Decay of Endoplasmic Reticulum-Localized mRNAs During the Unfolded Protein Response

See allHide authors and affiliations

Science  07 Jul 2006:
Vol. 313, Issue 5783, pp. 104-107
DOI: 10.1126/science.1129631


The unfolded protein response (UPR) allows the endoplasmic reticulum (ER) to recover from the accumulation of misfolded proteins, in part by increasing its folding capacity. Inositol-requiring enzyme–1 (IRE1) promotes this remodeling by detecting misfolded ER proteins and activating a transcription factor, X-box–binding protein 1, through endonucleolytic cleavage of its messenger RNA (mRNA). Here, we report that IRE1 independently mediates the rapid degradation of a specific subset of mRNAs, based both on their localization to the ER membrane and on the amino acid sequence they encode. This response is well suited to complement other UPR mechanisms because it could selectively halt production of proteins that challenge the ER and clear the translocation and folding machinery for the subsequent remodeling process.

The ER is responsible for the structural maturation of proteins entering the secretory pathway (1). When the folding burden on the ER exceeds its capacity, a collection of transcriptional and translational mechanisms termed the UPR is initiated (2). The UPR is essential for survival during ER stress and for development in metazoans (3, 4), especially for the differentiation of secretory cells (57). A key sensor of the folding status of the ER is IRE1, a conserved transmembrane protein with an ER luminal sensor domain and cytosolic kinase and ribonuclease domains (8, 9). Accumulation of misfolded proteins in the ER leads to the autophosphorylation and activation of IRE1, which cleaves specific sites in the mRNA encoding the transcription factor X-box–binding protein 1 (XBP-1) (10, 11). This cleavage initiates an unconventional splicing reaction, leading to production of an active transcription factor. Although the sole output of IRE1 activation in yeast appears to be the splicing of the XBP-1 homolog HAC1 (12, 13), several studies in metazoans suggest that IRE1 has additional functions not mediated by XBP-1 (1416). Overexpression of IRE1 can also promote cleavage of the 28S ribosomal RNA (rRNA) (17) and the mRNA encoding IRE1 itself (18), which raises the possibility that, under some circumstances, the IRE1 nuclease can act on a broader range of substrates.

To systematically test the XBP-1 dependence of IRE1 outputs, we compared the changes in expression profiles associated with the UPR in Drosophila S2 cells depleted of either IRE1 or XBP-1 by RNA interference (RNAi) (Fig. 1, A and B). We induced the UPR with the reducing agent dithiothreitol (DTT), which prevents disulfide-linked folding, and analyzed the changes in expression of ∼5000 genes using spotted DNA microarrays (19). Among strongly induced genes, there was excellent overlap between the requirements for IRE1 and XBP-1 (Fig. 1A). Consistent with previous results (12, 15, 20, 21), the IRE1- and XBP-1–dependent induced genes were enriched for ER-related processes, including folding, glycosylation, protein trafficking, and lipid metabolism (table S1).

Fig. 1.

Expression changes associated with the UPR in Drosophila S2 cells. (A and B) Hierarchical clustering of genes (A) induced and (B) repressed (by at least a factor of 2) by DTT (2 mM, 7.4 hours). Shown is the average log2 (induced/uninduced) ratio from three independent experiments. (C) Genes from the cluster shown in (B) with the effects of depletion of UPR components. Localization information was derived from GO annotations in FlyBase (30); if no annotation was available, protein sequences were used to predict transmembrane helices and signal sequences (19). (D) Northern blots of total RNA probed with the loading control, ribosomal protein L19 (Rpl19), and either an IRE1-dependent repressed target [Tetraspanin 42Ee (Tsp42Ee) or SPARC] or the IRE1- and XBP-dependent induced target translocation protein 1 (Trp1).

In addition to this classic induction pathway, our studies revealed an unanticipated branch of the UPR in which IRE1 mediated the rapid and selective repression of ER-targeted mRNAs. Specifically, we observed a cluster of genes whose repression was dependent on IRE1 but not on XBP-1 or the other UPR components ATF6, PERK, and ATF4 (Fig. 1, B to D and fig. S1). We also observed repression of several of these targets by tunicamycin, which induces the UPR by inhibiting glycosylation (fig. S2). Within the repressed targets, there was a strong enrichment for genes encoding plasma membrane and other secreted proteins (Fig. 1C and table S2), suggesting that these proteins traffic through the ER but are not directly involved in ER function. The repression was fast compared with expression changes mediated by XBP-1, which typically displayed a lag phase of ∼2 hours (Fig. 2A). Thus, this repressive response is well-suited to relieve acute ER stress, because it would specifically prevent the translation of proteins targeted to the ER before the protective mechanisms of the XBP-1–dependent pathway could take effect (22).

Fig. 2.

IRE1-dependent repression involves rapid mRNA destabilization. (A) Changes in the RNA abundance of several targets over time after DTT (2 mM) addition. Red circles, SPARC; red triangles, Tsp42Ee; red squares, CG3488; blue circles, Akap200; blue triangles, Stat92E; blue squares, dacapo; green triangles, Trp1; and green squares, CG8286. Results are representative of time-course data measured by microarray. (B) Changes in RNA abundance over time after DTT addition, in the presence or absence of the transcription inhibitor actinomycin D. Cells received no treatment (open circles), actinomycin D (open triangles), DTT (black circles), or both (black triangles). (C) Repression of endogenous transcripts (white bars) and transcripts of transfected reporters (gray bars) containing the coding sequences of the corresponding targets under the control of the metallothionein (MT) promoter, after 4.5 hours of DTT treatment. The averages and SDs of two to four independent transfections are shown. Measurements for (A to C) were done by quantitative reverse transcription polymerase chain reaction (RT-PCR).

The observed down-regulation appears to be mediated by mRNA degradation rather than transcriptional repression. Blocking transcription with actinomycin D (Fig. 2B and fig. S3) or 5,6-dichloro-1-βd-ribofuranosylbenzimidazole (DRB) (fig. S4) had no effect on the DTT-dependent repression of the target mRNAs. Furthermore, expressing the coding sequence under the control of the copper-inducible metallothionein promoter was sufficient to allow wild-type levels of repression of three target mRNAs (Fig. 2C), ruling out a role for the natural promoters or 5′ and 3′ untranslated regions.

We next investigated whether this mRNA degradation was initiated by internal cleavage, such as one mediated by the IRE1 endonuclease. If this were the case, the resulting RNA fragments would be subject to degradation by housekeeping machinery, including XRN1 and the Ski2-3-8 complex, which are involved in cytoplasmic 5′ to 3′ and 3′ to 5′ degradation, respectively (23). Consistent with such a mechanism, depletion of XRN1 by RNAi led to the IRE1- and DTT-dependent accumulation of 3′ fragments of the target mRNAs SPARC (Fig. 3, A and B) and TMS1 (fig. S5). Similarly, depletion of Ski2 led to the IRE1- and DTT-dependent accumulation of 5′ fragments (Fig. 3, A and C, and fig. S5). Primer extension analysis of mRNA derived from XRN1-depleted cells revealed two putative cleavage sites between positions 421 and 422 (codon 104) and 320 and 321 (codon 71) of the SPARC transcript (Fig. 3D). The two sites have similar sequences (Fig. 3E) and are consistent with the sizes of the 5′ SPARC fragments seen in the Ski2 depletion, which suggests that the fragments were generated by two distinct endonucleolytic cleavage events.

Fig. 3.

mRNA decay proceeds through endonucleolytic cleavage. (A) Northern blot of total RNA derived from S2 cells depleted of IRE1, XRN1, and/or Ski2, and either untreated or treated with DTT (2 mM, 1.5 hours), hybridized with a probe complementary to the 5′ UTR and coding sequence of SPARC. The triangle indicates the 3′ mRNA fragment, and asterisks indicate the 5′ fragments. (B and C) Northern blots as in (A) hybridized with probes complementary to the 5′ or 3′ end of the SPARC transcript. Numbers refer to the nucleotide position within the transcript (total length, 1170 nucleotides). (D) Primer extension analysis of poly(A)+ RNA isolated from cells depleted of XRN1 and either untreated or treated with DTT as in (A and B), using a radiolabeled primer for SPARC. The two DTT-specific bands (positions T321 and T422 in the transcript) were seen in two independent experiments. (E) Sequences surrounding the sites identified in (D).

Exploration of the cis elements necessary for targeting of mRNAs to this decay pathway revealed a critical role for an ER-targeting signal sequence. Deletion of the region encoding residues 1 to 40 in a reporter expressing SPARC abolished DTT-dependent degradation (Fig. 4A). This defect was due to the removal of the N-terminal signal sequence, because replacement of these residues with the signal sequence of BiP, an ER chaperone that is not subject to IRE1-mediated degradation, completely restored the regulation (Fig. 4A). Replacement of three hydrophobic residues in the SPARC signal sequence with charged residues also abolished DTT regulation, whereas proximal silent mutations had no effect (Fig. 4B and fig. S6). Examination of the effects of these mutations on localization of the protein, using C-terminal green fluorescent protein (GFP) fusions to the first 138 amino acids of SPARC [SPARC(Δ139–304)], confirmed that the charged, but not the silent, mutations disrupted signal sequence function (Fig. 4C). Similar results were seen with a second target, CG9917 (Fig. 4B), which suggested that localization of the mRNA and nascent chain to the ER membrane is an important aspect of the degradation process. Apart from the signal sequence, it appears that cis elements within the SPARC mRNA are distributed and redundant, such that no discrete loss of regulation was seen on removal of individual RNA sequence elements (Fig. 4A and SOM Text).

Fig. 4.

Analysis of the cis elements required for IRE1-mediated mRNA decay. (A) Repression of the endogenous SPARC transcript (endog.) and of reporter constructs, in cells treated with DTT (2 mM, 4.5 hours). Reporters express the coding sequence of SPARC with the indicated codons deleted (FL, full-length, 304 amino acids). The Δ1–40+BiP ss (signal sequence) construct contains the coding sequences of BiP (amino acids 1 to 23) followed by SPARC (amino acids 41 to 304). Reporter repression in (A and B) was measured by real-time quantitative PCR; shown are the averages and SDs of three to five independent transfections. (B) Repression of reporter constructs in cells treated as in (A). Reporters express the FL coding sequence of either SPARC or CG9917, with either the wild-type signal sequence (wt ss) or with mutated signal sequences. Charged mut constructs replace three or four signal sequence residues with charged residues; silent mut constructs contain silent mutations at similar sites (see fig. S6). (C) Images of live S2 cells expressing SPARC(Δ139–304)–GFP with the signal sequence mutations in (B). All reporters were regulated similarly to their full-length, untagged counterparts. Scale bar, 10 μm. (D) Northern blots of SPARC(Δ139–304)-GFP with and without insertions of one to three nucleotides after the indicated amino acid. Numbers below refer to the average (± SD) repression of these constructs in three to five independent transfections, measured by quantitative RT-PCR. (E) Images of cells expressing reporters from (D). Scale is the same as in (C).

The requirement for a functional signal sequence suggests that IRE1-mediated mRNA decay occurs during cotranslational translocation. This led us to ask whether the nascent protein was involved in transcript degradation. To address this possibility, we took advantage of the fact that for the SPARC(Δ139–304) reporter, insertion of one nucleotide after codon 63 or two nucleotides after codon 88 gave rise to frame-shifted messages that fortuitously lacked downstream stop codons. We fused these to the GFP coding sequence in the appropriate frame. Despite the fact that these constructs expressed ER-targeted proteins (Fig. 4E), the presence of the frame-shift mutations rendered the transcripts largely resistant to IRE1-dependent degradation (Fig. 4D and fig. S7). This effect was unlikely to be caused by the disruption of RNA sequence or structure, because constructs with three nucleotide insertions retained their regulation (Fig. 4D). Thus, it appears that, in addition to being necessary for targeting to the ER membrane, correct translation of the polypeptide plays an important role in the regulation of this mRNA.

The features of the IRE1-mediated mRNA decay pathway described here are interesting in light of the known properties of IRE1. Signal sequence–mediated targeting of the messages would bring them to the cytosolic face of the ER membrane, perhaps in proximity to IRE1's nuclease domain. Because degradation appears to be initiated by an endonucleolytic cleavage, IRE1 may act directly on the target messages, showing a reduced specificity similar to its closest homolog, ribonuclease (RNase) L (24). Alternatively, IRE1 may rapidly recruit or activate a second ribonuclease or may promote translational stalling and cleavage by no-go decay (25). More speculatively, the requirement for in-frame translation in the degradation of these mRNAs suggests that there is an active discrimination among translocating polypeptides, based on their physical properties or propensity for translational stalling. For example, binding of translocating peptides to activated IRE1 via its luminal hydrophobic cleft (26) could mediate the recruitment of translating ribosomes, which would in turn expose the mRNAs to the nuclease domain of IRE1. A targeting mechanism based on direct peptide recognition or translational effects would potentially allow IRE1 to focus on messages that present the most immediate challenge to the translocation and folding machinery.

How might the mRNA degradation pathway described here complement the previously characterized transcriptional and translational branches of the UPR in allowing for a coherent response to misfolded proteins? Under stress conditions, the burden on the ER exceeds its capacity to fold proteins, yet the solution, to synthesize more ER folding machinery, will temporarily add to that burden. Our studies reveal that IRE1 initiates two distinct cascades. The rapid, XBP-1–independent pathway, together with the generalized translational inhibition mediated by PERK (27), can immediately relieve the burden on the ER and partially clear the translocation machinery, freeing up resources for the subsequent up-regulation of genes directly involved in folding and trafficking by the slower, IRE1- and XBP-1–dependent pathway. More broadly, given the increasing evidence for large-scale intracellular localization of mRNAs (28), the coupling of RNA degradation to local stimuli may prove to be an efficient and widely used strategy for rapidly altering the local distribution of proteins.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S7

Tables S1 and S2


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article