I. INTRODUCTION

A. The Endoplasmic Reticulum

The endoplasmic reticulum (ER) is an essential subcellular compartment responsible for the synthesis and folding of proteins that traffic through the secretory pathway. This organelle is the site for executing and regulating many posttranslational modifications that ensure protein function (34, 46). The ER serves as the major intracellular calcium store, and also plays a crucial role in the biosynthesis of cholesterol, steroids, and other lipids, regulating second messenger signaling.

After passing the ER quality control process, properly folded proteins traffic in vesicular membranes to various organelles, the cell surface, or the extracellular space. After translocation to the ER, nascent proteins fold assisted by a complex network of foldases, chaperones, and cofactors (34, 72, 146). Some of the better described folding catalysts are immunoglobulin binding protein BiP (also known as glucose-regulated protein 78, Grp78) and Grp94, the protein disulfide isomerases Erp57 and PDI, calnexin, and calreticulin, among many other enzymes (40, 126, 137).

Although large families of distinct chaperones and foldases are expressed in the ER lumen, little is known about how protein folding networks are arranged at the ER, and what controls the substrate specificity for these protein folding pathways. The best described folding/quality control network at the ER is the calnexin and calreticulin cycle. In this pathway, ERp57 associates with calnexin and/or calreticulin to assist the folding and quality control of a subset of glycoproteins. PDI and ERp57 are part of a group of oxidoreductases responsible for catalyzing the formation, reduction, and isomerization of disulfide bonds (158). In theory, most glycol-polypeptides are released from calnexin/calreticulin/ERp57 in a folded state. After passing this quality control step, proteins undergo glucose trimming (49), to undergo transport to other compartments. A percentage of the newly synthesized glycoproteins do not reach a final folded state and enter into an additional folding cycle likely to consist of a disulfide reshuffling. Terminally unfolded/misfolded proteins are delivered for proteasome-mediated degradation by the ER-associated degradation (ERAD) pathway (Fig. 1) (11, 181). Although the calnexin cycle is proposed to be essential for the folding of glycoproteins, different studies using genetic manipulation suggest that this pathway may be relevant for the folding and quality control of only a small subset of glycoproteins (158, 188). In general, most of the folding components and networks expressed at the ER remain poorly characterized.

B. Protein Folding Stress at the ER

The extremely high concentration of proteins within the ER (100 mg/ml) renders this organelle's environment very susceptible to protein aggregation (126, 170). Different physiological and pathological perturbations interfere with protein folding processes in the ER lumen, leading to accumulation of unfolded or misfolded proteins, a cellular condition termed “ER stress.” Protein folding stress triggers the activation of an adaptive reaction to cope with ER stress termed the unfolded protein response (UPR). The UPR is a complex signal transduction pathway that conveys information about protein folding status in the ER lumen to increase protein folding capacity and decrease unfolded protein load. If these mechanisms of adaptation and survival are insufficient to recover ER homeostasis, cells undergo cell death by apoptosis.

Physiological fluctuations in the demand for protein synthesis and secretion as part of the development and differentiation of specialized secretory cells lead to the occurrence of ER stress. Physiological levels of ER stress engage the UPR to maintain protein homeostasis, and directly contribute to the development and maintenance of a differentiated and functional state of specialized secretory cells such as pancreatic beta cells, B cell lymphocytes, and salivary gland cells (reviewed in Ref. 109). ER stress is also triggered by many conditions that alter protein homeostasis networks. These perturbations include altered protein maturation, modification of chaperone function, expression of disease-related mutant proteins, decreased ER calcium content, and redox alterations among others (9, 162). There is growing biomedical interest in investigating the regulatory mechanisms underlying UPR signaling and the development of strategies to target this pathway, since there is substantial evidence for the involvement of chronic ER stress in many diseases, including neurodegeneration (54, 127), diverse forms of cancer (83, 122), diabetes (101), and proinflammatory conditions (176). Here, we summarize current thinking on this topic and provide our view of the impact of ER stress in physiological and pathological settings.

II. UPR SIGNALING

A. UPR Stress Sensors

The major impact of the UPR is the maintenance of protein homeostasis in the context of high unfolded protein load. To accomplish this, the UPR reprograms the transcriptome modulating the expression of a vast number of secretory pathway-related genes. Gene expression profiling studies indicate that the UPR regulates a variety of genes involved in specific secretory pathway-related processes including protein entry into the ER, folding, glycosylation, redox metabolism, protein quality control, protein degradation, lipid biogenesis, and vesicular trafficking (Fig. 2).

In yeast, the UPR is controlled by only one signaling pathway, mediated by a type I transmembrane ER protein known as IRE1p (inositol-requiring transmembrane kinase/endonuclease) (22, 123, 166). In higher eukaryotes, the UPR is mediated by at least three classes of stress sensors including IRE1α and IRE1β, PERK (PKR-like ER kinase), and ATF6α and ATF6β (activating transcription factor 6) (Fig. 3). These three UPR branches control the expression of specific transcription factors and signaling events that modulate a variety of UPR downstream responses, orchestrating adaptation to ER stress. In this review we address in detail the current proposed mechanisms underlying UPR signaling, stress sensing, and downstream effectors as well as their relevance to physiology and disease. We also speculate about possible mechanisms that explain the transition from adaptive UPR responses to apoptosis under chronic ER stress. We will provide a brief general overview of UPR signaling and then focus primarily on the IRE1α branch of the UPR, probably the best characterized of the three arms of the ER stress pathway.

B. PERK and ATF4 Signaling

PERK is a type I ER transmembrane protein kinase that upon activation inhibits general protein translation into the ER through the inactivation of the initiation factor eIF2α by serine 51 phosphorylation. This phosphorylation inhibits the guanine nucleotide exchange factor eIF2B, a complex that recycles eIF2α to its active GTP-bound form. This inhibitory effect of translation helps alleviate ER stress by decreasing the overload of misfolded proteins (162). In addition, eIF2α phosphorylation is reversible, but the identity of the phosphatases involved in this process is not clear. However, two components of this negative regulation pathway include growth arrest and DNA-damage inducible protein-34 (GADD34) and constitutive repressor of eIF2α phosphorylation (CReP) (162).

In addition, eIF2α phosphorylation preferentially increases the translation of selective mRNAs that contain inhibitory upstream open reading frames (uORFs) within their 5′ untranslated region (UTR) that prevent translation in unstressed cells. The most studied of these genes is activating transcription factor 4 (ATF4) (162). ATF4 is a transcription factor that upregulates a subset of UPR genes that function preferentially in amino acid import, glutathione biosynthesis, and resistance to oxidative stress (43, 110, 167, 208) (Fig. 2). Attempts to define the universe of PERK-dependent UPR target genes in mammalian cells revealed that nearly half of the PERK-dependent targets are ATF4 independent (15, 43), suggesting the existence of other PERK downstream effectors. It is not clear if other PERK substrates exist, but PERK may also phosphorylate nuclear factor (erythroid-derived 2)-like 2 (NRf2) (26, 27). Nrf2 is a transcription factor that controls the regulation of oxidative stress responses (27, 47). Although controversial (155), it has been suggested that PERK may also control the expression of NF-κB in an ATF4-independent manner. In addition to its response to ER stress, eIF2α is also phosphorylated by other kinases linked to amino acid starvation and double-stranded RNA accumulation, indicating that the cellular responses controlled by this UPR signaling branch are not solely restricted to protein folding stress.

C. ATF6 Signaling

Another UPR pathway is mediated by ATF6α and ATF6β, type II ER transmembrane proteins which encode bZIP transcription factor domains in the cytosolic region of the protein (45, 196). ATF6 is synthesized as an inactive precursor, retained at the ER by a transmembrane spanning segment. Under ER stress conditions, ATF6 translocates to the Golgi where it is processed, first by site 1 protease and then in an intramembrane region by site 2 protease. This proteolytic processing releases its cytoplasmic domain, ATF6f (a fragment of ATF6), which operates as a transcriptional activator that upregulates many UPR genes related to ERAD and protein folding, among other processes (45, 92, 192) (Fig. 2). Interestingly, there are many other putative ATF6 homologs identified, which are modulated by ER stress in specific tissues, including CREBH (160, 199), OASIS (79, 124), CREB4 (171), LUMAN/CREB3 (96), and BBF2H7 (80, 159). All of these ATF6-related bZip factors are processed at the Golgi as described for ATF6, but their function, if any, in the UPR is poorly characterized.

D. IRE1α Signaling

Although IRE1α initiates the most conserved signaling pathway of the UPR, little is known about its biochemical regulation. IRE1α is a Ser/Thr protein kinase and endoribonuclease that catalyzes the unconventional processing of the mRNA encoding the transcriptional factor X-Box binding protein-1 (XBP1) (18, 92, 195) (Fig. 4A). A 26-nucleotide intron of xbp1 mRNA is spliced out by IRE1α in mammalian cells, leading to a shift in the codon reading frame of the mRNA that generates a new COOH-terminal end that contains a potent transactivation domain (Fig. 4A).

The spliced version of XBP1 (termed XBP1s) controls the upregulation of a general pool of UPR-related genes involved in different processes including protein folding, protein entry to the ER, and ERAD (91, 164). In addition, XBP1s indirectly regulates the biogenesis of the ER and Golgi by enhancing the activity of enzymes related to phospholipid biosynthesis (164, 168, 169). The universe of XBP1s target genes differs in different tissues or under different conditions of ER stress (1), possibly reflecting the fact that XBP1 can interact and heterodimerize with other transcription factors (see below).

III. CHRONIC ER STRESS AND THE UNFOLDED PROTEIN RESPONSE: THE APOPTOSIS PHASE

Under chronic ER stress, cells undergo cell death by apoptosis (35, 153), where several pro-apoptotic members of the BCL-2 family of proteins are essential for the elimination of irreversibly damaged cells (Fig. 5). The BCL-2 family of proteins is formed of pro- and anti-apoptotic members classified by the presence of up to four BCL-2 homology (BH) domains (28). The initiation of intrinsic apoptosis is mediated by the activation of pro-apoptotic “multidomain” members (i.e., BAX and BAK) at the mitochondria, leading to the release of cytochrome c and activation of the caspase cascade (198). Upstream regulators of BAX and BAK contain only the BH3 domain termed “BH3-only proteins” (17, 150, 198). Initial studies identified the transcriptional upregulation of two BH3-only proteins, PUMA and NOXA, in cells undergoing ER stress, which operate as relevant pro-apoptotic inducers (Fig. 5) (95, 147). Furthermore, upregulation of BIM at the transcriptional and posttranslational level also contributes to apoptosis by irreversible ER stress (144). IRE1α also activates the ASK1 and JNK pathway, which is a relevant factor for the induction of apoptosis by ER stress (69, 74, 116).

Sustained PERK has a bifunctional role in adaptation to ER stress and apoptosis (98). Expression of ATF4 and possibly ATF6 controls the upregulation CCAAT/enhancer binding protein (C/EBP) homologous protein (CHOP), also termed growth arrest and DNA damage-inducible gene (GADD153). CHOP has a pro-apoptotic activity with an unclear mechanism of action. CHOP may trigger cell death by downregulation of BCL-2 (117) and the transcriptional upregulation of BIM (144), PUMA (36), and GADD34 (110, 144) (Fig. 5) . Many other regulators of apoptosis in the setting of ER stress have been described and are reviewed elsewhere (165). We and others have speculated that the transition between adaptive UPR programs and the elimination of irreversibly damaged cells by apoptosis depend in part on the duration of ER stress stimulation (191).

IV. REGULATION OF IRE1α ACTIVATION AND SIGNALING

Recent studies have uncovered many important aspects of how protein misfolding is detected by UPR stress sensors. In this section we discuss the key findings that have revealed the biochemical basis of IREα activation by ER stress, the dynamics of this process in terms of assembling IRE1 oligomers, and its inactivation under prolonged ER stress conditions. In addition, we summarize recent interesting models proposed for the targeting of XBP1/HAC1 to IRE1 in the cytosol. As the reader will notice, there are striking differences in the way IRE1 signals in yeast versus mammalian cells.

A. Mechanism of IRE1α Activation and Stress Sensing

Many different models have been proposed to explain the mechanism of IRE1α activation. The initial model suggests that under normal conditions, the ER chaperone BiP binds to the luminal domain of IRE1α or the yeast homolog IRE1p, maintaining the protein in an inactive state as a monomer (13, 76) (Fig. 6A). Conversely, in cells undergoing ER stress, BiP is released and binds to unfolded proteins. This event allows IRE1α multimerization and autophosphorylation, activating the RNase domain through a putative conformational change (Fig. 6A). Unexpectedly, mutagenesis analysis of the BiP binding site did not alter the ability of IRE1p to detect the accumulation of misfolded proteins, although it ablated BiP (Kar2 in yeast) binding (77). Additional information about the possible mechanism of IRE1α/IRE1p activation was provided by structural studies of the ER luminal domain of yeast and human IRE1 proteins. Analysis of the yeast IRE1p structure through bioinformatic and mutagenesis analysis suggested a model where misfolded proteins may actually bind the NH₂-terminal region of IRE1p, promoting its oligomerization through a structure similar to an MHC groove (Fig. 6B) (23). In this scenario, misfolded proteins may directly bind to the luminal region of yeast IRE1p. Although this model was not directly tested in the study, the idea was reinforced by a study using recombinant luminal domain of yeast IRE1p, observing an association with unfolded proteins in a cell-free system (75). IRE1p activation was then proposed to be initiated by BiP dissociation from IRE1p, leading to IRE1p dimerization and cluster formation possibly by binding misfolded proteins (Fig. 6, B and C) (75). However, no studies so far have demonstrated the binding of unfolded proteins to IRE1p in vivo as a mechanism for unfolded protein recognition and UPR activation in yeast.

Many other studies have also explored the impact of the binding of human IRE1α to BiP. In contract to yeast IRE1p, recombinant fragments of the luminal domain of IRE1α do not interact with unfolded proteins in a cell-free system (135). This model correlates well with the prediction that the MHC-like groove observed in the human IRE1α ER luminal domain may not be compatible with the binding of a peptide as indicated in the crystal structure through in silico analysis (206). These biochemical data suggest that the mechanism of sensing ER stress by yeast and mammalian IRE1 may be different (Fig. 6) (135). These limited biochemical observations highlight the need for more protein-protein interaction studies to fully address the possible direct impact of the MHC-like groove of IRE1α on the activation of this stress sensor.

A recent report provides a new interpretation for the role of BiP in IRE1p activation. Careful kinetic analysis of IRE1p signaling in a mutant yeast strain where IRE1p partially binds BiP revealed a higher susceptibility to undergo activation under mild or low ER stress conditions (143). Remarkably, the kinetics of IRE1p inactivation were significantly delayed in the IRE1p mutants. The authors proposed a model where three pools of IRE1p exist, an inactive subpopulation in equilibrium with an active unfolded protein-bound pool and a third inactive set sequestered by BiP for inactivation (Fig. 6D) (138, 143). Thus BiP binding to or release from IRE1p is not instrumental for activating the UPR as previously proposed, but it may modulate the sensitivity and dynamics of IRE1p activity, adding more complexity to our understanding of IRE1p signaling. Finally, two recent, elegant studies suggested that the purpose of IRE1p auto-phosphorylation is to turn off IRE1p under conditions of prolonged ER stress rather than to activate it (20, 157). Inactivation of the IRE1p phosphorylation site leads to prolonged activation of IRE1p, chronic ER stress, and decreased yeast viability. Although IRE1α is an essential component of the mammalian UPR, structure-functional analysis of its phosphorylation sites in mammals is far from complete.

B. Cluster Formation by IRE1

Analysis of the tridimensional structure of the cytosolic domain of IRE1p also revealed interesting insights about its mechanism of signaling (82, 93). Peter Walter's group was able to determine the structure and assembly of IRE1p oligomers and observed an unexpected high-order rod-shaped organization (82). This oligomeric assembly is critical for IRE1p signaling because different forms of IRE1p dimers may position the kinase domain for trans-autophosphorylation, generating the RNase active site. Interestingly, the authors also predicted that this oligomeric assembly may also generate an additional surface for mRNA binding (82).

Other studies suggested that IRE1α predominantly forms dimers upon activation in mammalian cells, in contrast to PERK activation that is associated with the generation of high-molecular-weight complexes (13). Structural studies of the cytosolic domain of human IRE1α indicate the presence of dimeric structures, not large oligomers (2, 93). However, a recent study using overexpression of a GFP-tagged form of IRE1α indicates that mammalian IRE1α also oligomerizes in the ER membrane, correlating with the onset of IRE1α phosphorylation and XBP1 mRNA splicing (94). The authors suggested that the kinase and RNase domains activate cooperatively, containing more than four IRE1α molecules in the complex for activation (94). Inactivation of IRE1α over time is associated with the dissolution of IRE1α clusters, IRE1α dephosphorylation, correlating with a decline of XBP1 mRNA splicing (94). Assembly of higher order IRE1α clusters may represent a conserved mechanism for its signaling. Nevertheless, all experiments were performed in an overexpression system, and it remains to be determined if endogenous IRE1α clusters in physiological and/or pathological models of ER stress.

C. mRNA Targeting to IRE1

Overall, mechanistic differences are proposed for the targeting of XBP1 and HAC1 mRNAs to IRE1α and IRE1p, respectively. In yeast, in nonstressed cells most of the unspliced HAC1 mRNA is cytoplasmic and remains bound to the ribosome, but it is not translated. Upon induction of ER stress, the unspliced HAC1 mRNA colocalizes with IRE1p clusters in an IRE1p-dependent manner (4). In addition, HAC1 mRNA targeting to IRE1p requires a specific secondary structure of HAC1 mRNA (4). The splicing of XBP1 mRNA also occurs in the cytosol (179).

In mammals, unspliced XBP1 protein (XBP1u) is expressed but it is very unstable and is degraded by the proteasome, and hence, it is in general undetectable in normal cells (18, 128). However, it was shown that XBP1u may have a function in targeting of the XBP1u mRNA to the ER membrane (193) (Fig. 4B). Although unstable, XBP1u could interact with membranes, and possibly the ER, bringing XBP1 and IRE1α in close proximity. The membrane association of XBP1u was mediated by the presence of a well-conserved hydrophobic region at its COOH terminus (193) (Fig. 4A). Remarkably, a recent study from Kohno's group (194) demonstrated that transient ribosomal translational pausing is important for membrane targeting of XBP1u and further XBP1 mRNA splicing. This process was also mediated by an evolutionarily conserved sequence in the XBP1u COOH-terminal region (194). In contrast to the yeast study, IRE1α expression is not necessary for the colocalization of the XBP1 mRNA with the ER membrane. Only two groups have addressed the possible mechanism of mRNA targeting to IRE1 in the cytosol. More studies are needed to understand this important event and to define its relevance in vivo. It is interesting to note that previous studies using XBP1-deficient animals described a hyperactivation of IRE1α in liver and intestinal tissue associated with splicing of the truncated XBP1 mRNA (70, 89), suggesting that expression of a functional XBP1u is not absolutely required for splicing.

V. XBP1s BIOLOGY AND REGULATION

XBP1 mRNA splicing is a major event mediating UPR responses. Data from gene expression profile analyses indicate that the universe of XBP1s target genes and downstream effects may vary depending on the cell type and the nature of the stress stimuli. Although XBP1s orchestrates many essential processes for adaptation to ER stress, it is only recently that studies have uncovered possible regulatory effects at the posttranslational level, in addition to the interaction with other important transcription factors in the control of gene expression. This emerging topic is discussed below.

A. XBP1-Dependent Transcriptional Reprogramming

As mentioned, XBP1s regulates a subset of UPR-induced genes that participate in folding, quality control, and ERAD. XBP1s target genes initially identified in mammalian MEF cells were chaperones such as p58IPK, ERdj4, HEDJ, and the protein disulfide isomerase PDI-P5; the ERAD-related gene EDEM; and genes involved in glycosylation such as RAMP4 (91). Interestingly, in this study XBP1s did not affect the upregulation of classical UPR-target genes such as Grp78/BiP and Grp94. Other studies in plasma cells confirmed the role of XBP1 in secretory cell function (164). In terminally differentiated B cells, additional XBP1s target genes were identified that operate in protein targeting to the ER, ER translocation of proteins, folding, the ERAD pathway, glycosylation, and vesicular trafficking through the secretory pathway (164).

Interestingly, gene expression profile analysis in neurons indicates that XBP1s may control a distinct set of genes in different cell types (68). We and others have also investigated the impact of XBP1 in the expression of target genes in neuronal cultures, and identified Grp58, PDI, and Herp (73), as well as GABAergic markers as XBP1s target genes (44). Using genome-wide promoter binding assays, another study described a network of genes regulated by XBP1 using cultures of skeletal muscle and B cells. A core group of genes were confirmed as targets of XBP1s that are related to secretory pathway function and folding (1). This screen also identified a group of unexpected targets that relate XBP1 to diseases affecting the brain and muscle, in addition to genes related to DNA damage and repair. Overall, one may propose that XBP1s regulates a broad array of genes involved in almost every aspect of ER function and physiology. Based on the variability observed in most gene expression profile analyses, further studies are required to define the universe of XBP1 target genes and to identify the stimuli (pathological and physiological) that modulate them.

B. XBP1s Transcriptional Partners and Regulation

Combinatorial interactions among transcription factors are critical to directing gene expression in a tissue-specific manner. Despite the fundamental role of XBP1 in UPR responses, current knowledge of the regulation of XBP1 activity is limited. Some studies indicate a specific requirement for XBP1 expression in the transcriptional activity of ATF6f through the formation of heterodimers (192), although XBP1s was also proposed to operate mostly as a homodimer in the control of transcription (192). Recent studies identified an interaction between the p85α regulatory subunit of phosphatidylinositol 3-kinase (PI3K) with XBP1s in an ER stress-dependent manner (141, 189). This physical association is relevant for metabolic control in diabetes models (141, 189). In addition, XBP1s was proposed to negatively control the expression levels of the transcription factor Forkhead box O1 (FOXO1) through a physical interaction, which also modulated glucose metabolism (207). These three studies defined the impact of XBP1 interaction partners at the biochemical level in addition to addressing their physiological impact in vivo. Future efforts are needed to uncover the possible impact of FOXO1 and p85α on adaptation to protein folding stress in more classical models of ER stress.

Other reports suggest that p300/CBP-associated factor (PCAF) interacts with XBP1s through its COOH-terminal domain, which may be relevant for host-virus interactions between a cellular factor, XBP1, and the transcriptional regulation of the HTLV-1 virus (86). XBP-1s also binds estrogen receptor α in a ligand-independent manner (32). However, the relevance of these interactions in vivo is still unclear.

Posttranslational modifications of XBP1s were recently proposed to control its activity as a transcription factor (Fig. 4B). XBP1s is a target of acetylation and deacetylation mediated by p300 and SIRT1, respectively (183). p300 increases acetylation and protein stability of XBP1s and enhances its transcriptional activity (183). In addition, XBP1s is SUMOylated, mainly by PIAS2 (protein inhibitor of activated STAT2) at two lysine residues located in the COOH-terminal transactivation domain (21). Ablation of these SUMOylation events significantly enhances the transcriptional activity of XBP1s (21) (Fig. 4B). Although these preliminary findings are interesting, it is still not known if XBP1s activity is altered by posttranslational modification under conditions of ER stress.

Another point of regulation was proposed at the level of XBP1s degradation, where XBP1u markedly accumulates after prolonged ER stress and forms a heterodimeric complex with XBP1s that is rapidly degraded by the proteasome (197). This regulatory loop may help to turn off downstream UPR responses (Fig. 4B). This model remains to be confirmed.

C. XBP1 and Organelle Biogenesis

In activated B cells, induction of XBP1s results in an increase of ER and Golgi content (164). The effects of XBP1s on organelle biogenesis are also observed in other organelles including lysosomes and mitochondria as evidenced by an increase in cell size (164). Consistent with these in vitro phenotypes were the effects observed in vivo when XBP1 was deleted in exocrine cells of the salivary glands, pancreas, intestinal Paneth cells, and gastric epithelial mucous neck cells. Loss of XBP1 in these tissues resulted in a massive disorganization of the elaborate ER network normally present in these cell types and subsequent failure of cell development (60, 90).

It is still unknown how ER and Golgi expansion is induced under conditions of ER stress. The main barrier in identifying a mechanism arises from the fact that XBP1s target genes identified to date do not include major genes related to membrane phospholipid biosynthesis, with the exception of Chkb, which is involved in the biosynthesis of phosphatidylcholine (164, 168, 169).

VI. ADDITIONAL IRE1α DOWNSTREAM SIGNALING COMPONENTS

In addition to upregulating UPR target genes, mammalian IRE1α signals through additional pathways including c-Jun NH₂-terminal kinase (JNK) and NF-κB. In addition, a new function for IRE1α, IRE1-mediated or regulated decay (RIDD), has recently been described to exist in both Drosophila and mammalian cells. Here, IRE1 degrades many mRNAs, some of which encode secretory pathway-related proteins. These XBP1-independent signaling events are proposed to modulate a vast spectrum of physiological processes ranging from apoptosis/survival, macroautophagy, proliferation, and metabolism, to inflammatory processes. This is an active area of research, and recent work has begun to uncover the detailed mechanisms by which IRE1α governs these pathways. Additional work is directed to understanding their relevance in adaptation to stress in vivo.

A. Control of Stress Kinases and Alarm Signaling Pathways

IRE1α controls the initiation of several downstream signaling pathways in addition to processing XBP1 mRNA (Fig. 2). The cytosolic domain of activated IRE1α recruits the adaptor protein TNFR-associated factor 2 (TRAF2), which then activates the apoptosis signal-regulating kinase 1 (ASK1) pathway (131, 180). Other reports indicate that IRE1α may also initiate the activation of stress pathways downstream of p38, ERK (130), and NF-κB (59). Cell-based studies suggested that these effects were mediated by the binding of the SH2/SH3 containing adaptor proteins Nck and the inhibitor κB kinase, respectively. However, the possible impact of these downstream UPR signaling branches in the context of ER stress is still not well defined, and most of the reports monitor these kinases as stress markers only.

B. Regulation of Autophagy

IRE1α controls the levels of macroautophagy possibly through the activation of JNK under ER stress conditions (25, 133) and also by proteasome inhibition (33). Macroautophagy, here referred to as autophagy, is a survival pathway essential for nutrient starvation conditions through the recycling of intracellular components (85, 121). It has been speculated that autophagy may serve as a mechanism to eliminate damaged organelles and abnormal protein aggregates under ER stress conditions (10, 25, 108). Interestingly, an initial report indicated that activation of autophagy by ER stress in MEFs is dependent on the kinase domain of IRE1α, and not affected by the RNAse/XBP1 signaling branch (133). An RNAi screen using fly cells revealed that knocking down XBP1 or its target genes increases basal autophagy levels in the absence of any stress (6). We recently reported that knocking down XBP1 in neuronal cells leads to increased basal autophagy flux without any additional stimuli (54). Similar observations were reported in the central nervous system of xbp1-deficient mice (54). A major XBP1s-target gene includes EDEM1, an essential component for ERAD. The impairment of ERAD activity by XBP1 deficiency was associated with the enhancement of autophagy in neurons. In this scenario, accumulation of abnormally folded proteins at the ER due to ERAD impairment may operate as the signal to induce autophagy (115).

C. ER-Associated Degradation

As mentioned, some XBP1s target genes are related to the ERAD pathway and the ER translocon, including HERP, EDEM, and Sec61. A protein-protein interaction screen revealed a physical association between the ubiquitin specific protease (USP) 14 and IRE1α (125). Interestingly, IRE1α interacts not only with USP14, but also with other ERAD components including DERLIN-1, DERLIN-3, SEL1, and HRD1 (125). These data suggest that IRE1α may form a protein complex with the ERAD machinery. However, a limitation of this study was the reliance on in vitro experiments using cell lines and protein overexpression. The functional impact of these interactions remains to be established.

D. IRE1α-Mediated mRNA Decay

Attempts to identify new substrates of IRE1p (132) or IRE1α (129) RNAse activity yielded only HAC1 and XBP1 mRNAs as positive hits. However, another study revealed that active IRE1α in insect cells controls the degradation of mRNAs of genes encoding ER proteins that are predicted to be difficult to fold under stress conditions (55, 56). A subset of genes was shown to be downregulated during ER stress in an IRE1α-dependent and XBP1-independent manner, which were proposed to be direct targets of the IRE1α ribonuclease activity (55, 56). The authors speculated that the selective mRNA degradation by IRE1α occurs in a dynamic way, where the misfolding of a nascent protein during its translation may locally activate IRE1α's RNase domain to degrade mRNAs being translated (Fig. 7). Two additional studies confirmed the occurrence of IRE1α-dependent mRNA decay in mammalian cells (41, 55). Interestingly, artificial dimerization of IRE1α in the absence of ER stress did not trigger mRNA decay, but initiated XBP1 mRNA splicing, suggesting that these two functions of IRE1α are regulated by different factors (55). This alternative function of IRE1α may operate to selectively decrease the production of proteins that challenge the ER at the folding level and hence alleviate stress. These three reports used cell-based assays to uncover IRE1α-dependent mRNA decay. It remains to be determined if this pathway has any physiological relevance in vivo in the context of ER stress.

Interestingly, the reports describing IRE1α-dependent mRNA decay suggested that the pool of mRNAs targeted by IRE1α depends on the cell type analyzed and not the nucleotide sequence of the mRNA. The authors failed to find any consensus sequence for cleavage in all of the RNAs identified after diverse bioinformatic analyses. In contrast, a recent study further investigated a possible mechanism for the recognition of target mRNAs by IRE1α in vitro using a recombinant IRE1α. The authors employed a combination of an in vitro cleavage assay with an exon microarray analysis, in addition to a genome-wide screening for IRE1α cleavage targets (136). Thirteen novel mRNAs were identified as candidate targets of IRE1α. Sequence analysis suggested the presence of a putative consensus sequence that when accompanied by a stem-loop structure is associated with IRE1α-mediated cleavage (136). These observations still need to be corroborated in living cells.

VII. THE UPROSOME: MODULATION OF IRE1α ACTIVITY BY COFACTORS

Although IRE1α and PERK ER luminal stress-sensing domains are structurally similar and functionally interchangeable (105), the kinetics of IRE1α and PERK signaling are very different. This fundamental difference in their temporal signaling patterns is proposed to have a substantial impact on cell fate decisions that balance the rate of adaptation/survival to elimination of damaged cells by apoptosis under stress conditions (191). These differences may now be explained in part by the discovery of novel specific regulators of IRE1α that modulate the downstream outcomes of its signaling. These regulators control the rate of IRE1α activation/inactivation possibly through the formation of a dynamic scaffold referred to conceptually as the UPRosome. In this section we discuss in detail recent findings that have uncovered unanticipated regulators of IRE1α signaling.

A. Positive Regulation of IRE1α Signaling by Apoptosis-Related Proteins

Accumulating evidence indicates that IRE1α activation is specifically regulated by a set of different interaction partners that may determine the threshold of activation and inactivation (Fig. 8). We have reported that IRE1α signaling is induced by the expression of some pro-apoptotic BCL-2 family members such as BAX and BAK (50). BAX and BAK expression at the ER modulates the intensity of IRE1α signaling but does not alter PERK signaling (50). This regulation is proposed to be mediated by the formation of a protein complex between the cytosolic domain of IRE1α and BAX/BAK (Fig. 3). In addition, BAX and BAK modulated IRE1α signaling in vivo in liver upon ER stress induction (50). Similarly, another report suggested that the enforced expression of the pro-apoptotic BH3-only proteins BIM and PUMA at the ER membrane triggers the activation of the JNK pathway in an IRE1α- and BAK-dependent manner (78).

The expression of the pro-apoptotic protein ASK1-interacting protein 1 (AIP1) also increases the amplitude of IRE1α signaling as demonstrated in cellular and in vivo models of ER stress (106). AIP1 expression did not affect the signaling of the PERK pathway, ruling out possible general effects on protein folding at the ER lumen (106). These data suggest an alternative function of certain pro-apoptotic genes in the regulation of early UPR signaling events, in which they play a bifunctional role in cell death (late event) and adaptation to stress (early event) (103) (Figs. 5 and 8). It still remains to be determined if all of these IRE1α regulators are part of the same pathway or if they modulate the UPR through independent mechanisms.

B. Regulation of IRE1α by Other Activators

IRE1α signaling is also instigated by the expression of the ER-located protein-tyrosine phosphatase 1B (PTP-1B) in cellular (38) and animal models (31). Similar to AIP1 and BAX/BAK, PTP-1B deficiency does not affect PERK-related signaling in cell culture models. The cytosolic chaperone heat shock protein 72 (Hsp72) was recently shown to decrease cell death under ER stress conditions (39). Unexpectedly, Hsp72 enhances XBP1 mRNA splicing and its downstream responses (Fig. 8) (39). Regulation of the UPR by Hsp72 is mediated by the formation of a protein complex between Hsp72 and IRE1α. Remarkably, Hsp72 enhances the RNase activity of IRE1α in a cell-free system, suggesting direct binding and regulation of its activity (39). These results provide for the first time an interconnection between cytosolic chaperones and the UPR. Moreover, the modulation of XBP1 mRNA splicing by Hsp72 had a substantial impact on the efficiency of protein secretion.

C. Inactivation of IRE1α

As mentioned, XBP1 mRNA splicing levels are decreased after chronic or prolonged ER stress. In contrast, PERK signaling is sustained over time (97), which may negate the pro-survival effects of the IRE1α and XBP1 pathway (98, 144). Correlative data initially indicated that the IRE1α and ATF6 pathways are negatively modulated by the expression of the ER-located protein BAX inhibitor-1 (BI-1) in vivo (8, 84). Under ischemic conditions in liver and kidney, BI-1-deficient mice displayed increased levels of XBP1s, ATF6f, and JNK phosphorylation, without altering eIF2α phosphorylation (8). We reported an unexpected function of BI-1 in turning off IRE1α signaling. BI-1-deficient cells displayed sustained XBP1 mRNA splicing and enhanced downstream responses (104). Remarkably, a marked alteration was observed in the inactivation phase of IRE1α signaling over time in BI-1-deficient cells. The inhibition of IRE1α by BI-1 was validated in vivo in fly and mouse models of ER stress (104) to affect physiological processes driven by XBP1. The regulatory activity of BI-1 on IRE1α was also corroborated in the setting of obesity and diabetes in vivo (7).

BI-1 interacts with the cytosolic domain of IRE1α, and this association can be reconstituted in vitro where BI-1 reduces the endoribonuclease activity of IRE1α in a cell-free assay (104). The ER-associated RING-type E3 ligase bifunctional apoptosis regulator (BAR) interacts with BI-1, promoting its proteasomal degradation (156), which specifically enhances IRE1α signaling (Fig. 8) (156).

Another study indicates that IRE1α is specifically phosphorylated on Ser⁷²⁴ by glucose stimulation (145), and unlike the activation of IRE1α by inducers of ER stress, glucose-induced phosphorylation does not cause a shift of the IRE1α protein as detected by Western blot (100, 145) and does not release the inhibitory interactions with BiP (100). The scaffold protein receptor for activated C-kinase 1 (RACK1) interacts with IRE1α in a glucose-stimulated or ER stress-dependent manner (145). RACK1 mediates the assembly of a protein complex containing IRE1α, RACK1, and the protein phosphatase 2A (PP2A) (145). This complex regulates the dephosphorylation of IRE1α by PP2A (Fig. 8) (145), thereby inhibiting glucose-stimulated IRE1α activation and attenuating IRE1α-dependent increases in insulin production.

D. IRE1p Allosteric Site

Overall the data described in the previous sections suggest that IRE1α is actively regulated by the binding of cofactors that modulate its enzymatic activity. It is not known if all these regulators (i.e., BAX, BAK, Hsp72, BI-1, RACK1, AIP1, etc.) bind to the same IRE1α domain, and how this interaction relates to IRE1α enzymatic activity at the structural level. Interestingly, a pharmacological screen identified the existence of a possible allosteric site on yeast IRE1p (190). The flavonol quercetin was shown to activate the RNase domain in vitro and potentiate activation by ADP, a natural ligand that binds to the IRE1p nucleotide-binding cleft (190). Interestingly, enzyme kinetic studies and the visualization of the structure of a cocrystal of IRE1p bound to ADP and quercetin identified a new ligand-binding pocket (190). It remains to be determined if this possible allosteric site is relevant for the regulation of IRE1p in physiological conditions and if it is functional in mammalian IRE1α.

E. IRE1α Stability

In addition to the regulation of IRE1α RNAse activity, its protein stability is also controlled, which may affect the amplitude of UPR responses. HSP90 interacts with the cytosolic domain of IRE1α, and inhibition of HSP90 interfered with the association of HSP90 and IRE1α. Dissociation of HSP90 and IRE1α leads to its degradation by the proteasome (111). HSP90 also regulates the stability of many other protein complexes having an important role in cell signaling. It remains to be determined if HSP90 regulates the dynamics of the IRE1α interactome. In addition, the impact of HSP90 on IRE1α stability needs to be confirmed in other cell types and in vivo.

VIII. ROLE OF THE IRE1/XBP1 PATHWAY IN ORGAN PHYSIOLOGY

Generation of genetically modified mice has uncovered fundamental functions of the IRE1α/XBP1 axis of the UPR in many physiological processes and diseases. Most of the evidence available points out a key role of XBP1 in mastering secretory cell function in diverse tissues, in addition to orchestrating lipid metabolism, glucose homeostasis, and inflammatory processes. In agreement with these observations, manipulating IRE1α signaling has already proven efficacious in altering disease severity and progression in diabetes, cancer, autoimmunity, neurodegeneration, and other pathologies in rodent models (see Table 1). In the next section we discuss studies that have defined the impact of the UPR in vivo.

A. Overlapping But Unique Functions of IRE1 and XBP1 In Vivo

Initial studies of mice with deficiencies in ER-signaling components revealed the function of specific pathways in various organs and tissues. In mice, IRE1α inactivation results in widespread developmental defects leading to embryonic death at 12.5 days of gestation (201). Analysis of IRE1α conditional knockout mice revealed that the early embryonic lethality of the germline knockout was caused by the loss of IRE1α in the placenta (66). Gene expression studies in IRE1α- and XBP1-deficient placenta suggested that the IRE1/XBP1 pathway contributes to the placental expression of the carcinoembryogenic antigen (CEA) family of proteins (134). The role of these proteins in the placenta is unclear, and it is unknown whether CEA expression by the IRE1/XBP1 pathway is required to sustain placental functions. The placental trophoblast produces many secretory molecules including lactogens and growth factors; it is therefore possible that the placental defects observed in IRE1α-deficient animals are caused by impaired secretion of factors supporting embryonic development. Interestingly, the deletion of IRE1α under the control of the Mox2 promoter leads to viable mice that are born at near-Mendelian ratios (66). In these mice, CRE expression and subsequent IRE1α deletion are effective in all epiblast-derived cells, leading to IRE1α deficiency in virtually all embryonic and adult cells except extraembryonic tissues such as the placenta, demonstrating that aside from extraembryonic tissues, IRE1α is dispensable for proper embryonic development.

While both IRE1α and XBP1 are partners that function in the same pathway, early studies examining XBP1-deficient mice revealed phenotypes distinct from IRE1 deletion. Inactivation of XBP1 results in embryonic lethality at 12–13.5 days of gestation caused by severe liver hypoplasia and a resulting fatal anemia (148). To rescue embryonic death, selective expression of an XBP1 transgene in liver of XBP1-deficient mice was generated using a liver-specific promoter (88). Mice lacking XBP1 in virtually all organs except the liver died shortly after birth from a severe impairment in the production of pancreatic digestive enzymes leading to hypoglycemia and death (88). The expansion of the ER in salivary glands and pancreatic exocrine cells was severely impaired resulting in a complete disorganization of the ER network and impaired production and release of zymogen granules (88), demonstrating the crucial role of XBP1 in the development of highly secretory exocrine cells. Interestingly, mice in which IRE1α deficiency is limited to epiblast-derived cells have normal salivary glands and only a partial defect in the pancreas (65). Because IRE1α is believed to be essential for XBP1 activation, it is unclear why IRE1α deficiency does not mirror the phenotypes observed in XBP1-deficient mice. Differences in experimental conditions such as background strain and usage of different tissue-specific Cres may be partly responsible. More likely is the possibility that these differences indicate that IRE1α has functions independent of XBP1 and, vice versa, that XBP1 has functions independent of IRE1α. IRE1α is known to engage multiple pathways beyond XBP1 activation (51); however, the activation of XBP1 is believed to be exclusively IRE1α dependent. It is therefore possible that some aspects of the phenotypes observed in XBP1-deficient animals are caused by the deficiency in the unspliced form of XBP1 that can bind ATF6 and therefore regulate expression of a subset of genes independently of IRE1α. Alternatively, XBP1 could be activated independently of IRE1α, although there are no data that support that idea. Finally, we cannot exclude the possibility that part of the phenotype observed in XBP1 deficiency is caused by constitutive IRE1 hyperactivation. It is known for example that IRE1-mediated RNA decay may promote death of stressed cells (41). Therefore, in the absence of XBP1, stressed cells, unable to cope with the stress, might activate IRE1-mediated cell death and tissue disorganization. This model would predict more severe abnormalities in XBP1-deficient compared with IRE1α-deficient animals.

B. XBP1 Function in Secretory Cells and Lipogenesis

The role of XBP1 in the maintenance and differentiation of secretory cells in the salivary gland and exocrine pancreas of XBP1-deficient mice is consistent with its role in ER/Golgi biogenesis and phospholipid synthesis (164, 168, 169). Indeed, the first evidence that XBP1 was involved in the differentiation of secretory cells came 10 years ago with the finding that XBP1-deficient B cells are unable to differentiate into antibody-producing plasma cells (149). Antibody secretion in vivo in response to antigenic challenge is impaired in XBP1-deficient mice, an activity later shown to be directly dependent on XBP1 splicing in stimulated B cells (63, 177). Interestingly, XBP1 activation in B cells lacking IgM is still present (58), suggesting that secretory cells activate XBP1 as a part of the differentiation program rather than as a consequence of massive immunoglobulin synthesis and secretion. Of note, IRE1α deficiency in B cells led to a much earlier block in B-cell differentiation at the pre-B cell stage, emphasizing the lack of complete concordance in pathway components.

XBP1 is also required for Paneth cell function (70). When exposed to bacteria or bacterial antigens, Paneth cells become secretory and release a number of antimicrobial enzymes and proteins into the lumen of the crypt, thereby contributing to maintenance of the gastrointestinal epithelial barrier. XBP1 deletion in intestinal epithelial cells triggered spontaneous enteritis secondary to Paneth cell dysfunction leading to increased susceptibility to induced colitis (70). In agreement with these reports, conditional IRE1α deficiency caused structural abnormalities of the pancreatic acinar and salivary tissues, in addition to attenuated serum levels of immunoglobulin (65).

XBP1 is also involved in the homeostasis and function of nonsecretory cells. XBP1 expression in the liver is required for normal fatty acid and sterol synthesis (37, 89). The deletion of XBP1 in the liver led to significant decrease of triglycerides, cholesterol, and free fatty acids without causing fatty liver (89). XBP1 activation upon high-carbohydrate diet feeding triggers the transcription of key lipogenic genes in hepatocytes, suggesting that XBP1 directly promotes synthesis of lipids in the liver (89). Although liver-specific deletion of IRE1α had minor effects on basal liver physiology (65, 200), the mice developed severe hepatic steatosis upon ER stress induction, expression of a misfolding-prone human blood clotting factor VIII, or after partial hepatectomy (200).

The XBP1/IRE1 pathway in the liver may be regulated by circadian rhythms. XBP1 and IRE1α have been shown to be activated rhythmically every 12 h in hepatocytes (24). Animals lacking a circadian clock exhibit constitutive activation of the IRE1/XBP1 pathway. These findings are consistent and correlate with asynchronous expression of enzymes involved in lipid metabolism and triglyceride accumulation in the liver (24).

C. XBP1 and Immunity

Beyond its role in the maintenance of secretory cells and lipid metabolism, XBP1 was shown to modulate immune responses. XBP1 is important for the development and survival of dendritic cells (64) and in the immune response to challenge by pathogens that activate the Toll-like receptors (TLRs) (112). Toll-like receptors (TLRs) are single, membrane-spanning receptors that recognize structurally conserved molecules derived from microbes to orchestrate immune responses. In the absence of detectable ER stress, TLR4 and TLR2 activation by microbial products specifically promotes the phosphorylation of IRE1α and the activation of XBP1 (112). IRE1α activation by TLR engagement does not induce ER stress target genes, but is required for optimal and sustained production of proinflammatory cytokines in macrophages (112). Consistent with these findings, XBP1 deficiency markedly increases bacterial burden in mice infected with the TLR2-activating pathogen Francisella tularensis. Similarly, infection of Caenorhabditis elegans with pore-forming toxins harboring bacteria leads to the activation of XBP1 and ATF6 to promote immune defense (14). The functional synergy demonstrated between TLR activation and classic pharmacological stressors like tunicamycin in IRE1 activation suggests that discovery of nontoxic pharmacological ER stressors may be useful in augmenting vaccine efficacy. TLR signals also actively repress the other UPR branches, a strategy that may prevent prolonged, damaging ER stress (173). Another study demonstrated that C. elegans infected with Pseudomonas aeruginosa activate IRE1 and XBP1 through the innate immune kinase PMK-1 (151). In this model it was shown that XBP1 loss of function decreases the survival of infected worms, probably due to aberrant activation of PMK-1 (151).

Altogether these studies clearly demonstrate that XBP1 and IRE1 are key in the maintenance of various physiological processes including lipid metabolism, maintenance of secretory cell function, and innate immunity. A better understanding of the mechanisms and specific roles of the various ER-signaling pathways in these processes will be key in appreciating the link between stress responses and physiological processes.

IX. ACTIVATION OF THE IRE1/XBP1 PATHWAY AND DISEASE PATHOLOGY: FROM CANCER TO ATHEROSCLEROSIS

A. Cancer

Although the engagement of the ER stress response is essential to adapt to alterations in the ER, abnormal and sustained ER-stress can contribute to development of pathologies such as neurodegenerative diseases. Moreover, the ER stress response may modulate the pathological state of preexisting diseases such as some cancers.

Cancer development is often associated with a range of cytotoxic conditions like hypoxia, nutrient deprivation, and pH changes caused by poorly vascularized solid tumor cells. These conditions trigger a set of cellular stress response pathways including the ER-stress response that helps the cell to cope with the stress. Many aspects of the ER-stress response are cytoprotective, and several studies indicate that this response has a crucial role in tumor growth (48, 107).

Studies have suggested that IRE1 and XBP1 are involved in cancer progression. XBP1 overexpression has been demonstrated in numerous human cancers such as breast cancer, hepatocellular carcinoma, and pancreatic adenocarcinomas (81). Transformed MEFs or HT1080 that are deficient in XBP1 have an impaired ability to grow as tumor xenografts in SCID mice compared with XBP1 proficient cells (154). Moreover, sustained active XBP1 overexpression in a transgenic mouse model suggested that XBP1 is capable of neoplastic transformation of plasma B cells into multiple myeloma (19). The recent identification of a putative IRE1 RNase inhibitor that displays significant anti-myeloma activity in a model of human multiple myeloma xenografts (140) further suggests that the IRE1/XBP1 pathway is a promising target for anti-cancer therapy. Additional IRE1α inhibitory compounds have also been recently described (182).

Folding problems in the ER can compromise the traffic and function of a variety of proteins resulting in the degradation of the unfolded protein and/or the activation of ER stress. Many human diseases are associated with perturbations in the ER machinery and the accumulation of unfolded proteins in the ER (142). At least some of these pathologies may involve the engagement of the IRE1/XBP1 pathway.

B. Neurodegenerative Diseases

Multiple studies suggest that the UPR may be involved in the modulation of neurodegenerative diseases (114). A common feature of many neurodegenerative diseases is the accumulation and deposition of misfolded proteins leading to decreased neuronal function and viability (61). Upregulation of ER stress markers has been observed in post mortem brain tissues and cell culture models of disorders such as Parkinson's disease, amyotrophic lateral sclerosis (ALS), Alzheimer's disease, Huntington disease, and Creutzfeldt-Jacob disease (5, 99, 113). Whether ER stress is a cause or a consequence of neurodegeneration and how ER stress modulates disease progression are still poorly understood. On one hand, the ER stress response may be protective and provide relief in cases of mild injury, while on the other hand, it could contribute to neuronal toxicity in cases of sustained stress and continuous accumulation of misfolded proteins. Consistent with a model in which components of the ER stress response favor neurotoxicity, XBP1 deficiency in the nervous system was shown to be protective in a mouse model of ALS due to an enhanced clearance of mutant SOD1 aggregates by macroautophagy, a cellular pathway involved in protein degradation (54, 127). In contrast, the specific deletion of XBP1 in neurons has no detectable effect on the development of the central nervous system and did not trigger any obvious impairment or enhance the progression of a prion disease model (53). In contrast, PERK deficiency exacerbated ALS disease progression (184), indicating that defining the participation of the UPR in neurodegeneration is very complex.

C. Diabetes

ER stress may be involved in the development of congenital diabetes. Neonatal diabetes, for example, is a rare genetic disorder developing in the first weeks of life characterized by insulin-demanding hyperglycemia (172, 187). An autosomal dominant form of this disease is caused by mutations in one of the two alleles of the insulin gene. These mutations cause improper folding of the proinsulin probably leading to sustained ER stress and β-cell loss of function. A similar mutation affecting mouse insulin in the Akita diabetes model has been shown to cause insulin misfolding leading to β-cell disruption and diabetes (67). How ER stress affects the overall integrity of β-cells is unclear. IRE1α enhances proinsulin synthesis upon acute exposure to high glucose concentrations (100). Intriguingly IRE1α engagement by high glucose in β-cells does not result in XBP1 mRNA splicing, but correlates with the induction of WSF1. This activation is believed to be beneficial to β-cells. In contrast, chronic exposure of β-cells to high glucose causes ER stress and hyperactivation of IRE1, leading to the suppression of insulin gene expression (100). Insulin suppression in stressed β-cells has been proposed to be directly caused by IRE1-mediated degradation of the insulin mRNA (102). Therefore, IRE1-mediated loss of insulin production may constitute a hallmark of ER stress-mediated diabetes and may explain how ER stress mediated by misfolding of one of the two insulin alleles dramatically contributes to the development of insulin-demanding hyperglycemia.

ER stress has also been linked to type 2 diabetes. Obesity-induced ER stress in the liver plays a central role in the development of insulin resistance and type 2 diabetes by promoting JNK activity via IRE1α and dampening insulin receptor signaling (139). XBP1^+/− mice exhibited increased ER stress and increased susceptibility to develop insulin resistance upon high-fat diet-induced obesity (139). It has also been suggested that deficient insulin signaling might directly affect XBP1 activity by dampening nuclear translocation of active XBP1 in the hepatocytes of obese mice (141). This defect in nuclear localization of XBP1 may contribute to the inability to cope with increased stress observed in obese mice. Finally, interaction of XBP1 with other transcription factors may also be important for regulating glucose metabolism and insulin secretion (141, 189, 207).

D. Inflammatory Diseases

A growing number of reports have suggested that ER stress may contribute to inflammation and inflammation-related diseases (57). Rats expressing a misfolding-prone and ER stress-promoting human HLA-B27 protein in macrophages exhibit inflammation in the joints and the intestine (29, 178). Humans with HLA-B27 are at increased risk of developing seronegative spondyloarthropathies such as ankylosing spondylitis (AS), a potentially disabling form of arthritis characterized by spinal inflammation and enthesopathy (174). These observations are consistent with the fact that activation of ER stress has been associated with hyperinflammatory responses in macrophages. Indeed, recent findings demonstrate that ER-stressed macrophages are hyperresponsive to TLR stimulation in an XBP1-dependent manner (112). These findings, in addition to the observation that XBP1 is involved in TLR signaling pathways in the absence of ER stress (112), support the notion that XBP1 is a positive regulator of TLR gene induction and inflammatory programs. The observation that ER stress and the IRE1/XBP1 pathway per se may promote inflammation by regulating the intensity and duration of inflammatory responses is quite intriguing and directs our attention to a broad range of diseases that are associated with inflammation and the upregulation of ER stress-responsive genes (57, 112). Among those diseases, cardiovascular diseases and diseases involving intestinal inflammation are the best characterized.

Increasing evidence indicates that the ER stress response is activated in atherosclerotic lesions including in macrophages and endothelial cells (120, 173). Various pathological mediators of atherosclerosis can trigger ER stress pathways including oxidative stress, oxysterols, pathological levels of intracellular lipids such as cholesterol and saturated fatty acids characteristic of diseases, and conditions associated with obesity (173). Whether ER stress-mediated outputs such as hyperinflammatory reactions and cell death contribute to the development of atherosclerosis in vivo is an exciting question that requires further investigation (57, 173).

The intestinal epithelium is exposed to a diverse array of pathogens and commensal bacteria as well as numerous metabolic products derived from the host and the microbial community. At the same time, the intestinal epithelium secretes via specialized secretory cells such as Paneth cells and goblet cells antimicrobial and regulatory molecules that are key in maintaining tissue homeostasis. As such, the intestine is a key organ where both inflammatory pathways and ER stress pathways are present and crucial. It is therefore not surprising that deficiencies in ER stress pathways can lead to intestinal inflammation (71, 118). IRE1β is a homolog of IRE1α mainly expressed in colonic and gastric epithelial cells whose deficiency is associated with increased susceptibility to dextran sodium sulfate (DSS)-mediated colitis (12). Similarly, mice with XBP1 deficiency in intestinal epithelium develop a spontaneous enterocolitis and display increased susceptibility to DSS-mediated colitis (70). Moreover, in humans, XBP1 polymorphisms were identified as risk factors for the human inflammatory bowel diseases Crohn's disease and ulcerative colitis (70). Other ER stress-causing mutations have been associated with intestinal inflammation (71, 118). One example is illustrated by the study of an ER-residing protein disulfide isomerase (PDI) family member, anterior gradient 2 (AGR2). AGR2 deficiency results in the accumulation of misfolded proteins in the ER (203). Similarly to XBP1-deficient mice, AGR2 mice exhibit severe intestinal inflammation. Part of the inflammation observed in XBP1- and AGR2-deficient mice is probably caused by deficiencies in Paneth cells and goblet cells; however, it is likely that ER stress per se can intersect with the immune response and contribute to the excessive, pathological inflammation observed in these diseases.

X. DISCUSSION

A. The UPRosome as an Integrator of Physiological and Pathological Stress Responses

The UPR is a complex signal transduction pathway, essential for the survival and function of specialized secretory cells. Genetic evidence in different animal models revealed a vital function of UPR components in the development and function of plasma B cells, pancreatic exocrine and endocrine cells, and salivary glands. Remarkably, new functions of the UPR have emerged in other tissues where this pathway modulates lipid and cholesterol levels in addition to regulating the innate immune system. The UPR also participates in balancing global homeostasis by controlling the levels of insulin and its consequences in the liver, in addition to metabolic control at the hypothalamus level. To add more complexity to all these findings, genetic studies targeting specific UPR components indicate that the function and impact of distinct UPR signaling branches may differ drastically when the same tissue is analyzed. In addition to operating in diverse physiological processes, chronic ER stress is linked to a variety of diseases related to abnormal protein folding and ER dysfunction. Thus the UPR is becoming a relevant target for therapeutic intervention in a wide variety of disease conditions.

Little is known about the regulation of UPR signaling by specific stimuli or at the levels of distinct tissues/organs, and how the kinetics and amplitude of signaling of each UPR branch are controlled. Since PERK and IRE1α share functionally similar luminal sensing domains and they are interchangeable without affecting cytosolic signaling (105), we speculate that the specific activation of ER stress sensors in different tissues may be explained by the presence of specific regulators of their activities. We propose a model where an intricate signaling platform docks at IRE1α and maybe other UPR stress sensors to fine-tune its activation threshold to modulate signaling intensity and kinetics of activation/inactivation. This fine tuning of UPR signaling responses is particularly relevant for cell fate decisions by controlling downstream programs that regulate either adaptation to stress or the initiation of apoptosis to eliminate irreversibly injured cells. We refer to this regulatory and signaling complex as the UPRosome (52, 103). This scaffold initiates multiple signaling responses in a highly regulated manner (Fig. 8). It remains to be determined if PERK and ATF6 are also regulated by the binding of specific modulators, forming other distinct UPRosomes. Defining the static and dynamic composition of tissue specific UPRosomes is of particular relevance due to the divergent and fundamental roles of the UPR in cell physiology, in addition to its participation in many important diseases including cancer, neurodegeneration, diabetes, and autoimmunity. A deeper understanding of how the UPR signals and is regulated may provide new therapeutic targets to modulate ER stress levels in human disease.

GRANTS

This work was supported by the following: FONDECYT Grant 1100176, Millennium Institute Grant P09-015-F, Michael J. Fox Foundation for Parkinson's Research, ICGEB, Alzheimer's Disease Foundation, CHDI Foundation, and FONDAP Grant 15010006 (to C. Hetz); FONDECYT Grant 3100033 (to D. Rodriguez); Swiss National Science Foundation Grant 31003A_130476 (to F. Martinon); and National Institute on Aging Grant AI-32412, the Leila and Harold Mathers Foundation, and an anonymous foundation (to L. H. Glimcher).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.