Proapoptotic BAX and BAK Modulate the Unfolded Protein Response by a Direct Interaction with IRE1α

See allHide authors and affiliations

Science  28 Apr 2006:
Vol. 312, Issue 5773, pp. 572-576
DOI: 10.1126/science.1123480


Accumulation of misfolded protein in the endoplasmic reticulum (ER) triggers an adaptive stress response—termed the unfolded protein response (UPR)—mediated by the ER transmembrane protein kinase and endoribonuclease inositol-requiring enzyme–1α (IRE1α). We investigated UPR signaling events in mice in the absence of the proapoptotic BCL-2 family members BAX and BAK [double knockout (DKO)]. DKO mice responded abnormally to tunicamycin-induced ER stress in the liver, with extensive tissue damage and decreased expression of the IRE1 substrate X-box–binding protein 1 and its target genes. ER-stressed DKO cells showed deficient IRE1α signaling. BAX and BAK formed a protein complex with the cytosolic domain of IRE1α that was essential for IRE1α activation. Thus, BAX and BAK function at the ER membrane to activate IRE1α signaling and to provide a physical link between members of the core apoptotic pathway and the UPR.

Cell viability depends on the functional and structural integrity of intracellular organelles. Multidomain proapoptotic BAX and BAK proteins function in concert as essential gateways to intrinsic cell death pathways operating at mitochondria (1). Several anti- and proapoptotic BCL-2 family members also localize to the ER and modulate steady-state calcium homeostasis (24). In higher eukaryotes, ER stress stimulates three distinct UPR signaling pathways through sensors that include IRE1α (also described as inositol-requiring transmembrane kinase and endonuclease 1α), PERK (protein kinase–like ER kinase), and ATF6 (activation of transcription factor 6) (5, 6). IRE1α is a serine-threonine protein kinase and endoribonuclease that, on activation, initiates the unconventional splicing of the mRNA encoding X-box–binding protein 1 (XBP-1) (79). Spliced XBP-1 is a potent transcriptional activator that increases expression of a subset of UPR-related genes (10). The cytosolic domain of activated IRE1α binds the adaptor protein TRAF2 [tumor necrosis factor (TNF)–associated factor 2], and triggers the activation of the c-Jun N-terminal kinase (JNK) signaling pathway (11, 12). Activated PERK directly phosphorylates and inhibits the translation initiation factor eIF2α (thus decreasing protein loading into the ER) and induces expression of the transcription factor ATF4, which increases expression of certain UPR genes such as Chop or GADD153 and Grp78 or BiP (5, 7). BiP is a chaperone that maintains PERK and IRE1α in an inactive state. However, in cells undergoing ER stress, BiP preferentially binds to misfolded proteins, thereby releasing the stress sensors to undergo activation by homodimerization and autophosphorylation (13, 14).

Double knockout (DKO) cells from BAX-BAK–deficient mice are resistant to proapoptotic agents that induce the UPR (1) and also show a defect in steady-state ER calcium homeostasis under nonapoptotic conditions (3). A validated in vivo model for ER stress uses intraperitoneal injection of tunicamycin (Tm, an inhibitor of N-linked glycosylation) (1517). This treatment triggers a stress response in the liver and kidney that causes extensive cell death in these organs after several days of treatment. Most DKO mice (more than 90%) die during embryogenesis (18), as a result of developmental defects. We generated a conditional BAX-BAK DKO model in which a bax allele flanked with LoxP sites was targeted in bak-null embryonic stem cells (19). To achieve inducible deletion of Bax in adulthood, MxCre+baxfl/–bak–/– and control MxCre+bax+/–bak–/– mice were treated with poly(IC). Mice were subsequently injected with Tm (1 μg/g body weight) and killed 6, 12, or 24 hours after injection. Similar expression levels of phospho-eIF2α, CHOP, and BiP were observed in control or MxCre+baxfl/–bak–/– livers after Tm injection (Fig. 1A), indicating unperturbed PERK signaling. However, JNK phosphorylation was decreased in MxCre+baxfl/–bak–/– livers (Fig. 1A). Six hours after Tm injection, increased abundance of XBP-1 was observed by Western blot analysis of lysates prepared from nuclear fractions of control liver extracts, but only a small increase in the amount of XBP-1 was detected in extracts from MxCre+baxfl/–bak–/– liver (Fig. 1B), and almost no XBP-1 was evident 12 hours after injection (Fig. 1B). Expression of XBP-1 target genes (10) edem, erdj4, and ero1L, as well as protein disulfide isomerase pdi-P5, but not bip (a negative control), was reduced in MxCre+baxfl/–bak–/– livers compared with control livers (Fig. 1C). The number of cells that showed terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling (TUNEL-positive cells) was significantly decreased in MxCre+baxfl/–bak–/– liver compared with those of control animals (Fig. 1D), consistent with the lack of mitochondrial expression of BAX and BAK. Despite decreased apoptosis, histological analysis of the liver (Fig. 1E) and elevated alanine aminotransaminase and aspartate aminotransferase activities in serum (Fig. 1F) revealed increased tissue damage in MxCre+baxfl/–bak–/– animals after Tm injection, consistent with impaired XBP-1 expression (20). Thus, the phenotype observed in BAX-BAK–deficient mice after Tm injection is indicative of cellular dysfunction rather than cell death. A normal stress response was observed in the kidney, where low deletion efficiency of the floxed allele provides an internal control (fig. S1). Overall, these data suggest that BAX and BAK have a crucial role in adaptation responses to ER stress in vivo through modulating IRE1α to XBP-1 signaling.

Fig. 1.

Diminished XBP-1 induction and JNK phosphorylation in DKO livers of mice injected with Tm. MxCre+bax+/–bak–/– (control) and MxCre+baxfl/–bak–/– (DKO) mice were treated with poly(IC) to induce bax deletion. Animals were injected with Tm and killed after 6, 12, or 24 hours. (A) Expression levels of phospho-JNK, JNK1, CHOP/GADD153, BiP (Grp78), BAX, BAK, phospho-eIF2α, eIF2α, and actin were analyzed by Western blot in total liver extracts. WT livers from untreated mice were used as a positive control for BAK expression. (B) XBP-1 expression levels were analyzed in nuclear extracts of liver samples obtained in (A). A nonspecific band from the same Western blot is shown as a loading control. (C) Relative mRNA levels of edem, erdj4, ero1, pdi-P5, and bip were quantified by real-time polymerase chain reaction (PCR) in total cDNA obtained from the liver of control and DKO mice injected with Tm for 12 hours. (D) TUNEL staining of liver sections from control and DKO mice 4 days after Tm injection. Representative images of tissue from Tm-treated animals are presented. Green fluorescence, TUNEL staining; red fluorescence, propidium iodide (PI) staining. Graphically displayed results represent means and SD from the quantification of five different animals per group. (E) Hematoxylin-and-eosin staining of samples analyzed in (D). Of note, increased cytosolic vacuolation was observed in MxCre+baxfl/–bak–/– mice compared with control animals (n = at least five animals per group). (F) In parallel, serum alanine aminotransaminase (ALT) and aspartate aminotransferase (AST) levels were determined. Results represent means and SDs from three to four different animals before and after Tm injections. Where indicated, P values for statistical analysis were obtained using Student's t test (ns, not significantly different).

To explore further the signaling pathways by which BAX and BAK modulate the UPR, we initiated an ER stress response with Tm in wild-type (WT) and DKO murine embryonic fibroblasts (MEFs) (1) and analyzed the time course of signaling events initiated by IRE1α and PERK. XBP-1 expression was decreased in DKO cells undergoing ER stress compared with WT MEFs. Little or no phosphorylation of JNK at Tyr183 and Tyr185 and of its upstream kinase ASK1 (apoptosis signal–regulating kinase) was observed after Tm treatment of DKO cells (Fig. 2A). In contrast, tumor necrosis factor–α (TNF-α) (21) caused normal JNK phosphorylation in DKO cells (fig. S2A). In summary, DKO cells displayed a phenotype similar to that of IRE1α-deficient cells (fig. S2B) (7, 10).

Fig. 2.

BAX and BAK regulate IRE1α signaling in cells undergoing ER stress. (A) WT and DKO MEFs were treated with Tm (2.2 μg/ml) for indicated time points, and expression levels of XBP-1, phospho-JNK, phospho-ASK1, and PTP-1B were determined by Western blot. As a loading control, total JNK and ASK1 levels were assessed. (B) PERK signaling events were studied in cells treated as above with Tm. Expression levels of phospho-eIF2α, total eIF2α, phospho-PERK, CHOP, BiP, and actin were determined by Western blot. (C) DKO cells stably expressing WT BAX (wtBAX) or mtBAX were treated with the indicated concentration of Tm or etoposide, and cell viability was determined after 24 hours of incubation by the soluble tetrazolium salt, MTS, cytotoxicity assay. Differences in Tm treatment between cell types were statistically significant (P < 0.001), as analyzed by two-way analysis of variance (ANOVA). (D) BAKcb5 was transiently expressed in DKO cells for 48 hours in the presence of zVAD-fmk (10 μM) and then treated with thapsigargin (Thap) for 4 hours. Expression levels of XBP-1, phospho-JNK, BAKcb5, CHOP, and actin were determined by Western blot. (Left) Controls—the subcellular localization of BAKcb5 was assessed by immunofluorescence (red), costaining the ER with a KDEL antibody (green) and the nucleus with Hoechst (blue). All results are representative of at least three independent experiments.

BAX and BAK deficiency did not alter PERK activation, as observed from comparable phosphorylation of PERK (Thr980) and eIF2α (Ser51) in WT and DKO cells (Fig. 2B). In agreement with this result, the expression of PERK downstream targets, such as CHOP and BiP, were similar in WT and DKO cells (Fig. 2B) as previously described (22, 23). As an additional control, UPR activation was assessed in cells lacking the antiapoptotic protein BCL-2. No defect in XBP-1 and CHOP induction was observed in Tm-treated BCL-2–deficient cells (fig. S2C). Thus, the defects in IRE1α activity observed in DKO cells are likely not a consequence of imbalanced expression of anti- and proapoptotic proteins.

BAX and BAK regulate apoptosis initiated at the mitochondria. To assign the site of BAX and BAK action in modulating IRE1α signaling, we performed organelle-specific reconstitution assays. Expression of a mitochondria-targeted form of BAX (mtBAX) (3) did not restore activation of the IRE1α pathway as measured by JNK activation or XBP-1 expression (fig. S3A), although it did restore the susceptibility of DKO cells to ER stress–mediated apoptosis (3) (fig. S3B). DKO cells reconstituted with WT BAX or mtBAX sensed apoptotic signals at the mitochondria as reflected by similar rates of apoptosis induced by the DNA damage agent etoposide (Fig. 2D) or by overexpression of tBID (truncated, active fragment of BID) (fig. S3C). Cells expressing mtBAX were more sensitive to ER stress than were cells expressing WT BAX, which suggests that BAX expression at the ER membrane may have a positive effect on adaptation to ER stress (Fig. 2C and fig. S3D). We reconstituted DKO cells with BAKcb5, a BAK mutant in which the transmembrane domain is replaced by the ER targeting sequence of cytochrome b5 (23), in the presence of caspase inhibitors to minimize toxic effects. BAKcb5 expression increased amounts of XBP-1 and phospho-JNK in DKO cells undergoing ER stress and did not affect CHOP expression (Fig. 2D). These findings indicate that BAK expression at the ER membrane is required for IRE1α signaling. Altered calcium homeostasis appeared not to be responsible for the defects in IRE1α signaling, because restoration of ER calcium content by SERCA (sarcoplasmic or endoplasmic reticulum calcium adenosine triphosphatase) overexpression (3) did not restore stress-induced expression of XBP-1 or JNK activation in DKO cells (fig. S3A), although these cells recovered susceptibility to calcium-mediated apoptosis (3).

In plasma cells, XBP-1 induced the expression of multiple secretory pathway genes, increased cell size, initiated biogenesis of the ER, and elevated total protein synthesis (2426). We assessed ER morphology in DKO cells expressing a cytochrome b5–green fluorescent protein (GFP) fusion protein that allows visualization of the ER (27) (fig. S4A). BAX and BAK deficiency prevented the appearance of morphological changes in the ER such as vacuolization and redistribution of the ER triggered in response to thapsigargin treatment (Fig. 3A). This effect was independent of caspase activation or calcium release (fig. S4B). Secretory pathway expansion was quantified by fluorescence-activated cell sorting (FACS) in living cells stained with a red fluorescent version of brefeldin A (24) that labels the ER and Golgi compartments (Fig. 3B, left). ER-Golgi content was similar in WT and DKO cells under unstressed conditions (Fig. 3B, right). However, increased brefeldin A–BODIPY (boron dipyrromethene difluoride) staining was observed in WT but not DKO cells undergoing ER stress (Fig. 3C and fig. S3C). The phenotype of IRE1α knockout cells resembled that observed in DKO cells (Fig. 3D and fig. S3D). DKO cells retained the ability to expand the ER-Golgi network, because transient expression of the spliced form of XBP-1 in DKO cells triggered this process (Fig. 3E), further implicating BAX and BAK as upstream regulators of the IRE1α and XBP-1 pathway. Moreover, expression of BAKcb5, but not mtBAX or SERCA (fig. S3E), restored the ability of DKO cells to expand the secretory pathway under ER stress conditions (Fig. 3F).

Fig. 3.

Impaired expansion of the secretory pathway in DKO cells undergoing ER stress. (A) WT and DKO cells stably expressing cytochrome b5–GFP were treated with thapsigargin (1.5 μM) for 2.5 hours or left untreated, and the morphology of the ER (GFP fluorescence) was assessed. (B) DKO cells expressing cytochrome b5–GFP (green) were stained with brefeldin A–BODIPY (red), and fluorescence emission was analyzed (left panel). WT or DKO cells were stained with brefeldin A–BODIPY, and basal fluorescence emission was analyzed by FACS (right). (C) WT or DKO cells were treated with Tm (2.2 μg/ml) for 2.5 hours and then analyzed by FACS after staining with brefeldin A–BODIPY. (D) For comparison, IRE1α knockout (KO) cells or WT control cells were analyzed as described in (C). (E) DKO cells were transduced with a retroviral expression vector for spliced XBP-1 or empty vector (mock) and, after 48 hours of selection with hygromycin, cells were stained with brefeldin A–BODIPY and analyzed by FACS. Autofluorescence of unstained cells is shown. (Right) Expression levels of XBP-1 and actin were determined by Western blot in the same samples. (F) DKO cells transiently expressing BAKcb5 for 48 hours and then treated with Tm (2.2 μg/ml) for 2.5 hours were analyzed by FACS after brefeldin A–BODIPY staining. Experiments were performed in the presence of zVAD-fmk (10 μM). All results are representative of at least three independent experiments performed in duplicates.

The expression of proteins that regulate IRE1α activation, such as PTP-1B (Fig. 2A) and BiP (Fig. 2B), was not altered in DKO cells, which raises the possibility that BCL-2–related proteins might directly interact with IRE1α. We therefore tested a His-tagged version of the cytosolic region of human IRE1α (His-cytIre1α, amino acids 468 to 977) for its ability to bind recombinant BAX, tBID, a fusion protein of BCL-2 with glutathione S-transferase (GST), or BCL-xL–GST. Isolation of complexes with nickel-agarose revealed an interaction between His-cytIRE1α and BAX (Fig. 4A and fig. S5A). No binding was observed between His-cytIRE1α and BCL-2, BCL-xL, or proapoptotic tBID (Fig. 4A).

Fig. 4.

Protein complex formation between BAX/BAK and IRE1α. (A) The binding of recombinant BCL-2–related proteins with a His-tagged version of the cytosolic domain of IRE1α (His-IRE1α) was investigated by pull-down assay with Ni-agarose and analyzed by Western blot. Control experiments were performed in the absence of IRE1α. As positive controls, total BCL-2 recombinant proteins (0.5 μg) were loaded where indicated. (B) Cell extracts from WT MEFs stably expressing full-length IRE1α-HA, the N-terminal region (IRE1αΔC-HA), the C-terminal region (IRE1αΔN-HA), or empty vector (mock) were immunoprecipitated with agarose containing antibody against hemagglutinin (HA), and association with endogenous BAX was assessed by Western blot. (C) WT MEFs were treated with Tm (2.2 μg/ml) for the indicated times or left untreated, and BAX was immunoprecipitated. Association with endogenous IRE1α was assessed by Western blot. Control immunoprecipitation experiments were performed with DKO cell extracts (Ctrl DKO). (D) DKO cells stably expressing IRE1α-HA were transiently infected with retroviruses encoding BAK WT, BAKmBH3, BAKmBH1, or empty vector (mock) for 48 hours in the presence of zVAD-fmk (10 μM), and their coimmunoprecipitation with IRE1α-HA was assessed with HA antibody coupled to agarose. (E) In parallel, DKO cells reconstituted with vectors as described in (D) were treated with Tm for 8 hours, and expression levels of XBP-1, phospho-JNK, BAK, and actin were determined by Western blot. (F) WT and DKO cells stably expressing HA-IRE1α were treated with brefeldin A (20 μM) for the indicated times, and the electrophoretic pattern of IRE1α was analyzed by Western blot (phosphorylated IRE1α, P-IRE1α). (G) WT and DKO cells stably expressing IRE1α-HA or empty vector (Ctrl) were treated with Tm for 4 hours or left untreated, and IRE1α was immunoprecipitated for analysis of BiP binding. (H) DKO cells stably expressing IRE1α-HA were transiently infected with retroviruses encoding BAK WT (WT), BAKmBH3 (BH3), BAKcb5 (cb5), or empty vector (m) for 48 hours in the presence of zVAD-fmk (10 μM), and the coimmunoprecipitation of BiP with IRE1α-HA was assessed. Control immunoprecipitation experiments were performed in WT cells not expressing IRE1α-HA.

Coimmunoprecipitation experiments using lysates from cells transfected with IRE1α hemagglutinin-tagged cDNA (IRE1α-HA) showed an association of IRE1α with endogenous BAX (Fig. 4B) and BAK (fig. S5B). A similar result was observed when endogenous BAX was immunoprecipitated, and the presence of IRE1α-HA was assessed (fig. S5C). These interactions appeared to be specific because IRE1α-HA showed little or no association with BCL-2 (fig. S5D), BCL-xL (fig. S5E), or PERK (fig. S5F). The IRE1α-HA and BAX interaction required the cytosolic C-terminal region of IRE1α, which encodes the kinase and endoribonuclease domains (Fig. 4B). The interaction of BAX with endogenous IRE1α was also detected (Fig. 4C). Interestingly, the association of BAX with IRE1α was significantly increased in cells undergoing ER stress, which suggests that BAX preferentially interacts with active IRE1α (Fig. 4C).

BH3 and BH1 domains of BAK control its proapoptotic activity through interaction with other BCL-2–protein partners (28). Analysis of point mutants in the BH3 and BH1 domains (BAKmBH1, W122A, G123E, R124A; and BAKmBH3, L75E) (29, 30), showed that these domains were also required for binding of BAK to IRE1α-HA (Fig. 4D). These same mutations also altered, to different extents, the ability of BAK to modulate IRE1α signaling (Fig. 4E). Taken together, these results suggest that the interaction between IRE1α and BAX and BAK is specific; requires the BH3 and BH1 domains; and, if disrupted, abolishes the effect of BAX/BAK on IRE1α signaling.

To assess whether impaired JNK phosphorylation and diminished XBP-1 expression indicated altered IRE1α activation in DKO cells, we examined the phosphorylation and dimerization status of IRE1α. DKO cells treated with brefeldin A for 2 to 4 hours did not show the mobility shift of IRE1α indicative of its autophosphorylation (Fig. 4F) (8). In addition, we detected the oligomerization of IRE1α in WT, but not in DKO, cells undergoing ER stress, by using nondenaturing gels or sedimentation in a sucrose gradient (fig. S6A and B). These data are consistent with diminished IRE1α signaling. Immunoprecipitation experiments revealed that more BiP was bound to IRE1α in DKO cells than in WT cells, under both basal and stress conditions (Fig. 4G), consistent with a decreased activation state of IRE1α in the DKO cells. Reconstitution of DKO cells with WT BAK or the ER-targeted mutant BAKcb5 decreased the basal binding of BiP to IRE1α, but expression of BAKmBH3 did not (Fig. 4H). Deletion of the cytosolic domain of IRE1α leads to basal binding of BiP that is not affected by BAX and BAK deficiency, which reinforces the idea that BAX and BAK regulate IRE1α activation through interaction with its cytosolic domain (fig. S6C). Finally, partial knockdown of BiP with small interfering RNA (siRNA) increased the rate of IRE1α activation in DKO cells as evidenced by increased expression of spliced XBP-1 (fig. S6D). Thus, BAX and BAK modulate IRE1α signaling through direct binding, possibly by the stabilization of the active form of the protein.

Our results reveal the proapoptotic proteins BAX and BAK as essential components of the UPR, a signaling system that allows secretory cells to handle the stress of protein folding, protein quality-control, and protein secretion. This function is independent of the proapoptotic function of BAX and BAK at the mitochondria and depends on the presence of the BH3-BH1 binding pocket. Under stress conditions, BAX and BAK interact with the cytosolic region of IRE1α, which is required for the modulation of IRE1α signaling (fig. S7). Members of the BCL-2 family are found in regulatory complexes at diverse organelles (4, 29, 3133). Thus, members of the BCL-2 protein family may act as stress sentinels that connect stress signals to the proapoptotic core when cellular homeostasis is irreversibly altered.

References and Notes

Stay Connected to Science

Navigate This Article