miR-216b regulation of c-Jun mediates GADD153/CHOP-dependent apoptosis

The ability of the unfolded protein response, UPR, to regulate cell homeostasis through both gene expression and protein synthesis has been well documented. One primary pro-apoptotic protein that responds to both PERK and Ire1 signalling is the CHOP/GADD153 transcription factor. Although CHOP deficiency delays onset of cell death, questions remain regarding how CHOP regulates apoptosis. Here, we provide evidence demonstrating that CHOP/GADD153-dependent apoptosis reflects expression of micro-RNA, miR-216b. MiR-216b accumulation requires PERK-dependent induction of CHOP/GADD153, which then directly regulates miR-216b expression. As maximal expression of miR-216b is antagonized by Ire1, miR-216b accumulation reflects the convergence of PERK and Ire1 activities. Functionally, miR-216b directly targets c-Jun, thereby reducing AP-1-dependent transcription and sensitizing cells to ER stress-dependent apoptosis. These results provide direct insight into the molecular mechanisms of CHOP/GADD153-dependent cell death.

T he endoplasmic reticulum (ER) serves as a nexus for the folding and maturation of proteins that transit the secretory pathway. Proteins are imported co-translationally whereupon they undergo additional energy intensive posttranslational modifications before folding and maturation. Protein folding is both regulated and sensed by ER resident chaperones, such as Grp78/BiP and Grp94 (refs 1-5). Under conditions where protein production out paces protein folding capacity, for example, under nutrient-deficient conditions as might occur during neoplastic growth, misfolded proteins accumulate in the ER triggering a stress sensing/adaptive pathway referred to as the unfolded protein response (UPR).
Mammalian cells contain three ER transmembrane proteins that function as proximal effectors of the UPR. Inositol-requiring enzyme, Ire1, isoforms (a, ubiquitously expressed; b tissue restricted) are composed of a luminal domain that senses stress, a single transmembrane domain, and a cytosolic tail that contains both a protein kinase domain and an RNase domain 6,7 . Ire1 triggers increased expression of numerous ER chaperones through activation of the X-box-binding protein 1 (Xbp1) transcription factor 8,9 . Accumulation of Xbp1 is mediated by the RNase and splicing function of Ire1; stress-induced splicing generates a shorter Xbp1 mRNA that is more efficiently translated 10,11 . Protein kinase RNA-like ER kinase (PERK) is activated in a manner analogous to the Ire1; it catalyses serine 51 phosphorylation of eIF2a resulting in repression of protein synthesis [12][13][14] . The third signalling component are the transmembrane transcription factors ATF6a/b (refs 15,16). Although normally tethered to the ER, upon stress, ATF6 migrates to the trans-Golgi, where it is processed by S1P and S2P proteases to release the N-terminal DNA-binding transcription factor domain [17][18][19] . Activation of PERK, Ire1 and ATF6 is mediated by sequences within their respective luminal domains to which the ER chaperone, BiP, binds 20 . Activation is triggered by increased unfolded proteins, which compete for BiP binding.
The ability of the UPR to regulate protein synthesis, gene expression and contribute to cell homeostasis has been well documented. However, an added layer of communication that entails the regulation of small non-coding RNAs, micro-RNAs, was recently revealed. Micro-RNAs are 20-22 nucleotide molecules that generally impact protein expression by virtue of their capacity to degrade target mRNAs or reduce their efficiency of translation 21,22 . During the past 4-5 years, work from several groups has revealed that all three branches of the UPR regulate specific subsets of micro-RNAs [23][24][25][26][27] . The modes of regulation include Ire1mediated micro-RNA degradation 25 , ATF4-dependent and ATF6-dependent transcription 27 . The consequences of micro-RNA induction include the modulation of protein expression as would be expected, while certain micro-RNAs also directly contribute to regulated gene expression 27 .
The UPR can communicate both pro-adaptive and proapoptotic signals. Ambiguities remain regarding whether an adaptive or apoptotic signal is generated given that all three transducers have been associated with both survival and apoptosis under varying conditions ( 9,28-34 ). One primary pro-apoptotic protein that responds to both PERK and Ire1 signalling is the CHOP/GADD153 transcription factor. Although CHOP deficiency delays onset of cell death, questions remain regarding how CHOP regulates apoptosis. The data presented here reveal evidence for CHOP-dependent micro-RNA, miR-216b, which subsequently suppresses c-Jun expression thereby, triggering cell death.

Results
MiR-216b is induced by the UPR. To identify micro-RNAs that are induced following prolonged ER stress, RNA harvested from NIH3T3 cells challenged with tunicamycin (Tu) was subjected to micro-array analysis. MiR-211, not represented in this figure, induction was assessed as an internal control. MiR-216b and miR-217 exhibited increased expression throughout the time course (Fig. 1a). Several groups have evaluated miR-216b expression and potential function in various cancers including colorectal cancer, hepatocellular carcinoma and nasopharyngeal carcinoma [35][36][37] ; we therefore chose to assess miR-216b regulation and function during ER stress. To verify the microarray data, RNA was collected from Tm-challenged NIH3T3 cells, assessed for CHOP induction as a control for UPR activation and for miR-216b. CHOP induction was noted by 2 h and consistent with our microarray data, we observed an accumulation of miR-216b (Fig. 1b). Accumulation of miR-216b was evident following the exposure of cells to an independent inducer of ER stress, thapsigargin (TG; Fig. 1c). MiR-216b expression was also induced in response to chemotherapeutic agents that generate ER stress 38 and physiologically relevant low glucose (Fig. 1d,e).
PERK-dependent induction of miR-216b. Because PERK was previously implicated as a regulator of micro-RNA accumulation during ER stress, we assessed its contribution to the regulation of miR-216b expression. Wild-type or PERK deficient murine embryonic fibroblasts 39 were treated with TG (500 nM), and expression of miR-216b was assessed by quantitative PCR (qPCR). PERK deficiency was confirmed by western blot analysis using a PERK-specific antiserum and through assessment of CHOP accumulation following stress (Fig. 2a). MiR-216b accumulation was noted in wild type but not PERK À / À cells (Fig. 2a). To corroborate the dependence of miR-216b expression on PERK activity, we used GSK2606414 (GSK414), a highly specific PERK small-molecule inhibitor 40,41 . GSK414 effectively inhibited PERK activation and miR-216b induction (Fig. 2b).
PERK-dependent regulation of gene expression is often dependent upon increased accumulation of the ATF4 transcription factor, whose accumulation is in turn dependent upon eIF2a phosphorylation 13,14,[42][43][44] . To address the contribution of eIF2a and ATF4, wild-type mouse embryonic fibroblasts (MEFs) or MEFs isolated from eIF2aS51A knockin embryos 45 or from ATF4 À / À embryos 46 were challenged with TG. MiR-216b induction was absent in both eIF2aS51A and ATF4 À / À backgrounds but was induced in wild-type MEFs (Fig. 2c,d) demonstrating a role for ATF4 in the regulation of miR-216b expression.
As ATF4 accumulates within the first 2-4 h of ER stress 46,47 , and miR-216b accumulation occurs primarily between 5 and 10 h, we considered the possibility that ATF4-dependent regulation was indirect. We focused our attention on CHOP, a direct transcriptional target of ATF4 (ref. 43). MiR-216b induction was abolished in chop knockout MEFs (Fig. 2e). Let7a accumulation was assessed as a nonspecific control ( Supplementary Fig. 1). Consistent with CHOP regulation, restoring CHOP expression in chop À / À MEFs rescued stress-dependent miR-216b expression (Fig. 2f). These data demonstrate that ER stress-dependent induction of miR-216b reflects a pathway minimally composed of PERK, eIF2a, ATF4 and CHOP.
We also assessed miR-216b expression in a panel of MMTV-Neu tumours that were derived in mice with either wild-type PERK or from mice in which PERK was excised from mammary epithelium with Cre recombinase 23,48 . In this cohort, miR-216b expression was significantly reduced in PERK-deficient tumours and this correlated with c-Jun expression levels, providing evidence for PERK-dependent induction of miR-216b in vivo (Fig. 2g,h).
MiR-216b is a direct CHOP transcriptional target. Interestingly, ATF4 and CHOP share a set of target genes, with the preference of binding to the similar motifs (GCATCAT/G) 47 . Using an unbiased motif search of B2 kbp upstream of mouse miR-216b, we found a potential binding site for ATF4 and/or CHOP at À 266 to À 259 ( Supplementary Fig. 2a). To evaluate potential binding to this motif, we performed chromatin immunoprecipitation (ChIP). ChIP with either anti-CHOP (Fig. 3a, right and left panel) or anti-ATF4 ( Supplementary   Fig. 2b) followed by qPCR revealed that TG treatment triggered a significant enrichment of CHOP on the miR-216b promoter relative to unstressed cells and control IgG. Under the same conditions, there was no apparent enrichment of ATF4 on the miR-216b promoter, although binding of ATF4 to the CHOP promoter was confirmed ( Supplementary Fig. 2b,c).
To further evaluate whether the upstream region from À 266 to À 259 region is responsible for miR-216b regulation, we used a reporter assay system where 1 kbp of upstream miR-216b   ARTICLE sequence was cloned upstream of the firefly luciferase gene (Fig. 3b). Following expression in 293T cells, luciferase activity was induced following TG treatment (Fig. 3c). Luciferase activity was abolished by GSK414 (Fig. 3c, black bars) demonstrating PERK dependence. Alteration of the CHOP-binding site through site-directed mutagenesis abrogated UPR-dependent induction ( Fig. 3c-white bars). Actinomycin D (2 mg ml À 1 ) treatment before exposure to TG treatment suppressed miR-216b expression, consistent with transcription dependence of miR-216b induction (Fig. 3d). To determine (a) PERK þ / þ and PERK À / À MEFs were treated with 500 nM TG for indicated times. MiR-216b was assessed by qPCR (left graph); PERK and CHOP were assessed by immunoblot (right). (b) MiR-216b levels were quantified by qPCR following exposure of cells to thapsigargin and a small-molecule PERK inhibitor (left). PERK, eIF2a-p and CHOP induction were assessed by immunoblot (right). (c-e) MEFs of the indicated genotype were treated with TG (500 nM) for indicated intervals. Protein extracts from these cells were immunoblotted for the proteins as indicated (lower panels) and miR-216b levels were quantified by qPCR (upper panels; n ¼ 3). (f) CHOP À/ À MEFs were transfected with vector or CHOP and 2 days later treated with TG (500 nM) for indicated intervals. Protein extracts from these cells were immunoblotted for CHOP and miR-216b levels quantified by qPCR (n ¼ 3). (g) MiR-216b expression and (h) c-Jun mRNA levels were analysed in MMTV-Neu tumours from either PERK þ / þ or a PERK À / À background. Data represent mean ± s.d. of three independent observations. Statistical significance was analysed analysed by Student's t-test.
Ire1a inhibits miR-216b expression. Mammalian Ire1a initiates downstream signalling through unconventional splicing of the transcription factor Xbp1 (ref. 10), phosphorylation of downstream targets or via regulated Ire1-dependent decay (RIDD) of multiple substrates 50 . Ire1a RIDD activity also contributes to the regulation of micro-RNA accumulation during ER stress 25 . To determine the contribution of Ire1a in miR-216b expression, we treated Ire1a knockout MEFs (Ire1a À / À ) and wild-type MEFs with TG. A significant increase in miR-216b induction in Ire1a knockout cells relative to wild type was observed (Fig. 4a). As an independent analysis, we used three different short hairpin RNAs (shRNAs) targeting Ire1a; notably, all three shRNAs reduced Ire1a expression thereby permitting increased accumulation of miR-216b following ER stress (Fig. 4b).
MiR-216b regulates c-Jun accumulation during ER stress. Before searching for targets of miR-216b, a luciferase reporter gene containing three tandem miR-216b-binding seeds in its 3 0 untranslated region (UTR) was used to determine whether miR-216b induction was of functional consequence. Treatment of cells expressing the reporter with TG resulted in a 50% reduction in luciferase activity ( Fig. 5a; left two columns). Co-transfection with miR-216b mimic also resulted in an B50% decrease in reporter gene expression; inclusion of TG reduced expression further by B80% ( Fig. 5a; middle columns). Luciferase activity increased 1.5-fold when co-transfected with an antagomir, A-miR-216b ( Fig. 5a; right columns).
To identify miR-216b targets, a computational algorithm was utilized to search 3 0 UTR regions of annotated cDNAs for miR-216b seed sequence matches; TargetScan (http://www.targetscan.org), revealed c-Jun as a high relevance target (Fig. 5b). TG exposure resulted in a time-dependent decrease in c-Jun levels ( Fig. 5c; quantification Supplementary Fig. 4a). In contrast preincubation with a PERK inhibitor ablated c-Jun regulation (Fig. 5c). To assess the contribution of miR-216b to stressdependent inhibition of c-Jun, U20S cell lines that stably overexpressed an shRNA vector encoding a miR-216b mimic or an anti-miR-216b were generated and subsequently exposed to a time-course of TG treatment. Cells expressing miR-216b mimic exhibited reduced basal c-Jun levels and increased kinetics of c-Jun loss relative to control ( Fig. 5d; quantification cJun Supplementary Fig. 4c, cleaved caspase Supplementary Fig. 4b).
Conversely, expression of A-miR-216b resulted elevated basal expression of c-Jun and a significant maintenance of c-Jun levels during ER stress (Fig. 5d).
In combination with c-Fos, c-Jun forms the AP-1 early response transcription factor 53,54 . It is expected that miR-216b induction should trigger decreased AP-1 activity. To test this hypothesis, a reporter construct wherein luciferase expression is driven by a minimal promoter and 3X AP-1-binding sites (TGAACTCA) was utilized. AP-1-dependent activity was reduced by ER stress (Fig. 5e). The inhibition of AP-1dependent reporter gene expression was amplified by coexpression of ectopic miR-216b and attenuated by A-miR-216b (Fig. 5e). We also reasoned that if miR-216b induction is CHOP dependent, CHOP deficiency should ablate c-Jun inhibition during ER stress. Consistently, CHOP À / À exposed to ER stress failed to suppress c-Jun after TG treatment (Fig. 5f). Finally, assessment of expression of c-Jun target genes revealed that ER stress and miR-216b reduced expression, whereas A-miR-216b rescued expression (Fig. 5g-j; quantification of cJun Supplementary Fig. 4d). These results revealed a direct relationship between miR-216b expression and c-Jun downregulation following ER stress.
Suppression of c-Jun sensitizes cells to apoptosis. As c-Jun protects cells from ultraviolet and tumour-necrosis factor-a-induced apoptosis through cooperation with NF-kB 55,56 , we considered the possibility that miR-216b might regulate ER stress-dependent apoptosis through regulation of c-Jun. Initially, we determined whether modulation of miR-216b altered ER stress-dependent ARTICLE apoptosis. Consistent with miR-216b as a regulator of cell survival, overexpression of miR-216b increased apoptosis induced by TG (Fig. 6a). In contrast, A-miR-216b-expressing cells were largely refractory to the TG-induced apoptosis (Fig. 6a). As an independent measure of cell survival, we exposed control, miR-216b-and A-miR-216b-expressing U2OS to TG for 0 or 3 h following which cells were supplied with fresh medium lacking TG. The cells were allowed to recover for 12 days, at which point surviving cells/colonies were quantified. miR-216b-expressing U2OS cells were significantly more sensitive to stress than control, whereas A-miR-216b-expressing cells were refractory to TG-induced death (Fig. 6b). The effect of miR-216b on recovery from ER stress was also observed with a second ER stressor, Tm ( Supplementary Fig. 5). Cell doubling times were not affected by the expression of either A-miR-216b or overexpression of miR-216b mimic confirming that miR-216b is regulating cell viability not cell proliferation (Fig. 6c).
To investigate whether the effect of miR-216b is c-Jundependent, c-Jun was suppressed using three independent shRNAs targeting distinct regions of c-Jun. c-Jun loss triggered a significant increase in cleaved poly (ADP-ribose) polymerase (PARP) that correlated with the degree of c-Jun suppression (Fig. 7a). We compared the impact of overexpression of sh-c-Jun versus overexpression of miR-216b by serial increases in viral supernatants encoding either sh-c-Jun or miR-216b. This experiment revealed that while the artificial sh-c-Jun was significantly more potent in reducing c-Jun than miR-216b, the degree of cleaved caspase 3 triggered by either correlated with the degree of c-Jun knockdown (Fig. 7b). If loss of c-Jun triggers apoptosis, overexpression of c-Jun should over-ride ER stressdependent and miR-216b-dependent apoptosis as assessed by caspase 3 cleavage. U2OS cells were infected with retrovirus coding c-Jun, lacking a 3 0 UTR, followed by TG treatment. Enforced c-Jun expression in U20S cells rescued cells from cell death induced by miR-216b expression and TG treatment (Fig. 7c,d). Finally, we reasoned that if miR-216b mediates CHOP-dependent apoptosis, enforced expression of miR-216b should sensitize CHOP À / À MEFs to UPR-dependent cell killing. Consistently, enforced miR-216b expression resulted in increased cleaved caspase 3 in CHOP À / À MEFs (Fig. 7e).

Discussion
Activation of the UPR under varying stress conditions triggers an alteration in cellular protein content through direct modulation of gene expression, micro-RNA biosynthesis and protein translation rates; this in turn impacts cell division and ultimately enforces cell fate determination 57 . The final outcome, cell adaptation or cell death, is influenced by both stress intensity and duration. Importantly, how stress intensity is sensed by the UPR transducers, PERK/Ire1/ATF6, and transmitted into a fate decision remains to be elucidated. The complexity of this is increased by genetic evidence revealing that both genetic and pharmacologic inhibition of anyone branch can either sensitize cells to apoptosis or delay onset of apoptosis 33,42,58 . Some of these paradoxical outcomes can be resolved when one considers either the type of stress or the cell/tissue type being stressed. However, such context still requires that downstream effectors of the primary transducers be selectively activated to achieve specific fates. The elucidation of how effectors interpret signals and regulate cell fate requires a detailed understanding of both prosurvival and pro-apoptotic signalling.
Accumulating evidence supports a role for differential regulation of micro-RNAs by the UPR in the regulation cellular fate decisions under conditions of stress. Indeed, all three UPR transducers have been associated with micro-RNA regulation [24][25][26][27] . Thus far, Ire1 activity has been generally associated with inhibition of micro-RNA accumulation through its RNase function. Targeted micro-RNAs include miRs-17, 34a, 96 and 125b (ref. 25). In contrast, PERK-dependent activities appear to be associated with both micro-RNA induction, miR- It is of interest that all of the PERK-dependent micro-RNAs have been linked to the pro-survival activities of PERK. In the work described here, miR-216b is demonstrated to accumulate via the coordinated activities of both PERK and Ire1. Strikingly, in contrast to other PERK-regulated micro-RNAs, miR-216b contributes to UPR-associated apoptosis through regulation of a newly identified target, c-Jun. Although c-Jun, a component of the AP-1 transcription factor 54 , has not previously been demonstrated to contribute to gene expression patterns associated with ER stress, activation of Jun kinase pathway (JNK), was previously linked with Ire1-dependent activation of apoptotic signalling kinase-1 (refs 59,60). Accordingly, activation of JNK is associated with increased apoptosis, a consequence of JNK-dependent inactivation of Bcl-2 (refs 61,62). Induction of miR-216b is not associated with alterations in JNK signalling; however, it also contributes to increased apoptosis through downregulation of c-Jun. miR-216b accumulation peaks several hours into the stress response, suggesting it functions as cells commit to either survival or death. This conclusion is supported by a number of experiments. First, miR-216b expression is directly dependent upon CHOP activity, another pro-apoptotic factor. Second, reduction in miR-216b activity increases cell resistance to UPRdependent cell death. Third, overexpression of a miR-216b increases apoptosis. Finally, we demonstrate that modulation of c-Jun, a bona fide target of miR-216b during ER stress, also has the expected outcome on cell fate during an ER stress response including loss increasing cell death during ER stress.
One of the more vexing issues associated with UPR-triggered apoptosis has been the CHOP transcription factor. Genetic evidence from both cell and animal models has revealed CHOP to be a pro-apoptotic transcription factor during cellular stress, whether ER stress or DNA damage 62,63 . Several mechanisms have been suggested to mediate the pro-apoptotic activity of CHOP. These include suppression of anti-apoptotic Bcl-2 expression 64 , inducing pro-apoptotic Bim and DR5 (ref. 64), increasing translation and thereby compromising cellular bio-energetics and redox homeostasis 65 . Critically, many of these activities are not only dependent upon PERK-dependent activation of CHOP expression, but also depend upon ASK1-dependent activation of p38 MAPK , which phosphorylates and induces CHOP 66 demonstrating key points of Ire1 and PERK convergence on the establishment of an apoptotic fate under high stress conditions. Our data, which demonstrates CHOP-dependent induction of miR-216b, reveal yet another point of pro-apoptotic convergence wherein miR-216b antagonizes c-Jun accumulation and thereby promotes cell death.
The collective data suggest that CHOP-dependent apoptosis is unlikely to reflect a single functional target. Rather, as with many transcription factors, it is the additive function of several downstream targets that facilitate the establishment of cell fate. In this model, CHOP serves as a focal point downstream of PERK and Ire1 action. CHOP-dependent transcription of targets such as Bcl-2, BIM and miR-216b, the latter of which then targets an increasing number of transcripts, c-Jun as one example, thereby amplifying the pro-apoptotic activity of CHOP (Fig. 7f). As the action of miR-216b is delayed, by virtue of the time needed for its induction and targeting its substrates, this latency likely provides a needed opportunity for cells to repair and adapt if possible prior to initiating apoptosis.
MicroRNA microarray and qPCR. NIH 3T3 cells were challenged with 2 mg ml À 1 Tm for 2, 5 or 10 h. Total RNA was purified using miRNeasy Mini Kit (Qiagen). Microarray was carried out using GeneChip miRNA Arrays (Affymetrix) and results were analysed using GENESPRING software. analysis of variance and t-tests were used to calculate fold change and P-values. Spotfire (Somerville) software was used to generate heat maps. For qPCR, total RNA was purified after appropriate treatments using Ambion microRNA purification kit mirVANA (Ambion) as per the manufacturer's instructions. All microarray data were deposited and can be accessed via GSE79455. RNA was reverse transcribed using specific primers with microRNA reverse transcription kit (Applied Biosystems). qPCR was performed on an Applied Biosystems 7900 qRT-PCR machine (Applied Biosystems). MicroRNA was reverse transcribed using TaqMan MicroRNA Reverse transcription kit with specific Taqman RT primers for mmu-miR-216b and miR-202 (Cat. # 4427975, Applied Biosystems) as per the manufacturer's protocol. qPCR was performed with mmu-miR-216b and miR-202 Taqman  Luciferase assays. Oligonucleotides containing three copies of the miR-216b seed sequences were cloned into the 3 0 UTR of the firefly luciferase gene in pRLTK. The 1,000 bp upstream of mmu-miR-216b promoter was amplified by PCR and cloned into pGL4.11 using primers: forward 5 0 -GATCTCGAGGTTCCTTACTGGCTT GCT-3 0 and reverse 5 0 -CTGAAGCTTTTGTCAGACTGAAGAATC5-3 0 . The mutant mmu-miR-216b promoter mutant was generated using the QuikChange kit (Stratagene). Mutagenic primers used were as follows: forward 5'-CAGCTCTCTC ACCTATATTGACCACACTGCAAAC-3'; reverse 5'-GTTTGCAGTGTGGTCA ATATAGGTGAGAGAGCTG-3'. The 3X AP-1 reporter contains three canonical AP-1-binding sites (TGACTCA) upstream of a minimal promoter fragment containing a TATA box in the luciferase reporter site luciferase 67 . For luciferase assays, cells were transfected with the constructs and pRLTK (Addgene) vector, to normalize for transfection efficiency and protein synthesis, with Lipofectamine plus method and treated with TG (500 nM) for the time as indicated.
Clonogenic survival assay. Clonogenic assays were performed in U2OS stable cells 69 . Briefly, cells were plated at 2 Â 10 3 cells per 60 mm plates. Twenty-four hours later, they were exposed to 400 nM TG or 2.5 mg ml À 1 Tm for the indicated times, then returned to growth medium. Fresh media were replenshished every 2 days. Viable colonies were stained with 3 ml of 0.5% Giemsa (v/v) for 5-10 min at room temperature at 12 days. Colony counts were normalized to plating efficiency and represented as fraction surviving compared with control (dimethylsulphoxide).