Notch-Driven Cholangiocarcinogenesis Involves the Hippo Pathway Effector TAZ via METTL3-m6A-YTHDF1

Abstract

Background & Aims

Notch and TAZ are implicated in cholangiocarcinogenesis, but whether and how these oncogenic molecules interact remain unknown.

Methods

The development of cholangiocarcinoma (CCA) was induced by hydrodynamic tail vein injection of oncogenes (Notch1 intracellular domain [NICD]/AKT) to the FVB/NJ mice. CCA xenograft was developed by inoculation of human CCA cells into the livers of SCID mice. Tissues and cells were analyzed using quantitative reverse transcription polymerase chain reaction, Western blotting analyses, immunohistochemistry, chromatin immunoprecipitation-quantitative polymerase chain reaction and WST-1 cell proliferation assay.

Results

Our experimental findings show that TAZ is indispensable in NICD-driven cholangiocarcinogenesis. Notch activation induces the expression of methyltransferase like-3 (METTL3), which catalyzes N6-methyladenosine modification of TAZ mRNA and that this mechanism plays a central role in the crosstalk between Notch and TAZ in CCA cells. Mechanistically, Notch regulates the expression of METTL3 through the binding of NICD to its downstream transcription factor CSL in the promoter region of METTL3. METTL3 in turn mediates N6-methyladenosine modification of TAZ mRNA, which is recognized by the m6A reader YTHDF1 to enhance TAZ protein translation. We observed that inhibition of Notch signaling decreased the protein levels of both MELLT3 and TAZ. Depletion of METTL3 by short hairpin RNAs or by the next generation GapmeR antisense oligonucleotides decreased the level of TAZ protein and inhibited the growth of human CCA cells in vitro and in mice.

Conclusion

This study describes a novel Notch-METTL3-TAZ signaling cascade, which is important in CCA development and progression. Our experimental results provide new insight into how the Notch pathway cooperates with TAZ signaling in CCA, and the findings may have important therapeutic implications.

Graphical abstract

Summary.

TAZ is important in Notch1 intracellular domain-driven cholangiocarcinogenesis. Notch1 intracellular domain induces the expression of RNA methyltransferase methyltransferase-like protein 3, which catalyzes N6-methyladenosine modification of TAZ mRNA. This study provides novel insight into how the Notch pathway cooperates with TAZ in cholangiocarcinoma. The Notch-methyltransferase-like protein 3-TAZ signaling cascade is critically implicated in the regulation of cholangiocarcinoma development/progression.

Cholangiocarcinoma (CCA) is a highly malignant human cancer with poor prognosis. The incidence and mortality of CCA are increasing worldwide and currently there is no effective chemoprevention or treatment.1, 2, 3 The prognosis of CCA is poor due to the inability of early diagnosis, the aggressive nature of the tumor growth, and the lack of effective target therapy. Therefore, a better understanding of the molecular mechanisms of cholangiocarcinogenesis is needed to develop more effective target therapy.

The pathogenetic mechanisms for CCA involve various risk factors as well as genetic and epigenetic aberrations.¹ Studies in recent years have shown that activation of Notch signaling and Hippo pathway molecules YAP/TAZ are critical during cholangiocarcinogenesis, sparking interest in these pathways as potential therapeutic targets. Notch signaling pathway is known to be implicated in the development and morphogenesis of the intrahepatic biliary system and in cholangiocarcinogenesis.⁴ During CCA development, Notch signaling promotes the conversion of hepatocytes into biliary lineage cells.^5,6 Regarding Hippo-YAP/TAZ pathway, studies have shown that both YAP and TAZ are implicated in CCA initiation and progression.7, 8, 9 Emerging evidence suggests that the interplay between Notch and Hippo-YAP/TAZ pathways plays an important role in cholangiocarcinogenesis.10, 11, 12 For example, YAP- or TAZ-driven CCA displayed the activation of Notch pathway, as characterized by the increased level of Notch receptors (Notch1/2) and its ligands (Jag1/2).^8,12 Blocking the canonical Notch cascade delayed TAZ/AKT-induced CCA formation and suppressed the cholangiocellular phenotype of YAP/AKT-induced CCA.^8,12 Furthermore, Notch-driven CCA showed YAP activation which mediated Notch-induced cholangiocarcinogenesis.^10,11 However, despite the documented role of YAP in Notch-mediated cholangiocarcinogenesis, it remains unknown whether TAZ, the other Hippo pathway effector, is implicated in Notch-driven CCA.

Although YAP and TAZ are homologous transcriptional cofactors regulated by the Hippo kinase cascade, they have unique structural and physiological characteristics and differ in their regulation and downstream functions.¹³ Accumulating evidence suggests that the structural differences between YAP and TAZ may confer unique functions on each protein in tumorigenesis.^14,15 For example, it has been reported that TAZ can form a homodimer or a heterodimer (with YAP) to initiate transcription, whereas YAP is only able to form YAP-TAZ heterodimers^16,17; these findings suggest that TAZ may mediate functions independent of YAP. Functionally, Wang and colleagues’ study revealed that TAZ, but not YAP, is indispensable for c-Myc-driven hepatocellular carcinoma (HCC) progression.¹⁸ In mouse model of cholangiocarcinogenesis induced by hydrodynamic tail vein (HDTV) injection of oncogenes, tail vein delivery of Notch intracellular domain (NICD) or TAZ in combination with AKT vector is well known to robustly induce CCA development,^8,19 whereas tail vein delivery of YAP in combination with AKT is known to induce mixed hepatocellular and cholangiocellular carcinoma.¹¹ When a single oncogene is delivered via tail vein, HDTV injection of YAP alone fails to induce CCA development,²⁰ whereas HDTV injection of TAZ alone is able to induce small numbers of CCA tumors with characteristic glandular and tubular structures (similar to the small numbers of CCA tumors induced by HDTV injection of NICD alone^8,19).

As the role of TAZ in Notch-regulated cholangiocarcinogenesis has not been defined, we carried out studies to delineate the cross interactions between Notch and TAZ in CCA. Given that TAZ-induced CCA exhibits a longer latency (40 weeks) than NICD-induced CCA (20 weeks) under mouse model of HDTV injection,^8,19 we first explored the possibility whether TAZ might serve as a regulator in Notch-mediated cholangiocarcinogenesis. Our experimental findings reveal that TAZ is indispensable in NICD-driven cholangiocarcinogenesis. Further mechanistic analyses show that NICD induces the expression of methyltransferase like-3 (METTL3), which catalyzes N6-methyladenosine (m6A) modification of TAZ mRNA and that this mechanism plays a crucial role in the crosstalk between Notch and TAZ in CCA cells.

Results

The Association Between Notch Pathway and TAZ in CCA

We first utilized an established mouse model of CCA¹⁹ induced by tail vein delivery of NICD and myr-Akt plasmids in conjunction with the hyperactive sleeping beauty (SB) transposon system to the livers of wild-type mice (FVB/NJ) via HDTV injection (outlined in Figure 1A). Under this system, the CCA lesions occupied almost the entire liver 5 weeks following HDTV injection, and the tumor cells stained positive for the biliary marker SOX9 (Figure 1B). We observed that the CCA tumor tissues harvested from the mice receiving HDTV injection of NICD/AKT showed increased TAZ expression when compared with the bile duct epithelial cells and the hepatocytes in CCA-adjacent liver tissues as determined by immunohistochemical (IHC) staining (Figure 1C). Increased TAZ expression in CCA tissues was further confirmed by Western blotting (Figure 1D). Because TAZ and YAP are homologous transcriptional cofactors regulated by the Hippo kinase cascade, we also examined the expression of YAP by IHC staining in NICD/AKT-induced CCA. Our data showed that the expression of YAP is not significantly different between CCA cells and bile duct epithelial cells (Figure 1C).

Figure 1.

The expression of TAZ is increased in CCA. (A) Outline of study design. FVB/NJ mice were subjected to HDTV injection of NCID, AKT and transposase plasmids, and the animals were sacrificed 5 weeks after tail vein injection. (B) Representative gross photographs of livers from the mice with or without HDTV injection (upper panels). Histological images (H&E stain) from the corresponding livers are shown in the middle panels. The lower panels show the immunohistochemical staining for the biliary epithelial marker SOX9. (C) IHC staining for TAZ (left) and YAP (right) in the liver with NICD/AKT-induced CCA tumor. (D) Western blot analysis for NICD (Myc tag), AKT (HA tag), and TAZ proteins in the liver tissues from mice with or without HDTV injection. (E) Statistical graphs show upregulated expression of TAZ mRNA in human CCA samples from TCGA database and GEO datasets (∗∗P < .01, compared with control samples). (F) Analysis of TAZ protein expression in human CCA tissues from The Human Protein Atlas (HPA). Left: TAZ staining intensity in human cholangiocarcinoma tissues. Right: Representative immunohistochemical staining for TAZ protein in human CCA tissues. (G) Western blotting for TAZ protein in human CCA cell lines (CCLP1, SG231, HuCCT1), BECs, and a nonneoplastic human biliary epithelial cell line (H69).

We then analyzed the expression of TAZ mRNA in human CCA tissues using publicly available databases, including The Cancer Genome Atlas (TCGA) database and 2 Gene Expression Omnibus (GEO) datasets (GSE107943 and GSE76297). Our analyses reveal that TAZ mRNA is upregulated in human CCA from both unpaired CCA/non-CCA samples (TCGA and GSE107943) and paired CCA/non-CCA samples (GSE76297) (Figure 1E). We then examined the expression of TAZ protein in The Human Protein Atlas database and observed moderate or high expression of TAZ protein in 23 of 25 samples of human CCA (92%) (Figure 1F). By Western blotting analysis of TAZ protein in human CCA cells, we found that the level of TAZ protein is higher in CCA cells (CCLP1, SG231, HuCCT1) when compared with primary human biliary epithelial cells (BECs) and a non-neoplastic biliary epithelial cell line (H69) (Figure 1G).

As HDTV injection of NICD/AKT increased TAZ expression in CCA tissues (see Figure 1C), we sought to further delineate the role of NICD vs AKT in the induction of TAZ expression. To this end, we treated NICD/AKT-induced CCA cells with 2 Notch inhibitors (targeting γ-secretase), Avagacestat/PF-03084014, and two AKT inhibitors, Afuresertib/MK-2206. Specifically, CCA cells (MC22891) isolated from mice induced by HDTV injection of NICD/AKT (recently established in our lab²¹) were treated with the Notch inhibitors or AKT inhibitors, in vitro. Similar studies were also performed in cultured human CCA cells (CCLP1). We observed that inhibition of Notch or AKT significantly suppressed the growth of both mouse and human CCA cells (Figure 2A); these findings are consistent with the documented roles of both Notch and AKT signaling pathways in CCA cell growth. However, Western blotting analysis revealed that inhibition of Notch, but not inhibition of AKT, was able to decrease the level of TAZ protein (Figure 2B). The latter observations implicated the role of Notch, but not AKT, in the regulation of TAZ expression in CCA cells. To validate these findings, we designed small interfering RNAs (siRNAs) targeting AKT1, AKT2, and the major transcriptional effector CSL (CBF1/Suppressor of Hairless/LAG-1) of Notch signaling pathway. Our results showed that knockdown of CSL, but not AKT1 and AKT2, was able to reduce the level of TAZ protein in MC22891 cells (Figure 2C). Accordingly, quantitative real-time polymerase chain reaction (qRT-PCR) assay revealed that inhibition of Notch decreased the expression of Hippo-TAZ pathway downstream genes (including CTGF, CYR61, EDN1, and ZEB1) in CCA cells (Figure 2D). By analyzing the reverse phase protein array (RPPA) data from the TCGA CCA database, we observed that the expression of Notch1 protein is positively correlated with the expression of TAZ protein (Figure 2E) in human CCA tissues.

Figure 2.

The association between Notch pathway and TAZ in CCA. (A) The effects of Notch inhibitors (Avagacestat [40 μM], PF-03084014 [40 μM]) or Akt inhibitors (Afuresertib [15 μM], MK-2206 [10 μM]) on the growth of mouse MC22891 cells and human CCA cells (CCLP1). The cells were treated with Notch inhibitors or Akt inhibitors for 48 hours (∗∗P < .01, compared with vehicle DMSO). (B) The effects of Notch inhibitors or Akt inhibitors on the level of TAZ protein in MC22891 and CCLP1 cells. (C) The effects of CSL, Akt1, and Akt2 siRNAs on the level of TAZ protein in MC22891 cells. (D) qRT-PCR assay was performed to examine the effects of Notch inhibitors on Hippo pathway downstream genes in MC22891 and CCLP1 cells (∗∗P < .01, compared with DMSO). (E) Informatics analysis for the correlation of Notch1 protein and TAZ protein in human CCA tissues (TCGA database). (F–G) The effect of TAZ knockdown on Notch pathway activity. MC22891 cells with or without shRNA depletion of TAZ were transfected with CSL luciferase reporter vector for 48 hours (with pRL-TK Renilla luciferase vector as internal control). The cell lysates were then collected to measure the luciferase reporter activity with a luminometer (F). The mRNA levels of Notch downstream genes were analyzed by qRT- PCR in MC22891 cells with or without TAZ depletion (G).

Previous studies suggest that TAZ signaling may be involved in the regulation of Notch pathway in certain cell types.^8,22 In our system, we transfected MC22891 cells with TAZ short hairpin RNA (shRNA) vector or control shRNA vector and the cells were subjected to CSL luciferase reporter activity assay and qRT-PCR analysis to determine Notch pathway activity. Our results showed that depletion of TAZ by shRNA significantly decreased CSL transcription activity and reduced the expression of Notch pathway downstream genes (including CCND1, ERBB2, HES5, HEY1, NRARP, SOX9) (Figure 2F and G). These results suggest that TAZ signaling is implicated in the activation of Notch pathway in CCA cells. The findings suggest the existence of a bidirectional crosstalk between Hippo pathway and Notch signaling in CCA cells.

Taken together, the above findings provide important evidence for the association between Notch pathway and TAZ in CCA.

TAZ is Important in NICD-driven Cholangiocarcinogenesis

Based on the above-described findings, we sought to further assess the role of TAZ in NICD-driven cholangiocarcinogenesis. To this end, we included TAZ shRNAs or control shRNA in the plasmid solution containing NCID, AKT, and transposase vectors for HDTV injection using FVB/NJ mice. Six weeks after HDTV injection, the mice were sacrificed to examine CCA tumor burden (outlined in Figure 3A). Our results showed that depletion of TAZ by shRNA significantly delayed the development of CCA. Upon sacrifice, the livers from TAZ shRNA-injected mice developed much smaller tumor nodules when compared with the control shRNA-injected group (Figure 3B). The TAZ shRNA-injected mice exhibited significantly lower liver/body weight ratio than the control mice (Figure 3C). Western blotting analysis confirmed decreased expression of TAZ protein in CCA tissues harvested from the mice receiving TAZ shRNA injection (Figure 3D). Decreased TAZ protein in CCA tissue from TAZ shRNA-injected mice was further confirmed by immunohistochemical staining (Figure 3E). Immunohistochemical staining for SOX9 showed decreased CCA cell mass in TAZ shRNA-injected mice (Figure 3E). Although depletion of TAZ did not lead to increased apoptosis of CCA cells by terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay (Figure 3F), IHC staining for the cell proliferation marker Ki-67 revealed decreased tumor cell proliferation capability in TAZ-depleted CCA (Figure 3G). We observed that the survival time of TAZ shRNA-injected mice was significantly longer than that of control shRNA group (Figure 3H). To further validate the effect of TAZ depletion, we performed HDTV injection of NCID/AKT/transposase vectors plus CRISPR/Cas9-mediated TAZ knockout vector in FVB/NJ mice. Our data showed that deletion of TAZ by CRISPR/Cas9 also significantly inhibited CCA development (Figure 3I–K). Together, these findings demonstrate that TAZ is important in NICD-driven cholangiocarcinogenesis in mice. Consistent with the fact that TAZ and YAP are homologous coactivators in the transcriptional complex of Hippo signaling pathway, in our system we observed that depletion of YAP also delayed the development of NICD/AKT-induced CCA development (Figure 3B and C).

Figure 3.

TAZ is important in NICD/AKT-induced CCA growth. (A) Gene delivery strategy via HDTV injection. FVB/NJ mice were subjected to HDTV injection of NCID, AKT, and transposase plasmids with or without control/TAZ/YAP shRNA vectors (shCntrl/shTAZ/shYAP) or CRISPR/Cas9 knockout vectors (sgCntrl/sgTAZ), and the animals were sacrificed 6 weeks after tail vein injection. (B) Gross photographs of livers recovered from mice receiving TAZ/YAP shRNA or control vector. (C) Knockdown of TAZ/YAP by shRNA decreased the liver/body weight ratio (n = 8 for each group, 1 shTAZ#2 liver was not available because that mouse died after HDTV injection). (∗∗P < .01, compared with shCntrl). (D) Western blotting analysis confirmed TAZ protein reduction in CCA tissues from TAZ shRNA group. (E) IHC staining for TAZ and SOX9 in NICD/AKT-induced CCA with or without TAZ knockdown. (F) TUNEL staining assay in NICD/AKT-induced CCA with or without TAZ knockdown. Liver tissue from mouse model of Jo2-induced hepatocyte apoptosis was used as positive control. (G) IHC staining for the cell proliferation marker Ki67. (∗∗P < .01, compared with shCntrl). (H) The survival time of mice following HDTV injection of NCID/AKT and transposase plasmids along with TAZ-shRNA vector (n = 9) or control-shRNA vector (n = 10). (I) Gross photographs of livers recovered from mice receiving CRISPR/Cas9-mediated TAZ knockout vector or control vector. (J) Knockout of TAZ by sgRNA reduced the liver/body weight ratio (n = 7 for each group). (∗∗P < .01, compared with shCntrl). (K) Western blotting analysis were performed to detect the expression of TAZ in CCA tumors with control or TAZ sgRNA.

Notch-regulated TAZ Expression Involves m6A Methyltransferase METTL3

It is well-documented that Notch signaling activation depends on γ-secretase-mediated proteolytic cleavage of Notch receptor and the release of NICD (Notch1 intracellular domain), which then translocates to the nucleus where it binds the transcription factor CSL to activate transcription.^23,24 To determine the mechanism of Notch-regulated TAZ expression, we next examined whether the Notch pathway might regulate TAZ mRNA expression. For this purpose, mouse and human CCA cells were treated with Notch inhibitors, and the levels of TAZ mRNA were measured by qRT-PCR assay. Our data showed that inhibition of Notch pathway did not affect the level of TAZ mRNA (Figure 4A). Although analyses of the TCGA CCA datasets showed a positive correlation between Notch1 protein and TAZ protein (as shown in Figure 2E), we observed that high Notch1 protein level did not correlate with TAZ mRNA (Figure 4B). These data suggest that TAZ mRNA may not be a direct transcriptional target of Notch signaling pathway in CCA.

Figure 4.

Notch signaling pathway regulates the expression of METTL3 m6A methyltransferase. (A) The cells were treated with the Notch inhibitors Avagacestat (40 μM) or PF-03084014 (40 μM) for 48 hours. The levels of mouse and human TAZ mRNA were measured by qRT-PCR assay. (B) Informatics analysis for the correlation between Notch1 protein and TAZ mRNA in CCA tissues from TCGA database. (C) The effects of MG132 and CHX on Notch-regulated TAZ expression. The CCLP1 cells with or without PF-03084014 treatment were treated with MG132 (5 μM) or CHX (20 μM) for 8 hours. The level of TAZ protein was measured by Western blotting. (D) Western blotting assay was performed to analyze the effects of Notch inhibitor Avagacestat (40 μM) or PF-03084014 (40 μM) on the level of METTL3 and TAZ protein in mouse MC22891 cells and human CCLP1 cells. (E) The effect of the Akt inhibitors on the level of METTL3 protein in MC22891 and CCLP1 cells. The cells were treated with the Akt inhibitors Afuresertib (15 μM) and MK-2206 (10 μM) for 48 hours. Western blotting analysis was performed to determine the effect of Akt inhibitors on the level of METTL3 protein in MC22891 and CCLP1 cells. (F) The effects of Notch inhibitor on the level of METTL3 mRNA in MC22891 cells and CCLP1 cells (∗∗P < .01, compared with DMSO). (G) Analysis of the correlation between Notch1 protein and METTL3 mRNA in human CCA from TCGA database. (H) The correlation between Notch signature score and METTL3 mRNA or TAZ protein from TCGA-CCA datasets. (I) Prediction of transcription factor binding sites using Eukaryotic Promoter Database revealed 2 potential binding sites for the NICD binding partner CSL in the promoter region of METTL3 gene. (J) ChIP-qPCP assay was performed to examine the binding of NICD/CSL complex to the promoter region of METTL3 gene in CCLP1 cells (P1/P2: primer#1 or primer#2) (∗∗P < .01, compared with IgG). (K) Treatment of CCLP1 cells with the Notch inhibitor PF-03084014 (40 μM) decreased the binding of NICD or CSL to METTL3 promoter (∗∗P < .01, compared with DMSO). (L) Western blotting assay was used to assess the effects of METTL3 (left) or TAZ (right) overexpression on the level of TAZ or METTL3 protein under Notch inhibitor treatment.

We next investigated whether the Notch pathway might regulate TAZ protein expression through a mechanism involving post-transcriptional regulation. To delineate whether proteolytic degradation or mRNA translation might be involved in Notch-regulated TAZ expression, we treated CCA cells with the proteasome inhibitor MG132 or the translation elongation inhibitor cycloheximide (CHX). Our results showed that treatment of CCA cells with MG132 evidently increased the level of TAZ protein, whereas the TAZ protein increase under MG132 treatment was still partly reduced by the Notch pathway inhibitor PF-03084014 (Figure 4C); these findings suggest that proteolytic degradation pathway is unlikely involved in Notch-regulated TAZ protein expression. In contrast, we observed that Notch inhibition was able to cooperate with the translation inhibitor CHX to further downregulate the level of TAZ protein (Figure 4C), suggesting that regulation of protein translation may be responsible for Notch-induced TAZ expression.

Accumulated evidence in recent years has revealed the role of m6A in determining the fates of mRNAs and the levels of proteins.^25,26 Given that the m6A methyltransferase METTL3 has been reported to sustain the expression of TAZ protein by promoting the translation of TAZ mRNA,²⁷ we evaluated whether Notch-regulated expression of TAZ protein in CCA cells might be mediated through regulation of METTL3-mediated m6A methylation. We began by treating mouse and human CCA cells with Notch or AKT inhibitors. Our data showed that the protein level of METTL3 was dramatically reduced by the Notch inhibitors (Figure 4D), but not by the AKT inhibitors (Figure 4E). qRT-PCR assay revealed that Notch inhibition significantly decreased the expression of METTL3 mRNA (Figure 4F). Analysis of the TCGA CCA dataset confirmed that Notch1 protein was positively correlated with METTL3 mRNA (Figure 4G). To further validate the correlation between Notch signaling and the expression of METTL3 mRNA or TAZ protein, we utilized a defined Notch signature score system²⁸ by taking the average of the normalized mRNA expression values of major Notch signaling components using the CCA expression profiling datasets from TCGA database. In this system, the major Notch signaling components include the Notch ligands (JAG1, JAG2, DLL1, DLL3, DLL4), receptors (NOTCH1, 2, 3, 4), NICD binding factors (CSL, MAML1), and several downstream targets (CCND1, CHUK, DLGAP5, DTX1, EPHB3, ERBB2, HES1, HES5, HEY1, HEY2, NRARP, PTCRA, SOX9). Our analysis indicates that the Notch signature score is positively correlated with the expression of both METTL3 mRNA or TAZ protein (Figure 4H).

Based on the above findings, we carried out further studies to determine whether Notch pathway might regulate METTL3 gene transcription. Indeed, by using the Eukaryotic Promoter Database, we identified 2 binding sites of CSL in the promoter region of METTL3 gene (Figure 4I). As CSL is central for the activation of Notch signaling, we next performed chromatin immunoprecipitation (ChIP)-qPCR assay to determine the association of NICD and CSL with the METTL3 gene promoter. Our data showed that both NICD and CSL were enriched in the promoter region of the METTL3 gene (Figure 4J); these enrichments were significantly reduced by treatment with the Notch inhibitor PF-03084014 (Figure 4K). Restoration of METTL3 expression in Notch inhibitor-treated CCA cells was able to reinstate the expression of TAZ protein, whereas restoration of TAZ expression did not re-establish the expression of METTL3 (Figure 4L); these findings suggest that METTL3 is upstream of TAZ in regulation of CCA cell functions.

The Oncogenic Role of METTL3 in CCA

We analyzed the expression of METTL3 in human CCA tissues by using TCGA database and 2 GEO datasets (GSE107943 and GSE76297). Our analyses unveiled a global dysregulation of m6A modifiers, including m6A methyltransferases (writers), m6A demethylases (erasers), and m6A binding proteins (readers) in human CCA. Among all 3 datasets, we observe that the dysregulated m6A modifiers in either unpaired CCA/non-CCA samples (Figure 5A, B) or paired CCA/non-CCA (Figure 5C) are highly consistent. As the catalytic core in the m6A methyltransferase complex,²⁹ we found that METTL3 is upregulated in human CCA from both unpaired CCA/non-CCA samples and paired CCA/non-CCA samples (Figure 5A–D). By qRT-PCR and Western blotting analyses, we found that the METTL3 mRNA in CCA cells is significantly higher than in noncancerous human biliary epithelial cells (BECs and H69) (Figure 6A). Accordingly, we observe that the level of METTL3 protein is also higher in CCA cells when compared with non-neoplastic biliary epithelial cells (Figure 6B).

Figure 5.

Up-regulation of METTL3 in human CCA. (A–C) Heatmap displaying global dysregulation of m6A regulators in CCA based on the data from TCGA database and GEO datasets, including the unpaired tumor/non-tumor tissues (A, B) and the paired tumor/non-tumor tissues (C). (D) Statistical graphs show upregulated METTL3 mRNA expression in human CCA samples from TCGA database and GEO datasets (∗∗P < .01, compared with control samples).

Figure 6.

The up-regulated METTL3 facilitates the growth and the invasion of CCA cells. (A) qRT-PCR was performed to examine the level of METTL3 mRNA in human CCA cell lines (CCLP1, SG231, HuCCT1), and the data were compared with the normal biliary epithelial cells (BECs and H69) (∗∗P < .01, compared with H69 cells). (B) METTL3 protein levels in human CCA lines (CCLP1, SG231, HuCCT1) were higher than in human biliary epithelial cells (BECs and H69). (C) The METTL3 shRNA knowdown efficiency was analyzed by Western blotting. (D) Depletion of METTL3 by shRNA inhibited cell proliferation, as measured by WST-1 assay (∗∗P < .01, compared with shCntrl). (E) shRNA depletion of METTL3 inhibited the colony formation ability of CCA cells. (F) Transwell assay indicated that shRNA mediated depletion of METTL3 inhibited the invasion ability of CCA cells. (G) Experimental study design for CCA xenograft in mouse livers. CCLP1 cells with or without METTL3 stable knockdown were mixed with Matrigel and then the cells were injected into the livers of SCID mice. After 6 weeks of injection, the mice were sacrificed, and the livers were collected. (H) Depletion of METTL3 inhibits CCA in vivo. Upper: Images of the livers bearing xenograft tumors recovered from 7 mice in each group (1 control liver was not available because that mouse died prior to sacrifice). Lower left: Depletion of METTL3 reduced the liver/body weight ratio of SCID mice. Lower right: the tumor volume of CCA xenograft (∗∗P < .01, compared with shCntrl). (I) IHC for METTL3 and Ki67 using CCA xenograft tissues. The statistical graph presents the Ki67-positive cells in control tissues and METLL3 knockdown tissues (∗∗P < .01, compared with shCntrl). (J) TUNEL staining in CCA xenograft tissues. Liver tissue from mouse model of Jo2-induced hepatocyte apoptosis was used as positive control.

We then established CCA cell lines with stable knockdown of METTL3 by shRNAs to assess the functional impact of METTL3. Satisfactory depletion of METTL3 protein in CCA cells transfected with METTL3 shRNAs was confirmed by Western blotting (Figure 6C). As shown in Figure 6D, knockdown of METTL3 by shRNA significantly decreased the proliferation of CCLP1, SG231, and HuCCT1 cells. Furthermore, our data indicated that shRNA depletion of METTL3 inhibited the colony formation ability of CCA cells (Figure 6E). We also evaluated the effect of METTL3 knockdown on CCA cell invasiveness by Transwell cell invasion assay and observed that METTL3 shRNA significantly decreased CCA cell invasion (Figure 6F).

Following the above-described in vitro studies, we assessed the effect of METTL3 knockdown on the progression of CCA in SCID mice. For this purpose, CCLP cells with or without stable knockdown of METTL3 were inoculated into the livers of SCID mice. Six weeks after the inoculation, the animals were sacrificed, and the livers were harvested (Figure 6G). As shown in Figure 6H, METTL3 knockdown in CCA cells led to significant reduction of tumor growth in SCID mice, as reflected by decreased tumor volume and reduced ratio of liver-to-body weight. IHC analysis confirmed decreased METTL3 protein in CCA tumors derived from METTL3-shRNA cells (Figure 6I). The tumor cells with METTL3 depletion exhibited decreased staining for the cell proliferation marker Ki-67 (Figure 6I) and no alteration of apoptosis under TUNEL assay (Figure 6J). These findings demonstrate that knockdown of METTL3 suppressed CCA cell growth, in vitro and in vivo.

METTL3 Mediates m6A Modification of TAZ mRNA

METTL3 functions as a RNA methyltransferase implicated in the regulation of mRNA biogenesis, decay, and translation through m6A modification.³⁰ To confirm METTL3-mediated m6A modification of TAZ mRNA, we performed m6A methylated RNA immunoprecipitation sequencing (MeRIP-seq) to identify m6A-methylated RNA profiles in CCA cells (CCLP1). Our MeRIP-seq results showed that 5427 genes (including 7324 transcripts) exhibited high level of m6A peaks (fold enrichment [IP/input] >10) (Figure 7A). Among these genes, there were 1120 genes showing decreased m6A peaks when METTL3 was knocked down (Figure 7A). Functional enrichment analysis of the above 1120 genes reveal that these genes are significantly enriched in cellular functions relevant to “cell cycle arrest,” “cell adhesion,” “angiogenesis,” and “cell migration,” which are critically important for CCA tumor progression (Figure 7B). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis indicates that these genes are associated with several cancer-associated signaling pathways, including the Hippo signaling pathway and the Wnt signaling pathway (Figure 7C).

Figure 7.

Global profiling of m6A in CCLP1 cells confirms METTL3-mediated m6A modification of TAZ mRNA. (A) Venn diagram showing the genes with decreased level of m6A upon METTL3 knockdown (fold change >2). MeRIP sequencing assay was used to profile the m6A methylome in control or METTL3-depleted cells. (B) Gene Ontology functional enrichment analysis using Database for Annotation, Visualization, and Integrated Discovery (DAVID). (C) KEGG pathway enrichment analysis using DAVID online tool. (D) Western blotting was used to evaluate the effects of METTL3 knockdown on some key molecules of Hippo pathway in CCLP1 cells. (E) qRT-PCR results reveal that METTL3 knockdown decreases Hippo pathway downstream genes in CCLP1 cells (∗∗P < .01, compared with shCntrl). (F) m6A enrichment in TAZ mRNA was confirmed by MeRIP-qPCR assay in both CCLP1 and HuCCT1 cells (∗∗P < .01, compared with IgG). (G) Analysis of the effects of METTL3 knockdown on TAZ m6A modification in CCA cells by MeRIP-qPCR assay (∗∗P < 0.01, compared with shCntrl). (H) The effect of METTL3 knockdown on the level of TAZ protein in the nuclei of CCA cells as determined by Western blotting analysis.

We then performed Western blotting assay to analyze the effect of METTL3 knockdown on Hippo pathway molecules in CCA cells. We observed that depletion of METTL3 decreased the level of TAZ protein, but not the protein levels of other Hippo pathway molecules (Figure 7D). By RT-qPCR assay, we further confirmed that depletion of METTL3 by shRNA reduced the level of Hippo-TAZ pathway downstream genes (including CTGF, CYR61, EDN1, and ZEB1) in CCLP1 cells (Figure 7E). The MeRIP-seq data were further validated by MeRIP-qPCR assay, which showed that TAZ mRNA enriched by m6A-specific antibody was significantly higher when compared with the IgG control antibody in CCA cells (Figure 7F). Consistent with these findings, we observed that depletion of METTL3 by shRNA significantly reduced the level of m6A enrichment in TAZ mRNA (Figure 7G). Furthermore, we examined the effect of METTL3 knockdown on the level of TAZ protein in the nuclei of CCA cells and found that METTL3 depletion led to reduction of nuclear TAZ protein (Figure 7H).

METTL3 GapmeR Antisense Oligonucleotides Inhibit TAZ signaling and Decrease the Malignant Characteristics of CCA Cells

To further validate the role of METTL3-mediated regulation of TAZ in CCA cells, we designed and synthesized locked nucleic acid (LNA) GapmeR antisense oligonucleotides (ASOs) specific for METTL3. The produced METTL3 GapmeRs were assessed for their effects in CCA cells. Western blotting analysis revealed that METTL3 GapmeR ASOs effectively decreased the level of METTL3 protein in CCA cells (Figure 8A). Treatment with the METTL3 GapmeR ASOs led to decreased level of TAZ protein in CCA cells (Figure 8A). RT-qPCR assay showed that the METTL3 GapmeR ASOs significantly decreased the level of Hippo-TAZ pathway downstream genes in CCA cells (Figure 8B). By using WST-1 cell proliferation assay, we found that treatment with the METTL3 GapmeR ASOs significantly decreased the proliferation of CCA cells (Figure 8C). Colony formation assay showed that METTL3 GapmeR ASOs reduced the clonogenicity of CCA cells (Figure 8D).

Figure 8.

Knockdown of METTL3 by LNA (Locked Nucleic Acid)-GapmeR ASO in human CCA cells (CCLP1 and HuCCT1). The cells were transfected with METTL3 GapmeR ASOs for 48 hours. (A) Western blotting assay to examine the effect of METTL3 GapmeR ASOs on the levels of METTL3 and TAZ protein. The cells were transfected with 2 sets of METTL3 GapmeR ASOs (#1 and #2) for 48 hours. (B) The effect of METTL3 GapmeR ASO on Hippo pathway downstream genes as assessed by qRT-PCR assay. The cells were transfected with METTL3 GapmeR ASO#1 for 48 hours. (C) CCA cell proliferation was measured by WST-1 assay after transfection with METTL3 GapmeR ASOs for 120 hours. (D) Colony formation assay in CCA cells transfected with scramble control or METTL3 GapmeR ASOs. (∗∗P < .01, compared with scramble control [SC]).

METTL3 Overexpression Enhances TAZ Signaling in METTL3-depleted CCA Cells and in Non-neoplastic Biliary Epithelial Cells

We next performed rescue experiments to determine whether forced overexpression of METTL3 in METTL3-depleted CCA cells were able to reverse the effects seen with METTL3 depletion. As the level of METTL3 is low in non-neoplastic biliary epithelial cells, we also performed experiments with forced overexpression of METTL3 in the non-cancerous biliary epithelial cell line, H69. Successful METTL3 restoration or overexpression was confirmed by Western blotting analysis (Figure 9A). We observed that METTL3 restoration or overexpression in these cells upregulated the level of TAZ protein (Figure 9A). RT-qPCR assay showed that METTL3 restoration or overexpression significantly induced the expression of TAZ downstream genes (Figure 9B). We next performed WST-1 assay to assess the effects of METTL3 restoration or overexpression on cell proliferation. Our data showed that restoration of METTL3 expression in CCA cells rescued the reduction of cell proliferation caused by METTL3 depletion. Meanwhile, we observed that forced overexpression of METTL3 in H69 cells enhanced their proliferation (Figure 9C). Transwell cell invasion assay showed that METTL3 restoration rescued the reduction of CCA cell invasion caused by METTL3 depletion, whereas forced overexpression of METTL3 enhanced the migration of H69 cells (Figure 9D). To further validate that TAZ is a downstream target of METTL3, we performed rescue experiments in METTL3-depleted cells with overexpression of TAZ. Satisfactory transfection of TAZ was confirmed by Western blotting (Figure 9E). By performing WST-1 cell proliferation assay and transwell cell invasion analysis, we observed that restoration of TAZ expression rescued the reduction of cell proliferation and invasion caused by METTL3 depletion (Figure 9E, F). These findings further support the oncogenic role of METTL3 in biliary carcinogenesis.

Figure 9.

The effects of METTL3 restoration or forced overexpression. (A) Western blotting analysis showed that restoration of METTL3 expression in METTL3 stable knockdown cells (CCLP1 and HuCCT1) or forced overexpression of METTL3 in H69 cells upregulated TAZ protein. (B) The mRNA levels of TAZ target genes were analyzed by qRT-PCR in METTL3 stably depleted CCLP1 cells or in H69 cells with or without METTL3 re-expression/overexpression. (C) WST-1 cell proliferation assay were used to evaluate the effect of METTL3 re-expression or overexpression on cell growth. (D) Transwell cell invasion assay was performed to assess the effect of METTL3 re-expression or overexpression on cell invasiveness. (E) The effect of TAZ overexpression on the growth of METTL3-depleted CCA cells was measured by WST-1 cell proliferation assay. (F) Transwell cell invasion analysis was performed to determine the effect of TAZ overexpression on the invasion of METTL3-depleted CCA cells. ∗∗P < .01, compared with the empty vector for overexpression (oeVector).

METTL3 Enhances the Expression of TAZ Protein via the m6A Reader Protein YTHDF1

The functional impact of RNA m6A modification depends on its specific readers. Recent studies have identified different m6A readers which are known to regulate multiple RNA metabolic process, including translation, stability, degradation, etc.^25,31 To explore the mechanism by which METTL3 regulates TAZ, we performed qRT-PCR analysis to evaluate the effect of METLL3 knockdown on TAZ mRNA. As shown in Figure 10A, knockdown of METTL3 did not affect the level of TAZ mRNA, suggesting that METTL3-mediated m6A modification is not involved in regulating the stability or degradation of TAZ mRNA.

Figure 10.

METTL3 regulates TAZ protein and cell growth via a YTHDF1-dependent mechanism in human CCA cells. (A) The effect of METTL3 knockdown by shRNA on TAZ mRNA in CCLP1 cells and HuCCT1 cells, as measured by qRT-PCR assay. (B) The effects of MG132 and CHX on METTL3-mediated regulation of TAZ protein. CCLP1 cells with or without METTL3 stable knockdown were treated with MG132 (5 μM) or CHX (20 μM) for 8 hours. Western blotting assay was performed to examine the changes of TAZ protein. (C, D) Increased expression of the m6A readers, YTHDF1 (C) and YTHDF3 (D), in human CCA tissues. Informatics analysis of the expression of m6A readers in TCGA and GEO tissue samples (GSE107943) revealed upregulated expression of YTHDF1 and YTHDF3 in human CCA (∗∗P < .01, compared with control samples). (E, F) Western blotting analysis showing the effects of YTHDF1 (E) or YTHDF3 (F) knockdown on the level of TAZ protein in CCLP1 and HuCCT1 cells. (G) Western blotting analysis revealed that re-expression of METTL3 in METTL3 depleted CCA cells restored the level of TAZ protein and that the effect was reversed by YTHDF1 knockdown. (H) WST-1 cell proliferation assay showed that re-expression of METTL3 in METTL3-depleted CCA cells restored cell proliferation and that the effect was reversed by YTHDF1 knockdown (∗∗P < .01, compared with scramble control [SC]). (I) WST-1 cell proliferation assay was used to analyze the effect of TAZ overexpression on the growth of YTHDF1-depleted CCA cells (∗∗P < .01, compared with oeVector). (J) Transwell cell invasion assay was performed to analyze the effect of TAZ overexpression on the invasion of YTHDF1-depleted CCA cells.

We then assessed the potential role of proteolytic degradation or mRNA translation in METTL3-regulated TAZ expression by treating CCA cells with MG132 or CHX. Our data showed that METTL3 depletion was still able to decrease the level of TAZ protein under MG132 or CHX treatment (Figure 10B), suggesting that mRNA translation, but not proteolytic degradation, may account for METTL3-regulated TAZ expression.

We next explored the possibility of whether METTL3-regulated expression of TAZ protein might require translation-associated m6A reader proteins. Given that YTHDF1 and YTHDF3 are m6A reader proteins implicated in regulation of translation as previously documented,^25,32 we performed bioinformatics analyses from the TCGA and GEO CCA database. Our analyses revealed that the m6A reader proteins, YTHDF1 and YTHDF3, were upregulated in human CCA tissues (Figure 10C, D). Therefore, we carried out further studies to evaluate the potential effect of YTHDF1 and YTHDF3 on the expression level of TAZ protein by using specific siRNAs against YTHDF1 or YTHDF3. Our data showed that knockdown of YTHDF1 decreased the level of TAZ protein (Figure 10E), whereas knockdown of YTHDF3 exhibited no significant effect (Figure 10F). In CCA cells with stable knockdown of METTL3 by shRNA, METTL3 re-expression mediated upregulation of TAZ protein was abolished by siRNA knockdown of YTHDF1 (Figure 10G). Furthermore, our data showed that restoration of METTL3 expression in METTL3-depleted CCA cells rescued the reduction of cell proliferation caused by METTL3 depletion and that this effect was inhibited by knockdown of YTHDF1 (Figure 10H). We then performed rescue experiments in YTHDF1 knockdown cells with overexpression of TAZ. Although knockdown of YTHDF1 by siRNA inhibited the proliferation and the invasion of CCA cells, the effects were rescued by restoration of TAZ overexpression (Figure 10I, J). Together, our results demonstrate that METTL3-mediated m6A modification maintains the expression of TAZ protein via a YTHDF1-dependent mechanism in CCA cells.

The Role of METTL3 in Notch-driven Cholangiocarcinogenesis

Given our experimental findings that TAZ is important in Notch-driven cholangiocarcinogenesis and that Notch-regulated TAZ expression is dependent on METTL3-mediated m6A modification, we further assessed the functional impact of METTL3-regulated TAZ expression in Notch-driven CCA development. Our data showed that the CCA tumor tissues harvested from the mice receiving HDTV injection of NICD/AKT showed an increased METTL3 expression when compared with the controls as determined by Western blotting (Figure 11A). Increased METTL3 expression in CCA tissues was further confirmed by IHC staining (Figure 11B).

Figure 11.

Expression and function of METTL3 in NICD/AKT-induced CCA. (A) Western blot analysis for METTL3 protein in the liver tissues from mice with or without HDTV injection. (B) IHC for METTL3 in normal liver and the liver with NICD/AKT-induced CCA tumor. (C) Experimental study design. FVB/NJ mice were subjected to HDTV injection of NCID, AKT, and transposase plasmids with or without control/METTL3 shRNA vectors. Five weeks after HDTV injection, the animals were sacrificed to determine CCA tumor burden. (D) Gross photographs of the livers showing that inhibition of METTL3 by shRNA prevented CCA development (n = 7 for each group). (E) Knockdown of METTL3 by shRNAs reduced the liver/body weight ratio of FVB/NJ mice. (F) The survival time of mice following HDTV injection of NCID/AKT and transposase plasmids along with METTL3-shRNA vector (n = 9) or control-shRNA vector (n = 10). (G, H) IHC staining for SOX9 (G) and Ki67 (H) in NICD/AKT-induced CCA with or without METTL3 knockdown. (I) IHC staining for METTL3 and TAZ. (J) Western blotting analysis showed the decreased METTL3 and TAZ in NICD/AKT-induced CCA with METTL3 shRNA injection. (∗P < .05; ∗∗P < .01, compared with shCntrl)

We then utilized mouse model of HDTV injection-induced CCA to evaluate the effect of METTL3 knockdown on Notch-driven CCA development. To this end, wild type mice were subjected to HDTV injection of NCID, AKT, and transposase plasmids with control or 2 METTL3 shRNA vectors. Five weeks after HDTV injection, the mice were sacrificed to examine CCA tumor burden (Figure 11C). We observed that depletion of METTL3 by shRNAs noticeably inhibited the tumor growth in this model of cholangiocarcinogenesis (Figure 11D). The liver/body weight ratio in METTL3 shRNA-injected mice was significantly lower than that in the control mice (Figure 11E). The survival time of METTL3 shRNA-injected mice was significantly longer than that of control mice (Figure 11F). IHC staining for the biliary marker SOX9 further confirmed decreased CCA cell mass in METTL3 shRNA-injected mice (Figure 11G). Our data also showed that METTL3 knockdown reduced the numbers of Ki67-positive tumor cells (Figure 11H). IHC staining analysis showed decreased expression of METTL3 and TAZ proteins in METTL3-depleted CCA tumors (Figure 11I); these results were further confirmed by Western blotting analysis (Figure 11J). Together, our findings demonstrate an important role of METTL3-mediated TAZ expression in cholangiocarcinogenesis.

Discussion

Our experimental results presented in this study provide novel evidence that TAZ is indispensable in NICD-driven cholangiocarcinogenesis. Mechanistically, our data indicate that NICD induces the expression of METTL3 which catalyzes m6A modification of TAZ mRNA. m6A-modified TAZ mRNA is then recognized by the m6A reader protein YTHDF1 leading to enhanced TAZ protein translation in CCA cells. Our experimental results indicate this mechanism plays an important role in the interplay between Notch pathway and TAZ, which coordinately regulates cholangiocarcinogenesis.

Notch signaling pathway is a highly evolutionarily conserved cascade which plays a vital function in biliary fate and morphogenesis.^4,33 Studies have shown that the cooperation between YAP and Notch pathway plays an important role in cell fate decision and cholangiocarcinogenesis.^11,34 Although YAP-induced conversion of hepatocytes to CCA is controlled by Notch signaling,^8,12 Notch-driven CCA formation is further accelerated by YAP.¹⁰ However, it remains unknown whether TAZ interacts with the Notch pathway in CCA cells. Our experimental results presented in this study provide the first evidence that Notch pathway activation leads to increased TAZ protein expression in CCA cells. TAZ protein increase is not directly driven by transcription of the TAZ gene in CCA cells. Instead, our findings show that Notch pathway activation induces the expression of m6A methyltransferase METTL3 via transcriptional regulation, which in turn leads to m6A modification of the TAZ mRNA and thus increased TAZ protein.

TAZ is a paralog of YAP, which interacts with YAP and TEA domain (TEAD) DNA binding proteins in the nucleus to initiate the transcription of the Hippo pathway downstream genes involved in tumorigenesis.³⁵ In the current study, we observed that depletion of TAZ or YAP showed similar effects in suppressing the development of NICD/AKT-induced CCA. This is not surprising considering that TAZ and YAP are homologous transcriptional cofactors regulated by the Hippo kinase cascade. However, the contribution of YAP and TAZ in Notch-induced cholanciocarcinogenesis may involve different mechanisms, as evidenced by the experimental results presented in the current study that METTL3-mediated m6A modification is implicated in Notch-regulated TAZ expression but not in the expression of YAP in CCA cells. Also of note, previous studies have demonstrated substantial differences between TAZ and YAP in their structure and regulatory mechanisms.^13,14

Recent studies report that the activation of TAZ is associated with chromosomal instability and CCA progression.36, 37, 38 Notably, HDTV injection of TAZ has been shown to successfully induce the formation of CCA,⁸ indicating that this gene is a key driver in cholangiocarcinogenesis. The current study discloses an important connection between the Notch pathway and TAZ. Our experimental findings reveal that the expression of TAZ is importantly regulated by METTL3 and that METTL3/TAZ expression plays a vital role in Notch-driven CCA. Our further results indicate that depletion of METTL3 or TAZ significantly delays Notch-induced CCA development and inhibits CCA cell proliferation, invasion, and colony formation. Thus, it is conceivable that the METTL3-TAZ cascade may be implicated in hepatocyte transformation into intrahepatic CCA as well as in the proliferation, survival, and invasion of CCA cells.

Notch activation upregulates the expression of TAZ protein through METTL3-mediated m6A modification. As the most abundant mRNA modification in eukaryotes, m6A has been shown to play a predominant role in regulating multiple physiological processes in the life cycle of mRNA, including splicing, nuclear export, degradation, and translation.^30,39 To date, the mechanism for METTL3 upregulation in cancer is largely unknown. Our data presented in the current study provide novel evidence for upregulation of METTL3 by the Notch pathway, which is mediated through NICD binding to its downstream transcription factor CSL in the promoter region of the METTL3 gene.

The roles of METTL3 mediated m6A in carcinogenesis and tumor progression are complex. Upregulated expression of METTL3 has been observed in various human cancers including CCA,^22,40 which supports an oncogenic role of METTL3. In contrast, some studies report that METTL3 is down-regulated in certain cancer types including lung adenocarcinoma⁴¹ and triple-negative breast cancer,⁴² which suggest a tumor-inhibiting role of METTL3 in these cancers. In the present study, we observe that the expression of METTL3 is significantly increased in both human and mouse CCA. The carcinogenic role of METTL3 in CCA is demonstrated by the findings that knockdown of METTL3 by shRNA inhibits CCA cell proliferation, colony formation, and migration/invasion in vitro and prevents CCA development/progression in mice.

Mechanistically, our data show that METTL3-mediated m6A modification establishes a crosstalk between Notch and Hippo-TAZ pathways, which coordinately regulate cholangiocarcinogenesis. The core Hippo kinases include MST1/2 and LATS1/2, which, along with NF2, are well-documented tumor suppressors. These kinases phosphorylate and inactivate the downstream effector proteins YAP and/or TAZ through proteasomal degradation.⁴³ Our experimental findings presented in this study reveal that Notch activation stimulates the transcription of METTL3, which leads to m6A modification of TAZ mRNA. The m6A modified TAZ mRNA is recognized by the m6A reader protein YTHDF1, leading to increased TAZ protein translation in CCA cells. Consistent with these findings, our data reveal that depletion of METTL3 inhibits CCA development and progression through downregulation of TAZ protein in an m6A-YTHDF1-dependent manner.

In summary, the current study discloses an essential role of METTL3-mediated m6A modification in the interplay between the Notch pathway and TAZ in cholangiocarcinogenesis. Our results demonstrate that TAZ is important in Notch-driven CCA development/progression, which is dependent on METTL3-mediated m6A modification. Detailed mechanistic investigations depict a novel Notch-METTL3-TAZ signaling axis, which is critically implicated in the regulation of cholangiocarcinoma cell growth. It is conceivable that therapeutic strategies aimed at this signaling axis may have important implication in CCA treatment.

Materials and Methods

Materials

Dulbecco’s modified minimum essential medium (DMEM) and fetal bovine serum (FBS) were purchased from Sigma. Epithelial cell medium (EpiCM) was purchased from ScienCell Research Laboratories. RT-qPCR, MeRIP-qPCR and ChIP-qPCR primers were synthesized at Thermo Fisher Scientific. The pLKO-control shRNA vector (#10879) and pLKO-TAZ shRNA vectors (mouse, #TRCN0000095953 and #TRCN0000095951), pLKO-YAP vector (mouse, #TRCN0000238436), pLKO-METTL3 shRNA vectors (human, #TRCN0000289814, #TRCN0000034717; mouse, #TRCN0000039111 and #TRCN0000039112) were purchased from Addgene and MilliporeSigma, respectively. The METTL3 GampeR antisense oligonucleotides (ASOs, #LG00785443 and #LG00785444) and the control ASO (#LG00000002) were synthesized at Qiagen. The plasmid constructs, including pT3-EF1a-NICD1 (#46047), pT3-myr-AKT-HA (#31789), pcDNA3/Flag-METTL3 (#53739), pcDNA3-HA-TAZ (#32839) were purchased from Addgene. The siRNA duplexes were synthesized at Integrated DNA Technologies. The sequences of GapmeR ASOs and siRNAs are listed in Table 1. Lipofectamine 3000 reagent, glutamine, and antibiotics were obtained from Invitrogen. Rabbit polyclonal antibodies against METTL3, TAZ, YAP, YTHDF1, YTHDF3, and HA-tag were purchased from Proteintech. Rabbit monoclonal antibodies against HA-tag/LATS1/LATS2 and mouse monoclonal antibodies against Myc-tag/Ki67/ GAPDH were purchased from Cell Signaling Technology. Mouse monoclonal antibody against β-actin was purchased from Sigma. IRDye goat anti-mouse/rabbit IgG secondary antibodies were purchased from LI-COR Biosciences.

Table 1.

The Sequences of siRNAs and GapmeR ASOs Used in this Study

siRNAs/ASOs	Sequence
hYTHDF1 siRNA#1	Sense: rArUrUrUrArGrArGrUrArUrUrCrUrGrArUrArArArArUrCTC Antisense: rGrArGrArUrUrUrUrArUrCrArGrArArUrArCrUrCrUrArArArUrGrA
hYTHDF1 siRNA#2	Sense: rGrUrCrUrArGrUrUrGrUrUrCrArUrGrArArGrCrArUrGrUCG Antisense: rCrGrArCrArUrGrCrUrUrCrArUrGrArArCrArArCrUrArGrArCrGrC
hYTHDF3 siRNA#1	Sense: rCrGrArGrCrCrUrGrArUrUrGrCrUrArUrCrArUrGrArArGTA Antisense: rUrArCrUrUrCrArUrGrArUrArGrCrArArUrCrArGrGrCrUrCrGrCrA
hYTHDF3 siRNA#2	Sense: rGrGrArUrUrArArArUrCrArGrUrArUrCrUrArArGrUrGrAAT Antisense: rArUrUrCrArCrUrUrArGrArUrArCrUrGrArUrUrUrArArUrCrCrArU
mCSL siRNA#1	Sense: rGrCrArUrUrUrUrArCrCrUrUrArArGrGrArUrArCrArGrAAA Antisense: rUrUrUrCrUrGrUrArUrCrCrUrUrArArGrGrUrArArArArUrGrCrArC
mCSL siRNA#2	Sense: rGrGrCrArArGrGrArUrArArArGrUrGrArArCrArArUrCrUTT Antisense: rArArArGrArUrUrGrUrUrCrArCrUrUrUrArUrCrCrUrUrGrCrCrUrU
mAKT1 siRNA#1	Sense: rGrCrGrUrGrGrUrCrArUrGrUrArCrGrArGrArUrGrArUrGTG Antisense: rCrArCrArUrCrArUrCrUrCrGrUrArCrArUrGrArCrCrArCrGrCrCrC
mAKT1 siRNA#2	Sense: rCrArCrGrCrUrUrArCrUrGrArGrArArCrCrGrUrGrUrCrCTG Antisense: rCrArGrGrArCrArCrGrGrUrUrCrUrCrArGrUrArArGrCrGrUrGrUrG
mAKT2 siRNA#1	Sense: rArGrCrUrArCrUrCrUrCrUrUrGrArUrUrCrUrCrArArUrAAA Antisense: rUrUrUrArUrUrGrArGrArArUrCrArArGrArGrArGrUrArGrCrUrUrG
mAKT2 siRNA#2	Sense: rGrArGrArUrGrUrGrGrUrGrUrArCrCrGrUrGrArCrArUrCAA Antisense: rUrUrGrArUrGrUrCrArCrGrGrUrArCrArCrCrArCrArUrCrUrCrUrC
METTL3 ASO#1	T∗G∗C∗G∗T∗T∗G∗C∗A∗G∗T∗T∗G∗A∗T∗T
METTL3 ASO#2	T∗G∗G∗T∗C∗A∗G∗C∗A∗T∗A∗G∗G∗T∗T∗A

ASO, Antisense oligonucleotide; siRNA, small interfering RNA.

Cell Culture and Transfections

Human CCA cell lines (CCLP1, SG231, HuCCT1) and human bile duct epithelial cell line (H69) were cultured according to our previous described methods.44, 45, 46, 47 Mouse CCA cells (MC22891) isolated from CCA-bearing mice induced by HDTV injection of NICD, AKT, and transposase plasmids, as we recently described,²¹ were also utilized. Human primary BECs were purchased from ScienCell Research Laboratories and cultured in EpiCM supplemented with 2% FBS, 1% epithelial cell growth supplement, and 1% penicillin/streptomycin in a humidified atmosphere of 5% CO₂ incubator at 37 °C.

For transfections, METTL3 GapmeR ASOs, YTHDF1/YTHDF3 siRNAs, and their respective scramble controls were transfected into CCA cells using Lipofectamine 3000 reagent. Following transfections, the cells were analyzed for proliferation, invasion, colony formation, and other parameters as described in the manuscript.

For establishment of CCA cells with stable depletion of METTL3, CCLP1, HuCCT1, and SG231 cells were transfected with METLL3 shRNA or control vectors. After 48 hours of transfection, the cells were cultured in DMEM medium containing 1 μg/mL Puromycin (Calbiochem). The selection medium was replaced every 3 days for the next 4 weeks. Subsequently, distinct colonies of surviving cells were transferred onto 6-well plates and the cultures were maintained under the same selection medium. For METTL3 overexpression, pcDNA3/Flag-METTL3, pcDNA3-HA-TAZ, and pcDNA3 control vector were transfected into CCA cells using Lipofectamine 3000 reagent. Following transfections, the cells were analyzed for proliferation, invasion, and specific protein levels.

Cell Proliferation WST-1 Assay and Clonogenicity Assay

Cell proliferation WST-1 assay was performed according to the manufacturer’s instruction. For cells with stable depletion of METTL3, 2 × 10³ cells were seeded onto each well of 96-well plates and cultured for 5 days. For GapmeR ASO transfection, 2 × 10³ cells were seeded onto each well of 96-well plates and the cells were transfected with GapmeR ASOs in the presence of Lipofectamine 2000 reagent for 6 hours; the cells were then continued cultured in fresh DMEM medium for additional 5 days. For Notch/Akt inhibitor treatment, 4 × 10³ cells were seeded onto each well of 96-well plates, and the cells were treated with Notch inhibitors (40 μM Avagacestat/40 μM PF-3084014) or AKT inhibitors (15 μM Afuresertib/10 μM MK-2206) for 48 hours. To determine cell proliferation, 10 μl WST-1 reagent was added to each well, and the cells were incubated for 1 hour at 37 °C and 5% CO₂. A450 nm was measured using an automatic enzyme-linked immunosorbent assay (ELISA) plate reader.

For cell colony formation assay, CCA cells transfected with METTL3 GapmeR ASOs or shRNA and their respective controls were seeded onto 6-well plates (500 cells/well). After culture for 14 days, the cells were fixed with methanol and stained with 0.1% crystal violet.

Transwell Invasion Assay

Cell invasion assays were performed using 24-well Corning BioCoat Matrigel transwell chambers with 8-μm pore size PET membrane according to the manufacturer's instructions (Fisher Scientific). Briefly, 0.5 mL DMEM medium containing 1% FBS and 3 × 10⁴ CCLP1 cells or 5 × 10⁴ SG231/HuCCT1/H69 cells were added to the upper chamber and 0.5 mL DMEM medium with 10% FBS was added to the bottom well. After 24 hours (for CCLP1/HuCCT1/H69 cells) or 36 hours (for SG231 cells) of incubation, the non-invaded cells were removed by a cotton tip from the upper side of the chamber. The attached cells at the lower section were stained with 0.1% crystal violet.

Protein Extraction and Western Blotting

For whole cell protein extraction, the cells were washed twice with ice-cold phosphate buffered saline (PBS) and lysed in RIPA buffer containing 10 mM Tris-Cl (pH 8.0), 1 mM EDTA, 1% Triton X-100, 0.1% sodium deoxycholate, 0.1% SDS, 140 mM NaCl, and 1 mM PMSF. After sonication on ice, the cell lysates were centrifuged at 13,000 rpm for 10 minutes at 4 °C, and the supernatants were collected for Western blotting. The protein concentrations were measured using the Bio-Rad Protein Assay Kit (Bio-Rad).

For Western blotting analysis, samples were boiled for 5 minutes in protein loading buffer with 2-mercaptoethanol and subjected to 10% sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE). The proteins were then transferred onto the nitrocellulose membrane (BioRad, Hercules, CA). Non-specific binding was blocked by incubating the membranes in PBST (0.1% Tween 20 in PBS) containing 5% nonfat milk for 1 hour at room temperature. The membranes were then incubated overnight at 4 °C with individual primary antibodies at the dilutions recommended by the manufacturers in PBST containing 5% nonfat milk. Following 4 washes with PBST, the membranes were incubated with the IRDye secondary antibodies at 1: 5000 dilutions in PBST containing 5% nonfat milk for 1 hour at room temperature. After 4 washes with PBST, the ODYSSEY Infrared Imaging System (Licor) was used to visualize protein bands.

qRT-PCR

Total RNA was extracted using Tri-zol Reagents (Invitrogen) following the manufacturer’s instructions. Reverse transcription was performed with iScript Supermix (Bio-Rad). Quantitative PCR was performed with the Bio-Rad SYBR Green Supermix in a C1000 thermal cycler (Bio-Rad). The PCR conditions were 15 minutes at 95 °C, followed by 35 cycles of 15 seconds at 94 °C, 30 seconds at 55 °C and 30 seconds at 72 °C, and finally 10 minutes at 72 °C. The PCR primer sequences are listed in Table 2. β-actin was used as the internal control. Results were analyzed by using CFX Manager Software version 3.1 (Bio-Rad). The expression level of mRNA was normalized to the internal control gene, and relative change was calculated by using the 2^−ΔΔCT method.

Table 2.

Primer Sequences for qRT-PCR

Gene	Sense (5′ - 3′)	Antisense (5′ - 3′)
hMETTL3	CAAGCTGCACTTCAGACGAA	GCTTGGCGTGTGGTCTTT
hTAZ	CCGTTTCCCTGATTTCCTTG	GGGATCAGGTCTTCAGATTCC
hCTGF	CTTGCGAAGCTGACCTGGAA	AAAGCTCAAACTTGATAGGCTTGGA
hCYR61	AGCCTCGCATCCTATACAACC	TTCTTTCACAAGGCGGCACTC
hEDN1	AGCCTCCTCTGCTCTTTCTGCTGGA	CTTTTGTCTATGCCCCTGCAGCCTT
hZEB1	GCCAATAAGCAAACGATTCTG	TTTGGCTGGATCACTTTCAAG
hβ-actin	CATGTACGTTGCTATCCAGGC	CTCCTTAATGTCACGCACGAT
mMETTL3	CTTTCTACCCCATCTTGAGTG	CCAACCTTCCGTAGTGATAGTC
mTAZ	GTCCATCACTTCCACCTC	TTGACGCATCCTAATCCT
mCTGF	CACTCTGCCAGTGGAGTTCA	AAGATGTCATTGTCCCCAGG
mCYR61	CGAGTTACCAATGACAACCCAG	TGCAGCACCGGCCATCTA
mEDN1	GCACCGGAGCTGAGAATGG	GTGGCAGAAGTAGACACACTC
mZEB1	TGGCAAGACAACGTGAAAGA	AACTGGGAAAATGCATCTGG
mCCND1	GCGTACCCTGACACCAATCTC	ACTTGAAGTAAGATACGGAGGGC
mERBB2	GAGACAGAGCTAAGGAAGCTGA	ACGGGGATTTTCACGTTCTCC
mHES5	mAGTCCCAAGGAGAAAAACCGA	GCTGTGTTTCAGGTAGCTGAC
mHEY1	CCGACGAGACCGAATCAATAAC	TCAGGTGATCCACAGTCATCTG
mNRARP	TTCAACGTGAACTCGTTCGGG	TTGCCGTCGATGACTGACTG
mSOX9	GAGCCGGATCTGAAGAGGGA	GCTTGACGTGTGGCTTGTTC
mβ-actin	ACCCTAAGGCCAACCGTGA	ATGGCGTGAGGGAGAGCATA

qRT-PCR, Quantitative real-time polymerase chain reaction.

ChIP

Cells were cross-linked by 1% formaldehyde for 10 minutes. Chromosome DNA was extracted according to the manufacturer's instructions provided by SimpleChIP Plus Enzymatic Chromatin IP Kit (Cell Signaling) and precipitated by using specific anti-NICD rabbit antibody and anti-CSL rabbit antibody. Rabbit IgG was used as negative control. The PCR conditions were 15 minutes at 95 °C, followed by 35 cycles of 15 seconds at 94 °C, 30 seconds at 55 °C, and 30 seconds at 72 °C, and then 10 minutes at 72 °C. The ChIP-PCR primer sequences are listed in Table 3.

Table 3.

Primer Sequences for ChIP-qPCR

Gene	Sense (5′ - 3′)	Antisense (5′ - 3′)
ChIP-METTL3#1	TCGGAACCCGATTTTAATAGAT	ACGTGATCTCTCTCTTAGGCT
ChIP-METTL3#2	GGGTGACAGAGTGAGTGAGA	ATACAAGGTTGGTGGTGGTG

ChIP, Chromatin immunoprecipitation; qPCR, quantitative polymerase chain reaction.

MeRIP-seq

Total RNA was extracted using Trizol reagent (Invitrogen) following the manufacturer’s procedure. The total RNA quality and quantity were analysis using the Bioanalyzer 2100 and RNA 6000 Nano LabChip Kit (Agilent). Following purification, the poly(A) mRNA fractions are fragmented into ∼100-nt-long oligonucleotides using divalent cations under elevated temperature. Then the cleaved RNA fragments were subjected to incubation for 2 hours at 4 °C with m6A-specific antibody (Synaptic Systems) in IP buffer supplemented with BSA (0.5 μg/μl). The mixture was then incubated with protein-A beads and eluted with elution buffer. Eluted m6A-containing fragments (IP) and untreated input control fragments are converted to final cDNA library in accordance with a strand-specific library preparation by deoxyuridine triphosphate (dUTP) method. And then the purified RNA fragments from m6A meRIP were sequenced with Illumina Novaseq 6000 platform following the vendor’s recommended protocol. Library preparation, high-throughput sequencing, and data analysis were done by LC Sciences.

MeRIP-qPCR assay

Total RNA was extracted by Trizol reagent (Invitrogen). MeRIP assays were performed using the EpiQuik CUT&RUN m6A RNA Enrichment (MeRIP) Kit (Epigentek) according to the manufacturer’s protocols. Target m6A-containing fragments were pulled down using a beads-bound m6A capture antibody. The enriched RNA was then released, purified, and eluted. The purified RNA was used for reverse transcription; qPCR was performed to evaluate the m6A modified RNA levels. The primer sequences of TAZ used in MeRIP-qPCR assay were F, 5′-GAGACACAAGCGGACCCC-3′; R, 5′-CTGTTCCAGTTGCCGGATCA-3′.

IHC

The CCA tumors and liver tissues were fixed in 10% buffered formalin and embedded in paraffin. Sections of 4-μm thickness were deparaffinized and processed for hematoxylin and eosin (H&E) staining and IHC. Primary antibodies against METTL3, TAZ, SOX9, or Ki67 were diluted in 1× PBS containing 4% horse serum, 0.4 mg/mL methiolate, and 0.2% Triton-X100. (The titers for primary antibodies were: anti-METTL3, 1:700; anti-SOX9, 1:1000; anti-HA-tag, 1:600, anti-TAZ, 1:500 and 1:300.) After blocking with Peroxidazed 1 (Biocare Medical) for 5 minutes, the slides were incubated with primary antibodies at room temperature for 1 hour. The slides were then washed with tris-buffered saline (TBS) and incubated with horseradish peroxidase-conjugated second antibody at room temperature for 1 hour. After washing with TBS, the slides were incubated for 5 minutes at room temperature with 3,3′-diaminobenzidine (DAB) for chromogenic development.

TUNEL Assay

The DeadEnd Colorimetric TUNEL System (Promega Corporation) was used for the examination of apoptotic cell death of the liver tissue. Slides with formalin-fixed sections (5-μm thick) were washed twice in xylene (5 minutes each time), hydrated in 100% ethanol for 2 minutes, and washed in decreasing concentrations of ethanol (100%, 95%, and 80%; 2 minutes each time). The slides were then immersed in water for 2 minutes and processed for TUNEL staining, according to the manufacturer’s guidelines. For positive control, liver tissues recovered from mice 4 hours after intraperitoneal injection of Jo2 (anti-Fas antibody, 0.35 μg/g of body weight) (BD Bioscience) were used for TUNEL staining.

Bioinformatics Analyses

The gene expression profiles of human CCA and non-cancerous tissues were downloaded from the GEO datasets (GSE107943 and GSE76297) and TCGA database. GSE107943 consisted of 30 CCA samples and 27 non-CCA samples.⁴⁸ GSE76297 consisted of 90 CCA samples and 90 paired non-CCA samples.⁴⁹ TCGA database consisted of 36 CCA samples and 9 non-CCA samples. Fragments per kilobase of exon model per million mapped fragments (FPKM) values from the TCGA database were used for gene expression.

Animals

The 5-week-old male athymic nude NOD CB17-Prkdc/SCID mice and 6- to 8-week-old male FVB/NJ mice were obtained from Jackson Laboratory. For all animal studies, the procedures were carried out in strict accordance with the National Institutes of Health Guidelines for the Care and Use of Laboratory Animals. The handling of the mice and all experimental procedures were approved for this study by the Institutional Animal Care and Use Committee of Tulane University (Protocol #: 4159).

Studies in CCA Xenograft

The CCA xenografts were established by inoculating 1 × 10⁶ human CCA cells (CCLP1) into the livers of 5-week-old male athymic nude NOD CB17-Prkdc/SCID mice (obtained from Jackson Laboratory). To develop CCA xenografts with METTL3 depletion, 1 × 10⁶ CCLP1 cells stably transfected with the METTL3 shRNA or control vector were inoculated into the livers of the SCID mice. Six weeks after inoculation, the mice were sacrificed, and the livers were collected.

NICD/AKT induced-CCA studies

We employed a mouse model of CCA induced by SB transposase-mediated integration of oncogenes (NICD/AKT) via HDTV injection. The HDTV injection procedure was performed as previously described,^19,50 with modifications. Briefly, 15 μg NICD, 5 μg AKT, and 1.25 μg transposase plasmids were diluted in 2 mL PBS, sterile filtered, and injected into a lateral tail vein of 6- to 8-week-old male FVB/NJ mice within 5 to 7 seconds. For TAZ, YAP, and METTL3 knockdown, specific shRNAs were released from pLKO-TAZ shRNA vectors (#TRCN0000095953 and #TRCN0000095951), pLKO-YAP shRNA vector (#TRCN0000238436), or pLKO-METTL3 shRNA vectors (#TRCN0000039111 and #TRCN0000039112); and the released shRNAs were then subcloned to pT2-GFP4 vector.⁵¹ Twenty μg pT2-shRNA vector or control vector was used along with 15 μg NICD, 5 μg AKT, and 1.25 μg transposase plasmids for HDTV injection to establish NICD/AKT-induced CCA with or without knockdown of METTL3 or TAZ. The injected mice were closely monitored and sacrificed at indicated times after tail vein injection. The liver/body weight was recorded, and the livers were harvested for protein analysis and H&E staining. For survival analysis, the mice were injected with control-shRNA (n = 10) or TAZ-shRNA (n = 9) or METTL3-shRNA (n = 9) in conjunction with NICD/AKT/transposase vectors.

Statistics

Results are presented as mean ± standard error (SE) from a minimum of 3 replicates. Difference between groups was evaluated by SPSS 13.0 statistical software with one-way analysis of variance, Student’s t-test, and log-rank test. Statistical graphs were plotted by GraphPad Prism 7.0 software and SigmaPlot statistical software. P value < .05 was considered as statistically significant.