STIM1 Knockdown Reveals That Store-operated Ca2+ Channels Located Close to Sarco/Endoplasmic Ca2+ ATPases (SERCA) Pumps Silently Refill the Endoplasmic Reticulum*

Stromal interaction molecule (STIM) proteins are putative ER Ca2+ sensors that recruit and activate store-operated Ca2+ (SOC) channels at the plasma membrane, a process triggered by the Ca2+ depletion of the endoplasmic reticulum (ER). To test whether STIM1 is required for ER refilling, we used RNA interference and measured Ca2+ signals in the cytosol, the ER, and the mitochondria of HeLa cells. Knockdown of STIM1 (mRNA levels, 73%) reduced SOC entry by 73% when sarco/endoplasmic Ca2+ ATPases (SERCA) were inhibited by thapsigargin but did not prevent Ca2+ stores refilling when cells were stimulated by physiological agonists. Stores could be fully refilled by increasing the external Ca2+ concentration above physiological values, but no cytosolic Ca2+ signals were detected during store refilling even at very high Ca2+ concentrations. [Ca2+]ER measurements revealed that the basal activity of SERCA was not affected in STIM1 knockdown cells and that [Ca2+]ER levels were restored within 2 min in physiological saline following store depletion. Mitochondrial inhibitors reduced ER refilling in wild-type but not in STIM1 knockdown cells, indicating that ER refilling does not require functional mitochondria at low STIM1 levels. Our data show that ER refilling is largely preserved at reduced STIM1 levels, despite a drastic reduction of store-operated Ca2+ entry, because Ca2+ ions are directly transferred from SOC channels to SERCA. These findings are consistent with the formation of microdomains containing not only SOC channels on the plasma membrane and STIM proteins on the ER but also SERCA pumps and mitochondria to refill the ER without perturbing the cytosol.

Stromal interaction molecule (STIM) proteins are putative ER Ca 2؉ sensors that recruit and activate store-operated Ca 2؉ (SOC) channels at the plasma membrane, a process triggered by the Ca 2؉ depletion of the endoplasmic reticulum (ER). To test whether STIM1 is required for ER refilling, we used RNA interference and measured Ca 2؉ signals in the cytosol, the ER, and the mitochondria of HeLa cells. Knockdown of STIM1 (mRNA levels, 73%) reduced SOC entry by 73% when sarco/endoplasmic Ca 2؉ ATPases (SERCA) were inhibited by thapsigargin but did not prevent Ca 2؉ stores refilling when cells were stimulated by physiological agonists. Stores could be fully refilled by increasing the external Ca 2؉ concentration above physiological values, but no cytosolic Ca 2؉ signals were detected during store refilling even at very high Ca 2؉ concentrations. [Ca 2؉ ] ER measurements revealed that the basal activity of SERCA was not affected in STIM1 knockdown cells and that [Ca 2؉ ] ER levels were restored within 2 min in physiological saline following store depletion. Mitochondrial inhibitors reduced ER refilling in wild-type but not in STIM1 knockdown cells, indicating that ER refilling does not require functional mitochondria at low STIM1 levels. Our data show that ER refilling is largely preserved at reduced STIM1 levels, despite a drastic reduction of store-operated Ca 2؉ entry, because Ca 2؉ ions are directly transferred from SOC channels to SERCA. These findings are consistent with the formation of microdomains containing not only SOC channels on the plasma membrane and STIM proteins on the ER but also SERCA pumps and mitochondria to refill the ER without perturbing the cytosol. Ca 2ϩ signals generated by the release of Ca 2ϩ ions from the endoplasmic reticulum (ER) 2 regulate essential cellular func-tions such as secretion, contraction, and gene transcription. The depletion of ER Ca 2ϩ stores, in turn, activates Ca 2ϩ -permeable channels at the plasma membrane to ensure long term signaling. This mechanism of store-operated Ca 2ϩ entry was described 20 years ago (1), and the prototypic store-operated channel CRAC (for Ca 2ϩ release-activated Ca 2ϩ channel) was extensively characterized (2). However, the mechanism that activates SOC and CRAC channels upon ER depletion has long remained elusive (3).
Two protein families, STIM and Orai, were identified recently as essential for SOC activity (4,5). STIM1 is a highly conserved type I ER membrane protein containing a luminal EF-hand domain and several cytosolic protein-protein interaction domains. Several evidences indicate that STIM1 is the ER Ca 2ϩ sensor that regulates the activity of SOC channels: STIM1 redistributes into ER puncta located 10 -25 nm from the plasma membrane upon Ca 2ϩ store depletion, and puncta formation precedes the activation of CRAC channels by several seconds, consistent with a causal role of STIM1 in SOC activation (6). STIM1 EF-hand mutations constitutively activate SOC and induce puncta formation (4,7). Finally, STIM1 associates with the CRAC channel pore subunit Orai1, an interaction increased by store depletion (8), and also with the hTRPC1 channel (9).
The other protein, Orai1 (also known as CRACM1), was identified by genetic linkage in a subset of patients with severe combined immunodeficiency, who lack functional CRAC channels (10). As observed for STIM1, knockdown of Orai1 drastically reduced both store-operated Ca 2ϩ entry and CRAC current (11,12). Co-expression of STIM1 and Orai1, but not expression of either protein alone, reconstituted store-operated Ca 2ϩ entry and generated massive CRAC currents (12)(13)(14)(15). All three mammalian Orai homologues synergized with STIM1 to augment store-operated Ca 2ϩ entry in the potency order Orai1 3 Orai2 3 Orai3 (15). Mutagenesis studies then conclusively showed Orai1 to be the CRAC channel pore as point mutations in Orai1 transformed the Ca 2ϩ -selective, inwardly rectifying channel into an outwardly rectifying channel permeable to monovalent cations (8,16,17).
During store depletion, STIM1 and Orai1 move in a coordinated fashion to form closely apposed ER-plasma membrane clusters, and the clusters are associated with highly localized increases in subplasmalemmal [Ca 2ϩ ] (18). The STIM-Orai interaction thus restricts Ca 2ϩ influx to specific regions of the plasma membrane located 10 nm away from the ER. This cellular structure resembles the synaptic cleft and creates a diffusion barrier that prevents the escape of Ca 2ϩ ions from the cleft. In these conditions, Ca 2ϩ influx should be barely detectable with cytosolic Ca 2ϩ dyes. Yet cytosolic Ca 2ϩ dyes are used routinely to measure store-operated Ca 2ϩ entry. One possible explanation is that SERCA inhibitors are often used to activate SOC. With SERCA inhibited, Ca 2ϩ ions entering across SOC channels cannot be transferred to the ER and are thus more readily detected in the cytosol. Alternatively, STIM1 might bring together the ER and plasma membrane only in cells that express CRAC channels, such as Jurkat cells.
Here, we tested the hypothesis that SOC influx occurs in membrane clusters closely apposed to the ER by measuring the impact of STIM1 levels on cytosolic, ER, and mitochondrial Ca 2ϩ handling. STIM1 knock-down with siRNA markedly decreased SOC activity in HeLa cells but had surprisingly little impact on ER Ca 2ϩ homeostasis because all incoming Ca 2ϩ ions were directly taken up by SERCA pumps. These findings are consistent with the formation of clusters containing SOC channels on the plasma membrane and both STIM and SERCA proteins on the closely apposed ER membrane to enable efficient ER refilling with minimal changes in cytosolic Ca 2ϩ .

EXPERIMENTAL PROCEDURES
Reagents-Minimum essential medium, fetal calf serum, penicillin, streptomycin, and Lipofectamine 2000 transfection reagent were obtained from Invitrogen. Histamine, thapsigargin, oligomycin, and rotenone were obtained from Sigma, CGP-37157 was from Calbiochem, and UTP was from GE Healthcare. YC3.6 cyto and D1 ER were kindly provided by Drs. Amy Palmer and Roger Tsien, and YC2.1 2mit was provided by Dr. Tullio Pozzan.
Cell Culture and Transfection-HeLa cells were grown in minimum essential medium containing 10% heat inactivated fetal calf serum, 2 mM L-glutamine, 50 units/ml penicillin, 50 g/ml streptomycin at 37°C and 5% CO 2 . For all experiments, cells were plated on 25-mm-diameter glass coverslips and cotransfected with plasmids (2 g) coding for YC probes and dsRNA (80 nM) using Lipofectamine 2000. Cytosolic, ER, and mitochondrial recordings were performed using YC3.6 cyto , D1 ER , and YC2.1 2mit , respectively. All experiments were performed 2 days after transfection.
Quantitative RT-PCR-Two day after co-transfection, cells were harvested by trypzination, washed twice with phosphatebuffered saline, resuspended in phosphate-buffered saline, and subjected to cytofluorometric analysis. Green fluorescent protein-positive cells were sorted using a FACStarϩ (BD Biosciences). Total RNA was isolated from the sorted cells using the NucleoSpin RNA II kit (Macherey-Nagel, Düren, Germany), and 0.5 g of Dnase-treated RNA was used to synthesize cDNA using QuantiTec reverse transcription kit (Qiagen, Hombrechtikon, Switzerland). RT-PCR assays were carried out in an iCycler (Bio-Rad Laboratories) using the TaqMan system in a final volume of 25 l. The reaction mix included 12.5 l of ABsolute QPCR mixes (ABgene), 0.5 M primers, and 0.1 M specific Taqman probe. The sequences of the primer used were as followed: GAPDH, 5Ј-GAAGGTGAAGGTCGGAGTC-3Ј and 5Ј-GAAGATGGTGATGGGATTTC-3Ј; STIM1, 5Ј-TGA-CAGGGACTGTGCTGAAG-3Ј and 5Ј-AAGAGAGGAGGC-CCAAAGAG-3Ј; GAPDH fluorogenic probe, 5Ј-FAM-CAAG-CTTCCCGTTCTCAGCC-TAMRA-3Ј; STIM1 fluorogenic probe, 5Ј-FAM-ACAGACCGGAGTCATCGGCAAGAAG-BHQ1-3Ј. Taqman fluorogenic probe are labeled with 6-carboxyfluorescein (FAM) at the 5Ј-end and with the fluorescent black hole quencher 1 (BHQ1) or with tetramethylrhodamine (TAMRA) used as fluorescent quencher at the 3Ј-end. For quantification, relative standard curves were created for each gene product, and a housekeeping gene, GAPDH, was used for normalization of the concentration. Relative expression was calculated using the 2 -⌬⌬C T method. Statistics-The significance of differences between means was established using the Student's t test for unpaired samples. The level of significance was defined as p Ͻ 0.05.

Effect of STIM1 Knockdown on Calcium Release and Influx-
To decrease cellular STIM1 levels, double-stranded RNAs designed against STIM1 were transiently transfected in HeLa cells, together with a "cameleon" Ca 2ϩ -sensitive fluorescent protein (YC3.6) that was used as a marker of transfection. Fluorescent cells were sorted by flow cytometry to select cells that presumably received the dsRNA, and quantitative RT-PCR was performed. As shown in Fig. 1A, STIM1 mRNA levels were decreased by 73% in fluorescent cells exposed to STIM1 dsRNA when compared with cells exposed to control, scramble dsRNA. To test whether STIM1 knockdown altered Ca 2ϩ handling, we measured the cytosolic Ca 2ϩ responses elicited by agonists and by SERCA inhibitors using ratio imaging of the YC3.6 fluorescence. As shown in Fig. 1B, the initial peak of the histamine response was identical in cells that received control or STIM1 dsRNA, whereas the subsequent plateau phase, which reflects Ca 2ϩ influx, was nearly abrogated in STIM1 knockdown cells. As a result, the integrated Ca 2ϩ response to histamine was inhibited by 72% in STIM1 knockdown cells (Fig.  1C). Consistent with a decreased Ca 2ϩ influx but preserved Ca 2ϩ store content, the addition of thapsigargin in Ca 2ϩ -free medium to passively deplete Ca 2ϩ stores elicited similar responses in control and STIM1 knockdown cells, whereas the response elicited by the further readdition of Ca 2ϩ was severely inhibited in STIM1 knockdown cells, the area under the curve being reduced by 73% (Fig. 1, B and C). These observations confirm previous results showing that Ca 2ϩ influx, but not Ca 2ϩ release from stores, is affected by STIM1 silencing.
Effect of STIM1 Knockdown on the Refilling of Ca 2ϩ Stores-The strongly reduced SOC influx but preserved Ca 2ϩ store content suggested that STIM1 knockdown cells were still able to refill their internal Ca 2ϩ stores. This observation is surprising because the primary role of SOC influx is to sustain ER refilling during physiological stimulations. To study the impact of STIM1 knockdown on the refilling of Ca 2ϩ stores, we extensively depleted Ca 2ϩ stores without altering the activity of SERCA. For this purpose, cells were repeatedly stimulated with histamine in the absence of external Ca 2ϩ ( Fig. 2A, phase i). As shown in Fig. 2, the amplitude and kinetics of the agonist-induced Ca 2ϩ release were equivalent in control and STIM1 knockdown cells, confirming that STIM1 knockdown cells retained a normal Ca 2ϩ store content (Fig. 2, B and C, phase i). After this extensive store depletion, Ca 2ϩ was transiently readded to allow store refilling ( Fig. 2A, phase ii), and cells were stimulated again with histamine to assess the content of Ca 2ϩ stores ( Fig. 2A, phase iii). Remarkably, no Ca 2ϩ changes were observed in STIM1 knockdown cells during Ca 2ϩ readmission, whereas the expected increase was observed in control cells (Fig. 2B, phase ii). As a result, the integrated "Ca 2ϩ influx" response was nearly abrogated in STIM1 knockdown cells (Ϫ99%, Fig. 2C, phase ii). Despite the lack of visible Ca 2ϩ influx, however, the final histamine stimulation elicited a large cytosolic Ca 2ϩ increase in STIM1 knockdown cells (Fig. 2B, phase iii), the integrated Ca 2ϩ response averaging 58% of control cells (Fig. 2C, phase iii). To verify that histamine efficiently depleted Ca 2ϩ stores, we repeated this experiment using ATP and UTP as Ca 2ϩ -mobilizing agonists. The Ca 2ϩ responses elicited by UTP and ATP were smaller than the responses induced by histamine, and Ca 2ϩ influx remained undetectable in STIM1 knockdown cells (data not shown). When added to cells previously stimulated with histamine, UTP elicited a very small Ca 2ϩ response, indicating that the agonist mobilizable store was efficiently depleted by histamine (supplemental Fig. S1). Importantly, Ca 2ϩ influx remained abrogated in STIM1 knockdown cells, even when UTP was added in combination with histamine (supplemental Fig. S1). Thus, although no Ca 2ϩ changes were observed in the cytosol during Ca 2ϩ readmission, Ca 2ϩ stores were able to refill efficiently in STIM1 knockdown cells.
Effect of External Ca 2ϩ Concentration on the Refilling of Ca 2ϩ Stores-The ability of STIM1 knockdown cells to remobilize Ca 2ϩ from internal stores after the depletion/readmission protocol indicated that a "silent" supply of Ca 2ϩ ions sustained the activity of SERCA pumps. To reveal this silent influx pathway, we varied the external Ca 2ϩ concentration applied during the 5-min readmission phase, from 0 to 50 mM. As expected, when no Ca 2ϩ was present during the refilling period, histamine could not remobilize Ca 2ϩ from stores (Fig. 3D). Interestingly, at 0.2 mM [Ca 2ϩ ], the responses were identical in control and STIM1-invalidated cells. In both conditions, Ca 2ϩ influx was undetectable, and histamine could remobilize an identical amount of Ca 2ϩ from internal stores (Fig. 3, C and D). The responses diverged markedly at higher Ca 2ϩ concentrations, however. In control cells, Ca 2ϩ responses increased in a dose-dependent manner with the external [Ca 2ϩ ], both during Ca 2ϩ readmission and during Ca 2ϩ remobilization from stores (Fig. 3,  C and D). In STIM1 knockdown cells, only minimal changes were observed during the readmission phase even at the highest Ca 2ϩ concentration, but the amount of Ca 2ϩ that could be remobilized by histamine increased with the amount of external Ca 2ϩ supplied (Fig. 3, C and D). The response obtained in STIM1 knockdown cells with 50 mM Ca 2ϩ approached the response obtained in control cells with 2 mM Ca 2ϩ (Fig. 3D), indicating that Ca 2ϩ stores could be completely replenished despite the lack of visible influx. Thus, increasing the external [Ca 2ϩ ] enabled complete refilling of Ca 2ϩ stores in STIM1 knockdown cells, but the entering Ca 2ϩ ions remained undetectable in the cytosol.
Effect of STIM1 Knockdown on ER Ca 2ϩ Homeostasis-The major intracellular Ca 2ϩ store is the ER, where the STIM1 protein resides (6). The experiments shown in Figs. 2 and 3 indicate that external Ca 2ϩ ions can reach the ER without altering cytosolic Ca 2ϩ levels in STIM1 knockdown cells. To verify this observation, we directly measured the changes in the free ER Ca 2ϩ concentration, [Ca 2ϩ ] ER , using a cameleon probe targeted to the ER, D1 ER (kindly provided by Dr. Amy Palmer, Boulder, CO). As illustrated in Fig. 4, A and B, the D1 ER recordings showed that [Ca 2ϩ ] ER levels decreased upon histamine stimulation and returned to resting levels upon Ca 2ϩ readmission, both in wild-type and in STIM1 knockdown cells. The [Ca 2ϩ ] ER changes correlated temporally with the changes in cytosolic Ca 2ϩ measured simultaneously with Fura-2 during our depletion/readmission protocol (supplemental Fig. S2). The amplitude of the [Ca 2ϩ ] ER changes during Ca 2ϩ release and uptake was similar in all conditions (Fig. 4C), but detailed analysis of the D1 ER responses revealed that the kinetics of ER Ca 2ϩ refilling were slower in cells transfected with STIM1 siRNA ( ϭ 31 s) than in cells transfected with scramble siRNA ( ϭ 11 s, Fig. 4, D and F). In contrast, the kinetics of histamine-induced Ca 2ϩ release were identical for both conditions (supplemental Fig. S3). Further stimulation with histamine elicited a second decrease in [Ca 2ϩ ] ER , whereas a final stimulation with thapsigargin to extensively deplete Ca 2ϩ stores elicited a slower and monotonic decrase in [Ca 2ϩ ] ER . Importantly, [Ca 2ϩ ] ER decreased with identical kinetics in wild-type and STIM1 knockdown cells exposed to thapsigargin (Fig. 4E). This indicates that the Ca 2ϩ permeability of the ER was not affected, and by inference, that the activity of SERCA that counterbalances the ER Ca 2ϩ leak was also unaffected. Thus, although Ca 2ϩ ions reach the ER more slowly in STIM1 knockdown cells, SERCA activity is normal at low STIM1 levels, and resting [Ca 2ϩ ] ER levels are restored in 2 min upon Ca 2ϩ readmission.
Effect of STIM1 on Mitochondrial Ca 2ϩ Handling-The silent ER refilling revealed by STIM1 knockdown suggested that an organelle could relay Ca 2ϩ ions from membrane channels to SERCA pumps on the ER. Mitochondria rapidly accumulate and release Ca 2ϩ and thereby can provide such a Ca 2ϩ relay mechanism. In addition, Ca 2ϩ buffering by mitochondria has been shown to favor SOC influx by preventing the Ca 2ϩdependent inhibition of SOC channels (19). To investigate whether mitochondria participate in Ca 2ϩ influx and ER refilling in STIM1 knockdown cells, we used a combination of rotenone (an inhibitor of complex I from the respiratory chain) and oligomycin (an inhibitor of the H ϩ ATPase) to dissipate the mitochondrial membrane potential and prevent mitochondrial Ca 2ϩ uptake. In wild-type cells, the histamine-induced Ca 2ϩ influx was severely inhibited by the FIGURE 2. Effect of STIM1 knockdown on the refilling of Ca 2؉ stores. HeLa cells were co-transfected with YC3.6 and either control or STIM1 siRNA, and Ca 2ϩ responses were measured by single cell imaging. Cells were stimulated three times with histamine (Hist) in Ca 2ϩ -free medium to completely deplete Ca 2ϩ stores. Ca 2ϩ was then added transiently, and the cells were stimulated again with histamine to assess store refilling. mitochondrial inhibitors, which were applied shortly before Ca 2ϩ readmission (Fig. 5A). The response during Ca 2ϩ readmission was reduced by 69%, and the refilling efficiency, assessed as in Fig. 2, was reduced by 37% (Fig. 5, B and C). In contrast, in STIM1 knockdown cells, the refilling efficiency was essentially insensitive to mitochondrial inhibitors (Fig,  5C). As previously reported (20), mitochondrial inhibition significantly decreased thapsigargin-induced Ca 2ϩ influx in control cells (Fig. 6, A and B). In this case, however, SOC influx was nearly abolished in STIM1 knockdown cells exposed to mitochondrial inhibitors, the integrated Ca 2ϩ response averaging 11% of the response of control, untreated cells (Fig. 6, A and  B). Thus, the residual channels of STIM1 knockdown cells do not appear to require functional mitochondria when SERCA are active but are sensitive to mitochondrial inhibitors when SERCA are inhibited by thapsigargin. To test whether the effect of oligomycin and rotenone reflected reduced Ca 2ϩ uptake by mitochondria, we used CGP-37157, an inhibitor of mitochondrial Na ϩ /Ca 2ϩ exchange. This compound greatly reduces mitochondrial Ca 2ϩ buffering without altering the ability of mitochondria to produce ATP (21). As shown in Fig. 6, C and D, CGP-37157 prevented thapsigargin-induced Ca 2ϩ influx as efficiently as the combination of oligomycin and rotenone in wild-type cells but had no effect on SOC entry in STIM1 knockdown cells.
To check whether mitochondria were taking up the Ca 2ϩ ions entering across SOC channels, we directly measured Ca 2ϩ signals within the mitochondrial matrix, [Ca 2ϩ ] mit , using a cameleon bearing a mitochondrial targeting sequence (YC2.1 2mit , kindly provided by Dr. T. Pozzan). As shown in Fig. 6, E and F, large changes in [Ca 2ϩ ] mit were observed upon Ca 2ϩ readmission in control cells treated with thapsigargin. In STIM1 knockdown cells, the [Ca 2ϩ ] mit response was much smaller, consistent with the cytosolic responses shown in Fig. 1 These experiments indicate that mitochondria are located close to SOC channels and that SOC influx depends not only on STIM1 levels but also on the activity of SERCA and of neighboring mitochondria.

DISCUSSION
In this study, we decreased the levels of the protein STIM1, a key activator of store-operated Ca 2ϩ influx, to study how Ca 2ϩ ions entering across membrane channels are transferred to the ER. We confirmed that the STIM1 protein is an important regulator of SOC influx; reducing STIM1 mRNA levels by 73% with siRNA caused a 73% decrease in the integrated Ca 2ϩ response to thapsigargin, a robust and very sensitive assay for SOC activity. The excellent correlation between STIM1 mRNA levels and the magnitude of the thapsigargin response confirms previous studies showing that the flux of Ca 2ϩ ions entering across SOC channels is limited by the amount of STIM1 proteins (4,5). Despite the drastic reduction in SOC influx, however, Ca 2ϩ handling was essentially normal in STIM1 knockdown cells, which maintained normal resting cytosolic Ca 2ϩ levels, released a normal amount of Ca 2ϩ from internal stores, and were able to replenish their Ca 2ϩ stores efficiently. The major difference observed was that at low STIM1 levels, store refilling occurred without detectable changes in cytosolic [Ca 2ϩ ]. Store reloading at resting [Ca 2ϩ ] cyt was reported previously when the amplitude of Ca 2ϩ influx was reduced pharmacologically with SOC blockers (22) or with receptor antagonists (23) or when Ca 2ϩ influx was spatially restricted to one side of the cell with a patch pipette (24). Our data confirm these early studies and show that a decrease in cellular STIM1 levels "silences" Ca 2ϩ entry but has minimal impact on store refilling.
The efficiency of Ca 2ϩ stores refilling was quantified by two distinct approaches: 1) by remobilizing Ca 2ϩ from stores after a brief refilling period and 2) by measuring Ca 2ϩ changes within the lumen of the endoplasmic reticulum. The first approach revealed that store refilling was reduced by 42% in STIM1 knockdown cells kept for 5 min in physiological saline. Stores could be fully refilled by increasing the Ca 2ϩ concentration from 2 to 50 mM during the Ca 2ϩ readmission period, but even at this very high Ca 2ϩ concentration, Ca 2ϩ stores refilled without any visible cytosolic signal. When measured from within the ER, store refilling was very rapid, and [Ca 2ϩ ] ER returned to resting levels within 2 min upon Ca 2ϩ readmission. ER refilling proceeded more slowly in STIM1 knockdown cells than in wild-type cells, but in both cases, [Ca 2ϩ ] ER returned within 2 min to prestimulatory levels in physiological saline. The kinetics of histamine-and thapsigargin-induced Ca 2ϩ release were not altered by STIM1 knockdown, indicating that the permeability and InsP3 sensitivity of Ca 2ϩ stores were normal in these cells. These observations indicate that cells can maintain normal levels of Ca 2ϩ within the ER even with drastically reduced SOC activity. When Ca 2ϩ influx is limiting, the few Ca 2ϩ ions entering across the plasma membrane are rapidly carried to the ER without impacting on cytosolic Ca 2ϩ homeostasis.
In cells containing a normal amount of STIM1, however, the large flux of Ca 2ϩ ions through SOC channel clusters exceeds the capacity of subplasmalemmal SERCA. In this case, a significant fraction of the entering Ca 2ϩ is taken up by nearby mitochondria and redirected to SERCA located farther from the plasma membrane (25,26). This trans-mitochondrial Ca 2ϩ flux ensures that only a minimal fraction of the Ca 2ϩ entering via SOC channels diffuses in the cytosol (Fig. 7A). The combined capacity of mitochondria and SERCA is fairly large, and in wild-type cells, cytosolic flooding is not observed when the external Ca 2ϩ concentration is reduced to 0.2 mM. This mechanism ensures that during stimulation with physiological agonists, nearly all the entering Ca 2ϩ ions are used by SERCA to refill the ER, either directly or indirectly via mitochondria. The situation is very different when influx through SOC channels is limiting because in STIM1 knockdown cells, the silent ER refilling was insensitive to mitochondrial inhibitors. This indicates that all the entering Ca 2ϩ ions were directly captured by subplasmalemmal SERCA and not relayed by mitochondria. Mitochondria, however, were still located close to the active channels because a mitochondrial Ca 2ϩ signal was detected when SERCA were inhibited by thapsigargin (Fig. 6). This indicates that, at low STIM1 levels, the Ca 2ϩ uptake capacity of SERCA exceeds the flux of Ca 2ϩ ions across the remaining functional SOC channels (Fig. 7B).
Our data thus highlight a major role of SERCA in buffering subplasmalemmal Ca 2ϩ ions, a role that is intrinsic to their Ca 2ϩ pumping activity but was not fully appreciated before. SOC channels inactivate rapidly at high cytosolic Ca 2ϩ concentrations, and their sustained activity requires efficient buffering systems to prevent the buildup of Ca 2ϩ near the mouth of the channel. Mitochondria were shown to prevent the Ca 2ϩ -dependent inactivation of SOC channels by buffering subplasmalemmal Ca 2ϩ (19). SERCA can perform the same function, as shown in patch clamp studies (24,27) and in cells expressing FIGURE 4. Effect of STIM1 knockdown on ER Ca 2؉ release and uptake. Cells were transfected with the ER-targeted cameleon probe D1 ER to measure ER Ca 2ϩ changes. Cells were kept for 2 min in the absence of external Ca 2ϩ to monitor resting ER Ca 2ϩ levels (␣), stimulated with 50 M histamine (H) to mobilize Ca 2ϩ from stores (␤), and Ca 2ϩ added back transiently to refill ER Ca 2ϩ stores (␥). Cells were then stimulated sequentially with histamine and thapsigargin (Tg) to extensively deplete Ca 2ϩ stores (␦). A and B, representative recordings of ER Ca 2ϩ responses in control (A) and STIM1 knockdown cells (B). C, mean amplitude of the [Ca 2ϩ ] ER changes during Ca 2ϩ release and ER Ca 2ϩ refilling. The changes in D1 ER ratio elicited by histamine (␣-␤) and by Ca 2ϩ readmission (␥-␤) are shown. Bars are mean Ϯ S.E. of 30 -32 cells. D and E, time course of the [Ca 2ϩ ] ER responses during ER refilling (D) and during passive store depletion with thapsigargin (E). The D1 ER responses were fitted with an exponential function to extract the time constants (). F, averaged time constants () of the ER Ca 2ϩ refilling and ER Ca 2ϩ leak processes, derived from the fits shown in panels D and E. Data are mean Ϯ S.E. of 6 -7 experiments comprising 23-32 cells; **, p Ͻ 0.0001 versus control. transient receptor potential channels (28). Our data show that, at low STIM1 levels, SERCA take up all the Ca 2ϩ ions entering across SOC channels, thereby preventing channel inactivation and enabling efficient ER refilling. A similar conclusion was reached by Malli et al. (29) using high K ϩ to reduce the electrochemical driving force for Ca 2ϩ and thus the magnitude of Ca 2ϩ influx. When Ca 2ϩ influx exceeds the capacity of subplasmalemmal SERCA, mitochondria take up the excess Ca 2ϩ and relay it to deeper ER regions. Finally, when both SERCA and mitochondria are saturated, the Ca 2ϩ -dependent inactivation of SOC channels shuts down the supply of Ca 2ϩ ions. Thus, as depicted in the diagram of Fig. 7, three mechanisms contribute to minimize the diffusion of Ca 2ϩ ions in the cytosol in the vicinity of SOC channels: 1) the presence of active SERCA on the juxtaposed ER membrane, which under physiological conditions take up most of the entering Ca 2ϩ ions; 2) the presence of neighboring mitochondria, which scavenge the remaining Ca 2ϩ ions; and 3) the Ca 2ϩ -dependent inactivation of SOC channels, which terminates Ca 2ϩ entry when both Ca 2ϩ scavenging systems are saturated. At low STIM1 levels, SOC influx did not require mitochondrial Ca 2ϩ buffering but nevertheless required functional mitochondria (Fig. 6), indicating that local ATP production is also required for SOC channel activity. Because SERCA were inhibited by thapsigargin in this experiment, the mitochondrial ATP was not required as energy supply but most likely as a mobile Ca 2ϩ buffer, as recently shown in lymphocytes (30). The mitochondria located near the ER-PM junction are thus particularly important because they supply ATP used to buffer the entering Ca 2ϩ ions and to energize SERCA and scavenge the excess of Ca 2ϩ ions when SOC influx exceeds the capacity of SERCA. Our STIM1 knockdown study thus highlights the important role of the mitochondria located near the narrow and extended ER-PM junctions.
Our [Ca 2ϩ ] ER measurements further show that removing STIM1 proteins is well tolerated by the ER Ca 2ϩ handling machinery. This is somewhat unexpected given the dominant role of STIM1 in the regulation of SOC entry. STIM1 contains a lumenal EF-hand domain that acts as a [Ca 2ϩ ] ER sensor and a cytosolic domain that is both necessary and sufficient for the activation of SOC channels (31). STIM1 interacts with Orai1 and TRP Ca 2ϩ channels at the plasma membrane via its cytosolic ERM domain, but it is not known whether the only partners of STIM1 are plasma membrane Ca 2ϩ channels or whether STIM1 also interacts with Ca 2ϩ transporters on the ER membrane and modulate their activity. Our data indicate that FIGURE 5. Effect of mitochondrial inhibitors on ER Ca 2؉ refilling. A, the protocol of Fig. 2 was used to deplete and refill Ca 2ϩ stores, and mitochondrial inhibitors (oligo/rot, 5 g/ml oligomycin and 25 M rotenone) were added 2 min before Ca 2ϩ readmission. B and C, integrated Ca 2ϩ responses measured during Ca 2ϩ readmission (B, phase ii) and during Ca 2ϩ remobilization with histamine (C, phase iii). Bars are mean Ϯ S.E. of 58 -156 control cells and 59 -70 STIM1 knockdown cells; *, p Ͻ 0.001, and **, p Ͻ 0.0001 versus control. AUC, area under the curve. STIM1 has minimal impact on the basal activity of ER Ca 2ϩ transporters because decreasing STIM1 levels did not significantly alter resting [Ca 2ϩ ] ER levels or the passive Ca 2ϩ permeability of the ER. Since the [Ca 2ϩ ] ER homeostasis is maintained by a "pump and leak" mechanism, this implies that the basal activity of SERCA Ca 2ϩ pumps and of the Ca 2ϩ leak pathway (which reflects the basal Ca 2ϩ permeability of InsP3 receptors and other pathways, see Ref. 32) is not affected by the decrease in STIM1 levels. Thus, our data indicate that removing STIM1 proteins has minimal impact on ER Ca 2ϩ homeostasis, confirming a recent study in Caenorhabditis elegans (33). In this model system, STIM1 knockdown inhibited store-operated Ca 2ϩ entry but had no effects on Ca 2ϩ oscillations and waves and did not induce ER depletion. Our data confirm these findings and show that STIM1 regulates the activity of SOC channels at the plasma membrane without modifying the activity of Ca 2ϩ transporters located on the ER. The structure formed by the juxtaposition of the ER and the plasma membrane ensures the optimal delivery of Ca 2ϩ ions to the reticulum and minimizes the Ca 2ϩ contamination of the cytosol, creating a privileged pathway for ER refilling.