SAT0423 Molecular profiles associate with clinical disease activity and inform patient subsetting in adult systemic lupus erythematosus

Background Systemic lupus erythematosus (SLE) is characterised by remarkable clinical and pathophysiological diversity, hindering diagnosis, treatment and treatment development. Subsetting of patients based upon clinical presentations alone has not identified homogeneous groups of patients best treated with a directed therapeutic. Objectives To cluster SLE patients into more homogeneous subsets with common molecular pathway signatures, and to assess the clinical, therapeutic, and demographic features enriched in each cluster. Methods Serial or single plasma, serum and RNA samples (n=290) were collected from 198 SLE patients who met ACR classification. Disease activity was assessed by modified SELENA-SLEDAI at an average of 17 visits per patient. Transcriptional co-expression signature module scores were calculated from Illumina Beadchip Microarray gene expression data for 29 immune pathway related modules. Plasma soluble mediators (n=23) and 12 antinuclear autoantibodies (anti-dsDNA, chromatin, ribosomal P, Sm, Sm, SmRNP, RNP, Ro/SSA, La/SSB, centromere B, Scl-70 and Jo-1) were assessed by multiplex bead-based assay and ELISA. Spearman correlations were used for univariate and multivariate analysis using R. Patients were clustered on module signature scores and soluble mediators using random forest and tSNE. Results SLEDAI scores strongly correlated with interferon modules and were modestly correlated with plasmablast and select cell cycle signatures in this adult lupus collection. SLEDAI scores also correlated with soluble levels of IFNa, IL21, IL1a, IL17A, IP10 and MIG. Random forest defined seven clusters of SLE patients with unique molecular phenotypes based upon gene co-expression module signatures and soluble mediators. Inflammation and interferon (IFN) signatures were elevated in Clusters 1 (moderately) and 4, with decreased T cell signatures in Cluster 4. The other clusters had lower IFN and inflammation signatures, but differed in their monocyte, plasmablast and T cell signatures. Clusters 1 and 4 had the highest SLEDAI scores, with high rates of anti-dsDNA, low complement, proteinuria and hematuria; these features were also prominent in Cluster 3, which lacked the IFN and inflammation signatures. Cluster 6 had the highest plasmablast module score, highest IL1a levels, and SLEDAI scored rashes, but only moderate IFN and inflammation module scores. Cluster 2 had higher rates of SLEDAI scored alopecia, a slightly elevated inflammation signature, but lower interferon signature and lower soluble mediator levels. Conclusions SLE subsets can be distinguished by a range of molecular profiles encompassing IFN, T cell, neutrophil, plasmablast, and inflammation co-expression signatures, as well as soluble mediators that vary with disease activity. Prospective longitudinal studies of these molecular profiles may inform clinical trial design and personalised disease management. Acknowledgements This work was supported in part by grants from the National Institutes of Health: U19AI082714, U01AI101934, U54GM104938, and P30AR053483. Disclosure of Interest None declared