Direct Association of Unfolded Proteins with Mammalian ER Stress Sensor, IRE1β

IRE1, an ER-localized transmembrane protein, plays a central role in the unfolded protein response (UPR). IRE1 senses the accumulation of unfolded proteins in its luminal domain and transmits a signal to the cytosolic side through its kinase and RNase domains. Although the downstream pathways mediated by two mammalian IRE1s, IRE1α and IRE1β, are well documented, their luminal events have not been fully elucidated. In particular, there have been no reports on how IRE1β senses the unfolded proteins. In this study, we performed a comparative analysis to clarify the luminal event mediated by the mammalian IRE1s. Confocal fluorescent microscopy using GFP-fused IRE1s revealed that IRE1β clustered into discrete foci upon ER stress. Also, fluorescence correlation spectroscopy (FCS) analysis in living cells indicated that the size of the IRE1β complex is robustly increased upon ER stress. Moreover, unlike IRE1α, the luminal domain of IRE1β showed anti-aggregation activity in vitro, and IRE1β was coprecipitated with the model unfolded proteins in cells. Strikingly, association with BiP was drastically reduced in IRE1β, while IRE1α was associated with BiP and dissociated upon ER stress. This is the first report indicating that, differently from IRE1α, the luminal event mediated by IRE1β involves direct interaction with unfolded proteins rather than association/dissociation with BiP, implying an intrinsic diversity in the sensing mechanism of mammalian sensors.


Introduction
The endoplasmic reticulum (ER) is responsible for the structural maturation of proteins entering the secretory pathway. To ensure the fidelity of protein folding and maturation, cells turn on a network of signaling pathways, collectively termed the unfolded protein response (UPR) [1]. Protein folding is monitored by three distinct sensors: inositol-requiring enzyme 1 (IRE1) [2][3]; protein kinase RNA (PKR)-like ER kinase (PERK) [4]; and activating transcription factor 6 (ATF6) [5]. IRE1 is a conserved transmembrane protein that has an ER luminal sensor domain and cytosolic kinase and ribonuclease domains. The luminal domain of IRE1 senses the accumulation of unfolded proteins and then the activated ribonuclease domain cleaves specific exon-intron sites in the mRNA encoding the transcription factor XBP1 (X-box binding protein 1) [6][7]. This cleavage initiates an unconventional splicing reaction, leading to the production of active XBP1 and induction of various adaptive genes [8].
The mechanism explaining how IRE1 senses the unfolded proteins is best understood in yeast. A series of studies identified the two-step sensing mechanism of yeast Ire1, consisting of a BiPdeprivation step and a direct association step. Under normal conditions, ER chaperone BiP is associated with Ire1. Under stressed conditions, excess unfolded proteins deprive Ire1 of BiP, and the resulting BiP-free Ire1 forms homomeric associations (Step 1) [9][10][11][12]. Then the homomeric Ire1 associates directly with unfolded proteins, which may elicit conformational change in the luminal domain, leading to the reorientation of the cytosolic domain and possible autophosphorylation of the kinase domain (Step 2) [13][14]. Moreover, recent studies indicate that the activated Ire1 clusters into discrete foci and forms dot-like assemblies [14][15]. Since the target RNA of yeast Ire1, Hac1 mRNA, is co-localized with these foci in cells, the high-ordered assemblies are believed to provide a concentrated, specialized molecular microenvironment, which could attract low-affinity binders with high avidity [16].
Similar to the yeast Ire1, mammalian IRE1a also senses the unfolded proteins via BiP dissociation and following homomeric association [17][18][19][20]. However, the mechanism is different from that of yeast Ire1. Compared to yeast Ire1 activation, which is dually regulated by BiP deprivation and direct association with unfolded proteins, the activation of mammalian IRE1a is mainly regulated by the BiP deprivation step [21]. It was recently reported that mammalian IRE1a also forms high-ordered assemblies upon ER stress [22].
In mammals, there are two IRE1 paralogues, IRE1a and IRE1b [23][24][25]. The major difference is their expression pattern. While IRE1a is expressed ubiquitously, the expression of IRE1b is restricted to the epithelium of the gastrointestinal tract [26]. Also, there is considerable divergence in their downstream events. Contrary to the survival effect mediated by IRE1a [27], IRE1b is involved in apoptotic cell death [25]. One reason behind this would be the diverse characteristics in the cytosolic domain, and the completely different targets: IRE1a cleaves XBP1 or insulin mRNA [28][29][30], while IRE1b targets ribosomal RNA [25] [31] or MTP mRNA [32].
Thus, even though the downstream effects mediated by the cytosolic domains of the two mammalian IRE1s have been extensively studied, their luminal events, especially in IRE1b, have not been elucidated. In this study, we performed a comparative analysis using the two mammalian IRE1s to clarify the luminal event mediated by IRE1b.

Cell Culture, Transfection, and Treatment
HeLa cells and HEK293T cells were cultured at 37uC in DMEM supplemented with 100 U/ml penicillin, 100 mg/ml streptomycin, and 10% fetal bovine serum, in an atmosphere containing 5% CO 2 . EffecteneH Reagent (QIAGEN) was used to introduce plasmid DNA into HeLa cells, and the calciumphosphate-DNA precipitation method was used for HEK293T cells. To induce ER stress, cells were treated with tunicamycin (2.5 mg/ml) or thapsigargin (1 mM). The assays with IRE12/2 MEFs (Fig. S1A) were performed as described previously [21].

Live Cell Imaging
Before imaging, cells were washed twice with HBSS, and replaced in phenol red-free DMEM (Invitrogen) supplemented with 25 mM Hepes-NaOH (pH 7.4) and 10% FBS. Images were collected by LSM 510 META confocal microscope equipped with a C-Apochromat 40x/1.2NA UV-VIS-IR Korr. water immersion objective lens (Carl Zeiss).

FCS Measurements
FCS measurements were performed with a ConfoCor 2 system and C-Apochromat 40x/1.2NA UV-VIS-IR Korr water immersion objective lens (Carl Zeiss) [37]. GFPs were excited at 488 nm. Confocal pinhole diameters were adjusted to 70 mm. Emission signals were detected with a 505 nm long-pass filter, and measured at 37uC in 5% CO 2 , 95% air-humidified atmosphere. The fluorescence autocorrelation function, G(t), from which the average residence time (t) and the absolute number of fluorescent proteins in the detection volume are calculated, was obtained as follows: where I(t+t) is the fluorescence intensity obtained by the single photon counting method in a detection volume at a delay time t (brackets denote ensemble averages). Curve fitting for the multicomponent model is given by: where F i and t i are the fraction and diffusion time of component i, respectively; N is the average number of fluorescent molecules in the detection volume defined by the beam waist w 0 and the axial radius z 0 ; s is the structural parameter representing the ratio of w 0 and z 0 ; T is the triplet fraction and t t is the relaxation time of the triplet state. In living cell analysis, a two-component model was used for the determination of diffusion time. G(t)s in living cells were measured for 30 s. The relationships between diffusion time and structural parameters were determined using a 10 27 M Rhodamine 6G (Rh6G) solution as a standard before measurement. The values of structural parameters were 4.5-6.0.

Recombinant Protein Technique
Expression and purification of recombinant proteins were performed as described previously [14] [21] [33]. Escherichia coli strain BL21 CodonPlus TM (DE3)-RIL (Strategene) was used for the expression of each protein. Each MBP-fused protein was induced by 0.3 mM IPTG for 1 h at 37uC, and purified by Ni-NTA (QIAGEN). Each purified protein was subjected to SDS-PAGE (8% polyacrylamide) and CBB staining.
Anti-aggregation assays using purified recombinant proteins were performed as described in our previous reports [14] [21]. Citrate synthase (Roche) or luciferase (Promega) was dissolved at 70 mM in 20 mM HEPES (pH 7.2), 50 mM KCl, 2 mM MgCl 2 , 6 M Gdn-HCl. After denaturation for 1 h at room temperature, samples were diluted out of the denaturant to 1.5 mM in the case of citrate synthase or 1 mM for luciferase in 100 ml of 20 mM HEPES (pH 7.2), 50 mM KCl, 2 mM MgCl 2 , with or without 1.5 mM (CS) or 4 mM (Luc) of recombinant proteins. Turbidity of the sample mixtures was monitored by measuring absorbance at 320 nm and normalized against the maximum value of the buffer sample.

Luciferase Assay
In the dual luciferase assay with the XBP1-Luc reporter, ATF4-Luc reporter, or ER-Luc, phRL-TK (Promega) was used as an internal control. HEK293T cells were seeded in 24-well plates, then transfected with plasmid DNA. At 24 h after transfection, cells were lysed for a luciferase assay. Reporter activity was measured using the dual luciferase assay system (Promega) and a luminometer (Berthold). The results are shown as means 6 SEM from triplicate experiments. Each value is shown as a fold induction normalized to that of mock transfectant (for overexpression) or nontreatment (for drug treatment), the value of which was set at 1.0. In Figure S1B, the value was shown as a Relative activity (RLU).

RT-PCR
Total RNA was prepared from cells using Isogen reagent (Nippon Gene). A SuperScriptH First-Strand Synthesis System (Invitrogen) was used to synthesize the cDNA, according to the manufacturer's instructions. cDNA for XBP1 was amplified by 35 cycles of PCR using the following primers: human Xbp1 sense primer, 59-AGAACCAGGAGTTAAGACAGC-39; human Xbp1 antisense primer, 59-AGTCAATACCGCCAGAATCC-39.
The lysates and immunoprecipitates were denatured in SDS sample buffer (50 mM Tris-HCl, pH 6.8, 2% SDS, 50 mM DTT, 10% glycerol, and 1 mg/ml bromophenol blue). SDS-PAGE was performed to resolve the proteins in the lysate. After electrophoresis, the proteins were electrotransferred onto a polyvinylidene fluoride microporous membrane and immunodetection was performed using an anti-HA antibody (HA.11; Covance), anti-KDEL antibody (Stressgen) or anti-Flag antibody (Sigma) according to standard procedures.

Clustering of IRE1a and IRE1b upon ER Stress
First, to elucidate the within-cell dynamics of the mammalian sensors, we fused monomeric green fluorescent protein tags (mEGFP) at the C terminus of IRE1s (Fig. 1A). As shown in Figure 1B, IRE1s were distributed on the ER, and could be merged with ER-localized markers (ER-mKate2). The functionality of GFP-fused IRE1s was confirmed with the induction of XBP1-Luc (IRE1a; Fig. S1A), and with the specific attenuation of ER-Luc (IRE1b; Fig. S1B). Under normal conditions, these GFPfused IRE1s were diffusely distributed on the ER. However, upon ER stress, both IRE1a (Li et al., 2010) and IRE1b clustered into discrete foci (Fig. 1C).
Surprisingly, there was a slight time-lag in the clustering between IRE1a and IRE1b. While IREb showed a distinct dotlike structure just 1 h after Tunicamycin (Tun) treatment, the IRE1a cluster began to appear 2 h after the treatment (Fig. S2). Moreover, the contributing domain to their clustering was different between IRE1a and IREb. While IRE1a required its luminal function as the homomeric defective mutation (D123P), not kinase defective mutation (K599A), inhibited the clustering, IRE1b seemed to require its kinase activity because K547A (kinase defective) mutant did not shown any dot-like structures even under ER stressed conditions ( Fig. 1D and 1E). These results imply that the molecular mechanism underlying the foci-formation is distinct in IRE1a and IREb.

FCS Analysis with IRE1a and IRE1b
Next, to scrutinize the within-cell dynamics of IRE1s more precisely, we employed a dynamic imaging method, fluorescence correlation spectroscopy (FCS). FCS measures the fluorescent molecules within a confocal-detection volume at near single molecule sensitivity, and estimates their molecular number, or diffusion coefficient which reflects the size of the containing complex [38]. In this study, this technique was applied to the GFPfused IRE1s in living cells (Fig. 2).
In FCS analysis, both of IRE1a and IRE1b showed fast movement, indicating that the IRE1s are dynamic on the ER membrane. Compared with the case of IRE1a, the autocorrelation curve of IRE1b significantly and robustly shifted to slower movement upon ER stress ( Fig. 2A). Curve fitting analysis of this data (performed by a two-component diffusion model which is best fitted for IRE1s) revealed that the curve shift in IRE1b was caused by the increment in relative content of slow component rather than the decrement in the diffusion coefficient itself (Fig. 2B). Although it is not clear which IRE1 situation or condition is reflected by each component (fast or slow), this data absolutely indicates that compared to IRE1a, the size of IRE1b-complex is robustly enlarged upon ER stress. As shown in Figure S3, the FCSdetected enlargement of IRE1b-complex did not require kinase activity, because such shift was also detected in the K547A mutant.
What cause the stress-dependent shift of IRE1b? Generally, a smaller diffusion coefficient indicates that the diffusional mobility of a molecule is decreased. Indeed, the increase in the amount of IRE1b showing smaller diffusion coefficient suggested that the amount of IRE1b homomeric-association and/or binding with other proteins to make a huge complex (whose diffusion coefficient is , 0.4 mm 2 /s) would be increased. To identify the IRE1bassociated factors that are not associated with IRE1a, a series of assays was performed as follows.

Association of IRE1b with Model Unfolded Proteins
One hint came from the study of yeast Ire1. Yeast Ire1 has been shown to associate with unfolded proteins directly [14] [39], which may explain the IRE1s' larger size-shift observed in the FCS analysis. To investigate this possibility for mammalian IRE1b, its luminal domain was prepared as MBP-fused fragments (Fig. 3A  and 3B), and was subjected to in vitro anti-aggregation assay.
Strikingly, the luminal fragments of IRE1b exhibited robust anti-aggregation activity in vitro, by inhibiting the aggregation of denatured citrate synthase (Fig. 3C) or luciferase (Fig. 3D). Such activity was also detected in yeast Ire1, but not in mammalian IRE1a [21]. This indicates the possibility that, upon ER stress, IRE1b directly associates with unfolded proteins, and forms a larger complex.
Next, to evaluate the association of IRE1b with unfolded proteins in cells, immunoprecipitation was performed using Amy1 or TCRa-GFP as model substrates (Fig. 4). Previous reports indicated that the overexpression of Amy1 induces significant ER stress [40], and TCRa-GFP is known as an ERAD (ER-associated degradation) target [41]. As shown in Fig. 4A and 4B, the overexpression of these proteins caused considerable activation of UPR pathways, with the drastic activation of reporters including the XBP1-Luc (IRE1a-XBP1 branch; Fig. 4A), and ATF4-Luc (PERK-ATF4 branch; Fig. 4B). Also, the cytosolic splicing of XBP1 mRNA was induced by the overexpression of these proteins (Fig. 4C). The partially lowered splicing of XBP1 mRNA in Figure 4C (compared to Fig. 4A) would be attributed to transfection efficiency. In the immunoprecipitation with these model unfolded proteins (Fig. 4D and 4E), the expression of Amy1 or TCRa-GFP was partially attenuated by the expression of IRE1b (wildype, not K547A mutant), due to the intrinsic RNase activity of IRE1b [36]. As expected, IRE1b showed strong coprecipitation signals, both with Amy1 (Fig. 4D) and TCRa-GFP (Fig. 4E), which was not detected in IRE1a. Also, since these associations were not inhibited by K547A mutation, kinase activity does not appear to be required for direct association with unfolded proteins.

Different Pattern in the Association with BiP
Finally, we evaluated the association of IRE1b with ER chaperone BiP. IRE1a showed the BiP-association signal under normal conditions, and the signal weakened upon ER stress (Fig. 5A) [21]. However, and surprisingly, IRE1b did not show any BiP-association signal irrespective of ER stress (Fig. 5A). This implies the possibility that the luminal events mediated by IRE1b do not involve BiP association/dissociation.

Discussion
In this study, we performed a comparative analysis using two mammalian IRE1s to clarify the luminal event mediated by IRE1b. One finding is that both IRE1a and IRE1b clustered into discrete foci upon ER stress, though the molecular mechanisms seem to be distinct in IRE1a and IRE1b (Fig. 1). FCS analysis indicated that IRE1b significantly and robustly shifts to a slower diffusion state upon ER stress, contrary to the IRE1a (Fig. 2). In agreement with this, the luminal domain of IRE1b showed antiaggregation activity in vitro (Fig. 3), and IRE1b was coprecipitated with model unfolded proteins (Fig. 4). Another striking difference was found in the BiP-association pattern. While IRE1a was associated with BiP and dissociated upon ER stress as previously reported [17] [21], any association signals was not detected in IRE1b (Fig. 5A), which might be caused by their sequence differences on the region corresponding to the BiP-binding site of IRE1a (Fig. 5B) [21]. These results indicate that, differently from IRE1a, the luminal event mediated by IRE1b directly interacts with unfolded proteins. This study provides the significant information about the luminal event mediated by IRE1b, and also suggests the sensing mechanism of mammalian sensors may involve the specific pathway on signal transition during UPR. To visualize the within-cell dynamics of IRE1s, we fused monomeric green fluorescent protein tags (mEGFP) at the C terminus of IRE1s. Monomeric substitution (A206K) may inhibit artificial oligomerization via intermolecular disulfide-bond between GFP. As previous studies [16] [22] inserted GFP-tag between the transmembrane and the cytosolic-effector domain, our C terminus adding of the GFP-tag might have eliminated the activity. However, this possibility was denied by the robust stress response of our GFP constructs (Fig. S1). Because all of these GFP-IRE1 (irrespective of the location of GFP-tag) showed stressdependent clustering, it seems common activating mechanism to cluster into discrete foci rather than making small oligomer. However, the molecular mechanism behind the clustering seems different between the two IRE1 molecules, as IRE1b showed clearer and faster foci than IRE1a (Figs. 1 and S2; especially with Tun treatment), and as the kinase activity was only required for IRE1b not for IRE1a to clustered into discrete foci (Fig. 1).
Importantly, though primitive, this study contains trial experiment with FCS measurement of ER stress sensors. The only report evaluating the molecular diffusion of an UPR-involving factor is a FRAP (fluorescence recovery after photobleaching) analysis of the ER chaperone, BiP [42]. Molecular diffusion of BiP is decreased upon ER stress, which might be caused by the direct association with unfolded proteins to increase the size of the BiPcontaining protein complex. Similarly, the molecular diffusion of IRE1b was decreased upon ER stress in our FCS analysis (Fig. S3), supporting the idea that IRE1b luminal events involve direct association with unfolded proteins under stressed conditions. The clustering detected by fluorescent microscopy and the shift of autocorrelation curve in FCS analysis would reflect different molecular events, respectively, because FCS is detectable to only mobile molecules with a single molecule sensitivity [38]. Our results in Figure 1 showed that the cytosolic domain is important for the clustering of IRE1b. However, its shifts of autocorrelation curve did not need the cytosolic kinase activity (Fig. S3). Also, while both IRE1a and IRE1b clustered into foci upon ER stress, the robust shift to a slower diffusion was only detected in IRE1b (Figs. 1, 2 and S2). These observations would be implicated as two possibilities; (i) IRE1s (beta?) possess an ability to assemble without forming visible clusters, (ii) IRE1b is associated with unfolded proteins in the ER with a mobile state to rapidly recognize the substrates regardless the kinase activity in the cytosol.
The newly found differences between IRE1a and IRE1b at the activating step may involve the specific pathway on signal transition during UPR. It is thought that while IRE1a is transiently activated and attenuated soon to perform survival effect for cells [22] [27], the activation of IRE1b is continual to elicit apoptotic cell death [25], as implied from the sustained repression of MTP mRNA [43] and chronic change in intestinal lipid absorption [44]. Such difference in their activating timing or continuousness might contribute to their different downstream effects. Also, the substrates sensed by IRE1a and IRE1b might be different each other, resulting from their intrinsic sensing mechanism (Fig. 5C). The unfolded or malfolded status that elicits the deprivation BiP of IRE1a could differ from that directly binds to IRE1b, and the difference would determine which IRE1 molecule should be activated. As showon in Figure 5B, the amino acid-sequence between IRE1a and IRE1b is not so conserved on their luminal domain, especially in the adjacent region to the transmembrane domain that is the BiP-binding site in case of IRE1a [21]. This might be one reason behind the different activating step between IRE1a and IRE1b. Also, such selectivity might be one reason why the two sequential steps in yeast Ire1, the BiP-deprivation step and the direct association step, are evolutionally divided into each IRE1 molecules in mammal (Fig. 5C).
Still, some questions remain to be solved. One question is how IRE1b forms homo-associates or dot-like assemblies. Previous structure analysis reported that yeast Ire1 has multiple homomeric interfaces in its lumen and forms polymeric oligomers [45]. On the contrary, the luminal domain of mammalian IRE1a has a single interface and forms dimers or small oligomers [46]. At this time, we have no clear answer to whether the luminal domain of IRE1b forms dimers (small oligomers) or robust high-molecular oligomers, because there are no structural information about the luminal domain of IRE1b, and because the amino-acid sequence is not so conserved on their luminal domain (Fig. 5B). We could not estimate the oligomer size of the MBP-fused IRE1b (Fig. 2), due to several technical difficulties. Alternatively, the cytosolic domain which contains dimmer-forming interface in IRE1a [47] might contribute to the dot like assemblies also in IRE1b, as the cytosolic kinase activity was important for the clustering of IRE1b (Fig. 1). Another question is how the stress-dependent clustering of IRE1b is regulated. Although an association with BiP was hardly detected in our immunoprecipitation (Fig. 5A), we cannot exclude the possibility that BiP is involved in the clustering of IRE1b by a different manner than with IRE1a, as it has been reported that IRE1b is co-immunoprecipiated with BiP from extracts of mice stomach mucosa (a tissue rich in IRE1b) [26]. Alternatively, there may be other regulating factors. Recent reports on IRE1a describe various regulating factors associated with the cytosolic domain, including BI-1 [48] or RACK1 [49][50]. Such factors might exist to regulate the clustering or activation of IRE1b. Also, it is still unknown how the luminal events (clustering, or direct association with unfolded proteins) links to the cytosolic activation (or phosphorylation). Does direct association function as a regulating step for the activation of IRE1b? Moreover, we could not examine the activity of IRE1b in detail. Only a small number of targets specific to IRE1b have been reported, and a system for precise evaluation of their changes has not been developed. To overcome these problems, and to fully elucidate the sensing mechanism of IRE1b, further research is needed.