Renal and Peripheral Blood Transcriptome Signatures That Predict Treatment Response in Proliferative Lupus Nephritis—A Prospective Study

Abstract

Mechanisms contributing to non-response to treatment in lupus nephritis (LN) are unclear. We characterised the transcriptome of paired peripheral blood mononuclear cells (PBMCs) and renal tissues in LN before and after cyclophosphamide (CYC) treatment and identified markers that predicted treatment response. Total RNA isolated from paired PBMCs (n = 32) and renal tissues (n = 25) of 16 proliferative LN before CYC treatment, 6 months post-treatment, and during renal flare, was sequenced on Illumina Novaseq-6000 platform. Post-treatment, eight patients were clinical responders (CR), of whom four flared (FL), and eight were non-responders (NR). Comparative transcriptomic analyses before and after treatment within CR, NR, and FL groups was performed using DESeq2. Weighted gene co-expression network analysis (WGCNA) and ROC analysis was performed to identify and validate hub genes predictive of treatment response. Based on this, we observed that pathways such as degradation of cell cycle proteins, expression of G0 and G1 phase proteins, and apoptosis, were upregulated in CR PBMCs post-treatment, while IFN-γ signalling and ECM organisation were downregulated. In NR PBMCs, ECM molecules, neddylation and BCR signalling were upregulated post-CYC treatment, while in NR renal tissue, TLR, IFN and NF-κB signalling pathways were upregulated. In FL PBMCs, neutrophil degranulation and ROS and RNS production in phagocytes were downregulated following treatment, whereas, in the corresponding renal tissue, cell-ECM interactions and ISG15 antiviral mechanism were downregulated. After WGCNA and subsequent ROC analysis, TENM2, NLGN1 and AP005230.1 from PBMCs each predicted NR (AUC-0.91; p = 0.03), while combined model improved prediction (AUC-0.94; p = 0.02). AP005230.1 from renal tissue also predicted non-response (AUC-0.94; p = 0.01) and AC092436.3 from PBMCs predicted renal flare (AUC-0.81; p = 0.04). Our study identified significant DEGs/pathways specific to different treatment outcomes and hub genes that predicted non-response and renal flare.