Several forms of inflammatory arthritis are triggered by infection with bacteria, including reactive arthritis (ReA) following certain gastrointestinal or urogenital infections 1, Lyme disease 2 and the arthritis associated with streptococcal or neisseria infection. Each of these differs from classical septic arthritis in that bacteria cannot readily be isolated from the affected joints, although occasionally Borrelia burgdorferi has been cultured from the synovial fluid (SF) of some patients with Lyme arthritis 23, and gonococci or meningococci can be cultured from joints in the early phase of disease 4. Because the joint is usually sterile, these forms of arthritis have been considered to be the result of some form of autoimmune reaction that was triggered by an infection at a site distant from the joint. However, in more recent studies bacterial antigens 567 and, in several cases, bacterial DNA and/or RNA have been identified in synovium or SF from selected patients with ReA 8910 and Lyme arthritis 11. This implies that organisms associated with these forms of arthritis do reach the joint, although they may be present there in an uncultivable state, and raises the possibility that joint inflammation is driven by immune responses to bacterial antigens without the need to evoke autoimmune mechanisms. In Lyme disease both mechanisms have been suggested for different phases of disease 1213: arthritis driven by bacterial antigens in early disease that is responsive to antibiotic treatment; and autoimmunity in antibiotic-resistant, chronic disease caused by cross-reactive T cells that are responsive to Borrelia OspA protein and self-antigens such as CD11a. However, the precise role of bacterial antigens in pathogenesis has not been defined, and it is unclear whether bacterial antigens reach the joint in sufficient quantities to account for the major sustained inflammation that can characterize ReA.

Using a reverse transcription (RT)-PCR technique to detect bacterial 16S rRNA, we recently demonstrated that transcriptionally active bacteria are present in synovial tissue from patients with rheumatoid arthritis (RA) and other arthropathies 14. These bacteria were mainly derived from commensals that are normally present in the skin and gut, and we concluded that bacteria engulfed by macrophages can reach the joint when macrophages are recruited to the synovial membrane, particularly when recruitment is increased by joint inflammation. We have now applied this general technique for the detection of bacteria to SF samples from patients with ReA and other forms of postinfectious arthritis. The data show that the same commensal bacteria can also be detected in SF from these patients, whereas specific disease associated bacteria can also be demonstrated in some cases. However, bacterial rRNA sequences from disease associated bacteria were detected relatively infrequently as compared with those from other bacteria also detected in the joint. These findings have implications for our understanding of the pathogenesis of postinfectious arthritis.

SF samples were collected from 12 patients with various forms of postinfectious arthropathy; 10 had clinical features of ReA and two had postmeningococcal or post-streptococcal arthritis. An additional two patients were studied with undifferentiated spondyloarthropathy in which a diagnosis of chronic ReA was considered possible. Patient details are given in Table 1, along with the basis on which the diagnosis was made. Total RNA was extracted from the SF samples and approximately half was then treated with DNase I, and then aliquots of both were subjected to RT-PCR using universal primers R1 and R2 for detection of bacterial 16S rRNA sequences. Of DNase treated samples 11 out of 14 tested positive, as compared with 12 out of 14 of the untreated samples. SF from patient C was the only sample that tested positive without DNase treatment and negative following DNase treatment, indicating that bacterial 16S DNA was most likely being detected rather than RNA. Although samples from patients M and N were negative with and without DNase treatment, all other samples had evidence of bacterial 16S rRNA, indicating the presence of 'live' transcribing bacteria in the inflamed joint.

Sequencing of the PCR products was then performed. Twenty clones were picked from each PCR product and sequenced. Only good quality, full-length (350–400 bp) sequences were analyzed in detail; short sequences (less than 200 bp) were ignored because the species from which they came could not be accurately identified. A summary of the total number of sequences analyzed from each patient is detailed in Table 2. (Note that the primers R1 and R2 fail to amplify Chlamydia 16S rRNA, and so Chlamydia sequences would not be expected to be found in this analysis; their presence is investigated below when Chlamydia-specific primers were developed and used.)

As previously noted in the study of bacterial rRNA sequences amplified from synovium of RA patients 14, each patient's SF contained a diverse range of bacteria with no one species dominating the population detected. Interestingly, with the possible exception of patient D, in whom Streptococcus spp. were detected, no evidence of the bacteria thought to be the causative agent for arthritis was detected in any of the samples. Although there was a reasonable correlation between the bacteria detected with and without DNase treatment (e.g. patient K), some samples showed the presence of quite different bacteria in the DNase treated and untreated samples (e.g. patient H). This probably reflects the relatively small number of sequences generated from each sample. In our previous study in which 46 sequences from each tissue sample were analysed 14, there was only approximately 80% correlation between the bacteria detected when the same PCR product was cloned and sequenced on separate occasions. This indicates that even with a large amount of sequencing only a relatively small proportion of the total bacteria present within the joint are detected.

Although bacterial antigens, DNA and RNA from the arthritis-triggering organism have previously been reported in SF and tissue from ReA patients, the presence of a wide variety of bacterial species was unexpected, but we reported very similar findings in our previous study of RA and other forms of chronic arthritis 14. The source of these organisms is not known but may be environmental or from the indigenous microflora. As noted, we took stringent precautions to avoid laboratory contamination, and the organisms identified cannot all be accounted for by skin bacteria introduced into the SF at the time of percutaneous aspiration. We did not surgically remove the skin at the aspiration site to avoid contamination with skin flora, but our results are rather similar to those obtained by others who took this precaution 15. As tests of the effectiveness of our precautions, reagent controls were always negative, and when normal synovium was tested (in a previous study) no bacterial rRNA was obtained. In the present study two SF samples were also negative when DNase was used. In addition, the sequences obtained varied between patients (including the E. coli sequences), arguing against laboratory contamination. Together, this constitutes strong evidence of the ability of commensal organisms such as E. coli to colonize inflamed joints; the gut in different patients would be expected to contain a range of E. coli 'subspecies', with minor variant rRNA sequences.

Our results are similar to those detailed in other recently published reports 161718, although our use of RT-PCR rather than PCR is likely to be responsible for the high proportion of SF samples in which we could detect bacteria. Nevertheless, all of the studies have emphasized that bacterial nucleic acids can be detected in many forms of inflammatory arthropathy (RA, inflammatory osteoarthritis, crystal arthropathy 18) in addition to ReA, and that organisms associated with ReA can be found in synovium/SF of patients with other diagnoses 16. Many species that we found in ReA SF (e.g. Propionibacterium acnes and Streptococcus epidermidis) were seen in our own previous study, implying that their presence in the joint is not disease specific and that they are likely to be opportunistic colonizers of inflamed joints. P. acnes has also be implicated by some investigators in the pathogenesis of SAPHO (synovitis, acne, pustulosis, hyperostosis, osteitis) 19; our data confirm that the organism can access joints, but if it plays a role in SAPHO then this must reflect an abnormal response to the organism in that disease because it is clearly present in synovium from other conditions. Likewise, although streptococci were identified in a patient with seronegative erosive arthritis and persistent high titres of antistreptococcal antibodies, the role of streptococci in this case must remain speculative because streptococcal rRNA was amplified from six other ReA and several RA patients in the previous study.

The present analysis of SF from ReA patients indicates that, even in patients in whom an infectious agent is known to be involved in the pathogenesis of the arthritis, detection of that infectious agent by universal PCR can be masked by the high level of commensal bacteria present. Their presence suggests trafficking from sites such as the gut (an idea strengthened by the detection of E. coli sequence in all samples). In enteric ReA active bowel inflammation will inevitably affect the barrier function of the gut wall, allowing systemic access by gut flora. Nonsteroidal drugs have also been reported to impair gut permeability and mucosal competence 2021, and most of our patients were taking such drugs. The fact that sequences from species of bacteria that definitely play a role in triggering arthritis are in a distinct minority among all of the sequences detected within ReA SF raises questions about the role of intra-articular bacteria in the pathogenesis of ReA, and questions the practical use of universal PCR to diagnose ReA. Although the technique shows increased sensitivity over detection of 16S rDNA, this increased sensitivity may also mask detection of the causative agent. However, the use of specific primers showed complete specificity (Yersinia and Chlamydia rRNAs were only detected in patients in whom other evidence pointed to the involvement of these organisms) but low sensitivity in the case of Yersinia, because only one out of three patients with ReA thought to be related to Yersinia infection had evidence of the specific rRNA in SF.

The findings from investigations in patients with Campylobacter-induced ReA were unexpected. Although Campylobacter sequences were isolated, Campylobacter jejuni was never detected in any of the samples tested. Our data are unable to distinguish clearly between the possibility that certain Campylobacter spp. other than Campylobacter jejuni can sometimes be associated with ReA, and the dissemination of commensal Campylobacter spp. to inflamed joints along with other components of gut flora. The latter is more likely in view of the isolation of Campylobacter sequences from patients with diagnoses other that ReA, but it is not inconceivable that a normal member of the gut flora could, under the appropriate circumstances, be involved in triggering ReA; ReA secondary to Clostridium difficile infection falls into this category 22.

In conclusion we have shown that, in forms of arthritis associated with preceding infection, rRNA from the organism responsible can sometimes be demonstrated within SF. However, rRNAs from the commensal flora of gut and skin are even more easily detected. What are the implications of these findings for our understanding of the pathogenesis of postinfectious arthritis? First, the fact that rRNA sequences from disease associated organisms are a minority of those detected in the joint does not necessarily imply that the disease associated organisms have no role in pathogenesis locally. They may synthesize antigens that evoke vigorous T-cell mediated immune responses, as suggested by the consistent finding of prominent responses to triggering organisms in ReA SF T cells 2324. In contrast, the immune system is usually tolerant of gut flora 25. It is clear that the traffic of organisms to the joint does not lead to their replication and sepsis, because the joints remain sterile when culture studies are carried out. Presumably, they are controlled by innate immune mechanisms, including macrophages and polymorphs; indeed, it is likely that many of the organisms detected were engulfed by phagocytes in the periphery that were then recruited to the inflamed joint. Nevertheless, although the organisms may be rapidly eliminated or contained, and may not elicit specific immune responses, they may have a proinflammatory effect because of the effects of component such as lipopolysaccharide, bacterial DNA and bacterial heat shock proteins on the innate immune system 26272829. These effects may amplify inflammation initially triggered by an immune response to an ReA-associated organism. The duration and severity of arthritis following infection could reflect the efficiency with which these generic proinflammatory mechanisms are brought under control.

RT-PCR using universal primers reveals the presence of bacterial 16S rRNA in SF from patients with ReA or postinfectious arthritis. However, the majority of the sequences detected are derived from commensal bacteria that traffic, probably in macrophages, to the inflamed joint. 16S rRNA from organisms that are responsible for causing arthritis can also be detected but only in a minority of cases, and only by using species-specific primers for amplification. This might reflect low abundance of these bacteria in SF, but truly quantitative studies targeting organism chromosomal and DNA and using real-time PCR would be required to confirm this idea.

bp = base pairs; RA = rheumatoid arthritis; ReA = reactive arthritis; RT = reverse transcription; SAPHO = synovitis, acne, pustulosis, hyperostosis, osteitis; SF = synovial fluid.