Next generation sequencing identifies differences in the transcriptomic profile of pretreatment early rheumatoid arthritis synovial tissue from responders and non-responders targeted to adalimumab

Background: Despite the diverse range of mechanisms currently targetable with treatments for rheumatoid arthritis (RA), predicting consistent clinical response to treatment remains a challenge.

Objectives: To investigate the mechanisms of treatment response and the ability to predict treatment-responsiveness in early CCP+ RA patients, targeted to biologic DMARDs.

Methods: Synovial tissue (ST) was obtained from ACPA+, DMARD-naïve, early RA patients (N=30) prior to treatment initiation (<12 months of symptom onset, fulfilling 2010 ACR/EULAR criteria). ST sections were stained for CD45RO and TNF by immunohistochemistry (IHC) and patients were prescribed abatacept or adalimumab respectively, based on the dominance of these markers within ST. In addition, the transcriptomic profiles of these samples along with 26 paired post-treatment RA ST biopsies were examined by bulk RNA-sequencing after 6 months of therapy. Total RNA was isolated using a Qiagen RNeasy mini Plus kit and RNA libraries produced with the Tecan Universal Plus Total RNA library preparation kit. Samples were sequenced on the NovaSeq using paired end 100bp reads (Illumina). Differential gene expression analysis (DGE) was completed using EdgeR. Responders were defined as patients achieving remission (DAS28-CRP<2.6) at 3 months.

Results: Despite targeting of patients to biologic DMARDs based on pretreatment ST characteristics (21 to adalimumab and 9 to abatacept), 48% of patients were non-responders to either treatment at 3 months. Transcriptomic profiling indicated that >3000 genes were found to be differentially expressed between the ST of patients targeted to either treatment at baseline (p<0.05, FDR<0.05). Utilizing t-SNE unsupervised clustering, two distinct clusters (c1 and c2) were identified. Intriguingly, 91% of adalimumab responders were in c1 and all these patients remained in remission at 6 months. Interestingly at baseline, there were no significant differences between c1 and c2 in clinical parameters including DAS28 (p=0.8031), age (p=0.3793), CCP (p=0.3298), RF (p=0.3327), and CRP titres (p=0.5888). When comparing the DGE in patients receiving adalimumab in c1 and c2, unique CD markers upregulated in c2 include CD19, CD79A and B, CD266 and CD27 suggesting an increased lymphocyte profile in c2 compared to c1. To further investigate the cell populations involved in treatment response, we performed cellular deconvolution using CIBERSORT. 22 cell populations were identified in the pre-treatment ST but no significant differences in the cell populations in each cluster were identified. Analysis of post-treatment transcriptomic profiles revealed that 553 genes were differentially expressed post-abatacept (p<0.05, FDR<0.05) while 244 were differentially expressed post-adalimumab (p<0.05, FDR<0.05). 13 genes were found to differ between baseline and 6 months in adalimumab responders (p<0.05, FDR<0.05). These included HST6ST2, SOD2, COL1A1, CPXM1, GRIN2A, FNDC1, CTXN1, NCS1, GDF7, APCDD1L-DT and 3 long non-coding RNAs.

Conclusion: Transcriptomic analysis of ST may enable more accurate prediction of differential treatment responses in early RA patients. Treatment response may thus be more dependent on baseline ST gene expression rather than treatment choice. IHC of ST may be insufficient to identify the heterogeneity in RA ST and further in-depth characterisation may be required to fully comprehend the diversity of RA ST before treatment commences. Further work may allow the development of treatment-predictive biomarkers using a reductive approach which could improve future clinical outcomes for patients with RA.