Notch signaling is activated in human hepatocellular carcinoma and induces tumor formation in mice

Abstract

Background & aims: The Notch signaling pathway is activated in leukemia and solid tumors (such as lung cancer), but little is known about its role in liver cancer.

Methods: The intracellular domain of Notch was conditionally expressed in hepatoblasts and their progeny (hepatocytes and cholangiocytes) in mice. This was achieved through Cre expression under the control of an albumin and α-fetoprotein (AFP) enhancer and promoter (AFP-Notch intracellular domain [NICD]). We used comparative functional genomics to integrate transcriptome data from AFP-NICD mice and human hepatocellular carcinoma (HCC) samples (n = 683). A Notch gene signature was generated using the nearest template prediction method.

Results: AFP-NICD mice developed HCC with 100% penetrance when they were 12 months old. Activation of Notch signaling correlated with activation of 3 promoters of insulin-like growth factor 2; these processes appeared to contribute to hepatocarcinogenesis. Comparative functional genomic analysis identified a signature of Notch activation in 30% of HCC samples from patients. These samples had altered expression in Notch pathway genes and activation of insulin-like growth factor signaling, despite a low frequency of mutations in regions of NOTCH1 associated with cancer. Blocking Notch signaling in liver cancer cells with the Notch activation signature using γ-secretase inhibitors or by expressing a dominant negative form of mastermind-like 1 reduced their proliferation in vitro.

Conclusions: Notch signaling is activated in human HCC samples and promotes formation of liver tumors in mice. The Notch signature is a biomarker of response to Notch inhibition in vitro.

Figure 1. Activated Notch induces liv er oncogenesis in vivo

(a) Schematic representation of the generation of bigenic mice over-expressing a constitutively active form of Notch specifically in the liver (AFP-NICD). (b) Tumor incidence in AFP-NICD bigenic and control NICD monogenic mice. (c) Kaplan-Meier curves of mice survival. (d) Representative H&E images of control liver, a dysplastic nodule, HCC with different degrees of differentiation, HCC with atypical pattern (i.e., infiltration of duct-like cells with marked architectural distortion), and intense ductular proliferation (as observed in CK19 staining image).

Figure 2. Notch activation and de-regulation in human HCC

Prediction of the Notch signature (red in the first row) in HCCs of the HCV-related dataset (n=91). Each square represents data of each individual sample. One third of the samples (31.8%) had activation of Notch based on a significant prediction for the presence of the signature (FDR<0.05, Nearest Template Prediction method). Tumors with the Notch signature were significantly enriched in the Proliferation class of our molecular classification of HCC (Proliferation=green). In addition, HCCs with activated Notch were significantly enriched in different markers of IGF pathway activation including phophorylated(p)-Akt, p-IGF1R, p-RPS6 (immunohistochemistry), as well as high expression levels of IGF2. The bottom panel shows Notch Pathway Gene List (see methods for details) genes found significantly deregulated between Notch-activated HCC and those HCC without the signature as well as with normal liver (FDR<0.05, red and blue bar in the first row, respectively). Deregulation magnitude is graded based on a red (over-expression) and green gradient (down-regulation) normalized to the median expression value in normal liver.

Figure 3. De-regulation of the Notch target gene SOX9 in HCC

(a) Representative images of nuclear SOX9 staining. Top panels show SOX9 staining in mice: bile duct staining in livers from control monogenic mice (upper left), hepatocytes and cholangiocytes in murine tumors (upper right). Bottom panels show human HCC samples with negative (bottom left) and positive (bottom right) nuclear staining. (b) Boxplots of SOX9 mRNA levels from HCC samples (n=51) with and without the Notch signature (left) or SOX9 nuclear staining (right).

Figure 4. Signature-based activation of Notch across HCC datasets

(a) Performance of the Notch signature across HCC datasets. Top rows in each dataset represent Notch predictions, bottom rows show our molecular classification of HCC (Proliferation=green) for each sample. In all except one dataset (HCC-Izuka), tumors with Notch activation are enriched in patients of the Proliferation class. (b) Heatmap of Cramer’s V coefficient showing correlation between the Notch signature and other previously reported signatures that predict clinical aggressive behavior^,. Briefly, Cramer’s V statistic values range from 0 to 1, being 0.36–0.49 substantially correlated, >0.5 strongly correlated and 1 identical.

Figure 5. Notch signature predicts response to Notch inhibition in vitro

(a) Prediction of the Notch signature in 318 cancer cell lines from the Cancer Biomedical Informatics Grid, caBIG® (dark green: solid tumors’ cell lines, light green: hematological malignancies’ cell lines). (b) Notch signature prediction in 21 liver cancer cell lines (Broad-Novartis Cancer Cell Line Encyclopedia). Cell lines used in further analysis are highlighted in bold. (c) Cell viability upon 5-day incubation with a γ-secretase inhibitor. Decrease in viability was restricted to cells harboring the Notch signature; standard deviation is shown. (d) Representative images of pHH3 staining in DN-MAML-GFP and GFP-only (control) transfected cells, white arrowheads highlight transfected proliferating cells (GFP+ and pHH3+). Quantification of cell proliferation was evaluated by enumerating pHH3+ GFP+ cells as a percentage of total GFP+ cells in liver cancer cell lines 48h post-infection; standard deviation is shown.