This mouse model of Rheumatoid Arthitis (RA) was developed completely by accident during studies of thymocyte differentiation that had nothing to do with autoimmunity. The authors inserted on the C57BL6 background (I-Ab) a T cell receptor (TCR) transgene that recognized a peptide from bovine RNase in the context of I-Ak(KRN TCR transgene). By chance this transgenic mouse, which was healthy and without any apparent abnormal phenotype, was crossed with a NOD mouse (I-Ag7). Surprisingly all F1 mice, named K/BxN, developed severe joint swelling that started at about 7-8 weeks of age and gradually progressed until the overall motility of the animal was impaired. The arthritic phenotype, which was observed in a very large number of animals (around 800), reminded the authors of the human disease RA.

In a first series of experiments the authors showed how similar the mouse disease is to the human RA. Like RA, the joint disease in the mouse was chronic and progressive, affected symmetric joints, and had a proximal-to-distal gradient of severity. Histologically, the joints showed synovitis (infiltration of the synovial by inflammatory cells), pannus formation (a mass of synovium consisting of inflammatory cells, granulomatous tissue and fibroblasts which causes erosion of the underlying cartilage) and ultimately bony ankylosis. In serum, there was overproduction of IL-1, IL-6 and TNF, and a hypergammaglobulinemia mainly due to increase in IgG1. Unlike RA, the distal interphalangeal joints were affected, there were no extra-articular manifestations, with the exception of splenomegaly, the disease course was more rapid, and rheumatoid factor was not present (rheumatoid factor is an antibody directed against the Fc portion of an autologous IgG, and it is present in about 80% of RA patients). So this mouse is not a perfect model of RA, and probably models only of a subset of patients with RA, which is itself a heterogenous disease.

In the second series of experiments the authors asked what activates the transgenic T cells. In fact, the F1 cross that develops arthritis lacks the specific antigen against which the T cells were directed (bovine RNase), and also the restricting I-Ak molecule (the F1 cross expresses co-dominantly the parental class II alleles: I-Ab and I-Ag7). So, evidently, the transgenic T cells recognize something different from bovine RNase in an MHC context that is not I-Ak. The authors indeed showed in vitro that the TCR transgenic T cells were activated by NOD antigen presenting cells that carry a self peptide on their MHC class II molecules (I-Ag7).

The authors next asked what are the MHC class II antigen-presenting cells that activate the transgenic T cells (dendritic cells, macrophages or B cells)? Using bone marrow chimeras, they found surprisingly that the cells required for disease production are the B cells (not the dendritic cells or the macrophages). The main and best characterized function of B cells is to produce immunoglobulins. B cells, however, are also involved in formation of lymphoid structures (through lymphotoxin production), cytokine production, and antigen presentation through the MHC class II pathyway. B cells actually present peptides on their class II MHC molecules much more efficiently when the presented peptide is the same as the peptide recognized by their B cell receptor (and thus by the secreted antibody). The authors demonstrated that the key B cell function responsible for disease is indeed the production of pathogenic immunoglobulins. This was shown by transferring serum from the arthritic mice into NOD mice that have the KRN T cell receptor transgene but that do not have mature B cells (owing to a disruption of the m chain gene, mMT mutation). The simple transfer of serum in the B cells deficient mice was capable of producing severe arthritis, in a very rapid (within 1 day after transfer) and effective (as little as 100 ul of serum required) fashion. Not only were the B cells not needed in the recipient for disease development, but also the transgenic T cells were not needed. In fact, the authors could transfer the arthritic serum into C57BL6 mice that do not have mature B and T cells (owing to a disruption of the recombinase activating gene), and still showed the same disease in the recipients. This tells us that the arthritogenic immunoglobulins are the mediators of disease. In other words, once these arthritogenic immunoglobulin are formed they can mediate disease without further input from the adaptive immune response (T or B cell arm). Actually this transfer system works in all sorts of recipient strains, so it is a very convenient system to test hypotheses and involvement of specific pathways.

In the next series of experiments the authors asked what self peptide is recognized on the surface of MHC class II expressing B cells by the transgenic KRN T cells, recognition that then leads to the production of immunoglobulin directed against the same peptide. With an open-minded approach, the authors looked for this peptide not only in joint tissue, but also in many other tissues (such as kidney, liver, spleen, etc). They extracted proteins from these tissues, separated them in a polyacrylamide gel electrophoresis, and transferred them to solid membranes. The membranes were then incubated with sera from the arthritic mice and from control mice. They consistently observed that the arthritic sera bound to a 60 kDa protein band present in all tissues (not only the joints), whereas the control sera did not. To characterize further this 60 kDa band, protein extracts from the kidneys was run through a column that had bound immunoglobulins from the arthritic mice (or immunoglobulins from control mice). The protein bound to the column was then eluted and run on a gel. They showed that the 60 kDa protein was coming only from the arthritic column, and not from the control column. The 60 kDa band was cut out of the gel, the proteins from that band were eluted and digested, and the resulting peptides submitted to sequencing. They obtained four peptides, all coming from the same protein: glucose-6-phosphate isomerase (GPI), an enzyme of the glycolytic pathway which reversibly catalizes the conversion of glycose-6 phosphate to fructose-6 phospate. To confirm that GPI was the autoantigen recognized by the KRN transgenic T cells, and subsequently by the immunoglboulins, GPI was expressed in bacteria by recombinant DNA technology. The recombinant GPI was shown to be recognized specifically by the arthritic sera. To prove that the anti-GPI activity was pathogenic, the authors ran arthritic sera over a column that had bound recombinant GPI or over a control column, and used the flow throughs to inject into mice. The flow though from the GPI column lost the ability to transfer arthritis in the recipient mice, activity that instead was still present in the flow through from the control column.

Thus, the kinetics of arthritis are as follows: transgenic KRN T cells recognize a GPI peptide presented on the MHC class II molecules of B cells. This T-B cell interaction stimulates B cells to produce antibodies directed against GPI, which are the mediators of disease. GPI is expressed in the cytoplasm of all cells and is required for life (in fact mice knock out for GPI are embryonic lethals).

GPI is expressed in all B cells, but only the B cells whose B cell receptor is specific for GPI will proliferate very strongly upon interaction with the transgenic KRN T cells.

In the next series of experiments the authors investigated the actual mediators that, when activated by the GPI antibodies, induce arthritis. Using the serum transfer model, they showed that the arthritis caused by anti-GPI antibodies requires several mediators, such as IL1 and TNFa, neutrophils, mast cells (there is actually specific mast cells degranulation within two hours post transfer), the FcgRIII pathways and the alternative pathway of complement. Indeed the involvement of the complement pathway, and C5 in particular, is critical to explain the joint specificity. C5 is central in the complement cascade: the classical pathway leads from C5a and the alternative pathway leads from C5b, resulting finally in the membrane attack complex. It turned out that it is the alternative pathway of complement that is absolutely required for disease development. The alternative pathway, basically a self/nonself discrimination system, is initiated through spontaneous hydrolysis of C3. C3 is abundant in plasma, and it is constantly hydrolyzed to produce C3b (it is constantly "ticking over"). C3b is rapidly inactivated unless it binds to host cells or pathogen surfaces, thus becoming much more stable. Factor I and factor H also can bind to surfaces and thus compete with C3b. Whether the alternative pathway goes to completion or not depends on what prevails: C3b on one hand and H and I on the other. On prokaryotic surfaces the binding of C3b is favored over the binding of H and I. In contrast, on eukaryotic surfaces H and I bind much better, and also eukaryotic cells express complement-regulatory proteins, such as decay-accelerating factor and membrane cofactor of proteolysis, which block the activation of the alternative pathway. The authors showed that pre-treatment of recipient mice with anti-C5 antibodies before transfer completely block the development of arthritis, and that anti-C5 antibodies correct an already established disease.

The authors produced monoclonal antibodies against GPI to test whether arthritis could be transferred using these monoclonal antibodies, rather than whole serum. And it was easy to make many of such monoclonal because they the arthritic mice have a lot of B cells producing GPI antibodies. Hundreds of anti-GPI monoclonal antibodies were pooled in groups of 10 and injected into recipient mice. These antibodies were able to induce arthritis that was identical to the one observed when the whole serum was transferred. The authors then tried to separate in each pool the single anti-GPI monoclonal antibody responsible for the disease transfer. Surprisingly, the singly injected monoclonal antibodies were not able to induce arthritis. And it was the same for about 50 monoclonal antibodies tested alone. They then tried injections of monoclonal in pairs, triplets or quadruplets, and saw that arthritis could be induced by some pairs, but was mild; it was more severe with triplets and very severe with quadruplets. The difference between the pairs of monoclonal that could induce arthritis and the pairs that could not, was that you must have a pair that sees two distinct epitopes on the GPI molecule. If the antibodies in the pair see the same region of GPI they cannot induce arthritis. Overall the authors concluded that what is important is lattice formation. Probably lattice formation is more effective in mobilizing the downstream mediators of disease, such as cytokines, mast cells and complement.

However, knowning what the antigen is and knowing the mechanisms by which GPI antibodies cause arthritis still does not explain the disease specificity: if GPI is expressed in all cells, and the effector cells (neutrophils and mast cells) and molecules (IL1, TNFa, complement alternative pathway) are present everywhere in the body, why the disease is limited to the joints? The answer came from analysis of the joint histopathology. When the authors stained several types of cells with anti-GPI antibodies, they observe a nice cytoplasmic positivity. When the same staining was performed on the joints of a normal mouse, in addition to the cytoplasmic positivity in the cells of the joint, such as the condrocytes, the authors strikingly saw a staining also on the articular (acellular) surface of the cartilage. When the joints of an arthritic mouse were analyzed with the same anti-GPI staining, the positivity was very strong. This positivity was due to immune complex formation, because a triple staining with anti-GPI reagent, anti-immunoglobulin reagent, and anti-C3 reagent showed that all these three stains colocalize. So there is the accumlation of GPI-anti GPI complexes and the productive activation of the complement alternative pathway only on the articular surfaces. The anti-GPI antibodies also bind to GPI that is present in other cells of the body, but here they do not activate complement. The difference between joints and other tissues is that only the joints have extracellular GPI that is recognized by the GPI antibodies. This recognition activates C3 leading to production of C5a, a mediator of inflammation and phagocyte recruitment. The authors believe that this is the key to joint specificity in this mouse model of arthritis. These findings are supported by Wipke et al who used positron emission tomography to show elegantly that purified anti-GPI immunoglobulins, but not control immunoglobulins, specifically localize to distal joints in the front and rear limbs of normal mice witin minutes after their intravenous injection.

Finally the authors investigated the relevance of autoantibodies to GPI in patients with RA. An initial report by Schaller et al from the Scripps Research Institute showed that 64% of patients with RA, but none of controls, have circulating antibodies directed against native rabbit GPI. In contrast, Schubeert et al and Kassahn et al showed that only a small proportions (~3%) of patients with RA and other autoimmune diseases have antibodies againts recombinant human GPI, suggesting that GPI is not an autoantigen in RA. To study this discrepancy the authors tested RA sera against native rabbit GPI or recombinant human GPI. The authors confirm that some patients with RA have anti-GPI antibodies. These findings, however, at the moment rest as a simple correlation. What needs to be done is to show the pathological relevance of GPI antibodies in human RA.