Mutational landscape of a chemically-induced mouse model of liver cancer

Abstract

Background & aims: Carcinogen-induced mouse models of liver cancer are used extensively to study the pathogenesis of the disease and are critical for validating candidate therapeutics. These models can recapitulate molecular and histological features of human disease. However, it is not known if the genomic alterations driving these mouse tumour genomes are comparable to those found in human tumours. Herein, we provide a detailed genomic characterisation of tumours from a commonly used mouse model of hepatocellular carcinoma (HCC).

Methods: We analysed whole exome sequences of liver tumours arising in mice exposed to diethylnitrosamine (DEN). Mutational signatures were compared between liver tumours from DEN-treated and untreated mice, and human HCCs.

Results: DEN-initiated tumours had a high, uniform number of somatic single nucleotide variants (SNVs), with few insertions, deletions or copy number alterations, consistent with the known genotoxic action of DEN. Exposure of hepatocytes to DEN left a reproducible mutational imprint in resulting tumour exomes which we could computationally reconstruct using six known COSMIC mutational signatures. The tumours carried a high diversity of low-incidence, non-synonymous point mutations in many oncogenes and tumour suppressors, reflecting the stochastic introduction of SNVs into the hepatocyte genome by the carcinogen. We identified four recurrently mutated genes that were putative oncogenic drivers of HCC in this model. Every neoplasm carried activating hotspot mutations either in codon 61 of Hras, in codon 584 of Braf or in codon 254 of Egfr. Truncating mutations of Apc occurred in 21% of neoplasms, which were exclusively carcinomas supporting a role for deregulation of Wnt/β-catenin signalling in cancer progression.

Conclusions: Our study provides detailed insight into the mutational landscape of tumours arising in a commonly used carcinogen model of HCC, facilitating the future use of this model to better understand the human disease.

Lay summary: Mouse models are widely used to study the biology of cancer and to test potential therapies. Herein, we have described the mutational landscape of tumours arising in a carcinogen-induced mouse model of liver cancer. Since cancer is a disease caused by genomic alterations, information about the patterns and types of mutations in the tumours in this mouse model should facilitate its use to study human liver cancer.

Keywords: Cancer genomics; Carcinogen mouse model; Hepatocellular carcinoma; Hras; Mutational signatures.

Graphical abstract

Fig. 1

DEN-initiated carcinogenesis in mouse hepatocytes. (A) Overview of experimental design. Cohorts of C3H male mice aged 14–16 days were either administered DEN or left untreated. Liver samples were collected during the first 24 h after DEN exposure for histopathological analysis, or mice were aged to develop tumours. Dysplastic nodules and HCCs were collected from cohorts of mice 24–26 weeks and 26–40 weeks, respectively, after administration of DEN; spontaneous tumours were collected from mice aged 37–76 weeks. (B) In situ detection and quantification of DEN-induced DNA damage. Representative photomicrographs of immunohistochemistry for O⁶-ethyl-2-deoxyguanosine and gamma-H2AX in untreated mice and 8 h post-DEN injection. All scale bars = 200 μm. Original magnification ×200. Automated image quantification was used to evaluate the dynamics of DNA damage for O⁶-ethyl-2-deoxyguanosine and gamma-H2AX over the course of 24 h (n = 5 samples per timepoint; bar indicates mean; Welch two-sample t test: *p <0.05, **p <0.01, ***p <0.001, ****p <0.0001). (C) Histology of murine hepatocellular neoplasms. Representative photomicrographs of serial sections of normal liver tissue and liver tumours arising in DEN-treated and untreated mice. H&E staining demonstrates tissue morphology; reticulin staining is used to assess architecture (normal: staining around each cord of hepatocytes; DN: loss of regular architecture; HCC: thickened trabeculae and corresponding reduction in staining); and Ki67 identifies mitotic cells (normal adult liver is mostly quiescent; DN and HCC show increasing numbers of dividing cells). All scale bars = 200 μm. Original magnification ×200. CV, central vein; DEN, diethylnitrosamine; DN, dysplastic nodule; H&E, haematoxylin and eosin; HCC, hepatocellular carcinoma; PV, portal vein.

Fig. 2

Independent evolution of DEN-initiated liver tumours is revealed by their unique SNV profiles. (A) Experimental design. Liver neoplasms were generated by intraperitoneal injection of DEN into 14–16-day old mice, which were then aged for 24–26 weeks. We performed whole exome sequencing of nine separate nodules isolated from a single mouse liver and of single nodules from livers of seven other mice. To evaluate the noise associated with library preparation, triplicate sequencing libraries were prepared in three separate batches for a single nodule. (B) Phylogenetic analysis of DEN-initiated tumours. A phylogenetic tree was constructed using the ape package in R, where branch lengths correspond to the number of unshared SNVs. Long branch lengths indicate no relatedness among the nodules within a single mouse, whereas three replicate libraries from the single tumour had short branches, indicating few SNV differences. Branches are labelled using mouse and tumour identification codes. DEN, diethylnitrosamine; SNV, single nucleotide variant.

Fig. 3

DEN-initiated neoplasms have a high SNV burden and few indels or copy number variations. (A) Estimated SNV mutation rates per megabase (Mb) in mouse and human liver tumour cohorts. The point mutation frequencies are shown for DEN-induced DN (n = 16) and HCC (n = 18) and for spontaneous tumours (n = 25) arising in untreated mice. Previously reported human HCC (LICA, n = 224) and hepatic adenoma (LIAD, n = 38) samples are shown for comparison. Each point represents a single sample. Bars indicate the median number of SNVs per Mb. (B) Comparison of frequencies of insertions and deletions (indels, 1–50 base pairs) in each cohort of mouse and human liver tumours. Each cohort had few indels, regardless of tumour histology or aetiology. Bars indicate median number of indels. (C) The fraction of the genome altered by cancer-associated CNVs in each cohort of mouse and human liver tumour samples. Mouse tumours had a lower fraction of their genomes present in CNVs larger than 10 Mb compared with human HCCs (LICA). Bars indicate median genomic fraction with CNVs. (D) Distribution of VAFs in mouse liver tumours. Plots show the density of VAFs in DEN-induced DNs (n = 16) and HCC (n = 18) and in spontaneous tumours arising in untreated mice (n = 25). Each line represents the distribution of VAFs from a single tumour. Separate plots are shown for SNVs classified as either protein-altering or as other (intergenic, intronic, or protein-coding synonymous). DEN-initiated tumours typically have higher VAFs (mean 0.32) than spontaneous neoplasms (mean 0.14). CNV, copy number variant; DEN, diethylnitrosamine; DN, dysplastic nodule; HCC, hepatocellular carcinoma; SNV, single nucleotide variant; VAF, variant allele frequency.

Fig. 4

The exomes of DEN-initiated tumours have distinct and reproducible mutational profiles. (A) Frequencies of substitution mutations in mouse and human liver tumour cohorts. Mutational profiles are shown for DEN-induced mouse tumours (combined DN and HCC samples, n = 34); for spontaneous tumours (n = 25) arising in untreated mice; and for human liver tumours (LICA, n = 50). The profiles are displayed using the 96-substitution classification, which is defined by reporting the specific base substitution combined with the immediate neighbouring 5′ and 3′ nucleobases. The arrow indicates one example of a trinucleotide context mutational bias observed in DEN-initiated tumours. (B) Heat map of the occurrence of mutational profiles of individual mouse tumours samples (rows) classified by substitution and trinucleotide context (columns). The right panel shows a phylogenetic tree quantifying the clustering observed from the individual mouse sample mutational profiles. A circle indicates an HCC sample; no circle indicates a DN sample. Neoplasms clustered by mutational profile, revealing a clear grouping into DEN-induced vs. spontaneous tumours. (C) Mutational portraits of individual mouse and human liver tumours reconstructed using COSMIC mutational signatures. Each column shows the composition of signatures in an individual sample. DEN-induced mouse neoplasms showed reproducible portraits largely composed of six component COSMIC signatures (shown in the left panel). In contrast, mouse spontaneous and human tumour portraits were more heterogeneous. DEN, diethylnitrosamine; DN, dysplastic nodule; HCC, hepatocellular carcinoma.

Fig. 5

DEN-initiated and spontaneous mouse liver tumours carry recurrent activating mutations in Hras, but only carcinogen-induced tumours acquire a diversity of consequential SNVs in many cancer genes. (A) Proportions of predicted protein-coding and non-coding variants in DEN-induced and spontaneous liver tumours. Pie charts show the proportions of each variant type in DEN-induced DN (n = 16) and HCC (n = 18) and in spontaneous liver tumours (n = 25). The observed (OBS) distribution of each substitution type was as expected (EXP), regardless of tumour histology or aetiology. The total area of the pie charts reflects the median SNV load within each sample set; spontaneous tumours had extremely low mutational loads. (B) Predicted consequential mutations in oncogenes and tumour suppressor genes for individual tumours. Each column in the table is a mouse tumour sample and each row is a cancer gene showing the occurrence of non-synonymous substitutions found in individual samples. Only genes mutated in at least two samples are shown (see Table S1 for the complete list in each sample of somatic non-synonymous mutations in cancer genes). DEN, diethylnitrosamine; DN, dysplastic nodule; HCC, hepatocellular carcinoma; SNV, single nucleotide variant.

Fig. 6

DEN-initiated and spontaneous mouse liver tumours acquire different recurrent mutations in putative driver genes. (A) Prevalence and location of somatic mutations in Hras, Braf and Egfr in DEN-initiated and spontaneous mouse liver tumours. Activating mutations in Hras were found at different hotspots in DEN-induced tumours (codon 61) compared with spontaneous tumours arising in untreated C3H mice (codon 117). Hotspot mutations in Braf (codon 584) and Egfr (codon 254) were recurrent in DEN-induced tumours, in contrast to spontaneous tumours which rarely carried non-synonymous SNVs in Braf or Egfr. (B) Prevalence and location of somatic mutations in Apc in DEN-induced HCCs compared with DEN-induced DNs. Truncating mutations in Apc were common in DEN-initiated tumours, exclusively in carcinoma samples. The spontaneous samples did not carry any non-synonymous mutations in Apc (not shown, see Table S1). (C) Aberrantly elevated nuclear β-catenin protein expression in tumours with a nonsense mutation in Apc. Representative photomicrographs of serial tissue sections of DEN-induced HCCs. H&E staining demonstrates similar tumour morphology in tumours with wild-type Apc (upper panels) and in tumours with a nonsense Apc mutation (lower panels). Immunohistochemistry for β-catenin protein demonstrates aberrant strongly positive nuclear staining in Apc mutant HCC (lower panels). Codon lengths of genes are shown at the right of each gene schematic. All scale bars = 100 μm. Original magnification ×100. DEN, diethylnitrosamine; DN, dysplastic nodule; HCC, hepatocellular carcinoma.