AUTHOR=Dai Dongjun , Guo Yinglu , Shui Yongjie , Li Jinfan , Jiang Biao , Wei Qichun 

TITLE=Combination of Radiosensitivity Gene Signature and PD-L1 Status Predicts Clinical Outcome of Patients With Locally Advanced Head and Neck Squamous Cell Carcinoma: A Study Based on The Cancer Genome Atlas Dataset

JOURNAL=Frontiers in Molecular Biosciences

VOLUME=Volume 8 - 2021

YEAR=2021

URL=https://www.frontiersin.org/journals/molecular-biosciences/articles/10.3389/fmolb.2021.775562

DOI=10.3389/fmolb.2021.775562

ISSN=2296-889X

ABSTRACT=Aim: The aim of our study was to investigate the potential predictive value of the combination of radiosensitivity gene signature and PD-L1 expression for the prognosis of locally advanced HNSCC.
Methods: The cohort was selected from TCGA and classified into radiosensitive (RS) group and radioresistant (RR) group by a radiosensitivity-related gene signature. The cohort was also grouped as PD-L1-high or PD-L1-low based on PD-L1 mRNA expression. The least absolute shrinkage and selection operator (lasso) based cox model was used to select hub survival genes. An independent validation cohort was obtained from GEO database.
Results: We selected 288 locally advanced HNSCC patients for analysis. The Kaplan–Meier method found that RR & PD-L1-high group had a worse survival than others (p = 0.033). The differentially expressed gene (DEG) analysis identified 553 upregulated genes and 486 downregulated genes (p < 0.05, fold change > 2) between RR & PD-L1-high group and others. The univariate Cox analysis of each DEG and subsequent lasso-based Cox model revealed 5 hub survival genes (POU4F1, IL34, HLF, CBS and RNF165). A further hub survival genes-based risk score model was constructed, which was validated by external cohort. It found that higher risk score predicted worse prognosis (p = 0.0013). The AUC plots showed that this risk score model had well prediction value (1-year-AUC = 0.684, 2-year-AUC = 0.702, 3-year-AUC = 0.688). The 5 different deconvolution methods all showed the B cells were lower in RR & PD-L1-high group (p < 0.05). Finally, connectivity mapping analysis showed that the histone deacetylase inhibitor trichostatin A might have the potential to reverse the phenotype of RR & PD-L1-high locally advanced HNSCC (p < 0.05, false discovery rate < 0.1). 
Conclusion: The combination of 31-gene signature and the PD-L1 mRNA expression had potential predictive value for the prognosis of locally advanced HNSCC who had RT. The B cells were lower in the RR & PD-L1-high group. The identified risk gene signature of locally advanced HNSCC and the potential therapeutic drug trichostatin A for RR & PD-L1-high group worth to be further studied in a prospective homogenous cohort.