Antibody elevations to Klebsiella pneumonias in AS patients were first reported in 1983  and those original observations have since been confirmed in various centres . Furthermore, antibodies to other gram negative organisms such as Escheric-hia coli, Proteus mirabilis, Pseudomonas aeruginosa, Salmonella spp, Shigella spp, and Yersinia enterolitica have not been found to be elevated in AS patients . Additionally, a strong association between HLA-B27 and AS has been known for over 20 years . These two observations were linked in 1987  in the form of an amino acid homology between B*2705 and K. pneu-moniae nitrogenase. The data presented in this paper suggests a new sequence homology between B*2705 and another protein present in K. pneumoniae, the pulD secretion protein of the starch debranching pullulanase enzyme system (Table 1). Whilst the pulD sequence does not display the 100% homology shown with K. pneumoniae nitrogenase a similarity is nonetheless present in the form of a tetramer, (residues 596-599) (DRDE) and B*2705 (DRED). The pulD sequence (residues 596-599) only differs from the DRED sequence in B*2705 (74-77) and K. pneumoniae nitrogenase (190-193) by two conservative substitutions, one, a D (aspartic acid) for an E (glu-tamic acid) at position 598 and the second, of an E (glutamic acid) for a D (aspartic acid) at position 599. It is clear to see that this sequence (underlined) falls within the region of homology previously published by Schwimmbeck et al.  (Table 1).
Various other proteins from enteric bacteria have been shown to have sequence homology with HLA-B27 . These include sequences from common gram negative organisms such as E. coli, Salmonella, Shigella, Pseudomonas and Klebsiella. It has been suggested that the mechanism for the apparent autoimmune condition in AS occurs from the breakdown in tolerance of the host to its own HLA-B27 molecule and the endogenous peptides bound therein. Once the host is exposed to enteric peptides of similar sequence to that of HLA-B27 an immune response may be evoked. It is this immune response that may breach tolerance and so initiate a chronic inflammatory condition such as that seen in AS . However, antibody levels against various organisms, (such as those shown to have similar sequences to HLA-B27) in AS patients revealed no specific elevation in immunoglobulin levels with respect to controls with the exception of K. pneumoniae .
The pulD is the terminal secretion protein of the pullulanase system and is known to have regions said to form strongly amphiphatic/? sheets which confer membrane spanning properties common to outer membrane proteins . This suggests that the protein should be on the surface of the organism and freely accessible to the surveillance of the immune system. The data obtained from the ELISA studies with synthetic peptides demonstrate that the AS patients have antibodies to the 16 mer B*2705 and pulD peptides. The AS patients showed significant elevations in both IgA and IgG antibody classes to the B*2705 peptide. This binding to a peptide of a self protein demonstrates the presence of autoantibodies in the sera of AS patients. It is, therefore, possible that these antibodies may bind HLA-B27 positive cells in patients, fix complement and thereby initiate inflammation. The microbial source, evoking these antibodies could be in the form of the pulD protein of K. pneumoniae. The ELISA data with the pulD synthetic peptide yielded similar data to that obtained with B*2705. The pulD peptide antibody elevations were seen in all three antibody classes whilst, with the control peptide (scrambled pulD peptide), no antibody elevations were observed in any patient group to any antibody class. The data derived from the B*2705 and pulD peptide binding suggests that there is some degree of molecular mimicry or cross reactivity between the two sequences, one self and one microbial.
The presence of the pullulanase system in K. pneumoniae allows the organism to cleave a(l-6) bonds in a(l-4) and a(l-6) branched starches such as glycogen and amylopectin. The effect of this property is that the organism gains access to another carbon source, which can alter the type of antigens presented on the surface of the bacteria leading to the development of the polysaccharide capsule layer and so alters the level of patients serum antibody binding. This can be seen in Fig. 2A, where K. pneumoniae has been grown in a minimal medium with 0.005% yeast extract. When organisms grown in this media were exposed to sera from AS patients and healthy and disease (RA patients) controls an elevation in antibody litre is observed. However, if the organism is grown on the same media supplemented with pullulan (0.3% w/v) as the carbon source the degree of antibody binding in AS patients is greatly increased. This increase also shows that the capsule is highly antigenic and is one of the predominant antigens recognised by AS patients, an observation also made by other workers . In the control experiment, where P. mirabilis is grown in the same conditions, with and without pullulan, antibody elevations are only seen in
B n.s. p< 0.001 n.s. p< 0.001
Fig. 4. IgA (A) and IgG (B) antibody litres (mean ± S.E.M.) in 25 controls (C), 25 rheumatoid arthritis (RA) patients and 97 ankylosing spondylitis (AS) patients when tested by ELISA against type I and type IV collagen. OD = optical density; n.s. = not significant.
against the extracellular enzyme pullulanase (pulA) in AS patients when compared to controls. The pulA enzyme would appear to have a very important property in relation to arthritic disease. At the N-terminal end of the pulA protein in some strains of K. pneumoniae there are repeats of the tripeptide Gly-X-Pro, characteristic of collagens type I, type III and type IV . IgA and IgG antibody elevations to both type I and type IV collagens have been demonstrated in AS patients compared to controls. It is interesting to note that type I collagen is present predominantly in tendons and bone . The possibility of antibodies cross reacting between pulA and type I collagen may be important in the deposition of fibrous tissue in the axial skeleton and the development of inflammation at the entheses (enthesitis).
Histological changes in thigh muscle biopsies of AS patients have previously been reported  and type III collagen is reported to be present in muscle tissue  and type IV collagen is present in basement membranes, basal lamina, retina and cornea. AS patients often suffer from acute anterior uveitis (AAU), which could be triggered by some of these anti-collagen antibodies. The antibody binding to pulA, type I and type IV collagen demonstrated here, as well as the suggested cross reactivity between pulA and collagen type III may explain some of the pathological features of AS.
The potential role of B*2705 and the other subtypes of HLA-B27 in the initiation of an immune response, in tissue typed patients and tissue typed controls, with an intestinal infection with a gram negative bacteria such as K pneumoniae needs to be investigated further. However, studies performed on tissue typed AS patients against a homologous MHC background in tissue typed control subjects, have shown elevations in antibody to K. pneumoniae when compared to the control patients . The fact that both the disease and control populations were tissue typed for HLA-B27 precluded the potentiality for the responses seen in the test groups to be due to the humoral immune response against K. pneumoniae to be influenced by the MHC class.
AS patients are usually treated by the administration of anti-inflammatory drugs. However, other treatments such as special diets have been proposed. Patients treated with a low starch diet in an open study over a 9-month period demonstrated a decrease in acute phase reactants (ESR and C reactive protein levels) and serum IgA levels . It is possible that this low starch diet may affect the bowel flora by substrate depletion. This may decrease or even prevent the induction of substrate dependent enzymes and proteins such as those in the pullulanase system thereby possibly modifying the disease outcome. The involvement of inducible enzymes and proteins in bacteria related to rheumatic disorders such as AS, may explain the observed patterns of exacerbations and remissions.
RA patients not AS or healthy controls. Importantly, there is no significant difference in the antibody binding capacity of P. mirabilis when grown in the presence or the absence of pullulan. regardless of whether the sera is from patients or healthy controls (Fig. 2B).
IgA and IgG antibody litres were also found to be elevated
Acknowledgements: M.F. acknowledges the support of the BBSRC. We would like to thank Prof. R.G. Price for helpful discussions during the preparation of this manuscript.