Attenuation of yeast UPR is essential for survival and is mediated by IRE1 kinase

In eukaryotic cells, folding and modification of nascent membrane and secreted proteins take place in the ER (Wickner and Schekman, 2005). The increased demand of the ER activates the unfolded protein response (UPR) (Kaufman, 1999; Mori, 2000; Patil and Walter, 2001; Harding and Ron, 2002; Ron and Walter, 2007). In yeast, the UPR is initiated by activation of the ER transmembrane sensor Ire1 (Cox et al., 1993; Mori et al., 1993), leading to Ire1 autophosphorylation and oligomerization (Shamu and Walter, 1996; Lee et al., 2008; Aragón et al., 2009) and activation of Ire1 endo-RNase. Ire1 RNase-mediated excision of the translation-inhibitory intron in HAC1 mRNA is essential for the expression of the Hac1 protein (Cox and Walter, 1996; Sidrauski and Walter, 1997; Kawahara et al., 1998; Rüegsegger et al., 2001). The Hac1 protein, a UPR-specific transcription factor, then induces UPR target gene expression, including ER resident chaperones and critical protein-folding enzymes (Travers et al., 2000).

Although the initial phase of the UPR pathway has been well studied, subsequent events leading to the reestablishment of ER homeostasis and details of how UPR is finally attenuated have only been partially studied. Upon activation, the extent of Ire1 RNase activity must be tightly controlled as both ectopic expression of the spliced form of HAC1 mRNA (Kawahara et al., 1997) and the overexpression of Ire1 cause cells to grow poorly, highlighting the importance of UPR attenuation. Here, we establish a UPR attenuation assay in vivo and find an important role for the Ire1 kinase domain in attenuating HAC1 splicing beyond its known role in activation.

To test the molecular steps involved in the recovery phase of the UPR, we profiled Ire1 activity by monitoring HAC1 mRNA splicing upon UPR activation. Wild-type (WT) yeast cells were incubated with the well-characterized UPR inducer tunicamycin (Tm) for an extended length of time. Tm inhibits glycosylation in the ER, resulting in the accumulation of unglycosylated proteins, which leads to the activation of the UPR. Initially, spliced HAC1 mRNA appeared upon Tm incubation (Fig. 1, A–I) as reported previously (Cox and Walter, 1996; Kawahara et al., 1998). Under the experimental conditions used, the presence of spliced HAC1 mRNA was sustained for ≤6 h, suggesting that Ire1 RNase remained activated (Fig. 1, A–I). We reasoned that metabolic states of cells might affect either the activation or recovery kinetics of the UPR. Thus, we performed another UPR time course in which Tm was added to the same WT yeast cell except when they were at their earlier point of log phase. Although kinetic activation of Tm-induced HAC1 mRNA splicing was similar to that shown in Fig. 1 (A–I), we found that most of the spliced HAC1 mRNA disappeared after 6 h and the majority of HAC1 mRNA was of the unspliced form (Fig. 1, A–I). A possible explanation for the disappearance of the spliced HAC1 mRNA after 6 h may be the restoration of the ER protein-folding capacity and, thus, the termination of the UPR by inactivating Ire1 RNase. In a further extended time course (Fig. 1 B), the kinetics of appearance and the following disappearance of spliced HAC1 mRNA correlated well with the level of KAR2 mRNA, an immediate Hac1 transcription factor target, suggesting the potential recovery of ER protein-folding homeostasis at the later stages.

However, another explanation for decreased levels of spliced HAC1 mRNA could be the loss of Tm activity over time. We added a second dose of fresh Tm (1 µg/ml) to the yeast culture that had already been incubated with Tm for 8 h (Fig. 1 D, lane 4, fresh Tm). Cells were no longer responsive to Tm the second time (Fig. 1 D, lanes 5 and 6), whereas the initial exposure to Tm effectively induced Ire1 (Fig. 1 D, lanes 2 and 3). Furthermore, Tm in the yeast media was still capable of activating the UPR. When Tm-containing media were taken from those grown for 8 h and added into fresh cells, HAC1 mRNA splicing occurred efficiently (Fig. 1 E, lanes 5 and 6). In addition to the potency of Tm, these results also revealed that no “toxic” agent preventing HAC1 mRNA splicing was secreted in the media during the initial incubation of cells with Tm. Collectively, these results suggest that decreased levels of spliced HAC1 mRNA at later times were unlikely to be caused by a loss of the ability of Tm to induce the UPR.

Finally, we measured ER protein-folding capacity at a later time by examining the recovery of the carboxypeptidase Y (CPY) function. CPY is a vacuolar peptidase that traverses the ER, where it undergoes multiple modifications, including glycosylation, to become an enzymatically active mature form. If the later disappearance of spliced HAC1 mRNA reflects the recovery of the ER function, glycosylation and overall folding of proteins should be restored. Thus, we monitored the CPY enzymatic activity during a Tm time course to test whether the ER protein-folding state was later recovered. CPY activity present in WT and ire1Δ cells that were measured by a cleavage assay using the colorimetric substrate N-benzoyl-l-tyrosine p-nitroanilide (BTPNA) in vitro (Fig. 1 F; Stevens et al., 1986; Ohi et al., 1996; Lahav et al., 2004) was decreased upon UPR induction as anticipated.

Subsequently, at 7 h, Ire1-WT cells were able to regain CPY activity at levels similar to what was present before UPR activation (Fig. 1, F and G). The recovery in CPY activity was in inverse correlation with the level of the spliced form of HAC1 mRNA and occurred in a WT Ire1-dependent manner because CPY activity remained at low levels in ire1Δ UPR-deficient cells. These results further supported the idea that the disappearance of the spliced HAC1 mRNA reflected the recovery of ER protein-folding capacity.

To further test the relationship between ER functional recovery and disappearance of the spliced HAC1 mRNA, we performed a UPR recovery assay. If disappearance of the spliced HAC1 mRNA reflects the ER functional recovery, then removal/washing of Tm followed by incubation with fresh media should make the kinetics of the spliced HAC1 mRNA disappearance faster. To this end, cells were washed and placed in fresh media after a 2-h Tm treatment (Fig. 2 C), and we monitored the level of spliced HAC1 mRNA. At 2 h after Tm washout (and 4 h from the beginning of the assay), spliced HAC1 mRNA declined by 50% (Fig. 2, C [lane 4] and D) and was mostly undetectable with the majority of HAC1 mRNA present in the unspliced form by 3 h (Fig. 2, B, [open squares], C [lane 5], and D). The faster decline of the spliced form of HAC1 mRNA was also observed when cells were washed after 1, 1.5, or 0.6 h of Tm treatment, at the time before reaching the maximum levels of spliced HAC1 mRNA (Fig. 2, E–J), providing further support that the decline of spliced HAC1 mRNA represents the recovery phase of the UPR (Fig. 2, A and B). The initial 1-h delay before the levels of spliced HAC1 mRNA began to decline presumably reflects the time required to replenish functional glycosylation stores (Fig. 2 F), as N-acetylglucosamine phosphotransferase, the primary target of Tm, works at an early step in glycosylation.

Because Ire1 undergoes autophosphorylation of the activation loop during UPR induction (Shamu and Walter, 1996, Lee et al., 2008), we reasoned that its dephosphorylation might constitute at least one of the Ire1 deactivation events. We, therefore, predicted that a phosphomimetic Ire1 mutant in which three of the previously described phosphorylation sites on the activation loop (Shamu and Walter, 1996) were mutated to aspartate (Ire1-S840D/S841D/T844D) should be constitutively active or should not be able to recover once activated. Cells expressing Ire1-S840D/S841D/T844D were not constitutively active for the splicing of HAC1 mRNA, but upon Tm addition, these cells spliced HAC1 mRNA efficiently (Fig. 2, K and L). Thus, phosphorylation at these three residues alone is not sufficient to activate Ire1, but either phosphorylation of the fourth residue on the recently identified activation loop (Lee et al., 2008) or additional events beyond phosphorylation, such as sensing of the ER environment via the Ire1 sensor domain, must be required for activation of Ire1 (Okamura et al., 2000; Credle et al., 2005; Zhou et al., 2006; Kimata et al., 2007; Oikawa et al., 2007). Importantly, in marked contrast to WT cells, Ire1-S840D/S841D/T844D cells continued to display the spliced form of HAC1 mRNA at later times or even after removal of Tm (Fig. 2, M–P), suggesting that dephosphorylation is required for attenuating Ire1. Collectively, the disappearance of the spliced HAC1 mRNA at the later stage of the UPR time course reflects the reestablishment of ER protein-folding homeostasis, providing an experimental system to study the recovery phase of the UPR.

Using the UPR recovery assay, we examined whether any cytoplasmic domain mutants, specifically kinase domain mutants, would be defective in the transmission of the UPR attenuation signals from the ER luminal domain and display defects in decline of spliced HAC1 mRNA upon removal of Tm. Most kinase domains, including Ire1, have critical residues that are important for ATP binding and phosphorylation. Specifically, the aspartate at 828 (D828) is within the conserved DFG kinase motif and is predicted to be involved in the coordination of β and γ phosphates of ATP with Mg²⁺ (De Bondt et al., 1993; Johnson et al., 1996; Huse and Kuriyan, 2002; Lee et al., 2008) in the kinase pocket. In an in vitro kinase assay, we found that Ire1-D828A was unable to undergo autophosphorylation despite its ability to bind to ATP, confirming that it is an inactive kinase (Fig. S1, C and D). Mutating to the alanine of this residue (Ire1-D828A) resulted in the sustained presence of spliced HAC1 mRNA upon ER stress induction for ≤8 h. Only after that was there a gradual, small decline (Fig. 3, A and B). In parallel, the transcript of KAR2 was also sustained at later times. The observed differences in spliced HAC1 and KAR2 mRNA were unlikely to be caused by the expression levels of Ire1-WT or Ire1-D828A proteins, as they were detected at similar levels (Fig. S1 A), and an increase in Ire1-D828A expression by a 2µ plasmid did not change the profile of sustained HAC1 mRNA splicing (Fig. S1 B). Furthermore, sustained HAC1 mRNA splicing was also observed with Ire1-D828N (unpublished data). In the recovery assay, although washing/removal of Tm and the subsequent growth in fresh media facilitated the disappearance of spliced HAC1 mRNA in Ire1-WT cells, Ire1-D828A cells showed essentially no recovery where a spliced form of HAC1 mRNA was persistently present, even after Tm removal. The presence of spliced HAC1 mRNA was sustained regardless of the length of Tm treatment before removal of Tm (Fig. 3, C–J), which is similar to cells with the phosphomimetic Ire1 mutant (S840D/S841D/T844D). We also tested the involvement of the other conserved kinase domain residues, lysine at 702 (K702) and aspartate at 797 (D797), that are predicted to be important for ATP binding and phosphotransfer, respectively (De Bondt et al., 1993; Johnson et al., 1996; Huse and Kuriyan, 2002; Lee et al., 2008). However, neither of these mutants activated Ire1 RNase, making it impossible to analyze the disappearance of HAC1 mRNA (Fig. S2). Thus, these results suggest that the Ire1 kinase domain plays an important role in terminating the UPR in addition to the activation of HAC1 mRNA splicing.

Because phosphorylation of the activation loop is known to induce conformational changes, including movement of the loop itself, the DFG aspartate may play a role in establishing the position of the activation loop, depending on the state of ATP binding and/or autophosphorylation. Thus, D828A may allow the Ire1 RNase domain to mimic an active conformation for splicing HAC1 mRNA, which is normally achieved by autophosphorylation of the activation loop. If this were the case, it might rescue the minimal HAC1 mRNA-splicing levels of the Ire1-S840A/S841A/T844A activation loop mutant. Indeed, HAC1 mRNA splicing in Ire1-D828A/S840A/S841A/T844A cells increased significantly to levels similar to that of Ire1-D828A at 2 h after UPR induction (Fig. 4 A), suggesting that the D828A mutation could convert the nonphosphorylated activation loop into a conformation suitable for RNase activation.

Finally, beyond the important phosphorylation sites on the activation loop, we reasoned that additional residues might exist to participate in such conformational changes. We inspected the presence of a conserved phenylalanine residue at position 842 (Fig. 4 B) and examined the effect of mutating F842 to alanine on HAC1 splicing. We found that Ire1-F842A was activated to WT levels but also that, like Ire1-D828A, it showed sustained HAC1 mRNA splicing over an extended time course (Fig. 4 C). The effect of the mutation was specific to phenylalanine at 842 as the mutation of arginine at position 843 on the activation loop to alanine had no effect on Ire1 RNase activity, demonstrating that not all Ire1 mutants with significant RNase activity result in the phenotype found in Ire1-D828A. Collectively, these results suggest that the activation loop of Ire1 is important not only during activation of Ire1 but also during the attenuation of HAC1 mRNA splicing. In addition to simple dephosphorylation, Ire1-D828A and -F842A phenotypes have revealed the presence of another critical event for the Ire1 attenuation process.

Proper attenuation of HAC1 splicing during UPR signaling must be important for cell physiology, although very little has been described so far. To determine the functional consequences of the sustained UPR, we examined the ability of either WT or D828A cells to grow under the UPR-induced conditions. Both WT and D828A cells grew similarly in the absence of Tm (Fig. 4 D). However, D828A cells were reduced ∼25-fold in their ability to survive when grown for an extended amount of time on plates with growth medium containing 0.4 µg/ml Tm. Thus, the inability of D828A cells to attenuate HAC1 splicing in a timely manner diminished the ability of the cells to survive the extended activation of UPR.

Although the importance of Ire1 kinase domain functions in the activation of the UPR has been demonstrated previously (Tirasophon et al., 2000; Papa et al., 2003; Lee et al., 2008), we report here that Ire1 kinase function plays an important role in the attenuation of the activated Ire1 RNase. In addition to the initial responses for the increased demands for ER functions, Ire1 must also remain sensitive to the information that ER functional capacity is reestablished to attenuate the activating signal in a timely manner. The Ire1 kinase domain plays a role in transmitting the ER luminal information to Ire1 RNase during the attenuating phase of the UPR. We have found that mutating several residues within the Ire1 kinase domain, D828, F842, S8440, S841, and T844, resulted in an essentially similar phenotype of cells becoming insensitive to the reestablishment of the ER functional capacity and having the sustained presence of the spliced HAC1 mRNA. This was in contrast to WT cells in which removal of Tm, an ER stress agent, facilitated the disappearance of the spliced HAC1 mRNA. The importance of the timely attenuation of the UPR was supported by the inability of Ire1-D828A cells to sustain their growth, even though these cells are able to mount HAC1 mRNA splicing at the initial phase.

One interesting finding is that mutations in multiple different residues of the Ire1 kinase domain showed similar phenotypes of the sustained presence of the spliced HAC1 mRNA. Simultaneously, in addition to the mutants we report here, a new study independently shows that another kinase-inactive mutant Ire1 (D797N or K799N) is unable to attenuate Ire1 RNase (see Rubio et al. in this issue). As these residues are also projected toward the kinase catalytic cleft, they may provide the framework for conformational changes needed for attenuating Ire1 RNase activity. For example, the involvement of the DFG motif in a more direct catalytic function beyond nucleotide binding (Johnson et al., 1996; Huse and Kuriyan, 2002) has been predicted from structural studies with insulin receptor kinase (IRK) and Src and Abl kinases (Till et al., 2001; Levinson et al., 2006). The ability of the DFG motif to affect activation loop conformation is also supported by the finding that the D828A mutation converts the nonphosphorylatable activation loop mutant (Ire1-S840A/S841A/T844A) into an active RNase (Fig. 4 A).

Conversely, in IRK (Till et al., 2001), alteration of the IRK activation loop residue (IRK-D1161A) induced a structural disorder in the IRK nucleotide-binding pocket, mimicking a transition state. Similarly, the conserved activation loop phenylalanine (F842) on the Ire1 activation loop may be involved in the establishment of certain critical hydrophobic interactions that allow a unique conformational link with the nucleotide-binding pocket. Thus, a specific conformational link between the residues within the nucleotide binding and the activation loop, including the DFG motif and phosphorylation sites, may provide the molecular bases for using the evolutionarily conserved kinase domain for both activation and attenuating signals in response to changes in the luminal domain of Ire1.

Several signaling pathways use multiple strategies for signal attenuation (Dikic and Giordano, 2003). In mammalian cells, a recent study has shown that the spliced form of XBP1 mRNA goes away after 12 h of a Tm incubation time course (Lin et al., 2007). In contrast, other UPR inducers, such as PERK or ATF6, remain active for a while, suggesting that Ire1 was inactivated even before the ER capacity recovery. In addition to the splicing of XBP1 mRNA, however, the UPR in mammalian cells appears to be negatively regulated by the XBP1 protein coded from the unspliced form of XBP1 mRNA (pXBP1(U); Yoshida et al., 2006). As pXBP1(U) concentration increases at later times, heterodimer formation with a spliced form (pXBP1(S)) causes pXBP1(S) to be transported to the cytoplasm where it undergoes proteasome-mediated degradation. Such a mechanism may allow the cell to inactivate the transcriptional response even before the inactivation of Ire1. In yeast, although no evidence for the production of pHAC1(U), presumably caused by the presence of the translational inhibitory intron within unspliced HAC1, is found, other negative factors generated during the UPR may facilitate UPR down-regulation.

Human diseases that correlate with insufficient UPR activation are rapidly increasing. For example, the loss of pancreatic β cells in juvenile diabetic patients (Wolcott–Rallison syndrome) is thought to be caused by the loss of the functional PERK protein, another ER transmembrane kinase involved in UPR induction (Harding et al., 2001). Our finding that Ire1 can be trapped in an activated state for an extended length of time by modulation of the kinase domain may provide a basis for the development of novel strategies for augmenting UPR insufficiency in some human diseases.

Yeast strains carrying various forms of IRE1 were generated by the standard yeast transformation protocol (Elble, 1992). Mutant forms of IRE1 were generated using the QuickChange kit (Agilent Technologies) and confirmed by sequencing the entire coding region. Tables I and II summarize the yeast strains and plasmids used in this study. The KAR2 plasmid was a gift from D. Ng (National University of Singapore, Singapore).

All yeast strains were grown in synthetic complete media supplemented with inositol at a final concentration of 50 µg/ml. Tm (EMD) was added to a final concentration of 1 µg/ml to induce the UPR when required. Yeast strains were grown overnight to saturation with an OD₆₀₀ of ∼1.5. The following morning, cells were diluted to 0.75–0.8 × 10⁷ cells/ml and allowed to recover for one doubling time in fresh media to reach an OD₆₀₀ of ∼0.5. Upon addition of Tm, cells were collected for analysis of RNA at the time points indicated in the figures.

Total RNA was isolated according to the hot phenol method (Köhrer and Domdey, 1991). RNA was separated by electrophoresis on 1.5% agarose gels containing 6.7% formaldehyde and transferred to a Duralon-UV membrane (Agilent Technologies) as previously described (Cox and Walter, 1996). Hybridization was performed at 65°C overnight in church buffer containing 0.5 M Na₂HPO₄, 1 mM EDTA, and 7% SDS, pH 7.5. Quantitation of Northern blots was performed on an imager system (Typhoon 9400; GE Healthcare). All probes were labeled with α-[³²P]deoxy-CTP using a DNA-labeling system (Megaprime; GE Healthcare). The probe used for detecting both spliced and unspliced forms of HAC1 mRNA was as described previously (Papa et al., 2003). The KAR2 probe was generated by PCR of a 0.5-kb fragment of the coding region. The percentage of HAC1 splicing was calculated as the percentage of spliced over total HAC1 mRNA.

CPY catalyzed the hydrolysis of its substrate BTPNA (Sigma-Aldrich) to give the yellow product p-nitroaniline, whose production was followed by an increase in absorbance at 410 nm. Cells were permeabilized in a buffer containing 100 mM Tris, pH 7.6, and 0.2% Triton X-100 for 30 min at 37°C in the presence of BTPNA. Upon permeabilization, the BTPNA gained access to mature vacuolar CPY for the cleavage reaction, which was measured by absorbance value in 410 nm. The absorbance value obtained from UPR-uninduced cells was set to 100%, and the percentage of functional CPY activity was calculated.

The GST-tagged WT Ire1-(linker kinase tail [LKT])p containing the entire cytosolic domain of Ire1 (aa 556–1,115) was expressed and purified as described previously (Sidrauski and Walter, 1997). The fusion protein was purified using glutathione–Sepharose beads (GE Healthcare). All the mutant forms of Ire1 were prepared essentially in a similar manner. In vitro autophosphorylation of various forms of the recombinant Ire1p was performed as described previously (Nock et al., 2001). In brief, the recombinant Ire1 was incubated in kinase buffer (10 mM Hepes, pH 7.5, 5 mM Mg(OAc)₂, 25 mM KOAc, and 1 mM DTT) containing 200 µM of unlabeled ATP and 167 µCi γ-[³²P]ATP (7,000 Ci/mmol; ICN) at 30°C for 30 min before being analyzed on 7% SDS-PAGE. Levels of various forms of the recombinant Ire1 proteins were examined by staining the gel with SYPRO ruby staining based on the manufacturer’s instructions (Invitrogen).

Pellets from cells with an OD₆₀₀ of 3 were resuspended in 100 µl SUME (1% SDS, 8 M urea, 10 mM MOPS, pH 6.8, and 10 mM EDTA) with protease inhibitors (1 mM PMSF, 10 µg/ml tosyl phenylalanyl chloromethyl ketone, leupeptin, and pepstatin). The total cell extract was prepared by vortexing in the presence of 100 µl of 0.5-mm acid-washed glass beads at 4°C. 100 µl of 2× USB (75 mM MOPS, pH 6.8, 4% SDS, 200 mM DTT, and 8 M urea) was then added, and the samples were incubated at 65°C for 10 min. Proteins were isolated on SDS-PAGE and were transferred on to a nitrocellulose membrane, which was probed for either Ire1-WT or Ire1-D828A using an anti-HA antibody (Santa Cruz Biotechnology, Inc.).

An increasing amount (5–120 µM) of the fluorescent nucleotides N-methylanthraniloyl nucleotide (MANT)-ADP and MANT-ATP (Invitrogen) were mixed with 1.4 µM recombinant Ire1(LKT) in 25 mM Tris, pH 7.5, 0.4 M NaCl, 5% glycerol, and 10 mM MgCl₂ before being excited at 280 nm. The emission was measured at 440 nm using a spectrofluorometer (Fluoromax; Horiba Scientific) as described previously in detail (Ni et al., 2000). The equilibrium dissociation constants were determined by plotting the changes in fluorescence at 440 nm as a function of nucleotide concentration (either MANT-ATP or MANT-ADP, 5–120 µM) and are the concentrations of a MANT nucleotide at which 50% of the maximum fluorescence changes was measured as described previously (Adams and Taylor, 1992; Ni et al., 2000).

Fig. S1 shows that Ire1-D828A is expressed and binds ATP at a level similar to Ire1-WT but is an inactive kinase. Fig. S2 shows that nucleotide binding and phosphorylation are important for the full activation of Ire1 RNase.

We are grateful to Drs. Douglass Forbes, Susan Taylor, Randy Hampton, Tony Hunter, and Tracy Johnson for scientific discussions and comments on the manuscript. We also thank Dr. Davis Ng for the pMR713 KAR2 plasmid. We also thank Dr. Joseph Adams for his help with the MANT nucleotide-binding assays.

This work was supported by the American Cancer Society and National Institutes of Health (General Medicine) to M. Niwa.