Mitochondria control store‐operated Ca²⁺ entry through Na⁺ and redox signals

Abstract

Mitochondria exert important control over plasma membrane (PM) Orai1 channels mediating store‐operated Ca²⁺ entry (SOCE). Although the sensing of endoplasmic reticulum (ER) Ca²⁺ stores by STIM proteins and coupling to Orai1 channels is well understood, how mitochondria communicate with Orai1 channels to regulate SOCE activation remains elusive. Here, we reveal that SOCE is accompanied by a rise in cytosolic Na⁺ that is critical in activating the mitochondrial Na⁺/Ca²⁺ exchanger (NCLX) causing enhanced mitochondrial Na⁺ uptake and Ca²⁺ efflux. Omission of extracellular Na⁺ prevents the cytosolic Na⁺ rise, inhibits NCLX activity, and impairs SOCE and Orai1 channel current. We show further that SOCE activates a mitochondrial redox transient which is dependent on NCLX and is required for preventing Orai1 inactivation through oxidation of a critical cysteine (Cys195) in the third transmembrane helix of Orai1. We show that mitochondrial targeting of catalase is sufficient to rescue redox transients, SOCE, and Orai1 currents in NCLX‐deficient cells. Our findings identify a hitherto unknown NCLX‐mediated pathway that coordinates Na⁺ and Ca²⁺ signals to effect mitochondrial redox control over SOCE.

Introduction

The store‐operated Ca²⁺ entry (SOCE) pathway mediating the highly Ca²⁺ selective Ca²⁺ release‐activated Ca²⁺ (CRAC) current is the major Ca²⁺ influx route in non‐excitable cells (Putney, 1986; Potier & Trebak, 2008). The pathway is activated by interaction between the plasma membrane Ca²⁺ entry channel, Orai1 (Feske et al, 2006; Vig et al, 2006; Zhang et al, 2006), and the endoplasmic reticulum (ER) Ca²⁺ sensor protein, STIM1 (Liou et al, 2005; Zhang et al, 2005). Numerous studies have revealed that mitochondria are also key regulators of SOCE activity (Hoth et al, 1997; Deak et al, 2014). However, the exact mechanisms by which mitochondria facilitate SOCE are poorly understood. With the steep mitochondrial membrane potential as its driving force, Ca²⁺ permeates into the mitochondrial matrix via the recently identified mitochondrial Ca²⁺ uniporter (MCU; Baughman et al, 2011; De Stefani et al, 2011; Patron et al, 2013). Ca²⁺ is extruded from the mitochondria by an electrogenic mitochondrial 3Na⁺/1Ca²⁺ exchanger (NCLX; Palty et al, 2010) which is powered by the same mitochondrial membrane potential. This constant cycling of Ca²⁺ in and out of mitochondria functions as a dynamic and important signaling link between cytosolic Ca²⁺ signals and metabolic activity maintained by the regulation of key Krebs cycle enzymes as well as the F1/F0‐ATPase pump (Szabadkai & Duchen, 2008). The NCLX is a member of superfamily of Na⁺/Ca²⁺ exchangers and shares with them the hallmark of α1/α2 repeats that form the catalytic cation transport domain. It belongs to a distinct branch of this family termed CAX and is the single representative of this group in mammalian cells (Lytton, 2007; He & O'Halloran, 2014). It also has a unique selectivity to monovalent cations being capable of transporting either Li⁺ or Na⁺ in exchange for Ca²⁺, while other members of the family are inert to Li⁺ (Palty et al, 2010). Because the mitochondrial Ca²⁺ efflux rate is 2‐3 orders of magnitude slower than the rate of mitochondrial Ca²⁺ influx, the efflux represents a rate limiting step controlling the duration and magnitude of mitochondrial Ca²⁺ transients (Rudolf et al, 2004). In addition, by virtue of transporting at least 3 Na⁺ per 1 Ca²⁺, NCLX is the major mitochondrial Na⁺ influx pathway (Baysal et al, 1994; Jung et al, 1995; Nita et al, 2014a). In fact, the affinity of the NCLX for Na⁺ is ~10 mM and is therefore highly tuned to minor changes in cytosolic Na⁺ (Nita et al, 2014a).

Studies utilizing metabolic inhibitors and NCLX blockers suggest that mitochondria exert crucial regulation of CRAC channel activity (Malli et al, 2003; Ardon et al, 2009; Feldman et al, 2010). In immune cells, mitochondria are closely associated with the ER‐PM junctions in which CRAC channels function (Hoth et al, 1997; Rizzuto et al, 1998, 2012; Alvarez et al, 1999; Berridge et al, 2003; Singaravelu et al, 2011). Because of their spatial proximity and Ca²⁺ shuttling capacity, it was proposed that mitochondria buffer Ca²⁺ at the vicinity of CRAC channels thus eliminating the well‐established Ca²⁺‐dependent inactivation of the CRAC pathway (Parekh, 2008; Demaurex et al, 2009). However, in other cell types, for example, endothelial cells, mitochondria are more remotely located and are therefore less likely to directly buffer Ca²⁺ in the vicinity of CRAC channels (Demaurex et al, 2009; Giacomello et al, 2010; Naghdi et al, 2010; Nunes & Demaurex, 2014). Yet, remarkably, even in these cell types, the mitochondria can still impinge on CRAC activity through an as yet unidentified mechanism.

Interestingly, SOCE is also closely associated with the regulation of cytosolic Na⁺ and many studies have demonstrated that Na⁺ influx, mediated by non‐selective transient receptor potential canonical (TRPC) channel members, occurs in response to receptor‐mediated activation of CRAC channels (Lemos et al, 2007; Baryshnikov et al, 2009; Lee et al, 2010). Thus, Na⁺ signaling events are closely associated with SOCE and communication with mitochondria as recently reported (Reyes et al, 2013); however, the regulatory basis of this interaction remains poorly understood. SOCE is also strongly regulated by redox signals. These control several key steps including the interaction between STIM1 and Orai1 (Hawkins et al, 2010), and Ca²⁺ permeation by the Orai1 channel through modification of a redox‐sensitive cysteine located on the Orai1 channel (Bogeski et al, 2010). Despite the fact that mitochondria are the major effectors of cellular redox, it is still not understood how mitochondrial redox impacts SOCE activity and mitochondrial Ca²⁺ signaling.

In the current study, we have sought to understand the cross talk between NCLX and SOCE. We reveal that expression of NCLX is required for SOCE and CRAC channel regulation through a mitochondrial proximity‐independent pathway. We show that such regulation does not involve the elimination of Ca²⁺‐dependent inactivation of CRAC channels through mitochondrial buffering of Ca²⁺. We found instead that integrated Ca²⁺ and Na⁺ signals act synergistically to promote mitochondrial Ca²⁺ shuttling that modulates mitochondrial redox status, thereby controlling SOCE activity.

Results

SOCE and CRAC currents are regulated by mitochondrial NCLX

To determine the role of the mitochondrial Na⁺/Ca²⁺ exchanger, NCLX, we reduced its expression in HEK293T cells using previously characterized siNCLX (Palty et al, 2010). Consistent with previous results, HEK293T cells transfected with siNCLX displayed over twofold reduction in expression of NCLX protein compared to siControl cells with no significant change in the expression of STIM1 and Orai1 (Fig 1A). To determine the effect of NCLX knockdown on SOCE, we measured Ca²⁺ influx in Fura‐2‐loaded cells after depleting ER Ca²⁺ stores with Ca²⁺‐free Ringer containing ATP (100 μM) and thapsigargin (TG; 1 μM) followed by superfusion with Ca²⁺‐containing solutions (1.8 mM) while monitoring the rate of Ca²⁺ influx. Silencing of NCLX expression led to a 2.4‐fold reduction in Ca²⁺ influx rate (*P < 0.05) compared to cells transfected with siControl (1.2E‐02 versus 5E‐03; Fig 1B). As an additional strategy, we have employed the same experimental paradigm in cells expressing the plasma membrane‐targeted Ca²⁺ probe, GCamp5 (Akerboom et al, 2012) to determine the localized change in Ca²⁺ concentration at the vicinity of the plasma membrane. This localized Ca²⁺ influx signal in siNCLX‐transfected cells was also strongly (10‐fold) reduced compared to Ca²⁺ signals observed in siControl‐transfected cells (1.16E‐03 versus 1.1E‐04; *P < 0.05; Fig 1C). However, NCLX knockdown had no effect on STIM1 and Orai1 protein expression determined by Western blotting (Fig 1A). To directly establish whether NCLX has a role in controlling SOCE thereby ruling out indirect effects of cellular Ca²⁺ pumping or changes in membrane potential, we measured store depletion‐activated CRAC currents by whole‐cell electrophysiological recordings (Fig 1D–F). CRAC currents were recorded in Ca²⁺‐containing bath solutions and further amplified by exchanging bath solutions with divalent‐free (DVF). Such replacement immediately after break‐in (first DVF exchange) led to the appearance of a relatively small linear Na⁺ current, reflecting cell membrane or seal leak due to the absence of divalent cations (Fig 1D). During subsequent DVF exchanges, Ca²⁺ store depletion had taken place and the DVF inward current increased steadily while the outward current remain unchanged. In such experiments, the leak subtracted store depletion‐activated CRAC current recorded in DVF bath solution, displayed inward rectification typical of CRAC currents measured in Ca²⁺‐containing bath solutions. These CRAC Na⁺ currents also showed the typical rapid depotentiation in DVF bath solutions (Fig 1D). The current–voltage (I–V) relationships revealed inward rectification with a highly positive reversal potential and complete block by low concentrations of lanthanides (5 μM Gd³⁺; Fig 1D), consistent with the properties of CRAC currents recorded in many other cells. Just as NCLX knockdown reduced the store‐operated Ca²⁺ entry response, it also led to a twofold–threefold inhibition of both CRAC‐mediated Ca²⁺ and Na⁺ currents (Fig 1E and F). To determine whether the NCLX‐dependent changes in SOCE and CRAC are mediated by a change in STIM1‐Orai1 interactions, we monitored co‐localization of STIM1‐YFP and Orai1‐CFP. We determined that puncta formation after store depletion (Fig 2A and B) as well as enhanced STIM1‐Orai1 FRET (Fig 2C) was not affected by NCLX knockdown. Importantly, the time course of STIM1/Orai1 FRET interaction upon store depletion is indistinguishable between siControl and siNCLX cells (Fig EV1 and Movies EV1 and EV2). In contrast to the inhibitory effects that NCLX knockdown had on Ca²⁺ signals and CRAC current, the knockdown of MCU did not inhibit SOCE and CRAC currents in HEK293T cells (Appendix Fig S1).

Figure 1. NCLX controls SOCE activity.

1. Upper panel: Western blot was performed with 100 μg of protein samples obtained from HEK293T cells transfected with either siControl or siNCLX. Lower panel: Densitometry analysis of immunoblots of siControl‐transfected cells (n = 3) versus siNCLX‐transfected cells (n = 3) representing normalized expression levels (% of control) of NCLX, STIM1 and Orai1.
2. Upper panel: HEK293T cells were transfected with either scrambled siRNA construct (siControl, black) or siNCLX (red) and loaded with Fura‐2. SOCE was triggered by ATP (100 μM) and thapsigargin (TG, 1 μM), and Fura‐2 fluorescence was monitored as described in Materials and Methods. Lowe panel: Averaged rates of Ca²⁺ rise in siNCLX‐silenced cells from several independent recordings (n = 6) versus siControl‐transfected cells (n = 6).
3. Upper panel: Fluorescence of the membrane calcium sensor, GCamp5, targeted to the plasma membrane following treatment as described above using confocal microscopy. The scale bar represents 10 μm. Middle panel: Traces of membrane‐localized Ca²⁺ responses were measured in HEK293T cells co‐transfected with either the siNCLX (red) or siControl (black) and GCamp5‐expressing plasmid. The lower panel shows the averaged rates of membrane‐localized Ca²⁺ rise either in siControl cells (n = 6) or siNCLX cells (n = 10).
4. Electrophysiological recordings were performed on the same batch of transfected cells used for Western blot and shown in panel (A). Representative time courses of whole‐cell CRAC currents activated by dialysis of 20 mM BAPTA through the patch pipette and taken at −100 mV from cells transfected either with siControl (black trace) or siNCLX (red trace).
5. Representative I–V relationships are taken from traces in (D) where indicated by color‐coded asterisks.
6. Statistical analysis on Na⁺ CRAC currents measured at −100 mV is shown.

Data information: The results are presented as the means ± SEM. *P < 0.05; **P < 0.01; ***P < 1E‐03. P‐values indicate the results of an unpaired Student's t‐test.Source data are available online for this figure.

Figure 2. NCLX knockdown has no effect on Orai1 and STIM1 interaction following store depletion.

• A, B
  HEK293T cells stably expressing Orai1‐CFP and STIM1‐YFP were transfected with either control siRNA (A) or NCLX siRNA (B) and incubated for 72 h. E‐FRET experiments were then performed. The scale bar represents 20 μm.
• C
  Silencing of NCLX by siNCLX (red) has no effect on Orai1 and STIM1 colocalization after store depletion induced by applying 2.5 μM ionomycin to the bath solution, compared to siControl (black; means ± SEM).

Figure EV1. NCLX knockdown has no effect on Orai1/STIM1 interactions measured with FRET.

HEK293T cells expressing Orai1‐CFP and STIM1‐YFP were transfected with either control siRNA or NCLX siRNA and effect on Orai1/STIM1 corrected FRET after store depletion with 2.5 μM ionomycin (Iono) were compared. Bar scale: warmer colors indicate high FRET. The scale bar represents 10 μm.

Cultured primary aortic vascular smooth muscle cells (VSMCs) display robust SOCE as previously characterized in detail (Potier et al, 2009). We sought to determine whether similar cross talk between SOCE and mitochondrial NCLX exists in these cells. Silencing of NCLX expression led to a fourfold reduction in store‐operated Ca²⁺ influx rate measured with Fura‐2 (*P < 0.05) compared to cells transfected with siControl (2.1E‐03 versus 5.2E‐04; Fig 3A). The electrophysiological recordings of CRAC currents in VSMCs also showed twofold–threefold reduction in CRAC current density upon NCLX knockdown (Fig 3B and C). Together, these results indicate that mitochondrial NCLX is required for optimal activation of SOCE and CRAC currents.

Figure 3. Effect of NCLX knockdown on SOCE in primary vascular smooth muscle cells (VSMCs).

• A
  Upper panel: Fluorescence traces of cytosolic Ca²⁺ responses in VSMC cells. Lower panel: Averaged rates of cytosolic Ca²⁺ influx in siControl cells (n = 5) and siNCLX cells (n = 6).
• B, C
  Electrophysiological CRAC recordings in cells as in (A) using the same protocol used in Fig 2. Statistical analysis on Na⁺ CRAC currents measured at −100 mV is shown in (C).

Data information: The results are presented as the means ± SEM. *P < 0.05; ***P < 1E‐03. P‐values indicate the results of an unpaired Student's t‐test.

Mitochondrial control of store‐operated Ca²⁺ signaling is known to occur as a result of buffering the Ca²⁺ entering through CRAC channels by virtue of the close proximity of mitochondria to the channels (Glitsch et al, 2002). This mode of regulation is revealed when CRAC channel activation is measured in the presence of low Ca²⁺ buffer in the patch pipette (e.g. 0.1 mM EGTA). In contrast, other studies suggested that an elusive proximity‐independent mechanism is mediating mitochondrial regulation of CRAC channels (Giacomello et al, 2010; Pizzo et al, 2012). To address this, we determined whether NCLX regulation of CRAC current is dependent on Ca²⁺ buffering by analyzing CRAC current under varying cytosolic Ca²⁺ buffering conditions. Although the inhibitory effect of NCLX knockdown on CRAC channels persisted when CRAC channels were activated by a high concentration of a fast Ca²⁺ buffer (20 mM BAPTA; Figs 1 and 3), we considered that NCLX might be more intimately coupled to CRAC channels and therefore higher buffering capacity would be needed to recapitulate this close interaction. As shown in Fig 4A and B, the inhibitory effect of NCLX knockdown was essentially maintained whether 20 mM or 50 mM BAPTA was used indicating that NCLX is unlikely to be acting only through buffering cytosolic Ca²⁺ to relieve the Ca²⁺‐mediated negative feedback on CRAC channels. Thus, our results indicate that the regulation of SOCE by mitochondrial NCLX is at least partially mediated by a mechanism distinct from Ca²⁺ buffering in the vicinity of the mouth of CRAC channels.

Figure 4. NCLX does not control CRAC channels through buffering of cytosolic Ca²⁺ or mitochondrial energization.

• A, B
  Analysis of Ca²⁺ (A) and Na⁺ (B) CRAC currents measured at −100 mV in HEK293T cells transfected with either siControl or siNCLX using the indicated concentrations of BAPTA (20 mM versus 50 mM) in the pipette solution to activate CRAC and buffer cytosolic Ca²⁺.
• C–E
  Electrophysiological CRAC recordings were performed on cells where store depletion was induced by the use of a pipette solution containing low buffering capacity (0.1 mM EGTA) with IP₃ (30 μM) and a cocktail to energize the mitochondria whose composition is listed in the Results section. Representative time courses of whole‐cell current development at −100 mV (D, E) and measured in Ca²⁺‐containing and DVF solutions, respectively, by comparison with siControl. At the end of recordings, 5 μM Gd³⁺ was used to inhibit CRAC currents. Representative I–V relationships of Ca²⁺ (D) and Na⁺ (E) CRAC in cells transfected with siControl or siNCLX are taken from traces in (C) where indicated by color‐coded asterisks. Statistical analysis on Ca²⁺ and Na⁺ CRAC currents measured at −100 mV (D, E).

Data information: The results are presented as the means ± SEM. *P < 0.05; **P < 0.01 ***P < 1E‐03. P‐values indicate the results of a one‐way ANOVA test followed by Tukey post hoc analysis (A, B) or unpaired Student's t‐test (C–E).

A similar inhibitory effect of NCLX knockdown was observed in cells dialyzed with a mitochondrial‐energizing solution (see Materials and Methods). Thus, when the mitochondrial‐energizing cocktail was used, CRAC currents were 2.8‐fold bigger than those recorded with conventional pipette solution (Fig EV2). Importantly, and consistently with the results described in Fig 4A and B, we found that even under conditions of energized mitochondria (and hence enhanced mitochondrial Ca²⁺ buffering), NCLX knockdown still inhibits CRAC current (Fig 4C–E), further supporting our conclusion that NCLX acts through a pathway independent of mitochondrial Ca²⁺ buffering.

Figure EV2. CRAC channels are enhanced under physiological conditions of energized mitochondria.

• A, B
  Electrophysiological CRAC recordings were performed on HEK293T cells where store depletion was induced by the use of a pipette solution containing 0.1 mM EGTA with IP₃ with (B) and without (A) inclusion of a cocktail to energize the mitochondria. At the end of recordings, 5 μM Gd³⁺ was used to inhibit CRAC currents.
• C, D
  Representative I–V relationships of Ca²⁺ (C) and Na⁺ (D) CRAC in are taken from traces in (A) and (B) where indicated by color‐coded asterisks.
• E, F
  Statistical analysis on Ca²⁺ and Na⁺ CRAC currents measured at −100 mV.

Data information: The results are presented as the means ± SEM. ***P < 1E‐03. P‐values indicate the results of an unpaired Student's t‐test.

Sodium transport by NCLX controls Ca²⁺ influx via SOCE

One important distinction between NCLX and MCU is that the latter is solely a Ca²⁺ transporter while the former mediates exchange of three Na⁺ per Ca²⁺ ion and therefore exerts dual control of Ca²⁺ and Na⁺ transport (Pitts, 1979; Reeves & Hale, 1984). We asked whether Na⁺ plays a role in regulating SOCE by comparing the SOCE responses in cells superfused with bath solutions containing either physiological concentrations of Na⁺ (140 mM) or an equal concentration of NMDG⁺, a large monovalent cation that is not transported. Notably, replacing extracellular Na⁺ with NMDG⁺ led to a threefold decrease in store‐operated Ca²⁺ influx rate (Fig 5A) [5E‐03 versus 1.53E‐02; ***P < 1E‐03]. However, when Na⁺ ions were replaced with Li⁺, a substrate ion for NCLX but not of plasma membrane Na⁺/Ca²⁺ exchanger (NCX; Palty et al, 2004), the store‐dependent Ca²⁺ influx was rescued and was similar to that of the Na⁺‐containing Ringer solution (1.13E‐02 versus 1.49E‐02, respectively; Fig 5B, lower panel). Furthermore, silencing of NCLX expression combined with replacing Na⁺ ions in the Ringer's solutions with NMDG⁺ led to eightfold reduction in Ca²⁺ influx [1.49E‐02 and 1.8E‐03, respectively; ***P < 1E‐03 (Fig 5B)]. To determine whether intracellular Na⁺ rise is required for SOCE activation, cytosolic Na⁺ rise was triggered by blocking the Na⁺/K⁺ ATPase by preincubation with ouabain (100 μM; Fig 5C) in the presence of Na⁺ to load the cytosol with Na⁺. As expected, application of ouabain in the presence of extracellular Na⁺ led to an increase in cytosolic Na⁺ concentrations (Appendix Fig S2). We then compared the rates of store‐dependent Ca²⁺ influx in cells initially pretreated with ouabain in the presence of Na⁺ and then superfused with NMDG⁺ (without Na⁺) versus untreated cells superfused with either NMDG⁺ or Na⁺. Ouabain pretreatment in cells subsequently superfused with NMDG⁺‐containing bath solution had a similar effect of supporting SOCE to that of extracellular Na⁺ (Fig 5C, upper panel). Thus, preloading the cytosol with Na⁺ is sufficient for restoring store‐dependent Ca²⁺ influx rate in NMDG⁺‐containing bath solutions to similar values as in Na⁺‐containing bath solutions (1.78E‐02 and 1.41E‐02, respectively; Fig 5C, lower panel), but not in cells superfused with NMDG⁺ without preincubation with ouabain (8.5E‐03, **P < 0.01). This set of experiments indicates that the cytosolic rise in Na⁺ is required for full activation of SOCE.

Figure 5. Extracellular Na⁺ signaling controls Ca²⁺ influx through SOCE and CRAC channels.

• A, B
  Upper panel: Cytosolic Ca²⁺, Ca²⁺ [Cyto], responses monitored in HEK293T cells after store depletion by ATP and TG as in Fig 1B and upon extracellular Na⁺ ions replacement by either NMDG⁺ or Li⁺ ions. Lower panel: Averaged rates of cytosolic Ca²⁺ influx. The combined effects of NMDG⁺ together with silencing of NCLX on cytosolic Ca²⁺ influxes via SOCE are presented in (B).
• C
  Upper panel: After being loaded with Fura‐2, intracellular Na⁺, [Na⁺]_i was preloaded by preincubation for 20 min with ouabain (100 μM) in the presence of Na⁺, following loading of intracellular Na⁺ SOCE protocol was applied as in Fig 1 using NMDG⁺‐containing Ringer. Lower panel: Averaged Ca²⁺ influx rates in NMDG⁺ [Na⁺]_i preloaded cells (n = 6), Na⁺ control cells (n = 14), and cells perfused with NMDG⁺ alone (n = 13).
• D–H
  Representative Ca²⁺ CRAC current recordings and corresponding color‐matched I–V relationships measured in HEK293T cells with different monovalent cation‐based bath solution is shown in (D) for Na⁺, in (E) for NMDG⁺, and in (F) for Li⁺; 20 mM Ca²⁺ was included in all bath solutions. Statistical analysis on Ca²⁺ CRAC currents measured at −100 mV in all three conditions is shown in (G). The effects of cytosolic Na⁺ concentration on Ca²⁺ CRAC currents. (H) The whole‐cell patch clamp recordings were performed in HEK293T cells transiently expressing eYFP‐STIM1 and CFP‐Orai1. Scatter plots represent Ca²⁺ CRAC currents activated by dialysis through the patch pipette of a solution containing 20 mM BAPTA and the different concentration of Na⁺. Currents were measured at −100 mV and shown as current density pA/pF.

Data information: The results are presented as the means ± SEM. *P < 0.05; **P < 0.01 ***P < 1E‐03. P‐values indicate the results of a one‐way ANOVA test followed by Tukey post hoc analysis.

We then asked whether a rise in cytosolic monovalent cations will have a similar effect on CRAC currents in HEK293T cells. Bath solutions containing Ca²⁺ (20 mM) in the presence of the indicated cations (Na⁺, NMDG⁺ or Li⁺) were used to measure Ca²⁺ CRAC currents. The pipette solution contained 0.1 mM EGTA, IP₃ to deplete stores, together with the mitochondrial‐energizing cocktail described above. As shown in Fig 5D–G, the presence of either Na⁺ or the NCLX substrate Li⁺ in the bath solution led to development of inwardly rectifying Ca²⁺ CRAC currents that were blocked by 5 μM Gd³⁺. However, replacing Na⁺ in the bath solution with NMDG⁺ inhibited Ca²⁺ CRAC current activation in response to store depletion by IP₃. A scatter blot representation of these data showing individual recordings with mean ± SE is shown in Fig EV3.

Figure EV3. Substitution of Na⁺ by NMDG⁺ inhibits CRAC currents in HEK293T and RBL cells.

• A–C
  Scatter plots representing Ca²⁺ CRAC currents activated either by dialysis through the patch pipette of a solution containing either 0.1 mM EGTA+IP₃ (A) or 20 mM BAPTA (B) in HEK293T cells, or in RBL cells activated by dialysis of a solution containing 20 mM BAPTA (C). P‐values (A‐C) indicate the results of an unpaired student's t‐test.

Data information: The results are presented as the means ± SEM. **P < 0.01 ***P < 1E‐03. P‐values indicate the results of an unpaired Student's t‐test.

We obtained similar results when CRAC currents were activated in either HEK293T cells (Fig EV3A and B) or rat basophilic leukemia cells (RBL; Fig EV3C). Recordings in HEK293T cells were performed under two conditions: either with store depletion achieved by inclusion of IP₃ and mitochondrial‐energizing cocktail together with 0.1 mM EGTA in the patch pipette (Fig EV3A) or with 20 mM BAPTA in the patch pipette (Fig EV3B) in the presence of either extracellular Na⁺ or NMDG⁺. Although CRAC current was consistently inhibited by extracellular NMDG⁺ under all conditions, extracellular NMDG⁺‐mediated inhibition of CRAC current was more pronounced when CRAC currents were activated by 0.1 mM EGTA together with 30 μM IP₃ (Fig EV3A and B). Note that the slightly smaller potency of Li⁺ versus Na⁺ in stimulating CRAC and SOCE is consistent with the lower exchange rate of Li⁺ versus Na⁺ mediated by NCLX (Carafoli et al, 1974). This corresponds to the functional fingerprint of NCLX and reveals that NCLX is the integrating component between Na⁺ signaling and SOCE (Palty et al, 2004).

To determine the role of intracellular Na⁺ in mediating the action of CRAC current, we performed another set of experiments in which HEK293T cells were co‐transfected with STIM1 and Orai1 to generate large CRAC currents. Using these cells, whole‐cell patch clamp recordings were performed after store depletion induced by dialysis with 20 mM BAPTA through the patch pipette together with bath solution containing 135 mM Na⁺ and 20 mM Ca²⁺ in the presence of varying concentrations of Na⁺ in the patch pipette (Fig 5H). HEK293T cells co‐expressing STIM1 and Orai1 exhibited enhanced CRAC currents when the pipette solution contained 10 mM or higher Na⁺ concentrations (Fig 5H), providing additional support for the role of Na⁺ in optimal CRAC current activation.

Additional support for the cross talk between Na⁺ and Ca²⁺ ions on CRAC channel function was derived from direct monitoring of Na⁺ influx, applying the SOCE protocol to cells loaded with the fluorescent Na⁺ dye Asanta Natrium (Fig 6A and B). Depletion of Ca²⁺ stores triggered a 4.5‐fold rise in Na⁺ influx rate compared to control cells with intact Ca²⁺ stores [2.60E‐04 and 5.87E‐05, respectively, **P < 0.01] (Fig 6A, lower panel). This finding is consistent with previous studies documenting store‐dependent Na⁺ influx (Poburko et al, 2009). Na⁺ influx was observed only in the presence of extracellular Na⁺ indicating that it reflects Na⁺ permeation into cells and not release form putative intracellular Na⁺ stores (Fig 6B, upper panel) [2E‐04 versus ‐3E‐05, **P < 0.01] (Fig 6B, lower panel). We further asked whether the cytosolic Na⁺ rise is propagating into the mitochondria by monitoring changes in mitochondrial Na⁺ in cells preloaded with the mitochondrial Na⁺ reporter, CoroNa red (Jayaraman et al, 2001a,b). As shown in Fig 6C, SOCE activation was linked to a strong mitochondrial Na⁺ influx, consistent with the major role of NCLX in mediating mitochondrial Na⁺ uptake (Palty et al, 2010; Nita et al, 2014a). Mitochondrial Na⁺ uptake was however reduced in cells superfused with NMDG⁺ Ringer compared to cells superfused with Na⁺‐containing Ringer, (2.3E‐04 and 5E‐04, respectively, *P < 0.05; Fig 6C), indicating that mitochondrial Na⁺ influx is preceded by Na⁺ influx into the cell across the plasma membrane. Previous studies demonstrated that NCLX is tuned to sense small changes in cytosolic Na⁺ and is therefore strongly activated by a rise in cytosolic Na⁺ triggered, for example, by voltage gated Na⁺ channels in pancreatic β cells (Nita et al, 2014a). We therefore asked whether Na⁺ influx is activating SOCE by triggering activation of NCLX. We monitored mitochondrial Ca²⁺ extrusion via the mitochondrial NCLX (Palty et al, 2010) by performing the SOCE protocol as described in Fig 1 while monitoring mitochondrial Ca²⁺ in the absence or presence of extracellular Na⁺. The presence of Na⁺ ions in the Ringer solution strongly accelerated mitochondrial Ca²⁺ efflux (Fig 6D, upper panel). In contrast, mitochondrial Ca²⁺ efflux rate was reduced by 2.5‐fold in the presence of NMDG⁺ compared to cells that were superfused with Na⁺‐containing Ringer (2.26E‐04 and 5.7E‐04, respectively, **P < 0.01, Fig 6D, lower panel).

Figure 6. Depletion of intracellular Ca²⁺ stores triggers cytosolic Na⁺ influx which drives NCLX‐associated mitochondrial Na⁺ influx and Ca²⁺ efflux.

1. Upper panel: Fluorescence traces of cytosolic Na⁺ responses, Na⁺ [Cyto], in ATP‐depleted cells (n = 7) versus control non‐depleted HEK293T cells (n = 5) loaded with Asanta Natrium. Na⁺ influx averaged rates are shown in the lower panel.
2. Upper panel: Fluorescence traces of cytosolic Na⁺. Lower panel: Averaged rates of cytosolic Na⁺ influx in cells superfused with either Na⁺‐ (n = 11) or NMDG⁺ (n = 7)‐containing Ringer's solution.
3. Upper panel: Traces of mitochondrial Na⁺ transport in cells preloaded with the mitochondrial Na⁺ dye, CoroNa red in HEK293T cells. Lower panel: Mitochondrial Na⁺ influx rates in the presence (n = 7) or absence (n = 5) of Na⁺.
4. Upper panel: Traces of mitochondrial Ca²⁺, Ca²⁺ [Mito], recorded by monitoring RP‐mt fluorescence following Ca²⁺‐store depletion in HEK293T cells by ATP and TG as in Fig 1B in the presence of Na⁺ or NMDG⁺. Lower panel: Mitochondrial Ca²⁺ efflux in cells in the presence of NMDG⁺ (n = 8) compared to Na⁺ (n = 11).
5. Upper panel: Traces of Ca²⁺ [Mito] of cell superfused with ATP alone in Ca²⁺‐free Ringer followed by superfusion with Ca²⁺ Ringer with or without Na⁺ in shControl cells versus shNCLX cells. Lower panel: Rates of traces shown in the upper panel in the presence (n = 4) or absence (n = 3) of Na⁺ ions in cells transfected with shControl (n = 3) or shNCLX (n = 3).
6. Upper panel: Cytosolic Na⁺‐dose dependence of Ca²⁺ [Mito] efflux. Lower panel: Mitochondrial Ca²⁺ efflux rates at the indicated concentrations of Na⁺ [mM] [0 (n = 4); 1, 5 and 20 (n = 6)].

Data information: The results are presented as the means ± SEM. *P < 0.05; **P < 0.01; ***P < 1E‐03; ****P < 1E‐04. P‐values indicate the results of an unpaired Student's t‐test (A, B and D) or one‐way ANOVA test followed by Tukey post hoc analysis (C, E and F).

Thus far, we determined that the mitochondrial Ca²⁺ efflux triggered by robust store depletion using ATP+TG is Na⁺ dependent. Next, we sought to determine whether this is also the case when SOCE is activated using more physiological means (i.e. with purinergic ATP stimulation alone; Fig 6E, upper panel). As expected, absence of Na⁺ led to twofold lower ATP‐activated mitochondrial Ca²⁺ efflux than the control (with Na⁺; 1.13E‐04 and 5.13E‐05, respectively, **P < 0.01, Fig 6E, lower panel) and the combination of using NMDG⁺ Ringer together with silencing NCLX resulted in the lowest Ca²⁺ efflux rate (−1.02E‐06). Silencing of NCLX alone shows 3.6‐fold decrease in the Ca²⁺ efflux rate versus control (7.85E‐04 and 2.18E‐04, respectively; Fig EV4). As shown in Fig 6F, mitochondrial Ca²⁺ efflux following mitochondrial Ca²⁺ entry is strictly dependent on the presence of Na⁺, consistent with our previous studies (Nita et al, 2014a). Na⁺ dose dependence analysis of mitochondrial Ca²⁺ efflux revealed an apparent Km of 9.7 mM in HEK293T cells (Fig 6F), in agreement with values previously reported for cardiac myocytes (Cai et al, 2016) and consistent with Na⁺ physiological concentrations, which are typically within 4–16 mM range (Bers et al, 2003). Altogether our results indicate that Ca²⁺ influx through the SOCE pathway requires a cytosolic Na⁺ rise to propagate within mitochondria through the exchange activity of NCLX which activates mitochondrial Ca²⁺ efflux. The results further suggest that this Na⁺ is required for activating mitochondrial Ca²⁺ shuttling by NCLX hence ensuring optimal activation of SOCE and CRAC currents.

Figure EV4. Knockdown of NCLX expression inhibits mitochondrial Ca²⁺ efflux and Na⁺ uptake.

1. Upper panel: Mitochondrial Ca²⁺ was monitored in cells expressing the mitochondrial Ca²⁺ sensor RP‐mt. Mitochondrial Ca²⁺ transient was evoked by application of ATP and TG, added when indicated. Lower right panel: Rates of mitochondrial Ca²⁺ efflux derived from the upper panel. Lower left panel: The basal Ca²⁺ level (arbitrary units, AU) of shControl (n = 3) versus shNCLX (n = 4).
2. Upper panel: Mitochondrial Na⁺ uptake in HEK293T cells preloaded with the mitochondrial Na⁺ sensor, CoroNa red. Na⁺ signals were evoked by ATP and TG added when indicated. Lower panel: Rates of mitochondrial Na⁺ influx derived from the upper panel.

Data information: The results are presented as the means ± SEM. *P < 0.05. P‐values indicate the results of an unpaired Student's t‐test.

SOCE and CRAC current are regulated by mitochondrial redox

Mitochondrial redox state is elicited by mitochondrial Ca²⁺ transients and in turn regulates SOCE. We reasoned that the NCLX knockdown effect on SOCE is mediated by mitochondrial redox (De Marchi et al, 2014). To determine whether NCLX knockdown affects mitochondrial redox state, we applied the SOCE protocol as described in Fig 1 in cells expressing the redox sensor roGFP1 targeted to the mitochondrial matrix. Store depletion by ATP and thapsigargin (TG) in the absence of external Ca²⁺ followed by addition of Ca²⁺ to the bath solution, triggered transient changes in the mitochondrial redox state, temporarily reducing mitochondrial free oxygen radical levels in the mitochondrial matrix (Fig 7A). In contrast, knockdown of NCLX expression was followed by a twofold inhibition of this redox change (Fig 7A). Consistent with the role of NCLX in controlling the mitochondrial redox state, the knockdown of NCLX was followed by a fivefold lower NADH/NAD⁺ ratio, monitored by intrinsic fluorescence (2.4E‐04 and 1.2E‐03, respectively, *P < 0.05; Fig 7B, lower panel). These results indicate that the redox transients maintained by NCLX limit free radical bursts and thereby prevent oxidant‐mediated inhibition of SOCE/CRAC.

Figure 7. Mitochondrial catalase (m‐catalase) rescues SOCE, CRAC, and mitochondrial redox responses after NCLX knockdown.

• A
  Upper panel: Traces of mitochondrial roGFP1 fluorescence in HEK293T cells co‐transfected with either siControl or siNCLX and Ca²⁺‐ store depleted as described in Fig 1B. Changes in roGFP1 410/480 ratio fluorescence were determined using minimal and maximal values obtained by application of 100 mM DTT and 10 mM H₂O_2, respectively, as described in Materials and Methods. Lower panel: ΔF ratio of redox change after Ca²⁺ restoration to the external milieu in siControl cells (n = 8) compared to siNCLX cells (n = 12).
• B
  Upper panel: Effect of NCLX on redox state determined by monitoring NAD(P)H intrinsic fluorescence in HEK293T cells, transfected with either siNCLX or siControl, and treated as described in Fig 1B. Oligomycin or FCCP were used for calibration and added where indicated. Lower panel: Averaged autofluorescence rates after adding Ringer's solution containing Ca²⁺.
• C
  Upper panel: Effect of mitochondrial catalase (pZeoSV2 + mCat) expression. Lower panel: Rates of redox change in siNCLX + pZeoSV2 + (pZeo) empty vector transfected cells (n = 4) compared to siNCLX cells transfected with m‐catalase (n = 6).
• D
  Upper panel: Fluorescence traces of cytosolic Ca²⁺ responses in HEK293T cells co‐transfected with either siControl or siNCLX together with pZeo or m‐catalase. Lower panel: Averaged rates of Ca²⁺ influx in cells transfected with siNCLX (n = 9) or cells co‐transfected with siNCLX and either m‐catalase (n = 9) or pZeo (n = 5) compared to cells transfected with siControl (n = 6) or co‐transfected with siControl and m‐catalase (n = 5).
• E–H
  CRAC electrophysiological recordings in HEK293T cells activated by dialysis of 20 mM BAPTA through the patch pipette under the same conditions of transfection used in (B). I–V relationships are shown in (G) and statistical analysis on Na⁺ CRAC currents measured at −100 mV is shown in (H). m‐Catalase is routinely co‐transfected with a plasmid encoding eGFP for identification of transfected cells.

Data information: The results are presented as the means ± SEM. *P < 0.05; **P < 0.01 ***P < 1E‐03. P‐values indicate the results of an unpaired Student's t‐test (A–C) or one‐way ANOVA test followed by Tukey post hoc analysis (D–H).

Hence, the SOCE/CRAC activity suppressed by NCLX knockdown could be rescued by mitochondrial targeted expression of an H₂O₂‐metabolizing enzyme, mitochondrial catalase (m‐catalase). We determined whether m‐catalase can restore the mitochondrial redox response in NCLX‐silenced cells. Note that the m‐catalase expression fully rescued the SOCE‐dependent increase in reduced state of the mitochondrial matrix (Fig 7C). We then compared SOCE rate in siControl versus siNCLX cells with or without m‐catalase expression. Remarkably, overexpression of m‐catalase in NCLX knockdown cells was sufficient to recover SOCE activity and Ca²⁺ influx rate (Fig 7D). Further support for the link between Na⁺ and NCLX in controlling SOCE by redox is demonstrated by our finding that expression of m‐catalase rescued SOCE even in the absence of extracellular Na⁺ (Fig EV5).

Figure EV5. Na⁺ has no effect on SOCE in the presence of m‐catalase.

• A
  HEK293T cells were transfected with either siControl or siNCLX with or without m‐catalase. Cytosolic Ca²⁺ responses were monitored in HEK293T cells after store depletion by ATP and TG as in Fig 1B in the presence or absence of Na⁺ ions.
• B, C
  Averaged rates (means ± SEM) of Ca²⁺ influx in either siControl cells (n = 4), siNCLX (n = 4), siNCLX+m‐catalase with Ringer containing Na⁺ (n = 4), or NMDG⁺ (n = 4) are shown in (B). P‐values indicate the results of a one‐way ANOVA test followed by Tukey post hoc analysis. *P < 0.05. The basal redox levels of siNCLX+pZeo cells (n = 6) versus siNCLX+m‐catalase (n = 4) are shown in (C).

We then asked whether the rescue of the mitochondrial redox transient by m‐catalase could restore CRAC currents suppressed by knockdown of NCLX expression. Consistent with the above SOCE results (Fig 7C and D), we observed that knockdown of NCLX expression was followed by a strong inhibition of CRAC current (Fig 7E–H). Remarkably, targeting catalase to the mitochondria was sufficient to rescue CRAC currents in cells in which NCLX was knocked down (Fig 7F and H). We compared the basal redox state of siNCLX cells vs siNCLX cells expressing m‐catalase and found that siNCLX cells are slightly more oxidized, although this difference was not significant (Fig EV5C). Together, our results strongly support a novel paradigm whereby NCLX control of the mitochondrial redox state participates in mitochondrial communication with SOCE.

Since our results (Fig 2) indicate that the NCLX effect on SOCE is not mediated via disruption of STIM1 and Orai1 interactions, we reasoned that the mitochondria redox response evoked by NCLX is targeting a redox‐sensitive residue on Orai1. Orai1 possesses three cysteine residues potentially modified by oxidants, Cys126, Cys143, and Cys195. One of the Orai1 cysteine residues, Cys195, has recently emerged as a redox responsive element within this channel (Bogeski et al, 2010). We therefore focused on Cys195 within Orai1 and compared the wild‐type SOCE response (Fig 8A) to the SOCE response of either Orai1 or Orai1 C195S coexpressed with STIM1 in either siControl‐ or siNCLX‐transfected HEK293T cells (Fig 8B and C). Remarkably, while the silencing of NCLX was followed by reduction in native as well overexpressed Orai1‐dependent Ca²⁺ influx, mutating Cys195 to serine (C195S) fully rescued Orai1‐dependent SOCE in NCLX knockdown cells (Fig 8C). To establish the role of this residue in mediating the NCLX‐dependent redox effect on Orai1, we analyzed Orai1‐mediated CRAC currents in HEK293T cells expressing STIM1 with either Orai1 or the Orai1‐C195S mutant (Fig 8D–I) while depleting internal stores and strongly buffering cytosolic Ca²⁺ with 20 mM BAPTA, as defined in Fig 1. Consistent with data obtained with SOCE measurements using Fura‐2, the inhibitory effect of NCLX knockdown was also observed on CRAC currents in cells over‐expressing STIM1 and Orai1. In contrast, mutating the redox‐sensitive cysteine 195 to serine (C195S) fully rescued Orai1‐mediated CRAC currents following the knockdown of NCLX. Mutation of all three cysteines Cys126, Cys143, and Cys195 in Orai1 to serine also fully rescued Orai1‐mediated CRAC currents following the knockdown of NCLX (Fig 8J–L). CRAC currents mediated by this triple Orai1 mutant were significantly bigger than those mediated by either Orai1 or Orai1‐C195S (Orai1; 17.62, C195S‐Orai1; 16.52, C126S/C143S/C195S‐Orai1; 51.59). Taken together, our results indicate that the mitochondrial redox transients controlled by NCLX prevent CRAC channel inhibition mediated by oxidation of cysteine 195 on Orai1.

Figure 8. C195S mutation in Orai1 prevents the inhibitory effect of SOCE and CRAC current caused by NCLX knockdown.

• A–C
  HEK293T cells were transfected with either control siRNA (black) or NCLX siRNA (red) and incubated for 72 h (A). Cells were then transfected again with siRNA along with either plasmids, STIM1 and Orai1 (B), or plasmids, STIM1 and C195S‐Orai1 (C), and incubated for another 24 h. The numbers in bar graphs (i.e., (x, y)) represent the total number of independent recording (x) and the total number of cells from all recordings (y). SOCE was triggered as described in Fig 1B, and Fura‐2 fluorescence was monitored as described in Materials and Methods. Averaged rates of Ca²⁺ influx were shown in the lower panels.
• D–L
  Representative time course traces (D and G) of Orai1‐ and Orai1‐C195S‐mediated CRAC currents activated by dialysis through the patch pipette of a solution containing 20 mM BAPTA and recorded in a bath solution containing 20 mM Ca²⁺. I–V curves are shown in (E and H) which are taken from traces as indicated by color‐coded asterisks. Scatter blots (F and I) depict maximal CRAC currents values taken at −100 mV and represented as current densities (pA/pF). Similar CRAC current recordings from HEK293T cells expressing STIM1 and C126S/C143S/C195S‐Orai1 triple mutant are shown in (J–L).

Data information: The results are presented as the means ± SEM. *P < 0.05; ns, not significant. P‐values indicate the results of an unpaired Student's t‐test.

Discussion

Although a role for mitochondrial Ca²⁺ signaling in regulating SOCE is well established, the mode of communication between these two domains is not understood. The recent identification of the mitochondrial Ca²⁺ transporters enables more selective molecular tools to specifically control their activity and study their role in controlling SOCE. In this study, we have focused on NCLX whose rate of mitochondrial Ca²⁺ removal is two orders of magnitude slower than the mitochondrial Ca²⁺ influx. Thus, NCLX is the rate limiting component in mitochondrial Ca²⁺ shuttling. NCLX is also a Na⁺ transporter that mediates the exchange of 3 Na⁺ per 1 Ca²⁺ making it the major mitochondrial Na⁺ uptake pathway (Baysal et al, 1994; Jung et al, 1995). Na⁺ transport was previously linked to SOCE. Thus, it was proposed that Na⁺ influx occurs through concomitant activation of non‐selective TRPC cation channels in response to pharmacological or receptor‐mediated store depletion (Parekh & Putney, 2005). However, the role of cytosolic Na⁺ in regulating the highly Ca²⁺ selective Orai1 channel is still poorly understood. In this study, we sought to address the following questions. Does NCLX control SOCE‐dependent Ca²⁺ signaling? And, if so, what is the mechanism of NCLX‐mediated regulation of SOCE? Our results indicate that NCLX is required for SOCE and support a redox effect of NCLX on CRAC channels that is independent of mitochondrial Ca²⁺ buffering.

An important mode of cross talk between CRAC channels and mitochondria is based on the ability of mitochondria to control Ca²⁺ concentrations at the vicinity of the CRAC channel thereby decreasing their Ca²⁺‐dependent inactivation. Such interaction between the mitochondria and CRAC channels is particularly prominent in immune cells (Rizzuto et al, 1998, 2012; Alvarez et al, 1999; Berridge et al, 2003; Singaravelu et al, 2011), in which inhibition of MCU and, more recently, molecular knockdown of MCU have demonstrated that SOCE is inhibited by the block of mitochondrial Ca²⁺ uptake in immune cells (Gilabert et al, 2001; Ma & Beaven, 2011). In other cell types, however, the proximity of the mitochondria to CRAC channels is less pronounced (Naghdi et al, 2010). Studies employing organellar Ca²⁺ reporters targeted, for example, to the outer mitochondrial membrane and the plasma membrane suggest that mitochondria are not mediating strong Ca²⁺ changes at the vicinity of the CRAC channel (Giacomello et al, 2010). Furthermore, inhibition of mitochondrial efflux by the NCLX inhibitor CGP‐37157b enhances mitochondrial Ca²⁺ uptake rather than decreasing it and yet leads to similar inhibition of SOCE. These findings indicate that in addition to the Ca²⁺‐dependent inactivation, there is an additional missing link that participates in the cross talk between the mitochondria and SOCE. The first question that we have addressed is whether a Ca²⁺‐dependent inactivation or a distinct regulatory mode is solely responsible for the cross talk between NCLX and SOCE. If Ca²⁺‐dependent inactivation is the predominant mode at play, then buffering of cytosolic Ca²⁺ should rescue CRAC currents when NCLX is silenced. Our results indicate that Ca²⁺ buffering is not involved in NCLX‐mediated control of SOCE. While it may be argued that chelation by EGTA is not sufficiently strong or rapid, we have used the stronger and faster Ca²⁺ chelator BAPTA at high concentrations (up to 50 mM) and failed to rescue CRAC currents upon NCLX knockdown.

Our results further indicate that store‐associated activation of Na⁺ influx across the plasma membrane is essential for NCLX function thereby maintaining SOCE and CRAC currents (Fig 9) based on our following findings. First, consistent with previous results (Rosker et al, 2004; Eder et al, 2007; Lemos et al, 2007; Demaurex et al, 2009), we find that store depletion is followed by cytosolic Na⁺ influx which is then transported by NCLX to the mitochondria. Second, we find that the SOCE‐associated cytosolic Na⁺ influx is required to activate NCLX‐dependent mitochondrial Ca²⁺ extrusion. The requirement for cytosolic Na⁺ rise for activation of NCLX is consistent with the low affinity of this exchanger for Na⁺ such that under resting cytosolic Na⁺ levels, NCLX will be inactive while a small increase in cytosolic Na⁺ will activate a large mitochondrial Ca²⁺ efflux by NCLX. The rise of cytosolic Na⁺ in response to store depletion can occur, for example, by the well described Ca²⁺‐dependent activation of non‐selective TRPC cation channels (Lemos et al, 2007; Lee et al, 2010). Thus, NCLX can be physiologically considered as a Na⁺‐activated Ca²⁺ exchanger. Consistent with this hypothesis, we also found that extracellular Na⁺ is required to fully activate CRAC channels. Thus, Na⁺‐dependent activation of NCLX is required for full activation of SOCE. Perhaps most remarkably, Li⁺, the substrate cation of NCLX but not of plasma membrane NCX members, can effectively replace Na⁺ in supporting SOCE and CRAC currents. Notably, SOCE and CRAC channel activity monitored in the presence of Li⁺ is slightly lower than that of Na⁺ thus bearing the exact monovalent cation finger print of NCLX selectivity (Carafoli et al, 1974; Palty et al, 2004). The latter finding also provides strong evidence for a distinct role of NCLX in the regulation of CRAC currents because plasma membrane NCX members are inert to Li⁺. However, it is important to note that the reverse mode of the plasma membrane NCX triggered by a SOCE‐associated Na⁺ rise (e.g. through TRPC6) is also contributing to cytosolic Ca²⁺ rise (Borin et al, 1993; Lemos et al, 2007, 2007). Unlike NCLX, the NCX effect on cytosolic Ca²⁺ is not mediated through direct action on CRAC conductance. Therefore, our current and previous studies (Borin et al, 1993; Lemos et al, 2007; Nita et al, 2014a,b) indicate that NCLX and NCX concertedly amplify SOCE‐dependent cytosolic Ca²⁺ signals but through different mechanisms. For NCX, the SOCE‐dependent Na⁺ rise triggers its reverse mode leading to increased cytosolic Ca²⁺ while for NCLX it triggers forward activation that can lead to maintenance of CRAC channel activity (Fig 9). How does the powering of NCLX by cytosolic Na⁺ culminate in the activation of SOCE and CRAC currents? Mitochondria are not only the metabolic engine and a major hub for Ca²⁺ signaling, they function as major regulators of cellular redox potential (Szabadkai & Duchen, 2008; Nunes & Demaurex, 2014) that link between activation of NCLX to SOCE and CRAC currents.

Figure 9. Na⁺ and Ca²⁺ signals are regulated by NCLX and mediate mitochondrial redox control of SOCE.

A schematic illustration of the Na⁺ and Ca²⁺ signaling networks involving NCLX and SOCE cross talk. Ca²⁺ depletion of intracellular stores activates store‐operated Ca²⁺ entry (SOCE) pathway mediated through the highly Ca²⁺ selective, Ca²⁺ release‐activated Ca²⁺ (CRAC) currents. SOCE activation involves interaction upon store depletion of the endoplasmic reticulum (ER) Ca²⁺ STIM1 with the plasma membrane channel protein Orai1. Upon store depletion, rise of cytosolic Na⁺ also occurs, apparently, reflecting the activation of non‐selective transient potential channel canonical (TRPC) channel isoforms. Cytosolic Na⁺ is required for powering NCLX to activate mitochondrial Ca²⁺ efflux, thus ensuring proper mitochondrial Ca²⁺ shuttling. This tight regulation by NCLX of mitochondrial matrix Ca²⁺ homeostasis prevents the excessive Ca²⁺ rise in mitochondrial matrix that would otherwise lead to increases in mitochondrial reactive oxygen species (ROS) and subsequent inhibition of SOCE via oxidation of a reactive cysteine at position C195 on an extracellular loop of Orai1.

Our results indicate that NCLX regulation of SOCE persists following the knockdown of MCU. Although the reason for this observation is not entirely clear, it is important to note that mitochondrial Ca²⁺ transients initiated by MCU reflect only ~5% of the total mitochondrial Ca²⁺. The rest, ~95% of the mitochondrial Ca²⁺ pool, is an insoluble/soluble pool consisting mainly of the phosphate Ca²⁺ salt. The solubility of the latter pool is strongly pH dependent and therefore any change with matrix pH; for example, metabolic activity will trigger a matrix‐free Ca²⁺ change and thereby a change in mitochondrial redox. Affinity and sensitivity of the mitochondrial Ca² sensitive dyes are, however, less than optimal (Pizzo et al, 2012). Therefore, these changes are often left unnoticed. Consistent with this hypothesis, our pervious and current results show that the apparent duration of mitochondrial Na⁺ fluxes (which in contrast to Ca²⁺ is unbuffered) is much longer than mitochondrial Ca²⁺ transients (Palty et al, 2010; Nita et al, 2014a), indicating that NCLX is pumping out Ca²⁺ for much more extensive periods than the “apparent” mitochondrial Ca²⁺ transients. Finally, in contrast to MCU, NCLX is dually linked to Ca²⁺ and Na⁺ signaling which we show here is critical for redox regulation and the SOCE response.

Regulation by redox has recently emerged as a critical mode of SOCE regulation (Bogeski et al, 2010; Hawkins et al, 2010; Nunes & Demaurex, 2014).

There are several studies which show regulation of m‐catalase (Naziroglu, 2012; Littlejohns et al, 2014). However, the H₂O₂ clearing rate of catalase is mainly dependent on the levels of expression of catalase (Rodriguez et al, 2000). Therefore, over‐expressed catalase targeted to the mitochondria (as described here) is expected to enhance H₂O₂ detoxification in the mitochondria generated by abnormal mitochondrial Ca²⁺ rise.

Our previous results indicate that expression and activity of NCLX are critical in triggering a mitochondrial redox signal (De Marchi et al, 2014). Our current results demonstrating that catalase rescues the mitochondrial redox response and CRAC/SOCE in NCLX‐deficient cells, suggest that diffusible H₂O₂ is the link between NCLX and SOCE (Fig 9). While buffering of cytosolic Ca²⁺ failed to rescue SOCE and CRAC currents in NCLX knockdown cells, expression of mitochondrial catalase leads to a full recue of SOCE in NCLX knockdown cells. It may be argued that this effect is indirect and associated, for example, with ER Ca²⁺ release. However, our electrophysiological analysis in cells expressing the mitochondrial catalase also show rescue of CRAC currents. Redox potential is a key regulator of SOCE that can act on multiple targets in this pathway. The oligomerization of STIM1 and its interaction with Orai1 are redox‐sensitive. However, our results indicate that the oligomerization of STIM1 and its interaction with Orai1 (as well as their expression) are not affected by knockdown of NCLX, arguing against such a scenario. Instead, our results suggest that the target of the mitochondrial redox response is an Orai1 cysteine residue (Cys195). Oxidation of the latter by H₂O₂, leads to CRAC channel inhibition (Bogeski et al, 2010). Furthermore, our finding showing that substitution of cysteine 195 in Orai1 with serine (C195S) is sufficient to rescue SOCE and CRAC currents indicate that redox‐dependent regulation of SOCE by mitochondria is mediated through cysteine 195 of Orai1 (Fig 9).

In conclusion, the results of this study identify a long sought and novel mode of interaction between the mitochondria and SOCE. This pathway integrates Na⁺ and Ca²⁺ signals that converge on the mitochondrial NCLX Na⁺/Ca²⁺ exchanger, to modulate cellular redox which controls SOCE and CRAC currents, mediated by a redox‐sensitive site on Orai1 (Fig 9).

Materials and Methods

Cell culture and transfection

HEK293T cells (human embryonic kidney cell line) were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 1% penicillin/streptomycin, 2 mM l‐glutamine. Transfection of HEK293T cells was performed using the CaPO₄ precipitation protocol in cultures of 30–50% confluence, as previously described (Palty et al, 2004). Vascular smooth muscle cells (VSMCs) were cultured on glass coverslips and transfected with DharmaFECT 1 (Dharmacon, T‐2001). SiRNA NCLX or siControl was diluted in DharmaFECT siRNA transfection reagent, incubated ~20 min at room temperature and then added in the antibiotics‐free media. The efficiency of transfection was assessed by visualizing Dharmacon siGLO Red (Dharmacon, D‐001630‐02‐05) transfection particles according to the protocol provided by manufacturer. The transfection efficiency, for siNCLX delivery as determined by siGlo fluorescent marker was high, ~90%. VSMCs were isolated as previously described (Potier et al, 2009) and maintained in culture 45% DMEM and 45% Ham's F12 with 10% FBS supplemented with l‐glutamine) at 37°C, 5% CO₂, and 100% humidity, passaged and used within 1–5 passages. Rat basophilic leukemia (RBL‐2H3) mast cells were obtained from ATCC, cultured in EMEM with 2 mmol/l l‐glutamine and 10% FBS and maintained in a 37°C, 5% CO₂ humidified incubator as described previously (Abdullaev et al, 2008).

Plasmid and siRNA preparation

Double‐stranded siRNAs used to silence NCLX expression were obtained from Applied Biosystems as previously described (Palty et al, 2010). The human NCLX shRNA plasmid was obtained from Sigma (Mission TRC shRNA Target Set TRCN‐5045). The shMCU and shControl plasmids were a gift from Fabiana Perocchi (TRCN0000133861; Baughman et al, 2011). The membrane‐targeted GCAMP5 was obtained from Addgene [pN1 Lck‐GCaMP5G (Plasmid #34924)]. The RP‐mt plasmid was provided by Atsushi Miyawaki (Wako, Japan; Nagai et al, 2001). The roGFP1 was kindly gifted by Roger Y. Tsien (UCSD; Dooley et al, 2004). The pZeoSV2+ mCat and pZeoSV2+ empty vector were provided by J. Andres Melendez (Albany, USA; McCarthy et al, 2013). The plasmid of eYFP‐STIM1 was kindly gifted by Tobias Meyer; the plasmid of CFP‐Orai1 was provided by James W. Putney; and the plasmids of C195S‐Orai1‐eGFP and C126S/C143S/C195S‐Orai1‐eGFP were constructed by Mohamed Trebak's laboratory. All those tagged or mutant of STIM1 and Orai1 plasmids are used for patch clamp experiments.

Immunoblot analysis

Immunoblot was performed with 100 μg of protein samples obtained from HEK293T cells transfected with either siControl or siNCLX. Detection for NCLX, STIM1, and Orai1 proteins was performed 96 h post‐transfection using specific antibodies; HSC 70 loading control is also shown. Anti‐NCLX polyclonal antibody was generated in our laboratory as previously described (Palty et al, 2010) while anti‐STIM1 monoclonal antibody was from BD Biosciences, anti‐Orai1 polyclonal antibody was from Alomone Labs, and anti‐HSC 70 was from Santa Cruz Biotechnology.

Fluorescent Ca²⁺ and Na⁺ imaging

The imaging system consisted of an Axiovert 100 inverted microscope (Zeiss, Oberaue, Germany), Polychrome V monochromator (Till Photonics, Planegg, Germany) and a Sensi‐Cam cooled charge‐coupled device (PCO, Kelheim, Germany). Fluorescence images were acquired with Imaging WorkBench 6.0 software (Axon Instruments, Foster City, CA, USA). Ca²⁺ imaging was performed in HEK293T and VSMCs cells that were attached onto coverslips and superfused with Ringer's solution containing (in mM): 126 NaCl, 5.4 KCl, 0.8 MgCl₂, 20 HEPES, 1.8 CaCl₂, and 15 glucose; pH was adjusted to 7.4 with NaOH or NMDG⁺ in Na⁺‐free Ringer's solutions (Palty et al, 2010). For cytosolic Ca²⁺ measurements, HEK293T and VSMCs cells were loaded with Fura‐2 AM (2 μM) for 25 min at room temperature or 4 μM for 45 min at 37°C, respectively (Bisaillon et al, 2010). Cell were excited with 340/380 nm wavelength light and imaged using a 510‐nm long‐pass filter, as described previously (Jonkers & Henquin, 2001). For measurements of plasma membrane‐localized Ca²⁺ signals, HEK293T cells were transfected with 1 μg of GCamp5 which is targeted solely to the plasma membrane. The membrane‐targeted Gcamp5 was excited at 485 nm, and fluorescence was monitored at 510 nm emission (Akerboom et al, 2012). Mitochondrial Ca²⁺ measurements were performed in HEK293T cells expressing ratiometric mitochondrial pericam (mitopericam), which is targeted solely to the inner membrane of mitochondria. The mitochondrial pericam fluorescence in HEK293T cells was acquired at 430 nm excitation and 550 nm emission, as described previously (Nagai et al, 2001). Cytosolic Na⁺ levels were recorded in Asanta Natrium Green‐2 (Teflabs, Jackson Springs, NC, USA)‐loaded HEK293T cells excited with 485 nm and imaged at 510‐nm long‐pass filter. It is well known that CoroNa Red is localized to the mitochondria (Yang et al, 2004; Baron et al, 2005; see in Appendix Fig S3). Therefore, mitochondrial Na⁺ signals were monitored in cells loaded with CoroNa Red (Molecular Probes) at excitation of 568 nm and emission at 590 nm, respectively (Lemos et al, 2007). Note that at 50 s, Na⁺ fluorescence falls a bit. This change is apparent in the presence or absence of Na⁺ and therefore a fluorescence artifact due to addition of solution.

For all single‐cell imaging experiments, traces of averaged responses were analyzed and plotted using Origin Labs software. The fluorescent Na⁺ or Ca²⁺ signals were normalized once again to the averaged baseline signal (F/F ₀ or R/R ₀), obtained at the beginning of the measurements. The influx and efflux rates were derived from a linear fit of the fluorescence change during 30 s (Palty et al, 2010; Nita et al, 2012). Averaged rates of the fluorescent Na⁺ or Ca²⁺ responses, over n (indicated in the figure legends) experiments, are presented in the bar graphs and analyzed using Origin software.

Mitochondrial redox state measurements

Ratiometric measurements of the mitochondrial redox state were performed on using the same instrument as described above using the mitochondrial targeted, genetically encoded sensor roGFP1. Cells were exited at 410 and 480 nm and emission was collected at 535 nm. Images were acquired every 2 s. The 410/480 fluorescence ratios were normalized by obtaining the maximum and minimum ratios (obtained after addition of 1 mM H₂O₂ and of 10 mM DTT, respectively; De Marchi et al, 2014). The fluorescence signals of roGFP1 were normalized to the averaged baseline signal (F/F ₀ or R/R ₀), obtained at the beginning of the measurements.

Measurements of NADH fluorescence

NAD(P)H intrinsic fluorescence in HEK293T cells was fluorescently monitored (360 nm excitation and 440 nm emission) in an inverted microscope (see Fluorescent Ca²⁺ and Na⁺ imaging), as previously reported (Nita et al, 2012) and calibrated by superfusing the cell with a Ringer's solution containing 2 μM FCCP, protonophore carbonyl cyanide 4‐(trifluoromethoxy)phenylhydrazone (Ascent Laboratories, Asc‐081) and 1 μM oligomycin (Merck Millipore, MBS495455) at the end of the experiments.

Confocal microscopy

HEK293T cells were grown onto six‐well chamber slides (Nunc, Rochester, NY, U.S.A.) and transfected with membrane‐targeted GCamp5 (0.2 μg of DNA/well). Forty‐eight hours following transfection, samples were examined by fluorescence confocal microscopy (LSM510 system, Carl Zeiss, Jena, Germany) using 480 nm (Ca²⁺ insensitive) excitation laser and 505‐nm long‐pass emission filter as previously described (Feldman et al, 2010). Briefly, baseline picture was captured followed by incubation with 2 μM of thapsigargin (TG) for 3 min. Then, cells were perfused by Ringer's solution containing 5 mM of CaCl₂ before starting recordings.

Föster resonance energy transfer measurements

These measurements were done as described previously (Navarro‐Borelly et al, 2008; Wang et al, 2014; Zhou et al, 2015; Cai et al, 2016). To determine resonance energy transfer (FRET) signals between Orai1‐CFP and STIM1‐YFP, we used the Leica DMI 6000B fluorescence microscope equipped with CFP (438Ex/483Em), YFP (500Ex/542Em), FRET (438Ex/542Em) filters. Fluorescence intensities captured with the CFP, YFP, and FRET filters were collected at room temperature using a 40× oil objective (N.A.1.35; Leica) and processed using Slidebook 6.0 software (Intelligent Imaging Innovations). Three‐channel corrected FRET was calculated using the following formula: F_C = I_DA− Fd/Dd * I_DD− Fa/Da * I_AA, where I_DD, I_AA, and I_DA represent the background‐subtracted CFP, YFP, and FRET images, respectively. And F_C represents the corrected energy transfer, Fd/Dd represents measured bleed‐through of CFP through the FRET filter (0.457), and Fa/Da is measured bleed‐through of YFP through the FRET filter (0.19). Although the CFP signal was constant in Orai1‐CFP stable cells, there was some variation in transiently expressed STIM1‐YFP levels. To minimize the variation, only cells with similar YFP intensity were analyzed. Also, the above FRETc values were normalized against acceptor fluorescence (FYFP) to generate normalized FRET signals, which compensated for these variations in YFP expression. Thus, normalized FRET (NFRET) = FRETc/I_AA. Average FRET measurements shown are typical of at least three independent analyses ± SEM. Same data pool was also calculated for EFRET with following formula (Zal & Gascoigne, 2004): E_app = Fc / (F_C + G*I_DD), where the F_C = I_DA− Fd/Dd * I_DD − Fa/Da * I_AA. G is the instrument‐specific constant and was determined as described (Zal & Gascoigne, 2004; Navarro‐Borelly et al, 2008). G value is 1.9 ± 0.1 (n = 32 cells). Only cells with similar YFP/CFP ratio were used for EFRET analysis.

Patch clamp electrophysiology

Conventional whole‐cell patch clamp recordings were carried out using an Axopatch 200B and Digidata 1440A (Molecular Devices, LLC, Sunnyvale, CA) as previously published (Gonzalez‐Cobos et al, 2013; Zhang et al, 2013, 2014). Pipettes were pulled from borosilicate glass capillaries (World Precision Instruments, Inc., Sarasota, FL) with a P‐97 flaming/brown micropipette puller (Sutter Instrument Company, Novato, CA) and polished with DMF1000 (World Precision Instruments). Resistances of filled glass pipettes were 2–4 MΩ. Series resistances were < 10 MΩ. The liquid‐junction potential offset was around 5 mV and was corrected. Only cells with tight seals (> 13 GΩ) were selected for break‐in. Cells were maintained at a 0 mV holding potential during experiments and subjected to voltage ramps from −140 to +100 mV lasting 250 ms every 2 s. All experiments were performed at room temperature (20–25°C). Clampfit 10.1 software was used for data analysis. The solutions employed for patch clamp recordings are as follows.

Bath solution

115 mM Na⁺‐methanesulfonate, 10 mM CsCl, 1.2 mM MgSO₄, 10 mM HEPES, 20 mM CaCl₂, and 10 mM glucose (pH was adjusted to 7.4 with NaOH). For Li⁺ based bath, we replaced Na⁺‐methanesulfonate with Li⁺‐methanesulfonate; for N‐methyl‐D‐glutamine (NMDG⁺)‐based bath, we replaced Na⁺‐methanesulfonate with NMDG⁺.

Pipette solution (20 mM BAPTA)

115 mM Cs‐methanesulfonate, 20 mM Cs‐1,2‐bis‐(2‐aminophenoxy)ethane‐N,N,N′,N′‐tetraacetic acid (Cs‐BAPTA), 8 mM MgCl₂, and 10 mM HEPES (pH adjusted to 7.2 with CsOH). For high buffer, we replaced 20 mM Cs‐BAPTA with 50 mM in pipette solution.

Pipette solution (0.1 mM EGTA)

135 mM Cs⁺‐methanesulfonate, 8 mM NaCl, 1.0 mM MgCl₂, 0.1 mM EGTA, and 10 mM HEPES, 0.03 mM IP₃ (pH adjusted to 7.2 with CsOH). For cocktail, we added 2 mM pyruvic acid, 2 mM malic acid, 1 mM NaH₂PO₄, 0.5 mM cAMP, 2 mM ATP, 0.5 mM GTP to the pipette solution.

To energize the mitochondria a pipette solution containing a cocktail (2 mM pyruvic acid, 2 mM malic acid, 1 mM NaH₂PO₄, 0.5 mM cAMP, 0.5 mM GTP, 0.5 mM MgCl₂) was used. Note, that the mitochondria also led to enhanced CRAC current only when activated in the presence of weak to moderate concentrations of EGTA (< 0.6 mM EGTA; Gilabert & Parekh, 2000; Parekh, 2008), presumably through enhanced mitochondrial Ca²⁺ buffering and reduced Ca²⁺‐dependent negative feedback on CRAC channels. When indicated, we performed CRAC recordings using a pipette solution containing 0.1 mM of the chelator EGTA combined with IP_3.

Divalent‐free (DVF) bath solution

155 mM Na‐methanesulfonate, 10 mM HEDTA, 1 mM EDTA, and 10 mM HEPES (pH 7.4, adjusted with NaOH).

Statistical analysis

The results of the experiments are presented as the means ± SEM of ≥ 4 experiments (n), using 5–30 cells in each condition. Statistical significance for all experiments was determined using either unpaired two‐tailed Student's t‐test or one‐way ANOVA test followed by Tukey post hoc analysis. Values of P < 0.05 were considered significant.

Author contributions

IS, MT, TBKN, XZ, and DLG wrote the manuscript; TBKN, XZ, AE, SR, YZ, RKM, and MG performed the experiments; JAS, NH, MH, DLG, MT, and IS provided experimental oversight. IS, MT, TBKN, XZ, and AE designed the experiments; IS, MT, TBKN, XZ, MH, and DLG interpreted the data.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Movie EV1

Movie EV2

Source Data for Appendix

Review Process File

Source Data for Figure 1A

The EMBO Journal (2017) 36: 797–815