Article Text

Download PDFPDF

  1. C. Triaille1,
  2. T. Sokolova1,
  3. S. De Montjoye2,
  4. A. Nzeusseu Toukap2,
  5. L. Meric de Bellefon2,
  6. A. Loriot3,
  7. B. Lauwerys1,
  8. P. Durez1,
  9. N. Limaye4
  1. 1UCLouvain Brussels Woluwe, IREC RUMA, Woluwe-Saint-Lambert, Belgium
  2. 2Cliniques universitaires Saint-Luc (UCLouvain), Service de rhumatologie, Bruxelles, Belgium
  3. 3De Duve Institute UCLouvain, Computational Biology, Woluwe-Saint-Lambert, Belgium
  4. 4De Duve Institute UCLouvain, Genetics of Autoimmune Diseases and Cancer, Woluwe-Saint-Lambert, Belgium


Background Synovitis is the common feature across all individuals with a diagnosis of rheumatoid arthritis (RA). Yet, cellular and transcriptomic alterations occuring in RA synovium are highly variable amongst patients. So far, most data on clinical-tissue correlations either rely on hypothesis-driven approaches or are potentially biased by heterogeneous clinical characteristics (e.g. disease duration or disease-modifying antirheumatic drugs).

Objectives We used transcriptomic profiling of synovial tissue from early, untreated rheumatoid arthritis patients (ERA) to 1/ identify the genes with the most variable expression amongst patients and 2/ explore the ability of unbiased (data-driven) approaches to define clinically relevant ERA subgroups.

Methods Synovial biopsies were harvested from clinically involved joints of ERA patients using needle arthroscopy or ultrasound-guided biopsy. Data on disease activity were collected at inclusion. For each sample, 350ng total RNA was sent for RNAsequencing using a standardized protocol (Macrogen Europe). After quality control (Fast QC) and genome alignement (HiSat2), normalized read counts were analyzed on Qlucore Omics Explorer. To focus on inter-sample heterogeneity, genes were filtered based on variance (σ/σmax). Unbiased approaches (Principal Component Analysis, Unsupervised Clustering) were applied to define patients’ clusters. Pathway enrichment analysis were performed on Metascape. CibersortX was used to extrapolate the immune cell subsets relative composition from gene expression data. All other statistical analyses were performed on GraphPad Prism v9.

Results Total RNA was obtained from synovial biopsies from 74 patients. We first applied variance filtering to identify the genes whose expression showed the greatest variation between patients (n = 894 most variable genes). PCA analysis on the level of expression of these genes did not divide samples into distinct groups, instead yielding a continuous distribution broadly associated with baseline disease activity, as measured by DAS28CRP. Consequently, we used unsupervised clustering to allow for unbiased definition of two patient clusters (PtC): PtC1 (n=52) and PtC2 (n=22) based on their expression of these 894 genes. Pathway analysis of these genes revealed significant enrichment of immune system genes, in the Inflammatory response and Rheumatoid Arthritis pathways (gene cluster 1: GC1), B cell & plasma cell-related pathways (GC2) and metabolic processes-related genes (GC3). Interestingly, PtC1 and PtC2 were characterized by very different clinical features. More specifically, patients from the group with a strong B & plasma cell signature (PtC1) displayed higher baseline indices of all disease activity score components (median DAS28CRP: 5.56 vs 4.09; p-value = 0.0003). They also had higher rates of baseline radiological erosions (erosive disease in 34.6 % vs 10%; p-value = 0.0252) but similar rates of seropositive disease. In line with our pathway analyses, we found a higher signature (inferred relative frequency) of B & plasma cells, T cells and M1-like macrophages in PtC1 compared to PtC2 synovia. PtC2 synovia instead had relatively higher M2-like macrophage and resting mast cell signatures.

Conclusion In this large synovial biopsy study, we found that synovial transcriptomic profiles in ERA patients distribute continuously based on the expression of inflammatory and immune cell transcriptomic pathways. These synovial transcriptomic signatures correlate strongly with systemic disease activity.

Acknowledgements This work was funded in part by unrestricted grants from Cap48 (RTBF), the Fonds de la Recherche Scientifique (FNRS), and the Fund for Scientific Research in Rheumatology (FWRO/FRSR), managed by the King Baudouin Foundation. CT is funded by the FNRS and Fondation Saint-Luc (Cliniques Universitaires Saint-Luc). NL is a chercheur qualifiée of the FNRS.

Disclosure of Interests Clément Triaille: None declared, Tatiana Sokolova: None declared, Stéphanie de Montjoye: None declared, Adrien Nzeusseu Toukap: None declared, Laurent Meric de Bellefon: None declared, Axelle Loriot: None declared, Bernard Lauwerys Shareholder of: BL owns shares (<15000€) in DNALytics, Employee of: BL is currently employed at UCB Biopharma, Patrick Durez: None declared, Nisha Limaye: None declared.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.