Salmonella as an Inducer of Autoimmunity

MARK J. SOLOSKI¹* AND ELEANOR S. METCALF²

[SECTION EDITORS: FERRIC FANG AND MARTIN KAGNOFF]

Posted April 19, 2007

Division of Rheumatology, Department of Medicine, and The Graduate Program in Immunology, The Johns Hopkins University School of Medicine, Baltimore, MD 21205,¹ and Department of Microbiology and Immunology, F. Edward Hébert School of Medicine, Uniformed Services University of the Health Sciences, Bethesda, MD 20814²

*Corresponding author. Mailing address: Division of Rheumatology, Department of Medicine, and The Graduate Program in Immunology, The Johns Hopkins University School of Medicine, Baltimore, MD 21205. Phone: (410)550-8493, Fax: (410) 550-2072, E-mail: This e-mail address is being protected from spambots. You need JavaScript enabled to view it .

A clear etiological link has been established between infection with several gram-negative enteric pathogens, including Salmonella spp., and the incidence of reactive arthritis (ReA), an autoimmune disease that largely affects the joints (24). ReA is sometimes referred to as Reiter’s syndrome, particularly when accompanied by uveitis and urethritis. At this time, the understanding of the precise mechanisms that underlie the induction of disease remains incomplete but is undergoing a significant paradigm shift. Numerous models have been proposed to explain the link between bacteria and host autoimmune diseases: (i) the presence of persistent bacteria and/or antigen, (ii) the induction of an antibacterial immune response that is cross-reactive with self-antigens, and (iii) the involvement of novel forms of HLA-B27 (99). Clearly, no one model is sufficient to completely explain the incidence of ReA disease and its pathogenesis. It is likely that an understanding of how bacteria induce ReA will require an in-depth understanding of bacterial virulence factors that govern in vivo growth, spread, and susceptibility to host immune elements as well as an understanding of the host immune elements that respond to infection and contribute to autoimmune pathology.

In this chapter, we will review the evidence etiologically linking Salmonella infection with autoimmune disease and address the roles that bacterial and host elements play in controlling disease outcome.

ETIOLOGICAL LINK BETWEEN SALMONELLA INFECTION AND ReA: GUILT BY ASSOCIATION

ReA is an autoimmune disease that largely consists of painful joint inflammation but also can include inflammation of the eye, gastrointestinal tract, and skin. ReA is a member of a broad spectrum of chronic inflammatory disorders termed the seronegative spondyloarthropathies (SNSpAs) that includes ankylosing spondylitis (AS), psoriatic arthritis, and enteropathic arthritis (24). The clinical manifestations of ReA, which is generally described as an episode of aseptic peripheral arthritis, occur within 2 to 4 weeks after the onset of intestinal or genitourinary bacterial infection. The results of infection can range from severe acute disease to colonization. Therefore, a clear etiological link between infection and the development of ReA exists. The idea that SNSpAs occur as a consequence of a recent urogenital or gastrointestinal infection is not new. However, no definitive infectious trigger as yet has been found for AS, despite early studies suggesting that Klebsiella can induce AS (21). In contrast, an infectious origin for other SNSpAs, such as ReA, has been widely documented (98). The most common infectious agents associated with ReA are the gram-negative enteric pathogens Salmonella, Shigella, Yersinia, and Campylobacter, as well as Chlamydia. In addition, ReA has been associated with enteropathogenic Escherichia coli as well as E. coli urinary tract infections (58, 66). A variety of new infectious agents have also been implicated in ReA, but the incidence and prevalence of ReA caused by these organisms (98) have yet to be documented. This chapter will focus on Salmonella-induced ReA.

One of the first reports of an association between Salmonella and ReA followed an outbreak of Salmonella food poisoning in Sweden due to the ingestion of contaminated meat (7). In this study, ReA occurred in 1.4% of hospitalized patients and the organism identified was Salmonella enterica serovar Typhimurium. Subsequently, a number of reports have associated the ingestion of Salmonella-contaminated food and the development of ReA, with an incidence of ReA ranging from 1.2% to as high as 29% of infected individuals (20, 34, 40, 61, 65, 67, 99, 110).

Investigations that link Salmonella infection with ReA are frequently of two different types. Some studies have used patients hospitalized following a Salmonella outbreak, while most others monitor an identified outbreak with questionnaires and clinical exams, irrespective of hospitalization (summarized in reference 34). It is interesting that the latter studies tend to give higher incidences of ReA. Assuming that hospitalized patients have more severe disease, such a finding would suggest that the severity of Salmonella-induced disease is a contributing factor in the incidence of ReA, with milder forms potentially even favoring the development of ReA. However, this conclusion is not uniformly accepted. In one study, new episodes of joint pain were still being reported up to 4 to 5 months after an outbreak of salmonellosis associated with a banquet (20), leading to the suggestion that individuals with more severe or prolonged gastroenteritis are at increased risk for ReA (20, 44). Such a viewpoint is further supported by a recent study of an outbreak of Salmonella serovar Enteritidis-linked ReA in which the duration of diarrhea correlated with ReA incidence (67).

While the studies mentioned above reached different conclusions, these studies as well as the wide range in the incidence of ReA following infection point out that genetic variations in bacterial pathogenicity and/or persistence need to be considered in discussions of ReA immunopathogenesis. Indeed, these variations must play a role since Salmonella serovars Enteritidis and Typhimurium have been implicated in the induction of postinfection ReA whereas the closely related enteric pathogen serovar Typhi has not (20, 25, 40, 98).

As with all diseases, the incidence of disease is determined by many factors, both genetic and environmental. In the case of SNSpAs that appear to have an infectious origin, such as ReA, the dynamics that define host-parasite interactions are critical to our understanding of disease pathogenesis. Thus, it is imperative that both sides of this equation be analyzed and understood if one is to truly understand the pathogenic mechanisms operative in this group of diseases. Both the virulence levels and the natures of the bacterial pathogens, as well as immunogenetic factors of the host, must be understood to make progress in the development of therapeutic approaches. The following sections will consider specific variables that may result in Salmonella persistence and ReA.

ONE SIDE OF THE EQUATION: THE PATHOGEN

Salmonella species, as well as other enteric pathogens associated with postgastroenteritis ReA, are facultative intracellular gram-negative bacteria (76). In humans, Salmonella serovars Typhimurium and Enteritidis cause primarily localized enteritis whereas serovar Typhi causes a systemic syndrome called typhoid fever (41). These pathogens can invade the host either via the ileal portion of the small bowel or via the colon (16, 80). Salmonella-induced gastroenteritis generally results from the ingestion of bacterially contaminated food or water. Invasion of the intestinal follicle-associated epithelium-associated M cells may be an important factor in ReA induction by providing the organisms with access to lymphatic and circulatory drainage. After passing through M cells, organisms are generally phagocytized by resident or recruited subepithelial dendritic cells (DC) or macrophages (16, 80). A second route by which Salmonella can reach the submucosa is via DC capture. Recent studies have suggested that subepithelial DC extend dendrites into the gut lumen, where they capture bacteria and retract the phagocytized organisms into the submucosa (87, 88, 117). Bacterium-containing mucosal DCs or macrophages may induce a localized host response in the Peyer’s patches or disseminate throughout the body via the bloodstream or mesenteric lymph nodes (MLN).

As mentioned earlier, serovar Typhi can disseminate systemically whereas serovar Typhimurium and serovar Enteritidis do not normally pass in significant numbers beyond the MLN. Recent studies have shown that the oral infection of genetically Salmonella-resistant mice with serovar Typhimurium results in the establishment of a persistent infection that lasts up to a year (77, 78). Importantly, the authors of these studies demonstrated that viable serovar Typhimurium resides in macrophages in the MLN, where it is protected from host immune mechanisms (77). The bacterial load in these macrophages was determined to be low, which may be related to the organisms’ ability to survive. This study suggests that the MLN can act as an extra-articular source of Salmonella antigens for extended periods of time. In humans, serovar Typhi can also establish a persistent infection but the host organ of the "carrier" state for serovar Typhi appears to be the gall bladder (3, 76). It should be reiterated here that serovar Typhi has not yet been shown to be an etiologic agent of ReA, but this conclusion may be revisited should a significant cohort of typhoid fever patients be evaluated in the future. It is also possible that Salmonella-infected mucosal DC could migrate to the MLN and become a source of persistent Salmonella antigens; to date, we are unaware that a study of this possibility has been undertaken, although it has recently been shown that gut commensal organisms are phagocytized by mucosal DC and migrate to the MLN (69).

Studies with the murine model have also shown the existence of Salmonella mutants that are able to persist in vivo. The persistence of such mutants may also act to establish an extra-articular niche for Salmonella and/or Salmonella antigens. Salmonella mutants that lack the ability to synthesize aromatic amino acids de novo (e.g., aroA mutants) are minimally virulent and are attenuated in their ability to replicate; however, these organisms are able to persist significantly longer in murine spleens than virulent Salmonella (79, 106, 119). A second class of persistent mutants that maintain their virulence in mice has been identified (17, 107). The mutation responsible for the ability of these organisms to persist in the host was recently identified in the housekeeping gene encoding polynucleotide phosphorylase (17). Apparently, these mutants exhibit increased expression of a subset of virulence genes that act to increase invasion of and replication within intestinal epithelial cells. While it is not clear that these or other mutations favoring persistence occur naturally, it is possible that such a mechanism would permit the establishment of an extra-articular source of antigen that could subsequently seed the joint synovium and activate host cells, thereby leading to the symptoms of ReA.

A related but separate mechanism that may result in the persistence of Salmonella organisms in the host is the establishment of antibiotic-resistant strains of Salmonella in vivo. Recent studies have suggested that the acquisition of antibiotic resistance traits by nonpathogenic commensals in food animals may increase the transfer of resistance traits to human gut flora (6). Subsequently, a report suggested that antibiotic-resistant commensals could transfer antibiotic resistance to human enteric pathogens under conditions present in the ileum (8). As the use of antibiotics in food animals increases, the potential for establishing antibiotic-resistant Salmonella species increases. How might this scenario be relevant to Salmonella-induced ReA? A recent analysis has indicated that antimicrobial-resistant strains of Salmonella are somewhat more virulent than antibiotic-susceptible strains and can cause more prolonged or more severe illness than antimicrobial-susceptible strains (111). Particularly startling are the results from a model developed by Travers and Barza that suggested increased hospitalization for individuals infected with antibiotic-resistant strains of non-typhoid Salmonella is required (111). Taken together with observations by Dworkin et al. and Locht et al. (20, 67) indicating that individuals with more severe gastroenteritis may be at the greatest risk of developing ReA, these data suggest that the incidence of Salmonella-induced ReA may be on the rise.

The previous discussion has assumed that persistent Salmonella exists in an extra-articular niche, whether the MLN, intestine, or spleen. If such is the case and bacterial components are the inducers of the host response in ReA, which bacterial virulence factors are the culprits? To date, little is known about the identity of Salmonella antigens that may play a role in ReA. Investigators at one laboratory have shown the presence of Salmonella lipopolysaccharide (LPS) in the synovial cells of ReA patients (30), and both LPS and nucleic acids have been shown to occur in the joints of individuals with ReA associated with other enteric bacteria (reviewed in reference 98). Nonmethylated CpG oligonucleotides from bacterial DNA have also been shown to induce arthritis when injected intra-articularly (19) or systemically (121). The former study also showed that subthreshold concentrations of LPS and CpG can synergistically induce arthritis in a mouse model (19).

Studies with Yersinia have provided support for the notion that not only bacterial components can enter a joint to induce a host response but viable bacteria can actually enter joints as well (28). However, studies demonstrating the presence of viable Salmonella organisms in joints are more controversial because they fail to distinguish between the transient trafficking of organisms through joints versus the intra-articular persistence of viable organisms (40, 98). Salmonella may be carried to joints by macrophages or DCs arising from extra-articular sites (40). Salmi and Jalkanen showed that multiple human mucosal leukocyte subpopulations from patients with inflammatory bowel disease, a disease accompanied by arthritis that resembles ReA in some respects, bind avidly to venules in the joint synovial membrane (96). Interestingly, each subset of mucosal leukocytes used distinct adhesion molecules to bind to the synovial endothelial cells. These studies not only suggest a mechanism by which Salmonella can reach the synovium and induce a synovial host response but also provide a means for activated T cells to enter the synovium and promote the inflammatory response characteristic of ReA. In addition to these studies, others have shown that synovial fibroblasts can be infected with Salmonella in vitro and support the possibility of intracellular replication (45, 75). These observations suggest that Salmonella organisms initially multiply in vacuoles within fibroblasts but that, over time, the bacteria may disintegrate until only "ghosts" remain. Immunogold electron microscopy using anti-LPS antibodies showed the staining of both bacterial ghosts and amorphous material persisting for weeks after infection. Together, these studies suggest a mechanism for increasing the concentration of bacterial antigens within the synovium as well as allowing Salmonella antigens to persist therein for weeks, thus serving as a nexus for chronic synovitis.

Cano et al. have recently described an intriguing aspect of the ability of serovar Typhimurium to survive in synovial fibroblasts (14). These investigators and others reported that the growth of wild-type serovar Typhimurium in fibroblast cell lines is restricted and results in the establishment of a latent, nongrowing state (14, 70). Further analysis demonstrated that the prolonged intracellular residence of serovar Typhimurium in fibroblasts allows the emergence of small-colony variants (SCV) (15). A genotypic analysis of SCV revealed mutations conferring auxotrophy for aromatic amino acids (mutations in aroD and aroA), consistent with results of earlier studies demonstrating the ability of aroA and purA mutants to persist in vivo (120). Although these studies were conducted with fibroblast cell lines, the results are consistent with the notion that serovar Typhimurium may establish a persistent infection in synovial fibroblasts through the development of SCV. These results also suggest that the failure to recover viable Salmonella by culturing specimens from joints of ReA patients in earlier studies may have been due to alterations in the bacterial phenotype that impaired detection.

Collectively, the studies described above suggest that Salmonella or at least Salmonella antigens can persist in the joints of persons with ReA. Persistent exposure to antigenic stimuli would be likely to induce a host immune response that results in the clinical entity known as Salmonella-induced ReA. The host response component of the dynamic interaction between host and pathogen will now be considered.

THE HOST RESPONSE: LINKAGE TO HLA-B27

As described by others, ReA is a consequence of the interaction of arthritogenic bacteria (e.g., Salmonella species) and a predisposed host. By "predisposed host," we mean someone who is genetically predisposed to generate a host response that is cross-reactive with a self-response. Many studies have analyzed the association of the HLA class I molecule, HLA-B27, with SNSpAs. Whereas B27 has been shown to be present in 90 to 95% of cases of AS (11, 12, 82), the association of the B27 haplotype with other SNSpAs is more tenuous. Although Salmonella-induced ReA has been extensively studied with regard to the host expression of B27, the data are divergent. Previous studies suggested that hospital-based cases showed the highest correlation between ReA and the presence of the HLA-B27 haplotype. In a retrospective hospital study of post-Salmonella-infection ReA, Leirisalo-Repo et al. found that 88% of the patients were B27 positive (62). Hakansson et al. showed that 69% of patients hospitalized with symptoms of ReA were HLA-B27 positive (31). However, in these studies, the high correlation may be due to the fact that these patients were hospitalized and therefore were the most severely afflicted. Clinical studies generally suggest that most people with post-Salmonella-infection ReA do not require hospitalization. In a series of nonhospital cohort studies of Salmonella-induced ReA, the frequency of B27 positivity ranged from 0 to 60% (reviewed in reference 61). Note that in the two studies in which the percentages of B27 positivity were the highest, the total numbers of individuals that had ReA were only four (34) and five (97). These cohort studies also suggested that the risk of developing ReA and HLA-B27 expression may be related to the strain of Salmonella and the age of the individual (40). Insufficient data are currently available to substantiate these observations. Some long-term studies of individuals with Salmonella-induced ReA have established that a correlation exists between B27 positivity and the development of a more severe and prolonged course of ReA (22, 62). However, chronic arthritis has also been observed in B27-negative patients with post-Salmonella-infection ReA (60) and in other studies a correlation between B27 and the severity of disease was not observed (34, 61). Collectively, these data show that the association between B27 and ReA remains controversial. B27 does not appear to predispose to the initiation of Salmonella infections but may be related to the severity and length of the ReA episode. Additional genetic complexities in the relationship between B27 expression and ReA may need to be considered. Tuokko et al. analyzed the HLA class II association with ReA. They found a positive correlation between the expression of B27 and DRB1*0408 and between B27 and DQG1*0301 in ReA patients but not controls (113). These findings indicate that susceptibility to ReA is a complex genetic trait and that additional genetic markers remain to be identified.

HLA-B27 AND BACTERIAL INVASION

As another approach to understanding the role of HLA-B27 in post-Salmonella infections and ReA, investigators at several laboratories have analyzed the capacity of Salmonella to invade nonphagocytic and phagocytic cells that express HLA-B27. As those of the clinical studies, the results of these in vitro analyses are controversial. Ortiz-Alvarez and colleagues investigated the effect of B27 expression on the invasion of B27-transfected HeLa epithelial cells by serovar Typhimurium (81). Their results showed that the percentage of internalized Salmonella organisms is not different in cells that express B27 and those that do not. In contrast, Saarinen et al. determined that the expression of B27 by Henle epithelial cells enhances the invasiveness of Salmonella (93). Ikawa et al., like Ortiz-Alvarez et al., showed that the expression of B27 on HeLa cells does not influence the invasiveness of serovar Typhimurium, but the presence of B27 on HeLa cells did allow the expression of c-fos and subsequently the transcription factor AP-1 after invasion by serovar Typhimurium, whereas transfection with HLA-A1 did not (43). Studies with other nonphagocytic cells, human fibroblasts, also showed no influence of the B27 haplotype on the invasiveness and survival of serovar Enteritidis (42). Laitio et al., using murine L cells, showed that serovar Enteritidis organisms invade B27- and B7-transfected L cells to a similar extent but that the bacterial load in B27-positive cells remains elevated in comparison to that in B7-transfected cells (59). In contrast, Kapasi and Inman showed that the susceptibility to invasion of B27-transfected L cells decreased relative to that of cells transfected with other, non-B27 alleles (47). Collectively, these studies do not definitively delineate the role of HLA-B27 in nonphagocytic host cells. Given the variability of the results to date, it is not possible to conclude whether HLA-B27 plays a major role in the initial stages of Salmonella-induced gastroenteritis, i.e., invasion and traversal of the intestinal epithelium.

The role of HLA-B27 in the interactions of Salmonella with mononuclear cells is incompletely understood, and only a few studies by a small number of investigators have been conducted to date. Laitio and colleagues used a transfected human macrophage-like cell line, U937, to show that the expression of B27 does not alter the ability of these cells to phagocytize serovar Enteritidis (59). However, the expression of B27 did alter the capacity of U937 cells to eliminate ingested bacteria. In a subsequent report, this group confirmed the results of their earlier studies and also showed that B27-transfected U937 cells produce a distinct cytokine profile after exposure to serovar Enteritidis, although the decreased capacity of B27-transfected U937 cells to kill Salmonella could not be attributed to observed differences in interleukin-10 (IL-10) and tumor necrosis factor alpha (TNF-α) production (23). Ortiz-Alvarez et al. found no differences in the phagocytic capacities of B27-transfected and nontransfected U937 or peripheral blood mononuclear cells (81). Together, the results of these studies suggest that B27 expression on macrophages may enhance the intracellular survival of Salmonella. If this is the case, then B27-positive macrophages may serve as an extra-articular source of Salmonella antigens for extended periods of time. In this regard, it is interesting that serovar Typhimurium can persist within macrophages of MLN in 129Sv mice for up to a year (77). Although the relationship between murine H-2K and human HLA-B27 is unknown, it would be worthwhile to evaluate the persistence of serovar Typhimurium MLN macrophages in other strains of mice that are genetically resistant to Salmonella.

In light of studies showing that LPS is one of the Salmonella antigens demonstrated to be present in the joints of ReA patients (40, 49, 98), Penttinen et al. studied the effects of treating B27-transfected U937 cells with Salmonella-derived LPS (83). Their results showed that LPS-stimulated transfected U937 cells exhibit pronounced and sustained increases in NF-κβ activation compared to A2-transfected cells. Investigators have also shown that B27-transfected U937 cells produce significantly more IL-10 and TNF-α than those transfected with HLA-A2 (23). These cytokine measurements are consistent with the results of a previous study demonstrating that peripheral blood mononuclear cells from patients with ReA produce elevated levels of IL-10 and TNF-α (13, 54). Synovial fluid mononuclear cells in patients with ReA were also found to display increased production of IL-10 and gamma interferon (101). Interestingly, the presence of IL-10G microsatellites characterize promoter haplotypes that are associated with protection against the development of ReA but no such association for AS has been observed (29, 46).

THE ADAPTIVE IMMUNE RESPONSE TO SALMONELLA INFECTION

The clear association between ReA and infection with Salmonella or other gram-negative enteric pathogens has led to the suggestion that the adaptive immune response to infection has an autoimmune component. Unfortunately, relatively little is known about the human immune response to infection with serovar Typhimurium. Studies on the nonarthritogenic serovar Typhi have found that oral infection with attenuated serovar Typhi elicits serovar Typhi-reactive CD4⁺ and CD8⁺ T cells (68, 94, 95, 108). Interestingly, the CD8⁺ T cells can recognize serovar Typhi-infected autologous targets and appear to be the predominant T-cell source of gamma interferon in antigen-driven ex vivo responses (94).

In contrast to humans, murine models have yielded information on the adaptive immune response to serovar Typhimurium. Studies using these models have demonstrated that both CD4⁺- and CD8⁺-T-cell responses are needed for full protection against lethal infection with serovar Typhimurium (39, 63, 64, 73). In the murine model system, the CD4⁺ T cells appear to play a large role since CD4-deficient mice are several orders of magnitude more susceptible than CD8-deficient hosts (39, 64). Interestingly, characterization of the T-cell response to serovar Typhimurium infection reveals the presence of a CD8⁺-T-cell population that recognizes a nonamer peptide derived from the Salmonella GroEL (Hsp60) protein (63). These CD8⁺ T cells also recognize an analogous peptide, identical at seven of nine positions, derived from the mammalian Hsp60 counterpart of GroEL. Furthermore, these serovar Typhimurium-induced Hsp60-reactive CD8⁺ cytotoxic T lymphocytes (CTLs) recognize uninfected targets that have been placed under stress conditions. Collectively, these results imply that following serovar Typhimurium infection a subset of bacterial antigen-reactive T cells develops with the potential to recognize and damage uninfected cells undergoing physiological stress. Such cells may contribute to an autoimmune disease process (102). Interestingly, the antigen-presenting structure identified in the presentation of GroEL (Hsp60) peptides to CD8⁺ CTLs is the murine class Ib molecule Qa-1. Human HLA-E is the functional counterpart of Qa-1, raising the intriguing possibility that analogous bacterial antigen-induced T cells that cross-recognize self-epitopes develop in the human setting.

IMMUNE ACTIVATION AND ReA

ReA is an immunopathologic disease, and phagocytes and lymphocytes from ReA patients have been reported to exhibit an activated phenotype compared with controls. For example, the increased expression of the monocyte markers CD11b, CD11c, and CD14 in ReA patients has been reported (53). However, in another study no change over controls was seen (57). When lymphocytes were examined, Konttinen et al. showed that lymphocytes in ReA patients display an activated phenotype as indicated by the increased expression of IL-2R, the transferrin receptor, and the gp40/80 glycoprotein 4F2 on synovial mononuclear cells (56).

If there is in fact an etiological link between infection and ReA, bacterial antigen-reactive T cells would be predicted to be present at the sites of disease. Indeed, this is the case, as CD4⁺ and CD8⁺ T cells directed against bacterial antigens have been identified in the synovial fluid and synovium of patients with ReA. Although serovar Typhimurium is associated with a significant number of ReA cases, there have been few studies investigating the involvement of bacterial antigen-reactive T cells in Salmonella-induced disease. In one study, CD4⁺ T cells recognizing an unknown antigen derived from Salmonella or Campylobacter were cloned from the synovial fluid of a single patient with Salmonella-induced ReA (37). In contrast, there have been several reports of bacterial antigen-reactive CD4⁺ and CD8⁺ T cells being recovered from patients with Yersinia-induced ReA (5, 36, 74, 84, 118). Strong CD4⁺-T-cell responses against peptides derived from Hsp60 and the β-subunit of the urease protein have been identified (74). Interestingly, cross-reactivity with self-antigens was demonstrated in at least one report, raising the question of whether such bacterial antigen-induced CD4⁺ T cells are relevant to ReA (74).

Several studies have also shown that bacterial antigen-reactive CD8⁺ T cells can be derived from ReA patients. Given the proposed role (discussed above) of HLA-B27, it was predicted that HLA-B27-restricted CD8⁺ CTLs would be recovered from ReA patients. In fact, this has proven to be the case. In an initial report, B27-restricted CD8⁺ CTLs were recovered from a patient with Salmonella-induced ReA (38). These CTLs recognized Salmonella- and Yersinia-infected targets and displayed autoreactivity. CTLs with similar properties have been found in patients with Yersinia-induced ReA (35). In another study, HLA-B27-restricted CTLs from patients with Yersinia-induced ReA (114) were found to recognize a peptide derived from Yersinia Hsp60 (amino acids 321 to 329). However, it was not reported whether these CTLs exhibited self-recognition and the identified Yersinia sequence does not have a homologous counterpart in human Hsp60. Nevertheless, the finding that Hsp60-reactive T cells can be recovered from ReA patients is highly significant since there have been numerous reports documenting Hsp60 to be a target for autoreactive T cells (103, 122).

POSSIBLE ROLE OF HLA-B27 AS AN ANTIGEN PRESENTATION STRUCTURE

A role for HLA-B27 in the pathogenesis of bacterium-induced arthritis such as ReA is therefore supported not only by genetic linkage data but also by the identification of HLA-B27-restricted CTLs that recognize bacterial and/or self-antigens obtained from patients with ReA associated with bacterial infections. Collectively, these observations support a model in which bacterium-induced effector T cells recognize a cross-reactive arthritogenic self-peptide presented by HLA-B27 on uninfected targets (99). In this manner, HLA-B27 may function as an antigen presentation structure and target for autoimmune effector cells.

In an effort to identify peptides bound to HLA-B27 that are relevant to ReA, the peptide repertoire presented by the B27 molecule has been analyzed. Ramos and colleagues eluted peptides from B27 of Salmonella-infected and uninfected cells and found the chromatographic profiles to be minimally different (85). In separate studies, Ringrose et al. concluded that the total numbers of peptide species eluted from B27 molecules of infected and uninfected cells were not significantly different (89, 90, 91). However, the latter investigators noted an increase in the frequency of specific self-peptides in Salmonella-infected B27-positive cells (90). In these studies, the investigators used a B-cell line (C1R-B27) transfected with HLA-B*2705. It was concluded that a dominant arthritogenic peptide could not be identified by this approach. However, it is possible that the B27-bound peptides in Salmonella-infected and uninfected cells would have been different if classic antigen-presenting cells had been utilized for these studies.

A central role for HLA-B27 in ReA was solidified through the generation of HLA-B27 transgenic rats and mice. In 1990, Hammer et al. reported that rats expressing the HLA-B27 molecule through a transgene develop a spontaneous inflammatory disease that shares many features with the spondyloarthropathies, including ReA (32). Although spontaneous, the disease was dependent on the gut flora, especially species of Bacteroides (86). Interestingly, the severity of disease correlated with the HLA-B27 transgene copy number (109). A similar disease could be induced in HLA-B27-positive mice deficient in endogenous β₂ microglobulin (β₂m), a subunit of class I molecules required for normal major histocompatibility class I (and HLA-B27) surface expression (50, 52). The involvement of a bacterial trigger in this model was also suggested, since mice raised under pathogen-free conditions did not develop disease. Collectively, the results from these animal models support a central role for HLA-27 in the pathogenesis of the spondyloarthropathies.

Recently, the role of HLA-B27 has been reconsidered and alternative models have been offered. The first indication that HLA-B27 may have functions in addition to a role as a "conventional" antigen presentation structure of CD8⁺ T cells was the dependency on high gene copy numbers in the B27 transgenic rat model (109). This finding suggested that the abnormal expression of HLA-B27 is related to disease. This notion was further supported by the finding that mice coexpressing HLA-B27 and human β₂m instead of murine β₂m also develop disease while expressing significant levels of a cell surface HLA-B27 heavy chain lacking β₂m (50). Therefore, novel cell surface forms of HLA-B27 may themselves represent targets, an idea further supported by the discoveries that the induction of murine disease does not require the Tap-dependent loading of peptides and that CD8⁺ T cells are not required in the rat model of ReA (51, 71).

Studies on the biosynthesis and expression of HLA-B27 have revealed that HLA-B27 can be expressed as a free heavy chain devoid of β₂m, which can exist as homodimers and oligomers. Multimeric forms of HLA-B27 are also expressed in the rat and murine HLA-B27 transgenic models, suggesting that these multimeric forms may be immune targets responsible for the induction and/or progression of disease (2, 4, 18). Employing dimeric HLA-B27 tetramers devoid of β₂m, Kollnberger et al. demonstrated that dimeric HLA-B27 is a ligand for several KIR and leukocyte Ig-like receptors (55). Another hypothesis suggests the involvement of HLA-B27-recognizing CD4⁺ T cells (9, 10). This idea is supported by the CD8⁺-T-cell independence of the rodent models, the observation that novel forms of HLA-B27 can bind peptides considerably longer than nonamers (16 to 33 amino acids) (26, 115), and the recovery of HLA-B27-restricted CD4⁺ T cells from patients with AS (9) and HLA-B27 transgenic mice (92). Collectively, these results suggest various ways in which HLA-B27 may serve as a target for pathological immune responses. The relevance of these mechanisms to ReA remains to be determined.

SHIGELLA SPECIES AND THE INDUCTION OF ReA

In addition to various Salmonella species, other gram-negative enteric pathogens have been linked to the development of ReA. Given their close relationship to Salmonella, we will consider the involvement of Shigella species in ReA. There are four major species of Shigella: S. dysenteriae, S. flexneri, S. boydii, and S. sonnei. The incidence of these species in causing human enteric disease varies among developed and third-world countries. From the results of early studies, it was thought that only S. flexneri was linked to the induction of ReA (48, 100). However, recent work has found that infection with S. sonnei and occasionally even S. dysenteriae can also be linked to the development of ReA (33, 72). The failure to previously document the involvement of S. sonnei in ReA has been attributed to several factors, including the small size of outbreaks, the inclusion in study groups of children and young adults who are thought to be less susceptible to ReA, the history of some study groups of high exposure to enteric pathogens, and the broad definition of ReA symptoms used in the recent surveys (27). Despite these issues, it is now evident that several species of Shigella are associated with the induction of ReA and that, as in the case of Salmonella-induced ReA, the induction of autoimmune disease is dependent on both bacterial and host factors.

It is interesting that studies of S. flexneri strains associated with ReA led to the proposal that strains carrying the pHS-2 plasmid are arthritogenic (104, 105). This possibility was solidified when other, nonarthritogenic Shigella species were found to lack the pHS-2 plasmid, and it was proposed that the presence of the plasmid was a bacterial marker for arthritogenicity (105). The pHS-2 plasmid was shown to encode a protein that immunologically cross-reacts with HLA-B27, and this finding led to the development of a molecular mimicry hypothesis (see above) (112, 116). However, interest in this hypothesis has waned since a clear correlation between the presence of the HLA-B27 cross-reactivity and ReA has not been found and ReA has subsequently been linked to Shigella species not believed to harbor pHS-2 (S. sonnei and S. dysenteriae). Interestingly, a recent study examining a range of Shigella species found that pHS-2 is widely distributed among Shigella species and exhibits significant sequence microheterogeneity (1). This finding resurrects the possibility that the pHS-2 plasmid is a marker for Shigella-induced ReA, but elements other than the HLA-B27 cross-reacting protein carried by the plasmid may play a role in disease induction.

SUMMARY AND CONCLUSIONS

Our understanding of how enteric bacteria, such as Salmonella, initiate an autoimmune disease process leading to ReA is presently quite incomplete. Notably, all organisms etiologically linked with ReA are intracellular pathogens that initiate infection by invading the urogenital or intestinal mucosa. Also, entry via an epithelial barrier and the involvement of host mucosal immune elements are likely to be important. Studies of ReA initiated by Yersinia or Chlamydia infection have clearly demonstrated the presence of bacterium-reactive T cells in the joints of affected individuals. How these cells initiate and/or sustain disease is not clear, and further study is needed to identify the recognized antigens and relevant effector functions. Although a role for HLA-B27 in ReA and related diseases is clear, the precise mechanism by which HLA-B27 promotes the development of ReA is uncertain. A controversial role for HLA-B27 in affecting bacterial pathogenesis has been proposed. Initial work supported a role for HLA-B27 in the presentation of an arthritogenic peptide to CD8⁺ T cells. However, this view is undergoing revision with the appreciation that alternative forms of HLA-B27 can be targets for CD4⁺ T cells as well as Ig-like receptors expressed on a broad range of cell types. In the initial report of HLA-B27-restricted CD8⁺ T cells, only 4 of 158 CD8⁺ T-cell receptor αβ (TCRαβ)-T-cell clones isolated from ReA patients displayed HLA-B27-restricted cytotoxicity toward infected targets while a similar number displayed HLA-B27-restricted cytotoxicity toward uninfected autologous targets (38). In addition, CD4⁺ TCRαβ- as well as TCRγδ-T-cell clones were recovered but their reactivities were not reported (38). This raises the possibility that T-cell subsets with noncytolytic effector functions (regulation and cytokine production, etc.) and additional HLA-B27-independent and -dependent recognition patterns may be involved in disease onset or progression. It is equally important that many cases of ReA occur in B27-negative individuals, so an increased understanding of HLA-B27-independent phenomena is needed (reviewed in reference 55). Further studies of the initial host response to infection, the definition of antigens driving Salmonella-associated autoimmune disease, and the development of suitable animal models will likely yield new insights of value not only in the diagnosis of ReA and the determination of prognosis but also in the development of improved strategies to modulate the disease process and restore healthy tissue.