Triadin and CLIMP-63 form a link between triads and microtubules in muscle cells

ABSTRACT In skeletal muscle, the triad is a structure comprising a transverse (T)-tubule and sarcoplasmic reticulum (SR) cisternae. Triads constitute the basis of excitation–contraction coupling as the cradle of the Ca2+ release complex. We have shown previously that triadin, a member of this complex, has shaping properties on reticulum membrane and is indirectly involved in a link between triads and microtubules. We have identified here that CLIMP-63 (also known as CKAP4), as the partner of triadin, is responsible for this association of triads and microtubules. Triadin and CLIMP-63 interact through their respective luminal domains and the shaping properties of triadin depend on the capacity of CLIMP-63 to bind microtubules with its cytosolic portion. In skeletal muscle, CLIMP-63 is localized in the SR, including triads, and is associated with the Ca2+ release complex through its interaction with triadin. Knockout of triadin in muscles results in the delocalization of CLIMP-63 from triads, its dissociation from the Ca2+ release complex and a disorganization of the microtubule network. Our results suggest that the association of triadin and CLIMP-63 could be involved in the shaping of SR terminal cisternae and in the guidance of microtubules close to the triads. Summary: CLIMP-63 is associated with the Ca2+ release complex of skeletal muscles through its interaction with triadin, and is involved in the shaping of triads and in the guidance of microtubules.


INTRODUCTION
Skeletal muscle contraction is initiated by a burst of Ca 2+ out of its major store, the sarcoplasmic reticulum (SR). This rise in cytosolic Ca 2+ concentration is triggered by an action potential that propagates along the plasma membrane and reaches invaginations, the transverse tubules (T-tubules). T-tubules make close and periodic contacts with the SR to form peculiar membrane structures called triads (Flucher and Franzini-Armstrong, 1996;Wagenknecht et al., 2002). Triads, defined as two terminal cisternae of SR surrounding a T-tubule, are highly organized membrane structures, precisely distributed in the muscle fiber. Excitation-contraction coupling, that is, the transformation of the action potential into a massive Ca 2+ release, is performed at the triads by a macromolecular complex, the Ca 2+ release complex (CRC) (Franzini-Armstrong and Jorgensen, 1994). The main players of CRC are the dihydropyridine receptor (DHPR), a L-type Ca 2+ channel of the T-tubule membrane, and ryanodine receptor (RyR1), the SR Ca 2+ channel. Although these channels are anchored in two different membranes, they are associated (Block et al., 1988;Marty et al., 1994). Thus, the correct function of the CRC relies on the precise alignment of these two membranes, and on the formation and maintenance of triads.
A number of proteins associate with either the DHPR or RyR1 and regulate their activity. Among them is the triadin. Triadin is a multi-protein family, with four isoforms, denoted Trisk 95, Trisk 51, Trisk 49 and Trisk 32, which arise from the alternative splicing of a unique gene, the TRDN gene (Thevenon et al., 2003). Triadins are specifically expressed in striated muscles, Trisk 95 and Trisk 51 being the major skeletal muscle isoforms and Trisk 32 the major cardiac isoform (Marty, 2015). Triadins are type II transmembrane proteins that are completely identical in their cytosolic and transmembrane domains. They diverge because of the length of their luminal domain and the sequence of their C-terminal end. They are localized in the SR membrane, and are associated to RyR1 at the skeletal muscle triad, or to RyR2 at the cardiac dyad (Marty, 2015). Mutations in the TRDN gene have been implicated in severe human cardiac arrhythmia with sudden death (Roux-Buisson et al., 2012).
We have previously shown that expression of triadin Trisk 95 in COS-7 cells induces deformation of the endoplasmic reticulum (ER) and of the microtubule network, creating a link between ER and microtubules . However, this link is indirect, as we observed that Trisk 95 domains implicated in its shaping properties are luminal . Therefore, in ER membrane there must be another transmembrane protein, on the one hand able to bind Trisk 95 luminal domains and, on the other hand, cytoplasmic microtubules. Because of its ability to shape ER and SR membrane , triadin could be involved in the overall structuration of SR cisternae at the triad. In addition, triadin knockout leads to deformations of part of the triads and to a reduction in excitation-contraction coupling and muscle strength (Oddoux et al., 2009;Shen et al., 2007). Therefore, the ER and SR membrane-shaping properties of triadin might favor the RyR1-DHPR coupling.
In the present study, using co-immunoprecipitation coupled to mass-spectrometry-based proteomics, we identified cytoskeletonlinking membrane protein of 63 kDa (CLIMP-63; also known as CKAP4) as a Trisk 95 partner. CLIMP-63 was discovered in the early 1990s (Mundy and Warren, 1992;Schweizer et al., 1993). It is a non-glycosylated type II transmembrane protein expressed in all tissues of higher eukaryotes. It is localized in the ER membrane, and more specifically concentrated in ER sheets where it would have the role of a luminal spacer (Klopfenstein et al., 2001;Schweizer et al., 1995;Shibata et al., 2010). The cytosolic domain of CLIMP-63 binds to microtubules and therefore creates a link between the ER and the microtubule network (Klopfenstein et al., 1998;Vedrenne and Hauri, 2006). After a detailed characterization of the molecular interaction between Trisk 95 and CLIMP-63, we show that the protein pair Trisk-95-CLIMP-63 could be involved in the guidance of the microtubule network in the vicinity of the triads in muscle cells.

CLIMP-63 is a partner for the triadin isoform Trisk 95
To look for a protein able to mediate the link between Trisk 95 and the microtubules, we analyzed proteins associated with Trisk 95 using co-immunoprecipitation combined with mass-spectrometrybased proteomics. Trisk 95 coupled to GFP, or GFP alone as a control, were expressed in HEK293 cells and immunoprecipitated using anti-GFP antibodies. The semi-quantitative analysis of the proteins co-immunoprecipitated with Trisk-95-GFP and GFP allowed us to identify proteins enriched with Trisk 95 (Table S1). From these potential binding partners, we selected the ER proteins that could bind microtubules and focused our attention on CLIMP-63. CLIMP-63 is an ER membrane protein whose main function is to shape and anchor the ER onto the microtubule network. CLIMP-63 binds microtubules through its N-terminal cytosolic domain and exhibits a long intra-luminal coiled-coil domain (Klopfenstein et al., 2001;Vedrenne and Hauri, 2006). It is a ubiquitous protein expressed in cell lines and muscle cells (Razzaq et al., 2003;Schweizer et al., 1995). Furthermore, CLIMP-63 has been described to induce ER rope-like structures (RLSs) when overexpressed in proliferating cells (Klopfenstein et al., 1998;Vedrenne and Hauri, 2006), which is the same phenotype as that observed upon Trisk 95 overexpression . Therefore, we raised the hypothesis that CLIMP-63 could be the missing link between triadin and the microtubule network.
To validate the association between CLIMP-63 and Trisk 95, Trisk 95 without the GFP moiety was expressed in HEK293 cells, and an immunoprecipitation using an anti-CLIMP-63 antibody was performed, which demonstrated the specific co-immunoprecipitation of Trisk 95 with CLIMP-63 (Fig. 1A).
We have shown previously that Trisk 95, when expressed in COS-7 cells, is located in the ER and can induce ER membrane deformation and microtubule network disorganization in 16% of transfected cells . We wondered whether CLIMP-63 could colocalize in the ER with Trisk 95 and be associated with the Trisk 95-induced phenotypes. To assess this possibility, COS-7 cells were transfected with Trisk 95 and immunolabeled for triadin, CLIMP-63 and tubulin. In control non-transfected cells, endogenous CLIMP-63 was mainly localized in ER sheets, and microtubules exhibited a normal radial pattern (Fig. 1B, cell 1). In every cell transfected with Trisk 95, CLIMP-63 colocalized with Trisk 95 in the ER (Fig. S1A). In the Trisk 95transfected cell shown in Fig. 1B (cell 2), the ER shape was altered, with the formation of RLSs, and CLIMP-63 colocalized with Trisk 95 in these RLSs. In addition, microtubule arrays were disorganized and co-aligned with the ER RLSs (Fig. 1B, cell 2). Hence, Trisk 95 expression could induce CLIMP-63 and microtubule redistribution, suggesting a connection between the three.
Thus, co-immunoprecipitation and colocalization experiments show that CLIMP-63 and Trisk 95 are associated in cell lines. We next wondered which regions of both proteins were responsible for this interaction. We have shown previously that it is the luminal domain of Trisk 95, and more specifically the so-called RLS domain comprising the coiled-coil region 306-341 with the two cysteine residues C270 and C649, which is involved in the formation of RLSs . We therefore hypothesized that the luminal domains of Trisk 95 and CLIMP-63 would be involved in the interaction. We investigated this by using different triadin mutants (Fig. 2C), either as previously described  or generated for this study: (1) a mutant without the cytosolic N-terminal domain (Δcyto-T95), (2) a mutant without the cytosolic and transmembrane domains but with ER targeting and retention sequences (T95lumen), (3) a mutant deleted of most the luminal domain (T95-mini), and finally (4) a mutant in which the specific sequences responsible for the RLS phenotype are removed (T95-ΔRLS). Wild-type CLIMP-63 (CLIMP, Fig. 2C) was co-transfected with each of the above deletion mutants in COS-7 cells. As previously described (Klopfenstein et al., 1998), CLIMP-63 overexpression by itself induced ER deformation in 78% of the transfected cells (Figs S1B and S2). When CLIMP-63 was co-transfected with Trisk 95 (T95), Δcyto-T95 or T95-lumen, the cells exhibiting RLSs showed a perfect colocalization between CLIMP-63 and Trisk 95 proteins in these ER structures ( Fig. 2Aa-Ac,B). In contrast, cells co-transfected with CLIMP-63 and either T95-mini or T95-ΔRLS exhibited RLSs due to CLIMP-63 overexpression, but the Trisk 95 mutants were not localized in these RLSs (Fig. 2Ad,Ae,B). The ER localization of T95-mini or T95-ΔRLS mutants was verified in our previous study . Immunoprecipitations showed that T95 and Δcyto-T95 could associate with CLIMP-63, whereas T95-ΔRLS or T95-mini could not (Fig. 2D,E). These experiments demonstrate that the luminal domain of Trisk 95 is responsible for association with CLIMP-63.
Triadin is a family with different members, differing only in their luminal domains (Fig. S3A). To see whether other triadin isoforms associate with CLIMP-63, HEK293 cells were transduced with different adenoviral vectors encoding the different isoforms and immunoprecipitations were performed with an antibody against the common N-terminal end of triadin. In each case, CLIMP-63 was co-immunoprecipitated, indicating that all the triadin isoforms are able to interact with CLIMP-63 (Fig. S3B).
To study domains of CLIMP-63 involved in the interaction, two mutant forms of CLIMP-63 were used (Fig. 3A): one with the luminal part replaced by GFP (CLIMP-Δlumen) and one with the cytoplasmic   When expressed in cell lines, both CLIMP-63 and Trisk 95 induce ER deformations, the RLSs, which are concomitant with microtubule array disorganization (see Fig. 1 and Fig. S1). Because it seems from the above results that Trisk 95 binds to microtubules through CLIMP-63, we hypothesized that the capacity of Trisk 95 to form RLSs was due to its ability to bind and recruit CLIMP-63. To test this, endogenous CLIMP-63 was knocked down by RNA interference (RNAi) in COS-7 cells before transfection of the cells with Trisk 95, and the extent of Trisk 95-induced RLSs was evaluated. Unfortunately, the knockdown of CLIMP-63 was not complete ( Fig. S4A-C) and Trisk 95 was still able to induce RLSs in cells (Fig. S4D,E). A similar situation has already been described in neurons where the knockdown of CLIMP-63 is incomplete (Cui-Wang et al., 2012).
We then used a mutant form of CLIMP-63, referred to as CLIMP-MT(−) because of its inability to bind to microtubules (Vedrenne et al., 2005). This mutant is a phospho-mimetic form of CLIMP-63, with four serine residues of the cytosolic domain mutated to glutamic acid residues (Fig. 4A). As a control, a phospho-deficient mutant of CLIMP-63 was used. This mutant, referred to as CLIMP-MT(+), is always bound to microtubules and has three phospho-sites mutated to alanine residues ( Fig. 4A) (Vedrenne et al., 2005). Trisk 95 was co-expressed in COS-7 cells with either CLIMP-MT(−) or CLIMP-MT(+) and the formation of RLSs was quantified. All of the cells co-transfected with CLIMP-MT(+) and Trisk 95 exhibited RLSs (Fig. 4B, upper panel, 4C). By contrast, cells co-expressing CLIMP-MT(−) and Trisk 95 never exhibited RLSs (Fig. 4B, lower panel, 4C). Rather, these cells presented a fenestrated ER that did not resemble RLSs, and microtubule arrays in these cells appeared normal. This absence of RLSs was not due to the incapacity of Trisk 95 to bind to the CLIMP-MT(−) mutant: indeed, CLIMP-MT(−) was still able to co-immunoprecipitate Trisk 95 (Fig. 4D, lane 5). Hence, CLIMP-MT(−), being unable to bind microtubules, acted as a dominant-negative mutant and abolished the capacity of Trisk 95 to induce RLSs. These experiments show that the Trisk 95-CLIMP-63 association is not dependent on the CLIMP-63-microtubule interaction but that the ER-remodeling induced by Trisk 95 is dependent on the ability of CLIMP-63 to bind microtubules.
In muscle fibers, CLIMP-63 is a partner for Trisk 95 at triads After the detailed characterization of the interaction between Trisk 95 and CLIMP-63 in cell lines, we examined the situation in skeletal muscle. Immunoprecipitation was performed on mouse muscle homogenates (SK, Fig. 5A, lane 1) with an antibody against Trisk 95. As shown in Fig. 5A, lane 2, CLIMP-63 was coimmunoprecipitated by Trisk 95 in these muscle extracts. We further studied the localization of CLIMP-63 using immunofluorescence labeling of CLIMP-63 and Trisk 95 on isolated mouse extensor digitorum longus (EDL) muscle fibers (Fig. 5B). Triads labeled by the anti-Trisk 95 antibody showed characteristic double rows of dots (Fig. 5B, red). CLIMP-63 also appeared as dots with a striated pattern (Fig. 5B, green). A partial overlap between the stainings of Trisk 95 and CLIMP-63 could be detected on the merged image as yellow dots (Fig. 5B). A colocalization analysis was performed with ImageJ software: the binary mask images represent the colocalized points between the green and the red stainings (Fig. 5B, right panel) and quantification from these binary images showed that 35.7±1.2% (mean±s.e.m.) of CLIMP-63 colocalized with Trisk 95 in WT muscles. Thus CLIMP-63, as a reticulum protein present in the ER and SR, partially colocalizes with the triad-specific protein Trisk 95. In this SR subdomain, the triad, these proteins might interact as shown by co-immunoprecipitation experiments.
CLIMP-63 localization at triads is altered in triadin KO muscles CLIMP-63 might be specifically localized at the triad through its interaction with Trisk 95. Therefore, the localization of CLIMP-63 was studied in EDL fibers of triadin-knockout (KO) mice, devoid of all isoforms of triadin. Triads were stained using an anti-RyR1 antibody. In triadin KO EDLs, RyR1 staining (Fig. 6A, red) revealed double rows of dots and was unmodified compared to wild type (WT) EDLs, as observed previously (Oddoux et al., 2009). In contrast, CLIMP-63 staining (Fig. 6A, green) was less striated in KO fibers (Fig. 6A, lower panel), less concentrated at the I-band and less well colocalized with triads (see line scans in Fig. 6B). Quantification of the relative amount of CLIMP-63 colocalized with RyR1 in WT and triadin KO muscles showed a significant reduction from 37.5±1.3% in WT to 13.8±0.8% (mean±s.e.m.) in KO muscles (Fig. 6C, quantified from binary mask images as shown in Fig. 6A). Quantitative western blots performed on muscle homogenates from WT and triadin KO mice demonstrated that the total amount of CLIMP-63 was not altered by triadin depletion (Fig. 6D). Rather, in triadin KO muscles, CLIMP-63 seems to redistribute in the whole ER-SR. Immunoprecipitations were performed using anti-RyR1 antibody on WT and triadin KO muscle homogenates. Whereas in WT muscles, immunoprecipitation of RyR1 resulted in the coimmunoprecipitation of both Trisk 95 and CLIMP-63, both proteins were absent in the immunoprecipitation made from triadin KO muscles (Fig. 6E). Therefore, in the absence of triadin, CLIMP-63 localization at triads is reduced and CLIMP-63 is not associated with the Ca 2+ release complex, highlighting the triadin-CLIMP-63 link in muscles.
The organization of transversal microtubules is disrupted both in triadin KO muscles and in muscles overexpressing CLIMP-MT(−) We have previously shown that in skeletal muscle some microtubule bundles come into close proximity to triads . Because the protein pair triadin-CLIMP-63 is most probably involved in the link between triads and microtubules, we compared the organization of microtubules in isolated EDL muscle fibers from WT and triadin KO mice. In WT EDLs, the microtubule network appears as an orthogonal grid with longitudinal and transversal microtubule bundles, as previously described (Oddoux et al., 2013;Ralston et al., 1999Ralston et al., , 2001. Although the organization of transversal microtubules is not as precise as that of the triads, these microtubules presented a striated pattern with the same 2-µm step as triads and they were usually localized within the double rows of RyR dots (Fig. 7A, WT panel). These transversal microtubules apparently make periodic contacts with triads, as suggested by a 34.7±1.4% (mean±s.e.m.) colocalization between tubulin and RyR1 stainings (quantified from binary mask images in Fig. 7A, WT panel). In triadin KO EDLs, transversal microtubules appeared less organized and less bundled, and often single oblique microtubules escaped from the double row of triads (Fig. 7A, KO panel). The percentage of tubulin staining colocalized with RyR1 was significantly reduced from 34.7±1.4% in WT animals to 26.4±1.4% in triadin KO mice (quantified from binary mask images in Fig. 7A). To assess the overall microtubule disorganization in triadin KO EDLs, a blind test (see Materials and Methods) was performed by two different persons to sort randomized images from WT and KO fibers and classify them in three categories: organized (MT org +), less organized (MT org +/−) and disorganized microtubules (MT org −, representative images in Fig. 7B). Results showed that WT fibers were mostly sorted out as 'organized' (65%) while KO fibers were evenly distributed in the three categories, the main pattern (38%) being 'MT org +/−' (Fig. 7B). Hence, transversal microtubules were less organized and less bundled between rows of triads in triadin KO muscles. The absence of the triadin-CLIMP-63 pair in triadin KO muscles seems to induce a reduction in the co-alignment between triads and microtubules.
In order to examine further the influence of CLIMP-63 on microtubule organization in muscle cells, the dominant negative CLIMP-MT(−) mutant was expressed in flexor digitorum brevis (FDB) muscles. GFP alone (as a control) or GFP-CLIMP-MT(−) was electroporated in FDBs of WT mice, and microtubules visualized by immunofluorescent staining on isolated fibers. The microtubule network in FDBs is slightly different from that in EDLs, being less bundled and less regularly organized in orthogonal grid. Nevertheless, FDB transversal microtubules are usually localized within the double rows of RyR dots as in EDL fibers (Fig. 8A, control panel). In FDBs expressing CLIMP-MT(−), a disorganization of the microtubule network could be observed, with a reduced localization of transverse microtubules between RyR dots and with many single microtubules running obliquely or longitudinally [ Fig. 8A, CLIMP-MT(−) panel]. The overall microtubule organization was tested as in experiments on EDLs, using a blind test. We found that control FDB fibers expressing GFP were mostly sorted out as 'organized' (MT org +, 47%) whereas fibers expressing GFP-CLIMP-MT(−) were mainly classified as 'disorganized' (MT org −, 45%, Fig. 8B). Therefore, as observed for triadin KO muscles, expression of the microtubule-binding deficient mutant CLIMP-MT(−) alters muscle microtubule arrays.

DISCUSSION
In this study, we identify CLIMP-63 as a partner of triadin in muscle cells, and show that it is involved in the link between triads and the microtubule network. This protein pair, triadin and CLIMP- 63, might play an important role in the shaping of triads. The ER-SR in muscle cells forms a highly structured 3D network surrounding the myofibrils and is organized in subdomains, the longitudinal ER-SR and the junctional SR (the terminal cisternae) (Franzini-Armstrong, 1980;Ogata and Yamasaki, 1997). The terminal cisternae in association with the T-tubule form the triads, which are the sites of Ca 2+ release and play a major role in excitation-contraction coupling. The triads are exclusively localized at the interface of A and I bands in mammalian skeletal muscles (Ogata and Yamasaki, 1997). The mechanisms leading to the formation and maintenance of these subdomains are unknown, as well as the mechanisms leading to the exclusive localization of proteins of the CRC at the triads.
CLIMP-63 is an ER transmembrane protein that has been mainly studied in proliferating mono-nucleated cells (Vedrenne and Hauri, 2006). CLIMP-63 is concentrated in ER sheets where it appears to have a role in the shaping of these ER subdomains (Klopfenstein et al., 2001;Shibata et al., 2010). Our study is the first description of CLIMP-63 in mature muscle fibers, where CLIMP-63 staining appears as bright dots localized in the I-bands, resembling the staining described in muscles for rough ER markers like BiP or PDI (Kaisto and Metsikko, 2003). However, in addition, a portion of the dots of CLIMP-63 (∼40%) colocalize with Trisk 95 at the triads where the ryanodine receptor RyR1 is also present. This partial colocalization of CLIMP-63 and triadin reflects the fact that CLIMP-63 is a protein present in the whole ER-SR of muscle cells whereas Trisk 95 is specific to the junctional SR at the triads. CLIMP-63 is probably able to diffuse more freely than Trisk 95 in the different ER-SR subdomains, and it might be maintained in the junctional SR by Trisk 95. Therefore, in the absence of Trisk 95, CLIMP-63 is partly delocalized (Fig. 6). Based on the observations in COS-7 cells, we can speculate on the functional impact of the triadin-CLIMP-63 interaction at the triad. We show in COS-7 cells that triadin and CLIMP-63 interact through their luminal domain, most probably through their respective luminal coiled-coil domains.
Besides, both proteins are known to multimerize in the ER and SR membrane to form oligomers Schweizer et al., 1994). Thus, CLIMP-63 and Trisk 95 could form hetero-oligomers in the SR cisternae membrane and these hetero-oligomers could shape the SR membrane. Indeed, in WT muscles, the SR terminal cisternae in triads appear as small rounded bags in electron microscopy images. In triadin KO mouse muscles, the SR terminal cisternae are flattened, an observation which has been attributed to calsequestrin depletion, but could also reflect the depletion of both triadin and CLIMP-63 (Oddoux et al., 2009). Along this line, Rapoport and co-workers describe CLIMP-63 as an ER luminal spacer and have found that CLIMP-63 depletion by RNAi in COS-7 cells results in a collapse of the ER sheet luminal width from 55±9 nm to 28±4 nm (mean±s.e.m.) (Shibata et al., 2010). Consistent with this, in triadin KO muscle, the luminal size Fig. 7. Partial disorganization of microtubule network in triadin KO muscles. (A) Isolated EDL muscle fibers from WT or triadin KO mice were double-stained with antibodies against tubulin to label microtubules (green) and RyR1 to label triads (red). Each image represents one single confocal plane. Scale bars: 4 µm. Merged panels and zooms are shown. The binary images (masks) represent in black the microtubule staining area that colocalizes with the RyR staining in the zoomed images. Using such image binarization, the tubulin staining colocalized with RyR1 was found to be significantly reduced from 34.7±1.4% (WT, n=16 quantified EDL fibers) to 26.4±1.4% (KO, n=17). P<0.0001, Mann-Whitney two-tailed test, values are mean±s.e.m. (B) Images of WT and KO EDL fibers were classified blindly in three categories: organized (MT Org +), less organized (MT Org +/−) and disorganized microtubules (MT Org −). Representative images are shown in the lower panels, arrowheads point to disorganized microtubules. The bar graph represents the percentages for each category (n=54 and 83 images examined, for WT and KO conditions, respectively).
of the SR terminal cisternae is reduced from 68±3 nm to 26±1 nm (mean±s.e.m.) (Oddoux et al., 2009). Thus, it is likely that the formation of triadin-CLIMP-63 hetero-oligomers does not compete with the CLIMP-63 properties as a luminal spacer, but rather maintains the SR cisternae structure.
CLIMP-63 is also a microtubule-binding protein. Its microtubulebinding domain has been mapped to amino acids 62-101 in the cytosolic N-terminal part of the protein (Klopfenstein et al., 1998;Vedrenne et al., 2005). When overexpressed, CLIMP-63 has a strong effect on the microtubule network, inducing bundles of stabilized microtubules (Klopfenstein et al., 1998;Vedrenne and Hauri, 2006;Vedrenne et al., 2005). In our work, we observed that the ability of Trisk 95 to cause a reorganization of the microtubule network in COS-7 cells is mediated by CLIMP-63. Indeed, the pseudo-phosphorylated CLIMP-MT(−) mutant, which cannot bind microtubules (Vedrenne et al., 2005), is still able to bind Trisk 95 but abolishes the effects of Trisk 95 on the microtubule network. It is likely that Trisk 95 expressed in COS-7 cells oligomerizes with endogenous CLIMP-63 and induces its aggregation or re-localization. These Trisk 95-CLIMP-63 aggregates would induce the formation of ER RLSs and bundled arrays of microtubules, a mechanism already suggested by the group of Hauri (Vedrenne and Hauri, 2006).
The roles of microtubules in skeletal muscle have not been extensively studied. In muscle cells, microtubules are organized in bundles exhibiting a transversal and longitudinal orientation and forming a grid-like network (Oddoux et al., 2013;Ralston et al., 1999Ralston et al., , 2001. This network is dynamic, with microtubules polymerizing and depolymerizing continuously along a bundled framework (Oddoux et al., 2013). Microtubule bundles in muscles are involved in the stretch-activated production of reactive oxygen species (ROS), and alteration in this microtubule-dependent ROS production is a major defect in Duchenne muscular dystrophy (Kerr et al., 2015;Khairallah et al., 2012). Dynamic microtubule bundles might also be necessary for the morphogenesis of organelle subdomains such as the ER or SR subdomains (Cui-Wang et al., 2012;Vedrenne and Hauri, 2006). In neurons, local zones of increased ER complexity are observed underneath dendritic spines, which could favor the growth of spines as well as the local synthesis and export of synaptic-specific proteins. The shaping of these complex ER zones appears to rely on CLIMP-63 and its interconnection with dendritic microtubule bundles (Cui-Wang et al., 2012). In muscle cells, our work shows that microtubule arrays are partly disorganized in the absence of triadin or after overexpression of the microtubule-binding-deficient CLIMP-MT (−) mutant. Hence, triadin-CLIMP-63 hetero-oligomers at triads could favor the organization of at least some of the transversal microtubules in muscle cells, and could guide these microtubules close to the terminal SR cisternae. This precise microtubule localization could be of major importance for the targeting of CRC proteins to the triads as has been shown for the targeting of the DHPR to cardiac T-tubules (Hong et al., 2010). The triadin-CLIMP-63 protein pair is most probably not the only microtubule organizer in muscle fibers, and the creation of orthogonal grid of microtubules would be expected to require numerous other proteins. For example, dystrophin as well as the complex ankyrin-Bdynactin-4 have been proposed as being responsible for microtubule anchoring at costameres (Ayalon et al., 2008;Belanto et al., 2014;Prins et al., 2009).
In conclusion, absence of triadin observed in human patients (Roux-Buisson et al., 2012) and triadin KO mice (Oddoux et al., 2009) could have wider consequences than the dysregulation of RyR1 and alteration of Ca 2+ release: it could also affect the overall microtubule organization in muscle fibers.

Ethics statement
All procedures using animals were approved by the Institutional ethics committee and followed the guidelines of the National Research Council Guide for the care and use of laboratory animals.

Viruses and transduction
In this study, we used adenoviral-mediated gene transfer to express Trisk 95, Trisk 51, Trisk 49 or Trisk 32 in HEK293 cells. These viruses were engineered and produced either by the Gene Vector Production Network at Genethon III (Evry, France) or by the Atlantic Gene Therapies (Nantes, France). Lentiviral vectors carrying either the rat Trisk 95 cDNA, the human CLIMP-63 cDNA or the cDNAs of the mutants CLIMP-MT(+) or CLIMP-MT(−) were produced in the laboratory by a triple transfection of pWPXLd, psPAX2 and pCMV-VSV-G (Addgene, Cambridge, MA) into HEK293T cells, as described previously (Bovia et al., 2003;Zufferey et al., 1997). Titers of viruses in cell culture supernatants were determined as previously described (Nguyen et al., 2005). For all transduction experiments, cells were transduced by overnight incubation with the chosen virus at a multiplicity of infection of 40 for adenoviruses and 5 for lentiviruses. Control cells were infected with the control viruses (adenovirus DsRed or lentivirus-GFP).

Antibodies
The primary antibodies used in this study are presented in Table S2. Secondary antibodies used for immunofluorescence studies were coupled to Alexa Fluor 488 or Alexa Fluor 546 (Molecular Probes); or to Cy3 or Cy5 (Jackson ImmunoResearch Laboratories). Secondary antibodies used for western blotting were horseradish peroxidase (HRP)-coupled anti-rabbit-IgG polyclonal antibodies (Jackson ImmunoResearch Laboratories). For western blotting after immunoprecipitation, TrueBlot ® HRP-coupled antirabbit-IgG polyclonal antibodies were used (Rockland).

Proteomic analyses
Proteomic experiments were performed as previously described (Milbradt et al., 2014). Briefly, proteins were loaded on a 4-12% NuPAGE gel (Invitrogen) before R-250 Coomassie Blue (BioRad) staining and in-gel digestion using modified trypsin (Promega, sequencing grade). Resulting peptides were analyzed by online nanoliquid chromatography coupled to tandem mass spectrometry (nanoLC-MS/MS with Ultimate 3000 and LTQ-Orbitrap Velos pro, Thermo Scientific) using a 120-min gradient. Peptides and proteins were identified through concomitant searches against the Uniprot databank (September 2014 version, Homo sapiens taxonomy, 138,560 sequences) and a classical contaminants database (in-house, 260 sequences) and the corresponding reversed databases using Mascot (version 2.5). The IRMa software, version 1.31.1 (Dupierris et al., 2009), was used to filter the results: conservation of rank one peptides, peptide identification with a false discovery rate (FDR) <1% (as calculated on peptide scores by employing the reverse database strategy), and minimum of one specific peptide per identified protein group. Filtered results were uploaded into a relational mass spectrometry identification database before performing a compilation, grouping and comparison of the protein groups from the different samples using a in-house tool (hEIDI). Differential analysis of GFP and Trisk 95-GFP co-immunoprecipitation eluates was performed using extracted spectral counts and specific spectral counts (SSC). A protein was considered as a potential binding partner if it was identified with a minimum of 10 SSC and enriched at least fourfold in the Trisk-95-GFP sample compared to GFP one.
Cell culture and transfection COS-7 and HEK293 cells were cultured in Dulbecco's modified Eagle medium (DMEM) supplemented with 10% fetal bovine serum and 1% penicillin-streptomycin (Life Technologies). Transient transfections were carried out using Exgen 500 reagent (Euromedex), according to the manufacturer's protocol. After 28 h of transfection, cells were treated for either immunoprecipitation or immunofluorescence. For immunoprecipitation, cells grown in 60-mm diameter plates were collected by trypsinization, washed with PBS and stored at −20°C until used. For immunofluorescence, cells grown on coverslips were fixed for 6 min in methanol at −20°C.
Electroporation and preparation of muscle fibers EDL muscle fibers were manually dissociated as described in Prins et al. (2009). Electroporations of FDB muscles were performed according to DiFranco et al. (2009). 10 µl of a 2 mg/ml hyaluronidase (Sigma) solution was injected subcutaneously in each footpad of the anesthetized mice, followed by a 20 µl injection of the endotoxin-free pEGFP-N1 or pEGFP-CLIMP-MT(−) (3-5 mg/ml) plasmids. Electroporation pulses (20×20 ms at 100 V) were delivered with acupuncture needles and a BTX ECM830 electroporator. Five days after electroporation, FDB muscles were dissected, dissociated enzymatically for 45 min at 37°C in 1.5 mg/ml collagenase (Sigma) and subjected to mechanical dissociation.
Single EDL and FDB fibers were fixed for 45-60 min at room temperature in PBS containing 4% paraformaldehyde and permeabilized for 90 min in PBS with 1% Triton X-100 at 30°C before being assessed for immunofluorescence as described below.

Immunofluorescence microscopy
Cells and muscle fibers were saturated in PBS supplemented with 0.1% Triton X-100, 0.5% bovine serum albumin and 2% goat serum for 30 min. They were incubated with primary antibodies in PBS with 0.1% Tween 20 either at room temperature for 1 h (COS-7 cells) or at 4°C overnight (muscle fibers). After incubation for 1-3 h at room temperature with fluorescent secondary antibodies, coverslips were mounted on microscope slides with FluorSave™ reagent (Calbiochem). Images were captured on a Zeiss LSM710 confocal microscope or Leica SPE confocal microscope.

Immunoprecipitation, SDS-PAGE and western blotting
Mouse skeletal muscle homogenates or homogenized transfected HEK293 cells were solubilized for 30 min at 4°C in a buffer composed of 0.9 M NaCl, 1.6% CHAPS, 0.1% phospholipids (Avanti Polar Lipids), 20 mM Pipes ( pH 7.1), 200 µM PMSF and 1 mM DFP. Immunoprecipitations were then performed as described previously (Marty et al., 1994) with the chosen antibody and Dynabeads ® Protein G (Life Technologies). All Triadin or CLIMP-63 constructs fused with GFP were immunoprecipitated from transfected HEK293 cell extracts using GFP-Trap ® (Chromotek), according to the manufacturer's protocol. The immunoprecipitated proteins were then analyzed by SDS-PAGE (5-15% gradient gels, Bio-Rad Laboratories) and western blotting performed with antibodies described above and with ECL substrate reagent (Thermo Scientific). Acquisitions were done using a ChemiDoc XRS apparatus (Bio-Rad) and quantification was performed using Quantity One software.

Colocalization analyses and quantification of microtubule disorganization on immuno-fluorescence images of EDL muscle cells
To quantify CLIMP-63 colocalization with Trisk 95 or RyR1, or microtubule colocalization with RyR1, a binary mask corresponding to the overlap between the green and the red stainings was generated with a inhouse ImageJ macro (available from the corresponding author on request). This mask corresponded to the surface of colocalization on one confocal plane. A ratio was calculated between the surface of colocalization and the total green surface (corresponding either to CLIMP-63 or microtubules). Results were given as percentages and averaged for several EDL muscle fibers from at least three independent experiments.
To quantify microtubule disorganization, A.O. and M.S. first captured images with the confocal microscope, either on EDLs from WT and KO mice (first experiment) or on FDBs from pEGFP-CLIMP-MT(−)-or EGFPelectroporated mice (second experiment). For each experiment, the names of images were then randomly replaced by numbers, thanks to a 'name_randomizer' macro in ImageJ. Classification of images in three categories (MT org +, MT org +/−, MT org −) was performed blindly and independently by A.F.-L. and I.M.

Statistics
All statistical analyses were performed using Prism 6.0 (GraphPad Software, La Jolla, USA).