MicroRNA induction by copy number gain is associated with poor outcome in squamous cell carcinoma of the lung

Abstract

Copy number gains in cancer genomes have been shown to induce oncogene expression and promote carcinogenesis; however, their role in regulating oncogenic microRNAs (onco-miRNAs) remains largely unknown. Our aim was to identify onco-miRNAs induced by copy number gains in human squamous cell carcinoma (Sq) of the lung. We performed a genome-wide screen of onco-miRNAs from 245 Sqs using data sets from RNA-sequencing, comparative genomic hybridization, and the corresponding clinical information from The Cancer Genome Atlas. Among 1001 miRNAs expressed in the samples, 231 were correlated with copy number alternations, with only 11 of these being highly expressed in Sq compared to adenocarcinoma and normal tissues. Notably, miR-296-5p, miR-324-3p, and miR-3928-3p expression was significantly associated with poor prognosis. Multivariate analysis using the Cox proportional hazards model showed that miRNA expression and smoking were independent prognostic factors and were associated with poor prognosis. Furthermore, the three onco-miRNAs inhibited FAM46C to induce MYC expression, promoting proliferation of Sq cells. We found that copy number gains in Sq of the lung induce onco-miRNA expression that is associated with poor prognosis.

Figure 1

Flow diagram showing the miRNA selection process used in this study.

Figure 2

Survival probability and ROC curve for Sq patients according to miRNA expression. (a) Kaplan-Meier curve showing that higher miRNA expression was associated with poor prognoses. P-values were calculated by log-rank test. (b,c) The ROC curve indicates that higher expression of individual miRNAs is a good indicator to separate Sq from normal tissues (N) (b) or Sq from Ad (c).

Figure 3

Characteristics of the three miRNAs associated with poor prognoses. (a) Density scatter plots showing the association between miRNA (miR-296-5p, miR-324-3p, and miR-3928-3p) expression and CNV in the precursor locus, particularly for copy number gains. The curved line was estimated with R function “lowess”. (b) Boxplots showing the higher expression of the three miRNAs of interest in Sq samples relative to those in Ad and normal (N) samples. P-values were calculated with the Wilcoxon rank sum test: (*) P-value < 0.05; (**) P-value < 0.01.

Figure 4

Target prediction and experimental validation. (a) Comparison of the common genes downregulated by miR-296-5p, miR-324-3p, and miR-3928-3p in PC10 or ACC-LC-73 cells. (b) Barplot demonstrating the ability of miRNA mimics to reduce FAM46C mRNA expression in PC10 or ACC-LC-73 cells. (c) Boxplot showing the decrease in FAM46C expression in Sq compared to Ad and normal tissues (N). P-values were calculated with the Wilcoxon rank sum test: (**) P-value < 0.01. (d) Barplot showing the ability of each miRNA to bind to AGO2 protein. (e) Barplot showing the ability of FAM46C mRNA to bind to AGO2 protein via miRNA. (f) Barplot demonstrating the ability of FAM46C siRNA to inhibit FAM46C mRNA expression in PC10 cells (left). Barplot demonstrating the ability of FAM46C siRNA to enhance MYC mRNA expression in PC10 cells (right). Error bars show standard deviations. (g) Western blot showing the decrease in FAM46C protein (f: full-length protein, s: short-length protein, ref.) expression and the increase in MYC protein expression upon transfection with FAM46C siRNA. (h) Polygonal line demonstrating the ability of the combination of the three miRNA inhibitors to suppress the proliferation of PC10 cells. Western blot showing the decrease in MYC protein expression upon transfection of the combination of the three miRNA inhibitors miRNAs in PC10 cells (see also Supplementary Fig. S8).