Overexpression of TAZ alone rarely induces the development of intrahepatic cholangiocarcinomas in mice
To determine whether TAZ deregulation is oncogenic in the liver, we overexpressed an activated form of human TAZ (TAZ S89A, which avoids the phosphorylation by LATS proteins inducing the proteasomal degradation of TAZ) in the mouse liver via hydrodynamic gene delivery (Fig. 1A). Mice injected with TAZ S89A (which will be referred to as TAZ mice) were harvested 10 and 40 weeks post-injection. Macroscopically and histologically, livers harvested 10 weeks post-injection (n= 15) were completely normal (Fig. 1C), indistinguishable from livers injected with the empty vector (Fig. 1B). Similarly, most TAZ mice sacrificed 40 weeks post-injection exhibited a normal, unaltered liver at the macroscopical and microscopical levels. However, detailed microscopical analysis of these mice revealed the presence of a small hepatic lesion in 1 of 15 (6.7%) and large, multiple lesions occupying most of the liver surface in 2 of 25 (8%) TAZ-injected mice harvested 10 and 40 weeks post-injection, respectively (Fig. 1 D, E). These lesions exhibited the histological features of intrahepatic cholangiocarcinoma (iCCA), consisting of glandular and tubular structures. The cholangiocellular origin of the lesions was further confirmed by the positive immunoreactivity for the biliary markers cytokeratin (CK) 7 and 19. The lesions expressed the TAZ protein homogeneously in the nuclei of iCCA cells. Surprisingly, we did not find any immunoreactivity for TAZ in the surrounding livers of those mice other than in biliary structures (not shown), thus indicating the disappearance of the transfected cells in these mice. Similarly, no TAZ-positive cells were detected in the remaining mice. Therefore, it is likely that TAZ-injected cells are eliminated early after hydrodynamic injection, resulting in a very low rate of iCCA development in TAZ-injected mice. In light of these findings, we hypothesized that other molecular alterations occurring in the three TAZ mice showing iCCA lesions might be responsible for the survival of TAZ-injected cells. We focused specifically on AKT, a central player in cell survival and a protooncogene often activated in mouse and human iCCA [37]. The lesions displayed strong immunoreactivity for activated/phosphorylated AKT (p-AKT), whereas the surrounding parenchyma exhibited faint or absent immunolabeling for the same protein (Fig. 1 D, E). Based on these results, we hypothesized that activation of survival pathways such as AKT might be necessary for TAZ-transfected cells to develop iCCA.
Cooperation of TAZ with AKT induces rapid iCCA development in mice
Based on the data obtained in TAZ mice, we determined whether the co-expression of TAZ with the AKT protooncogene accelerates tumorigenesis and/or increases the incidence of iCCA in mice. Previous data from our laboratory indicate that YAP, the TAZ paralog, cannot induce cholangiocarcinogenesis alone but cooperates with AKT to drive iCCA development in mice [37]. Thus, we injected the TAZ S89A plasmid together with a plasmid containing an activated/myristoylated form of AKT (HA-tagged; the combination will be referred to as TAZ/AKT) into mice (Fig. 2A). Mice were harvested at 2 (n=5), 6 (n=5), and 10-11 (n=15) weeks post-injection. At the latest time point, mice displayed significant abdomen enlargement and required euthanasia. Mice injected only with AKT (n=8) were instead sacrificed between 29- and 33-weeks post-injection (Fig. 2B). Macroscopically, the livers of these mice appeared normal at 2 weeks after injection (Fig. 2C). However, small, white, cyst-like lesions were present on the liver surface after 6 weeks. These lesions rapidly expanded and, by 10-11 weeks, had replaced most of the normal liver tissue, and mice rapidly deteriorated and either succumbed or needed to be euthanized. No extrahepatic metastases developed in these mice. At the microscopical level, transfected cells were already visible and formed clusters 2 weeks after hydrodynamic injection (Fig. 2C). Of note, two distinct cell types resulted from the simultaneous transfection of AKT and TAZS89 constructs. Specifically, one cluster consisted of enlarged, clear cells due to the intracellular accumulation of lipids and glycogen, often displacing the nucleus to the cell's periphery. These cells were indistinguishable from those generated by overexpressing AKT alone in the mouse liver [38]. The second cluster was characterized by small cells with scarce basophilic cytoplasm and a high nucleus/cytoplasm ratio. As expected, both cell types displayed robust immunoreactivity for HA-tagged AKT and TAZ antibodies, implying their origin from the injected constructs (Fig. 3). Importantly, when diluting the anti-TAZ primary antibody, it was evident that the levels of TAZ immunolabeling were higher in the cholangiocyte-like small cells than in the enlarged, clear-cell hepatocytes (Supplementary Figure 1). Small cells also displayed the highest levels of SOX9, a Notch effector and a critical player in hepatocyte reprogramming and biliary commitment [39] (Supplementary Figure 1). Consequently, only the small cell type exhibited strong positivity for the biliary markers CK19 (Fig. 3) and CK7 (not shown). In addition, the proliferation marker Ki67 was significantly higher in the small cells (Fig. 3), suggesting their higher growth capacity than the enlarged, lipid-rich cells. Significantly, several small cells displaying elevated TAZ and CK19 levels were surrounded by an inflammatory infiltrate in AKT/TAZ livers two weeks after hydrodynamic injection (Fig. 4). These inflammatory cells consisted of lymphocytes (characterized by immunoreactivity for CD4 and CD45 markers) and macrophages (immunoreactivity for the F4/80 marker). Some small cells encircled by the inflammatory reaction showed signs of apoptosis (as assessed morphologically and by positive immunoreactivity for cleaved Caspase 3 and cleaved PARP), implying their destruction by the inflammatory response. Importantly, this inflammatory event was not detectable at later time points (not shown), indicating exhaustion/loss of the inflammatory response during AKT/TAZ-driven cholangiocarcinogenesis. Because such an inflammatory response does not occur in livers from mice injected only with AKT (unpublished observation), transfection of the TAZ protooncogene likely is responsible for it, for unknown reasons, leading to the elimination of many transduced cells in AKT/TAZ mice. However, despite the described inflammatory reaction, the fast-proliferating small cells rapidly replaced the lipid-rich large cells in the hepatic parenchyma, and tumors already at 6 weeks post-injection were composed only of CK19/CK7-positive cells. However, tiny clusters of large cells could still be appreciated in the mouse liver. The same pattern was observed at 10-11 weeks post-injection, with pure iCCA CK19-positive occupying most of the liver surface (Fig. 3).
AKT/TAZ lesions resemble the histomolecular features of human intrahepatic cholangiocarcinoma
Next, we determined the molecular features of AKT/TAZ tumors by immunohistochemistry (Fig. 5A). As expected, AKT/TAZ lesions but not the surrounding non-tumorous livers displayed activation of the Notch pathway (as evaluated by NOTCH1, NOTCH2, and JAGGED1 immunoreactivity), a canonical downstream effector of the Hippo cascade responsible for iCCA development, and the AKT/mTOR signaling (phosphorylated/activated AKT or p-AKT, phosphorylated/activated ribosomal protein S6 or p-RPS6, and phosphorylated/inactivated glycogen synthase kinase 3 beta or p-GSK-3β). Increased mRNA levels of TAZ targets, including Ccn1, Ccn2, and Notch2, were also detected in AKT/TAZ tumors compared with livers injected with empty vector by quantitative real-time RT-PCR (Fig. 5B). Consistent with their cholangiocellular features, all tumor cells were positive for the bile duct markers CK19 and CK7 and the cholangiocyte-specific transcription factor HNF1B (Fig. 6). When examining hepatocellular characteristics, immunoreactivity for hepatocyte nuclear factor 4 alpha (HNF4α), CCAAT enhancer-binding protein alpha (CEBPA), carbamoyl-phosphate synthase 1 (CPS1 or Hep Par-19), cytochrome P450 2E1 (CYP2E1), cytochrome P450 3A4 (CYP3A4), glutamine synthetase (GLUL), and liver arginase (ARG1) were uniformly downregulated in the lesions when compared with non-neoplastic liver counterparts. In contrast, levels of forkhead box A1 and A2 (FOXA1 and FOXA2) transcription factors were equivalent in tumorous and non-tumorous livers (Fig. 6). Furthermore, AKT/TAZ tumors displayed positive staining for markers of tissue desmoplasia and collagen deposition/synthesis, prominent features of human iCCA, such as S100 calcium-binding protein A4 (S100A4), vimentin, platelet-derived growth factor receptor beta (PDGFRβ), periostin, osteopontin, alpha-smooth muscle actin (α-SMA), and hydroxyproline, in the stromal cells (Supplementary Figure 2). Also, the progenitor/stemness markers SOX9, CD44v6, CD133, and EPCAM were highest in the tumor lesions. In addition, activation of the TAZ paralog YAP (as assessed by its nuclear accumulation) was observed in AKT/TAZ tumors (Supplementary Figure 2). Furthermore, pronounced immunoreactivity for the angiogenic marker CD34 and the lymphatic vessel marker podoplanin and low levels of apoptosis (as determined by cleaved caspase 3 immunolabeling) was readily detected in AKT/TAZ cholangiocellular lesions (Supplementary Figure 3).
Taken together, the present findings indicate that activated TAZ and AKT cooperation induces cholangiocellular tumors recapitulating various histopathologic and molecular features of human iCCA.
The interaction of TAZ with TEAD transcription factors is necessary for AKT/TAZS89A-driven cholangiocarcinogenesis
TAZ is a transcriptional coactivator that interacts with TEAD DNA-binding proteins to drive downstream gene expression. Nevertheless, TAZ also possesses functions that do not depend on its interaction with TEAD factors [10,11,12]. To determine whether TAZ can trigger cholangiocarcinogenesis independent of TEADs, we co-expressed TAZS89AS51A and activated/myristoylated AKT plasmids into the mouse liver by hydrodynamic injection (n=6) (Fig. 7A). The S51A mutation prevents the binding of TAZ with TEAD proteins [33]. All AKT/TAZS89AS51A mice appeared healthy and were harvested 18 weeks after injection. Grossly, no tumor nodules could be observed in the mouse livers. Microscopically, AKT/TAZS89AS51A livers were full of large clear-cell hepatocytes due to high lipid content, thus recapitulating the phenotype of AKT-overexpressing mice [38] (Figure 7B). We further confirmed the inability of the TAZS89AS51A plasmid to induce TEAD-mediated downstream effectors in vitro. Indeed, forced overexpression of either TAZS89A or TAZS89AS51A triggered similar overexpression of TAZ in the human Hucct1 iCCA cell line (Supplementary Figure 4B). This cell line was selected for overexpression experiments because it displays low basal levels of TAZ (Supplementary Figure 4A). However, transient transfection of TAZS89A but not TAZS89AS51A could trigger the mRNA up-regulation in the same cells of CCN1 and CCN2 specific targets of TAZ (Supplementary Figure 4 D, E), whose induction is mediated by TEAD factors [10,11,12]. Furthermore, only forced overexpression of TAZS89A significantly increased the proliferation of HuCCT1 cells in culture (Supplementary Figure 4 F).
Overall, the present data demonstrate that the interaction between TAZ and TEAD transcription factors is required for TAZ-induced iCCA development in mice and TAZ-dependent proliferation in vitro.
The canonical Notch cascade is indispensable for cholangiocellular commitment in AKT/TAZ mice
Previous reports indicate that YAP, the TAZ paralog, induces hepatocyte-cholangiocyte transdifferentiation via the Notch pathway [22, 37]. To determine whether the same applies to TAZ, we co-injected myr-AKT1 and TAZS89A plasmids with a dominant-negative form of the Notch transcriptional activator RBP-J (dnRBP-J; these mice will be referred to as AKT/TAZ/dnRBP-J) (Fig. 8A). Previously, we have shown that dnRBP-J effectively suppresses the canonical Notch pathway in vitro and in vivo [33, 34]. Strikingly, the simultaneous injection of myr-AKT1, TAZS89A, and dnRBP-J did not hamper tumor development in mice, which required to be sacrificed by 13 weeks post-injection due to high tumor burden (Fig. 8B). Macroscopically, numerous nodules could be appreciated on the liver surface of AKT/TAZ/dnRBP-J mice. However, when analyzing the lesions at the histopathological level, AKT/TAZ/dnRBP-J livers were characterized solely by the presence of lipid-rich preneoplastic lesions, identical to those detected in mice injected exclusively with AKT [38], and pure HCC. The hepatocellular nature of the neoplastic lesions was further confirmed by their widespread immunoreactivity for HNF4α and CPS1 hepatocyte markers and negative immunolabeling for CK19 (Fig. 8C).
Therefore, the canonical Notch pathway influences tumor cell differentiation but does not inhibit AKT/TAZ–induced carcinogenesis.
Knockdown of YAP delays AKT/TAZ-driven cholangiocarcinogenesis
Because we detected pronounced nuclear localization of the TAZ paralog YAP in AKT/TAZ lesions (Supplementary Figure 2), we tested the importance of YAP in AKT/TAZ dependent on iCCA development. For this purpose, TAZ was co-injected into the mouse liver with a plasmid containing myristoylated AKT and the short hairpin against YAP (pT3-EF1a-AKT-shYap-TAZ mice). The AKT-shLuc-TAZ combination was used as a scrambled control (Supplementary Figure 5A). Simultaneous injection of myr-AKT1, TAZS89A, and shLuc resulted in a high tumor burden, and AKT-shLuc-TAZ mice required sacrifice between 12 and 14 weeks post-injection (Supplementary Figure 5B). On the other hand, suppression of YAP by shYAP significantly delayed tumorigenesis, and AKT-shYAP-TAZ mice were euthanized significantly later, by 30 weeks post-injection (Supplementary Figure 5B). Both AKT-shLuc-TAZ and AKT-shYAP-TAZ mice developed pure iCCA, histomorphologically indistinguishable (Supplementary Figure 5C). No extrahepatic metastases developed in the two models. As assessed by immunohistochemistry, AKT-shLuc-TAZ livers exhibited nuclear YAP accumulation in the neoplastic lesions, and faint/absent cytoplasmic immunoreactivity for YAP in the non-tumorous counterpart. AKT-shYAP-TAZ tumors were completely YAP negative, and low/absent cytoplasmic YAP staining was detected in AKT-shYAP-TAZ non-tumorous livers (Supplementary Figure 5C). In addition, low levels of YAP1 mRNA characterized AKT-shYAP-TAZ tumors compared to AKT-shLuc-TAZ corresponding lesions (Supplementary Figure 5D). Similar liver weight/body weight ratios between AKT-shLuc-TAZ and AKT-shYAP-TAZ were found at sacrifice (Supplementary Figure 5E).
Overall, the data indicate that YAP contributes to AKT/TAZ tumor progression without affecting the cholangiocellular phenotype of the lesions.
Mature hepatocytes are the source of cholangiocarcinoma cells in AKT/TAZ mice
Mounting evidence indicates that iCCA lesions can originate in mice from mature hepatocytes via a transdifferentiation process [35, 40, 41]. In agreement with these findings, a recent investigation analyzed the gene expression programs in AKT/NICD-induced hepatocyte-derived CCA at the single-cell level using scRNASeq technology [42]. Intriguingly, the authors identified an Epcam+Alb+Krt19- epithelial cluster consisting of cells with both hepatocyte and cholangiocyte features. These cells gradually evolved from normal hepatocytes to iCCA, further confirming the origin of these tumors from mature hepatocytes. In light of these findings, we determined whether the same applies to AKT/TAZ mice. By electron microscopy, we found that enlarged and irregular nuclei and loss of cytoplasmic glycogen characterized the hydrodynamically transfected cells. Nonetheless, the transfected cells were connected to the adjacent hepatocytes through cell junctions, implying their hepatocellular origin (Fig. 9). The origin of iCCA cells from mature hepatocytes in AKT/TAZ mice was further confirmed by lineage tracing experiments. Specifically, AAV8-Tbg-Cre was injected into R26R-EYFP mice, which would initiate hepatocyte-specific expression of EYFP in one week. Subsequently, the HA-tagged AKT and TAZ plasmids were delivered into the liver by hydrodynamic tail vein injection to trigger tumor development (Supplementary Figure 6). Interestingly, as assessed by co-immunofluorescence (IF) staining of the tumor sections, all AKT/TAZ tumor cells expressed CK19, a biliary cell marker, and HA-tag. In addition, the cells were positive for EYFP, indicating the hepatocyte origin of these iCCA cells (Supplementary Figure 7). Moreover, immunohistochemistry conducted on mouse livers two weeks after AKT/TAZ hydrodynamic injection revealed the presence of "hybrid" cells displaying concomitantly hepatocellular and cholangiocellular features, indicating various phases of the transdifferentiation process. For instance, some hydrodynamically transfected mature hepatocytes expressed the cholangiocellular markers CK19 and CD133, while the small cells (malignant cholangiocytes) uniformly expressed the hepatocellular marker HNF4A (Supplementary Figure 8).
Therefore, AKT/Taz-induced cholangiocellular tumors originate from mature hepatocytes.
Ubiquitous activation of TAZ in human intrahepatic cholangiocarcinoma
Next, we assessed TAZ levels in human iCCA. We observed a significant upregulation of the WWTR1 gene (encoding TAZ) levels in the tumor lesions vs. non-tumorous surrounding livers in an iCCA cohort for which clinicopathological and survival data were available (n=50; Supplementary Figure 9A). In these specimens, WWTR1 expression did not correlate with the survival length of the patients, although a trend of lower survival in tumors with elevated WWTR1 could be observed. Noticeably, a significant correlation between TAZ levels and the presence of lymph node metastases (p = 0.014) was detected. A similar correlation occurred between TAZ expression and lung metastases; however, it did not reach significance (p = 0.076), presumably due to the low number of cases with lung metastases. No significant correlations between WWTR1 levels and clinicopathological data such as age, gender, liver cirrhosis, tumor grade, tumor stage, etc., were detected (Supplementary Material). Next, we analyzed the data on human cholangiocarcinoma specimens from The Cancer Genome Atlas (TCGA; http://ualcan.path.uab.edu/cgi-bin/TCGAExResultNew2.pl?genenam=WWTR1&ctype=CHOL). Similar to our collection's findings, the TCGA data indicate the lack of a correlation between WWTR1 mRNA levels and patients' survival length. Notably, the TCGA data analysis revealed a significant correlation between WWTR1 gene levels and lymph node infiltration by the tumor, in accordance with our data (Supplementary Figure 10). These concordant findings might suggest a role of WWTR1/TAZ in human iCCA dissemination and metastasis. No other clinicopathological data showed a significant correlation with WWTR1 expression. Subsequently, we evaluated the TAZ protein levels in a vast collection of human iCCA specimens (n=182) by immunohistochemistry. Consistent with previous data [31, 32], robust nuclear immunoreactivity for TAZ was observed in almost all iCCA lesions (179/182, 98.3%) (Fig. 10). In contrast, TAZ cytoplasmic immunoreactivity characterized hepatocytes from normal livers and surrounding non-neoplastic livers, whereas nuclear TAZ accumulation occurred in cholangiocytes (Supplementary Figure 11). Notably, phosphorylated/inactivated TAZ (p-TAZ) levels were highest in the non-tumorous surrounding livers and normal cholangiocytes, indicating that TAZ is efficiently inactivated in the extra-tumoral areas (Supplementary Figure 11). Consistently, the immunoreactivity for activated/phosphorylated LATS1 and LATS2 proteins, which trigger the inactivation of TAZ, was more pronounced in the non-neoplastic surrounding counterparts, indicating that the control of TAZ degradation by LATS1/2 proteins is hampered along with iCCA development (Fig. 10). In addition, strong TAZ nuclear immunoreactivity was detected in the totality of preinvasive lesions (n=15, consisting of 9 intra-ductal papillary biliary neoplasms or IPBN and 6 biliary epithelial neoplasias or BilIN), implying that TAZ activation occurs before iCCA progression (Supplementary Figure 12). Furthermore, we revealed positive nuclear staining for TAZ in 9 of 9 (100%) mixed HCC/iCCA tumors (Supplementary Figure 13). In these tumors, TAZ nuclear immunoreactivity was present in the cholangiocellular and hepatocellular compartments, but it was always more pronounced in the cholangiocellular component. Finally, we evaluated the status of TAZ and interacting proteins in benign (non-neoplastic) conditions, namely ductular reactions (DR; Supplementary Figure 14). These are reactive processes that arise, in disease and injury, at the interface of the portal (or septal) and parenchymal compartments, in human livers [43, 44]. Significantly, DR showed a robust induction of TAZ, p-LATS1/2, and p-TAZ, implying that TAZ is activated in these lesions and counteracted/balanced by an efficient control system. Instead, the regulatory mechanisms are lost in the neighboring tumor lesions (Supplementary Figure 14).
Overall, the present data indicate that TAZ activation is a predominant event in human iCCA.
Role of TAZ in iCCA in vitro
Next, we further assessed the effects of TAZ on the proliferation and survival of iCCA cell lines. For this purpose, TAZ was silenced either alone or in combination with its paralog YAP using specific small interfering RNAs in KKUM-213 and RBE cell lines (Supplementary Figure 15). In both cell lines, silencing of YAP triggered a compensatory upregulation of TAZ, whereas TAZ knockdown did not affect YAP expression. This finding suggests a compensatory mechanism inducing upregulation of TAZ when YAP is suppressed in iCCA cells. Similar data were previously obtained in hepatoblastoma cell lines [33]. Consistent with the latter hypothesis, simultaneous silencing of TAZ and YAP resulted in a significantly more pronounced reduction of cell proliferation than the silencing of either TAZ or YAP alone (Supplementary Figure 15). Furthermore, the combined knockdown of TAZ and YAP reduced the mRNA levels of Hippo targets (CCN1 and CCN2) more profoundly than the single silencing of either gene (Supplementary Figure 16).
Thus, the current data indicate that TAZ promotes iCCA growth in vitro and suggest that TAZ and YAP might have overlapping and distinct functions in this process.