Rheumatoid arthritis (RA) is a complex inflammatory disorder associated with synovitis and joint destruction that affects an estimated 1·3 million Americans and causes significant morbidity, a reduced life-span and lost work productivity. The use of biological therapies for the treatment of RA is costly, and the selection of therapies is still largely empirical and not guided by the underlying biological features of the disease in individual patients. The synovitis associated with RA is characterized by an influx of B and T cells, macrophages and neutrophils and the expansion of fibroblast-like synoviocytes, which form pannus and lead to cartilage and bone destruction. RA is associated with synovial production of rheumatoid factor (RF) and anti-citrullinated protein autoantibodies (ACPA) and with the production of inflammatory cytokines, including interleukin (IL)-1, IL-6, IL-17 and tumour necrosis factor (TNF)-α, which are targets for RA therapeutics. Recent ideas about the pathogenesis of RA emphasize a genetic predisposition to develop RA, a preclinical phase of disease that is associated with the production of ACPA and the development of symptomatic disease following inflammatory initiating events that are associated with expression of citrullinated epitopes in the joints of patients. However, we still have a limited understanding of the cytokine and intracellular pathways that regulate ACPA levels. In humans, therapy with biological agents affords a unique opportunity to better understand the cytokine and signalling pathways regulating ACPA levels and the impact of ACPA level changes on disease activity. In this study we summarize the effect of RA therapies on ACPA levels and B cell responses.
The shared epitope and anti-citrullinated protein antibodies (ACPA)
During the past decade, research and clinical care of RA has been revolutionized by the discovery that many rheumatoid arthritis (RA) patients have ACPA . The discovery of ACPA led to the development of a test [anti-cyclic citrullinated peptide (anti-CCP)] for the diagnosis of RA; the anti-CCP test is highly specific and has comparable sensitivity to rheumatoid factor (RF) for the diagnosis of RA [2-4]. Two human leucocyte antigen (HLA) class II alleles, DRB1*0401 and DRB1*0404, account primarily for the DR4 association with RA in Caucasians. In addition, a DR1 allele is present in many Caucasian RA patients who are negative for DR4 alleles. All the RA-associated HLA alleles share a region of highly similar amino acid sequences (amino acids 67–74) called the shared epitope (SE), which has been postulated to control susceptibility to disease . Among the important potential pathogenic peptides presented by the SE are citrullinated peptides. Studies in mice carrying the HLA DRB1*0401 transgene have demonstrated that converting an amino acid residue from arginine to citrulline leads to enhanced T cell activation and increased binding of peptides by the SE .
In the model of RA pathogenesis proposed by Klareskog and colleagues , smoking plays a prominent role as an inducer of protein citrullination that fosters the development of ACPA in patients expressing HLA-D-related (DR) alleles with the SE, although smoking has also been reported to be associated with the production of ACPAs in SE-negative individuals, so the relationship appears complex . This hypothesis is based on studies indicating that smoking is associated with the development of RA [9-11], and continued smoking after the onset of RA is associated with worse disease outcomes . The hypothesis proposed by Klareskog and colleagues is supported by studies from their group and others which have shown that SE HLA alleles and smoking are associated primarily with the development of RA in patients who are anti-CCP-positive [11, 13, 14]. Recently, their group has refined their hypothesis further to indicate that smoking and the development of RA is linked to autoimmunity to a specific citrullinated antigen (α-enolase) in the context of SE HLA-DRB1*04 alleles .
In addition to α-enolase, several potential candidate proteins have been proposed as citrullinated autoantigens in RA patients. Although citrullinated filaggrin was the first citrullinated protein to be identified that bound autoantibodies in RA sera [1, 3], recent studies have emphasized the potential role of autoantibodies that bind citrullinated fibrinogen, collagen, vimentin and α-enolase . Immune complexes of ACPA and fibrinogen have been shown to induce macrophage tumour necrosis factor (TNF)-α production, and these immune complexes were found in the serum and joints of RA patients [17-19]. Collagen is also an attractive autoantibody target in RA by virtue of its prominence in the joint, the presence of anti-citrullinated collagen antibodies in the serum of RA patients and because citrullinated collagen peptides have been found in the synovial fluid of RA patients [20, 21]. Citrullinated and mutated isoforms of the intermediate filament protein vimentin have been identified in the joints of patients with RA . This led to the development of a commercial diagnostic test for RA [anti-mutated citrullinated vimentin (MCV) test] that has a sensitivity and specificity similar to that of the anti-CCP test. Citrullinated α-enolase is expressed abundantly in the joints of RA patients but not osteoarthritis (OA) patients . It has also been postulated that anti-citrullinated α-enolase autoantibodies develop in the context of periodontal infection with Porphyromonas gingivalis, which produces the enzyme peptidylarginine deiminase, which citrullinates α-enolase .
Pathogenic role of ACPA in rheumatoid arthritis
Studies in mouse models of arthritis currently provide the best evidence for a direct pathogenic role of ACPA in RA. While ACPA do not appear to be the sole mediators of synovitis , ACPA augment arthritis significantly when animals have existing low-level joint inflammation. For example, studies with passive transfer of mouse ACPA into mice with pre-existing synovial inflammation support a model for inflammation-induced exposure of citrullinated epitopes within synovium and ACPA-mediated disease progression . In addition, purified human immunoglobulin (Ig)G from RA patients induced arthritis in mice when transferred passively . More recently, the arthritogenicity of ACPA from RA patients has been underscored further by the observations of Cairns and colleagues, who reported arthritis induction in FcγRIIb-deficient mice following passive transfer of affinity-purified ACPA . This same group has also shown that arthritis could be induced in C57BL6 mice transgenic for human HLA-DRB1*0401 using citrullinated fibrinogen; importantly, arthritis could not be induced using unmodified fibrinogen or in wild-type non-transgenic mice .
While direct evidence for a role of ACPA in the pathogenesis of human disease is lacking, there is good evidence that clinical outcomes are worse in RA patients who are anti-CCP-seropositive [3, 30, 31]. As noted above, citrullinated peptides bind to HLA-DRB1 and there is a strong correlation between HLA-DRB1 or SE binding and anti-CCP seropositivity, suggesting that the pathogenic correlation of the SE with disease severity is related to HLA-DRB1 presentation of citrullinated peptides [6, 7, 13]. Finally, as discussed below, data from the analysis of anti-CCP levels in the context of rituximab treatment also provides support for the notion that ACPA are pathogenic in RA [32-35].
The effect of targeted RA therapies on ACPA levels and B cell responses
The development of biologics that target specific mediators of inflammation has led to several highly successful therapies for the treatment of RA [36, 37]. Biological therapies for RA are directed at neutralizing TNF-α, interleukin (IL)-1, IL-6 or IL-17, blocking T cell co-stimulation [cytotoxic T lymphocyte antigen-4 (CTLA-4)-Ig] or depleting B cells (Fig. 1). Among these biological therapies, B cell depletion  and neutralization of TNF-α [38-49] and IL-6 [50, 51] have been studied most successfully for their effects on ACPA levels and B cell populations. There are some limited data on the effects of oral disease-modifying anti-rheumatic drugs (DMARDs) on ACPA levels  and very few data on the effects of CTLA-4-Ig and IL-17 neutralization on ACPA levels, although there is good reason to believe that T cell co-stimulation and IL-17 also regulate ACPA levels.
B cell depletion (rituximab)
Recent studies have indicated that responsiveness to rituximab therapy (anti-CD20) is better in RA patients who are anti-CCP-positive, suggesting a possible relationship between the pathogenic capacity of ACPA and the effectiveness of rituximab . The effectiveness of B cell depletion therapy with rituximab (anti-CD20) strongly supports a role for B cells in the pathogenesis of RA [52-54]. Besides providing evidence of therapeutic efficacy for B cell depletion in RA, these studies provided evidence for the existence of CD20-negative long-lived plasma cells in humans. In these studies, it was noted that depletion of CD20+ naive and memory B cells had little effect on total serum immunoglobulin levels and no effect on anti-tetanus antibody levels . However, autoantibody levels of RF and anti-CCP decreased significantly with anti-CD20 therapy and anti-CCP levels were associated with improvement and relapse in RA patients treated with rituximab  (Table 1). These data suggested that most serum antibodies, and in particular vaccine-induced antibodies such as tetanus toxoid antibodies, are produced by CD20-negative long-lived bone marrow plasma cells. In contrast, autoantibodies in RA may be produced primarily by short-lived CD20+ memory B cells that differentiate into short-lived antibody-secreting plasmablasts [55-57]. Studies have shown that short-lived plasmablasts in synovium secrete autoantibodies and are an important source of ACPA and RF [58, 59]. Short-lived synovial plasmablasts are reduced by rituximab therapy, and plasmablast depletion during rituximab therapy correlates with responsiveness to rituximab therapy [34, 35]. The reduction of short-lived synovial plasmablasts by rituximab provides a potential pathophysiological mechanism for the ability of rituximab to reduce ACPA levels and to improve arthritis symptoms in RA patients.
Table 1. Summary of studies on the effects of B cell depletion with rituximab on anti-cyclic citrullinated peptide (CCP).
Effect of non-TNF biologic on ACPA levels
aNot stated: ‘Ten patients with active RA, unresponsive to methotrexate …’. RA: rheumatoid arthritis; ACPA: anti-citrullinated protein autoantibodies; Ig: immunoglobulin; i.v.: intravenous.
B cell depletion with rituximab in RA also provided important information about the role of memory B cell subsets in the pathogenesis of RA. Leandro et al.  reported that disease relapse in rituximab-treated RA patients was associated with a higher frequency of B cells, which had a memory (CD27+) phenotype at the time of B cell repopulation. In this study, patients with greater than 3 × 106/l CD27+ memory B cells at the time of B cell repopulation had an earlier relapse than patients with fewer CD27+ memory B cells [60, 61].
Other researchers have suggested that B cell depletion may have other effects on B cells that lead to indirect reductions of autoantibody levels. For example, some researchers have suggested that B cell depletion may mimic TNF-α blockade by eliminating lymphotoxin-alpha (LT-α) and TNF-α-secreting B cells .
In contrast to the data on rituximab therapy, there are less consistent data on the correlation between declines in anti-CCP levels and clinical responsiveness to TNF-antagonists in RA patients. Although TNF antagonists reduce RF levels [39-49], these same studies have not established conclusively whether oral DMARDs and/or TNF-antagonists reduce anti-CCP levels. As summarized in Table 2, some studies have identified reductions in anti-CCP levels with TNF-antagonist therapy [38, 39, 43-46, 48], while others have seen no effect of TNF-antagonists on anti-CCP levels [40-42, 47, 49]. In some of these studies TNF-antagonist therapy was compared to oral DMARD therapy [43, 45]; TNF antagonists, but not oral DMARDs, decreased anti-CCP levels, although in these studies subjects receiving only oral DMARDs did not have as great a reduction in disease activity. All these studies examined the effect of TNF antagonists on anti-CCP levels and did not examine the effect of TNF antagonists on specific ACPA levels.
Table 2. Summary of studies on the effects of tumour necrosis factor (TNF)-antagonists on anti-cyclic citrullinated peptide (CCP) levels.
aNot stated (n.s.): ‘failed therapy with at least one prior disease-modifying anti-rheumatic drug (DMARD)’; n.s.: ‘refractory RA’. MTX: methotrexate; AZA: azathioprine; LEF: leflunomide.
There are several factors that may be confounding the analysis of anti-CCP levels during TNF antagonist treatment. For example, differences in disease duration may affect the ACPA response during TNF antagonist therapy; a reduction in anti-CCP levels with anti-TNF therapy was more likely in RA patients with a disease duration of less than or equal to 1 year [38, 40]. Although all anti-CCP2 assays are derived from the same source, some have suggested that the inadequate dilution of serum samples makes the anti-CCP test too sensitive, thereby preventing the detection of variations in the antibody titre during treatment . Other confounders may also affect ACPA levels, including cigarette smoking and periodontal infections with P. gingivalis , which probably provide citrullinated antigenic sources for ACPA production. In addition, many of the published studies did not control for SE status, which may influence ACPA levels due to the potential for stronger and sustained autoimmune responses in those with the SE.
There are several potential mechanisms whereby TNF-α may regulate ACPA levels. One postulate is that TNF antagonists can down-regulate the production of inflammatory cytokines, thereby modulating autoantibody generation, particularly in the synovial compartment . In support of this hypothesis, Anolik et al.  found that lymphoid architecture was altered in patients on etanercept, an anti-TNF therapy that also blocks LT-α. RA subjects treated with etanercept had a significant decrease in follicular dendritic cell (FDC) staining and germinal centres were reduced significantly in number and size . This suggested that TNF antagonists altered B cell populations and probably impacted the ability of B cells to enter or survive a germinal centre reaction.
The changes in lymphoid architecture mediated by TNF antagonists may contribute to the reductions in memory B cells noted in RA patients treated with these therapies [62, 65, 66], which may be important, as memory B cells in RA patients express anti-CCP autoantibodies . In addition, TNF antagonists decrease the proportion of memory B cells expressing CD86 after 6 months of therapy . This is probably important because CD86 and the related molecule CD80 are co-stimulatory proteins that are up-regulated on the cell surface during B cell activation [69, 70]. RA patients have a higher proportion of naive and memory B cells expressing CD86 than healthy controls , which probably favours increased ACPA production.
Immune complexes formed from citrullinated fibrinogen and human ACPA stimulate TNF-α production from macrophages , and co-ligation of the FcγR and Toll-like receptor-4 (TLR-4) by immune complexes containing citrullinated fibrinogen also stimulate TNF-α production . In RA patients, Catalan and colleagues found reduced FcγRIIB expression on memory B cells and plasmablasts compared to healthy controls , suggesting that TNF-α or other downstream cytokines may influence the expression of FcγRIIb on B cells. In addition, a polymorphism in the FcγRIIB gene has been linked to susceptibility to systemic lupus erythematosus (SLE) in humans, and altered function of FcγRIIB could also affect the humoral response against citrullinated proteins in RA patients . For example, Chen et al.  demonstrated an association between a functional polymorphism in FcγRIIB and anti-CCP-positive RA in an Asian cohort.
Although the emphasis in this section has been on the negative effect of TNF antagonism on ACPA levels, studies have shown that therapy with TNF antagonists, but not therapy with other biologics or DMARDs, is associated with the development of anti-nuclear (ANA) and anti-dsDNA autoantibodies [40, 72]; some patients treated with TNF antagonists develop a clinical syndrome that resembles SLE . The relatively selective induction of ANA and anti-dsDNA autoantibodies during treatment with TNF antagonists may be due to the effects of TNF antagonists on apoptosis and enhanced nuclear antigen presentation on the surface of apoptotic cells [40, 72]. A recent report indicated that TNF antagonists increased levels of cytoplasmic Lyn preferentially and up-regulated B cell surface CD20 expression . Lyn plays a role in the initiation of the B cell receptor (BCR)-mediated pathway . Stimulation of B cells via the BCR results in the immediate activation of Lyn, which phosphorylates tyrosine residues of Ig-α/β rapidly and activates a number of downstream signalling proteins, including Syk . This results ultimately in the expression of antibodies.
IL-6 blockade (tocilizumab)
A recent study by Roll et al.  indicated that IL-6 blockade (Fig. 2) with tocilizumab in RA patients led to reductions in total IgG and IgA levels but not reductions in RF levels; results of anti-CCP levels were not reported. In this study, IL-6 inhibition in RA patients led to decreased frequencies of pre- and post-switch memory B cells and to an increase in the percentage and numbers of transitional B cells. Parallel studies by the same group indicated that there were lower frequencies of somatic mutation in memory B cells during IL-6 inhibition . IL-6 has important effects on B cell differentiation and the production of antibodies by B cells . Besides its direct effects on plasma cells, IL-6 signals via signal transducer and activation of transcription-3 (STAT-3), and defects in this pathway lead to the development of defects in T helper type 17 (Th17) development and IL-17 production . Importantly, a study by Doreau et al.  indicated that B cell development and differentiation of B cells into immunoglobulin-secreting cells is regulated by IL-17A and B cell activating factor (BAFF), suggesting a possible interconnection between the effects of IL-6 on IL-17 production and B cell production of autoantibodies.
Recent clinical trials have suggested that blocking IL-17 may be an effective treatment for RA [78, 79]. IL-17 is produced by Th17 cells, which are also important producers of TNF-α . IL-6 drives Th17 cell differentiation by activating the transcription factor STAT-3 [81, 82]. A recent study in Nature Immunology by Doreau et al.  indicated that B cell development and differentiation of B cells into immunoglobulin-secreting cells is regulated by IL-17A and BAFF. IL-17 alone or in combination with BAFF influences the survival, proliferation and differentiation of human B cells directly, and the two in combination are more efficient than either cytokine separately in promoting persistence of self-reactive B cells .
B cell depletion by rituximab followed by increased BAFF mRNA expression in human monocyte-derived macrophages of patients with RA is believed to represent a homeostatic attempt to replenish B cells . BAFF transgenic animals express high levels of BAFF and develop autoantibodies . In the study by Doreau et al. , IL-17A serum levels were elevated in SLE patients and disease severity in SLE patients was correlated directly with IL-17A levels. IL-17A serum levels are also elevated in RA patients and, like SLE patients, BAFF serum levels are elevated in RA patients [78, 79, 85-88] and correlate with disease activity [88, 89]. The differentiation of B cells into immunoglobulin-secreting B cells was regulated by nuclear factor-kappa B (NF-κB_ and the NF-κB-regulated transcription factor TWIST1, which induced the expression of TWIST2 and BCL2A1. TWIST2 expression was up-regulated in B cells from patients with SLE and was correlated directly with SLE disease severity and IL-17A levels; BCL2A1 transcript levels in B cells correlated with IL-17A levels. Doreau et al.  also found that IL-17A and BAFF induced expression of MEF2C, which is an important mediator of BCR-induced proliferation . Importantly, IL-17A and BAFF expression are regulated by TNF [91-94] and TWIST1 is over-expressed in the synovium of patients with RA .
Blockade of T cell co-stimulation with CTLA-4-Ig (abatacept)
There are no direct studies that have determined whether blockade of T cell co-stimulation in humans results in reduced levels of ACPA. However, CTLA-4 signalling suppresses Th17 generation and, as discussed above, IL-17 may have a direct role in stimulating B cell autoantibody production, suggesting that treatment of RA with CTLA-4-Ig may lead to reduced ACPA levels .
Tofacitinib (CP-690,550) is a novel oral Janus kinase (JAK) inhibitor (Fig. 2) that was approved recently by the American Food and Drug Administration (FDA) for the treatment of rheumatoid arthritis. To date, studies in humans on the effects of tofacitinib treatment on ACPA and RF levels have not been reported. However, results from studies of animal models of arthritis indicate that IL-6 levels decrease following administration of tofacitinib [97, 98]. Furthermore, Tanaka and Yamaoka reported that tofacitinib inhibited human IL-17 expression in synovial tissue [97, 98]. Taken together, because tofacitinib suppresses IL-6 and IL-17, and as IL-6 and IL-17 are associated with ACPA and RF production, it seems likely that tofacitinib reduces RF and anti-CCP levels. Clinical studies will be needed to assess this possibility.
Although oral DMARDs reduce RF levels , these studies have not established conclusively whether oral DMARDs reduce anti-CCP levels. As noted above, in studies comparing TNF-antagonist therapy to oral DMARD therapy [43, 44] (Table 3) TNF antagonists, but not oral DMARDs, decreased anti-CCP levels, but in these studies subjects who received only oral DMARDs did not have significant reductions in disease activity. In a study by Mikuls et al. , in which subjects received methotrexate (MTX), sulfasalazine (SSZ) and hydroxychloroquine (HCQ) and had substantial reductions in disease activity, the authors found that reductions in anti-CCP levels were greatest in subjects with early disease while RF reductions were dependent upon disease activity reductions. Besides the Mikuls study, no other published studies have identified an effect of oral DMARDs on anti-CCP levels. Although oral DMARDs reduced both RF and anti-CCP levels in the study by Mikuls et al.  (Table 3), it remains unclear what mechanisms are involved in the reduction of autoantibody levels by oral DMARDs, given that less effective combinations of oral DMARDs did not reduce anti-CCP levels [43, 44].
Table 3. Summary of studies on the effects of oral disease-modifying anti-rheumatic drugs (DMARDs) on anti-cyclic citrullinated peptide (CCP) levels.
Effect of DMARD on ACPA level
ACPA: anti-citrullinated protein autoantibodies; MTX: methotrexate; HCQ: hydroxychloroquine; SSZ: sulfasalazine.
No significant reduction in anti-CCP levels in MTX group
Other mechanisms potentially regulating ACPA levels in RA patients
To date, more than 30 RA susceptibility loci have been identified . Notably, the majority of RA susceptibility loci have been described as risk factors for ACPA-positive RA [13, 100-103]. Direct comparisons between disease subgroups revealed that different genetic association patterns exist between ACPA-positive and ACPA-negative RA . Thus, expansion of the genetic study population(s) is needed to validate the existing genetic risk factors and to understand the implication of genetic heterogeneity among RA populations, as it relates to the regulation of ACPA levels.
The high specificity of ACPA combined with the presence of ACPA early in the disease process suggests an important role for ACPA in the pathogenesis of RA. It is clear that a number of signalling pathways and cytokines are involved in the regulation of ACPA levels in RA patients. Studies in humans treated with different biological therapies suggest key roles for TNF-α, IL-6 and IL-17 in the regulation of ACPA levels in RA patients.
Despite the impressive overall clinical impact of biologics, more than one-quarter of RA patients still have a poor clinical and radiological response to all biological therapies, which emphasizes the need for reliable predictive indices of the response to different biological therapies. Once we understand the complex interconnected regulatory pathways that lead to the generation and persistence of ACPA in RA we can select therapies for individual patients rationally, and we can design new therapies that target all the pathways that lead to ACPA production, synovitis and joint destruction.
First two authors have no financial disclosure to declare. Marc Levesque: Genentech, grant support and consultant; UCB, consultant; Baxter Healthcare, consultant; AbbVie, expert witness; Crescendo, consultant.