Dysregulated YAP1/TAZ and TGF-β signaling mediate hepatocarcinogenesis in Mob1a/1b-deficient mice

We generated liver-specific Mob1a/1b-deficient mice (LMob1DKO mice) by crossing Mob1a^flox/flox;Mob1b^−/− mice to AlbCre-Tg mice (35). Southern blotting confirmed disruption of the Mob1a^flox allele in most liver cells of LMob1DKO mice, leaving the Mob1a^Δ allele (Fig. S1A). Crossing of LMob1DKO mice with Rosa26-LSL-YFP reporter mice (36) confirmed that most cholangiocytes and hepatocytes had undergone Mob1a/1b deletion, as determined by YFP-positivity and PCR analysis following laser microdissection (Fig. S1 B and C).

From postnatal day (P) 10 (P10) onward, LMob1DKO mice displayed abdominal swelling due to massive hepatomegaly (Fig. 1 A and B). The total number of liver cells in LMob1DKO mice was five times that in controls at P14 (Fig. 1C). The first LMob1DKO mice died at P14, and more than half were dead by P21 (Fig. 1D). The skin of the mutant mice was jaundiced, and their serum was yellow in color due to elevated bilirubin (Fig. 1E and Fig. S1D), suggesting that defective liver function was likely the cause of death. Histological examination revealed no obvious structural abnormalities in livers of LMob1DKO mice at P0. However, immature cholangiocyte-like cells increased in mutant liver from P7, and massive accumulations of these cells were observed from P10, especially in periductal and intraductal locations (Fig. 1F). Although some mutant cholangiocyte-like cells started to form luminal structures, most were too immature to do so. The penetrance of these phenotypes was complete in LMob1DKO mice, and immature cholangiocyte-like cells were never observed in control mice (Mob1a^flox/flox;Mob1b^−/−), AlbCre-Tg mice, or in single KO mice with systemic deletion of Mob1a (Mob1a^Δ/Δ) or Mob1b (Mob1b^−/−) (Fig. S2 A and B). Immunostaining revealed that the cholangiocyte-like cells abundant in LMob1DKO liver expressed the cholangiocyte lineage markers KRT19, KRT7 and CDH1, and that some also bore the oval cell markers A6, TACSTD2, and DLK (Fig. 1G). Oval cells, which are stem/progenitor cells of hepatocytes and cholangiocytes (37), are not observed in healthy adult liver but appear in response to liver injury (38). Immature AFP⁺ hepatocytes were also present in LMob1DKO liver, as were infiltrating inflammatory cells and fibrosis (Fig. 1H). However, because the oval cell/cholangiocyte accumulations occurred in portal and intraductal locations where oval cells usually reside (39), and did not form a tumor mass at P14, LMob1DKO livers were deemed to exhibit “oval cell and immature cholangiocyte hyperplasia.”

Most LMob1DKO mice surviving beyond P21 exhibited mild hepato-cholangiocellular hyperplasia. Although these animals had a long lifespan, they all eventually developed hepatocellular adenomas, HCCs, ICCs or cHCC-CCs (Fig. 2). About 80% (8/10) of surviving LMob1DKO mice showed these tumors by age 15 wk, and all such mutants were dead by 60 wk due to multifocal cHC-CCs (59%; 13/22), ICCs only (23%; 5/22), or HCCs only (18%; 4/22). Tumors containing transitional cells intermediate between HCC and ICC (Fig. 2C), and occurrences of HCC and ICC within the same liver (Fig. 2B), resembled cases of human cHC-CC and so were classified as cHC-CCs. About 15% (2/13) of tumor-bearing mutants >40 wk of age showed lung metastasis (Fig. 2C). Equivalent but milder histological changes, including the presence of cHC-CCs and ICCs, occurred in livers of inducible AlbCreER;Mob1a^flox/flox;Mob1b^−/− mice (Materials and Methods) treated with tamoxifen (+Tmx) at 5 wk of age (Fig. S2 C and D).

To determine whether Mob1a/1b-deficient liver cells showed increased proliferation, we administered BrdU to WT and LMob1DKO mice at P10 and analyzed BrdU incorporation by liver cells 1 h later. An increase in BrdU⁺ proliferating cells occurred among KRT19⁺ oval cells/immature cholangiocytes (four times control levels) and HNF4A⁺ (hepatocyte nuclear factor-4 alpha) hepatocytes (three times control levels) of LMob1DKO mice (Fig. 3A), but TUNEL⁺ apoptotic cells were very rare both in control and LMob1DKO liver (Fig. S3A).

We next examined a Tmx-inducible murine Mob1a/1b DKO bipotential liver progenitor cell clone (imMob1DKOLP) that arose by spontaneous immortalization of liver cells from Rosa26CreER;Mob1a^flox/flox;Mob1b^−/− mice cultured without Tmx (−Tmx) (Fig. S3B). These cells are HNF4A⁺, ALB^dim, KRT19⁺, KRT7^dim, CDH1⁺, TACSTD2^dim, DLK^dim, KIT⁺, and THY1^dim, but CD45⁻ and CD31⁻ (Fig. S3C). imMob1DKOLP cells undergo biliary cell differentiation when cultured in collagen-1 gel containing TNFα but become hepatocytes when cultured in matrigel containing Oncostatin M (Fig. S3 D and E). After Tmx administration (+Tmx; induces Mob1a/1b deletion), the proliferation and saturation density of imMob1DKOLP cells increased (Fig. 3B) with no difference in cell size (Fig. S3F), suggesting that imMob1DKOLP+Tmx cells have a defect in cell contact inhibition. Moreover, these cells formed numerous multilayered foci after prolonged culture (Fig. 3C) and rapidly generated tumors in vivo after xenografting into nude mice (Fig. S3G).

RT-PCR examination revealed that imMob1DKOLP+Tmx cells showed increased mRNA levels of the stem cell markers Tacstd2 and Dlk1 and the cholangiocellular markers Krt7, Krt19, Intb4, and Aqp1 but decreased mRNA levels of the mature hepatocyte markers Hnf4α, Alb, Ttr, and Tat (Fig. 3D). Because this cloned cell line maintained a homogeneous cell shape and uniform staining, we speculated that the increase in immature cells in LMob1DKO liver might be due to hepatocyte dedifferentiation into oval cells and immature cholangiocytes.

RT-PCR examination of P14 LMob1DKO liver cells showed enhanced expression of mesenchymal cell markers (Vim, Acta2, S100a4, and Des) and transcriptional factors (Zeb1, Snal1, and Twist2) important for the epithelial–mesenchymal transition (EMT) (Fig. 3E). Immunostaining demonstrated that spindle-shaped YFP⁺VIM⁺ and YFP⁺ACTA2⁺ EMT cells were present in P14 livers of AlbCre;Mob1a^flox/flox;Mob1b^−/−;Rosa26-LSL-YFP (LMob1DKO-YFP) mice but rare in control AlbCre;Mob1a^+/+;Mob1b^−/−;Rosa26-LSL-YFP (Control-YFP) livers (Fig. 3F). In addition, the migration of imMob1DKOLP+Tmx cells in Transwell chambers was enhanced compared with imMob1DKOLP-Tmx cells (Fig. 3G). Finally, although not statistically significant, expression levels of epithelial markers such as Cdh1 and Tjp1 tended to be decreased in LMob1DKO liver compared with controls (Fig. S3H). This lack of significance may be due to the limited area of enhanced EMT present in LMob1DKO liver. In any case, our data suggest that elevated cell proliferation, impaired contact inhibition, augmented dedifferentiation, enhanced EMT, and increased cell migration may underlie the higher numbers of oval cells/immature cholangiocytes and liver tumor development seen in LMob1DKO mice.

We next investigated the effects of Mob1a/1b loss on Hippo components using imMob1DKOLP cells cultured at high density to achieve saturation. As expected, imMob1DKOLP+Tmx cells showed reduced phosphorylation of YAP1 (Ser127) and TAZ (Ser89) and increased total YAP1 and TAZ proteins under these culture conditions (Fig. 4A). Quantitative RT-PCR analysis established that these changes were not due to altered YAP and TAZ mRNA levels (Fig. S4A). Phospho-LATS1 (T1079) was below the assay detection limit but total LATS1 was reduced in imMob1DKOLP+Tmx cells (Fig. 4 A and B). No differences in phospho-MST1, total MST1, or total SAV1 were detected. We then stimulated imMob1DKOLP ± Tmx cells with okadaic acid (OA), which activates MST1/2 (40), and performed immunoblotting after adjusting lysates to equality of LATS1 protein. Whereas OA-stimulated imMob1DKOLP-Tmx cells showed vigorous LATS1 phosphorylation, OA-stimulated imMob1DKOLP+Tmx cells did not (Fig. 4B).

imMob1DKOLP-Tmx cells cultured at high cell density showed cytoplasmic localization and inactivation of YAP1, as expected (Fig. 4C). However, YAP1 remained activated and positioned in the nucleus in high-density imMob1DKOLP+Tmx cells. Similarly, most oval cells/immature cholangiocytes and hepatocytes in LMob1DKO livers showed nuclear YAP1, unlike controls (Fig. 4D). Thus, MOB1 promotes density-induced cytoplasmic localization of YAP1, and MOB1A/1B deficiency drives nuclear localization of YAP1.

We compared control and LMob1DKO livers using microarray and found a greater than fivefold increase in several growth factors and chemokines. High mRNA levels of Ctgf, Spp1, Jag1, Tgfb2/3, and Ccl2, which are all downstream YAP1/TAZ transcriptional targets, occurred in MOB1A/1B-deficient liver (Fig. 4E). Immunoblotting confirmed up-regulation of these proteins (except CCL2) in LMob1DKO liver, as well as elevation of phospho-SMAD2 (Fig. 4F). However, activated NOTCH1 (NOTCH1IC), NOTCH2IC, and their downstream target HES1, which are all important for the conversion of hepatocytes into primitive ductular cells (41), were not altered even though JAG1 (Notch ligand) was increased (Fig. 4F). Immunostaining of Mob1a/1b-deficient livers revealed that CTGF, JAG1, and SPP1 (Fig. 4G), as well as phospho-SMAD2 and nuclear SMAD2/3 (Fig. 4H), were all enhanced in oval cells/immature cholangiocytes. TGF-β2/3, phospho-SMAD2, and nuclear SMAD2/3 were increased in mutant hepatocytes, and TGF-β3 was also elevated in inflammatory cells. Thus, heightened activation of YAP1/TAZ, CTGF, TGF-βs–SMADs, JAG1, and SPP1 may contribute to the abnormalities in LMob1DKO liver. Alternatively, the observed increase in growth factors such as CTGF, JAG1, and SPP1, especially in immature cholangiocytes, may be secondary to the increased number of oval cells in LMob1DKO liver.

To determine whether Mob1a/1b-deficient liver phenotypes depended on activation of YAP1/TAZ, TGF-βs–SMAD2/3, and/or CTGF, we generated triple knockout (TKO) mice: Mob1a/1b plus Yap1 (AlbCre;Mob1a^flox/flox;Mob1b^−/−;Yap1^flox/flox); Mob1a/1b plus Taz (AlbCre;Mob1a^flox/flox;Mob1b^−/−;Taz^−/−); Mob1a/1b plus Tgfbr2 (AlbCre;Mob1a^flox/flox;Mob1b^−/−;Tgfbr2^flox/flox); and Mob1a/1b plus Ctgf (AlbCre;Mob1a^flox/flox;Mob1b^−/−;Ctgf^flox/flox). The increases in liver weight, numbers of oval cells/immature cholangiocytes, and liver fibrosis seen in LMob1DKO mice were dramatically suppressed by additional Yap1 loss (Fig. 5 A and B). However, at P14, these triple mutants still exhibited some oval cell/cholangiocellular hyperplasia and fibrosis, indicating that Yap1 deletion cannot rescue the LMob1DKO phenotype completely. All LMob1a/1b;Yap1 TKO mutants survived for at least 20 wk without developing liver cancers, whereas >80% of LMob1DKO mice show tumors or lethality by this age (Fig. 1D). LMob1a/1b;Taz TKO (Fig. 5C) and LMob1a/1b;Tgfbr2 TKO (Fig. 5D) mice displayed partial suppression of abnormalities associated with Mob1a/1b loss. However, LMob1a/1b;Ctgf TKO (Fig. 5E) mice showed the same range of liver abnormalities as LMob1DKO mice. Thus, the liver phenotypes of LMob1DKO mice are strongly dependent on YAP1 and partly dependent on TAZ and TGF-βs, but not significantly dependent on CTGF.

Because YAP1 and TAZ are coactivators of TEAD1–4, we suspected that the proliferation occurring in the absence of MOB1A/1B was TEAD-mediated. siRNA analysis of imMob1DKOLP+Tmx cells showed that their enhanced proliferation was suppressed by treatment with siYap1/Taz or siTead1/4 (Fig. S4 B–E). Addition of TGF-β2 did not alter the proliferation of imMob1DKOLP+Tmx cells but did increase their expression of mesenchymal/cholangiocyte markers while decreasing hepatocyte markers, consistent with a previous report (42) and their enhanced YAP1 activation (Fig. S5 A–D). In LMob1a/1b;Tgfbr2 TKO mice, the additional loss of Tgfbr2 decreased the YAP1 activation induced by Mob1a/1b deficiency in vivo (Fig. S5 E and F). Consistent with this observation, treatment in vitro of imMob1DKOLP+Tmx cells with the TGF-βRI inhibitor SB431542 also decreased YAP1 activation (Fig. S5G).

Thus, considering that (i) TGF-βs production is increased by YAP1/TAZ activation (11), (ii) nuclear SMAD2/3 activity is enhanced by nuclear translocation of YAP1/TAZ (33), and (iii) YAP1 is activated by TGF-βs, we believe that the TGF-β pathway contributes to YAP1 activation via a positive feedback loop. The elevated TGF-β2 and TGF-β3 in LMob1DKO liver may promote EMT and drive hepatocyte dedifferentiation toward immature cholangiocytes, and it is these cells that invade the liver parenchyma.

Abnormalities in Hippo components occur in many types of human tumors (43). We immunostained liver samples from patients with cHC-CC, ICC, or HCC to detect YAP1 and SMAD2 and observed significant YAP1 activation (greater than grade 2, as defined in Table S1) in 70% of cHC-CCs and 60% of ICCs, but in only 23% of HCCs, 14% of nontumorous cholangiocytes (NT-Chol), and 6% of nontumorous hepatocytes (NT-Hep) (Fig. 6A). Furthermore, YAP1 activation showed a significant correlation with SMAD2 activation in these cHC-CCs and ICCs (Fig. 6B).

We next investigated whether chemical compounds or natural products able to inhibit YAP1 might have anticancer effects in the liver. To visualize Hippo pathway activity, we devised a sensitive system based on the H1299 human lung cancer cell line, in which endogenous YAP1 is highly activated. We engineered H1299 cells to express 10 copies of the CTGF TEAD-binding sequence plus the luciferase reporter gene (Fig. S6A) and screened 1,282 natural compounds and 10,240 synthetic compounds. Ivermectin (Fig. S6B) and its derivative milbemycin D (Fig. S6C) were identified as potential YAP1 inhibitors. Ivermectin is a well-tolerated antiparasitic macrocyclic lactone commonly used to treat river blindness (44), and both ivermectin and milbemycin D activate glutamate-gated chloride channels (45). Treatment of our modified H1299 cells in vitro with ivermectin or milbemycin D inhibited YAP1/TAZ activity and decreased CTGF expression in a dose-dependent manner (Fig. 7 A–C and Fig. S6 D–F). Dasatinib, a c-YES inhibitor that reduces YAP1 nuclear localization and activation, served as a positive control for these experiments. Ivermectin and milbemycin D also strongly suppressed the in vitro growth of both imMob1DKOLP+Tmx cells and an NF2-deficient human mesothelioma cell line (NCI-H290) in which YAP1/TAZ is constitutively activated (Fig. 7 D and E and Fig. S6G). Cell growth inhibition by ivermectin was similar to that observed following application of siYAP1/TAZ (Fig. 7 D and E). Importantly, ivermectin treatment of LMob1DKO mice for 5 d (P4–P8) reduced their hepatomegaly, accumulations of oval cells/immature cholangiocytes, excessive hepatic YAP1 activation (Fig. 7 F and G), and expression of YAP1/TAZ target proteins (Fig. 7H). Finally, we transfected the KKU-M213 line of human CC cells with Dox-dependent shYAP1/TAZ, xenografted these cells into nude mice, and allowed tumors to establish. Treatment of these mice for 18 d with cyclodextrin-conjugated ivermectin [ivermectin (−Dox); 10 mg/kg daily by i.p. injection] significantly inhibited tumor growth compared with control treated with DMSO (−Dox) (Fig. 7 I and J). This effect of ivermectin was comparable to that of shYAP1/TAZ treatment alone [DMSO (+Dox); Fig. 7I], and ivermectin did not further suppress the growth of shYAP1/TAZ-treated tumors [ivermectin (+Dox); Fig. 7I]. Finally, ivermectin significantly reduced the growth of tumors established by xenografting imMob1DKOLP+Tmx cells into nude mice (Fig. S6H). These results suggest that antiparasitic macrocyclic lactones may aid in the treatment of liver cancers by blocking Hippo signaling.

Base mouse strains used in this study were as follows: AlbCre-Tg (Jackson), Mob1a^flox/flox;Mob1b^−/− (28), Rosa26-LSL-YFP reporter (36), AlbCreER-Tg (54), Rosa26CreER-Tg (55), Yap1^flox/flox (Knockout Mouse Project Repository), Taz^−/− (56), Tgfbr2^flox/flox (57), and Ctgf^flox/flox (58).

Liver-specific Mob1a/1b double-homozygous mutant mice (AlbCre;Mob1a^flox/flox;Mob1b^−/− mice; designated LMob1DKO) were generated by mating AlbCre-Tg mice to Mob1a^flox/flox;Mob1b^−/− mice. Mob1a^flox/flox;Mob1b^−/−, AlbCre;Mob1a^+/+;Mob1b^−/−, AlbCre;Mob1a^flox/flox;Mob1b^+/+ and AlbCre;Mob1a^+/+;Mob1b^+/+ mice were indistinguishable in pilot experiments examining body weight, histology, and lifespan (discussed below). AlbCre-Tg mice were of the C57BL/6 background, and Mob1a^flox/flox;Mob1b^−/− mice were back-crossed to C57BL/6 for more than six generations. Mob1a^flox/flox;Mob1b^−/− mice without the AlbCre-Tg allele were usually chosen to serve as controls. The deletion frequency of the Mob1a^flox allele was confirmed by genomic Southern blotting using the probe indicated in Fig. S1A. Primers used for genotyping PCR are listed in Table S2. The frequency of Mob1A deletion in hepatocytes and cholangiocytes (Fig. S1C) was determined by PCR analysis of cells microdissected using a Leica LMD6000 instrument.

To visualize cells with Mob1a/1b deletion, LMob1DKO mice were crossed to Rosa26-LSL-YFP reporter mice. To induce Cre-mediated deletion of Mob1a/1b in liver cells of adult mice, we generated AlbCreER;Mob1a^flox/flox;Mob1b^−/− and Rosa26CreER;Mob1a^flox/flox;Mob1b^−/− mice using AlbCreER-Tg mice or Rosa26CreER-Tg mice, respectively, and treated them with tamoxifen as described below.

AlbCre;Mob1a^flox/flox;Mob1b^−/−;Yap1^flox/flox, AlbCre;Mob1a^flox/flox;Mob1b^−/−;Taz^−/−, AlbCre;Mob1a^flox/flox;Mob1b^−/−;Tgfbr2^flox/flox, and AlbCre;Mob1a^flox/flox;Mob1b^−/−;Ctgf^flox/flox, TKO mice were generated by mating AlbCreMob1a^flox/flox;Mob1b^−/− mice with Yap1^flox/flox, Taz^−/−, Tgfbr2^flox/flox, or Ctgf^flox/flox mutant mice, respectively. All protocols for animal experiments were approved by the Animal Research Committee of Kyushu University. All animal experiments adhered to the criteria outlined in ref. 59.

Genomic Southern blots were performed as described (60). Agarose gels were transferred to nylon membranes (GE Healthcare) and hybridized to the probe described in Fig. S1A.

Hepatocytes were isolated from mice on various days after birth using an in situ two-step collagenase liver perfusion method as previously described (61). Residual liver cells containing Glisson’s sheath were further exposed to 0.05% collagenase/ 0.1% Pronase for 30 min to generate a single-cell suspension. This suspension was mixed with the isolated hepatocytes to create a preparation deemed to be “total liver cells,” which was used for analysis of total cell numbers and for primary cultures.

Liver tissues were fixed in phosphate-buffered 4% (wt/vol) paraformaldehyde (PFA) and embedded in paraffin, and sections (5 μm) were mounted on slides for H&E staining or immunohistochemical (IHC) analysis. IHC analysis was performed using the antibodies (Abs) and protocol described below. Liver tissues were also stained using Elastica Masson and Sirius Red according to standard protocols.

Tissues were fixed in 4% (wt/vol) PFA or neutral-buffered 10% (vol/vol) formalin and embedded in paraffin, or fixed in 1% (wt/vol) PFA and snap-frozen in optimal cutting temperature compound and sectioned (5 µm). Fixed sections or cells fixed in 4% (wt/vol) PFA were incubated overnight at 4 °C with primary Abs. Primary Abs used for IHC and immunocytochemistry recognized: HNF4A (sc-8987; Santa Cruz), KRT19 (TROMA-III-c; Hybrydomabank), KRT7 (MAB3226; Millipore), ALB (A90-134A; Bethyl), CDH1 (E-Cadherin; 610181; BD Transduction Laboratories), TACSTD2 (TROP2; BAF1122; R&D), A6 (62), DLK1 (AM26511AF-N; Acris), AFP(688031; MP Biomedicals), KIT (55353; BD), THY1 (11-0902-85; eBio), CD45 (12-0451-83; eBio), CD31 (553371; BD), VIM (Vimentin; V2258; Sigma), ACTA2 (αSMA; ab5694; Abcam), YFP/GFP (ab6673; Abcam), YAP1 (WH0010413M1; Sigma), SMAD2 (sc-101153; Santa Cruz), SMAD2/3 (sc-8332; Santa Cruz), TGF-β2 (ab36495; Abcam), TGF-β3 (BS1363; Bioworld Technology), CTGF (sc-14939; Santa Cruz), JAG1 (Jagged1; sc-8303; Santa Cruz), or SPP1 (AF808; R&D). Secondary Abs were tagged with Alexa Fluor488 or Alexa Fluor568 (Molecular Probes). Anti-rabbit/mouse-HRP was used for DAB staining. Nuclei were visualized using DAPI (Dojindo) or Giemsa staining.

Spontaneous tumor formation was monitored in 69 AlbCre;Mob1a^flox/flox;Mob1b^−/− and 26 Mob1a^flox/flox;Mob1b^−/− mice over a 60-wk observation period. Mice were killed when they became morbid (humane endpoint). The onset of liver tumors was confirmed by macroscopic analysis, and sections of tumors were prepared for H&E staining.

To induce Cre-mediated deletion of Mob1a/1b in liver cells of adult mice, AlbCreER;Mob1a^flox/flox;Mob1b^−/− mice (5 wk of age) received i.p. injection of 0.5 mg Tmx (5 mg/mL; i.p. 0.1 mL/d for 7 d). Mice were monitored daily for the onset of tumors.

Mice at P10 were i.p. injected with BrdU (50 μg/g body weight; Roche). Mice were killed 1 h later and livers were harvested and fixed in 100% methanol for 12 h. Fixed livers were embedded in paraffin and cut into 5-μm sections. BrdU incorporation was visualized by immunostaining with anti-BrdU Ab (5292; Cell Signaling). Hepatocytes and cholangiocytes were identified by immunostaining with anti-HNF4A (sc-8987; Santa Cruz) and anti-KRT19 (TROMA-III-c; Hybrydomabank) Abs, respectively. BrdU incorporation was assessed in at least 1,000 HNF4A⁺ hepatocytes and 200 KRT19⁺ cholangiocytes per mouse.

Apoptosis of liver cells was determined by TUNEL assay using the In Situ Cell Death Detection Kit (Roche) according to the manufacturer’s instructions. Nuclei were visualized using DAPI.

To generate a Tmx-inducible murine Mob1a/1b-deficient liver progenitor cell line, liver cells were isolated from 5-wk-old Rosa26CreER;Mob1^flox/flox;Mob1b^−/− mice (without Tmx) using the collagenase perfusion technique described above. Liver cells were cultured (3 × 10⁵ per 6-cm dish) in Williams’ Medium E (Sigma) supplemented with 10% (vol/vol) FCS. The medium was changed twice per week and the culture was maintained for 3 mo until a single spontaneously transformed colony appeared. This colony was isolated, recloned using limiting dilution, and expanded to generate the imMob1DKOLP cell line. Inducible loss of Mob1a/1b in these cells was confirmed by treating them with Tmx (1µM) for at least 5 d, followed by analyses of mRNA expression by RT-PCR and protein expression by immunoblotting.

To induce hepatocyte differentiation, imMob1DKOLP cells (500 cells per 20 μL) were cultured for 2 d using the hanging drop method, followed by culture for an additional 9 d in multiwell plates that were coated with Matrigel (BD Bioscience) and filled with serum-free Williams’ Medium E containing 1 µM dexamethasone and 10 ng/mL OncostatinM (R&D). To induce cholangiocyte differentiation, imMob1DKOLP cells (500 cells per 20 μL) were precultured in hanging drops and then cultured for 14 d in 3D cultures (63) containing 0.24 mg/mL type I collagen (Cellmatrix type I-A; Nitta Gelatin), 10% (vol/vol) FCS, and 10 ng/mL TNFα (R&D).

For cell growth analysis, imMob1DKOLP cells (2.5 × 10⁴ per well in 24-well plates) were cultured for up to 9 d in Williams’ E medium supplemented with 10% (vol/vol) FCS, with or without Tmx (1 µM).

Total RNA was extracted from imMob1DKOLP cells cultured with or without Tmx (1 μM) for 6 or 10 d, or from whole livers of AlbCre;Mob1^flox/flox;Mob1b^−/− (LMob1DKO) and control mice at P14, using RNAiso (Takara) according to the manufacturer’s protocol. Total RNA (1 μg) was reverse-transcribed using the Transcriptor First Strand cDNA Synthesis Kit (Roche) after DNase (Toyobo) treatment. Sequences of specific primer sets used for RT-PCR are listed in Table S3. The mRNA level of a specific gene in the mutant was expressed as the fold increase over the corresponding control value.

Immunoblotting was carried out using a standard protocol and primary Abs recognizing: pMST1/2 (3681; Cell Signaling), MST1 (3682; Cell Signaling), MOB1A/1B (13730; Cell Signaling), pYAP1 (4911; Cell Signaling), YAP1 (4912; Cell Signaling), pTAZ (sc-17610; Santa Cruz), TAZ (4883; Cell Signaling), pLATS1 (9159; Cell Signaling), LATS1 (3477; Cell Signaling), CTGF (sc-14939; Santa Cruz), SPP1 (AF808; R&D), JAG1 (sc-8303; Santa Cruz), TGF-β2 (ab36495; Abcam), TGF-β3 (BS1363; Bioworld Technology), CCL2 (NBP2-22115S; Novus), NOTCH1 IC (ab8952; Abcam), NOTCH2 IC (ab8926; Abcam), HES1 (ab71559; Abcam), pSMAD2 (AB3849; Millipore), SMAD2/3 (sc-8332; Santa Cruz), ACTIN (A2066; Sigma), or GAPDH (sc-25778; Santa Cruz). Primary Abs were detected using HRP-conjugated secondary Abs (Cell Signaling).

Migration of imMob1DKOLP cells was assayed using Transwell chambers (5-µm pore size; Costar). imMob1DKOLP cells precultured with or without Tmx (1μM) for 6 d were seeded in the upper Transwell chamber in 0.5% BSA/Williams’ Medium E (1 × 10⁴ cells per well), and 10% (vol/vol) FCS/Williams’ Medium E was added to the bottom chamber. After 24 h, cells remaining in the upper chamber were removed and cells in the bottom chamber were fixed with 4% (wt/vol) PFA, stained with crystal violet, and counted under a light microscope. Four fields per chamber were evaluated.

siRNA targeting of Yap1, Taz, Tead1, Tead2, Tead3, Tead4, Smad2, or Smad3 expression was performed using the specific siRNA oligonucleotides listed in Table S4. Transfection of siRNA oligonucleotides (10 nM) into exponentially-growing imMob1DKOLP cells or NF2-deleted NCI-H290 human mesothelioma cells (64) (3 × 10⁴ per well in 24-well plates) was performed using Lipofectamine RNAiMAX (Invitrogen) following the manufacturer’s protocol. At 72 h posttransfection, total cell numbers were evaluated and total RNA was subjected to RT-PCR to detect inhibition of mRNA expression.

Surgically obtained human cHC-CCs (n = 30), ICCs (n = 30), and HCCs (n = 30) were acquired from the Department of Surgery of Kyushu University Beppu Hospital and affiliated hospitals. Resected cancer tissues were fixed in formalin and stained with anti-YAP1 Ab (WH0010413M1; Sigma) or anti-SMAD2 Ab (sc-101153; Santa Cruz) using a standard protocol. Resected nontumorous healthy cholangiocytes (n = 30) and hepatocytes (n = 30) were examined as controls. Grades of YAP1 and SMAD2 activation were defined as described in Table S1. The clinical sample study design was approved by the Kyushu University Institutional Review Board for Clinical Research and informed consent was obtained from each patient.

H1299 cells were engineered to express a reporter construct containing 10 copies of the TEAD-binding sequence from the CTGF gene plus the firefly luciferase gene (Fig. S6A). Reporter cells were seeded in 384-well microplates (1,000 cells per well) and cultured overnight in a CO₂ incubator. A library of natural compounds (1,282 samples) and a library of synthetic compounds (10,240 samples) were screened for the ability to inhibit Hippo signaling by dispensing each natural compound (final concentration 10 μg/mL) or synthetic compound (final concentration 25 μM) into cell cultures. After 24-h treatment with screening compounds, reporter luciferase activity and cell viability were measured using the Picagene BrilliantStar-LT Luminescence Kit (307–15373; Toyo B-Net) and Cell Counting Kit-8 (CK04; Dojindo), respectively, according to the manufacturers’ instructions. Compounds that markedly suppressed reporter activity but exerted low cell toxicity were considered to be positive hits. Positive hits included the antiparasitic macrocyclic lactones ivermectin and milbemycin D.

To determine the effects of positive hit agents on YAP1 activation in vitro, H1299 cells (3 × 10⁵ per well in six-well plates) were cultured for 24 h in RPMI1640 medium containing ivermectin (10 μM; Sigma), milbemycin D (10 μM; BioAustralis Fine Chemicals), vehicle (DMSO, negative control), or Dasatinib (positive control; Abcam). Expression of YAP1 protein and mRNA encoding the YAP1 downstream target CTGF were analyzed by immunoblotting, immunocytochemistry, and RT-PCR. To determine the effects of agents on cell growth, imMob1DKOLP ± Tmx cells or NCI-H290 cells (3 × 10⁴ per well in 24-well plates) treated with siYAP1/TAZ or siScramble were cultured for 3 d in RPMI1640 medium containing ivermectin (10 μM) or milbemycin D (10 μM).

To determine the in vivo effects of ivermectin on the liver phenotypes of LMob1DKO mice, we treated control and LMob1DKO mice (P4; n = 7 per group) with ivermectin (2 mg/kg, i.p.) daily for 5 d. Liver weight and histology were analyzed at P10.

Human cholangiocellular cancer cells (KKU-M213; 1 × 10⁵) that had been transfected with Dox-dependent shYAP1/TAZ, or MOB1-deficient murine liver progenitor cells (imMob1DKOLP cells; 2 × 10⁵), were injected s.c. into the flanks of 8- to 10-wk-old female BALB/cAJcl-nu/nu mice (CLEA Japan). After visual detection of tumors (usually by 8 d postinjection), mice were treated with 45% (wt/vol) cyclodextrin-conjugated ivermectin (10 mg/kg, i.p., daily) or cyclodextrin carrier alone and supplied with normal drinking water or water containing Dox (2 mg/mL). Tumor volumes were measured every 2 d using calipers.

Where appropriate, results are expressed as the mean ± SEM of the indicated number of mice or cultures per group. Statistical differences were determined using the Student’s t test or Welch’s t test. Survival curves were calculated using the Kaplan–Meier method and analyzed with the log-rank test. For clinical samples, statistical differences between groups were determined using the χ² test and the JMP software package (SAS Institute).