The inactivation domain of STIM1 acts through intramolecular binding to the coiled-coil domain in the resting state

Ca²⁺ signaling plays a vital role in various cellular processes, such as proliferation, differentiation and cell death (Clapham, 2007). Store-operated Ca²⁺ entry (SOCE) is a major Ca²⁺ entry mechanism, which is activated by the depletion of endoplasmic reticulum (ER) Ca²⁺ stores (Putney, 1986). SOCE is mediated by two proteins, stromal interaction molecule 1 (STIM1) and Orai1 (Liou et al., 2005; Feske et al., 2006). STIM1 senses ER store depletion (Roos et al., 2005; Zhang et al., 2005) and further opens the Orai1 Ca²⁺ channel (Prakriya et al., 2006; Vig et al., 2006; Luik et al., 2008). Abnormal activation of STIM1 directly influences Ca²⁺ influx, resulting in severe diseases such as Stormorken syndrome (Nesin et al., 2014), York platelet syndrome (Markello et al., 2015) and tubular aggregate myopathy (Böhm et al., 2013). Therefore, STIM1 should be tightly maintained in its inactive state under resting conditions.

To understand the molecular mechanisms underlying STIM1 regulation, significant efforts have identified functional domains within STIM1 (Prakriya and Lewis, 2015). In the cytosolic C-terminus of STIM1, three coiled-coil domains (CC1, CC2 and CC3) are located behind a transmembrane domain. Among these domains, CC2 and CC3 constitute a Ca²⁺ release-activated Ca²⁺ (CRAC) activation domain [CAD; amino acids (aa) 342–448, also known as SOAR or CCb9], which binds directly to and activates Orai1 (Park et al., 2009; Kawasaki et al., 2009; Yuan et al., 2009). Recent studies have explored communications between cytosolic domains. Inactive STIM1 is reported to form a closed conformation through an intramolecular interaction between CC1 and CAD (Korzeniowski et al., 2010, 2017; Muik et al., 2011; Yang et al., 2012; Cui et al., 2013; Yu et al., 2013; Zhou et al., 2013; Fahrner et al., 2014; Ma et al., 2015). Under resting conditions, hydrophobic residues in CC1 and CC3 have been revealed to play a dominant role in retaining the closed conformation (Muik et al., 2011; Zhou et al., 2013; Fahrner et al., 2014; Ma et al., 2015). Just behind CC3, a conserved negatively charged region called the inactivation domain of STIM1 (IDstim, aa 470–491, also known as CMD or CTID) has been identified as essential for Ca²⁺-dependent inactivation of STIM1 (CDI) soon after Orai1 activation (Mullins et al., 2009; Derler et al., 2009; Srikanth et al., 2010). However, despite its functional importance and position, it is unknown whether IDstim plays a role in maintaining the inactive state of STIM1 before Orai1 activation.

In this study, we reveal that IDstim helps STIM1 keep inactive through a binding to the coiled-coil domain, retaining the hydrophobic interaction between CC1 and CC3. By identifying a short conserved linker whose extension or mutation (P445W) constitutively activates STIM1, we found that the linking between CAD and IDstim is critical for the inactive state of STIM1. Moreover, we identified the binding between IDstim and the coiled-coil domain, which is critical for the inactive state of IDstim. Binding of IDstim requires hydrophobic interaction between CC1 and CC3, and either mutations or deletions that disrupt the CC1–CC3 interaction impair the inhibitory function and binding of IDstim. This study expands our mechanistic understanding of how STIM1 maintains its inactive state to avoid spontaneous activation of SOCE under resting conditions.

IDstim (aa 470–491, 22 amino acids in length) consists of a conserved sequence of negatively charged amino acids and is essential for CDI soon after Orai1 activation (Mullins et al., 2009; Derler et al., 2009). However, the role of IDstim in keeping the inactive state of STIM1 before Orai1 activation is unknown. Therefore, we generated STIM1-ΔIDstim by deleting IDstim from STIM1 (Fig. 1A) and then measured the concentration of intracellular Ca²⁺ ([Ca²⁺]_i) in HEK293 cells co-expressing Orai1 with either wild-type STIM1 or STIM1-ΔIDstim using the fluorescent Ca²⁺ indicator Fura-2 (Fig. 1B–D). Cells expressing wild-type STIM1 did not show elevation of [Ca²⁺]_i (Fig. 1B,D). In contrast, the expression of STIM1-ΔIDstim elevated [Ca²⁺]_i, in a manner that was dependent on extracellular Ca²⁺ concentrations (Fig. 1B,D) even though the expression level of the constructs is similar (Fig. 1C). These results indicate that IDstim helps STIM1 keep inactive in the resting state as well as having a role in CDI after activation.

Next, to test whether [Ca²⁺]_i elevation by STIM1-ΔIDstim occurs through the STIM1–Orai1 complex, we monitored the location of mCherry-tagged STIM1 variants in cells co-expressing GFP–Orai1. We found that wild-type STIM1 did not colocalize with Orai1 before store depletion (Fig. 1E,F), but colocalization of these proteins occurred after store depletion (Fig. 1G,H) induced by thapsigargin [TG; a sarco/endoplasmic reticulum Ca²⁺-ATPase (SERCA) blocker]. In contrast, STIM1-ΔIDstim colocalized with Orai1 in cells both before (Fig. 1E,F) and after store depletion (Fig. 1G,H). These results reveal that STIM1-ΔIDstim is a constitutive activator of Orai1 and induces spontaneous Ca²⁺ influx through store-independent coupling with Orai1. Furthermore, reduction of [Ca²⁺]_i by means of treatment with the Orai1 blocker 2-APB (50 µM) supported that idea that the [Ca²⁺]_i elevation mediated by STIM1-ΔIDstim occurred through Orai1 (Fig. 1B,D).

Sustained [Ca²⁺]_i elevation induces various Ca²⁺-dependent cellular processes by transcriptionally upregulating nuclear factor of activated T-cells (NFAT) target genes including interleukin-2 (Feske et al., 2001). To check whether abnormal activation of STIM1-ΔIDstim alters the cellular signaling, we examined whether the [Ca²⁺]_i increase from STIM1-ΔIDstim activates an NFAT-dependent luciferase reporter (NFAT-luc). In the presence of the protein kinase C (PKC) activator phorbol 12-myristate 13-acetate (PMA), cells expressing STIM1-ΔIDstim showed three times higher luciferase activity than did cells expressing wild-type STIM1 (Fig. S1A). This indicates that the [Ca²⁺]_i rise from STIM1-ΔIDstim is sufficient to induce SOCE downstream signaling for the expression of NFAT target genes.

Interestingly, we realized that there was a conserved 21-aa linker resides between the inactivation domain IDstim and the activation domain CAD (Fig. S2A). To check whether this linker might be involved in the regulation of STIM1 activity, we tried to mutate the conserved amino acids in the linker. Serendipitously, we constructed a STIM1 mutant that contains an additional 44 aa after P445 through insertion of primer repeats (Fig. 2A). To our surprise, this 44 aa insertion mutant elevated [Ca²⁺]_i depending on extracellular Ca²⁺ concentrations (Fig. 2B–D) and coupled to Orai1 without store depletion (Fig. 2E–H). Moreover, the expression of the 44 aa inserted mutant led to constitutive translocation of GFP–NFAT1 (Fig. S2B,C) and showed a higher luciferase activity with NFAT-luc compared to wild-type STIM1 in the presence of PMA (Fig. S2D).

This observation led us to hypothesize that the intramolecular proximity between the two opposite domains is critical for inactive STIM1. To test this, we increased the length of the linker by inserting various repeats of the elastin-like polypeptide [ELP; (VPGVG)_n] between G462 and S463 (Fig. 3A) and measured [Ca²⁺]_i in the cells co-expressing these ELP-inserted variants with Orai1. A short extension (20 aa) produced no elevation of [Ca²⁺]_i despite an external Ca²⁺ elevation, similar to what is seen with wild-type STIM1 (Fig. 3B–D). However, longer extensions (40 aa and 60 aa) disrupted the inactive state of STIM1 and elevated [Ca²⁺]_i when the external Ca²⁺ levels were elevated (Fig. 3B–D; Fig. S2E,F). We also found that the 40 aa extension resulted in STIM1 coupling to Orai1 without store depletion (Fig. 3E–H). To determine whether it is possible to activate Ca²⁺ signaling by lengthening the linker between the two domains, we measured NFAT transcriptional activity in cells co-expressing the ELP-inserted STIM1 variants via NFAT-luc. Luciferase assays revealed that the longer extensions (40 aa and 60 aa) elevated NFAT transcriptional activity compared to that of the wild-type or 20 aa extension (Fig. S2G).

In the linker, small amino acids (glycine, alanine and serine), which impart a flexible structure, are frequent and conserved (Fig. S2A). In addition, the linker contains highly conserved proline residues, which have a unique structure that usually introduces kinks (Fig. S2A). These unique features might allow this short and conserved neck-linker domain between CAD and IDstim to harmonize a functional interaction between CAD and IDstim to maintain STIM1 in an inactive confirmation in resting conditions. To test this hypothesis, we mutated three conserved proline residues in the neck linker to the bulkiest amino acid, tryptophan, to introduce a structural distortion. We then checked the activity of the STIM1 mutants (P445W, P458W and P468W) and found that STIM1 P445W induced a spontaneous Ca²⁺ influx (Fig. 3I–K). To understand the significance of P445 in the inactive state of STIM1, we mutated P445 to various amino acids from small amino acids (P445G and P445A) to bulky amino acids (P445V, P445F and P445W). We then checked [Ca²⁺]_i in the cells expressing these mutants (Fig. S2H,I). We found that the STIM1 mutations to small amino acids (P445G and P445A) did not alter its activity and it behaved like wild-type STIM1. However, the bulky amino acid STIM1 mutants (P445V and P445F) caused spontaneous Ca²⁺ influx in resting conditions, like STIM1 P445W. This suggests that bulky hydrophobic residues at position 445 impair the inactive state of STIM1, perhaps by altering the interactions of IDstim.

To investigate the inhibition mechanism occurring in the cytosolic portion of STIM1, we continued our study using STIM1 cytosolic fragments. We first determined whether IDstim could inhibit CAD directly by constructing a CAD-IDstim fragment (aa 342–491; Fig. 4A). The expression of CAD-IDstim raised [Ca²⁺]_i at 2 mM extracellular Ca²⁺ to levels similar to those observed with the expression of CAD (Fig. 4B–D). This result implies that IDstim alone is not sufficient to inhibit the activity of CAD. To eliminate the possibility that the length of the inhibitory domain matters and to confirm the inability of the inhibitory effect of IDstim on CAD, we assessed the effect of another extended inhibitory domain, the C-terminal inhibitory domain (CTID, aa 448–530) (Jha et al., 2013). We constructed an extended STIM1 fragment (CAD-CTID, aa 342–530; Fig. S3A) and assessed its activity. As with CAD-IDstim, CAD-CTID elevated [Ca²⁺]_i, to a similar level to that observed with CAD expression alone (Fig. S3B,C). These results indicate that IDstim and CTID failed to inhibit the activity of CAD, suggesting that IDstim might require an additional domain between the transmembrane domain and CAD to act a co-inhibitory domain of IDstim. To test this hypothesis, we extended CAD-IDstim to include CC1 to form CC1-CAD-IDstim (C-C-I) and assessed its activity. In contrast to the results with CAD-IDstim, CC1-CAD-IDstim did not lead to a rise in [Ca²⁺]_i, indicating that it was in the inactive state (Fig. 4B–D). Moreover, we found that CAD-IDstim colocalized with Orai1 in the vicinity of the cell membrane, whereas CC1-CAD-IDstim did not colocalize with Orai1 and was located in the cytosol (Fig. 4E,F). These results validate that CC1 is necessary for the inhibitory function of IDstim.

Next, to validate whether the inhibitory function of IDstim requires CC1 and is sufficient for cellular signaling, we compared the nuclear translocation of GFP–NFAT1 with mCherry–CAD-IDstim or mCherry–CC1-CAD-IDstim after co-transfection in HEK293T cells. The fluorescence imaging revealed that 66% of cells expressing CAD-IDstim showed nuclear translocation of GFP–NFAT1 (Fig. S3D,E). In contrast, cells expressing CC1-CAD-IDstim showed a similar level of GFP–NFAT1 nuclear localization as that of control cells (11% of cells) (Fig. S3D,E). In addition to the nuclear translocation of NFAT, we assessed NFAT transcriptional activity through the previously mentioned NFAT-luc. In the presence of PMA, NFAT transcriptional activity was lower with CC1-CAD-IDstim than with CAD-IDstim (Fig. S3F). These results confirmed that IDstim requires CC1 to keep the cytosolic domains inactive.

Inactive STIM1 retains the closed conformation through hydrophobic interaction between CC1 and CAD (Korzeniowski et al., 2010; Muik et al., 2011; Ma et al., 2015; Hirve et al., 2018). Thereafter, does the CC1-CAD interaction alone keep STIM1 inactive, or does it require IDstim to maintain an inactive state? To address this question, we constructed two cytosolic STIM1 fragments, CC1-CAD-linker (similar to OASF; Muik et al., 2011) and CC1-CAD, by deleting IDstim and linker-IDstim from CC1-CAD-IDstim, respectively (Fig. 5A). We co-expressed these STIM1 fragments with Orai1 in HEK293 cells and measured the [Ca²⁺]_i. Cells transfected with either STIM1 fragment showed elevated [Ca²⁺]_i (Fig. 5B–D), indicating that both constructs are in an active state. However, elevation of [Ca²⁺]_i was not observed in cells expressing CC1-CAD-IDstim (Fig. 5B–D) or in longer cytosolic fragments containing IDstim (aa 230–513, 230–535 and 230–685; Fig. S4A,B). We also assessed the colocalization of both CC1-CAD or CC1-CAD-IDstim with Orai1 via fluorescence imaging. We co-expressed mCherry-tagged cytosolic fragments in cells with GFP–Orai1. The cells co-expressing mCherry–CC1-CAD and GFP–Orai1 showed colocalization of these proteins (Fig. 5E,F). In contrast, mCherry–CC1-CAD-IDstim did not colocalize with GFP–Orai1 (Fig. 5E,F). These results indicate that CC1 and IDstim are required for the inactive state of the cytosolic domain of STIM1.

To check the effect of CC1-CAD-IDstim on the downstream signaling pathway, we transfected HEK293T cells with GFP–NFAT1 and either mCherry–CC1-CAD-IDstim or mCherry–CC1-CAD. We found that 87% of the cells expressing CC1-CAD showed nuclear translocation of NFAT1, similar to the percentage observed with cells expressing CAD (Fig. S4C,D). In contrast, 11% of the cells expressing CC1-CAD-IDstim showed nuclear translocation of NFAT1, similar to the percentage observed with cells expressing mCherry (Fig. S4C,D). We also confirmed NFAT transcriptional activity using NFAT-luc. Consistent with the nuclear translocation data, in the presence of PMA, NFAT-luc was activated in cells expressing CC1-CAD, but not in cells expressing CC1-CAD-IDstim (Fig. S4E). These results indicate that IDstim prevents the CC1-CAD-mediated downstream signaling pathway.

Next, we confirmed whether the interaction between CC1 and CC3, a component of CAD, is altered in the presence of IDstim. We performed a fluorescence resonance energy transfer (FRET) assay with CC1-CAD and CC1-CAD-IDstim. We fused mRuby3 and mClover3 to the N-termini and C-termini of the STIM1 cytosolic fragments, respectively. First, we measured [Ca²⁺]_i in cells co-expressing Orai1 and either mRuby3–CC1-CAD-IDstim–mClover3 or mRuby3–CC1-CAD–mClover3 (Fig. 5G,H) and confirmed that introducing mRuby3 and mClover3 at both ends did not affect the activity of each fragment. We next expressed either mRuby3–CC1-CAD-IDstim–mClover3 or mRuby3–CC1-CAD–mClover3 in cells and performed the intramolecular FRET assay. The FRET efficiency was not different in cells expressing the CC1-CAD-IDstim or CC1-CAD construct (∼0.3 for both, Fig. 5I), similar to that of the positive internal FRET control, mRuby3–12aa-mClover3 (Fig. S4F). This result indicates that the N- and C-termini of the cytosolic fragments are spatially close both in the absence and presence of IDstim. This suggests that IDstim does not alter the arrangement of the interacting CC1 and CC3. Thus, IDstim might induce a subtle structural change in CC1-CAD for stabilizing the inactive state without causing a dramatic transition in the closed conformation.

We next considered how IDstim could inhibit the cytosolic domains. We assessed the binding of IDstim to CAD or other cytosolic domains. We labeled the cytosolic domains and IDstim with Flag and YFP tags, respectively (Fig. 6A). We then expressed these proteins in HEK293T cells to investigate their binding via co-immunoprecipitation. Interestingly, immunoprecipitation of Flag-tagged CC1-CAD, a whole coiled-coil domain, resulted in co-immunoprecipitation of YFP-IDstim in cells expressing both proteins (Fig. 6B; Fig. S5A). However, both Flag-tagged CC1 and CAD failed to co-immunoprecipitate with YFP–IDstim (Fig. 6B; Fig. S5A). Similarly, the extended IDstim variants (aa 470–513 or 470–535) could bind CC1-CAD, but not CC1 or CAD (Fig. S5B,C). These results reveal that IDstim binds to the cytosolic domain only in the presence of the whole coiled-coil domain containing CC1 through CAD.

To validate the correlation between the binding and function of IDstim, we co-expressed IDstim with either CC1-CAD or CC1-CAD-linker in HEK293 cells and measured [Ca²⁺]_i. Cells co-expressing IDstim with either CC1-CAD or CC1-CAD-linker showed reduced [Ca²⁺]_i compared to the [Ca²⁺]_i in cells expressing either CC1-CAD or CC1-CAD-linker alone (Fig. 6C,D). These results indicate that IDstim inhibits the activity of CC1-CAD or CC1-CAD-linker. Because the expression levels of CC1-CAD (and CC1-CAD-linker) were similar in both conditions (Fig. 6E), the reduced [Ca²⁺]_i was not caused by differences in protein levels. Considering that IDstim expression inhibited the activity of CC1-CAD, which does not contain the linker, the linker seems to be important in the intramolecular interaction of IDstim rather than the interaction of the external IDstim. We next expressed IDstim with CAD, which failed to bind with IDstim. We found that the [Ca²⁺]_i in cells co-expressing CAD and IDstim was similar to that in the cells expressing CAD alone (Fig. 6F,G), indicating that IDstim failed to suppress the activity of CAD. Considering that IDstim does not bind with CAD, the binding target of IDstim matches its functional target.

Several lines of evidence from our experiments indicate that CC1 is necessary for the inhibitory function (Fig. 4) and binding of IDstim (Fig. 6). CC1 is composed of short three coiled-coil domains: CC1α1, CC1α2 and CC1α3. To further map the CC1 residues for the inhibitory function of IDstim, we prepared several constructs by sequentially deleting N-terminal residues from CC1-CAD-IDstim (i.e. CC1-CAD-IDstim-Δ34aa, -Δ58aa, -Δ80aa, and -Δ96aa; Fig. 7A). We then assessed the activity of these constructs by measuring [Ca²⁺]_i. CC1-CAD-IDstim-Δ34aa (aa 264–491) and constructs with larger deletions (CC1-CAD-IDstim-Δ58aa, -Δ80aa, and -Δ96aa) produced higher Ca²⁺ levels than did CC1-CAD-IDstim (Fig. 7B–D). Furthermore, we found that CC1-CAD-IDstim-Δ34aa colocalized with Orai1 in the vicinity of the cell membrane, like CAD-IDstim (Fig. 7E,F). To determine the effect of CC1-CAD-IDstim-Δ34aa on downstream Ca²⁺ signaling, we examined whether CC1-CAD-IDstim-Δ34aa increases transcription from NFAT-luc. Consistent with the increased [Ca²⁺]_i, CC1-CAD-IDstim-Δ34aa enhanced luciferase activity in the presence of PMA, compared with that of CC1-CAD-IDstim (Fig. S6A). Overall, we found that CC1α1 (238–271) containing the first 34 aa is a necessary domain for the inhibitory function of IDstim. To specify the necessary CC1α1 residues for the binding of IDstim, we deleted the N-terminus residue of CC1α1 by 5 amino acid units. We then found that deletion of 245–250aa significantly reduces the binding affinity to IDstim and the deletion of 255–260aa completely removes the binding affinity to IDstim (Fig. 7G), implying that 245–260aa might be necessary residues for the binding of IDstim. Considering that 245–260aa contains the leucine residues that are important for CC1-CAD binding (Muik et al., 2011; Ma et al., 2015), it seems that CC1α1 must bind to CAD in order for IDstim to bind.

STIM1 has been known to form a hydrophobic interaction between CC1α1 and CC3 in the resting state (Muik et al., 2011; Zhou et al., 2013; Ma et al., 2015). Considering that IDstim requires CC1α1 to maintain cytosolic domains in an inactive state (Fig. 7), this hydrophobic interaction may be necessary for the inhibitory function of IDstim. In CC1α1, two leucine residues (L251 and L258) are known to be critical for the hydrophobic interaction between CC1 and CC3 (Muik et al., 2011; Ma et al., 2015). The mutation of these leucine residues to glycine or serine (L251S or L258G) abolishes the hydrophobic interaction, resulting in the constitutive activation of STIM1 (Muik et al., 2011; Ma et al., 2015). To verify whether this hydrophobic interaction in CC1α1 is essential for the inhibitory function of IDstim, we introduced either the L251S or L258G mutation in CC1-CAD-IDstim (Fig. 8A). We first assessed whether each leucine mutation sufficiently disrupts the hydrophobic interaction between CC1 and CC3 in CC1-CAD-IDstim using the intramolecular FRET assay. We found that CC1-CAD-IDstim L251S showed reduced FRET efficiency compared to that of wild-type CC1-CAD-IDstim, when we fused mRuby3 and mClover3 to the N-terminus and C-terminus, respectively (Fig. 8B). The results indicate that the L251S mutation impairs the hydrophobic interaction in CC1-CAD-IDstim. We measured [Ca²⁺]_i after expressing these CC1-CAD-IDstim mutants. We found that expression of either CC1-CAD-IDstim L251S or CC1-CAD-IDstim L258G induced a sustained elevation in [Ca²⁺]_i, compared to that induced by CC1-CAD-IDstim (Fig. 8C–E). Moreover, these mutations induced CC1-CAD-IDstim to colocalize with Orai1 (Fig. 8F,G) and increased the interaction between CC1-CAD-IDstim and Orai1 (Fig. 8H). To confirm whether CC1-CAD-IDstim L251S and CC1-CAD-IDstim L258G induce downstream Ca²⁺ signaling, we co-transfected GFP–NFAT1 with either mCherry–CC1-CAD-IDstim L251S or mCherry–CC1-CAD-IDstim L258G in HEK293T cells and assessed the nuclear translocation of NFAT1. We found that the L251S or L258G mutation in CC1-CAD-IDstim increased the percentage of cells with NFAT1 nuclear translocation from 16% to 65% (Fig. S6B,C). We also confirmed NFAT transcriptional activity in cells expressing CC1-CAD-IDstim mutants via NFAT-luc. As a consequence, in the presence of PMA, cells expressing either CC1-CAD-IDstim L251S or CC1-CAD-IDstim L258G showed a higher luciferase activity than cells expressing wild-type CC1-CAD-IDstim (Fig. S6D). These results indicate that the hydrophobic interaction in CC1α1 is needed for the inhibitory function of IDstim.

CC3 (aa 408–437) within the CAD is known to act as a binding partner of CC1α1 during the hydrophobic interaction, and L416G mutation in CC3 abolishes this hydrophobic interaction (Muik et al., 2011; Ma et al., 2015). By introducing this mutation into CC1-CAD-IDstim, we further assessed whether the hydrophobic interaction is required for the inhibitory function of IDstim. We first examined the interaction between CC1α1 and CC3 in CC1-CAD-IDstim L416G using the intramolecular FRET assay. We found the reduced FRET efficiency of CC1-CAD-IDstim L416G compared to that of wild-type CC1-CAD-IDstim (Fig. 8B), indicating that L416G impairs the hydrophobic interaction in CC1-CAD-IDstim. We co-expressed CC1-CAD-IDstim L416G and Orai1 in HEK293 cells and measured [Ca²⁺]_i. We found that [Ca²⁺]_i was higher with CC1-CAD-IDstim L416G than that with CC1-CAD-IDstim (Fig. 8C–E). Next, we compared the binding affinity of Orai1 to wild-type CC1-CAD-IDstim or CC1-CAD-IDstim L416G. For the comparison, we labeled the CC1-CAD-IDstim variants and Orai1 with YFP and Flag tags, respectively. We then examined their binding in co-immunoprecipitation assays. CC1-CAD-IDstim L416G immunoprecipitated more Orai1 than wild-type CC1-CAD-IDstim did (Fig. 8H). Thus, the L416G mutation in CC3 disrupts the inhibitory function of IDstim, confirming that the hydrophobic interaction between CC1 and CC3 is a requisite for the inhibitory function of IDstim.

We found that IDstim keeps the cytosolic domains inactive through binding (Fig. 6). Furthermore, to determine if the hydrophobic interaction is related to the binding of IDstim, we verified whether abolishing the hydrophobic interaction disrupts the binding of IDstim. We labeled the CC1-CAD mutants and IDstim with Flag and YFP tags, respectively. We then assessed their interaction by co-immunoprecipitation. As a result, Flag-tagged CC1-CAD L251S and L258G showed decreased binding to YFP-IDstim compared with that of wild-type CC1-CAD (Fig. 8I). Thus, the hydrophobic interaction is important for the proper binding of IDstim.

To conclude, we investigated how STIM1 maintains its inactive state with IDstim. We found that IDstim binds to the coiled-coil domain and requires the hydrophobic interaction between CC1 and CC3 to keep CC1-CAD inactive. Our finding is likely to contribute to the understanding of how the unintended activation of SOCE is prevented under physiological and resting conditions.

Our study reveals the mechanism by which IDstim helps STIM1 keep inactive through maintaining an inactive STIM1 conformation. Even though IDstim was described to be essential for CDI after Orai1 activation, it was unknown whether IDstim is important for maintaining the inactive state of STIM1 before Orai1 activation. Under resting conditions, the hydrophobic interaction between CC1 and CC3 was identified capable of maintaining the closed conformation of STIM1 (Muik et al., 2011; Ma et al., 2015; Fahrner et al., 2014). However, truncated CC1-CAD is constitutively active despite maintaining its head-to-tail proximity (Fig. 5), and this observation is consistent with previous reports (Derler et al., 2009; Park et al., 2009; Muik et al., 2009, 2011). Based on these results, the hydrophobic interaction alone is not sufficient to prevent inappropriate activation of the fragment, implying an additional control to prevent STIM1 activation. The present study reveals that IDstim binds and inhibits CC1-CAD fragment, and that this interaction may be a process to maintain STIM1 in an inactive state following the hydrophobic interaction between CC1 and CAD.

To the best of our knowledge, the present study is the first to identify the intramolecular target of IDstim. Until now, the molecular mechanism underlying the involvement of IDstim in the regulation of SOCE had been derived from studies on the extrinsic role of IDstim and the STIM1 partners, Orai1, SARAF and CaM (Mullins et al., 2009; Jha et al., 2013; Mullins and Lewis, 2016). Our study showed the intramolecular interaction of IDstim within the cytosolic portion of STIM1. We revealed that IDstim requires CC1 to inhibit CAD (Fig. 4) and binds with the coiled-coil domain (Fig. 6). Furthermore, our results show that IDstim requires hydrophobic interaction within the coiled-coil domain (Fig. 8). By discovering a new role for IDstim, this study advances our mechanistic understanding of how STIM1 functions.

IDstim contains conserved negatively charged amino acids. Previously, we found that the mutation of six negative charged amino acids to alanine or glycine (4A2G) abrogates CDI (Mullins et al., 2009). In this study, we found that the 4A2G mutation also activates CC1-CAD-IDstim (Fig. S7A,B) and disrupts the binding of IDstim to CC1-CAD (Fig. S7C). Thus, the negatively charged amino acids seem to be critical for the inhibitory effect and binding of IDstim as in CDI.

A previous report identified SOCE-associated regulatory factor (SARAF) as an inhibitor of STIM1 (Palty et al., 2012). Another study demonstrated that CTID (aa 448–530) is a binding site for SARAF (Jha et al., 2013). Interestingly, the report suggested that aa 448–490, which encompasses IDstim, suppresses the binding of SARAF and that the binding site for SARAF occurs at aa 490–530, after IDstim. In our study, the cytosolic STIM1 fragments were in an inactive state when they contained IDstim, although they did not include the SARAF-binding site (Fig. 4). Based on these observations, the intramolecular interaction of IDstim seems to inhibit STIM1 in a different way compared to SARAF.

The atomic resolution structure of the full cytosolic portion of STIM1 has not been obtained to date. Yang et al. revealed the crystal structure of a smaller STIM1 cytosolic fragment (SOAR, aa 345–444). However, three point mutations were required because purification of the wild-type form was challenging (Yang et al., 2012). Very recently, Rathner et al. reported an optimized protocol for one-step purification of cytosolic OASF_ext (aa 234–491), which is similar to CC1-CAD-IDstim (Rathner et al., 2018). Thus, a structural study to demonstrate the intramolecular interaction of IDstim should be carried out in the near future.

This study shows that deleting IDstim induces the translocation of STIM1 to the vicinity of Orai1, thus elevating intracellular Ca²⁺ levels (Fig. 1). A previous report examined STIM1Δ475-490 mutant clusters in the ER–plasma membrane junction in the absence of ER Ca²⁺ depletion (Jha et al., 2013). However, the same report demonstrated that STIM1Δ475-490, despite its colocalization with Orai1, did not elevate the intracellular Ca²⁺ levels. This conflicting result can be explained by the differences in methodology, the co-expression of Orai1. Indeed, we found that the single expression of STIM1-ΔIDstim gave similar [Ca²⁺]_i levels as STIM1, whereas STIM1-ΔIDstim co-expression with Orai1 resulted in much higher [Ca²⁺]_i levels (Fig. S1B,C). Similarly, a previous study has shown that a STIM1 cytosolic domain, CCb7 (aa 339–475), induced spontaneous Ca²⁺ influx with the co-expression of Orai1 as in our study, but failed to induce the Ca²⁺ influx with CCb7 expression alone (Kawasaki et al., 2009).

In this study, we proceeded in our experiments with the cytosolic STIM1 fragments. However, the behavior of IDstim in the context of full-length STIM1 is a remaining question. In previous studies, we had shown that STIM1 1–448 and STIM1 1–469 were not spontaneously active even though they lack IDstim (Mullins et al., 2009; Park et al., 2009). Therefore, we examined the significance of IDstim with the ER-anchored STIM1 fragments. When the expression levels were low or moderate, STIM1 1–448 and STIM1 1–469 did not show elevation of [Ca²⁺]_i (Fig. S8). However, when expression levels were high, STIM1 1–448 and STIM1 1–469 showed higher [Ca²⁺]_i levels compared to STIM1 1–491, similar to that of CAD (Fig. S8). These results suggest that, even in ER-anchored fragments, deletion of IDstim makes the STIM1 fragments more active, whereas the ER-anchored fragments are more resistant to activation than cytosolic fragments are. These results imply that IDstim contributes to the stabilization of the inactive state of STIM1.

A gain-of-function mutation (R304W) that causes constitutive STIM1 activation has been reported in patients with Stormorken syndrome (Nesin et al., 2014). Recently, it was reported that this R304W mutation stiffens the linker between CC1α2 and CC1α3, inducing CC1–CC1 intermolecular interaction and interfering with CC1–CC3 intramolecular interaction (Fahrner et al., 2018). The current study reveals that interfering with the CC1–CC3 interaction impairs the binding of IDstim to the coiled-coil domain (Fig. 7). Based on our results, we speculate that the constitutive STIM1 activation seen in the R304W mutant is due to a reduction in the intramolecular interaction of IDstim. Further testing and structural validation are needed for a new approach to reveal the IDstim-related physiological responses.

HEK293 and HEK293T cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS) at 37°C and 5% CO₂. For transient transfection, the cells were transfected at 70% confluency with 0.2–1 μg DNA using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions.

Human STIM1 or human Orai1 were amplified by PCR using the primer sets shown in Table S1. The PCR products were cloned into the pCR8/GW/TOPO cloning vector (Invitrogen) to yield the entry vectors, pCR8-STIM1, and pCR8-Orai1. The expression constructs were generated by transferring the pCR8-PCR products into custom-designed Gateway destination vector using enzyme mix (Gateway LR Clonase II; Invitrogen, Carlsbad, CA) according to the manufacturer's manual. All constructs were verified by sequencing with GW1 primer.

STIM1 (L251S, L258G, or L416G) mutations were constructed by mutating nucleotides by PCR using the primer sets shown in Table S1 (Quickchange XL; Stratagene).

STIM1 truncation constructs were generated by PCR using the primer sets shown in Table S1. The PCR products were cloned into the pCR8/GW/TOPO cloning vector (Invitrogen) to yield the entry vectors. The expression constructs were generated by transferring the pCR8 PCR products into custom-designed Gateway destination vector using enzyme mix (Gateway LR Clonase II; Invitrogen, Carlsbad, CA) according to the manufacturer's manual. All plasmids were sequenced to verify the constructs.

Fura-2/AM and Lipofectamine 2000 were obtained from Invitrogen. Thapsigargin (TG) and phorbol 12-myristate 13-acetate (PMA) were from Santa Cruz Biotechnology. Anti-Flag M2 affinity gel was obtained from Sigma. Antibodies targeting GFP (1:1000, 598, MBL), Flag (1:1000, M2, F1804, Sigma) were purchased from the indicated vendor.

HEK293T cells were cultured on poly-ornithine-coated 15 mm round cover glass at 37°C under 5% CO₂. Cells were transfected with GFP–NFAT1 and other indicated constructs by using Lipofectamine 2000; and the culture medium was changed after 2 h and cells incubated for 18–24 h. The cells were washed with PBS three times and fixed with 4% paraformaldehyde and mounted with the Aqua polyMount solution (PolyScience) at room temperature. The cells were imaged using the IX83 microscope equipped with an Olympus ×40 objective lens (oil, NA 1.30), fluorescent lamp (Olympus), a stage controller (LEP) and a CCD camera (ANDOR). Images were analyzed using Fiji or ImageJ software. Confocal fluorescence images were acquired on an LSM780 (NLO; Zeiss) confocal microscope with a Zeiss 100× oil objective lens (NA 1.46). mCherry and eGFP were excited simultaneously at 594 nm and 488 nm, respectively. Fluorescence emission was collected at 615–840 nm (mCherry) and 510–570 nm (eGFP). Zen software (Zeiss) was used for image acquisition.

HEK293T cells were transfected with the NFAT reporter gene and the indicated constructs. Co-transfection with the Renilla luciferase gene (pRL-TK) was used as an internal control for cell number and transfection efficiency. After 12-18 h incubation, cells were treated with dimethyl sulfoxide (DMSO), or phorbol 12-myristate 13-acetate (PMA; 1 μM) for 4–8 h. Assays were performed by the Dual Luciferase Reporter Assay System (Promega). For each condition, luciferase activity was measured with samples taken from duplicate wells with a 96-well automated luminometer (Bio-Rad). Results are calculated as the ratio of firefly to Renilla luciferase activity.

HEK293T cells were transfected with the indicated constructs for 12–24 h. Transfected cells were washed with PBS three times and lysed with 50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1% Triton X-100. Lysates were centrifuged at 9700 g for 10 min, and the supernatant was incubated at 4°C with anti-Flag M2 agarose beads (Sigma) overnight. Lysates (3% of total lysates) and immunoprecipitated samples (precipitated from 30% of total lysates) were run on an SDS-PAGE gel and electro-transferred onto PVDF membranes. The PVDF membranes were blocked with 7% skim milk for 2 h at room temperature, probed overnight at 4°C with specific primary antibodies in 3% BSA solution, incubated for 1 h at room temperature with horseradish peroxidase (HRP)-conjugated secondary antibody, and detected by enhanced chemiluminescence (Pierce).

Cells were loaded with 1 μM Fura-2/AM in DMEM at 37°C for 30 min. Ratiometric Ca²⁺ imaging was performed at 340 and 380 nm in 0 or 2 mM Ca²⁺ Ringer's solution with a IX81 microscope (Olympus) equipped with an Olympus ×40 oil (NA 1.30) objective equipped with a fluorescent arc lamp (LAMDA LS), an excitation filter wheel (SUTTER, LAMBDA 10-2), a stage controller (ASI, MS-2000) and a CCD camera (Hamamatsu, C10600) at room temperature. Images were processed with Metamorph and analyzed with Igor Pro.0.

FRET imaging experiments were performed at room temperature ∼24–48 h after cells were seeded on coverslips using an IX73 inverted fluorescence microscope (Olympus, Japan) equipped with an iXon Ultra electron-multiplying CCD (EM-CCD) camera (Oxford Instruments, UK) as previously described (Alsina et al., 2017; Park et al., 2012). An Olympus IX-73 inverted fluorescence microscope with an oil-immersion objective (NA=1.49, 100×, Olympus) with a filter cube containing a dual bandpass dichroic mirror (ZT488/561rpc, Chroma, USA) and a dual bandpass emission filter (ZET488/561m, Chroma) was used. A 488 nm laser (Coherent Inc., USA) and a 561 nm laser (Coherent Inc.) were used to excite mClover3 and mRuby3, respectively, using the same power of lasers. The images of donors and acceptors were finally acquired on a custom-made dual-view based FRET setup, in which a dichroic mirror (T535lp, Chroma) was placed near the microscope side-port to divide the emission signals into two spectral-distinct parts (donor or acceptor channel). Throughout the imaging experiments, cells were superfused either with Ca²⁺-free Tyrode solution (129 mM NaCl, 5 mM KCl, 1 mM MgCl2, 30 mM glucose, and 25 mM HEPES pH 7.4) containing 1 μM thapsigargin, or with a Tyrode solution containing 2 mM Ca²⁺ without thapsigargin.

Student's t-test was used to determine statistical significance between two groups. All results were analyzed with GraphPad Prism5 (GraphPad Software). Results are presented as means±s.e.m. A P-value of <0.05 is considered significant. P-values are presented as *P<0.05, **P<0.01 and ***P<0.001.

Authors thank K.M.K., Y.C.K., Y.Y.L., K.H.J. and S.G.M. for reading and comments for clarifying the manuscript, and thank J.H.H. of UOBC for advice and technical support.