Oligomerization of STIM1 couples ER calcium depletion to CRAC channel activation

Ca²⁺ release-activated Ca²⁺ (CRAC) channels generate sustained Ca²⁺ signals that are essential for a range of cell functions, including antigen-stimulated T lymphocyte activation and proliferation^{1, 2}. Recent studies³ have revealed that the depletion of Ca²⁺ from the endoplasmic reticulum (ER) triggers the oligomerization of STIM1, the ER Ca²⁺ sensor, and its redistribution to ER-plasma membrane (ER-PM) junctions^4–8 where the CRAC channel subunit Orai1 accumulates in the plasma membrane and CRAC channels open^9–12. However, how the loss of ER Ca²⁺ sets into motion these coordinated molecular rearrangements remains unclear. Here we define the relationships among [Ca²⁺]_ER, STIM1 redistribution, and CRAC channel activation and identify STIM1 oligomerization as the critical [Ca²⁺]_ER–dependent event that drives store-operated Ca²⁺ entry (SOCE). In cells expressing an ER-targeted Ca²⁺ indicator, CRAC channel activation and STIM1 redistribution follow the same function of [Ca²⁺]_ER, with a K_1/2 of ~200 μM and a Hill coefficient of ~4. Because STIM1 binds only a single Ca²⁺ ion⁵, the high apparent cooperativity suggests that STIM1 must first oligomerize to enable its accumulation at ER-PM junctions. To assess directly the causal role of STIM1 oligomerization in SOCE, we replaced the luminal Ca²⁺-sensing domain of STIM1 with the rapamycin-binding proteins FRB or FKBP. A rapamycin analog oligomerizes the fusion proteins and causes them to accumulate at ER-PM junctions and activate CRAC channels without depleting Ca²⁺ from the ER. Thus, STIM1 oligomerization is the critical transduction event through which Ca²⁺ store depletion controls store-operated Ca²⁺ entry, acting as a switch that triggers the self-organization and activation of STIM1-Orai1 clusters at ER-PM junctions.

The defining feature of store-operated channels is their activation in response to ER Ca²⁺ ([Ca²⁺]_ER) depletion. However, their sensitivity to [Ca²⁺]_ER and the factors that determine this sensitivity have never been established, largely because of the technical difficulty of quantifying [Ca²⁺]_ER. To address this issue we generated a Jurkat T cell line stably expressing the Ca²⁺-sensitive cameleon protein, YC4.2er. YC4.2er is selectively retained in the ER, as shown by its colocalization with the resident ER protein calnexin but not with mitochondrial or Golgi markers, and by its functional response to agents that deplete ER Ca²⁺ (Fig. 1a and Supp. Fig. 1). In situ calibration of the YC4.2er FRET signal indicates a responsivity to [Ca²⁺]_ER in the range of ~1 μM to >1 mM (Supp. Fig. 2).

Figure 1. The [Ca²⁺]_ER-response relation for the CRAC channel.

Simultaneous measurements of [Ca²⁺]_ER and I_CRAC in individual Jurkat T cells. a, Non-stationary measurements of I_CRAC and [Ca²⁺]_ER. Store depletion with 20 μM CPA induces an increase in I_CRAC (top) that follows a decrease in [Ca²⁺]_ER (middle) monitored with YC4.2er. The I-V relationship shows the inward rectification typical of I_CRAC (bottom). In this cell, a small inward current through outwardly-rectifying Cl⁻ channels is also present initially but disappears before I_CRAC is induced. b, Recordings of I_CRAC (top) and [Ca²⁺]_ER (middle) under steady-state conditions. Each cell was treated with the indicated CPA concentration for 8–15 min prior to recording, and CPA was maintained throughout the experiment. I-V relations are typical for I_CRAC (bottom). c, Steady-state I_CRAC and [Ca²⁺]_ER are plotted for 40 cells after treatment with 0.5–20 μM CPA. A fit of the Hill equation with a K_1/2 of 169 μM and Hill coefficient of 4.2 is superimposed on the data. Squares: mean ± s.e.m. of 3–12 cells. Circles: single cells (see Supplementary Information).

To determine the dependence of CRAC channel activation on [Ca²⁺]_ER, we measured I_CRAC in perforated-patch recordings from Jurkat YC4.2er cells treated with cyclopiazonic acid (CPA), a reversible SERCA inhibitor. CPA evokes a time-dependent decline in [Ca²⁺]_ER in parallel with the activation of I_CRAC measured in the same cell (Fig. 1a). However, because I_CRAC responds slowly to rapid changes of [Ca²⁺]_ER, non-stationary measurements like these will distort estimates of the true [Ca²⁺]_ER dependence of the CRAC channel. For this reason we determined instead the [Ca²⁺]_ER-I_CRAC relationship under steady-state conditions, by pretreating cells with 0.5–20 μM CPA for 8–15 min in the absence of extracellular Ca²⁺ to generate a range of constant [Ca²⁺]_ER values. This passive depletion approach also minimizes spatial variations of [Ca²⁺]_ER, allowing the [Ca²⁺]_ER dependence of SOCE to be determined from whole-cell YC4.2er measurements. Following readdition of 20 mM Ca²⁺ to the bath, current was monitored during brief hyperpolarizations from the resting potential of +30–50 mV at constant [Ca²⁺]_ER (Fig. 1b). The current was identified as I_CRAC based on its delayed response to extracellular Ca²⁺ reflecting Ca²⁺-dependent potentiation (Fig. 1b), an inwardly rectifying I/V relation (Fig. 1b), and extremely low current noise¹⁷. Measurements from 40 cells show that I_CRAC is a steep function of [Ca²⁺]_ER with a K_1/2 of 169 μM and a Hill coefficient (n_H) of 4.2 (Fig. 1c). Interestingly, a decline of >100 μM from the resting [Ca²⁺]_ER of ~400 μM is required to initiate CRAC channel opening in these cells, which may help explain the ability of IP₃ to release small amounts of ER Ca²⁺ without activating I_CRAC in some cells¹⁸.

We next addressed the source of the CRAC channel’s steep dependence on [Ca²⁺]_ER. Because STIM1 is known to be the Ca²⁺ sensor for SOCE^{6, 7} and its redistribution to ER-PM junctions is linked to I_CRAC activation^{6, 8, 11}, we measured the dependence of STIM1 redistribution on [Ca²⁺]_ER. Exposure to 0.5–3 μM CPA for > 8 min causes a partial redistribution of Cherry-STIM1 to the cell periphery, which can be seen by widefield imaging at the cell equator (Fig. 2a). We quantified the redistribution of Cherry-STIM1 as the ratio of the mean peripheral fluorescence to the mean total fluorescence (Fig. 2b); this method gives results that agree quantitatively with TIRF measurements of STIM1 puncta (Supp. Fig. 3) while facilitating the separation of the cameleon and Cherry fluorescence signals (see Supplementary Information). Measurements from 41 cells show that STIM1 redistribution is a steep function of [Ca²⁺]_ER that closely resembles that of I_CRAC activation, with a K_1/2 of 187 μM and a Hill coefficient of 3.8 (Fig. 2c). The value of K_1/2 is close to the binding affinity of the recombinant EF-hand/SAM domain of STIM1 measured in vitro (K_D = 200–600 μM; ref 5), consistent with its role as an ER Ca²⁺ sensor. Importantly, the close correspondence between the STIM1 and I_CRAC curves indicates that CRAC channels open in direct proportion to the concentration of STIM1 at ER-PM junctions and that the CRAC channel derives its highly nonlinear dependence on [Ca²⁺]_ER from the ER Ca²⁺ dependence of STIM1 redistribution. A recent study of HeLa cells found a similar dependence of STIM1 redistribution on [Ca²⁺]_ER (ref ¹³). In that study, the homolog STIM2 redistributed to ER-PM junctions at higher [Ca²⁺]_ER (K_1/2 = 406 μM) than did STIM1 (K_1/2 = 210 μM), and it was proposed that STIM2 functions as a homeostatic ER Ca²⁺ sensor by activating Orai1. Our findings that I_CRAC and STIM1 redistribution follow the same function of [Ca²⁺]_ER implies that in Jurkat cells STIM2 activates at most a minor fraction of endogenous CRAC channels, consistent with its low level of expression in T cells¹⁴.

Figure 2. The [Ca²⁺]_ER dependence of STIM1 redistribution determines the [Ca²⁺]_ER-response relation of the CRAC channel.

a,Widefield epifluorescence images of a cell expressing Cherry-STIM1 at rest (top) and following store depletion with 3 μM CPA (middle) and TG (bottom). The redistribution of Cherry-STIM1 in single cells was monitored as the ratio of the mean Cherry fluorescence in the most peripheral 0.5 μm of the cell (F_P, right) to the mean fluorescence of the entire cell (F_TOT, left). Scale bar = 2μm. b, In the same cell, STIM1 redistribution represented by F_P/F_TOT normalized to the maximum ratio with TG (blue). F_P/F_TOT increases as [Ca]_ER (red) declines. Individual data points (open symbols) and the mean response (bars) are shown. c, STIM1 redistribution (F_P/F_TOT, blue) plotted against [Ca²⁺]_ER after treatment with 0–3 μM CPA (means ± s.e.m. of 3–4 cells; 41 cells total). A fit of the Hill equation (blue line) indicates a K_1/2 of 187 μM and Hill coefficient of 3.8. Steady-state I_CRAC data fitted with the Hill equation are re-plotted from Fig. 1 (red).

The shape of the STIM1 redistribution curve has important implications for the mechanism underlying SOCE. The Hill coefficient of ~4 shows that puncta formation is a nonlinear process with respect to [Ca²⁺]_ER without necessarily indicating a cooperative mechanism or that the active form of STIM1 is a tetramer. However, the high Hill coefficient implies that STIM1 puncta at ER-PM junctions do not form by the independent accretion of STIM1 monomers, which contain only a single luminal Ca²⁺ binding site⁵, but suggests instead that only oligomers of STIM1 can accumulate at these sites. There are two ways in which STIM1 is known to oligomerize. In resting cells STIM1 self-associates with an undetermined stoichiometry via its cytosolic coiled-coil domains^{15, 16}; in addition, removal of Ca²⁺ from the EF-hand of STIM1 drives further oligomerization in vitro⁵ and in vivo⁴. Store-dependent oligomerization of STIM1 occurs within seconds, slightly in advance of puncta formation, and a causal role in SOCE has been hypothesized but never tested^{4, 5}.

To address the possible role of STIM1 oligomerization in SOCE, we adopted an approach based on rapamycin-induced protein heterodimerization^{17, 18}. We replaced the EF-SAM domain of Cherry-STIM1 with a tandem dimer of FK506-binding protein (FKBP12) or a variant of the rapamycin-binding domain of mTOR (FRB) to generate STIM1 chimeras that will heterodimerize when bound to a rapamycin analog (AP21967, or rapalog). Given that STIM1 is known to self-associate at rest^{15, 16}, rapalog would thus be expected to link multimers containing FRB with those containing FKBP to form extended oligomers of STIM1 (Fig. 3a). We assayed oligomer formation in HEK293 cells expressing Cherry-FRB-STIM1 and Cherry-FKBP-STIM1 (abbreviated hereafter as F-STIM1) using blue native polyacrylamide gel electrophoresis (BN-PAGE)¹⁹. The >2-fold increase in apparent mass following rapalog treatment confirms its ability to oligomerize F-STIM1, and because crosslinking of monomers would be expected to at most double the mass, indicates that the resting state of FRB-STIM1 and FKBP-STIM1 is at least a dimer (Fig. 3b).

Figure 3. STIM1 oligomerization induces the accumulation of STIM1 at ER-plasma membrane junctions.

a, The cartoon depicts the oligomerization of F-STIM1 induced by rapalog (R). At rest, FKBP-STIM1 and FRB-STIM1 are expected to form homo- and heterodimers; only intermolecular crosslinks between homodimers are shown here. Abbreviations: EF (EF hand), SAM (sterile-α motif), CC (coiled-coil), S/P (serine-proline-rich), K (lysine-rich), CH (mCherry). b, BN-PAGE and Western blot of transiently expressed F-STIM1 harvested from HEK293 cells. Untreated (left) and rapalog-treated (right) F-STIM1 was detected using a monoclonal anti-STIM1 antibody. c, Rapalog induces a time-dependent peripheral redistribution of F-STIM1 (n=10 cells). d, Peripheral redistribution of Cherry-STIM1 by TG (red bars; n=31 for each) and redistribution of F-STIM1 by rapalog (black bars; n=39, rest; n=42, rapalog). Values expressed as mean ± s.e.m. (c, d). e–h, TIRF images of Jurkat cells, scale bar = 2 μm. e, F-STIM1 before (left) and after (right) incubation with rapalog. f, Cherry-STIM1 before (left) and after (right) store depletion with TG. g, Cherry-STIM1 before (left) and after (right) incubation with rapalog. h, F-STIM1 (top row), GFP-STIM1 (middle row) and merged images (bottom row) from a single cell after rapalog treatment (left column) and subsequent store depletion with TG (right column). Cherry and GFP intensities are scaled to the maximal intensity of each fluorophore after TG treatment.

We first examined the effects of rapalog on the localization of F-STIM1 in Jurkat cells. Rapalog evoked a redistribution of F-STIM1 to the cell periphery that was complete within several minutes (Fig. 3c). Quantitative analysis shows that rapamycin triggers the redistribution of F-STIM1 as effectively as Ca²⁺ store depletion induces the redistribution of wild-type STIM1 (Fig. 3d). When examined by TIRF microscopy, the rapalog-driven peripheral accumulations of F-STIM1 (Fig. 3e) resemble the puncta of wild-type STIM1 that form in response to store depletion (Fig. 3f). Similar results were obtained in HEK293 cells. Rapalog did not affect the localization of wild-type STIM1 (Fig. 3g), nor did it deplete Ca²⁺ stores (see below). Finally, rapalog-induced F-STIM1 puncta co-localize with store depletion-induced GFP-STIM1 puncta in the same cell, confirming that rapalog causes F-STIM1 to accumulate at the same ER-PM junctions where STIM1 and Orai1 are known to interact. Thus, we conclude that oligomerization of STIM1 is sufficient to drive the redistribution of STIM1 to ER-PM junctions.

Heterodimerization of FRB-STIM1 and FKBP-STIM1 also activates endogenous CRAC channels. Rapalog elevated the mean resting [Ca²⁺]_i in Jurkat cells expressing F-STIM1 (Cherry-positive cells) from 170 ± 11 nM (untreated; n = 61) to 388 ± 45 nM (Fig. 4a; n = 45), but did not affect [Ca²⁺]_i in untransfected Jurkat cells. The elevated basal [Ca²⁺]_i was dependent on extracellular Ca²⁺ (Fig. 4a) and was inhibited by 2-APB and low concentrations of La³⁺ (Supp. Fig. 4), consistent with constitutive Ca²⁺ entry through open CRAC channels. Importantly, TG released similar amounts of ER Ca²⁺ in rapalog-pretreated and resting cells, indicating that rapalog stimulates Ca²⁺ entry without depleting Ca²⁺ stores (Fig. 4a). Whole-cell recordings with a high-[Ca²⁺] pipette solution designed to minimize store depletion confirmed that heterodimerization of FRB-STIM1 and FKBP-STIM1 directly activates I_CRAC. In untreated Jurkat cells expressing F-STIM1, I_CRAC was negligible upon breaking in to the whole-cell configuration and developed slowly to a small amplitude, presumably in response to partial store depletion. In contrast, in rapamycin-pretreated cells with visible puncta, large inward currents were evident immediately upon breaking in (Fig. 4b) and displayed essential features of I_CRAC, including a dependence on extracellular Ca²⁺, inwardly rectifying current-voltage relation (Fig. 4c), low current noise, rapid Ca²⁺-dependent inactivation, and inhibition by 2-APB and La³⁺ (Supp. Fig. 4). The mean current amplitude (2.6 ± 0.6 pA/pF, n= 9) was similar to that produced by Ca²⁺ store depletion in Jurkat cells overexpressing Cherry-STIM1 (ref ¹¹), consistent with the comparable degrees of STIM1 and F-STIM1 redistribution in response to TG or rapalog, respectively (Fig. 3). Together, these results suggest that F-STIM1 oligomers at ER-PM junctions are fully active and provide direct evidence that the oligomerization of STIM1, independently of changes in [Ca²⁺]_ER, is sufficient to evoke CRAC channel activation.

Figure 4. STIM1 oligomerization activates Ca²⁺ entry through CRAC channels.

a, In rapalog-treated cells expressing F-STIM1 (black, n=45), resting [Ca²⁺]_i is elevated and sensitive to the removal of extracellular Ca²⁺, indicating constitutive Ca²⁺ entry. In contrast, resting Ca²⁺ influx was largely absent in untreated F-STIM1-expressing cells (red, n=61) and in wt Jurkat cells with (green, n=617) or without (blue, n=517) rapalog. TG-induced Ca²⁺ release in rapalog-treated cells was similar to that of untreated cells. b, I_CRAC development during whole-cell recording from rapalog-treated (black, n=9) and untreated (red, n=9) cells expressing F-STIM1. I_CRAC was measured beginning within 5 s of break-in. c, I–V relations upon break-in, showing the inward rectification typical of I_CRAC in the rapalog-treated cell (black) and the absence of current in the untreated cell (red). Values expressed as mean ± s.e.m. (a,b).

We have shown that STIM1 redistribution and I_CRAC share a steep dependence on [Ca²⁺]_ER and that oligomerization of F-STIM1 is sufficient to drive puncta formation and CRAC channel activation. These results define the input-output relation of the CRAC channel and identify STIM1 oligomerization as the primary transduction event through which this relation is determined. The EF hand and SAM domains of STIM1 appear to serve primarily to control the extent of oligomerization, considering that removal of Ca²⁺ causes a recombinant EF-SAM peptide to oligomerize in vitro⁵, and that the FRB and FKBP modules in F-STIM1 can effectively substitute for the EF-SAM domain and activate I_CRAC when crosslinked by rapalog. That the latter occurs without Ca²⁺ store depletion suggests that once STIM1 oligomerizes, all subsequent steps leading to SOCE occur independently of ER Ca²⁺. Thus, we propose that the oligomerization of STIM1 acts as a switch to trigger the self-organization of STIM1 and Orai1 complexes at ER-PM junctions and the consequent activation of CRAC channels.

How might this oligomerization “switch” operate? In its resting state, Ca²⁺-bound STIM1 moves freely throughout the ER membrane⁴ but following store depletion, STIM1 oligomers accumulate in ER subregions located 10–25 nm from the PM, close enough to allow trapping by binding to targets in the PM⁸. These targets have not yet been positively identified, but suggested candidates include Orai1 (refs 20, 21) or an associated protein²², and PM phospholipids^{4, 23}. Once localized at ER-PM junctions STIM1 then promotes the accumulation of Orai1 at apposed sites, leading to channel activation^{11, 12, 24}. Oligomerization may promote the binding of STIM1 to its targets in two ways: an affinity-based mechanism in which a conformational change exposes a previously masked cytosolic binding domain, and an avidity-based mechanism in which clustering of the binding domains increases their local concentration at ER-PM junctions. Both of these mechanisms are likely to contribute to the assembly and function of CRAC channel complexes that constitute the final stage of the SOC activation process.

METHODS SUMMARY

[Ca²⁺]_ER measurements

[Ca]_ER was measured in a Jurkat E6-1 cell line stably expressing a modified YC4er (V68L and Q69M; refs 25–27). Cells were pre-treated with CPA (0.5–20 μM) in Ca²⁺-free Ringer’s for 8–15 min, and emission intensities at 485 and 535 nm were averaged across the cell to yield a raw emission ratio. Ratios were calibrated in situ for every cell as described (Supplementary Information).

Heterodimerizer experiments

To generate F-STIM1, mutant FRB and tandem FKBP sequences were substituted for the EF-SAM domain (wtSTIM1 aa 35–207) in Cherry-STIM1 using plasmids provided by Ariad Pharmaceuticals (Cambridge, MA). F-STIM1 was crosslinked using 1 μM rapalog (AP21967, Ariad Pharmaceuticals). Unless indicated otherwise, cells were pre-incubated in full medium at 37°C for 30 min, with or without rapalog, and subsequent measurements were performed at 22–25°C in standard Ringer’s solutions. Timelapse imaging was performed at 37°C in full medium with or without rapalog. Only cells with ~3–10% of the fluorescence of the brightest cells in each experiment were analyzed. BN-PAGE was performed essentially as described¹⁹. A mAb against the STIM1 C-terminus (1:250; Abnova, Taipei City, Taiwan) and an alkaline phosphatase-conjugated 2° antibody (1:30,000; Sigma) were used for Western blotting.

Perforated-patch and whole-cell recording

I_CRAC in YC4.2er cells (Fig. 1) was recorded in the perforated-patch configuration²⁸ with 20 mM extracellular Ca²⁺, using a stimulus of a 50-ms step to −100 mV followed by a ramp from −100 to +100 mV, delivered from a holding potential of +30 or +50 mV. Whole-cell recording of I_CRAC (ref ²⁹; Fig. 4) was performed with 20 mM Ca²⁺ Ringer’s, with stimuli consisting of a 100-ms step to −112 mV followed by a 100-ms voltage ramp from −112 to +88 mV applied from the holding potential of +38 mV beginning within 5 s of break-in.