Role of SERCA1 Truncated Isoform in the Proapoptotic Calcium Transfer from ER to Mitochondria during ER Stress

SUMMARY

Among the new players at the endoplasmic reticulum (ER)-mitochondria interface regulating interorganelle calcium signaling, those specifically involved during ER stress are not known at present. We report here that the truncated variant of the sarcoendoplasmic reticulum Ca²⁺-ATPase 1 (S1T) amplifies ER stress through the PERK-eIF2α-ATF4-CHOP pathway. S1T, which is localized in the ER-mitochondria microdomains, determines ER Ca²⁺ depletion due to increased Ca²⁺ leak, an increased number of ER-mitochondria contact sites, and inhibition of mitochondria movements. This leads to increased Ca²⁺ transfer to mitochondria in both resting and stimulated conditions and activation of the mitochondrial apoptotic pathway. Interestingly, S1T knockdown was shown to prevent ER stress, mitochondrial Ca²⁺ overload, and subsequent apoptosis. Thus, by bridging ER stress to apoptosis through increased ER-mitochondria Ca²⁺ transfer, S1T acts as an essential determinant of cellular fate.

INTRODUCTION

Calcium stored in the endoplasmic reticulum (ER) is crucial for posttranslational processing, folding, and export of proteins synthesized on ER-associated ribosomes, as well as for calcium (Ca²⁺) signaling (Berridge et al., 2000). Accumulation of unfolded proteins in the ER lumen and ER Ca²⁺ depletion induce ER stress, eliciting unfolded protein response (UPR), which is mediated by three pathways under the control of ER resident molecular “sentinels”: IRE1α (inositol-requiring transmembrane kinase and endonuclease 1α), PERK (protein kinase-like ER kinase), and ATF6 (activation of transcription factor 6) (Schroder and Kaufman, 2005). These sensors activate specific transcription factors (such as spliced XBP1, ATF6, and Activating Transcription Factor 4 [ATF4]), triggering the expression of UPR-related genes (such as C/EBP homologous protein [CHOP] and Glucose-Regulated Protein 78 [GRP78]) and determining the cellular fate. By reducing misfolded protein levels, UPR promotes cell survival, while unmitigated ER stress triggers apoptotic death (Boyce and Yuan, 2006). ER stress-dependent apoptosis was reported to be mediated by murine caspase-12/human caspase-4 and activation of proapoptotic members of the Bcl-2 family (i.e., Bax, Bak, Bim, Puma, etc.) and of CHOP (Hetz et al., 2006; Li et al., 2006; Puthalakath et al., 2007; Szegezdi et al., 2006).

Calcium has been studied for a long time as a regulator of apoptosis following its physiological or physiopathological release from the ER (Rizzuto et al., 2003). Recently, ER-mitochondria physical interactions have drawn attention as pivotal routes for Ca²⁺ transfer from the ER to the mitochondria, mediating interorganelle Ca²⁺ signaling and potentially leading to mitochondrial Ca²⁺ overload and apoptotic cell death (Pizzo and Pozzan, 2007). A few proteins have been involved in the ER-mitochondria contacts, including the ER membrane resident Ca²⁺-releasing channel, the inositol 1,4,5-triphosphate receptor (IP₃R); an outer mitochondrial membrane resident protein, the voltage-dependent anion channel (VDAC); the sorting protein PACS-2; and the molecular chaperones grp75 and Sigma-1R (Hayashi and Su, 2007; Pizzo and Pozzan, 2007; Simmen et al., 2005; Szabadkai et al., 2006).

In ER stress conditions, mitochondrial Ca²⁺ overload has been shown to be involved in mitochondrial apoptosis (Deniaud et al., 2008; Ferreiro et al., 2008). However, the molecular mechanism allowing increased Ca²⁺ transfer from the ER to the mitochondria under ER stress is not completely understood.

Active Ca²⁺ transport into the ER from the cytosol is ensured by sarcoendoplasmic reticulum calcium ATPases (SERCAs), creating a high [Ca²⁺] gradient between the cytosol and the ER lumen (~0.1 μM versus ~400 μM). SERCAs are encoded by three homologous genes (SERCA1-SERCA3) giving rise to 11 alternative spliced isoforms (Bobe et al., 2005; Vangheluwe et al., 2005; Zhang et al., 1995). We have previously described a SERCA1 variant (S1T) characterized by exon 11 splicing, encoding a C-terminally truncated protein unable to pump Ca²⁺ and which induces ER Ca²⁺ leak (Chami et al., 2001).

In this work, we provide evidence that S1T is induced upon ER stress and acts by increasing Ca²⁺ transfer from the ER to mitochondria. Thus, S1T is revealed as an ER stress protein directly involved in Ca²⁺-dependent mitochondrial apoptosis.

RESULTS

S1T Induction Is Involved in ER Stress Response

First, we investigated the role of S1T in ER stress induction. Unlike full-length SERCA1 isoform, overexpression of S1T native protein (S1T) as well as S1T-myc-His (S1T-myc-His) or GFP (S1T-GFP)-fused proteins in HeLa cells induced a set of ER stress proteins (Figures 1A and S1A and S1B [available online]).

Figure 1. S1T Is an ER Stress Protein.

(A) Detection of ER stress proteins (phosphorylated alpha-subunit of eukaryotic Initiation Factor 2 [P-eIF2α], GRP78, SERCA2b, Calreticulin [CRT], CHOP, GRP94, and ATF6 p50) in pcDNA3.1 (Control) or S1T-GFP-transfected HeLa cells for 40 hr. (B and C) Expression of endogenous S1T and of ER stress proteins in HeLa cells treated for 20 hr with 5 μg/ml BFA, 10 μg/ml TUN, 0.5 μM TG, 10 μM TBU, or 2 mM EGTA (B), and in hepatitis C virus “replicon” (HCV) and in control cell lines. (C) HCV core antibody was used as control.

(D and E) Detection of S1T and of ER stress proteins in HeLa cells treated or not treated with 0.5 μg/ml BFA for 20 hr (D) and in HCV “replicon” cells (E) transfected or not transfected with S1T_RNAi or Control_RNAi.

S1T, S1T-GFP, and SERCA2b were detected on microsomal fraction in (A) and (C) and in total extracts in (B), (D), and (E). All the other proteins were detected on total extracts. GFP antibody was used to detect S1T-GFP protein.

Analysis of endogenous S1T protein expression under various ER stress conditions revealed a parallel induction of S1T and of ER stress markers in HeLa cells upon ER Ca²⁺ depletion by thapsigargin (TG) or tert-butylhydroquinone (TBU), as well as after buffering extracellular Ca²⁺ by EGTA or blocking either ER to Golgi protein transport by brefeldin A (BFA) or N-linked glycosylation by tunicamycin (TUN) (Figure 1B). ER stress-induced S1T expression was observed in a series of human-derived transformed cell lines in a dose-dependent manner (Figures S1C and S1D). A similar extent of S1T and of ER stress markers expression was observed upon ER stress induction and S1T ectopic expression (Figure S1E). Interestingly, the full-length SERCA1 isoform was not induced upon BFA treatment (Figure S1F), thus demonstrating the specific induction of S1T isoform by ER stressors.

Importantly, induction of endogenous S1T was also observed in a physiopathological ER stress model, the hepatitis C virus “replicon,” i.e., HuH7 cell line stably expressing the HCV genome (Benali-Furet et al., 2005) (Figure 1C).

All in all, the above results suggest an amplifying role of S1T in ER stress signaling. Indeed, transfection of HeLa cells with an RNAi blocking the expression of endogenous S1T (S1T_RNAi) efficiently reduced the expression of P-eIF2α as well as that of GRP94 and CHOP in BFA-induced ER stress and that of GRP94 in the HCV replicon model (Figures 1D and 1E). Since S1T_RNAi transfection efficiency averaged 40% in HeLa cells (Figure S2A), we assumed that S1T is a key component of ER stress induction and amplification in these models. As controls we showed that (1) a control RNAi (Control_RNAi) did not interfere with BFA-induced expression of S1T and ER stress markers (Figure 1D), and (2) neither S1T_RNAi nor Control_RNAi modified the expression of proteins involved in Ca²⁺ homeostasis (SERCA2b, IP₃R, VDAC) and in ER stress (GRP78, P-eIF2α) in nontreated cells (Figure S2B).

S1T Is Induced by ER Stress through the PERK-eIF2α-ATF4 Pathway

To elucidate the mechanisms underlying S1T induction upon ER stress, kinetic studies of the expression of S1T, as well as of known ER stress proteins, were performed in BFA-treated HeLa cells. S1T expression was shown to occur in a biphasic manner: a rapid increase after 40 min of treatment following eIF2α phosphorylation and ATF4 expression (Figure 2A, left panel), a transient decline between 2 hr and 8 hr, and a pronounced reinduction at 10 hr preceded by eIF2α phosphorylation, ATF6 cleavage, and ATF4 and CHOP expression and accompanied by an increase of GRP78 and GRP94 protein level (Figure 2A, right panel). These data suggest that the initial induction of S1T expression might occur through ATF4 activation, while the late reinduction might involve the ATF4, ATF6, and CHOP transcription factors. Comparative promoter analysis of human SERCA1 gene and its orthologs revealed the presence of (1) an ATF4-responsive element (ATF/Cre) analogous to those previously described in different ER stress genes (GRP78 and GADD34) (Luo et al., 2003) (Figure 2B, upper panel) and (2) a CHOP-responsive element responsible for the induction of ER and mitochondrial stress genes (Ohoka et al., 2005)(Figure 2B, lower panel). However, no conserved ATF6-responsive element (ERSE) (Yoshida et al., 1998) was found, arguing against the direct implication of ATF6 in the induction of S1T upon ER stress.

Figure 2. S1T Is Induced through the PERK-eIF2α-ATF4 Pathway.

All the experiments were performed in HeLa cells treated or not treated with 0.5 μg/ml BFA.

(A) Induction of S1T and of ER stress proteins in cells treated for the indicated time points. Graphs: time course induction. Where indicated (asterisk), traces are referring to right scale.

(B) (Upper panel) Alignment of ATF/Cre-responsive element found in SERCA1, GRP78, and Growth Arrest and DNA Damage-inducible protein 34 (GADD34) genes. (Lower panel) Alignment of CHOP-responsive element found in SERCA1, carbonic anyhdrase VI (CA VI), tribbles-related protein 3 (TRB3), caseinolytic serine protease (ClpP), mitochondrial heat shock protein DnaJ (mtDnaJ), and chaperonins 60/10 (Cpn 60/10) genes. The base positions of the consensus are shaded and indicated 5′→3′. Positions are relative to transcription start site or where indicated (asterisk) to initiation codon.

(C) Electrophoresis mobility shift assay with ATF/Cre and CHOP probes derived from the SERCA1, the GADD34, and the Cpn60/10 promoters performed on nuclear extracts from cells left nontreated (NT) or treated with BFA for 40 min (for ATF/Cre probes) or 20 hr (for CHOP probes). The shifted ATF4 and CHOP complexes are indicated by arrows a, a′, and a″ and b, b′, b″, respectively.

(D) Oligonucleotide pull-down assay with biotin-labeled probes used in (C), and western blot analyses of DNA-protein complexes in nuclear extracts (input) and on pull-down complexes (PD).

(E) Detection of PERK, P-eIF2α, and S1T in HeLa cells treated or not treated with 0.5 μg/ml BFA for 1 hr transfected or not transfected with Control_RNAi or PERK_RNAi. In (A) and (E), total extracts were used for western blot analyses.

We then explored the functional activity of the identified responsive elements by using electrophoretic mobility shift assay (EMSA). An increase of DNA/protein complexes was observed with human SERCA1 ATF/Cre probe upon 40 min of BFA treatment, and with human CHOP probe upon 20 hr of BFA treatment. Similar DNA/protein complexes were observed with the murine GADD34 ATF/Cre and the human Cpn60/10 CHOP probes used as positive controls for ATF4 and CHOP binding, respectively (Figure 2C).

We then demonstrated that ATF4 and CHOP transcription factors belong to the protein complexes respectively bound to SERCA1 ATF/Cre and CHOP elements upon BFA treatment by using DNA-protein pull-down assay followed by western blot analysis (Figure 2D). As controls, we showed the absence of both peroxisome proliferator-activated receptor γ coactivator 1 (PGC1) and P53 transcription factors in the protein complexes bound to both probes (Figure 2D).

Finally, S1T induction was shown to occur through the PERK-eIF2α-ATF4 pathway, since RNAi-driven knockdown of PERK (PERK_RNAi) significantly reduced eIF2α phosphorylation and abrogated S1T induction upon BFA treatment (Figure 2E, upper panel). Control_RNAi transfection did not modify PERK-mediated eIF2α phosphorylation in BFA-treated cells (Figure 2E, lower panel).

S1T Expression Links ER Stress to Deregulation of ER Ca²⁺ Homeostasis

We have previously shown that S1T triggers Ca²⁺ leak by forming homodimers in the ER membrane and depletes the ER Ca²⁺ pool (Chami et al., 2001). Here we showed, by using cytosolic and ER-targeted recombinant aequorin probes, that similarly to S1T, ER stressors (20 hr of treatment) induced a reduction of steady-state ER [Ca²⁺] ([Ca²⁺]_er) and agonist-evoked cytosolic Ca²⁺ peak ([Ca²⁺]_cyt) (Figures 3A and S3A; Tables S1 and S2). We deliberately used a concentration of SERCA inhibitors that does not provide complete inhibition of the SERCA, to study the effect of S1T on ER Ca²⁺ content and leak. Note also that the decreased agonist-evoked cytosolic Ca²⁺ signal observed in S1T-transfected cells is largely due to the decreased [Ca²⁺]_er, since no modification of extracellular Ca²⁺ influx was observed (Figure S3B).

Figure 3. S1T Mediates Ca²⁺ Deregulation Related to ER Stress.

In (A), (C), and (D), [Ca²⁺]_er was measured using erAEQ probe 40 hr posttransfection. (A and C) Steady-state [Ca²⁺]_er (presented as a percentage of control considered as 100%) in nontreated HeLa cells (Control), 0.5 μg/ml BFA, 1 μg/ml TUN, 2.5 nM TG-treated cells for 20 hr, in S1T-pcDNA3.1-transfected cells for 40 hr (S1T) (A) (Student's t test, *p < 0.05, **p < 0.005 versus Control) and in nontreated HeLa cells (Control), in BFA-treated cells for 20 min or 2 hr (Student's t test, ns = nonsignificant, *p < 0.005 versus Control) (C). (B) Expression of S1T in HeLa cells treated with 0.5 μg/ml BFA for the indicated time points. (D) Representative traces of [Ca²⁺]_er in nontreated HeLa cells (Control) and in 0.5 μg/ml BFA, 10 μM TBU, or 2.5 nM TG-treated cells for 20 hr transfected or not transfected with S1T_RNAi. The effect of S1T_RNAi was recorded in 40% of aequorintransfected cells and thus underestimated (see the Experimental Procedures). Data in (A), (C), and (D) are represented as mean ± SEM. Numbers of experiments are indicated in Table S1.

Interestingly, BFA treatment for 2 hr induced the expression of the endogenous S1T and a parallel decrease of [Ca²⁺]_er, while neither S1T induction nor [Ca²⁺]_er decrease were observed after 20 min of treatment (Figures 3B and 3C). These results suggest that S1T is likely to be involved in ER Ca²⁺ depletion upon ER stress.

Indeed, analyses of ER Ca²⁺ homeostasis in cells subjected to different ER stressors after knockdown of S1T expression revealed that BFA-, TBU-, and TG-induced reduction of ER Ca²⁺ steady state and subsequent agonist-evoked cytosolic signal were significantly reverted by S1T_RNAi (Figures 3D and S3C; Tables S1 and S2). Decreased [Ca²⁺]_er was demonstrated to be likely due to increased ER Ca²⁺ leakage (Figures S4A and S4B). In addition, ER Ca²⁺ pool was progressively depleted upon S1T expression (Table S1), indicating that S1T-dependent ER Ca²⁺ leak is persistent over time. Control tests showed that (1) S1T_RNAi but not Control_RNAi transfection reversed the reduction of [Ca²⁺]_er induced by S1T overexpression, and (2) S1T_RNAi did not modify the increased [Ca²⁺]_er induced by SERCA2b overexpression (Figure S2C; Table S1).

All in all, these data demonstrate that ER Ca²⁺ depletion upon ER stress is mediated by S1T.

S1T Is Localized to Mitochondria-Associated Membranes and Determines Mitochondrial Immobilization and Docking at the ER Surface

We then showed by using western blot and subcellular fractionation that S1T (endogenous and upon ectopic expression) is enriched both in the microsomal (Mic) and the mitochondria-associated membrane (MAM) fractions (Figure 4A). Accordingly, high-resolution imaging demonstrated that S1T colocalized with mitochondria (Figure 4B). This finding suggests that S1T-dependent ER Ca²⁺ leak may preferentially occur in the ER-mitochondria microdomains.

Figure 4. S1T Is Localized to Mitochondria-Associated Membranes and Determines Mitochondrial Immobilization and Docking at the ER Surface.

(A) Subcellular fractionation. Tot, total; mito, crude mitochondria; Mic, microsomes; MAM, mitochondrial-associated membranes. All endogenous proteins except S1T-GFP were detected in nontransfected HeLa cells. S1T and GFP antibodies were used to detect endogenous S1T and S1T-GFP 40 hr posttransfection, respectively. SERCA2b, VDAC, and cytochrome c oxydase (COX) were used as microsomal, MAM, and mitochondrial fractions markers, respectively.

(B) Quantification of colocalization of S1T with mitochondria (labeled with MitoTracker Deep Red) in living HeLa cells transfected with S1T-GFP (S1T-GFP, n = 20) or with erGFP probe (Control, n = 10). Representative magnified image shows S1T-GFP signal in green, mitochondria in red, and colocalization in white. Student's t test, *p < 0.0005 versus Control.

(C) Mitochondrial movements analyzes in HeLa cells treated with 0.5 μg/ml BFA (3 hr) transfected or not transfected with S1T_RNAi-fluorescein or transfected with mtGFP or S1T-GFP. In S1T_RNAi conditions, quantification was performed exclusively in S1T_RNAi-fluorescein-positive cells. Mitochondrial movements was evaluated before and after 100 μM histamine stimulation (used as control of ER Ca²⁺ release-related mitochondrial immobilization) or upon BAPTA-AM (5 μM, 20 hr) application. Student's t test, ns^‡ = nonsignificant, ^‡p < 0.005 versus nontreated cells, ^Δp < 0.005 versus BFA-treated cells, ns* = nonsignificant, *p < 0.05, and **p < 0.0005 versus mtGFP-transfected cells, ^#p < 0.0005 versus S1T-transfected cells. Numbers of experiments are indicated in Table S3.

(D) Graph: quantitative analysis of the colocalization of ER and mitochondria (as percentage of total mitochondrial volume) in HeLa cells cotransfected with erGFP and S1T-pcDNA3.1 (S1T 8 hr, n = 16; S1T 40 hr, n = 11) or SERCA1-pcDNA3.1 (SERCA1, n = 9) or in cells transfected with erGFP left untreated (Control, n = 30) or treated with 5 μM BAPTA-AM for 3 hr (Control + BAPTA-AM, n = 17) or with 0.5 μg/ml BFA (BFA, n = 19), or with BFA and BAPTA-AM (BFA + BAPTA-AM, n = 8). Representative images of ER (erGFP, green), mitochondria (MitoTracker Deep Red, red) and magnified overlay images showing colocalization between ER and mitochondria (white). Student's t test, ns = nonsignificant, *p < 0.005, **p < 0.0005 versus Control, and ^#p < 0.05 versus BFA.

(E) Representative fields of electron micrographs and ER-mitochondria distance (graph) in HeLa cells transfected for 40 hr with pcDNA3.1, S1T-pcDNA3.1, or S1T-GFP (Control, n = 31; S1T, n = 46; S1T-GFP, n = 44). M, mitochondria; ER, endoplasmic reticulum. Arrows indicate proximity of ER to mitochondria. Student's t test, *p < 0.0005 versus Control.

Data in (B)-(E) are represented as mean ± SEM.

It was recently shown that elevated Ca²⁺ signals in microdomains between mitochondria and ER Ca²⁺ release sites immobilize mitochondria close to the ER, thus increasing ER-mitochondria physical association (Brough et al., 2005; Yi et al., 2004). We then applied time-lapse imaging of the mitochondrial network and 3D-digital microscopy to quantify mitochondrial movements and ER-mitochondria colocalization, respectively. BFA treatment (3 hr) as well as S1T overexpression (40 hr) were shown to induce inhibition of mitochondrial movements (Figure 4C; Table S3) and to increase the extent of association between ER and mitochondria (Figure 4D). These effects were reverted in BFA-treated cells by S1T_RNAi (Figure 4C; Table S3) and upon Ca²⁺ chelation by BAPTA-AM in both BFA-treated and S1T-overexpressing cells (Figures 4C and 4D).

Interestingly, docking of mitochondria to the ER was shown to occur in S1T but not in SERCA1-expressing cells and to increase progressively upon S1T expression (Figure 4D).

Finally, reduction of ER-mitochondria distance was also demonstrated by transmission electron microscopy in S1T and S1T-GFP-transfected cells for 40 hr as compared to control cells (Figure 4E).

These data allowed us to identify S1T as a protein of the ER-mitochondria contact sites that controls mitochondrial movement and docking to the ER in a Ca²⁺-dependent manner.

ER Stress Increases Ca²⁺ Transfer to Mitochondria through S1T Induction

Since MAM regulate Ca²⁺ signaling between ER and mitochondria (Hajnoczky et al., 2000; Rizzuto et al., 1993), S1T may play a role at MAM in regulating Ca²⁺ signaling between these two organelles.

Analysis of basal mitochondrial Ca²⁺ load by using X-rhod-1, AM dye, showed that S1T overexpression (8 hr) as well as BFA treatment (20 hr) determined a huge increase of basal mitochondrial Ca²⁺ level, which was corrected by S1T_RNAi transfection in BFA-treated cells (Figure 5A; Table S4). Furthermore, no modification of resting cytosolic Ca²⁺ level (as measured with fluo-4, AM dye) was observed 8 hr post S1T transfection (Table S2), arguing that S1T-dependent Ca²⁺ leak occurs preferentially in the ER-mitochondria contact sites.

Figure 5. S1T Overexpression Enhances Basal and Agonist-Evoked Mitochondrial Ca²⁺ Load, Leading to Mitochondrial Structure Alteration.

(A) Mitochondrial basal Ca²⁺ level as evaluated with X-rhod-1, AM dye in HeLa cells transfected with S1T-pcDNA3.1 (S1T) for 8 hr and 40 hr or in cells treated with 0.5 μg/ml BFA for 20 hr and transfected or not transfected with S1T_RNAi. Calibrated values of basal mitochondrial Ca²⁺ load are presented as percentages of controls (pcDNA3.1-transfected or nontreated cells) considered as 100%. The effect of S1T_RNAi was recorded in S1T_RNAi-positive cells (fluorescein-positive cells). Values of basal mitochondrial Ca²⁺ level are provided in μM in Table S4 and were determined as described in the Supplemental Experimental Procedures. Student's t test, *p < 0.01, **p < 0.0001 versus Control, ^#p < 0.005 versus BFA.

(B) Mitochondrial agonist-evoked Ca²⁺ uptake (peak value) after 100 μM histamine application, as measured with mitAEQmut probe, in HeLa cells transfected or treated as in (A), and presented as percentage of controls considered as 100%. The effect of S1T_RNAi was recorded in 40% of aequorin-transfected cells and thus underestimated. Student's t test, *p < 0.005, **p < 0.0001 versus Control, ^#p < 0.005 versus BFA.

(C) Percentage of cells with tubular mitochondrial network and those with altered mitochondrial network in HeLa cells (see representative images) cotransfected with mitBFP and pcDNA3.1 (Control, n = 58) or S1T-GFP (S1T 8 hr, n = 83; S1T 40 hr, n = 105) or SERCA1-GFP (SERCA1, n = 50) and in Control and S1T cells (40 hr) treated with 5 μM BAPTA-AM for 20 hr posttransfection (n = 64 and n = 74, respectively).

(D) Relative fluorescence intensity changes (dF/F) following 100 μM histamine stimulation in HeLa cells 40 hr after cotransfection of mitochondrial pericam probe and pcDNA3.1 (Control) or S1T-pcDNA3.1 [S1T(T) with tubular-like mitochondrial network and S1T(A) with altered mitochondrial network]. Table: representative images of the mitochondrial network of S1T (T) and (A) cells, percentage of each population, and maximal rate of mitochondrial Ca²⁺ uptake evaluated from the ascending phase of dF/F kinetics versus Control considered as 100%. The Control trace shows the mean ± SEM of cells with tubular versus altered mitochondrial network (Table S5). Data in (A)-(D) are represented as mean ± SEM. For (A), (B), and (D), numbers of experiments are indicated in Tables S4 and S5.

These microdomains constitute a preferential pathway leading to mitochondrial Ca²⁺ accumulation upon Ca²⁺ release by IP₃ receptor (Rizzuto et al., 1993).

Indeed, S1T transfection (8 hr) and BFA treatment (20 hr) led to a vast increase of agonist-evoked mitochondrial Ca²⁺ accumulation ([Ca²⁺]_mt) as measured with mitAEQ probe, which was corrected by S1T_RNAi transfection in BFA-treated cells (Figure 5B; Table S4).

No modification of agonist-evoked [Ca²⁺]_cyt was observed in S1T-transfected cells 8 hr posttransfection (Table S2), indicating that S1T-dependent increased ER-mitochondria apposition enhances IP₃-mediated Ca²⁺ transfer to mitochondria, leading to increased mitochondrial signals in stimulated cells.

Long-term expression of S1T (40 hr post-S1T-transfection) determined a reduction of basal and agonist-evoked mitochondrial Ca²⁺ load (Figures 5A and 5B; Table S4). Under these conditions, we observed a significant alteration of mitochondrial network, which was reverted by BAPTA-AM treatment (Figure 5C), while mitochondrial structure was mostly tubular-like (normal) 8 hr post-S1T-transfection (Figure 5C).

Accordingly, simultaneous evaluation of Ca²⁺ uptake rate and mitochondrial structure in single cells by using mitochondria-targeted pericam probe and digital imaging, performed 40 hr after S1T transfection, allowed us to identify two different cell populations: (1) cells with tubular-like mitochondrial network and a higher Ca²⁺ uptake rate and amplitude [S1T(T), 20%], and (2) cells with an altered mitochondrial network and a lower Ca²⁺ uptake rate and amplitude [S1T(A), 80%] (Figure 5D; Table S5).

These data suggest that mitochondrial structure alteration and decreased Ca²⁺ content follow mitochondria Ca²⁺ overload.

S1T Overexpression and ER Stress Induce Ca²⁺-Dependent Apoptosis through the Mitochondrial Gateway

As mitochondrial Ca²⁺ overload is one of the major factors triggering mitochondrial apoptosis (Kroemer et al., 2007), we lastly investigated the involvement of S1T protein in the activation of mitochondrial apoptotic pathway upon ER stress induction. We assumed that mitochondrial apoptosis is a late effect of S1T expression and thus performed experiments 40 hr after transfection of S1T. First, by using digital imaging and 3D reconstitution, we showed that mitochondrial volume (a parameter of mitochondrial swelling), but not ER volume, was significantly increased in S1T-overexpressing cells as compared to control and SERCA1-overexpressing cells (Figure 6A). Mitochondrial swelling induced by S1T was further confirmed by transmission electron microscopy as demonstrated by increased mitochondrial cross-section area (Figure 6B). A decrease of mitochondrial potential (ΔΨ_m) (TMRM intensity) was also noticed in S1T-GFP-transfected cells as compared to control and SERCA1-GFP-overexpressing cells (Figure 6C). Moreover, the loss of ΔΨ_m was accompanied by the mitochondrial release of cytochrome c in S1T-transfected cells as compared to control cells (low and diffuse versus high and compartmentalized cytochrome c signal) (Figure 6D). We then demonstrated that S1T overexpression induced two specific hallmarks of apoptosis, the cleavage of caspase-3 and poly (ADP-ribose) polymerase (PARP), which were blocked by the application of the pan-caspase inhibitor ZVAD-fmk (Figure 6E). Lastly, S1T-related cell death was significantly abolished by ZVAD-fmk, BAPTA-AM, and the inhibitor of mitochondrial permeability transition pore cyclosporin A (CsA). Interestingly, no cumulative effect was observed with ZVAD-fmk and BAPTA-AM cotreatment (Figure 6F). No effect on the viability of SERCA1-overexpressing cells was noted with the same treatments (Figure 6F). These findings indicate that S1T overexpression leads to mitochondria-mediated apoptosis bearing all the principal features of the process and that this is driven by both mitochondrial Ca²⁺ overload and caspase activation.

Figure 6. S1T Alters Mitochondrial Structure and Function.

All these analyses were performed in HeLa cells 40 hr posttransfection. (A) Representative high-magnification 3D-reconstructed images of cells transfected as described in Figure 4D. Mitochondrial and ER volume were evaluated (as compared to Control cells considered as 100%) (Control, n = 18; S1T, n = 18; SERCA1, n = 10). (B) Representative fields of electron micrographs of cells transfected with pcDNA3.1 (Control, n = 109) or with S1T-pcDNA3.1 (S1T, n = 110). M, mitochondria. Graph: mean of mitochondria cross-sectional area. (C) Mitochondrial potential change was evaluated after transfection of GFP empty vector (Control, n = 326), S1T-GFP (S1T, n = 155), or SERCA1-GFP (SERCA1, n = 201). (D) Representative images and graph showing cytochrome c fluorescence intensity in cells transfected with pcDNA3.1 (Control, n = 262) or S1T-GFP (S1T, n = 189). (E) Western blot analyses of cells transfected with pcDNA3.1 or with S1T-pcDNA3.1 and treated or not treated with 100 μM ZVAD-fmk for 20 hr. As positive control, cells were treated with 1 μM staurosporine (STS) for 2 hr. (F) Cellular viability in S1T-GFP or SERCA1-GFP-overexpressing cells treated (20 hr posttransfection) or not treated with 12 μg/ml cyclosporine A (CsA), 5 μM BAPTA-AM, and 100 μM ZVAD-fmk, presented as percentages of GFP-fluorescent cells among the whole cell population (determined by Hoechst staining) versus the nontreated transfected cells considered as 100%. Student's t test, ns = nonsignificant, *p < 0.05, and **p < 0.0005 versus Control. Data are represented as mean ± SEM.

If S1T induction took part in ER stress signaling, we would expect that stress conditions evoke the same apoptotic pathway. Indeed, BFA treatment for 20 hr, apart from inducing increased mitochondrial Ca²⁺ accumulation (Figures 5A and 5B), also led to apoptotic changes analogous to S1T, i.e., mitochondrial network alterations and swelling (Figures 7A and 7B), cytochrome c release (Figure 7C), translocation of proapoptotic factor Bax to the mitochondria (Figure 7D), and cleavage of caspase-3 and PARP (Figure 7E). Importantly, BFA-induced changes were reverted by S1T_RNAi or by BAPTA-AM (Figures 7A-7E). Finally, we demonstrated that silencing of S1T significantly reduced apoptosis mediated by ER stressors (Figure 7F). As control we showed that these drugs used in the same conditions did not induced necrotic cell death (Figure 7F, inset). Overall, these data pointed out the key role of S1T in the control of ER stress-related apoptosis.

Figure 7. S1T Knockdown Reduces BFA-Induced Apoptosis.

All the experiments were performed in HeLa cells left untreated or treated with 0.5 μg/ml BFA for 20 hr and transfected or not transfected with S1T_RNAi. (A) Percentage of cells with tubular or altered mitochondrial network in Control (n = 428), BFA (n = 293), and BFA + S1T_RNAi (n = 197) cells. (B) Mitochondrial volume recorded as in Figure 6A in Control (n = 19), in BFA (n = 24), and in BFA cells cotreated with 5 μM BAPTA-AM (BFA + BAPTAAM; n = 12). (C) Quantification of cytochrome c fluorescence intensity evaluated as in Figure 6D in Control (n = 262), in BFA (n = 335), and in BFA + S1T_RNAi (n = 242) cells. (D) Percentage of cells with activated Bax (intense mitochondrial staining in representative images) in Control (n = 747), BFA (n = 596), and BFA + S1T_RNAi cells (n = 300). (E) Western blot analyses performed as in Figure 6E. (F) Percentage of apoptotic bodies in nontreated cells (Control) or treated for 20 hr with 0.5 μg/ml BFA, 5 μg/ml TUN, 0.5 μM TG, and 10 μM TBU after being transfected or not transfected with S1T_RNAi. Inset: Percentage of necrotic cells (propidium-iodide-positive cells) recorded in the same conditions. (A, E, and F) Effect of S1T_RNAi is recorded in total cell population and thus underestimated. (C and D) Quantification was performed exclusively in S1T_RNAi-fluorescein-positive cells. (B-D and F) Student's t test, ns = nonsignificant, *p < 0.0001 versus Control, ^#p < 0.0001 versus BFA, TUN, TG, or TBU. Data are represented as mean ± SEM.

DISCUSSION

This work provides evidence that the SERCA1 truncated variant (S1T) promotes cell death through ER stress amplification, CHOP activation, ER-mitochondria Ca²⁺ transfer, and Ca^2+-driven sensitization of the mitochondrial apoptotic pathway (Figure S5).

Throughout this study we disclosed the mechanisms underlying the induction of S1T under ER stress conditions. Kinetic changes of the expression profiles of known ER stress signaling components and S1T revealed that S1T is induced early through the PERK-eIF2α-ATF4 pathway. Full-length SERCA1 was not induced under ER stress, suggesting that this regulation occurred in a spliced variant-specific manner. Interestingly, increase of SERCA2b isoform expression and activity has already been reported in cells subjected to drug-induced ER stress (Caspersen et al., 2000; Hojmann Larsen et al., 2001). Since S1T and SERCA2b have opposite effects on [Ca²⁺]_er, the balanced expression of both Ca²⁺ transporters is expected to be finely tuned during ER stress to keep cell fate under control. Accordingly, induction of SERCA2b and calreticulin as well as GRP78 is likely to enable protein translation during chronic ER stress through sufficient ER Ca²⁺ storing capacity and reduction of PERK-dependent translation inhibition, respectively (Brostrom and Brostrom, 2003).

Furthermore, our data demonstrated that S1T achieves a positive feedback on the PERK-eIF2α-ATF4 pathway, since eIF2α phosphorylation and CHOP activation upon BFA treatment were drastically reduced by S1T silencing. Recently, it was shown that excessive phosphorylation of eIF2α potentiates fatty-acid-induced ER stress and subsequent apoptosis in pancreatic b cells (Cnop et al., 2007) and that ATF4-mediated CHOP induction leads to cell-cycle arrest and/or apoptosis (Oyadomari and Mori, 2004). Finally, BH3-only Bcl-2 family members (Bim, Puma, and Noxa) have been recently identified as ER stress-related apoptotic effectors downstream of CHOP and P53 activation (Li et al., 2006; Puthalakath et al., 2007). It appears that, in ER stress signaling, S1T acts early as a key component of the proapoptotic cascade, downstream of ATF4 and in concert with CHOP, to control the lethal outcome of ER stressed cells.

We show here that S1T expression leads to a progressive depletion of ER Ca²⁺ pool. Since S1T is localized within the ER-mitochondria contact sites and determines Ca²⁺ leak, we assume that S1T-dependent Ca²⁺ leak occurs within these sites channeling leaky Ca²⁺ toward the mitochondria. As a consequence, basal mitochondrial Ca²⁺ level increases upon S1T expression.

It has been previously reported that Ca²⁺ regulates both mitochondrial motility and association between mitochondria and ER (Brough et al., 2005; Wang et al., 2000; Yi et al., 2004). Accordingly, we showed that S1T induces mitochondria immobilization in a Ca²⁺-dependent manner and leads to a progressive increase of ER-mitochondria connections. Moreover, S1T-expressing cells displayed an increased agonist-evoked mitochondrial Ca²⁺ signals. These results are in agreement with recent studies demonstrating that ER-mitochondria contacts are increased under ER stress conditions (Csordas et al., 2006; Tiwari et al., 2006) and that IP₃-mediated Ca²⁺ transfer toward mitochondria is enhanced by stable interaction between these organelles (Filippin et al., 2003) and upon artificial linkage between the ER and the outer mitochondrial membranes (Csordas et al., 2006).

Recent studies highlighted that chaperone machinery at both mitochondria and ER sides (grp75 and Sigma-1R, respectively) regulate IP₃R-mediated mitochondrial Ca²⁺ signaling through the stabilization of IP₃R on ER-mitochondria contact sites (Hayashi and Su, 2007; Szabadkai et al., 2006). Further studies would be deserved in order to determine if S1T interacts with chaperones or Ca²⁺ proteins located within ER-mitochondria contacts.

Artificial increase of ER-mitochondria contacts led to Ca^2+-induced cell death in rat peritoneal mast cells (Csordas et al., 2006), while depletion of PACS-2, maintaining ER-mitochondria contacts, conferred resistance to apoptosis (Simmen et al., 2005). As S1T expression persists during ER stress, mitochondrial Ca²⁺ signals decrease as a consequence of mitochondrial Ca²⁺ overload-induced mitochondrial permeabilization and swelling (Kroemer et al., 2007).

Finally, it is noteworthy that both proapoptotic S1T and antiapoptotic Bcl-2 proteins induce ER Ca²⁺ leak. Unlike S1T, Bcl-2, which has never been shown to be located at MAM, was found to reduce mitochondrial Ca²⁺ uptake (Pinton et al., 2000) and to inhibit mitochondrial Ca²⁺ waves preceding apoptosis (Magnelli et al., 1994).

Overall, we identified S1T as a key molecular player in ER stress-related apoptosis. Since ER stress has been identified as a common physiopathological process in a number of major diseases, our finding are expected to promote future research avenues with potential clinical impact.

EXPERIMENTAL PROCEDURES

Additional information regarding cell culture, plasmids, transient transfection, aequorin measurements, protein preparation, western blot, imaging, electron microscopy, search for putative sites for transcriptions factors, and cellular viability and death analyses can be found in the Supplemental Data.

Inhibition of PERK and S1T Protein Expression

PERK/EIF2AK3 (Hs_EIF2AK3_5 HP Validated siRNA) (PERK_RNAi) was purchased from QIAGEN SA. HeLa cells were exposed to RNAi for 48 hr using HiPerFect Transfection Reagent (QIAGEN SA) according to the manufacturer's instructions. GripNA oligo (peptide nucleic acid, Active Motif) was used for S1T knockdown experiments. S1T_RNAi: 5′- AGCTAGAGCGGCCTCCAT-3′. Control_RNAi (5′-ATGGAGGCCGCTCATGCT-3⁰) was used as a negative control. Tagged S1T_RNAi (Fluorescein-S1T_RNAi) was used in imaging experiments to identify RNAi-transfected cells. Chariot II (Active Motif) was used to transfect the gripNAs, according to the manufacturer's instructions. Transfection efficiency of S1T_RNAi was estimated to 40%. For aequorin measurements, cells were transfected first with aequorin and then with S1T_RNAi. By using this protocol, S1T_RNAi was shown to be efficiently transfected (~40%) in aequorin-expressing cells (Figure S2D).

Aequorin Measurements

Cytosolic, mitochondrial, and ER Ca²⁺ analyses were performed by using aequorin-targeted probes. For details see the Supplemental Experimental Procedures.

Imaging Analyses

Morphological, Ca²⁺ imaging, and mitochondrial potential analyses were performed as previously described (Szabadkai et al., 2004) (Supplemental Experimental Procedures). For measurement of mitochondrial movements, images were acquired on a Zeiss LSM 510 confocal microscope (Carl Zeiss) after loading cells with 10 nM Mitotracker Red CMX Ros in KRB/Ca²⁺ at 37 C for 15 min. Time-series images were taken with a time interval of 10 s between each image. Images were 2D deconvolved, median filtered, thresholded, and clipped to 8 bit binary images using Metamorph (Universal Imaging) software. Using the stack-T-functions/Delta-F-down plugin of the WCIF ImageJ software (http://www.uhnres.utoronto.ca/facilities/wcif/), pixels in each frame were subtracted from the next frame. The resulting images were quantified by measuring the total area of object on the binary and ΔF images by the Integrated Morphometry Analysis function of the Metamorph software. Data were normalized as the ratio of ΔF area values over the total area of the original binary images for each cell. This parameter is referred to as the mitochondrial movement index (MI). Binarization and normalization were applied to avoid artifacts eventually arising from changes of intensity and focal plane.

Protein Preparation and Western Blot Analyses

Total and microsomal protein preparation and subcellular fractionation were performed as previously described (Chami et al., 2001; Vance, 1990) (Supplemental Experimental Procedures). Protein loading was normalized as compared to actin signal. Band intensity was quantified versus a control sample (arbitrarily assessed as 1), using ImageJ software.

EMSA and Oligonucleotide Pull-Down Assay

Nuclear extracts were purified as already described (Ma et al., 2002). EMSA was performed by using the LightShift Chemiluminescent EMSA Kit (Pierce). Nuclear extract (10 μg) was incubated with 20 fmol of biotin-labeled probe (Supplemental Experimental Procedures) in a nondenaturing buffer containing 1 μg/μl poly dI•dC and 1 μg/μl poly dA•dT. Reaction mix was electrophoresed in a 5% polyacrylamide-TBE gel and transferred to a nylon membrane, and biotin signal was detected. Specificity of the reaction was assessed using 200x excess of nonlabeled probe. The oligonucleotide pull-down assay was achieved with 250 μg of nuclear extracts and 1 μg of biotin-labeled probe. DNA-protein complexes were captured on immobilized neutravidin beads (Pierce), then washed and separated by SDS-PAGE. Western blot analyses were performed on nuclear extracts and on pull-down complexes.

Statistical Analyses

Results are reported from at least three different experiments. Statistical analyses were performed using a Student's t test. A p value ≤0.05 was considered to be significant. All data are reported as means ± standard error of the mean (SEM).