Genetic variation and expression levels of tight junction genes identifies association between MAGI3 and inflammatory bowel disease

Inflammatory bowel disease (IBD), including Crohn’s disease (CD) and ulcerative colitis (UC), are complex diseases thought to result from loss of homeostasis between the intestinal microbial milieu, the immune system, and a genetic predisposition [1]. Genetic association in IBD has been the focus of much research [2, 3] leading to the recent identification of a large number of susceptibility loci [3‐5]. Still, the hitherto identified genetic associations only account for a small fraction of the heritability of CD and UC [3].

IBD has been linked to increased paracellular permeability [6, 7]. The intestinal permeability is further increased in both CD patients and their relatives, indicating underlying hereditary factors [8]. However, it remains controversial whether this affected permeability is primary, caused by genetic factors, and/or secondary to inflammation or environmental factors. The tight junction (TJ) structure is critical for the permeability properties of the intestine [9] and several studies provide support for a genetic basis in increased permeability [10].

Wapenaar et al. [11] have described an association between celiac disease and single nucleotide polymorphism (SNP) markers in MAGI2 and PARD3. This MAGI2 marker was also associated with UC [11]. Additionally, McGovern et al. [12] identified a link between genetic variation in MAGI2 and both CD and UC.

The aim of this study was to investigate relations between IBD and MAGI2 and PARD3, as well as other TJ genes (F11R, MAGI1, MAGI3, PTEN, and TJP1) encoding products interacting with each other (Fig. 1) [13‐15]. Additionally, PTPN22 was included since it is located in the same genetic region as MAGI3 (Fig. 2) and has previously been described in relation to IBD [5, 16]. To gain a better understanding of the pathogenic mechanism of IBD we further analyzed ileal and colonic gene expression in relation to genotype, inflammatory status, phenotype, and ongoing medical treatment.

IBD has been associated with an increased paracellular permeability of the intestinal epithelium and a dysfunctional barrier is considered important in the pathogenesis of IBD [6, 7]. The TJ structure is critical for the paracellular barrier properties [9]. Although a few IBD susceptibility markers have been linked to barrier related genes, e.g. CLDN2 and MAGI2 [11, 12, 20], the extent to which a dysfunctional barrier is a consequence of genetic factors remains unclear.

The large numbers of susceptibility loci identified in relation to IBD [3‐5] only explain 10.9% and 7.7% of the heritability of CD and UC, respectively [3]. Regional heterogeneity of allele frequencies among different sub-populations has been observed [4, 5, 21], and indicates the importance of identifying associations in homogenous populations to increase study power [21] and to explain the genetic variance in IBD. To further clarify the role of the paracellular barrier in the etiology of IBD, a Swedish population of IBD patients and non-IBD controls was genotyped for several TJ related genes and analyzed for expression in the intestinal mucosa.

One strong association was observed between the T-allele of the MAGI3 SNP marker rs1343126 and IBD, CD, and UC. Further, the C-allele of another MAGI3 SNP marker (rs6689879) was found to be significantly associated with UC. Presence of this marker also resulted in an increased MAGI3 expression in ileal non-IBD mucosa, which might contribute to the restriction of inflammation in UC to the colon. To the best of our knowledge we have for the first time demonstrated an association between MAGI3 and IBD. However, Jostins et al. have previously demonstrated association between this genetic region and CD, but primarily highlighted PTPN22 [5].

In our study, carriage of at least one IBD associated allele at rs1343126 was nominally associated to a decreased colonic expression of PTPN22 (located in the same genetic region as MAGI3) in the non-IBD subgroup. The corresponding relationship was not identified in inflamed or non-inflamed mucosa of IBD patients. In fact, all relationships between gene expression levels and the susceptibility alleles were observed in mucosal biopsies from non-inflamed, non-IBD controls. Even though the expression of several genes were affected in inflamed mucosa, it is unlikely that genotype – gene expression relationships were concealed in the IBD patients by clinical or subclinical inflammation in the non-inflamed biopsies, since these biopsies were obtained from IBD mucosa that were assessed as histopathologically normal. It is, however, possible that the investigated genes were affected by other factors (e.g. microRNA) that persist in non-inflamed IBD-mucosa [22‐24].

Jostins et al. previously identified the C-allele of rs6679677 (MAGI3-PTPN22) as a susceptibility allele of CD, but a protective allele with respect to UC [5], which was not confirmed by our study. Nevertheless, it is possible that our study missed a possible association due to a comparatively small sample size.

Significant associations were observed between an SNP marker in MAGI2 (rs6962966) and IBD, CD, as well as UC. Wapenaar et al. [11] previously described an association between this marker and UC. Their study revealed, however, an association to the opposite allele (G) compared to our study (A). Presently we have no explanation for this discrepancy.

In our association study we applied a significance level of p < 0.05, and the findings were further followed up by analyzing for a genotype-gene expression correlation in a second independent cohort. However, even if we apply a more stringent p-value, by making a Bonferroni adjustment based on the number of analyzed genes (n = 8; p < 0.006), both MAGI3 (rs1343126) and MAGI2 (rs6962966) would still remain significantly associated to IBD.

One limitation of our study was that MAGI1 and MAGI3 were only investigated for SNP markers of exons and exon-intron-boundaries, due to the large size of these genes, and that the selection of SNP markers of MAGI2, MAGI3-PTPN22, PARD3, and TJP1 were based on reports in the literature.

MAGI3 was expressed to a lower level in inflamed, compared to non-inflamed, colonic mucosa from IBD and UC patients. MAGI3 is involved in suppression of the PI3K/Akt pathway due to interaction with PTEN [25], and also in the Wnt/β-catenin pathway [26, 27]. T cells with a high β-catenin level induce T cell activation and intestinal inflammation [28]. Thus, regulation of these pathways might constitute a mechanism by which decreased MAGI3 expression promotes intestinal inflammation. Furthermore, the PTPN22 gene expression was shown to be increased in inflamed colonic IBD mucosa, compared to non-inflamed mucosa, which corroborates previous results reported by Chen et al. [29]. The PTPN22 gene is expressed at a higher level in immune cells compared to non-immune cells [30], and further PTPN22 negatively regulates T cell receptor signaling [31]. In our study gene expression was determined using biopsies, and therefore we are unable to determine the exact cellular origin of the increased expression in inflamed mucosa. It may, however, be due to increased quantities of PTPN22-expressing immune cells during inflammation.

Decreased PTEN expression was observed in inflamed colonic mucosa of IBD, CD, and UC patients, compared to non-inflamed mucosa. Furthermore, there was no overlap between the PTEN expression levels in inflamed colonic mucosa of CD patients, compared to non-inflamed CD mucosa, suggesting PTEN as a marker for inflammatory response in colonic CD mucosa. PTEN is described to play a role in the regulation of intestinal permeability [32, 33], and is also defined as a tumor suppressor molecule inhibiting inflammatory response, cell migration, and proliferation via PI3K/Akt pathway [34, 35]. An increased level of microRNA-21 has been described in colonic biopsies from UC patients, compared to normal colonic mucosa, and expression of microRNA-21 in the Caco-2 cell line, a colonic epithelial cell model, resulted in an increased permeability [36]. The mechanism of microRNA-21 has been investigated in Caco-2 cells, and may involve an effect on paracellular permeability via the PTEN level and Akt phosphorylation [33]. Moreover, activation of PI3K/Akt pathway has been demonstrated to inhibit the differentiation of T cells towards regulatory T cells [37] and previously a decreased PTEN expression level was observed in intestinal mucosal lymphocytes in CD patients, compared to controls [38].

The decreased TJP1 expression level observed in inflamed IBD and UC mucosa, compared to non-inflamed mucosa, is consistent with a loss of ZO-1 protein (encoded by TJP1) in dextran sulfate sodium induced colitis in mice [39]. Poritz et al. further observed a correlation between loss of ZO-1 and increased paracellular permeability [39].

Gene expression was not affected by ongoing medical treatment (thiopurines, aminosalicylates, or glucocorticosteroids), irrespective of the presence of inflammation. Nevertheless, other studies have shown that infliximab, an anti-TNF-antibody, can impact the expression of genes associated with intestinal permeability [40] and mucosal cytokine expression [41]. Among our patients, too few were treated with infliximab to allow meaningful analysis with this drug.