NONINVASIVE HIGH-THROUGHPUT SERUM PROTEOMICS FOR DISTINGUISHING SUBTYPES OF LUPUS NEPHRITIS

PT006 / #521 Topic: AS15 - Lupus Nephritis-Clinical POSTER TOUR 02: RECENT INSIGHTS ON THE PATHOGENESIS OF LUPUS NEPHRITIS 23-05-2025 10:00 AM - 10:40 AM Background/Purpose Lupus nephritis (LN) treatment decisions are commonly guided by histopathological classifications based on the ISN/RPS and NIH activity and chronicity indices. Since LN class and activity may shift over time, treatment adjustments are often necessary. However, repeated kidney biopsies are invasive and impractical, highlighting the need for noninvasive biomarkers to inform LN classification and guide therapy. In this study, we analyzed serum proteomic profiles to identify noninvasive biomarkers reflective of histological class, activity, and chronicity indices. Methods This study recruited 196 SLE patients with lupus nephritis (LN) as part of the AMP RA/SLE network. Each patient underwent a kidney biopsy evaluated by a renal pathologist for LN classification using the ISN/RPS system and NIH activity and chronicity indices. Serum samples were collected at biopsy to explore noninvasive biomarkers. High-throughput proteomic analysis was conducted using the Olink Explore HT platform to identify protein expression patterns linked to LN class, activity, and chronicity. Multivariate logistic regression, adjusted for age, gender, and genetic ancestry, along with random forest algorithms, were used to pinpoint potential biomarkers to guide LN treatment decisions. Results Compared to healthy controls, LN patients upregulated multiple pathways related to the innate and adaptive immune systems, including TNF, IL-10, efferocytosis, and antigen processing and presentation pathways. Patients with pure proliferative LN (class III or IV) showed further upregulation in B cell receptor signaling, Th1/Th2 differentiation, neutrophil degranulation, Th17 differentiation, and leukocyte chemotaxis pathways compared to those with minimal disease (class I/II), membranous (V), or mixed proliferative (III/IV+V) LN. Machine learning models using a decision-tree-based boost algorithm achieved high accuracy for distinguishing healthy controls (95.3% [86.9%-99%]) and LN patients (99.5%, [97% - 100%]), as well as advanced sclerosing (class VI), compared to other classes (AUC, 0.85 ± 0.11; accuracy, 88.1% ± 0.7%). When distinguishing membranous vs pure proliferative classes, the ML model showed a modest prediction performance with an AUC of 0.75 ± 0.06 with a cross-validation accuracy of 71.1% ± 0.6%. When compared to healthy controls, there are 862 upregulated proteins, including interferons, IL-10, and lymphocyte surface receptors, shared among patients with membranous, proliferative, and mixed classes and 92 downregulated proteins, including C2, C4, and C8 (Figure 1C). In addition, the expression of 398 and 2252 proteins was associated with the NIH activity and chronicity indices, respectively (Figure 1D). Specifically, proteins involved in IL-18, TNF, and IL-1 pathways and intracellular proteins from multiple organ systems with prominent enrichment in immune cells positively correlated with the activity index (Figure 1D). Interestingly, proteins enriched in interferon, growth factor and neurotrophin receptor pathways and intracellular proteins from multiple organ systems, particularly the nervous system, correlated with the chronicity index (Figure 1E). Figure 1. Conclusions This study revealed that lupus nephritis (LN) patients exhibited significant upregulation of immune pathways, including TNF and IL-10, compared to healthy controls, particularly in proliferative LN. A machine learning model effectively distinguished LN patients from healthy controls and showed moderate performance in differentiating membranous from proliferative LN. Proteomic analysis identified proteins associated with NIH activity and chronicity indices, underscoring the potential of serum proteomics as a noninvasive tool for LN classification and monitoring.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.