STIM2 regulates both intracellular Ca2+ distribution and Ca2+ movement in skeletal myotubes

Stromal interaction molecule 1 (STIM1) along with Orai1 mediates extracellular Ca2+ entry into the cytosol through a store-operated Ca2+ entry (SOCE) mechanism in various tissues including skeletal muscle. However, the role(s) of STIM2, a homolog of STIM1, in skeletal muscle has not been well addressed. The present study, first, was focused on searching for STIM2-binding proteins from among proteins mediating skeletal muscle functions. This study used a binding assay, quadrupole time-of-flight mass spectrometry, and co-immunoprecipitation assay with bona-fide STIM2- and SERCA1a-expressing rabbit skeletal muscle. The region for amino acids from 453 to 729 of STIM2 binds to sarcoplasmic/endoplasmic reticulum Ca2+-ATPase 1a (SERCA1a). Next, oxalate-supported 45Ca2+-uptake experiments and various single-myotube Ca2+ imaging experiments using STIM2-knockdown mouse primary skeletal myotubes have suggested that STIM2 attenuates SERCA1a activity during skeletal muscle contraction, which contributes to the intracellular Ca2+ distribution between the cytosol and the SR at rest. In addition, STIM2 regulates Ca2+ movement through RyR1 during skeletal muscle contraction as well as SOCE. Therefore, via regulation of SERCA1a activity, STIM2 regulates both intracellular Ca2+ distribution and Ca2+ movement in skeletal muscle, which makes it both similar to, yet different from, STIM1.


Results
In skeletal muscle, STIM2 binds to SERCA1a via a region of amino acids from 453 to 729. To find unique STIM2-binding proteins among proteins mediating or regulating the contraction and relaxation of skeletal muscle, first, cDNA of the variable region between STIM1 and STIM2 (amino acids from 453 to 729 in STIM2, i.e., STIM2-UI in Fig. 1a) was constructed in a GST vector. The region was referred to as STIM2-UI because the role of the region was unidentified (UI). The GST-fused STIM2-UI protein was expressed in E. coli. The bacterial cell lysate was subjected to immunoblot assay with anti-GST antibody (Fig. 1b). The GST-fused STIM2-UI protein was successfully expressed (approximately 56 kDa). Next, affinity beads were prepared by immobilizing the GST-fused STIM2-UI proteins on GST beads, and the affinity beads were incubated with the solubilized triad sample from 'rabbit' skeletal muscle (i.e., binding assay). The triad sample is an enriched one with triad proteins that mediate the contraction and relaxation of skeletal muscle, as described in the Materials and Methods section. The proteins that were bound to the affinity beads were separated on a SDS-PAGE gel and were stained with Coomassie Brilliant Blue in order to find the proteins that could specifically bind to the GST-fused STIM2-UI protein (Fig. 1c). The bands for the proteins that bound to GST itself were excluded from consideration. Eight bands appeared as proteins that were bound to the GST-fused STIM2-UI protein.
In identifying the binding proteins, the eight bands were subjected to in-gel digestion and to qTOF mass spectrometry (qTOP-MS). Supplementary Fig. 1 and Table 1 show the results of qTOF-MS and database searches. Bands 1 to 6 were identified as non-specifically bound proteins that originated from E. coli. Band 7 was identified as SERCA1a that originated from rabbit skeletal muscle, which suggested that SERCA1a could SCientifiC RepoRts | (2017) 7:17936 | DOI: 10.1038/s41598-017-18256-3 be a STIM2-binding protein in skeletal muscle. The binding of full-length STIM2 to SERCA1a was accessed using bona-fide STIM2-and SERCA1a-expressing rabbit skeletal muscle and co-immunoprecipitation assay, and SERCA1a was successfully co-immunoprecipitated with full-length STIM2 (Fig. 1d). Therefore, based on the two different approaches (binding assay and qTOF-MS, and co-immunoprecipitation assays), we suggest that STIM2 binds to SERCA1a in skeletal muscle via its region from amino acids from 453 to 729.
It is possible that, according to the expected protein size, band 8 could be responsible for an oligomeric STIM2-UI (approximately 224 kDa in the case of a tetramer). However, there was no matching signal to the band 8 in the known databases, suggesting that the STIM2-UI region is irrelevant to the self-oligomerization of STIM2. It is possible that the signal from the band 8 could be obtained from experimental shortcomings, for example, an incomplete digestion of the band 8 by trypsin, which could result in a mixed signal that does not match to the known databases. STIM2 in skeletal myotubes attenuates SERCA1a activity, which contributes to the Ca 2+ distribution between the cytosol and the SR. To examine how STIM2 is related to SERCA1a activity, STIM2 in mouse primary skeletal myotubes was knocked down using siRNA. Three different siRNAs for mouse STIM2 were used to knock down STIM2 (Table 2, upper panel). Immunoblot assay with anti-STIM2 antibody using the lysate of STIM2-knockdown myotubes by each siRNA showed that STIM2 expression was reduced up to 90% by #3 siRNA compared with that of the untransfected or scrambled siRNA control (Fig. 2a). Therefore, #3 siRNA was used for the further experiments to knock down STIM2. On the other hand, fully differentiated myotubes on Figure 1. Schematic primary sequences of the STIM2 and STIM2-UI region, binding assay of GST-fused STIM1-UI protein with a triad sample, and co-immunoprecipitation of full-length STIM2 with SERCA1a. (a) The position of each domain in STIM2 is presented according to previous reports: the overall diagram 92,93 , and CAD/SOAR 94 . Numbers indicate the sequence of amino acids. UI means unidentified region. S, signal peptide; cEF, canonical EF-hand; hEF, non-functional hidden EF-hand; SAM, sterile α-motif; T, transmembrane domain; C, coiled-coil domain; CAD/SOAR, Ca 2+ release-activated Ca 2+ (CRAC)-activating domain/STIM1-Orai1-activating region; P, proline/serine-rich domain; K, lysine-rich domain. (b) Immobilized GST-fused STIM2-UI proteins on GST beads were separated on a 10% SDS-PAGE gel, and the gel was subjected to immunoblot with anti-GST antibody. GST or GST-fused STIM2-UI proteins are indicated by asterisks. (c) The bound proteins in the binding assay of GST-fused STIM2-UI protein with the triad sample from rabbit skeletal muscle were separated on a 10% SDS-PAGE gel, and the gel was stained with Coomassie Blue. GST was used as a negative control. GST-fused STIM2-UI proteins are indicated by an asterisk. The eight proteins that were bound to the GST-fused STIM2-UI protein are indicated on the right side of the figure (bands 1 to 8). (d) The triad sample obtained from rabbit skeletal muscle (30 μg of total proteins) was subjected to a co-immunoprecipitation assay with anti-STIM2 antibody, and the immunoprecipitant was subjected to immunoblot analysis with anti-STIM2 or anti-SERCA1a antibodies. Triad sample indicates a simple immunoblot of the triad sample. Without Ab indicates a reaction without anti-STIM2 antibody. Three independent experiments were conducted. IB, IP, or Ab means immunoblot, immunoprecipitation, or antibody, respectively. The immunoblot data were cropped from the immunoblot images of different gels and were grouped. The full-length blots are presented in Supplementary Fig. 4. SERCA1a was successfully co-immunoprecipitated with full-length STIM2.
SCientifiC RepoRts | (2017) 7:17936 | DOI:10.1038/s41598-017-18256-3 differentiation day 5 were much longer and thicker and more multi-nucleated than the untransfected immature myotubes on differentiation day 3 (left image of Fig. 2b). siRNA-transfected myotubes were not distinguishable from the untransfected or scrambled siRNA-transfected control. This suggests that myotube formation (i.e., the terminal differentiation) was not significantly affected by the STIM2-knockdown, and STIM2 is not significantly involved in the terminal differentiation of skeletal muscle.
To examine SERCA1a activity in the STIM2-knockdown myotubes, an oxalate-supported 45 Ca 2+ -uptake assay was conducted. The Ca 2+ -uptake activity of SERCA1a was not changed at a resting cytosolic Ca 2+ concentration (70 nM of free Ca 2+ , Fig. 3a) by the STIM2-knockdown. However, at a higher cytosolic Ca 2+ concentration, such as that found during skeletal muscle contraction (1 μM of free 45 Ca 2+ ), the Ca 2+ -uptake activity of SERCA1a was significantly increased by the STIM2-knockdown (more than 50% increase compared with the untransfected or scrambled siRNA control, Fig. 3a, and Table 3). This suggests that STIM2 attenuates SERCA1a activity at micromolar cytosolic Ca 2+ concentrations in skeletal muscle (i.e., during skeletal muscle contraction).
To access the increased SERCA1a activity by the STIM2-knockdown from a different angle, the cytosolic Ca 2+ level at rest and the amount of Ca 2+ in the SR in the STIM2-knockdown myotubes were measured using single-myotube Ca 2+ -imaging experiments (Fig. 3b,c). First, the cytosolic Ca 2+ level at rest was significantly reduced by the STIM2-knockdown ( Fig. 3b and Table 4). This result agreed with the prediction that more cytosolic Ca 2+ could be taken up into the SR by the increased activity of SERCA1a via the STIM2-knockdown, which, subsequently, lowers the cytosolic Ca 2+ level. Second, to measure the amount of Ca 2+ in the SR (that is, to estimate how much Ca 2+ is releasable from the SR to the cytosol), the SR of the STIM2-knockdown myotubes was depleted with thapsigargin (TG) in the absence of extracellular Ca 2+ . Extracellular free Ca 2+ allows the avoidance of extracellular Ca 2+ entry and an assessment of the amount of Ca 2+ exclusively in the SR ( Fig. 3c and Table 4). There was a significant increase in the amount of Ca 2+ in the SR via STIM2-knockdown. Thus, STIM2-mediated attenuation of SERCA1a activity could account for the balanced Ca 2+ distribution between the cytosol and the SR at rest in skeletal muscle. STIM2 in skeletal myotubes also contributes to SOCE. To examine the role of STIM2 in the SOCE of skeletal muscle, Ca 2+ in the SR was depleted with TG in the absence of extracellular Ca 2+ in the STIM2-knockdown myotubes, and extracellular Ca 2+ was applied to the myotubes (Fig. 4a). Both the area under the peaks (i.e., the amount of Ca 2+ entry, Fig. 4b) and the slope at the rising phase of the peaks (i.e., the rate of

PCR primers for the construction of GST-fused STIM2-UI region
Forward primer (with EcoR1 enzyme site) 5′-CGGAATTCATGAGCCTGACCTCTTCCCTTTATTC-3′ Backward primer (with Sal1 enzyme site) 5′-CGTCGACTCACTCTCCATTATGACAAAGGTCATG-3′ Ca 2+ entry, Fig. 4c) that represent the overall degree of extracellular Ca 2+ entry during SOCE were decreased by STIM2-knockdown. Therefore SOCE was significantly reduced via STIM2-knockdown ( Fig. 4 and Table 4). This result suggests that, in addition to the primary role of STIM1 in the SOCE, STIM2 also contributes to SOCE in skeletal muscle. It is possible that the STIM2-knockdown could alter the expression levels of Orai1 and STIM1 that mainly mediate SOCE in skeletal muscle, and their altered expression could be the cause of the reduced SOCE in the STIM2-knockdown myotubes. The expression levels of either Orai1 or STIM1 were accessed by immunoblot assay (see the results of the immunoblot assays below). There was no significant change in the protein content levels of either Orai1 or STIM1 via STIM2-knockdown. Therefore, the decrease in SOCE via STIM2-knockdown was not due to a simple change in the expression levels of either Orai1 or STIM1. STIM2 in skeletal myotubes regulates the intracellular Ca 2+ release from the SR to the cytosol through RyR1 during skeletal muscle contraction. To increase the cytosolic Ca 2+ level to a certain point at a given time is the key step in skeletal muscle contraction, and RyR1 is the only channel that is responsible for intracellular Ca 2+ release from the SR to the cytosol for skeletal muscle contraction. Intracellular Ca 2+ releases through RyR1 in response to caffeine, a direct agonist of RyR1 in skeletal muscle 55 , were examined in the STIM2-knockdown myotubes using Ca 2+ -imaging experiments (Fig. 5a). Responses to caffeine were increased via STIM2-knockdown ( Fig. 5a and Table 4). Responses to KCl, a membrane depolarizer (that can induce coupling between DHPR and RyR1 and then intracellular Ca 2+ release from the SR to the cytosol through RyR1 for skeletal muscle contraction 1-4 , i.e., the mimicry of intracellular Ca 2+ release during skeletal muscle contraction) were also increased by the STIM2-knockdown to an approximately equal response to caffeine (Fig. 5b, Table 4). Therefore, STIM2 in skeletal muscle participates in intracellular Ca 2+ release during skeletal muscle contraction.
To find additional possible factors for the changes by the STIM2-knockdown above, thirteen proteins that mediate or regulate Ca 2+ movements or Ca 2+ handling in skeletal muscle were examined by immunoblot assays using the lysate of STIM2-knockdown myotubes (Fig. 6a). There was no significant change in the protein content levels of the three main proteins that mediate Ca 2+ movements during the contraction and relaxation of skeletal muscle: DHPR, RyR1 and SERCA1a. No change in SERCA1a protein content suggests that the increased SERCA1a activity in the STIM2-knockdown myotubes was not simply due to an increase in the protein content levels of SERCA1a. There also was no significant change in the protein content levels of Orai1 and STIM1 or in the proteins that are responsible for extracellular Ca 2+ entry in skeletal muscle, TRPC1, TRPC3, and TRPC4, but there was a significant decrease in the protein content levels of TRPC6 (Fig. 6b). There also was a significant decrease in the protein content levels of JP1, which mediates the proper formation of the junctional membrane complex (Fig. 6b). However, based on transmission electron microscopy observations, there was no significant change in the junctional membrane complex by the STIM2-knockdown (data not shown), suggesting that the reduced amount of JP1 was sufficient to properly form the junctional membrane complex. There was no change in the protein content levels of JP2 or calsequestrin. Interestingly, via the STIM2-knockdown, there was a significant decrease in the protein content levels of calmodulin 1 (CaM1), which is a ubiquitous Ca 2+ -binding protein in various cells (Fig. 6b). These results suggest that TRPC6 and CaM1 could be involved in the changes that were found in the STIM2-knockdown myotubes.

Discussion
In the present study, we searched for STIM2-binding proteins from among proteins that mediate or regulate skeletal muscle functions, and we examined the functional role of the STIM2-binding protein along with STIM2  Table 4. Cytosolic Ca 2+ level at rest was decreased by STIM2-knockdown. (c) To measure the amount of Ca 2+ in the SR, TG was applied to the STIM2-knockdown myotubes in the absence of extracellular Ca 2+ to avoid extracellular Ca 2+ entry. A representative trace for each group is shown, and the results are summarized as bar graphs in the right-hand panel. *Significant difference compared with the untransfected control (p < 0.05). The values are presented as the mean ± S.E. for the number of myotubes shown in the parentheses of  Table 3. 45 Ca 2+ -uptake from the cytosol into the SR through SERCA1a in the STIM2-knockdown mouse primary skeletal myotubes. Either an untransfected or a scrambled siRNA-transfected control was used as a negative control. The results are presented as the mean ± S.E. of three independent experiments. *Significant difference compared with the untransfected control (p < 0.05).
in skeletal muscle. STIM2 binds to SERCA1a via its C-terminal region of amino acids from 453 to 729, and attenuates SERCA1a activity during skeletal muscle contraction. It seems that STIM2 could be involved in a moderate level of Ca 2+ -uptake to the SR through SERCA1a by attenuating SERCA1a activity in order to maintain skeletal muscle contraction for longer periods. During SOCE, three dimers of STIM1 (i.e., six STIM1s) are required to activate 'a' functional hexameric Orai1 that contains a central pore 56,57 . This stoichiometry seems not to be in the interaction between STIM2 and SERCA1a because a functional SERCA1a is monomeric and there is no symmetry in the three-dimensional structure of SERCA1a 58,59 . However, considering that STIM2 shares homologies with STIM1 in the amino acid sequences and domains, it is still possible that a dimeric or oligomeric STIM2 (possibly via EF-SAM, first coiled-coil domain, and/or CAD that are known to be involved in the self-oligomerization of STIM1 20,21,33 ) participates in the interaction with a SERCA1a.
Previously, we reported that STIM1 binds to SERCA1a in skeletal muscle and is required for the full-activity of SERCA1a during skeletal muscle relaxation, which means STIM1 is a positive regulator of SERCA1a 45 . In the present study, we found another in-situ SERCA1a regulator in skeletal muscle, STIM2, but its regulation is the reverse of STIM1, which means STIM2 is a negative regulator of SERCA1a. The relationship between STIM1 and STIM2 in terms of regulating SERCA1a is akin to a tug-of war that results in SERCA1a that is neither too much nor too little.
Interruption of STIM1-SERCA1a interaction by an over-expression of the SERCA1a-binding region of STIM1 in mouse primary skeletal myotubes induced effects similar to the properties of STIM2-knockdown myotubes in the present study, and resulted in a decrease in cytosolic Ca 2+ levels at rest, an increase in the amount of Ca 2+ in the SR, but no effect on terminal differentiation 45 . However, under the disrupted interaction of STIM1 or STIM2 with SERCA1a, the regulation of intercellular Ca 2+ movement through RyR1 in response to either caffeine or KCl differed: Ca 2+ movement through RyR1 was decreased by a disrupted STIM1-SERCA1a interaction 45     it was increased by a disrupted STIM2-SERCA1a interaction, as shown via the use of STIM2-knockdown in the present study. The binding region of STIM1 or STIM2 to SERCA1a belongs to the variable region between STIM1 and STIM2 (amino acids from 449 to 671 in STIM1, and from 453 to 729 in STIM2), which possibly creates the differences in the roles of STIM1 and STIM2 in skeletal muscle. Different properties in the sensing of Ca 2+ and in the self-oligomerization of STIM1 and STIM2 could also account for the different roles of STIM1 and STIM2 in skeletal muscle 39,40,60,61 . Considering that the regulation of SERCA2a activity by phospholamban depends on the phosphorylation status of phospholamban 62 , possible phosphorylation sites were searched in the SERCA1a-bindgin regions using NetPhos 3.1 (that predicts both generic-and kinase-specific phosphorylation sites), GPS 3.0 (that predicts kinase-specific phosphorylation sites in a proteome-wide level), or PhosphoSVM (that is a non-kinase-specific prediction tool) ( Supplementary Fig. 2). There were no similarities between the predicted possible phosphorylation sites in the SERCA1a-binding regions of STIM1 and STIM2, in terms of the position, number, or pattern of the predicted phosphorylation sites. Therefore, it seems that the binding of STIM2 to SERCA1a could be mediated in a manner that differs from that of STIM1. It has reported that the higher luminal Ca 2+ content in the SR enhances the activity of RyR1 in mouse, frog, or rabbit skeletal muscle 63 . In addition, the presence of luminal Ca 2+ in the SR is crucial for the coupled gating of RyR1 and DHPR during skeletal EC coupling 64,65 . Therefore the increase in the releasable Ca 2+ content from the SR (which reflects the luminal Ca 2+ content in the SR, Fig. 3c) is a clue to explain a possible mechanism whereby the intercellular Ca 2+ movement through RyR1 in response to either caffeine or KCl is significantly increased by STIM2-knockdown. Meanwhile, unlike the results in the present study, STIM2-knockdown in human skeletal myotubes in a related study reduces the response to a membrane depolarizer such as KCl 66 . In that study, repeated KCl applications that mimicked the state of skeletal muscle fatigue were used, instead of a single KCl application Figure 5. Enhanced intracellular Ca 2+ release through RyR1 in response to caffeine or KCl by STIM2knockdown. Caffeine (a specific agonist of RyR1) (a), or KCl (a membrane depolarizer) (b), was applied to the STIM2-knockdown myotubes, and the intracellular Ca 2+ release from the SR to cytosol through RyR1 in the myotubes was measured. Histograms are shown for the normalized peak amplitude to the mean value of those from the untransfected control. The results are presented as the mean ± S.E. for the number of experiments in the parentheses in Table 4. *Significant difference was compared with the untransfected controls (P < 0.05). Intracellular Ca 2+ releases in response to either caffeine or KCl is significantly increased by STIM2-knockdown.
SCientifiC RepoRts | (2017) 7:17936 | DOI:10.1038/s41598-017-18256-3 that mimics a single twitch, which was used in the present study. Based on these two studies, it is possible that STIM2 could affect the response of RyR1 to membrane depolarization in a frequency-dependent manner.
Less is known about the roles of STIM2 in skeletal muscle. It is proposed that STIM1 and STIM2 are functionally redundant because the over-expression of either one of them corrects most of the defects on the terminal differentiation of human myoblasts to myotubes by STIM2-or STIM1-knockdown 43,66 . STIM2, however, is required for the terminal differentiation of human myoblasts to myotubes 66 . On the other hand, no significant change in the terminal differentiation by either STIM1-or STIM2-knockdown was found in either the present study (Fig. 2b) or in our previous research 45 . These discrepancies in the effects of STIM2 on terminal differentiation seem to be caused by different experimental procedures that were used to knock down STIM2. In the present study, instead of the siRNA transfection to myoblasts (i.e., totally undifferentiated forms) that was used by others, siRNA for STIM2 was transfected to immature myotubes on differentiation day 3 (Fig. 2b), because we tried to examine the interplay between STIM2 and SERCA1a, and SERCA1a was expressed from the middle to the end stage during terminal differentiation 67 . Taken together, these studies provide interesting input and show that, in skeletal muscle, STIM2 could be a multiplayer in a stage-dependent manner, both as a member for the terminal differentiation at the early stage of the terminal differentiation, and as a regulator of SERCA1a and a SOCE-mediator at the late, or mature, stage of terminal differentiation.
Ca 2+ -charged CaM1 (Ca 2+ -CaM1) exerts an inhibitory effect on RyR1 channel activity 68,69 . Less availability of Ca 2+ due to decreases in SOCE by the STIM2-knockdown in the present study could induce a switch of Ca 2+ -CaM1 to aop-CaM1 that could enhance RyR1 activity. In addition, it is possible that the decreased protein content level of CaM1 by the STIM2-knockdown (Fig. 6a) could have reduced the availability of Ca 2+ -CaM1, which could also enhance RyR1 activity, and this could be a way to compensate for the reduced cytosolic Ca 2+ level caused by the STIM2-knockdown. This is well supported by a previous report showing that CaM1 regulates cytosolic Ca 2+ levels by regulating Ca 2+ release from the SR in skeletal muscle 69 . Indeed, RyR1 activity in response to a specific agonist of RyR1, caffeine, was increased in the present study (Fig. 5a). CaM1 is known to bind to STIM2 70 . Therefore, our results again support the finding that STIM2 regulates cytosolic Ca 2+ levels in skeletal muscle, and suggest the possibility that the regulation of cytosolic Ca 2+ levels by STIM2 occurs in conjunction with CaM1.
In addition to the major role of Orai1 as a Ca 2+ entry channel during SOCE 4,71 , canonical-type transient receptor potential cation channels (TRPC) have also been proposed as Ca 2+ entry channels for SOCE in skeletal muscle [72][73][74] . Skeletal muscle expresses mainly four types of TRPCs: TRPC1, TRPC3, TRPC4, and TRPC6 75 . TRPC1, TRPC3, and TRPC4 are known to mediate SOCE in skeletal muscle 76,77 . However, little is known about TRPC6 in skeletal muscle, although it is reported that myoblasts from mdx mice, a mouse model of Duchenne muscular dystrophy, shows a reduced expression of TRPC6 78 . In the present study, TPRC6 protein content was significantly decreased by the STIM2-knockdown (Fig. 6), and STIM2 bound to TRPC6 in skeletal muscle ( Supplementary  Fig. 3). This suggests the possibility that TRPC6 could also have participated in the reduction of SOCE or in reduced cytosolic Ca 2+ levels at rest via the STIM2-knockdown. However, the role of TRPC6 along with STIM2 in skeletal muscle should be resolved with a clear working mechanism. Here we suggest that similar and different roles of STIM1 and STIM2 exist in the skeletal muscle. First, both STIM1 and STIM2 are responsible for SOCE, albeit possibly for different reasons. Second, STIM2 is only now being thought of as an in-situ regulator of SERCA1a. The regulatory activity of STIM2 on SERCA1a is the reverse of STIM1 26 . Third, the differences in the roles of STIM1 and STIM2 in intracellular Ca 2+ movements in skeletal muscle could attribute to divergent and coordinated Ca 2+ -handling in skeletal muscle. Therefore, STIM2 could be a necessary partner of STIM1 in skeletal muscle function, and skeletal muscle could utilize STIM1 and/or STIM2, depending on the cellular context or need.

Materials and Methods
Ethics statement. All   Preparation of a triad sample, and binding assay of GST-fused STIM-UI protein with the triad sample. The triad vesicles that are enriched portions with proteins mediating or regulating skeletal muscle functions, including SERCA1a 3,4,81 , were prepared and solubilized to obtain the triad sample, as previously described 45,81,82 . Binding assays were performed, as previously described 45,80,83,84 . Briefly, affinity beads were prepared by immobilizing GST-fused STIM2-UI proteins on GST beads (Amersham, GE Healthcare Biosciences, Pittsburgh, PA). The affinity beads were then incubated with 150 μg of the solubilized triad sample for 8 h at 4 °C. The proteins that were bound to the affinity beads were separated on a 10% SDS-PAGE gel, and the gel was stained with Coomassie Brilliant Blue in order to obtain the protein bands.
In-gel digestion, protein identification by qTOF-MS, and database search. Protein bands obtained from the binding assay were subjected to in-gel digestion with trypsin 80,81 . The digested peptide solution was desalted and concentrated, and was eluted using a homemade C18 nano-column (100-300 nl with trypsin of POROS reverse-phase R2 material (20-30 μm in bead size, PerSeptive Biosystems, Foster City, CA)) and 1.5 ul of 50% MeOH, 49% H 2 O, and 1% HCO 2 H. qTOF-MS of the eluted peptides was performed using a Hybrid Quadrupole-TOF LC/MS/MS Mass Spectrometer (AB Sciex Instruments, Framingham, MA) equipped with an electrospray ionization (ESI) source. The quadrupole analyzer was used to select precursor ions for fragmentation in the hexapole collision cell. The produced ions were analyzed using an orthogonal TOF analyzer and fitted with a reflector, a micro-channel plate detector, and a time-to-digital converter.
Cell culture. Mouse primary skeletal myoblasts were derived from mouse skeletal muscle using a single-cell cloning method, and these proliferated and differentiated to myotubes, as previously described 15,26,45,83,[85][86][87] . For the terminal differentiation of the primary skeletal myoblasts to myotubes, myoblasts were re-plated either on 10-cm plates (either for the preparation of myotube lysate or for an oxalate-supported 45 Ca 2+ -uptake experiment) or on 96-well plates (for the single-myotube Ca 2+ imaging experiments) coated with Matrigel (BD Biosciences, San Jose, CA, USA). All reagents for the cell cultures were obtained from Invitrogen (ThermoFisher Scientific).
Knockdown of STIM2 in mouse primary skeletal myotubes. Three different siRNAs for mouse STIM2 (GenBank accession number NM_001081103.2) were selected using two forms of siRNA design software from Ambion (ThermoFisher Scientific) and from Sigma-Aldrich (Table 2, upper panel). A scrambled siRNA was used as a negative control. Immature myotubes on differentiation day 3 were transfected with one of the siR-NAs in a mixture containing 600 μl of low-glucose DMEM, 60 μl of X-tremeGENE siRNA Transfection Reagent (Roche Applied Science, Penzberg, Upper Bavaria, Germany), and 200 nM of the specific synthetic siRNAs, as previously described 15,26,45 . After the transfection, fully differentiated myotubes on differentiation day 5 were subjected to further experiments.

Co-immunoprecipitation and immunoblot assays.
For the co-immunoprecipitation assay, the solubilized triad sample (50 µg of total protein) was incubated with anti-STIM2 antibody (Sigma-Aldrich) overnight at 4 °C, as previously described 26,80,81,83 . Anti-STIM1 (Abcam, Cambridge, MA, USA), anti-SERCA1a (Thermo Scientific Inc., Rockford, IL, USA), or anti-TRPC6 (Alomone Laboratories, Jerusalem 9104201, Israel) antibody was used for immunoblot assay. For the immunoblot assay, fully differentiated mouse primary skeletal myotubes on 10-cm plates on differentiation day 5 were solubilized, and the solubilized lysate (5 or 10 μg of total protein) was subjected to SDS-PAGE (8, 10, or 12% gel) and immunoblot assay, as previously described 15,26,45,80,83,87 . Anti-RyR1, anti-CSQ, anti-CaM1, anti-JP1, and anti-JP2 antibodies were obtained from Thermo Scientific Inc. Anti-TRPC1, anti-TRPC3, and anti-TRPC4 antibodies were obtained from Alomone Laboratories. Anti-DHPR, In silico approach. The possible phosphorylation site was predicted using NetPhos 3.1 89 , GPS 3.0 90 , or PhosphoSVM 91 . Under NetPhos 3.1, both generic-and kinase-specific predictions are performed, and the kinase-specific predictions by NetPhos 3.1 cover the prediction by NetPhosK 1.0. GPS 3.0 predicts kinase-specific phosphorylations with a large-scale prediction of >13,000 mammalian phosphorylation sites and a proteome-wide prediction of Aurora-B specific substrates including protein-protein interaction information. PhosphoSVM is a non-kinase-specific prediction tool which detects possible phosphorylation sites for which the associated kinase is unknown or the number of known substrate sequences of the associated kinase is few. For the prediction using NetPhos 3.1 or PhosphoSVM, predictions with more than 0.7 in score were considered. For the predictions using GPS 3.0, threshold was set at high and predictions with more than 20 in score were considered. Statistical analysis. The results are presented as the mean ± S.E. for the number of myotubes shown in the parentheses of Tables or in the legends of either the Tables or Figures. The values were normalized to the mean value from the corresponding controls. The significant differences were analyzed using an unpaired t-test (GraphPad InStat, v2.04, GraphPad Software, La Jolla, CA, USA). The differences were considered to be significant at p < 0.05. The graphs were prepared using Origin v7 software.