Listening to the quiescence of single STIM1 dimers

STIM1 plays a central role in store-operated calcium entry, a major component of cellular calcium signaling in mammalian cells (1). STIM1 senses a transient reduction in the level of free calcium in the endoplasmic reticulum (ER) lumen, conveys a signal from the ER to the plasma membrane, and gates the plasma membrane ORAI calcium channel. Inactive STIM1 is a dimer of identical subunits, with each subunit possessing an ER-luminal EFSAM domain; a single transmembrane segment (TM); and a cytoplasmic region comprising CC1, SOAR/CAD, a predicted unstructured region, and a polybasic tail (Fig. 1). Cell-biological and protein-chemical studies have shown that CC1 retains SOAR/CAD and the polybasic tail near the ER when STIM is inactive, and that ER calcium depletion triggers a rearrangement of EFSAM, TM, and CC1 that physically extends the STIM cytoplasmic domain and allows SOAR/CAD and the polybasic tail to interact with the plasma membrane at specialized ER-plasma membrane contact sites (2). Intensive efforts applying X-ray diffraction or NMR have yielded insight into elements of the STIM1 protein structure, most notably the SOAR/CAD domain and the calcium-bound EFSAM domain, but maddening gaps have persisted, including the lack of a comprehensive structural view of the inactive and active STIM1 cytoplasmic domain in cells.

Fig. 1.

The inactive STIM1 dimer. STIM1 regions relevant to this commentary are labeled at Left. Conformational restraints described in ref. 3 are indicated at Right.

Now, a masterly paper by Qiu and Lewis (3) describes the structural organization of the CC1-SOAR/CAD region of purified full-length STIM1, reconstituted into a lipid bilayer, under conditions that mimic the Ca²⁺-replete ER. The authors measured FRET efficiencies between donor and acceptor labels introduced at 15 selected STIM1 residue pairs in the cytoplasmic domain, for example 242:400′ FRET between residue 242 near the base of CC1 and residue 400′ in the SOAR/CAD domain of the paired STIM1 monomer. (The authors denote FRET pairs by specifying the residues labeled with FRET donor and acceptor fluorophores, where a prime indicates a residue of the paired monomer.) The FRET efficiencies were converted to estimated distances in the range 3.42 to 7.16 nm, which were used as constraints in silico to arrive at a model for the predominant geometry of the inactive STIM1 cytoplasmic domain. Certain predicted intradimer contacts were then confirmed by biochemical crosslinking. Taken together with an earlier analysis of the isolated STIM1 cytoplasmic domain (4), the report offers a beautiful and instructive portrait of resting STIM1 in the region from residue 242 to 444. This brief commentary will focus specifically on its insights into structural restraints that prevent the STIM1 C terminus from unfurling in resting cells independently of calcium dissociation from the STIM1 ER-luminal domain.

To investigate these restraints that maintain STIM1 quiescence, the paper exploits the fact that inactive STIM1 is not locked into a single conformation. The main 242:242′ FRET peak (FRET efficiency 0.28), representing ~70% of STIM1 dimers, is accompanied by a second peak (FRET efficiency 0.68). The presence of a minority of molecules that differ in conformation from the predominant resting population attracted special interest, because several activating STIM1 mutants increased occupancy of the additional peak(s). This is not to say that the minor population(s) represent full physiological activation. It would take special pleading, for example, to explain why ~30% of resting wildtype STIM1 dimers were fully active in 2 mM Ca²⁺ [(3), figure 2A], and the minor 242:242′ peak in the FRET histogram of STIM1 in 2 mM Ca²⁺ is not a precise match to new 242:242′ peaks in the FRET histogram of STIM1 activated by depleting Ca²⁺ with EGTA [(3), figure 2B]. Nonetheless, the finding that activating mutations correlate with displacement of the resting FRET distribution away from the canonical resting conformation [(3), figures 4 and 5 and S2 and S5] suggests that the increased 242:242′ FRET might reflect a configuration with a propensity for full activation, and perhaps an intermediate along a path to active STIM1.

Another feature seems to mark the higher-FRET 242:242′ peak as corresponding to an inactive configuration. Wildtype STIM1 in calcium-replete conditions has no detected population in an extended configuration that would bridge the ~15 nm distance from ER to plasma membrane (5). The most extended form observed by 242:400′ FRET [(3), figures 1C and S2] corresponds to a peak with FRET efficiency ~0.27, whereas 242:400′ FRET efficiency for the AlexaFluor 555–AlexaFluor 647 pair at full STIM1 extension would fall to near zero. However, it is conceivable that the extended conformation in cells depends on interaction with other STIM1 dimers, or other proteins, or STIM1 C-terminal tail engagement with plasma membrane lipids, and so would not be observed here. This is a secondary point that will be resolved by data on whether wildtype STIM1 dimers in isolation extend fully when calcium is depleted. Whatever the eventual answer, 242:242′ FRET has been an informative parameter to define the restraints on resting STIM1.

Now, a masterly paper by Qiu and Lewis (3) describes the structural organization of the CC1-SOAR/CAD region of purified full-length STIM1, reconstituted into a lipid bilayer, under conditions that mimic the Ca²⁺-replete ER.

The major message of the new study is its mapping of several inbuilt structural restraints in STIM1 that tend to stabilize the inactive configuration(s) of the STIM1 cytoplasmic domain (Fig. 1). A first class of conformational restraints is typified by L251S (ref. 6). The L251S substitution breaks a key contact between CC1α1 and CC3, and thereby releases CAD from CC1 and enables extension of the C-terminal domain (7). Release of that interaction is the critical effect. The CC1 coiled coil does not form as a stable secondary structure in vitro in the L251S mutant of the isolated STIM1 cytoplasmic domain, even though SOAR/CAD is released, and CC1(L251S) monomers do not form a coiled coil when their N termini are brought together covalently (7). The failure of CC1(L251S) to assemble as a coiled coil does not preclude STIM1 activation by the mutation in cells, presumably because the span of an unstructured or partially structured CC1 segment is still sufficient to bring CAD into the proximity of the plasma membrane. While the overall conclusion that interaction with CC1 retains SOAR/CAD near the ER is unchanged, the new STIM1 structural model, supported by new crosslinking data, defines precisely the CC1α1-CC3 contacts that underlie the interaction. This structural picture makes tangible the mechanism of STIM1 activation by other previously reported CC1 mutations in this class (at M244, L248, L258, L261, and L265) and activating mutations in the complementary surface of SOAR/CAD (at I409, L416, V419, and L423).

A second class of conformational restraints turned out to be linked to the CC1-SOAR/CAD restraint. The ³¹⁸EEELE³²² > AAALA (4EA) mutation is a well-known activating mutation (8), whose structural basis had remained undefined. Qiu and Lewis showed that the 4EA mutant exhibits substantial changes in both 242:242′ FRET and 309:309′ FRET, and this triggered a chain of logic and experiment that led to the mechanism. Residue 309 is in the loop where CC1 folds back to pack its α3 helix against its α2 helix, and a mutation at nearby residue 304 is the root cause of the pathology in Stormorken syndrome (9). The Stormorken mutation R304W had been studied in the original paper characterizing the structure of the isolated STIM1 cytoplasmic domain (4). That paper determined that the R304W substitution destabilizes the dimer of STIM1 CC1α2–CC1α3 hairpin structures in STIM1, and thus in turn destabilizes the CC1α1–CC3 interaction. The idea that 4EA might act by a similar mechanism was buttressed by examining STIM1 dimers with engineered changes suggested by the new structural model. Indeed, replacements in the CC1α2-CC1α3 hairpin structures that are expected to disrupt either CC1α3–CC1α3′ nonpolar contacts or CC1α2–CC1α3 salt bridges all perturbed the conformation of inactive STIM1, pointing to the CC1α2–CC1α3 hairpin as a key determinant stabilizing the resting conformation of STIM1.

A novel and compelling insight is that there is STIM1-lipid coupling at the cytoplasmic face of the ER. The ³⁸²KIKKK³⁸⁶ > QIQQQ (4KQ) replacements or similar mutations that eliminate several positively charged sidechains in the region have been implicated in STIM1 interaction with the plasma membrane or with ORAI1, but a role for this region in inactive STIM1 has not been elucidated previously. Qiu and Lewis show that the 4KQ replacement or a similar 4KA replacement, or changes in membrane lipid composition that eliminate the net negative charge at the bilayer surface, all act on the STIM1 conformation as assessed by 242:242′ FRET. Note that unlike L251S or 4EA—mutations that defined the first two restraints— the 4KQ mutant does not drive extension of isolated STIM1^CT (ref. 7). The fact that this mutant acts only in the context of STIM1 inserted into a lipid bilayer is best interpreted as saying that the forced apposition of the 4KQ or 4KA mutant CAD apex with the membrane has an energetic cost, and that the favorable interaction of the clustered positive charges of KIKKK in wildtype CAD with negatively charged lipids mitigates the cost.

There may be a parallel in the effects of the ²³⁴NRYS²³⁷ > LRYL replacement, which also does not drive extension of isolated STIM1^CT, although it activates full-length STIM1 in cells (10). The constitutive activation has been attributed in this instance to the expected ability of the substitutions to perfect a CC1 coiled coil, which of course assumes that the CC1 N-terminal segments can get together in a resting cell. But what brings the helices together? There are two plausible possibilities— either normal “breathing” of the resting conformation sweeps SOAR/CAD out of the way from time to time and affords opportunities to nucleate the coiled coil, or the CC1 juxtamembrane regions (roughly residues 230 to 241 in wildtype STIM) are involved in stabilizing the predominant resting conformation. In the latter scenario, mutating ²³⁴NRYS²³⁷ > LRYL would provide both an unlatching that permits the CC1 N-terminal segments to come together and the coiled-coil nucleation site. There is some indirect, although not decisive, support for the second view. First, releasing the CC1-SOAR/CAD clamp either by an R304W replacement or by deletion of CC1 in isolated STIM1^CT alters 389:389’ and 400:400’ FRET, suggesting that an adjacent part of CC1 stabilizes the conformation of the SOAR/CAD apex (4). And, second, the Y236G substitution within the NRYS segment causes mild constitutive activity of full-length STIM1 in cells (11).

The remaining conformational restraint investigated was the ER-luminal EFSAM domain itself. The authors propose that EFSAM in inactive STIM1 forms a steric barrier that prevents apposition of the two TM segments of the STIM dimer, since removal of the STIM1 luminal domain by deletion allows weak constitutive activation of STIM1. Physical distance between the TM segments is necessarily a major control on activation, but other experimental evidence points to a more nuanced narrative. Disulfide crosslinking has documented the occasional adjacency of L216C with L216′C and S219C with S219’C in full-length STIM1 under calcium-replete conditions (10). Crosslinking might well capture an infrequent geometry, but it is nonetheless a geometry that is not amenable to TM–TM pairing in the fully active configuration, since crosslinking between cysteine residues introduced into the outer half of the TM segments is negligible under the same conditions. This suggests that EFSAM–EFSAM dimerization during physiological STIM1 activation not only brings the TM helices together, but brings them together in a preferred orientation where contact of the paired TM segments along their entire length is favored. If a structured active EFSAM brings TM segments together stably (12), the simplest possibility is that productive EFSAM–EFSAM dimerization is sufficient by itself to accomplish the necessary alignment of TM helices. An interesting alternative possibility—in line with the hints that juxtamembrane residues near the cytoplasmic face of the ER membrane participate in extension of the STIM1 cytoplasmic domain—is that the documented structural readjustments within EFSAM upon calcium dissociation (12) might cause a subtle rearrangement of residues 202 to 216 linking EFSAM to the TM segment, and that this readjustment is needed to support activation.

The rigorous demands of cellular calcium signaling require that inactive STIM1 maintain its silence under resting conditions, and yet be poised to release SOAR/CAD and the STIM polybasic tail upon depletion of ER calcium stores. The new paper by Qiu and Lewis highlights weak conformational restraints that maintain the silence. The next steps with this approach will be to delineate the full constellation of physiological activated conformations of the cytoplasmic domain and the overall movements of the ER-luminal EFSAM domains at rest and upon activation. The finer conformational changes that have been detected within EFSAM (ref. 12) and any conformational adjustments in the confined space of the membrane-proximal segments are likely to require additional creative approaches. It will be of prime importance to extend the conformational analysis to the kinetic pathways that facilitate the inactive > active and active > inactive transitions, and to develop a broader picture of STIM1 activation in the cellular context of plasma membrane, ORAI, and other cellular proteins. There is much still to learn.