The recognition of autophagy related 16-like 1 (ATG16L1) as a genetic risk factor has exposed the critical role of autophagy in Crohn’s disease1. Homozygosity for the highly prevalent ATG16L1 risk allele, or murine hypomorphic (HM) activity, causes Paneth cell dysfunction2,3. As Atg16l1HM mice do not develop spontaneous intestinal inflammation, the mechanism(s) by which ATG16L1 contributes to disease remains obscure. Deletion of the unfolded protein response (UPR) transcription factor X-box binding protein-1 (Xbp1) in intestinal epithelial cells, the human orthologue of which harbours rare inflammatory bowel disease risk variants, results in endoplasmic reticulum (ER) stress, Paneth cell impairment and spontaneous enteritis4. Unresolved ER stress is a common feature of inflammatory bowel disease epithelium4,5, and several genetic risk factors of Crohn’s disease affect Paneth cells2,4,6,7,8,9. Here we show that impairment in either UPR (Xbp1ΔIEC) or autophagy function (Atg16l1ΔIEC or Atg7ΔIEC) in intestinal epithelial cells results in each other’s compensatory engagement, and severe spontaneous Crohn’s-disease-like transmural ileitis if both mechanisms are compromised. Xbp1ΔIEC mice show autophagosome formation in hypomorphic Paneth cells, which is linked to ER stress via protein kinase RNA-like endoplasmic reticulum kinase (PERK), elongation initiation factor 2α (eIF2α) and activating transcription factor 4 (ATF4). Ileitis is dependent on commensal microbiota and derives from increased intestinal epithelial cell death, inositol requiring enzyme 1α (IRE1α)-regulated NF-κB activation and tumour-necrosis factor signalling, which are synergistically increased when autophagy is deficient. ATG16L1 restrains IRE1α activity, and augmentation of autophagy in intestinal epithelial cells ameliorates ER stress-induced intestinal inflammation and eases NF-κB overactivation and intestinal epithelial cell death. ER stress, autophagy induction and spontaneous ileitis emerge from Paneth-cell-specific deletion of Xbp1. Genetically and environmentally controlled UPR function within Paneth cells may therefore set the threshold for the development of intestinal inflammation upon hypomorphic ATG16L1 function and implicate ileal Crohn’s disease as a specific disorder of Paneth cells.
The UPR and autophagy are integrally linked pathways10. To investigate their relationship in the intestinal epithelium, we stably transduced the small intestinal epithelial cell line MODE-K with a short hairpin Xbp1 (shXbp1) lentiviral vector (Extended Data Fig. 1a)4. We observed increased levels of ATG7, beclin 1 (Extended Data Fig. 1b) and phosphatidylethanolamine (PE)-conjugated microtubule-associated protein 1 light chain 3-β (LC3-II) relative to LC3-I compared to control-silenced (shCtrl) cells (Extended Data Fig. 1c). Increased autophagic flux accounted for this, given increased levels of LC3-II relative to LC3-I observed after inhibition of autophagosome–lysosome fusion by bafilomycin11 (Extended Data Fig. 1d). Increased GFP–LC3 punctae were seen in shXbp1 compared to shCtrl MODE-K cells transfected with a green fluorescent protein (GFP)–LC3-expressing vector12 (Extended Data Fig. 1e, f), as well as increased numbers of autophagic vacuoles in shXbp1 compared to shCtrl cells (Extended Data Fig. 1g, h). Accordingly, isolated primary intestinal epithelial cells from the small intestine of villin (V)-cre+;Xbp1fl/fl (hereafter called Xbp1ΔIEC)4 mice backcrossed onto C57BL/6 (B6) background exhibited nearly complete consumption of LC3-I and a relative increase in LC3-II (Fig. 1a and Extended Data Fig. 1i), stable amounts of ATG7 (presumably reflecting a combination of increased production and consumption), elevated levels of ATG16L1 and beclin 1 (Fig. 1b and Extended Data Fig. 1j), and autophagosomes and degradative autophagic vacuoles consistent with autophagy induction in hypomorphic4 Paneth cells, and to a lesser extent goblet cells (data not shown), compared to V-cre–;Xbp1fl/fl (hereafter called wild type) mice (Fig. 1c and Extended Data Fig. 1k). To gain temporal control of Xbp1 deletion and the ability to monitor autophagy in situ and exclude the role of chronic inflammation4 in this induction, we generated V-creERT2;Xbp1fl/fl (hereafter called Xbp1T-ΔIEC) mice on a B6 background crossed to GFP-LC3 transgenic mice12. Three days after tamoxifen-induced Xbp1 deletion (Extended Data Fig. 2a), although mature Paneth cells remained present with little detectable inflammation (data not shown), punctate GFP signal accumulation was greatest at the bottom of the crypts of Lieberkühn (Fig. 1d, e), and co-localized with lysozyme-positive Paneth cells (Extended Data Fig. 2b). Purified crypts of Xbp1T-ΔIEC mice revealed increased LC3-I/II conversion and reduced p62 compared to wild-type mice (Extended Data Fig. 2c). Thus, Xbp1 loss in intestinal epithelial cells induced autophagy most notably in Paneth cells.
Although Xbp1-deficient intestinal epithelial cells exhibited broad evidence of ER stress4 (Extended Data Fig. 2d–j), an examination of shXbp1, relative to shCtrl, MODE-K cells demonstrated a particularly significant increase in phosphorylated (p)-PERK and its substrate p-eIF2α (Fig. 1f), with the latter reversed by Perk (also called Eif2ak3) silencing, identifying it as the factor responsible for p-eIF2α formation (Extended Data Fig. 3a). Increased p-eIF2α was also detected in primary intestinal epithelial cells of Xbp1ΔIEC (Fig. 1g and Extended Data Fig. 3b) and Xbp1T-ΔIEC mice (Extended Data Fig. 3c). Consistent with the involvement of PERK–eIF2α in autophagy induction, ATF4, a transcriptional effector of this pathway, and its transcriptional target, C/EBP-homologous protein (CHOP; encoded by Ddit3), were increased in primary intestinal epithelial cells of Xbp1ΔIEC mice (Fig. 1g and Extended Data Fig. 3b), and chromatin-immunoprecipitation (ChIP) with anti-ATF4 demonstrated increased binding to the Map1lc3b (LC3b) (Fig. 1h) and Atg7 (Fig. 1i) promoters, both of which contain ATF4 binding sites13, in shXbp1 relative to shCtrl MODE-K cells. ATG7 is essential for the formation of the ATG12–ATG5 conjugate during autophagy10,14. shXbp1 MODE-K cells showed increased LC3b and Atg7 expression compared to shCtrl MODE-K cells (Extended Data Fig. 3d), and Perk co-silencing abrogated ATG7 induction observed in shXbp1 compared to shCtrl MODE-K cells (Extended Data Fig. 3a). Salubrinal, a selective inhibitor of eIF2α dephosphorylation15 (Extended Data Fig. 2a), increased the accumulation of GFP–LC3 punctae primarily in Paneth cells, in both Xbp1-sufficient and -deficient intestinal epithelial cells (Fig. 1j and Extended Data Fig. 2b and 3e), and provoked increased levels of LC3-II relative to LC3-I (Extended Data Fig. 3f) and CHOP (Extended Data Fig. 3f) in Xbp1T-ΔIEC mice, along with, importantly, an amelioration of the acute enteritis (Fig. 1k and Extended Data Fig. 3g). Similarly, silencing of growth arrest and DNA-damage-inducible protein 34 (Gadd34; also called Ppp1r15a), part of the protein phosphatase 1 complex that dephosphorylates eIF2α16, led to increased eIF2α phosphorylation and ATG7 expression in shXbp1 compared to shCtrl MODE-K cells (Extended Data Fig. 3h, i). Xbp1ΔIEC;Gadd34+/− mice with hypomorphic GADD34 function exhibited increased p-eIF2α and ATG7 in purified crypt epithelial cells compared to Xbp1ΔIEC;Gadd34+/+ mice (Extended Data Fig. 3j). Thus, PERK–p-eIF2α is a critical mediator of UPR-induced autophagy primarily in Paneth cells consequent to XBP1 deficiency.
These studies led us to propose that autophagy may function as a compensatory mechanism in intestinal epithelial cells upon sustained ER stress. We therefore generated V-cre+;Atg7fl/fl;Xbp1fl/fl (Atg7/Xbp1ΔIEC) mice14. Intestinal epithelial cells from Atg7ΔIEC mice lacked LC3-II formation and the ATG12–ATG5-ATG16L1 complex (Extended Data Fig. 4a). Atg7/Xbp1ΔIEC mice demonstrated a complete absence of UPR-induced autophagy (Fig. 2a and Extended Data Fig. 4a–c), and a remarkable worsening of ileitis compared to Xbp1ΔIEC mice. In notable contrast to Xbp1ΔIEC mice, where inflammation was limited to the mucosa, >70% of Atg7/Xbp1ΔIEC mice developed discontinuous submucosal or transmural inflammation, characterized by acute and chronic inflammation extending in an abrupt knife-like fashion to muscularis propria and serosa, closely resembling the early fissuring ulcerations and fistulous tracts observed in human Crohn’s disease (Fig. 2b, c and Extended Data Fig. 4d). In contrast to Xbp1ΔIEC mice, enteritis in Atg7/Xbp1ΔIEC mice progressed over the 18-week observation period such that at this time point all animals exhibited submucosal or transmural disease (Extended Data Fig. 4e, f).
ATG16L1 is a major genetic risk factor for Crohn’s disease1,17, especially ileal Crohn’s disease18. Complex formation of ATG16L1 protein with ATG12–ATG5 defines the site of LC3 PE conjugation during autophagosome formation19,20. ATG16L1 protein expression was markedly increased in Xbp1ΔIEC compared to wild-type primary intestinal epithelial cells (Fig. 1b and Extended Data Fig. 1j), presumably consequent to PERK/eIF2α/ATF4-dependent transactivation of Atg7 and LC3b promoters and stabilization by the ATG7-induced ATG12–ATG5 complex21. We therefore developed mice with a floxed Atg16l1 allele that would allow for intestinal-epithelial-cell-specific deletion via V-cre (Atg16l1ΔIEC; Extended Data Fig. 4g–i). Paneth cells in Atg16l1ΔIEC mice demonstrated a reduction in their overall size and number of granules, similar to gene-trap-targeted Atg16l1HM mice2,3 (Extended Data Fig. 4j–n). Intestinal epithelial cells from Atg16l1ΔIEC mice, compared to wild-type mice, exhibited reduced expression of ATG7 and the ATG12–ATG5 conjugate (Extended Data Fig. 5a), along with disruption of the secretory pathway with a distended ER, reduced size and number of secretory granules, a loss of homeostatic autophagy (Fig. 2d and Extended Data Fig. 5b, c) and increased p62 immunoreactivity in crypts (Extended Data Fig. 5d). To address the role of ATG16L1 under ER stress conditions, we generated V-cre+;Atg16l1fl/fl;Xbp1fl/fl (Atg16l1/Xbp1ΔIEC) mice. Atg16l1/Xbp1ΔIEC mice, which lacked UPR-induced autophagy (Extended Data Fig. 5b, c), developed severe spontaneous ileitis compared to Xbp1ΔIEC or Atg16l1ΔIEC mice, with discontinuous submucosal or transmural inflammation in >70% of 18-week-old animals (Fig. 2b, e and Extended Data Fig. 5e, f) with features similar to those observed in Atg7/Xbp1ΔIEC mice, and present in two distinct animal facilities (Fig. 2b, e and Extended Data Fig. 5g). This phenotype highlights the important compensatory role played by autophagy and in particular ATG16L1 in defending against inflammation arising from unabated ER stress precisely in the small intestinal epithelium as a consequence of XBP1 deficiency.
ATG7 hypofunction in hepatocytes can induce ER stress22, raising the possibility that cross-talk between the UPR and autophagy may be bi-directional in intestinal epithelial cells. Isolated Atg16l1ΔIEC crypts exhibited increased Xbp1 splicing compared to wild type (Fig. 2f, g) and increased Grp78 (also called Hspa5) expression (Extended Data Fig. 5h) localized to the crypt bottom (Fig. 2h). Dextran sodium sulphate, a colitis model involving ER stress in intestinal epithelial cells that can be treated with ER stress-relieving chaperones23,24, induced more inflammation in Atg16l1ΔIEC compared to wild-type mice (Extended Data Fig. 5i–l), similar to Atg16l1HM mice3. Thus, disturbances in autophagy within intestinal epithelial cells also affect the UPR, and autophagy-associated factors such as ATG16L1 endow intestinal epithelial cells with the ability to mitigate ER stressors that are commonplace at the mucosal surface25.
We next turned our attention to mechanisms by which autophagy counteracts ER stress and synergistically increases intestinal inflammation when absent. Increased numbers of TdT-mediated dUTP nick end labelling (TUNEL)+ cells in Atg16l1/Xbp1ΔIEC and Atg7/Xbp1ΔIEC mice (Extended Data Fig. 6a, b) correlated with enteritis severity in double-mutant, in contrast to Xbp1ΔIEC, mice (Extended Data Fig. 6b), and, as demonstrated for Atg16l1/Xbp1ΔIEC mice, concomitantly with increasing age (Extended Data Fig. 6c and Supplementary Video 1). Silencing of Atg16l1 in shXbp1 MODE-K cells significantly increased the proportion of apoptotic cells in vitro (Fig. 3a and Extended Data Fig. 6d), indicating that increased apoptosis could function as an initial event in intestinal inflammation. Furthermore, intestinal-epithelial-cell-associated NF-κB activation, a critical player in intestinal inflammation26, was absent in primary intestinal epithelial cells from wild-type mice and gradually increased in Atg16l1ΔIEC, Xbp1ΔIEC and Atg16l1/Xbp1ΔIEC mice (Fig. 3b and Extended Data Fig. 6e, f). shXbp1 MODE-K cells exhibited increased NF-κB activation when stimulated with tumour-necrosis factor (TNF) (Fig. 3c and Extended Data Fig. 6g, h) or Toll-like receptor ligands (data not shown) relative to shCtrl MODE-K cells, demonstrating increased sensitivity of XBP1-deficient intestinal epithelial cells to inflammatory and environmental stimuli. Inhibition of NF-κB by treatment with BAY11-7082 decreased intestinal epithelial cell death in Xbp1T-ΔIEC mice (Fig. 3d and Extended Data Fig. 6i) and protected from enteritis in both Xbp1T-ΔIEC and Xbp1ΔIEC mice (Fig. 3e and Extended Data Fig. 6j) relative to vehicle-treated mice. A progressive increase in intestinal epithelial cells of total and phosphorylated IRE1α (encoded by Ern1), the sensor of ER stress upstream of XBP1 and known to control NF-κB27, was observed in Atg16l1ΔIEC, Xbp1ΔIEC and Atg16l1/Xbp1ΔIEC mice in comparison to wild-type mice (Fig. 3f) and mirrored the escalating elevations in NF-κB (Fig. 3b and Extended Data Fig. 6e, f). Indeed, increased NF-κB activity and ileitis were governed by IRE1α, as the increased epithelial NF-κB phosphorylation observed in Xbp1ΔIEC mice was abrogated in Ern1/Xbp1ΔIEC mice (Fig. 3g), Ern1 co-silencing of shXbp1 MODE-K cells abolished the increased expression of the NF-κB target gene Nfkbia in response to TNF stimulation relative to shCtrl MODE-K cells (Extended Data Fig. 6k), and enteritis was diminished in Ern1/Xbp1ΔIEC compared to Xbp1ΔIEC mice (Fig. 3h). Enteritis was also reversed by germline deletion of Tnfrsf1a in Xbp1ΔIEC mice (Fig. 3i), and by re-derivation of Xbp1ΔIEC mice into a germ-free environment (Fig. 3j), which was associated with reduced p-IκBα immunoreactivity in Xbp1-deficient intestinal epithelial cells compared to mice housed under specific pathogen-free (SPF) conditions (Fig. 3k). These studies together demonstrate that enteritis in this model is driven by TNF—the cytokine targeted by the most potent therapeutics of human Crohn’s disease25—and microbes in a pathway that derives from intestinal epithelial cell death and IRE1α-dependent activation of NF-κB, with ATG16L1-dependent autophagy serving to restrain the inflammatory nature of the latter, probably through the removal of IRE1α.
Accordingly, treatment of Xbp1T-ΔIEC mice with the mTOR inhibitor rapamycin10 induced autophagy primarily in crypts (Extended Data Fig. 7a, b), diminished intestinal-epithelial-cell-associated NF-κB activation and number of TUNEL+ intestinal epithelial cells in Xbp1ΔIEC mice (Fig. 4a and Extended Data Fig. 7c–e), and markedly reduced the severity of enteritis (Fig. 4b). These beneficial consequences of rapamycin in the setting of unabated ER stress in intestinal epithelial cells were not observed in Atg7/Xbp1ΔIEC and Atg16l1/Xbp1ΔIEC mice (Extended Data Fig. 7e–i), demonstrating that these effects required intact autophagy within intestinal epithelial cells.
The spatial convergence of the consequences of hypomorphic UPR and autophagy function in Paneth cells prompted us to test the hypothesis that these pathways were interdependent in this cell type. We developed a defensin 6 alpha promoter28-driven cre transgenic line (D6-cre), and confirmed the exclusive activity of Cre recombinase in lysozyme+ Paneth cells (Fig. 4c). Paneth-cell-specific deletion of Xbp1 in D6-cre;Xbp1fl/fl (Xbp1ΔPC) mice resulted in autophagy activation (Fig. 4d and Extended Data Fig. 8a, b) and structural defects in granule morphology in Paneth cells (Fig. 4e, f). Crypts of Xbp1ΔPC compared to wild-type mice exhibited increased GRP78 expression (Fig. 4d and Extended Data Fig. 8b) and p-eIF2α immunoreactivity (Fig. 4g), accompanied by increased conversion of LC3-I/II and reduced p62 levels (Fig. 4d and Extended Data Fig. 8b), demonstrating ER stress and autophagy induction. Notably, 75% of Xbp1ΔPC mice developed spontaneous enteritis (Fig. 4h, i) that shared the histological inflammatory characteristics observed in Xbp1ΔIEC mice (Fig. 2b). Xbp1ΔPC mice exhibited increased cell death in crypts (Extended Data Fig. 9a, b) and intestinal epithelial cell turnover (Extended Data Fig. 9c–f), whereas goblet cells remained unaffected (Extended Data Fig. 9g, h). Notably, Atg7/Xbp1ΔPC mice exhibited transmural disease as early as 8 weeks of age (Extended Data Fig. 9i). We conclude that deletion of Xbp1 specifically in Paneth cells results in unresolved ER stress, UPR activation and induction of autophagy, which can serve as a nidus for the emergence of spontaneous intestinal inflammation that evolves into transmural disease in the absence of the compensation provided by autophagy.
Our studies thus support a mechanistic model (Extended Data Fig. 10a–c) for how ATG16L1-associated genetic risk may convert into a disease phenotype. The competence of the UPR probably sets the threshold for susceptibility of the host with hypofunctional autophagy to interacting genetic and environmental factors capable of inducing inflammation. Consistent with our model, patients with Crohn’s disease carrying the ATG16L1T300A risk variant, which impairs autophagosome formation29, frequently exhibit ER stress in their Paneth cells, in contrast to those harbouring the normal variant30. Finally, our studies also unequivocally establish that these inflammatory susceptibilities can emerge directly from highly secretory Paneth cells and suggest that small intestinal Crohn’s disease may be a specific disorder of this cell type.
Villin (V)-cre+;Xbp1fl/fl (Xbp1ΔIEC)4, V-creERT2+;Xbp1fl/fl (Xbp1T-ΔIEC)4, Atg7fl/fl (ref. 14), Atg16l1fl/fl (generated as detailed in Methods) and Gadd34–/– mice16, all backcrossed >8 generations onto B6, and GFP-LC3 (ref. 12), Ern1fl/fl and Tnfrsf1a–/– mice were used to generate experimental mouse strains. Experiments were performed with sex- and age-matched animals using littermates where technically possible. Generation of Paneth-cell-specific Defa6-cre;Xbp1fl/fl (Xbp1ΔPC) mice is described in Methods. Murine norovirus (MNV) status is reported in Extended Data Fig. 9j. Stable clones of the Xbp1-silenced (shXbp1) small intestinal epithelial cell line MODE-K were generated via a lentiviral shRNA vector and selected by hygromycin4. The sources of antibodies and reagents are listed in the Methods. Immunoblot, chromatin immunoprecipitation, qRT–PCR, histological analysis, transmission electron microscopy and confocal microscopy techniques are described in Methods. Statistical significance was calculated using an unpaired two-tailed Student’s t-test or a Mann–Whitney U-test and considered significant at P < 0.05. In experiments where more than two groups were compared, Kruskal–Wallis test followed by Mann–Whitney U and post-hoc Bonferroni Holm’s correction or one-way ANOVA/Bonferroni were performed as appropriate. Data were analysed using GraphPad Prism software.
Xbp1fl/fl mice were backcrossed >8 generations onto B6 and crossed with V-cre-ERT2+ (B6) mice provided by N. Davidson and S. Robine to generate mice with tamoxifen-inducible Xbp1 deletion in the intestinal epithelium (Xbp1T-ΔIEC), and with V-cre (B6) mice to generate mice with constitutively active Xbp1 deletion in the intestinal epithelium (Xbp1ΔIEC). V-cre-ERT2+ recombinase was activated by 3 or 5 daily intraperitoneal administrations of 1 mg tamoxifen (MP Biomedicals) as indicated in figure legends. V-cre+;Xbp1fl/fl (Xbp1ΔIEC) mice were crossed with Atg7fl/fl mice14 provided by M. Komatsu to obtain V-cre+;Atg7fl/fl;Xbp1fl/fl (Atg7/Xbp1ΔIEC) mice. Mice with a floxed Atg16l1 allele were generated in collaboration with GenOway. Briefly, a proximal loxP site was introduced within the promoter region of the Atg16l1 gene upstream of exon 1, a distal loxP site was introduced with an FRT flanked neomycin selection cassette within intron 1. The resultant mouse line was bred with deleter-mice constitutively expressing Flp recombinase to remove the neomycin selection cassette, creating an Atg16l1fl/+ mouse in which Atg16l1 exon 1 was flanked by two loxP sites (Extended Data Fig. 4g). After backcrossing onto B6, these mice were crossed with V-cre+ mice31 resulting in V-cre;Atg16l1fl/fl mice with intestinal-epithelial-cell-specific Atg16l1 deletion (Atg16l1ΔIEC). Atg16l1fl/fl mice were crossed with Xbp1ΔIEC mice to develop V-cre+;Atg16l1fl/fl;Xbp1fl/fl (Atg16l1/Xbp1ΔIEC) mice. GFP-LC3 transgenic mice12 (gift of N. Mizushima) were crossed with V-cre-ERT2+;Xbp1fl/fl to generate GFP-LC3;V-cre-ERT2+;Xbp1fl/fl mice (GFP-LC3;Xbp1T-ΔIEC). For the generation of Paneth-cell-specific Defa6-cre;Xbp1fl/fl mice, a 1.1-kb cDNA fragment encoding improved Cre (iCre)32 recombinase was subcloned downstream of nucleotides −6,500 to +34 of mouse cryptdin-2 gene (Defa6) in the BamHI site of the pCR2-TAg-hGH plasmid33,34 to replace the DNA fragment containing the simian virus 40 large antigen (SV40). A linearized 10.2-kb fragment containing the Defa6 promoter, iCre and hGH (Defa6-iCre-hGH) was removed by EcoRI digestion, agarose gel-electrophoresed and purified with the Qiaex Gel Extraction kit (Qiagen), and used for pronuclear injection of B6 mice. Six founders from 22 live-born mice were identified by screening tail DNA using iCre-specific primers, and two lines were further characterized and crossed to Xbp1fl/fl, Atg7fl/fl and EYFP-Rosa26 reporter mice35, provided by K. Rajewsky. Gadd34–/– mice16 were crossed with V-cre;Xbp1fl/fl mice to generate V-cre;Xbp1fl/fl;Gadd34+/– (Xbp1ΔIEC;Gadd34+/–) mice. Ern1fl/fl (ref. 36) and Tnfrsf1a–/– (Jackson) mice were crossed with V-cre;Xbp1fl/fl mice to generate V-cre;Ern1fl/fl;Xbp1fl/fl (Ern1/Xbp1ΔIEC) and V-cre;Xbp1fl/fl;Tnfrsf1a–/– (Xbp1ΔIEC/Tnfrsf1a–/–) mice, respectively. Cre transgenes were maintained in the hemizygous state in all experimental strains with a floxed allele to generate littermate controls. Xbp1ΔIEC mice were re-derived in a germ-free environment and housed in sterile isolators at the Taconic Farms breeding facility. Tail or ear biopsy genomic DNA was used for genotyping of respective mouse strains as described previously4. Primer sequences are available on request. Mice were housed in specific pathogen free (SPF) barrier facilities at Harvard Medical School (Xbp1ΔIEC, Xbp1T-ΔIEC, Atg7ΔIEC, Atg7/Xbp1ΔIEC, Atg7/Xbp1ΔPC, Xbp1ΔPC, GFP-LC3;Xbp1T-ΔIEC, Ern1/Xbp1ΔIEC mice and their respective controls), University of Cambridge (Xbp1ΔIEC, Xbp1T-ΔIEC, Atg16l1ΔIEC, Atg16l1/Xbp1ΔIEC, Xbp1ΔIEC/Gadd34+/–, Ern1/Xbp1ΔIEC mice and their respective controls), Innsbruck Medical University (Xbp1ΔIEC, Xbp1T-ΔIEC, Atg16l1ΔIEC, Atg16l1/Xbp1ΔIEC, Xbp1ΔIEC/Tnfrsf1a–/– mice and their respective controls), and Christian-Albrechts-Universität zu Kiel (Atg16l1ΔIEC mice and their respective controls). Colonies maintained at Boston and Innsbruck were murine norovirus (MNV) positive by Taqman qRT–PCR (Extended Data Fig. 9j). Xbp1ΔIEC, Atg16l1ΔIEC, Ern1ΔIEC and their associated double-mutant strains were re-derived from the Innsbruck colony into the MNV-free enhanced barrier Cambridge facility, and colonies confirmed MNV-negative by PCR (Extended Data Fig. 9j) and serology (data not shown), as were Atg16l1ΔIEC mice held at the Kiel facility. MNV Taqman qRT–PCR was performed as described37. The phenotype of single- and double-mutant colonies that had been re-derived from the MNV+ Innsbruck facility into the MNV– Cambridge facility were indistinguishable, in particular relating to qualitative and quantitative measures of enteritis and the reciprocal induction of autophagy and ER stress. Mice were handled and all experiments performed in accordance with institutional guidelines and with the approval of the relevant authorities. 8–10-week-old mice were used for all experiments unless stated otherwise in the figure legend, and were randomly allocated into treatment groups.
Antibodies and reagents
The following antibodies and reagents were used for immunoblotting. Sigma Aldrich: anti-LC3B (L7543); Cell Signaling Technology: anti-β-actin (4970; 13E5) anti-GAPDH (2118; 14C10), anti-eIF2α (9722), anti-phospho-eIF2α (3597; 119A11), anti-PERK (3192; C33E10), anti-phospho-PERK (3179; 16F8), anti-JNK (9252), anti-phospho-JNK (4668; 81E11), anti-ATG5 (8540; D1G9), anti-beclin 1 (3495; D40C5), anti-ATG7 (8558; D12B11), anti-CHOP (5554; D46F1), anti-ATG12 (4180; D88H11), anti-p62 (5114), anti-IKK1 (2682), anti-IKK2 (2370; 2C8), anti-phospho-IKK1/2 (2697; 16A6), anti-IRE1α (3294; 14C10), anti-phospho-NF-κB p65 (3033; 93H1), anti-NF-κB p65 (4764; C22B4) and anti-rabbit/mouse HRP antibodies (7074, 7076); Abcam: anti-phospho-IRE1α (48187); MBL: anti-ATG16L1 (M150-3; 1F12); Stressgen: anti-haem-oxygenase-1 (ADI-SPA-895); Novus Biologicals: anti-GRP78 (NBP1-06274); Santa Cruz Biotechnology: anti-ATF4 (sc-200; C20). Immunoprecipitation antibody: anti-IRE1α (Santa Cruz Biotechnology, 20790; H190). Immunohistochemistry antibodies: Santa Cruz Biotechnology: anti-lysozyme (27958; C19); MBL: anti-ATG16L1 (M150-3; 1F12); Cell Signaling Technology: anti-ATG16L1 (8089; D6D5), anti-phospho-IκBα (2859; 14D4), anti-phospho-eIF2α (3597; 119A11); Abcam: anti-Ki67 (15580), anti-GRP78 (21685); Progen: anti-p62 (GP62-C); BD Bioscience: anti-BrdU (551321).
The following reagents were used: TNF (Peprotech, 315-01A), bafilomycin A1 (Sigma Aldrich, B1793), rapamycin (LC Laboratories, R-5000), JNK inhibitor (Sigma Aldrich, SP600125), NF-κB inhibitor (Calbiochem, BAY11-7082) and salubrinal (Alexis Biochemicals, ALX-270-428) were dissolved in DMSO as recommended. N-acetyl cysteine (Sigma Aldrich, A9165) and glutathione (Calbiochem, NOVG3541) were used at final concentration of 1 mM. Ambion siRNA for Atg16l1 (94892, sense 5′-GAACUGUUAGGGAAGAUCATT-3′, antisense 5′-UGAUCUUCCCUAACAGUUCCA-3′), Perk (65405, sense 5′-CCCGAUAUCUAACAGAUUUTT-3′, antisense 5′-AAAUCUGUUAGAUAUCGGGAT-3′; 65406, sense 5′-CGAAGAAUACAGUAAUGGUTT-3′, antisense 5′-ACCAUUACUGUAUUCUUCGTG-3′), Gadd34 (70230, sense 5′-CCAUAGCUCCGGGAUACAATT-3′, antisense 5′-UUGUAUCCCGGAGCUAUGGAA-3′; 70231, sense 5′-AGACAACAGCGAUUCGGAUTT-3′, antisense: 5′-AUCCGAAUCGCUGUUGUCUTC-3′), Ern1 (95857, sense 5′-GUUUGACCCUGGACUCAAATT-3′, antisense 5′-UUUGAGUCCAGGGUCAAACTT-3′; 95858, sense 5′-GGAUGUAAGUGACCGAAUATT-3′, antisense 5′-UAUUCGGGUCACUUACAUCCTG-3′; 95859, sense 5′-GCUCGUGAAUUGAUAGAGATT-3′, antisense 5′-UCUCUAUCAAUUCACGAGCAA-3′) and scrambled control were used at a final concentration of 10 μM.
ChIP with anti-ATF4 (Santa Cruz Biotechnology) and control IgG rabbit antibody was performed in Xbp1 and control silenced MODE-K cells according to ChIP protocol by Agilent. To determine the presence of ATF4 binding sites in the Atg7 promoter, a 4-kb region proximal to the transcription start site identified with the Eukaryotic Promoter Database Primers was analysed using MatInspector (Genomatix). Immunoprecipitated DNA was subject to quantitative PCR (qPCR) to determine enrichment of ATF4 binding to respective promoters and results were normalized to input chromatin DNA. Primers used for qPCR were as follows for Map1LC3b (ref. 13), for Atg7 forward 5′-GCGCTTCCGCGTTTGTGTGG-3′ and reverse 5′-CTGCTCCGCAACCACGGCTT-3′.
Salubrinal, rapamycin and BAY11-7082 treatment in vivo
Salubrinal (1 mg kg−1 d−1), rapamycin (1.5 mg kg−1 d−1) or vehicle (DMSO) was administered intraperitoneally (i.p.) 24 h before the first tamoxifen administration to GFP-LC3;V-creERT2;Xbp1fl/fl (GFP-LC3;wild type) and GFP-LC3;Xbp1T-ΔIEC mice. 3-day treatment was used for evaluation of accumulation of GFP–LC3 punctae in the intestinal epithelium, whereas a 5-day combined tamoxifen and salubrinal treatment followed by two daily salubrinal injections was used in experiments with enteritis assessment as an end point (Extended Data Figs 2a and 3g). To assess the effects of rapamycin on ER stress-induced intestinal inflammation in XBP1 deficiency, V-cre;Xbp1fl/fl (Xbp1ΔIEC), V-cre;Atg16l1fl/fl;Xbp1fl/fl (Atg16l1/Xbp1ΔIEC) or V-cre;Atg7fl/fl;Xbp1fl/fl (Atg7/Xbp1ΔIEC) and the respective control mice were treated with rapamycin or vehicle for 14 consecutive days i.p. and inflammation was evaluated. BAY11-708238 or vehicle (DMSO) was administered i.p. every other day at 5 mg kg−1 for 14 consecutive days in Xbp1ΔIEC mice, or for 5 consecutive days at 20 mg kg−1 in Xbp1T-ΔIEC mice concomitant with i.p. tamoxifen.
Transmission electron microscopy
Small intestinal tissue from mice was handled by standard methods to be fixed with 1.25% glutaraldehyde, 4% formaldehyde in 0.1 M cacodylate buffer at pH 7.4 at room temperature for electron microscopy. The detailed procedures for electron microscopy was previously described39 and the tissue was observed with a JEOL 1400 transmission electron microscope at 120 kV operating voltage. For quantification of autophagy, number of autophagic vacuoles was manually counted by a TEM expert (J.H.) blinded to sample identity in 10 consecutive Paneth cells per sample. ImageJ software was used to measure average size of autophagic vacuoles.
Formalin-fixed and paraffin-embedded intestinal tissue was sectioned and stained with haematoxylin and eosin as previously described4. A semi-quantitative composite scoring system was used for the assessment of spontaneous intestinal inflammation, computed as a sum of five histological subscores, multiplied by a factor based on the extent of the inflammation. Histological subscores (for each parameter: 0, absent; 1, mild; 2, moderate; 3, severe): mononuclear cell infiltrate (0–3), crypt hyperplasia (0–3), epithelial injury/erosion (0–3), polymorphonuclear cell infiltrates (0–3) and transmural inflammation (0, absent; 1, submucosal; 2, one focus extending into muscularis and serosa; 3 up to five foci extending into muscularis and serosa; 4, diffuse). Extent factor was derived according to the fraction of bowel length involved by inflammation: 1, <10%; 2, 10–25%; 3, 25–50%; and 4, >50%. Ileal inflammation was assessed by an expert gastrointestinal pathologist (J.N.G.) who was blinded to the genotype and experimental conditions of the samples. No spontaneous colonic inflammation was detected in any of the reported genotypes.
Reactive oxygen species, cell death detection by flow cytometry and NF-κB activity assays
Stable clones of the Xbp1-silenced (shXbp1) small intestinal epithelial cell line MODE-K were generated via a lentiviral shRNA vector and selected by hygromycin, as previously described4. To evaluate oxidative stress, Xbp1- and control-silenced MODE-K cells40 were incubated with 5 μM 5-(and-6)-chloromethyl-2′,7′-dichlorodihydrofluorescein diacetate, acetyl ester (CM-H2DCFDA) (Molecular Probes) for 30 min41. After washing with PBS, cells were further incubated with complete medium for 2 h. Reactive oxygen species generation was determined using flow cytometry. To evaluate cell death, shXbp1 and shCtrl cells were co-silenced for Atg16l1 using siRNA or scrambled control (Ambion). After 4 days, cells were collected and stained for annexin V (Biolegend) in staining buffer (Biolegend) and mode of cell death was determined by flow cytometry after addition of propidium iodide (PI). To assess NF-κB signalling pathway activation, Xbp1- or control-silenced MODE-K cells were stimulated with 50 ng ml−1 TNF for indicated periods of time, followed by immunoblotting (using NE-PER (Thermo Scientific) isolated cytoplasmic extracts), qRT–PCR, and chemiluminescent detection of NF-κB consensus sequence binding activity (in nuclear extracts isolated with NE-PER) with the NF-κB p65 transcription factor assay kit (Thermo Scientific). Ern1 or scrambled siRNA (Ambion) was used for co-silencing as indicated.
Immunohistochemistry, BrdU and TUNEL labelling
Formalin-fixed paraffin-embedded sections were stained according to standard immunohistochemistry protocols and manufacturer’s recommendations as described previously4. Cell death was assessed by TdT-mediated dUTP nick end labelling (TUNEL) of formalin-fixed paraffin-embedded slides of the respective genotypes using the TUNEL cell death detection kit (Roche). Entire slides were analysed for TUNEL+ cells and numbers normalized to intestinal length on the slide. Proliferation of the intestinal epithelium was assessed after a 24-h pulse with 5-bromodeoxyuridine (BrdU; BD Pharmingen) and incorporated BrdU was detected by the BrdU in situ detection kit (BD Pharmingen). BrdU+ nuclei per total intestinal epithelial cells along the crypt villus axis are shown. Toluidine blue staining and Periodic acid-Schiff (PAS) reaction was performed according to standard protocols.
Confocal microscopy for detection of GFP–LC3 and EYFP
For detection of GFP–LC3 or EYFP, mice were euthanized, followed by transcardiac perfusion with PBS (2–3 min) and 3.7% formaldehyde (3–4 min). Small intestine was dissected and promptly washed with PBS. The tissue was fixed in formalin for an additional 12–18 h. Fixed tissue was embedded in OCT and sectioned on a cryotome into 5 μm sections. Slides were washed with PBS and mounted with Prolong Gold Antifade reagent with DAPI (Invitrogen). Images of the sections were collected using Olympus semi-confocal system. MetaMorph software was used for image analysis. For the detection of autophagosome formation in vitro, Xbp1 and control silenced MODE-K cells were transfected with a GFP–LC3 plasmid12 (gift from N. Mizushima) with use of Lipofectamine LTX (Invitrogen) following the manufacturer’s instructions. Accumulation of GFP–LC3 punctae or EYFP signal was assessed using LSM510 META confocal microscopy (Carl Zeiss).
Intestinal epithelial cell purification and crypt isolation
Mice were euthanized and the intestine was washed with ice-cold PBS after being cut open longitudinally. Peyer’s patches were removed and the intestine was cut into small pieces. Mucus was removed by shaking the intestine in 1× HBSS containing 1 mM DTT for 10 min at room temperature. After washing with PBS, pieces were digested with dispase (1 U ml−1 in RPMI with 2% FCS) for 30 min at 37 °C with shaking (250 r.p.m.). Cells were collected and debris removed with a 100 μm cell strainer, and centrifuged for 5 min at 1,500 r.p.m. Intestinal epithelial cells were collected in the top layer after 40–100% Percoll gradient centrifugation. Purity of the population was determined by staining with anti-EpCAM antibody and flow cytometry analysis. Intestinal epithelial cells were lysed with RIPA buffer and equal amounts of protein were used for western blot analysis as indicated in figure legends. To isolate small intestinal crypts, the intestine was flushed, cut open longitudinally and incubated on ice for 30 min in 2 mM EDTA/PBS. Two sedimentation steps and application of a cell strainer separated crypts from villi42, which were then used for RNA isolation (Qiagen), Xbp1 splicing assay, and for protein lysis with RIPA buffer and subsequent immunoblotting.
Intestinal epithelial scrapings
Mice were euthanized, intestines collected and longitudinally opened, and immediately washed with ice-cold PBS. Intestinal epithelium was collected by scraping with glass slides and snap frozen into liquid nitrogen for further analysis. For protein analysis, intestinal epithelial scrapings were homogenized in RIPA buffer using a 25G needle with a syringe. Lysates were cleared by centrifugation and aliquots of protein were used for protein assessment using standard western blot or immunoprecipitation protocols as indicated. Isolation of mRNA from intestinal epithelial scrapings or MODE-K lysates and RT-qPCR was performed as described4 using following pairs of primers. grp78 5′-ACTTGGGGACCACCTATTCCT-3′ and 5′-ATCGCCAATCAGACGCTCC-3′; Gadd34 5′-CCCGAGATTCCTCTAAAAGC-3′ and 5′-CCAGACAGCAAGGAAATGG-3′; LC3b 5′-GCGCCATGCCGTCCGAGAAG-3′ and 5′-GCTCCCGGATGAGCCGGACA-3′; Xbp1 5′-AGCAGCAAGTGGTGGATTTG-3′ and 5′-GAGTTTTCTCCCGTAAAAGCTGA-3′; Atg7 5′-CCTTCGCGGACCTAAAGAAGT-3′ and 5′-CCCGGATTAGAGGGATGCTC-3′; Nfkbia 5′-TGAAGGACGAGGAGTACGAGC-3′ and 5′-TTCGTGGATGATTGCCAAGTG-3′; β-actin 5′-GACGGCCAGGTCATCACTATTG-3′ and 5′-AGGAAGGCTGGAAAAGAGCC-3′.
Dextran sodium sulphate induced colitis
Acute colitis was induced by adding 4% DSS (TdB Consultancy) to drinking water ad libitum for five consecutive days. Daily disease activity index (DAI) was assessed evaluating weight loss, stool consistency and rectal bleeding according to Supplementary Table 1. A high resolution mouse endoscopic system (Hopkins) was used. Colitis severity was assessed by a semi-quantitative score consisting of two subscores; endoscopic tissue damage (0–3, where 0 is no damage, 1 is lymphoepithelial lesions, 2 is surface mucosal erosion or focal ulceration, and 3 is extensive mucosal damage with expansion into deeper structures of the bowel wall) and inflammatory infiltration (0–3, where 0 is occasional inflammatory cells in the lamina propria, 1 is increased numbers on inflammatory cells in lamina propria, 2 is confluence of inflammatory cells extending into the submucosa, and 3 is transmural extension of the infiltrate).
Xbp1 splicing assay and densitometric quantification
Xbp1 splicing was assessed as described previously4. Briefly, RNA was isolated, reverse transcribed and amplified by RT–PCR with the following primers: Xbp1 sp. forward 5′-ACACGCTTGGGAATGGACAC-3′; Xbp1 sp. reverse 5′-CCATGGGAAGATGTTCTGGG-3′. The PCR product of 171 (unspliced) and 145 (spliced) bp were resolved on a 2% agarose gel. Densitometric analysis for splicing assay and immunoblots was performed with ImageJ.
Statistical significance was calculated as appropriate using an unpaired two-tailed Student’s t-test or a Mann–Whitney U-test and considered significant at P < 0.05. In experiments where more than two groups were compared, Kruskal–Wallis test followed by Mann–Whitney U-test and post-hoc Bonferroni Holm’s correction or one-way ANOVA/Bonferroni was performed. Grubb’s test was used as appropriate to identify outliers. Data were analysed using GraphPad Prism software. Experimental group sizes were based on the goal of achieving desired effect sizes typically of ≤2.0 standard deviations and a power of 0.9 on the assumption of a normal distribution, and therefore typically involved n = 6–10.
We thank L. Glimcher, A. Goldberg, J. Yuan, M. Parkes, A. Franke, H. Tilg, M. Pasparakis, K. Vlantis, A.-H. Lee and C. L. Bevins for discussion of the project, are grateful to J. Gordon, L. Hooper and K. Rajewsky for providing critical reagents, and thank O. Will for initial handling of the Atg16l1 colony and help with DSS colitis. A. Kaser began work for this study at the Department of Internal Medicine II, Innsbruck Medical University, A-6020 Innsbruck, Austria. This work was supported by NIH grants DK044319, DK051362, DK053056, DK088199, the Harvard Digestive Diseases Center (HDDC) (DK0034854) (R.S.B.); the European Research Council under the European Community’s Seventh Framework Programme (FP7/2007-2013)/ERC Grant agreement no. 260961 (A.K.); the National Institute for Health Research Cambridge Biomedical Research Centre (A.K.); the Austrian Science Fund and Ministry of Science P21530-B18 and START Y446-B18 (A.K.); the Addenbrooke’s Charitable Trust (A.K. and L.N.); BMBF NGFN Animal Model grant (P.R.), the DFG Cluster of Excellence Inflammation at Interfaces (S.S. and P.R.); EU SysmedIBD grant (P.R), the Hans-Dietrich Bruhn Memorial Foundation (R.B.); DFG grants RO2994/5-1 (P.R.) and SFB 877 project B9 (P.R. and S.S.); fellowships from Inflammatory Bowel Disease Working Group (M.F.T.), Crohn’s and Colitis Foundation of America (M.B.F.), European Crohn’s and Colitis Organization (T.E.A.), Crohn’s in Childhood Research Association (A.K. and L.N.), National Research Foundation of Korea funded by the Korean government KRF-2008-357-E00022 no. 2011-0009018 (H. -J.K.).
Extended data figures
Three-dimensional linear least square regression analysis for the correlation of enteritis histology score with cell death and age of animals by genotype. Each dot represents a single animal (grey, Wt; yellow, Atg16l1ΔIEC; blue, Xbp1ΔIEC; red, Atg16l1/Xbp1ΔIEC mice) and the plane represents the linear regression for the histological score as a function of age and TUNEL labeling for Atg16l1/Xbp1ΔIEC mice. Note that the severity of inflammation significantly correlates with numbers of TUNEL+ IECs and age in Atg16l1/Xbp1ΔIEC mice (R2=0.602, p=0.016) but not any other genotype (n=6/12/12/12). Regression analysis was performed using the R package lessR (http://cran.r-project.org/web/packages/lessR/index.html), last accessed May 2013.