The oncogenic potential of YAP requires Y357 phosphorylation in cholangiocytes but not in hepatocytes

Abstract

Purpose

The transcriptional cofactor yes-associated protein (YAP) is activated in primary liver cancers (PLCs). YAP is canonically inhibited through serine 127 (S127) phosphorylation via the Hippo pathway but activated by phosphorylation at tyrosine 357 (Y357) by SRC family kinases. Due to liver plasticity, PLCs can originate from different cell types. However, it is unknown whether YAP activation favors tumorigenesis from specific cells of origin or promotes lineage commitment during tumorigenesis. This study investigates how YAP Y357 phosphorylation influences tumor phenotype and cell of origin.

Methods

C57BL/6 mice underwent biliary transfection with myr-AKT and YAP S127A (YAP-S) or YAP S127A/Y357F (YAP-SY). RNA sequencing and pathway analysis were performed on tumors. To determine the cell of origin, tumors were generated in lineage-tracing mice by biliary transfection or hydrodynamic tail vein injection. Two tumor-derived cell lines were isolated and characterized.

Results

YAP-S mice developed more frequent intrahepatic cholangiocarcinoma (iCCA), whereas a shift to hepatocellular carcinoma (HCC) was identified in YAP-SY mice. Transcriptome analysis revealed differential activation of pathways linked to this phenotypic switch. Lineage tracing demonstrated iCCA originated from cholangiocytes and HCC from hepatocytes in the biliary transfection, whereas iCCA originated from hepatocytes in the hydrodynamic tail vein injection. Two novel, transplantable syngeneic cell lines with mixed HCC/iCCA features were established. Tumors and cell lines commonly exhibited differential TGF-β signaling.

Conclusions

YAP Y357 phosphorylation modifies the tumor phenotype and the cell of origin in liver cancer models. Differential TGF-β signaling suggests a potential therapeutic avenue for iCCA and mixed tumors.

Clinical trial number

Not applicable.

Supplementary Information

The online version contains supplementary material available at 10.1007/s13402-025-01133-x.

Introduction

Globally, primary liver cancers (PLCs) rank as the sixth most prevalent malignancy and are the third leading cause of cancer-related mortality [1]. Hepatocellular carcinoma (HCC) and intrahepatic cholangiocarcinoma (iCCA) constitute the major subtypes, accounting for 75–85% and 10–15% of cases, respectively [1–3]. Additionally, combined HCC-CCA (cHCC-CCA) or mixed tumors are relatively rare yet aggressive subtypes that exhibit both HCC and iCCA differentiation in varying proportions within the same tumor [4]. Consequently, PLCs form a heterogeneous spectrum, with HCC and iCCA representing the opposite ends and cHCC-CCA representing an intermediate subtype with a variable molecular profile [5]. In fact, the continuous molecular spectrum of liver cancers necessitates careful evaluation in clinical settings, as subtle changes might reflect distinct radiopathologic features [5, 6]. Chronic viral hepatitis, alcohol intake, smoking, and metabolic disorders are common risk factors for all three subtypes of PLC and promote cellular transformation and lineage plasticity by causing liver injury [7–11]. The regenerative capacity of the liver is closely tied to hepatobiliary cellular plasticity, where hepatocytes and cholangiocytes can transdifferentiate under conditions of regeneration, inflammation, and other conditions [12, 13]. Considering tumorigenesis as a maladaptive regenerative response, exploring the molecular mechanisms guiding cell fate decisions is critical.

Investigating the cell of origin in PLCs has been a prominent research area in the field. It was previously thought that iCCA originates from cholangiocytes and that HCC develops from mature hepatocytes [14]. However, recent evidence has challenged the traditional view by demonstrating that mature hepatocytes can give rise not only to HCC but also to other PLC subtypes [15]. Notably, Notch signaling has been identified as a key mechanism responsible for hepatocyte-derived iCCA formation [16–23]. Yes-associated protein (YAP) is a transcriptional cofactor and is activated in a wide variety of malignancies [24–34]. Among PLCs, increased YAP activation is more frequently associated with CCA than with HCC, suggesting that CCA is the predominant YAP-activated cancer subtype within PLCs [29, 35, 36]. The Hippo pathway mediates its inhibitory canonical regulation via a serine kinase relay module. Specifically, serine 127 phosphorylation (S127) sequesters YAP in the cytoplasm by binding 14-3-3 proteins and increasing nuclear export [37, 38]. However, SRC family kinases (SFKs) also regulate YAP and lead to enhanced activation with increased nuclear localization as a result of tyrosine 357 (Y357) phosphorylation [39–42]. Importantly, tyrosine phosphorylation of YAP is required for its full activation even in the absence of Hippo repression [39]. Furthermore, SFK inhibitors can restrain the growth of CCA patient-derived xenografts [39, 43]. Despite these insights, how the activation status of YAP, which is mediated by S127 and Y357 phosphorylation, governs the cell of origin and lineage commitment during hepatobiliary oncogenesis has yet to be explored.

Herein, by adapting from a previously established mouse model [44], we investigated the role of YAP tyrosine phosphorylation in liver tumorigenesis. These results demonstrated that preventing YAP tyrosine phosphorylation shifted the distribution of PLCs toward HCC, indicating that variable YAP activation is associated with different PLC phenotypes. We further explored the changes in the tumor transcriptomic profile and signaling associated with this phenotypic switch. Moreover, we utilized a lineage tracing approach to detect the origin of the tumors in our model. Finally, we established two murine PLC cell lines with mixed features.

Materials and methods

Mice

All animal experiments were performed in accordance with a protocol approved by the Mayo Clinic Institutional Animal Care and Use Committee (A00004395). C57BL/6J male wild-type and ROSA26^mT/mG reporter mice (B6.129(Cg)-Gt(ROSA)26Sort^{m4(ACTB−tdTomato,−EGFP)Luo}/J; Cat#007676) were purchased from the Jackson Laboratory and housed under a 12-hour light-dark cycle with unrestricted access to food and water.

Biliary transfection model. The biliary transfection model was utilized for liver-specific, cholangiocyte-enriched transgene delivery, following a previously described method [44]. For surgical interventions, anesthesia was induced and maintained with 2.5-3% isoflurane. Once a deep anesthetic state was confirmed, a 1 cm midline incision was made below the xiphoid process to access the abdominal cavity. The left lobe of the liver was selectively targeted by positioning a loose ligature around the left liver duct. The common bile duct was clamped to block solution flow into the duodenum. Lipid-complexed plasmids were injected into the biliary tree via gallbladder injection. Plasmid mixes included Sleeping Beauty (SB) transposase and transposons encoding murine constitutively active AKT (myr-AKT) and either the single mutant form of human YAP (YAP S127A, referred to as YAP-S) or the double mutant form (YAP S127A/Y357F, referred to as YAP-SY), indicating that phosphorylation was prevented at serine 127th or serine 127th and tyrosine 357th of YAP, respectively. Specifically, for each mouse, the following DNA constructs were utilized: 11.25 µg of AKT-PT3EF1α, 11.25 µg of YAP2-PT3EF1α (YAP S127A or YAP S127A/Y357F) [39], and 2.5 µg of pCMV-SB transposase. The constructs were combined in a total volume of 100 µl and complexed with jetPEI (Cat#101000053; Polyplus, Sartorius) according to the manufacturer’s instructions. Following plasmid injection, the left lateral bile duct was ligated, and the common bile duct was unclamped. Gallbladder ligature and cholecystectomy were performed. The abdominal wall and skin were closed in separate layers via absorbable Vicryl 4 − 0 or 5 − 0 gut sutures. Postoperatively, the mice received intraperitoneal (i.p.) injections of IL-33 (1 µg per mouse; Cat#3626-ML-010; R&D Systems) for three consecutive days. Necropsy was performed 10 weeks after surgery, and liver tissues were harvested for further analyses.

Hydrodynamic tail vein injection (HDTVI) model. The HDTVI model was utilized for liver-specific, hepatocyte-enriched transgene delivery, following a previously described method [45]. For each mouse, 11.25 µg of AKT-PT3EF1α, 11.25 µg of YAP2-PT3EF1α (YAP S127A/Y357F), and 2.5 µg of pCMV-SB transposase diluted in 2 ml of lactate’s ringer solution were injected into a lateral tail vein within 5–7 s.

Lineage tracing model. ROSA26^mT/mG reporter mice [46] received a tail-vein injection with adeno-associated virus serotype 8 (AAV8-Tbg-Cre; Addgene viral prep #107787-AAV8; http://n2t.net/addgene:107787; RRID: Addgene_107787), a gift from James M. Wilson. This was followed by either biliary transfection or HDTVI, as described above. During the 14th to 16th weeks post-surgery, necropsy was performed, and liver tissues were collected for fluorescence imaging.

Orthotopic tumor model. A superficial incision was made beneath the xiphoid process to reach into the abdominal cavity while under deep anesthesia (2.5-3% isoflurane). SY2 cells (1 × 10⁶) in 40 µl of standard media supplemented with 40% Matrigel^® Matrix (Cat#354234; Corning) were instilled into the median lobe of the liver. A sterile cotton-tipped applicator was utilized as a support to prevent media leakage and blood loss. The internal organs were repositioned, and the abdominal wall and skin were closed in separate layers via absorbable Vicryl 4 − 0 or 5 − 0 gut sutures.

Immunohistochemistry

Liver sections from sacrificed mice were fixed in 4% paraformaldehyde (PFA; Cat# 50-980-487; Fisher Scientific) for 48 hours and embedded in paraffin. Blocks were sectioned into 3–5 µm slices. Paraformaldehyde-fixed, paraffin-embedded mouse tumors and adjacent liver sections were deparaffinized, rehydrated, and incubated with primary antibodies overnight at 4°C. The sections were stained with antibodies against CK7 (1:8000; Cat#ab181598; Abcam), CK17/19 (1:250; (D4G2) XP^®; Cat#12434S; Cell Signaling Technology), SOX9 (1:500; Cat#82630; Cell Signaling Technology), GPC3 (1:400; Cat#MA5-17083; Thermo Fisher Scientific), HNF4α (1:200; Cat# 3113; Cell Signaling Technology), or HepPar1 (1:100; clone OCH1E5; Cat#MA5-12417; Thermo Fisher Scientific). Horseradish peroxidase (HRP)-conjugated secondary antibodies (Cat# MP-7500; Vector Laboratories) and 3,3’-diaminobenzidine (DAB; Ca# GV82511-2; Agilent Dako) were used for detection of the primary antibodies. Hematoxylin staining was applied as a counterstain.

Immunofluorescence

Fresh liver tissues were collected during necropsy and frozen in Tissue-Tek^® O.C.T. compound (Cat#4583; Sakura Finetek). Sections in 5 μm thickness were prepared via a cryostat (CM1950; Leica Microsystems). The sections were allowed to equilibrate at room temperature for 1 h before being fixed with 4% PFA (Cat#50-980-487; Fisher Scientific) for 20 min. The permeabilization step was performed using 0.1% Triton^® X-100 (Cat#H5141; Promega) for 5 min. The sections were gently dried, and a hydrophobic barrier was created via a PAP pen liquid blocker (Cat#6506; Newcomer Supply). Blocking was carried out with Rodent Block M (Cat#RBM961H; Biocare Medical) for 1 h at room temperature, followed by overnight incubation at 4 °C with primary antibodies against either Keratin 17/19 (1:50; Keratin 17/19 (D4G2) XP^®; Cat#12434S; Cell Signaling Technology), CD45 (1:50; Cat#70257S; Cell Signaling Technology), or alpha-smooth muscle actin (α-SMA; 1:250; Cat#ab124964; Abcam). The next day, the sections were incubated with a fluorophore-conjugated secondary antibody (1:200; goat anti-rabbit IgG (H + L) cross-adsorbed secondary antibody, Alexa Fluor™ 633; Cat# A-21070; Thermo Fisher Scientific) for 1 h at room temperature in the dark. Nuclear staining was performed using DAPI (1:800; Cat#62248; Thermo Fisher Scientific) for 10 min. The sections were mounted with ProLong™ Gold Antifade Mountant (Cat#P36934; Thermo Fisher Scientific) and allowed to dry overnight in a slide holder before imaging. Fluorescence images were acquired via a confocal microscope (LSM 980; Zeiss).

RNA sequencing

Total RNA was isolated from the mice tumor nodules and cell lines using the miRNeasy Tissue/Cells Advanced Micro Kit with on-column DNase I treatment (Cat#217684; Qiagen) according to the manufacturer’s protocol. The RNA integrity and concentration were assessed using the Agilent TapeStation 4200 with the high-sensitivity RNA ScreenTape^® assay. Samples with an RNA Integrity Number (RIN) > 8.0 were used for sequencing. For mouse samples, macroscopically visible tumor nodules present at the experimental endpoint (n = 6 per genotype) were harvested without selection or histologic classification. Tumor identity (HCC-predominant, iCCA-predominant, or mixed) was later determined based on unsupervised clustering using principal component analysis (PCA) of whole transcriptome profiles, combined with the expression of lineage markers (HNF4A, ARG1, CPS1, AFP, and GPC3 for HCC; SOX9, KRT8, KRT19, and KRT7 for CCA) and YAP signature genes, as described in Supplementary Fig. 1A-D. RNA sequencing (RNAseq) libraries were prepared using the TruSeq mRNA Library Prep Kit v2 (Set A; Cat#RS-122-2001; Illumina). Whole-transcriptome sequencing was performed on an Illumina HiSeq 4000, generating 100 bp paired-end reads. The sequencing run was conducted at the Mayo Clinic Genome Analysis Core, Rochester, MN. The raw paired-end FASTQ reads were processed via the Mayo RNA-Seq bioinformatics pipeline MAP-RSeq (v3.1.3) [47]. Read alignment was performed via the splice-aware STAR aligner [48] against the Mus musculus mm10 reference genome. Gene and exon expression levels were quantified via the Subread package [49], which generates both raw and normalized reads per kilobase per million (RPKM) values. Differential expression analysis was performed via edgeR (Bioconductor v2.6.2) [50]. Genes with significant differential expression were identified using a false discovery rate (FDR) threshold of < 5% and were reported along with their magnitude of change (log₂ scale). Functional enrichment analysis was conducted using Ingenuity Pathway Analysis (IPA) to identify the top canonical pathways and upstream regulators among the compared groups.

Heatmaps of differentially expressed genes (DEGs) (padj < 0.05, |log2FC| ≥ 1.5) were generated via the “pheatmap” package in R, where gene expression values were row scaled to highlight expression patterns across conditions. Venn diagrams illustrating the overlap of DEGs (padj < 0.05, |log2FC| ≥ 1.5) across comparisons were generated via the “ggVennDiagram” package. The volcano plot was constructed via “ggplot2”, where significant DEGs (padj < 0.05, |log2FC| ≥ 1) were color-coded. The genes with the most significant differences on the basis of adjusted p values and the genes with the most differential expression on the basis of log2-fold changes were annotated via “ggrepel” for clarity. Canonical pathway enrichment results were visualized via bubble plots with the “ggplot2” package, where pathways were manually grouped on the basis of functional commonality. The p values are represented by bubble size, whereas the Z-scores are mapped to a gradient color scale, highlighting the most significantly enriched pathways across conditions. Heatmaps for upstream regulators were constructed via “ggplot2”, where Z-scores were mapped to a color gradient to indicate regulatory activity.

Real-time quantitative polymerase chain reaction

Total RNA was extracted from cultured cells by using the RNeasy Plus Mini Kit (Cat#74136; Qiagen) according to the manufacturer’s instructions. cDNA was synthesized from 2 µg of purified RNA using Moloney Murine Leukemia Virus Reverse Transcriptase (M-MLV RT; Cat#28025013; Thermo Fisher Scientific) and random primers (Cat#58875; Thermo Fisher Scientific). Real-time quantitative polymerase chain reaction (RT-qPCR) was conducted on a LightCycler^® 480 Instrument II system (Roche) with LightCycler 480 SYBR Green I Master (Cat#04707516001; Roche) used as the fluorescence reporter. Gene expression levels were quantified via the Δ-Δ Ct method, with normalization to ACTB as the reference gene. Technical replicates were completed for each run, and 6 biological replicates were completed for each condition except for the SB6 cells (3 biological replicates). The primer sequences are provided in Supplementary TableS1.

Establishment of cell lines

Tumor nodules from two different YAP-SY mice at 10 weeks following the biliary transfection method were dissociated using the mouse Tumor Dissociation Kit (Cat#130-096-730; Miltenyi Biotech) and gentleMACS™ Dissociator (Cat# 130-093-235; Miltenyi Biotech) following the manufacturer’s instructions. The dissociated tumor material was then filtered through a sterile 60-µm mesh and seeded into 6-well plates containing Dulbecco’s Modified Eagle’s Medium (DMEM; Cat#11995065; Gibco™, Thermo Fisher Scientific) supplemented with 10% fetal bovine serum (FBS; Cat#A5256701; Gibco™, Thermo Fisher Scientific) and Primocin^® (100 µg/ml; Cat#ant-pm-1; InvivoGen). Once the cells adhered, the medium was changed every 3–4 days. When the attached cells approached confluence, they were detached via trypsin and reseeded. The propagated cell lines were named SY1 and SY2.

Cell culture

The newly established mouse SY1 and SY2 cell lines, the previously established murine nonmalignant, immortalized cholangiocyte cell line 603B [51, 52], and the SB1 and SB6 cell lines, which are iCCA cell lines from the YAP S127A/AKT model [53], were cultured in DMEM (Cat#11995065; Gibco™, Thermo Fisher Scientific) supplemented with 10% FBS, penicillin (100 U/mL; Cat#15140122; Gibco™, Thermo Fisher Scientific), and streptomycin (100 µg/mL; Cat#15140122; Gibco™, Thermo Fisher Scientific) in a humidified 5% CO₂ incubator at 37 °C. All the cell lines were periodically tested for mycoplasma by Myco-Sniff™ (Cat#093050201; MP Biomedicals).

Statistical analysis

All the statistical analyses were performed via GraphPad Prism version 9.5 (GraphPad Software, Inc.) and R version 4.4.2 (R Foundation for Statistical Computing; URL: https://www.R-project.org/). Normality was tested via the Shapiro-Wilk and Kolmogorov-Smirnov tests. Summary statistics are presented as the means and standard errors of the means (SEMs) for normally distributed data and as medians with interquartile ranges (IQRs) for nonnormally distributed data. For analyzing normally distributed data, one-way ANOVA was used; for analyzing nonnormally distributed data, the Mann-Whitney test was used to compare two groups, and the Kruskal-Wallis test with post hoc Dunn’s test was used for multiple comparisons. The differences in the distribution of tumor phenotypes between the mouse groups were analyzed via the chi-square test. In the whole study, p values less than 0.05 were considered statistically significant.

Results

YAP Y357 phosphorylation is associated with increased cholangiocarcinoma formation

Previously, our group developed a model of iCCA initiated by YAP-S/AKT [44]. To investigate the requirement for YAP Y357 phosphorylation in liver tumor formation, we adapted this model to compare tumorigenesis in YAP-SY versus YAP-S mice. Accordingly, mice underwent surgical biliary transfection with SB transposase and transposons expressing myr-AKT and YAP S127A (YAP-S mice) or myr-AKT and YAP S127A/Y357F (YAP-SY mice), followed by left liver lobe ligation (Fig. 1A). Tumor formation was promoted by IL-33 injection on postoperative days 1–3, according to the established model [44]. To confirm successful genetic integration of the constructs, genomic DNA was isolated from tumors and adjacent liver tissues and evaluated with polymerase chain reaction (PCR) and Sanger sequencing for the presence of point mutations introduced into the YAP gene (data not shown).

Fig. 1.

Blocking YAP Y357 phosphorylation causes increased HCC and decreased iCCA formation. (A) Schematic representation of the in vivo experimental design. (B) Quantification of total tumor nodules formed in the ligated liver lobe of YAP-S and YAP-SY mice at 10 weeks. (C) Liver weight-to-body weight ratios in YAP-S and YAP-SY mice at 10 weeks. The data are expressed as the median ± IQR (n = 6). (D) Representative macroscopic liver images, H&E staining, and immunohistochemistry images of mouse liver tissues collected at 10 weeks. Tumor regions (T) and non-tumor regions (NT) are indicated. Scale bar: 100 μm. (E) Proportional distribution of tumor types and (F) quantification of tumor surface area in representative sections from YAP-S and YAP-SY mice. **p < 0.01; ****p < 0.0001; ns, not significant

At 10 weeks post-surgery, no significant differences were observed in the tumor bulk or total number of nodules between the two groups (Fig. 1B, C). Samples from tumor nodules were prepared for histological evaluation via hematoxylin and eosin (H&E) staining and immunohistochemistry with a hepatocyte marker, HepPar-1, and a cholangiocyte marker, CK7 (Fig. 1D). Histology was reviewed by a dedicated hepatobiliary pathologist, and two distinct morphologic tumor types (iCCA and HCC) were noted. To confirm the phenotype of the observed tumors, additional staining with CK17/19 and SOX9 for cholangiocytic differentiation and GPC3 and HNF4α for hepatocytic differentiation was applied (Supplementary Fig. 2). The tumor nodules morphologically consistent with HCC exhibited polygonal cells with granular eosinophilic cytoplasm and abundant fat globules. In contrast, tumors morphologically consistent with iCCA had low columnar/cuboidal cells with abundant glandular structures and dense desmoplastic stroma. A subset of tumor nodules displayed both morphologies intermingled with each other, which is consistent with cHCC-CCA (mixed) tumors [4]. Tumor nodules with all three morphologies were observed in YAP-S mice: iCCA (25.3%), mixed (10.7%), and HCC (64%). In YAP-SY mice, there was a significant increase in HCC tumor nodules, representing 86.7% of the tumors, with rare iCCA (8.8%) and mixed tumors (4.5%) (Fig. 1E). The evaluation of the total tumor area in representative sections revealed a similar pattern, with a markedly decreased area of iCCA in YAP-SY mice (Fig. 1F).

Collectively, our results indicate that preventing YAP Y357 phosphorylation and thus decreasing YAP activation does not reduce overall tumor formation; rather, the phenotype of the tumor nodules changes, and iCCA formation becomes a rare event.

YAP-S and YAP-SY tumors demonstrate distinct transcriptional profiles

Next, to elucidate the differences in cell signaling underlying the differential tumor formation observed in YAP-SY and YAP-S mice, we performed bulk RNA sequencing of tumor nodules from both groups (n = 6 per genotype). By using PCA to cluster tumors based on all expressed genes (Supplementary Fig. 1A), the expression of HCC (HNF4A, ARG1, CPS1, AFP, and GPC3) and CCA (SOX9, KRT8, KRT19, and KRT7) lineage markers (Supplementary Fig. 1B) and YAP signature genes [54, 55] (Supplementary Fig. 1C), YAP-S tumors were classified as iCCA predominant (YAP-S CCA) (n = 3) or mixed tumor predominant (YAP-S Mixed) (n = 3), whereas YAP-SY tumors were classified as HCC predominant (YAP-SY HCC) (n = 3) or mixed tumor predominant (YAP-SY Mixed) (n = 3; Supplementary Fig. 1A-C). Consistent with the introduced mutations, YAP-SY HCC predominant tumors presented lower expression of YAP signature genes, whereas YAP-S CCA predominant tumors showed an increased YAP target gene expression (adjusted p = 1.6 × 10⁻¹⁰; ANOVA; Supplementary Fig. 1D). Interestingly, irrespective of the mutation type, mixed tumors clustered together, representing an intermediate level of YAP activation.

Differential gene expression analysis identified YAP-SY HCC and YAP-S CCA as the most distinct groups, with 273 upregulated and 1826 downregulated genes in YAP-SY HCC tumors compared to YAP-S CCA tumors, based on an FDR threshold of 5% and an absolute log₂ fold-change ≥ 1.5 (Fig. 2A, B). Between YAP-SY HCC vs. YAP-SY Mixed, we identified 17 upregulated and 296 downregulated genes (Supplementary Fig. 3A). Between YAP-S Mixed vs. YAP-S CCA tumors, we identified 95 upregulated and 417 downregulated genes (Supplementary Fig. 3B). The most similar tumors were mixed tumors arising from different genotypes. The YAP-SY Mixed and YAP-S Mixed tumors presented only 21 DEGs (Supplementary Fig. 3C). Consequently, the preponderance of downregulated genes in tumors with loss of YAP tyrosine phosphorylation is consistent with the loss of the YAP transcriptional coactivator program.

Fig. 2.

Transcriptome analysis of tumor nodules from YAP-S and YAP-SY mice. (A) Heatmap depicting differentially expressed genes (DEGs) between YAP-S CCA (n = 3) and YAP-SY HCC (n = 3). Tumors were classified as iCCA-predominant, HCC-predominant, or mixed based on bulk transcriptome principal component analysis (PCA), followed by sample cluster identification by expression of lineage-specific markers (HNF4A, ARG1, CPS1, AFP, and GPC3 for HCC; SOX9 and KRT7/8/19 for CCA) and YAP target gene signatures. (B) Volcano plot of DEGs in YAP-SY HCC vs. YAP-S CCA tumors. Genes are color-coded by significance (adjusted p < 0.05, |log₂FC| ≥ 1). The top 10 most statistically significant and top 10 highest absolute log₂ fold-change DEGs were annotated for both upregulated and downregulated genes in YAP-SY HCC versus YAP-S CCA mice. (C) Venn diagram illustrating the number of shared and unique DEGs in the same direction across three pairwise comparisons: YAP-SY HCC vs. YAP-S CCA, YAP-S Mixed vs. YAP-S CCA, and YAP-SY Mixed vs. YAP-SY HCC. (D) Canonical pathway enrichment was performed using Ingenuity Pathway Analysis (IPA) to identify functionally relevant signaling networks associated with the DEGs from each comparison. The bubble plot illustrates the top significantly enriched canonical pathways. Each bubble represents one pathway: bubble size corresponds to the statistical significance (− log₁₀ p-value), and bubble color indicates predicted activation status based on IPA Z-score (red = activated, purple = inhibited, gray = significant but no directionality predicted). Pathways were manually grouped into functional categories. Note that, since only 21 DEGs were detected between YAP-S Mixed and YAP-SY Mixed tumors, indicating high transcriptional similarity, we did not perform IPA for this comparison. Overall, the data reveals that YAP-SY HCC tumors exhibit broad suppression of inflammatory and matrix-related pathways compared to YAP-S CCA tumors, while mixed tumors consistently show intermediate activation states. (E) Upstream regulator analysis was performed using IPA to identify key transcription factors, cytokines, and signaling molecules. The heatmap displays the top significantly enriched upstream regulators. Color intensity reflects the IPA-predicted activation state (red = activated, purple = inhibited, gray = significant but no directionality predicted). The directionality of these regulators’ activity was suppressed in YAP-SY HCC, elevated in YAP-S CCA, and intermediate in mixed tumors, highlighting a potential gradient along the PLC spectrum

Furthermore, the unique and common DEGs across the different conditions were visualized (Fig. 2C). We identified a set of 396 DEGs associated with the iCCA phenotype that were common to YAP-S CCA vs. YAP-SY HCC and YAP-S CCA vs. YAP-S Mixed. The HCC phenotype was associated with a set of 291 genes common to YAP-SY HCC vs. YAP-S CCA and YAP-SY HCC vs. YAP-SY Mixed. Unique gene sets were identified under each condition, with 1497 genes specific to YAP-SY HCC vs. YAP-S CCA, 112 genes in YAP-S Mixed vs. YAP-S CCA, and 18 genes in YAP-SY Mixed vs. YAP-SY HCC.

To gain more insight into biological relevance, pathway analysis was performed on the DEGs of each comparison (Fig. 2D). The majority of the top canonical pathways showed a consistent pattern of inhibition in HCC tumors from YAP-SY mice compared with iCCA tumors from YAP-S mice. Similarly, the mixed tumors were noted to have an intermediate phenotype, with activation of the majority of the top canonical pathways compared with HCC and inhibition compared with iCCA. Notably, the primary functions of the differentially activated pathways were related to the immune response and inflammation, cancer-related mechanisms, extracellular matrix remodeling, and cellular signaling and adhesion. Among these pathways, a pathway associated with inflammatory cytokine signaling and immune activation (“Pathogen Induced Cytokine Storm Signaling”) emerged as the most significantly altered. This process is driven primarily by key cytokines and immune mediators, such as TNF, CXCL10, CCL2, IL1RN, TGFB1, and STAT1, which are known to orchestrate proinflammatory and regulatory immune responses.

Corroborating with enriched pathways, upstream regulator analysis was performed on YAP-SY HCC vs. YAP-S CCA. Key molecules influencing tumorigenesis were identified (Fig. 2E), with cytokines such as TGFB1, TNF, IL1B, IL4, and IFNG emerging as the top upstream regulators, suggesting differences in immune modulation. Additionally, the presence of “IMMUNOGLOBULIN (complex)” among the top significant regulators suggests the involvement of humoral immune responses. The same set of regulators was also significantly changed in the other comparisons, and the directionality was consistent across comparisons and tumor types: a trend toward inhibition in HCC tumors compared to iCCA tumors with intermediate activation in mixed tumors (Fig. 2E).

Taken together, the transcriptomic analysis results align with the spectrum model of PLCs: while iCCA and HCC represent the two ends with the most divergent gene expression, mixed tumors from both genotypes have the most similar RNA expression. The common DEGs identified across comparisons may reflect core cell signaling pathways associated with PLC phenotypes.

iCCA originates from cholangiocytes, and HCC originates from hepatocytes in YAP-S and YAP-SY mice

To determine the cell of origin of tumors in the biliary transfection model, a hepatocyte-specific fate-tracing model was employed, as previously described [46, 56]. This model utilized Rosa26^mT/mG double-fluorescent Cre-reporter mice, which possess a loxP-flanked stop codon preventing mGFP expression while allowing constitutive mTomato expression at the ubiquitously expressed Rosa26 locus. To induce hepatocyte-specific recombination, the mice were administered adeno-associated virus 8, encoding Cre recombinase under the control of the thyroxine-binding globulin (Tbg) promoter (AAV8-Tbg-Cre), via tail vein injection (Fig. 3A). This resulted in mGFP expression in hepatocytes and therefore enabled us to distinguish them from all other nonparenchymal cell types in the liver with high efficiency (Supplementary Fig. 4) [46, 57, 58].

Fig. 3.

Cellular origins of iCCA and HCC in YAP-S and YAP-SY mice in the biliary transfection model. (A) Schematic representation of the hepatocyte-specific fate-tracing model employed to determine the cell of origin. (B) H&E, native fluorescence (mTomato and mGFP), and immunofluorescence staining of DAPI, CK17/19, CD45, and alpha-smooth muscle actin (α-SMA) in an iCCA and HCC nodule. Tumor regions (T) and non-tumor regions (NT) are indicated. Scale bar: 100 μm

Two weeks after AAV8-Tbg-Cre administration, the mice underwent biliary transfection with myr-AKT and YAP-S or myr-AKT and YAP-SY constructs, as illustrated in Fig. 1A. Liver sections collected at necropsy were cryosectioned for visualization of native fluorescent protein expression. Adjacent sections were fixed and stained with markers of the cholangiocyte lineage (CK17/19), immune cells (CD45), or fibroblasts (α-SMA) and visualized via confocal microscopy. Formed nodules were carefully examined, and it was consistently demonstrated that iCCA nodules originated from cholangiocytes (mTomato- and CK-positive), and HCC nodules arose from hepatocytes (GFP-positive, with infiltration of mTomato-positive cells). The mTomato-positive infiltrating cells in the HCC nodule were found to represent immune cells (CD45+) and fibroblasts (α-SMA+) but were negative for cholangiocyte markers (CK17/19; Fig. 3B). In contrast, when the myr-AKT and YAP-SY constructs were delivered at the same concentration via HDTVI, rather than biliary transfection, iCCA formation was found to originate from hepatocytes (Supplementary Fig. 5), consistent with prior reports [17, 18].

These findings suggest that cholangiocytes and hepatocytes are the dominant cells of origin for iCCA and HCC development, respectively, in the biliary transfection model initiated by YAP/AKT. This observation implies that the change in tumor phenotypes was driven by the relative oncogenicity of YAP mutants in different cells of origin rather than an effect of lineage commitment during tumorigenesis mediated by cellular plasticity.

SY cell lines from YAP-SY mice exhibit mixed features of iCCA and HCC

The SY1 and SY2 cell lines were established from tumor nodules harvested from two distinct YAP-SY mice during necropsy at 10 weeks post-biliary transfection (Fig. 1A). PCR and Sanger sequencing of genomic DNA confirmed that both cell lines retained the introduced YAP point mutations (data not shown). To characterize the phenotypic profile of the cell lines, we assessed the expression of HCC and CCA markers using normal murine cholangiocyte cells (603B) as a reference [51, 52]. To compare the expression levels of markers in SY1 and SY2, two other cell lines with iCCA features previously established by our group derived from iCCA nodules in YAP-S mice, namely, SB1 and SB6, were used [53].

We evaluated the expression of HCC differentiation markers (AFP, GPC3, ARG1, and HNF4A) by RT-qPCR. Surprisingly, the majority of HCC lineage genes were not differentially expressed, although SY1 cells were found to have significantly higher expression of GPC3 than SB1 and SB6 cells did (Fig. 4A). SY2 cells unexpectedly expressed significantly less AFP compared to SB1 cells (Fig. 4A). Regarding CCA marker genes (KRT7, KRT8, KRT19, and SOX9) (Fig. 4B), the cancer cell lines (SB1, SB6, SY1, and SY2) all had comparable levels of KRT7, KRT8, and KRT19 expression. SOX9 was most highly expressed in SB6. Overall, both SY1 and SY2 cells demonstrated CCA differentiation, while with strong GPC3 expression, SY1 cells were more closely associated with cHCC-CCA. We next examined the expression of YAP target genes across the groups (Fig. 4C). In contrast to our expectations, SY cell lines (in particular SY2) expressed similar or higher levels of several YAP target genes than SB cells did. This is opposite to the YAP inhibition observed in SY primary tumors, which might suggest that the process of cell line derivation favored the selection of cells with active YAP signaling under in vitro conditions.

Fig. 4.

Phenotypic and functional characteristics of SY1 and SY2 cell lines. (A-C) RT-qPCR analysis of (A) HCC markers (AFP, GPC3, ARG1, HNF4A), (B) CCA markers (KRT7, KRT8, KRT19, SOX9), and (C) canonical YAP target genes (CTGF, CYR61, NUAK2, ANKRD1, MCL1) across four murine cell lines: SY1, SY2 (derived from YAP-SY tumors), SB1, and SB6. Expression was normalized to ACTB. For all genes except AFP and GPC3, results were further normalized to the 603B normal murine cholangiocyte cell line. AFP and GPC3 were excluded from this step because their expression was undetectable in 603B cells. The data are expressed as the means ± SEMs (n = 6 for SB1, SY1, SY2, and 603B; n = 3 for SB6). *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001; ns, not significant. (D) Representative images of H&E staining and immunohistochemistry of mouse liver tissues (10x magnification) collected 10 weeks post-SY2 orthotopic implantation. Tumor regions (T) and non-tumor regions (NT) are indicated. Scale bar: 100 μm. (E) Heatmap depicting DEGs between SY1 and SB1 cells (n = 3 biological replicates per cell line). (F) Venn diagram showing the number of shared and unique DEGs between the SY1 vs. SB1 cell lines and the YAP-SY HCC vs. YAP-S CCA tumor masses. The 222 overlapping DEGs represent genes that are significantly differentially expressed in both comparisons and in the same direction. (G) Upstream regulator analysis was performed using Ingenuity Pathway Analysis (IPA) on the 222 overlapping DEGs. The heatmap displays predicted activation or inhibition states of top regulators (red = activated, purple = inhibited). These findings support consistent TGF-β signaling suppression in both SY1 cells and YAP-SY HCC tumors and relative activation in SB cells and YAP-S CCA tumors

To further assess the functional outcome of the SY2 cell line, an orthotopic tumor model was established by implanting SY2 cells into the livers of C57BL/6 mice. Liver sections were collected at 16 weeks post-surgery and analyzed histologically using CK7 and HepPar1 staining. Consistent with the expression profile, most tumor nodules exhibited iCCA characteristics; however, foci of HepPar1+/CK7- tumors were observed, indicating some mixed cHCC-CCA phenotype tumors (Fig. 4D). The phenotype of the mixed tumors was further characterized using additional CK17/19 and SOX9 markers for cholangiocytic differentiation and GPC3 and HNF4α markers for hepatocytic differentiation. Regional expression of all markers was observed in an intermingled pattern (Supplementary Fig. 6).

SY1 cells, which had a more mixed HCC-CCA phenotype than SY2 cells, were compared to SB1 cells by RNAseq to evaluate their differential transcriptional programs. We identified 2206 DEGs, of which 1277 were upregulated and 929 were downregulated in the SY1 cell line when compared to the SB1 cell line (Fig. 4E). To determine whether the cell lines were similar to primary tumors with the same genotypes, a Venn diagram was plotted showing the common DEGs in the same direction in both comparisons (Fig. 4F). The total number of DEGs was found to be similar (2099 and 2206), and 222 DEGs were found to be either commonly upregulated or downregulated in each comparison. Pathway analysis of the 222 shared DEGs was performed through IPA, and molecules related to TGF-β signaling (TGFB1, SMAD3, TGFB2, TGF BETA (family), TGFB3, and STAT3) were predominant and consistent with TGF-beta inhibition in SY1 cells compared to SB1 cells (Fig. 4G). Importantly, since TGF-β signaling components are among the most significant upstream regulators in both iCCA primary tumors and cell lines, they may serve as potential targets.

Discussion

YAP co-transcriptional activity is inhibited by canonical Hippo pathway serine phosphorylation [59, 60] and activated by SFK tyrosine phosphorylation [39–41]. However, the influence of activating YAP Y357 phosphorylation on the development of distinct subtypes of PLCs remains unexplored. Our results suggest that YAP tyrosine phosphorylation favors increased YAP co-transcriptional activity, skewing the distribution of PLCs toward iCCA on the spectrum. This observation might potentially be explained by the degree of YAP activity. In fact, accumulating evidence has supported the idea of “graded” YAP activation and its influence on liver cancer phenotypes. For example, increasing YAP activity was reported to disrupt hepatocyte identity, induce SOX9 + ductal features, and drive aggressive HCC in mice [61]. Similarly, ectopic YAP activation in differentiated hepatocytes robustly induced a ductal/progenitor-like cell fate [62]. While Fitaman et al. linked increased YAP activity to the promotion of dedifferentiated HCC [61], our results may indicate that Y357 phosphorylation may push this plasticity further toward the biliary fate. In support of our hypothesis, Van Haele et al. demonstrated that the nuclear localization of YAP and TAZ increased progressively from HCC to cHCC-CCA and CCA, which suggests that YAP/TAZ activity is linked to biliary lineage commitment [29]. Additionally, major disruption of the Hippo pathway in a Mob1a/1b knockout mouse model resulted in sustained YAP/TAZ activation, which led predominantly to iCCA and cHCC-CCA tumors [63]. They also evaluated YAP activation in human liver cancer samples and reported significant YAP activation in cHCC-CCA (70%) and iCCA (60%) compared with HCC (23%) [63]. Therefore, our results, in the context of prior literature, suggest a model in which intermediate levels of YAP activation support HCC development, whereas higher levels of YAP activation are required for iCCA tumorigenesis.

PLCs represent a spectrum between HCC and iCCA, with various mixed and intermediate phenotypes in between, highlighting both intra- and inter-tumor heterogeneity [4]. This inherent heterogeneity poses an active challenge for precise diagnosis, stratification for treatment, and prediction of patient outcomes, which often contribute to therapeutic resistance [64, 65]. The current study explored the influence of YAP Y357 phosphorylation in this heterogeneous model. However, additional levels of YAP regulation may impact the PLC phenotype. In addition to phosphorylation, other post-translational modifications of YAP and the crosstalk of the Hippo pathway with other major signaling pathways play important roles in tumorigenesis. One particular example is the Wnt/β-catenin pathway, which is known for its critical role in liver development and HCC carcinogenesis and has been reported to have a significant interaction with YAP signaling [66]. Indeed, β-catenin can physically interact with YAP and become essential for its full oncogenic activity in iCCA, which thus suggests functional codependence in cholangiocarcinogenesis [67]. Moreover, many β-catenin-driven cancers also show a dependency for YAP activation [68]. Therefore, deciphering the intricate network of YAP regulation and a holistic approach to its activation potential is crucial for framing PLC heterogeneity and for developing tailored and personalized therapeutic strategies.

The concept of the cell of origin in PLCs has been a prominent topic in the field, as emerging evidence suggests that diverse cell types can contribute to tumor formation. HPCs, cholangiocytes, and even mature hepatocytes that undergo transdifferentiation may contribute to the development of iCCA [17]. In our biliary transfection model, we identified cholangiocytes as the predominant cell of origin for iCCA. Interestingly, in the HDTVI model using YAP-SY and the SB transposon system, we observed that iCCA arose from hepatocytes, consistent with the findings of Wang et al. [18]. Focusing on hepatocyte-derived iCCA, previous reports have identified Notch signaling as a key driver of its development [16, 18–21]. Together, our results and prior literature support the notion that the cell of origin in iCCA in experimental models depends on the plasmid delivery method [69]. One study suggested that the cell of origin in iCCA development may be influenced by the type of risk factor; specifically, during chronic liver injury, hepatocytes are more likely to undergo transdifferentiation into cholangiocytes [18]. Consequently, the HDTVI model, which enables widespread liver transfection predominantly targeting hepatocytes [70], may better mimic iCCA development in the context of chronic liver disease, whereas the biliary transfection model may more closely resemble idiopathic iCCA. Importantly, although the biliary transfection method is designed to favor cholangiocyte transfection [44], the anatomical continuity between bile canaliculi and periportal hepatocytes, together with increased hydrostatic pressure during biliary transfection, might explain the tumors arising from hepatocyte transfection that predominated in the YAP-SY mice. Therefore, to the best of our knowledge, we are the first to report the cell of origin in YAP/AKT-driven biliary transfection-induced PLC models. Significantly, the differential distribution of PLC subtypes in YAP-S and YAP-SY mice suggests that, compared to hepatocytes, cholangiocytes rely more on YAP Y357 phosphorylation to give rise to efficient tumor formation.

Establishing a reliable and reproducible model in PLC research, particularly for cHCC-CCA, is challenging because of the extreme heterogeneity and rarity of the disease. To the best of our knowledge, there are only a few cHCC-CCA cell lines: KMCH-1 and KMCH-2 from human tumor specimens and CC-62 and LCSC from chemically induced rat models [71–74]. In this study, we established two murine PLC cell lines with mixed features. However, there are important caveats to consider. Both primary tumor cell lines established in this study strongly expressed CCA differentiation markers, and SY2 cells did not exhibit increased levels of any of the HCC markers tested in our study. A possible explanation might be related to decreased differentiation markers in cell lines derived from original tumors. Additionally, YAP target gene expression was not lower in SY cell lines than in SB cell lines, in contrast to our expectations. YAP plays a crucial role in organ growth and cell proliferation [75, 76], and it is possible that our in vitro culture conditions selectively favored rapidly dividing tumor cells with heightened YAP activation. Similar to our observations, KMCH-1 exhibited features of cholangiocarcinoma in vitro and in vivo but not of HCC, possibly due to the favorable selection of cells that differentiated toward iCCA [74]. Nevertheless, we confirmed that SY2 cells are able to form mixed tumors when implanted in vivo and that both original tumors and derived cells retain the expected mutations. Interestingly, the expression of TGF-β signaling members was also found to be significantly downregulated in SY1 cells as compared to SB1 cells, similar to the trend observed in YAP-SY HCC vs. YAP-S CCA tumors.

The role of TGF-β signaling in liver cancers remains controversial. While it is well established that TGF-β contributes to liver fibrogenesis and creates a pro-tumorigenic microenvironment, it also performs tumor-suppressive functions by inhibiting cell proliferation [77]. Previous whole-exome sequencing studies of human CCA have identified TGF-β signaling as one of the most frequently mutated pathways [78, 79]. In contrast, genetic alterations in individual TGF-β signaling genes are relatively rare in HCC [79, 80]. Consistent with these findings, our analysis revealed that several key upstream regulators associated with the TGF-β pathway (TGFB1, STAT3, STAT1, SMAD3, RELA, and TGF BETA (family)) were significantly upregulated in iCCA compared to HCC. From the translational perspective of our study, further research on the use of TGF-β inhibitors in PLCs, especially for iCCA, is warranted. Clinical trials of galunisertib, an oral small molecule inhibitor of the TGF-β receptor I kinase, have been performed in advanced HCC as monotherapy after progression or combined with sorafenib in untreated patients [81–83]. These studies showed a subset of patients had TGF-β1 and AFP response to treatment, which was associated with increased overall survival [82, 83]. In support of the findings of others [84, 85], consequently, we speculate on the potential benefit of testing TGF-β inhibitors in iCCA and cHCC-CCA patients on the basis of our findings. Indeed, clinical trials have investigated TGF-β pathway inhibition in biliary tract cancers, including CCA [86, 87]. Early phase studies of bintrafusp alfa, a bifunctional fusion protein targeting both TGF-β and PD-L1, showed clinical activity in pretreated patients with biliary tract cancer [86], though it did not demonstrate superiority over the standard of care in the first-line setting [88]. With a similar approach, another bifunctional agent targeting TGF-β and PD-L1, SHR-1701, was combined with famitinib in a phase II trial in advanced biliary tract cancers, with an objective response rate of 28% [87]. Importantly, this study developed an “immune/metabolism” scoring system as a biomarker for response prediction, rendering more selective and personalized treatment strategies [87]. A multi-faceted approach, potentially incorporating baseline TGF-β levels and pathway activation, in addition to an immune/metabolism scoring system, might represent a clinical stratification strategy to identify patients most likely to benefit from TGF-β inhibitors. Supporting this, Robbrecht et al. demonstrated that SAR439459, a second-generation anti-TGF-β inhibitor, could remodel an immune-cold microenvironment, enhancing responses to immunotherapy [89]. Their study stratified patients based on baseline TGF-β levels and transcriptomic profiling of tumor biopsies. Therefore, biomarker-guided strategies are critical to ensure maximum benefit from TGF-β inhibitors, especially when used in combination with other treatment modalities, including immunotherapies. Patient subsets with more active pro-tumorigenic TGF-β signaling are most likely to benefit, enabling more personalized management. Together with our findings, these studies are encouraging for the identification and refinement of robust biomarkers to utilize TGF-β inhibition more effectively, particularly in iCCA and cHCC-CCA.

However, the tumor microenvironment in iCCA is governed not only by TGF-β but also by other inflammatory pathways and cellular components. This interplay of inflammatory pathways and stromal interactions leads to the defining histologic characteristics of iCCA: a highly desmoplastic structure enriched with various infiltrating immune and stromal cell types, pro-inflammatory cytokines, and increased secretion of extracellular matrix proteins [90]. Our results indeed highlight the pro-inflammatory milieu of the iCCA tumor microenvironment, revealing a significant increase in cytokines such as TNF, IL1B, IL4, and IL6, which are known to be abundantly secreted by tumor-associated macrophages (TAMs) and cancer-associated fibroblasts (CAFs) [91]. Consistently, pathways related to inflammation, extracellular matrix organization, tissue remodeling, and the tumor microenvironment were significantly enriched in CCA.

There are notable differences in the tumor immune microenvironment (TIME) across the PLC spectrum. For example, the immune dysfunction observed in HCC is exacerbated by the underlying chronic inflammatory disease etiology that characterizes most HCC cases, as well as by liver-resident tolerogenic cells, especially Kupffer cells [92]. In contrast, dense fibroinflammatory stroma with abundant CAFs and extracellular matrix proteins may physically impede infiltration of immune cells in CCA [93]. Therefore, distinct immunosuppressive mechanisms are highlighted. In HCC, metabolic competition and cytokine-mediated suppression dominate, whereas in CCA, stromal exclusion and PD-L1 expression are the primary mechanisms [92, 93]. However, both tumor types use multiple levels and overlapping suppressive mechanisms with a significant heterogeneity that necessitates comprehensive and comparative further studies. Together with these insights, the interplay between oncogenic signals and the TIME has been increasingly recognized as a determinant of tumor phenotype and evolution. TIME composition may be dictated by tumor histology but also shapes tumor phenotype, lineage commitment, and differentiation [94]. These findings reinforce the plasticity of the liver and the dynamic interaction between tumor cells and the microenvironment. Moreover, it has been shown that YAP activity can modulate cytokine expression and immune cell recruitment in other contexts [95, 96]. Specifically, YAP activation can be enhanced in macrophages by pro-inflammatory cytokines such as TNF and IL-1β, which in turn drive inflammatory responses and create a self-reinforcing loop [96]. Therefore, crosstalk between YAP signaling and TIME is suggested. In our study, a key question is whether YAP phosphorylation and activation directly drive immune-related transcriptional differences that shape the TIME, or if cell of origin and lineage commitment are selected by early changes in the TIME. While our data is consistent with the hypothesis that YAP activation status may alter both intrinsic tumor programs and the composition of the TIME, further comparative studies investigating the formation of PLCs and TIME differences at various stages are warranted.

While our findings offer mechanistic insights into the role of YAP phosphorylation in liver tumorigenesis, several limitations of the biliary transfection model should be acknowledged when interpreting its relevance to human PLCs. Human PLCs frequently arise in the context of chronic liver diseases such as viral hepatitis, metabolic dysfunction-associated steatotic liver disease, cirrhosis, or primary sclerosing cholangitis, which are not recapitulated in our model. However, we aimed at mimicking the increased inflammatory stress by consecutive IL-33 injections. Moreover, since our model was designed to investigate YAP-driven tumorigenesis, it utilizes defined genetic constructs and is expected to have a low tumor mutational burden, in contrast to the genetic heterogeneity and complex mutational landscapes observed in human tumors. Another caveat to consider when interpreting the findings is that the experiments were conducted exclusively in male C57BL/6 mice. While this approach helps ensure consistency and minimizes variability, both mouse strain and sex are known to influence tumor development and phenotype in oncogene-driven models. For example, Xue et al. [97] demonstrated that the same oncogenic construct led to purely cholangiocytic tumors in C57BL/6 mice, whereas mixed hepatocytic and cholangiocytic tumors were observed in FVB mice. Therefore, the experimental design presented here may not fully capture the range of biological responses that could occur in other genetic backgrounds or in female mice. Finally, as with all murine studies, species-specific differences in liver architecture, immune responses, and YAP signaling must be considered. To reduce this, we employed human AKT and YAP constructs to enhance translational relevance. Another approach to strengthen the translational significance of our findings would be direct validation of increased YAP Y357 phosphorylation in human iCCA clinical samples. For instance, Sugihara et al. demonstrated increased nuclear localization of tyrosine-phosphorylated YAP in patient-derived xenograft (PDX) tumors [39]; nonetheless, comprehensive and comparative assessment in cohorts of liver cancer patients with appropriate normal controls is warranted in future studies. In addition, future studies using a time-course approach could further elucidate the temporal progression of tumor development, liver architectural remodeling, and differential pathway activation across distinct genetic contexts. Such analyses may offer dynamic insight into mechanistic changes and represent a valuable direction for future research.

In summary, our study offers several novel contributions to our understanding of PLC pathogenesis and heterogeneity. First, we investigated the role of YAP Y357 phosphorylation as a driver of the PLC phenotype for the first time in a murine AKT/YAP model. Notably, upstream regulatory pathways involved in inflammation and extracellular matrix-associated signaling were identified as the most differentially activated between YAP-S and YAP-SY mice, highlighting distinct mechanistic drivers. Using a lineage-tracing mouse, we further demonstrated that, in contrast to HDTVI, the biliary transfection modelled the development of HCC from hepatocytes and iCCA from cholangiocytes, providing critical insights into the cellular origins of these malignancies. Additionally, we successfully established two distinct cell lines: SY1, which exhibits characteristic mixed tumor phenotypic features, and SY2, which, while predominantly displaying iCCA-like features, retains the functional capacity to give rise to mixed tumors. Finally, our findings present clear therapeutic implications. By comparing the transcriptomic data of SB1 vs. SY1 cell lines and tumor masses, we found that TGF-β signaling molecules were commonly enriched both in the original tumor masses and in the cell lines, potentially suggesting a critical mechanism underlying a therapeutic avenue.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

α-SMA

Alpha-smooth muscle actin

AAV8

Adeno-associated virus serotype 8

BSA

Bovine serum albumin

CAFs

Cancer-associated fibroblasts

cHCC-CCA

Combined HCC-CCA

DAB

3,3’-diaminobenzidine

DEGs

Differentially expressed genes

DMEM

Dulbecco’s Modified Eagle’s Medium

FBS

Fetal bovine serum

FDR

False discovery rate

H&E

Hematoxylin and eosin

HCC

Hepatocellular carcinoma

HDACs

Histone deacetylases

HDTVI

Hydrodynamic tail vein injection

HPCs

Hepatic progenitor cells

HRP

Horseradish peroxidase

iCCA

Intrahepatic cholangiocarcinoma

i.p.

Intraperitoneal

IPA

Ingenuity Pathway Analysis

IQR

Interquartile range

M-MLV

Moloney Murine Leukemia Virus Reverse Transcriptase

OS

Overall survival

PCA

Principal component analysis

PCR

Polymerase chain reaction

PBS

Phosphate-buffered saline

PDX

Patient-derived xenograft

PFA

Paraformaldehyde

PLCs

Primary liver cancers

RNAseq

RNA sequencing

RPKM

Reads per kilobase per million

RT-qPCR

Real-time quantitative polymerase chain reaction

SB

Sleeping Beauty

SEM

Standard error of the mean

SFKs

SRC-family kinases

TAMs

Tumor-associated macrophages

Tbg

Thyroxine-binding globulin

TIME

Tumor immune microenvironment

YAP

Yes-associated protein

Author contributions

Conceptualization: G.J.G., R.L.S., and C.B.C.; Methodology: C.B.C.; Validation: M.D.A., J.T.M., N.W.W., and C.B.C.; Formal analysis: M.D.A., C.E.M., and C.B.C.; Investigation: M.D.A., J.T.M., R.D.W., J.A.Y., E.H.B., N.W.W., D.R.O., R.P.G., and C.B.C.; Resources: J.T.M., N.W.W., G.J.G., R.L.S., and C.B.C.; Data curation: M.D.A., C.E.M., and C.B.C.; Writing – original draft: M.D.A. and C.B.C.; Writing - Review & Editing: M.D.A., G.J.G., R.L.S., and C.B.C.; Visualization: M.D.A., C.E.M., and C.B.C.; Supervision: G.J.G., R.L.S., and C.B.C.; Project administration: C.B.C.; Funding acquisition: G.J.G. and C.B.C.

Funding

This work was supported by the Fifth District Eagles Cancer Telethon (5th District Eagles Cancer Telethon) the Cholangiocarcinoma Foundation Fellowship, and NIH grant P30DK084567 (Center for Cell Signaling in Gastroenterology, C-Sig) (all awarded to C.B.C.); and by NIH grant R01 DK124182 (awarded to G.J.G.).

Data availability

The RNA sequencing data supporting the findings of this study have been deposited in the Gene Expression Omnibus (GEO) with the accession number GSE297845. This data will be publicly available upon publication of this article.

Declarations

Ethics approval and consent to participate

All animal experiments were performed in accordance with a protocol approved by the Mayo Clinic Institutional Animal Care and Use Committee (A00004395).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.