A Conserved Structural Determinant Located at the Interdomain Region of Mammalian Inositol-requiring Enzyme 1α*

Inositol-requiring enzyme 1α (IRE1α), an endoplasmic reticulum-resident sensor for mammalian unfolded protein response, is a bifunctional enzyme containing kinase and RNase domains critical for trans-autophosphorylation and Xbp1 mRNA splicing, respectively, in response to endoplasmic reticulum stress. However, the amino acid residues important for its function and activation remain largely unexplored. Here, through analysis of IRE1α mutants associated with human somatic cancers, we have identified a highly conserved proline residue at position 830 (Pro830) that is critical for its structural integrity and hence, the activation of both kinase and RNase domains. Structural analysis revealed that Pro830 may form a highly conserved structural linker with adjacent tryptophan and tyrosine residues at positions 833 and 945 (Trp833 and Tyr945), thereby bridging the kinase and RNase domains. Indeed, mutation of Pro830 to leucine (P830L) completely abolished the kinase and RNase activities, significantly decreased protein stability, and prevented oligomerization of IRE1α upon ER stress; similar observations were made for mutations of Trp833 to alanine (W833A) and to a lesser extent for Y945A. Our finding may facilitate the identification of small molecules to compromise IRE1α function specifically.

Inositol-requiring enzyme 1␣ (IRE1␣), an endoplasmic reticulum-resident sensor for mammalian unfolded protein response, is a bifunctional enzyme containing kinase and RNase domains critical for trans-autophosphorylation and Xbp1 mRNA splicing, respectively, in response to endoplasmic reticulum stress. However, the amino acid residues important for its function and activation remain largely unexplored. Here, through analysis of IRE1␣ mutants associated with human somatic cancers, we have identified a highly conserved proline residue at position 830 (Pro 830 ) that is critical for its structural integrity and hence, the activation of both kinase and RNase domains. Structural analysis revealed that Pro 830 may form a highly conserved structural linker with adjacent tryptophan and tyrosine residues at positions 833 and 945 (Trp 833 and Tyr 945 ), thereby bridging the kinase and RNase domains. Indeed, mutation of Pro 830 to leucine (P830L) completely abolished the kinase and RNase activities, significantly decreased protein stability, and prevented oligomerization of IRE1␣ upon ER stress; similar observations were made for mutations of Trp 833 to alanine (W833A) and to a lesser extent for Y945A. Our finding may facilitate the identification of small molecules to compromise IRE1␣ function specifically.
The unfolded protein response (UPR), 2 a highly conserved endoplasmic reticulum (ER)-to-nucleus signaling pathway is critical for maintaining ER homeostasis and has been implicated in the pathogenesis of many human diseases, including diabetes, cancer, and lung and heart diseases. UPR is initiated by the activation of three major sensors at the ER membrane: inositol-requiring enzyme 1 (IRE1), PKR-like-ER kinase, and activating transcription factor 6. Activation of UPR leads to the induction of chaperones, and ER-associated degradation components, global translational attenuation, is required for the maintenance of ER function, clearance of misfolded proteins from the ER, and if stress persists, induction of apoptosis (1). Among the three branches, the IRE1␣-initiated pathway is the most evolutionarily conserved and represents the only UPR in yeast (2,3). Recent studies have shown that this pathway plays important roles in a wide range of physiological and disease conditions including B cell and adipocyte differentiation, secretory function for pancreas and salivary glands, neurodegeneration, and obesity and insulin resistance (1,4). Hence, a comprehensive understanding of the regulatory mechanisms underlying mammalian IRE1␣ activation is critical to the development of new therapeutic approaches.
In addition to a kinase domain, the cytoplasmic tail of the IRE1␣ protein also possesses endoribonuclease (RNase) activity which cleaves Hac1/Xbp1 mRNA (5,6). The Hac1/Xbp1 mRNA encodes a potent transcription factor responsible for the up-regulation of many genes involved in protein folding, degradation and trafficking (7)(8)(9)(10). The ER luminal domain of IRE1␣ protein is critical for sensing ER stress and subsequent IRE1␣ activation. Although earlier studies of human IRE1␣ have suggested that IRE1␣ dimers are sufficient for its activity (11)(12)(13)(14), recent studies showed that the formation of stressinduced IRE1␣ foci is conserved between yeast and mammals (15,16). The formation of IRE1␣ foci in response to ER stress is believed to juxtapose the IRE1␣ kinase and RNase domains, thereby allowing for a more efficient way to relay signals emanating from the ER lumen.
One outstanding question is the importance of and the molecular mechanism underlying the activation of mammalian IRE1␣ RNase and kinase domains. Some recent studies have shed light on this question. First, kinase-defective K599A mutant abolishes the RNase activity of IRE1␣, suggesting that kinase activity and trans-autophosphorylation of IRE1␣ protein are important for the activation of its RNase domain (15). This is likely achieved by phosphorylation-induced conformational changes. Moreover, a previous study reported the failure to express stable forms of individual kinase and RNase domains (18). Together, these studies imply that the two cytosolic domains of IRE1␣ protein are functionally linked via autophosphorylation. However, two recent studies suggested a dispensable role of IRE1␣ phosphorylation in activation of its RNase domain (19,20). Thus, further studies are required to elucidate the relationship between these two domains.
Through analysis of IRE1␣ mutants associated with human somatic cancers (21), our study identifies the interdomain linker region of the cytosolic domain of IRE1␣ as an important structural determinant for its function. Thus, this region may be used as a potential drug target for small molecules to specifically regulate IRE1␣ signaling.   (University of California, San Francisco). HEK293T, T-REx293, and Phoenix cells as described (22) were maintained in DMEM supplemented with 10% FBS (Hyclone) and 1% penicillin/streptomycin. Thapsigargin (Tg; EMD Calbiochem) and stock cycloheximide (Sigma) were dissolved in dimethyl sulfoxide and ethanol, respectively. Cells were treated with Tg at indicated concentrations for indicated times and immediately snap-frozen in liquid nitrogen. Phos-tag was purchased from the NARD Institute (Japan).

Transfection, Retroviral Transduction, and Stable Cell Lines-
HEK293T cells were transfected with plasmids using polyethylenimine (Sigma) as we recently described (24). Cells were snap-frozen in liquid nitrogen 24 h post-transfection followed by Western blotting. To avoid experimental variations due to transfection efficiency, stable IRE1␣ Ϫ/Ϫ MEF lines expressing various IRE1␣ constructs were generated using retroviral transduction as described (22). Stable cell lines were selected in hygromycin (VWR) at 125 g/ml. Stable cell lines were made and tested independently at least twice.
Analysis of IRE1␣ Foci Formation-Performed as described (16) with the following modifications. T-REx293 cells were transfected with 0.2-0.5 g of mutant IRE1␣-3F6HGFP plasmid and 1 g of OG44 (a gift from Dr. Peter Walter) for 24 h followed by selection by 62.5 g/ml hygromycin for 2-3 weeks. Cells were treated with doxycycline at 5 g/ml for 24 h to induce expression of IRE1-3F6HGFP followed by Western blotting. In some experiments, cells were plated on a polylysine-coated coverslip in a 6-well plate overnight and transfected with pDsRed2-ER plasmids for 12 h followed by doxycycline (VWR) treatment for 24 h. In some experiments, cells were then treated with Tg at 300 nM for 4 h to induce ER stress. The Prolong Gold Antifade Reagent with DAPI (Invitrogen) was used as mounting medium. Fluorescent microscopic pictures of the T-REx293 were taken with the Zeiss 710 confocal microscope using a 63ϫ/1.4 objective (Cornell Imaging Core Facility).
Structure Analysis of Cytosolic Domains of Yeast and Human IRE1␣ Proteins-The crystal structure of the cytosolic domain of yeast IRE1␣ (Protein Data Bank accession number 3FBV) (15) and human IRE1␣ (Protein Data Bank accession number 3P23) (26) were used for structural analysis of IRE1␣ mutants.
Cycloheximide Treatment-Stable MEF cell lines with various IRE1␣ constructs were treated with 100 g/ml cycloheximide for 0, 6, 9, 12, 16, and 24 h. Cells were then snap-frozen in liquid nitrogen and analyzed by Western blotting.
Western Blotting, Phosphatase Treatment, and Image Quantitation-Preparation of cell lysates, nuclear extract, and Western blotting were performed as we described previously (22,24). Antibodies used in this study included XBP1 and HSP90 (Santa Cruz Biotechnology), IRE1␣ (Cell Signaling) and PARP (a gift from Lee Kraus, University of Texas Southwestern Medical Center). Phos-tag gel and phosphatase (PPase) treatment were performed as described (22,27,28). Membranes were routinely strip-reprobed for HSP90 as a position control. Band density was quantitated using the Image Lab software on the ChemiDOC XRSϩ system (Bio-Rad) and presented as mean Ϯ S.E. from several independent experiments or as representative data from at least two independent experiments.
RNA Extraction and Xbp1 mRNA Splicing Assay-These experiments were performed as described previously (22).
Statistical Analysis-Results are expressed as mean Ϯ S.E. Comparisons between groups were made by an unpaired twotailed Student t test. p Ͻ 0.05 was considered statistically significant. All experiments were repeated independently at least twice.

RESULTS
The S769F and P830L Mutations Abolish IRE1␣ Phosphorylation and Activation-In a recent proteomic screening study, seven mutations (N244S, L474R, R635W, S769F, Q780-stop, and P830L) were identified to be associated with various human cancers (21) (Fig. 1A). As they present in tumor cells in one copy (21), their significance in cancer is unclear. Nonetheless, the identification of these mutants allowed us to address how cancer-associated mutations may affect IRE1␣ structure and function. The relative positions of these mutations in the kinase domain of the active form of yeast IRE1␣ protein (15), based on homology, are shown in Fig. 1B.
To determine the impact of these mutants on IRE1␣ activity, we generated IRE1␣ Ϫ/Ϫ MEF cells stably expressing the mutant proteins and assessed IRE1␣ phosphorylation in response to ER stress using Phos-tag gels as we recently described (22,27,28). Our previous studies have shown that the slower migrating band on the Phos-tag gel represents the phosphorylated form of IRE1␣, and the ratio of phosphorylated to total IRE1␣ correlates with the amount of ER stress (27,28). Strikingly, mutation of IRE1␣ at either S769F or P830L, but not R635W or L474R, abolished IRE1␣ mobility shift upon Tg treatment, indicative of defective IRE1␣ kinase activity ( Fig. 2A, lanes 7-14, and Fig. 2B,  lanes 1-8). Of note, we were not able to detect N244S and Q780-stop mutants due to technical reasons. Quantitation of the percent of p-IRE1␣ in total IRE1␣ from at least two independent experiments is shown in Fig. 2D.
It is worth noting that it is necessary to use Phos-tag gels to analyze phosphorylation of IRE1␣ mutants because p-IRE1␣ did not separate well in regular gels under the same running conditions (Fig. 2, B and C). The kinase-dead K599A mutant (17) exhibited no IRE1␣ phosphorylation whereas the dimerization-defective D123P mutant (14) showed a 70% reduction (lanes 7, 8, 10, and 11, and Fig. 2C). As additional controls, mutations at S548D and T973D had no effect on IRE1␣ phosphorylation upon ER stress ( Fig. 2A, lanes 1-6). Supporting the notion that the slower migrating band is due to phosphorylation, phosphatase treatment abolished the bandshift in WT IRE1␣ (Fig. 2C, lane 3); by contrast, there was no change in the electrophoretic mobility of P830L IRE1␣ following phosphatase treatment (Fig. 2C, lanes 4 -6).
Pro 830 Is Located in a Linker Region Bridging the Two Cytosolic Domains of IRE1␣ Protein-We then examined the potential effects of the two loss-of-function mutants on IRE1␣ structure (Fig. 3A). Based on the structure of human IRE1␣ (26), Ser 769 is located in the kinase domain (Fig. 3A). Replacement of the serine with a bulky phenylalanine residue causes sterical collisions with at least two neighboring residues, Tyr 765 and Cys 794 (Fig.  3B), which may affect the stability and activation of the kinase domain. Although Cys 794 is not conserved, both Ser 769 and Tyr 765 are highly evolutionally conserved (Fig. 3D).
Interestingly, Pro 830 is located at the junction between the kinase and RNase domains (Fig. 3A). Further analysis revealed that Pro 830 may form a highly hydrophobic patch with Trp 833 and Tyr 945 or with Leu 941 on the other side (Fig. 3C). Based on this, mutation of Pro 830 to Leu was not favored as the side chain of leucine would contact the Trp 833 and Tyr 945 residues (Fig.  3C), or Leu 941 in the other direction (data not shown), and hence, disrupt this structural element. All four residues involved in this structural motif (Pro 830 , Trp 833 , Leu 941 , and Tyr 945 ) were highly conserved in both IRE1␣ and IRE1␤ proteins through evolution from yeast to humans (Fig. 3E), suggesting that the Pro 830 -containing structural determinant bridging the two domains of IRE1␣ may be critical for IRE1␣ activation in a highly conserved manner. We speculated that understanding the effect of P830L may provide insight into the importance of the region linking the kinase and RNase domains of IRE1␣, hereafter termed the "linker." Therefore, in the remainder of this study, we focused on the Pro 830 residue.
The Structural Element Consisting of Pro 830 -Trp 833 -Tyr 945 Is Critical for IRE1␣ Function-We next attempted to delineate the effect of this hydrophobic patch consisting of Pro 830 , Trp 833 , and Tyr 945 or Leu 941 on IRE1␣ activation by mutagenesis. Interestingly, W833A caused a dramatic defect on IRE1␣ phosphorylation similar to that of P830L in response to ER stress (Fig. 4A, lanes 7 and 8), and to a lesser extent, for Y945A (Fig. 4A, lanes 11 and 12), but not L941A (Fig. 4A, lanes 9 and  10). To exclude the possibility that the P830L effect was caused by random mutations generated at other sites during mutagenesis, we mutated P830L back to P (P830L3 P). Indeed, the P830L3 P mutation fully restored the WT phenotype (Fig. 4B,  lanes 13 and 14). Pointing to the importance of the Pro 830 position, mutation at the nearby P821L, also a highly conserved residue (Fig. 3E), had only a minor effect (ϳ20%) on IRE1␣ phosphorylation (Fig. 4B, lanes 5 and 6). Moreover, mutation of Pro 830 to alanine (P830A) caused a much milder defect in IRE1␣ phosphorylation relative to the P830L mutant, with ϳ70% of IRE1␣ being phosphorylated under ER stress (Fig. 4, B  and C, lanes 1 and 2).
Next, we asked whether mutating Tyr 945 or Trp 833 to a smaller residue such as alanine could reverse the detrimental effect of P830L by accommodating the leucine residue. However, the double mutants of P830L/Y945A or P830L/W833A exhibited the same phenotype as P830L with no IRE1␣ phosphorylation in response to ER stress (Fig. 4B, lanes 7, 8, 11, and  12). Quantitation of IRE1␣ activation in all mutants is shown in Fig. 4C.
We then checked the RNase activity of these mutants by assessing the protein levels of XBP1s using Western blotting. Indeed, XBP1 protein levels correlated well with the extent of IRE1␣ phosphorylation: the P830L, W833A, and Y945A mutations led to no XBP1 production in response to ER stress and to a much lesser extent for L941A and P830A (Fig. 4D). As a positive control, RNase-dead K907A mutant (25) with normal IRE1␣ phosphorylation (Fig. 4A, lanes 5 and 6) was defective in XBP1 production in response to ER stress (Fig. 4D). Taken together, our data showed that three residues, Pro 830 , Trp 833 , and Tyr 945 , not Leu 941 , likely form a structural linker that is critical for IRE1␣ function.
The Pro 830 -containing Structural Linker of IRE1␣ Is Important for Its Stability-To shed further light on the defects caused by P830L, we examined the stability of IRE1␣ mutants under basal conditions. IRE1␣ Ϫ/Ϫ MEFs stably expressing var- Western blot analysis of XBP1s and IRE1␣ followed overnight transfection of HEK293T cells with near 100% transfection efficiency. Cells were treated with 300 nM Tg for 4 h prior to harvest. PARP and HSP90, loading controls for XBP1s and IRE1␣, respectively. Representative of two independent experiments. Numbers shown below the XBP1s blot refer the relative abundance of XBP1s protein (after normalized with PARP). Arrow next to the IRE1␣ blot indicates the endogenous IRE1␣ band. The ratio of endogenous to exogenous IRE1␣ protein levels is 1:10.
ious mutant proteins were treated with cycloheximide, an inhibitor of protein synthesis, for the indicated time periods followed by Western blot analysis of IRE1␣ protein levels. Although K599A had a half-life similar to that of WT protein (t1 ⁄ 2 ϭ 12 h; Fig. 5, A and B), the P830L, W833A, and Y945A mutations dramatically reduced the half-life of IRE1␣ to 6 h or less (Fig. 5, D-F). Pointing to the specificity of these three residues, L941A had no effect on IRE1␣ stability (Fig. 5C). Taken together, these results suggested that P830L, W833A, and Y945A mutations at the linker region had a significant impact on IRE1␣ protein stability, an effect distinct from the kinasedead K599A mutation.
The Pro 830 -containing Structural Linker of IRE1␣ Is Important for Its Oligomerization-As mammalian IRE1␣ oligomerizes upon ER stress (16), we next tested the effect of P830L on foci formation. To this end, we used an inducible system to detect IRE1␣ foci formation in T-REx293 cells (16) (Fig. 6A). Using co-expression of ER-localized dsRed encoded in the pDsRed2-ER plasmid, we first confirmed that exogenous WT and mutant IRE1␣ proteins were predominantly localized to the ER (supplemental Fig. S2). In line with a previous study (16), foci formation peaked in cells expressing WT-IRE1␣ protein upon 4-h treatment of 300 nM TG ( Fig. 6B and data not shown). Interestingly, unlike WT IRE1␣, P830L failed to form foci upon ER stress (Fig. 6C) as did the dimerization-defective D123P IRE1␣ (Fig. 6D). Thus, the Pro 830 -containing structural linker of IRE1␣ affects its oligomerization in response to ER stress.
To test whether mutant IRE1␣ protein would have a dominant negative effect on endogenous IRE1␣ protein, we transfected mutant proteins into HEK293T cells and tested for an ER stress response. Surprisingly, despite being expressed 10-fold more than endogenous protein levels, exogenous mutant proteins failed to reduce IRE1␣ activity significantly as assessed by XBP1 protein levels (Fig. 6E). Thus, the P830L mutant does not act as a dominant negative mutation but rather may serve as a loss-of-function mutation.

DISCUSSION
This is the first study addressing how cancer-associated IRE1␣ mutations affect its structure and function. Here, we have identified an important structural element that is critical for the function of IRE1␣. This structural linker consisting of Pro 830 , Trp 833 , and Tyr 945 bridges the cytosolic kinase and RNase domains of IRE1␣ protein. Disturbance of this linker structure abolishes the activities of both functional domains and prevents IRE1␣ foci formation in response to ER stress. This mechanism is likely to be evolutionarily conserved as these three residues are highly conserved from yeast to humans. Although it remains unclear how P830L may affect tumor development in vivo (which is not the focus of this study), this study demonstrates that the P830L mutation is deleterious for IRE1␣ stability and function.
How the folding and activation of the two cytosolic domains of IRE1␣ are coupled in vivo remains unclear. Currently, one prevailing model is that the activation of the RNase domain of IRE1␣ depends on the trans-autophosphorylation of its adjacent kinase domain following dimerization or oligomerization (1,18). Efforts to understand this phosphorylation-mediated regulatory event further are hampered by an incomprehensive knowledge of all of the specific phosphorylated residues on mammalian IRE1␣ protein. In yeast, an estimated 17 Ser/Thr residues located at both the activation loop and the adjacent ␣EF-␣F loop are believed to be phosphorylated under ER stress (15). Although we predict that trans-autophosphorylation of mammalian IRE1␣ protein is not likely to be as extensive as in yeast, as indicated by a single shift in the Phos-tag-based Western blots (22,27,28), the identities and the importance of these possible phosphorylation events on mammalian IRE1␣ activation remain an interesting question.
Our data suggest that the Pro 830 -Trp 833 -Tyr 945 residues in the IRE1␣ hydrophobic core are important for its folding and activation. This linker region affects not only the activities of the kinase and RNase domains, but also its half-life and the ER stress-induced oligomerization. The effect of Pro 830 -related mutations on IRE1␣ function is indeed more severe than the kinase-dead K599A or the dimerization-defective D123P mutants, suggesting that Pro 830 -related mutations may interfere with the folding process of IRE1␣ protein. How we can translate this finding to therapeutic purposes require further studies.