Collagen-induced arthritis (CIA) is an experimental model of arthritis inducible in susceptible strains of mice, for example DBA1/J, by immunization with bovine collagen type II (CII) in complete Freund's adjuvant (CFA) 1234. The development of CIA is known to depend on T cells, and disease susceptibility is linked to the major histocompatibility complex (MHC) region 5. After T cell activation an inflammatory cascade involving T cells, macrophages/monocytes, B cells, and activated synoviocytes, is triggered. Different types of leucocytes and synovial cells produce a complex array of cytokines and other soluble mediators, which are thought to be responsible for cartilage destruction and bone erosion 6789. Some of the main features of disease are synovial hyperplasia and mononuclear cell infiltration. Factors contributing to this phenomenon are unknown; however, an imbalance between rates of cell proliferation and cell death (apoptosis) has been suggested by recent studies of rheumatoid synovium demonstrating that apoptosis of synovial cells and infiltrating lymphocytes was common in situ 1011. In the immune system, apoptosis is involved in development and in the negative selection of lymphocytes. It is also crucial in downregulating immune responses to foreign antigens. Cytotoxic T lymphocytes and other killer cells can eliminate their targets by the induction of cell death. All of these functions are primarily mediated through the receptors Fas/APO-1, tumor necrosis factor receptor 1 (TNFR1; p55 TNFR) and the perforin (pfp)/granzyme pathway 1213.

Perforin is expressed mainly in activated cytotoxic T lymphocytes (CTLs) and natural killer (NK) cells, although some reports suggest its expression in microglia as well 14. In CTLs, pfp is stored in cytoplasmic granules and is a major effector of cytolysis by these cells. On pfp release it inserts itself into the plasma membrane of target cells and polymerizes into pore forming aggregates. Pores of pfp lead to osmotic lysis of target cells and induce apoptosis by allowing granzymes to enter the target cells. Perforin-deficient mice have confirmed its function as an effector molecule and in the immune response to viruses and tumors as well as in other aspects of immune regulation such as activation-induced cell death (AICD), antibody production and spontaneous autoimmunity 151617. NOD mice, an animal model of insulin-dependent diabetes mellitus with a mutation of the pfp gene (NOD/pfp-mice), develop diabetes with highly reduced incidence and markedly delayed onset, pointing to a role of the pfp death pathway in tissue damage in this disease 18.

The role of pfp in arthritis is not clear, although some observations suggest a role in disease pathogenesis, for example pfp-expressing CTLs has been demonstrated in the rheumatoid synovium, and CD8-deficient mice seem to be less susceptible to induction of collagen-induced arthritis 19.

It is conceivable that the pfp/granzyme pathway could contribute to the pathology of arthritis in at least two ways: promotion of autoimmunity by blocking peripheral tolerance and AICD or destruction of target tissues. In this study we attempted to evaluate the role of the pfp-mediated death pathway in the pathogenesis of CIA with the use of pfp-deficient mice (pfp^-/-) by examining the effect of the mutation on the clinical course of disease, immune response to collagen and on joint pathology.

The primary aim of this study was to investigate whether the delayed onset and mild arthritis observed earlier in CD8^-/- mice 19 is due to a lack of cytotoxic ability or a lack of some other aspects of T cell function such as the secretion of cytokines. Indeed, we found that pfp-deficient mice develop disease with reduced incidence, a slightly delayed onset, and significantly decreased severity, suggesting that CTL activity is important for the initiation and maintenance of arthritis in the CIA model. This is in agreement with earlier observations in the insulin-dependent diabetes mellitus (NOD) mouse model. In that model, as in CIA, pfp is a susceptible rather than a protective gene. This is in contrast to other disease models, for example experimental autoimmune encephalomyelitis (EAE), systemic lupus erythematosus (SLE) and autoimmune pancreatitis, in which pfp clearly protects against the disease 16171822. The protective effects of pfp could be explained by its immune regulatory function, namely the killing of autoreactive B cells and functional self-lysis of activated T cells through AICD. Hence, its role in regulation of peripheral tolerance and its role in autoimmunity, which is distinct from, but overlaps, the role of the FasL/Fas and TNF/TNFR pathways, might be the reason for accelerated autoimmunity in pfp-deficient mice in the EAE and SLE models.

In the present study we obtained evidence that pfp can also act as a susceptibility gene rather than a protective gene. A lack of the molecule led to reduced severity throughout the disease, a slightly delayed onset, and a reduced incidence of CIA. However, we also observed a strong anti-collagen B cell response with similar anti-collagen-specific IgG levels in control and pfp^-/- mice, and significantly increased T cell proliferation in response to collagen. This might have been due to a reduced killing and accumulation of autoreactive T cells after activation because of impaired AICD. These findings indicate that a diminished adaptive response to collagen was not responsible for the reduced arthritis.

Previous studies demonstrated a strong involvement of pfp in the control of CD8⁺ T cell homeostasis. Perforin deficiency resulted in enhanced CD8⁺ T cell expansion because of decreased killing of antigen-presenting cells and consequential prolonged stimulation by antigen 2324. Our results support these propositions: we observed a significantly increased T cell proliferation in pfp-deficient mice in comparison with controls. These data also suggest a role of pfp in the regulation of T cell homeostasis. Nevertheless, cytokines produced by cytotoxic T cells are also involved in the regulation of the T cell response. TNF-α, for example, can mediate AICD of CD8⁺ T cells through TNFR1 and TNFR2 25.

The histopathology of CIA in wild-type pfp^+/+ mice of the present study was similar to that seen in previous studies 26. The data are therefore not shown again here. In the present study there were no significant differences between histopathological changes of pfp^+/- and pfp^-/- mice, most probably because of a severe disease in individual pfp^-/- mice. This result argues for an additional involvement of pfp-independent death pathways in joint destruction. Indeed, pfp was upregulated in the inflamed arthritic joints of wild-type mice as shown by RT-PCR (Additional file 1b).

This is similar to our recent finding showing that the FasL/Fas pathway has a proinflammatory role in CIA and an activating function on fibroblasts in vitro 27. Fas-deficient mice developed arthritis that was less severe, probably through a reduced IL-1R1/Toll-like receptor-4 signaling that might contribute to a decreased expression of other cytokines, chemokines and matrix metalloproteinases potentially regulated by this pathway 28. Previous studies also reported a strong involvement of the TNF/TNFR pathway because CIA only developed with a reduced disease incidence, and the severity and neutralization of TNF led to the prevention of arthritis 2930.

Taken together, the results show that in CIA the disease-promoting effect of pfp prevails. It is therefore tempting to speculate that pfp could contribute to arthritis in at least two ways. First, it could promote tissue damage by direct cytotoxic effects through CD8⁺ T cells and NK cells. Second, pfp might have some activating functions on fibroblasts or macrophages, leading to the production of proinflammatory cytokines. Indeed, there are reports indicating that fibroblasts and monocytes can be activated by granzyme A to secrete the proinflammatory cytokines IL-6, IL-8 and TNF-α, which could subsequently severely regulate the inflammatory response 3132.

There are other indications that argue for a prominent role of pfp in arthritis. Perforin was found to be differentially expressed in lymph nodes and joints of DBA/1J and FVB/N (CIA-resistant strain) mice (SM Ibrahim and D Koczan, unpublished observations) and pfp-expressing cytotoxic T lymphocytes and increased apoptosis were observed in the synovia of patients with rheumatoid arthritis 103334. The targeted pfp gene is likely to have a major role in the observed effects, on the basis of the absence of pfp production, although it is possible that other polymorphic genes in the linked region could also contribute to CIA reduction. Indeed, the pfp gene is mapped to a CIA-susceptible locus, Cia8 on chromosome 10. This quantitative trait locus is covered by the 12 cM C57BL/6J fragment including the mutant pfp gene. The Cia8 region contains several candidate genes that have been implicated in the modulation of CIA susceptibility, for example the macrophage migration inhibitory factor Mif 35 and the autoimmune regulator Aire 36. However, the observation that heterozygous littermates and pfp^+/+ mice that were also backcrossed and contained the same or smaller C57BL/6J fragments do not show effects on the disease argues against a major role for another gene.

In summary, the CIA in pfp-deficient mice was mild and showed delayed onset and reduced incidence, but some individual pfp^-/- mice also developed a severe disease. These results suggest that pfp-dependent cytotoxicity is involved in the initiation of tissue damage in arthritis, but that pfp-independent cytotoxic death pathways, for example the FasL/Fas pathway, might also contribute to CIA.

We report that arthritis developed only with reduced incidence, severity and delayed onset in pfp-deficient DBA/1J mice. These findings suggest that pfp-dependent cytotoxicity is involved in the initiation of tissue damage in arthritis but also that one or several other pfp-independent mechanisms, possibly involving FasL/Fas, contribute to the early phase of joint destruction in CIA.

AICD = activation-induced cell death; CFA = complete Freund's adjuvant; CIA = collagen-induced arthritis; CII = collagen type II; CTL = cytotoxic T lymphocyte; EAE = experimental autoimmune encephalomyelitis; ELISA = enzyme-linked immunosorbent assay; FITC = fluorescein isothiocyanate; H & E = hematoxylin and eosin; IFN = interferon; IL = interleukin; MHC = major histocompatibility complex; NK = natural killer; RT-PCR = reverse transcriptase polymerase chain reaction; pfp = perforin; SLE = systemic lupus erythematosus; TNF = tumor necrosis factor.

The author(s) declare that they have no competing interests.

KB, AK and HTR did the experimental work and contributed to writing the manuscript. KB and AK contributed equally to the work. DK, HJT, RH and SMI participated in the design and co-ordination of the study and drafted the manuscript. HJK and HN did the histopathology. EM carried out the T cell proliferation and cytokine assays. All authors read and approved the final manuscript.