Over the past two decades, RA patients have benefitted from development of highly active therapies including biologic agents that block cytokines and cellular signaling pathways. The availability of such treatments along with MTX has resulted in high response rates, so that guideline recommendations are to aim for remission or sustained low disease activity [24]. Despite these advances, significant numbers of RA patients do not respond satisfactorily to available therapies. Delay in initiating early effective intervention has been identified as a major risk factor for a refractory disease course [19]. At least 30% of patients who are initially treated with MTX do not have adequate disease control and will need addition of another agent, usually a biologic DMARD [25]. If it were possible to predict at the outset which patients were likely to require more than MTX, this information could be used to justify addition of a second agent earlier in the disease course, which would improve the likelihood of a good clinical outcome [5, 18].
Tools that assist with monitoring responses to therapy and drive treat-to-target strategies are available for use in clinical practice. The commercially-available Vectra DA diagnostic, for example, incorporates 12 different serum markers into a single reported score that correlates with responses to MTX and other DMARDs [9]. It has been used to assess responses to therapy over time and can confirm maintenance of low activity states. Other measures that incorporate clinical assessments, such as the DAS-28 scores that include acute phase reactant blood tests as well as instruments such as the CDAI or RAPID3, which do not require any laboratory tests, also are useful in driving remission-inducing treatment strategies [8, 26].
However, correlation with clinical status and response to treatment is more straightforward than prediction of the response to an intervention using information obtained pre-treatment or very early in the course of therapy. Developing a risk profile for those RA patients who are likely to have a more difficult treatment course is especially needed to optimize use of available therapeutics [20]. Some elements of such a risk profile, including demographic and environmental variables, clinical measures and laboratory tests have been recognized. One model that was developed for prediction of inadequate response to MTX in the first year of disease included higher disease activity and current smoking as contributing factors [18]. It is also generally accepted that female sex predicts a course that is more refractory to treatment [19, 20]. Patients positive for the HLA-DRB1 shared epitope generally have more severe disease, and appear to also have a poorer response to MTX [20]. However, these and other identified predictors have only modest utility in an individual patient, and are not sufficiently robust to be used for a personalized approach to therapy. Levels of autoantibodies that are routinely measured in clinical practice (RF, CCP), have been shown to not be predictive of response to treatment [11, 20]. Other proposed serum biomarkers also have been shown not to be useful, including VCAM-1, which when measured at pre-treatment baseline did not add to clinical predictors of treatment response to MTX [27] and S100A9 which did not predict responses to etanercept [28].
We have previously shown that a large number of genes that are dysregulated in RA patients return toward levels seen in HC subjects after treatment with MTX [12]. Furthermore, in vitro responses to MTX in cell culture occur very rapidly [12], suggesting that such changes might be detected early in patients who initiate treatment with MTX. Analysis of the patients in the current study used the available data to distinguish between effects of disease activity and MTX treatment. While some similarities were observed in the disease activity groups regardless of MTX status, it was also found that patients who achieved low levels of disease activity with MTX have the greatest changes towards normal. Furthermore, patients in whom MTX treatment had not produced a low activity state showed greater dysregulation of the gene expression profile (Fig. 4). The data collected here are cross-sectional and not longitudinal, but are suggestive that a patient in whom these normalizing trends in gene expression were observed would be predictive of MTX responsiveness and this might be expected to occur early, prior to definitive changes in the clinical variables. A similar approach has been recently reported by others, showing that changes in whole blood gene expression profiles at 4 weeks were predictive of MTX nonresponsiveness [29]. The gene expression classifier in this study was found to be superior to models that were generated with clinical measures. The 4-week timepoint alone had very good predictive capability, while the pretreatment point alone had some limited value in predicting response. Since many patients are kept on MTX as monotherapy for up to 6 months before considering adding therapies, this type of early test result might be utilized in an individual patient to make changes in therapy that would lead to better outcomes.
Pathway analyses in this same previous study highlighted genes involved in the response to Type I interferon and the Type I interferon signaling pathway as being relatively enriched in the nonresponders [29]. This result is similar to that reported in the current study, in which nonresponders to MTX showed higher levels of gene specificities in interferon-related pathways. The Type I IFN pathway and the signature of genes that are involved in IFN responses have been shown to be elevated in patients with RA, though it appears that both the alpha and beta IFN subtypes are represented, while in other autoimmune conditions such as systemic lupus erythematosus (SLE), the alpha type predominates [30]. In general, the Type I IFN signature in patients with RA was found not to be suppressed by MTX [31]. Of interest is the previously reported finding that the ratio of serum activity of the Type I IFNs beta/alpha was predictive of reponse to anti-TNF therapeutics [32]. Another study has reported that an elevated IFN gene expression signature in neutrophils from RA patients, including alpha, beta and gamma subtypes, was correlated with a good response to treatment with TNF inhibitors [33]. Involvement of IFN-gamma in enhancing invasiveness of synovial cells has been recently reported and thus expression of this cytokine might predict a more damaging course [34]. Taken together with these previous results, our findings suggest that the patients in whom IFN-related gene expression remains elevated may require treatment with TNFi or other agents agents such as JAK-STAT inhibitors to achieve remission [35].
The present study is limited by the cross-sectional rather than longitudinal nature of the data, and it is not known how rapidly changes toward normal gene expression patterns occur in patients who are started on MTX. If changes in the signature occur as early as 4 weeks of MTX treatment, as reported by others [29], the finding that gene patterns are not showing normalizing changes could be utilized to indicate likely MTX nonresponse status and accelerate the addition of a second agent to the treatment. If predictive signatures were shown to be present at the baseline, when MTX is initiated, those would be even stronger signals to be utilized in rapid acceleration of therapy in given individuals.
In addition to the cross-sectional design, another limitation of this study is that the dose and mode of MTX administration was not controlled, as treatment was carried out according to each rheumatologist’s preference. In addition, the study population was small and predominantly non-Hispanic Caucasian; findings might differ in other racial or ethnic groups.