Local Cytosolic Ca²⁺ Elevations Are Required for Stromal Interaction Molecule 1 (STIM1) De-oligomerization and Termination of Store-operated Ca²⁺ Entry

Background: STIM1 oligomerization upon endoplasmic reticulum (ER) Ca²⁺ depletion activates store-operated Ca²⁺ entry (SOCE) channels, but whether this mechanism is reversible is unknown.

Results: STIM1 de-oligomerization upon ER Ca²⁺ refilling requires concomitant cytosolic Ca²⁺ elevations near STIM1 membrane clusters.

Conclusion: Cytosolic Ca²⁺ elevations dissociate STIM1 oligomers during SOCE termination.

Significance: Ca²⁺ acts on a cytosolic target to control the disassembly of STIM1 membrane clusters.

Abstract

The Ca²⁺ depletion of the endoplasmic reticulum (ER) activates the ubiquitous store-operated Ca²⁺ entry (SOCE) pathway that sustains long-term Ca²⁺ signals critical for cellular functions. ER Ca²⁺ depletion initiates the oligomerization of stromal interaction molecules (STIM) that control SOCE activation, but whether ER Ca²⁺ refilling controls STIM de-oligomerization and SOCE termination is not known. Here, we correlate the changes in free luminal ER Ca²⁺ concentrations ([Ca²⁺]_ER) and in STIM1 oligomerization, using fluorescence resonance energy transfer (FRET) between CFP-STIM1 and YFP-STIM1. We observed that STIM1 de-oligomerized at much lower [Ca²⁺]_ER levels during store refilling than it oligomerized during store depletion. We then refilled ER stores without adding exogenous Ca²⁺ using a membrane-permeable Ca²⁺ chelator to provide a large reservoir of buffered Ca²⁺. This procedure rapidly restored pre-stimulatory [Ca²⁺]_ER levels but did not trigger STIM1 de-oligomerization, the FRET signals remaining elevated as long as the external [Ca²⁺] remained low. STIM1 dissociation evoked by Ca²⁺ readmission was prevented by SOC channel inhibition and was associated with cytosolic Ca²⁺ elevations restricted to STIM1 puncta, indicating that Ca²⁺ acts on a cytosolic target close to STIM1 clusters. These data indicate that the refilling of ER Ca²⁺ stores is not sufficient to induce STIM1 de-oligomerization and that localized Ca²⁺ elevations in the vicinity of assembled SOCE complexes are required for the termination of SOCE.

Introduction

Store-operated Ca²⁺ entry (SOCE)² is a cellular mechanism that couples the Ca²⁺ depletion of the endoplasmic reticulum (ER) to the activation of plasma membrane Ca²⁺-permeable channels (1). SOCE is required for the generation of the sustained Ca²⁺ signals that mediate secretion, metabolism, and proliferation in both excitable and nonexcitable cells (2), and disruptions in the SOCE pathway lead to severe immune disorders (3–6) and to altered function of skeletal and smooth muscle cells as well as endothelial cells (5, 7–9). The concept that Ca²⁺ depletion of the ER activates membrane channels was proposed 25 years ago (10), but the molecules that mediate this process were unveiled only in 2005. Two families of proteins were identified by genome wide interfering RNA screens: the stromal interaction molecules (STIM), ER transmembrane proteins that sense the Ca²⁺ filling state of the ER to control SOCE (11, 12), and the plasma membrane-spanning proteins Orai1, Orai2, and Orai3 (13–15) that are the pore-forming subunits of SOCE channels.

STIM1 and its homologue STIM2 are single pass type I ER transmembrane proteins bearing a luminal EF-hand Ca²⁺-binding motif fused to a sterile α motif (SAM), and a cytosolic tail containing two proximal coiled-coiled domains and a terminal lysine-rich domain. Ca²⁺ unbinding from the EF-SAM domain (STIM1: K_d 0.2–0.6 mm (16)) induces STIM1 oligomerization and exposes interacting domains in the STIM1 cytosolic tail that bind to the C terminus of Orai channels and induce their activation (17–20). A region of 100 amino acids containing the STIM1 second coiled-coil domain, the CRAC (Ca²⁺ release-activated Ca²⁺ current) activation domain (CAD), also known as the STIM-Orai activating region or Orai1-activating small fragment, is sufficient to activate Orai channels (17–19, 21). In resting cells, the high free Ca²⁺ concentration within the ER lumen ([Ca²⁺]_ER) of 300–600 μm (22) stabilizes the protein in an inactive state via intra-molecular interactions between the two coiled-coils of its cytosolic tail, the membrane-proximal first coiled-coil acting as an autoinhibitory domain that masks the CAD/STIM-Orai activating region domain and prevents its interaction with Orai channels (23, 24). When cells are stimulated with Ca²⁺-mobilizing agonists, the decrease in [Ca²⁺]_ER causes Ca²⁺ to dissociate from the STIM luminal EF-hand, and the resulting conformational change promotes the formation of high order STIM oligomers (16) and their translocation toward the plasma membrane. The STIM oligomers accumulate in ER regions juxtaposed to the plasma membrane (25, 26), where they overlap with Orai1 clusters and Ca²⁺ entry sites (27, 28). The ER structures induced by STIM1 appear on the electron microscope as thin ER sheets devoid of chaperones that are still connected to the conventional ER, and we termed these specialized ER compartments “cortical ER” because of their proximity to the PM (29).

Two lines of evidence indicate that the Ca²⁺-dependent oligomerization of STIM1 is the rate-limiting event in the activation of SOCE. First, the enforced heterodimerization of STIM1 chimeras bearing luminal rapamycin-binding motifs induces STIM1 translocation and Orai1 activation in the absence of Ca²⁺ store depletion, indicating that the oligomerization of STIM1 is sufficient to initiate the cascade of events leading to the activation of plasma membrane channels (30). Second, a point mutation in the STIM1 EF-hand domain that impairs Ca²⁺ binding leads to constitutively SOCE activation, indicating that [Ca²⁺]_ER detection is a crucial step for STIM1 function (12, 31). A [Ca²⁺]_ER value of 169 μm was calculated for half-maximal SOCE activation in situ, by measuring the steady-state [Ca²⁺]_ER and the corresponding CRAC currents in cells exposed to different concentrations of the sarco/endoplasmic reticulum Ca²⁺-ATPase (SERCA) inhibitor cyclopiazonic acid (30). A comparable [Ca²⁺]_ER value (193 μm) was obtained for half-maximal STIM1 oligomerization, measured by fluorescence resonance energy transfer (FRET) between STIM1 fusion proteins, during application of histamine to endothelial cells (32). Interestingly, in this study a lower [Ca²⁺]_ER value was required for half-maximal STIM1 de-oligomerization during store refilling, suggesting that the dissociation process has a higher sensitivity to [Ca²⁺]_ER than the oligomerization process.

The molecular and cellular mechanisms that control SOCE activation have been characterized in details, but the mechanisms that control the retrieval of STIM1 from the plasma membrane and the deactivation of SOCE channels have been comparatively less studied. The oligomerization of STIM is a reversible event (28), and store replenishment is accompanied by STIM1 dissociation from puncta and by its redistribution to deeper ER regions. By analogy with the mechanism of STIM1 activation, it is assumed that Ca²⁺ replenishment of the ER controls the termination of SOCE by inducing the de-oligomerization of STIM1 molecules and their dissociation from the SOCE complex. However, previous studies did not verify that the binding of Ca²⁺ to the EF-hand domain of STIM1 is sufficient to induce the de-oligomerization of STIM1 molecules, their dissociation from the active SOCE complex, and the termination of SOCE. Thus, we do not know whether STIM1 oligomerization and de-oligomerization in vivo are exclusively controlled by changes in the ER Ca²⁺ concentration, as inferred from in vitro studies.

In this study, we correlate [Ca²⁺]_ER to the extent of STIM1 oligomerization during Ca²⁺ depletion and refilling of the ER in cells treated with agonists or reversible SERCA inhibitors. To manipulate [Ca²⁺]_ER without altering the Ca²⁺ concentration in other cellular compartments, we devised a procedure to refill ER Ca²⁺ stores without adding exogenous Ca²⁺, using a membrane-permeable Ca²⁺ chelator that provides a large reservoir of buffered Ca²⁺ in the cytoplasm. Our data indicate that although the degree of STIM1 oligomerization exclusively depends on [Ca²⁺]_ER, an increase in [Ca²⁺]_ER is not sufficient to induce STIM1 de-oligomerization. In addition to ER Ca²⁺ refilling, a localized cytosolic Ca²⁺ elevation is required for the dissociation of STIM1 complexes and their retrieval from the plasma membrane.

EXPERIMENTAL PROCEDURES

Chemicals

Thapsigargin, histamine, gadolinium, SK&F 96365, ionomycin, and digitonin were purchased from Sigma. 2,5-Di-tert-butyl-1,4-benzohydroquinone (BHQ) was purchased from Aldrich Calbiochem. Lanthanum chloride, potassium dihydrogen phosphate, carbonyl cyanide m-chlorophenylhydrazone, and sulfinpyrazone were purchased from Fluka. Fura-2/AM and BAPTA-AM were purchased from Molecule Probes Europe (Leiden, The Netherlands). Quest Fluo-8/AM was purchased from AAT Bioquest. D1_ER was kindly provided by Drs. Amy Palmer and Roger Tsien (University of California, San Diego, CA).

Plasmids

Human STIM1 cDNA was from OriGene. YFP-STIM1 was a gift from Dr. A. B. Parekh (University of Oxford, United Kingdom). mCherry-STIM1 was a gift from Dr. R. S. Lewis (Stanford University). For CFP-STIM1 construction, CFP was cloned by PCR from D1_ER using the primers: forward, 5′-GAG AAC CGG TCG TGA GCA AGG GCG-3′ and reverse, 5′-GCA CGC TGC CGT CCT CGA TGT TGT G-3′. STIM1 was cloned using the primers: forward, 5′-CAC AAC ATC GAG GAC GGC AGC GTG C-3′ and reverse, 5′-GGG GGA ATT CCT ACT TCT TAA GAG GCT TCT TAA AG-3′. Full-length CFP-STIM1 were amplified by recombinant PCR, and cloned into the same vector as YFP-STIM1 between AgeI and EcoRI sites.

Cell Culture and Transfection

HeLa cells were maintained at 37 °C in 5% CO₂ in minimal essential medium containing 10% FCS, 10 μg/ml of streptomycin, and 10 units/ml of penicillin. Cells were seeded on 25-mm diameter glass coverslips and transfected at 70–80% confluence with Lipofectamine 2000 (Invitrogen) by adding 2 μg of plasmid/coverslip, except for co-transfection of CFP-STIM1 and YFP-STIM1, where 0.75 μg of CFP-STIM1 and 2.25 μg of YFP-STIM1 were used. Cells were imaged 48 h after transfection.

Calcium and FRET Measurements

The extracellular solution used for all experiments contained (in mm): 140 NaCl, 5 KCl, 1 MgCl₂, 2 CaCl₂, 10 glucose, and 20 Hepes (pH 7.4 with NaOH), with CaCl₂ replaced by 1 mm EGTA for Ca²⁺-free solutions. For cytosolic Ca²⁺ measurements, cells were loaded with 2 μm Fura-2/AM for 30 min at room temperature, washed twice, and kept 10–15 min to allow de-esterification. Glass coverslips were inserted in a thermostatic chamber (Harvard Apparatus, Holliston, MA) and experiments were conducted at 37 °C. Cells were imaged using a ×40 oil-immersion objective on an Axiovert microscope (Zeiss) equipped with a 16-bit cooled charge-coupled device camera (MicroMax, Princeton Instruments). Cells were alternatively excited at 340 and 380 nm, using a monochromator (DeltaRam, Photon Technology International Inc.), and emission was collected through a 430DCLP dichroic mirror and a 510WB40 emission filter (Omega Optical, USA). For [Ca²⁺]_ER measurements, cells were transiently transfected with the cameleon probe targeted to the ER (D1_ER). The oligomerization/de-oligomerization of STIM1 was followed by measuring the FRET signal between CFP- and YFP-STIM1. For D1_ER or FRET recordings, cells were excited at 430 nm through a 455-nm dichroic mirror (455DRLP, Omega Optical), and emission was collected alternatively at 480 and 535 nm (480AF30 and 535DF25, Omega Optical) using a filter wheel (Ludl Electronic Products). In situ calibration of the D1_ER was performed in medium containing (in mm): 10 NaCl, 135 KCl, 1 MgCl₂, 20 sucrose, 20 Hepes, 0.01 digitonin, 0.01 ionomycin, 0.005 carbonyl cyanide p-chlorophenylhydrazone (pH 7.1 with NaOH). 5 mm EGTA and 5 mm HEDTA were used to buffer free [Ca²⁺] below 100 μm, and 10 mm HEDTA to buffer free [Ca²⁺] between 100 μm and 1 mm (Max Chelator Winmaxc version 2.51). To establish the apparent Ca²⁺ dependence of STIM1 oligo/de-oligomerization, STIM1-FRET and D1_ER responses obtained in parallel experiments were aligned according to the time of drug application, and FRET change versus [Ca²⁺]_ER was fitted using a sigmoid function. For experiments using cytosolic Ca²⁺ buffer, cells were incubated with 20 μm BAPTA-AM for 30 min prior to the measurements. Confocal images (Figs. 1–3) were acquired on an Axiovert 200M (Zeiss) using a spinning wheel disk (Nipkow disc; QLC-100, Visitech Int., UK) with a ×63 oil-immersion objective, and collected by a CCD camera (Photometrics). For cameleon or FRET measurements, cells were excited with a 440 nm diode-pumped solid-state (DPSS) laser line (Coherent), and emission fluorescence was collected simultaneously at 480 and 535 nm using a 450-nm dichroic mirror and a 505 dcxr-D480/30m-D535/40m beam splitter. Image acquisition and processing were done with MetaFluor (version 6.1; Molecular Devices). All the images were background corrected and ratios were divided by the mean ratio recorded during the initial 2 min to minimize the variability between cells (R/R_o).

FIGURE 1.

Apparent [Ca²⁺]_ER dependence of STIM1 oligomerization. A, changes in FRET signal between CFP-STIM1 and YFP-STIM1 (upper panel) and in D1_ER ratio fluorescence (lower panel) were measured by confocal microscopy in HeLa cells exposed to 1 μm TG to induce store depletion. Traces are average of 43 and 53 FRET and D1_ER recordings, respectively. B, in situ Ca²⁺ calibration of the D1_ER probe in HeLa cells (n = 33–90). C, apparent [Ca²⁺]_ER dependence of STIM1 oligomerization (K_d = 306 μm) induced by TG, established after converting the D1_ER ratio values to [Ca²⁺]_ER.

FIGURE 2.

[Ca²⁺]_ER dependence of STIM1 oligomerization during rapid and slow store depletion. A, upper panel, average FRET responses evoked by 1 μm TG (black squares, n = 43), 15 μm BHQ (gray triangles, n = 21), and 50 μm histamine (light gray circles, n = 27), applied in Ca²⁺-containing (TG and BHQ) or Ca²⁺-free medium (histamine). Lower panel, average D1_ER responses evoked by TG (n = 53), BHQ (n = 18), and histamine (n = 40). Note the different time courses of FRET increase and store depletion. B, statistical evaluation of the FRET (upper panel) and [Ca²⁺]_ER (lower panel) changes evoked by TG, BHQ, and histamine. The amplitude of the FRET increase was comparable but the three agents decreased by [Ca²⁺]_ER to different extents. C, apparent [Ca²⁺]_ER dependence of STIM1 oligomerization evoked by the three protocols of store depletion. TG data are from Fig. 1.

FIGURE 3.

Apparent [Ca²⁺]_ER dependence of STIM1 de-oligomerization. A, upper panel, average FRET response (n = 12) during transient application of the reversible SERCA inhibitor BHQ (15 μm), removed in Ca²⁺-containing medium to promote stores refilling. Lower panel, average D1ER response (n = 10) evoked by this protocol. B, [Ca²⁺]_ER-FRET relationship during Ca²⁺ readmission to cells depleted with BHQ (closed circles) or histamine (open circles, n = 15–22). C, [Ca²⁺]_ER-FRET relationship during readmission of 2 mm Ca²⁺ (closed circles, n = 15–22) or 20 mm Ca²⁺ (open circles, n = 12–18) to cells depleted with histamine. Data at 2 mm are reported from panel B. D, aggregate data of all the [Ca²⁺]_ER-FRET relationships shown in Fig. 2 (depletion) and this figure (refilling), showing that STIM1 de-oligomerization occurs at lower [Ca²⁺]_ER than oligomerization.

Rapid Ca²⁺ Imaging

Ca²⁺ microdomain images were acquired using a ×60 oil objective (CFI Plan Apo) of a Nikon A1 microscope (Nikon Instruments, Japan) with the Perfect focus system and equipped with a PMT detector. Prior to the experiment, cells expressing mCherry-STIM1 were incubated with standard solution containing 20 μm BAPTA-AM, 4 μm Fluo-8/AM, and 0.25 mm sulfinpyrazone for 30 min at room temperature. Cells were excited at 488 and 561 nm of a LU4 4-laser unit, and the emission was collected through a 405/488/561/638 dichroic mirror, 525/50 and 595/50 filters. The acquisition was controlled by NIS-Elements software (Nikon).

TIRF Imaging

TIRF (total internal reflection fluorescence) images were obtained on an Axiovert microscope (100M, Zeiss) equipped with a combined epifluorescence/TIRF adapter (TILL Photonics, Germany) using a ×100 oil immersion objective. Cells expressing GFP-STIM1 were excited at 488 nm of a 50 milliwatt tunable laser (Sapphire LP, Coherent) and the emission was collected by a 12-bit CCD camera (Orca 4742-95-ER, Hamamatsu Photonics). Acquisition of images was conducted with Openlab software (version 3.1.7, PerkinElmer Life Sciences).

Statistics

Mean ± S.E. are shown with significance of differences determined by two-tailed Student's t test for unpaired samples. p < 0.05 was considered to be significant.

RESULTS

[Ca²⁺]_ER Dependence of STIM1 Oligomerization

Several studies have shown that the formation of STIM1 higher order oligomers is the initial step in the signaling cascade that couples ER Ca²⁺ store depletion to SOCE activation (16, 30, 33), but whether the reversible SOCE process is solely controlled by changes in [Ca²⁺]_ER is not known. To address this question, we correlated the changes in free luminal ER Ca²⁺ concentrations ([Ca²⁺]_ER) to the extent of STIM1 oligomerization in HeLa cells treated with SERCA pump inhibitors or with Ca²⁺ mobilizing agonists to irreversibly or reversibly deplete their ER Ca²⁺ stores. The degree of STIM1 oligomerization was quantified as the increase in FRET signal in cells coexpressing CFP-STIM1 and YFP-STIM1 (Fig. 1A, upper panel), and [Ca²⁺]_ER changes were measured in parallel experiments with the ER-targeted cameleon probe D1_ER (Fig. 1A, lower panel). Fig. 1A shows the parallel changes in [Ca²⁺]_ER and in FRET, aligned to the time of drug addition, during irreversible depletion of ER Ca²⁺ stores with the SERCA pump inhibitor thapsigargin (TG). In situ calibration of the D1_ER probe in cells equilibrated at fixed [Ca²⁺]_ER concentrations with ionophores confirmed that the D1_ER F₅₃₅/F₄₈₀ ratio fluorescence increased linearly with [Ca²⁺]_ER over a wide range of Ca²⁺ concentrations (Fig. 1B), and enabled to convert the D1_ER ratio values to [Ca²⁺]_ER. This enabled us to express the STIM1 FRET change as a function of the Ca²⁺ concentration within the ER lumen (Fig. 1C). A sigmoid fit of the FRET-[Ca²⁺]_ER relationship obtained during irreversible store depletion with TG yielded an apparent [Ca²⁺]_ER dependence of 306 ± 10.2 μm for the oligomerization of STIM1 (Fig. 1C, dotted line).

During the oligomerization process, STIM1 molecules are recruited to subplasmalemmal cortical ER domains where they bind and activate Orai1 channels (12, 17, 19, 20, 31). As a result of this translocation, different confocal cellular sections can be either enriched or depleted of STIM1 molecules during store depletion, depending on their proximity to the plasma membrane. To test whether the choice of the confocal plane affected our FRET signal, we compared the responses obtained by imaging a cellular section encompassing the nucleus to the responses obtained by imaging a TIRF-like plane at the cell-substrate interface. Upon TG addition, the fluorescence of both CFP-STIM1 and YFP-STIM1 decayed in the trans-nuclear plane (supplemental Fig. S1A), whereas in the near membrane plane the YFP-STIM1 signal increased and the CFP-STIM1 signal slightly was decreased (supplemental Fig. S1B). Despite the opposite changes in fluorescence intensities, the changes in the FRET ratio had identical kinetics when measured at the cell center or near the plasma membrane (supplemental Fig. S1C). These experiments confirm that the fluorescent STIM1 molecules are recruited to the plasma membrane, and indicate that translocation of STIM1 does not perturb the FRET signal. The FRET changes exhibited a higher dynamics at the plasma membrane (supplemental Fig. S1B), but the transnuclear plane was less prone to motion artifacts during long-term recordings and was therefore selected for all subsequent confocal FRET recordings.

To test whether the Ca²⁺ dependence of STIM1 oligomerization varies depending on the speed of the ER depletion, we treated cells with BHQ, a slow-acting SERCA inhibitor, and with histamine, an IP₃-generating agonist that rapidly releases Ca²⁺ from stores. As shown in Fig. 2A, BHQ induced a slower and less extensive decrease in [Ca²⁺]_ER than TG, whereas histamine, applied in Ca²⁺-free medium to maximize store depletion, decreased [Ca²⁺]_ER more rapidly than TG (Fig. 2A, bottom panel). The different [Ca²⁺]_ER kinetics were mirrored by the kinetics of the FRET change between CFP-STIM1 and YFP-STIM1 (Fig. 2A, top panel). Interestingly, although the three compounds depleted the stores to different levels (Fig. 2B, lower panel), the maximal amplitude of the FRET increase was similar (Fig. 2B, upper panel). As shown in Fig. 2C, the FRET-[Ca²⁺]_ER relationships of cells treated with TG and BHQ were similar and overlapped with the FRET-[Ca²⁺]_ER relationship of cells exposed to histamine, which had a steeper slope. These data indicate that the [Ca²⁺]_ER dependence of STIM1 oligomerization is not related to the speed of ER depletion and is largely independent of the stimulus used to deplete the ER.

[Ca²⁺]_ER Dependence of STIM1 De-oligomerization

Having established the [Ca²⁺]_ER dependence of STIM1 oligomerization, we took advantage of the reversibility of histamine and BHQ to study the Ca²⁺ dependence of STIM1 de-oligomerization. For this, cells were exposed to histamine or BHQ in Ca²⁺-free medium to decrease [Ca²⁺]_ER and increase STIM1 FRET levels, and the drugs were then washed off with a saline solution containing 2 mm Ca²⁺ to promote store refilling. As shown in Fig. 3A, Ca²⁺ readmission to cells depleted with BHQ led to a rapid ER refilling and a concomitant decrease in the STIM1 FRET signal. More rapid [Ca²⁺]_ER and FRET responses were observed when histamine was used to transiently deplete the stores (supplemental Fig. S2). As shown in Fig. 3B, the FRET-[Ca²⁺]_ER relationships of the ER refilling phase were similar in cells exposed to the agonist or the SERCA inhibitor (K_d: 152 ± 6.6 and 172 ± 5.5 μm for histamine and BHQ, respectively). This suggests that the [Ca²⁺]_ER dependence of STIM1 de-oligomerization does not depend on the type of stimulus used to deplete the stores. To verify that the [Ca²⁺]_ER dependence of STIM1 de-oligomerization did not vary with the speed of ER refilling, we increased the extracellular Ca²⁺ concentrations from 2 to 20 mm to facilitate the entry of Ca²⁺ into cells. Readmission of 20 mm Ca²⁺ to cells depleted with histamine caused a more extensive decline in the STIM1 FRET signal than readmission of 2 mm Ca²⁺, the FRET signal decreasing below basal levels (supplemental Fig. S2). However, the FRET-[Ca²⁺]_ER relationships obtained at low and high external Ca²⁺ were nearly identical (K_d: 152 ± 6.6 and 177 ± 2.3 μm for 2 and 20 mm, respectively; Fig. 3C). Altogether, these data indicate that [Ca²⁺]_ER dependence of STIM1 de-oligomerization does not depend on the speed of ER replenishment.

Because the FRET-[Ca²⁺]_ER relationships appeared to be independent on the stimulus used, we aggregated all the data collected during store depletion and refilling to better compare the [Ca²⁺]_ER dependence of the oligomerization and de-oligomerization of STIM1. Fig. 3D shows the aggregate data obtained under various store depletion conditions (filled circles: 1 μm TG or 15 μm BHQ in Ca²⁺ containing medium; 15 μm BHQ or 50 μm histamine in Ca²⁺-free medium) and store refilling conditions (open circles: readmission of 2 or 20 mm Ca²⁺ following histamine or BHQ exposure). The results highlight the different [Ca²⁺]_ER dependence of the STIM1 oligomerization and de-oligomerization processes previously reported in endothelial cells (32) and confirm that during store refilling STIM1 de-oligomerizes at lower [Ca²⁺]_ER levels than those required for STIM1 oligomerization (K_d, 281 ± 2.6 and 165 ± 3.2 μm for depletion and refilling, respectively; Fig. 3D).

ER Calcium Replenishment Is Not Sufficient to Induce STIM1 De-oligomerization

The left-shifted [Ca²⁺]_ER dependence of STIM1 de-oligomerization could reflect a higher sensitivity of the STIM1 dissociation process to changes in luminal Ca²⁺, or the presence of additional factors promoting de-oligomerization during store refilling. Elevations in cytosolic Ca²⁺ have been shown to prevent the recruitment of STIM1 to the plasma membrane (32), and might therefore increase the STIM1 de-oligomerization process. [Ca²⁺]_Cyt elevations are unavoidable during store refilling, because the Ca²⁺ ions flowing across SOCE channels must enter the cytosol to be taken up by SERCA. To minimize the cytosolic Ca²⁺ elevations and promote ER refilling in the absence of Ca²⁺ readmission, we loaded cells with BAPTA-AM, a membrane-permeable Ca²⁺ chelator that accumulates inside cells and provides a large reservoir of buffered Ca²⁺ in the cytoplasm (34, 35). Predictably, loading the cells with BAPTA-AM prevented the cytosolic Ca²⁺ elevations evoked by our Ca²⁺ depletion and readmission protocol (supplemental Fig. 2A), indicating that cytosolic Ca²⁺ was effectively buffered. To test whether the exogenous Ca²⁺ buffer could provide sufficient Ca²⁺ for ER store refilling, we transiently exposed cells to the reversible inhibitor BHQ in the absence of external Ca²⁺. As shown in Fig. 4A, BHQ removal led to a rapid and complete refilling of the stores in BAPTA-loaded cells (gray trace, *2). In contrast, in cells not exposed to BAPTA the stores remained depleted upon BHQ removal and only refilled upon readmission of external Ca²⁺ (black trace, *3). Statistical evaluation confirmed that BAPTA loading increased [Ca²⁺]_ER as efficiently as Ca²⁺ readmission (Fig. 4B, *2 and *3). These data indicate that, as expected, the SERCA pumps were able to refill the store by extracting the Ca²⁺ buffered by the cytosolic Ca²⁺ chelator. Therefore, this protocol enabled us to test whether ER Ca²⁺ refilling alone, in the absence of external Ca²⁺ readmission, was sufficient to induce STIM de-oligomerization. Strikingly, despite the complete replenishment of the ER in BATPA-loaded cells, the FRET signal between STIM1 molecules remained elevated, suggesting that STIM1 stayed oligomerized (Fig. 4, C and D, *2). Subsequent readmission of Ca²⁺ evoked a rapid decrease in the FRET signal, whose kinetics and amplitude were undistinguishable from the decrease observed in control cells (Fig. 4, C and D, *3). The increased STIM1 FRET signal of cells with BAPTA-refilled stores might reflect the persistence of oligomerized STIM1 at the cortical ER or the failure of STIM1 to de-oligomerize during its retrieval from the cortical ER. To distinguish between these two possibilities, we followed the formation and disappearance of YFP-STIM1 clusters in cells loaded with BAPTA-AM. As shown in Fig. 4E, upon BHQ addition YFP-STIM1 formed visible membrane puncta that persisted upon washout of the inhibitor (bottom panel, *2) and that disappeared upon Ca²⁺ readmission (*3). Thus, in cells loaded with BAPTA, STIM1 remained in cortical ER structures and its de-oligomerization and disappearance from the cortical ER appear to be coordinated events induced by the readmission of Ca²⁺ to cells.

FIGURE 4.

ER refilling does not initiate STIM1 de-oligomerization in BATPA-loaded cells. Cells were loaded with loaded 20 μm BAPTA-AM to provide a cytosolic reservoir of buffered Ca²⁺ for the refilling of ER Ca²⁺ stores. A, average D1_ER responses of cells transiently stimulated with BHQ (15 μm) in Ca²⁺-free medium and subsequently switched to a Ca²⁺-containing medium. [Ca²⁺]_ER increased immediately upon BHQ removal in BAPTA-loaded cells (gray traces, n = 21) and only increased upon Ca²⁺ readmission in control cells (black traces, n = 30). B, average decrease in D1_ER ratio fluorescence at the time points indicated in A; *, p < 0.001. Note that [Ca²⁺]_ER increased by the same extent in BAPTA-loaded cells kept in Ca²⁺-free (*2, BAPTA) than in control cells exposed to 2 mm Ca²⁺ (*3, ctrl). C, average FRET responses evoked by the same protocol. The FRET signal remained elevated in BAPTA-loaded cells kept in Ca²⁺-free cells (gray traces, n = 20) and decreased upon subsequent readmission of 2 mm Ca²⁺ with kinetics similar to control cells (black traces, n = 20). D, average increase in FRET at the three time points indicated in A. E, TIRF images of BATPA-loaded cells expressing YFP-STIM1 exposed to the same protocol, taken before stimulation and at the three time points indicated in A. Note that the STIM1 puncta persisted after ER refilling (*2) and disappeared upon Ca²⁺ readmission (*3). Images are representative of 4 independent experiments.

Cytosolic Calcium Is Required for STIM1 De-oligomerization

Having established that the de-oligomerization of STIM1 requires the readmission of Ca²⁺ to cells with refilled stores, we aimed to precise the site of action of this exogenous Ca²⁺. To test whether Ca²⁺ was acting on an extracellular or intracellular site, we prevented the entry of Ca²⁺ into cells with a maximally effective concentration of La³⁺, a SOCE channel inhibitor (1). As shown in Fig. 5, A and B, 50 μm La³⁺ did not affect the [Ca²⁺]_ER increase evoked by BHQ removal in BAPTA-loaded cells (*2, gray trace), indicating that the Ca²⁺ trapped by cytosolic BAPTA was efficiently used for ER refilling in the presence of the SOCE inhibitor. In contrast, [Ca²⁺]_ER did not recover when Ca²⁺ was readmitted to control cells in the presence of 50 μm La³⁺ (*3, black trace), indicating that SOCE channels were effectively inhibited by La³⁺. Importantly, the STIM1 FRET signal remained elevated upon Ca²⁺ readmission in both control and BAPTA-loaded cells (Fig. 5, C and D), indicating that a high extracellular Ca²⁺ concentration is in itself not sufficient to induce STIM1 de-oligomerization. Identical results were obtained with 10 μm Gd³⁺ (Fig. 5, E and F), whereas BTP2 and SK&F 96365 did not prevent ER refilling upon Ca²⁺ readmission (supplemental Fig. S4) indicating that these two SOCE inhibitors inhibit Ca²⁺ entry less efficiently than La³⁺ and Gd³⁺. These results strongly suggest that Ca²⁺ is acting from the intracellular, cytosolic side, and therefore that a cytosolic Ca²⁺ elevation is necessary to initiate the retrieval of STIM1 from the cortical ER.

FIGURE 5.

SOCE channel inhibition prevents Ca²⁺-induced STIM1 de-oligomerization. Cells were treated as described in the legend to Fig. 4, using bath solutions containing 50 μm La³⁺ (A–D) or 10 μm Gd³⁺ (E and F) to inhibit SOCE channels. A, average D1_ER responses of control (n = 36) and BAPTA-loaded cells (n = 35). B, average decrease in the D1_ER ratio fluorescence at the time points indicated in A; *, p < 0.001. C, average FRET responses evoked by the same protocol. The FRET signal remained elevated during Ca²⁺ readmission to control (n = 11) and BATPA-loaded cells (n = 14). D, average increase in FRET at the three time points indicated in A. E, average FRET responses recorded with 10 μm Gd³⁺ instead of La³⁺. The FRET signal remained elevated in both control (n = 5) and BATPA-loaded cells (n = 29). F, average increase in FRET at the three time points indicated in E.

To confirm that a localized cytosolic Ca²⁺ elevation occurred during Ca²⁺ readmission to cells loaded with BAPTA, we performed confocal imaging with Fluo-8, a Ca²⁺-sensitive dye whose fluorescence increases markedly upon Ca²⁺ binding. To localize the Ca²⁺ entry sites, cells were transfected with mCherry-STIM1 and the changes in mCherry and Fluo-8 fluorescence were measured concomitantly during transient depletion of stores with BHQ. As shown in Fig. 6, store depletion induced the apparition of mCherry fluorescence clusters that persisted upon removal of the reversible inhibitor (top, second column). Ca²⁺ readmission to these BAPTA-loaded cells expressing mCherry-STIM1 evoked localized [Ca²⁺]_Cyt elevations that co-localized with the mCherry fluorescent aggregates but that were of slightly smaller spatial dimension (bottom, third column). Line scan confocal imaging revealed that the red STIM1 clusters broke apart at the peak of the [Ca²⁺]_Cyt elevations and disappeared entirely shortly after the [Ca²⁺]_Cyt elevations had waned (Fig. 6B). Quantification of these images showed that the disappearance of the fluorescent STIM1 clusters correlated with the presence of a local Ca²⁺ elevation (Table 1). Nearly two-thirds of the STIM1 clusters located in the vicinity of a Ca²⁺ hot spot disappeared during the readmission period, whereas the same proportion of STIM1 clusters devoid of Ca²⁺ hot spots remained stable. Due to the small number of STIM1 clusters lacking Ca²⁺ hot spots the relative risk could not be calculated, but the presence of a Ca²⁺ hot spot was associated with an increased probability of STIM1 cluster disappearance (odds ratio: 3.6). These experiments confirm that, in cells loaded with Ca²⁺ chelators, local [Ca²⁺]_Cyt elevations occur at Ca²⁺ entry sites and coincide temporally with the disappearance of STIM1 clusters.

FIGURE 6.

Local cytosolic Ca²⁺ elevations evoked by Ca²⁺ readmission to BATPA-loaded cells. Cells expressing mCherry-STIM1 were loaded with Fluo-8/AM and BAPTA-AM and imaged by confocal microscopy during transient BHQ application in Ca²⁺-free and subsequent Ca²⁺ readmission. A, mCherry fluorescence (top) illustrating the apparition of the STIM1 cluster upon BHQ-induced store depletion and corresponding changes in Fluo-8 fluorescence (bottom). Ca²⁺ readmission evoked a local [Ca²⁺]_Cyt elevation that overlapped with the mCherry fluorescence cluster. B, line scan confocal image of the mCherry and Fluo-8 fluorescence changes evoked by the readmission of 0.5 mm Ca²⁺. Distance along the 39-μm scan line is depicted vertically in the image, time runs from left to right, and increasing fluorescence intensities (corresponding to increasing free [Ca²⁺]_Cyt for the bottom panel) are depicted by increasingly brighter colors. Note that the disappearance of the mCherry fluorescence cluster coincides temporally with the local [Ca²⁺]_Cyt elevation.

TABLE 1.

Ca²⁺ hot spots and STIM1 stability

STIM1 clusters shown in Fig. 6 were imaged for 10 min after Ca²⁺ readmission and classified as “stable” or “unstable” if they remained visible or if they had disappeared at the end of the recording. The presence of a Ca²⁺ hot spot in the same spatial domain during the 10-min Ca²⁺ readmission period was then determined to establish the association between Ca²⁺ hot spots and cluster stability. All the Ca²⁺ hot spots were associated with a STIM1 cluster, and no new STIM1 clusters were formed during the readmission period. Two-thirds of the STIM1 clusters associated with a Ca²⁺ hot spot disappeared during the readmission period, whereas two-thirds of the STIM1 clusters lacking Ca²⁺ hot spots remained stable (n = 27 cells from 10 independent recordings, odds ratio: 3.6).

No. of events		STIM1 puncta
		Stable	Unstable	Total
Ca2+ hot spot	Yes	12	26	38
	No	5	3	8
	Total	17	29	46

DISCUSSION

In this study, we show that during the activating phase of SOCE, [Ca²⁺]_ER levels are inversely correlated with the degree of STIM1 oligomerization irrespective of the kinetics of store depletion evoked by agonists or SERCA inhibitors. In sharp contrast, an increase in [Ca²⁺]_ER was not sufficient to initiate STIM1 de-oligomerization, which required a local elevation of the cytosolic Ca²⁺ concentration (Fig. 7).

FIGURE 7.

Schematic representation of the STIM1 movement in BAPTA-AM loaded cells. 1) at resting [Ca²⁺]_ER, STIM1 is in the Ca²⁺-bound state and evenly distributed on ER membranes. 2) upon SERCA inhibition, the decrease in [Ca²⁺]_ER induces STIM1 oligomerization and accumulation in PM-associated clusters, where it binds and activates Orai channels. In BAPTA-loaded cells, most of the Ca²⁺ released from the ER is chelated by the cytosolic BAPTA (blue crescents) and provides an intracellular reservoir of buffered Ca²⁺. 3) relief of SERCA inhibition enables ER refilling as SERCA extract the Ca²⁺ bound to the cytosolic BAPTA, but the restoration of resting [Ca²⁺]_ER levels does not trigger STIM1 de-oligomerization. 4) upon Ca²⁺ re-admission a local cytosolic Ca²⁺ elevation triggers the de-oligomerization of STIM1 and its retrieval from PM-associated clusters.

Previous studies have shown that the extent of SOCE activation is inversely related to [Ca²⁺]_ER levels (30, 32). Accordingly, we observed a strict inverse correlation between [Ca²⁺]_ER and the degree of STIM1 oligomerization, which increased as soon as [Ca²⁺]_ER decreased below basal levels and reached half-maximum at a [Ca²⁺]_ER of ∼280 μm. Lower [Ca²⁺]_ER values (∼200 μm) were reported for half-maximal I_CRAC current activation in Jurkat T cells (30) and for half-maximal STIM1 oligomerization in endothelial cells (32), but in these two studies the resting [Ca²⁺]_ER levels were also significantly lower (∼450 versus ∼520 μm in our HeLa cells). Our higher [Ca²⁺]_ER values do not appear to reflect the poor sensitivity of the D1_ER probe (K_d ∼60 μm) at high micromolar Ca²⁺ concentrations, because the calibration curve was linear in the pCa range 3–6 (Fig. 1B). Despite the discrepancies in absolute [Ca²⁺]_ER values, we also observed that the extent of STIM1 oligomerization was directly proportional to the degree of depletion of ER Ca²⁺ stores, with half-maximal oligomerization occurring when the stores were depleted by 50%. By using agonists or SERCA inhibitors to actively or passively deplete Ca²⁺ stores, we further demonstrate that the [Ca²⁺]_ER dependence of STIM1 oligomerization is independent of the kinetics and mechanisms of ER Ca²⁺ release. More importantly, we show that the [Ca²⁺]_ER dependence of the association and dissociation processes differ markedly, a difference that we could attribute to the need for cytosolic Ca²⁺ elevations to initiate the de-oligomerization process and the disassembly of STIM1 puncta.

The influx of Ca²⁺ across PM channels is both the justification of existence of SOCE and a pre-requisite for its termination, as it provides the ions used for Ca²⁺ replenishment of the ER. The opening of Ca²⁺ permeable channels therefore precedes ER refilling and conveys the signals that subsequently trigger the termination of SOCE. To separate these two intricate processes, we designed a protocol that enables refilling of the ER without adding exogenous Ca²⁺, thereby bypassing the Ca²⁺ entry phase. We expected that the restoration of a normal ER Ca²⁺ content by the provision of buffered cytosolic Ca²⁺ ions would induce the disaggregation of STIM1 molecules and the disassembly of STIM-Orai1 complexes, because Ca²⁺ binding to STIM1 EF-SAM domains reverse the conformational change induced by the unbinding of Ca²⁺ that leads to STIM1 oligomerization, the key step in the activation process that culminates in the opening of PM channels (30). Early mutagenesis studies showed that Ca²⁺ binding to STIM1 EF-SAM domains stabilizes the molecule in an inactive state (12, 31), whereas subsequent structural studies revealed that the combined EF-SAM domains form a compact structure in the Ca²⁺-bound state (33) and that Ca²⁺ unbinding leads to a more open configuration that initiates the oligomerization of the luminal domain (16, 36). The reversibility of the conformational changes leading to STIM1 oligomerization suggest that the whole SOCE process is reversibly regulated by Ca²⁺ sensitivity of the luminal STIM domains, and hence by changes in [Ca²⁺]_ER (36). Contrary to this prediction, our experiments in BAPTA-loaded cells showed that the full restoration of normal high [Ca²⁺]_ER levels is not sufficient to initiate the dissociation of STIM1-STIM1 oligomers or the retrieval of STIM1 clusters from the PM (Fig. 7). Despite full replenishment of the ER, the exogenous STIM1 fusion proteins were still associated as evidenced by their high FRET signal, and remained within the same membrane-associated puncta located within a TIRF plane of 100 nm width (Fig. 4). Dissociation was only observed upon readmission of external Ca²⁺ (Fig. 7) and this effect was prevented by the SOC channel blocker La³⁺, indicating that Ca²⁺ acts on the cytosolic side. In BAPTA-loaded cells, Ca²⁺ readmission did not cause global cytosolic Ca²⁺ elevations (supplemental Fig. S3) but evoked localized Ca²⁺ elevations restricted to STIM1 puncta (Fig. 6). BAPTA loading was used previously to resolve Ca²⁺ entry sites in Jurkat T cells, revealing discrete hot spots of high [Ca²⁺]_Cyt localized in the immediate vicinity of STIM1 puncta (27, 28). Our data now indicate that these highly localized increases in cytosolic Ca²⁺ can trigger the dissociation of STIM1 oligomers and that de-oligomerization of STIM1 occurs coordinately with its retrieval from the plasma membrane. This indicates that the Ca²⁺ ions entering across SOCE channels act on target molecule(s) located in immediate proximity to the STIM1-Orai membrane complexes.

Although local elevation in the cytosolic Ca²⁺ concentration is a pre-requisite for the dissociation of STIM1 oligomers, cytosolic Ca²⁺ elevations are not sufficient to terminate SOCE. It is well known that TG, the drug most commonly used to activate SOCE and study STIM1 translocation, evokes large and long-lasting elevations in cytosolic Ca²⁺ when applied in Ca²⁺-rich medium (1, 37). Despite persistent [Ca²⁺]_Cyt elevation, treatment with TG induces a massive formation of subplasmalemmal STIM1 clusters and cortical ER (29). Accordingly, STIM1 remained in membrane-associated clusters and the FRET signal did not decrease when Ca²⁺ was readmitted to cells treated with TG, despite a massive global increase in cytosolic Ca²⁺ (data not shown). These data indicate that an increase in [Ca²⁺]_ER is required, but is not sufficient, to terminate SOCE and reverse the conformational changes induced by the unbinding of Ca²⁺ from the STIM1 EF-SAM luminal domain. Our data now show that Ca²⁺ acts not only on the luminal domain of STIM1 but also on a cytosolic target, and that both effects are required for proper termination of SOCE (Fig. 7).

The small size of the local Ca²⁺ elevations occurring in BAPTA-loaded cells indicates that the cytosolic Ca²⁺ targets are likely to be molecules associated with the STIM1-Orai1 complex. CRACR2, a cytosolic Ca²⁺-binding protein, was reported to stabilize STIM1-Orai1 complexes and dissociate upon Ca²⁺ binding (38), making this molecule a potential target for the Ca²⁺-dependent inactivation process that we report here. However, the mRNA levels of the two CRACR isoforms were either undetectable (CRACR2A) or very low (CRACR2B) in our HeLa cells, and CRACR2B silencing did not promote STIM1 de-oligomerization during BAPTA-mediated ER refilling (data not shown). Besides CRACR2, SERCA were shown to co-localize with STIM upon store depletion (39, 40) and calmodulin to bind to the polybasic cytosolic tail of STIM1 in a Ca²⁺-dependent manner (41). The most likely molecular target of the local cytosolic Ca²⁺ elevations, however, is the STIM1 protein itself, because several STIM1 C-terminal domains are involved in productive electrostatic interactions with either the plasma membrane or PM channels. The requirements for exogenous Ca²⁺ persisted in STIM1 mutants lacking the lysine-rich C-terminal domain (data not shown), suggesting that the interactions between the lysine-rich C terminus and membrane phospholipids are not the target of the Ca²⁺ inactivation. Other candidates that remain to be tested include the CRAC modulatory domain (amino acids 474–485), a sequence of seven negatively charged amino acids whose neutralization diminished the fast Ca²⁺-dependent inactivation of the Orai channel (42). The acidic region was situated on the first coiled-coil domain of STIM, which has an autoinhibitory function by masking the CAD/STIM-Orai activating region domain, and may also contribute to the inactivation of SOCE (23). Finally, the electrostatic interactions between the CAD/STIM-Orai activating region domain and the C terminus of the Orai1 channels might be disrupted by a local elevation in cytosolic Ca²⁺. Future mutagenesis studies of STIM1 and Orai1 cytosolic domains will precisely indicate the target of the Ca²⁺ inhibition on the assembled SOCE complex.

One potential caveat of [Ca²⁺]_ER measurements with genetically encoded Ca²⁺ indicators is that the specialized domains derived from the ER that form during SOCE are not accessible to proteins bearing KDEL ER-retention signals. We previously showed that GFP fusion proteins targeted to the ER by virtue of KDEL retention signals do not co-localize with STIM1 and that chaperones such as BiP are excluded from the cortical ER (29). The D1_ER probe that we used here to measure [Ca²⁺]_ER contains a KDEL ER retention signal and is also excluded from subplasmalemmal ER regions enriched in STIM1.³ Despite this limitation, our spatially averaged D1_ER measurements likely respond to [Ca²⁺]_ER changes occurring within the cortical ER as this specialized ER compartment is connected with the conventional ER (29). Ca²⁺ moves rapidly within the ER lumen (43, 44), thus the local [Ca²⁺]_ER within the cortical ER is expected to equilibrate with the global [Ca²⁺]_ER of the bulk ER. Given its proximity to the PM, however, the cortical ER might be affected by the Ca²⁺ drag exerted by neighboring plasma-membrane Ca²⁺ pumps, whose function is to extrude Ca²⁺ from cells. During ER refilling with BAPTA, the cortical ER might therefore remain in a depleted condition as plasma-membrane Ca²⁺ pumps compete with SERCA for the extraction of the Ca²⁺ ions bound to the exogenous cytosolic buffer. [Ca²⁺]_ER measurements with a Ca²⁺ probe that can access the cortical ER will be required to clarify this issue.

In a previous study, cytosolic Ca²⁺ elevations were shown to inhibit the formation of subplasmalemmal STIM1 clusters (32). Interestingly, the high Ca²⁺ concentrations had no effects on the kinetics of oligomerization or on the formation of STIM1 aggregates in deep ER regions, but specifically prevented the formation and/or translocation of STIM1 clusters at ER-PM junctions, an effect that was attributed to the reduced velocity of STIM1 comets at high cytosolic Ca²⁺. The lack of effects of [Ca²⁺]_Cyt on oligomerization are consistent with our observation that [Ca²⁺]_ER dependence of STIM1 oligomerization is similar in cells depleted with agonists or SERCA inhibitors, agents that cause cytosolic Ca²⁺ responses of different amplitude and duration. The inhibitory effects of [Ca²⁺]_Cyt on subplasmalemmal STIM1 clustering does not appear to be a dominant feature of the SOCE process because, as mentioned above, STIM1 forms membrane clusters and is recruited to the cortical ER in cells treated with TG, despite large elevations in cytosolic Ca²⁺. Our finding that a cytosolic Ca²⁺ elevation is required to induce the de-oligomerization of STIM1 might account for the delayed apparition of new clusters at high [Ca²⁺]_Cyt reported in this study (32). Accordingly, these authors also observed that [Ca²⁺]_Cyt elevations promoted the disassembly of pre-existing STIM1 clusters. The main effect of cytosolic Ca²⁺ elevations is therefore not to inhibit the formation of subplasmalemmal STIM1 clusters but to induce the de-oligomerization of STIM1.

In summary, we show here that the restoration of physiological [Ca²⁺]_ER levels is not sufficient to induce de-oligomerization of STIM1 or the disaggregation of STIM1 membrane clusters, and that both processes require a local increase in cytosolic Ca²⁺ in the vicinity of plasma membrane STIM1 clusters. Ca²⁺ act at two distinct sites on both sides of the ER membrane to control the retrieval of STIM1 from the plasma membrane and the termination of SOCE.