Journal list menu

Volume 585, Issue 23 p. 3681-3688
Open Access

Citrullination of autoantigens: Upstream of TNFα in the pathogenesis of rheumatoid arthritis

Anne-Marie Quirke

Anne-Marie Quirke

Kennedy Institute of Rheumatology, Imperial College London, London W6 8RF, UK

Search for more papers by this author
Benjamin A.C. Fisher

Benjamin A.C. Fisher

Kennedy Institute of Rheumatology, Imperial College London, London W6 8RF, UK

Search for more papers by this author
Andrew J. Kinloch

Andrew J. Kinloch

Section of Rheumatology, Department of Medicine and Knapp Center for Lupus and Immunological Research, University of Chicago, 5841 S. Maryland Ave, Chicago, IL 60637, USA

Search for more papers by this author
Patrick J. Venables

Corresponding Author

Patrick J. Venables

Kennedy Institute of Rheumatology, Imperial College London, London W6 8RF, UK

Corresponding author. Address: Kennedy Institute of Rheumatology, Charing Cross Campus, Imperial College London, London W6 8LH, UK. Fax: +44 20 8383 4499.Search for more papers by this author
First published: 20 June 2011
Citations: 43


The discovery of autoimmunity to citrullinated protein/peptide antigens (ACPA) has led the concept that ACPA may be the essential link between disease susceptibility factors and the production of TNFα, which ultimately accounts for the disease phenotype. In this review we will consider (1) the mechanisms of citrullination, both physiological and pathological, (2) how known genetic and environmental factors could drive this peculiar form of autoimmunity and (3) how the immune response could lead to excessive production of TNFα by the synovial cells and ultimately to the disease phenotype (Fig. 1).

1 Protein citrullination

Citrullination (also commonly referred to as deimination) is the post-translational modification of the positively charged amino acid arginine, to a neutral citrulline. Citrullination in the context of a peptide backbone is catalysed by peptidylarginine deiminase (PAD) enzymes. The conversion of a positively charged peptidylarginine sidechain to a comparatively neutral peptidylcitrulline can alter the three-dimensional structure of the protein and its solubility in water. This is important in generating structural proteins, but in the context of RA, may lead to the breaching of immunological tolerance as the neo-epitopes that could be generated may not be expressed in the thymus or bone marrow during lymphocyte selection.

Citrullination has numerous essential physiological roles in a variety of cells and tissues in the body. Citrullination of structural proteins such as pro-filaggrin and keratin in the skin facilitate proteolysis and cross-linking of the proteins which contributes to cornification [1, 2]. In the nervous system, citrullination of myelin basic protein (MBP) is essential for the electrical insulation offered by myelin sheaths [2, 3]. When trichohyalin, a structural protein that bundles cytokeratin filaments, is citrullinated, it contributes to the maturation of hair cuticle cells [1, 4]. Citrullination also controls the functions of histones as support structures and transcriptional control elements for DNA [5, 6]. Furthermore, hypercitrullination of histones is needed for formation of neutrophil extracellular traps, part of the innate immune system response to bacterial infection [7, 8].

The activity of PAD enzymes is dependent on high concentrations of calcium (Ca2+) [9]. As the Ca2+ concentrations required for PAD activity are 100 fold higher than those present in intact unstimulated cells, citrullination is likely to occur in conditions which lead to mobilisation of free intracellular calcium, such as chemokine receptor ligation, apoptosis, necrosis and cellular differentiation [10]. The high calcium concentration needed in vitro also suggests activating mechanisms that may modulate this requirement in vivo [11]. However peptidylarginine deiminases may preferentially deiminate peptidylarginine bearing extracellular substrates such as cytokines, collagen, fibronectin and fibrinogen in the extracellular environment, where calcium concentrations are more optimal for their activity.

Five PAD enzymes (PAD1–4 and PAD6) have been identified in humans, which are all encoded by a single gene cluster on chromosome 1p35–36. PAD homologs for some or all of these enzymes have also been found in other eukaryotes, with similar genomic organisation across species [3, 12]. However, a prokaryotic enzyme with PAD activity has only been described in one bacterium to date, Porphyromonas gingivalis [13, 14], which is a major pathogen in periodontitis. This enzyme shows very limited sequence homology with human PADs, but shares a common membership of the guanidino-group modifying enzyme superfamily [15]. Unlike the human PAD enzymes, P. gingivalis PAD (PPAD) is not Ca2+ dependent, can convert free l-arginine [15] and only citrullinates carboxy-terminal arginines [13, 14].

Human PAD enzymes show a characteristic tissue distribution. PAD1 is predominantly expressed in the epidermis. PAD2 is the most widely expressed, with the highest levels found in skeletal muscle, secretory glands, brain and spleen. PAD3 is expressed in hair follicles and the upper epidermal layer. PAD4 is expressed in white blood cells and most well characterised in granulocytes. PAD6 is only found in eggs, ovaries and in early embryos [3, 16]. With the exception of PAD4, which has a nuclear localisation sequence, the intracellular presence of PAD enzymes is restricted to the cytoplasm.

PAD2 and 4 are the family members that have been studied most extensively in the context of RA. This is because their tissue distribution would predict that both enzymes would be found within the joint, which is enriched for myeloid cells. Indeed, both PAD2 and PAD4, together with citrullinated proteins, have been demonstrated in the synovial fluid and the synovial membrane [17, 18]. Therefore, at present, PAD2 and PAD4 are the strongest candidates for generating the citrullinated antigens that are targeted in the rheumatoid joint. Several reports have demonstrated that citrullinated proteins accumulate at sites of inflammation [19-21]. Citrullinated fibrin and vimentin have been observed in the synovium (outlined in [22]). We demonstrated citrullinated α-enolase detectable at similar levels in the synovial fluid from patients with RA and spondyloarthritis [18], but undetectable in non-inflamed osteoarthritis samples. Importantly the autoantibody response to the citrullinated α-enolase was restricted to patients with RA.

2 Induction of the autoantibody response to citrullinated proteins

The classical paradigm for induction of autoreactivity in any autoimmune disease is the interaction of genetic susceptibility factors with environmental factors to produce a surprisingly limited repertoire of disease-specific antibodies, which cause the associated pathology. In the case of RA, characterisation of the autoimmune response to specific citrullinated proteins has done much to unravel this gene/environment/autoimmunity triad.

2.1 Autoantibodies to citrullinated proteins in RA

The presence of rheumatoid factor – RF (an antibody reactive with the Fc portion of IgG) in RA individuals contributed towards RA being termed an ‘autoimmune’ disease, and has been a component of the classification criteria for many years [23]. However, the presence of RF is not specific for RA, but is thought to be the consequence of immune activation [24, 25]. The second-generation cyclic citrullinated peptide (CCP) assays, now widely used in diagnostic laboratories and included in the new 2010 ACR/EULAR classification criteria, evolved by selecting randomly generated citrulline-containing peptides from a large panel which were tested against RA and control serum, with the sequences giving the best discrimination in diagnostic sensitivity and specificity being adopted for clinical use. Anti-CCP antibodies have been detected prior to the development of clinically apparent RA [26] and are associated with more severe and erosive disease (reviewed by Zendman et al. [27]). In spite of being a powerful diagnostic tool, the cyclic citrullinated peptides used in the CCP assay are of limited use for understanding the disease aetiology and pathogenesis of RA as they do not correspond to in vivo generated citrullinated proteins. However, four citrullinated proteins that are targeted by anti-citrullinated protein antibodies, and are present in the joint, are now well established as autoantigens: citrullinated fibrinogen/fibrin [28], vimentin [17], collagen type II [29], and α-enolase [30], with further proteins awaiting identification and characterisation (reviewed by Wegner et al. [14]).

2.2 Genetic factors contributing to the autoantibody response to citrullinated proteins

Amongst the major and best-studied genetic risk factors identified so far for the development of RA is a group of MHC class II alleles, namely HLA-DR4, -DR1 and -DR10, principally DRB10401, 0404, 0408, 0405, 0101, 0102, 1001 and 1402 [31]. All share variants of the Q/R-K/R-R-A-A amino acid motif, termed the ‘shared epitope’ (SE), present in the third hypervariable region of the DRβ1 chain and which constitutes part of the P4 pocket of the peptide binding groove [32]. The SE has been shown by many to be associated with the ACPA positive subset of RA [33, 34]. However, there is much speculation in the literature regarding the underlying mechanism of the SE-RA association. Hypotheses include a direct role of the SE on increased affinity and presentation of autoantigens and subsequent activation of self-reactive T cells [35], decreased activation of regulatory T cells [31], altered thymic T cell repertoire selection [36], and the SE being an innate immune system activator [37, 38]. Another genetic risk factor for ACPA positive RA, but also for other autoimmune diseases, is the susceptibility allele 620W of PTPN22, a gene which encodes a tyrosine phosphatase involved in T and B cell signaling [39, 40]. A gene in the tumor necrosis factor receptor-associated factor 1-C5 (TRAF1-C5) region has also been shown to associate with ACPA-positive RA [41]. Polymorphisms in the human PADI4 gene were first associated with RA in a Japanese cohort [42]. Patients with the susceptibility haplotype were more likely to be anti-CCP antibody positive. The association with RA has been confirmed in a number of other Asian cohorts [43-45], although independent of anti-CCP status in a large Korean cohort [46]. Conflicting findings have been reported with Caucasian cohorts [47-51] and a large case-control study of over 5500 UK Caucasian patients showed no association [52]. The reasons for this discrepancy are unclear.

2.3 Environmental factors contributing to the autoantibody response to citrullinated proteins

Environmental factors are considered to contribute to the onset of RA in a genetically predisposed individual. Evidence mainly stems from the low disease concordance rate (15%) in monozygotic twins [53] and the declining incidence of RA in genetically predisposed populations such as the Pima Indians [54, 55]. Although a number of environmental exposures have been linked to RA including smoking, periodontitis, hormonal factors and exposure to silica [56, 57], smoking is the most clearly established [58-61].

2.3.1 Smoking

The link between RA and smoking was first recognized in 1987 [62] as an unexpected finding in a study investigating the association between RA and the use of oral contraceptives, and later confirmed in a number of case-control and cohort studies (reviewed by Sugiyama et al. [63]). The most striking results were from the Arthritis and Rheumatism Council Twin Study, where 13 pairs of monozygotic twins, discordant for RA and smoking, were identified, and in 12 out of 13 cases the RA patient was also the smoker [59].

Further evidence of a link between smoking and RA has been provided by studies of gene-environment interactions. Klareskog and colleagues were the first to report that both smoking and the SE were risk factors for anti-CCP antibody positive RA but not anti-CCP negative and that the presence of both factors was associated with a striking odds ratio for RA development [33]. They proposed that smoking, in the presence of HLA-DR SE alleles, activates an antigen specific autoimmune response to citrullinated proteins, which ultimately leads to the development of RA. We recently observed that this gene-environment interaction was even stronger for the subset of anti-CCP positive patients that also carried antibodies to the immunodominant epitope of citrullinated α-enolase (CEP-1) [64]. However others have reported similar findings with an overlapping subset defined by antibodies to a citrullinated vimentin peptide. It therefore seems likely that the smoking and shared epitope interaction facilitates immunity to multiple different citrullinated epitopes and so broadens the immune response in RA [65].

Exactly how smoking contributes to RA, and why smoking is linked to the development of ACPA positive RA in a particular genetic background, is the subject of ongoing investigations (Fig. 1 ). Cigarette smoke contains thousands of toxic compounds including cyanide [66] which may cause tissue damage and inflammation and modulate immune cell function and cytokine production [67, 68]. An increased frequency of citrullinated protein in bronchoalveolar lavage (BAL) cells has been shown in the lungs for healthy smokers and smokers with pulmonary inflammation [69]. The increase of citrullinated protein was associated with increased expression of PAD2 in the bronchial mucosal and alveolar compartment, which leads to the possibility of smoking increasing citrullination due to the increased expression of PAD enzyme in BAL cells. The cyanide present in tobacco smoke could also contribute to the onset of RA by increasing tissue thiocyanate ions as a result of its metabolism, or its reaction with protein disulphide bonds and the subsequent removal of thiocyanate from intermediate under basic conditions [70]. Thiocyanate can undergo enzymatic reactions catalysed by myeloperoxidase that eventually lead to the formation of homocitrulline. Homocitrulline is homologous to citrulline in terms of shape and charge, the only difference being an extra carbon in the side chain. It is frequently found at sites of inflammation and atherosclerotic plaques, both of which are more common in smokers [71]. Although homocitrullination is an attractive hypothesis for explaining a direct relationship between RA and smoking, the remarkable disease specificity of ACPA in RA has not yet been replicated using homocitrullinated peptides as substrates for detecting autoantibodies in human disease. Importantly, one should also bear in mind that in vitro smoking studies use different compounds, cells, and experimental conditions to mimic smoking, and may only partly reproduce the true short- and long-term physiological events accompanying tobacco smoking.

figure image
Hypothetical etiological model for the development of ACPA-positive RA illustrating mechanisms of citrullination and tolerance breakdown in the lungs and oropharynx followed by the effector phase of pathogenesis in the joint.

2.3.2 Porphorymonas gingivalis infection

Besides smoking, P. gingivalis is another environmental agent that has been linked to RA [72]. Periodontitis, in which P. gingivalis is a major pathogen, arises as a result of inflammatory responses to the accumulation of bacteria on tooth surfaces adjacent to the supra- and sub-gingival tissues. Periodontitis and RA share a number of common predisposing genetic and environmental risk factors. For example, the DRB1∗04 SE subtypes (0401, 0404, 0405 and 0408) that predispose to ACPA positive RA have also been associated with the development of severe, rapidly progressive periodontitis [73]. Furthermore, in patients with coexistent rheumatoid arthritis and periodontitis, both articular and periodontal bone erosions are associated with the SE [74]. Periodontitis and RA also share common inflammatory mechanisms.

In addition to the genetic and environmental susceptibility factors shared by RA and periodontitis, the action of the P. gingivalis PAD enzyme (PPAD) elevates levels of citrullinated proteins and may have a role in priming autoimmunity in a subset of patients with RA [75]. P. gingivalis produces many virulence factors including extracellular cysteine proteases (called gingipains), haemagglutins, Lipopolysaccharide (LPS), and fimbrae, which enable the bacterium to colonize and invade periodontal pockets [76]. The proteolytic processing of P. gingivalis gingipains generates carboxy-terminal arginine residues that facilitates the generation of endogenous c-terminal citrullinated peptides by PPAD [77]. P. gingivalis is also capable of citrullinating host human peptides at the site of gingival inflammation [77]. Wegner et al. (2010) demonstrated in this study that PPAD citrullinates human fibrinogen and α-enolase peptides after proteolytic cleavage by P. gingivalis gingipains. P. gingivalis gingipains can also cleave and activate the human proteinase-activated receptor-2 (PAR-2) on human neutrophils, which increases intracellular calcium concentrations [78]. The human PAD enzymes, which require a calcium rich environment for activation, could subsequently citrullinate peptidyl arginines.

We have recently tested the hypothesis that citrullination of peptides from endogenous proteins generated by P. gingivalis could be one source of antigen that prime ACPA production in individuals who subsequently develop RA. Preliminary data suggest that these peptides generated by P. gingivalis are not as immunoreactive with autoantibodies as their equivalent internally citrullinated counter-peptides (Quirke AM et al., unpublished observations). However, it remains possible that these peptides could breach tolerance to citrullinated proteins long before the development of RA.

2.4 Mechanisms of tolerance breakdown

What distinguishes RA from other inflammatory joint diseases is not the presence of synovial citrullination but the immune response to citullinated proteins. The conditions required to trigger this immune response in a physiological context are not well known. It is often assumed that a T cell response to citrullinated peptides is necessary. This would require the uptake and processing of citrullinated antigens by antigen presenting cells (APCs) with subsequent presentation by HLA-DR molecules containing the SE binding motif. The interaction of the peptide/SE complex with specific CD4+ T cells in the context of co-stimulation would lead to the perpetuation of inflammation.

Hill et al. [79] tested the hypothesis that citrullination might evoke an autoimmune response by studying the T cell response to citrulline-containing peptides in HLA-DRB1∗0401 transgenic (DR4-IE tg) mice. They found that citrullination of a vimentin-related peptide, one of the potential target autoantigens found in the joint, dramatically increased peptide-0401 binding (100-fold) and led to the activation of CD4+ T cells. The target arginine/citrulline of vimentin was positioned at the P4 peptide-anchoring pocket. Feitsma et al. [80] identified two naturally processed peptides from citrullinated vimentin that induced a citrulline specific T cell response following immunisation, and which also stimulated T cells from RA patients. These T cells might in turn provide B cell help and stimulate the production of ACPA. Another study using the same DR4-IE tg mice found that citrullinated peptide sequences of the α- and β-chains of a second target autoantigen, fibrinogen, had a higher binding affinity to DR4 compared to the uncitrullinated peptides. The significance of this was demonstrated in vivo in an animal model where immunisation of the DR4-IE tg mice with fibrinogen citrullinated in vitro induced a progressive arthritis, which was not seen in the wild type mice or with the transgenic mice immunised with native fibrinogen [81]. Citrulline-specific-DR4 restricted T cell responses against the naturally processed vimentin peptides present in the inflamed joint were also demonstrated in the study. These studies suggest that citrullinated peptides, capable of binding to HLA-DR SE can elicit an autoimmune response which is then maintained by reactivity with citrullinated proteins within the joint.

Verification of the pathogenicity of citrulline specific T cells in mice is complicated by several factors. Firstly, citrullinated proteins do not accumulate in the normal healthy joint. Secondly, murine MHC and autoantigens vary between mouse and man, which means that different T cell clones exist in a naïve mouse and a healthy patient prior to a breach in immunological tolerance. Therefore, the absence of arthritis in the DR4-IE mice immunised with citrullinated vimentin is likely to be, at least in part, due to the absence of citrullinated vimentin in healthy mouse joints. However, in the study by Hill et al. [81], the authors overcame both of these hurdles by transferring T cells from DR4-IE mice (immunised with in vitro citrullinated human fibrinogen) into naïve recipients, who had been given intra-articular human or mouse fibrinogen. The arthritis could only be transferred when the recipient mouse received intra-articular citrullinated fibrinogen. This supports the previously suggested hypothesis [10] that in order for either citrulline specific T or B cells to become pathogenic, either the synovium must itself become inflamed, or citrullinated proteins produced elsewhere must be deposited there.

P. gingivalis may play a role in tolerance breakdown by promoting ACPA production to self citrullinated peptides and host citrullinated peptides. Lundberg et al. [75] showed that RA patients with a humoral immune response to CEP-1 had antibodies also reactive with a CEP-1 enolase orthologue from P. gingivalis. This suggested that P. gingivalis infection could break tolerance by citrullinating bacterial enolase, with the subsequent antibody response cross-reacting with citrullinated human α-enolase and initiating ACPA production [75]. Hitchon et al. [82] revealed an association between the IgG immune response to P. gingivalis lipopolysaccharide and ACPA in a genetically predisposed population of North American Native individuals with RA. Our research group (Kinloch et al. [83]) recently tested the arthritogenicity of citrullinated enolase from both humans and P. gingivalis using the same DR4-IE transgenic mouse model used by Hill et al. [81] and control mice (MHC class II knockout and C57BL/6). Both the human and P. gingivalis citrullinated enolase induced arthritis in the DR4-IE mice. However, antibodies reacting with citrullinated and uncitrullinated enolase, as well as antibodies to CEP-1 and the arginine control peptide, were found in all groups apart from the MHC class II knockout mice. This suggests that enolase is different in that it is the uncitrullinated molecule that may break tolerance in the context of the DR4 risk alleles and provides a novel model for investigating an etiological role for P. gingivalis in RA.

Smoking may breach tolerance through the generation of homocitrullinated proteins. Recent evidence in support of this hypothesis has shown that immunisation of mice with homocitrulline-containing peptides induced anti-homocitrulline antibodies as well as the proliferation and chemotaxis of CD4+ T cells and the production of proinflammatory cytokines [84]. The authors also found increased levels of both homocitrullinated peptides and citrullinated-peptides in patients with erosive arthritis and importantly the level of homocitrullinated peptides was found to be greatest in patients with erosive arthritis that were also anti-CCP positive. It has also been shown that some positive results with anti-CCP2 assays possibly reflect cross-reactivity with homocitrulline, since animals immunised with citrullinated and homocitrullinated proteins both developed anti-CCP antibodies [85].

3 Effector mechanisms

We have indicated how known genetic and environmental factors could drive citrulline autoimmunity. However, smoking and/or P. gingivalis infection are not sufficient for the mature autoimmune response. Production of antibodies to citrullinated peptides occur before onset of disease and are a major feature in substantially increasing the risk of developing RA. This has been supported by mouse data where administration of ACPA (raised against citrullinated fibrinogen antibodies) alone did not cause arthritis. However, mice subjected to mild CIA developed a more severe arthritis after administration of ACPA [86]. Clinical studies have demonstrated how patients with an undifferentiated arthritis that are ACPA positive are much more likely to progress to RA (as defined in the 1987 ACR criteria) than ACPA negative patients [87, 88]. An important step following tolerance breakdown to specific citrullinated peptides generated outside the joint may be epitope-spreading to other host citrullinated proteins in the inflamed joint, ultimately giving rise to immune complexes with ACPA and RF. Epitope spreading describes an increase or shift in antigen recognition during the course of an immune response [89]. It has been observed in the DR4 RA mouse model induced by immunisation with citrullinated fibrinogen [79, 90], and in a study of pre-RA sera [91] .

3.1 Immune complex formation in the inflamed joint

Recent work has elucidated many of the mechanisms by which ACPA could lead to downstream TNF production and synovial inflammation. Immune complexes formed from solid-phase citrullinated fibrinogen and ACPA from human sera, were able to stimulate TNF production from macrophages [92]. Fibrinogen was immobilised in this model to replicate the synovial fibrin deposits observable in RA. However, circulating immune complexes containing citrullinated fibrinogen can also be identified in more than half of patients with anti-CCP antibodies [93]. Whether these are capable of stimulating macrophages/monocytes in the same way is unclear, since Mathsson reported that only immune complexes derived from synovial fluid and not from blood were able to elicit TNF from peripheral blood mononuclear cells [94]. They did not identify the complexed antigens however. The effect on TNF production in these systems appears dependent on Fcγ receptor IIa (FcγRIIa) binding, which is consistent with the increased susceptibility to collagen-induced arthritis in mice transgenic for human FcγRIIa [95]. It seems likely that other citrullinated antigens besides fibrinogen are involved and, using a proteonomic approach, van Steendam also identified citrullinated vimentin in synovial fluid immune complexes [96].

Macrophages may also be activated by Toll-like receptor (TLR) binding. TLRs comprise part of the innate immune system and are capable of recognising conserved microbial sequences as well as endogenous ligands such as RNA and DNA. Fibrinogen has previously been demonstrated to bind to TLR4 [97, 98], but citullination of fibrinogen enhances its ability to stimulate TNF production from macrophages. Co-ligation of both FcγR and TLR4 by immune complexes containing citrullinated fibrinogen further increases TNF [99]. Interestingly, RA synovial macrophages appear more susceptible to activation by TLR 2 and 4 binding, than macrophages from patients with other arthritides or healthy controls [100], and there is preliminary evidence that blockade of TLR4 may improve the signs and symptoms of RA [101].

ACPA may also stimulate macrophages independent of Fcγ and TLR4 binding. Lu and colleagues identified the chaperone glucose-regulatory protein 78 (Grp78, also known as binding immunoglobulin protein; BiP) as being present in a citrullinated form on the surface of monocytes. This was a target for antibodies present in affinity-purified ACPA from pooled RA sera which, when bound, induced TNF expression [102].

3.2 TNFα production leading to joint destruction and inflammation

ACPA mediated immune complexes are capable of causing imbalances in the regulation of cytokines and other inflammatory mediators leading to the inflammatory bone erosion characteristic of RA. This osteo-destructive feature of chronic inflammatory arthritis is a major cause of disability in patients with RA. Besides TNF-α, several other cytokines play a role in RA pathophysiology, such as interleukin-1 (IL-1), IL-6, IL-15, IL-17, IL-18, IL-32 and IL-33, leading to activation and recruitment of inflammatory cells (reviewed in [103]). However the importance of TNF-α in the mechanism of rheumatoid inflammation was indicated in a study by Brennan et al. [104] where addition of TNF-α-specific antibodies at the beginning of cell culture abrogated IL-1 production (which has been shown to be involved in joint damage in experimental situations). Consistent with this dominant role of TNF-α in vitro, Kollias and colleagues (1991) demonstrated in vivo that the overexpression of TNF-α in transgenic mice was sufficient to cause arthritis [105]. The widespread regulatory effect of anti-TNF antibodies, including the downregulation of expression of other pro-inflammatory cytokines, such as granulocyte-macrophage colony-stimulating factor (GM-CSF), IL-6, and IL-8 have been shown in subsequent experiments [106-108].

As well as increasing expression of other pro-inflammatory cytokines and perpetuating the inflammatory process, TNF-α production (in association with the other proinflammatory cytokines) activates synovial fibroblasts and macrophages to produce cartilage-destructive enzymes, such as metalloproteinases (MMPs) and expression of bone-destruction related molecules, such as RANKL (receptor activator of NFκ B ligand) causing the cartilage and bone erosion characteristic of RA [103]. The role of two MMPs, MMP1 and MMP3, are of particular significance in RA as they can degrade all the important structural proteins in the extracellular matrix of cartilage [109], and are found at elevated levels in the serum of RA patients, which is thought to originate from the synovium [110].

TNF-α also plays an important role in other pathways that contribute to bone erosion in RA. One mechanism involves activated T cells, long known to stimulate osteoclast formation [111]. The requirement of osteoclasts in joint destruction in RA was demonstrated in an experiment where transgenic mice that expressed human TNF-α and that developed a severe and destructive arthritis were crossed with osteopetroric mice completely lacking osteoclasts [112]. Although the mice exhibited severe inflammatory changes, they did not experience any bone destruction, demonstrating that TNF-dependent bone erosion is mediated by osteoclasts. Another driver of bone erosion in RA involves TNF-α in association with IL-1 and probably IL-6 driving RANKL expression and release from T-cells, osteoblasts and synovial fibroblasts [109]. RANKL stimulates the differentiation of monocytes to osteoclasts, which can remove mineral as well as matrix and subsequently cause bone destruction. RANKL activates the receptor antagonist of NFκ B (RANK) on the surface of osteoclast precursors, and this interaction is crucial for osteoclast formation [113]. Dougall et al. [114] showed that mice unable to express either RANKL or RANK are osteopetrotic because of a complete lack of osteoclasts. TNF also regulates synovial angiogenesis, which may be critical in maintaining an invasive mass of inflammatory synovial tissue, and anti-TNF agents downregulate both vascularity and the pro-angiogenic vascular endothelial growth factor (VEGF) [115].

4 Concluding remarks

The cardinal features of RA, inflammation and joint destruction, can thus be accounted for by the chronic over production of TNF-α. The dominance of TNF in the hierarchy of the cytokine cascade and the remarkable therapeutic advances with the introduction of TNF-α inhibitors has led to a focus on effector mechanisms when considering the pathogenesis of RA. We have summarised recent evidence that overproduction of TNF-α can be caused by immune complexes containing citrullinated antigens and their antibodies. This implies that cirullination itself is critical to the cause of the disease and that these events, well upstream of TNF-α in the pathogenic pathways of RA hold the key to our understanding of aetiology and ultimately prevention and cure of this disease.