Seronegative rheumatoid arthritis as an evolving entity

For classification purposes, seronegative rheumatoid arthritis (RA) has become something of a moving target, reflecting both an evolving understanding of disease-associated autoantibodies and recent modifications in RA classification overall . Such factors inevitably impact upon the clinical spectrum and natural history of the syndrome described. RA ‘seropositivity’ historically referred only to the presence of immunoglobulin M (IgM) rheumatoid factor (RF), but this term must now also take account of circulating anti-citrullinated protein antibodies (ACPAs) whose presence is independently associated with accelerated joint destruction . Consequently, by excluding from the definition of ‘seronegative RA’ the 10% of RF-negative patients now determined ACPA positive, a subtly more benign disease subset may be being described. ACPA status is typically measured using the now widely available anti-cyclic citrullinated peptide 2 (CCP2) assay , and efforts to develop comparably disease- specific but more disease- sensitive ACPA tests may further advance this trend . Moreover, a novel system of RA-associated autoantibodies was recently described, directed against carbamylated proteins (anti-CarP); 16% of ACPA-negative RA patients were found to be anti-CarP positive, having an adverse disease course compared with their ‘double-negative’ counterparts . Although the classification criteria for RA are not yet informed by the new antibody system, they were modified in 2010 to incorporate consideration of the ACPA status alongside RF in an effort to enhance their ability to identify early RA . Interestingly, this replacement of the pre-existing 1987 ACR classification criteria results in an increased prevalence of classifiable RA in the early arthritis clinic . Although no difference in the proportion of patients seropositive for either RF or ACPA has ensued, the modified criteria appear to describe an early RA cohort with a milder disease course than before – a phenomenon likely to apply to both seronegative and seropositive subgroups.

Such a rapidly changing landscape complicates efforts to define seronegative RA, raising legitimate questions as to whether a stable phenotype for the condition exits . We here argue that it does, starting with a consideration of the clinical spectrum of the condition, and then moving through pathogenetic and therapeutic aspects. Although challenging, we conclude that improving our understanding of seronegative RA is a worthwhile endeavour, with the potential to improve disease outcomes through the development of robust diagnostic tools and more targeted treatment strategies.

Prevalence and clinical spectrum

Depending on the population studied, the proportion of classifiable RA patients reportedly seronegative for both RF and ACPA ranges widely between 0.10 and 0.48 . A recent meta-analysis provided a pooled ‘double-seronegativity’ rate estimate of 0.22 (95% confidence interval (CI) 0.20–0.24) in this setting . Even allowing for slight diminution of this figure attributable to the availability of anti-CarP testing in the future, an overall population prevalence for seronegative RA that approaches 20 per 10,000 may be extrapolated from epidemiological data – comparable, for example, to that of coeliac disease and twice that of psoriatic arthritis .

The literature has long recognised particular clinical presentations of RA as being associated with autoantibody seronegativity, particularly in older patients. Indeed, employing a cut-off of 60 years of age, lower rates of both RF and ACPA have been observed in late-onset RA (LORA) compared with young-onset RA . A syndrome of explosive-onset, seronegative LORA with polymyalgic features and a soaring acute phase markers is well recognised , which has been described as having a relatively mild course . Distinguishing seronegative LORA from polymyalgia rheumatica (PMR) may therefore be challenging at disease onset, as the latter condition can also present with evidence of peripheral joint synovitis , but longitudinal study has identified particular patterns of joint involvement as being discriminatory for LORA in this context . Importantly, >30% of RF-negative LORA patients, and no PMR patients, developed bony erosions at follow-up, justifying their classification as distinct entities and giving the lie to the notion that seronegative RA is a ‘benign’ condition. Patients with remitting symmetrical seronegative synovitis with pitting oedema (RS3PE) may also describe myalgic symptoms alongside characteristic peripheral soft tissue oedema and/or tenosynovitis . Although initially considered a form of RA , the condition’s unique associations with raised circulating serum vascular endothelial growth factor (VEGF), distinct tissue types and, in some cases, particular malignancies have increasingly led experts to classify it separately .

Of course, many seronegative RA patients do not map to any of the above descriptions and, in practical terms, are essentially indistinguishable from their seropositive counterparts at presentation . Yet diagnosing seronegative RA with confidence remains particularly challenging , and the introduction of new classification criteria for RA has had the effect of reducing the prevalence of ACPA seropositivity amongst undifferentiated arthritis (UA) patients from 24% to 10% . The conclusion that diagnostic delay may increasingly be a feature of seronegative disease is borne out by a review of >130 early RA patients at our own unit, amongst whom a diagnosis was achieved within 6 weeks of referral for proportionately fewer ‘double-seronegative’ patients than for RF and/or ACPA seropositive individuals (51% vs. 73%, p = 0.02, unpublished). It is therefore anticipated that renewed efforts to understand aspects of RA pathogenesis, and the extent to which they are concordant or discordant between serotypes, should yield tools of clinical utility for diagnosis as well as treatment.

Prevalence and clinical spectrum

Depending on the population studied, the proportion of classifiable RA patients reportedly seronegative for both RF and ACPA ranges widely between 0.10 and 0.48 . A recent meta-analysis provided a pooled ‘double-seronegativity’ rate estimate of 0.22 (95% confidence interval (CI) 0.20–0.24) in this setting . Even allowing for slight diminution of this figure attributable to the availability of anti-CarP testing in the future, an overall population prevalence for seronegative RA that approaches 20 per 10,000 may be extrapolated from epidemiological data – comparable, for example, to that of coeliac disease and twice that of psoriatic arthritis .

The literature has long recognised particular clinical presentations of RA as being associated with autoantibody seronegativity, particularly in older patients. Indeed, employing a cut-off of 60 years of age, lower rates of both RF and ACPA have been observed in late-onset RA (LORA) compared with young-onset RA . A syndrome of explosive-onset, seronegative LORA with polymyalgic features and a soaring acute phase markers is well recognised , which has been described as having a relatively mild course . Distinguishing seronegative LORA from polymyalgia rheumatica (PMR) may therefore be challenging at disease onset, as the latter condition can also present with evidence of peripheral joint synovitis , but longitudinal study has identified particular patterns of joint involvement as being discriminatory for LORA in this context . Importantly, >30% of RF-negative LORA patients, and no PMR patients, developed bony erosions at follow-up, justifying their classification as distinct entities and giving the lie to the notion that seronegative RA is a ‘benign’ condition. Patients with remitting symmetrical seronegative synovitis with pitting oedema (RS3PE) may also describe myalgic symptoms alongside characteristic peripheral soft tissue oedema and/or tenosynovitis . Although initially considered a form of RA , the condition’s unique associations with raised circulating serum vascular endothelial growth factor (VEGF), distinct tissue types and, in some cases, particular malignancies have increasingly led experts to classify it separately .

Of course, many seronegative RA patients do not map to any of the above descriptions and, in practical terms, are essentially indistinguishable from their seropositive counterparts at presentation . Yet diagnosing seronegative RA with confidence remains particularly challenging , and the introduction of new classification criteria for RA has had the effect of reducing the prevalence of ACPA seropositivity amongst undifferentiated arthritis (UA) patients from 24% to 10% . The conclusion that diagnostic delay may increasingly be a feature of seronegative disease is borne out by a review of >130 early RA patients at our own unit, amongst whom a diagnosis was achieved within 6 weeks of referral for proportionately fewer ‘double-seronegative’ patients than for RF and/or ACPA seropositive individuals (51% vs. 73%, p = 0.02, unpublished). It is therefore anticipated that renewed efforts to understand aspects of RA pathogenesis, and the extent to which they are concordant or discordant between serotypes, should yield tools of clinical utility for diagnosis as well as treatment.

Aspects of pathogenesis

Although genetic variation has long been supposed to account for around 60% of RA risk in Caucasian populations , recent work suggests that this figure for heritability , may be considerably lower in seronegative RA . It follows that environmental factors should play a more important role in seronegative RA, but much has yet to be understood in this area . Interestingly, whilst smoking is considered a major environmental determinant of seropositive disease, in turn interacting with genetic risk, it plays a lesser, and perhaps negligible, role in the induction of seronegative RA . Nonetheless, the likelihood that a first-degree relative of a seronegative patient will share the diagnosis is more than twice the RA population prevalence, implying a substantial contribution of genetic variation to disease risk in this group. The increasingly well-characterised genetic architecture of RA now supports a paradigm in which disease serotypes delineate aetiologies with both distinct and overlapping features.

The strongest and most established genetic association with RA is with ‘shared epitope’ (SE) alleles at the highly polymorphic human leucocyte antigen (HLA) DRB1 locus, which encode a conserved amino acid sequence within the peptide-binding groove of major histocompatibility complex (MHC) class II molecules, which are expressed by antigen-presenting cells . The contribution to the RA risk of two additional amino acid positions in the binding groove, encoded at HLA-B and HLA- DPB1 loci, respectively, has since been confirmed . Consistent with the notion that the SE allelic expression provides a mechanism for citrullinated peptides to generate antibody responses , it was initially thought that the HLA- DRB1 association represented a primary risk factor for ACPA positivity rather than RA per se . A more nuanced reality has recently been confirmed by increasingly well-powered studies, clearly demonstrating HLA association in both serotypes, albeit with diminished effect sizes in ACPA-negative disease . Indeed, recent work suggests that individual HLA alleles confer differing risk according to ACPA status, and that distinct antigens may therefore be involved in the pathogenesis of both conditions . A similar pattern of association with seropositive and seronegative RA is seen for missense mutation at the common autoimmune susceptibility locus protein tyrosine phosphatase, non-receptor type 22 ( PTPN22 ), encoding lymphoid protein tyrosine phosphatase (Lyp) – with a diminished effect size in ACPA-negative individuals . The functional role of this variant in conferring RA risk may vary within and between cellular contexts , but important regulatory effects on T-cell-receptor-mediated signalling seem beyond doubt .

A differential role of T cells between RA serotypes is also implied by scrutiny of synovial tissue in established disease. van Oosterhout et al. studied 34 ACPA-positive and 23 ACPA-negative patients (median 9 years duration), and identified significantly more fibrosis and synovial lining thickening, with less lymphocyte infiltration, in the ACPA-negative group . A non-significant trend towards a reduced likelihood of ectopic germinal centre formation in synovial infiltrates of ACPA-negative patients was also noted. Intriguingly, Cantaert et al. by contrast observed comparable numbers of infiltrating synovial T cells between serotypes, but demonstrated a relatively restricted T-cell repertoire in the synovial compartment of ACPA-positive patients compared with ACPA-negative counterparts – consistent with a role for T cells in both disease subsets, but a more classical ‘antigen-driven’ pathology in seropositive disease. Gomez-Puerta et al. found that synovial fluid from ACPA-positive patients contained higher levels of T-cell-derived pro-inflammatory cytokines, including interleukin (IL)-17, than seronegative patients .

These data must now be considered alongside recently highlighted cluster of differentiation (CD)4+ T-cell-mediated mechanisms of ACPA-negative disease induction. For example, the most compelling non-HLA genetic association with seronegative RA thus far identified, with an effect size equivalent to that seen in seropositive disease, is with an intronic variant of the ankyrin repeat domain-55 ( ANKRD55 ) locus on chromosome 5 – a gene highly expressed in CD4+ T cells but of unknown functional significance . Intriguingly, the locus lies ∼185 kilobases upstream of the gp130-encoding IL-6ST transcriptional start site, and within a transcription factor-binding domain for that gene . Glycoprotein 130 (gp130) is a ubiquitously expressed co-receptor for numerous cytokines including the pro-inflammatory cytokine IL-6. It is notable that polymorphism within several other components of this cytokine’s signalling pathway , including the IL-6 receptor ( IL-6R ), tyrosine kinase-2 ( TYK2 ) and signal transduction and activator of transcription-3 ( STAT3 ) genes, have been associated with RA – in some cases specifically ACPA-negative disease . Moreover, we recently described a STAT3 target-enriched gene signature in circulating CD4+ T cells of drug-naïve, early arthritis patients, which predicted progression to RA most effectively in ACPA-negative disease . As this was validated in an independent cohort, we found that the phenomenon most likely reflects increased circulating IL-6 in early ACPA-negative RA, together with a maximal capacity for STAT3 phosphorylation in response to the cytokine amongst CD4+ T cells (manuscript submitted). The extent to which these observations may be mechanistically linked to genetic risk remains the subject of ongoing investigation, but the precedent for the importance of this pathway in the induction of inflammatory arthritis continues to emerge from animal models .

Accumulating evidence now implicates T-cell-mediated immune dysregulation in RA pathogenesis, irrespective of autoantibody status. It may be that qualitatively distinct cellular mechanisms during disease induction not only define serotype but also drive parallel pathways towards apparently indistinguishable clinical phenotypes. Hence, T cells non-specifically activated by cytokines as a feature of generalised ‘immune ageing’ may be relatively important in sustaining chronic joint inflammation without autoantibody production, and auto-antigen-mediated adaptive immune induction may represent a potent additional disease mechanism under particular environmental and/or genetic conditions . Such a model is highly speculative and has much about it yet to be explained ( Fig. 1 ). It nonetheless illustrates how a fuller appreciation of seronegative RA pathogenesis should inform our understanding of RA as a whole, and could ultimately contribute to more ‘personalised’ strategies for managing the condition in the future.

RA serotypes delineate aetiologies with distinct, but overlapping, features. Select environmental and genetic risks are depicted in the upper part of the figure, with individual factors, respectively, positioned to the left (light grey) or right (dark grey) according to their postulated importance in the development of seronegative versus seropositive disease. In our speculative model, these elements interact to promote qualitatively distinct mechanisms of T-cell dysregulation, which lead to a common pathway of persistent joint inflammation, and define serological status. Pathogenetic mechanisms of seronegative disease may, in fact, apply to both serotypes. See text.

Therapeutic considerations

The ability to use biomarkers for defining disease subtypes that reflect targetable pathobiology, and so permit stratification by therapeutic responsiveness, has become a major research goal in rheumatology, as in other fields . Attempts to determine the extent to which RA autoantibody serotype may represent such a biomarker are summarised here, and have generally arisen from observational studies and retrospective analyses of clinical trials.

Methotrexate (MTX) has emerged as the widely accepted ‘first-line’ disease-modifying anti-rheumatic drug (DMARD) for the management of newly diagnosed RA, and its use in the management of inflammatory UA has also been advocated . In the PROMPT study, which followed newly presenting UA patients, rates of progression to classifiable RA, radiographic progression and drug-free remission were all comparable in MTX- and placebo-treated arms . However, MTX did delay RA progression uniquely in the ACPA-positive UA subgroup, perhaps suggesting its relatively diminished effect on the natural history of ACPA-negative disease. In established RA, neither RF nor ACPA status have in general proved effective in predicting responsiveness to MTX therapy , despite observations that drug-free remission is less likely, and radiographic progression more likely, in ACPA-positive individuals . The introduction of combination DMARDs, such as sulphasalazine and hydroxychloroquine, alongside MTX with corticosteroid ‘induction’, has been advocated for the treatment of newly diagnosed RA . It seems that the advantages of initial DMARD combination (more rapid reductions in DAS28) should be weighed up alongside potential disadvantages (need for dose adjustments due to side effects) on a case-by-case basis, as a recent randomised controlled trial (RCT) has reported no difference between them in terms of radiological or functional outcome at 12 months providing a responsive, treat-to-target approach is maintained . A reanalysis of the Combination Anti-Rheumatic Drugs in Early RA (CARDERA) trial suggests that the autoantibody status may yet prove a useful biomarker for guiding such decisions, as the use of combination strategies (including the use of oral steroids) significantly reduced radiologic progression over 2 years in ACPA-positive, but not ACPA-negative patients . Whilst such data seem to favour a more ‘aggressive’ combination DMARD approach for the treatment of ‘poor-prognosis’ ACPA-positive early RA, radiological progression in the ACPA-negative sub-cohort of this study was not trivial, and the need to identify alternative therapeutic approaches for this disease subtype is therefore emphasised.

As regards biologic use, retrospective analyses of registry and cohort data have suggested superior efficacy of anti-tumour necrosis factor (TNF) therapy for the treatment of seronegative RA compared with seropositive disease . In such studies, the contribution of autoantibody status to the variability of response nonetheless remains small – perhaps influenced by protracted disease duration at the time of recruitment and/or analysis. Indeed, a prospective study found an association between high baseline ACPA and RF titres with good responses to infliximab, but concluded that the predictive value of autoantibodies in this setting was unlikely to be clinically useful . Although focussed pharmacogenetic analyses have yet to conclusively identify genetic associations with anti-TNF responsiveness , an intriguing report that polymorphism at the tumour necrosis factor receptor II (TNFRII) locus predicts therapeutic response preferentially in seronegative RA awaits full publication .

Despite a dearth of large prospective studies addressing the issue for anti-TNF, a meta-analysis of phase II or III clinical trials for the B-cell-depleting agent rituximab demonstrated increased efficacy in seropositive patients – including those naïve to MTX and with relatively short disease duration . A similar enhanced efficacy amongst ACPA-positive RA patients has been reported in an extensive registry study of the cytotoxic T-lymphocyte-associated protein 4 (CTLA4)–Ig fusion protein abatacept , perhaps reflecting differential effects of co-stimulation blockade on T-cell function according to autoantibody serotype . By contrast, no difference in the efficacy of the IL-6R-targeting monoclonal antibody tocilizumab has to date been demonstrated with respect to autoantibody serotype .