ABSTRACT
-
Background/Aims
Intrahepatic cholangiocarcinoma (iCCA) represents an unmet clinical need due to its increasing incidence, aggressive biology, and limited treatment options. The extremely low-response rates to current systemic regimens and the emergence of adaptive resistance to targeted therapies underscore the urgent need for alternative therapeutic strategies. Given that the lineage-defining transcription factors SOX9 and YAP1 are central regulators of cholangiocyte and iCCA identity, we investigated their functional roles as potential therapeutic vulnerabilities across multiple preclinical models.
-
Methods
Patient tissue-microarray analysis, Sleeping Beauty hydrodynamic tail vein injection–based iCCA models, and Cre-mediated inducible gene deletion systems were used to investigate the roles of Sox9 and Yap1. Deep-learning–based prediction, RNA-seq, chromatin immunoprecipitation sequencing and immunohistochemistry analyses were performed to delineate transcriptional networks and downstream effectors associated with SOX9/YAP1 signaling.
-
Results
Dual deletion of Sox9 and Yap1 effectively eradicated advanced iCCA while preserving intrahepatic bile ducts, regardless of oncogenic drivers. Mechanistically, SOX9 and YAP1 transcriptionally compensated for each other when one was absent, and ILF2 and MGAT5 were identified as key downstream effectors mediating this compensatory mechanism. Loss of Ilf2 and Mgat5 suppressed iCCA, whereas overexpression of Ilf2 following Sox9/Yap1 co-deletion restored tumor development, indicating that ILF2 can functionally substitute for YAP1 and SOX9 in sustaining iCCA.
-
Conclusions
Co-targeting SOX9 and YAP1 offers a promising and safe broad-spectrum preventive/therapeutic approach for iCCA, potentially overcoming resistance to YAP1 inhibition. The adaptive resistance mechanism identified may extend to other malignancies, providing insights for addressing the advanced resistance to YAP1-TEAD-directed therapies.
-
Keywords: Liver cancer; Neoplasm drug resistance; Bile duct cancer; Interleukin enhancer binding factor 2; Transcriptional coactivator with PDZ-binding motif (TAZ) proteins
Study Highlights
• Co-repression of Sox9 and Yap1 eradicates established intrahepatic cholangiocarcinoma across multiple oncogenic mouse models while preserving normal intrahepatic bile ducts.
• SOX9 and YAP1 transcriptionally compensate for each other, sustaining malignant biliary identity when either factor is individually suppressed.
• Integrated RNA-seq, ChIP-seq, deep-learning prediction, and functional validation identify ILF2 and MGAT5 as key downstream effectors of this compensatory network.
• Dual SOX9/YAP1 targeting may provide a broad-spectrum therapeutic strategy to overcome adaptive resistance to YAP1/TEAD-directed therapies.
Graphical Abstract
INTRODUCTION
Intrahepatic cholangiocarcinoma (iCCA) is the second most common liver cancer, with steadily increasing incidence worldwide [
1]. Diagnosed in ~8,000 new patients annually in the US, iCCA accounts for ~15% of all liver malignancies and has a 5-year survival rate of only 15%, underscoring its heavy disease burden and status as a major unresolved health challenge [
2]. Surgical resection and liver transplantation, the most effective treatments for liver cancers, are feasible only for stage 1–2 iCCA patients, who represent about 30% of all cases [
2,
3]. Unresectable patients mainly rely on gemcitabine/cisplatin (Gem/Cis) therapy, which yields a response rate under 15% and extends survival by merely a year in responders [
3–
5]. Unlike hepatocellular carcinoma (HCC), immune checkpoint inhibitors (ICIs) show extremely limited efficacy in iCCA [
6]. Combining ICIs with Gem/Cis produces only marginal improvement and remains insufficient for most advanced-stage patients [
3,
5–
7]. Even approved targeted agents, including FGFR-fusion and IDH1/2 inhibitors, provide modest benefits due to rapid adaptive resistance [
5,
8–
10]. Thus, alternative therapeutic options are urgently needed, especially for advanced-stage iCCA patients lacking effective targeted treatments.
SOX9 and YAP1 are central regulators of Notch signaling in bile duct development, particularly cholangiocyte maturation and morphogenesis [
11–
14]. Both are expressed in over 90% of human iCCA, with levels correlating positively with tumor grade [
15,
16]. Murine studies have shown that forced expression of
NICD (active Notch) or
YAP1S127A in hepatocytes (HCs), together with proto-oncogenes such as
myristoylated (
myr) Akt or
KRASG12D, reprograms HCs into iCCA [
17,
18]. Using well-established Sleeping Beauty transposon/transposase-mediated hydrodynamic tail vein injection, we previously found that deleting either
Sox9 or
Yap1 markedly delays but does not prevent tumor formation [
19]. Despite their roles in bile duct pathophysiology and iCCA progression, the downstream effectors of SOX9 and YAP1 and their compensatory crosstalk remain poorly defined. Moreover, while YAP1-TEAD inhibitors have emerged as promising anticancer agents, the molecular basis of adaptive resistance to YAP1 inhibition is still unclear.
Here, we show that cholangiocyte-specific co-deletion of Sox9 and Yap1 eradicates fully developed iCCA across multiple genetic models, irrespective of oncogenic drivers, while preserving normal bile duct physiology. We further reveal how SOX9 and YAP1 engage in transcriptional compensation to sustain biliary malignancy when either factor is suppressed, involving persistent activation of pathways and regulators such as ILF2 and MGAT5. These findings suggest that dual targeting of SOX9 and YAP1, along with key compensatory effectors, may offer a broad-spectrum therapeutic strategy for unresectable iCCA, particularly in cases unresponsive to Gem/Cis or existing targeted therapies.
MATERIALS AND METHODS
Animals
All procedures followed University of Pittsburgh Institutional Animal Care and Use Committee guidelines. OPN-CreERT2 mice (gift from Dr. Frédéric Lemaigre) were crossed with Sox9(f lox/ f lox)(JAX#013106);Yap1(f lox/ f lox) (JAX#032192), and Yap1(flox/flox);Wwtr1(flox/flox) (JAX#030532) to generate OPN-CreERT2;Sox9(flox/flox);Yap1(flox/flox) and Yap1(flox/flox);Wwtr1(flox/flox) lines. Both sexes (6–12 weeks) were analyzed. FVB/NJ mice (JAX#001800) were used for CRISPR/Cas9 knockouts. Serum biochemistry was performed at the UPMC clinical chemistry lab.
Patient data
Human CCA samples were obtained from the PLRC Bio-repository (P30DK120531; IRB STUDY19070068). Tissue microarrays (TMAs) from 108 UPMC patients (two 1 mm tumor cores each) were stained for p-AKT-S473, SOX9, and YAP1. Slides were scanned on the Aperio XT and scored by a pathologist (A.S.) as SOX9/YAP1 0–2+ and p-AKT 0–3+. Mean ≥1.5=“High”; <1.5=“Low/Negative.” Demographic and clinical data are summarized in
Supplementary Table 1.
RESULTS
Single deletion of Yap1 or Sox9 is insufficient to prevent iCCA formation
In our previous study [
19], we demonstrated that individual deletion of either
Sox9 (SKO) or
Yap1 (YKO) significantly delays
myrAkt-NICD (
AN)-iCCA formation. Despite this delay, iCCA nodules still developed in SKO and YKO mouse livers (
Fig. 1A, 1B) ultimately resulting in mortality, indicating that the loss of either factor alone is insufficient to fully block Notch-dependent HC-to-iCCA transformation [
19]. Immunohistochemistry (IHC) revealed tumor nodules positive for only one of SOX9 or YAP1, confirming that these factors function independently under Notch signaling in
AN-iCCA (
Fig. 1C). Indeed, analysis of proliferation and cell death markers (
Supplementary Fig. 1) suggests that each plays a distinct role in
AN-iCCA development. To assess the clinical relevance of SOX9 or YAP1 singular positive murine iCCA subsets, we analyzed patient TMA samples and identified YAP1
+/SOX9
− (4.6%) and SOX9
+/YAP1
− (3.7%) CCA cases (
Fig. 1D, 1E). These findings provide evidence of distinct iCCA subsets driven by either SOX9 or YAP1, potentially characterized by unique molecular signatures.
Given the observed insufficient prevention of single-gene deletion of
Sox9 or
Yap1 on
AN-iCCA tumorigenesis (
Fig. 1B, 1C), we next investigated the impact of concurrently ablating
Sox9 and
Yap1. To achieve tumor-specific dual knockout/down of
Sox9 and
Yap1 (dKO), we injected
NICD,
myrAkt-shYap1 and
Cre recombinase (
Cre) into
Sox-9flox/flox mice (
Fig. 2A). As controls, we injected
NICD,
pCMV-Empty, and
myrAkt-sh-Luciferase plasmids into the
Sox9flox/flox mice to generate dual WT (dWT) mice. Consistent with previous findings [
17,
19],
AN-dWT mice developed lethal iCCA, necessitating euthanasia around 5 weeks post-HDTVI, whereas
AN-dKO mice remained asymptomatic at both 5 weeks and 3 months post-HDTVI, demonstrating significantly improved survival (
Fig. 2B). Liver weight to body weight (LW/BW) ratio and gross liver assessment revealed markedly reduced tumor burden in
AN-dKO mice at both 5 weeks and 3 months compared to
AN-dWT mice at 5 weeks (
Fig. 2C,
2D). Histological analysis at 5 weeks using IHC for HA-tag (myrAKT), MYC-tag (NICD), and panCK confirmed that
AN-dWT livers developed iCCA, whereas
AN-dKO livers showed complete absence of any tumor formation (
Fig. 2E). These data demonstrate that simultaneous suppression of
Yap1 and
Sox9 completely abrogates
AN-dependent HC-derived iCCA development. The compensatory roles of SOX9 and YAP1 suggest that their functional redundancy may be critical for sustaining iCCA development, highlighting their synergistic contribution and potential for adaptive resistance when one is inhibited.
Given the clinical context of iCCA, where most cases are diagnosed at an advanced and unresectable stage, we next investigated the effect of simultaneous
Sox9 and
Yap1 deletion on established iCCA. To achieve iCCA-specific inducible gene deletion at the established tumors, we employed the
Osteopontin (
OPN)-
CreERT2 strain, a well-established, tamoxifen (TM)-inducible
Cre expression system specific to cholangiocytes [
20,
21]. We generated the
OPN-CreERT2;
Sox9(flox/flox);
Yap1(flox/flox) (
OPN-
SY) strain enabling inducible co-elimination of
Sox9 and
Yap1 in the intrahepatic bile ducts and iCCA. As previously validated [
22], three intraperitoneal (i.p.) injections of TM (100 mg/kg) effectively induced
Cre-mediated recombination, resulting in
Sox9 and
Yap1 deletion without causing hepatobiliary injury (
Supplementary Fig. 2). To delete
Sox9 and
Yap1 (di-SYKO) at the established stage of
AN-iCCA, when tumor burden occupied approximately 25% of each liver lobe, as evident by HA-tag IHC (
Supplementary Fig. 3), we administered 6 doses of 100 mg/kg TM (i.p.) between 3 and 4 weeks post-HDTVI, followed by assessments at 8 and 12 weeks (
Fig. 3A). Littermate controls (di-SYWT) received corn oil injections and were sacrificed at the same time points as the TM-treated group. As expected, di-SYWT mice developed lethal iCCA requiring euthanasia 6–8 weeks post-HDTVI. In contrast, TM injection significantly prolonged survival, with over 50% of the di-SYKO mice still alive at 3 months post-HDTVI (
Fig. 3B). The gross and LW/BW ratio assessment revealed a significant reduction in tumor burden in the di-SYKO mice at both 8 weeks and 3 months compared to the controls at the same stages (
Fig. 3C, 3D). Notably, 5 out of 12 TM-treated livers exhibited a completely normal LW/BW ratio with no visible tumors, while the remaining 7 displayed a slightly elevated LW/BW ratio with small but progressing tumor foci. These findings underscore the potent tumor-eliminating effects of
Sox9 and
Yap1 co-deletion in
AN-iCCA. Given that some tumor nodules were still observed in the livers of di-SYKO mice following TM administration, we further characterized residual tumors using IHC for the HA-tag (myrAKT), SOX9, YAP1, CK19, and HNF4α (
Fig. 3E, 3F). Consistent with the gross and LW/BW ratio assessments, HA-tag
+ iCCA nodules were detected only in 7 out of 12 di-SYKO livers, whereas the remaining 5 showed complete tumor elimination. In the 7 di-SYKO livers harboring residual tumors, the majority of iCCA nodules retained expression of either SOX9 or YAP1 (
Fig. 3E), indicating incomplete
Cre-mediated deletion of both genes despite the 6 TM injections (
Fig. 3F). Interestingly, in 2 out of the 7 mice, a small subset of SOX9
− YAP1
− tumors co-expressed the mature biliary marker CK19 and the HC lineage marker HNF4α, indicating a mixed HCC/ iCCA phenotype (
Fig. 3F and
Supplementary Fig. 4). These rare resistance subclones may reflect stochastic integration of the SB transposase, leading to incomplete targeting of cholangiocyte lineage instead of elimination. In summary, aside from cases with incomplete
Cre-mediated deletion, concurrent loss of
Yap1 and
Sox9 effectively eradicated established
AN-iCCA tumors. These findings highlight the therapeutic potential of
Sox9 and
Yap1 co-deletion, particularly in the 16–30% of clinical iCCA cases we previously identified [
19].
Next, to further evaluate the therapeutic efficacy of
Sox9 and
Yap1 co-repression across genetically heterogeneous iCCA subtypes, we investigated the co-deletion effects on multiple established-stage murine iCCA models. Specifically, we induced iCCA in
OPN-SY mice by delivering
Akt-Fbxw7D (
AF) [
23] or
KRASG12D/G12V-sg-p19 (
KP19) [
24] plasmids, which model distinct molecular subclasses of iCCA. These murine models exhibited heterogeneous molecular signatures and unique pathological features, including distinct immune tumor microenvironments, thereby representing distinct clinical iCCA classes [
25]. As described previously in the established
AN-iCCA model, TM was administered between 3 and 6 weeks post-HDTVI. Remarkably, co-deletion of
Sox9 and
Yap1 resulted in significant eradication of
AF or
KP19-driven iCCA tumors (
Supplementary Figs. 5,
6). Histologically, the few tumor nodules that remained were predominantly SOX9
+, similar to those observed in
AN-di-SYKO models (
Supplementary Figs. 5E,
6E), indicating that these lesions likely arose from a minor subset of cells due to incomplete
Cre-mediated gene deletion.
Given that robust expression of biliary factors is essential for bile duct integrity [
26], we evaluated the effects of
Sox9 and
Yap1 co-deletion on biliary injury. In contrast to
Yap1/Taz deletion, which disrupts biliary integrity despite eliminating various iCCA subtypes (
Supplementary Fig. 7), serum chemistry analyses revealed no significant differences in markers of liver or biliary injury, including alanine transaminase, aspartate aminotransferase, alkaline phosphatase, and total bilirubin—between di-SYKO and di-SYWT mice (
Supplementary Fig. 2B). Together, these findings demonstrate that concurrent
Sox9 and
Yap1 deletion eradicates iCCA across molecular subtypes while preserving the integrity of non-malignant cholangiocytes and HCs, underscoring its therapeutic potential.
To elucidate the distinct role of SOX9 and YAP1 in
AN-iCCA, we established YAP1
+ SOX9
− (
AN-SKO) or SOX9
+;YAP1
− (
AN-YKO)
AN-iCCA using
AAV8-TBG-Cre approach to ensure complete deletion of target gene in the entire HC population (
Supplementary Fig. 8A). As controls,
AAV8-TBG-GFP virus-injected
AN-iCCA (
AN-GFP) livers were used. Consistent with previous studies [
19],
AN-GFP mice developed a lethal iCCA burden, requiring euthanasia at 3–5 weeks post-HDTVI (
Supplementary Fig. 8B). In contrast,
AN-SKO and
AN-YKO exhibited delayed disease progression and improved survival until animals were sacrificed for analysis at 5–8 or 11–13 weeks. All animals were sacrificed upon exhibiting severe morbidity, including immobility, at which point gross examination revealed ascites and extensive tumor liver burden, suggesting iCCA-driven lethality in
AN-SKO and
AN-YKO (
Supplementary Fig. 8B, 8C). Microscopic observation of
AN-GFP,
AN-YKO, and
AN-SKO livers by IHC for SOX9 and YAP1 confirms complete target gene deletions and successful establishment of single-positive iCCAs for further investigations (
Supplementary Fig. 8D). Next, we performed bulk RNA-seq on
AN-GFP,
AN-YKO, and
AN-SKO tumors and compared differentially expressed genes (DEGs) in three independent comparisons:
AN-GFP vs. healthy liver (HL),
AN-YKO vs. HL, and
AN-SKO vs. HL (
Supplementary Fig. 8E–8L). Principal component analysis demonstrated clear clustering of HL samples distinct from the three iCCA groups, confirming distinct transcriptomic profiles between each tumor and non-tumor liver tissue (
Fig. 4A). Within-group transcriptomic similarity was observed for HL,
AN-GFP, and
AN-SKO, whereas
AN-YKO displayed greater heterogeneity despite efficient
Yap1 deletion. Comparison of DEGs across the iCCA models identified 3,281 upregulated and 2,097 downregulated genes relative to HL, regardless of
Sox9 or
Yap1 deletion (
Fig. 4B). Pathway enrichment analysis using the DEGs identified 149 activated and 89 suppressed pathways in
AN-iCCAs (
Fig. 4C,
Supplementary Tables 2,
3). To identify potential upstream regulators of the DEGs, we performed EnrichR analysis [
27], revealing that top predicted regulators of upregulated genes included biliary-specific transcription factors and tumor suppressors such as SOX9, TP53, SOX2, and TEADs (
Supplementary Fig. 9A, 9B). Taken together, our analysis reveals that while
Yap1 or
Sox9 deletion induces distinct transcriptomic profiles, the core gene expression networks required to sustain
AN-iCCA malignancy remain intact. These findings provide insight into compensatory mechanisms underlying iCCA progression and potential vulnerabilities for therapeutic targeting.
To investigate the compensatory mechanisms underlying
AN-iCCA malignancy mediated by SOX9 and YAP1 transcriptional regulation, we performed chromatin immunoprecipitation sequencing (ChIP-seq) to identify genomic regions occupied by SOX9 and YAP1 in
AN-GFP, AN-SKO, and
AN-YKO livers, along with pooled input controls (N=3). As a supplement, we included TEAD1 because TEAD1 is the most highly expressed TEAD family member in
AN-iCCA (GSE200472), and its inhibition effectively blocks
AN-iCCA progression [
19]. We also included TAZ given the validated YAP1 and TAZ compensation in our model (
Supplementary Fig. 7). The analysis identified 2,206, 80, 2,685, and 207 binding peaks for SOX9, YAP1, TAZ, and TEAD1, respectively, in
AN-GFP livers, with approximately 50% of peaks located within <2 kb of promoter regions, suggesting their primary role in transcriptional regulation (
Supplementary Fig. 10). To confirm whether the genes whose promoters were bound by SOX9 or YAP1 were induced, we integrated ChIP-seq data with transcriptomic profiles from the corresponding livers. Among the 3,281 genes consistently upregulated across all three iCCAs compared to HL, TAZ binding was the most prevalent, associated with 1,537 genes, followed by SOX9 (197 genes), TEAD1 (120 genes), and YAP1 (47 genes) (
Fig. 4D). This suggests that TAZ plays a dominant role in upregulating this gene signature. Next, we examined compensatory transcriptional interactions by identifying genes originally bound by either SOX9 or YAP1 in
AN-GFP livers that became bound by YAP1 in the absence of SOX9 (
AN-SKO) or vice versa (
AN-YKO). This analysis revealed 82 and 37 genes, respectively, among the 3,281 upregulated genes across all three iCCAs (
Fig. 4E, 4F,
Supplementary Tables 4,
5). Interestingly, the 82 newly YAP1-bound genes in
AN-SKO overlapped with TEAD1-bound genes, suggesting that YAP1 regulates these genes through a transcriptional complex with TEAD1 (
Supplementary Fig. 9C, 9D). EnrichR [
27] upstream regulator analysis of the 82 newly YAP1-bound genes predicted 721 potential regulators, including SOX9, and TEAD4 while the same analysis of the 37 newly SOX9-bound genes predicted 654 potential regulators, including SOX2, TCF3, and HNF1A (
Fig. 4E, 4F). Consistent with this compensatory model, ChIP-seq analysis revealed a significantly enhanced YAP1 binding signal at the promoter regions of
Runx1 in
AN-SKO livers, where SOX9 had previously been the dominant binding factor in
AN-GFP livers. This suggests that YAP1 compensates for SOX9 loss by
de novo binding to maintain transcriptional regulation (
Supplementary Fig. 9E, 9F). Collectively, our findings identify 82 and 37 novel functional downstream targets of SOX9 and YAP1, respectively, highlighting their compensatory roles in transcriptional networks essential for iCCA in murine models.
To investigate the clinical relevance of the SOX9-YAP1 compensatory relationship in human CCA, we checked their expression pattern in publicly available transcriptome datasets from large clinical cohorts, comprising a total of 319 CCA patients and 101 non-tumor controls: The Cancer Genome Atlas (TCGA-CHOL, 9 controls and 36 CCAs) [
28], GSE26566 (59 controls and 104 CCAs) [
29], GSE33327 (6 controls and 149 iCCAs) [
30] and GSE107943 (27 controls and 30 CCAs) [
31]. Among the 82 newly YAP1-bound genes (
Fig. 4E), expression patterns for 78 were available in the TCGA dataset, 77 in GSE26566, 76 in GSE33327, and 80 in GSE107943, while expression data for all 37 newly SOX9-bound genes (
Fig. 4F) were available across all three datasets (
Supplementary Fig. 11). Notably, of the 82 and 37 candidates (
Fig. 5A, 5B), 20 and 15 genes, respectively, showed consistent increases in expression across at least three out of four datasets (
Fig. 5C and
Supplementary Tables 6,
7). Next, to assess the therapeutic potential of our candidate genes, we applied a deep learning method, DeepDEP [
32], to predict how knocking out each gene would impact tumor cell viability using a data set from the TCGA cohort [
28]. We found that most of the differentially expressed candidates (77 out of 82 and 33 out of 37) were predicted to be significantly associated with tumor cell viability, suggesting the functional relevance of these candidate genes in human CCA (
Fig. 5D, 5E). We then selected three genes based on the following criteria: consistent expression across at least three independent clinical CCA cohorts (
Fig. 5C and
Supplementary Tables 7,
8), novelty within the CCA literature, and predictive essentiality identified by our Deep-DEP approach. These genes were
MIDN and
ILF2. In addition, we manually selected
MGAT5 from the pool of SOX9- and TAZ-bound genes in
AN-GFP, as
Mgat5 expression is significantly enhanced in both mouse and human CCA (
Supplementary Fig. 12), and prior studies have implicated it in pancreatic ductal adenocarcinoma (PDAC), a malignancy closely related to CCA [
31]. To functionally assess these candidates in iCCA progression as downstream of SOX9 and YAP1 in iCCA, we validated their loss-of-function effects in iCCA using a
CRISPR/Cas9-based platform (
Fig. 6A, 6B) [
22]. This system features a tumor-specific, inducible knockout strategy, utilizing a
Cre-activated, GFP-tagged
Cas9 along with gene-specific guide RNAs (
Fig. 6B) [
22]. Using this approach, we specifically deleted
Midn,
Mgat5, or
Ilf2 in
AN-iCCA livers (
Fig. 6C). At four weeks post-HDTVI, loss of
Mgat5, or
Ilf2 led to a significant reduction in
AN-iCCA tumor size and LW/BW ratios. In contrast, no significant impact on tumor growth was observed in livers with
Midn deletion. Given prior evidence implicating ILF2 in bile duct development in zebrafish [
33], we tested whether re-expression of
Ilf2 could likewise restore tumor growth in the same model (
Fig. 6D). Remarkably, overexpression of
Ilf2 restored tumor burden in
Sox9/
Yap1-DKO livers, as demonstrated by LW/BW ratio, and gross analyses (
Fig. 6E). These findings confirm that MGAT5 and ILF2 play crucial roles in HC-to-iCCA transformation and that ILF2 can functionally substitute for the simultaneous loss of SOX9 and YAP1 in iCCA development.
DISCUSSIONS
Liver cancer, particularly iCCA, is among the most genetically and pathologically heterogeneous tumors [
4,
29,
30]. This heterogeneity results from diverse hepatic exposures to metabolites, toxins, or injury-induced mutations that drive malignant transformation and complicate treatment. From this standpoint, dual targeting of SOX9 and YAP1 provides a promising strategy for unresectable iCCA by eliminating resistant tumor clones that survive single-target therapy. We demonstrate that this combinatorial approach acts safely and specifically in cholangiocytes when restricted to the liver. Yet, efficient gene-therapy delivery to the mammalian bile duct remains challenging. Achieving CCA-selective toxicity while preserving normal ducts could expand delivery options and address a major limitation in current therapies.
Previous work showed that co-repression of YAP1/TAZ suppresses iCCA [
34], but its efficacy in advanced disease was unclear. We found that simultaneous inhibition of YAP1 and TAZ eliminates established iCCA and HCC, although dual blockade induces bile-duct damage and acute cholestasis, underscoring the need for tumor-specific targeting (
Supplementary Fig. 7). Our data indicate that TAZ binds more dominantly than YAP1 or TEAD1 to key iCCA genes, suggesting functional divergence within Hippo effectors. Given the limited efficacy of current YAP1-TEAD inhibitors due to adaptive resistance, improved pharmacologic disruption of TEAD–TAZ interaction may enhance clinical translation. These findings shift the paradigm from a YAP1-centric to a TAZ-TEAD-focused view of biliary pathophysiology. TAZ-specific binding genes identified by ChIP-seq may represent essential regulators of bile duct viability and warrant further investigation. Integrating chromatin accessibility and co-factor mapping in future work will be important to define the mechanistic basis for locus-specific Hippo effector switching.
Importantly, identifying SOX9 as a safe co-target to overcome YAP1 resistance could extend to other SOX9-positive, YAP1-active cancers such as gastric [
35] and intestinal malignancies [
36]. Several compensatory genes co-regulated by SOX9/YAP1 also displayed strong anti-CCA effects. Despite decades of Hippo pathway research, their downstream functional mediators remain largely undefined, warranting further mechanistic and therapeutic exploration. Consistent with this, treatment with the TEAD inhibitor VT104 led to a slight reduction in tumor burden in
Sox9-deleted AN-iCCA (
Supplementary Fig. 13). However, this preliminary observation is based on a limited sample size (n=2) with a very short treatment window and therefore will require further validation in future studies.
To validate functional candidates from integrated transcriptomic and ChIP-seq analyses, we established a
CRISPR/ Cas9-based inducible, tumor-specific gene-disruption platform using
OPN-CreERT2 mice [
22]. This system enables both preventive and therapeutic interrogation of target genes
in vivo without the need for floxed strains and can be extended to HCC using
AAV8-TBG-Cre. We found that strict matching of control-vector dosage is essential for accurate tumor-suppression assessment.
Using this model, we screened three candidates in
AN-iCCA and identified
Ilf2 and
Mgat5 as functional contributors to tumor formation (
Fig. 6C). Although
Midn was upregulated in murine and clinical datasets, its loss did not reduce tumor burden [
37]. ILF2 regulates cancer-cell proliferation and apoptosis [
38], but its link to CCA remains unreported, identifying a new therapeutic opportunity. MGAT5, a key N-glycosylation enzyme, was found to suppress iCCA development under SOX9/YAP1 control. Altered glycosylation profoundly affects tumor signaling, metastasis, and immune evasion [
39]. Increased β1–6 branching driven by MGAT5 is common in many cancers [
40] and represents a therapeutic target in HCC [
41] and PDAC [
42,
43]. Our data show that
CRISPR/Cas9-mediated
Mgat5 deletion blocks iCCA formation, suggesting potential preventive and therapeutic applications across immune-classified iCCA models. Interestingly, our unbiased bioinformatic analysis suggests that
Wwtr1 may act in a context-dependent compensatory manner in the absence of
Sox9. In WT iCCA, SOX9 occupies the
Wwtr1 promoter region, whereas
Sox9 loss is associated with inducing increased TEAD1/YAP1 binding at the promoter (data not shown). These observations suggest that
Sox9 may help maintain
Wwtr1 transcription in the AN-iCCA context. Further comprehensive investigation will be required to define SOX9-Hippo pathway crosstalk under specific genetic conditions that contribute to iCCA malignancy.
In summary, dual targeting of SOX9 and YAP1 offers a potent and safe therapeutic strategy for iCCA by abolishing transcriptional compensation between these fate-determining factors and preventing resistant tumor outgrowth. This study reveals adaptive resistance mechanisms underpinning Hippo-YAP1-dependent malignancies and provides a foundation for developing combination therapies that integrate gene-specific strategies.
FOOTNOTES
-
Authors’ contribution
Conception and design: SK. Development of methodology: SH, SL, SK, MK, LJW, YCC. Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): LM, SH, MK, AS, SL, YC, YP, YCC. Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): SK, MK, SH, LM, AS, LW, JJL, SL, YC, LJW, JJL and YCC. Writing, review, and/or revision of the manuscript: SK, MK. Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): SH, SL, AS, JK. Study supervision: SK.
-
Acknowledgements
This project used the UPMC Hillman Cancer Center and Tissue and Research Pathology/Pitt Biospecimen Core shared resource, which is supported in part by award P30CA047904. NIH grant 1P30DK120531-01 to the Pittsburgh Liver Research Center. We would like to thank the Center for Biologic Imaging (funded by S10MH126905) at the University of Pittsburgh.
This work was supported by NIH grants R01CA258449, R01CA300059 and by PLRC Pilot & Feasibility grant PF 2019-05 to S.K., and Innovation in Cancer Informatics Discover grant (https://www.the-ici-fund.org) to S.K and S.L., and NIH grant 1P30DK120531-01 to Pittsburgh Liver Research Center (PLRC), and in part by R35GM154967 and R00CA248944 to Y.C. This work was supported in part by the University of Pittsburgh Center for Research Computing through the resources provided. Specifically, this work used the HTC cluster, which is supported by NIH award number S10OD028483. This project was also partly supported by NIH grant R35GM159862, the Competitive Medical Research Fund (CMRF) of the UPMC Health System to S.L.
-
Conflicts of Interest
The authors have no conflicts to disclose.
SUPPLEMENTARY MATERIAL
Supplementary material is available at Clinical and Molecular Hepatology website (
http://www.e-cmh.org).
Supplementary Figure 1.
Distinct roles of SOX9 and YAP1 in regulation of viability of myrAkt-NICD-mediated iCCA. (A) Experimental scheme showing plasmids used for HDTVI, mouse genotypes, and analysis time points. (B) Representative IF for SOX9 (red), YAP1 (red), PCNA (green), and DAPI (blue) in 5-week myrAkt–NICD WT, Sox9 KO (SKO), and Yap1 KO (YKO) livers. (C) IHC for TUNEL showing markedly increased cell death in SKO, with minimal and comparable death in WT and YKO at 5 weeks. (D) Quantification of PCNA⁺ nuclei relative to total tumor cells from (B) showing significantly reduced proliferation in SKO and YKO versus WT. (E) Quantification of TUNEL⁺ nuclei from (C) showing significantly increased cell death in SKO versus WT and YKO. Error bars, SEM; *P<0.05, **P<0.01, ****P<0.0001.
cmh-2025-1170-Supplementary-Fig-1.pdf
Supplementary Figure 2.
Co-deletion of Sox9 and Yap1 preserves cholangiocytes, maintaining intact bile ducts. (A) Experimental scheme showing plasmids used for HDTVI, tamoxifen treatment, mouse genotypes, and analysis time points. (B) Serum ALT, AST, and total bilirubin showing comparable levels, with slightly but significantly decreased ALP in tamoxifen-treated versus oil-treated controls; all values within normal ranges indicate intact biliary structure without hepatic injury. (C) Serial IHC for SOX9, YAP1, and GFP in tamoxifentreated OPN-CreERT2;Sox9f/f;Yap1f/f;Rosa26-stopf/f–eYFP livers confirming efficient CreERT2-mediated deletion of Sox9, Yap1, and the stop cassette. Error bars, SEM; ***P<0.001.
cmh-2025-1170-Supplementary-Fig-2.pdf
Supplementary Figure 3.
AN-iCCA at 3 weeks post-HDTVI shows an established tumor stage. (A) Experimental scheme showing plasmids used for HDTVI, mouse genotypes, and analysis time points. (B) Representative whole-slide IHC for HA-tag showing established AN-iCCA at 3 weeks post-HDTVI in OPN-CreERT2;Sox9f/f;Yap1f/f (OPN-SY) mice.
cmh-2025-1170-Supplementary-Fig-3.pdf
Supplementary Figure 4.
Simultaneous deletion of
Sox9 and
Yap1 very rarely reprograms
myrAkt-NICD iCCA into mixed HCC/iCCA (related to
Fig. 3F). (A) Experimental scheme showing plasmids used for HDTVI, tamoxifen treatment, mouse genotypes, and analysis time points. Serial IHC for SOX9, YAP1, CK19, HNF4A, and Ki-67 showing SOX9⁺/YAP1⁻/CK19⁺/HNF4A⁻/Ki-67⁺ proliferating iCCA (upper, B) and SOX9⁻/YAP1⁻/CK19⁺/HNF4A⁺/Ki-67⁺ proliferating mixed HCC/iCCA nodules (lower, C) at 8 weeks post-HDTVI.
cmh-2025-1170-Supplementary-Fig-4.pdf
Supplementary Figure 5.
Simultaneous deletion of
Sox9 and
Yap1 abrogates fully developed
myrAkt-Fbxw7D iCCA (related to
Fig. 3). (A) Experimental scheme showing plasmids used for HDTVI, tamoxifen treatment, mouse genotypes, and analysis time points. (B) Kaplan– Meier curve showing significantly improved survival of
Sox9/Yap1 inducible double KO (
di-SYKO) compared with
di-SYWT. (C) Representative gross liver images showing multiple large tumors in
di-SYWT and only a few small tumors in
di-SYKO at 3 months post- HDTVI. (D) LW/BW ratios showing significantly lower values in
di-SYKO than in
di-SYWT at 3 months. (E) Whole-slide IHC for HA-tag and SOX9 showing markedly reduced tumor nodules in
di-SYKO versus
di-SYWT, with remaining SOX9⁺ nodules indicating incomplete
Yap1 deletion. Error bars, SEM.
cmh-2025-1170-Supplementary-Fig-5.pdf
Supplementary Figure 6.
Simultaneous deletion of
Sox9 and
Yap1 abrogates advanced
KRAS G12D-sg-p19 iCCA (related to
Fig. 3). (A) Experimental scheme showing plasmids used for HDTVI, tamoxifen treatment, mouse genotypes, and analysis time points. (B) LW/BW ratios showing significantly lower values in
Sox9/Yap1 inducible double KO (
di-SYKO) than in
di-SYWT at 7–8 weeks. (C) Representative gross liver images showing multiple large tumors in
KRASG12D/G12V-sg-p19 di-SYWT and only a few small tumors in
di-SYKO at the same stage. (D) Whole-slide IHC for panCK and SOX9 showing markedly reduced tumor nodules in
di-SYKO versus
di-SYWT, with remaining SOX9⁺ nodules indicating incomplete
Yap1 deletion. Error bars, SEM.
cmh-2025-1170-Supplementary-Fig-6.pdf
Supplementary Figure 7.
Co-deletion of Taz and Yap1 abrogates advanced iCCA and biliary tree, leading to lethal cholestasis. (A) Experimental scheme showing plasmids used for HDTVI, tamoxifen treatment, mouse genotypes, and analysis time points. (B) Body weight significantly decreased in tamoxifen-treated mice versus corn-oil controls at 6–10 weeks post-HDTVI. (C) LW/BW ratios showing significantly lower values in tamoxifen-treated mice versus controls at the same time points. (D) Representative gross liver images showing multiple large tumors in tamoxifen-treated mice and only occasional small tumors in controls at 6–10 weeks. (E) Gross images of yellow abdomen, tail, ear, and serum in tamoxifen-treated mice versus controls indicate jaundice. (F) Serum ALP, ALT, AST, and total bilirubin levels significantly elevated in tamoxifen-treated versus control mice, confirming jaundice. Error bars, SEM; *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001.
cmh-2025-1170-Supplementary-Fig-7.pdf
Supplementary Figure 8.
Establishment and molecular characterization of pure
Sox9- or
Yap1-deleted
AN-iCCAs (related to
Fig. 4). (A) Experimental scheme showing plasmids used for HDTVI, AAV8-TBG-Cre/GFP treatment, mouse genotypes, and analysis time points. (B) Kaplan–Meier survival curve showing significantly improved survival in AAV8-TBG-Cre–mediated
Sox9 (
AN-SKO) or
Yap1 (
AN-YKO) knockout
AN-iCCAs compared with AAV8-TBG-GFP–treated WT controls (
AN-GFP). (C) Representative gross liver images showing multiple large tumors in
AN-GFP,
AN-SKO, and
AN-YKO at 3–12 weeks, when mice exhibited morbidity. (D) Whole-lobe IHC for YAP1 and SOX9 confirming efficient
Sox9 and
YAP1 deletion in respective KO livers. (E–I) Heatmaps of hepatic cell type–specific gene signatures in healthy liver,
AN-GFP,
AN-SKO, and
AN-YKO: hepatocyte (E), cholangiocyte (F), hepatoblast (G), hepatic stellate cell (H), and macrophage (I). (J–L) RNA-seq read count plots showing expression of cholangiocyte (J), hepatocyte (K), and HCC-specific (L) genes in healthy liver,
AN-GFP,
AN-SKO, and
AN-YKO. Error bars, SEM.
cmh-2025-1170-Supplementary-Fig-8.pdf
Supplementary Figure 9.
Analysis of Chip-seq binding patterns within genome annotations (related to
Fig. 4). (A) Upstream regulator analysis showing the top 20 predicted transcriptional regulators for the 3,281 genes in
Fig. 4B. (B) Upstream regulator analysis showing top 20 predicted transcriptional regulators for the 2,097 genes in
Fig. 4B (C) Venn diagram showing 462 overlapping genes bound by SOX9 in
AN-GFP but by TEAD1 in
AN-SKO among the 3,281 upregulated genes with top 10 predicted upstream regulators of these genes. (D) Venn diagram showing 28 overlapping genes bound by YAP1 in
AN-GFP but by TEAD1 in
AN-SKO among the 3,281 upregulated genes with the top 10 predicted upstream regulators of these genes. (E) Visualization of YAP1 and SOX9 ChIP-seq peaks aligned with transcript mapping at the
Runx1 and
Wwtr1 loci in each iCCA group. (F) qPCR showing significantly increased
Runx1 and
Wwtr1 mRNA expression in all iCCA groups versus healthy liver.
cmh-2025-1170-Supplementary-Fig-9.pdf
Supplementary Figure 10.
Additional data for
Figure 4. (A) Annotation of ChIP-seq peaks for SOX9, YAP1, TAZ, and TEAD1 relative to RefSeq transcripts showing enrichment at promoter regions. (B–D) Enrichment of ChIP-seq binding within ±2.5 kb of transcription start sites (TSS) in respective CCA groups.
cmh-2025-1170-Supplementary-Fig-10.pdf
Supplementary Figure 11.
Expression of 82 and 37 candidate genes across clinical CCA cohorts (related to
Fig. 5). Heatmaps showing expression of 82 and 37 candidate genes (from
Fig. 5A, 5B) in GSE26566 (A, B), GSE33327 (C, D), and TCGA (E, F) datasets, comparing CCA versus non-tumor regions in large patient cohorts.
cmh-2025-1170-Supplementary-Fig-11.pdf
Supplementary Figure 12.
SOX9- and TAZ-bound Mgat5 in iCCA. (A) Visualization of ChIP-seq peak signals for Mgat5 at promoter regions. (B) Expression data from TCGA, GSE76297, GSE107943, and GSE26566 showing elevated MGAT5 expression in CCA.
cmh-2025-1170-Supplementary-Fig-12.pdf
Supplementary Figure 13.
TEAD inhibitor VT104 reduces tumor burden in Sox9-deleted AN-iCCA (A) Experimental scheme showing plasmids used for HDTVI, VT104 dosing schedule, and analysis time points. As tumors developed at 4–5 weeks after HDTVI in Sox9-deleted AN-iCCA, VT104 was administered daily for four consecutive doses at week 5 and mice were scarified at week 6. (B) LW/BW ratio and representative gross images show that VT104 treatment reduces the tumor burden in Sox9-deleted AN-iCCA.
cmh-2025-1170-Supplementary-Fig-13.pdf
Figure 1Single deletion of Yap1 or Sox9 is insufficient to prevent iCCA formation. (A) Experimental scheme showing plasmids used for HDTVI, mouse genotypes, and analysis time points. (B) Representative HA-tag (myrAKT) IHC images showing papillary and cystic iCCA in WT, Sox9 KO (SKO), and Yap1 KO (YKO) at 5 weeks (scale bar=1 mm). IHC for HA-tag, SOX9, YAP1, and panCK in SKO and YKO CCA. (D) Correlation of SOX9 and YAP1 nuclear staining in human CCA TMA (scale bar=100 μm). (E) Representative TMA sections stained for SOX9 and YAP1; enlarged images highlight nuclear localization (red arrows, YAP1−; black arrows, SOX9+; red empty arrows, YAP1+; black empty arrows, SOX9−). iCCA, intrahepatic cholangiocarcinoma; IHC, immunohistochemistry; HDTVI, hydrodynamic tail vein injection; TMA, tissue microarray.
Figure 2Simultaneous suppression of Yap1 and Sox9 prevents myrAkt-NICD-driven iCCA formation. (A) Experimental scheme showing plasmids used for HDTVI, mouse genotypes, and analysis time points. (B) Kaplan–Meier survival curve showing markedly improved survival of Sox9/Yap1 double KO (dKO) compared with dWT in the myrAkt–NICD model. (C) LW/BW ratios showing significantly reduced tumor burden in dKO at 5 weeks and 3 months post-HDTVI. (D) Representative gross liver images showing multiple large tumors in dWT and the absence of visible tumors in dKO at both time points. (E) IHC for HA-tag, MYC-tag, and panCK in dWT and normal histology in dKO at 5 weeks. HDTVI, hydrodynamic tail vein injection; iCCA, intrahepatic cholangiocarcinoma; IHC, immunohistochemistry; LW/BW, liver weight to body weight. Error bars, standard deviation; *P<0.05, **P<0.01.
Figure 3Simultaneous deletion of Yap1 and Sox9 eliminates established myrAkt-NICD-driven iCCA. (A) Experimental scheme showing plasmids used for HDTVI, tamoxifen treatment, mouse genotypes, and analysis time points. (B) Kaplan–Meier survival curve showing significantly improved survival of Sox9/Yap1 inducible double KO (di-SYKO). (C) Representative gross liver images showing multiple large tumors in di-SYWT and only occasional small tumors in di-SYKO at 8 weeks and 3 months. (D) LW/BW ratios showing significantly lower values in di-SYKO at both time points. (E) Whole-lobe IHC for HA-tag, YAP1, and SOX9 showing HA-tag+/SOX9+/YAP1− iCCA (red dashed line) and HA-tag+/SOX9−/YAP1− tumors (green dashed line) in di-SYKO (scale bar=1 mm). (F) IHC for SOX9, YAP1, CK19, and HNF4A showing HA-tag+/SOX9+/YAP1− iCCA (upper) and HA-tag+/SOX9−/YAP1− mixed HCC/iCCA nodules from (lower). HDTVI, hydrodynamic tail vein injection; iCCA, intrahepatic cholangiocarcinoma; IHC, immunohistochemistry; LW/BW, liver weight to body weight. Error bars, standard deviation; ***P<0.001, ****P<0.0001.
Figure 4SOX9 and YAP1 compensate for each other during HC transformation into iCCA. (A) Principal component analysis of bulk RNA-seq between iCCAs (AN-GFP, AN-SKO, AN-YKO) and healthy livers. (B) Venn diagram showing overlap of up- and down-regulated genes among DEG comparisons of healthy livers versus AN-GFP, AN-SKO, or AN-YKO. (C) Venn diagram showing overlap of activated and suppressed pathways among the same comparisons. (D) UpSetR plot showing ChIP-seq binding patterns of YAP1, TEAD1, SOX9, and TAZ within the 3,281 overlapping upregulated genes from (B). (E) A Venn diagram showing 82 genes bound by SOX9/YAP1 in AN-GFP but only YAP1 in AN-SKO among the 3,281 upregulated genes, with top 10 predicted upstream regulators of these genes. (F) Venn diagram showing 37 genes bound by SOX9/YAP1 in AN-GFP but SOX9-only in AN-YKO, with top 10 predicted upstream regulators of these genes. ChIP-seq, chromatin immunoprecipitation sequencing; DEG, differentially expressed genes; HC, hepatocyte; iCCA, intrahepatic cholangiocarcinoma.
Figure 5SOX9–YAP1 compensatory targets identified in mouse iCCA are consistently upregulated in human CCA. (A, B) A Venn diagrams showing overlap of upregulated genes in clinical CCA datasets (TCGA, GSE26565, GSE33327 and GSE107943) among the 82 and 37 candidates from
Figure 4. (C) Graphs showing significantly increased expression of 10 genes across all datasets shown in (A) and (B). (D, E) Predicted gene essentialities of the 82 (
Fig. 4E) and 37 (
Fig. 4F) candidates in TCGA CCA tumors, based on DeepDEP gene-effect scores, where negative values indicate reduced viability upon knockout. iCCA, intrahepatic cholangiocarcinoma. **
P<0.01, ***
P<0.001.
Figure 6MGAT5 and ILF2 functionally contribute to transcriptional compensation between SOX9 and YAP1 in iCCA development. (A) Experimental scheme showing plasmids used for HDTVI, mouse genotypes, and analysis time points. (B) Schematic of SB-HDTVI–CRISPR/Cas9 system for inducible gene knockout. (C) LW/BW ratios showing significantly lower tumor burden in sgMgat5, and sgIlf2, and representative gross liver images showing multiple large tumors in sgEmpty, and sgMidn, with only small occasional tumors in sgMgat5 and sgIlf2. (D) Experimental scheme showing plasmids used for HDTVI, mouse genotypes, and analysis time points. (E) (Left) LW/BW ratios showing significantly increased tumor burden following Ilf2 overexpression in Sox9- and Yap1-deleted mice. (Right) Representative gross images showing minimal tumor nodules by Sox9/Yap1 co-deletion and widespread iCCA upon Ilf2 overexpression. iCCA, intrahepatic cholangiocarcinoma; LW/BW, liver weight to body weight; SB-HDTVI, Sleeping Beauty hydrodynamic tail vein injection. Error bars, standard deviation; *P<0.05, **P<0.01, ***P<0.001.
Abbreviations
chromatin immunoprecipitation sequencing
differentially expressed genes
intrahepatic cholangiocarcinoma
immune checkpoint inhibitor
liver weight to body weight ratio
pancreatic ductal adenocarcinoma
Sleeping Beauty transposon/transposase-mediated hydrodynamic tail vein injection
REFERENCES
- 1. Cancer Stat Facts: Liver and Intrahepatic Bile Duct Cancer. National Cancer Institute web site. <https://seer.cancer.gov/statfacts/html/livibd.html>. Accessed 27 Oct 2025.
- 2. Ilyas SI, Affo S, Goyal L, Lamarca A, Sapisochin G, Yang JD, et al. Cholangiocarcinoma - novel biological insights and therapeutic strategies. Nat Rev Clin Oncol 2023;20:470-486.
- 3. Kam AE, Masood A, Shroff RT. Current and emerging therapies for advanced biliary tract cancers. Lancet Gastroenterol Hepatol 2021;6:956-969.
- 4. Banales JM, Marin JJG, Lamarca A, Rodrigues PM, Khan SA, Roberts LR, et al. Cholangiocarcinoma 2020: the next horizon in mechanisms and management. Nat Rev Gastroenterol Hepatol 2020;17:557-588.
- 5. Elvevi A, Laffusa A, Scaravaglio M, Rossi RE, Longarini R, Stagno AM, et al. Clinical treatment of cholangiocarcinoma: an updated comprehensive review. Ann Hepatol 2022;27:100737.
- 6. Greten TF, Schwabe R, Bardeesy N, Ma L, Goyal L, Kelley RK, et al. Immunology and immunotherapy of cholangiocarcinoma. Nat Rev Gastroenterol Hepatol 2023;20:349-365.
- 7. Kalyan A, Khosla H, Kim RD. Immunotherapy in biliary tract cancers: where are we? Curr Oncol Rep 2022;24:1821-1828.
- 8. Wu MJ, Shi L, Merritt J, Zhu AX, Bardeesy N. Biology of IDH mutant cholangiocarcinoma. Hepatology 2022;75:1322-1337.
- 9. Wu Q, Ellis H, Siravegna G, Michel AG, Norden BL, Fece de la Cruz F, et al. Landscape of clinical resistance mechanisms to FGFR inhibitors in FGFR2-altered cholangiocarcinoma. Clin Cancer Res 2024;30:198-208.
- 10. Wu Q, Zhen Y, Shi L, Vu P, Greninger P, Adil R, et al. EGFR inhibition potentiates FGFR inhibitor therapy and overcomes resistance in FGFR2 fusion-positive cholangiocarcinoma. Cancer Discov 2022;12:1378-1395.
- 11. Hrncir HR, Goodloe B, Bombin S, Hogan CB, Jadi O, Gracz AD. Sox9 inhibits Activin A to promote biliary maturation and branching morphogenesis. Nat Commun 2025;16:1667.
- 12. Xu WP, Cui YL, Chen LL, Ding K, Ding CH, Chen F, et al. Deletion of Sox9 in the liver leads to hepatic cystogenesis in mice by transcriptionally downregulating Sec63. J Pathol 2021;254:57-69.
- 13. Antoniou A, Raynaud P, Cordi S, Zong Y, Tronche F, Stanger BZ, et al. Intrahepatic bile ducts develop according to a new mode of tubulogenesis regulated by the transcription factor SOX9. Gastroenterology 2009;136:2325-2333.
- 14. Molina LM, Zhu J, Li Q, Pradhan-Sundd T, Krutsenko Y, Sayed K, et al. Compensatory hepatic adaptation accompanies permanent absence of intrahepatic biliary network due to YAP1 loss in liver progenitors. Cell Rep 2021;36:109310.
- 15. Yuan X, Li J, Coulouarn C, Lin T, Sulpice L, Bergeat D, et al. SOX9 expression decreases survival of patients with intrahepatic cholangiocarcinoma by conferring chemoresistance. Br J Cancer 2018;119:1358-1366.
- 16. Wu H, Liu Y, Jiang XW, Li WF, Guo G, Gong JP, et al. Clinicopathological and prognostic significance of Yes-associated protein expression in hepatocellular carcinoma and hepatic cholangiocarcinoma. Tumour Biol 2016;37:13499-13508.
- 17. Fan B, Malato Y, Calvisi DF, Naqvi S, Razumilava N, Ribback S, et al. Cholangiocarcinomas can originate from hepatocytes in mice. J Clin Invest 2012;122:2911-2915.
- 18. Chen X, Calvisi DF. Hydrodynamic transfection for generation of novel mouse models for liver cancer research. Am J Pathol 2014;184:912-923.
- 19. Hu S, Molina L, Tao J, Liu S, Hassan M, Singh S, et al. NOTCH-YAP1/TEAD-DNMT1 axis drives hepatocyte reprogramming into intrahepatic cholangiocarcinoma. Gastroenterology 2022;163:449-465.
- 20. Di-Luoffo M, Pirenne S, Saandi T, Loriot A, Gérard C, Dauguet N, et al. A mouse model of cholangiocarcinoma uncovers a role for tensin-4 in tumor progression. Hepatology 2021;74:1445-1460.
- 21. Lesaffer B, Verboven E, Van Huffel L, Moya IM, van Grunsven LA, Leclercq IA, et al. Comparison of the Opn-CreER and Ck19-CreER drivers in bile ducts of normal and injured mouse livers. Cells 2019;8:380.
- 22. Park Y, Hu S, Kim M, Oertel M, Singhi A, Monga SP, et al. Context-dependent distinct roles of SOX9 in combined hepatocellular carcinoma-cholangiocarcinoma. Cells 2024;13:1451.
- 23. Wang J, Wang H, Peters M, Ding N, Ribback S, Utpatel K, et al. Loss of Fbxw7 synergizes with activated Akt signaling to promote c-Myc dependent cholangiocarcinogenesis. J Hepatol 2019;71:742-752.
- 24. D’Artista L, Moschopoulou AA, Barozzi I, Craig AJ, Seehawer M, Herrmann L, et al. MYC determines lineage commitment in KRAS-driven primary liver cancer development. J Hepatol 2023;79:141-149.
- 25. Martin-Serrano MA, Kepecs B, Torres-Martin M, Bramel ER, Haber PK, Merritt E, et al. Novel microenvironment-based classification of intrahepatic cholangiocarcinoma with therapeutic implications. Gut 2023;72:736-748.
- 26. Verboven E, Moya IM, Sansores-Garcia L, Xie J, Hillen H, Kowalczyk W, et al. Regeneration defects in YAP and TAZ mutant mouse livers are caused by bile duct disruption and cholestasis. Gastroenterology 2021;160:847-862.
- 27. Kuleshov MV, Jones MR, Rouillard AD, Fernandez NF, Duan Q, Wang Z, et al. Enrichr: a comprehensive gene set enrichment analysis web server 2016 update. Nucleic Acids Res 2016;44:W90-97.
- 28. Farshidfar F, Zheng S, Gingras MC, Newton Y, Shih J, Robertson AG, et al. Integrative genomic analysis of cholangiocarcinoma identifies distinct IDH-mutant molecular profiles. Cell Rep 2017;18:2780-2794.
- 29. Andersen JB, Spee B, Blechacz BR, Avital I, Komuta M, Barbour A, et al. Genomic and genetic characterization of cholangiocarcinoma identifies therapeutic targets for tyrosine kinase inhibitors. Gastroenterology 2012;142:1021-1031e15.
- 30. Sia D, Hoshida Y, Villanueva A, Roayaie S, Ferrer J, Tabak B, et al. Integrative molecular analysis of intrahepatic cholangiocarcinoma reveals 2 classes that have different outcomes. Gastroenterology 2013;144:829-840.
- 31. Ahn KS, O’Brien D, Kang YN, Mounajjed T, Kim YH, Kim TS, et al. Prognostic subclass of intrahepatic cholangiocarcinoma by integrative molecular-clinical analysis and potential targeted approach. Hepatol Int 2019;13:490-500.
- 32. Chiu YC, Zheng S, Wang LJ, Iskra BS, Rao MK, Houghton PJ, et al. Predicting and characterizing a cancer dependency map of tumors with deep learning. Sci Adv 2021;7:eabh1275.
- 33. Cheung Y, Wu Z, Garcia-Barcelo MM, Tam PKH, Ma ACH, Lui VCH. Deletion of interleukin enhancer binding factor 2 (ILF2) resulted in defective biliary development and bile flow blockage. J Pediatr Surg 2021;56:352-359.
- 34. Ko S, Kim M, Molina L, Sirica AE, Monga SP. YAP1 activation and Hippo pathway signaling in the pathogenesis and treatment of intrahepatic cholangiocarcinoma. Adv Cancer Res 2022;156:283-317.
- 35. Cao Z, An L, Han Y, Jiao S, Zhou Z. The Hippo signaling pathway in gastric cancer. Acta Biochim Biophys Sin (Shanghai) 2023;55:893-903.
- 36. Hong AW, Meng Z, Guan KL. The Hippo pathway in intestinal regeneration and disease. Nat Rev Gastroenterol Hepatol 2016;13:324-337.
- 37. Li F, Chen Q, Xue H, Zhang L, Wang K, Shen F. LncRNA MNX1-AS1 promotes progression of intrahepatic cholangiocarcinoma through the MNX1/Hippo axis. Cell Death Dis 2020;11:894.
- 38. Du H, Le Y, Sun F, Li K, Xu Y. ILF2 directly binds and stabilizes CREB to stimulate malignant phenotypes of liver cancer cells. Anal Cell Pathol (Amst) 2019;2019:1575031.
- 39. RodrÍguez E, Schetters STT, van Kooyk Y. The tumour glycocode as a novel immune checkpoint for immunotherapy. Nat Rev Immunol 2018;18:204-211.
- 40. Pinho SS, Reis CA. Glycosylation in cancer: mechanisms and clinical implications. Nat Rev Cancer 2015;15:540-555.
- 41. Liu H, Wu Q, Liu Y, Liu W, Zhang W, Pan D, et al. Prognostic significance of β1,6-N-acetylglucosaminyltransferase V expression in patients with hepatocellular carcinoma. Jpn J Clin Oncol 2015;45:844-853.
- 42. Hollander EE, Flock RE, McDevitt JC, Vostrejs WP, Campbell SL, Orlen MI, et al. N-glycosylation by Mgat5 imposes a targetable constraint on immune-mediated tumor clearance. JCI Insight 2024;9:e178804.
- 43. Greco B, Malacarne V, De Girardi F, Scotti GM, Manfredi F, Angelino E, et al. Disrupting N-glycan expression on tumor cells boosts chimeric antigen receptor T cell efficacy against solid malignancies. Sci Transl Med 2022;14:eabg3072.
