Background Variability of clinical course and response to therapy in RA suggests differences in molecular mechanisms depending on stage and sub-entity of the disease. Most dominant molecular discriminators are HLA-DR4 shared epitope (SE), and the status of rheumatoid factor (RF) and anti-citrullinated protein antibody (ACPA) development, all indicating subentities with influence on severity and outcome.

Objectives Early RA prior to methotrexate (MTX) treatment was investigated to exclude major therapeutic influence. By comparing whole blood transcriptomes between subsequent responders and non-responders we aimed to unravel molecular mechanisms with influence on MTX outcome of RA.

Methods Whole blood of 52 patients with early RA was obtained in the HITHARD and in the ArthroMark study. Clinical characteristics assessed at baseline and after 16 weeks included RF, ACPA, DAS28 and EULAR response for classification in good (GR), moderate (MR) and non-responders (NR). Total RNA was globin reduced and processed according to standard protocols for hybridization to U133 Plus 2.0 microarrays. Biostatistical methods included MAS5.0 and RMA algorithms with limma, lasso and Wilcoxon tests. For functional interpretation, expression of candidate genes was tested in various blood cell types and stimulation conditions using own and GEO reference transcriptomes. Marker selection was validated by qPCR with commercial primers and four different housekeeping genes.

Results Different biostatistical approaches favored MAS5.0 algorithms for marker selection. Comparison between good and non-responders in unselected patients revealed insufficient discriminatory power. Generating different subgroups defined by gender, status of RF, ACPA, SE, HLA-DRB4 or DQA1 identified molecular patterns with a strong discrimination between GR and NR in HLA-DRB4^– patients. All other subgroups did not improve pattern selection. In HLA-DRB4^– patients, genes increased in GR were related to phagocytes and bone marrow activation whereas genes increased in NR were associated with lymphocyte activity. In HLA-DRB4⁺ patients, patterns of adaptive immunity were again related to non-response but frequently combined also with innate patterns. Receiver-operating-characteristics (ROC) with the top 20 markers in HLA-DRB4^– and DRB4⁺ subgroups revealed AUC<0.96 and AUC<0.91 for response and AUC<0.98 and AUC<0.99 for non-response associated genes, respectively. Technical validation by qPCR with the 16 top candidates confirmed AUC-values up to 0.97 for prediction in the HLA-DRB4^– group and reached AUC up to 0.89 in HLA-DRB4⁺ patients.

Conclusions Genetic stratification revealed transcriptional patterns that indicate different molecular pathomechanisms in RA depending on response to MTX therapy. Interestingly, the group responding to MTX was dominated by phagocyte but not lymphocyte activity, which may indicate an important contribution of innate triggers to the pathomechanisms in RA.

Disclosure of Interest None declared