Evolutionary trajectories of immune escape across cancers

Abstract

Immune escape is a critical hallmark of cancer progression and underlies resistance to multiple immunotherapies. However, it remains unclear when the genetic events associated with immune escape occur during cancer development. Here, we integrate functional genomics studies of immunomodulatory genes with a tumor evolution reconstruction approach to infer the evolution of immune escape across 38 cancer types from the Pan-Cancer Analysis of Whole Genomes dataset. Different cancers favor mutations in different immunomodulatory pathways. For example, the antigen presentation machinery is highly mutated in colorectal adenocarcinoma, lung squamous cell carcinoma, and chromophobe renal cell carcinoma, and the protein methylation pathway is highly mutated in bladder transitional cell carcinoma and lung adenocarcinoma. We also observe different timing patterns in multiple immunomodulatory pathways. For instance, mutations impacting genes involved in cellular amino acid metabolism were more likely to happen late in pancreatic adenocarcinoma. Mutations in the glucocorticoid receptor regulatory network pathway tended to occur early, while mutations in the TNF pathways were more likely to occur late in B-cell non-Hodgkin lymphoma. Mutations in the NOD1/2 signaling pathway and DNA binding transcription factor activity tended to happen late in breast adenocarcinoma and ovarian adenocarcinoma. Together, these results delineate the evolutionary trajectories of immune escape in different cancer types and highlight opportunities for improved immunotherapy of cancer.

Keywords: Immune escape; cancer evolution; mutation timing.

Figure 1.. Genetics of antigen presentation machinery across cancer types.

(A) Prevalence of antigen presentation machinery mutations in PCAWG-TCGA samples. (B) Prevalence of antigen presentation machinery mutations in each PCAWG-TCGA cancer type. (C) Tumor mutation burden with antigen presentation machinery mutation in each cancer type. (D) Different tumor microenvironments inferred by CIBERSORT. The dot size shows the mean ratio of cell score between samples with APM and the ones without APM mutations. The color shows the FDR for adjusting every comparison using the Wilcoxon test. * show the significance of the comparison. (E) Differential enrichments by antigen presentation machinery mutations in multiple cancer types. The red box presents the up-regulation of the gene set in the samples with APM mutations and the blue box shows the down-regulated ones. The gray box indicates insignificant enrichments.

Figure 2.. Timing of mutations in the antigen presentation machinery across cancer types.

(A) Definition of early/late mutations in cancer samples. CN: Copy number; SNV: single nucleotide variant. (B) Timing of driver genes in stomach adenocarcinoma. Green represents early timing, transitioning gradually to purple for late timing. The color intensity reflects the proportion of early or late events out of 250 samplings, indicating the likelihood of an event occurring early or late. (C) An example of the timing of antigen presentation machinery mutations in bladder transitional cell carcinoma. (D) Comparison of the timing of antigen presentation machinery point mutations and the background timing.

Figure 3.. Identification of genetic pathways with immunomodulatory effects.

(A) Workflow of identifying CRISPR screen studies for genetic pathways with immunomodulatory effects. (B-C) Summary of CRISPR screen studies reporting the enrichments of MHC-I regulators (B) and T cell killing to tumor cells (C). (D) Prevalence of mutations in the key pathways for T cell killing to tumor cells in all PCAWG cancer samples. (E) Prevalence of mutations in the key pathways for T cell killing to tumor cells in different cancer types. (F-G) Prevalence of point mutations in the representative pathways (autophagy and IFNγ signaling) for the regulation of sensitivity of tumor cells to T cell killing in all PCAWG samples.

Figure 4.. Timing of mutations in genetic pathways with immunomodulatory effects.

(A) Positive selection of mutations in regulators for the response of cancer cells to T-cell killing across cancers. (B) Positive selection of mutations in regulators for the response of cancer cells to T-cell killing in each cancer. (C-D) Timing of mutations in genetic pathways with immunomodulatory effects in pancreatic adenocarcinoma (C) and B-cell non-Hodgkin lymphoma (D). Green represents early timing, transitioning gradually to purple for late timing. The color intensity reflects the proportion of early or late events out of 250 samplings, indicating the likelihood of an event occurring early or late. Right Panel: The event is classified as “early” or “late” at the cancer-type level. If the ratio < 0.5, the event is classified as “early”. If the ratio > 2, the event is classified as “late”. Ratio = (Number of late + 1) / (Number of late + 1).