Type I interferon response gene expression in established rheumatoid arthritis is not associated with clinical parameters

Rheumatoid arthritis (RA) is a systemic autoimmune disease characterized by chronic joint inflammation. It manifests as a heterogeneous disease that is partly reflected at the level of gene expression. Genome-wide gene expression analysis revealed evidence for molecular differences between patients with RA, in particular in the type I interferon (IFN) system. Approximately 50% of patients with RA display a peripheral blood IFN signature (i.e., relatively high expression of interferon response genes [IRGs]) [1].

Type I IFNs were initially known for their antiviral effects, but increasing insight into their activities revealed their role as pleiotropic cytokines with a critical role in modulating immune responses, such as cellular activation, major histocompatibility complex upregulation, induction of apoptosis, and inhibition of angiogenesis [2]. It is thought that type I IFNs contribute to autoimmunity by initiating a break of tolerance (e.g., by the induction of dendritic cell maturation and inhibition of regulatory T cells) [3]. The exact role of the IFN signature in RA is yet unknown, although it was shown to have potential clinical relevance. That is, (1) the presence of the IFN signature was shown to be a risk factor for arthritis development in preclinical disease [4], and (2) the presence of the IFN signature in established RA was found to be associated with the clinical response to treatment with rituximab [5] and tocilizumab [6].

Earlier studies have addressed whether the IFN signature in RA could be associated with clinical parameters, which inconclusively revealed a potential relationship of the IFN signature with anticitrullinated protein antibody (ACPA) titers [7, 8]. However, these study cohorts were rather small (35 subjects or less) and therefore highly subject to a lack of power. Hence, the relationships between the peripheral blood IFN signature and disease- and inflammation-related clinical parameters have never been thoroughly assessed. In the present study, we used a larger cohort of patients with established RA (n = 182) in combination with a random sampling algorithm to systematically investigate whether the peripheral blood IFN signature in RA could be associated with parameters such as disease activity, laboratory parameters, and the use of immunosuppressive treatment.

We studied the association between the peripheral blood IFN score and the following parameters: disease duration, DAS28 and its individual components, the occurrence of erosions and nodules, autoantibody positivity, and immunosuppressive treatment. As demonstrated in Table 2, prednisone use and dose, hydroxychloroquine (HCQ) use, and sulfasalazine (SSZ) use were the only variables that showed a significant result, of which only HCQ use remained significant after correction for multiple testing. Similar results were obtained in the cross-validation using the random sampling algorithm, which revealed significance for HCQ use only: a p value <0.05 was detected in both sets in 521 of the 1000 iterations, in 1 of the 2 sets for 479 of 1000 iterations, and never in none of the sets (median p value 0.015). A slight trend was also observed for prednisone use and dose and SSZ use (median p values 0.090–0.14, median coefficient for prednisone dose −0.18), although significance was never found in both sets for these variables. Each IRG was also analyzed individually, which revealed similar results (data not shown).

For prednisone, HCQ, and SSZ treatment, the IFN score was lower in the treated group than in the untreated group. Combining HCQ use, prednisone use, and SSZ use also revealed a significantly lower IFN score in patients using HCQ and/or prednisone and/or SSZ than in patients not treated with any of these agents, both in the analysis of the complete cohort and in the cross-validation (p = 0.0080 with Benjamini-Hochberg correction, median cross-validation p ≤ 0.012) (see Table 2 and Fig. 1a). Moreover, the suppressive effect appeared larger for patients treated with two or more of those agents than for patients treated with one agent (Fig. 1b). No association was found between IFN score and methotrexate (MTX) treatment or dose. The relation between IFN score and treatment did not appear to be confounded by DAS28 or disease duration.

Because the suppression of IFN score in HCQ-, prednisone-, and/or SSZ-treated patients could have a masking effect on other associations between IFN signature and clinical parameters, we also performed the analyses for the selection of patients who were not treated with prednisone, HCQ, and/or SSZ (n = 95) at the moment of blood collection. This did not result in any significant associations between the IFN score and other variables (uncorrected p value ≥0.099, median p value cross-validation ≥0.23) (see Additional file 1: Table S2).

The present study is the first use of a systematic approach in a relatively large cohort to study the relationship between the IFN signature in established RA and clinical parameters. We demonstrated that the IFN signature was suppressed in patients treated with HCQ, prednisone, and/or SSZ, but not with MTX. Furthermore, we did not observe any associations between the IFN signature and the other clinical parameters.

Van der Pouw Kraan et al. showed that a subgroup of patients with RA displays a common pathogen-response program, which was characterized by a higher incidence of the IFN signature as well as higher ACPA titers, suggesting that these parameters might be associated with one another [7]. However, a causal relationship was not established, and our data indicate that this is not the case. The IFN signature was not significantly different between ACPA-negative and ACPA-positive patients, nor did it significantly correlate with ACPA titers. Possibly, the IFN signature and ACPA positivity are independently associated with activation of the common pathogen response program, because they are both implied to be induced via certain pathogens [12, 13].

Remarkably, the IFN scores were decreased in HCQ-, prednisone-, and/or SSZ-treated patients, even though the beneficial effects of these treatments were supposedly diminished. Moreover, cotreatment with these agents appeared to have an additive suppressive effect. Interference of both prednisone and HCQ with type I IFN signaling has been described before [14, 15], but the influence of SSZ remains to be elucidated. It has been shown that SSZ reduces the levels of RA-related cytokines, such as interleukin-1β and tumor necrosis factor-α [16], suggesting that SSZ might function through overall suppression of inflammatory cytokines, including type I IFNs. Furthermore, it was demonstrated that SSZ is able to accelerate apoptosis of neutrophils [17], which we have recently shown to be major inducers of the type I IFN response in RA [18]. Consequently, suppression of the type I IFN response by SSZ might be mediated via an increase in neutrophil apoptosis.

As previously described, suppression of the IFN score by certain treatment could affect the applicability of the IFN signature as a biomarker for therapy response, particularly to rituximab [5, 19]. That is, the treatment-related suppression of IFN score might impair the discriminative capacity of the biomarker, which would consequently lead to more false predictions. Because multiple studies have demonstrated that the extent of the IFN signature is highly variable between patients [1, 20, 21], we considered it important to assess the IFN score as a continuous variable rather than as a dichotomous variable. However, inclusion of a healthy control population would allow us to determine whether patients with prednisone, SSZ, and/or HCQ treatment would still display an IFN response above the levels of healthy control subjects. This could also give more insight into the applicability of the IFN response as a biomarker in these patients. Future studies should be done to elucidate the effect of each individual treatment, as well as combinatory therapy, on the IFN signature and the corresponding response prediction. Alternatively, presence of the IFN signature in individuals with arthralgia was shown to be associated with a higher risk for developing arthritis [4]. It would be interesting to study whether early treatment with one of the implied suppressors of the IFN response could delay or even prevent disease onset.

Our data indicate that there are no evident associations between the peripheral blood IFN signature in established RA and clinical parameters. This suggests that the IFN signature is not an indication of disease activity per se, but its presence could indicate a potential difference in pathology or immune pathway activation compared with patients without this signature. Consequently, this could influence the response to therapy, particularly to biologics because these are specific modulators of these immune pathways.