ANKYLOSING SPONDYLITIS IS CAUSED BY KLEBSIELLA
Evidence from Immunogenetic, Microbiologic, and Serologic Studies
Alan Ebringer, BSc, MD, FRCP, FRACP
From the Division of Biomolecular Sciences, King's College, University of London; and Department of Rheumatology, UCM School of Medicine, Middlesex Hospital, London, England.
DEFINITION OF THE PROBLEM: THE LINK BETWEEN KLEBSIELLA AND ANKYLOSING SPONDYLITIS
Ankylosing spondylitis (AS) is caused by Klebsiella pneumoniae infection. The reason for proposing this particular bacterium, as the causative agent of this disease, is that specific antibodies to this microbe, but not to twelve others, have been demonstrated by two groups, initially by us in England63-44 and subsequently by another group in Finland."
Patients suffering from AS, during active phases of the disease, as defined by elevations in acute phase reactants, have specific antibodies to Klebsiella pneumoniae, but not to other bacteria, such as:
or the yeast Candida albicans.
It was in 1976, that for the first time, an association was suggested between the gram-negative microbe Klebsiella and the chronic, remitting inflammatory arthritic disorder, AS.22
Using three separate approaches, involving immunogenetic, microbiologic, and serologic criteria, this article reviews the background and evidence that links the presence of Klebsiella to the development of recurrent inflammation in AS patients, especially during active phases of the disease. If an association can be clearly established between Klebsiella and AS, then this has important therapeutic implications. It could provide a novel approach to treating or at least controlling severe and active inflammatory phases of the disease, especially in younger patients, before deformities or severe ankylosis of the spine has developed, thereby avoiding or preventing the severe sequelae of AS. Furthermore, a therapeutic response, would by itself, have important theoretical implications in that it could provide additional evidence that Klebsiella is the causative agent of this disease.
The presence of specific anti-Klebsiella antibodies in AS patients, during acute phases of the disease, clearly indicates that these patients have recently been exposed to these specific bacteria, and the speculative proposal is being made that the consequent immune response leads to secondary pathologic complications of "reactive arthritis," especially in a genetically susceptible host, such as one carrying HLA-B27." The test of this theory is clear and relatively simple: removal of anti-Klebsiella antibodies. Elimination of the implicated bacteria, using antibiotics or whatever means, should lead to reduction in inflammation, which can be assessed readily by a drop in levels of acute phase reactants.
If the level of acute phase reactants drops, then the clinical question would arise, whether elimination of specific antibacterial antibodies would lead to patient improvement and arrest in the progression of the disease. It is evident that damage that had already occurred, such as arthritic deformities, contractures, or "bamboo spine," could not be altered by such therapeutic procedures. The clinical end-point that would be expected if this theory is correct is that no new damage would occur in AS patients when specific anti-Klebsiella antibodies had disappeared and acute phase reactants had returned to normal levels.
If the disease continued to progress despite the disappearance of the specific antibodies following antibiotic or dietary therapy, then this would indicate that the specific antibacterial antibodies are interesting epiphenomena occurring in AS, but the bacteria evoking these responses are not the causative agents of this disease.
Although an arthritic skeleton from presumably an AS patient was first described by Connor, in Paris, in 1692, it was not until the end of the 19th century that AS was considered as a distinct inflammatory disorder of the spine and large joints by European physicians such as Bechterew in Russia, Fagge in England, Marie in France, and Strumpell in Germany.10 In the American literature, this condition was thought to be a variant of rheumatoid arthritis (RA) and known as "rheumatoid spondylitis." It was not until the publication of the Rome and New York criteria in 1961 and 1966, respectively, that consensus was achieved, and AS recognized as a distinct entity, quite separate from RA.
Within a few years, however, this clinical separation of AS from RA was eminently confirmed from an entirely unexpected quarter, when it was demonstrated by American57 and European' workers that both diseases occur more frequently in individuals carrying particular HLA antigens. HLA-B27 was found to occur more frequently in AS patients, whereas a significant proportion of RA patients subsequently were found to carry HLA-DR4,5'-60 compared to the frequencies of these HLA antigens found in the healthy, control population. These HLA associations have altered significantly our perception of rheumatology in that no viable model of either of the two diseases, AS or RA, can be proposed without, at the same time, providing some explanation for these immunogenetic observations. Can these immunogenetic observations, such as HLA-B27 in AS, be linked to gram-negative bacteria, such as Klebsiella?
HLA-B27 AND GRAM-NEGATIVE REACTIVE ARTHRITIS
It is the association between HLA antigens and arthritic complications occurring after enteric infections that has provided the vital clues to identifying the specific bacteriologic agents involved in AS and related diseases. The three main gram-negative microorganisms associated with reactive arthritis following enteric infections are Salmonella, Shigella, and Yersinia, and most of the affected patients are carriers of HLA-B27.3 These observations provide a possible clue to any putative infectious agent that may be involved in AS. The agent could probably be a gram-negative microbe, related to Salmonella, Shigella, or Yersinia, but not having such overt enteropathic effects. Such a microbe could readily be a member of the normal, commensal bowel flora. Thus it would not be that unexpected for AS to also be a form of reactive arthritis and the microbe should be related to the enteric pathogens, causing HLA-B27-associated reactive arthrindes.
Because reactive arthritis is also present in Reiter's syndrome and rheumatic fever,40 as well as in gram-negative reactive arthritis,2-62 to entertain the possibility that a similar process operates in either AS or RA is not such a novel or revolutionary concept and quite compatible with existing pathologic precedents and examples; however, it should be apparent that the concept of "reactive arthritis" breaks one of Koch's postulates for incriminating a microbe as the cause of an infection. Clearly, the causative microbe is not to be found at the sites of pathologic lesions but somewhere else, and it is immunologic cross-reactivity mediated by antibodies and immunologically competent cells that cause the secondary tissue damage, characteristic of "reactive arthritis" or rheumatic fever. Thus if AS and RA are forms of reactive arthritis, then it would explain why despite an intensive search for causative agents in the joints and synovial fluids over the last 100 years none has so far been unequivocally identified. The concept of reactive arthritis, when applied to AS and RA, predicates that no viral or bacterial agent will be found in the arthritic joints of AS and RA patients, an expectation more or less in agreement with current negative observations.
Where should one look for such a microbiologic trigger factor? In rheumatic fever, the Streptococcus is located in the tonsils and upper respiratory tract; in Reiter's disease, the agent is in the urethra, and in Salmonella. Shigella, and Yersinia reactive arthritis, the microbe is found in the gut. A possible answer to the question of where one should look for the putative viral or bacterial agent may be provided by the immunogenetic dimension-the association of HLA-B27 with AS.
HLA-B27-THE IMMUNOGENETIC DIMENSION
It is well established that HLA-B27 is associated with AS in all racial groups examined, whether they be European, Japanese, or Asians from the Indian subcontinent.41-42 Furthermore, AS is rare in groups in which HLA-B27 is almost absent, such as the African blacks, and many years ago, Morris Ziff6* noted the lower prevalence of AS in American blacks. By contrast, in ethnic groups in which the frequency of HLA-B27 is found to be above 15% or 20%, AS is a common, almost endemic disease, as observed in Eskimos and Athabascan Indians of North America. The highest prevalence of definite AS is observed among Haida Indians, who have the highest frequency of HLA-B27; 50% of the Haida population have HLA-B27 and virtually 100% of the AS patients are HLA-B27 positive.42 Thus prevalence of AS correlates with the presence of HLA-B27 in all ethnic groups, but it must be noted that the predominant majority of HLA-B27 carriers are healthy individuals, free from any disease. Although the frequency of HLA-B27 is high in AS patients, it rarely reaches 100%, and this raises the question of HLA-B27-negative AS.
Recently, evidence was presented that strongly suggests that HLA-B27-positive AS and HLA-B27- negative AS were different entities. In an extensive literature survey of clinical features in B27-positive and B27-negative AS patients, it was found that in B27-negative patients, the mean age of onset was in the thirties, incidence of the disease in both sexes appeared to be the same, and associated Crohn's disease or psoriasis was frequently present.48
Some years ago, it had been observed that there is a greater prevalence of acute-anterior unveitis (AAU) in B27-positive patients in the United States,43 and subsequently, similar observations were made in the Netherlands.49 Furthermore, it is unusual to observe families with two or more first-degree relatives affected with B27-negative primary AS in the absence of psoriasis or inflammatory bowel disease.71 In general, HLA-B27-negative AS patients rarely develop severe disease, at least in Dutch14 and English populations,71 in the absence of inflammatory bowel disease or psoriasis. Clearly, the presence of HLA-B27 is important in defining severe AS, and thus the B27 gene would appear to be the "arthritogenic gene." This can be tested by transferring the HLA-B27 gene into experimental animals and assess whether such transgenic animals develop arthritis if exposed to gram-negative bacteria, such as Yersinia, one of the microbes involved in HLA-B27-associated reactive arthritis. Several groups have reported studies with transgenic mice, all showing some form of arthritis, and when the animals were exposed to gram-negative bacteria, such as Yersinia, then the occurrence of arthritis rose to almost 100%.39-54
Almost two decades ago, it had been suggested that the HLA-B27 gene was only a marker for a second gene present in the vicinity of HLA-B27, which was coding for a putative "ankylosing spondylitis" gene." The AS gene was thought to be in linkage disequilibrium with the HLA-B27 gene, and this was called a "two gene theory"18; however, the lack of dissociation between AS and HLA-B27 in family studies,71 the association of AS with HLA-B27 in many different racial groups,42 the evidence showing that B27-positive and B27-negative AS are different entities,4* and the demonstration by transgenic studies that HLA-B27 is an arthritogenic gene54 clearly indicate that the "two gene theory" is no longer tenable and does not provide an explanation for the association of HLA-B27 with AS.21
The fundamental assumption of "one gene theories" is that only one gene is involved, thereby suggesting that the HLA-B27 gene itself is the predisposing AS gene.18 At the moment, the two main theories vying for the role to explain the association of HLA-B27 with AS are both based on the assumption that the HLA-B27 gene is the arthritogenic gene, which in association with an environmental factor causes the disease. The evidence suggesting the interaction of an environmental factor comes from identical twin studies, which show that the frequency of the second twin developing AS when the first twin has the disease, that is to say the concordance rate, is well below 50%.w
THEORETICAL MODELS TO EXPLAIN THE ASSOCIATION OF HLA-B27 WITH ANKYLOSING SPONDYLITIS
The two main theories trying to explain the association of HLA-B27 with AS are the receptor theory and the molecular mimicry theory. Any theory trying to answer the question of what causes AS, or what causes RA, must provide an explanation for the specific facts known about these diseases:
1. The male/female ratio in AS is 3/1, and in RA, it is the reverse, 1/3.
2. In AS, 96% of the patients possess HLA-B27, whereas in RA, 70% of the patients possess HLA-DR4.
3. The concordance rate in identical twins in AS is around 40%, whereas in RA, it is about 20%.
It is clear that both AS and RA possess different specificity factors, which can be used to test the validity and viability of any theoretical model proposed to explain the association of HLA with disease. Following are the theoretical models that have been proposed.
The receptor theory states that a class I major histocompatibility complex (MHC) molecule acts as a receptor cavity that binds some as yet undefined, external, environmental polypeptide, which then somehow increases the susceptibility of the subject to develop the disease. The receptor theory, at the moment, appears to be favored by most researchers.7 It is based on the idea, initially proposed by Zinkema-gel and Doherty,73 that class 1 MHC molecules present antigens. The recent crystallographic studies by Bjorkman and co-workers8 that the HLA-A2 molecule consists of two alpha domains separated by a floor of eight antiparallel beta strands-thereby forming a groove, which would contain or present an antigenic polypeptide-provides a structural basis for the antigen-presenting function of class I MHC molecules (Fig. 1). At this stage, however, the nature of the "arthritogenic" polypeptides present in the HLA-B27 groove have not, as yet, been identified. Until such a polypeptide is found, the validity of the receptor theory will remain untested, and therefore, the ability of this theoretical model to answer the specificity questions relating to AS will be unknown. Clearly, at the moment, the receptor theory does not offer any testable therapeutic possibilities, either in AS or RA.
Molecular Mimicry Theory
The molecular mimicry theory states that the HLA antigen has a sequence, which resembles biochemically and immunologically a similar sequence, located in some environmental agent, such as a bacterium.17 The molecular mimicry theory was first proposed in 1976, when it was suggested that several gram-negative microorganisms, such as Klebsiella, Shigella, and Yersinia carry antigens cross-reacting with HLA-B27.24-70 In 1977, increased isolation of Klebsiella from AS patients was reported from London.30
In 1980, anti-B27 allogeneic human tissue typing sera were found to bind preferentially to Klebsiella antigens using three different
Figure 1. Schematic representation of the first two domains of an HLA-B27 molecule. The "molecular mimicry" theory proposes that an anti-Klebsiella antibody binds to part of the hexamer sequence, present on the outer lip of the a-1 domain and when present in high concentration will damage the HLA-B27-positive lymphocyte. The "receptor theory" pro-poses that a bacterial or viral polypeptide, some 20 to 30 amino acids long, fits into the groove between the a-1 and a-2 domains and thereby, probably, through T-cell activation, win produce inflammation and AS.
techniques: hemagglutination, radiobinding assay, and radiolabelled antigen competition assay.4 Subsequently, reports appeared from several different centers, confirming the presence of cross-reactivity between HLA-B27 and gram-negative bacteria.
In 1984, a group from Amsterdam reported that mouse monoclonal anti-HLA-B27 sera showed increased binding for Klebsiella, Shigella, and Yersinia antigens.*5
In 1985, a group from Los Angeles carried out the reverse experiment, when they described an anti-Yersinia monoclonal antibody, which reacted with 12 out of 12 HLA-B27 1ymphoblastoid cell lines, but with only 4 out of 31 B27-negative ones. However, three of the four reactive B27-negarive cell lines carried HLA-B7, an antigen that cross-reacts with HLA-B27.44
In 1986, it was shown that the anri-HLA-B27 (M2) monoclonal antibody bound specifically to a 70 kD component of Klebsiella pneumoniae, whereas no such reactivity was demonstrated with five other monoclonal antibodies.55
In 1987, studies from Oldstone's group identified a hexamer amino acid sequence, Gln-Thr-Asp-Arg-Glu-Asp (QTDRED), which was present in both the HLA-B27.5 molecule (residues 72-77) and Klebsiella pneumoniae nitrogenase reductase enzyme (residues 188-193).** Furthermore, antibodies to this shared sequence were found in the sera of the AS patients." In the case of HLA-B27, this hexamer epitope is located on the outer lip of the alpha-1 domain and therefore accessible to circulating anti-Klebsiella cross-reacting antibodies (Fig. 1). The identification of such an epitope raises the question whether antibodies to it would also recognize HLA-B27-positive cells.
In 1989, Husby et al37 showed that rat antisera raised against Klebsiella nitrogenase enzyme polypeptides, containing the shared QTDRED hexamer, reacted with 13 of 14 synovial biopsies obtained from HLA-B27- positive AS patients, but with none of seven biopsies obtained from HLA-B27-negative individuals suffering from either RA or osteoarthritis.
It would seem that molecular mimicry between HLA-B27 and Klebsiella bacteria has been defined down to the level of similarity of six amino acids, found in both the suspect bacteria and the genetically susceptible population. Thus the term "molecular mimicry" could not be said to be inappropriate when studying cross-reactivity between bacterial antigens and HLA molecules. Furthermore, this molecular mimicry would appear to provide a simple model for the pathogenesis of AS in genetically susceptible individuals.
MOLECULAR MIMICRY AND PATHOLOGY OF ANKYLOSING SPONDYLITIS
If there is cross-reactivity between HLA antigens and bacteria, then infection by such microorganisms will lead to the production of antibodies, which will have both antimicrobial, as well as antiself or autoimmune activity. Only a small proportion or a subset of the antimicrobial antibodies will also have antiself or autoimmune activity. The bacterial antigens, carrying the shared sequences, will be immunogenic, especially around the edges of the shared sequences, because it is at the sites that the immune system will not recognize that it is dealing with a self antigen. If for instance, the bacterial shared sequence is spanning residues 188-193 of the nitrogenase molecule then the edge antigenic pentamer, 187-188-189-190-191, will not be recognized as a self antigen, because of the haptenic effect of the nonself amino acid at position 187.
The antibodies produced against such an edge antigenic determinant, however, may well bind to HLA-B27. There is thus no breakdown of tolerance, and the evocation of bacterial antibodies having antiself or autoimmune activity is part of the normal immune response when encountering partially cross-reacting antigens. A similar process occurs in rheumatic fever.40 When present in small quantities, such antibodies would not activate the complement cascade, and therefore, no cytotoxic event and no inflammation would ensue (Fig. 2A). When such antibodies are present in high concentrations, however, they would activate the complement cascade with consequent stimulation of inflammation that eventually may result in localized tissue damage and fibrosis, especially in the entheses around the lumbar spine and sacroiliac joints6 (Fig. 2B).
If the bacteria were to be present in the gut or the bowel mucosa, it could be expected that the related lymph nodes, which are present in the mesentery of the gut and the pelvis, would be closely linked anatomically to both the lumbar spine and sacroiliac joints, and therefore, high titers of antibodies would be expected in these areas, which are the main pathologic sites of disease expression in AS. Continued production of high titers of antibodies would lead to inflammation around the lumbar spine and sacroiliac joints, and therefore, presence of clinical signs and symptoms related to those areas would be expected in such patients.
In situations of persistent infections, the antibodies would enter from the mesenteric and pelvic lymph nodes into the general circulation, and thus such antibodies should not only be detectable in the blood, but would also affect more distal sites, such as the neck, uvea, and peripheral large joints."
Figure 2. A—Anti-Klebsiella antibody present in low concentration, binds to HLA-B27 lymphocyte but compliment cascade is not activated, cell is not damaged, and therefore no inflammation occurs. B—Anti-Klebsiella antibodies present at high concentrations bind to HLA-B27 lymphocyte. Proximity of 2 Fc parts of immunoglobulins leads to compliment activation with cell death and therefore inflammation occurs. The result of inflammation is the deposition of fibrous tissue, thereby limiting joint mobility and producing spinal ankylosis.
.The persistent infection would have to be in the gut, or more precisely in the bowel mucosa. Elegant ileocolonoscopic studies by Mielants et al53 have shown that over 50% of active AS patients have histologic evidence of gut inflammation. If the bowel mucosa is affected, then abnormalities should also be present in the local mesenteric, presacral, and pelvic lymph nodes. Some years ago French workers described abnormalities in pelvic and sacral lymph nodes in AS patients using lymphangiography and observed that inflammation and sclerosis of lymph nodes were not only present in AS but also the lymph node changes seemed to precede the development of radiologjc changes in the lumbar spine and sacroiliac joints."
The pathogenetic mechanism in AS, using the molecular mimicry model, could be described briefly as follows:
1. INFECTION occurs in the bowel mucosa of the ileum and ascending colon, in which the bacteria replicate and the quantity of bacteria will depend on dietary starch intake of the patient. This infection is at a different site from the eventual location of the arthritic lesions of AS.
2. ANTIBODY against Klebsiella is produced in the lymph nodes draining the lower ileum, and they are located in the mesentery, which is attached to the anterior aspect of the lumbar spines. The high titer of anti-KIebsiella antibodies are being produced by lymph nodes that are anatomically and geographically located near the main pathologic sites in AS, namely the lumbar spine and sacroiliac joints. A subset of these antibodies will bind to cells carrying HLA-B27.
3. INFLAMMATION will be triggered by those anti-Klebsiella antibodies, binding to HLA-B27 molecules and because of complement cascade activation, cell death and tissue damage will occur (Fig 2B). The result of inflammation is fibrosis, arid this is the main pathologic lesion in advanced AS.
4. CHRONIC DISEASE will supervene, because repeated cycles of infection will produce recurrent waves of anti-Klebsiella antibodies that will cause tissue damage. Eventually, this will manifest itself as a generalized fibrosis around the lumbar spine, and this disease is called ankylosing spondylitis.
Is there any evidence that suggests that the pathologic processes in AS occur through such a molecular mimicry mechanism?
Antisera raised against Klebsiella nitrogenase polypeptides, containing the shared QTDRED hexamer, were found to bind to synovial biopsies from B27-posi-tive AS patients but not to biopsies obtained from B27-negative patients, thereby suggesting that this particular epitope could be the target of immunological activity.37
In 1990, Baines and co-workers5 showed that rabbit anti-Klebsiella sera bound to a greater extent to lymphocytes obtained from 30 B27-positive AS patients and 21-B27 positive healthy controls compared to the binding by lymphocytes obtained from 33 B27-negative healthy controls. Thus HLA-B27-positive lymphocytes obtained from either AS patients or healthy controls reacted equally well with anti-Klebsiella sera, suggesting that there was no difference in the HLA-B27 molecules found in the two groups. This agrees with molecular biology studies that indicate that the HLA-B27 gene itself does not differ between AS patients and healthy controls," whereas Ivanyi31 could not find any dysfunction in HLA-B27 that could explain its role in AS. Finally, Ewing and co-workers,37 using synthetic peptides, demonstrated that higher antibody-binding activity was found around the edges of the shared QTDRED hexamer sequence.
Because allogeneic anti-HLA B27 antibodies bind to Klebsiella antigens4 and xenogeneic Anti-Klebsiella antibodies bind preferentially to HLA-B27-positive cells,5 it is not too speculative to suggest that in presence of high anti-Klebsiella antibody titers, the characteristic lesions of AS may arise. If the hypothesis is to be considered as a valid model for the pathogenesis of AS, then a clear logical expectation is that both antigen (Klebsiella microbe) and antibody (anti-Klebsiella immunoglobulin) should be detectable in AS patients during active phases of the disease.
In 1976, when our experimental studies suggested that there might be cross-reactivity between HLA-B27 and some gram-negative bacteria such as Enterobacter, Klebsiella, Salmonella, Shigella, and Yersinia,1* a research AS Clinic was set up at the Middlesex Hospital in London. It was discovered quickly that only Klebsiella but not the other four cross-reacting microorganisms could be isolated from fecal samples obtained from AS patients.
In 1977, we reported in a study involving 63 AS patients that Klebsiella could be isolated more frequently during active phases of the disease.30 In a second sequential study, involving 163 AS patients, it was shown that a clinical relapse was preceded by the appearance of Klebsiella in fecal samples,31 and active inflammatory disease was associated with elevation in total serum IgA, suggesting that a microbial agent was acting across a mucosal surface, such as the gut.13
Uveitis occurs in approximately 30% of AS patients, but because AAU is seen more often in AS patients with severe disease than in HLA-B27 relatives with or without sacroiliitis, has led Van der Linden et al" to argue that AAU is associated with severity of disease rather than HLA-B27. If this is the case, then presence of fecal Klebsiella should also correlate with presence of AAU. In a study of AS patients with AAU, as diagnosed by an ophthalmologist, Klebsiella was found in 13 of 17 episodes (76%) compared with an overall isolation rate of 30% in AS patients without uveitis, involving altogether 782 cultures.29
A similarly high isolation rate was found by the Leeds group, who reported that eight of nine AS patients with AAU had Klebsiella in their stools,15 but this association was not found when other parameters of disease activity were examined.72
Three other groups have found an association between isolation of Klebsiella and active disease in AS patients. In a study involving 17 AS patients, the Winnipeg group found an association between presence of fecal Klebsiella and elevation in serum C-reactive protein levels, but no association with elevation in erythrocyte sedimentation rate.3* By contrast, in a study involving 56 AS patients, Van Kregten et al67 found an association between presence of fecal Klebsiella and elevation in erythrocyte sedimentation rate (P < 0.001), but no association with serum C-reactive protein levels. Because erythrocyte sedimentation rate and serum C-reactive protein levels are correlated in AS patients, if the number of subjects examined is sufficiently large," this clearly indicates that greater numbers of patients should be studied, especially during active phases of the disease. In a study involving 43 AS and Reiter's syndrome patients, there was an association between fecal Klebsiella and presence of clinically active disease.45
These studies on AS patients from Europe and North America, involving several centers, clearly leads to the suggestion, but not conclusive proof, that Klebsiella is somehow involved during active phases of the disease. Nevertheless, the clear demonstration that there is cross-reactivity between HLA-B27 and a Klebsiella protein, and the independent demonstration that the Klebsiella microbe is found in AS patients during active phases of the disease, is consistent with the predictions of the molecular mimicry model.
The molecular mimicry model, however, makes a further prediction in that it is not sufficient for the antigen (Klebsiella microbe) to be present, but antibodies to that antigen must also be demonstrated, because the pathogenetic process involves antibodies produced against cross-reactive antigens and together with immunocompetent cells this leads to autoimmune damage.
It has been known for a long time that serum IgA is elevated in AS47-6* and that it is usually associated with inflammatory phases of disease activity."30 Elevation in serum IgA indicates antigenic stimulation across a mucosal surface, and the largest mucosal surface in the body is the gut. Plasma cells in the gut mucosa are the main source of serum IgA,M and therefore, gram-negative gut bacteria could have been responsible for this immunoglobulin elevation.20
The demonstration of specific antibodies to a particular antigen is crucial to any theory, purporting to claim that the antigen in question is the etiologic trigger factor of the disease. Therefore, this problem has been investigated by our group, using six different methods of measuring antibodies and in each case employing different microbial controls. All studies have been carried out under coded conditions, in that the person carrying out the assay did not know the source of the sera, whether it came from AS or other patients, whether the patients were active, or whether the sera came from control subjects. The results of these investigations, spanning a period of 8 years, give an unequivocal and clear answer only antibodies to Klebsiella and to no other bacteria can be demonstrated in AS patients,10 and similar results have been reported recently from Finland."
Briefly, the results of our investigations are as follows:
1. In 1983, in a collaborative study between Guy's Hospital and Middlesex Hospital, the mean IgA anti-Klebsiella titer, measured by ELJSA, was elevated significantly in 43 active AS patients when compared to 39 inactive AS patients, 57 healthy controls, 13 patients with psoriatic arthritis, and 38 RA patients.*4 Control studies with E. coli and Candida albicans showed no elevation in any of the five groups. It is relevant to note that in this study, the mean serum C-reactive protein level in the RA patients was twice as high as the level found in active AS patients, thereby indicating that the elevation in anti-Klebsiella antibodies found in AS patients was not the result of some nonspecific inflammation. Furthermore, the lack of elevation in anti-E. coli antibodies, clearly argues against some nonspecific leakage of bacterial antigens across the bowel mucosa.
2. In 1984, in a second collaborative study with Guy's Hospital, specific antibodies to Klebsiella, Pseudomonas, Salmonella, and Yersinia were measured in 12 AS patients, whose total serum IgA was in excess of 4.6 g/L (i.e., 2 SD above the mean for controls) and compared to the mean levels in 81 controls. Although there are many shared antigens between these gram-negative bacteria, only antibodies to Klebsiella and not to the other three microbes were found in the AS patients."
3. In 1985, absorption studies were carried out with Klebsiella pneumoniae, Proteus mirabilis, and Streptococcus pyogenes on sera obtained from 91 AS patients and 20 healthy controls. In 47 AS patients, each with a total serum IgA greater than 3 g/L, there was an increased absorption of IgA antibodies by Klebsiella, whereas absorption by equal numbers of Proteus or Streptococcus microorganisms did not exceed background levels.22
4. In a second study carried out in 1985, using indirect Coombs bacterial agglutination, elevation in mean titer of anti-Klebsiella antibodies was found in 24 active AS patients, whereas no such elevation was found in 28 inactive AS patients, 30 RA patients attending a gold clinic, and 41 healthy controls.2* By contrast, when the same sera were tested against Proteus mirabilis, using the same indirect Coombs bacterial agglutination technique, elevation in titer of anti-Proteus antibodies was found only in the RA patients, but not in the other three groups. This was the first time that specific anti-Proteus antibodies had been reported to be elevated in RA, but subsequently this has been confirmed by two other groups, one from Dublin** and the other one from Newcastle.15 It is relevant to note that in this study each disease group acted as an inflammatory control for the other group and in each group, a specific antibacterial antibody titer could be demonstrated: AS patients had antibodies to Klebsiella but not to Proteus, and RA patients had antibodies to Proteus but not to Klebsiella, whereas inactive AS patients and controls had no antibodies against either microbe. Subsequent studies from our group have extended these observations, and we have suggested that RA is a reactive arthritis, similar to AS, but evoked by a different microbe, namely Proteus mirabilis, probably located in the upper urinary tract.27
5. In 1986, in collaboration with the Toivanen group from Finland," the absence of anti-Yersinia antibodies in English active AS patients was confirmed, using their immunoblotting anti-Yersinia assay. Coded samples were sent from London, containing English active AS patients, English RA patients, English controls, and four sera from Finnish Yersinia reactive arthritis patients. Specific antibodies to Yersinia were detected only in the Yersinia reactive arthritis patients from Finland."
6. In 1987, using a radiobinding assay, an elevation in anti-Klebsiella antibodies was found in eight active and 15 probably active AS patients when compared to the level found in 20 control subjects, whereas no such elevations were found with Shigella microorganisms.26
7. In 1988, IgA antibodies to Klebsiella were measured by immunoblotting in 29 active and 26 inactive AS patients and compared to the levels found in 19 RA patients and 30 controls. Highest levels in number of anti-Klebsiella bands were found in active AS patients, and intermediate levels were found in inactive AS patients, whereas there was no significant difference between RA patients and controls.25
8. In 1989, antibodies to Klebsiella and Proteus were measured by immunofluorescence in 33 active AS patients, 102 RA patients, and 20 controls. Again,.elevated anti-Klebsiella antibody titers were found in AS patients but not in the RA patients and controls, whereas elevated anti-Proteus antibody titers were found in the RA patients but not in the AS patients and controls.1
Other groups have also found elevated titers of anti-Klebsiella antibodies in AS patients. In 1988, Cooper et al11 reported elevated anti-Klebsiella antibody titers not only in AS patients but also in RA patients and patients with inflammatory bowel disease, compared to the levels found in controls, thus favoring a leaky gut hypothesis, but no control studies were done with other bacteria.
Further studies from other centers are required to resolve the problem of the presence or absence of specific antibacterial antibodies in both AS and RA patients.
More recently, in an extensive study from Finland, antibodies to Salmonellae, Yersinia, Campylobacter jejuni, Borrelia burgdorferi, Klebsiella pneumoniae, Escherichia coli, Proteus mirabilis, and Chlamydia trachomatis were measured by ELISA in the sera of 99 AS patients and 100 controls.51 Increased prevalence of IgA and IgG antibodies were observed only against Klebsiella, whereas some slight elevation in IgA antibodies was also observed with E. coli. In their discussion, the authors comment, "IgA class antibodies against E. coli were also observed to be slightly elevated in AS, this can perhaps be explained by cross-reactive antigens with Klebsiella.""
The presence of specific anti-Klebsiella antibodies, demonstrated by several techniques and from different centers, clearly suggests that no pathogenetic model of AS can be considered without taking into account this highly reproducible, serologic phenomenon.
THE CONCEPT OF KLEBSIELLA REACTIVE ARTHRITIS AND TEST OF THE THEORY
All clinicians are aware that early cases of any disease are difficult to diagnose, but this applies especially to AS in which a characteristic delay to diagnosis may take several years, if not decades. Once deformities have occurred, any therapy will be palliative or at most stop current inflammation, but severe arthritic changes such as Bechterew stoop or bamboo spine are essentially irreversible; however, if one could identify early AS or pre-AS, before the Rome or New York criteria had been satisfied, therapeutic intervention may stop or even prevent the severe sequelae of advanced AS.
The body of extensive evidence presented in this article, involving immunogenetic, microbiologic, and serologic studies, each group of investigations being independent from one another and confirmed by other centers, leads to the following conclusions: (1) Klebsiella cross-react with HLA-B27; (2) Klebsiella microbes can be isolated from active AS patients; and (3) Anti-Klebsiella antibodies can be identified in the sera of active AS patients.
In view of these empirical observations, which are consistent with the molecular mimicry model, the tentative hypothesis or theory can be proposed: that AS is a reactive arthritis following Klebsiella infection of the gut. Although the data presented earlier is consistent with the theory or hypothesis that "AS is caused by Klebsiella," it does not prove it. One independent way of proving this theory would be to stop the process of disease progression or abort the disease in its early stages. Thus a logical consequence of the theory that "AS is caused by Klebsiella" would be that removal of the cause should lead to the arrest or even prevention of the disease. Therefore, therapeutic intervention, aimed at removal of Klebsiella microbes and reduction of anti-K!ebsiella antibodies, should lead to a reduction in inflammation and arrest in the progression of the disease. Thus therapeutic intervention is a test of the theory and should be carried out in prospective studies. If AS is the result of repeated episodes of "Klebsiella reactive arthritis," then treatment of such patients, before they develop the stigmata of AS, may abort or even prevent the development of the disease.
If one defines "Klebsiella reactive arthritis"-in a similar manner to "Yersinia reactive arthritis"-as an arthritic condition, occurring in an HLA-B27-positive individual, who has elevated levels of acute phase reactants and elevated titers of anti-Klebsiella antibodies," then objective laboratory parameters, together with"' clinical criteria, can be used to follow response to therapeutic intervention.
Salazopyrine is known to be an effective drug in treating AS,33 which also has antibiotic properties against Klebsiella. Preliminary studies from Finland indicate that the titer of anti-Klebsiella antibodies falls following treatment with salazopyrime.
Other methods of reducing Klebsiella could be by lowering the amount of starch in the diet, which is known to be a substrate for Klebsiella proliferation,12 and this could explain the higher rate of Klebsiella isolation observed in healthy males.18 Whether substrate availability for bacterial growth could explain the male prevalence of AS, awaits further studies; however, if removal of Klebsiella microbes and reduction in anti-Klebsiella antibodies did not arrest the clinical progression of the disease, then this would be a critical Popperian test of the theory and disprove the hypothesis that "AS is caused by Klebsiella."
The results of these therapeutic studies are awaited with eager anticipation, not only as a critical test of the theory but it may also provide a new approach to treatment.
CONCLUSIONS1. Evidence from several laboratories indicates that there is molecular mimicry between HLA-B27 and a number of gram-negative bacteria, especially Klebsiella.
The author gratefully acknowledges the support of the Trustees of the Middlesex Hospital and the Arthritis and Rheumatism Council.
REFERENCES1 Abuljadayel I, Ebringer A, Cox NL- Antibodies to Klebsiella measured by immunofluorescence. Clin Rheum 8:25, 1989
Address reprint request to
Alan Ebringer, BSc, MD Division of Biomolecular Sciences King's College Campden Hill Road Kensington, London W8 England