Transgenic strategies to generate heterogeneous hepatic cancer models in zebrafish

Intratumor heterogeneity describes phenomena that cancer cells from a single focal population can be highly diverse in morphology, gene expression, metabolism, motility, proliferation, metastatic potential, and drug sensitivity (McGranahan and Swanton, 2017). The heterogeneity is the consequence of cancer evolution and the major cause of therapeutic failure due to drug resistance (McGranahan and Swanton, 2017). Theoretically, multi-target combination therapy may eliminate heterogeneous cancer cells but may also induce severe side effects and promote the genesis of novel malignancy. Therefore, further investigations on the pattern and mechanism of cancer heterogeneity are required, with the hope to better refine the treatment strategies. However, there is lack of appropriate in vivo cancer models for simulating multidimensional heterogeneity. Patient/cell-derived tumor xenografts (P/CDX) and spontaneous models induced by toxic substances are naturally heterogeneous in cellular composition during the time but are also highly personalized or unrepeatable. By contrast, transgenic cancer models provide definite genetic backgrounds via direct manipulation of oncogenic pathways; however, the tissue-specific modifications are usually on the whole organ level and poorly mimic the focal lesion(s) in the real cancer patients.

Zebrafish is a highly visualized vertebrate model animal, and liver is a conserved organ between zebrafish and human in morphology and anatomy (Supplementary Figure S1A) (Yao et al., 2017). The hepatocyte-specific promoter fabp10a has been extensively used to drive gene expression in zebrafish liver (Supplementary Figure S1B) (Yao et al., 2018). We and other groups have employed zebrafish to generate several transgenic models of hepatic carcinoma as preclinical models, among which, a diversity of oncogenes, including kras (Yao et al., 2018), swordtail fish egfr (XMRK) (Meierjohann et al., 2006), mouse Myc (Yao et al., 2018), human CTNNB1 (Yao et al., 2018), and HBX (Shieh et al., 2010), were expressed in zebrafish hepatocytes and induced hepatic hyperplasia and eventually carcinoma. Here we propose three different strategies to model heterogeneous hepatic tumorigenesis in zebrafish to better mimic the focal lesion(s) in human cancer patients.

Firstly, we created heterogeneity by performing vector injection at single/double-cell stages. Zygote microinjection has been previously employed to generate chimera zebrafish for modeling tumor heterogeneity or for manipulating oncogenic driver (Ung et al., 2015; Wang et al., 2018). Here we injected a Tol2-flanked conditional expression vector hsp70l:ABC-2A-tcf7l2 into Tg(fabp10a:tetOn; tre:eGFP-kras^v12;tcfsiam:mCherry) zebrafish at single-cell stage. fabp10a:tetOn expresses the transcriptional factor TetOn, which induces the universal expression of eGFP-fused Kras in all hepatocytes. tcfsiam:mCherry is a classic reporter of the canonical Wnt signaling pathway (Wang et al., 2012). hsp70l:ABC-2A-tcf7l2 expressed the active human β-catenin (ABC) and zebrafish Tcf7l2 and activated the Wnt reporter in cells carrying vectors (Figure 1A). After doxycycline (DOX) administration and heat shock, the cancerous liver in the injected larvae displayed a heterogeneous pattern that the Wnt signaling pathway was only activated in a proportion of the hepatocytes, which were labeled in red fluorescence (Figure 1B).

Figure 1.

Schematic diagrams and images of zebrafish heterogeneous hepatic cancer models. (A) Schematic diagram of the double-transgenic zebrafish Tg(fabp10a:tetOn; tre:eGFP-kras^v12; tcfsiam:mCherry) injected by hsp70l:ABC-P2A-tcf7l2. (B) Representative images of the whole mount view (left), left hepatic lobe (middle), and orthogonal views (optical sections, right) of a dissected liver of a 5 days post-fertilization (dpf) Tg(fabp10a:tetOn; tre:eGFP-kras^v12; tcfsiam:mCherry) injected by hsp70l:ABC-P2A-tcf7l2 (heat shock and DOX added at 3 dpf) (scale bar, 25 μm). (C) Schematic diagram of the triple-transgenic zebrafish Tg(fabp10a:tetOn; tre:Gal4; UAS:ABC-P2A-tcf7l2; tcfsiam:eGFP). (D) Representative images of the lateral view optical sections/pseudo-color of a 5 dpf Tg(fabp10a:tetOn; tre:Gal4; UAS:ABC-P2A-tcf7l2; tcfsiam:eGFP) larva (DOX added at 3 dpf, left) and high-magnification images of cryo-sections in fluorescence and HE staining of a 10 dpf transgenic liver (DOX added at 4 dpf, right) (scale bar, 60 μm). (E) Schematic diagram of the double-transgenic zebrafish Tg(fabp10a:tetOn; tre:Gal4; UAS:Oncobow). (F) Lateral views of a Tg(fabp10a:tetOn; tre:Gal4; UAS:Oncobow) zebrafish at 3 and 6 dpf in bright field and the liver under three different fluorescent channels (DOX added at 2 dpf) (scale bar, 150 μm). (G) Statistics of fluorescent populations in the liver of Tg(fabp10a:tetOn; tre:Gal4; UAS:Oncobow) zebrafish at 3 and 6 dpf (n = 5 per groups). (H) Representative images of a 10 dpf Tg(fabp10a:tetOn; tre:Gal4; UAS:Oncobow) in bright field (left above, DOX added at 4 dpf), the representative flow cytometry of the digested liver cells (left below), and high-magnification images of cryo-sections in fluorescence, HE staining, and SA-β-Gal staining (right) (scale bar, 150 μm). (I) mRNA expression of senescence-relevant genes in different hepatocyte populations of the 10 dpf Tg(fabp10a:tetOn; tre:Gal4; UAS:Oncobow) livers (n = 3 per assay). Data are represented as mean ± SD. Statistical differences were determined by one-way analysis of variance and Student’s t-test. ^⋆⋆P < 0.01; ^⋆⋆⋆P < 0.001.

Besides, the heterogeneity can also be implemented in the stabilized transgenic lines. For example, a mifepristone-inducible Cre recombination system was previously employed to generate transgenic zebrafish models of hepatic carcinoma with mosaic patterns (Nguyen et al., 2016). Here we provided an alternative strategy that the random silencing of the UAS promoter in the stabilized transgenic zebrafish can also produce a mosaic expression pattern (Yao et al., 2018). Based on this phenomenon, we have generated a triple transgenic line that displayed chimeric expression of human ABC and zebrafish tcf7l2 in the liver and carried a green fluorescent Wnt reporter tcfsiam:eGFP, Tg(fabp10:tetOn; tre:Gal4-VP16; UAS:ABC-2A-tcf7l2; tcfsiam:eGFP) (Figure 1C). The Wnt reporter tcfsiam:eGFP is activated by both the endogenous and ectopic/exogenous β-catenin and Tcf7l2. Compared to the normal zebrafish liver that barely has any Wnt activity (Yao et al., 2018), the triple transgenic zebrafish liver displayed significant mosaic green fluorescence/Wnt activity in hepatocytes after DOX administration, indicating the establishment of genetic heterogeneity in the partially cancerous liver (Figure 1D). Besides, the hematoxylin-eosin (HE) staining on the adjacent cryo-sections revealed that both the cell density and nucleo-cytoplasmic ratio were increased in the Wnt active regions.

In addition, other conditional transgenic designs can be modified to display heterogeneity in stabilized zebrafish lines. ‘Brainbow’ tool box provided high-level heterogeneity in fluorescence for cell lineages tracing at complex in vivo environments (Cai et al., 2013). Here we fused oncogenes kras^v12 and murine Myc before two fluorescent proteins of the brainbow cascades and generated a novel transgenic cascade termed ‘Oncobow’. Oncobow was then packed into an Tol2-flanked expression vector driven by a UAS promoter and integrated into Tg(fabp10:tetOn; tre:Gal4-VP16) background. The hepatocytes of the double transgenic larvae Tg(fabp10a:tetOn; tre:Gal4; UAS:Oncobow) expressed different oncogenes in the hepatocyte populations upon DOX administration, and the oncogenic heterogeneity was visualized by the colorful fluorescence, which may allow us to monitor the dynamics of tumorigenesis at single-cell resolution (Figure 1E). To investigate the behaviors of the heterogeneous populations, we took images on the same ‘Oncobow’ zebrafish at two different time points (3 and 6 dpf (days post-fertilization)) and measured the sizes of the different fluorescent populations (Figure 1F). The statistics indicated that the cell populations with different genetic backgrounds displayed a diversity of pattern changes, the phiYFP-positive/kras^v12-expressing population expanded faster than the mKate2-positive/Myc-expressing population and the cyan fluorescent protein (CFP)-positive population (Figure 1G). Previously, we reported that Ras oncogenic insults induced the accumulation of β-galactosidase in zebrafish hepatocytes (Yao et al., 2018); here, we also observed the increased SA-β-Gal staining in the liver regions with phiYFP fluorescence but not the regions with mKate2 and CFP fluorescence (Figure 1H). To further investigate the differences in senescence of distinct populations at transcriptional level, we performed cell sorting and harvested the hepatocytes carrying three different fluorescence for analyzing mRNA expression (Figure 1H). The results showed that the senescence-relevant genes glb1, cdkn1a, and mcm2 were significantly deregulated in the kras^v12-expressing hepatocytes as expected in senescence. However, the expression of cdkn2a/b (homologs of the human CDKN2A) was not detected, and lmnb1 was increased in the kras^v12-expressing hepatocytes, which is opposite to its downregulation in cells experiencing classical senescence. Therefore, the accumulation of β-galactosidase may be insufficient to define the status of the kras^v12-expressing hepatocytes as senescence.

Altogether, the three transgenic strategies in combination of the transparency and small size in zebrafish will allow us to better mimic and understand cancer heterogeneity, serving as novel tools for exploring the development of and therapeutic options for solid tumor focal environment in vivo.[Supplementary material is available at Journal of Molecular Cell Biology online. This work was supported by the National Natural Science Foundation of China (81402582) and the Natural Science Foundation of Shanghai (14YF1400600 and 18ZR1404500).]