 2013), hypoglycosylation of αDG could also contribute to the fiber detachments in MZtmem2 mutants, as seen in embryos with reduced glycosyltransferase activity (Kawahara et al., 2010; Lin et al., 2011).

Do the ECM and CMAC features of the MZtmem2 phenotype represent two separate functions of Tmem2, or are these roles of Tmem2 inter-related? Although composition of the ECM is not likely to have a direct impact on αDG glycosylation, prior studies have found that laminin organization can be influenced by DGC glycosylation state (Kanagawa et al., 2005; Michele et al., 2002). Alternatively, Tmem2 could influence the ECM and DGC through two independent mechanisms. In this regard, it is intriguing that the Tmem2 ectodomain can fulfill some, but not all, aspects of Tmem2 function: although the ectodomain can improve fiber attachment in MZtmem2 mutants, it seems less effective than full-length Tmem2 and it cannot rescue αDG glycosylation. Thus, our data suggest that Tmem2 can function in the extracellular environment, consistent with our prior finding that myocardial expression of tmem2 can non-autonomously rescue MZtmem2 endocardial phenotypes (Totong et al., 2011). At the same time, our results suggest that functions of Tmem2 at distinct subcellular locations are relevant to its influence on post-translational modification of αDG. We therefore favor a model in which independent activities of Tmem2, affecting ECM organization and αDG glycosylation, collaborate to enforce muscle fiber attachment.

The influence of Tmem2 on muscle fiber attachment suggests an interesting link to the etiology of muscular dystrophy. In particular, Tmem2 may be relevant to the set of congenital muscular dystrophies known as dystroglycanopathies, which feature aberrant glycosylation of αDG (Muntoni et al., 2008; Wells, 2013). Mutations in 18 genes have been shown to cause dystroglycanopathies and several of these genes encode characterized or putative glycosyltransferases (Bouchet-Séraphin et al., 2015; Godfrey et al., 2011). However, as many as half of the dystroglycanopathy patients examined do not present mutations in known genes, and the process of post-translational modification of the DGC is not fully understood. Future elucidation of the molecular mechanisms of Tmem2 function is likely to provide valuable perspective on its relationship to dystroglycanopathy, as well as further insight into how ECM organization and CMAC composition both contribute to the stability of cell-matrix interactions during muscle development.

To obtain MZtmem2 mutant embryos, we used germline replacement to generate chimeric female fish with a tmem2^sk38 mutant germline and we bred these females to male tmem2 heterozygotes, as previously described (Totong et al., 2011). MZtmem2 mutants were distinguished from their Mtmem2 siblings by morphological criteria and PCR genotyping (Totong et al., 2011). All zebrafish work followed protocols approved by the University of California, San Diego Institutional Animal Care and Use Committee (IACUC).

Whole-mount immunofluorescence was performed as previously described (Goody et al., 2012), using Rhodamine Phalloidin (Invitrogen, R415) and antibodies listed in Table S4. For cryosections, embryos were fixed overnight in 4% paraformaldehyde at 4°C, followed by cryoprotection, mounting, sectioning, staining, and treatment with SlowFade Gold with DAPI (Invitrogen), as described previously (Garavito-Aguilar et al., 2010).

In situ hybridization for ptc1 (ZDB-GENE-980526-196), ehh (ZDB-GENE-980526-135) and myl7 (ZDB-GENE-991019-3) was performed as previously described (Yelon et al., 1999).

Embryos were injected at the one-cell stage with 200 pg mRNA encoding either full-length Tmem2 (Totong et al., 2011) or a modified version of the Tmem2 ectodomain. In this fusion protein, we replaced the first 103 amino acids of Tmem2, corresponding to its cytoplasmic and transmembrane domains, with the first 23 amino acids of zebrafish Sonic Hedgehog (Ekker et al., 1995), which serve as a signal to target the ectodomain for secretion.

Fluorescent images are maximal intensity projections of confocal reconstructions, with the exception of the single optical slices shown in Fig. 3A-E. Z-stacks containing 120-140 slices (0.5 µm thick) were acquired with a 25× water objective on a Leica SP5 microscope and analyzed with Imaris software (Bitplane). Additional images were captured using Zeiss Axiozoom and Axioimager microscopes with a Zeiss AxioCam and processed using Zeiss AxioVision and Adobe Creative Suite.

We thank L. Pandolfo and K. Garske for expert zebrafish care, C. Henry for helpful input, and members of the Yelon lab for constructive discussions.