Introduction

The potential of the human microbiota to redefine our understanding of spondyloarthritic diseases or spondyloarthritides (SpAs) is discussed within this review. We begin with an overview of the microbiota, moving to potential relationships between the intestinal microbiota and SpA pathogenesis as this has been a major and ongoing focus of research efforts. This discussion includes consideration of different inflammatory pathways that may intersect with an altered gut microbiota, a phenomenon termed “dysbiosis”. We further consider how extraintestinal translocation of intestinal microbes or microbial products may contribute to SpA-related disease, in addition to microbiota-related immune pathways that may link gut and joint pathology. Finally, we review therapeutic manipulation of the microbiota and future research directions for both clinicians and basic scientists. The authors acknowledge the valuable contributions of many to this field, especially those that could not be cited in this review due to space constraints.

The human microbiome

The past decade has seen the advent of high-throughput sequencing approaches to characterize the human microbiota in increasing detail. The human body provides a plethora of habitats for the colonization of trillions of microbes. This is manifest in the high degree of inter-site variation in the community structure of microbiota. For instance, anaerobic Firmicutes/Bacteroidetes spp. dominate the intestine, whereas Actinobacteria and Proteobacteria spp. are found in high abundance on the skin . Indeed, even individual teeth may have a distinct microbial community structure . It is also evident that there is considerable interindividual variation in the microbiota at least at the species level. However, despite this species diversity, it appears that there is an aggregate selection towards species with similar functional gene profile at specific tissue sites .

The microbiota is rapidly acquired post partum and persists in a state of flux for the first few years of life. A myriad of factors influence the early structure of the microbiota including delivery method, parenteral nutrition, early infection, and antibiotic use in infancy. As the first few years of life represent a critical window of microbial exposure and immune education of the host, it is therefore feasible (albeit unproven) that early changes to the microbiota may be relevant to the pathogenesis of SpA-related diseases, despite their adult onset.

By roughly 3 years of age, the microbiota appears to stabilize, more closely resembling that of an adult. Nonetheless, this stability is relative. For instance, consumption of a strongly plant- or animal-based diet by human subjects can lead to changes within the microbiota within 24 h, although reversion to a pre-diet community structure is observed within days of returning to a “normal” dietary intake . Following antibiotic treatment, the microbiota is also rapidly impacted, but returns to a pre-antibiotic state consistent with the idea of ecological memory. The intrinsic stability of the microbiota remains an ongoing area of research, but it is possible that a failure to return to a “normal” gut ecosystem following a disturbance of the microbiota in SpA-susceptible individuals may be functionally relevant to disease pathogenesis.

The definition of what constitutes a “normal” gut microbiome, however, is far from resolved. Recently, many efforts are undertaken to study the composition and function of the human gut microbiome as a new strategy to understand IBD. The European MetaHIT consortium identified three different metagenomes from fecal samples from four European countries . These so-called enterotypes were not nation or gender specific. Each enterotype had a dominant bacterial genus: enterotype 1 was dominated by the genus Bacteroides , enterotype 2 by Prevotella , and enterotype 3 by Ruminococcus . These enterotypes appeared to be functionally different and showed differences in substrate fermentation . However, this concept has not been confirmed by others, such as the National Institutes of Health Human Microbiome Project.

The human microbiome

The past decade has seen the advent of high-throughput sequencing approaches to characterize the human microbiota in increasing detail. The human body provides a plethora of habitats for the colonization of trillions of microbes. This is manifest in the high degree of inter-site variation in the community structure of microbiota. For instance, anaerobic Firmicutes/Bacteroidetes spp. dominate the intestine, whereas Actinobacteria and Proteobacteria spp. are found in high abundance on the skin . Indeed, even individual teeth may have a distinct microbial community structure . It is also evident that there is considerable interindividual variation in the microbiota at least at the species level. However, despite this species diversity, it appears that there is an aggregate selection towards species with similar functional gene profile at specific tissue sites .

The microbiota is rapidly acquired post partum and persists in a state of flux for the first few years of life. A myriad of factors influence the early structure of the microbiota including delivery method, parenteral nutrition, early infection, and antibiotic use in infancy. As the first few years of life represent a critical window of microbial exposure and immune education of the host, it is therefore feasible (albeit unproven) that early changes to the microbiota may be relevant to the pathogenesis of SpA-related diseases, despite their adult onset.

By roughly 3 years of age, the microbiota appears to stabilize, more closely resembling that of an adult. Nonetheless, this stability is relative. For instance, consumption of a strongly plant- or animal-based diet by human subjects can lead to changes within the microbiota within 24 h, although reversion to a pre-diet community structure is observed within days of returning to a “normal” dietary intake . Following antibiotic treatment, the microbiota is also rapidly impacted, but returns to a pre-antibiotic state consistent with the idea of ecological memory. The intrinsic stability of the microbiota remains an ongoing area of research, but it is possible that a failure to return to a “normal” gut ecosystem following a disturbance of the microbiota in SpA-susceptible individuals may be functionally relevant to disease pathogenesis.

The definition of what constitutes a “normal” gut microbiome, however, is far from resolved. Recently, many efforts are undertaken to study the composition and function of the human gut microbiome as a new strategy to understand IBD. The European MetaHIT consortium identified three different metagenomes from fecal samples from four European countries . These so-called enterotypes were not nation or gender specific. Each enterotype had a dominant bacterial genus: enterotype 1 was dominated by the genus Bacteroides , enterotype 2 by Prevotella , and enterotype 3 by Ruminococcus . These enterotypes appeared to be functionally different and showed differences in substrate fermentation . However, this concept has not been confirmed by others, such as the National Institutes of Health Human Microbiome Project.

Links between SpA and bowel disease

SpA refers to a group of clinically and genetically related disorders, whose entities include ankylosing spondylitis (AS), psoriatic arthritis (PsA), juvenile SpA, reactive arthritis (ReA), and inflammatory bowel disease (IBD)-related arthritis. Dependent on the predominant symptoms, SpA can be classified as axial SpA, including both AS and non-radiographical axial SpA (nr-axSpA), or as peripheral SpA. The typical clinical features include inflammatory back pain, sacroiliitis, oligoarticular asymmetric synovitis, enthesitis, and frequent extra-articular symptoms such as anterior uveitis, psoriasis, and gut inflammation.

IBD including ulcerative colitis and Crohn’s colitis is arguably the disease category in which dysbiosis of the microbiome is most strongly implicated in disease causation. Notably, there is significant overlap between IBD and SpA.

In the 1980s, Mielants et al. performed colonoscopies in SpA patients without bowel symptoms, and they found that up to half of these patients showed microscopic signs of bowel inflammation. Two types were distinguished based on morphological characteristics: an acute type resembling infectious enterocolitis and chronic inflammation in which normal mucosal architecture is disturbed. This type of inflammation was in fact indistinguishable from early Crohn’s disease (CD). These pioneering findings have since then been confirmed by others, and include an association between CD-like inflammation and AS, ReA, or PsA.

A similar presence of microscopic gut inflammation was also seen in the GIANT (Ghent Inflammatory Arthritis and spondylitis) cohort, a prospective follow-up study including newly diagnosed patients fulfilling the Assessment of SpondyloArthritis International Society (ASAS) criteria for axial and/or peripheral SpA . Asymptomatic gut inflammation was present in 46.2% of patients including 16.9% acute inflammation and 29.2% chronic inflammation.

Based on this cohort, a predictive model was developed for axial SpA. This model showed that a younger age, progressive disease, male sex, high disease activity as measured by Bath Ankylosing Spondylitis Disease Activity Index (BASDAI), and restricted spinal mobility as measured by the Bath Ankylosing Spondylitis Metrology Index (BASMI) were all independently associated with microscopic gut inflammation . The prevalence of gut involvement in the GIANT cohort was higher in axial than in peripheral SpA, but similar in nr-axSpA and AS.

Long-term follow-up showed that 6.5% of SpA patients develop clinically overt IBD in a duration of 5 years. Patients with chronic inflammation at the initial biopsy, who were also human leukocyte antigen (HLA) B27 negative, had the highest risk (up to 20%) of developing IBD. This suggests that this type of inflammation represents a preclinical form of CD.

Conversely, 30% of patients with IBD develop articular symptoms with a pattern similar to those in SpA including oligoarticular peripheral arthritis, or even AS . Up to 9% of IBD patients have been reported to develop uveitis . The follow-up study by Mielants et al. also showed a clear clinical association between gut and joint: Remission of joint inflammation was associated with disappearance of gut inflammation and vice versa . Importantly, initial chronic gut inflammation also implied a higher risk of evolution to AS. Recent results from the GIANT cohort showed that chronic gut inflammation is linked with more extensive bone marrow edema (BME) of the sacroiliac joints on magnetic resonance imaging (MRI) . This underscores the impact on mucosal inflammation on disease extent and prognosis in SpA patients.

Links between bacteria and SpA

Beyond the links between IBD and SpA, multiple sets of observations implicate bacteria as being causally related to SpA pathogenesis.

Dysbiosis in IBD and SpA

A convincing body of data ports the concept that the gut microbiota is altered in IBD patients, although the identification of dysbiotic changes in SpA patients remains in its infancy. Nonetheless, given the significant disease overlap between these diseases, it is reasonable to hypothesize that they may share common changes to the intestinal microbiota. The most reproducible findings in CD patients are the outgrowth of enterobacteria and a reduction in anaerobes (reviewed by Ref. ). The loss of clostridial species is especially marked, particularly of the clades IV and XIVa during active inflammation. These findings are supported by the reduced abundance of clostridial commensals, Faecalibacterium and Roseburia , in ileal CD patients. Whether these changes are secondary to inflammation remains equivocal, yet a consensus has increasingly emerged that IBD is associated with decreased overall microbial diversity, with a concomitant loss of microbes that may promote gut homeostasis.

In contrast to organisms that elicit a regulatory response, the so-called colitogenic flora capable of inducing inflammation may be expanded. Of note, Prevotellaceae were amongst the expanded Bacteroidetes phyla responsible for increased severity of dextran sodium sulfate (DSS) colitis in NLPR6 inflammasome-deficient mice , and were also found to be linked to periodontal disease . Interestingly, Prevotella was recently also linked to disease in new-onset rheumatoid arthritis, which suggests that, even though gut inflammation is not associated with this disease, the intestinal microbiome could play a pivotal role in inflammatory diseases traditionally viewed as strictly autoimmune in nature . Despite the identification of promising candidates such as adherent/invasive Escherichia coli , Helicobacter spp., Bacteroides vulgatus , Klebsiella pneumoniae , and Proteus mirabilis , which either consistently expand with colitis or induce disease in rodent models; to date, no single pathogenic organism has been identified as inducing the development of IBD.

One of the earliest attempts to identify dysbiotic changes in SpA patients used the sequencing-independent technique of polymerase chain reaction and denaturing gradient gel electrophoresis (DGGE) to compare the gut microbiota of 15 AS patients and controls . This study did not find consistent differences between groups, with high interindividual variation, but it did find that a significantly higher proportion of AS fecal samples had sulfate-reducing bacteria. A similar approach was used in HLA-B27/β2m rats and healthy controls and found a number of unknown or uncultivable species present in the inflamed colon of HLA-B27+ animals . More recently, we have used high-throughput 16s sequencing techniques and identified the significant expansion of Prevotella spp. and an outgrowth of B. vulgatus in the cecum of HLA-B27+ve rats . This was accompanied by a loss of Rikenellaceae spp. Global dysbiotic changes in this system manifest with a loss of members of the dominant Bacteroidetes phylum, as seen in IBD patients, and expansion of less abundant proteobacterial and verrucomicrobial species (Ref. and unpublished observations). Importantly, these changes occurred in the absence of bowel inflammation, which indicates that dysbiosis may also be observed using high-resolution molecular approaches in sufficiently powered studies of SpA patients, irrespective of colitis phenotype.

Pathways of dysbiosis and inflammation

The main hypothesis in CD pathogenesis is an aberrant immune response to intestinal commensal bacteria due to environmental and genetic factors. Very convincing evidence for the importance of microbial–mucosal interactions in IBD was the discovery of nucleotide-binding oligomerization domain-containing protein 2 ( NOD2 )/caspase-associated recruitment domain 15 ( CARD15 ) polymorphisms that predispose to CD. NOD2 is a pattern recognition receptor (PRR) that recognizes bacterial peptidoglycans and, upon activation, leads to stimulation of the nuclear factor kappa B (NFκB) pathway, often associated with inflammation. The spectrum of genetic polymorphisms has increased continuously, expanding to other pathways such as autophagy and endoplasmic reticulum (ER) stress. These have enforced an increasingly important role of perturbed bacterial handling in IBD pathogenesis. Some of the findings have been extended to SpA. Herein, we discuss pathways that may intersect with dysbiosis and disease development.

Disrupted first defenses

The epithelial monolayer of the intestine is a dynamic frontier between gut-resident microbes and the underlying stromal and immune cells of the intestinal lamina propria. Beyond the physical barrier afforded by the intestinal epithelium, many mediators are secreted into the intestinal lumen including mucus, antimicrobial peptides (AMPs), and secretory immunoglobulin A (IgA). Studies of both SpA and CD intestinal tissue indicate that these first defenses may be disturbed relative to healthy controls.

The mucus layer is a major physical and chemical barrier that excludes luminal microbes from the intestinal epithelium, and dysregulated mucus production has been reported in the intestine of both CD and SpA patients . Mice deficient in mucus production also develop a spontaneous colitis phenotype, and altered mucus synthesis accompanies the phenotype of B27+ transgenic rats with an SpA phenotype . Mucus may modulate the host microbiota directly, acting as a nutrient source for some bacteria. Moreover, it has recently been reported that the major mucus protein muc2 is a bioactive mediator and can condition tolerogenic responses in both intestinal epithelial cells (IECs) and dendritic cells . Together, this indicates that disrupted mucus production impacts the microbiota and/or promotes a loss of immune tolerance to gut microbes.

AMPs such as defensins, C-type lectins, and cathelicidins are constitutively expressed in the intestine and act to contain the colonization of commensal microbes. Paneth cells, and to a lesser extent conventional enterocytes, are a significant source of intestinal AMPs whose expression is regulated by the recognition of microbial products by innate immune receptors such as NOD2. Interestingly, while Paneth cell deficiency and NOD2 polymorphisms have been associated with reduced production of defensins and ileal CD, these associations have not been observed in the majority of AS patients . Indeed, ileal biopsies from AS patients exhibit an upregulation of Paneth cell AMP expression, and myeloid cells from SpA-susceptible B27+ rats exhibit a transcriptional profile with markedly increased AMP expression . Further studies should examine whether elevated AMP expression either leads to the loss of protective commensal species in the gut of SpA-susceptible individuals and/or creates an ecological niche exploited by pathogenic/inflammatory gut microbes.

Secretory IgA production is another critical barrier defense mechanism. It is apparent that AS (and IBD patients) have increased levels of anti- Klebsiella and anti-enterobacterial secretory IgA (reviewed in Ref. ), which as mentioned potentially may cross-react with self-antigens in the joint such as HLA B27. Elevated IgA levels have also been reported in the blood and synovial tissue of AS patients . In light of numerous reports of the co-occurrence of AS and selective IgA deficiency (SIgAD), however, it is possible that these antibodies are not an etiopathogenic mechanism at all, but merely indicative of increased bacterial translocation from the gut in SpA patients. Indeed, given the role of sIgA in containing and shaping the composition of the intestinal microbiota, it is tempting to speculate that a defective sIgA response to undefined members of the microbiota could plausibly contribute to AS pathogenesis. Consistent with this hypothesis is the severe AS observed in some SIgAD patients .

Microbial sensing and barrier function

Constitutive sensing of local microbial products plays an intrinsic role in epithelial integrity and the maintenance of barrier function. Microbes stimulate signaling pathways such as NFκB, mitogen-activated protein (MAP) kinase, and inflammasome/caspase signaling, which modulate several aspects of IEC biology and the activation of local immune cells. These include epithelial proliferation and differentiation, expression of adhesion molecules/tight junction proteins, and secretion of chemokines and cytokines that are key local signals in the immune milieu of the gut.

These factors are plausibly dysregulated in SpA patients. For instance, the blunted and fused villi of ileal SpA inflammation are indicative of the perturbed proliferative responses of resident IEC . Cadherin/catenin molecules mediate cell–cell adhesion, and increased E-cadherin expression by IECs has been observed in SpA patients with subclinical bowel inflammation . Moreover, colonic T cells from AS patients exhibit enhanced expression of αEβ7, an E-cadherin ligand that helps mediate T cell–epithelial cell interactions. Conversely, however, the IBD phenotype of mice with disrupted N-cadherin function indicates that a diminished expression of adhesion molecules may be a primary pathogenic event , and these changes may be secondary to ongoing intestinal inflammation.

Further support for the hypothesis that the intestinal epithelium is perturbed in SpA patients is the observation that both SpA patients and their first-degree relatives exhibit increased bowel permeability . Increased intestinal permeability is also observed in B27+ transgenic rats, although no association was found between HLA-B27 status and permeability in a human study . This long-muted “leaky gut” hypothesis hinges on the idea that an intact epithelium limits the exposure of microbial antigen to mucosal immune cells, which may otherwise break oral tolerance. Furthermore, systemic translocation may also permit the accumulation of microbial antigens or adjuvants to sites of peripheral SpA inflammation (discussed further below).

Innate immune cells and IECs recognize invading microorganisms through microbe-associated molecular patterns (MAMPs). These are highly conserved microbial structures that are recognized by PRRs on these cells, such as TLR or receptors of the NOD family. With few exceptions, mice deficient in PRRs including Toll-like receptors (TLRs), NOD-like receptors (NLRs), RIG-I-like receptors (RLRs), and C-type lectin receptors (CLRs) and their downstream signaling components do not exhibit spontaneous bowel inflammation, but exhibit impaired epithelial repair responses and bowel inflammation following acute challenge to the intestinal epithelium. The nature of this acute challenge in IBD or AS patients remains to be determined, but is consistent with a multi-hit model of pathogenesis where environment and host genotype intersect to drive disease. The polymorphisms in NOD2 that predispose to CD are primarily in the leucine-rich repeat region of the protein, a region responsible for bacterial sensing. The NOD2 knockout mouse might thus be a model with features that resemble the effect of NOD2 polymorphisms predisposing to CD. Infection of mice lacking flagellin sensor TLR5 with adherent/invasive E. coli , a gut microbe itself associated with CD-like lesions, notably led to sustained inflammation even after clearance of the offending microbe . Importantly, this emphasizes that a failure to observe differences in the microbiota of SpA patients and controls may not preclude a role for gut microbes in SpA pathogenesis.

CARD9 has been identified as an AS and CD susceptibility gene in multiple studies and mediates signals from dectin-1 and dectin-2 receptors, which recognize β-glucan, a component of fungal and bacterial cell walls . CARD9 also augments muramyl-dipeptide activation of NOD2 (CARD15), in addition to signaling by the DNA damage/viral DNA sensor Rad50 . This underscores the potential of diverse prokaryotic, eukaryotic, and viral members of the microbiota to modulate host responses via CARD9. Interestingly, following acute disruption to the intestinal epithelium by chemical agents or intestinal pathogens, CARD9-deficient mice exhibited diminished epithelial restitution and repair and more severe intestinal inflammation . An inflammatory role for CARD9 signaling, however, is implicated by the spondyloarthritic phenotype of SKG mice following intraperitoneal administration of β-glucan . These dual roles of CARD9 signaling are indicative that inflammatory innate signaling pathways in the periphery may have been largely co-opted in the gut towards homeostatic functions to avoid chronic microbially induced inflammation.

Beyond microbial sensing, certain endogenous tissue-derived stress factors bind to the same PRRs. These factors are called damage-associated molecular patterns (DAMPs) and can be considered as “alarmins”, danger signals that indicate cell stress or damage. The fact that PAMPs and DAMPs share the same receptors for innate immune activation may also provide a link between gut (with microbial-derived PAMPs) and joint (with tissue-derived DAMPs) inflammation. An additional trigger for release of DAMPs in the joint or entheses might be mechanical stress .

An example of DAMPs are calgranulins, a set of cytosolic proteins from the S100 family that are expressed by phagocytic myeloid cells (granulocytes and monocytes). They consist of S100A8, S100A9, and S100A12. S100A8/9 form a complex that is also called calprotectin. As they represent initial cells undergoing stress, and are involved in the early developing inflammatory reaction, they could be promising markers for monitoring local disease activity .

Bacterial handling – autophagy and ER stress

In addition to defects in microbial sensing, diminished bacterial handling by host immune and nonimmune cells may also lead to an altered intestinal microbiota and/or the propagation of inflammatory responses to intestinal microbes.

Autophagy promotes cellular immunity by the degradation of intracellular pathogens and homeostasis by the degradation and recycling of cellular organelles. Autophagy inhibits reactive oxygen species generation, which may directly cause tissue damage itself or trigger inflammatory signaling pathways such as the NLRP3 inflammasome. Autophagy is also known to limit pyroptosis, an inflammatory form of caspase-1-dependent cell death. In chronically inflamed AS intestinal tissue, increased expression of multiple autophagy proteins has been reported, including LC3II, ATG5, and ATG12 . This may reflect that the handling of intestinal microbes by autophagic pathways is dysregulated in SpA patients. Epithelial cell autophagy also limits dissemination of invasive bacteria and commensals to extraintestinal sites . Whereas autophagy has been firmly implicated in the pathogenesis of IBD, including primary defects in bacterial handling leading to ileitis, the CD-associated autophagy genes ATG16L and immunity-related guanosine triphosphatase family M protein ( IRGM ) have not been associated with AS to date.

ER stress responses are another homeostatic pathway that may be disturbed in the inflamed intestine. The findings that polymorphisms in genes associated with or proximal to ER stress pathways are associated with CD and AS, and the increased susceptibility of mice with genetic lesions in the ER stress pathway such as XBP-1 and IRE1 to bowel inflammation indicate that perturbed ER stress responses may also be pertinent to IBD and SpA pathogenesis . Interestingly, extensive studies by Colbert and others have identified that HLA B27 is susceptible to protein misfolding and ER accumulation, serving as a potent trigger of the ER stress response (referenced in Ref. ). In B27-expressing macrophages, microbial signals such as lipopolysaccharide (LPS) augment this response, and elaborate the T helper 17 (Th17)-associated cytokine, IL-23. Thus, innate sensing of gut bacteria or their products may be particularly primed to intersect with ER-stress pathways promoting inflammation. ER-stress-induced IL-1α production has also been shown to promote osteoclast formation, which may be particularly relevant with respect to the joint . Nonetheless, a limited number of studies have not yet reached consensus as to how the ER stress response may be dysregulated in SpA patient samples.