Integrated analysis of gene expression and copy number identified potential cancer driver genes with amplification-dependent overexpression in 1,454 solid tumors

Abstract

Identification of driver genes contributes to the understanding of cancer etiology and is imperative for the development of individualized therapies. Gene amplification is a major event in oncogenesis. Driver genes with tumor-specific amplification-dependent overexpression can be therapeutic targets. In this study, we aimed to identify amplification-dependent driver genes in 1,454 solid tumors, across more than 15 cancer types, by integrative analysis of gene expression and copy number. Amplification-dependent overexpression of 64 known driver oncogenes were found in 587 tumors (40%); genes frequently observed were MYC (25%) and MET (18%) in colorectal cancer; SKP2 (21%) in lung squamous cell carcinoma; HIST1H3B (19%) and MYCN (13%) in liver cancer; KIT (57%) in gastrointestinal stromal tumors; and FOXL2 (12%) in squamous cell carcinoma across tissues. Genomic aberrations in 138 known cancer driver genes and 491 established fusion genes were found in 1,127 tumors (78%). Further analyses of 820 cancer-related genes revealed 16 as potential driver genes, with amplification-dependent overexpression restricted to the remaining 22% of samples (327 tumors) initially undetermined genetic drivers. Among them, AXL, which encodes a receptor tyrosine kinase, was recurrently overexpressed and amplified in sarcomas. Our studies of amplification-dependent overexpression identified potential drug targets in individual tumors.

Figure 1

Elevated expression levels of 64 oncogenes in 1,454 solid tumors across 15 cancer types. Cased with strong (fold change ≥10) and moderate (5≤ fold change <10) overexpression are indicated by dark red and dark blue bars, respectively. The two groups of oncogenes, 10 amplification-based (Amp-based), and 54 mutation-based, oncogenes, are indicated on the left. Since some probes recognize multiple splice variants, the number of microarray probes to identify the 64 oncogenes was 71, and there were 13 probes among these corresponding to 12 genes which showed no elevated expression in any tumors, which are not shown. The numbers on the right indicate the number of samples with strong (fold change ≥10) and moderate (5≤ fold change <10) overexpression, as well as the sum of these samples (fold change ≥5). As indicated at the top, samples are arranged by tumor type, including colorectal, lung, stomach, head and neck (HN), liver, breast, pancreas, sarcoma, uterus, kidney, esophagogastric junction (EGJ), esophagus, gastrointestinal stromal tumor (GIST), ovary, melanoma, and other types of tumors. The number of samples of each type of tumor is indicated in parentheses.

Figure 2

Gene expression and copy number of 64 oncogenes. (A) Gene expression levels of 12 oncogenes, including EGFR variant 1, ERBB2, MDM2, MYC, MYCL variants 1 and 2, MYCN, SKP2, NKX2-1, MET, HIST1H3B, EZH2, and CARD11, were linked to the genomic copy number. Pearson’s correlation coefficient (r with P-value) between copy number and mRNA expression is indicated at the bottom right. Red horizontal bars indicate 5-fold on the log₂ scale. (B) The relationship between gene overexpression and degree of genomic copy number gain. Tumors in which oncogenes were overexpressed by ≥5-fold were divided into three groups: those with high-level genomic amplification (copy number ≥6; dark blue bars), those with moderate genomic amplification (copy number 3–5; orange bars), and those without genomic amplification (copy number 1–2; light green bars). There were no cases with a copy number of zero among the indicated genes. Genes highlighted on the bottom correspond to the 12 oncogenes exhibited in panel A.

Figure 3

Radar charts comparing the 12 core signaling pathways involved in amplified and overexpressed oncogenes in individual cancer types. Each axis of the diagrams represents a percentage of frequencies of pathway activation derived from amplified and overexpressed oncogenes, as listed in Table 2. As listed in Supplementary Table 1, 138 driver genes were assigned to the 12 signaling pathways, including APC, cell cycle/apoptosis (CC/A), chromatin modification (CM), DNA damage control (DDC), Hedgehog (HH), NOTCH, PI3K, RAS, receptor tyrosine kinase (RTK), STAT, TGF-β, and transcriptional regulation (TR).

Figure 4

Genomic aberrations of cancer driver genes in 1,454 solid tumors. (A) Bar graph depicting the number of samples having somatic structural alterations in individual tumor types. Genomic alterations were grouped by cases of overexpression with gene amplification in 64 oncogenes (Exp/Amp), cases of mutations, insertions, and deletions in 138 driver genes, along with the presence of fusion genes (Seq), and cases without any alterations (Unknown). (B) Pie chart depicting the frequency of somatic structural alterations in all tumor samples.

Figure 5

Frequency of overexpression with gene amplification of 820 cancer-related genes in individual tumor types. Genes overexpressed with ≥5-fold and a genomic copy number of ≥3 are ordered by type of tumor, as shaded. In colorectal (CRC), squamous cell carcinoma of the lung (Lg-sq), stomach (St), head and neck (HN), liver (Liv), and breast (Bre) cancer, genes with ≥10% frequency are shown. For the remaining tumor types, including tumors from lung adenocarcinoma (Lg-ad), pancreas (Pan), sarcoma (Sar), uterus (Ute), kidney (Kid), esophagogastric junction (EGJ), esophagus (Eso), gastrointestinal stromal tumor (GIST), ovary (Ov), and melanoma (Mel), genes with frequencies approximately ≥20% are listed. Frequencies observed in all 1,454 tumor samples are indicated as “All”. The frequency is indicated as a heat map (range 0–100%).

Figure 6

Tumor-specific overexpression with gene amplification among 820 cancer-related genes in 1,454 solid tumors. Genes overexpressed ≥5-fold with high (copy number ≥6) and moderate (copy number 3–5) genomic amplification are indicated by dark blue and orange bars, respectively. The tumor specificity of each gene is indicated on the right. As indicated at the top, samples are arranged by tumor type, including tumors from colorectal, lung, stomach, head and neck (HN), liver, breast, pancreas, sarcoma, uterus, kidney, esophagogastric junction (EGJ), esophagus, gastrointestinal stromal tumor (GIST), ovary, melanoma, and other types of tumors. The number of samples of each type of tumor is indicated in parentheses.