Integration of Tumor Heterogeneity for Recurrence Prediction in Patients with Esophageal Squamous Cell Cancer

SIMPLE SUMMARY: This manuscript reports a deep sequencing study comprehensively analyzing the clinical impact of mutations considering the abundance of mutations. We built an eight-gene mutation predictor considering intratumoral heterogeneity to predict post-surgery recurrence in ESCC patients. Unlike previous studies that simply treated mutations as binary variables (mutant and wild type), we quantified mutations by the fraction of cancer cells carrying the mutations, and our results showed that the cancer cell fraction of mutations was more informative than the mutation status of genes in recurrence prediction. The predictor was further validated as a powerful recurrence indicator in our validation set and the TCGA-ESCC cohort. With the popularization of targeted deep sequencing in clinical work, our study will help clinicians make accurate predictions of recurrence for patients and will provide a new perspective in the clinical transformation of genomic findings. ABSTRACT: Esophageal squamous cell carcinoma (ESCC) is one of the deadliest malignancies in China. The prognostic value of mutations, especially those in minor tumor clones, has not been systematically investigated. We conducted targeted deep sequencing to analyze the mutation status and the cancer cell fraction (CCF) of mutations in 201 ESCC patients. Our analysis showed that the prognostic effect of mutations was relevant to the CCF, and it should be considered in prognosis prediction. EP300 was a promising biomarker for overall survival, impairing prognosis in a CCF dose-dependent manner. We constructed a CCF-based predictor using a smooth clipped absolute deviation Cox model in the training set of 143 patients. The 3-year disease-free survival rates were 6.3% (95% CI: 1.6–23.9%), 29.8% (20.9–42.6%) and 70.5% (56.6–87.7%) in high-, intermediate- and low-risk patients, respectively, in the training set. The prognostic accuracy was verified in a validation set of 58 patients and the TCGA-ESCC cohort. The eight-gene model predicted prognosis independent of clinicopathological factors and the combination of our model and pathological staging markedly improved the prognostic accuracy of pathological staging alone. Our study describes a novel recurrence predictor for ESCC patients and provides a new perspective for the clinical translation of genomic findings.