Poster #71: Thomas Westbrook

Plasma proteomic correlates of immune-related adverse events in a longitudinal checkpoint inhibitor cohort

Thomas M. Westbrook [1], Felix Li [2], Jaehyun Joo, PhD [1], Akira A. Nair [1] [2], Matthew Lee [1], Boqi Wang, MS [2], Yonghyun Nam, PhD [1], Terri M. Laufer, MD [2], Dokyoon Kim, PhD [1], Sokratis A. Apostolidis, MD [2]. [1] Department of Biostatistics, Epidemiology and Informatics, Perelman School of Medicine, University of Pennsylvania; [2] Division of Rheumatology, Department of Medicine, University of Pennsylvania.

Immune-related adverse events (irAEs) frequently complicate immune checkpoint inhibitor (ICI) therapy for cancer, yet reliable biomarkers to identify susceptible patients remain unavailable. Longitudinal plasma proteomics offers an accessible modality in which to discover baseline risk factors and track on-treatment toxicity. Here, we profiled 951 circulating proteins using the Nomic nELISA platform across approximately 900 longitudinal plasma samples and 275 ICI-treated cancer patients with clinically graded irAEs. To account for dependence between samples due to repeated sampling within patients and broad on-therapy proteomic shifts, statistical significance was evaluated against patient-permuted null distributions, and predictive models were cross-validated using nested feature selection. Unsupervised factorization of the pre-treatment proteome identified a coordinated axis characterized by circulating soluble immune-receptor ectodomains that varies inversely with structural, extracellular matrix, and liver-derived proteins. Patients with lower baseline levels of these soluble receptors exhibited a higher incidence of grade 2+ irAEs alongside longer progression-free survival (PFS), linking toxicity risk and therapeutic efficacy along a shared baseline proteomic axis. Baseline proteomic models significantly outperformed clinical covariates alone in predicting PFS. Furthermore, longitudinal analyses revealed that the chemokines CXCL9 and CXCL10 rose near irAE onset and remained elevated after adjusting for time on therapy, separating an active toxicity signal from general treatment effects. These findings suggest that baseline irAE susceptibility is captured by a distributed immune-receptor signature prior to treatment, while circulating CXCL9 and CXCL10 serve as dynamic markers of active irAE pathology.