Combination therapy for colorectal cancer with anti-PD-L1 and cancer vaccine: A multiscale mathematical model of tumor-immune interactions

TL;DR

A multiscale mathematical model of interactions among tumors, immune cells, and cytokines to investigate tumor evolutionary dynamics under different therapeutic strategies demonstrated that a multiple low-dose regimen significantly reduced advanced tumor burden compared to baseline treatment in anti-PD-L1 therapy.

Abstract

The tumor-immune system plays a critical role in colorectal cancer progression. Recent preclinical and clinical studies showed that combination therapy with anti-PD-L1 and cancer vaccines improved treatment response. In this study, we developed a multiscale mathematical model of interactions among tumors, immune cells, and cytokines to investigate tumor evolutionary dynamics under different therapeutic strategies. Additionally, we established a computational framework based on approximate Bayesian computation to generate virtual tumor samples and capture inter-individual heterogeneity in treatment response. The results demonstrated that a multiple low-dose regimen significantly reduced advanced tumor burden compared to baseline treatment in anti-PD-L1 therapy. In contrast, the maximum dose therapy yielded superior tumor growth control in cancer vaccine therapy. Furthermore, cytotoxic T cells were identified as a consistent predictive biomarker both before and after treatment initiation. Notably, the cytotoxic T cells-to-regulatory T cells ratio specifically served as a robust pre-treatment predictive biomarker, offering potential clinical utility for patient stratification and therapy personalization.

Figures (15)

• Figure 1: Framework flowchart of the dynamic regulatory network. Immune response initiation depends on tumor-specific antigens released by tumors and provided by cancer vaccines. Vaccine adjuvants enhance dendritic cell maturation and significantly improve their antigen-presenting capacity. Naïve T cells recognize tumor-specific antigens presented by dendritic cells through TCR-pMHC binding. Driven by various cytokines, naïve T cells differentiate into effector T cell subsets, which subsequently mediate tumor cell apoptosis through direct contact or cytokine secretion. Tumor cells exert immunosuppressive effects via PD-L1/PD-1 interactions with T cells. Solid black lines represent changes in cell state. Dashed black lines represent intercellular interactions. Colored dotted lines represent cytokine production and the action mechanism. Indigo and red colors represent the action mechanisms of cancer vaccines and PD-L1 inhibitors, respectively. Arrows denote promotion, proliferation, or activation, while blocking arrows indicate killing, blocking, or inhibition.
• Figure 2: Research framework overview. This study consists of five main components. (1) Experimental data: Primarily including treatment regimens, temporal changes in tumor volume, and generated virtual sample cohort data. (2) Biological mechanisms: Mainly encompassing regulatory mechanisms of cell-cell interactions, cytokine networks, and PD principles. (3) Mathematical modeling: Integrating cell dynamics, cytokine dynamics, and PK models. (4) Methodology: Comprising approximate Bayesian computation (ABC), qualitative and quantitative parameter analysis, and virtual patient generation. (5) Results: Model-experiment fitting, therapeutic efficacy evaluation, Bliss combination index, immune cell heterogeneity distributions, and ROC analysis.
• Figure 3: Tumor evolution dynamics and model validation under different treatment strategies. (A) Schematic of experimental design. (B) Experimental measurements (scatter points) versus model-predicted curves (solid lines) of tumor volume changes across treatment strategies. (C) $\sim$ (F) Individualized treatment response analysis: Tumor growth dynamics for 20 mice in control (C), cancer vaccine (D), anti-PD-L1 monotherapy (E), and combination therapy (F) groups. Scatter points represent experimental data derived from Liu.NatCancer.2022, while curves show numerical results from the tumor heterogeneity model. Parameter values corresponding to the tumor heterogeneity modeling framework are indicated in the upper-left text. Panel (B) corresponds to the baseline parameters in Table \ref{['Tab:Parameter']}. Panels (C)-(F) display the best-fit results for 1,000 virtual patients based on approximate Bayesian parameter selection.
• Figure 4: Parameter posterior distribution based on approximate Bayesian calculation. The diagonal elements represent the marginal posterior distributions calculated via weighted Gaussian kernel density estimation. The lower triangular region displays two-dimensional contour projections of different parameter combinations, with dark red areas corresponding to high probability density intervals.
• Figure 5: Global sensitivity analysis of the baseline system without treatment. (A) Sensitivity of system variables to model parameters was assessed using the Sobol method. Results are shown for (B) dendritic cells, (C) helper T cells, (D) regulatory T cells, (E) cytotoxic T cells, and (F) tumor cells.