Distinctive gut microbiomes of ankylosing spondylitis and inflammatory bowel disease patients suggest differing roles in pathogenesis and correlate with disease activity

Genetics research has provided important information as to the cause of the development of both ankylosing spondylitis (AS) and inflammatory bowel disease (IBD). There are strong genetic components in the risks of developing both conditions, with high heritability of AS [1, 2] and IBD [3] demonstrated in twin studies. The high disease heritability indicates that the environmental factors involved in the disease are likely to be ubiquitous. AS is strongly associated with HLA-B27, which contributes roughly 20% of the disease’s heritability. The carriage of HLA alleles (including HLA-B27) has been correlated with microbial dysbiosis even in healthy subjects, indicating that the gut microbiome may be a driver of disease, as opposed to being secondary to it [4, 5]. There are at least 163 loci reported so far to be associated with IBD, with a strong overlap between the genes associated with AS and IBD [6].

Up to 70% of AS patients have evidence of terminal ileitis resembling IBD, suggesting gut inflammation is important in disease pathogenesis [7, 8]. Increased gut permeability has been demonstrated in AS patients consistent with the hypothesis that ‘leakiness’ to gut microbes may drive inflammation in the disease [9‐12]. However, unlike ileal IBD, AS-associated terminal ileitis is very rarely trans-mural and does not cause strictures or perforation.

There is strong evidence to suggest that the gut microbiome or gut pathogens trigger the development of IBD [13]. Diversion of the faecal stream dramatically reduces the risk of disease recurrence in the neo-terminal ileum of IBD patients who have undergone ileal resection for severe disease, and antibiotics can similarly reduce the risk of post-operative recurrence of ileal disease [14, 15]. Genetic studies in IBD show a marked over-representation of genes which are also associated with the risk of developing mycobacterial diseases, including leprosy and tuberculosis [16]. Faecal microbiome transplants, correcting IBD-associated dysbiosis, are therapeutically effective in IBD [17].

To date, numerous studies have confirmed the differing microbial profiles of cases versus controls for both AS and IBD [13, 18‐25]. Studies have also shown a relationship between disease activity and microbial profile [22]. Whilst mounting evidence indicates a causal role for gut microbiome in driving AS and IBD, immunological and genetic evidence implies differing modes of pathogenicity. The recent demonstration that treatment with vedolizumab, which is effective for IBD through its effect on reducing lymphocyte migration in the gut mucosa, can induce de novo axial spondyloarthritis, indicates that the mechanisms by which the gut is involved in driving IBD and AS have significant differences [26]. However, direct comparison of the AS and IBD microbiomes has not been possible due to divergence of metagenomic approaches and differences in population demographics. In this study, we collected stool samples, terminal ileal biopsies, colonic biopsies and rectal biopsies from Australian AS patients, IBD patients and healthy controls, as well as stool samples from Italian and Swedish AS patients to (1) test the hypothesis that the gut microbiome in AS and IBD is different, (2) compare the AS microbiome across different populations, and (3) investigate the relationship between AS and IBD disease activity with microbiome composition.

It has been well established that the gut microbiome of AS and IBD patients are significantly different to healthy controls. However, it is less clear whether the AS and IBD microbiomes differ from each other. Utilising a combination of stool samples and biopsies of the terminal ileal, colonic and rectal mucosae collected from 33 AS patients (including 5 with concomitant IBD, ‘AS-IBD’), 59 IBD patients, and 105 controls, significant differences in beta diversity between AS patients and healthy controls, as well as IBD patients and healthy controls were reconfirmed (Fig. 1A, B). Also consistent with previous results [5, 24] the alpha diversity of the microbiome of AS patients was the same as healthy controls, whereas IBD patients exhibited a significant depletion of bacterial diversity (Fig. 1C). AS patients with concomitant IBD (‘AS-IBD’) also exhibited a reduced alpha diversity compared to AS patients, as well a significantly different microbiome composition (beta diversity) compared to both AS patients and IBD patients. A list of the specific bacterial genera associated with AS, AS-IBD and IBD in the most relevant sample types (terminal ileum and stool) is available in Supplementary Table 1 (p < 0.05). The detection of IBD using the findings from the stool microbiome of AS patients was achievable with moderate accuracy (AUC=0.754, 95% CI = 0.588–0.920), similar to the performance of faecal calprotectin (FCP) (AUC=0.752) and better than the DISQ (AUC=0.716). The stool microbiome also accurately distinguished IBD patients from healthy controls (AUC=0.757, 95% CI = 0.643–0.871) (Fig. 1D).

The composition of the AS microbiome detectable in stool was also compared across three geographic regions, Australia, Italy and Sweden (Fig. 2). As expected, significant differences in both alpha and beta diversity were observed, with AS patients from each region demonstrating unique microbiome composition. Furthermore, Swedish AS patients exhibited a less diverse microbiome overall (alpha diversity). A list of the bacterial taxa enriched or depleted in Italian and Swedish samples, relative to Australian samples is available in Supplementary Table 2. Whilst the sample collection, handling and sequencing, and bioinformatic and statistical analysis was consistent for these samples, a vast array of technical (e.g. duration of storage) and environmental variability (e.g. diet) exists between the cohorts, highlighting the challenges of comparing and interpreting AS microbiomes across different studies.

To investigate potential associations between the gut microbiome and disease characteristics of AS, we studied Australian intestinal biopsies from several sites as well as stool samples and excluding patients not undergoing TNFi treatment. Significant differences in beta diversity according to BASDAI (Fig. 3) were observed, however no significant correlation with Dialister was observed, contrary to previous findings [22]. Consistent with previous results, we also noted a significant difference in beta diversity for AS patients with elevated FCP (>100 μg/mg) (Fig. 4). The same effect was also noted for IBD patients in terms of beta diversity, as well as a notable increase in alpha diversity for patients with FCP greater than 100 μg/mg (Fig. 4C), potentially reflecting an increase in the number of inflammatory bacterial taxa.

Previous study of the AS stool microbiome using a GA-map™ Dysbiosis Test in the Swedish cohort revealed an association between dysbiosis and elevated faecal calprotectin levels, but no such association was observed for disease activity assessed by BASDAI [25]. In the current study, when considering changes in the overall microbiome composition, as opposed to association with individual species as conducted in the Swedish study, we note no significant association with FCP (Supplementary Fig. 1). We also noted a significant difference in microbiome composition between individuals with BASDAI 0 to 2.5 versus 5 to 7.5, yet no other BASDAI categories (Supplementary Fig. 1). Consistent with previous reports [25] we note a significant depletion of Faecalibacterium in patients with elevated FCP, but we also note an enrichment of several potentially inflammatory genera. A full list of differentially abundant taxa according to BASDAI and FCP are available in Supplementary Tables 3 and 4, respectively.

Bacterial genera which were significantly enriched or depleted in at least two of the following six categories are summarised in Fig. 5: AS patients, AS-IBD patients, IBD patients, patients with increased BASDAI, and AS or IBD patients with FCP >100 μg/mg. The potentially pathogenic genera Streptococcus and Haeomophilus were noted to be enriched in AS patients, IBD patients, and AS patients with concomitant IBD (AS-IBD). Also of note, Haeomophilus was also found to be enriched in patients with increased BASDAI and IBD patients with elevated FCP. Other than these two genera, no other similarities were noted between diseases. Four genera were noted to be enriched in IBD patients and IBD patients with elevated FCP, and Lachnospira was noted to be depleted in both IBD and IBD patients with elevated FCP.

Using a standardised analysis, this study reconfirmed the occurrence of gut dysbiosis in AS and IBD patients relative to healthy controls, as well as demonstrating a significant difference between the AS and IBD microbiomes. Differences were not only limited to the overall microbial composition (beta diversity), but also the species richness (alpha diversity), with IBD patients exhibiting a significantly less diverse microbiome compared to AS patients (Fig. 1). Clinically evident IBD has been observed in 6–14% [36] of AS patients. In the current study, AS patients with concomitant IBD (AS-IBD) exhibited a further differentiation of the microbiome compared to AS patients, both in terms of the overall composition and by reduced species richness, acknowledging that the sample size for this analysis was small and that these findings need replication. Furthermore, the presence of IBD in AS patients was also able to be identified with moderate diagnostic accuracy (AUC 0.754) in the stools of AS patients. It follows that the differing pathogenicity of AS compared to IBD corresponds with unique microbiome composition, highlighting the potential multifarious function of the gut microbiome.

The AS microbiome has been studied across a variety of geographically and ethnically diverse populations. In the current study, Australian, Swedish and Italian AS patients all exhibited unique microbiomes from each other (Fig. 2). Whilst the array of environmental and technical factors which differentially influence microbiome composition in each region can be numerous and varied, these results highlight the degree of inter-cohort variability occurring in AS studies internationally. Therefore, careful consideration is required when interpreting AS microbiome studies in isolation, and future meta-analysis of multiple cohorts is likely to require standardised recruitment and sampling protocols.

Disease activity, measured by BASDAI and FCP, was found to correlate with microbiome composition for both Australian AS and IBD patients (Figs. 3 and 4), indicating that therapies which serve to normalise the microbiome may be useful in strategy for patient treatment. Association between disease activity assessed by BASDAI and the stool microbiome has also been reported in a study using 16S ribosomal RNA amplicon sequencing of colonic mucosal samples from patients with spondyloarthritis [22], and normalisation of the gut microbiome in patients with AS on TNF-inhibitors and suppressed disease activity was detected in stool samples via shotgun sequencing [24]. IBD patients with elevated FCP exhibited an increase in alpha diversity, potentially highlighting an increase in inflammatory genera. Consistent with previous results, elevated FCP was also correlated with a depletion of Faecalibacterium, a bacterial genus with known anti-inflammatory properties. Interestingly, these genera were not correlated with elevated FCP in AS patients highlighting differences between the microbiomes associated with these diseases (Fig. 5). The correlation between FCP and microbiome composition was not reflected in Swedish AS patients; similarly, the correlation between BASDAI and microbiome composition was notably weaker. The stronger results for the Australian cohort may be due to the use of intestinal biopsies which captured the mucosal microbiome interfacing with the immune system, as opposed to stool samples which retain a relatively low level of mucosal microbiome amongst a background of transient luminal microbiome.

Comparison of the bacterial genera significantly associated with at least two of the following disease categories was performed: AS patients, AS-IBD patients, IBD patients, patients with increased BASDAI, and AS or IBD patients with elevated FCP (>100 μg/mg) (Fig. 5). Consistent with the observed microbiome differences between AS and IBD patients, many of the differentially abundant genera were not consistent between AS and IBD, except for enrichment of two potentially pathogenic genera, Haeomophilus and Streptococcus. Interestingly, enrichment of Haeomophilus was also associated with increased BASDAI and IBD patients with elevated FCP. Case reports have suggested that Haemophilus sp. are associated with reactive arthritis [37‐39]. Similarly, it is well established that Streptococcus genus infection is associated with reactive arthritis [40], although it is not HLA-B27-associated, and it appears not to be a form of spondyloarthritis. It follows that despite the divergence of the AS and IBD microbiomes, a small number of pathogenic genera may be the shared/common factors initially triggering disease. However, further species- and strain-level characterisation, enabled by whole genome sequencing, is required to properly determine whether the enriched taxa are in fact pathogenic. Other than these, several genera were also depleted in IBD patients and in IBD patients with elevated FCP. These genera included Actinomyces, Rothia, and Lachnospira, a genus within each of the Carnobacteriaceae and Ruminococcaceae families.

The use of 16S sequencing in the current study not only limited the taxonomic depth from which conclusions were drawn, but it also prevented a suitable measurement of the functional/metabolic capabilities of the microbiome. Multi-omic, systems biology, approaches (incorporating shotgun metagenomics, transcriptomics, metabolomics and proteomics) are increasingly being used to comprehensively interrogate the microbial features associated with disease. Adopting these approaches in future is essential to move beyond the broad correlations noted in 16S sequencing studies and towards identification of precise microbial features and their causative impacts.

In conclusion, consistent with immunological and genetic evidence, the distinct microbiomes of AS and IBD patients indicate that the gut plays a different role in driving disease. However, enrichment of specific pathogenic genera indicates that the presence of shared/common microbial trigger of disease cannot be discounted. Further research utilising contemporary approaches is required to expand upon these initial observations and to identify precisely the microbial features driving different types of disease.