The Hippo effector TAZ cooperates with oncogenic β-catenin in experimental and human hepatoblastoma development

Backgrounds: Hepatoblastoma (HB) is the most common pediatric liver tumor. Though Wnt/β-catenin and Hippo cascades are implicated in HB development, there is no study on the crosstalk of β-catenin and Hippo downstream effector TAZ in HB. Methods: The expression of TAZ and of β-catenin in human HB specimens was assessed by immunohistochemistry (IHC). The functional interplay between TAZ and β-catenin was tested through delivering either an activated form of TAZ (TAZS89A) alone or co-delivering TAZS89A and an activated form of β-catenin (∆N90-β-catenin) to mouse liver using sleeping beauty transposase via hydrodynamic tail vein injection (SBT-HTVI). In addition, the role of transcriptional enhanced associate domain (TEAD) factors, canonical Notch cascade, Yap, and the tumor modifier heat shock transcription factor 1 (HSF1) along TAZ/β-catenin-driven HB development was studied in vivo and vitro. Results: Activation of TAZ often co-occurred with that of β-catenin in clinical specimens. While overexpression of TAZS89A alone was unable to promote liver tumorigenesis, the concomitant overexpression of TAZ and ∆N90-β-catenin induced the development of HB lesions exhibiting both epithelial and mesenchymal features. Mechanistically, HB development driven by TAZ/β-catenin required TAZ interaction with TEAD factors. Furthermore, TAZ/β-catenin overexpression induced HB development in conditional Yes-associated protein knockout (Yap KO) mice, indicating that Yap activation is dispensable in this model. Activation of the Notch signaling was observed in TAZ/β-catenin mouse lesions, consistent with that reported in human HBs. Blocking of the canonical Notch cascade using the dominant negative form of RBP-J (dnRBP-J) did not inhibit TAZ/β-catenin dependent HB formation in mice, although suppressed the mesenchymal as a critical oncogene in HB and are dispensable for TAZ/β-catenin induced HB development in YAP overlap preliminary results our that of and in the mouse tumor development The latter observation suggests that YAP and TAZ are not able to trigger liver tumor initiation in combination, at by the hydrodynamic gene delivery method. However, these findings do not exclude a collaborative (and non-redundant) role of the two protooncogenes in driving tumor progression. Further studies using TAZ and YAP conditional knockout mice should be conducted to address this important issue.

provided by Dr. Eric Olson from University of Texas Southwestern Medical Center, Dallas, TX [15]. The hydrodynamic injection procedure has been described in detail in our previous studies [16][17][18][19] .
Briefly, to determine the oncogenic potential of TAZS89A, 20μg of pT3-EF1α-TAZS89A, either alone or in combination with 20μg pT3-EF1α-ΔN90-β-catenin, were mixed together with pCMV/sleeping beauty transposase at a ratio of 25:1 and injected into 6-to 8-week-old FVB/N mice via the lateral tail vein.
To investigate if the oncogenic potential of TAZ was based on TAZ transcriptional function, 20μg of pT3-EF1α-TAZS89AS51A were introduced into adult FVB/N mice along with 20μg pT3-EF1α-ΔN90-βcatenin and 1.6μg pCMV/sleeping beauty transposase. To block the Notch cascade, high doses of dnRBP-J (60 μg) together with pT3-EF1α-TAZS89A (20 μg) and pT3-EF1α-ΔN90-β-catenin (20 μg) plasmids were injected. In addition, to assess the importance of HSF1 in TAZ/β-catenin dependent HB formation, we injected 20μg of pT3-EF1α-TAZS89A and 20μg of pT3-EF1α-ΔN90-β-catenin together with 60 μg of either pT3-EF1α-dnHSF1 or pCMV empty vector. Finally, to study if YAP is required for TAZ-dependent HB development, we injected 20μg of pT3-EF1α-TAZS89A and 20μg of pT3-EF1α-ΔN90-β-catenin, with 60 μg of either pCMV-Cre or pCMV empty vector, as well as pCMV/sleeping beauty transposase into adult Yap flox/flox mice. All mice used in the experiments were monitored continually and euthanized at specific time points, as indicated in the main text, or when they became moribund. Mice were maintained and monitored in accordance with protocols approved by the Committee for Animal Research at the University of California, San Francisco (San Francisco, CA).

Histology, Immunohistochemistry, Assessment of Proliferation, Apoptosis, and Microvessel
Density Liver specimens were harvested and fixed in 10% formalin overnight at 4°C and embedded in paraffin. Hematoxylin and eosin (ThermoFisher Scientific, Waltham, MA) staining was conducted using a standard protocol on human and mouse liver sections and slides analyzed by two expert liver pathologists (KE and ME) in accordance with the criteria described before [9,20]. Immunohistochemistry (IHC) was performed according to our previous studies [9,13]. Proliferation index was determined in mouse HB lesions by counting Ki-67 positive cells on at least 3000 tumor cells per mouse sample. Apoptosis index was determined in mouse HCC lesions by counting TUNEL positive cells on at least 3000 tumor cells per mouse using the TumorTACS TM In Situ Apoptosis Detection Kit (Trevigen, Gaithersburg, MD), following the manufacturer's instructions. Assessment of microvessel density (MVD) was conducted as previously reported [21]. Specifically, liver tumor samples from YAPS127A/∆N90-β-catenin and TAZS89A/∆N90-β-catenin mice were subjected to immunostaining with the mouse monoclonal anti-CD34 antibody (Abcam, United Kingdom). The tumors were first screened at low power (× 40) to identify the areas of highest microvessel density (MVD). The four highest MVD areas for each tumor were photographed at high power (× 200) and the size of each area standardized using the ImageJ software. Any brown-stained endothelial cell or endothelial cell cluster was counted as one microvessel, irrespective of the presence of a vessel lumen. MVD was expressed as the percentage (mean±SD) of the total CD34-stained spots per section area (0.94 mm 2 ). The primary antibodies used in the study are described in Table 2.

Western Blot Analysis
Frozen liver tissues and cell pellets were homogenized in Mammalian Protein Extraction Reagent (ThermoFisher Scientific) containing the phosphatase inhibitors. Protein concentration was determined using the Bio-Rad Protein Assay kit (Bio-Rad, Hercules, CA) using bovine serum albumin as a standard. Supernatants were denatured by boiling in Tris-Glycine SDS Sample Buffer (Life Technologies, Carlsbad, CA) for Western blot analysis. Proteins were separated by SDS PAGE and transferred onto nitrocellulose membranes (Life Technologies) by electroblotting. Membranes were blocked in 5% non-fat dry milk for 1 h and then incubated with proper primary antibodies (Table 2).
Subsequently, membranes were incubated with horseradish peroxidase-secondary antibodies (Jackson Immunoresearch Laboratories Inc., West Grove, PA, USA) diluted 1:5000 for 30 min. Proteins bands were revealed with the Super Signal West Femto (Pierce Chemical Co., New York, NY).

RNA extraction and quantitative real-time reverse transcriptase polymerase chain reaction (qRT-PCR)
Total mRNA from liver tissues and cells was extracted by using the Quick RNA Miniprep kit (Zymo Research, Irvine, CA, USA). Next, mRNA expression of the genes of interest was detected by quantitative real-time polymerase chain reaction (qRT-PCR) using validated Gene Expression Assays for human and mouse genes (ThermoFisher Scientific; instructions. All experiments were repeated at least three times in triplicate.

Statistical Analysis
All data are presented as mean ±SD for each group. Statistical differences between two groups were determined using the U-tests embedded in the Prism 6 software version 6.0 (Graph Pad Software Inc., La Jolla, CA). P < 0.05 was considered statistically significant.

Frequent activation of TAZ in human HB samples
First, we assessed the levels of TAZ protein in a collection of human HB specimens (n = 32) by IHC ( Figure 1). In non-tumorous surrounding liver, cytoplasmic and nuclear positive staining for TAZ was limited to biliary cells, whereas hepatocytes showed either absent or faint cytoplasmic immunolabeling for this protein, in accordance with previous data [22]. An equivalent staining pattern was detected for the TAZ paralog, YAP, as well as for NOTCH2, a TAZ target. As expected, β-catenin protein was expressed in the membrane of hepatocytes as well as in the membrane and cytoplasm of biliary cells. Noticeably, strong cytoplasmic and nuclear immunoreactivity for TAZ was detected in most HB specimens (22/32; 68.7%). The frequency was lower than that found for nuclear β-catenin (27/32, 84.4%) and the TAZ paralog YAP (26/32, 87.5%) but higher than nuclear NOTCH2 (19/32; 59.4%) in the same HB collection. Concomitant nuclear immunoreactivity for TAZ and β-catenin was detected in 18 of 22 TAZ-positive tumors (81.8%), while 15 of 22 (71.4%) human HB with TAZ activation displayed positive nuclear immunolabeling for YAP and NOTCH2. No significant association was found between the levels of TAZ, β-catenin, YAP, and NOTCH2 with clinicopathologic features of the patients, including age, gender, histology subtype, recurrence, or lung metastasis (data not shown).
Overall, the present data unravel the frequent activation of TAZ in human HB, which is frequently associated with induction/overexpression of YAP, β-catenin, and NOTCH2.

HB development in mice
To elucidate the functional role of TAZ along HB pathogenesis, we investigated whether activated TAZ can either induce liver carcinogenesis or cooperate with other oncogenes to promote HB formation in mice. For this purpose, we hydrodynamically transfected the constitutively activated form of TAZ (TAZS89A) and the N-terminal deleted activated mutant form of β-catenin (∆N90-β-Catenin), alone or in combination, in the mouse liver ( Figure 2A). TAZS89A is a mutant form allowing TAZ to escape phosphorylation-mediated degradation and, thus, leading to TAZ unrestrained activity [23].
Overexpression of TAZS89A alone did not lead to liver tumor formation in mice. Indeed, 21 weeks post hydrodynamic injection, all mice appeared to be healthy. Accordingly, upon dissection, no tumor nodule was detected. Subsequent histological evaluation revealed that the mouse liver was completely normal, lacking any histological alteration ( Figure 2B Microscopically, small cluster of cells could be visualized as early as ~1.5 weeks post injection in the liver of TAZS89A/∆N90-β-catenin mice ( Figure 3A). Positive immunoreactivity for Myc Tagged-∆N90-βcatenin and TAZ antibodies confirmed that tumor cells originated from the injected plasmids ( Figure   3B,C), while immunolabeling for Ki67 implied active cell proliferation ( Figure 3D). The first tumor nodules were visible in TAZS89A/∆N90-β-catenin mice by 3 weeks post injection ( Figure 3E). Tumors quickly occupied nearly completely the liver parenchyma by 6 weeks post injection ( Figure 3F).
Histologically, most of the lesions consisted of small cells with prominent nuclei, similar to those described in YAPS127A/β-catenin mice [9,12], and resembling human fetal or crowded fetal HB ( Figure 3G,H). Nonetheless, a second, distinct tumor entity was appreciable in the liver parenchyma of TAZS89A/∆N90-β-catenin mice. Indeed, besides the "pure" hepatoblast-like lesions, many tumors displayed a mixed phenotype, consisting of an epithelial (hepatoblast) and a mesenchymal component, with the latter being characterized primarily by cells with spindle morphology ( Figure   3J,K). To the best of our knowledge, this is the first reported mouse model showing features of epithelial and mesenchymal differentiation. Following this interesting observation, we sought to determine the molecular characteristics of the two tumor entities by using IHC (Figure 4). As and TAZ (connective tissue growth factor or CTGF, Cysteine-Rich Heparin-Binding Protein 61 or Cyr61, Jagged 1, Notch1, Notch2, and Hes1) canonical targets was more elevated in TAZS89A/∆N90-βcatenin livers when compared with the wild-type counterpart (Supplementary Figure 4).

TAZS89A/∆N90-β-catenin induced hepatoblastoma requires an intact TAZ interaction with TEAD transcription factors
TAZ is a transcriptional activator and it interacts with TEAD DNA binding proteins to induce downstream gene expression. Nonetheless, TAZ is also known to have functions independent of its binding to TEAD factors [10, 11, 29]. Thus, we sought to investigate whether TAZS89A/∆N90-β-catenin induced HB development requires an intact TAZ/TEAD interaction. To address this issue, we generated the TAZS89AS51A plasmid for hydrodynamic transfection. The S51A mutation prevents TAZ from binding to TEAD proteins [29]. Therefore, we co-expressed TAZS89AS51A and ∆N90-βcatenin plasmids into the mouse liver ( Figure 5A). All TAZS89AS51A/∆N90-β-catenin mice appeared to be healthy and were harvested at 5-8 weeks post injection ( Figure 5B). Grossly, no tumors nodules were detected on the mouse liver. Subsequent histological analysis revealed that the liver tissues of TAZS89AS51A/∆N90-β-catenin mice was completely normal, and the liver weight/body weight ratio of these mice was equivalent of that of wild-type, un-injected liver ( Figure 5C,D). Altogether, the present data demonstrate that TEAD proteins are required for TAZ driven HB development in mice.
However, mesenchymal lesions were absent in YAPS127A/∆N90-β-catenin mice (Supplementary Figure 5A). At the cellular level, the rates of proliferation, apoptosis, and angiogenesis (microvessel density; MVD) were compared between the epithelial lesions from TAZS89A/∆N90-β-catenin and YAPS127A/∆N90-β-catenin mice. No significant differences were observed in the rates of proliferation and MVD between the two mouse models, whereas apoptosis was significantly lower in TAZS89A/ Finally, we determined the levels of two transcription factors, whose role in liver cancer is becoming more and more relevant, namely nuclear factor erythroid 2 like 2 (NFE2L2/NRF2) and heat shock factor transcription factor 1 (HSF1) in TAZS89A/∆N90-β-catenin and YAPS127A/∆N90-β-catenin tumors[30-32]. Levels of NRF2, its regulator Kelch like ECH associated protein 1 (Keap1), and those of its target NAD(P)H quinone dehydrogenase 1 (Nqo1), were upregulated in both mouse models, without statistical differences between the two cohorts, when compared with wild-type mice (Supplementary Figure 8). In striking contrast, a robust overexpression of HSF1 was observed only in TAZS89A/∆N90-β-catenin mouse lesions both via RT-PCR and IHC (Supplementary Figure 8).

HSF1 is dispensable along TAZS89A/∆N90-β-catenin dependent hepatoblastoma development
Following the interesting finding of upregulation of HSF1 exclusively in TAZS89A/∆N90-β-catenin tumors, we assessed the levels of HSF1 in human HB specimens by IHC to determine a possible role of HSF1 in this tumor type. Of note, HSF1 was ubiquitously (32/32, 100%) upregulated and localized in the nucleus of HB cells in the collection examined ( Supplementary Figure 9), whereas low/absent immunoreactivity for HSF1 was detected in corresponding surrounding non-tumorous livers.
Subsequently, the effect of HSF1 inactivation was investigated in the TAZS89A/∆N90-β-catenin mouse model. Specifically, to achieve this goal, we co-injected TAZS89A and ∆N90-β-catenin plasmids with a dominant negative form of HSF1 (dnHSF1) in the mouse liver via hydrodynamic gene delivery (these mice will be referred to as TAZS89A/∆N90-β-catenin/dnHSF1) ( Figure 6A). This approach was found to be highly effective in completely suppressing liver carcinogenesis driven by either AKT or c-Myc protooncogene in mice [33,34]. Unexpectedly, HSF1 inactivation neither affected HB development ( Figure 6B) nor changed the histopathological features of TAZS89A/∆N90-β-catenin mice ( Figure 6C).
Taken together, the present data indicate that TAZS89A/∆N90-β-catenin HB development does not primarily depend on HSF1 in mice.

Loss of endogenous YAP does not affect TAZS89A/∆N90-β-catenin driven hepatoblastoma formation
Next, we asked whether activated YAP and TAZ have redundant and/or distinct roles in HB formation.
Instead, positive YAP staining was readily observed in the surrounding normal liver tissues as well as stromal cells within the HB lesions (Fig. 7D). Loss of YAP expression in TAZS89A/∆N90-β-catenin/Cre tumor tissues was further validated by Western blotting (Fig. 7C).
Overall, the present data indicate that activated TAZ induces HB formation in the absence of the endogenous YAP.

Canonical Notch signaling is not required for TAZS89A/∆N90-β-catenin HB development in vivo
Mounting evidence indicate the Notch pathway as a downstream effector of the Hippo cascade in liver cancer [35][36][37]. In addition, NOTCH2 has been found to be consistently overexpressed in human HB [38]. Furthermore, our present data indicate the upregulation of members of the Notch pathway (Jag1, Notch1, Notch2, and Hes1) in the lesions of TAZS89A/∆N90-β-catenin mice. Thus, we investigated whether the canonical Notch signaling is required for TAZS89A/∆N90-β-catenin driven HB formation in vivo. For this purpose, we hydrodynamically injected TAZS89A/∆N90-β-catenin plasmids together with the dominant negative form of RBP-J (dnRBP-J; these mice will be referred to as TAZS89A/∆N90-β-catenin/dnRBP-J) that has been shown to effectively inhibit the canonical Notch cascade ( Figure 8A) [37, 39]. Additional mice were injected with TAZS89A/∆N90-β-catenin with pT3 empty vector as control (TAZS89A/∆N90-β-catenin/pT3). We found that all mice develop high liver tumor burden and were required to be euthanized around 6 to 9 weeks post injection ( Figure 8B-D).
Histologically, HB lesions could be appreciated in both cohorts of mice. However, overexpression of dnRBP-J in the livers of TAZS89A/∆N90-β-catenin mice led to the disappearance of the mesenchymal compartment of HB lesions ( Figure 8C), indicating that the Notch pathway might contribute to the differentiation toward a mesenchymal phenotype of TAZS89A/∆N90-β-catenin HB lesions.
Overall, our study indicates that Notch signaling is activated in TAZS89A/∆N90-β-catenin tissues and influences the tumor cell differentiation. However, Notch activation is dispensable in TAZS89A/∆N90β-catenin driven HB development.

TAZ cooperates with β-catenin and YAP for the growth of human HB cells in vitro
The findings in human and mouse HB specimens support a major role of TAZ in HB formation. To further investigate the relevance of TAZ activation in this tumor type, we examined whether TAZ expression is required for human HB cell growth in vitro. Thus, we silenced TAZ in two human HB cell lines (HepG2 and Hep293TT cells) using specific siRNA against TAZ (Supplementary Figure 10, 11). Of note, TAZ knockdown resulted in a decrease of cell proliferation and induction of apoptosis in the two HB cell lines. These anti-growth effects were increased when the two cell lines were concomitantly subjected to TAZ and β-catenin silencing. At the molecular level, suppression of TAZ did not affect βcatenin and YAP mRNA and protein levels, whereas β-catenin knockdown resulted in downregulation of TAZ and YAP.
Finally, we assessed whether TAZ and YAP have redundant roles in human HB cells. For this purpose, we silenced TAZ and YAP genes using specific siRNAs, either alone or in combination, in HepG2 and Hep293TT cell lines (Supplementary Figure 12, 13). We found that concomitant knockdown of TAZ and YAP resulted in a significantly more pronounced growth restraint when compared with the silencing of either TAZ or YAP alone. At the molecular level, knockdown of TAZ did not affect the levels of YAP, whereas silencing of YAP triggered upregulation of TAZ. These findings suggest the existence of a compensatory mechanism triggering upregulation of TAZ when YAP is suppressed in HB cells.
Altogether, the present data indicate that TAZ acts in concert with β-catenin and YAP to induce HB growth in vitro. Nonetheless, some important differences occur between TAZS89A/∆N90-β-catenin and YAPS127A/ ∆N90-β-catenin HB lesions. In particular, histopathological analysis revealed that HB lesions from TAZS89A/∆N90-β-catenin display the presence of a mesenchymal component, which is instead absent in corresponding lesions from YAPS127A/∆N90-β-catenin mice. Although the molecular reason(s) for this histopathologic difference between the two HB models is not clear, our data suggests that the Notch pathway plays a major role in this phenomenon. Indeed, suppression of the canonical Notch cascade via the use of a dnRBP-J was accompanied by the disappearance of the mesenchymal phenotype in TAZS89A/∆N90-β-catenin HB lesions. Additional studies are needed to better define the molecular mechanisms and the specific Notch targets involved in this event.

Discussion
Another gene specifically upregulated in TAZS89A/∆N90-β-catenin tumors was the HSF1 transcription factor, whose oncogenic role in liver cancer is well-established [15,32,33]. Unexpectedly, inactivation of HSF1 in TAZS89A/∆N90-β-catenin mice had no effect on HB development. Previously, we showed that blockade of HSF1 activity using the same strategy is able to completely suppress hepatocarcinogenesis driven by either AKT or c-Myc overexpression in mice [15,33]. These contrasting data suggest that the addiction to HSF1 is presumably oncogene-and context dependent However, these findings do not exclude a collaborative (and non-redundant) role of the two protooncogenes in driving tumor progression. Further studies using TAZ and YAP conditional knockout mice should be conducted to address this important issue.
Our present data might also have potentially important therapeutic implications for this aggressive disease in humans. Indeed, the findings from mouse models and cell lines suggest that concomitant targeting of the Wnt/β-catenin and Hippo pathways using specific inhibitors might be highly detrimental for HB growth. Unfortunately, these two signaling cascades are not easily druggable and no effective drugs against them are commercially available. A possible strategy would be the use of