Objective: The contribution of low-penetrance single nucleotide polymorphisms to methotrexate efficacy in rheumatoid arthritis (RA) is inconsistent between studies. We sought to elucidate architecture of methotrexate response in three cohorts of patients with RA treated with methotrexate.
Methods: Single nucleotide polymorphism frequencies in genes from folate, purine, and pyrimidine pathways were measured to develop a model of gene-gene interactions using multifactor dimensionality reduction in 439 patients who received methotrexate in the USA and The Netherlands. A third cohort of 530 patients with RA from Sweden was used to replicate the findings. Methotrexate efficacy was assessed using the European League Against Rheumatism criteria in the majority of patients.
Results: Nonlinear patterns of gene-gene interactions between variants in aminoimidazole carboxamide ribonucleotide transformylase (C347G), reduced-folate carrier (G80A) and inosine-triphosphate pyrophosphatase (C94A) revealed a predisposing genetic attribute significantly associated with methotrexate response in the USA and Dutch cohorts [odds ratio (OR)=2.9, 95% confidence interval (CI): 1.9-4.2; P<0.001]. Although the finding was not replicated in the Swedish cohort (OR=0.9; 95% CI: 0.64-1.37; P=0.74) a multifactor dimensionality reduction analysis superimposing the predisposing genetic attribute with patient's age, sex, and anticitrullinated peptide antibodies positivity (ACPA) revealed a pattern of interaction significant in all three cohorts (OR=2.2, 95% CI: 1.6-2.9; P<0.01). The selective advantage toward response in the presence of the predisposing genetic attribute was lost in females and ACPA-positive patients, whereas older and male ACPA-negative patients tended to exhibit a greater likelihood of response in the absence of the predisposing genetic attribute.
Conclusion: Gene-gene interactions together with nongenetic attributes may contribute to methotrexate efficacy in RA.