In this report, we have shown that it is possible to identify the underlying driver genes of human cancer amplicons by screening appropriately selected pools of cDNAs for their ability to promote tumorigenesis in a mosaic mouse model. Through follow-up analysis of one of these tumor-promoting amplified genes, we established that FGF19 is an oncogene that is co-amplified with CCND1 in human tumors, and demonstrated its inhibition through RNAi or a potentially therapeutic monoclonal antibody can block the clonal growth and tumorigenicity of human HCC cells harboring the FGF19/CCND1 amplicon. Given that there are currently no genetically-targeted therapies for hepatocellular carcinoma, we believe these results represent an important biomedical advance.
Previously one of us (D.F.) showed that mice with FGF19
expressed in the skeletal muscle of transgenic mice eventually developed liver tumors through a poorly understood but presumably paracrine mechanism (Nicholes et al., 2002
), and that an anti-FGF19 monoclonal antibody prevented tumor formation in this model in addition to inhibiting xenograft tumor formation of some human colon cancer cell lines (Desnoyers et al., 2008
). However, these studies didn t establish the basis for how FGF19
was involved in human cancer, which clearly can involve a cell autonomous mechanism, nor did they provide a clear strategy for selecting a likely-to-respond subpopulation of patients for treatment with the monoclonal antibody.
It is not clear if there is a biological explanation for why CCND1
are invariably co-amplified in HCC, or if their co-amplification is a secondary consequence of their close proximity and a result of amplicon formation involving DNA breaks at specific regions (Gibcus et al., 2007
). Nevertheless, our data indicate that the two genes are functionally linked in that cyclin D1 levels in hepatocytes are dependent upon FGF19 signaling. Additionally, while the downstream effector of FGF19 in hepatocytes and HCC cells has been clearly established as FGFR4 (Wu X et al., 2010
; Pai et al., 2008
), which downstream effectors are involved in cyclin D1 in HCC cells is not clear. Cyclin D1 activates CDK4/6 kinase which in turn inactivates RB1 (Sherr, 1996
). Genetic lesions affecting RB1 pathway members - including the tumor suppressor p16/INK4A
-- can be mutually exclusive in certain cancers (Sherr, 1996
). However, in some cancers CCND1
amplification frequently co-occurs with p16/INK4A
loss (Okami et al., 1999
). Protein analysis of human HCC tumors suggests that this could be true with HCC (Azechi et al., 2001
) which implies that other proteins that cyclin D1 binds to and influences (e.g. MYB, STAT3, PPARγ) (Knudsen, 2006
) are involved in cyclin D1 oncogenic effects in HCC.
It is surprising that amplicons do not always have the same driver genes in different tumor types; there is a fundamental difference between the 11q13.3 amplicon in breast and liver cancers in that FGF19 is clearly overexpressed as result of amplification in liver cancer, but is not so in breast cancer. Thus, driver genes can be tissue-type-dependent making it important to obtain amplification and overexpression data for different tumor types, even in the case of well-validated oncogenes.
We are optimistic that forward-genetic screens can be used generally for genome-wide identification of oncogenic driver genes from DNA amplifications or other activating alterations identified by human cancer genome profiling. Most importantly, by performing follow-up experiments using RNAi in human cancer cell lines or mouse models, it should be possible to identify more oncogene dependencies and therapeutic targets. A key point about amplified driver genes is that they provide an immediate biomarker for identifying the patients that might benefit from treatment.