Modeling CISH-KO TIL therapy: T-cell persistence and endogenous competition determine response in gastrointestinal cancer

Abstract

BackgroundTumor-infiltrating lymphocyte (TIL) therapies have shown promise in murine models; however, early-phase clinical trials reveal substantial heterogeneity in treatment response across patients. Elucidating the biological drivers of this variability is critical for improving patient selection and therapeutic design. MethodsWe developed a mechanistic mathematical model of Cish-inactivated TIL therapy capturing interactions among tumor cells, endogenous TILs, and infused CISH-KO (Cish gene knockout) T-cells. The model was calibrated to longitudinal tumor size data from two murine studies and also fitted to clinical data from a first-in-human phase 1 trial enrolling 22 and treating 12 patients with metastatic gastrointestinal cancers (27 independently tracked tumor lesions). We used the calibrated model to identify patient-level determinants of response and to simulate combination and fractionation treatment strategies in silico. FindingsFour mechanistically distinct parameters emerged as dominant determinants of therapeutic outcome: CISH-KO T-cell persistence (r = 0.827, p < 10-4) was the single strongest predictor of therapeutic failure; other major determinants included endogenous T-cell proliferation rate (r = 0.648, p < 10-4) , CISH-KO T-cell killing rate, and intrinsic tumor growth rate. In silico simulations predicted that the addition of anti-PD-1 checkpoint blockade to standard-dose CISH-KO TIL therapy achieved comparable or superior tumor control relative to dose doubling, while maintaining lower peak T-cell level. Additionally, fractionating the same total dose across multiple delayed infusions further improved predicted tumor control by prolonging CISH-KO T-cell persistence, again without raising peak effector burden. InterpretationThe efficacy of current CISH-KO TIL protocols is influenced not only by the cytotoxic potency of infused T-cells, but also by their persistence within the tumor microenvironment and competitive pressure from the reconstituting host immune compartment. These findings support strategies that enhance the functional efficiency and prolong the availability of infused T-cells--such as concurrent PD-1 blockade or fractionated dosing of the same total cell product--rather than dose escalation alone and identify CISH-KO T-cell persistence and endogenous immune competition as actionable targets for improving TIL-based immunotherapies.