Tumor mutation burden analysis in a 5,660 cancer patient cohort reveals cancer type-specific mechanisms for high mutation burden.

Abstract

2589 Background: Tumor mutation burden (TMB), calculated by whole-exome sequencing (WES) or large NGS panels, has an important association with immunotherapy responses. Elucidating the underlying biological mechanisms of high TMB might help develop more precise and effective means for TMB and immunotherapy response prediction. Meanwhile, the landscape of TMB across different cancer types and its association with other molecular features have not been well investigated in large cohorts in China. Methods: Cancer patients whose fresh tissue (n = 1556), formalin-fixed, paraffin-embed (FFPE) specimen (n = 1794), and pleural fluid (n = 84) were profiled using 295- or 520-gene NGS panel. The association of the TMB status with a series of molecular features and biological pathways was interrogated using bootstrapping. Results: TMB, measured by 295- or 520-cancer-related gene panels, were correlated with WES TMB based on in silico simulation in the TCGA cohort. We compared the TMB landscape across 11 cancer type groups and found the highest average TMB in lung squamous cell carcinoma, whereas the lowest TMB was established in sarcoma. High microsatellite instability, DNA damage response deficiency, and homologous recombination repair deficiency indicated significantly higher TMB. The independent predictive power for TMB of twenty-six biological pathways was tested in 10 cancer groups. FoxO signaling pathway most commonly correlated with low-TMB; significant association was identified in four cancer groups. In contrast, no pathway was significantly correlated with high-TMB in more than two cancer groups. Overall, we discovered that the underlying pathways which may be the main drivers of TMB status varied greatly and sometimes had an opposite association with TMB across different cancer types. Moreover, we developed a 14- and 22-gene signature for TMB prediction for LUAD and LUSC, respectively, with only 10 genes shared by both signatures, indicating a histology-specific mechanism for driving high-TMB in lung cancer. Conclusions: The findings extended the knowledge of the underlying biological mechanisms for high TMB and might be helpful for developing more precise and accessible TMB assessment panels and algorithms in more cancer types.