IBD is a Disorder of Defective Autophagy and Innate Immunity

Abstract

Endoplasmic reticulum (ER) stress due to the presence of misfolded or unfolded proteins in the ER invokes a fundamental biological response, termed the Unfolded Protein Response (UPR). The UPR is orchestrated by three main proximal effectors, of which the IRE1/XBP1 pathway represents the evolutionary most conserved one. Due to its high secretory burden, the intestinal epithelium is highly susceptible to perturbations in the UPR as revealed through functional investigations such as in mice that lack Xbp1 expression specifically in the intestinal epithelial cells (IECs). Genetic studies have revealed several ER stress / UPR-associated genes, including XBP1, ORMDL3, AGR2, and MUC19 as risk factors for IBD, and specific functional, rare variants have been described for XBP1. Xbp1^ΔIEC mice spontaneously develop small intestinal enteritis with crypt abscesses reminiscent of human inflammatory bowel disease (IBD), while Agr2^−/− mice develop granulomatous ileocolitis. Mechanistic studies into Xbp1^ΔIEC mice revealed a major effect on Paneth cells associated with alterations in host – microbe interactions in the intestine, and activation of key pro-inflammatory pathways in the host directly associated with unresolved ER stress and hypomorphic Xbp1 function. Remarkably, the intestinal epithelium of IBD patients commonly exhibits evidence of marked ER stress, which cannot easily be attributed to those genetic risk factors alone and indicates that the paradigm of ER stress-related inflammation might be both, a primary originator as well as a potent perpetuator of intestinal inflammation in IBD.

The endoplasmic reticulum (ER) is the site of translation for secretory proteins. From an evolutionary perspective, a delicate mechanism has evolved that allows cells to adapt to different levels of protein translation, which is called the unfolded protein response (UPR) [1]. The UPR is instigated by the presence of mis- or unfolded proteins in the ER, which causes ER stress and leads to the activation of an adaptive programme. The proximal “sensors” of ER stress are the chaperone grp78 (also known as BiP) along with PERK, ATF6, and IRE1, and their down-stream transcriptional effectors ATF4, ATF6, and XBP1s, respectively [1, 2]. This adaptive programme involves a temporary halt in translation (via eIF2alpha), the degradation of mRNA (‘RIDD’, regulated IRE1-dependent mRNA decay), and the transactivation of genes involved in the expansion of the ER, the secretory machinery, protein folding, and protein quality control [1]. When all these mechanisms fail, unresolved ER stress leads to apoptotic death of stressed cells.

The intestinal epithelium, consisting of the four epithelial cell types differentiating from a common intestinal epithelial stem cell, can be considered a highly secretory organ [3]. Specifically, Paneth cells located at the crypt bottom secrete abundant amounts of antimicrobial peptides and play an important role in regulating the composition of the intestinal microbiome and the propensity to withstand intestinal pathogens [4, 5]. Goblet cells produce large amounts of mucins which form the basis of the intestinal mucus layer [6]. Even the absorptive epithelium per se has a considerable secretory burden. In addition to these cell-intrinsic functional aspects, the intestinal environment may pose particular challenges to the cells capacity to fold proteins properly, especially as protein folding is an energy-dependent, active process [2]. Specifically, the anaerobic milieu (in particular in the colon), low glucose, microbial toxins and metabolites, or the presence of reactive oxygen species impose challenges to proper protein folding [7]. Based on these aspects, we hypothesised that the UPR and hence ER stress might affect intestinal epithelial cell function.

Of the three proximal sensors of ER stress mentioned above, the IRE1/XBP1 axis is the evolutionary most conserved one [1]. Remarkably, only the intestinal epithelium expresses a second isoform of IRE1 (i.e. IRE1beta) in addition to ubiquitously expressed IRE1alpha [8]. We therefore were particularly interested in studying the IRE1/XBP1 axis of the UPR in intestinal epithelial cell function. To this end, we generated a mouse model (‘Xbp1^ΔIEC’) with a conditional genetic deletion of a floxed Xbp1 gene specifically in the epithelium of the small and large intestine by using a Villin promoter-driven Cre recombinase transgene [9].

Xbp1^ΔIEC mice indeed exhibited evidence of increased ER stress as detected by increased expression of the chaperone grp78 [9]. Remarkably, these mice spontaneously developed intestinal inflammation in the small intestine that included histological features typically seen in human IBD; these included crypt abscesses, leukocyte infiltration, and ulcerations [9].

Prompted by these results and earlier genetic linkage studies published by several groups that pointed to a locus on chromosome 22 close to the XBP1 gene as a potential risk locus for both forms of inflammatory bowel disease (IBD) [10–12], Crohn’s disease (CD) and ulcerative colitis (UC), we performed a candidate gene study covering the XBP1 locus and its vicinity by 20 HapMap-selected tagging single nucleotide polymorphisms (SNPs) [9]. These studies identified a locus signal in the index cohort, which was also found in two replication cohorts [9]. Of note was the complex genetic architecture of the XBP1 locus which, e.g., exhibited very little linkage disequilibrium (LD) as measured by R², and with the XBP1 gene flanked by recombination hot spots [9]. We therefore hypothesised that the locus association signal might be due to rare/private functional variants and therefore embarked on a deep sequencing effort of the XBP1 gene and its promoter involving a total of ~1000 IBD patients and healthy controls [9]. This revealed 3–4-fold more rare SNPs in IBD patients than in healthy controls, and identified several non-synonymous SNPs (nsSNPs) that were only found in IBD patients [9]. These IBD-associated nsSNPs when engineered into expression vectors of XBP1 exhibited evidence of hypomorphic induction of the UPR [9]. Altogether, these genetic studies revealed an association of the XBP1 locus with IBD and identified rare (private) coding variants of XBP1 that may represent the functional-genetic basis of this signal [9]. Moreover, this genetic insight along with the fact that Xbp1 conditional knock-out mice spontaneously develop intestinal inflammation that is reminiscent of human IBD posited a unique opportunity to investigate the in vivo mechanism how this genetic risk factor (and unresolved ER stress in general) may lead to pathology [9].

In this context, an intriguing observation was that Xbp1^ΔIEC mice lack Paneth cells, which was associated with a virtual absence of transcripts for antimicrobial peptides that are commonly secreted by these cells [9]. Using a model pathogen, Listeria monocytogenes, it was established that Xbp1^ΔIEC mice exhibit a ~2 log increase in the number of L. monocytogenes colonies that can be recovered from faeces 10h after oral infection [9]. Along the same lines, Xbp1 deficiency in the intestinal epithelium also resulted in increased translocation of L. monocytogenes to the liver, implying that absence of Xbp1 in the epithelium leads to an impairment of the intestinal barrier towards invading pathogens [9]. Of note is that a similar phenotype of Paneth cell dysfunction with increased L. monocytogenes translocation had previously been reported for Nod2^−/− mice [13], and decreased expression of the human alpha-defensins HD4 and HD5 has been found in ileal Crohn’s disease, with lowest HD5 levels in patients harbouring the major NOD2^3020insC risk allele [14]. This is important as NOD2 represents the genetic risk factor for CD with the largest effect size [15–17]. Further evidence for impairment of Paneth cell function as a common theme in IBD arose from the recent discovery by Stappenbeck and Virgin and colleagues that hypomorphic Atg16l1 function, another major genetic risk factor for Crohn’s disease [17, 18], results in a major impairment in Paneth cell granule morphology [19, 20]. Remarkably, a similar phenotype of Paneth cells is observed in CD patients homozygous for the ATG16L1^T300A variant [19]. The discovery of ATG16L1 as a genetic risk factor of CD by Schreiber and colleagues [18] has provided the first hint toward autophagy as a pathomechanism of CD, and more recent studies by the Simmons and Philpott groups have reported that NOD2 is a major regulator of autophagy as well [21, 22]. This supports that the observed phenotype in Paneth cells in Nod2^−/− and Atg16l1^HM mice might possibly have a common underlying mechanism.

Paneth cell dysfunction has been reported to lead to alterations in the structural composition of the intestinal microbiome (‘dysbiosis’) [5]. Dysbiosis induced by host genetic alterations may also affect the susceptibility to dextran sodium sulphate (DSS)-induced colitis [23], a barrier-breaching model of intestinal inflammation. Accordingly, conditional knock-out mice that lack either one or both Xbp1 alleles exhibit a gene-dosage-dependent increase in colitis severity in the DSS model [9]. Reduction of the microbial flora via broad-spectrum antibiotics substantially reduce the severity of colitis in this model and abrogated the genotype-related differences [9], indeed implying a prominent role of microbial alterations in the phenotype of Xbp1^ΔIEC mice. However, it is also of note that Paneth cell depletion [24] or inability to convert the pro-form of alpha-defensins to their active form [25], which both leads to microbial dysbiosis [5], does not per se result in spontaneous intestinal inflammation. This is in line with the lack of spontaneous intestinal inflammation in Nod2^−/− and Atg16l1^HM mice [13, 19], and argues for additional mechanisms that drive intestinal inflammation in the context of Paneth cell dysfunction and dysbiosis. In this context, specific aspects of the IRE1/XBP1 branch of the UPR deserve further consideration.

IRE1, an ER transmembrane protein that is activated and phosphorylated upon the presence of misfolded proteins in the ER, activates XBP1 via an unconventional mechanism that involves IRE1’s endoribonuclease activity to excise a 26nt fragment from Xbp1 mRNA (termed ‘XBP1u’ for unspliced) [26–29]. This excision leads to a frame-shift in the translated protein (termed ‘XBP1s’ for spliced), and only this protein encodes the transactivation domain at its C terminus in addition to the N-terminal DNA binding domain that is encoded by both Xbp1u and Xbp1s. In its activated state and in specific contexts of ER stress, active IRE1alpha may also recruit TRAF2 and may thereby directly activate downstream inflammatory pathways [26], like JNK and NFkappaB. Of note is that even a biologically typically minor (i.e. 50%) reduction in Xbp1 transcript levels by genetic deletion of one Xbp1 allele leads to massive overactivation of IRE1 in the small intestinal epithelium as evidenced by the predominant presence of Xbp1s over Xbp1u mRNA [9]. Using an in vitro model system (the murine small intestinal MODE-K cell line) where Xbp1 expression is silenced via a retroviral shRNA expression vector, we noted a marked overactivation of the NFkappaB pathway upon stimulation with TNF (L.N., T.-E.A., manuscript in preparation). This was evident at several levels, including increased phosphorylation of IKK, IkappaBalpha, and NFkappaB p65 proteins, NFkB p65 DNA binding activity and transcription of prototypical NFkB target genes (L.N., T.-E.A., manuscript in preparation). Of note, co-silencing of Ern1 transcription, the gene encoding IRE1alpha, in Xbp1-silenced MODE-K cells resulted in abrogation of NFkappaB hyperactivation upon stimulation with TNF (L.N., T.-E.A., manuscript in preparation). These data suggested that NFkappaB signalling could be an important consequence of unresolved ER stress due to Xbp1 hypofunction. To test this, we treated VillinCre-ER^T2;Xbp1^ΔIEC mice with a selective and irreversible inhibitor of IkappaBalpha phosphorylation, BAY11-7082 [30]. VillinCre-ER^T2;Xbp1^ΔIEC mice express Cre recombinase under the epithelium-specific Villin promoter as a fusion protein with a mutated estrogen receptor, which allows controlling the time-point of Xbp1 deletion [9, 31]. Indeed, BAY11-7082 significantly ameliorated small intestinal inflammation as measured by a histological enteritis score (L.N., T.-E.A., manuscript in preparation). Moreover, BAY11-7082 also prevented Paneth cell depletion in Xbp1-deficient mice (L.N., T.-E.A., manuscript in preparation). These data might indicate that NFkappaB signalling may play an important role in driving enteritis under condition of unresolved ER stress in Xbp1-deficient mice.

Using the VillinCre-ER^T2;Xbp1^ΔIEC mouse model system, we also noted that TNF expression is increased in the intestinal epithelium upon genetic deletion of Xbp1 [9]. TNF is a prototypical inflammatory mediator that is induced by several inflammatory signal transduction pathways (including NFkappaB), and activates a wide variety of downstream inflammatory effector mechanisms [32]. TNF signaling has been identified as a quintessential common mechanism of both forms of IBD based on the substantial efficacy of neutralising monoclonal antibodies in CD and UC [33, 34]. Remarkably, we barely understand at the moment how the genetic underpinning of IBD connects with TNF signalling [7, 17, 35–37]. TNF signals through two distinct receptors, TNFR1 and TNFR2, with the pro-inflammatory signals almost exclusively being relayed via the TNFR1 [38, 39]. Interestingly, TNFR1 can be recruited to active IRE1 and has been implied to mediate ER stress-induced activation of the MAP kinase JNK [40], which is activated in the Xbp1-deficient intestinal epithelium [9]. We were therefore interested whether TNFR1 might have a role in mediating intestinal inflammation in Xbp1^ΔIEC mice. Indeed, Tnfr1^−/−;Xbp1^ΔIEC double-knockout mice were almost completely protected from spontaneous small intestinal enteritis (L.N., T.-E.A., manuscript in preparation). Remarkably, this did not involve prevention of Paneth cell depletion (L.N., T.-E.A., manuscript in preparation). These experiments establish that TNFR1 signalling is required for the development of intestinal inflammation in the context of unresolved ER stress due to Xbp1 deletion in the epithelium.

Altogether, these published and unpublished studies establish that the intestinal epithelium appears particularly susceptible to perturbations in ER stress mechanisms, which may emanate from genetic variants or environmental triggers [3, 41, 42]. Consequently, a hypomorphic ER stress response in the intestinal epithelium can lead to organ-specific inflammation as in inflammatory bowel disease.

However, XBP1 is no longer the only IBD risk gene that maps to ER stress / UPR pathways. Genome-wide association studies (GWAS) have identified ORMDL3 as a genetic risk factor of CD [43] and UC (and asthma [44], another disease of a mucosal barrier). ORMDL3, which localises to the ER [45, 46], appears to be involved in protein folding and regulates the UPR [45, 46] and has a role in sphingolipid metabolism [47]. Another particularly interesting IBD risk gene identified through a candidate gene study is AGR2 [48]. The protein product of this gene is an ER-resident protein disulphide isomerase and as such involved in protein folding. Agr2^−/− mice have recently been reported to spontaneously develop severe granulomatous ileocolitis, closely resembling human CD [49]. Remarkably, Agr2^−/− mice exhibit a marked disruption of Paneth and goblet cell homeostasis [49]. A further potential ER stress-related risk gene is MUC19 [43], encoding a mucin gene (see below).

Irrespective of specific IBD risk genes that are related to the UPR (and which might be infrequent such as XBP1 [9]), it is noteworthy that several studies have reported unresolved ER stress in the intestinal epithelium as a very common phenomenon in IBD patients [9, 50–52]. Haller and colleagues have reported the first observation of increased grp78 expression in the epithelium in inflamed mucosa [50], and a more recent paper by Ogier-Deniz and colleagues showed several-fold increased levels of XBP1s in essentially all UC patients they had studied [52]. In this context it is interesting to note that McGuckin and co-workers have reported two mouse models (called Eeyore and Winnie) generated through forward-genetic approaches that spontaneously develop colitis resembling human UC, and which have been genetically mapped to distinct functional mutations of the Muc2 gene [51], which encodes the predominant mucin. These point-mutations cause aberrant MUC2 biosynthesis which leads to ER stress in goblet cells, which has been suggested by these authors as the underlying mechanism for the development of colitis [51]. Remarkably, goblet cells of UC patients exhibit evidence of MUC2 glycosylation defects [51]. Based on results from Xbp1^ΔIEC mice, goblet cells are susceptible to unresolved ER stress, which leads to their numeric reduction and ultrastructural changes with smaller granule size [9].

A further notable ER stress-related mechanism in IBD recently reported by Darfeuille-Michaud and colleagues is related to the ER stress chaperone Gp96 [53]. Gp96 is overexpressed on the intestinal epithelium of CD patients and regulates the invasion of adherent-invasive E. coli (AIEC) [53]. Increased numbers of mucosa-adherent AIEC are observed in patients with CD, in particular involving the ileal mucosa [54, 55].

In summary, the UPR has emerged as an important regulator of intestinal epithelial cell function, and as such affects the host – microbial mutual relationship. Unresolved ER stress, either due to genetic variants such as XBP1 or environmental factors that impede proper resolution of ER stress, can be an originator of intestinal inflammation in human IBD. Through in vivo mouse models it has been possible to establish a mechanistic framework that connects these genetic risk factors with the evolution of intestinal inflammation. The paradigm of ER stress-related inflammation not only furthers our understanding of the mechanistic basis of IBD, but might also allow for the development of novel therapeutic approaches.