S-glutathionylation activates STIM1 and alters mitochondrial homeostasis

Calcium is a ubiquitous second messenger that is tightly controlled inside the ER, where it can be rapidly mobilized to translate receptor-mediated signaling into a cellular response. After the initial receptor-initiated Ca²⁺ transient, a sustained Ca²⁺ influx from the extracellular milieu occurs that serves to provide prolonged Ca²⁺ signals and allow ER store refilling to permit subsequent signaling events (Berridge et al., 2003; Parekh and Putney, 2005; Feske, 2007; Deng et al., 2009). Activation of this capacitive Ca²⁺ entry by the calcium release–activated calcium (CRAC) channel involves a complex molecular choreography. In general, a decrease in ER luminal Ca²⁺ triggers Ca²⁺ dissociation from the ER-resident protein STIM1 (stromal interaction molecule 1), which oligomerizes and translocates into discrete junctions near the plasma membrane, where it binds to and activates members of the Orai family of CRAC channels (Luik et al., 2006; Prakriya et al., 2006; Vig et al., 2006; Yeromin et al., 2006; Smyth et al., 2007; Park et al., 2009; Yuan et al., 2009). In addition to buffering Ca²⁺ released from ER stores, mitochondria are also intimately involved in CRAC activation. Through their associations with the ER (Rizzuto et al., 1998; Csordás et al., 2006), mitochondrial Ca²⁺ buffering can enhance ER Ca²⁺ depletion and, thus, increase CRAC activation (Gilabert et al., 2001). Furthermore, mitochondrial Ca²⁺ release can enhance ER Ca²⁺ refilling and, thus, reduce the duration of CRAC channel activation (Parekh, 2008). Mitochondrial Ca²⁺ uptake also stimulates the production of reactive oxygen species (ROS), which can influence both ER Ca²⁺ release (Bootman et al., 1992; Huddleston et al., 2008) and reuptake by the ER (Kaplan et al., 2003; Ihara et al., 2005). A prominent posttranslational molecular target of ROS is protein cysteine residues, which reversibly form disulfide bridges or sulphenic acid (PSOH) or irreversibly produce sulphinic (PSO₂H) or sulphonic acid (PSO₃H) moieties (Veal et al., 2007). Additionally, a common modification of protein thiol is S-glutathionylation via the reversible reaction between protein cysteine residues and glutathione (GSH; Anathy et al., 2009; Dalle-Donne et al., 2009). However, the specific molecular targets of oxidants that affect Ca²⁺ signaling and mitochondrial function are not fully defined. In this study, we demonstrate that, in addition to being a sensor for intracellular Ca²⁺ stores, STIM1 functions as a redox sensor to constitutively activate CRAC channels under oxidative conditions. S-glutathionylation or C⁵⁶A mutation of STIM1 positively regulates CRAC channel activation, which leads to mitochondrial Ca²⁺ overload and alterations in cellular bioenergetics. Importantly, we find that GSH is a critical regulator of STIM1 signaling during oxidative stress.

To investigate the role of oxidant stress in Ca²⁺ homeostasis, DT40 B-lymphocytes were challenged with lipopolysaccharide (LPS), a component of the gram-negative bacterial cell wall that stimulates ROS production through the Toll-like receptor 4 (Asehnoune et al., 2004; Madesh et al., 2005). 1 µg/ml LPS evoked a time-dependent increase in cellular oxidative stress that peaked at 5 h and could be ablated by 100 µM of the antioxidant butylated hydroxyanisole (BHA; Fig. 1, A and B). In nonchallenged B cells, αIgM activation (1.5 µg/ml) of the B cell receptor stimulates robust Ca²⁺ mobilization that presented as a nonsynchronized oscillatory pattern (individual traces) and rapidly returned to baseline (Fig. 1 C, dashed line). In contrast, 5 h of LPS treatment resulted in a phenotypic shift toward a more elevated Ca²⁺ oscillation pattern after αIgM addition that remained elevated above baseline (Fig. 1 D). Restoration of an oscillatory Ca²⁺ signaling pattern could be accomplished by scavenging ROS with BHA (Fig. 1 E), indicating that the effect of LPS on Ca²⁺ signaling may be mediated by oxidants. Cells pretreated with 100 µM hydrogen peroxide (H₂O₂) for 20 min also resulted in a sustained Ca²⁺ mobilization pattern, although to a much higher degree than the LPS and with no oscillations observed (Fig. 1 F). In normal cells, oxidants are effectively scavenged through multiple endogenous antioxidants, including superoxide dismutase, catalase, and GSH peroxidase. In addition, the tripeptide GSH, composed of the amino acids γ-glutamic acid, cysteine, and glycine, is present in millimolar quantities in cells and serves as an important antioxidant and redox molecule. As expected, LPS led to a time-dependent depletion of the cellular GSH pool, as determined by the GSH cross-linking fluorophore monochlorobimane (MCB; Fig. 1 G) and by direct measurement of cellular GSH (unpublished data). Similar to BHA, GSH replenishment with cell-permeable GSH-ester restored Ca²⁺ oscillations in LPS-challenged DT40 cells in response to αIgM (Fig. 1 H). Furthermore, GSH efflux was unaltered in cells challenged with LPS, indicating a reduction in intracellular GSH levels rather than loss of GSH to the extracellular milieu (unpublished data).

Under normal physiological conditions, Ca²⁺ is efficiently taken up and released by mitochondria in concert with cytosolic Ca²⁺ levels (Fig. 2, A and B). However, the sustained elevation of cytosolic Ca²⁺ concentration mediated by oxidative stress (1 µg/ml LPS for 5 h) dramatically enhances mitochondrial Ca²⁺ uptake (Fig. 2, C and D). Interestingly, elevated mitochondrial Ca²⁺ uptake did not occur immediately after αIgM stimulation but, rather, increased incrementally over time. This observation indicates that during oxidative stress, mitochondrial Ca²⁺ uptake is tied not to the magnitude of the cytosolic Ca²⁺ levels but to the diastolic Ca²⁺ level between oscillatory peaks. In addition to ROS production, LPS targets multiple signaling cascades, including the activation of NF-κB and the production of various cytokines, which may also impact the Ca²⁺ mobilization pattern and must be considered. As supplementation of GSH restored Ca²⁺ oscillations in LPS-challenged cells (Fig. 1 H), we chose to focus on GSH to dissect the specific role of oxidants in modulating physiological Ca²⁺ signaling.

To induce oxidative stress secondary to a depletion of GSH, DT40 cells were treated with 200 µM of the chemosensitizing agent buthionine sulfoximine (BSO; for 24 h), which inhibits γ-glutamyl synthetase, the rate-limiting enzyme in GSH synthesis (Arrick et al., 1981; Madesh et al., 1998; Diehn et al., 2009). In contrast to untreated conditions, GSH depletion by BSO resulted in a dramatic alteration in the Ca²⁺ signaling phenotype from an oscillatory to a more sustained Ca²⁺ mobilization pattern after αIgM addition (Fig. 2 E). Furthermore, mitochondrial Ca²⁺ handling in BSO-treated cells was similar to that of DT40 cells treated with LPS (Fig. 2, F and G) and was proportional to the degree of GSH depletion (Fig. S1). To delineate whether the Ca²⁺ handling evoked by BSO was attributable to GSH depletion (Fig. 2, H–K), we next assessed the coordination between cytosolic and mitochondrial Ca²⁺. In contrast to untreated conditions (Fig. 2 L), BSO challenge triggered sustained mitochondrial Ca²⁺ uptake (Fig. 2 M) that could be reversed by GSH supplementation (Fig. 2 N). Similarly, supplementation of the antioxidants BHA or GSH-ester restored the oscillatory mitochondrial Ca²⁺-handling pattern in LPS-treated cells (unpublished data). These results suggest two important points: (1) the effects of LPS on cytosolic and mitochondrial Ca²⁺ handling are mediated through GSH oxidation and (2) BSO constitutes a physiologically relevant model in which to study the effects of oxidant stress on cellular Ca²⁺ signaling.

A plausible explanation for the elevation in diastolic Ca²⁺ between oscillatory peaks may be an influx of Ca²⁺ from the extracellular milieu via activation of the CRAC channel. The absence of extracellular Ca²⁺ did not affect the αIgM response in wild-type (WT) DT40 cells but stimulated robust Ca²⁺ entry upon the addition of 2 mM Ca²⁺ (Fig. 3 A). Interestingly, the BSO-mediated Ca²⁺ phenotype (Fig. 3 B) could be normalized by removing extracellular Ca²⁺ (Fig. 3 C). Like untreated cells, reintroduction of extracellular Ca²⁺ in BSO-treated DT40 cells resulted in CRAC activation, which is mediated by the ER-resident protein STIM1 (Liou et al., 2005; Roos et al., 2005). DT40 cells lacking STIM1 responded in a similar manner to WT cells in response to αIgM. However, unlike WT cells, STIM1 knockout (KO) cells fail to demonstrate capacitive Ca²⁺ entry upon store depletion (Fig. 3 D; Baba et al., 2006). Although DT40 cells express both STIM1 and STIM2, the role of STIM2 in IgM-mediated Ca²⁺ entry is less clear. Surprisingly, BSO challenge did not alter the Ca²⁺ mobilization pattern in STIM1-deficient DT40 cells either in the presence (Fig. 3 E) or absence (Fig. 3 F) of extracellular Ca²⁺, despite a near-complete depletion of cellular GSH levels (Fig. 3 H). We next chose to determine whether the sustained Ca²⁺ elevation pattern during oxidative stress requires Orai molecules, which are the plasma membrane pore-forming units of the CRAC channel. Elimination of Orai1 and -2 evoked an oscillatory Ca²⁺ mobilization pattern during oxidative stress (Fig. 3 G) that appears to be Orai1 dependent (Fig. S2), further verifying CRAC activation during BSO-induced oxidative stress. Oxidative stress did not alter expression of the plasma membrane Ca²⁺ ATPases (PMCA1/4), nor did it affect cellular ATP levels (unpublished data), further verifying activation of the CRAC channel and not ablated extrusion of Ca²⁺ from the cytosol.

Depletion of GSH is associated with an elevation in ROS production (Armstrong and Jones, 2002). However, only a slight increase in superoxide generation (the initial radical species) was observed in WT DT40 cells after addition of BSO, as detected by the superoxide indicator dye dihydroethidine (HE; Fig. 3 I). Superoxide anions will undergo reduction either spontaneously or enzymatically in the presence of superoxide dismutase to form the stable oxidant H₂O₂. Unlike superoxide, BSO challenge elicited a dramatic increase in H₂O₂ in WT DT40 cells, as detected by the general ROS indicator dichlorofluorescein (DCF), indicative of a cellular deficiency in H₂O₂ scavenging (Fig. 3 J). However, only a smaller increase in H₂O₂ was observed in STIM1 KO DT40 cells (Fig. 3 J), suggesting that loss of STIM1 attenuates H₂O₂ accumulation. Importantly, BSO-mediated ROS accumulation was not altered in the presence of 30 µM of the plasma membrane NADPH oxidase inhibitor diphenyleneiodonium (Fig. 3 K), excluding the possibility that BSO enhanced oxidant generation via NADPH oxidase. This important finding reveals two key features of Ca²⁺ signaling during oxidative stress: (1) either STIM1 or Orai1 are a potential target of oxidative stress, and (2) activation of the CRAC channel can be regulated by cellular redox status.

B cell receptor cross-linking results in coordinated cytosolic Ca²⁺ mobilization and mitochondrial Ca²⁺ uptake in both WT (Fig. 4 A, top) and STIM1 KO cells (Fig. 4 A, bottom). The coordination between cytosolic Ca²⁺ mobilization and mitochondrial Ca²⁺ uptake can be attributed to the close physical proximity between the mitochondria and the ER (Filippin et al., 2003). A previous study showed that mitochondria may in fact be tethered to the ER and that this tethering is important in both physiological and pathological signaling (Pinton et al., 2008). Our current findings suggest that it is Ca²⁺ from the extracellular milieu and the resultant elevation in diastolic Ca²⁺ between oscillatory peaks (i.e., temporal Ca²⁺ increase) that are key determinants of sustained mitochondrial Ca²⁺ uptake during oxidative stress (Fig. S3, A and B). Although BSO challenge dramatically altered mitochondrial Ca²⁺ uptake in WT cells (Fig. 4, B and C, top), STIM1 KO cells did not exhibit sustained mitochondrial Ca²⁺ uptake in response to αIgM (Fig. 4, B and C, bottom) even after reintroduction of extracellular Ca²⁺ (Fig. S3 C).

Sustained cytosolic Ca²⁺ or aberrant mitochondrial Ca²⁺ uptake will lead to irreversible mitochondrial dysfunction and bioenergetic collapse (Crompton, 1999; Duchen, 2000). Consistent with our findings that BSO challenge leads to sustained mitochondrial Ca²⁺ uptake (Fig. 4 C), BSO-pretreated WT DT40 cells consumed less oxygen after the sequential additions of the mitochondrial complex I substrates malate and pyruvate and the complex II/III substrate succinate compared with nontreated cells (Fig. 4 D). In contrast, BSO treatment did not affect oxygen consumption in STIM1 KO DT40 cells. Ultimately, this reduction in mitochondrial bioenergetics will decrease mitochondrial energy production and result in cell death. As reported previously (Takata et al., 1995; White et al., 2005), chronic αIgM challenge induced Ca²⁺-mediated cell death in WT DT40 cells that was markedly enhanced by BSO challenge at all time points studied (Fig. 4 E). In stark contrast, STIM1 KO cells were protected from Ca²⁺-mediated cell death throughout the duration of the experiment (Fig. 4 E). Supplementation of GSH reversed αIgM-induced cell death (1.5 µg/ml; 48 h) in DT40 cells challenged with either BSO or LPS (Fig. 4 F), indicating that the mechanism for both is mediated through GSH. Both STIM1 and Orai1 KO DT40 cells treated with either BSO or LPS were resistant to αIgM-induced cell death, as determined by nuclear incorporation of TOTO-3 (Fig. 4 F). Mechanistically, TOTO-3 staining was preceded by the loss of the mitochondrial membrane potential (ΔΨ_m) both in LPS (Fig. 4 G)- and BSO-treated (Fig. 4 J) cells, indicating a loss of mitochondrial function before cell death (Fig. 4, G and J, black dots). GSH supplementation restored ΔΨ_m in WT DT40 cells (Fig. 4, G and J, red dots). Neither LPS nor BSO alone altered ΔΨ_m in WT cells (unpublished data). Interestingly, ΔΨ_m was unaltered in STIM1 and Orai1 KO cells (Fig. 4, H, I, K, and L). These findings clearly establish that sustained mitochondrial Ca²⁺ uptake, mitochondrial dysfunction, and cell death during oxidative stress are dependent on GSH bioavailability and CRAC-mediated Ca²⁺ entry.

Upon store depletion, STIM1 oligomerizes, redistributes, and binds to the plasma membrane–localized protein Orai1/CRACM1 to form the CRAC channel and trigger Ca²⁺ entry and refilling of the ER stores (Prakriya et al., 2006). As our work demonstrates that oxidative stress enhances Ca²⁺ influx, we hypothesized that the STIM1 cellular distribution may be altered during oxidative stress. Because STIM1 is a dynamic protein, we chose to overexpress fluorescently tagged (mCherry) STIM1 in COS-7 cells to visualize its localization in live cells. In unstimulated COS-7 cells, STIM1 presents primarily as an ER-resident protein (Fig. 5 A). However, both 100 µM BSO challenge (Fig. 5 B) and H₂O₂ (20 min; Fig. 5 C) triggered redistribution from the ER to sites near the plasma membrane, demonstrating that oxidant stress promotes STIM1 oligomerization and may, perhaps, trigger CRAC activation. Addition of Ca²⁺ to the extracellular milieu evoked robust Ca²⁺ entry in BSO- and H₂O₂-treated, but not untreated, mouse embryonic fibroblasts (MEFs; Fig. 5 D). In contrast, Ca²⁺ entry was not observed in STIM1 KO MEFs, excluding the possibility that BSO or H₂O₂ may facilitate Ca²⁺ entry through other mechanisms other than the CRAC channel (Fig. 5 E). Although it is possible that oxidant stress may deplete ER Ca²⁺, addition of the SERCA (sarco-ER Ca²⁺ ATPase) of Ca inhibitor thapsigargin (Tg) resulted in the passive depletion of ER Ca²⁺ and subsequent capacitive Ca²⁺ entry upon reintroduction of extracellular Ca²⁺ (unpublished data), indicating intact ER Ca²⁺ stores and normal CRAC channel function. Furthermore, DT40 cells under 100 µM H₂O₂ challenge (30 min) exhibited similar levels of ER Ca²⁺ as the control (unpublished data). Although STIM1 overexpression by itself did not induce constitutive Ca²⁺ entry (Fig. 5 F), short-term exposure to 200 µM H₂O₂ triggered capacitive Ca²⁺ entry without the need for ER store depletion by Tg (Fig. 5 G) in DT40 STIM1 KO cells overexpressing STIM1. In these experiments, DT40 cells were incubated with H₂O₂ for 20 min in Ca²⁺-containing conditions and transferred to the experimental buffer lacking Ca²⁺. Interestingly, because of this experimental methodology, we observed higher basal cytosolic Ca²⁺ levels in STIM1-positive cells versus STIM1-negative neighboring cells (Fig. 5 G), indicating the rapid activation of oxidant-triggered, STIM1-mediated Ca²⁺ entry. Nontransfected neighboring cells did not exhibit store-independent Ca²⁺ entry, further supporting that the H₂O₂ effect did not reflect oxidant-mediated ER Ca²⁺ depletion but, rather, required the expression of STIM1. Furthermore, BSO challenge in Orai1 KO cells did not result in CRAC activation (Fig. S4); however, the formation of STIM1 clusters remained unaffected (Fig. 5 H), revealing that STIM1, not Orai1, may be a target for cellular oxidants. Further investigation revealed that STIM1 puncta formation was dependent on the ER-luminal portion of the protein (Fig. S5).

A prominent molecular target of H₂O₂ is the reactive thiol group of the amino acid cysteine. Protein thiol groups react with H₂O₂ to yield sulfenic acid (R-SOH), which undergoes nucleophilic attack by RSH, resulting in disulfide formation (RSSR). Under oxidative conditions, GSSG can react with reactive thiol groups on proteins in this fashion in a process known as S-glutathionylation, which results in protein-specific functional changes. Under BSO-induced stress, S-glutathionylation is effectively an irreversible process because of the lack of free GSH needed to remove GSH via glutaredoxins. To test whether STIM1 is a target of S-glutathionylation, 0.3 µg recombinant human STIM1 was subjected to H₂O₂ in the presence of 10 mM GSH in a cell-free system. H₂O₂ challenge resulted in a strong S-glutathionylation signature, as detected by Western blotting (Fig. 5 I). Furthermore, S-glutathionylation was associated with a marked shift in molecular mass, implying that S-glutathionylation of STIM1 results in oligomer formation, a key step in CRAC activation. Addition of the reducing agent DTT normalized protein mobility in response to H₂O₂, indicating the reversibility of STIM1 S-glutathionylation. Mechanistically, H₂O₂-mediated S-glutathionylation decreases the Ca²⁺-binding affinity of STIM1 (Fig. 5 J), thereby decoupling ER Ca²⁺ levels from CRAC activation.

STIM1 is an evolutionarily conserved protein that is composed of a single transmembrane domain with a Ca²⁺-binding domain (EF hand) and a sterile α-motif at the N terminus and a coiled-coil region and proline-rich C terminus (Baba et al., 2006). Analysis of the formation of active STIM1–Orai1 complexes revealed that CRAC formation is dependent on the STIM1 C terminus, which is activated in response to ER Ca²⁺ depletion (Park et al., 2009; Yuan et al., 2009). Although considerable heterogeneity exists between species, a defining feature of STIM1 is the presence of two highly conserved cysteine residues in the N-terminal region, which is located within the ER lumen near the EF hand (Fig. 6 A). Deletion of these residues along with the EF hand resulted in constitutive store-independent Ca²⁺ entry (Zhang et al., 2005). As cysteine residues are important targets for oxidants via either direct modification (e.g., intramolecular disulfide bonds or R-SOH formation) or S-glutathionylation, we therefore chose to investigate the role of these two conserved cysteine residues in STIM1-mediated Ca²⁺ entry (Fig. 6 B). Exposure of the STIM1 N-terminal region (amino acids 23–312), containing the EF hand and cysteine residues 49 and 56, to 100 µM H₂O₂ for 30 min did not alter protein mobility under nonreducing conditions (i.e., without DTT), suggesting that oxidant stress did not directly modify STIM1 (Fig. 6 C). Because STIM1 can be directly S-glutathionylated in the presence of oxidants (Fig. 5 I), we next sought to identify which reactive cysteine serves as molecular target of oxidants in intact cells. Although no S-glutathionylation signature was detected under normal conditions, significant S-glutathionylation in both STIM1 and the SM1 mutant, but not the SM2 mutant, were observed after H₂O₂ challenge (Fig. 6 D), revealing the cysteine at position 56 as exquisitely sensitive to oxidant stress. As in our cell-free conditions, S-glutathionylation also resulted in the formation of a large molecular mass protein under nonreducing conditions (Fig. 6 D, **). Importantly, the mechanism is reversible, as resolving these proteins under reducing conditions revealed no S-glutathionylation in any of the STIM1 mutants (unpublished data). To conclusively demonstrate the identity of the glutathionylated residues on STIM1, the recombinant N-terminal region (amino acids 23–312) of STIM1 was exposed to 100 µM H₂O₂ in the presence of 10 mM GSH. Tryptic digestion resulted in the separation of the cysteine residues at positions 49 and 56 into two distinct peptide fragments, with position 56 within the fragment containing amino acids 181–199 and a mass of 2,216.102 kD (Fig. 6 E, blue). In the H₂O₂-treated, but not the control sample, a mass spectra was found at 2,522.0 ± 0.4 kD (Fig. 6 F). This corresponds to the predicted fragment mass of 2,216.102 + 306 kD, which is the deprotonated molecular mass of GSH. Because GSH can only react with free thiol groups, the presence of GSH in this fragment can only be associated with the cysteine residue at position 56 (Fig. 6 E, green box). Together with protein mobility and immunoblotting, the detection of GSH clearly identifies that the two cysteine residues in the N-terminal region of STIM1 are not directly modified by oxidants, but rather, cysteine 56 is a site for S-glutathionylation in the presence of oxidants.

A fundamental question is whether cysteine 56 is crucial for CRAC channel activation. To test this, Ca²⁺ entry was measured in STIM1 KO DT40 cells after the reexpression of mCherry-tagged WT, SM2, and double-mutant (DM) STIM1 variants. Surprisingly, similar to BSO and H₂O₂ treatment, expression of STIM1 mutants (SM2 and DM) in STIM1 KO DT40 cells resulted in constitutive Ca²⁺ entry, as detected by confocal microscopy (Fig. 7, A and B). Reconstitution of WT STIM1 in STIM1 KO DT40 cells also rescued BSO-induced constitutive Ca²⁺ entry (Fig. 7, A and B), whereas nontransfected neighboring cells did not exhibit an increase in Ca²⁺ fluorescence (unpublished data). In addition, introduction of the SM2 mutant increased ROS production in DT40 STIM1 KO cells to similar levels as BSO challenge (unpublished data). Moreover, in HEK293T cells, SM2 or DM mutation resulted in a substantial redistribution of STIM1 toward the plasma membrane, where it colocalized with Orai1 (Fig. 7 C). Functionally, overexpression of SM2 resulted in the formation of puncta (Fig. 7 C) and a constitutive CRAC current that was independent of the ER Ca²⁺ stores (Fig. 7, D–F).

Molecular identification of STIM1 as the ER Ca²⁺ sensor in CRAC activation has greatly aided our understanding of Ca²⁺ signaling. However, a key question regarding how STIM1 is controlled remains unanswered; namely, what physiological cues can modulate STIM1-mediated Ca²⁺ entry. Oxidative stress has long been implicated in Ca²⁺ dysregulation. In particular, lymphocytes respond to oxidants by an increase in basal Ca²⁺ levels (Howe et al., 2004). Although it is possible that oxidants may affect ER Ca²⁺ release or inhibit reuptake by SERCA pumps, the present findings suggest that oxidants directly decrease STIM1 Ca²⁺-binding affinity via S-glutathionylation at cysteine 56, facilitating constitutive Ca²⁺ entry and elevating basal cytosolic Ca²⁺ levels. This is supported by our findings that STIM1 KO DT40 cells are immune to LPS- and BSO-induced oxidative stress and that reintroduction of STIM1 into STIM1 KO DT40 lymphocyte cells markedly elevated basal cytosolic Ca²⁺ levels when exposed to oxidant stress compared with nontransfected neighboring cells (Fig. 5 G). Furthermore, genetic removal of either STIM1 (Fig. 5 E) or Orai1 (Fig. S4) eliminated constitutive CRAC activation in response to oxidant challenge. Therefore, these data are the first to definitively link Ca²⁺ regulation and cellular redox status and establish STIM1 as an important sensor for oxidative stress.

LPS is a common inflammatory signal inducer that initiates ROS production through Toll-like receptor 4 (Park et al., 2004), leading to extensive proliferation and differentiation in leukocytes (Coutinho et al., 1974). In contrast, B cell receptor antigen cross-linking can trigger several outcomes, including activation, proliferation, or death, depending on the stage of B cell development (Khan, 2009). However, costimulation by both LPS and antigens acts synergistically to sensitize B cells (Minguet et al., 2008), possibly allowing the immune system to focus B cell maturation only on antigen-presenting B cells (Ruprecht and Lanzavecchia, 2006). After activation, unregulated lymphocyte proliferation would exacerbate the immune response and could lead to the development of autoimmune diseases such as rheumatoid arthritis (Busconi et al., 2007). Based on our findings, it is possible that chronic mitochondrial ROS generation serves to limit lymphocyte activation through a gradual reduction in mitochondrial energy production, thus acting as a stop signal that limits the immune response. Costimulation of inflammatory molecules that generate ROS and mobilize Ca²⁺ exist in several different tissues aside from lymphocytes, including LPS and CD14 in dendritic cells (Zanoni et al., 2009) and TNF and leukotriene B4 in endothelial cells (Qiu et al., 2006). Therefore, we postulate that enhancement of CRAC by S-glutathionylation activation is the mechanism whereby ROS and Ca²⁺ synergistically modulate inflammation. Because elevated basal Ca²⁺ was also noted in other cell types, including HEK293, COS-7 (unpublished data), and the ubiquitous expression of STIM1, oxidant-regulated, STIM1-mediated Ca²⁺ entry may be an important physiological process in many tissues. S-glutathionylation of STIM1 correlated with an approximately threefold decrease in GSH from normal cellular levels (Fig. 3 H).

The defining step in CRAC activation is STIM1 oligomerization that follows Ca²⁺ dissociation from the EF hand region (Smyth et al., 2008). Structurally, it is believed that Ca²⁺ dissociation disrupts interactions between the EF hand and sterile α-motif domains and results in STIM1 destabilization and partial unfolding, which facilitates oligomerization of STIM1 and CRAC activation (Stathopulos et al., 2008). Because of their proximity to the EF hand, the highly conserved cysteine residues at positions 49 and 56 were considered potential target sites of oxidants via intra- or intermolecular disulfide bonds or through S-glutathionylation. Initially, we hypothesized that replacing the cysteine residues with the nonreactive amino acid alanine would ablate oxidant-induced constitutive Ca²⁺ entry. Surprisingly, cells transfected with mutant STIM1 constructs displayed dramatic redistribution of STIM1 into discrete puncta near the plasma membrane and higher basal Ca²⁺ levels and CRAC activation, which is indicative of STIM1 oligomerization. Although unexpected, this finding is consistent with a previous study in which point mutation of single amino acids within the EF hand led to CRAC activation (Zhang et al., 2005). Likely, by virtue of its close proximity to the EF hand region of STIM1 (10 amino acids), either mutation or S-glutathionylation of cysteine 56 alters protein confirmation enough to decrease Ca²⁺-binding affinity and activate the protein. Of these mutants, we identified that WT STIM1 and the SM1 mutation (C⁴⁹A) mutants formed large molecular mass structures in the presence of H₂O₂ under nonreducing conditions (Fig. 6 D, **). Furthermore, this STIM1 S-glutathionylation appears to be a nonenzymatic process, as GSH was detected both by immunoblotting (Fig. 5 I) and mass spectrometry (Fig. 6 F) in a cell-free system. Reducing the samples with DTT removed GSH from the cysteine at position 56 (unpublished data) and disrupted the formation of these large molecular mass oligomers (Fig. 5 I), effectively demonstrating the reversibility of this mechanism. Functionally, STIM1 S-glutathionylation serves to reduce the Ca²⁺-binding affinity of the EF hand region (Fig. 5 J), effectively activating STIM1 independent of ER calcium levels. A recent study demonstrated that STIM1 oligomerization is the switch that links ER store depletion and CRAC activation (Luik et al., 2008). However, the study used an artificial system to oligomerize STIM1 independent of ER Ca²⁺ levels. Our experiments are the first to demonstrate reversible STIM1 oligomerization independent of ER Ca²⁺ stores using a physiological model of chronic autocrine-derived oxidant stress. In contrast, acute challenge with H₂O₂ (paracrine) triggered inactivation of Orai1-dependent CRAC activity (Bogeski et al., 2010). Although contradictory to this finding, we postulate that our model of chronic autocrine oxidant production serves a different cellular response than that of an acute oxidant challenge.

In contrast to transient Ca²⁺ elevations, which restrict the actions of Ca²⁺ to specific, high affinity targets, chronic elevations in cytosolic Ca²⁺ extend the duration for Ca²⁺ binding (Boulware and Marchant, 2008). Thus, sustained Ca²⁺ elevations can influence additional cellular processes such as transcription (Dolmetsch et al., 1997) and enzyme activation (Leslie, 1997). Our data suggests that sustained Ca²⁺ elevations also govern mitochondrial Ca²⁺ handling. Elevated cytosolic Ca²⁺ triggers mitochondrial Ca²⁺ uptake by the Ca²⁺ uniporter (Kirichok et al., 2004), which is then inactivated by sustained Ca²⁺ levels (Parekh, 2008). Both LPS- and BSO-induced oxidative stress elicited mitochondrial Ca²⁺ uptake in a step-wise manner (Fig. 1 A; Fig. 2, F and M; Fig. 4 C; Fig. S1; and Fig. S3). The magnitude of each of these step-wise increases was similar to physiological Ca²⁺ transients seen in nonstressed mitochondria (Fig. 2, A and L; Fig. 4 A; and Fig. S1), which is indicative of intact uniporter function. These results suggest that chronic STIM1-mediated elevation in cytosolic Ca²⁺ is likely sufficient to sensitize the mitochondria either by inhibiting the inactivation phase of the uniporter or altering mitochondrial Ca²⁺ extrusion. In addition to STIM1, S-glutathionylation of mitochondrial proteins can also occur and has been linked to both cell death and survival. Reversible glutathionylation of complex I increases mitochondrial superoxide formation that leads to oxidative damage and cell death (Taylor et al., 2003). Furthermore, deglutathionylation of complex II decreases mitochondrial electron transfer activity during myocardial ischemia and reperfusion injury (Chen et al., 2007). In contrast, S-glutathionylation of adenine nucleotide translocase via carbon monoxide prevents loss of mitochondrial membrane permeabilization and apoptosis (Queiroga et al., 2010). In this study, we demonstrate that the S-glutathionylation of the ER-resident protein STIM1 affects the mitochondrial homeostasis via CRAC activation and an elevation in cytosolic Ca²⁺.

In conclusion, these findings demonstrate a relationship between oxidative stress and STIM1-mediated Ca²⁺ entry. By reversibly targeting the highly conserved cysteine 56 residue near the EF hand, oxidant-induced S-glutathionylation decreases Ca²⁺ binding, triggering STIM1 oligomerization and CRAC activation independent of ER Ca²⁺ stores. Constitutive activation of the CRAC channel facilitates an increase in Ca²⁺ levels both at rest and after cellular activation (i.e., increases in diastolic Ca²⁺). This sustained Ca²⁺ increase enhances mitochondrial Ca²⁺ loading and influences mitochondrial function, which over time can trigger cell death. Because of the widespread tissue expression of STIM1, this mechanism is likely to be involved in many different cell types and may play an important role in both physiological and pathological signaling by ROS.

WT (DT40 WT), STIM1 KO (DT40 STIM1 KO), and Orai KO B cell lines were cultured in RPMI 1640 supplemented with 10% FCS, 1% chicken serum, and antibiotics. HEK293T, COS7, and MEF (WT and STIM1 KO) cell lines were cultured in DME supplemented with 10% FCS and antibiotics. To induce oxidative stress, LPS (Sigma-Aldrich) or BSO (Sigma-Aldrich) was added at the indicated concentration 5 and 24 h before experimentation, respectively. BHA (Sigma-Aldrich) was added to cells 1 h before LPS challenge.

DT40 cells were affixed to Cell-Tak–coated (BD) 25-mm glass coverslips and loaded with 5 µM Fluo-4/AM in extracellular medium (ECM) as described previously (Madesh et al., 2005). For simultaneous measurement of [Ca²⁺]_c and mitochondrial Ca²⁺ uptake, cells were loaded with 2 µM rhod-2/AM (Invitrogen) and 5 µM Fluo-4/AM in ECM followed by an additional 10-min incubation in a dye-free medium. Coverslips were placed in a chamber and mounted in an open perfusion microincubator (PDMI-2; Harvard Apparatus) and maintained at 37°C on an inverted microscope (TE300 [Nikon] and Axio Observer [Carl Zeiss, Inc.]). 5 µg/ml mouse anti–chicken IgM (SouthernBiotech) was added after 1 min of baseline recording, and images were recorded every 3 s using a confocal imaging system (Radiance 2000; Bio-Rad Laboratories) or a laser-scanning confocal system (510 Meta; Carl Zeiss, Inc.) equipped with an Argon ion laser source at 488- and 568-nm excitation using a 60× oil objective. Images were acquired using either Lasersharp or ZEN 2008 software (Carl Zeiss, Inc). Images were analyzed and quantitated using ImageJ (National Institutes of Health) and a custom-made software (Spectralyzer). To assess Ca²⁺ entry, Ca²⁺-free ECM was used in conjunction with 0.5 mM EGTA. 2 mM Ca²⁺ was added as indicated.

GSH concentration was assessed via the reaction of GSH with Ellman’s reagent using a spectrophotometer according to the manufacturer’s protocol (Cayman Chemical). GSH was measured in intact cells using the fluorescent compound MCB (Invitrogen). In brief, 10 µM MCB was added directly to DT40 cells in complete medium and allowed to incubate at 40°C for 20 min. MCB will enter cells and bind to GSH to form the fluorescent adduct GSH-MCB. MCB will not react with oxidized GSSG. After loading, cells were washed, resuspended in HBSS, and fluorescence recorded using a confocal system (510 Meta) with excitation at 488 nm.

WT and STIM1 KO DT40 cells were challenged with 200 µM BSO. 24 h after BSO addition, cells were stained with 10 µM HE or 5 µM 5-(and-6)-chloromethyl-2′7′-dichlorodihydrofluorescein diacetate acetyl ester (CM-H₂DCF-DA or DCF) to detect superoxide anion and H₂O₂, respectively. Cells were washed, spun, and placed on the stage of a confocal imaging system (510 Meta). Excitation was 561 nm and 488 nm for HE and DCF, respectively. Images were analyzed using ImageJ software. For a positive control, BSO-treated DT40 STIM1 KO cells were incubated with 1 mM H₂O₂ for 30 min during DCF staining.

Recombinant human GST-tagged STIM1 was purchased from Abnova. 0.3 mg protein (in buffer containing 10 mM GSH) was incubated with either 1.0 or 0.1 mM H₂O₂ for 30 min at 23°C. Samples were resolved on a 4–12% Bis-Tris gel in the absence of reducing agents and probed with α-GSH antibody (1:1,000; Virogen). The membrane was then stripped and probed with α-STIM1 (1:500; BD).

Calcium binding to STIM1 was assessed as the protein-bound radioactivity retained after ultrafiltration procedure as described previously with minor modification (Ames et al., 2000). In brief, full-length GST-tagged STIM1 recombinant protein (Abnova) was subjected to the S-glutathionylation reaction as described in the previous paragraph. Both control and 0.1 mM H₂O₂–treated WT recombinant STIM1 (600 ng/sample) protein was incubated with 0.2 mM ⁴⁵CaCl₂ in phosphate-buffered saline, pH 7.4, for 30 min at room temperature. After washing twice, the radioactivity in the protein-containing solution was determined by liquid scintillation counting.

Oxygen consumption was measured using MitoCell (MT200; Strathkelvin Instruments). In brief, 10⁷ DT40 cells were washed in ECM, pelleted, resuspended, permeabilized in 110 µl intracellular medium containing 40 µg/ml digitonin, and placed into the MT200 chamber at 40°C under constant stirring. The oxygen electrode was calibrated using air-saturated ddH₂O and the oxygen chelator Na₂SO₃ for 21% and 0% dissolved oxygen, respectively. After 30-s equilibration, the following additions were added in the following order: 5 mM malate/pyruvate, 100 nM rotenone, 5 mM succinate, 50 nM antimycin A, 5 mM ascorbate/0.25 mM TMPD (tetramethyl-p-phenylenediamine), and 1 µM CCCP. Representative traces are displayed and are indicative of three independent experiments.

Untreated and 200 µM BSO–treated cells were activated with 1.5 µg/ml αIgM every 24 h, and cell viability was determined via nuclear TOTO-3 (Invitrogen) incorporation. In brief, cells were pelleted, stained with TOTO-3 (1:1,400), and imaged at 633/660-nm excitation/emission using a confocal imaging system (510 Meta). TOTO-3 is normally membrane impermeable and only becomes incorporated into the nucleus when the plasma membrane is compromised. In some experiments, untreated cells and those treated with either LPS or BSO were simultaneously loaded with TOTO-3 and the mitochondrial membrane potential indicator tetramethylrhodamine, ethyl ester (TMRE) to determine cell viability and mitochondrial function, respectively. Cells were challenged with BSO as above and 1 µg/ml LPS every 24 h. Samples were loaded with TOTO-3 and 100 nM TMRE for 15 min and imaged at 633/660-nm and 561/590-nm excitation/emission, respectively. Images were collected from five independent microscopic fields for three independent experiments, and simultaneous fluorescence was measured using ImageJ. Values indicate mean ± SEM (n = 3). Simultaneous fluorescence scatter plots for at least 600 data points are represented for both TOTO-3 and TMRE.

Substitution of alanine for cysteine at positions 49 and 56 in STIM1-mCherry constructs was performed using a site-directed mutagenesis kit (Quikchange; QIAGEN). COS7 and HEK293T cell lines were transfected using TransIT reagent according to manufacturer’s protocol (Mirus Bio LLC). DT40 cells were transfected via electroporation of 0.5 × 10⁶ cells in 0.5 ml RPMI using a gene pulsar apparatus (340 V and 950 microfarads; Bio-Rad Laboratories). STIM1 puncta were assessed using the WT STIM1-mCherry construct in Orai1 KO DT40 cells. In brief, cells transfected via electroporation were fixed on Cell-Tak–coated coverslips and imaged at 561 nm both before and after Tg addition in the presence or absence of BSO. YFP-STIM1ΔK (1–666), GFP-SOAR-LQ347/348AA, and mCherry-Orai1 were N-terminally labeled and were provided by S. Muallem (University of Texas Southwestern Medical Center, Dallas, TX).

Recombinant N-terminal portion of the STIM1 protein tagged with 0.3 µg calmodulin (Novus Biologicals) was incubated with 100 µM H₂O₂ in the presence of 10 mM GSH for 30 min as previously described (Aracena-Parks et al., 2006). Control conditions were incubated with 10 mM GSH without H₂O₂. Proteins were precipitated in acetone from the buffer solution. Two enzymes were used for the protein digestion: sequencing grade modified trypsin (Promega) and endoproteinase Asp-N (Roche) at 35°C overnight. For matrix-assisted laser desorption/ionization (MALDI) time of flight peptide mass fingerprinting, 0.3 µl digested peptides and 0.3 µl matrix (CHCA; Fluka) were spotted onto a MALDI target plate and allowed to dry. Mass spectra were acquired with a mass spectrometer (Reflex IV; Bruker Daltonics) between 500 and 5,000 m/z in reflectron mode and peptide peaks internally calibrated using trypsin autolysis peaks.

Protein lysates were prepared from COS7 cells transfected with STIM1 WT and mutant plasmids 36 h after transfection. Samples were immunoprecipitated with an α-STIM1 antibody and divided into two aliquots: one prepared with loading buffers lacking DTT (nonreducing conditions), and one prepared in DTT-containing loading buffers. Proteins were resolved on a 4–12% Bis-Tris gel (Invitrogen) and blotted with α-STIM1 (BD) and α-GSH (Virogen) antibodies for reducing and nonreducing conditions, respectively. After initial blotting, the α-STIM1 was stripped and reprobed with α-GSH to determine the S-glutathionylation reversibility. Plasma membrane Ca²⁺ ATPase levels in DT40 cells were assessed by immunoblotting. Anti-PMCA1/4 was purchased from Santa Cruz Biotechnology, Inc. (PMCA1/4).

In brief, linear voltage ramps of 50-ms duration spanning the voltage range of −100 to 100 mV were delivered from a holding potential of 0 mV at a rate of 0.5 Hz. The currents were filtered at 6 kHz and sampled at 10–50-µs intervals. We used automatic capacitive and series resistance compensation of the amplifier (EPC-10; HEKA). The intracellular solution contained 145 mM CsGlu, 10 mM Hepes, 10 mM EGTA, 8 mM NaCl, 6 mM MgCl₂, 2 mM Mg-ATP (total 8 mM Mg²⁺), and 3 mM CaCl₂, pH 7.2. 8 mM Mg²⁺ and ATP were used to inhibit TRPM7. According to WEBMAXCLITE (http://www.stanford.edu/∼cpatton/webmaxc/webmaxclite115.htm), the free Ca²⁺ concentration was 100 nM. The extracellular solutions contained 145 mM NaCl, 10 mM CaCl₂, 10 mM CsCl, 2 mM MgCl₂, 2.8 mM KCl, 10 mM Hepes, and 10 mM glucose, pH 7.4. For Ca²⁺-free solution, 10 mM CaCl₂ was replaced by 10 mM MgCl₂.

Fig. S1 shows the dose-dependent alterations in αIgM-induced calcium mobilization and mitochondrial calcium uptake in response to increasing concentrations of BSO. Fig. S2 demonstrates that Orai1 is requisite for the phenotypic alteration in calcium mobilization observed during oxidative stress. Fig. S3 shows the restoration of transient mitochondrial calcium uptake by elimination of capacitative calcium entry. Fig. S4 shows that Orai1 KO cells do not exhibit calcium entry in response to BSO. Fig. S5 illustrates that puncta formation under oxidant stress is dependent on the STIM1 N-terminal region.

We would like to thank Drs. Anjana Rao, Richard Lewis, Tamas Balla, and Shmuel Muallem for the STIM1 KO MEFs, STIM1-mCherry, Orai1-GFP, and STIM1ΔK (1–666) and STIM1 SOAR-LQ347/348AA plasmids, respectively. We thank the Fox Chase Cancer Center Proteomics facility for mass spectrometry studies. We would also like to thank Dr. Sheldon Feinstein, Mark Bollman, Pranavi Vemuri, and Roma Rajput for their experimental assistance.

This work was supported by the National Institutes of Health (grants HL086699, HL086699-01A2S1, and 1S10RR027327-01) and the American Heart Aassociation (Scientist Development grant to M. Madesh). B.J. Hawkins was supported by the National Institutes of Health (grant K99HL094536) and a postdoctoral fellowship from the Pulmonary Hypertension Association. S.K. Joseph is supported by the National Institutes of Health (grant DK34804).