Genomic correlates and classification of PD-L1 status in non-small cell lung cancer.

Abstract

e21578 Background: Programmed cell death 1 (PD-L1) is the first FDA-approved predictive biomarker for non-small cell lung cancer (NSCLC) patients treated with PD-(L)1 blockade therapy. Herein, we aim to identify potential anti-PD-L1 treatment-related biomarkers through evaluating the correlation between the PD-L1 expression level, clinical characteristics, and the mutational profile of a large Chinese NSCLC cohort. Methods: Genomic profiling of tumor biopsies from a total of 808 Chinese NSCLC patients, including 651 adenocarcinomas (ADCs) and 157 squamous cell carcinomas (SCCs), was performed using next-generation sequencing by targeting 425 cancer-relevant genes. Immunohistochemical analysis was used to evaluate PD-L1 protein expression using PD-L1 antibodies including DAKO 22C3 ( N= 695) and DAKO 28-8 ( N= 113), respectively. Results: The PD-L1 positive ( > 1%) rate was 49.2% in ADCs and 52.9% in SCCs, respectively. PD-L1 expression (22C3) was associated with the male gender( p< 0.01) and lymph node metastasis ( p= 0.048) in ADCs but not in SCC patients. PD-L1 expression (22C3) was inversely correlated with KRAS wildtype ( p< 0.001) and EGFR exon 19 deletion( p< 0.01) in ADC, while it was negatively associated with TP53 oncogenic mutations ( p= 0.049) in SCC. Copy number variation analysis revealed that MDM2 amplification ( p= 0.027), 1q gain ( p= 0.012), and 5q deletion ( p< 0.01) negatively correlated with PD-L1 expression, whereas PD-L1 and PD-L2 amplification ( p< 0.001 and p< 0.0001) were positively associated with PD-L1 expression in ADCs. In SCCs, PD-L1 expression (22C3) was negatively associated with copy number gain in EGFR ( p= 0.040), MDM2 ( p= 0.044), 14q ( p= 0.032), and 20q ( p= 0.026), along with PTPRD loss (p = 0.015) and 19p deletion (p = 0.025). However, it was positively associated with 9p amplification ( p< 0.01) and 13q deletion ( p= 0.019). Plus, KIF5B- RET ( p= 0.006) appeared to be inversely related to the PD-L1 expression levels (22C3) in ADCs alone. In addition, these predicted biomarkers were used to delineate the receiver operating characteristic (ROC) calculation to discriminate between PD-L1 low and high (22C3, 50%) with an AUC score of 0.779. Lastly, PD-L1 expression (28-8) did not show significant correlation with any detected oncogenic mutations, but negatively correlated with NKX2-1 gain ( p= 0.0379) and 9q deletion ( p= 0.0379) in ADCs. Conclusions: This study revealed the correlation between PD-L1 protein expression, clinical features, and mutational traits in NSCLC patients, and provided a classifier for PD-L1 expression prediction.