Calcium sensing by the STIM1 ER-luminal domain

Stromal interaction molecule 1 (STIM1) monitors ER-luminal Ca2+ levels to maintain cellular Ca2+ balance and to support Ca2+ signalling. The prevailing view has been that STIM1 senses reduced ER Ca2+ through dissociation of bound Ca2+ from a single EF-hand site, which triggers a dramatic loss of secondary structure and dimerization of the STIM1 luminal domain. Here we find that the STIM1 luminal domain has 5–6 Ca2+-binding sites, that binding at these sites is energetically coupled to binding at the EF-hand site, and that Ca2+ dissociation controls a switch to a second structured conformation of the luminal domain rather than protein unfolding. Importantly, the other luminal-domain Ca2+-binding sites interact with the EF-hand site to control physiological activation of STIM1 in cells. These findings fundamentally revise our understanding of physiological Ca2+ sensing by STIM1, and highlight molecular mechanisms that govern the Ca2+ threshold for activation and the steep Ca2+ concentration dependence.

Typically two populations of cells are observed, a major fraction in which STIM2 EFSAM-TM-CC1 is predominantly localized in the plasma membrane, and a minor fraction with both plasma membrane and ER localization. c, Confocal micrograph of a representative cell, showing the progressive recruitment of mCherry-CAD to CC1 at the plasma membrane in response to sequential stepwise increases in extracellular Ca 2+ concentration. d, Ca 2+ concentration dependence of mCherry-CAD recruitment to CC1, plotted as the decrease in cytoplasmic mCherry-CAD (n=17). Figure 7. STIM2-2NQ fails to release SOAR/CAD in cells. a, ITC analysis of Ca 2+ binding to STIM2 EFSAM-GrpE, showing raw heat changes corresponding to 1 l injections of 10 mM Ca 2+ into a sample cell containing initially 100 M protein in the top panel, and integrated heat change after subtracting heats of dilution at varying Ca 2+ :protein molar ratios in the bottom panel. b, Schematic representation of the CC1-SOAR/CAD interaction assay. c-e, Confocal sections at the level of the nucleus of HeLa cells co-expressing WT STIM2(1-434)-GFP (c, upper panels) or the 2NQ variant (d, upper panels) or the D167A variant (e, upper panels), together with mCherry-CAD (middle panels, and at higher magnification in lower panels). Relocalization of mCherry-CAD was assessed after changing the extracellular solution from no Ca 2+ (left panels of each set) to 12 mM Ca 2+ (right panels of each set). Scale bars, 5 m. f, Bar graph of the fraction of total mCherry-CAD intensity in a circumferential region near the plasma membrane (PM) under the conditions of panels (c)-(e). The fluorescence signal in the circumferential 'plasma membrane' region may include a contribution from unbound cytoplasmic mCherry-CAD. Data from STIM2-WT, n = 11 cells; STIM2-2NQ, n = 10 cells; STIM2-D167A, n = 10 cells. ***, p < 0.05 for STIM2-WT; the differences for STIM2-2NQ and STIM2-D167A were not statistically significant. g, Bar graph showing the fold change in mCherry-CAD intensity near the plasma membrane, relative to its initial value, after changing extracellular solution from no Ca 2+ to 12 mM Ca 2+ . Figure 8. Additional data related to the CD measurements. a, The far-UV CD spectrum of EFSAM-GrpE exhibited a decrease in the negative bands characteristic of -helical secondary structure with increasing temperature in the range 25C to 65C. MRE is the mean residue ellipticity. b, The spectrum of GrpE over the same temperature range exhibits little change. The MRE values determined for GrpE have been multiplied by the scaling factor 174/370-which is (number of residues in GrpE) / (number of residues in EFSAM-GrpE)-to depict the expected contribution of GrpE to the CD spectrum of the full fusion protein. The concentration of protein in each sample was 5 M. Figure 9. EFSAM remains structured despite loss of Ca 2+ . a, Graphic diagram of the HDX-MS work-flow used in this study. b, Deuterium exchange plots of the full repertoire of observed EFSAM-GrpE peptides, including GrpE peptides, in 0 Ca 2+ and 30 M Ca 2+ , at 0.5, 1, 2 and 5 min time points. c, Differences in deuterium exchange for the same pairs of conditions examined in Fig. 5i with data plotted for the complete set of time points (0.5, 1, 2 and 5 min). Low-exchanging peptides in Ca 2+ -bound condition from the EFSAM region around EVIQWLIT are boxed (green dotted box). Peptic peptides derived from GrpE are enclosed by a red dotted box. Note that the conformational change of wildtype EFSAM propagates to the distant -sheet domain of GrpE. Deuterium exchange of GrpE fused to EFSAM(D76A) resembles that for Ca 2+ -free wildtype EFSAM-GrpE, and exchange of GrpE fused to EFSAM(2NQ) resembles that for Ca 2+ -bound wildtype EFSAM-GrpE, establishing that H-D exchange in the -sheet domain of GrpE is sensitive to the conformation of EFSAM rather than to the absence or presence of Ca 2+ . We have not explored the basis for the propagated conformational change, but it is likely that the EFSAM rearrangement exerts this effect through the contiguous -helices of GrpE. It is known that the N-terminal -helices of native GrpE are physically coupled to the -sheet domain in E coli GrpE, and that thermal sensing in T thermophilus GrpE likewise involves a conformational change that increases protease sensitivity in both regions [1][2][3] . Data are the averages of triplicate measurements. d, EFSAM structure (PDB ID: 2K60) colored according to the change in deuterium exchange rates in the WT protein between the Ca 2+ -free state and the Ca 2+ -bound state. Two rotational views shown. Tryptophan residues are marked.  Table 1 Primers utilized for PCR amplification and mutagenesis.

More detailed characterization of STIM1-2NQ
STIM1-2NQ, expressed at moderate levels, had an ER distribution in most resting cells (Fig. 4b). In a minor fraction of resting cells, STIM1-2NQ fluorescence was partially localized to small puncta. These few small puncta need not be given undue weight, since a fraction of wildtype STIM1 may localize to puncta in resting cells, especially when the recombinant protein is expressed at high levels. In sharp contrast to the wildtype protein, though, and regardless of expression level, STIM1-2NQ showed very little relocalization to ER-plasma membrane junctions after store depletion (Fig. 4b). Notably, whereas STIM1(D76A) localizes in puncta constitutively 4 , incorporation of the D76A mutation into STIM1-2NQ did not override its failure to form puncta (Supplementary Fig. 4). This finding ruled out the possibility that STIM1-2NQ binds Ca 2+ more tightly at the EF-hand site than wildtype STIM1 and shifts the threshold for activation to lower ER-luminal Ca 2+ concentrations that were not reached under our experimental conditions. STIM1-2NQ is simply very poorly responsive in cells.
EFSAM-2NQ far-UV CD spectra documented an -helix content comparable to wildtype EFSAM-GrpE and a secondary structure unaffected by the presence or absence of Ca 2+ (Fig. 5b), yet ITC showed that only a single Ca 2+ -binding site was occupied in the concentration range monitored (Fig.  4c). A straightforward explanation of the findings would be that negatively charged sidechains in region 2 are directly involved in coordinating Ca 2+ . An alternative explanation, keeping in mind the allosteric effect of the STIM1(D76A) replacement on Ca 2+ binding at other sites, would be that the negatively charged sidechains in region 2 are required for Ca 2+ binding elsewhere in EFSAM.

STIM1-2NQ and STIM2-2NQ maintain an inactive conformation despite Ca 2+ depletion
The simplest mechanism that would explain the failure to relocalize is that STIM1-2NQ is unable to transition to the active conformation upon loss of Ca 2+ from the EF-hand sites. We examined this question directly in ER membranes, using an oxidative crosslinking assay that measures apposition of the STIM1 transmembrane helices 5 . STIM1(A230C)-2NQ showed minimal crosslinking in this assay compared to 'wildtype' control STIM1(A230C), indicating that it maintained the inactive conformation even in the absence of Ca 2+ (Supplementary Fig. 5a,b). Introducing the D76A mutation into STIM1-2NQ did not override the failure to crosslink (Supplementary Fig. 5c), which established that the impairment was not a failure of Ca 2+ dissociation from the EF-hand, in line with the localization result for GFP-STIM1(D76A)-2NQ ( Supplementary Fig. 4).
A second direct indicator of the STIM active conformation is the release of SOAR/CAD from its intramolecular interaction with CC1. The clearest demonstration of this mechanism is that the ERtethered recombinant fragment EFSAM-TM-CC1 retains soluble SOAR/CAD near the ER in resting cells, and releases SOAR/CAD upon store depletion 6,7 . A modification of this assay takes advantage of the fact that a fraction of EFSAM-TM-CC1 expresses at the cell surface, with the luminal EFSAM domain exposed on the extracellular side, allowing precise experimental control over the Ca 2+ concentration 'seen' by the luminal domain and the resulting conformational change ( Supplementary  Fig. 6). We used STIM2 EFSAM-TM-CC1 for this experiment, because the STIM2 EFSAM-TM-CC1 fragment was more efficiently expressed at the cell surface than the corresponding fragment of STIM1. The Ca 2+ -dependent association of SOAR/CAD with STIM2 EFSAM-TM-CC1 is illustrated in Supplementary Figs. 6c and 6d. Note that the midpoint Ca 2+ concentration is ~2 mM, contrasting with ~400 M for STIM2 in the ER of intact cells 8 and affording further evidence that Ca 2+ binding to EFSAM is influenced by the context. 'Context' in this case could encompass variables including the entropic cost of capturing the independent SOAR/CAD fragment, differences in lipid composition or thickness between ER membrane and plasma membrane, and the presence or absence of specific protein partners.
In initial experiments, we verified by ITC that STIM2 EFSAM, like STIM1 EFSAM, has multiple Ca 2+binding sites (Supplementary Fig. 7a). We then introduced D>N and E>Q substitutions into STIM2 EFSAM-TM-CC1-YFP at positions corresponding to the sites mutated in STIM1-2NQ, expressed the protein in cells, and examined the recruitment of mCherry-SOAR/CAD to the plasma membrane by EFSAM-TM-CC1 under low-Ca 2+ and high-Ca 2+ conditions. Wildtype STIM2 EFSAM-TM-CC1-YFP and its D167A EF-hand mutant (corresponding to D76A in STIM1) served as controls. The results were clearcut (Supplementary Fig. 7b-g). Wildtype STIM2 EFSAM-TM-CC1 recruited SOAR/CAD poorly in the absence of extracellular Ca 2+ , but bound SOAR/CAD efficiently when the external Ca 2+ concentration was raised to 12 mM. The EF-hand mutant STIM2(D167A) failed to recruit appreciable SOAR/CAD regardless of Ca 2+ concentration. And STIM2-2NQ retained SOAR/CAD at the plasma membrane at both low and high Ca 2+ concentrations.
We conclude that STIM1-2NQ and STIM2-2NQ have minimal ability to assume an active conformation, even in the absence of Ca 2+ . Mechanistically, the 2NQ mutations might interfere directly or indirectly with the known EFSAM-EFSAM dimer interaction underlying activation, or might impair a still unidentified interaction, for example at an EFSAM-lipid interface.
Coincidentally, the four tryptophan residues of EFSAM are in the region marked by most prominent changes in the HDX-MS signals (Fig. 5j, Supplementary Fig. 9d). There are no tryptophan residues in GrpE, so we were able to monitor the intrinsic tryptophan fluorescence of EFSAM for additional confirmation that there is a Ca 2+ -dependent conformational change in this region. Indeed, wildtype EFSAM displayed an enhancement of tryptophan fluorescence compared to the denatured protein both in the absence and in the presence of Ca 2+ , indicating that the tryptophan sidechains are at least partly buried in both conditions. Further, an increase in fluorescence confirmed that the local environment of one or more tryptophan residues in EFSAM did change on binding of Ca 2+ (Supplementary Fig. 10a-c). The slight Ca 2+ -dependent alteration in peptide backbone dynamics of this region in EFSAM(D76A) was not detected as a tryptophan fluorescence change.
ANS (8-anilinonaphthalene-1-sulfonic acid) is a commonly used extrinsic probe of protein conformational change. Wildtype EFSAM-GrpE, EFSAM(D76A)-GrpE, and EFSAM-2NQ-GrpE all bound ANS as judged by the enhanced fluorescence and a blue shift in the fluorescence emission peak of ANS-protein samples compared to free ANS ( Supplementary Fig. 10d-f). The three protein variants showed distinct patterns of change in response to titration with Ca 2+ in the range 1 M to 1 mM, discussed here in order of increasing complexity. EFSAM(D76A) showed no change in ANS fluorescence when Ca 2+ was added. EFSAM-2NQ exhibited modest quenching of ANS fluorescence. Wildtype EFSAM exhibited both modest quenching, evident at somewhat lower Ca 2+ concentrations than for EFSAM-2NQ, and a further 12-nm blue shift of fluorescence emission in the presence of higher concentrations of Ca 2+ . The latter change in the emission spectrum indicates decreased polarity surrounding the ANS binding site(s) or availability of new nonpolar site(s). The blue shift is completed at ~300 M Ca 2+ , reminiscent of results from the D4 binding competition assay, and reinforcing the conclusion that Ca 2+ binding to EFSAM occurs in this concentration range.
These experiments with intrinsic and extrinsic fluorescence probes support the conclusion from CD and HDX-MS data that there are structured conformations of EFSAM at both endpoints of the high Ca 2+ -low Ca 2+ conformational change observed by FRET in Fig. 1g. The conformational change monitored by EFSAM-EFSAM FRET is substantially completed between 0 M Ca 2+ and 30 M Ca 2+ , the concentration used in the HDX-MS experiments. The D4 Ca 2+ sensor titration experiments (Fig.  3g,i) and ANS fluorescence measurements ( Supplementary Fig. 10d) indicate that further Ca 2+ binding to wildtype EFSAM occurs at Ca 2+ concentrations up to 300-400 M, and this may correspond to a further consolidation of the inactive conformation.