Association of Crohn’s disease-related chromosome 1q32 with ankylosing spondylitis is independent of bowel symptoms and faecal calprotectin

Introduction

In recent years a new paradigm has developed around ankylosing spondylitis (AS) in which AS is considered part of a spectrum of spondyloarthropathies that also includes Crohn’s disease (CD) (Raychaudhuri & Deodhar, 2014). These two conditions not only demonstrate many common phenotypic characteristics, but also share a common genetic risk. Furthermore, gut inflammation has been identified a frequent sub-clinical manifestation of AS (Mielants et al., 1988, 1991, 1995a, 1995b). Around 10% of AS patients go on to develop true CD, and a similar proportion of CD patients develop AS. Subclinical terminal ileitis is present in around 60% of patients with AS and 10–50% of patients with CD have radiographic evidence of sacroiliitis (Mielants et al., 1988, 1991, 1995a, 1995b; Stebbings et al., 2012).

The Icelandic genealogy database has demonstrated the coinheritance of CD and AS (Thjodleifsson et al., 2007). Not only do first and second degree relatives of a patient with CD have a heightened risk of developing CD themselves, but they are also at increased risk of developing AS, compared to the general population (Thjodleifsson et al., 2007). The reciprocal relationship is seen in families with AS. More recently genome-wide association studies (GWAS) have identified a plethora of risk genes for both conditions (Australo-Anglo-American Spondyloarthritis et al., 2010; Ellinghaus et al., 2016; The Australo-Anglo-American Spondyloarthritis Consortium (TASC) et al., 2011; Jostins et al., 2012). A subsequent meta-analysis of three CD GWAS identified further CD risk genes (Barrett et al., 2008). Danoy et al. (2010) took the 53 risk loci identified by this meta-analysis to be most strongly associated with CD, and genotyped them in a cohort of 2,773 patients with AS and 2,215 controls of Caucasian ancestry. Of these loci, eight were also found to associate strongly with AS (Danoy et al., 2010), indicating these loci are risk factors not only for CD, but also for AS.

Non-invasive markers of bowel inflammation have been piloted in several observational studies in AS. The Dudley Inflammatory Bowel Symptom Questionnaire (DISQ) (Stebbings et al., 2012) has been shown to identify patients with spondyloarthritis who have a high burden of bowel symptoms. Faecal calprotectin is widely used to identify patients with active bowel inflammation in CD and has been shown to be elevated in patients with AS who have histological evidence of bowel inflammation (Cypers et al., 2016).

We hypothesized that shared risk loci may influence the clinical spectrum of bowel inflammation in AS, and hence may associate with an excess of adverse bowel symptoms in patients who have AS but no evidence of overt CD. In order to investigate this hypothesis, the objective of our study was to assess whether the eight CD risk loci most strongly associated with AS were also associated with non-invasive indicators of bowel inflammation in patients with AS.

Materials and Methods

Results

In our dataset, the minor alleles of 1q32 SNP rs11584383 (p = 0.0006, OR = 0.68, 95% CI [0.53–0.84]), and IL23R SNP rs11209026 (p = 0.02, OR = 0.57, 95% CI [0.35–0.93]) both conferred protection against AS (Table 1). However, only the association of 1q32 SNP rs11584383 remained significant after Bonferroni correction for multiple testing (p = 0.006). With stratification by DISQ score, the association of 1q32 SNP rs11584383 remained significant in both AS patients who had minimal bowel symptoms (p = 0.004) and those with clinically significant bowel symptoms (p = 0.02) (Table 2). Whilst stratification by faecal calprotectin showed an apparent difference between genotype groups (Table 2), comparison of ORs demonstrated this difference was not significant.

Discussion

It is well recognized that many AS patients experience clinically significant bowel symptoms (Mielants et al., 1988, 1991, 1995a, 1995b). The aim of our study was to assess whether the eight CD risk loci previously shown to be most strongly associated with both CD and AS (Danoy et al., 2010) were associated with non-invasive markers of gut inflammation in patients with AS, thereby suggesting a potential role for these genes in the aetiology of bowel inflammation. Of the loci identified by Danoy et al. (2010), several have strong functional associations with inflammatory pathways. JAK2 (rs10758669) and STAT3 (rs744166) are associated with mucosal immuno-regulation and more widely in immune-mediated disease (O’Shea & Plenge, 2012). Loci associated with cytokines are also likely to be of relevance to mucosal inflammation, particularly IL12 and IL23; indeed, a combined monoclonal antibody therapy targeting these cytokines has proven to be an effective therapy in both IBD and psoriatic arthritis (Feagan et al., 2016; Lebwohl et al., 2015). Other loci identified have less obvious functional association with IBD.

In our study cohort two CD risk loci showed evidence of association with AS, but after Bonferroni correction for multiple testing only the SNP rs11584383 in the chromosome region 1q32 remained significant. The lack of association of the other confirmed CD loci with AS, may be a type II error. Power calculations using the ORs from Danoy et al. (2010), found our cohort had 79.5% power to detect an association with 1q32 with an effect size of 1.35, but only 30–40% power to detect association of the other SNPs with AS.

The minor allele of 1q32 SNP rs11584383 conferred significant protection against AS. This direction of effect was consistent with the direction previously reported in CD. Stratification of the 1q32 SNP rs11584383 genotypes of AS patients, firstly by DISQ score and then by faecal calprotectin had no effect, indicating the association of rs11584383 with AS is independent of non-invasive markers of bowel inflammation. Indeed it is possible that 1q32 SNP rs11584383 is not a marker of bowel inflammation at all, but rather a marker of axial inflammation. The 1q32 SNP rs11584383 was only identified as a risk gene for CD in a meta-analysis of 3,230 CD patients and 4,829 controls. It is estimated that 10–50% of CD patients have sacroiliitis. In this combined cohort the number of CD patients affected by sacroiliitis may have been sufficient for an association of 1q32 SNP rs11584383 with axial inflammation to be detected, but incorrectly attributed to bowel inflammation.

The 1q32 SNP rs11584383 is an intergenic variant (Barrett et al., 2008; Danoy et al., 2010). Therefore, how it may influence inflammation, whether it is axial or bowel-related, is a matter of conjecture. It is possible that 1q32 SNP rs11584383 is in linkage disequilibrium with a functional SNP in a nearby gene. The mostly likely candidate is Kinesin Family Member 21B. This gene is highly expressed in key immune cells, including CD4+ and CD8+ T cells, natural killer cells and B cells. The 1q32 region has also been implicated in AS in Asian populations. In a genome-wide association study investigating copy number variation (CNV) in Korean patients, a deletion type CNV of the 1q32 was found to confer significant risk of developing AS (Jung et al., 2014).

Another possibility is that the 1q32 SNP (rs11584383) is itself functional, and is involved in the regulation of one or more genes. Data from GWAS have shown disease associated intergenic SNPs are often located in or near regulatory elements, suggesting that these variants may to disrupt or alter the function of key regulatory regions of specific genes (1). To determine whether this is the case for the rs115843383 would be challenging (1). However, a number of recently developed bioinformatic techniques, such as Chromosome Conformation Capture (2), Hi-C (3), and chiA-PET (4); which identify long-range chromatin interactions may help to determine whether 1q32 SNP (rs11584383) indeed has a role in the regulation of one or more genes.

There is a wide body of evidence supporting the concept that defective gut mucosal immunity is a factor in initiating and perpetuating spondyloarthritis. The specific combination of genetic variants seems likely to determine whether an individual develops a more spondyloarthritis or IBD phenotype (Brown, Kenna & Wordsworth, 2016). In our study we were unable to demonstrate an association between shared risk genes and clinical features of bowel inflammation in patients with AS.

Our study has several strengths. The study population met strict classification criteria for AS, came from a predominant Northern European genetic population and was carefully phenotyped. In particular, this is the first study to correlate new non-invasive markers of gut inflammation with genetic risk loci, enabling us to look at a much larger population than invasive endoscopic studies would allow.

Potential weaknesses of our study include the small sample size which may have meant that modest associations were missed. However, the stringent phenotyping of our study participants and our decision to only look at loci which had previously been demonstrated to be strongly associated with AS in ethnically comparable populations likely reduced the impact of our small sample size. We chose not to use the DISQ as a continuous variable and this may have reduced our ability to detect a significant correlation with specific loci. We used categorical values to define clinical case definitions of significant bowel symptoms, based on the original validation of the DISQ in AS. Using DISQ scores rather than case defined grouping would be to assign undue significance to subjective self-reported questionnaire values, when compared with categorical objective genetic testing. Such case definitions based on questionnaire values are routinely used in the study of other human diseases, such as depression. Particularly when these are correlated with genetic testing (Taylor et al., 2014).

We did not stratify for ethnicity, although our population was overwhelmingly of Northern European genetic lineage. More recent GWAS in AS have identified further shared genes with SNPs having the same direction of effect in IBD and AS (International Genetics of Ankylosing Spondylitis Consortium (IGAS), 2013). Only around one quarter of the genetic contribution to AS has so far been identified with most of this (approximately 20%) attributable to HLA-B27 alone, an allele which is not associated with IBD risk (Brown, Kenna & Wordsworth, 2016). Therefore, although we did not demonstrate an association between symptoms and the risk genes we studied, it is possible that other genes alone or in combination may associate with the identified increase in non-invasive markers of bowel involvement.

To our knowledge, this study is the first to stratify confirmed CD risk genotypes with presence of non-invasive makers of bowel inflammation in AS patients, in order to determine whether the association of these genotypes with AS is primarily due to presence of bowel inflammation. Using this approach, we provide preliminary evidence that the association of 1q32 SNP rs11584383 with AS is independent of bowel inflammation as assessed by non-invasive markers. To strengthen and confirm this preliminary finding, further studies which incorporate histology or colonoscopy are needed.

Supplemental Information