Analysis of proliferating Treg cells as a predictor for immunotherapy response using systemic longitudinal immune profiling: Biomarker analysis of the phase 3 CONTINUUM trial.

140 Background: The discovery of predictive biomarkers for immunotherapy is hindered by the lack of control groups in most translational studies, making it impossible to discern whether the identified biomarkers are merely prognostic or truly predictive for treatment outcomes. This study reported on prospectively designed biomarker analyses of the phase 3 CONTINNUUM trial (NCT03700476), the first trial that demonstrated prolonged event-free survival (EFS) with the addition of anti-PD-1 (aPD1) to chemoradiotherapy (CRT) in locoregionally-advanced nasopharyngeal carcinoma (LA-NPC). Methods: Mass cytometry was performed to generate a dynamic single-cell atlas on longitudinal peripheral blood mononuclear cell (PBMC) samples from 12 pairs of matched patients with or without relapse in the aPD1-CRT arm. The machine-learning algorithm CellCnn was applied to identify a cell subset that was most strongly associated with disease relapse, which was further confirmed by flow cytometry (n = 120). Single-cell RNA sequencing data from matched PBMC and tumor samples were analyzed and multiplex immunohistochemistry was performed on baseline tumor samples (n = 249) to verify the predictive value of the cell subset within tumors. Results: Baseline circulating regulatory T cells (Tregs), with a Ki67 + proliferating phenotype, were found to display a higher frequency in LA-NPC patients who developed posttreatment relapse ( P < 0.001), which was further validated by flow cytometry. The intratumoral Ki67 + Tregs were also identified and correlated with an immunosuppressive tumor microenvironment. In patients with a low baseline level of Ki67 + Tregs, additional aPD1 significantly improved EFS compared with CRT alone (log-rank P = 0.011, HR = 0.22, 95% CI: 0.06−0.79), while EFS was almost identical in both treatment arms in patients with high Ki67 + Treg levels (log-rank P = 0.995, HR = 1.00, 95% CI: 0.49−2.05; interaction P = 0.047). This predictive value was further validated in datasets from two independent randomized trials in renal cancer. Conclusions: The frequency of Ki67 + Tregs can serve as a reliable predictive tool in selecting patients who are more likely to benefit from immunotherapy, potentially informing individualized immunotherapy strategies.