Molecular characterization of hepatocarcinogenesis using mouse models

Abstract

Hepatocellular carcinoma (HCC) is a deadly disease, often unnoticed until the late stages, when treatment options become limited. Thus, there is a crucial need to identify biomarkers for early detection of developing HCC, as well as molecular pathways that would be amenable to therapeutic intervention. Although analysis of human HCC tissues and serum components may serve these purposes, inability of early detection also precludes possibilities of identification of biomarkers or pathways that are sequentially perturbed at earlier phases of disease progression. We have therefore explored the option of utilizing mouse models to understand in a systematic and longitudinal manner the molecular pathways that are progressively deregulated by various etiological factors in contributing to HCC formation, and we report the initial findings in characterizing their validity. Hepatitis B surface antigen transgenic mice, which had been exposed to aflatoxin B1 at various stages in life, were used as a hepatitis model. Our findings confirm a synergistic effect of both these etiological factors, with a gender bias towards males for HCC predisposition. Time-based aflatoxin B1 treatment also demonstrated the requirement of non-quiescent liver for effective transformation. Tumors from these models with various etiologies resemble human HCCs histologically and at the molecular level. Extensive molecular characterization revealed the presence of an 11-gene HCC-expression signature that was able to discern transformed human hepatocytes from primary cells, regardless of etiology, and from other cancer types. Moreover, distinct molecular pathways appear to be deregulated by various etiological agents en route to formation of HCCs, in which common pathways converge, highlighting the existence of etiology-specific as well as common HCC-specific molecular perturbations. This study therefore highlights the utility of these mouse models, which provide a rich resource for the longitudinal analysis of molecular changes and biomarkers associated with HCC that could be exploited further for therapeutic targeting.

Keywords: Aflatoxin B1; Hepatitis B surface antigen; Hepatocellular carcinoma; Mouse models.

Fig. 1.

Experimental set-up and characterization of liver nodules. (A) Schematic showing the time points of mice collection as well as injections of either corn oil (as control) or AFB₁, into the various groups of mice. In the first group, 7-day-old (D7) mice were injected and the cohorts were collected at 3-month intervals over 15 months. In the adult group, 6- or 12-month-old (6 M or 12 M) mice were injected and collected at 3-month intervals until 15 months of age. (B) Work flow upon sacrifice of mice at collection. All organs were harvested and fixed for histological analysis by H&E staining. The liver was separated into nine parts, in triplicates, for DNA, RNA and protein isolation, in addition to the histological analysis. Liver nodules were separated from normal livers and collected similarly. Serum was obtained via blood collected through heart puncture upon sacrifice of the animals. (C,D) Liver samples were photographed prior to dissection at each stage of collection [shown as months (M)]. C shows livers from the various mice cohorts that were AFB₁/corn oil-injected at D7. D shows adult mice when injected at 6 M (left) or 12 M (right). Representative pictures are shown.

Fig. 2.

Characterization of susceptibility based on age and gender. (A,B) Liver samples were scored as described in the text, and the data are represented as number of liver nodules (≥0.5 cm) per mouse, over the time course of collection in the various groups of mice (top panel). The bottom panel shows the size of these liver nodules in the various categories of mice. Nodules <0.5 cm were scored as 0.2 cm for ease of enumeration, and were counted in this case. Data for mice injected at D7 (A) and 6 M or 12 M (B) are shown. Each dot represents a single mouse (top) or nodule (bottom). Horizontal bars represent average numbers of nodules (top panel). (C,D) Effect of gender on liver nodule development was determined by scoring the percentage of male or female mice with nodules (0.2 cm and above) at the various time points of collections, after injection at D7 (C) or at 6 M (D). *P≤0.05; ns, not significant.

Fig. 3.

Histological analysis of liver sections. (A-D) Liver samples were H&E-stained and analyzed by photomicrography from the various categories of mice injected at D7, at the different time points of collection. Representative pictures from at least six independent mice for each time point/category are shown. Liver tumors from each case, at 15 M of age, are shown in the last image of B-D (denoted by ‘Tumor - 15’). (E) Comparison of mouse liver nodules at 15 M with human HCC and adenoma.

Fig. 4.

Genomic characteristics of mouse HCC model. (A) Transcriptome profiling of normal livers and liver nodules from the various groups were performed by Affymetrix mouse arrays. PCA, shown in the left panel, revealed distinct clustering of tumor and normal samples, with adjacent normal clustered between tumor and normal samples. Each data point represents one sample and the ellipsoids represent the zone within 2 s.d. of each experimental group. Right panel shows the heat map of unsupervised hierarchical clustering based on probes with s.d.≥1, which reveals distinct clustering of tumor and normal samples. Within the tumor group, differential clustering was also observed based on genotype. Treatment parameters are indicated on the sides. (B) Heat map of 11 genes found to be differentially expressed between HBsAg_Oil_T and HBsAg_Oil_N in the mouse HCC model. Distinct clustering was observed as expected between tumor and normal samples, based on the expression pattern of these genes. Seven genes (underlined) could be mapped to human microarray chips.

Fig. 5.

HCC-specific gene expression signature. (A,B) Differential clustering between tumor and normal samples was observed when segregated by the HCC gene expression signature, similar to the mouse HCC model. (C,D) Heat map of unsupervised clustering based on the gene signature in human breast cancers (C) or colon cancers (D) showed less differential clustering of tumor and normal samples. (E) Summary of SVM model analysis of sensitivity of the gene expression signature to identify the human tumor types.

Fig. 6.

Molecular characterization of liver nodules from mouse models. (A,B) Venn diagram showing the genes (A) or upstream regulators (B) affected in the liver tumors in each of the categories and their overlap. The top five upstream regulators are highlighted on the sides. Numbers in parentheses indicate the genes/regulators identified through the differential gene expression. (C) Top canonical pathways involving the upstream regulators in the tumors from the various categories are listed. (D) Venn diagram showing upstream regulators perturbed in human HCCs and HBsAg_AFB_T versus WT_Oil_N.