THU0014 COMPARATIVE TRANSCRIPTOME ANALYSES ACROSS TISSUES AND SPECIES IDENTIFY TARGETABLE GENES FOR HUMAN SYSTEMIC LUPUS ERYTHEMATOSUS (SLE) AND LUPUS NEPHRITIS (LN)

Background: Systemic Lupus Erythematosus (SLE) is a complex disease associated with the dysfunction of multiple tissues and cells. The causal tissue for each disease phenotype is not known a priori. Despite improvements in diagnosis and treatment, major organ involvement (such as the kidneys) contributes significantly to morbidity and mortality that still remain increased. There is an unmet need for timely targeted therapy. Objectives: RNA-sequencing was performed to investigate the patterns of transcription variation across tissues between healthy and lupus-prone mice at different stages of lupus, and how these patterns associate with human Systemic Lupus Erythematosus (SLE). Methods: NZB/W-F1 lupus prone mice were sacrificed at the pre-puberty, pre-autoimmunity and nephritic stage. Age-matched C57BL/6 were used as controls. An “effector” tissue (spleen) and “end-organs” (kidneys, brain) were collected. Total RNA was isolated, and mRNA-sequencing was performed. A time-series analysis was developed and differentially expressed genes (DEGs) were analyzed with DESeq. Hierarchical clustering and functional enrichment analysis were performed with gProfiler. Human orthologs of mouse tissue DEGs were identified in the whole-blood RNA-sequencing dataset comprised of 55 lupus-nephritis (LN), 65 non-LN SLE patients and 58 healthy individuals (HI). Human orthologs were compared to human DEGs. Using machine learning, human orthologs identified in the mouse dataset were used to predict kidney involvement in the human dataset, which was split in training and validation sets. Results: Lupus susceptibility and progression signatures at different tissues and different stages of the disease were identified. Tissue-specific signatures and a common cross-tissue signature were also described. Previously described and novel biological processes and pathways were revealed. The comparative murine-human transcriptome analysis identified human orthologs from the mouse spleen-signature (including CCL5, IFIT and HLA genes) that are involved in systemic autoimmunity. It also identified human orthologs from the kidney- and brain-signature (including FCGR2A, C1Q, JAK1 and APOA2) that are involved in major “end-organ” damage and response mechanisms. Using a neural network model, 193 human orthologs accurately predicted LN patients vs HI (accuracy=0.86, sensitivity=0.82, specificity=0.91 in the validation set). Using a support vector machine model, 30 human orthologs and age and gender were the best predictors of LN vs non-LN SLE patients (accuracy=0.71, sensitivity=0.73, specificity=0.69 in the validation set). Conclusion: Murine tissue gene signatures identified by RNA-sequencing analysis revealed biological processes and pathways that could be potentially used as biomarkers or therapeutic targets in human SLE. Comparison of the murine tissue-transcriptome with the whole-blood human-transcriptome revealed common gene signatures, demonstrating similar biological processes and pathways. Machine learning identified a murine kidney lupus signature that can accurately predict kidney involvement in human SLE. Validation in other datasets is ongoing. References: [1]Panousis NI, et al. Ann Rheum Dis 2019;78:1079 Acknowledgments: This work was supported by FOREUM, SYSCID and ERC -Advanced Grant Disclosure of Interests: Eleni Frangou: None declared, Panayiotis Garantziotis: None declared, Maria Grigoriou: None declared, Aggelos Banos: None declared, Nikolaos Panousis: None declared, Emmanouil Dermitzakis: None declared, George Bertsias Grant/research support from: GSK, Consultant of: Novartis, Dimitrios Boumpas: None declared, Anastasia Filia: None declared