Cellular stress conditions evoking the UPR are cumulatively called ER stress; this has been generally believed to mean accumulation of unfolded proteins in the ER. We and others previously reported that the ER chaperone BiP associates with and deactivates Ire1 under nonstress conditions (Bertolotti et al., 2000 ; Okamura et al., 2000 ; Kimata et al., 2003 ). ER stress causes dissociation of BiP from Ire1, which then leads to clustering of Ire1 (Kimata et al., 2007 ). The formation of clusters of Ire1 contributes to gathering and efficient splicing of its target RNA (Aragón et al., 2009 ; Korennykh et al., 2009 ). The molecular mechanism by which BiP dissociates from Ire1 is still obscure, since some observations (reviewed in Kimata and Kohno, 2011 ) argue against a simple model in which unfolded proteins accumulating in the ER deprive Ire1 of BiP. Thus it is unclear whether unfolded proteins are directly involved in the dissociation of BiP from Ire1.

On the other hand, further studies argue for direct involvement of unfolded proteins in activation of yeast Ire1. As illustrated in Figure 1A, the BiP-binding site is located on subregion V of the luminal domain of Ire1 (Kimata et al., 2004 ). Since Ire1 mutants carrying deletions in subregion V are still regulated by ER stress almost as well as wild-type Ire1, BiP is unlikely to be the principal determinant of Ire1's activity (Kimata et al., 2004 ; Pincus et al., 2010 ). Credle et al. (2005) reported the x-ray crystal structure of the core stress-sensing region (CSSR; Figure 1B) of Ire1, which, when dimerized, forms a cavity that may capture unfolded proteins. Interaction of unfolded proteins with Ire1 has also been supported by our study showing a recombinant CSSR protein inhibits aggregation of denatured model proteins in vitro (Kimata et al., 2007 ). A truncation mutation, ΔIII (see Figure 1C for mutation location) abolished this ability of the CSSR. The direct interaction of unfolded proteins with Ire1 is likely to contribute to activation of Ire1, since ΔIII Ire1 was poorly activated by ER stress induced by treatment of cells with the thiol-reducing agent dithiothreitol (DTT) or N-glycosylation inhibitor tunicamycin (Kimata et al., 2007 ). We therefore propose that Ire1 is activated upon ER stress conditions via two sequential steps, namely, its clustering, followed by the dissociation of BiP and direct interaction of unfolded proteins with the Ire1 cluster.

Nevertheless, it is unlikely that the UPR is activated only in order to cope with unfolded proteins accumulated in the ER. Comprehensive gene expression studies showed that UPR transcriptionally induces not only ER chaperones and protein degradation factors, but also various proteins, including ER translocon components, COPII coat components, and enzymes for ER membrane biogenesis via the Ire1-HAC1 pathway (Travers et al., 2000 ; Kimata et al., 2006 ). Deletion of either the IRE1 or the HAC1 gene confers auxotrophy for inositol, an important component of phospholipids, on yeast cells. It should be noted that expression of Ire1 is induced by depletion of inositol from yeast culture (Cox et al., 1997 ). As reviewed in Rutkowski and Hegde (2010) and argued in the Discussion, the UPR signal pathway is activated and required under various physiological and pathological situations in mammals, some of which do not seem to be tightly related to protein load in the ER.

This property of ΔIII Ire1 allowed us to explore involvement of the unfolded-protein association in activation of Ire1 under various stress stimuli. We noticed that, unlike DTT or tunicamycin treatment, inositol depletion activated ΔIII Ire1 at almost the same level and time course as wild-type Ire1 (Figure 5C). On the basis of the results from gene deletions (Figures 8), we propose that two different types of stress stimuli activate the UPR in distinct manners.

Initially, stress stimuli causing accumulation of unfolded proteins in the ER activate Ire1 via its interaction with the aberrant proteins. Cellular expression of aberrant proteins (CPY*-GFP in this study); DTT or tunicamycin treatment; and deletion of any of SCJ1, SPC2, STE24, ALG3, EOS1, PMT2, ERD1, ERV14, and ERV25 seem to fall into this category, since these stress stimuli activate wild-type Ire1 more strongly than ΔIII Ire1. Given that Scj1, Spc2, Alg3, and Eos1 can each function as a BiP cochaperone, a subunit of the signal peptidase complex or factors in N-glycosylation, the loss of these proteins is likely to produce aberrant proteins in the ER. Also, Pmt2 was recently reported to be an O-mannosyltransferase that participates in protein quality control in the ER (Goder and Melero, 2011 ). Evocation of the UPR by ERD1 deletion may be explained by mislocalization of BiP (Hardwick et al., 1990 ), which is also likely to impair protein folding in the ER. Erv14 and Erv25, although not required for COPII-coated vesicle formation per se (Matsuoka et al., 1998 ), contribute to ER-to-Golgi transport through their physical interaction with cargo proteins (Muñiz et al., 2000 ; Powers and Barlowe, 2002 ). We therefore think that loss of Erv14 or Erv25 causes stacking of cargo proteins in the ER that is sensed by Ire1 in a manner similar to sensing of unfolded-protein accumulation. However, we can offer no explanation as to why the STE24 deletion exhibits such a UPR activation profile.

In contrast, deletion of certain other genes, as well as inositol depletion, activates wild-type and ΔIII Ire1 almost equally. We propose that these stress stimuli lead to membrane- or lipid-related aberrations, which activate Ire1 even without its interaction with unfolded proteins. OPI3 and ISC1 encode enzymes that metabolize phospholipids. Scs3 is a member of the ER-located FIT family of proteins, which are involved in fat storage (Kaderei et al., 2008 ). Although the function of Arv1 and its metazoan orthologues is still obscure, their loss is reported to perturb intracellular distribution of lipidic components (Kajiwara et al., 2008 ; Tong et al., 2010 ). Since Get1 is a member of the GET complex, which mediates insertion of tail-anchored proteins from cytosol to the ER membrane (Schuldiner et al., 2008 ), its loss is likely to damage primarily the membrane, rather than the lumen of the ER. It also seems reasonable that membrane homeostasis is perturbed when genes involved in glycosylphosphatidylinositol (GPI) anchor biogenesis, such as LAS21 or BST1, are deleted. Although it is possible that such mutations also adversely affect the integrity of GPI-anchored proteins, this is not directly sensed by Ire1, because the mutations activate wild-type Ire1 and ΔIII Ire1 equally. Interestingly, SEC28, but not ERV14 or ERV25, falls into this category, although all three genes are involved in intracellular vesicle transport. While a possible role of Erv14 and Erv25 is to function as cargo-protein receptors in the ER-to-Golgi transport (Muñiz et al., 2000; Powers and Barlowe, 2002), Sec28 is a component of the coatomer (Duden et al., 1998 ; Kimata et al., 1999 ), which per se is responsible for formation of transport vesicles in Golgi-to-ER retrieval transport. We therefore speculate that loss of Sec28 primarily impairs membrane composition, but not protein flux, in the ER.

Inositol is one of the main components of phospholipids, and we propose that UPR evocation by inositol depletion is also due to a membrane-related abnormality. In other words, although inositol depletion is reported to lead to altered conditions in the ER lumen (Merksamer et al., 2008 ), this is not the main factor that activates Ire1 upon this stress stimulus. This is because, as mentioned in the next paragraph, inositol depletion causes up-regulation of bZiP-Ire1, which lacks the authentic Ire1 luminal domain, as well as wild-type Ire1 (Figure 9C and D). Moreover, while DTT exposure produced BiP-containing protein aggregates, inositol depletion failed to exhibit such an effect (Figure 7A). This observation implies that inositol depletion does not damage protein folding in the ER lumen as potently as DTT. It also should be noted that myriocin, an inhibitor of sphingolipid biosynthesis, compromises activation of Ire1 by inositol depletion but not by DTT exposure (Figure 7B). Although the link between sphingolipids or biosynthetic intermediates of sphingolipids and the UPR is obscure, this finding strongly suggests a tight relationship between cellular membrane conditions and Ire1 activation by inositol depletion.

The activation steps of Ire1 upon DTT or tunicamycin treatment have been documented in our previous report (Kimata et al., 2007 ), which proposed that dissociation of BiP from Ire1 causes cluster formation of Ire1, while unfolded proteins interact with the Ire1 clusters for full activation. To explore the activation steps of Ire1 that are independent of unfolded proteins, we used inositol depletion as a model stress condition. BiP dissociation and cluster formation of Ire1 were also observed upon inositol depletion (Figure 6, A and B). The ΔIΔV mutation abolishes BiP binding and causes constitutive clustering of Ire1 (Oikawa et al., 2007 ; Kimata et al., 2007 ). In the present study, we noticed that inositol depletion up-regulates ΔIΔV Ire1 and, albeit slightly weakly, ΔIΔIIIΔV Ire1. We therefore conclude that, instead of interaction between unfolded proteins and Ire1, an undisclosed molecular event occurs to give full activation of clustered Ire1 molecules upon inositol depletion. Since bZIP-Ire1 responded well to inositol depletion (Figure 9, C and D), the luminal domain of Ire1 is unlikely to contribute to sensing this stress stimulus.

Unlike the quick activation of wild-type Ire1 observed upon treatment with DTT or tunicamycin (peak activation within 30 min or 1 h after stimulus onset; Figure 5, A and B), activation of wild-type Ire1, and also of ΔIII Ire1 and bZIP-Ire1, by inositol depletion is rather slow (Figures 5C and ​and9C).9C). Although this time lag may be due to residual cellular inositol stock, it is also possible that the cellular mechanism activating Ire1 upon inositol depletion per se is slow in responding. We therefore speculate that the cellular response to membrane-related ER stress need not be acute. Conversely, the response of Ire1 to unfolded-protein accumulation is so acute that cells can cope with such a stress condition quickly. It should be noted that DTT and tunicamycin somehow activated ΔIII Ire1 and bZIP-Ire1, albeit more slowly and weakly than observed for wild-type Ire1 (Figures 5, A and B, and ​and9C).9C). We speculate that these conventional ER stressors also disturb membrane homeostasis and then slowly activate Ire1, even when it carries a mutation abolishing its ability to associate with unfolded proteins. Excess accumulation of aberrant proteins in the ER may concomitantly damage the ER membrane.

Mammals carry two Ire1 paralogues, of which IRE1α is the major version expressed ubiquitously. According to the x-ray crystal structure reported by Zhou et al. (2006) , the luminal domain of IRE1α carries a cavity-like structure, which, however, is too narrow to capture unfolded proteins, unlike that of yeast Ire1. Moreover, we failed to demonstrate in vitro interaction between unfolded proteins and a recombinant luminal-domain fragment of IRE1α (Oikawa et al., 2009 ). While one explanation for these observations is that the size of the IRE1α cavity is somehow regulated and enlarged when it is required to capture unfolded proteins, it is also possible that IRE1α lacks the ability to associate with unfolded proteins. Meanwhile, BiP association/dissociation is not likely to be the sole determinant of IRE1α activity, since an IRE1α truncation mutant lacking the major BiP-binding region is still up-regulated by ER stress (Oikawa et al., 2009 ). We therefore think that the new mechanism of stress sensing presented here may also contribute to activation of IRE1α upon ER stress.

How does the cytosolic (or transmembrane) domain of Ire1 sense stress stimuli? It is an attractive idea, as proposed by Wiseman et al. (2009) , that small molecules interact with the cytosolic domain of Ire1 to up-regulate Ire1. However, the newly found ligand-binding pocket (Wiseman et al., 2009 ) is unlikely to be involved in activation of Ire1 upon inositol depletion (Figure S4). In the case of mammalian IRE1α, its activity has been reported to be modulated by association with proteins other than BiP (Hetz et al., 2006 ; Luo et al., 2008 ; Lisbona et al., 2009 ; Qiu et al., 2010 ). This matter may have to be addressed in order to further elucidate the new mode of stress sensing by Ire1.

It should be noted that various reports have touched upon evocation of the UPR by lipid- or membrane-related stimuli. Pineau et al. (2009) and Deguil et al. (2011) proposed that an imbalanced fatty acid composition could activate yeast Ire1. In contrast to the concept we are now presenting, Pineau et al. (2009) argued that activation of Ire1 by the fatty acid imbalance involves accumulation of unfolded proteins, since it is attenuated by the chemical chaperone 4-phenyl butyrate. In mammals, obesity induces ER stress and may lead to type 2 diabetes (Özcan et al., 2004 ), the symptoms of which are alleviated by chemical chaperones (Özcan et al., 2006 ). Cunha et al. (2008) and Wei et al. (2006) described lipid-induced ER stress followed by cellular damage of pancreatic β cells and liver cells. Furthermore, as reviewed by Zheng et al. (2010) , the UPR pathway is involved in lipogenesis of mammalian cells, as well as of yeast. It is widely accepted that IRE1α and its downstream target XBP1 are required for homeostasis maintenance of and/or differentiation to cells or tissues secreting high levels of proteins, such as antibody-producing plasma cells, pancreatic β cells, and placenta, in which the ER membrane is highly proliferated (Iwakoshi et al., 2003 ; Lipson et al., 2008 ; Iwawaki et al., 2009 , 2010). One possible explanation for this is that excess influx of proteins into the ER leads to activation of ER stress sensors, including IRE1α. However, this idea is not supported by the observation that mutant B lymphocytes engineered to lack antibody production also activate XBP1 upon their differentiation to plasma cells (Hu et al., 2009 ). We therefore think that upon various lipid- or membrane-related stimuli, membrane stress per se and concomitant accumulation of unfolded proteins are recognized by ER stress sensors, including Ire1, in a complex manner.