Pan-cancer noncoding genomic analysis identifies functional CDC20 promoter mutation hotspots

Abstract

Noncoding DNA sequences occupy more than 98% of the human genome; however, few cancer noncoding drivers have been identified compared with cancer coding drivers, probably because cancer noncoding drivers have a distinct mutation pattern due to the distinct function of noncoding DNA. Here we performed pan-cancer whole genome mutation analysis to screen for functional noncoding mutations that influence protein factor binding. Recurrent mutations were identified in the promoter of CDC20 gene. These CDC20 promoter hotspot mutations disrupt the binding of ELK4 transcription repressor, lead to the up-regulation of CDC20 transcription. Physiologically ELK4 binds to the unmutated hotspot sites and is involved in DNA damage-induced CDC20 transcriptional repression. Overall, our study not only identifies a detailed mechanism for CDC20 gene deregulation in human cancers but also finds functional noncoding genetic alterations, with implications for the further development of function-based noncoding driver discovery pipelines.

Keywords: Cancer; Cancer Systems Biology; Genetics; Genomics.

Graphical abstract

Figure 1

Summary of pan-cancer noncoding analysis data and workflow (A) Proportion of tumor samples by disease types. (B) Mutation count distribution of individual samples in 19 cancer types. (C) Workflow of the method to detect recurrently mutated noncoding regions that affect protein factor binding.

Figure 2

Noncoding mutation hotspots analysis (A) Shown is the probability (−log₁₀) of noncoding mutations in the 11-bp window (y axis) plotted against the number of times the noncoding region is found mutated (x axis). Bonferroni-adjusted p values are shown. (B) Typical noncoding mutation hotspots in regulatory regions of CDC20, DPH3/OXNAD1, and RPL18A genes are shown.

Figure 3

mRNA and CNV analysis of CDC20 in various human cancers (A) CDC20 mRNA expression levels were compared in multiple types of human cancers and corresponding normal control tissues based on The Cancer Genome Atlas (TCGA) database. The boxplot is bounded by the first and third quartiles with a horizontal line at the median. (B) CDC20 CNV levels in various cancers are shown based on TCGA datasets. The unit is Gistic2 copy number. (C) The correlation between CDC20 CNV and mRNA in TCGA melanoma samples (n = 367). Pearson correlation P and R values are shown. (D and E) Kaplan-Meier overall survival curves of patients with melanoma are shown. Patients are separated into two groups based on CDC20 mRNA (D) or CNV (E) values. n = 231 for both CDC20 mRNA high and low groups. n = 24 for CDC20 CNV amplified group and n = 333 for CDC20 CNV normal group. Log rank (Mantel-Cox) test p values are shown. BLCA: bladder cancer; CESC: cervical cancer; CHOL: bile duct cancer; ESCA: esophageal cancer; HNSC: head and neck cancer; LUAD: lung adenocarcinoma; LUSC: lung squamous cell carcinoma; PAAD: pancreatic cancer; PRAD: prostate cancer; SARC: sarcoma; SKCM: melanoma; STAD: stomach cancer; UCEC: endometrioid cancer. OV: ovarian cancer; TGCT: testicular germ cell tumors; UCS: uterine carcinosarcoma.

Figure 4

Functional consequence of CDC20 promoter hotspot mutations (A and B) Luciferase reporter assay was performed in 293 (A) and M14 cells (B) with wild-type (WT) or mutant CDC20 promoter driving luciferase vectors. Error bars represent mean ± SD from three experiments. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001. Unpaired Student's t test p values between each mutation and WT control are shown. (C) EMSA assays were performed with wild-type or mutant CDC20 promoter probes. CDC20 promoter mutation strongly abolish protein factor binding to DNA probes. (D) Kaplan-Meier overall survival curves of patients with melanoma with indicated CDC20 promoter mutations or control mutations. n = 25 for patients with clustered CDC20 promoter mutations (including G25A, G28A, G29A, and GG28/29AA), n = 17 for patients with other mutations in the background region. Log rank (Mantel-Cox) test p value is shown.

Figure 5

ELK4 binds to the hotspot mutation targeted sequence and represses CDC20 transcription (A) Screen for ETS proteins that bind the hotspot mutation targeted sequence “GGAAGG” and repress CDC20 transcription. shRNA experiments were performed in 293 cells; expression of each ETS and CDC20 mRNA was quantified by qPCR. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, Student's t test compared to sh-control. The results are an average of three independent experiments. Values are mean ± SD. (B) ENCODE ELK4 ChIP-seq data around the hotspot mutation target sequence “GGAAGG” in 293 and HeLa cells. (C) ChIP was performed with anti-FLAG antibody in M14 cells stably expressing FLAG-ELK4 or FLAG control. The DNA sequence around the hotspot mutation target sequence was quantified with qPCR. ∗∗∗p < 0.001, Student's t test compared with FLAG control. The results are an average of three independent experiments. Values are mean ± SD. (D) Luciferase reporter assay was performed in ELK4 knockdown 293 cells with wild-type or mutant CDC20 promoter driving luciferase vectors. The results are an average of three independent experiments. Values are mean ± SD.

Figure 6

Hotspot mutation targeted sequence mediates DNA damage-induced CDC20 transcriptional repression (A and B) Luciferase reporter assay was performed in 293 (A) or M14 (B) cells with wild-type or mutant CDC20 promoter driving luciferase vectors in the presence or absence of DNA damage drug 5-FU. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, Student's t test compared with wild-type. The results are an average of three independent experiments. Values are mean ± SD. (C) Luciferase reporter assay was performed in ELK4 shRNA knockdown 293 cells with wild-type or mutant CDC20 promoter driving luciferase vectors in the presence or absence of 5-FU. (D) Proposed function for the hotspot mutation targeted sequence in CDC20 transcriptional regulation.