Introduction

The endoplasmic reticulum (ER) is the entry site for proteins destined to the endo/exocytotic pathway, and provides an optimal and unique environment for protein folding, assembly, and disulfide bond formation prior to exposure to the extracellular space. The concentration of proteins within the ER lumen is extremely high, approximately 100 mg/ml.1 The protein synthesis rate is also astonishingly high in professional secretory cells, where it is estimated that hepatocytes and pancreatic exocrine cells synthesize approximately 13 and 2.6 million secretory proteins per minute, respectively.2 To accomplish such a thermodynamically unfavorable process in an overwhelmingly crowded environment, the cell expends a large amount of energy to ensure that this quantitative achievement does not come at the price of quality. Homeostasis within the ER lumen is meticulously monitored and elegantly maintained. A broad spectrum of insults can lead to the activation of a coordinated adaptive program called the unfolded protein response (UPR). In response to the accumulation of unfolded proteins in the ER, the rate of general translation initiation is attenuated, the expression of ER resident protein chaperones and protein foldases is induced, the ER compartment proliferates, and ER-associated degradation (ERAD) is activated to eliminate the irreparably misfolded proteins. When the prosurvival efforts are exhausted, ER-stress related apoptosis commences.

A number of insults lead to protein misfolding in the ER. These include nutrient deprivation, alterations in the oxidation–reduction balance, changes in calcium concentration, failure of post-translational modifications, or simply increases in secretory protein synthesis.3 Pharmacological reagents were initially employed to elucidate how cells cope with immediate and severe challenges to the protein folding quality control system. These approaches have led to our present knowledge of the molecular logic of UPR signaling pathways, which has been thoroughly reviewed.4 However, as analysis of increasing numbers of murine genetic models created with mutations in major UPR components proceeds, it is becoming apparent that a wide diversity of physiological fluctuations and pathological conditions can also disrupt the protein folding efficiency in the ER in a relatively subtle and chronic manner to activate the UPR (Table 1). For professional secretory cells that harbor an extensive, highly evolved ER structure, such as antibody-secreting plasma cells, collagen-secreting osteoblasts, and cells within the endocrine/exocrine organs, emerging evidence suggests that the UPR is indispensable and actively involved to ensure proper function and survival. In this review, we compare the regulation of the classic mammalian UPR upon acute ER stress to the diversified roles that each UPR subpathway plays in different tissues under various, more subtle chronic physiological and pathological conditions. We describe presently outstanding questions in the field and suggest possible experimental approaches to obtain needed answers.

Table 1 Genetic models studied for the physiological roles of the unfolded protein response

The Temporal Sequential Activation of UPR Subpathways upon Acute ER Stress

Response initiation: one master regulator

During the prestress state, nascent-folding-competent polypeptides are maintained in a soluble form by interaction with ER lumenal chaperones. BiP/GRP78 (immunoglobulin heavy chain-binding protein/glucose-regulated protein of molecular weight 78 kDa) is one of the most highly expressed ER resident chaperones. BiP is a member of the heat-shock protein (Hsp70) family. BiP is localized to the ER through an N-terminal cleaved signal peptide and a C-terminal ER retention motif (KDEL). The C-terminal domain of BiP binds to exposed hydrophobic patches on protein folding intermediates with rather low substrate specificity. The N-terminal domain of BiP functions as a peptide-dependent ATPase that uses ATP hydrolysis to promote a conformational change in the C-terminal region that promotes high-affinity peptide binding. Upon exchange of ATP for bound ADP, BiP reverts back to the low-affinity peptide-binding state. In this way, the cell couples substrate binding and release with the expenditure of cellular energy.5, 6, 7

To date there are three identified proximal sensors of the UPR: the PKR-like ER protein kinase/pancreatic eIF2α (eucaryotic translation initiation factor 2, α subunit) kinase (PERK/PEK); the activating transcription factor 6 (ATF6); and the inositol-requiring enzyme 1 (IRE1). (Figure 1) All of these sensors associate with BiP in their inactive states. It is proposed that when ER homeostasis is perturbed, BiP preferentially binds to and is sequestered by unfolded/misfolded proteins that accumulate in the ER lumen. As a consequence, BiP dissociates from the UPR transducers to permit their signaling. The major support for this hypothesis is that BiP is found in association with each stress transducer under nonstress conditions and is released upon accumulation of unfolded proteins.8, 9, 10, 11 In addition, overexpression of BiP inhibits signaling through all three subpathways.12 Unfortunately, there is little direct data to support this model where BiP inhibits UPR activation; however, it is an attractive hypothesis because this model implies that the peptide-dependent ATPase activity of BiP can function as an unfolded protein sensor to mechanistically link UPR activation with the cellular energy status and/or nutrient level.

Figure 1
figure 1

Adaptive signaling of the unfolded protein response. The three proximal UPR transducers, ATF6, IRE1, and PERK, all associate with BiP in their inactivate state. Upon accumulation of unfolded/misfolded proteins in the ER lumen, these sensors are released and activated. ATF6 transits to the Golgi where it is cleaved by S1P/S2P and the cytosolic fragment of ATF6 migrates to the nucleus. Both IRE1 and PERK are oligomerized and autophosphorylated. Phosphorylated IRE1 catalyzes the splicing of XBP1 mRNA, which generates a more potent transcription factor. Activated PERK phosphorylates eIF2α, which attenuates the general translation rate while inducing the translation of selective mRNAs with inhibitory uORFs in their 5′ UTR. The downstream effectors of these three subpathways combinatorially induce the expression of the genes encoding proteins that function to augment the ER protein-folding capacity. Meanwhile, ERAD is accelerated to remove terminally misfolded proteins.

In contrast to the simultaneous activation of the three UPR transducers upon severe, acute ER stress, selective UPR subpathways function under various physiological and pathological ER stresses (see Section The Role of the UPR in Professional Secretory Cell Development and Function). Two groups recently proposed that the ER stress transducers Ire1p and ATF6 themselves are actively involved in the dissociation from BiP and the consequent activation, contrary to being passively competitive deprived of BiP.13, 14 It is also conceivable that there are other regulators involved in response initiation, for example, co-chaperones such as the DnaJ family members that stimulate the ATPase activity of the Hsp70 family members, or other tissue-specific adaptor proteins that may facilitate activation of specific UPR subpathways in response to different physiological conditions. Alternatively, it is possible that the affinity of each UPR transducer for BiP varies from cell type to cell type.

Signal diversification: unique signal transduction from three proximal transducers

Although the three UPR subpathways are simultaneously activated upon severe ER stress, the immediate response occurs through the PERK/eIF2α pathway. PERK is a type I ER transmembrane protein kinase. Upon release from BiP, PERK dimerizes to promote its autophosphorylation and activation. Activated PERK phosphorylates eIF2α to attenuate the rate of general translation initiation.15, 16 The rapid and reversible regulation of mRNA translation provides an ‘emergency brake’ to prevent further synthesis of proteins when the ER lumen is compromised in its protein-folding capacity. Paradoxically, phosphorylation of eIF2α preferentially increases the translation of selective mRNAs that contain inhibitory upstream open-reading frames (uORFs) within their 5′ untranslated region (UTR). The best-studied example in mammalian cells is the eIF2α phosphorylation-dependent translation of activating transcription factor 4 (ATF4) mRNA.17, 18

Coincident with PERK activation upon BiP release, BiP release from IRE1 permits its dimerization and autophosphorylation to activate its site-specific endoribonuclease (RNase) activity.19 The RNase activity of IRE1 initiates splicing of a 26-base intron from X-box-binding protein 1 (XBP1) mRNA. This nonconventional splicing reaction introduces a translational frameshift into the XBP1 mRNA that alters the C-terminus of the protein to create a potent transcription factor.20, 21, 22, 23

ATF6 is a type II transmembrane protein of the ER. BiP release allows ATF6 transit to the cis-Golgi compartment, where it is cleaved by site-1 protease (S1P) and site-2 protease (S2P), the same enzymes that process the sterol-response-element-binding protein (SREBP) upon cholesterol deprivation.24, 25, 26 The cleaved cytosolic N-terminal fragment of ATF6 migrates to the nucleus and acts as an active transcription factor, together with ATF4 and spliced XBP1, to increase the expression of the genes encoding proteins that function to augment the ER protein folding capacity. These gene targets include those encoding ER chaperones such as BiP/GRP78, GRP94, calreticulin, and calnexin and proteins that catalyze protein-folding such as the protein disulfide isomerases PDI, ERP57, and ERP72. In addition, UPR-activated genes stimulate ER biogenesis to compensate for the increased demand for the protein-folding machinery and accelerate ERAD to remove terminally misfolded proteins.27, 28, 29, 30, 31 The specificity and diversity of downstream transcriptional responses upon different conditions is likely a result of combinatorial interactions between these transcription factors at the promoters of target genes.

This transcriptional adaptative program requires mRNA translation and, therefore, a mechanism to recover from the PERK-mediated translational suppression.

Termination of the UPR signaling: negative feedback regulators

Translational recovery is mediated through dephosphorylation of PERK and/or eIF2α. To preclude the destructive translation attenuation mediated by eIF2α phosphorylation upon cell stress, mammalian cells evolved both a constitutively active regulator of p-eIF2α phosphatase, CreP (constitutive repressor of eIF2α phosphorylation), and a stress-induced regulator of p-eIF2α phosphatase, GADD34 (growth arrest and DNA damage-inducible gene 34). Both proteins act as subunits of the protein phosphatase holoenzyme PP1. In addition, p58IPK (58 kDa PKR inhibitor) is a UPR-induced gene product that inhibits the eIF2α kinases PERK and the double-stranded (ds) RNA-activated protein kinase PKR.32, 33, 34, 35, 36 It is unknown how the mRNAs of GADD34 and p58IPK escape translational repression before dephosphorylation of eIF2α occurs, but CreP may mediate this function.

In yeast, inactivation of Ire1p/Hac1p pathway is achieved by an Ire1p phosphatase, Ptc2p.37 The mechanism of deactivation of the ATF6 and IRE1/XBP1 pathways in higher eukaryotes has not been established. However, as the UPR remains activated, the level of BiP increases so that it can bind and prevent further activation of ATF6, IRE1, as well as PERK. The mechanisms for inactivation of ATF6 and XBP1 transcriptional functions are also not known, but may involve targeted degradation by the proteasome.

ER stress-mediated apoptosis

When all the prosurvival efforts fail to correct the protein-folding defect, apoptosis is activated. It is not clear at what point and by which mechanism the cell commits to death in response to excessive ER stress. Both mitochondrial-dependent and -independent cell death pathways likely mediate apoptosis in response to ER stress. The ER might actually serve as a site where apoptotic signals are generated through several mechanisms including: Bak/Bax-regulated Ca2+ release from the ER; cleavage and activation of procaspase-12; and IRE1-mediated activation of ASK1 (apoptosis signal-regulating kinase 1)/JNK (c-Jun amino terminal kinase). (Figure 2)

Figure 2
figure 2

ER stress-mediated apoptotic pathways. Upon ER stress, Bak, and Bax in the ER membrane undergo conformational alteration and permit Ca2+ efflux, which activates m-Calpain in the cytoplasm and subsequently cleaves and activates ER-resident procaspase-12 and leads to activation of the caspase cascade. The Ca2+ efflux also leads to the activation of the mitochondria-dependent apoptosis. CHOP, one of the UPR downstream effectors, inhibits the expression of Bcl-2 and thus promotes apoptosis. Activated IRE1 binds to c-Jun-N-terminal inhibitory kinase (JIK) and recruits TRAF2, which leads to the activation of ASK1/JNK and also the release of the procaspase-12 from the ER.

The available data suggest that upon encountering ER stress, proapoptotic Bcl2-related proteins, Bak and Bax, undergo conformational alteration in the ER membrane to permit Ca2+ efflux into the cytoplasm.38, 39 In vitro experiments support the idea that the increase in Ca2+ concentration (from micromolar to millimolar) in the cytoplasm activates the calcium-dependent protease m-Calpain, which subsequently cleaves and activates the ER-resident procaspase-12.40 Activated caspase-12 cleaves and activates procaspase-9 and consequently leads to activation of the caspase cascade.41 The Ca2+ released from the ER enters mitochondria leading to depolarization of the inner membrane, cytochrome c release, and activation of Apaf-1 (apoptosis protease-activating factor 1)/procaspase-9-regulated apoptosis.42, 43 CHOP (CEBP homologous protein) is a downstream transcriptional target of ATF6 and PERK/eIF2α/ATF4. CHOP is a basic leucine zipper-containing transcription factor that inhibits the expression of Bcl-2 and thereby is proposed to promote apoptosis.44, 45

IRE1α is proposed to play a role in ER stress-mediated apoptosis by interaction with TRAF2 (TNF receptor-associated factor-2) and ASK1 leading to the activation of ASK1 and JNK, and subsequent cell death.46 It has also been suggested that IRE1/TRAF2 association releases procaspase-12 from TRAF2, which is required for the activation of procaspase-12.47

Analysis of gene-deleted mice has provided insight into ER stress-induced apoptosis. Cells from Apaf-1-deficient mice are susceptible to ER stress-induced apoptosis, indicating that nonmitochondrial death pathways exist.41 Bak/Bax double knockout, caspase-12−/− and chop−/− murine embryonic fibroblasts (MEFs) all show partial resistance to ER stress-induced apoptosis, further supporting the idea that they facilitate the apoptotic response upon ER stress.44, 48, 49 Although, caspase-12-deficient and CHOP-deficient mice show no developmental defects, they display protection to genetically imposed or environmentally imposed ER stress.49, 50

The Role of the UPR in Professional Secretory Cell Development and Function

Compared to apoptotic pathways, the adaptive aspects of UPR signaling are better understood. However, in vivo analyses of these pathways under physiological and pathological contexts constantly surprise researchers and have inspired the current popular hypothesis that UPR pathways are regulated in a more finely tuned manner with selectivity and additional specificity to meet the complexity of various developmental and metabolic demands. The most well-understood physiological requirements for the UPR were elucidated in studies performed in professional secretory cells, which are under greater demand to fold and process large quantities of proteins.

Lessons from plasma cell differentiation

Analysis of the differentiation of B lymphocytes into plasma cells has suggested that the UPR drives ER biogenesis in response to high-level secretory protein synthesis.51 Extensive studies have elucidated the relationship between the UPR and plasma cell differentiation. The results from these studies established the fundamental principle that the UPR provides an essential physiological role in professional secretory cell differentiation.

It was reported over 20 years ago that the expression level of chaperones in the ER lumen is increased during lipopolysaccharide (LPS)-induced B-cell differentiation.52, 53 BiP was originally characterized as a protein that interacts with immunoglobulin (Ig) heavy chains to prevent their secretion in pre-B lymphocytes prior to light chain gene rearrangement.54 Subsequent studies demonstrated that BiP acts as a chaperone to assist Ig heavy chains interact with Ig light chains. In parallel studies, it was observed that GRP78 is one of a number of gene products that reside in the ER and that are induced upon glucose deprivation. The products of these genes were thus termed glucose-regulated proteins. When it was discovered that BiP is the same protein as GRP78, it became evident that many of the GRPs provide protein chaperone function in the ER. At that point, investigators began to test whether the UPR may be involved in the process of B lymphocyte differentiation into high-level Ig-secreting plasma cells. B-cell lymphopoiesis consists of an antigen-independent phase in the bone marrow (pro-B cell, pre-B cell) and an antigen-driven phase of differentiation that is completed in the periphery (mature B-cell, plasma cell). In pre-B cells, the ER exists as a minimal structure, most of which is nuclear envelope. Plasma cell differentiation in the periphery is accompanied by a five-fold expansion of the ER compartment, into stereotyping ribosome-studded stacked membrane sheets, presumably to accommodate the high level of Ig secretion. The first factor identified to be required for B lymphocyte differentiation into plasma cells was the transcription factor XBP1. When it was discovered that the mammalian substrate for the endoribonuclease activity of IRE1 is XBP1 mRNA, it was proposed that the IRE1 subpathway of the UPR provides an important physiological role to drive plasma cell differentiation.

LPS- or interleukin-4-induced differentiation of B lymphoma cells into plasma cells provides a unique experimental system to study the general involvement of the UPR. Results from in vitro lymphoma or splenic B-cell differentiation assays showed that the classical UPR target genes, such as BiP and GRP94 are dramatically induced.55, 56 Stimulation of primary splenic B cells from different mouse genetic backgrounds can identify the unique requirements for individual gene products in this differentiation. In addition, the ability to reconstitute the B-cell lineage in immunodeficient mice (such as recombination activating gene RAG1- or RAG2-deficient mice) provides a method to directly test the role of different genes in B-cell differentiation. RAG-1 and RAG-2 are essential for the rearrangement of Ig heavy and light chain gene loci,57 as well as the T-cell receptor gene loci. As a consequence, RAG-2-deficient mice lack mature T cells and B cells. Analysis of chimeric mice created by injection of homozygous deficient embryonic stem cells into RAG-2-deficient blastocysts or of RAG-2-deficient mice transplanted with hematopoietic stem cells deleted in specific components of the UPR pathway has provided insights into the requirement of the IRE1α/XBP1 pathway in this system.

XBP1 transcription is induced during B-cell differentiation and xbp1−/− B cells fail to differentiate into antibody-secreting plasma cells in vivo.58 XBP1 splicing correlates with plasma cell differentiation, and ectopic expression of the spliced form of XBP1 restores Ig production in XBP1-deficient B cells in vitro.56 The important role that XBP1 plays in plasma cell differentiation suggested that IRE1α is also essential for this process. This idea was well supported by the observation that ire1α−/− B cells did not differentiate into antibody-secreting plasma cells.59 Unexpectedly, the same set of experiments discovered that IRE1α is also required at the first stage of B-cell lymphopoiesis to induce Ig gene rearrangements. More intriguing is that the cytoplasmic domain, but not the protein kinase or endoribonuclease catalytic activities, was required for this novel function.59 The IRE1α protein was required to activate expression of Rag1, Rag2, and terminal deoxynucleotidyl transferase (TDT).

These findings have led to several important questions. First, what signals activate IRE1α to promote Ig gene rearrangement? Second, is BiP dissociation from IRE1α required? Certainly, the analysis of plasma cell differentiation in B cells that express different mutants of IRE1α will provide important insight into this process.

Finally, what role does IRE1α/XBP1 play in the late stage of B-cell differentiation? To address these issues, it is important to elucidate the temporal hierarchy of the increase in antibody production to create ER load and the expansion of the protein-folding capacity. Does ER stress actually drive the ER differentiation process? Although this question has created a considerable debate, most reports support the hypothesis that expansion of the ER protein-folding capacity occurs at least simultaneously, if not, prior to initiation of Ig protein secretion,55, 60 suggesting that the UPR activation may not solely depend on BiP dissociation from IRE1 in this case. However, it should be noted that Ig gene rearrangements associated with B-cell differentiation produce a number of nonproductive alleles that would produce misfolded Ig chains. It is possible that expression of folding-defective Ig intermediates from incorrectly rearranged alleles potently activates the UPR to contribute to ER expansion at an early time in differentiation, after the first Ig heavy chain gene V-D rearrangement occurs.

The unsolved mysteries of B-cell differentiation also raise the question of why the PERK/eIF2α pathway is not as essential as the IRE1α/XBP1 pathway. In the absence of eIF2α phosphorylation, B cells can completely differentiate into functional plasma cells.59 Since BiP appears to associate with the lumenal domains of IRE1α and PERK in a similar manner to regulate the UPR upon severe ER stress,61 the in vivo disparateness of these two pathways supports the idea that a more sophisticated mechanism of regulation exists for the physiological UPR in specific cell types. The other UPR transducer, ATF6, is cleaved and activated during B-cell differentiation into plasma cells.55 In addition, inhibition of the ATF6 pathway by expression of a dominant-negative form of ATF6 reduced IgM production in differentiating B cells.62

The physiological function of ATF6 in the UPR and plasma cell differentiation is under question because knockdown of both isoforms of ATF (ATF6α and ATF6β) did not interfere with UPR gene induction,63 despite the finding that overexpression of ATF6 induced a significant set of UPR target genes. Cells that lack both XBP1 and ATF6α are defective in UPR activation,63 suggesting that these two pathways may overlap. To elucidate the function(s) of ATF6 in the UPR and plasma cell differentiation, it will be necessary to establish ATF6 knockout mouse models.

Another very appealing possibility is that the UPR is activated through interplay between the conventional UPR pathway and other signal transduction pathways. During plasma cell differentiation, XBP1 is downstream of BLIMP1 (B lymphocyte-induced maturation protein 1), a transcription factor required for plasma cell differentiation which is regulated by interleukin (IL)-4 at a transcriptional level.56, 64 It has also been proposed that XBP1 is upstream of IL-6 during plasma cell differentiation.56 Meanwhile, activated XBP1 increases the protein-folding machinery in the cell, at least in part by increasing the expression of ER chaperones and foldases. A recent report links XBP1 activation to ER membrane biogenesis.65

Pancreatic β cell function and survival

The pancreatic β cell has been under intense investigation ever since its pivotal role in the pathogenesis of diabetes mellitus was recognized.66 Extensive research is directed to understand what unique properties β cells have to confer glucose responsive insulin production and secretion. The hallmark of pancreas dysfunction in noninsulin-dependent diabetes mellitus is diminished glucose-responsive insulin secretion, which is regulated by signals derived from mitochondrial metabolism.67 In order to maintain adequate intracellular insulin stores, β cells must adapt their acute and chronic rates of insulin biosynthesis to compensate for insulin release. Chronic elevation in extracellular glucose concentration further stimulates the synthesis of insulin by increasing expression of the mRNA encoding preproinsulin, increasing proinsulin translation and processing, and induction of additional components of the secretory pathway to support processing, transport, and glucose-regulated secretion of insulin. Type II diabetes results from failure of β cells to adequately adapt to the increasing demand for insulin production as a consequence of peripheral insulin resistance.

ER overload leads to β cell dysfunction

As a professional secretory cell, the vulnerability of the β cell to ER stress was suggested in early studies that induced β-cell damage by overexpression of major histocompatibility complex (MHC) class II protein in islets of transgenic mice.68 Characterization of the Akita mouse further supported the hypothesis that ER stress could be one cause for β cell death. There are two insulin genes in the murine genome (Ins1 and Ins2). In the Akita mouse, a highly conserved cysteine in the Ins2 gene is replaced by a tyrosine, thereby disrupting formation of one essential disulfide bond in proinsulin-2. The mutant protein cannot be processed and secreted normally and is retained within the ER. Although the islets from newborn Akita mice do not show detectable defects, later in development, a progressive β cell loss occurs that correlates with early onset diabetes.69 Since the complete absence of Ins2 expression does not lead to a similar diabetic phenotype and the Akita phenotype is dominant, the Akita mutation represents a gain-of-function. It is believed that the mutant protein is toxic due to induction of the ER stress response.70 This notion is further supported by the findings that the ER in the Akita β cells is distended and the UPR target genes BiP and CHOP are induced. In addition, deletion of the CHOP gene delayed the onset of β-cell destruction and of hyperglycemia in heterozygous Akita mice.69, 71 The absence of CHOP was also reported to increase resistance to nitric oxide-induced apoptosis in pancreatic β cells.72 Since nitric oxide production is implicated in the pathogenesis of type I diabetes,73 development of specific inhibitors that either block CHOP expression or its transcription factor activity, for example preventing CHOP dimerization or DNA-binding activity, may yield potential therapeutics for the prevention of diabetes.

Defective regulation of the PERK/eIF2α pathway leads to β cell dysfunction

Recent studies also indicate that the UPR is required to maintain β cell function. Compared to other differentiated cells, the β cell requires periodic increases in the capacity to fold and secrete insulin in response to post-parandial increases in blood glucose. This regulation of insulin synthesis is immediate and reversible and may be mediated by the PERK/eIF2α UPR subpathway. PERK was first identified as a highly expressed eIF2α protein kinase in the pancreas. Deletion of PERK results in progressive destruction of pancreatic β cells in both humans and mice.74, 75 In humans, mutations in the PERK gene cause Wolcott–Rallison syndrome which manifests as an infantile-onset, insulin-requiring diabetes.74 perk−/− mice experience a progressive loss of β cells and develop diabetes within the first few weeks after birth.75 It is proposed that in the absence of PERK, mRNA translation cannot be attenuated so, the protein-folding load cannot be properly coupled with the protein-folding capacity of the ER. As a consequence, perk−/− mice have defects in β cells as well as in other highly specialized secretory cells, such as exocrine pancreatic acinar cells, hepatocytes, and osteoblasts.75 Islets isolated from perk−/− mice secrete more insulin when they are switched from low glucose to high glucose conditions.75 This is presumably due to the inability to downregulate protein synthesis in response to increased insulin production. The eventual outcome results in accumulation of unfolded proinsulin in the ER with loss of β cell secretion potential.

Mice that are resistant to regulation by all eIF2α kinases were generated by introducing a serine51 alanine point mutation, the residue phosphorylated by all eIF2α kinases, into the eIF2α gene.31 Homozygous eIF2α S51A knockin mice develop a more severe β-cell dysfunction prior to birth, compared to perk−/− mice.31 The more severe β cell defect in the eIF2α mutant mice may indicate that other eIF2α kinases, such as the general control of amino-acid biosynthesis kinase GCN2 or the ds RNA-activated protein kinase PKR, may partially compensate for eIF2α phosphorylation in the absence of PERK.76, 77, 78

Heterozygous eIF2α S51A mice, where eIF2α phosphorylation regulation is reduced to approximately 50%, have no significant phenotype. However, when these mice are challenged with a high-fat diet, they develop insulin resistance, obesity, and diabetes with pancreatic β cell failure associated with a loss of glucose-stimulated insulin secretion. The β cells in the heterozygous high-fat fed mice exhibit abnormal distension of the ER lumen, defective folding and trafficking of proinsulin, and a reduced number of insulin granules.79 The reduced number of insulin granules can account for the loss of glucose-stimulated insulin secretion. It is proposed that the partial defect in the PERK/eIF2α pathway compromises the ability to couple proinsulin synthesis with proinsulin folding in the ER, leading to defective insulin secretion.

Mice with a defect in downregulating PERK-mediated eIF2α phosphorylation by deletion of p58IPK also show increased pancreatic β cell apoptosis concurrent with a gradual onset of glucosuria and hyperglycemia.80 On the other hand, developmental or metabolic defects were not observed in mice with homozygous GADD34 mutation to prevent regulated p-eIF2α dephosphorylation.81 The divergence between the gadd34Δc/Δc- and p58IPK-deleted mice may indicate that these two negative regulators play different physiological roles in vivo.

The UPR in hepatocytes

The liver is one of the major secretory organs in the body. Its functions include regulation of glucose homeostasis, lipid metabolism, and drug detoxification. Although the role of the UPR in the liver has not been significantly studied, preliminary results suggest that the UPR is essential for hepatocyte function.

Both IRE1α- and XBP1-deficient mice display a hypoplastic fetal liver. Growth is severely diminished and prominent apoptosis occurs in XBP1-null hepatocytes.59, 82 UPR activation is observed in the livers of diet-induced or genetically obese and/or diabetic mice.83, 84 For example, PERK activation, eIF2α phosphorylation, and BiP expression were increased compared to lean controls. In addition, heterozygous XBP1-deleted mice were more susceptible to insulin resistance, associated with chronic ER stress and JNK1 hyperactivation.83 It has been reported that ER stress leads to JNK activation,85 and inhibition of JNK1 signaling was shown to increase insulin sensitivity.86 In hepatoma cells, it is proposed that the UPR initiates activation of JNK1 which subsequently phosphorylates serine residue 307 of the insulin receptor substrate-1 (IRS-1), consequently reducing insulin receptor signaling.83 Recent work has shown that protein-tyrosine phosphatase 1B (PTP-1B) interacts with the IRE1/XBP1 pathway.87 The absence of PTP-1B attenuates ER stress-induced apoptosis, JNK activation and XBP1 splicing. The potential involvement of PTP-1B in the physiological UPR is particularly interesting because of its predominant localization to the ER,88 and its role in insulin signaling and metabolism.89 Although data support a role for XBP1 and JNK1 in insulin signaling, it is presently not known whether signaling from these transducers emanates from ER stress and activation of IRE1.

Since an ATF6-deficient rodent model is not available to date, the particular role of this b-ZIP transcription factor in the liver function remains unknown. However, recent work has shown that ATF6 antagonizes the lipogenic functions of SREBP2 in the liver.90 Glucose deprivation to induce ATF6 cleavage and activation, as well as ectopic expression of the cleaved form of ATF6, attenuates the activity of the transcription factor SREBP2. This inhibitory effect can be reversed by overexpression of BiP, suggesting that the signal emanates from the ER.

The Caenorhabditis elegans homologue of Creb-H, a liver-specific basic leucine zipper (b-ZIP) transcription factor that shares significant homology with ATF6, is a UPR responsive gene and is regulated by ire-1 and xbp-1 in the nematode.91 Like ATF6, mammalian CREBH is cleaved upon ER stress by S1P/S2P, and the liberated N-terminal cytosolic fragment transits to the nucleus. Instead of inducing expression of the UPR target genes, CREBH acts with ATF6 to synergistically activate the transcription of the acute phase response genes, including serum amyloid P-component (SAP) and C-reative protein (CRP) in response to ER stress. CREBH knockdown in transgenic mice demonstrated that CREBH may not be required for hepatocyte differentiation, although there was a defect in the acute inflammatory response.92 These observations identify a novel link between ER stress and some previously thought-nonrelated physiological processes.

The mechanism by which misfolded proteins affect liver function is also illustrated in some cases of α1-antititrypsin deficiency. The Z allele of the α1-antitrypsin (α1-AT) (Glu342Lys mutation) produces a protein that polymerizes and is retained in the ER. Although this protein interacts with calnexin, there is question about whether it significantly interacts with BiP and activates the UPR.93 However, analysis of α1-AT expression in transfected fibroblasts isolated from affected patients suggests that defects in ERAD are associated with a greater severity of liver pathology.94

The physiological role of the UPR in osteoblasts

The osteoblast is the only cell type responsible for extracellular matrix deposition during bone formation. One of the major ER stress markers, ATF4, has been reported to regulate the onset of osteoblast differentiation, type I collagen synthesis, osteoblast-specific gene expression, and osteoblast terminal differentiation.95 Osteoblasts express high levels of ATF4 protein compared to most other tissues.96 This observation is surprising given that ATF4 mRNA is present in a rather broad range of organs.96 It is tempting to speculate that this discrepancy is related to stress-induced translational regulation. Mice and humans deficient in PERK have the same abnormal thinning of bone trabeculae that was observed in ATF4-deficient mice.74, 97 It will be informative to investigate whether tissue-specific knockin of the eIF2α S51A mutation in osteoblasts produces a similar defect as the ATF4-null and PERK-null mice. In contrast, osteoporosis and deficient bone mineralization occur during osteogenesis imperfecta (OI), which is a disease where misfolded mutant procollagen binds to BiP and activates the UPR.98 Therefore, a proper balance of UPR activation may be required for optimal osteoblast function.

The initial stage of tooth enamel development is a secretory event. The columnar ameloblast cells of the enamel organ are responsible for dental enamel development. During the secretory stage, the ameloblasts are tall, contain an extensive ER and secret large amounts of protein into the enamel matrix. It was reported that the UPR plays a role in the ameloblast, especially for its susceptibility to the toxic effects of fluoride exposure.99

The UPR also plays an essential role in neurons, especially in a large class of conformational diseases associated with accumulation of abnormal protein aggregates, which is discussed in detail by Lindholm in this perspective series.

Conclusions and Perspectives

Ire1p/Hac1p is the only UPR signaling pathway in yeast. 100, 101 Genomic analysis suggests that although atf-6 and pek emerged within the C. elegans UPR, the most significant UPR regulation occurs through the ire-1/xbp-1 pathway in the worm.91 In higher eukaryotes, the UPR expanded to a signal transduction nexus: IRE1 and ATF6 both contain α and β isoforms. Several b-ZIP transcription factors that share homology with ATF6, such as Luman,102 are reported to be ER stress responsive. To cope with different types of ER stress in different tissues, vertebrates have also evolved tissue-specific UPR regulators. Other than the liver-specific CREBH, OASIS and Tisp40 are also likely regulated in a manner similar to ATF6, but are selectively expressed in astrocytes and spermatids, respectively.103, 104, 105

Genetically manipulated rodent models harboring mutations in selective UPR components contributed significantly to our current knowledge of the physiological role of the UPR. Attempts to further define the physiological role of each UPR subpathway have been hampered by embryonic and neonatal lethality. One way to circumvent this problem is to create mice harboring conditional transgenes to rescue the null mutation or to create mice with conditionally targeted gene mutations. Then, the use of various Cre transgenic mice provides the ability to achieve temporal and/or tissue-specific gene-deleted mice. Also, the in vivo function of the third UPR proximal transducer, ATF6 will remain elusive until a deficient rodent model is obtained.