Primary liver cancers consist of hepatocellular carcinomas (HCCs) and intrahepatic cholangiocarcinomas (ICCs), which differ substantially in their histology, metastatic potential and prognosis. Studies in mice have indicated that ICC and HCC can share the same cell of origin, the hepatocyte. Yet, the mechanisms regulating the diversity in the commitment of transformed hepatocytes remain largely unknown. As primary liver cancers invariably develop in chronically damaged livers, this suggests that tissue context can determine the tumour subtype identity. Now, Seehawer, Heinzmann et al. have shown that hepatocytes harbouring the same oncogenic drivers can give rise to either HCC or ICC dependent on the type of cell death occurring nearby in the liver microenvironment.

Generating in vivo models of liver cancer, the authors used two different methods to deliver transposon vectors co-expressing oncogenic mouse Myc and NrasG12V (pCaMIN) or mouse Myc and human AKT1 (pCaMIA) into the livers of p19Arf–/– mice. When the genes were delivered by hydrodynamic tail vein injection (HDTV), the mice developed multifocal HCC with cancer cells strongly expressing the hepatocyte-specific transcription factor hepatocyte nuclear factor 4α (HNF4α). However, injection of the same vectors under the liver capsule followed by in vivo electroporation (Epo) resulted in the appearance of unilocular ICC or mixed HCC–ICC with cancer cells staining positive for biliary type keratin 19 (K19). Moreover, in vivo lineage tracing demonstrated that both tumour subtypes originated from differentiated hepatocytes.

Credit: Lara Crow/Springer Nature Limited

After eliminating the possibility that spontaneously acquired genetic mutations drive the differential lineage commitment to ICC or HCC, the authors hypothesized that variation in the liver microenvironment generated by the delivery method could be important. Following either HDTV or Epo treatment, damaged liver tissue with a concomitant inflammatory reaction could be observed, with similar numbers of hepatic stellate cells and infiltrating inflammatory and immune cells. Interestingly, while the number of dying hepatocytes was also equal between the two treatments, the type of cell death induced was different. Expression of the apoptosis marker cleaved caspase 3 was higher after HDTV whereas Epo-treated livers exhibited increased levels of phosphorylated MLKL and RIPK3, both markers of necroptotic cell death.

Necroptosis is typically associated with release of inflammatory cytokines from immune cells. Concordant with this, a change in the cytokine microenvironment was observed when mouse livers were subjected to Epo-treatment versus HDTV. To establish whether the necroptotic cell death occurring in Epo-treated livers was responsible for the increased induction of cytokines, such as CCL4 and CXCL13, mice were pre-treated with necrostatin 1, an inhibitor of necroptosis before Epo. This resulted in a decrease in the production of most Epo-related cytokines as well as a switch towards apoptotic cell death and the outgrowth of HNF4α+ HCCs.

Reasoning that the fate of transformed hepatocytes in liver tumorigenesis might be epigenetically controlled, the authors identified two transcription factors, TBX3 and PRDM5, as potential lineage commitment factors with reciprocal patterns of expression in HCC and ICC. Remarkably, HDTV delivery of pCaMIN vectors co-expressing full-length Prdm5 together with a Tbx3 short hairpin RNA switched the outgrowth of HCC towards the development of HNF4α–;K19+ ICCs.

This mechanism of lineage determination seems to be conserved in humans as analysis of 199 tumour samples revealed a necroptosis gene signature associated with patients with ICC and an apoptosis gene signature enriched in patients with HCC. Likewise, the gene expression patterns of TBX3 and PRDM3 in human HCC and ICC recapitulated the mouse models.

hepatocytes harbouring the same oncogenic drivers can give rise to either HCC or ICC dependent on the type of cell death occurring nearby in the liver microenvironment

Not only does this study provide a molecular basis for why common liver-damaging risk factors predispose to either HCC or ICC, but it also establishes valuable mouse models to investigate these early events in tumorigenesis.