Calmodulin dissociates the STIM1-Orai1 complex and STIM1 oligomers

Store-operated calcium entry (SOCE) is a major pathway for calcium ions influx into cells and has a critical role in various cell functions. Here we demonstrate that calcium-bound calmodulin (Ca2+-CaM) binds to the core region of activated STIM1. This interaction facilitates slow Ca2+-dependent inactivation after Orai1 channel activation by wild-type STIM1 or a constitutively active STIM1 mutant. We define the CaM-binding site in STIM1, which is adjacent to the STIM1–Orai1 coupling region. The binding of Ca2+-CaM to activated STIM1 disrupts the STIM1–Orai1 complex and also disassembles STIM1 oligomer. Based on these results we propose a model for the calcium-bound CaM-regulated deactivation of SOCE.

A s an intracellular second messenger, Ca 2+ has a pivotal role in a broad range of biological processes 1 . Storeoperated calcium entry (SOCE) is the major extracellular Ca 2+ influx process in excitable and, particularly, in non-excitable cells 2,3 . Previous studies have identified the molecular constituents of SOCE: an endoplasmic reticulum (ER) Ca 2+ sensor, the stromal interaction molecule (STIM) [4][5][6] and a plasma membrane (PM) pore-forming subunit, Orai [7][8][9] . By sensing Ca 2+ depletion within the ER, STIM undergoes oligomerization and translocates to ER-PM junctions, where it activates highly Ca 2+ -selective Orai channels 10 . Influx of Ca 2+ through Orai channels provides precise Ca 2+ signals that are essential for regulating long-term responses such as gene expression and cell growth 11 .
Ca 2+ entry through Orai channels modulates their activity through a process called Ca 2+ -dependent inactivation (CDI), which is an important negative feedback mechanism that maintains intracellular Ca 2+ homeostasis. CDI consists of fast CDI (FCDI) and slow CDI (SCDI), which have distinct kinetics and sites of action 2,12 . FCDI occurs within tens of milliseconds after channel activation and is controlled by Ca 2+ binding to a site located several nanometers from the pore 13,14 . The FCDI process was regulated by various factors, including the STIM1-to-Orai1 expression ratio 15 , the intracellular loop of Orai1 and the conserved negatively charged cluster in the STIM1 COOH terminus [16][17][18] . SCDI occurs within tens of seconds of channel activation and is driven by a global rather than local Ca 2+ rise 14,19,20 . The SOCE-associated regulatory factor (SARAF) has been reported to facilitate SCDI of SOCE 21 through interaction with the STIM1 C-terminal inhibitory domain (448-530) 22 . Interestingly, SARAF was not able to regulate SCDI after Orai1 channel activation by the constitutively active STIM1 mutant D76A 22 . Moreover, SARAF-dependent SCDI appears to be regulated by multiple factors, such as Caveolin, E-Syt1, Septin4 and PI(4,5)P 2 23 . Calmodulin (CaM) consists of an N-lobe and a C-lobe, each of which contains two Ca 2+ -binding sites. CaM adopts a closed conformation in the Ca 2+ -free state (apo-CaM) and an open conformation in the Ca 2+ -bound state (holo-CaM) 24 . Both states are able to bind various target proteins and consequently transduce different Ca 2+ signals 25 . Ion channels are major targets of CaM regulation. For example, the CDI of voltage-gated Ca 2+ channels regulated by CaM is well established 26 . Apo-CaM is preassociated with the IQ domain, whereas Ca 2+ influx induces subtle conformational rearrangements between CaM and the Ca 2+ -inactivation region, which facilitates CDI. For Orai channels, one previous study has shown that CaM might be the Ca 2+ sensor for FCDI 27,28 but the results of a more recent report have contradicted this argument 29 .
After more than 30 years of intensive studies, the activation mechanism of SOCE has been well documented 10,11,30,31 . However, the deactivation mechanism is largely unknown. In this study, we demonstrated that CaM associates with the core region of STIM1 in a Ca 2+ -dependent manner. Consequently, calciumbound CaM disrupts the STIM1-Orai1 complex and disassembles STIM1 oligomers, thereby inducing deactivation of the SOCE.

Results
Ca 2+ -CaM interacts with STIM1. Human STIM1 is a type I membrane protein that contains 685 amino acids. Its N-terminal region is located inside the ER and contains an EF-SAM domain, which senses Ca 2+11 . Its C-terminal region is located in the cytosolic portion of the protein and contains the core domain SOAR/CAD/OASF/Ccb9, which associates with the Orai1 channel [32][33][34][35] (for convenience, the term SOAR is used hereafter).
To investigate whether CaM is able to interact with the cytosolic C-terminal regions of STIM1 molecules, we designed various human STIM1 constructs and transiently transfected HEK293T cells. CaM-Sepharose beads were used for a pull-down experiment. We found that two constructs (SOAR and STIM1-c2) interacted with CaM in a Ca 2+ -dependent manner. Neither STIM1-c1, which perturbed SOAR formation nor STIM1-c3, STIM1-c4 and STIM1-c5, which contain residues after SOAR, associated with CaM (Fig. 1a), thus suggesting that both the correct SOAR conformation and exposed SOAR are necessary for the association of the C-terminal region of STIM1 with calciumbound CaM (Ca 2+ -CaM).
To measure the binding affinity of SOAR for CaM, recombinant SOAR and CaM proteins were purified, and isothermal titration calorimetry (ITC) experiments were carried out. Our results showed that in the presence of Ca 2+ , CaM interacted with SOAR with high affinity (K d = 0.23 µM). In the absence of Ca 2+ , CaM did not bind to SOAR. As a control, we also used CaM mutants (CaM-4EF) in which glutamate was mutated to glutamine to diminish Ca 2+ loading at each Ca 2+ -binding site 36 . The titration of CaM-4EF into SOAR in the presence of Ca 2+ showed no detectable binding signal (Fig. 1b). These results indicated that SOAR directly and tightly associates with Ca 2+ -CaM in vitro.
We then performed co-immunoprecipitation (co-IP) and fluorescence resonance energy transfer (FRET) experiments to verify that the interaction between SOAR and CaM occurs in cells. When coexpressed with myc-SOAR, CaM-YFP was able to pull down SOAR in the presence of Ca 2+ but not in its absence (Fig. 1c, d). Furthermore, when coexpressed with myc-CaM, the ER-localized STIM1 construct (1-444), which contained the entire SOAR, pulled down overexpressed CaM (Fig. 1e) and, more importantly, endogenous CaM (Fig. 1f) after cells were treated with thapsigargin. When co-expressed with Orai1, Ca 2+ influx induced significant apparent FRET efficiency (Eapp) increases between SOAR-YFP and CaM-CFP but not between SOAR-YFP and the mutant CaM-4EF-CFP (Fig. 1g). These results indicated that Ca 2+ is essential for the association of SOAR with CaM in vivo. Together, our results indicated that SOAR interacts with Ca 2+ -CaM both in vitro and in vivo.
Ca 2+ -CaM facilitates the SCDI of Orai1. To further understand the physiological significance of the Ca 2+ -dependent interaction between CaM and SOAR, we used electrophysiology to determine whether CaM participates in the regulation of Orai1 SCDI. Our results showed that including purified CaM in the pipette solution increased the extent of the SCDI of Orai1 current that was induced by SOAR and the extent of inactivation in a CaM concentration-dependent manner ( Fig. 2a and Supplementary  Fig. 1a). Including purified CaM-4EF mutant in the pipette solution had no effect on the Orai1 current ( Fig. 2b and Supplementary Fig. 1b). In the control experiment, 10 mM EGTA abolished the CaM-regulated SCDI of Orai1 current (Supplementary Fig. 2). Next, we tested whether CaM regulated the SCDI of Orai1 current that was induced by full-length STIM1. We found that CaM but not the CaM-4EF mutant facilitated the Orai1 SCDI (Fig. 2c, d and Supplementary Fig. 1c, d). We further tested whether CaM regulated the SCDI of Orai1 current that was induced by the constitutively active STIM1 mutant D76A. Our results showed that CaM, but not a CaM-4EF mutant, produced substantial increases in the SCDI of the Orai1 current ( Fig. 2e and Supplementary Fig. 1e). As CaM is notoriously difficult to knock down in cells, we instead used CaM inhibitors. W-7, a canonical chemical CaM inhibitor 37 Fig. 1f). We then used another type of CaM inhibitor to further verify this result. The Trp peptide (RRKWQKTGHAV-RAIGRL), a CaM peptide inhibitor 38 , was synthesized and included in the pipette solution. The Trp peptide significantly inhibited the SCDI of Orai1 current that was induced by the STIM1 WT ( Fig. 2g and Supplementary Fig. 1g). In summary, these data suggested that CaM promotes the SCDI of the Orai1 channel.
L390 and F391 are important in Ca 2+ -CaM binding. To address the molecular mechanism of CaM regulation of Orai1 SCDI, we sought to pinpoint the key residues that are responsible for the interaction between SOAR and Ca 2+ -CaM by using highresolution nuclear magnetic resonance (NMR) spectroscopy. Multiple human STIM1 constructs were screened, and one (STIM1 363-416) produced well-resolved peaks in the NMR 1 H-15 N HSQC spectrum. Each residue was assigned unambiguously ( Supplementary Fig. 3a). Titration of Ca 2+ -CaM into the solution of 15 N-STIM1 significantly attenuated the 15 N-HSQC cross-peaks for residues 382-398, thus indicating that these residues may be involved in Ca 2+ -CaM binding ( Fig. 3a and Supplementary Fig. 3b). This region contains several basic residues and several hydrophobic residues, which have been reported to regulate STIM1-Orai1 binding and Orai1 channel activation 30,39,40 . Within this region, residues L390 and F391 are good candidates to be further verified using site-directed mutagenesis, because Ca 2+ -CaM typically exposes the hydrophobic pocket for target binding 24 . The 15 N-HSQC experiment indicated that the interaction of Ca 2+ -CaM with the 15 N-labeled WT STIM1 (363-416) construct occurred with a slow-to-intermediate exchange rate, and that many cross-peaks were broadened with the addition of Ca 2+ -CaM ( Fig. 3b and Supplementary  Fig. 3b), thus indicating that these two proteins bind each other strongly (K d < 10 μM) 41 . In contrast, titrating Ca 2+ -CaM into an 15 N-labeled STIM1 (363-416) mutant, L390S-F391S, resulted in a faster exchange rate (Fig. 3b), thus indicating that mutation of L390 and F391 significantly reduced the binding affinity of the proteins. Using the cross-peak shifts of three residues (A369, G379 e Normalized averaged whole cell current at −100 mV measured in HEK293 cells that express Orai1 and the STIM1-D76A mutant alone, with CaM WT or with the CaM-4EF mutant at the indicated concentration. f, g Normalized averaged whole cell current at −100 mV measured in HEK293 cells expressing Orai1 and STIM1 WT with inhibitors W-7 and the TRP peptide. W-7 and the TRP peptide were introduced into the pipette solution at the indicated concentration. For all current recordings, 2 mM extracellular Ca 2+ was slowly diffused to the cell surface using a perfusion system. The number in the parenthesis indicated the number of cells that were analyzed. Error bars denote SEM and I383) as a reference, we determined the binding affinity of the STIM1 (363-416) mutant L390S-F391S with Ca 2+ -CaM to be 97 ± 10 μM, at least 10-fold lower than that of the WT (Fig. 3c, d).
It is noted that the dissociation constants that were determined by the chemical shift changes vary among different residues, perhaps reflecting the variation in local conformational changes in the association of STIM1 (363-416) with Ca 2+ -CaM.
We further tested the interaction between the intact SOAR domain and Ca 2+ -CaM by using CaM-Sepharose bead pull-down and co-IP experiments. Our results showed that the SOAR mutant L390S-F391S (LF-SS) eliminated the interaction of SOAR with Ca 2+ -CaM ( Fig. 3e and Supplementary Fig. 6a), which also occurred in the STIM1 (1-444) mutant L390S-F391S (Supplementary Fig. 6b). In summary, residues L390 and F391 are necessary sites for SOAR to bind to Ca 2+ -CaM.
We next asked whether the reduced affinity of the L390S-F391S mutant (SOAR-LF-SS) for Ca 2+ -CaM was due to an effect on SOAR dimer formation or the α-helix structural stability that is critical for SOAR function 30 Fig. 7a), far from the SOAR dimerization interface 30 . A size-exclusion chromatography (SEC)-coupled multiangle light scattering experiment showed that the molecular weights of WT SOAR and LF-SS mutant proteins were similar and corresponded to the size of a dimer ( Supplementary Fig. 7b, c). The co-transfection of C-terminal cyan fluorescent protein (CFP)-and yellow fluorescent protein (YFP)-labeled WT or mutant SOAR yielded comparable FRET Eapp values between WT and WT proteins, WT and mutant proteins, and mutant and mutant proteins ( Supplementary Fig. 7d). These data indicated that residues L390 and F391 are dispensable in SOAR dimerization.
To test whether the SOAR-LF-SS mutant affected SOAR's αhelix structural stability, circular dichroism was used. Our results showed that WT and mutant proteins are primarily α-helical, and there were no noticeable differences between the helicities (Supplementary Fig. 7e), which is consistent with the result that 15 N-STIM1 and 15 N-STIM1-L390S-F391S produced similar 15 N-HSQC spectra ( Supplementary Fig. 8). The thermal melting curves of the two variants also showed a similar Tm value (Supplementary Fig. 7f). These data indicated that the L390S-F391S mutant did not disturb the SOAR α-helix structural stability.
Ca 2+ -CaM disrupts SOAR-Orai1 interaction. Residues L390 and F391 are located adjacent to the polybasic patch (residues 382~387), which is the site where SOAR (and STIM1) binds to Orai1 30,40 . Therefore, the interference of residues L390 and F391 might affect SOAR-Orai1 binding. To test this hypothesis, we cotransfected Orai1-CFP and SOAR-YFP (WT or mutant) into HEK293T cells and measured the FRET Eapp values. We observed robust FRET signals between Orai1-CFP and the WT SOAR-YFP (SOAR-WT) at the PM. In contrast, the FRET signals between Orai1-CFP and the SOAR-YFP mutant (SOAR-LF-SS) were negligible (Fig. 4a, b). Moreover, live-cell confocal microscopy showed that SOAR-WT co-localized with Orai1 on the PM, whereas the mutant SOAR-LF-SS did not (Fig. 4c), thus indicating that the mutation of residues L390 and F391 abrogated the interaction of SOAR with Orai1. We further verified this reduced interaction by performing a co-IP experiment. When coexpressed with myc-Orai1, SOAR-WT (but not SOAR-LF-SS) was able to pull down myc-Orai1 (Fig. 4d). To test whether mutating residues L390 and F391 affected Orai1 channel activation, we measured intracellular Ca 2+ influx in HEK293T cells that expressed both SOAR-WT or the SOAR-LF-SS mutant and Orai1. SOAR-WT showed constitutive Ca 2+ influx, whereas the mutant SOAR-LF-SS did not induce Ca 2+ influx (Fig. 4e). These data suggested that residues L390 and F391 are important in SOAR-Orai1 binding.
Considering that residues L390 and F391 are also critical for Ca 2+ -CaM binding to SOAR, we tested whether Ca 2+ -CaM could directly disrupt the SOAR-Orai1 complex in vitro. GFP-Orai1 and myc-SOAR were co-transfected into HEK293T cells. The SOAR-Orai1 complex was purified by GFP-Trap and washed with purified CaM in the absence or presence of Ca 2+ . Our results showed that CaM-containing solution largely disrupted the SOAR-Orai1 complex in a Ca 2+ -dependent manner, compared with CaM-free solution (Fig. 4f).
We then investigated whether Ca 2+ -CaM affected the binding of SOAR to Orai1 by measuring the FRET Eapp value at the PM. Orai1-CFP, YFP-SOAR and CaM or CaM-4EF mutants were cotransfected into HEK293T cells. FRET Eapp values at the PM were measured after extracellular Ca 2+ was added. Compared with the control, the overexpression of WT CaM clearly decreased the FRET Eapp value, whereas the overexpression of the CaM-4EF mutant resulted in a similar FRET Eapp value (Fig. 4g). To inhibit endogenous CaM, Orai1-CFP and YFP-SOAR were co-transfected into HEK293T cells, and cells were treated with the inhibitor W-7 or a dimethyl sulfoxide (DMSO) control. After 5 min of extracellular Ca 2+ influx, the FRET Eapp value at the PM was significantly higher in W-7-treated cells than in DMSO-treated cells (Fig. 4h). These data together suggested that Ca 2+ -CaM dissociates the SOAR-Orai1 complex.
Ca 2+ -CaM disrupts the STIM1-Orai1 interaction. SOAR is the minimum activation domain of STIM1 that can bypass ER Ca 2+ -depletion and bind and activate Orai1 33 . Residues L390 and F391 are important in the SOAR-Orai1 interaction. We asked whether residues L390 and F391 are also important in the interaction of full-length STIM1 with Orai1. Live-cell confocal microscopy was used to monitor distribution changes of fulllength STIM1-GFP (WT and mutant) between the resting and activation states when it was co-expressed with mCherry-Orai1. In the resting state, both the STIM1-WT and STIM1 mutant (STIM1-LF-SS) were distributed across the ER, whereas Orai1 was uniformly distributed on the PM (Fig. 5a). After thapsigargin stimulation, STIM1-WT translocated and co-localized with Orai1. In contrast, the mutant STIM1-LF-SS did not co-localize with Orai1 after thapsigargin stimulation (Fig. 5a). We further transiently transfected STIM1-YFP (WT or mutant) into the HEK-Orai1-CFP-stable cells and then measured FRET signals. Ca 2+ depletion from the ER induced an approximately twofold increase of FRET Eapp between Orai1 and STIM1-WT, whereas no increase was observed between Orai1 and STIM1-LF-SS (Fig. 5c). Figure 5b shows the FRET Eapp change for the same cell before and after thapsigargin stimulation. We also verified the interaction by using a co-IP experiment. When co-expressed with myc-Orai1, STIM1-WT but not STIM1-LF-SS associated with myc-Orai1 after thapsigargin treatment (Fig. 5e). In HEK293Tcells, extracellular Ca 2+ influx was abolished with the coexpression of Orai1 and the mutant STIM1-LF-SS (Fig. 5d). These data showed that residues L390 and F391 are also necessary for STIM1-Orai1 binding and Orai1 activation.
We then tested whether Ca 2+ -CaM could disrupt the STIM1-Orai1 interaction. GFP-Orai1 and STIM1-myc were cotransfected into HEK293T cells. The cells were treated with ionomycin, and the STIM1-Orai1 complex was then purified by GFP-Trap. We found that CaM was able to wash STIM1 off the STIM1-Orai1 complex in a Ca 2+ -dependent manner (Fig. 5f). To further determine whether Ca 2+ -CaM affects the interaction of STIM1 with Orai1, Orai1-CFP, STIM1-YFP and WT CaM or the CaM-4EF mutant were co-transfected into HEK293T cells, and the FRET Eapp values were measured after 5 min of ionomycin treatment. As expected, the overexpression of WT CaM but not the CaM-4EF mutant largely decreased the FRET values compared with those of the control (Fig. 5g). When endogenous CaM was inhibited by the W-7 inhibitor, the FRET Eapp value increased significantly compared with that of the DMSO control (Fig. 5h). Together, these data indicated that Ca 2+ -CaM dissociates the STIM1-Orai1 complex.
Ca 2+ -CaM disassembles the STIM1 oligomer. Puncta formation at ER-PM junctions is a hallmark and key step for STIM1 activation [46][47][48] . We therefore determined whether mutation of residues L390 and F391 would affect the formation of STIM1 puncta. We monitored puncta formation by using live-cell confocal microscopy in HEK293T cells that expressed STIM1-GFP (WT or mutant) alone or with mCherry-Orai1. In cells expressing STIM1 alone, the WT and mutant were distributed across the ER in resting cells (Fig. 6a, upper panel). After stimulation by thapsigargin, STIM1 WT translocated and formed puncta, whereas no puncta were observed for cells with the mutant STIM1-LF-SS (Fig. 6a, lower panel). In cells that co-express Orai1, after thapsigargin stimulation, the STIM1 WT formed puncta that were co-localized with Orai1 near the PM. The mutant STIM1-LF-SS neither changed distribution nor colocalized with Orai1 (Fig. 6b).
STIM1 exists as a dimer in the resting state, and Ca 2+ depletion from ER store induces further oligomerization 2,46 . The STIM1-LF-SS mutant did not form puncta, thus suggesting that this mutant is unable to self-oligomerize after store depletion. To further examine this hypothesis, we used FRET microscopy h Steady-state FRET between Orai1-CFP and YFP-SOAR on the PM of HEK293T cells that were pretreated with 30 μM W-7 or DMSO control for 30 min. FRET values were measured after 2 mM extracellular Ca 2+ was added. Numbers of cells that were analyzed are indicated. *P < 0.05, **P < 0.01 (unpaired Student's t-test). Error bars denote SEM NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-01135-w ARTICLE methods. We co-transfected STIM1-CFP and STIM1-YFP (WT or mutant) into HEK293T cells. In the resting state, the STIM1 WT and mutant STIM-LF-SS showed similar FRET Eapp values. These values were higher than the control values, thus indicating that the mutation of residues L390 and F391 did not impair STIM1 dimer formation (Fig. 6c). Ca 2+ depletion from ER stores induced a robust increase in the FRET Eapp value for the STIM1 WT, owing to STIM1 oligomerization, in agreement with results from previous reports 44,46,49 . In contrast, the FRET Eapp value remained similar for the mutant STIM1-LF-SS, thereby indicating that the mutation of residues L390 and F391 prevented STIM1 oligomerization (Fig. 6c).
As residues L390 and F391 are important in STIM1 oligomerization, we hypothesized that Ca 2+ -CaM might have the capability to disassemble the STIM1 oligomer. STIM1-YFP and STIM1-myc were co-transfected into HEK293T cells. After ionomycin treatment, STIM1 oligomers were purified by GFP-Trap and incubated with Ca 2+ -CaM. We found that Ca 2+ -CaM largely reduced STIM1 oligomerization in our in vitro system (Fig. 6d). To investigate whether Ca 2+ -CaM would affect STIM1 oligomerization in cells, STIM1-CFP, STIM1-YFP and WT CaM or CaM-4EF mutants were co-transfected into HEK293T cells.   FRET values, whereas the overexpression of CaM-4EF mutant did not (Fig. 6e). The inhibition of endogenous CaM by W-7 led to FRET values that increased compared with those of the DMSO control (Fig. 6f). These data suggest that Ca 2+ -CaM can disassemble the STIM1 oligomer.

Discussion
SOCE is a major cell signaling path that has an essential role in various cell activities 3,11,50 . The process by which ER resident protein STIM1 senses Ca 2+ changes inside the ER and consequently regulates the PM-located Ca 2+ channel Orai1 has attracted the attention of many researchers for many decades. The molecular mechanism of STIM1 activation coupling to Orai1 has been elucidated, whereas the process from the activated STIM1-Orai1 complex to the resting state remains elusive. Our studies revealed that calcium-bound CaM associates with an activated STIM1 in which SOAR is exposed and consequently dissociates the STIM1-Orai1 complex and the STIM1 oligomer, thus reversing the process by which STIM1 activates Orai1. CaM has a crucial role in maintaining Ca 2+ homeostasis by regulating many Ca 2+ channels and pumps, such as voltagedependent Ca 2+ channels and PM Ca 2+ -ATPase 51, 52 . Our results showed that CaM interacts with SOAR in a Ca 2+ -dependent manner in vitro and in vivo. In particular, the STIM1 construct (1-444) was able to pull-down both overexpressed CaM and endogenous CaM after cells were stimulated by thapsigargin, thus providing good evidence that Ca 2+ -CaM has the capability to associate with exposed SOAR in vivo. Moreover, using a FRET technique, we observed that extracellular Ca 2+ influx induced CaM access to SOAR. The control CaM-4EF mutant was not able to access SOAR. These data suggest that after extracellular Ca 2+ influx causes cytosolic Ca 2+ elevation, CaM senses and loads with Ca 2+ and then accesses and binds to SOAR in activated STIM1.
Interestingly, the SOAR region does not contain a known CaM-binding motif. We then searched for the CaM-binding site in SOAR in a CaM target database 53 and found that the region around L390-F391 represents the most likely binding site. We further located the specific binding site through NMR titration and validated this binding site through immunoprecipitation methods. So far, two regions in STIMs were found to interact with Ca 2+ -CaM 54, 55 . One region is located in the STIM1 C-terminus and the other is located in the STIM2 SOAR domain. First, the very C-terminal polybasic region (667-685) in STIM1 was found through ITC and NMR methods to interact with Ca 2+ -CaM. This region (667-685) did not contain any known CaM-binding site and consists of several basic and hydrophobic residues. ITC experiments in which the polybasic peptide was titrated into CaM revealed a strong binding affinity between these two molecules (1 μM) 54 . Second, through pull-down and SPR methods, the region 459-482 in SOAR of STIM2, which corresponds with residues 368-391 in SOAR of STIM1, was found to interact with Ca 2+ -CaM 55 . This region is also rich in basic and hydrophobic residues and did not contain any known CaMbinding motif. Therefore, CaM seems to bind with STIM1 in an unconventional way.
SOAR is the basic unit of STIM1 that binds and activates the Orai1 channel. It has been demonstrated that the polybasic SOAR residues 384-387 are crucial for SOAR binding to Orai1. Interestingly, we identified residues L390 and F391 as necessary for SOAR binding to Ca 2+ -CaM. These residues are adjacent to the Orai1-binding site of SOAR, thus suggesting that Ca 2+ -CaM binding to SOAR will interfere with the SOAR-Orai1 complex or the STIM1-Orai1 complex. Furthermore, the affinity value that we measured between SOAR and Ca 2+ -CaM was~0.2 μM, 1,000-fold higher than the affinity (~200 μM) of SOAR for the Orai1 channel, according to previous reports 45,56 . This large difference in affinity allows Ca 2+ -CaM to compete with Orai1 for association with SOAR, thus leading to the breakdown of the SOAR-Orai1 complex or the STIM1-Orai1 complex.
Puncta formation represents STIM1 aggregation and activation. Our results showed that the mutation of residues Leu390 CaM undergoes a conformational change and binds to the exposed SOAR region of STIM1. The STIM1-Orai1 complex dissociates, the Orai1 channel is inactivated and the STIM1 oligomer is disassembled. As the cytosolic Ca 2+ concentration is decreased, CaM dissociates from STIM1. STIM1 returns to the resting state, owing to a filled ER Ca 2+ store and possibly other regulators and Phe391, which are part of the Ca 2+ -CaM binding site of SOAR or STIM1, leads to the failure of puncta formation and activation. Thus, we infer that Ca 2+ -CaM binding to the activated STIM1 oligomer may lead to a shift in equilibrium from an STIM1 oligomer to a less aggregated form. In our system that uses purified CaM, we indeed observed that the STIM1 oligomer became less aggregated after the addition of Ca 2+ -CaM. This result was further confirmed by FRET measurements in cells. The Orai1 channel is characteristic of both FCDI and SCDI. FCDI occurs within milliseconds, whereas SCDI occurs within minutes. Our results showed that after CaM responds to elevated cytosolic Ca 2+ levels, Ca 2+ -CaM disrupts the SOAR-Orai1 complex or the STIM1-Orai1 complex, resulting in the inactivation of the Orai1 channel. This process was consistent with the SCDI of the Orai1 channel. Indeed, including purified CaM in the pipette facilitated Orai1 SCDI. More importantly, when the Orai1 channel was activated by the constitutively active STIM1 mutant D76A, CaM was also able to facilitate SCDI. When inhibitors were used to inhibit endogenous CaM, the SCDI of the Orai1 current was reduced. Thus, Ca 2+ -CaM induced SCDI appears to differ from SARAF-induced SCDI. It has been reported that SARAF-induced SCDI may be mediated by the interaction of SARAF with residues 448-530 of STIM1, which are located after SOAR 22 . In addition, STIM1 and SARAF are localized in the ER and interact in the basal state with filled Ca 2+ stores 21, 23 , suggesting that SARAF functions by holding STIM1 in a resting state. This statement is consistent with data that SARAF does not inactivate the Orai1 channel if the constitutively active STIM1 mutant D76A or 4E-4A is used 22 . Therefore, we speculate that Ca 2+ -CaM has a dominant role in disrupting the STIM1-Orai1 complex and disaggregating the STIM1 oligomer. The inactive STIM1 conformation is presumably maintained by SARAF to avoid the spontaneous activation of STIM1. More evidence is required to test this possibility in the near future.
In summary, we propose a molecular mechanism of STIM1-Orai1 deactivation that is induced by Ca 2+ -CaM (Fig. 7). After Orai1 is activated by STIM1 and forms the STIM1-Orai1 complex, extracellular Ca 2+ influx through the Orai1 channel leads to the elevation of cytosolic Ca 2+ . As the cytosolic Ca 2+ increases to a certain level, CaM is charged with Ca 2+ and moves closer to the STIM1-Orai1 complex. Owing to the high affinity between Ca 2+ -CaM and the exposed SOAR, Ca 2+ -CaM pulls STIM1 from Orai1 and disrupts the STIM1-Orai1 complex. The Orai1 channel is then closed. Furthermore, the binding of Ca 2+ -CaM to STIM1 induces a change in the STIM1 oligomer that results in a less-aggregated form. By the action of ER Ca 2+ -ATPase and PM Ca 2+ -ATPase, the cytosolic Ca 2+ concentration is decreased, and the ER Ca 2+ store is refilled. Decreased cytosolic Ca 2+ concentrations lead to CaM dissociation from STIM1, whereas increased ER Ca 2+ concentrations load the N-terminal STIM1 region with Ca 2+ . Consequently, the N-terminal STIM1 returns to its monomeric form and presumably drives STIM1 back to the resting state. Of course, this process can be further regulated by many other factors, such as SARAF 23 , the orosomucoid-like 3 protein ORMDL3 57 and Golli 58 .
Our model is consistent with previously reported results. In RBL cells, SCDI is mainly regulated by cytosolic Ca 2+ concentration 14,19 , indicating that cytosolic Ca 2+ concentrations have a critical role in regulating the SCDI. Ca 2+ influx through the Orai1 channel negatively regulates channel activity by promoting dissociation of the STIM1-Orai1 complex and subsequent STIM1 deoligomerization to the resting state 46,59 . ER Ca 2+ replenishment is not sufficient to disassemble STIM1 oligomers and requires cytosolic Ca 2+ concentration elevation 59 . It has been reported that the ER protein ORMDL3 binds to and inhibits the SERCA pump, promoting cytosolic calcium accumulation and resulting in negatively regulated SOCE 60 . It is likely to be that increased cytosolic Ca 2+ levels activate CaM and facilitate the SCDI of the Orai1 channel. CaM has been proposed to be involved in FCDI through the interaction of CaM with the N-terminal residue W76 of the Orai1 channel 61,62 . A recent study has argued against this model because W76 is located in the pore of the Orai1 channel and is thus unlikely to be accessed by CaM 29 . Whether CaM is involved in FCDI may require further assessment. Thus, FCDI is beyond the scope of this manuscript.
In combination with previously described STIM1-Orai1 activation processes, our results now depict the entire SOCE at the molecular level. Ca 2+ has a central role in initiating, directing and terminating the SOCE.
Plasmids. Full-length complementary DNA of human STIM1 was inserted into pECFP-C1, pEGFP-C1 and pEYFP-C1 vectors (Clontech) between BglII and EcoRI sites to generate STIM1-CFP, STIM1-GFP and STIM1-YFP. The sequence of SOAR was amplified by PCR and cloned into pCMV-Myc (Clontech) vector between EcoRI and BglII sites to generate the Myc-SOAR. Full-length cDNA of human Orai1 was inserted into pmCherry-C1 vector (Clontech) between BglII and EcoRI sites to generate mCherry-Orai1. Full-length cDNA of human Orai1 was inserted into pCMV-Myc vector between EcoRI and BglII sites to generate Myc-Orai1. GFP-STIM1 truncations were generated by introducing stop codon at the specific position. Full-length cDNA of rat CaM was cloned into pET-33(b) vector (Novagen) between NcoI and XhoI. mCherry-Orai1, SOAR-GFP, myc-Orai1, myc-SOAR and MBP-SOAR have been previously described 30 . Mutations were generated by using the Quick-change mutagenesis kit (Agilent). Primers used in this study were listed in Supplementary Table 1.
Protein expression and purification. Human SOAR WT and LF-SS mutant were expressed in Escherichia coli. The cells were collected and suspended in lysis buffer (20 mM MES pH 6.0, 1 mM EGTA and 300 mM NaCl) and lysed by sonication. After centrifugation at 15,000 g for 1 h, the supernatant was loaded onto a MBP-Trap HP column (GE Healthcare) equilibrated with equilibration buffer (20 mM MES pH 6.0, 300 mM NaCl and 1 mM EGTA). The target protein was eluted using elution buffer (20 mM MES pH 6.0, 300 mM NaCl, 1 mM EGTA, 10 mM Maltose) and cleaved with Tobacco Etch Virus protease overnight at 4°C. The protein was applied to a HiTrap SP column (GE Healthcare) equilibrated with 20 mM MES pH 6.0, 50 mM NaCl, 1 mM dithiothreitol (DTT) and 1 mM EGTA and eluted by increasing the NaCl concentration. Target protein was collected and further purified by using a Superdex 200 10/300 GL column (GE Healthcare). Protein concentrations were determined using a BCA assay kit (Pierce). The purification of the LF-SS mutant followed the same protocol for the SOAR WT.
CaM and CaM-4EF were expressed in E. coli. The cells were collected and suspended in lysis buffer (50 mM Tris-HCl pH 7.5, 1 mM EGTA and 1 mM DTT) and lysed by sonication. For the WT CaM, after centrifugation at 15,000 g for 1 h, the supernatant was supplemented with NaCl and CaCl 2 to a final concentration of 500 mM NaCl and 5 mM CaCl 2 . The supernatant was loaded onto a HiPrep Phenyl FF (High sub) 16/10 column (GE Healthcare) equilibrated with equilibration buffer (50 mM Tris-HCl pH 7.5, 500 mM NaCl and 5 mM CaCl 2 ). Protein was eluted using elution buffer (50 mM Tris-HCl pH 7.5, 1 mM EGTA) and further purified by using a HiLoad 26/600 Superdex 200 column (GE Healthcare). For CaM-4EF, the supernatant was applied to a HiPrep Q FF 16/10 column (GE Healthcare) equilibrated with 50 mM Tris-HCl pH 7.5, 50 mM NaCl, 1 mM DTT and 1 mM EGTA. Protein was eluted by increasing the NaCl concentration. Target protein was collected and further purified by using a HiLoad 26/600 Superdex 200 column (GE Healthcare). Protein concentrations were determined using a BCA assay kit (Pierce).
The STIM1(363-416) construct was expressed in E. coli. The cells were collected and suspended in lysis buffer (50 mM Tris-HCl pH 7.5, 300 mM NaCl and 0.1 mM phenylmethane sulfonyl fluoride) and lysed by sonication. After centrifugation at 15,000 g for 1 h, the supernatant was loaded onto an Ni-NTA affinity column and washed extensively with lysis buffer. The target protein was eluted using elution buffer (50 mM Tris-HCl pH 7.5, 300 mM NaCl and 300 mM imidazole) and cleaved with PreScission protease overnight at 4°C. The protein was applied to a HiTrap SP column (GE Healthcare) equilibrated with 50 mM Tris-HCl pH 7.5, 50 mM NaCl, 1 mM DTT and 1 mM EGTA, and eluted by increasing the NaCl concentration. Target protein was collected and further purified by using a Superdex Peptide 10/300 GL column (GE Healthcare). Protein concentrations were determined using a BCA assay kit (Pierce). Mutants were expressed and purified by following the protocol for STIM1 (363-416).
For the 15 N-or 13 C-labeled protein, protein was expressed and purified by following the same protocol used for the unlabeled protein, whereas the Lysogeny broth (LB) medium was replaced with M9 medium supplemented with 15 N-NH 4 Cl or 13 C-glucose.
CaM pull-down assay. Transfected cells were washed twice with Tris-buffered saline (TBS) and lysed in chilled buffer containing 20 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1% Triton X-100, protease inhibitor cocktail (Roche) and 2 mM CaCl 2 or 1 mM EGTA. Cell lysates were centrifuged at 15,000 g for 30 min at 4°C. The supernatant was quantified with a BCA assay kit (Pierce). One milligram of cell lysate (1 mg ml −1 ) was incubated with 30 μl prewashed CAM-Sepharose 4B beads (GE Healthcare) for 2 h in 4°C. The beads were washed six times with lysis buffer and eluted in 30 μl gel-loading buffer by heating at 95°C for 10 min. Then, 10 μl of the elution was loaded onto 12-15% SDS-polyacrylamide gel electrophoresis (PAGE) gels or 10-16% tricine-SDS-PAGE (for small molecular weight protein) 63 and transferred to a polyvinylidene difluoride (PVDF) membrane for western blot analysis.
Intracellular Ca 2+ measurement. HEK293T was plated on glass-bottom dishes coated with poly-L-lysine (Sigma-Aldrich) and transfected at 60% confluence. Transfected cells were loaded with 2 μM Fura-2 AM (Invitrogen) in Ringer solution at room temperature for 30 min. After loading, cells were transferred to Fura-2 AM-free solution for 30 min. The Ringer solution contained (in mM): 140 NaCl, 5 KCl, 1 MgCl 2 , 10 D-glucose and 20 HEPES-NaOH, pH 7.4. Fluorescence imaging was undertaken at room temperature using a Leica DMI6000B microscope with a × 63 and 1.40 numerical aperture oil-immersion objective lens controlled by LAS software. Consecutive excitation occurred at 340 and 380 nm every 2 s, and emission was collected at 510 nm. Intracellular Ca 2+ concentration is shown as the 340/380 ratio. Data are shown as the mean ± SEM. FRET measurements. HEK293T or HEK Orai1-CFP were plated on glass-bottom dishes and transfected with ECFP (donor)-and EYFP (acceptor)-tagged constructs at 50% confluence. Imaging was performed at room temperature using a Leica DMI6000B microscope equipped with high-speed fluorescence-external filter wheels for a CFP-YFP index-based FRET experiment. Images were collected every 10 s using a ×63 oil objective (N.A.1.4; Leica) and processed using the Biosensor Processing Software 2.1 package in MATLAB R2014a. FRET was determined by pixel-by-pixel calculation using apparent FRET efficiency calculation methods 64 . Briefly, bleed-through and direct excitation containment were removed according to the equation Fc = I DA − aI AA − dI DD (assuming b = c = 0). The bleed-through factor "d" was determined for cells expressing ECFP alone (d = 0.715, 92 cells). The direct excitation factor "a" was determined for cells expressing EYFP alone (a = 0.173, 109 cells). The apparent FRET efficiency (Eapp) was calculated by the equation Eapp = Fc/(Fc + GI DD ), where Eapp represents the fraction of donor (ECFP) exhibiting FRET. The microscope-specific constant G was determined by using the D1ER cameleon (G = 3.458, 57 cells).
Solution NMR spectroscopy. All NMR spectra were recorded at 298 K with a Bruker Avance 600 MHz NMR spectrometer equipped with a QCI-cryoprobe. Protein samples were in 20 mM MES and 5 mM CaCl 2 at pH 6.5 unless otherwise indicated. The NMR data were processed with Topspin 2.1 and chemical shifts were assigned with the Sparky and Cara packages. The protein backbone assignments were completed with 3D CBCANH and CBCA(CO)NH spectra.
The interaction between CaM and STIM1 was analyzed by 15 N-HSQC spectra. The binding experiment was performed with 15 N-HSQC spectra of 15 N-STIM1 titrated by CaM. Cross-peak intensity of STIM1 was compared in the presence or absence of CaM. The chemical shift perturbations of STIM1 were plotted with CaM and the dissociation constant (K d ) was simulated with Origin.
Electrophysiological measurements. HEK293 cells were transiently transfected with CFP-Orai1 and SOAR-YFP, STIM1-YFP or the STIM1-YFP mutant in a ratio of 1:2 for 12-24 h at 37°C with 5% CO 2 . After trypsin digestion, transfected cells were plated onto 35 mm dishes with culture medium for at least 1 h to attach cells for electrophysiology. Experiments were performed at room temperature by using the tight-seal whole-cell configuration with the following voltage protocol: 50 ms voltage step to −100 mV from a holding potential of 0 mV, followed by a voltage ramp increasing from −100 to + 100 mV in 50 ms with a frequency of 0.5 Hz. To measure the SCDI, we used a pipette solution with 135 mM Cs aspartate, 8 mM MgCl 2 , 1.2 mM EGTA and 10 mM Cs-HEPES with CaM or its mutant or inhibitors (pH 7.2). The extracellular 0 Ca 2+ solutions contained the following: 130 mM NaCl, 4.5 mM KCl, 22 mM MgCl 2 , 10 mM TEA-Cl, 10 mM D-glucose and 5 mM Na-HEPES (pH 7.4). The extracellular Ca 2+ -containing solutions contained the following: 130 mM NaCl, 4.5 mM KCl, 20 mM MgCl 2 , 2 mM CaCl 2 , 10 mM TEA-Cl, 10 mM D-glucose and 5 mM Na-HEPES (pH 7.4). After whole-cell configuration in the 0 Ca 2+ extracellular solution, the first whole-cell currents that were recorded were denoted leak currents. Then, the 0 Ca 2+ extracellular solution was exchanged slowly with 2 mM Ca 2+ extracellular solution via a perfusion system to record Orai1 currents. Whole-cell currents were amplified with an Axopatch 700B and digitized with a Digidata 1550A system (Molecular Devices). All currents were low-pass filtered at 2 kHz and sampled at 10 kHz. pCLAMP software (Molecular Devices) was used for data acquisition and analysis. To reflect the SCDI process, the current was normalized as follows: the peak current density (pA pF −1 ) was set as −1.0 and the current densities during the measurement were displayed as the ratio of the peak current density to get normalized current densities and were plotted as a function of the recording time. Data points represent the mean ± SEM.
Competitive binding assay. Transfected cells were washed twice with TBS and lysed in chilled buffer containing 20 mM Tris-HCl (pH 7.5), 150 mM NaCl, 0.5% NP-40, protease inhibitor cocktail (Roche) and 2 mM CaCl 2 or 1 mM EGTA. For the STIM1-Orai1 and STIM1-STIM1 complexes, 3 μM ionomycin was used to induce Ca 2+ store depletion before lysis. Cell lysates were centrifuged at 15,000 g for 30 min at 4°C. The complex was pooled with 40 μl GFP-Trap (ChromoTek) for 2 h in 4°C. After samples were washed twice with lysis buffer, 0.3 mM Ca 2+ + CaM or EGTA + CaM was added to the complex for binding for 4 h in 4°C. The beads were washed three times with lysis buffer and eluted in 40 μl gel-loading buffer by heating at 95°C for 10 min. Then, 10 μl of the elution was loaded onto 7-15% SDS-PAGE gels or 10-16% tricine-SDS-PAGE (for low molecular-weight protein) 63 and transferred to a PVDF membrane for western blot analysis.
Isothermal titration calorimetry. Measurements were performed at 16°C using an ITC-200 microcalorimeter (GE Healthcare). Samples were buffered with 25 mM Tris-HCl (pH 7.5) containing 200 mM NaCl and 2 mM CaCl 2 or 1 mM EGTA. To determine the binding affinity of SOAR and CaM, 500 μM CaM or CaM-4EF was injected into 50 μM SOAR. Data were analyzed in the Microcal Origin software.
Circular dichroism spectroscopy. Protein secondary structure was recorded in a MOS-450 CD spectrometer (BioLogic) equipped with a temperature controller. Spectral data were measurement at 22°C using a 1 mm path length quartz cell with 10 μM protein between wavelength of 190-260 nm and using buffer for blank correction.
Multi-angle light scattering. Multi-angle light scattering was performed in-line with SEC using a DAWN HELLOS α instrument equipped with an Optilab Refractive Index Detector (Wyatt Technology). Protein (100 μM) was injected into a Superdex 200 10/300 column (GE Healthcare) equilibrated with 25 mM Tris-HCl pH 7.5, 200 mM NaCl at a flow rate of 0.5 ml min −1 . Protein concentration was calculated from the absorbance at 280 nm and the light scattering data were collected at 663 nm. Molecular weight was calculated using the ASTRA software.
Statistical analysis. Statistical significance was determined using unpaired Student's t-test with SPSS (IBM) software. No specific randomization or blinding protocols were used. Results are expressed as the mean ± SEM of at least three independent experiments. Differences with a P-value of < 0.05 were considered significant: *P < 0.05, **P < 0.01 and ***P < 0.001. Data availability. The authors declare that the data supporting the findings of this study can be found within the paper and its Supplementary Information files, and are available from the corresponding author upon reasonable request.