IRE1α activation protects mice against acetaminophen-induced hepatotoxicity

To investigate the function of the UPR regulated by IRE1α and XBP1 in APAP metabolism and hepatotoxicity, we used mice in which Xbp1 is inducibly deleted in the liver by polyinosinic:polycytidylic acid (poly[I:C]) treatment (Lee et al., 2008). Xbp1^f/f;Mxcre (Xbp1Δ) and the littermate control Xbp1^f/f (WT) mice were injected with poly(I:C) and used for experiments after more than 3 wk of acclimation. A single dose of 500 mg APAP/kg markedly increased serum alanine aminotransferase (ALT) levels in WT mice, peaking at 24 h, and then decreasing to near basal levels on day 3, indicating APAP-induced acute liver injury (Fig. 1 A). In contrast, Xbp1Δ mice displayed only a modest induction of serum ALT levels by APAP treatment. APAP-induced mortality was also dramatically decreased in Xbp1Δ mice (Fig. 1 B). Histological analysis of the liver revealed markedly reduced liver injury in Xbp1Δ mice compared with WT animals that exhibited severe centrilobular necrosis and hemorrhage (Fig. 1 C), demonstrating that Xbp1 deficiency protected mice from APAP-induced hepatotoxicity.

Although the generation of toxic metabolites by hepatocytes is the primary mechanism of liver injury from APAP overdose, inflammatory responses mediated by the innate immune system also contribute to APAP hepatotoxicity (Blazka et al., 1995; Liu et al., 2004; Liu and Kaplowitz, 2006). Xbp1^f/f;Mxcre mice expressed cre recombinase both in the liver and the hematopoietic cells in the bone marrow of upon poly (I:C) injection (Martinon et al., 2010). We recently demonstrated that XBP1 is required for optimal cytokine expression in macrophages and controls the inflammatory response (Martinon et al., 2010). To ascertain that XBP1 deficiency in hepatocytes and not in innate immune cells protected mice from APAP-induced hepatotoxicity, we crossed Xbp1^f/f mice with Alb-cre mice expressing Cre recombinase under the control of the mouse albumin promoter to achieve hepatocyte-specific deletion of XBP1 (Postic et al., 1999). Similar to Xbp1Δ mice that lacked XBP1 both in the liver and hematopoietic cells, Xbp1^f/f;Alb-cre (Xbp1^LKO) mice were viable and grossly normal. Xbp1^LKO mice were also resistant to APAP-induced hepatotoxicity, as indicated by lower serum ALT levels and decreased necrosis compared with littermate controls (Fig. 2). These results indicate that XBP1 deficiency in liver parenchymal cells is sufficient to protect mice against APAP-induced hepatotoxicity.

To investigate the mechanism by which XBP1 regulates APAP metabolism, we measured mRNA levels of genes involved in drug metabolism using the RT² Profiler PCR Array (SABiosciences). Notably, Cyp1a2 and Cyp2e1, which are crucial for APAP oxidation to the toxic NAPQI, were significantly decreased in the liver of Xbp1Δ mice (Table S1). Cyp2e1 plays a major role in the conversion of APAP into hepatotoxic NAPQI (Gonzalez, 2007). Cyp1a2 also contributes to APAP metabolism and hepatotoxicity, although it has a lower APAP affinity than Cyp2e1 (Raucy et al., 1989; Patten et al., 1993). Quantitative RT-PCR (qRT-PCR) confirmed the decrease of hepatic Cyp1a2 and Cyp2e1 mRNA levels by poly(I:C) administration in Xbp1^f/f;Mxcre mice that ablated XBP1, but not in the control Xbp1^f/f mice (Fig. 3 A). Western blotting and qRT-PCR demonstrated decreased expression of Cyp1a2 and Cyp2e1 in the liver of Xbp1^LKO mice as well (Fig. 3, B and C). In contrast, genes involved in GSH synthesis, such as GSH synthetase, glutamate-cysteine ligase catalytic, and glutamate-cysteine ligase modifier subunits, were unaffected by XBP1 deficiency (Fig. 3 D). Nuclear factor erythroid-2–related factor 2 (Nrf2) which is considered a key regulator of the antioxidant response (Chan and Kan, 1999; Chan et al., 2001), and its target genes such as GSH-S-transferase pi 1 (Gstpi) and UDP-glucuronosyltransferase 1a6 (Ugt1a6), were also unchanged in Xbp1Δ liver (Fig. 3, D and E). Collectively, these data indicate that XBP1 is required for the expression of Cyp1a2 and Cyp2e1, which are responsible for the generation of reactive NAPQI from APAP.

The reactive metabolite NAPQI generated from APAP by Cyp1a2 and Cyp2e1 P450 enzymes is normally removed by conjugation with GSH (James et al., 2003). Toxic doses of APAP deplete hepatic GSH, allowing covalent binding of NAPQI to intracellular macromolecules, which leads to cell death (Cohen and Khairallah, 1997). Hepatic GSH stores were rapidly decreased by APAP administration in both WT and Xbp1^LKO mice, reaching background levels within 2 h, indicating that the reactive NAPQI overwhelmed the GSH-mediated detoxification system of the liver (Fig. 4 A). Xbp1^LKO mice exhibited slightly faster recovery of GSH levels than WT mice, suggesting that the burden of toxic APAP metabolites was lower in the former (Fig. 4 A). In contrast, APAP-conjugated protein adducts were not substantially reduced in the XBP1-deficient liver, suggesting that the residual level of these P450 enzymes is sufficient to generate APAP adducts (Fig. 4, B and C).

JNK is activated by APAP-induced oxidative stress and plays a crucial role in the hepatotoxicity (Gunawan et al., 2006; Henderson et al., 2007; Kaplowitz et al., 2008; Nakagawa et al., 2008). To determine the effect of XBP1 ablation on JNK activation, we measured the level of JNK phosphorylation in liver tissues by Western blotting after APAP treatment. In WT mice, JNK phosphorylation was strongly induced for 6 h after APAP treatment, and then declined at the 24-h time point to low, but detectable levels, reflecting the prolonged oxidative stress (Fig. 4, B and C). Xbp1^LKO mice also exhibited marked induction of JNK phosphorylation at early time points (Fig. 4 B), consistent with GSH depletion and APAP adduct formation. Notably, JNK phosphorylation was decreased to baseline levels at the 24-h time point in Xbp1^LKO mice, in contrast to the markedly higher levels in WT mice (Fig. 4 C), suggesting that the lack of prolonged JNK activation might contribute to the protection of XBP1-deficient mice from APAP-induced hepatotoxicity. Given that Cyp2e1 generates reactive oxygen species by metabolizing a variety of substrates (Cederbaum et al., 2009), it is conceivable that the reduction of Cyp2e1 in XBP1-deficient liver might ameliorate oxidative stress and JNK activation.

Because XBP1 is a transcriptional activator, we first sought to determine if XBP1 directly induced the transcription of Cyp1a2 and Cyp2e1. Contrary to our prediction, adenoviral overexpression of XBP1s failed to induce these P450 enzyme mRNAs in primary hepatocytes (Fig. 5 A). Furthermore, tunicamycin injection decreased hepatic Cyp1a2 and Cyp2e1 mRNA levels, despite a strong induction of XBP1s and its target genes, BiP and ERdj4, arguing against a direct role of XBP1 in regulating these genes (Fig. 5, B and C).

Previously, we reported that the ablation of XBP1 led to feedback activation of its upstream enzyme IRE1α in the liver (Lee et al., 2008). IRE1α hyperactivation in XBP1-deficient liver appeared to be an ER stress–independent event, given that ER morphology was normal and activation of other UPR branches was unaltered. Interestingly, recent studies have demonstrated that IRE1α can induce the degradation of certain mRNAs, a process known as regulated IRE1-dependent decay (RIDD; Han et al., 2009; Hollien et al., 2009; Oikawa et al., 2010). In contrast to the efficient relegation of XBP1 mRNA after cleavage by IRE1α, in RIDD, mRNA substrates appear to be degraded after the cleavage, presumably by cellular ribonucleases (Lipson et al., 2008; Han et al., 2009; Hollien et al., 2009; Oikawa et al., 2010; Lee et al., 2011). These observations prompted us to test if the constitutively active IRE1α induced the degradation of Cyp1a2 and Cyp2e1 mRNAs in XBP1-deficient liver.

Western blot analysis of liver lysates revealed that IRE1α protein was induced and constitutively phosphorylated in the liver of Xbp1^LKO mice, consistent with what has been observed in Xbp1Δ mice (Fig. 3 C). Phos-Tag SDS-PAGE analysis demonstrated that IRE1α phosphorylation was dramatically induced in XBP1-deficient liver, indicating the constitutive activation of IRE1α in the mutant mice (Fig. 3 C). To investigate if the hyperactivated IRE1α down-regulated Cyp1a2 and Cyp2e1 in XBP1-deficient liver, we suppressed IRE1α expression in the liver of Xbp1^LKO mice using siRNA. Intravenous injection of lipidoid-formulated siRNA targeting IRE1α efficiently reduced IRE1α mRNA and protein levels in XBP1-deficient liver (Fig. 6). Notably, IRE1α siRNA markedly induced hepatic Cyp1a2 and Cyp2e1 at both mRNA and protein levels in the liver of Xbp1^LKO mice (Fig. 6). These data indicate that constitutively active IRE1α decreased Cyp1a2 and Cyp2e1 mRNAs in XBP1-deficient liver, presumably through RIDD.

Decreased expression of Cyp1a2 and Cyp2e1 mRNAs in tunicamycin-injected mice suggests that ER stress might also activate RIDD, leading to the degradation of these mRNAs (Fig. 5 C). To further investigate the link between the ER stress response and APAP metabolism, we characterized the phenotype of IRE1α-deficient mice, which were generated by crossing Ern1 flox mice (Iwawaki et al., 2010) with Mx1-cre transgenic mice. IRE1α deletion in the liver was achieved by poly(I:C) injection, as done previously to generate Xbp1Δ mice (Lee et al., 2008). Cre-mediated deletion of exons 20 and 21 of Ern1 caused an in-frame deletion of 64 aa (aa 844–907) in the ribonuclease domain of IRE1α, which completely abolished its ribonuclease activity, thus abrogating XBP1 mRNA splicing in the mutant mice (Fig. 7, A and B; Iwawaki et al., 2010). Consistent with what we observed in Xbp1Δ mice (Lee et al., 2008), ablation of IRE1α abolished the induction of XBP1-dependent UPR genes such as ERdj4 and Sec61α in the liver of tunicamycin-injected mice (Fig. 7 C). Expression of CHOP, which is known to be regulated primarily by the PERK pathway, was not significantly altered in IRE1α-deficient liver at the basal state or after tunicamycin treatment (Fig. 7 C). Notably, however, hepatic Cyp1a2 and Cyp2e1 mRNA levels were comparable between WT and Ern1Δ mice, in contrast to the significant reduction of these mRNAs in XBP1-deficient mouse liver. Further, Cyp1a2 and Cyp2e1 mRNAs were markedly decreased by tunicamycin treatment in WT but not IRE1α-deficient mice, indicating that these mRNAs were down-regulated by IRE1α in response to ER stress (Fig. 7 C). Consistent with the normal expression of Cyp1a2 and Cyp2e1 mRNAs in Ern1Δ mice, APAP-induced hepatotoxicity was comparable between WT and IRE1α-deficient mice, as assessed by serum ALT levels and histological analysis (Fig. 7, D and E). However, Ern1Δ mice were not more sensitive to APAP-induced hepatotoxicity than WT mice when challenged with a reduced dose of APAP, suggesting that basal IRE1α activity is not sufficient to confer a protective effect. In conclusion, removal of IRE1α stabilized Cyp1a2 and Cyp2e1 mRNAs and also removed protection from APAP-induced hepatotoxicity conferred by XBP1 deficiency.

To ascertain that Cyp1a2 and Cyp2e1 mRNAs are indeed degraded after cleavage by IRE1α, we first tested if IRE1α activation by transient overexpression suppressed the production of these mRNAs. We co-transfected HEK293T cells with Cyp1a2 and Cyp2e1 plasmids, along with WT or an inactive mutant IRE1α. Expression of Cyp1a2 and Cyp2e1 mRNAs was markedly suppressed by the co-transfected WT, but not mutant IRE1α (Fig. 8 A). Similarly, co-transfected IRE1α reduced expression of Bloc1s1 mRNA, which has been identified as an IRE1α substrate for degradation in previous studies (Han et al., 2009; Hollien et al., 2009).

To directly demonstrate the cleavage of mRNAs by IRE1α, we next generated Cyp1a2 and Cyp2e1 mRNAs by in vitro transcription and incubated them with recombinant IRE1α protein. In a control experiment, IRE1α efficiently cleaved XBP1 mRNA, generating two fragments with the expected sizes, indicating that the cleavage occurred at the correct sites (Fig. 8 B). Notably, Cyp1a2 and Cyp2e1 mRNAs were also cleaved by IRE1α to smaller fragments (Fig. 8 B). As a negative control, β-actin mRNA was not cleaved by IRE1α. To determine if IRE1α could cleave endogenously produced Cyp1a2 and Cyp2e1 mRNAs, we incubated recombinant IRE1α protein with total RNA isolated form mouse liver (Fig. 8 C). Northern blot analysis demonstrated that IRE1α cleaved Cyp1a2 and Cyp2e1 mRNAs at specific sites, yielding smaller fragments that were largely comparable to the cleavage products of in vitro–transcribed mRNAs (Fig. 8, B and C). XBP1 mRNA was also efficiently cleaved by IRE1α, generating a smaller fragment that was detected by a probe hybridizing to the 3′ end of the mRNA. In contrast, β-actin and Cyp2c29 mRNAs, which were normally expressed in Xbp1-deficient liver were unaffected by recombinant IRE1α protein (Fig. 8 C), indicating the specific cleavage of Cyp1a2 and Cyp2e1 mRNAs by IRE1α.

To determine IRE1α cleavage sites in Cyp1a2 and Cyp2e1 mRNAs, we performed 5′ rapid amplification of cDNA ends (RACE) PCR on in vitro–transcribed mRNAs after IRE1α cleavage. This procedure identified one cleavage site in Cyp1a2 and four sites in Cyp2e1 mRNAs (Fig. 8 D). DNA sequencing of PCR products revealed that IRE1α cleaves at G–C junctions within 5′-CUGCAG-3′ consensus sequences, which are conserved in XBP1 and in other RIDD substrate mRNAs (Oikawa et al., 2010). Although secondary structure prediction of individual IRE1α cleavage sites yielded dissimilar hairpin structures, sequence alignment of the cleavage sites generated a consensus secondary structure that was congruent with the XBP1 hairpin, suggesting that IRE1α recognizes a common secondary structure in target mRNAs (Fig. 8 D). Mutation of two G residues at −1 and +3 positions in the consensus sequence abolished the cleavage of Cyp1a2 and Cyp2e1 mRNAs by IRE1α (Fig. 8, E–G), indicating that these residues are essential for IRE1α cleavage. Collectively, these data demonstrate that hyperactivated IRE1α specifically cleaves Cyp1a2 and Cyp2e1 mRNAs in XBP1-deficient liver, leading to further degradation by cellular ribonucleases.

Recent studies demonstrated the association of ER stress response with a variety of liver diseases (Ji, 2008). APAP is known to activate ER stress response (Lorz et al., 2004; Nagy et al., 2007), an observation that invites the speculation that feedback activation of IRE1α by APAP may suppress the production of NAPQI driven by Cyp1a2 and Cyp2e1. However, we did not observe the activation of IRE1α or other UPR branches such as PERK and ATF6α in the liver of APAP-treated mice, arguing against a role of APAP in IRE1α activation (Fig. 9). Physiological and pathological conditions that activate IRE1α, leading to the degradation of Cyp1a2 and Cyp2e1 mRNAs, remained to be explored.

RIDD was first described in Drosophila melanogaster cells, where IRE1 was observed to induce the degradation of mRNAs encoding proteins that transit through the ER (Hollien and Weissman, 2006). Subsequent studies in mammalian cells reported that IRE1α suppressed the expression not only of secretory pathway cargo proteins, but also ER-resident proteins that handle the folding and trafficking of cargo proteins (Han et al., 2009). These observations raised the possibility that RIDD might act as a novel mechanism to alleviate ER stress by limiting the entry of cargo proteins into the ER. It was also possible that RIDD could exert opposing effects on cell fate, such that it might maintain homeostasis by acutely relieving ER stress, but promote apoptosis under conditions of prolonged, high level ER stress (Han et al., 2009; Hollien et al., 2009). However, the in vivo functions of RIDD in physiological or pathological situations remained to be explored. Here, we demonstrate that IRE1α hyperactivation in hepatocytes in vivo led to RIDD that resulted in protection from the stress of a hepatotoxic APAP metabolite. Our study reveals that the IRE1α–XBP1 UPR signaling pathway plays a key role in hepatic drug metabolism, at least in part by regulating the stability of certain p450 mRNAs by RIDD.

Important RIDD substrates in the liver include, but may not be limited to, mRNAs encoding the key p450 drug-metabolizing enzymes Cyp1a2 and Cyp2e1. Reduction of Cyp1a2 and Cyp2e1 mRNAs correlated with the protection of XBP1-deficient mice from APAP-induced hepatotoxicity, although it remains to be determined if the reduction of these mRNAs is sufficient for protection. Acute GSH depletion and APAP-protein adduct formation were not significantly altered in XBP1-deficient liver upon APAP treatment, raising the possibility that Cyp1a2- and Cyp2e1-independent mechanisms might contribute to the protection of XBP1-deficient mice from APAP-induced hepatotoxicity. There may well be additional RIDD substrates involved in APAP metabolism required to confer protection from APAP-induced hepatotoxicity.

JNK is activated by oxidative stress, and plays a central role in APAP-induced hepatotoxicity (Gunawan et al., 2006; Henderson et al., 2007; Kaplowitz et al., 2008; Nakagawa et al., 2008). IRE1α activates JNK under ER stress conditions (Urano et al., 2000). However, JNK was not induced in XBP1-deficient liver, despite the presence of hyperactivated IRE1α, indicating that IRE1α is not sufficient to activate JNK. Indeed, XBP1-deficient liver instead exhibited reduced JNK activation upon APAP treatment, which might contribute to the observed protection from APAP-induced hepatotoxicity.

In normal mouse liver, IRE1α is present mostly in an inactive form, as indicated by minimal phosphorylation of IRE1α and low levels of XBP1 mRNA splicing. Ablation of IRE1α did not significantly decrease Cyp1a2 and Cyp2e1 mRNA levels, suggesting that the basal IRE1α activity present in normal mouse liver is not sufficient to cause a significant reduction of its substrate mRNAs. Pathophysiological conditions that activate IRE1α to instigate RIDD in genetically intact animals remain to be identified. Interestingly, it has previously been shown that high-fat and high-cholesterol diets promote the degradation of MTP mRNA in the intestine by gut-restricted IRE1β protein, suppressing chylomicron production (Iqbal et al., 2008). IRE1α did not cleave MTP mRNA, indicating the divergent substrate specificities of IRE1α and IRE1β (Iqbal et al., 2008). Various insults to hepatocytes, such as fat accumulation, viral infection, inflammation, hyperhomocysteinemia, and oxidative stress, have been implicated in the ER stress response in the liver (Ji, 2008), and it would be interesting to see if RIDD is instigated under these conditions.

Hyperactivation of IRE1α either through elimination of its substrate XBP1 or through pharmacologically induced ER stress can instigate RIDD. The fact that XBP1 deficiency causes hyperactivation of IRE1α which can induce the degradation of certain mRNAs raises an interesting question. Ablation of XBP1 causes multiple abnormalities in many organ systems in mice, which include the deficiency of professional secretory cells (Reimold et al., 2001; Lee et al., 2005; Kaser et al., 2008; Huh et al., 2010), defective insulin secretion from β cells (Lee et al., 2011), decreased lipogenesis in the liver (Lee et al., 2008), and impaired cytokine production from macrophages (Martinon et al., 2010). What are the relative contributions of these two functions of IRE1α, one of them XBP1-independent (RIDD) and the other dependent on XBP1-driven gene regulation, to the phenotypes of XBP1-deficient mice? As we have done here, it will be critical to compare the phenotypes of XBP1 and IRE1α single and XBP1–IRE1α double-deficient mice in each organ to answer this question. Such studies are currently underway and should establish the contribution of indirect IRE1α hyperactivation (RIDD) induced by XBP1 deficiency versus a direct effect of XBP1 deficiency in vivo. We would predict, and our early studies support, that this balance will be different in different organ systems (Lee et al., 2008).

It is notable that there are both similarities and disparities between the phenotypes of XBP1 and IRE1α knock out mice. At the molecular level, Ern1 deletion abolishes the expression of both IRE1α and XBP1s. In contrast, Xbp1 deletion abolishes the expression of XBP1u and XBP1s, but activates IRE1α in certain organs and cell types, such as liver, intestinal epithelium, β cells, and macrophages; however, XBP1 does not activate IRE1α in neurons. Germ line deletion of Ern1 or Xbp1 caused embryonic lethality in mice because of impaired liver development (Reimold et al., 2000; Zhang et al., 2005). Interestingly, the embryonic lethal phenotypes of these mice were reversed by expression of IRE1α in the placenta or XBP1 in the liver (Iwawaki et al., 2009; Lee et al., 2005), indicating differential requirements for IRE1α and XBP1 in the development of these organs. XBP1- and IRE1α-deficient mice display abnormalities in secretory cells and organs, but the latter appear to have milder phenotypes (Reimold et al., 2001; Lee et al., 2005; Iwawaki et al., 2010). One exception to this, however, is in the B cell lineage, where IRE1α-deficient chimeras display a halt in B cell differentiation at an early stage of B cell development (Zhang et al., 2005) in contrast to XBP1 lymphoid chimeras or B cell–specific Xbp1^f/f;CD19^+/cre mice where B cell development proceeds normally until the terminal stage of B cell differentiation to plasma cells (Reimold et al., 2000; Todd et al., 2009). It remains to be determined if this difference in timing and severity of phenotype is caused by a contribution by RIDD or can be explained by different experimental settings, such as mouse genetic background.

Although basal IRE1α activity is low in normal mouse liver, it is sufficient to drive the production of measurable amounts of XBP1s protein. Remarkably, ablation of XBP1 induces a striking activation of its upstream endoribonuclease IRE1α in a feedback loop whose components are unknown. It is notable that ablation of XBP1 activated IRE1α, but not PERK or ATF6, in the liver, although a pharmacologic ER stress inducer, tunicamycin, simultaneously activated all three UPR branches (Lee et al., 2008), suggesting that the signal activating IRE1α might be distinct from those for other UPR branches. Identifying the molecular constituents of this regulatory circuit is an important goal of future studies.

Generation of Xbp1Δ mice (Xbp1^f/f;Mx1-cre mice injected with poly [I:C]) has been previously described (Lee et al., 2008). IRE1α-deficient Ern1Δ mice were generated by breeding Ern1^flox mice (Iwawaki et al., 2009) with Mx1-cre mice, and then i.p. injecting 250 µg of poly(I:C) twice into Ern1^f/f;Mx1-cre mice. B6.Cg-Tg(Alb-cre)21Mgn/J mice that contained an albumin promoter-driven Cre recombinase transgene were obtained from The Jackson Laboratory and crossed with Xbp1^flox mice to generate Xbp1^LKO mice. Animal studies and experiments were approved and performed according to the guidelines of the Animal Care and Use Committee of Harvard University.

APAP (Sigma-Aldrich) was dissolved in 1,2-propanediol-saline (50:50) solution at 50 mg/ml. After a 16-h fast, mice were injected i.p. with freshly prepared APAP solution at 500 mg/kg of body weight, unless otherwise indicated. Blood was drawn from tail vein to measure ALT activity using a commercial reagent (Bio-Quant). Plasma samples were serially diluted threefold in PBS to obtain the ALT level within the detection range of the reagent. Liver tissue samples were fixed in 10% neutral buffered formalin, embedded in paraffin, cut into 5-µm-thick sections, and stained with hematoxylin and eosin. Hepatic GSH content was measured using a Total GSH Quantification kit (Dojindo Molecular Technologies).

Lipidoid-siRNA formulations using C12-200 were prepared as described previously (Love et al., 2010). Lipidoid-formulated siRNAs were diluted to 0.1 mg/ml in PBS and administered to mice via tail vein at a dose of 0.5 mg/kg body weight.

Total RNA was isolated from liver tissue and transfected 293T cells using Qiazol lysis reagent (QIAGEN). SYBR green-based qRT-PCR was performed using the Mx3005P system (Stratagene), with the complementary DNA (cDNA) made using High Capacity cDNA Reverse Transcription kit (Applied Biosystems).

HEK293T cells were grown in DME supplemented with 10% fetal bovine serum. 150,000 cells were plated in a 12-well plate and transfected with total 1 µg DNA (450 ng pED-IRE1α, 10 ng pCMV-SPORT6-Angplt3, 10 ng pCMV-SPORT6-Bloc1s1, 10 ng pCMV-SPORT6-EGFP, and 520 ng pCMV-SPORT6) using Lipofectamine 2000 (Invitrogen). Cells were harvested 24 h after transfection to extract RNA, which were treated with DNase I (Invitrogen) and subjected to qRT-PCR analysis. Transgenic mRNA levels were normalized to the co-transfected EGFP level.

Liver lysates were prepared by homogenization in RIPA buffer (50 mM Tris–Cl pH 8.0, 150 mM NaCl, 1% NP-40, 0.5% deoxycholate, 0.1% SDS, 50 mM NaF) supplemented with protease inhibitor tablet (Roche). Liver nuclear extracts were prepared as described previously (Lee et al., 2008). Phos-tag acrylamide reagent (NARD Institute, Ltd.) and MnCl₂ were added to the gel to final concentrations of 10 and 20 µM, respectively, for Phos-Tag gel electrophoresis. Rabbit polyclonal antibody raised against mouse XBP1s, ATF6α, and Nrf2 (McMahon et al., 2003) were used to detect nuclear XBP1s, ATF6α(N), and Nrf2 proteins by Western blot. Anti-APAP (Matthews et al., 1996), anti-Cyp1a2 (Detroit R&D), anti-Cyp2e1 (Detroit R&D), anti-IRE1α (Cell Signaling Technology), anti–phospho-PERK (Cell Signaling Technology), anti-JNK (Cell Signaling Technology), and anti–phospho-JNK (Cell Signaling Technology) antibodies were used to detect specific proteins in whole-liver lysates by Western blot.

Recombinant adenoviruses expressing GFP or mouse XBP1s and infection of mouse primary hepatocytes were described previously (Lee et al., 2008). Primary hepatocytes were infected with recombinant adenoviruses at 10 PFU per cell and harvested 24 h later to isolate RNAs for gene expression analysis.

In vitro mRNA cleavage assays were performed as described previously (Lee et al., 2011). In brief, twofold serially diluted recombinant protein containing the cytoplasmic domain of human IRE1α (28, 14, 7, and 3.5 ng) was incubated with 1 µg of RNA generated by in vitro transcription using SP6 or T7 polymerase (Invitrogen) and specific MGC cDNA clones in pCMV-SPORT6 vector (Open Biosystems), or 3 µg of total liver RNA for 30 min at 37°C. IRE1α-mediated cleavage of in vitro–transcribed RNA was revealed by ethidium bromide staining of the agarose gel after electrophoresis. Liver RNA treated with IRE1α was resolved on denaturing agarose gels and subjected to Northern blot analysis. Radiolabeled probes were prepared using Rediprime II random prime labeling system (GE Healthcare). Templates for radiolabeled probes were generated by PCR amplification of the fragments of Cyp1a2 (+1 to +1093, BC064827), Cyp2c29 (+9 to +412, BC019908), and Cyp2e1 (+1 to +512, BC013451) cDNAs.

Cyp1a2 and Cyp2e1 mRNAs produced by in vitro transcription were incubated with recombinant IRE1α. IRE1α-cleaved RNAs were treated with T4 polynucleotide kinase, and then ligated with 5′-RACE adapter (5′-GCUGAUGGCG­AUGAAUGAACA­CUGCGUUUGCU­GGCUUUGAUGAAA-3′; Invitrogen) using T4 RNA ligase (New England BioLabs). After a purification step using RNeasy MinElute Cleanup kit (QIAGEN), ligated RNAs were subjected to cDNA synthesis followed by PCR amplification using RACEouter primer (5′-GCTGATGGCGATGAATGAACACTG-3′) and various reverse primers that anneal to Cyp1a2 and Cyp2e1 mRNAs with 300–400-bps intervals. PCR products were subjected to DNA sequencing.

Table S1 shows the expression of genes related to drug metabolism in XBP1-deficient liver.

We thank Dr. Dean Roberts for anti-APAP antibody, Dr. Michael McMahon for anti-Nrf2 antibody, Dorothy Hu for histological analysis, Kirsten Sigrist for animal maintenance, and Drs. Fabio Martinon and Eun-Sook Hwang for critical reading of this manuscript. We also thank Brian Bettencort and Greg Hinkle (design and bioinformatics); Akin Akinc and Alnylam formulations group; and Muthiah Manoharan, Martin Maier, Satya Kumchimanchi, Klaus Charisse, and the rest of the Alnylam chemistry team for siRNA synthesis.

This study was supported by National Institutes of Health grants AI32412 (L.H. Glimcher), DK082448 (L.H. Glimcher), and DK089211 (A.-H. Lee), and a grant from the American Heart Association (A.-H. Lee).

L.H. Glimcher is a Member of the Board of Directors of and holds equity in Bristol Myers Squibb. The other authors have no conflicting financial interests.