Clinical correlation and diagnostic value of circular RNA determined via next-generation sequencing in lung squamous cell carcinomas

There are investigations suggesting that circular RNAs (circRNAs) may play a critical role in tumor progression, and that their identification could serve as novel biomarkers to improve cancer diagnosis. In this study, we proposed to investigate the landscape of circRNAs in lung squamous cell carcinomas (LSCCs) and to identify those that can aid in lung cancer (LC) diagnosis. The circRNA expression landscape and its functional assessments were profiled in five surgical LSCCs along with five paired adjacent normal tissues via next-generation sequencing (RNAseq) and bioinformatics technology. Correlations were explored between expression levels of validated circRNAs and clinical features of participants. ROC curves and AUC were constructed to evaluate the diagnostic values. Among 256 co-differentially expressed circRNAs (co-DE circRNAs) in LSCCs with paired samples, there were 15 significantly co-upregulated and 93 significantly co-downregulated DE circRNAs. The GO function prediction analyses showed that protein binding, intracellular, and metabolic process were the top three functions of circRNA parental genes. Adherens junctions, lysine degradation, and ErbB signaling pathway were the three most enriched pathways according to KEGG pathway assessment. Of five predicted co-DE circRNAs, including circLIFR, circBRAF, circPTPRM, circEPB41L2, and circPVT1, four co-DE circRNAs except circPVT1 were identified by qRT-PCR with Sanger sequencing in 86 samples (43 surgical LCs with 43 paired adjacent normal tissues). Futhermore, circLIFR, circBRAF, circPTPRM, and circEPB41L2 were significantly associated with pathological subtypes, which had a statistical discrepancy regarding LSCCs discriminated from lung adenocarcinomas. CircLIFR and circEPB41L2 were also negatively correlated with blood SCC. ROC curves showed that circLIFR exhibited the highest AUC value (0.871) in LCs with corresponding high sensitivity and specificity (0.744 and 0.884), and circEPB41L2 had the highest AUC value (0.947) in LSCCs with higher sensitivity (0.867) and specificity (0.947). Combination of these four co-DE circRNAs did not present a high AUC value (AUC: 0.871 in LCs; 0.937 in LSCCs), without corresponding increase in sensitivity (0.767 in LCs, 0.933 in LSCCs) and specificity (0.837 in LCs, 0.867 in LSCCs). This study highlights downregulated four circRNAs (circLIFR, circBRAF, circPTPRM, and circEPB41L2) that might be promising diagnostic biomarkers for LCs, with circEPB41L2 being the best diagnostic biomarker for LSCCs.