Genome-wide mutation profiles of colorectal tumors and associated liver metastases at the exome and transcriptome levels

Abstract

To characterize the mutation profiles of colorectal cancer (CRC) primary tumors (PTs) and liver metastases (CLMs), we performed both whole-exome and RNA sequencing. Ten significantly mutated genes, including BMI1, CARD11, and NRG1, were found in 34 CRCs with CLMs. We defined three mutation classes (Class 1 to 3) based on the absence or presence of mutations during liver metastasis. Most mutations were classified into Class 1 (shared between PTs and CLMs), suggesting the common clonal origin of PTs and CLMs. Class 1 was more strongly associated with the clinical characteristics of advanced cancer and was more frequently superimposed with chromosomal deletions in CLMs than Class 2 (PT-specific). The integration of exome and RNA sequencing revealed that variant-allele frequencies (VAFs) of mutations in the transcriptome tended to have stronger functional implications than those in the exome. For instance, VAFs of the TP53 and APC mutations in the transcriptome significantly correlated with the expression level of their target genes. Additionally, mutations with high functional impact were enriched with high VAFs in the CLM transcriptomes. We identified 11 mutation-associated splicing events in the CRC transcriptomes. Thus, the integration of the exome and the transcriptome may elucidate the underlying molecular events responsible for CLMs.

Keywords: RNA sequencing; colorectal cancer; exome sequencing; liver metastasis; somatic mutation.

Figure 1. Molecular patterns of PTs and CLMs

A. The number of mutations and base substitutions detected in 19 CRCs with CLMs. B. Percentage of base substitutions (left panel) and proportion of transversions (Tv) and transitions (Ts) (right panel) detected during the progression from normal tissues to PTs and ultimately to CLMs. C. Distribution of LOH counts occurring at heterozygote SNPs of PTs and CLMs. D. Distribution of the standard deviations of gene expression levels in normal tissues, PTs, and CLMs.

Figure 2. Significantly mutated genes in 34 Korean CRC patients with CLMs

A. Significantly mutated genes analyzed by MutSig 1.4. B. NRG1, CARD11, and BMI1 mutations found in Korean and TCGA CRC cohorts.

Figure 3. Mutational classes and their clinical associations

A. Fraction of mutational classes in each patient. B. Correlation between clonality changes during metastasis and the proportion of shared mutations (Class 1). C. Correlation between clonality changes during metastasis and the proportion of CLM-specific mutations (Class 3). D. Percentage of Class 1 and 2 mutations detected in patients with or without LVI. Class 1_15 (30) indicates Class 1-H mutations that exhibit ≥ 15 (30)% higher VAFs in CLMs than PTs. E. Percentage of Class 1 and 2 mutations detected in patients in stage T3 or T4. F. Scatter plots for VAFs of CRC driver mutations presented as VAFs in PTs versus VAFs in CLMs.

Figure 4. Co-occurrence of Class 1-H or Class 2 mutations with chromosomal aberrations

A. Two types of molecular processes that potentially generate Class 1-H or Class 2 mutations. B. Frequency of the co-occurrence of chromosomal deletions (left) or amplification (right) with mutations. C. A LOH density plot presented as LOH counts at heterozygote SNPs per distance within ± 1-Mb regions relative to mutations. D. List of Class 1-H mutations superimposed with chromosomal deletions.

Figure 5. Integration of exome-seq with RNA-seq and functional implication of selective expression of mutant alleles

A. Intersection of exome-seq and RNA-seq mutations. B. VAFs of TP53 mutations in the exomes and transcriptomes of PTs and CLMs. C. Correlation between TP53 VAFs and the expression level of p53-target genes (BBC3, BAX, FAS, APAF1, CCNG1, CDKN1A, GADD45A, PTEN, SFN, TSC2, and TP53i3). D. Correlation between APC VAFs and the expression level of WNT-target genes (MYC, CCND1, HNF1A, LEF1, PPARD, JUN, FOSL1, MMP7, AXIN2, NRCAM, TCF4, CLDN1, VEGFA, FGF18, MYCBP, ID2, TERT, LGR5, and FZD7).

Figure 6. Enrichment of mutations exhibiting high functional impact in the transcriptomes of CLMs

A. Distribution of mutational functional impact scores according to increasing VAFs in RNA-seq (upper) and exome-seq (bottom) analysis of CLMs. B. Candidate CLM mutations enriched in the transcriptomes of CLMs.

Figure 7. Mutation-dependent splicing events occurring in CRCs

A. Exon 5 skipping associated with a GPR56 splice site mutation in patient 8804. Curved lines between exons indicate exon-exon junction reads. An arrow denotes the position of a splice-site mutation. B. A predicted GPR56 protein generated by a GPR56 splice-site mutation. C. Exon 4 skipping associated with a MTRF1 splice site mutation in patient 326. D. A predicted MTRF1 protein generated by a MTRF1 splice site mutation.