IRE1α Cleaves Select microRNAs During ER Stress to Derepress Translation of Proapoptotic Caspase-2

See allHide authors and affiliations

Science  09 Nov 2012:
Vol. 338, Issue 6108, pp. 818-822
DOI: 10.1126/science.1226191


The endoplasmic reticulum (ER) is the primary organelle for folding and maturation of secretory and transmembrane proteins. Inability to meet protein-folding demand leads to “ER stress,” and activates IRE1α, an ER transmembrane kinase-endoribonuclease (RNase). IRE1α promotes adaptation through splicing Xbp1 mRNA or apoptosis through incompletely understood mechanisms. Here, we found that sustained IRE1α RNase activation caused rapid decay of select microRNAs (miRs -17, -34a, -96, and -125b) that normally repress translation of Caspase-2 mRNA, and thus sharply elevates protein levels of this initiator protease of the mitochondrial apoptotic pathway. In cell-free systems, recombinant IRE1α endonucleolytically cleaved microRNA precursors at sites distinct from DICER. Thus, IRE1α regulates translation of a proapoptotic protein through terminating microRNA biogenesis, and noncoding RNAs are part of the ER stress response.

Various physiological events (e.g., secretory cell differentiation or polypeptide hormone production) and pathological insults (e.g., hypoxia, ischemia, or changes in intracellular pH) increase protein-folding demand on the secretory pathway, which, in turn, triggers ER stress (1). The tripartite unfolded protein response (UPR)–signaling arms of IRE1α, PERK, and ATF6α attempt to resolve ER stress through complementary homeostatic mechanisms, which include inhibiting cap-dependent protein translation, increasing ER chaperones, and enhancing ER-associated protein degradation (ERAD) (2). However, if ER stress cannot be remedied through these mechanisms, the UPR induces apoptosis through the mitochondrial BAX/BAK–dependent pathway (1). Excessive ER stress–induced cell loss contributes to numerous human degenerative diseases, including diabetes, neurodegeneration, and cardiovascular disease (1, 3, 4).

Severe ER stress activates the protease Caspase-2 (CASP2) as an early apoptotic switch upstream of mitochondria (37). Once activated, CASP2 cleaves the BH3-only protein BID, which then localizes to mitochondria to induce BAX/BAK–dependent apoptosis (8, 9). However, the molecular events leading from the detection of upstream ER stress to CASP2 activation remain unknown. To address this question, we challenged wild-type (WT) and apoptosis-resistant Bax−/−Bak−/− (DKO) mouse embryonic fibroblasts (MEFs) with brefeldin A (BFA), a drug that retards protein trafficking in the secretory pathway to cause ER stress (10). CASP2 protein, which is normally expressed at low levels in these cells, increased dramatically within 2 hours of BFA treatment and rose steadily over the next 12 to 18 hours (Fig. 1, A and B, and fig. S1). CASP2 then underwent internal cleavage at ~18 hours, concomitant with entry of WT—but not DKO—MEFs into the apoptotic pathway, as evidenced by activation of the downstream executioner CASP3 and annexin V staining (Fig. 1, A to C). CASP2 up-regulation and activation were preserved in DKO MEFs, which suggests that they occur upstream of the mitochondrial apoptotic pathway (Fig. 1, B and C). Furthermore, CASP2 was efficiently induced by ER stress in Atf6α−/− and Perk−/−, but not in Ire1α−/−, MEFs (Fig. 1, D and E), and CASP2-dependent proteolytic activation of BID in response to ER stress was absent in Ire1α−/− MEFs (fig. S2). Thus, IRE1α may represent the upstream ER sensor used by cells to up-regulate CASP2 protein. Consistent with this notion, Ire1α−/− MEFs were resistant to BFA-induced apoptosis, and provision of IRE1α reconstituted apoptosis in a BAX/BAK–dependent manner (figs. S3 and S4).

Fig. 1

IRE1α is necessary and sufficient for CASP2 up-regulation. (A and B) Immunoblot for full-length (FL) CASP2 and cleaved (Clvd) CASP2 in WT and DKO MEFs after BFA treatment. (C) Annexin V–directed fluorescence-activated cell sorting analysis of WT and DKO MEFs treated with BFA. Each data point represents the mean value ± SD from three independent experiments. (D) CASP2 immunoblot in UPR sensor–deficient MEFs treated with BFA. (E) CASP2 immunoblot of Ire1α+/+ and Ire1α −/− MEFs treated with tunicamycin (Tn) or thapsigargin (Tg).

IRE1α consists of an N-terminal sensor domain within the ER that detects misfolded proteins, a transmembrane region, and a cytosolic tail containing two distinct catalytic activities—a serine-threonine kinase and an endoribonuclease (RNase) (11, 12) (fig. S7). Accumulation of misfolded proteins within the ER leads to IRE1α oligomerization and subsequent trans-autophosphorylation, which allosterically activates its RNase. Isogenic T-REx-293 cell lines have been generated that express either WT or various mutant forms of IRE1α under doxycycline (Dox) control (13). Because activation of IRE1α requires oligomerization in the ER membrane, this process can be driven by mass action in the absence of ER stress (13). Activation of WT-IRE1α was sufficient for both robust up-regulation of CASP2 and its subsequent cleavage, similar to what occurs under irremediable ER stress (Fig. 2A and figs. S5 and S6).

Fig. 2

The RNase activity of IRE1α up-regulates CASP2 independently of XBP1. (A) CASP2 immunoblot upon Dox induction of WT-IRE1α in T-REx-293 cells. (B) CASP2 immunoblot in T-REx 293 cells that overexpress various IRE1α forms (C) CASP2 immunoblot in Xbp1−/− MEFs transfected with pcDNA5-WT-IRE1α. (D) CASP2 immunoblot in T-REx-293 cells before and after Dox induction of XBP1s.

To determine the contribution of the kinase and/or RNase activity of IRE1α in CASP2 up-regulation, we tested various IRE1α mutants (fig. S7). A kinase-active and RNase-dead variant of IRE1α (K907A) was unable to up-regulate CASP2 (Fig. 2B). We used a chemical-genetic tool that can selectively activate IRE1α’s RNase, IRE1α (I642G) (13). The IRE1α (I642G) mutant is deficient in phosphotransfer activity (13, 14), but, in the presence of the adenosine triphosphate analog 1NM-PP1, it undergoes a conformational change that activates its RNase (13, 14). Dox-induced expression of IRE1α (I642G) alone (RNase “OFF”) did not increase CASP2, but the addition of 1NM-PP1 (RNase “ON”) increased CASP2 as efficiently as WT-IRE1α (Fig. 2B). Unlike under WT-IRE1α activation, there was no CASP2 cleavage after 1NM-PP1 activation of IRE1α (I642G). Thus, CASP2 is subject to multistep regulation downstream of IRE1α, and activating the RNase of IRE1α (I642G) with 1NM-PP1 selectively activates the initial step in the process. Consistent with this notion, forcible activation of IRE1α (I642G) does not cause apoptosis (13). IRE1α-mediated CASP2 up-regulation occurred independently of its classical target, XBP1 (Fig. 2, C and D, and fig. S8).

Although Casp2 mRNA levels remained stable in response to BFA or WT-IRE1α (fig. S9), polyribosome-associated Casp2 mRNA significantly increased within 1 hour of BFA in Ire1α+/+, but not Ire1α−/−, MEFs (Fig. 3A and fig. S10). Moreover, CASP2 increased in response to BFA even when transcription was blocked with actinomycin D (Fig. 3B) but not when translation was blocked with cycloheximide (fig. S11). Thus, IRE1α’s RNase appears to up-regulate CASP2 expression posttranscriptionally. A bioinformatic analysis of Casp2 mRNA revealed several high-confidence matches for binding sequences of known miRNAs within its 3′ untranslated region (3′UTR), including miR-17, miR-34a, miR-96, and miR-125b (fig. S12). These four miRNAs rapidly and significantly decreased in Ire1α+/+, but not Ire1α−/−, MEFs upon BFA treatment (Fig. 3C); an unrelated miRNA, let-7a, did not. We could mimic this regulation using our chemical-genetic tools: WT-IRE1α or 1NM-PP1 activation of IRE1α (I642G) also caused rapid decreases in miR-17, -34a, -96, and -125b (Fig. 3D).

Fig. 3

Anti-Casp2 miRNAs decrease in IRE1α-dependent manner. (A) Quantitative polymerase chain reaction (QPCR) on polyribosome-associated Casp2 mRNA derived from Ire1α+/+ and Ire1α −/− MEFs before and after BFA treatment. (B) CASP2 Immunoblot of Ire1α+/+ MEFs treated with BFA plus or minus pretreatment with actinomycin A (ActD). QPCR of select miRNAs forms. (C) Ire1α+/+ and Ire1α −/− MEFs after BFA treatment and (D) T-REx-293 cells after overexpression of WT-IRE1α or 1NM-PP1 activation of IRE1α (I642G). Each data point represents the mean value ± SD from three independent experiments. Asterisks indicate a statistically significant change from the vehicle-treated controls (P < 0.05).

To test whether IRE1α-mediated reduction of select miRNAs can increase translation of a target mRNA, we devised a reporter system with an mCherry gene construct containing defined miRNA binding sites within its 3′UTR. As such, mCherry expression is low in the presence of the matching miRNA, but becomes up-regulated if the specific miRNA decreases sufficiently. The mCherry reporters for miR-17, -34a, -96, and -125b (but not let-7a) in T-REx-293 cells were all up-regulated after Dox induction of WT-IRE1α or IRE1α (I642G) plus 1NM-PP1 (fig. S13, A, B, G, and H). BFA caused similar increases in these miRNA reporters in Ire1α+/+, but not Ire1α−/−, MEFs (fig. S13, C to F). To examine the role of these miRNAs in regulating Casp2 translation, we transfected T-REx-293 cells with specific anti-miRNA oligonucleotides (1517). Anti–miR-17, -34a, -96, or -125b each modestly increased CASP2, but all four together led to CASP2 levels similar to those seen with BFA (Fig. 4A). Conversely, single anti-Casp2 miRs (-17, -34a, -96, and -125b) partially reduced, and the combination of all four effectively prevented, CASP2 up-regulation by IRE1α (Fig. 4B). To determine whether the binding sequences for miR-17, -34a, -96 and -125b within the 3′UTR of Casp2 mRNA are critical for translational control, we introduced the complete 3′UTR sequence into a dual Firefly (FLuc) and Renilla (RLuc) luciferase reporter system in the T-REx-293 cells (18). Similar to the endogenous Casp2 mRNA, Dox induction of WT-IRE1α or I642G-IRE1α plus 1NM-PP1 activated 3′UTR-dependent translation as indicated by increases in the RLuc/FLuc ratio, but not when the binding sequences for the anti-Casp2 miRNAs were mutated (fig. S14). Thus, IRE1α controls Casp2 translation via down-regulating these select anti-Casp2 miRNAs.

Fig. 4

IRE1α directly cleaves pre-miR-17. CASP2 immunoblot of T-REx-293 cells transfected with indicated (A) anti-miRNAs or (B) miRNA mimics after Dox induction of WT-IRE1α. (C) QPCR of pri-, pre-, and mature miR-17 after IRE1α activation in WT-IRE1α T-REx-293 cells. Each data point represents the mean value ± SD from three independent experiments. (D) Radioblot of 32P-labeled pre-miR-17 digestion products after incubation with indicated recombinant IRE1α proteins. (E) Mapping of IRE1α cleavage sites in pre-miR-17. (F) Illustration of the IRE1α cleavage sites within pre-miR-17.

To explore the mechanism through which IRE1α down-regulates these miRNAs, we examined the biogenesis of miR-17. IRE1α activation did not affect pri-miR-17, but it significantly reduced both pre-miR-17 and mature miR-17 (Fig. 4C), which suggested that IRE1α regulates its biogenesis at the precursor step. Indeed, incubation of radiolabeled pre-miR-17 with recombinant human WT-IRE1α or 1NM-PP1–activated IRE1α (I642G) (but not the K907A RNase mutant) resulted in two prominent fragments [of ~40 nucleotides (nt) and ~17 nt], as well as a slightly less abundant ~49-nt fragment (Fig. 4D). These IRE1α cleavage products all mapped to positions distinct from DICER sites but with some similarity to IRE1α’s scission sites in Xbp1 (Fig. 4, E and F). Thus, IRE1α appears to cleave select pre-miRNAs directly to prevent proper DICER processing of their mature forms, a mechanism recently described for how another ribonuclease, MCPIP1, terminates miRNA biogenesis (19).

Hyperactivated WT-IRE1α—but not 1NM-PP1 activated IRE1α (I642G)—endonucleolytically degrades hundreds of ER-localized mRNAs, which further compromises ER protein-folding capacity and hastens cell demise (13). Although activated WT-IRE1α and IRE1α (I642G) both decreased anti-Casp2 miRNAs to induce expression of pro-CASP2, only the former triggered its subsequent proteolytic processing, which suggested that ER-localized mRNA decay may be a critical “second signal” for full CASP2 activation and apoptosis. Unlike mRNAs directed to the ER membrane through a signal peptide leader sequence, it is currently unclear how pre-miRNAs gain proximity to IRE1α under conditions of ER stress. Possibilities include alterations in miRNA localization structures (e.g., P-bodies or stress granules), the involvement of miRNA-binding proteins, or the loss of other miRNA protective factors (2022). Alternatively, because the outer nuclear membrane is continuous with the ER, IRE1α may become exposed to pre-miRNAs as they transit through the nuclear pore.

There have been previous hints of a connection between miRNAs and ER stress signaling. For instance, the apoptotic proteins BIM, BAK, and PUMA are known to increase during ER stress-induced apoptosis (7, 23, 24) and to contain putative miRNA binding sites within their 3′UTRs. Furthermore, we recently found that IRE1α posttranscriptionally stabilizes the mRNA encoding the pro-oxidant thioredoxin-interacting protein (TXNIP), in part, through reducing miR-17 to activate the NLRP3 inflammasome in pancreatic beta cells under ER stress (25). Here, we provide evidence that the action of IRE1α on miRNA biogenesis is direct and that it antagonizes classical processing by DICER to derepress a translational block to entry into apoptosis. Thus, endonucleolytic cleavage of miRNA precursors by IRE1α adds to a growing list of extra-Xbp1 mRNA splicing functions controlled by this UPR sensor. Given their potential to alter expression of multiple mRNA targets simultaneously, noncoding RNAs are well-suited to govern complex cellular remodeling in response to ER stress signaling.

Supplementary Materials

Materials and Methods

Figs. S1 to S15

Reference (26)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: We thank R. Kaufman for Atf6α−/− MEFs and D. Ron for Perk−/− and Ire1α−/− MEFs. The work was supported by NIH: DP2OD001925 (F.R.P), R01CA136577 (S.A.O.), R01CA136717 (A.G.), R01DK080955 (F.R.P), R01CA154916 (D.R.), GM080783 (M.T.M.), DK063720 (M.T.M.), and R01CA140456 (D.R.); Leukemia & Lymphoma Society Scholar Award (D.R.); Howard Hughes Medical Institute Physician-Scientist Early Career Award (S.A.O.); American Cancer Society Research Scholar Award (S.A.O.); Burroughs Wellcome Foundation (F.R.P.); Hillblom Foundation (F.R.P.); Juvenile Diabetes Research Foundation (F.R.P); Partnership for Cures (F.R.P); National Science Foundation (E.S.W.); Susan G. Komen Foundation (A.G.); and an A*STAR Fellowship (L.L.). J-P.U., L.W., D.H., E.S.W., N.H., and M.T. designed and performed experiments and contributed to the manuscript. L.L., M.T.M., D.R., and A.G. contributed key ideas, reagents, and data interpretation. F.R.P. and S.A.O. designed the study and wrote the manuscript. There are no conflicts of interest.

Stay Connected to Science

Navigate This Article