Hypoxia Methylome Classifier (Hdmc) Outperforms Gene Signatures In Identifying Hpv-Negative Hnscc Patients At Risk For Locoregional Failure Post Primary Radiochemotherapy: A German Cancer Consortium Radiation Oncology Group (Dktk-Rog) Multicenter Trial

Tumor hypoxia is a negative prognostic factor in radiochemotherapy of head and neck squamous cell carcinoma (HNSCC). However, lack of robust and reliable biomarker surrogating tumor hypoxia levels limits its clinical implementation. We have evaluated three Gene Expression-based Signatures of Hypoxia (GESoH) in TCGA-HNSCC cohort. Validation studies in the present multicenter retrospective DKTK-ROG HNSCC primary RCHT trial revealed significant prognostic value for two GESoH (15- and 30-gene sets) in for smaller tumors (GTV<19ccm). Therefore, we sought to discover and validate the impact of the epigenetic fingerprint of tumor hypoxia on DNA-methylome level. To identify the prognostic impact of tumor hypoxia on increased risk for loco-regional recurrence (LR) and all event progression in Human Papilloma Virus DNA negative (HPV-) HNSCC, methylation data (450K Illumina, FFPE material) was obtained from the homogeneous primary RCHT DKTK-ROG validation cohort (total n = 117, HPV- n = 88). HDMC was trained in the TCGA-HNSCC cohort (total n = 278, HPV- n = 242) where matching RNA-seq based assignments for 15- and 30 GEsOH as well as methylome data were available. A random forest machine learning based HDMC was developed based on differentially methylated probes (5129, FDR<0.05) between tumors with high GEsOH (consensus, n = 96) vs those with intermediate-to low GEsOH (n = 146). HDMC was validated in the DKTK-ROG primary cohort. HDMC-predicted classes (high 53% vs low 47%) were associated with significant differences in overall survival (OS, p<0.05) in the TCGA training cohort. HDMC-based assignment (41% high vs 59% low) revealed significant correlation with OS (p<0.02), progression (p<0.002) and LR (p<0.0006) in HPV- patients of the DKTK-ROG validation cohort. HDMC remained significant on adjusted multivariate analysis for anatomical site, age, gender and T-stage for LR, progression and OS in the DKTK-ROG validation cohort (HRs 1.8-2.2, p<0.05). A methylation-based classifier of tumor hypoxia is successfully developed and validated to be prognostic for LR, progression and OS in HPV- HNSCC patients treated with primary RCHT.