The Association of Metabolic and Genetic Heterogeneity in Head and Neck Squamous Cell Carcinoma with Prognostic Implications: Integration of FDG PET and Genomic Analysis

360 Purpose: The linkage between genetic and phenotypic tumor heterogeneity has not been well evaluated. Herein, we investigated the association between tumor genetic and metabolic heterogeneity features based on 18F-fluorodeoxyglucose positron emission tomography (FDG PET) in head and neck squamous cell carcinoma (HNSC) and further assessed the prognostic significance of the features. Methods: Genomic and clinical data were obtained from the head and neck squamous cell carcinoma dataset of the cancer genome atlas (TCGA-HNSC) (n = 499). The FDG PET data of the patients included in TCGA-HNSC was obtained from cancer imaging archive (TCIA). Twenty-five patients were available for both genomic and FDG PET data. After primary tumor segmentation in FDG PET scans, two tumor metabolic features [maximum standardized uptake value (SUVmax), peak standardized uptake value (SUVpeak)], two metabolic-volumetric [total lesion glycolysis (TLG), metabolic tumor volume (MTV)] and two metabolic heterogeneity features [entropy, and coefficient of variation (COV)] were obtained. The mutant-allele tumor heterogeneity (MATH) value, which is a genetic heterogeneity signature, was calculated from somatic variants data. Genetic glycolysis signature (GlycoS) was obtained from RNA sequencing data using gene set enrichment analysis (Ref. 1). These genetic and FDG PET features were analyzed using Spearman’s correlation analysis, log rank test, and Cox regression analysis (Fig. A). Results: Metabolic heterogeneity features as well as metabolic-volumetric features showed significant association with the genetic heterogeneity feature, MATH (ρ = 0.521, P = 0.008 for MTV; ρ = 0.472, P = 0.017 for TLG; ρ = 0.488, P = 0.013 for entropy; ρ = 0.402, P = 0.047 for COV) (Fig. B). We also evaluated the association between GlycoS calculated by gene expression and FDG PET features to use GlycoS as a surrogate of tumor metabolic features. As a result, TLG, MTV and entropy from FDG PET showed significant association with the GlycoS (ρ = 0.590, P = 0.002 for MTV; ρ = 0.570, P = 0.004 for TLG; ρ = 0.519, P = 0.009 for entropy). The GlycoS showed a trend of positive correlation with COV, SUVmax and SUVpeak, while they did not reach statistical significance (ρ = 0.393, P = 0.057 for COV; ρ = 0.331, P = 0.114 for SUVmax; ρ = 0.272, P = 0.291 for SUVpeak). We assessed the prognostic value of the FDG PET features. SUVmax, MTV, TLG, and entropy were predictive of overall survival (OS) (P < 0.05 for the features). We further analyzed the prognostic value of MATH and GlycoS in 499 patients. We found that MATH and GlycoS were highly predictive of OS in univariate analysis (P = 0.002 for MATH; P = 0.0001 for GlycoS) (Fig. C). MATH and GlycoS were predictive of OS even after adjustment using clinicopathologic features (age, sex, and tumor stage) in multivariate Cox regression analysis. Furthermore, both MATH and GlycoS were still significant prognostic factors even after including both features and the clinicopathologic features in the same model. This result indicates that the both features have an additive role over each other to predict OS (P = 0.015 for MATH; P = 0.006 for GlycoS). Conclusions: Tumor metabolic heterogeneity as well as metabolic-volumetric features assessed by FDG PET were closely associated with genetic heterogeneity in HNSC. This finding sheds light on how genetic heterogeneity affects tumor metabolic heterogeneity and may strengthen the validity of both metabolic and genetic features for evaluation of tumor heterogeneity. Furthermore, the genetic heterogeneity (MATH) and glycolysis features (GlycoS) were independent prognostic factors in HNSC, which implies the potential of using both metabolic and heterogeneity features for precise prognostication after validation in prospective studies. (Ref. 1) Choi H, Na K. Pan-cancer analysis of tumor metabolic landscape associated with genomic alterations. Mol Cancer. 2018;17(1):150.