The production of mouse models to study particular human cancers has become something of a cottage industry. The difficulty in exploiting them arises from the underlying genetic heterogeneity of most cancers, which means that the molecular pathways that are disrupted in a particular mouse might be irrelevant to the human disease, despite an overall phenotypic similarity. A study in Nature Genetics addresses this issue by outlining a new method to identify the best-fit mouse model for hepatocellular carcinoma (HCC).

HCC is the fifth most common cancer in the world, causing at least 500,000 deaths annually. Although only three agents — hepatitis B virus, hepatitis C virus and aflatoxin B1 — are responsible for approximately 80% of all HCCs in humans, the genetic and epigenetic changes that follow exposure are varied and incompletely understood. Snorri Thorgeirsson and colleagues describe a comparative functional-genomic approach that compares the similarities in gene expression between 68 HCCs from 7 mouse models and 91 human HCCs from pre-defined sub-classes. The mouse models include two chemically induced, four transgenic and one knockout line.

The authors first applied hierarchical clustering analysis to previously generated microarray data from the mouse HCCs and identified three distinct groups. Using a subset of the genes that they had previously shown to be associated with survival in human HCC, they compared their expression between the three subgroups of mouse HCC and two subclasses of human HCC. In this clustering analysis, three of the transgenic mouse lines with targeted overexpression in the liver (Myc, E2f1 and Myc/E2f1) had the highest relative similarity to those in the better survival group of human HCC, whereas the Myc/Tgfa transgenic and one of the chemically induced models were most similar to the poor survival group of human HCC. Thorgeirsson and co-workers then found that the individual genes that differentiated the two human HCC subclasses showed a highly similar differential expression in the mouse models that compare best with them.

Do the similarities at the transcriptional level reflect phenotypic similarities between the mouse and human HCCs? By several criteria, the answer is yes. For example, the Myc/Tgfa line has a poor prognosis phenotype, with a higher rate of HCC development and higher mortality. Ultimately, this cross-species approach that is based on overall patterns of gene expression might help to integrate mouse models of cancer with mainstream cancer research more rapidly, and should enable the identification of new mechanisms and treatments.