Spondyloarthritis, Acute Anterior Uveitis, and Fungi: Updating the Catterall–King Hypothesis

Spondyloarthritis is a common type of arthritis which affects mostly adults. It consists of idiopathic chronic inflammation of the spine, joints, eyes, skin, gut, and prostate. Inflammation is often asymptomatic, especially in the gut and prostate. The HLA-B*27 allele group, which presents intracellular peptides to CD8+ T cells, is by far the strongest risk factor for spondyloarthritis. The precise mechanisms and antigens remain unknown. In 1959, Catterall and King advanced a novel hypothesis explaining the etiology of spondyloarthritis: an as-yet-unrecognized sexually acquired microbe would be causing all spondyloarthritis types, including acute anterior uveitis. Recent studies suggest an unrecognized sexually acquired fungal infection may be involved in prostate cancer and perhaps multiple sclerosis. This warrants reanalyzing the Catterall–King hypothesis based on the current literature. In the last decade, many links between spondyloarthritis and fungal infections have been found. Antibodies against the fungal cell wall component mannan are elevated in spondyloarthritis. Functional polymorphisms in genes regulating the innate immune response against fungi have been associated with spondyloarthritis (CARD9 and IL23R). Psoriasis and inflammatory bowel disease, two common comorbidities of spondyloarthritis, are both strongly associated with fungi. Evidence reviewed here lends credence to the Catterall–King hypothesis and implicates a common fungal etiology in prostate cancer, benign prostatic hyperplasia, multiple sclerosis, psoriasis, inflammatory bowel disease, and spondyloarthritis. However, the evidence available at this time is insufficient to definitely confirm this hypothesis. Future studies investigating the microbiome in relation to these conditions should screen specimens for fungi in addition to bacteria. Future clinical studies of spondyloarthritis should consider antifungals which are effective in psoriasis and multiple sclerosis, such as dimethyl fumarate and nystatin.

Spondyloarthritis is a common type of arthritis which affects mostly adults. It consists of idiopathic chronic inflammation of the spine, joints, eyes, skin, gut, and prostate. Inflammation is often asymptomatic, especially in the gut and prostate. The HLA-B*27 allele group, which presents intracellular peptides to CD8+ T cells, is by far the strongest risk factor for spondyloarthritis. The precise mechanisms and antigens remain unknown. In 1959, Catterall and King advanced a novel hypothesis explaining the etiology of spondyloarthritis: an as-yet-unrecognized sexually acquired microbe would be causing all spondyloarthritis types, including acute anterior uveitis. Recent studies suggest an unrecognized sexually acquired fungal infection may be involved in prostate cancer and perhaps multiple sclerosis. This warrants reanalyzing the Catterall-King hypothesis based on the current literature. In the last decade, many links between spondyloarthritis and fungal infections have been found. Antibodies against the fungal cell wall component mannan are elevated in spondyloarthritis. Functional polymorphisms in genes regulating the innate immune response against fungi have been associated with spondyloarthritis (CARD9 and IL23R). Psoriasis and inflammatory bowel disease, two common comorbidities of spondyloarthritis, are both strongly associated with fungi. Evidence reviewed here lends credence to the Catterall-King hypothesis and implicates a common fungal etiology in prostate cancer, benign prostatic hyperplasia, multiple sclerosis, psoriasis, inflammatory bowel disease, and spondyloarthritis. However, the evidence available at this time is insufficient to definitely confirm this hypothesis. Future studies investigating the microbiome in relation to these conditions should screen specimens for fungi in addition to bacteria. Future clinical studies of spondyloarthritis should consider antifungals which are effective in psoriasis and multiple sclerosis, such as dimethyl fumarate and nystatin.
Keywords: spondyloarthritis, acute anterior uveitis, reactive arthritis, ankylosing spondylitis, fungal infections introdUCtion More than four decades ago, major histocompatibility complex (MHC) class I allele group HLA-B*27 was identified as a potent risk factor for developing spondyloarthritis (1,2). Despite the strength of this association, mechanisms accounting for the link between HLA-B*27 and spondyloarthritis remain a mystery. The bacterial microbiome has attracted a great deal of attention as a possible explanation for this link (3). While we acknowledge the importance of bacteria, we describe a likely role for fungi and suggest that the prostate may be an important locale harboring microbes etiologically related to spondyloarthritis.
In this article, we review the evidence which allowed Catterall and King to postulate their hypothesis, and related studies published in the six intervening decades. We then review the links between fungi and spondyloarthritis, in part to determine if the sexually acquired prostatic infection sought by Catterall and King (28) could be the same as the putative sexually acquired fungal infection suspected of causing prostate cancer (4,15).

postUlation oF tHe Catterall-KinG HypotHesis
Between 1818 and 1948 many case reports were published describing a syndrome affecting men, consisting mainly of the simultaneous inflammation of the urethra, eyes, and joints (29). This syndrome was initially called Reiter's disease. It is now called ReA due to the genital and enteric infections which often immediately precede onset. The classic eye inflammation in ReA is bilateral conjunctivitis on the external surface of the eye and less frequently unilateral AAU (30).

Harkness' review of rea (1949)
After studying 126 cases, Harkness published a comprehensive review of ReA (29). He made several key observations ( Table 1). Harkness extended the definition of ReA to include incomplete cases in which either eye or joint inflammation was absent, and to similar cases in women where the main genital symptom was cervicitis. Harkness noted that generally no microbe could be observed or cultured to account for urethritis or arthritis symptoms, and that the presence of Neisseria gonorrhoeae in a subset of patients-previously thought to be causative-seemed to be coincidental. Some idiopathic urethritis cases in his series were likely caused by infections that were not as well characterized at the time, such as Chlamydia trachomatis (serology was positive in 15% of his cases) and Mycoplasmataceae species. Idiopathic urethritis remains frequent today: its prevalence is about 13% in healthy American men (31), and no infectious agent can be found in about half of American men presenting with urethritis at an STI clinic (32). Similarly, no causative microbe can be found in about half of sexually acquired ReA cases (22).
Harkness advanced a four-part hypothesis to explain the etiology of ReA: (a) a single as-yet-unrecognized infectious agent is necessary in all ReA cases; (b) some relapses are caused by nonsexual "additional factors" that reactivate the infectious agent; (c) the infectious agent cannot be cleared by the immune system in at least some cases and remains in a latent state between attacks; and (d) enteric microbes are not directly causative, but rather unleash the single causative infectious agent.
prostatic inflammation in as and rea (1958) The association between inflammation of the prostate-usually defined as a high concentration of leukocytes in expressed prostatic secretion obtained by prostate massage-and non-genital symptoms was firmly established by Visher in 1929 (41). Visher tested the expressed prostatic secretion of 500 consecutive young men admitted for any reason to the Veterans' Hospital in Waukesha, Wisconsin (41). 87 men (17%) had a high prostate leukocyte concentration (41). Of these 87 men, 36 men had radiographs taken of their spine and sacroiliac joints, of which 20 showed signs of AS (55%) (41). This greatly exceeded the expected rate, though control radiographs were not used.
In 1958, a Swedish group (42) and a British group (43) published independent studies which replicated the association between prostate, spine, and sacroiliac inflammation reported by Visher. Both studies reported prostatic inflammation in a third of controls ( were also included in these two studies, and a strong association with prostatic inflammation was found again (34/40 and 56/59, respectively) (42,43). Many other studies reported high rates of prostatic inflammation in AS and ReA patients (44)(45)(46)(47), though they did not include controls, so their results are difficult to interpret (48).
the Catterall-King Hypothesis (1959) By 1957, it was recognized that ReA often occurred in an incomplete form (29). King suspected that idiopathic AAU, a common comorbidity of ReA and AS, was part of the same syndrome even when it occurred alone. King thus predicted that prostatic inflammation would be associated with isolated AAU (49), as previously reported in small case series (50)(51)(52)(53)(54). To formally demonstrate this association, King asked Catterall to perform a prospective study of all new male uveitis cases from May 1957 to December 1958 at the Institute of Ophthalmology in London (49). Catterall's study confirmed that AAU was strongly associated with prostatic inflammation as compared to controls and to other patients in the series ( Table 2) (27,28). Based on these prostatic inflammation association studies, Catterall and King advanced a three-part hypothesis (26,28): (a) isolated AAU, ReA, and AS are part of the same syndrome; (b) an as-yet-unidentified genital infection is necessary for this syndrome; and (c) this infection is generally sexually acquired. The main marker of this genital infection in men was considered to be prostatic inflammation, as determined by counting the number of leukocytes in expressed prostatic secretion. Unfortunately, this was a low specificity marker because the vast majority of men with prostatic inflammation never develop the syndrome (43). This was thought to be due to rare genetic predispositions which were necessary for inflammation to occur outside the genital area (43). No assay available at the time managed to detect the putative infection (48). Four years later, one of King's coworkers stated (55): "The trouble is that we have not yet identified the organism. " replication by dark and Morton (1968) The association between prostatic inflammation and isolated AAU was not considered biologically plausible, and Catterall's results were deemed unlikely (56,57). All studies associating prostatic inflammation with isolated AAU, ReA, or AS used semiquantitative leukocyte per high-powered field methods, which could have been biased by improper blinding during analysis or sample collection, yielding a spurious association.
To test the association between prostatic inflammation and isolated AAU with the least possible bias and highest accuracy, Dark and Morton used total ejaculate rather than expressed prostatic secretion (eliminating sample collection bias), used cell counting chambers rather than plain microscope slides (a fully quantitative counting method), blinded their analysis (eliminating observer bias), and excluded men with a history of urethritis or ReA (eliminating bias toward sexually acquired cases) (57). Their results were unequivocal: the association between genital inflammation and isolated AAU is real (57).

Hla-b*27
The HLA-B*27 allele group is the strongest genetic risk factor for spondyloarthritis. Its association with AS exceeds odds ratios of 40 in Caucasian populations (58)(59)(60), and it is also strongly associated with ReA (36) and isolated AAU (61). Table 3 lists conditions associated with the HLA-B*27 allele group. Though hypotheses other than antigen presentation have been proposed, the association of ERAP1 polymorphisms with AS in HLA-B*27 carriers strongly suggests major histocompatibility complex (MHC) class I antigen presentation to CD8+ T cells is part of the causative pathway leading to spondyloarthritis (62). The discovery of HLA-B*27 and its association with isolated AAU, ReA, and AS gave much credence to the Catterall-King hypothesis by confirming that these conditions shared an underlying immune mechanism (56). It also provided evidence that the putative etiological infectious agent was intracellular and that a genetically determined immune response caused symptoms (43).
The expression of HLA-B in various tissues and cell types was recently measured by the Human Protein Atlas project (74). HLA-B exhibited highly variable expression in different tissues and surprisingly was either undetected or negligibly expressed in a number of tissues including hepatocytes, myocytes, and soft tissues. HLA-B was highly expressed in the secretory epithelial cells of the prostate (although less in the cervix), glandular cells of the small intestine and colon, as well as the skin. Its expression was not measured in the eyes or joints, though the strong expression of HLA-B*27 in synovial lining cells of AS patients was reported by an older study (75). HLA-B*27 presentation of intracellular antigens is thus consistent with HLA-B expression in all spondyloarthritis sites tested.

tHe UsUal sUspeCts
Enteric microbial epidemics and sexual risk factors have both been convincingly associated with the onset of ReA (76). Proving this link was relatively easy due to the short lag of a few weeks between enteric/genital symptoms and ReA symptoms. The list of confirmed or suspected precipitating infectious agents is very long, especially for enteric ReA (76)(77)(78). The relative incidence of enteric and genital ReA cases is not firmly established. The  (72) Ulcerative colitis largest prospective community-based study, performed in Oslo (Norway) between 1988 and 1990, found that half of the cases were idiopathic and the remaining cases were evenly split between enteric and sexually acquired types (36). A smaller study published 10 years earlier reported a similar distribution (37). Another small study published 10 years later found a similar fraction of idiopathic cases, but a much lower fraction of sexually acquired cases as defined by C. trachomatis seropositivity (38). Though various infections have been considered as possible triggers for isolated AAU and AS, fewer studies have investigated this link as compared to ReA. The onset of AS symptoms is gradual, so retracing infectious triggers which occurred years earlier is difficult.

enteric infections in rea
One of the best controlled studies of enteric ReA was conducted in 1962 on an American Navy ship after a sudden outbreak of Shigella-induced dysentery (40). Because the outbreak timing was circumscribed and the ship was at sea during the following months, all ReA cases coinciding with this outbreak could be traced (40). Out of a population of 1,276 male crew members, 602 proven cases of dysentery occurred, of which nine developed ReA (1.5%) (40). The sequence of symptoms varied between individuals, with six out of nine showing the complete triad and with urethritis preceding other symptoms in most cases (40). This matches ReA presentation described in sexually acquired cases (29,33). Less well-controlled communitybased outbreak studies also support enteric triggers for ReA, and have implicated many enteric infections beyond Shigella (76,77). These infections include not only bacteria such as Salmonella, Campylobacter, Yersinia, Clostridium difficile, and Escherichia coli but also protists such as Giardia lamblia (79). Idiopathic enteric symptoms, where no plausible causative infection can be found, are also common in ReA (79). There is no obvious pattern linking these infections other than enteric inflammation.

Genital infections in rea
Unlike enteric infections, genital infections like Candida albicans and STIs do not occur in large confined epidemics, making associations with ReA more difficult to prove. Prior to large prospective studies, sporadic ReA cases seemed concentrated in men who consulted for urethritis symptoms resembling gonorrhea and shortly thereafter developed inflammation of the eyes and joints (33). Sexually acquired ReA cases were thus initially (wrongly) deemed to be post-gonorrheal polyarthritis (47).
Sporadic ReA cases were mainly seen by urologists and venereologists, who considered urethritis a necessary symptom. In 1933, Harkness realized that a majority of urethral discharge cases were of non-gonococcal origin (80). The search for genital infections which could explain idiopathic urethritis and ReA began in earnest after World War II, and the presence of C. trachomatis (29), Ureaplasma urealyticum (29), Mycoplasma hominis (29), and Trichomonas vaginalis (26) was quickly demonstrated in some cases. However, even after accounting for these new genital infections, most urethritis cases remain unexplained (22,31,32).
C. trachomatis became widely recognized as an STI in the 1970s (81) and is currently the genital infection most convincingly associated with ReA (21,82). The strongest evidence of such an association can be found in case-control studies that measure either serological markers of past exposure to C. trachomatis or the presence of C. trachomatis itself in the genital tract (22,36,37,83). The largest prospective community-based study cultured C. trachomatis in 25 of 112 ReA cases (22%) which occurred in Oslo (Norway) between 1988 and 1990 (36). The largest serological study analyzed 323 ReA cases referred to the Diagnostic-Research Centre of Sexually Transmitted Diseases in Bialystok (Poland) between 2001 and 2012, finding C. trachomatis IgG seropositivity in 70 cases (22%) (83). Both studies considered these rates to be much higher than those of control populations.
Due to the lack of natural experiments such as confined epidemics, it is difficult to demonstrate that C. trachomatis directly causes a subset of ReA cases and is not acting as a surrogate for another genital infection-as occurred with N. gonorrhoeae before it. This was well understood in 1968, when early reports of C. trachomatis in ReA were met with skepticism because C. trachomatis rates in ReA series were similar to those seen in STI clinics (84), suggesting it was acting as a surrogate for sexual activity and was not directly involved (85). Exposure to C. trachomatis cannot be demonstrated in a majority of sexually acquired cases (22,82), let alone in all types of ReA cases combined (36,38,83), and two prospective STI clinic studies have confirmed that C. trachomatis is not associated with ReA in high STI risk populations (86,87).
N. gonorrhoeae and C. trachomatis can be excluded as a cause of most ReA cases, even in STI clinic series (22,29,82,84,86,87). Their association with ReA could be due to these two STIs acting as surrogates for an as-yet-unrecognized sexually acquired infection, as proposed by Catterall and King, or it could be due to a direct etiological role in a minority of sexually acquired ReA cases. N. gonorrhoeae and C. trachomatis are both associated with recent changes in sexual partners (88) and thus to genital exposure to a new set of microbes. In contrast with other STIs, their clearance by the immune system within a few months (89) makes them excellent markers of a recent change in sexual partners. If known STIs are acting as surrogates for an as-yet-unrecognized sexually acquired infection which causes ReA, then the strongest associations would be expected to be found with N. gonorrhoeae and C. trachomatis.

infections in as and aaU
Infectious triggers of AS and AAU have not been as widely studied as in ReA. The studies which have been run are small. Demonstrating associations with infections occurring a month before ReA onset is easy in comparison, whereas it often takes a decade for AS to be recognized (24): triggering infections will be hard to identify because causative microbes may have been cleared and seropositivity may have been lost in the interim. For example, C. trachomatis IgG seropositivity is lost within 6 years when using microimmunofluorescence assays (90).
Antibodies against Klebsiella have been associated with AS in many studies (91), and Klebsiella stool cultures have been associated with disease activity in AS (92) and AAU (93), though Klebsiella cannot be found in most patients (94). Antibodies against peptidoglycan, a common component of bacterial cell walls, have been associated with AS (95) and other spondyloarthritis types (96).
Two small studies found high rates of C. trachomatis in male AS patients, respectively, using cell culture (15/31) (97) and IgG/IgA solid phase enzyme immunoassay (20/32) (98). The selective reporting of C. trachomatis in only a subset of AS patients in the first study and the lack of comparison to a control group in both studies means these results are difficult to interpret. A study of genital infections in women with AS reported a similarly high C. trachomatis rate detected by cell culture (15/32), which was significantly higher than the rate measured in controls (5/33) (99). However, a similar study performed in men with AS using sensitive molecular methods found much lower rates of C. trachomatis (1/32), which were indistinguishable from controls (3/120) (98); note that this study found high rates of C. trachomatis antibodies in these AS patients (20/32) (98), as reported earlier in this paragraph. Finally, two small C. trachomatis serological studies of AS patients of both sexes did not find a statistically significant association, perhaps due to low power (100,101).
Two small serological studies did not find an association between AAU and C. trachomatis when comparing with agematched controls, nor did they find an association with any enteric infection suspected of causing ReA (107,108). Neither study measured ASCAs.
Strong consistent associations between AS/AAU and infections suspected of causing ReA are lacking. The moderate association between STI seropositivity and AS/AAU implied by the Catterall-King hypothesis cannot be confirmed nor refuted by current studies: higher powered studies are warranted.

aGe at onset
Ankylosing spondylitis is more common and has a more predictable course than ReA and AAU, making it easier to study from an epidemiological point of view. ReA and AAU typically occur in attacks lasting a few weeks or months, whereas AS is characterized by long-term inflammation of the sacroiliac joints and spine.
age at onset of as Ankylosing spondylitis risk before puberty is very low (23,24). Its risk of onset increases sharply around the age of 15 years, peaks during young adulthood (age: 18-29 years), and tapers off exponentially over the following 30 years (23,24). This distribution suggests environmental factors necessary for AS (if any) affect mainly adults, not children.
The observation that many HLA-B*27 carriers do not develop AS suggests that additional genetic and environmental factors contribute to the disease. AS monozygotic twin concordance does not reach 100% (109), strongly supporting the existence of environmental factors. Because the monozygotic twin concordance observed in AS exceeds 50% (109), environmental factors required for triggering the disease in genetically susceptible individuals must be ubiquitous (110). This is also consistent with the tapering off of AS onset risk after the age of 30 years, since by that age most individuals would presumably have been exposed to any environmental trigger and have already developed symptoms.
Sexually acquired infections match well with the age at onset of AS because their incidence is very low in children and peaks in young adulthood (25). However, no known sexually acquired infections other than all human papillomavirus (HPV) types combined are present in over half of the population. For STIs which can be cleared by the immune system, peak prevalence occurs between 18 and 24 years of age (89). STI prevalence in Americans of this age group was estimated to be 53.8% for HPV (all types), 3.9% for herpes simplex virus type 2, 1.6% for C. trachomatis, 0.9% for T. vaginalis, and 0.3% for N. gonorrhoeae (89). C. trachomatis and N. gonorrhoeae are both well-established risk factors of ReA; other STIs in this list have not been widely studied in ReA because they do not often cause urethritis. Lifetime risk of sexual exposure to HPV (all types) in men is estimated to be 91% (111), so this infection reaches a high enough fraction of the population to be able to cause AS through a hit-and-run mechanism. In contrast, it appears unlikely that well over half the male population could be exposed to either C. trachomatis or N. gonorrhoeae, though formally demonstrating this is difficult based on currently published studies (89).

age at onset of rea and aaU
Studies of the age at first attack of ReA and AAU are much smaller than those related to AS. AAU is rare in children, and the highest risk of onset occurs in young adults (66,112,113). Sexually acquired ReA generally does not occur in children due to a lack of sexual activity, and its rate peaks in young adults (34,35,114); this distribution is very similar to that of AS. Enteric ReA has a more even age distribution within adults (36) and is also rare in children (115)(116)(117). At first blush, the paucity of enteric ReA cases in children is difficult to explain.
The largest sexually acquired ReA cohort study demonstrated that relapses sometimes coincide with genital or enteric infections but in many cases seem unprovoked (33,39). Most of these relapses were attributed to flare-ups rather than to repeated exposure to triggering microbes (39). Urethritis is often the first symptom in enteric ReA, suggesting enteric infections are acting as one of many possible cofactors that can trigger flare-ups of a latent genital infection. This could explain why a wide range of enteric infections are associated with ReA, and why enteric ReA is rare in children (115)(116)(117) despite ample exposure to enteric microbes (115).

linKs WitH FUnGi
The association between ASCAs and spondyloarthritis (16,(102)(103)(104)(105) warrants analyzing links between fungal infections and conditions listed in Table 3. In a recent study, Maillet et al demonstrated that ASCAs are more strongly associated with peripheral symptoms (uveitis, arthritis, and inflammatory bowel disease) than axial symptoms (spondylitis and sacroiliitis), and with the absence of HLA-B*27 alleles (16). ASCAs are a biomarker of CD4+ T cell recognition of fungal mannoproteins, suggesting the recognition of fungi by CD4+ T cells may play an important role in peripheral spondyloarthritis symptoms, and in cases where antigen presentation to CD8+ T cells is less efficient due to the absence of HLA-B*27 alleles. CD4+ T cell recognition of Malassezia antigens resulting in a Th1 response has been reported in psoriasis (20), though it has not been studied in spondyloarthritis.
Rare homozygous mutations in either CARD9 or in the IL-17/IL-23 pathway cause chronic mucocutaneous candidiasis by impairing the immune response against fungi (118)(119)(120). Similarly, IL-17 inhibitors increase candidiasis risk (121). Genome wide association studies have linked prevalent functional CARD9 and IL23R polymorphisms to AS and associated conditions ( Table 3). These genes are involved in the inflammatory cascade downstream of phagocyte recognition of fungal cell wall components beta-glucan (mainly through Dectin-1) and mannan (mainly through Dectin-2) (110,122,123). These are the two primary antigens leading to phagocytosis of fungi such as Malassezia furfur (124) and Cryptococcus neoformans (125). These links with fungi suggest that the immune response against fungal antigens may be an important component of spondyloarthritis, and interventions targeting cytokines associated with fungal infections-for example, the anti-IL-17A drug secukinumab (126)-may improve symptoms.

Fungi and Uveitis
ASCAs are strongly associated with uveitis in spondyloarthritis patients (16), but antibodies against various bacteria suspected of causing ReA are not (107,108). Similarly, circulating CD4+ T cells in uveitis patients are more sensitive than controls to fungal antigens (C. albicans) after a 24-h antigen exposure period (127), whereas sensitivity to bacterial antigens (Staphylococcus aureus, Clostridium tetani, and Mycobacterium tuberculosis) and protist antigens (Toxoplasma gondii) are similar in both groups (127). Intermediate uveitis is associated with multiple sclerosis onset (128) and with the HLA-DRB1*1501 allele (129), which are both associated with the immune response to fungi (5). A small study reported that oral dimethyl fumarate, a fungicidal compound known to be effective in multiple sclerosis and psoriasis, improved chronic idiopathic uveitis symptoms in four patients (130).
Though most cases of uveitis are considered idiopathic, some cases can be attributed to infections: bacteria (especially M. tuberculosis, C. trachomatis, Treponema pallidum, and Borrelia burgdorferi), viruses (especially herpes viruses), protists (especially T. gondii), and fungi (especially C. albicans) can cause uveitis (131). In a very large study of uveitis etiology, infections were reported in 13% of anterior uveitis, 7% of intermediate uveitis, 40% of posterior uveitis, and 43% of panuveitis cases (131). The abovementioned studies indicate that increased immune sensitivity to fungal antigens may be an important risk factor of idiopathic uveitis, especially in association with spondyloarthritis and multiple sclerosis.

Fungi and inflammatory bowel disease
About 10% of spondyloarthritis patients also have inflammatory bowel disease (Crohn's disease or ulcerative colitis) (16, 24), a rate significantly higher than that of the general population (<1%) (132). Enteric inflammation can be found in about 60% of spondyloarthritis patients, though it is usually subclinical (133). Functional CARD9 and IL23R alleles are also associated with isolated Crohn's disease and ulcerative colitis ( Table 3), suggesting the immune response against fungi may play an important role here as well.
The recognition of fungal antigens is accentuated in human peripheral blood mononuclear cells (PBMC) from Crohn's disease patients as compared to controls (134,135); this was attributed to increased expression of Dectin-1, Dectin-2, and the mannose receptor (135). A similar study focused on bacterial antigens found that CD4+ T cell activation through antigen presentation by PBMC was elevated in inflammatory bowel disease patients as compared to controls when exposed to either C. albicans or nine bacterial species (136). Because CD4+ T cells recognize antigenic peptides which are more species specific than sugars recognized by monocytes, this study should be repeated with a wider panel of medically important fungi including Malassezia and Cryptococcus species. These two genera are taxonomically very distant from Candida and are expected to share few antigenic peptides despite sharing antigenic sugars such as beta-glucan and mannan. CD4+ T cell recognition of Candida and Malassezia antigens can be very different, as demonstrated in psoriasis (20). Crohn's disease is strongly associated with antibodies against conserved fungal antigen sugars (especially mannan, but also beta-glucan and chitin) (70), and recent gut microbiome studies suggest a fungal etiology (17,18,137,138).
Three recent studies using vedolizumab in Crohn's disease and ulcerative colitis patients reported that this intervention triggered peripheral arthritis, sacroiliitis, or psoriasis in about 10% of cases (139)(140)(141); these are the most common symptoms associated with spondyloarthritis ( Table 3). Vedolizumab blocks the α4β7 integrin homing receptor, preventing lymphocytes originating from the gut from returning there during recirculation (141). It is thus plausible that following activation due to antigen exposure in the gut, stray lymphocytes subsequently trigger inflammation upon reaching the peripheral joints, sacroiliac joints, and skin (142,143) by recognizing a similar or identical antigen in these sites (144). Since Crohn's disease is strongly associated with an immune response against fungi, these may be fungal antigens.
The distribution of the age at onset of isolated AS (24) is very similar to that of AS with inflammatory bowel disease (24) and to that of isolated Crohn's disease (145,146): risk is low before puberty, increases sharply around the age of 15 years, peaks during young adulthood (age: 18-29 years), and tapers off over the following 30 years (146). Though Crohn's disease is not as strongly associated with an MHC class I allele as AS, associations with ERAP alleles and interactions between ERAP alleles and MHC class I alleles have also been reported in Crohn's disease (72,73). It is thus plausible that exposure to the same sexually acquired intracellular fungal infection is necessary for both spondyloarthritis and isolated Crohn's disease, and that genetic predisposition determines which symptoms appear following exposure to this environmental factor.

Fungi and psoriasis
Approximately 15-35% of spondyloarthritis patients also have psoriasis (16,24), which is somewhat higher than the rate measured in the general population (<10%) (147). Many studies have associated psoriasis with fungi. Candida colonization of the gut (19,148) and antibodies against C. albicans (67) and M. furfur (67,68) are associated with psoriasis. The application of lysed M. furfur cells to the skin of psoriasis patients induces psoriasis-like lesions much more frequently than in controls (149). Psoriasis patients' PBMC readily produce a Th1 response when incubated with M. furfur antigens (20), whereas the same response does not occur in controls, nor when using C. albicans or Trichophyton rubrum antigens (20). The chitin binding protein YLK-40 is associated with psoriasis and is a marker of disease severity (150); chitin is a highly conserved fungal antigen which is not present in bacteria or in human cells. Finally, many fungicidal compounds have been shown to reduce psoriasis symptoms (151)(152)(153)(154)(155)(156)(157)(158), though such drugs can also induce psoriasis flare-ups (159)-such flare-ups have been tentatively attributed to a Jarisch-Herxheimer reaction (154,156).
Though the incidence of psoriasis peaks at about the same age as AS (24), incidence in children and older adults is much higher in psoriasis (160). The age at onset of isolated AS and AS with psoriasis is nearly identical (24). This suggests that psoriasis may be a heterogeneous disease, of which only a subset of cases share a common etiology with AS (161). A recent study of ERAP alleles in psoriasis supports this hypothesis: ERAP alleles are only associated with psoriasis onset in adolescents and young adults (162), suggesting that the same sexually acquired intracellular fungal infection may be necessary for both spondyloarthritis and this subset of psoriasis cases.
The efficacy of oral nystatin in reducing psoriasis symptoms (154)(155)(156)(157) [oral nystatin is not absorbed and thus limited to killing fungi in the gut (163,164)], the expression of αEβ7 integrin by CD8+ T cells in psoriatic lesions (165) (αEβ7 integrin is a marker of mucosal origin), and the triggering of psoriasis by vedolizumab (140,141) suggest that some psoriasis cases may be caused by lymphocytes originating in the gut. The many links with fungi described earlier and the associations between ERAP alleles and psoriasis (162) suggest that similar intracellular fungal antigens are present in the gut and on the skin, providing a common antigenic target for CD8+ T cells migrating from the gut to the skin. The only fungal genus currently known to be highly prevalent on the human skin is Malassezia (65), present in both healthy skin and psoriatic lesions (166). Several groups have proposed that a loss of immune tolerance to Malassezia may cause some psoriasis cases (149,151,153,167,168). Malassezia are found within skin keratinocytes (169) and have recently been detected in the gut (71,138,170).

Fungi and prostate disease
The presence of idiopathic prostatic inflammation affecting men has been recognized for a very long time, both in isolation and in association with spondyloarthritis (41,43,48). The association between prostatic inflammation and spondyloarthritis, combined with other genital symptoms and sexual risk factors in ReA, suggests that a chronic genital infection may reach the prostate and cause these conditions (26,27). Fungal infections of the prostate are considered rare in immune-competent individuals (171), and few studies have investigated possible fungal etiologies in prostate disease (4). The recent discovery of an abundant fungicidal protein in the prostate (172) indicates that a fungal infection reaches this site, hence necessitating such an antimicrobial protein. This protein is called either prostate secretory protein 94 (PSP94) or betamicroseminoprotein, and is encoded by the MSMB gene. High PSP94 concentration in the prostate protects men from prostate cancer in a dose-dependent manner (173)(174)(175)(176), suggesting prostatic fungi may be an important etiological component of prostate cancer (4). PSP94 concentration interacts with sexual risk factors in affecting prostate cancer risk, suggesting this fungal infection may be sexually acquired (15). Sexual risk factors have been widely studied in prostate cancer and are well recognized (6)(7)(8), though no specific causative infection has been strongly associated with prostate cancer (177). Due to inhibition by calcium ions (172), PSP94's fungicidal activity within the prostate is restricted to the cytosol of secretory epithelial cells, suggesting that targeted fungi must be intracellular and must invade these cells (4).
Interestingly, an aberrant truncated form of PSP94 lacking the fungicidal region is strongly associated with benign prostatic hyperplasia (BPH) (178,179), and prostate epithelial cells in BPH tissue do not stain for PSP94 in histological studies, as opposed to healthy tissue (180,181). This suggests that the cytotoxic T cell immune response against secretory epithelial cells observed in BPH (182) may be triggered by the loss of PSP94-mediated immunity against a ubiquitous (183) intracellular fungal infection. HLA-B*27 would efficiently bind to a peptide from an abundant protein present in or on an intracellular fungus, and then present this peptide to CD8+ T cells on the infected host cell's surface. In this example, a peptide from a fungal cell wall mannoprotein is presented to a CD8+ T cell. Cell wall mannoproteins are good antigen candidates due to their abundance, though presentation of peptides from other fungal proteins is also plausible.
Histological studies have found melanin inclusions within secretory epithelial cells of the prostate targeted by CD8+ T cells (184)(185)(186), though a fungal infection was not considered as a possible explanation. Fungi synthesize melanin as a defense mechanism when exposed to fungicides (187,188), so an intracellular fungus exposed to PSP94 within these cells could explain the presence of melanin.
PSP94 is also present in glandular epithelial cells of the cervix (189) [the main genital inflammation site of ReA in women (22)] and colon (189,190) (one of the main inflammatory bowel disease inflammation sites). It is not present in the skin (190). Its presence in the eyes and joints has not been tested.
PSP94's ortholog in pigs has conserved its fungicidal activity (172) despite rapid evolution (191) resulting in changes to half of the amino acids in the protein (192). Humans and pigs shared an ancestor about 90 million years ago, suggesting that PSP94 is coevolving with fungal infections in both species, and PSP94's fungicidal activity confers a selective advantage to host fitness (193).
animal Models PSP94's conserved fungicidal properties suggest that rodents may also be infected with fungal species targeted by this protein.
In 1956, Pearson noticed that the injection of Freund's adjuvant (lysed Mycobacterium suspended in oil and water) into the footpad of Wistar and Long-Evans rats resulted in symptoms resembling ReA: lymphocyte-mediated inflammation of the joints, genitals, skin, and eyes ensued 10-16 days later (194,195). Lipidic Mycobacterium extracts readily induced arthritis as well, suggesting that the adjuvant's causative antigens were not proteins or peptides (195). It was later shown that alpha beta T cells were the main mediators of inflammation (196). Because alpha beta T cells directly recognize peptides rather than lipids, offending antigens may have already been present at the site of adjuvant administration. Host and commensal microbe antigens were considered the most probable target of these T cells, both in the footpad and in the many other sites of inflammation (195). Such sites appeared to be sterile by cell culture (195), and no self-antigens were strongly associated with this condition (197): the underlying mechanisms in this animal model remain unresolved (3). Likewise, the antigenic target in the widely used HLA-B27/β2m transgenic rat model of spondyloarthritis remains to be identified. It has been demonstrated that development of genital tract inflammation precedes and is required for the development of arthritic symptoms in these animals (198). We therefore hypothesize that a fungal infection at this site could be a contributing agent to the ensuing pathology.
An animal model of arthritis based on the injection of Streptococcus pyogenes cell wall antigens into the knee joint of mice demonstrated that a minute quantity of C. albicans added to this injection skewed the T cell immune response toward Th17, markedly increasing the level of inflammation (201). Interestingly, C. albicans alone, even in large quantities, caused little inflammation (201). This suggests that the immune response to fungal infections is muted unless inflammatory cofactors such as bacterial antigens are also present, giving a plausible explanation for Pearson's animal model of ReA: Mycobacterium antigens may have stimulated an immune response against an elusive fungal infection already present in the rat footpad and other sites of inflammation. For example, Mycobacterium glycolipid trehalose dimycolate upregulates the expression of Mincle (202), and Mincle was recently shown to be a key receptor in an animal model of Mycobacterium-induced uveitis (203). Mincle has a very high affinity to alpha-mannose, an antigen found in Malassezia but not in other medically important fungal species (204). A very recent study reported that alpha-mannan induces uveitis in an animal model with an efficiency similar to Mycobacterium antigens (123). These findings suggest antigenic challenges may increase phagocytic activity directed against Malassezia, leading to activation of alpha beta T cells which recognize Malassezia peptides.
The hypothesis of an adaptive immune response against a fungal infection proposed in the previous paragraph is quite speculative and would be on more solid ground if there existed a disseminated ubiquitous fungal infection in an animal. One such infection was discovered in 2012 in mealworms (205): this fungus-like eukaryote is vertically transmitted to all individuals and is present throughout the mealworm body, though it is concentrated in the genitals and is sexually transmitted as well (205). To the best of our knowledge, no fungal infection with similar properties has been found in a rodent or other mammal. If such a fungal infection existed in humans, it could explain why a heterogeneous set of inflammatory stimuli can cause ReA: each stimulus can trigger the loss of immune tolerance to this disseminated fungal infection by acting as an immunological adjuvant which provokes an adaptive immune response against fungal antigens. This would explain why fungal colonization and fungal antigens are effective at breaking immune tolerance. If this elusive fungal infection was mainly sexually acquired in humans, this would explain why children and young teens are at very low risk of conditions associated with HLA-B*27, and why ReA has manifest sexual risk factors and genital symptoms. Finally, a fungal etiology would explain why functional polymorphisms in genes which are part of the immune response against fungi are associated with conditions listed in Table 3.

ConClUsion
The hypotheses put forward by Harkness in 1949 to explain the etiology of ReA (29), and expanded by Catterall and King in 1959 to include prostatic inflammation, AS, and isolated AAU (26,28), have remained consistent with the scientific literature published since. The evidence reviewed here supports the existence of an as-yet-unrecognized genital infection which is a necessary etiological factor in spondyloarthritis. While the existence of an as-yet-unrecognized genital infection may seem to be far-fetched at first, such a hypothesis has been proposed to explain the sexual risk factors of multiple sclerosis (9,14) and prostate cancer (7,8,13) based on evidence completely unrelated to spondyloarthritis ( Table 4). Recent studies have implicated fungal infections in multiple sclerosis (5,(206)(207)(208) and prostate cancer (4,15,172), and these infections appear to be intracellular in both cases ( Table 4) (4,209,210).
A commonly held view is that ReA-causing inflammation is aseptic, and chronic inflammation in the joints, spine, and eyes is due to autoimmunity (75). Associations with ERAP alleles and the wide variety of MHC class I alleles present in HLA-B*27-negative patients suffering from conditions listed in Table 3 suggest that many epitopes can be involved in the disease process. Thus, the hypothesis of molecular mimicry-induced autoimmunity triggered by an intracellular infection confined to the genitals cannot easily explain non-genital symptoms, unless this infection mimicked a wide range of human antigens (this seems unlikely). This suggests that the putative infection is not confined to the genitals, but rather spreads throughout the body and provides antigens necessary for inflammation in all affected sites, as proposed by Visher (41). This hypothesis is plausible because the only known microbial STI in humans which cannot be cleared by the immune system also spreads from the genitals and causes a wide variety of seemingly unrelated symptoms in infected sites: these symptoms are known as secondary and tertiary syphilis, and are caused by the bacterium T. pallidum (211).
The many links between fungi and conditions listed in Table 3 reviewed here suggest that a ubiquitous intracellular fungus, which usually reaches the genitals of adolescents and young adults through sexual activity, is a necessary etiological factor in these conditions. In a majority of individuals, this infection would remain asymptomatic or subclinical because immune tolerance is maintained. How and where immune tolerance is lost would be mostly genetically determined (43); this would explain why symptoms vary widely between infected individuals and why different conditions in Table 3 are associated with different genes. For example, HLA-B*27 molecules would have a high affinity to an antigenic protein in or on this fungus (Figure 1), frequently leading to a loss of immune tolerance in the sacroiliac joints and spine through the detection of this fungus in these sites. When immune tolerance is marginal, extraneous factors such as unrelated enteric or genital microbes would cause exacerbation of symptoms or flare-ups by acting as immunological adjuvants. These adjuvants are likely bacteria and unrelated fungi which infect or colonize mucosal surfaces, pushing lymphocytes in these sites to recognize the intracellular fungus, clonally expand, recirculate, enter non-mucosal tissue, and cause inflammation by detecting the intracellular fungus in such tissue.
The updated Catterall-King hypothesis proposed here has important implications for future studies. The microbiome of the prostate and cervix should be included in studies of conditions listed in Table 3. The many links with fungal infections described here highlight the limitations of 16S consensus microbiome techniques which can only detect bacteria. Techniques such as deep sequencing of total RNA (212)(213)(214) or DNA (166), while more expensive, should be used because they can detect microbes from the entire tree of life (including fungi). Recent studies have shown that many as-yet-unrecognized infections are present in humans (166,215), so the existence of a novel microbe present in the genitals and in other sites is plausible. For example, molecular studies of oral (216) and genital (217) fungi in healthy individuals found many novel species. It would be interesting to know if fungal colonization of the gut is a risk factor of spondyloarthritis, as demonstrated in Crohn's disease (17,18) and psoriasis (19,148), and if antifungal compounds such as dimethyl fumarate improve the course of spondyloarthritis, as demonstrated in psoriasis (158), psoriatic arthritis (218,219), and multiple sclerosis (220).
The possible association between Malassezia and psoriasis (20,149,151,153,167,168) suggests that particular attention should be given to Malassezia species in spondyloarthritis. Malassezia are ubiquitous facultative intracellular fungi which are difficult to detect. Because of their unique requirement for lipids, they do not grow in commonly used fungal culture media (221). Their DNA is difficult to extract (222), possibly because of their uniquely thick cell wall (221). ITS1 and ITS2, two of the most commonly used primers for fungal consensus PCR, have, respectively, two and one mismatches with Malassezia sequences, leading to underrepresentation in PCR products. Recent reports suggest that Malassezia are not limited to the skin and are present in the mouth (63), nose (223), gut (71,138,170), breast (224), brain (225,226), and lung (227).
Malassezia have the right properties to be able to cause spondyloarthritis: they are ubiquitous, intracellular, present on the skin (65,166), glans penis (64), mouth (63), and gut (71,138,170). Malassezia have beta-endorphin receptors which stimulate the secretion of lipases necessary for their growth (228)(229)(230), and high beta-endorphin levels are found in the prostate (231)(232)(233). Malassezia have been strongly associated with granulomatous Crohn's disease in a pediatric biopsy study (OR = 25.2, 95% CI: 2.45-259, P = 0.0025) (138) and with Dectin-1 SNPs in an inflammatory bowel disease study (18). Reads from unidentified Malasseziales were reported in several recent studies of the skin (166,234) and gut (18), suggesting some species and strains remain to be discovered. To the best of our knowledge, Malassezia's presence in the joints, eyes, and prostate has not been tested.
Because Malassezia are common on the human skin, they can be inadvertently inserted in clinical specimens, causing spurious findings. Much care must be taken to ensure that these fungi are actually present in sampled sites. Though our preliminary microbiome results suggest that they are present in the prostate, excluding contamination beyond all doubt has proven to be challenging. As opposed to most infection types, Malassezia are ubiquitous and are part of the normal human microbiome, making their association with disease very difficult to prove. For example, distinguishing chronic inflammation caused by autoimmunity against human antigens and immunity against ubiquitous commensal microbe antigens on the psoriatic skin is not trivial, as healthy sites and psoriatic lesions have about the same microbiome (166). Microbes present in more sterile sites such as the joints, eyes, or prostate could provide important clues as to which species may be involved in spondyloarthritis.
Since the evidence available at this time is insufficient to definitely confirm the Catterall-King hypothesis, microbiome studies similar to that performed by Kellermayer et al (138) should be performed to test this hypothesis.