The Sec61 translocon limits IRE1α signaling during the unfolded protein response

IRE1α is an endoplasmic reticulum (ER) localized endonuclease activated by misfolded proteins in the ER. Previously, we demonstrated that IRE1α forms a complex with the Sec61 translocon, to which its substrate XBP1u mRNA is recruited for cleavage during ER stress (Plumb et al., 2015). Here, we probe IRE1α complexes in cells with blue native PAGE immunoblotting. We find that IRE1α forms a hetero-oligomeric complex with the Sec61 translocon that is activated upon ER stress with little change in the complex. In addition, IRE1α oligomerization, activation, and inactivation during ER stress are regulated by Sec61. Loss of the IRE1α-Sec61 translocon interaction as well as severe ER stress conditions causes IRE1α to form higher-order oligomers that exhibit continuous activation and extended cleavage of XBP1u mRNA. Thus, we propose that the Sec61-IRE1α complex defines the extent of IRE1α activity and may determine cell fate decisions during ER stress conditions. DOI: http://dx.doi.org/10.7554/eLife.27187.001


Introduction
The majority of secretory and membrane proteins enter the endoplasmic reticulum (ER) through the Sec61 protein translocation channel (Rapoport, 2007). In the ER, folding enzymes and chaperones facilitate maturation of newly synthesized proteins. Proteins that fail to achieve their folded state are eliminated by ER-associated quality control pathways (ERAD) (Brodsky, 2012), while correctly folded proteins are transported to their intra or extracellular site of activity. When the influx of proteins exceeds the ER protein folding and quality control capacity, misfolded proteins accumulate in the ER leading to a condition known as ER stress. During ER stress, signaling pathways, collectively termed the Unfolded Protein Response (UPR), are activated in order to upregulate chaperones and folding enzymes, to reduce the influx of proteins into the ER, and to increase the capacity for ERassociated degradation (Walter and Ron, 2011). In this way, the UPR adapts cells to ER stress conditions and restores ER homeostasis. However, the UPR can also trigger apoptosis during chronic or severe ER stress conditions, suggesting that UPR activity is tightly controlled in order to elicit the appropriate cellular response, whether pro-adaptive or pro-apoptotic (Hetz, 2012). Indeed, inappropriate activation of UPR signaling is linked to a number of disease states, including pancreatic beta cell death in diabetes (Back and Kaufman, 2012) and neuronal cell death in certain neurodegenerative diseases (Wang and Kaufman, 2016).
Three transmembrane sensors, IRE1a, PERK, and ATF6, mediate the UPR. Upon ER stress, all three sensors become activated by changes in their oligomerization state. The most ancient UPR sensor is IRE1a, a transmembrane endonuclease/kinase that senses the accumulation of misfolded proteins in the ER lumen (Cox et al., 1993;Mori et al., 1993). When ER misfolded proteins are detected, IRE1a self-oligomerizes through its luminal domains. This, in turn, leads to cytosolic transautophosphorylation of the IRE1a kinase domain and subsequent activation of its RNase domain. The activated IRE1a restores the ER folding capacity by cleaving XBP1u mRNA (u; unspliced) to initiate splicing on the ER membrane (Yoshida et al., 2001;Calfon et al., 2002). Efficient cleavage of XBP1u mRNA requires an interaction between IRE1a and the Sec61 translocon as well as the SRP pathway-mediated recruitment of XBP1u mRNA to the Sec61 translocon (Plumb et al., 2015;Kanda et al., 2016). Subsequently, the cleaved fragments of XBP1 mRNA are ligated by the RtcB tRNA ligase (Lu et al., 2014;Jurkin et al., 2014;Kosmaczewski et al., 2014) with its co-factor archease (Poothong et al., 2017). The spliced XBP1 mRNA is translated into an active transcription factor, XBP1s, which induces UPR genes to alleviate ER stress (Lee et al., 2003;Acosta-Alvear et al., 2007). In addition, IRE1a also promiscuously cleaves ER-localized mRNAs including mRNAs encoding secretory and membrane proteins, a process known as IRE1a-dependent mRNA decay (RIDD) (Hollien and Weissman, 2006;Hollien et al., 2009;Han et al., 2009). RIDD is implicated in reducing the incoming protein burden on the ER during stress conditions as well as in mediating cell death (Hollien and Weissman, 2006;Ghosh et al., 2014;Tam et al., 2014).
Since the continuous activation of IRE1a is associated with cell death, the activation and inactivation of IRE1a must be properly regulated. Indeed, seminal studies from Peter Walter's group demonstrated that IRE1a activity is temporally and quantitatively attenuated during ER stress conditions (Lin et al., 2007). However, the mechanism by which IRE1a is inactivated in the presence of ER stress is unclear. Previous studies have provided important insights into how IRE1a activity can be regulated by its associated proteins (Bertolotti et al., 2000;Okamura et al., 2000;Lisbona et al., 2009;Eletto et al., 2014;Carrara et al., 2015;Morita et al., 2017). Interestingly, factors such as BiP and PDIA6 that are implicated in attenuating IRE1a activity also interact with PERK, which, in contrast to IRE1a, remains activated during prolonged ER stress conditions (Lin et al., 2007). We therefore tested the role of the Sec61 translocon in regulating IRE1a activity because it selectively interacts with IRE1a but not with the other ER stress sensors PERK, ATF6 and Ire1b (Plumb et al., 2015). We have used a Blue Native polyacrylamide gel electrophoresis (BN-PAGE) immunoblotting procedure to probe IRE1a complexes in cells during normal and ER stress conditions. Our studies reveal that IRE1a exists as preassembled hetero-oligomeric complexes with the Sec61 translocon and becomes activated during ER stress conditions with minor changes to its complexes. We find that the Sec61 translocon limits IRE1a oligomerization and thereby controls activation and inactivation of IRE1a activity during ER stress conditions. Indeed, either the loss of the IRE1a interaction with the Sec61 translocon or severe stress causes IRE1a to form higher-order oligomers that exhibit continuous activation of IRE1a and extended cleavage of XBP1u mRNA. Thus, our studies suggest that the IRE1a-Sec61 complex plays a critical role in controlling IRE1a signaling during ER stress.

IRE1a forms hetero-oligomeric complexes with the Sec61 translocon
We hypothesized that the Sec61 translocon may limit IRE1a oligomerization during ER stress and thus control IRE1a activity because of the following observations. First, our previous studies showed that nearly all the endogenous IRE1a is bound with the Sec61 translocon in the ER membrane during normal and ER stress conditions (Plumb et al., 2015). Second, the concentration of the Sec61 translocon vastly outnumbers the concentration of IRE1a in the ER (Plumb et al., 2015;Kulak et al., 2014), suggesting that it could provide a barrier to IRE1a oligomerization. To test this hypothesis, we searched for IRE1a mutants that either disrupt or increase the interaction with the Sec61 translocon. Our previous studies identified a ten amino acid region in the luminal domain proximal to the transmembrane domain of IRE1a that when deleted nearly abolished the interaction with the Sec61 translocon ( Figure 1-figure supplement 1A,B). We refer to the IRE1a D434-443 mutant as weakly interacting IRE1a or wIRE1a. Fortuitously, our previous mutagenesis studies also revealed that IRE1a S439A showed an increased binding to the Sec61 translocon. We then further significantly improved the interaction between IRE1a and Sec61 by combining S439A with the mutation of three hydrophilic residues in the transmembrane domain of IRE1a (Figure 1-figure supplement 1A,B) (Sun et al., 2015). We refer to this mutant (IRE1a S439A/T446A/S450A/T451A) as strongly interacting IRE1a (sIRE1a).
To investigate the role of the Sec61 translocon in regulating IRE1a oligomerization and activity, we complemented IRE1a, wIRE1a or sIRE1a into IRE1a-/-HEK 293 Flip-In T-Rex cells generated by CRISPR/Cas9 (Mali et al., 2013;Plumb et al., 2015). IRE1a expression is controlled by the tetracycline promoter in these complemented cells. Low expression levels as well as ER stress dependent activation of IRE1a were achieved through leaky expression in the absence of doxycycline ( Figure 1A; Figure 1-figure supplement 1C). To examine the oligomerization status of IRE1a in these different cells, we employed a BN-PAGE based immunoblotting procedure. This technique allows separation of large membrane protein complexes with minimal perturbation of native complexes using Coomassie G250 dye as the charged ion carrier (Wittig et al., 2006). The cells were treated with or without thapsigargin, which induces ER stress by inhibiting calcium import into the ER lumen, and analyzed by BN-PAGE immunoblotting. Surprisingly, BN-PAGE analysis of IRE1a complemented cells showed two forms of preassembled IRE1a complexes. Form A corresponds to ã 500 kDa complex, and Form B corresponds to a~720 kDa complex ( Figure 1A). Intriguingly, upon ER stress treatment, IRE1a Form B slightly increased in intensity, while IRE1a Form A was reduced. Probing phosphorylated IRE1a using the phos-tag reagent (Yang et al., 2010) further confirmed that IRE1a was activated upon ER stress as shown by stress dependent detection of phosphorylated IRE1a ( Figure 1A). To next determine the role of the Sec61 translocon in controlling IRE1a oligomerization, we performed BN-PAGE analysis with cells expressing either wIRE1a, which cannot interact with Sec61, or sIRE1a, which interacts strongly with Sec61 ( Figure 1-figure supplement 1B). In comparison to the wild-type IRE1a, wIRE1a predominantly existed in the Form B complex, whereas sIRE1a showed significantly more of the Form A ( Figure 1A). Unlike the wild type, the stress-dependent changes were less obvious for both wIRE1a and sIRE1a oligomers, but they were clearly activated as shown by their phosphorylation using phos-tag based immunoblotting ( Figure 1A).
Since we did not observe a significant change in IRE1a complexes upon ER stress, we asked if this result was due to a limitation of BN-PAGE to detect changes in IRE1a complexes. To examine this, we performed a BN-PAGE analysis of PERK, the luminal domain of which is structurally similar, and even interchangeable with IRE1a (Liu et al., 2000), but does not interact with Sec61 (Plumb et al., 2015). Similar to IRE1a, PERK existed as a preformed complex, though of~900 kDa, in cells under normal conditions. However, upon stress, PERK became a~1200 kDa complex ( Figure 1B). These results were recapitulated in HEK293 and insulin secreting rat pancreatic betacells (INS-1) treated with ER stress. Here, the endogenous IRE1a again presented as approximately 500 and 720 kDa complexes that changed little during ER stress conditions, while PERK exhibited a significant ER stress-dependent shift in complex size ( Figure 1-figure supplement 2).
We hypothesized that if the Sec61 translocon controls oligomerization of IRE1a, its depletion in cells should resemble wIRE1a, which exhibited predominantly~720 kDa complexes on BN-PAGE. Such a result would suggest that the mutation in wIRE1a does not cause secondary effects in IRE1a independent of the Sec61 translocon interaction disruption. To test this, we depleted the Sec61 translocon by treating cells with siRNA oligos against Sec61a and performed BN-PAGE analysis. Remarkably, wild-type IRE1a resembled wIRE1a in the Sec61 translocon depleted cells, as the 500 kDa Form A shifted to the 720 kDa Form B ( Figure 1C). In contrast, the Sec61 translocon depletion had little effect in cells expressing wIRE1a, which remained in Form B. Consistent with recent findings (Adamson et al., 2016), depletion of the Sec61 translocon partially activated IRE1a as shown by a slight increase in self-phosphorylation in the absence of ER stress compared to control siRNA treated cells. However, an efficient activation of IRE1a in these cells typically required treatment with the ER stress inducer thapsigargin ( Figure 1C). Intriguingly, the depletion of the Sec61 translocon specifically affected IRE1a complexes, as PERK complexes were less disrupted in Sec61 depleted cells relative to the control siRNA-depleted cells ( Figure 1D). To determine if the Sec61 translocon co-migrates with IRE1a complexes, we performed BN-PAGE immunoblotting with Sec61a antibodies. The Sec61 translocon, which is composed of a, b, and g subunits, ran predominantly as a~146 kDa form and a minor~350 kDa form on BN-PAGE ( Figure 1E, Figure 1-figure supplement 3), which is consistent with previous studies (Conti et al., 2015). Currently, it is unclear why we were not able to detect Sec61 co-migration with IRE1a, though it is likely that only a small population of the highly abundant Sec61 exists in a complex with IRE1a in cells. Collectively, these results suggest that IRE1a complexes in cells are regulated by an interaction with the Sec61 translocon. , or sIRE1a-HA (S439A/T446A/S450A/T451A) were treated with 2.5 mg/ml thapsigargin (Tg) for the indicated hours (hr), lysed with digitonin, and analyzed by BN-PAGE immunoblotting (top) as well as phos-tag based immunoblotting to probe phosphorylated IRE1a (bottom). A denotes a~500 kDa complex of IRE1a in BN-PAGE immunoblotting. B denotes a~720 kDa complex of IRE1a. (B) The cells expressing IRE1a-HA or wIRE1a-HA were treated with 2.5 ug/ml Tg for the indicated hours and analyzed by both BN-PAGE immunoblotting and standard immunoblotting with a PERK antibody. (C) IRE1a-HA or wIRE1a-HA expressing cells were treated with either control siRNA or Sec61a siRNA followed by treatment with 2.5 Figure 1 continued on next page We next sought to determine whether the Sec61 translocon co-migrates with the different complexes of IRE1a by performing BN-PAGE with the purified Sec61-IRE1a complex. To achieve this, we established stable cell lines expressing IRE1a and purified the IRE1a and Sec61 complex through a combination of affinity and ion exchange chromatography using digitonin, which preserves the interaction between IRE1a and the Sec61 translocon. The coomassie blue stained gel revealed that purified IRE1a associated with the Sec61 translocon and Sec63, a component of the translocon complex ( Figure 2A) (Meyer et al., 2000). As expected, wIRE1a lacked the Sec61 translocon complex, whereas sIRE1a associated with the Sec61 translocon complex ( Figure 2A). All three IRE1a proteins had a similar ability to cleave in vitro transcribed XBP1u mRNA substrate, though wIRE1a and sIRE1a showed slightly slower kinetics of cleavage (Figure 2-figure supplement 1).
We then analyzed these purified proteins by BN-PAGE immunoblotting to determine if the Sec61 translocon co-migrates with different IRE1a complexes. Similar to the IRE1a complexes in cells, purified IRE1a existed as complexes of both Form A and Form B when it associated with the Sec61 translocon ( Figure 2B). In contrast, the purified wIRE1a existed predominantly as Form B and as ã 240 kDa complex. The 240 kDa form of IRE1a was not obvious in cells, suggesting that IRE1a complexes may be labile during the purification procedure. sIRE1a closely resembled the wild-type IRE1a complexes because both purified IRE1a and sIRE1a proteins contained similarly enriched Sec61 translocon complex ( Figure 2C). Remarkably, BN-PAGE analysis with Sec61a antibodies revealed that Sec61 co-migrates with both Form A and Form B in purified IRE1a and sIRE1a ( Figure 2C). In contrast, Sec61a was not detectable in BN-PAGE with the purified wIRE1a. At present, the role of BiP, which is known to interact and inhibit IRE1a oligomerization (Bertolotti et al., 2000;Okamura et al., 2000;Oikawa et al., 2009;Carrara et al., 2015), in the Sec61 transloconmediated regulation of IRE1a complexes is unclear, since we could not detect BiP in our purified IRE1a complexes ( Figure 2D). Nevertheless, our results with purified IRE1a proteins are consistent with the results derived from cells. We find that IRE1a and sIRE1a exist in Forms A and B with the Sec61 translocon, while wIRE1a is predominantly in Form B but without the Sec61 translocon. Although further work is required to determine the precise copy numbers of IRE1a in these complexes, our data suggest that Sec61 is an intrinsic part of the IRE1a complexes under normal and ER stress conditions.

The Sec61 translocon inhibits formation of higher order IRE1a oligomeric clusters in cells
We next asked whether the Sec61 translocon-mediated regulation of IRE1a oligomerization can be observed by immunofluorescence. Previous studies reported that IRE1a forms higher-order oligomers or clusters upon ER stress, which correlate with IRE1a RNase activity and are proposed to be important for IRE1a signaling (Li et al., 2010). To determine whether the Sec61 translocon mediates regulation of IRE1a oligomerization, we looked for ER stress-dependent changes in IRE1a oligomerization in IRE1a-/-HEK293 cells complemented with IRE1a variants containing a C-terminal HA tag to facilitate immunostaining. Under normal conditions, IRE1a and wIRE1a were diffusely distributed in the ER membrane and colocalized with Sec61b, a subunit of the Sec61 translocon (  type IRE1a, we failed to observe clusters in sIRE1a expressing cells ( Figure 3A), supporting the idea that the IRE1a interaction with the Sec61 translocon limits cluster formation. wIRE1a clustering was not dependent on cell types since we obtained similar results when we analyzed IRE1a-/-mouse embryonic fibroblast (MEF) cells complemented with either wild-type or wIRE1a (Figure 3-figure supplement 2A). In addition, ER stress-mediated clusters of wIRE1a were not unique to this particular wIRE1a mutant, which has a ten amino acid deletion in IRE1a, but were also observed in cells expressing a wIRE1a mutant where two critical residues are mutated within the ten amino acid region (Figure 3-figure supplement 2B). Only after increasing the expression level of IRE1a using doxycycline, could we detect a small percentage of clusters in wild-type IRE1a expressing cells ( Figure 3B,C,D). In contrast, we detected robust wIRE1a clusters in ER stress treated cells even at low expression levels. Together, these results suggest that the Sec61 translocon interaction prevents Source data 1. Doxycycline titration and quantification of IRE1a clusters as described Figure 3C. the formation of IRE1a higher order oligomers or clusters in cells. In contrast with previous work (Li et al., 2010;Ghosh et al., 2014), we observe only a low percentage of cells containing wild-type IRE1a clusters. This difference may be due to the intensity of ER stress applied to monitor IRE1a clusters in cells. Nevertheless, the differences we observe between wild-type IRE1a and wIRE1a indicate that the Sec61 translocon inhibits the formation of these higher-order oligomers or clusters.

Proper activation of IRE1a relies on the interaction between IRE1a and the Sec61 translocon
Since wIRE1a robustly formed higher order oligomeric clusters under ER stress conditions, we predicted that it may be more quickly activated than wild-type IRE1a. To test this idea, we gradually increased the expression level of IRE1a by titrating the concentration of doxycycline and assayed for activation by probing for IRE1a phosphorylation ( Figure 4A,B). Consistent with previous findings, overexpressed IRE1a was partially activated as shown by phosphorylation even in the absence of ER stress (Li et al., 2010). Overexpressed wIRE1a exhibited an even larger amount of auto-phosphorylation and thus activation compared to wild-type IRE1a, while overexpressed sIRE1a showed The following source data is available for figure 4: Source data 1. Doxycycline titration and activation of IRE1a, wIRE1a or sIRE1a as described Figure 4B. DOI: 10.7554/eLife.27187.013 Source data 2. Activation of IRE1a, wIRE1a or sIRE1a in Tg-treated cells as described Figure 4D. DOI: 10.7554/eLife.27187.014 reduced auto-phosphorylation compared to wild-type IRE1a. Interestingly, all IRE1a variants required ER stress treatment, in this case thapsigargin, to achieve a full activation state, suggesting that the accumulation of misfolded proteins plays a major role in IRE1a activation ( Figure 4A,B). We next tested the role of the Sec61 translocon in IRE1a activation during ER stress treatment. Consistently, wIRE1a was more quickly activated as shown by auto-phosphorylation compared to the wild type, whereas sIRE1a was activated at slower rate during ER stress ( Figure 4C,D). As a control, we probed for the activation of PERK, which was activated similarly in all three IRE1a variants expressing cells. Taken together, our results suggest that the proper activation of IRE1a relies on an interaction with the Sec61 translocon.
The attenuation of IRE1a signaling requires an interaction with the Sec61 translocon We reasoned that higher order oligomers and clusters of IRE1a formed by disrupting the IRE1a-Sec61 translocon interaction might be altering the inactivation rate of IRE1a during ER stress. Therefore, we compared ER stress-induced inactivation of IRE1a and wIRE1a by probing for IRE1a phosphorylation. IRE1a was fully activated after two hours of ER stress treatment as demonstrated by all IRE1a shifting to the phosphorylated state ( Figure 5A,B). Spliced XBP1 (XBP1s) protein production peaked at five hours. During prolonged stress, IRE1a was gradually inactivated with a concomitant reduction in the production of spliced XBP1 protein (XBP1s) ( Figure 5A). Unlike IRE1a, PERK was activated through the duration of the stress period. This is consistent with previous studies which showed that IRE1a-mediated XBP1u mRNA splicing diminished within a few hours of stress despite the continuation of the ER stress treatment (Lin et al., 2007). In sharp contrast to wild-type IRE1a, the Sec61 interaction-defective mutant, wIRE1a, showed significantly reduced inactivation as well as extended production of XBP1s during prolonged ER stress ( Figure 5A,B). A similar difference in IRE1a and wIRE1a phosphorylation was observed with tunicamycin ( Figure 5C,D). Here, IRE1a was nearly completely inactivated, but wIRE1a was only partially inactivated during prolonged stress. The temporal inactivation of IRE1a during ER stress was not specific to the complemented recombinant IRE1a since we obtained a similar result with the endogenous IRE1a in HEK293 cells ( Figure 5figure supplement 1). Consistent with our previous work, under ER stress treatment conditions when both IRE1a and wIRE1a are equally activated, wIRE1a cells produced less XBP1s protein ( Figure 5A,C and five hour treatment) since the lack of the Sec61 translocon interaction prevented efficient XBP1u mRNA cleavage (Plumb et al., 2015). We therefore wondered whether the slow attenuation observed in wIRE1a expressing cells was due to decreased XBP1s production, which could cause a reduction in ER chaperone production. However, we found that production of XBP1sinduced proteins, such as BiP and the Sec61 translocon, were similar in both IRE1a and wIRE1a expressing cells ( Figure 5A,C). To further confirm that XBP1s levels were not causing the observed phenotype, we overexpressed XBP1s by transfecting an XBP1s expressing plasmid into both IRE1a and wIRE1a expressing cells. Despite the overexpression of XBP1s, wIRE1a was still attenuated significantly slower than wild-type IRE1a ( Figure 5E,F). We predicted that if the Sec61 translocon promotes IRE1a inactivation, sIRE1a, which interacts strongly with Sec61, should be more quickly inactivated than the wild type IRE1a. Indeed, sIRE1a showed a faster inactivation rate during prolonged ER stress conditions ( Figure 5G,H). Finally, we asked whether the presence of misfolded proteins in the ER is required for the continuous activation of wIRE1a during ER stress. Halting protein synthesis after removing ER stress allowed for complete inactivation of wIRE1a similar to IRE1a and PERK ( Figure 5-figure supplement 2). This result implies that the presence of misfolded proteins in the ER is required for the continuous activation of wIRE1a. Together, these results indicate that an efficient inactivation of IRE1a requires the IRE1a interaction with the Sec61 translocon.

Severe ER stress induces clusters and extended activation of wild-type IRE1a
Our results suggested that the Sec61 translocon limits IRE1a oligomerization and thereby controls activation and inactivation of IRE1a during ER stress. Therefore, we hypothesized that severe ER stress may overcome this restriction and induce higher-order oligomers as well as extended activation of IRE1a in wild-type IRE1a expressing cells, similar to that observed with wIRE1a. To test this, we examined IRE1a cluster formation after increasing the intensity of ER stress by adding four-fold more thapsigargin. This high concentration of thapsigargin, but not a lower concentration, induced clusters in IRE1a, wIRE1a, and sIRE1a expressing cells, though a higher percentage of wIRE1a cells presented clusters than wild-type IRE1a or sIRE1a cells ( Figure 6A,B). These results suggest that the Sec61-IRE1a complex plays a role in limiting IRE1a oligomerization under ER stress conditions, but increased misfolded protein accumulation during severe ER stress conditions overcomes the Sec61 translocon-mediated restriction of IRE1a oligomerization. Interestingly, the interaction between IRE1a and the Sec61 translocon was little changed during both medial and severe ER stress conditions ( Figure 6-figure supplement 1). We therefore hypothesized that the Sec61 translocon might be clustering with IRE1a during severe ER stress conditions. Consistent with our hypothesis, confocal imaging revealed that the endogenous Sec61 translocon co-localized with IRE1a clusters in both wild type IRE1a and sIRE1a expressing cells. However, wIRE1a clusters appear to lack the Sec61 translocon ( Figure 6-figure supplement 2).
Since wild type IRE1a resembles wIRE1a in forming higher order oligomers or clusters during severe ER stress conditions, we predicted that wild-type IRE1a deactivation might also resemble wIRE1a under such conditions. Indeed, the attenuation of wild-type IRE1a during severe stress was significantly delayed compared to less severe stress. Thus, the production of spliced XBP1 mRNA and its protein were continued ( Figure 6C,D and E). These results suggest that once IRE1a forms higher oligomers, due to either a defect in the interaction with Sec61 or under severe ER stress, it becomes resistant to inactivation.

Discussion
In this study, we addressed the question of how IRE1a activity is regulated during ER stress conditions. We find that IRE1a oligomerization and RNAse activity are limited by the Sec61 translocon during normal and remedial ER stress levels, but that severe ER stress overcomes this block, resulting in prolonged IRE1a activation. Our results point to an important role for the IRE1a-Sec61 complex in measuring ER stress levels and accordingly tuning IRE1a activity, which may determine cell fate during ER stress.
To determine the role of the Sec61 translocon in regulating IRE1a oligomerization in cells under normal and ER stress conditions, we employed a BN-PAGE immunoblotting protocol. To our surprise, we found that IRE1a appears to be in preassembled complexes during steady-state conditions. Upon ER stress, the IRE1a complexes showed little change, albeit the intensity of Form B slightly increased with stress. This result suggests that IRE1a activation is most likely caused by a conformational change induced within the preformed IRE1a complexes by binding with misfolded proteins in the lumen. In contrast, changes in the PERK complex were conspicuous upon ER stress, Source data 1. Attenuation of IRE1a and wIRE1a in Tg-treated cells as described in Figure 5B. DOI: 10.7554/eLife.27187.016 Source data 2. Attenuation of IRE1a and wIRE1a in TM-treated cells as described Figure 5D. DOI: 10.7554/eLife.27187.017 Source data 3. Attenuation of IRE1a and wIRE1a in XBP1s expressing cells as described Figure 5F. DOI: 10.7554/eLife.27187.018 Source data 4. Attenuation of IRE1a and sIRE1a in Tg-treated cells as described in Figure 5H.  Figure 6. Severe ER stress causes higher-order oligomer formation and extended activation of wild type IRE1a. (A) IRE1a-HA, sIRE1a-HA, and wIRE1a-HA complemented IRE1a -/-HEK293 cells were treated with 2.5 mg/ml Tg for 4 hr or 10 mg/ml Tg for 2 hr. Subsequently, cells were processed using an immunostaining procedure to label IRE1a (green) with rabbit anti-HA as well as a Hoechst stain to label nuclei (blue). Scale bars are 10 mm. (B) Images from A were analyzed to determine the number of cells containing IRE1a, wIRE1a clusters, or sIRE1a clusters. Error bar represents S.E.M. (C) IRE1a-HA or wIRE1a-HA expressing cells were treated with 10 mg/ml Tg for the indicated time points and analyzed by phos-tag immunoblotting for IRE1a and standard immunoblotting for the indicated antigens. (D) Quantification of IRE1a and wIRE1a phosphorylation from panel C. (E) IRE1a or wIRE1a expressing cells were treated with either 2.5 mg/ml Tg or 10 mg/ml Tg for 18 hr and analyzed by immunoblots as well as the XBP1 mRNA splicing assay. XBP1u -Unspliced XBP1 mRNA, XBP1s -spliced XBP1 mRNA. DOI: 10.7554/eLife.27187.022 The following source data and figure supplements are available for figure 6: Source data 1. Quantification of IRE1a clusters under sever stress as described Figure 6B. DOI: 10.7554/eLife.27187.023 Source data 2. Attenuation of IRE1a or wIRE1a under severe stress as described Figure 6D. as it moves from~720 kDa to~1200 kDa in size. These results led us to wonder what advantage preassembled complexes of IRE1a might have over ER stress-induced IRE1a oligomers. We propose that the extreme low-abundance of IRE1a (Plumb et al., 2015;Kulak et al., 2014) might result in a very slow rate of oligomer formation and activation. Thus, preassembled IRE1a complexes may be essential for the rapid and robust IRE1a activation observed in cells. Future work is required to determine how many IRE1a molecules are present in each complex of IRE1a on BN-PAGE.
We next investigated higher-order oligomerization of IRE1a by examining cluster formation during ER stress. It has been reported that IRE1a forms clusters upon ER stress that correspond to higher-order oligomers (Li et al., 2010). In accordance with previous reports, we find that wild-type IRE1a forms clusters, though we only observe significant cluster formation under severe ER stress conditions. In contrast, wIRE1a formed robust clusters during remediable ER stress conditions and exhibited a higher percentage of clusters than IRE1a during severe stress conditions. These data suggest that the Sec61 translocon limits IRE1a cluster formation and that the preassembled complexes of wIRE1a may collide and form clusters rapidly in the absence of the Sec61 interaction. Through this method, we observed large, ER stress dependent, IRE1a clusters that were not captured in our BN-PAGE assay. The precise reason for this is not well understood at this point, although we cannot exclude the limitation of BN-PAGE in detecting the transient and highly dynamic nature of higher-order oligomers or clusters of IRE1a (Li et al., 2010).
Severe ER stress drastically increases cluster formation in cells expressing wild-type IRE1a, suggesting that the Sec61 translocon-mediated restriction of IRE1a oligomerization may be overcome under these conditions. Exactly how severe stress precisely tempers the Sec61 translocon barrier warrants further investigation. One potential explanation is that severe ER stress increases the number of misfolded polypeptides in the ER, which may overcome the Sec61 barrier and drive the IRE1a and Sec61 complexes into clusters. This is supported by previous studies that indicate misfolded proteins can directly bind and activate yeast IRE1a (Kimata et al., 2007;Gardner and Walter, 2011), though the similar evidence is currently lacking in metazoans (Oikawa et al., 2012). Future structural and biochemical studies are needed to understand how the Sec61 translocon is precisely arranged with IRE1a to prevent IRE1a oligomerization and how this barrier is overcome under severe ER stress conditions.
Our studies also revealed that IRE1a interaction with Sec61 might be necessary to prevent inappropriate activation during physiological low levels of stress. This is apparent with wIRE1a, where a small population is constitutively activated in the absence of stress, and overall it presents increased ER stress sensitivity and exhibits prolonged activity. The constitutive activation of wIRE1a under basal conditions is consistent with the recent findings that IRE1a signaling is activated upon depletion of the Sec61 translocon in cells (Adamson et al., 2016). These findings fit with the intriguing model proposed where IRE1a may sense the Sec61 translocon level and accordingly upregulate Sec61 genes by cleaving XBP1u mRNA (Adamson et al., 2016). However, it remains to be understood how the low abundant IRE1a becomes activated by subtle quantity changes in vastly more abundant Sec61 translocon.
At present, the role of BiP in the Sec61 mediated regulation of IRE1a oligomerization is unclear. Similar to wIRE1a, earlier studies have shown that a small fraction of the BiP interaction defective IRE1a mutant is constitutively activated even under normal conditions as reflected by XBP1u mRNA cleavage (Oikawa et al., 2009). Therefore, it is likely that both BiP and the Sec61 translocon are required to maintain IRE1a in an inactive form under normal conditions. However, unlike BiP, which is released from IRE1a during ER stress (Bertolotti et al., 2000;Okamura et al., 2000;Oikawa et al., 2009;Pincus et al., 2010), the interaction with the Sec61 translocon is maintained throughout ER stress ( Figure 6-figure supplements 1 and 2). We therefore propose that the Sec61 translocon may play a crucial role during ER stress to limit IRE1a activity. The disparate effects of the wIRE1a and sIRE1a mutants, which either promote or prevent IRE1a oligomerization and activation, respectively, lend support for this model.
In comparison to wild-type IRE1a, the attenuation of wIRE1a activity is significantly delayed. One plausible explanation is that wIRE1a robustly forms large oligomers that sustain a longer activation period during ER stress. In contrast, sIRE1a was slower to phosphorylate and more quickly attenuated than wild-type IRE1a. It is likely that other IRE1a interacting proteins besides Sec61 also contribute to IRE1a activation and attenuation (Lisbona et al., 2009;Pincus et al., 2010;Rodriguez et al., 2012;Eletto et al., 2014;Morita et al., 2017), since wIRE1a attenuation is significantly delayed but not completely prevented during prolonged ER stress.
An alternative possibility for the observed phenotypes of wIRE1a and sIRE1a is that these mutations in IRE1a affect IRE1a homo-oligomerization and/or ER stress-dependent activation of IRE1a independently of Sec61. Although future work is required to rule out this possibility, current results strongly indicate that the Sec61 translocon limits IRE1a oligomerization since two independent wIRE1a mutants exhibit similar effects and increased oligomerization. Furthermore, sIRE1a exhibits the opposite phenotype of wIRE1a, with reduced IRE1a oligomerization and slow activation/quick de-activation kinetics compared to wild-type IRE1a.
Upon first glance, the effects of disrupting the IRE1a-Sec61 interaction on XBP1u mRNA cleavage observed by our previous study (Plumb et al., 2015) and the current study may appear contradictory. However, the data are reconciled by considering the number of activated IRE1a molecules under all conditions. When an equal number of IRE1a and wIRE1a proteins are activated, wIRE1a exhibits less XBP1u mRNA cleavage because its RNase domains cannot access XBP1u mRNA as efficiently (Plumb et al., 2015), and thus produces less XBP1s protein ( Figure 5A,C and five hour treatment). However, as we demonstrate in this study, disrupting the Sec61-IRE1a interaction results in a slight increase in the number of activated IRE1a molecules under normal conditions and a more dramatic increase during prolonged ER stress conditions (Figures 4 and 5). In this case, the difference in the ability of IRE1a and wIRE1a to access XBP1u mRNA becomes negligible, since so many more wIRE1a molecules are active compared to wild type IRE1a. Thus, the production of XBP1s protein is in fact greater in wIRE1a than wild type IRE1a expressing cells under such conditions.
Our data suggest that the intensity of ER stress determines whether IRE1a signaling is attenuated or remains active. We propose that the selective attenuation of IRE1a signaling may be beneficial for secretory cells such as pancreatic beta cells and plasma cells by providing a longer time window to resolve ER stress and avert inappropriate cleavage of ER-localized mRNAs, including mRNAs encoding secretory proteins such as insulin and immunoglobulin (Lipson et al., 2006;Benhamron et al., 2014). If ER stress is prolonged and irremediable, as shown by previous studies (Lin et al., 2007;Rutkowski et al., 2006;Lu et al., 2014), the PERK pathway remains active and induces CHOP-mediated cell death. However, severe ER stress may induce higher-order oligomers of IRE1a by overcoming the Sec61 translocon barrier, thus leading to a defect in the attenuation of IRE1a signaling. This continuous IRE1a activation might be beneficial for tumor growth (Cubillos-Ruiz et al., 2015) but may promote cell death in secretory cells such as pancreatic beta cells (Ghosh et al., 2014). In conclusion, the Sec61 translocon plays an essential role in controlling oligomerization and activity of IRE1a during ER stress. Thus, the IRE1a and the Sec61 translocon may be a prime target for small molecule manipulation to either enhance or suppress IRE1a signaling in diseases conditions. DNA constructs pcDNA5/FRT/TO (Invitrogen, Carlsbad, CA) containing IRE1a-HA, wIRE1a (IRE1a-D434-443 HA) and IRE1a-K907A-HA were described previously (Plumb et al., 2015). IRE1a-T446A-S450A-T451A-HA mutant was created using previously described primers (Sun et al., 2015). sIRE1a (IRE1a-S439A-T446A-S450A-T451A-HA), IRE1a-V437A-D443A-HA, IRE1a-D434-443A-K907A-HA, in pcDNA5/FRT/ TO were made by site-directed mutagenesis. Prl-His-2xstrep-IRE1a-FLAG constructs were generated by first inserting Prl-His-2xstrep into pcDNA5/FRT/TO using standard methods. Next, IRE1a-FLAG was amplified beginning from amino acid 29 and cloned into pcDNA5/FRT/TO Prl-His-2xstrep. Mouse spliced XBP1 plasmid (Addgene# 21833) is a kind gift from Dr. David Ron. All PCR reactions were performed with Phusion high fidelity DNA polymerase (NEB, Ipswich, MA), except for site directed mutagenesis, which used Pfu-Ultra polymerase (Agilent technologies, Santa Clara, CA). 3% DMSO was included in all PCR reactions to enhance amplification. The coding regions of all constructs were sequenced to preclude any sequence error. The Yale Keck DNA Sequencing Facility performed all sequencing services.

Cell culture
HEK 293-Flp-In T-Rex cells were purchased from Invitrogen and cultured in high glucose DMEM (Corning, Corning, NY) containing 10% FBS (Gibco, Gaithersburg, MD ), 100 U/ml penicillin and 100 mg/ml streptomycin (Gibco) at 5% CO 2 . IRE1aÀ/À HEK293-Flp-In T-Rex cells were previously described (Plumb et al., 2015). To establish stable cell lines, IRE1aÀ/À HEK293 cells were transfected with 1 mg of pOG44 vector (Invitrogen) and 0.1 mg of FRT vectors containing IRE1a or its mutants using Lipofectamine 2000 (Invitrogen). After transfection, cells were plated in 150 mg/ml hygromycin (Invitrogen) and 10 mg/ml blasticidin (InvivoGen, San Diego, CA). The medium was replaced every three days until colonies appeared. The colonies were picked and equal expression of the recombinant IRE1a or its mutants was evaluated by western blotting. The same protocol was applied in HEK 293-Flp-In T-Rex cells to generate Prl-His-2xstrep-IRE1a-FLAG stable cell lines of IRE1a, wIRE1a or sIRE1a. IRE1a À/À/FRT MEF cells (Hollien et al., 2009) are from Julie Hollien (University of Utah, USA) and they were complemented with either IRE1a, or wIRE1a as previously described (Plumb et al., 2015). INS-1 cells are from Richard Kibbey (Yale School of Medicine, USA) and were grown in RPMI (Sigma), 12.5% FBS (Gibco), 1 mM sodium pyruvate, 10 mM HEPES, 2 mM glutamine, and 50 mM and beta-mercaptoethanol. All the cell lines used in this study were not tested for mycoplasma, but many cell lines were used in immunofluorescence assays with Hoechst staining that should reveal presence of mycoplasma. Cells were assumed to be authenticated by their respective suppliers and were not further confirmed in this study. However, IRE1a knock out cell lines were verified by immunoblotting with IRE1a antibodies.

ER stress treatment
Cells were counted and plated in 24 well (1.5 Â 10 5 ) plates and grown overnight to reach a confluence of 70% prior to treatment. In the case of overexpression study, doxycycline was added overnight. ER stress was induced by treating cells with tunicamycin (TM) or thapsigargin (Tg). All the concentrations and treatment time were as indicated in either result or figure sections. After the treatment, cells were directly harvested by adding 100 ul of 2X SDS sample buffer and boiled for 5 min with intermittent mixing and analyzed by western blotting. For XBP1u mRNA splicing assay, the cells were harvested in Trizol (Ambion, Foster City, CA) and the splicing assay was performed as described previously (Calfon et al., 2002).
Samples were run using 3-12% BN-PAGE Novex Bis-Tris (Invitrogen) gel at 150 V for 1 hr with dark blue buffer (50 mM Tricine pH 7, 50 mM BisTris pH 7% and 0.02%% G250) at room temperature and then exchange with light blue buffer (50 mM Tricine pH 7, 50 mM BisTris pH 7% and 0.002%% G250) for 4 hr in the cold room. To probe the Sec61 translocon, the gels were run for 1 hr with dark blue buffer at room temperature and 2 hr 45 min with light blue buffer in the cold room. After electrophoresis, gel was gently shaken in 1x Tris-Glycine-SDS transfer buffer for 20 min to remove residual blue dye. Transfer was performed using PVDF membrane (EMD Millipore) for 1 hr and 30 min at 85V. After transfer, the membrane was fixed with 4% acetic acid and followed with a standard western blotting procedure.

Phostag assay
IRE1a phosphorylation was detected by previously described method (Yang et al., 2010). Briefly, 5% SDS PAGE gel was made containing 25 mM Phos-tag (Wako). SDS-PAGE was run at 100 V for 2 hr and 40 min. The gel was transferred to nitrocellulose (Bio-Rad, Hercules, CA) and followed with western blotting. The intensities of the Phos-tag bands were quantified with Image Quant TL software (GE HealthCare).

Western blotting
Protein extracts were electrophoresed under reducing conditions on Tricine (Sigma) based SDS-PAGE gel and electro blotted onto nitrocellulose membrane (Bio-Rad). Blots were incubated with primary antibodies prepared in 1XPBS/Tween containing 5% BSA/0.02% NaN 3 for 1 hr and 30 min at room temperature. The secondary antibodies prepared in 5% Milk with 1XPBS/Tween were incubated for 1 hr at room temperature. Proteins were detected with SuperSignal West Pico or Femto Substrate (Thermo Scientific), exposed to Film BioExcel (Worldwide Life Sciences, Irvine, California) and developed.
2x Strep IRE1a and associating Sec61 complex protein purification Stable cell lines expressing 2xStrep IRE1a, wIRE1a and sIRE1a were induced with 200 ng/ml doxycycline and grown in 15 cm plate until 100% confluence. Cells were pelleted and proceed with microsome preparation as described (Plumb et al., 2015).. Briefly, cells were re-suspended in buffer (10 mM Hepes pH7.4, 250 mM Sucrose, 2 mM MgCl 2 and 1x protease inhibitor cocktail (Roche) and lysed by passing through 25-gauge for three times followed by 27-gauge for five times in cold room. Lysed samples were spun at low speed 2800g for 30 min and supernatant was collected and spun at 75,000g for 1 hr at 4˚C using MLA80 rotor. Microsome pellet was re-suspended in buffer containing (50 mM Hepes pH7.4, 250 mM Sucrose, 2 mM MgCl 2 and 0.5 mM DTT) and homogenized carefully using 2 ml dounce. Microsome concentrations were measured using absorbance A 280 and flashfreeze stored at À80˚C until further analysis.
To remove free IRE1a, which is not bound to Sec61, the material was further purified by passing through SP Sepharose beads (GE Healthcare). Briefly beads were prepared in 2 ml Bio-Rad column and washed 5x using no salt buffer (20 mM Tris pH8, 2 mM MgAc and 0.4% digitonin). Purified protein was diluted 5x with no salt buffer and pass-through S-column. Beads were washed 5x column volume and eluted with 500 mM NaCl buffer (50 mM Tris pH8, 2 mM MgAc, 10% glycerol, and 0.4% digitonin). Purified IRE1a-translocon complex was quantified along with BSA standards.
For the mRNA cleavage assay, purified IRE1a (5 nM) was mixed with cleavage buffer (50 mM Tris pH8, 50 mM NaCl, 5 mM MgCl 2 , 0.4% digitonin, 1 mM ATP, 2 mM DTT and 2U RNAsin). The reaction was initiated by adding 2 ng of P32 CTP labeled XBP1u mRNA. Samples were incubated at 30˚C. At each time point, sample was collected and the reaction was stopped by incubating at 70˚C for 10 min in formamide (American Bioanalytical) sample loading buffer containing 5 mM EDTA and 1x bromophenol blue. Sample was loaded in a 6% Urea PAGE gel. Prior to actual samples running, gel was pre-run at 20W for 25 min. Actual sample running was performed at 9W for 35 min. Gel was fixed in 10% (methanol and acetic acid) for 20 min, dried at 55˚C for 1 hr and 30 min. Gel was exposed to film and developed.

Immunoprecipitation
To test the interaction between recombinant IRE1a and the endogenous Sec61 translocon, HEK 293 cells were transiently transfected with HA-tagged IRE1a constructs and expression induced with 100 ng/ml doxycycline. 24 hr after transfection, cells were harvested in 1xPBS and centrifuged for 2 min at 13,800g. Cell pellet was lysed in Buffer A (50 mM Tris pH 8, 150 mM NaCl and 1% digitonin) by rotating 30 min at 4˚C. The supernatant was collected by centrifugation at 20,000g for 15 min. For co-immunoprecipitation, supernatant was incubated with anti-HA-agarose (Roche) and anti-HA magnetic beads (Thermo Scientific). The beads were washed 3x with 1 ml of Buffer A containing 0.2% digitonin. The bound material was eluted from the beads by directly boiling in 50 ml of 2x SDS sample buffer and analyzed by immunoblotting.

Immunostaining assay
Cells (0.12 Â 10 6 ) were plated on 12 mm round glass coverslips (Fisher Scientific) coated with 0.1 mg/mL poly-lysine in 24-well plates. Expression of IRE1a constructs was induced with doxycycline (2 to 5 ng/ml) for 16 hr prior to treatment with ER stress inducers. For immunostaining, cells were fixed with 3.7% formaldeyhyde (J.T. Baker, Phillipsburg, New Jersey) for 10 min and permeabilized with 0.1% Triton X-100 (American Analytical, Akron, OH) for 5 min. The non-specific binding sites were blocked with Buffer A (1xPBS containing 10% Horse Serum and 0.1% Saponin) for 45 min. 100 mL of rabbit anti-HA, mouse anti-HA (Covance, Princeton, NJ), or anti-Sec61b primary antibodies were added at 1:100 dilution in Buffer A and incubated for 1 hr, then washed 5X for 5 min. 100 mL of the secondary antibodies anti-rabbit Cy3, anti-mouse Cy3, and anti-mouse Cy2 (Jackson Immuno Research) were added at 1:100 dilution in Buffer A and incubated for 1 hr before washing five times with Buffer A. Coverslips were then incubated with 5 mg/ml Hoechst stain in 1xPBS for 15 min, washed with 1xPBS, and mounted using Fluoromount G (SouthernBiotech).
Cells were imaged on Leica scanning confocals (provided by the West Campus Imaging Core and the Nanobiology Institute at Yale University) consisting of an inverted microscope (Leica SP6/SP8), and an HC PL APO 63X (CS2 No: 11506350) oil objective lens (Leica, Wetzlar, Germany), and was controlled by the Leica Application Suite X. Sequential image scanning at 1x zoom, 100 Hz, 1024 Â 1024 pixels, and with line averaging set at four was used to collect images for cluster analysis. Sequential image scanning at 1.5x zoom, 100 Hz, and 2048x2048 pixels and line averaging of 6 was used for displayed images. To quantify number of cells with IRE1a puncta, the total number of cells per frame was first determined by manually counting Hoechst-stained nuclei. Only cells with a clearly present ER signal were included in this count. Subsequently, the number of cells with IRE1a puncta were counted, with puncta being defined as concentrated fluorescence signal typically approximately 0.4 um in diameter (4 hr tunicamycin treatment, 0.5 hr thapsigargin treatment) or approximately 1.5 mm in diameter (2 hr thapsigargin treatment). FIJI was used for cell counting. Data was graphed using GraphPad Prism and represented with standard error of the mean.