Aberrant Notch and Wnt signaling are known drivers of cholangiocarcinoma (CCA), but the underlying factors that initiate and maintain these pathways are not known. Here, we show that the proline-rich homeodomain protein/hematopoietically expressed homeobox (PRH/HHEX) transcription factor forms a positive transcriptional feedback loop with Notch3 that is critical in CCA. PRH/HHEX expression is elevated in CCA, and depletion of PRH reduces CCA tumor growth in a xenograft model. Overexpression of PRH in primary human biliary epithelial cells is sufficient to increase cell proliferation and produce an invasive phenotype. Interrogation of the gene networks regulated by PRH and Notch3 reveals that unlike Notch3, PRH directly activates canonical Wnt signaling. These data indicate that hyperactivation of Notch and Wnt signaling is independent of the underlying mutational landscape and has a common origin in dysregulation of PRH. Moreover, they suggest new therapeutic options based on the dependence of specific Wnt, Notch, and CDK4/6 inhibitors on PRH activity.

Significance:

The PRH/HHEX transcription factor is an oncogenic driver in cholangiocarcinoma that confers sensitivity to CDK4/6 inhibitors.

Cholangiocarcinoma (CCA) is a tumor of the bile duct epithelium with an unmet clinical need as this disease is usually advanced at diagnosis. For nonresectable disease, median survival is less than 12 months, and 5-year survival is around 2% (1). New treatment options and markers that allow early detection and/or inform personalized cancer treatment are therefore urgently needed. Risk factors for CCA include viral infection (hepatitis B and C), liver fluke infection, primary sclerosing cholangitis, cholestasis (loss of bile flow), and exposure to dietary toxins or metabolites, all of which lead to inflammation in and around the biliary tree. Chronic inflammation can provoke uncontrolled wound-healing responses involving the generation of DNA-damaging reactive oxygen and nitrogen species, activation of immune cells (particularly macrophages) and stroma, and aberrant activation of autocrine signaling, as well as activation of signaling pathways that promote epithelial–mesenchymal transition (EMT) and angiogenesis. All these events ultimately lead to cancer development and progression (2). Many of the dysregulated pathways involved in CCA are also involved in embryonic liver morphogenesis, such as the Wnt, TGFβ, Notch, Hedgehog, and Hippo pathways. Several of these pathways (including Wnt and Notch) can transiently promote EMT or the acquisition of mesenchymal features, such as increased cell migration and matrix invasion or loss of the epithelial cell–cell adhesion molecule E-cadherin (3, 4). In many cancers including mammary cancers, EMT is associated with an increased proportion of tumor-initiating cells (the so-called cancer stem cells or CSCs) that show increased self-renewal properties and increased chemoresistance (3). EMT and the CSC phenotype can be induced by any one of a common set of aberrantly expressed EMT transcription factors (e.g., Snail, Twist, Zeb) that, among many other things, directly repress the expression of E-cadherin (reviewed in ref. 5). CCA patient samples exhibit decreased E-cadherin expression and contain CSCs (6), suggesting that EMT may also be a feature of CCA.

Canonical Wnt signaling is critical for cell proliferation and contributes to the development of CCA in a mouse CCA carcinogenesis model (7, 8). In canonical Wnt signaling, Wnt ligands activate Frizzled family receptors and stabilize the β-catenin protein, which then enters the nucleus and activates transcription of Wnt target genes through interaction with T-cell-specific transcription factor/lymphoid enhancer binding factor (TCF/LEF) family transcription factors and coactivators (4). β-Catenin, by acting as both a coactivator for TCF/LEF transcription factors and a structural component of adherens junctions via interaction with E-cadherin, facilitates cross-talk between canonical Wnt signaling and cell–cell adhesion. Although there are few mutations in Wnt signaling components in CCA, there are high levels of nuclear β-catenin in the majority of CCA (7, 9). Importantly, pharmacologic inhibition of Wnt signaling decreases CCA formation in mouse models (7).

Notch signaling is also of central importance to the development of CCA (10). Notch signaling involves four transmembrane NOTCH receptors and two families of ligands, Serrate/Jagged (JAG-1, -2) and Delta-like (DLL-1, -3, -4), along with other proteins that transduce and regulate the signal. Upon ligand binding, Notch receptors are sequentially cleaved by an ADAM family protease and the γ-secretase protease complex. Cleavage releases the NOTCH intracellular domain (ICD), which interacts with the DNA-binding protein RBP-J/CSL and MAML1 coactivator, resulting in transcriptional activation of NOTCH target genes, including the HES and HEY family of genes encoding bHLH transcriptional repressors (reviewed in ref. 11). In hepatic regeneration models, Wnt and Notch signaling promote different cell fates; Notch signaling promotes biliary fate in hepatic progenitor cells, and Wnt signaling promotes hepatic specification (12). Dysregulation of Notch signaling in mouse hepatocytes through constitutive Notch1 ICD or Notch2 ICD expression can result in either hepatocellular carcinoma or CCA depending on the cooperating oncogene. NOTCH3 is not expressed in adult liver, but Notch3 expression is elevated in samples of patients with CCA, and knockout of Notch3 in a mouse model of CCA abrogates tumor growth (10). Notch3 ICD is known to be a key driver of CCA in part through activation of PI3K/Akt signaling (10). However, it is not known which genes the Notch3 ICD regulates nor is the interplay between Notch3 ICD signaling and Wnt signaling understood.

The proline-rich homeodomain protein/hematopoietically expressed homeobox (PRH/HHEX) protein is a transcription factor encoded by the HHEX gene that is required in the development of many tissue types including the liver and bile duct (reviewed in ref. 13), where PRH regulates hepatic (HNF4a) and biliary (Onecut1) transcription factors, respectively (14). PRH plays a growth-inhibitory role in hepatic regeneration (15) and when overexpressed in hepatocellular carcinoma cells PRH inhibited tumor growth in a mouse xenograft model (16, 17). PRH also exhibits tumor-suppressive properties in other epithelial lineages and in some hematopoietic lineages (13). Similarly, in breast and prostate cells, PRH inhibits cell migration and loss or inactivation of PRH induces EMT-like changes in cell morphology and behavior (18) and increases the proportion of CSC-like cells (19, 20). In stark contrast, PRH is involved in oncogenic transformation in at least two leukemic subtypes in which cytogenetic rearrangements promote dysregulated PRH expression (21–23). As PRH potentiates Wnt signaling during early embryonic development (24) and early liver development (25), it is of interest to understand whether PRH plays a role in CCA and whether it is involved in regulating Wnt and Notch signaling.

Here, we show for the first time that PRH plays an essential role in the maintenance of CCA. We also demonstrate that PRH promotes multiple features of tumor initiation and spread in primary untransformed biliary epithelial cells (BEC; cholangiocytes) isolated from human liver. Our work shows that a PRH-Notch3–positive feedback loop is a novel driver of CCA as both proteins collaborate to promote Wnt signaling. Further, we demonstrate that PRH expression levels can determine sensitivity or resistance to novel CCA chemotherapeutic strategies.

Cell culture and plasmids

CCLP-1, CCSW-1, AKN-1, KKU-100, and KKU-M213 cells (26, 27) were grown in DMEM with l-glutamine (Sigma-Aldrich, D5796), supplemented with 10% FBS and MEM nonessential amino acids (Sigma-Aldrich, M1745). CCLP-1, CCSW-1, and AKN-1 cells were authenticated in house by phenotyping using flow cytometry. KKU-100 and KKU-M213 (28) were obtained from the Japanese Cell Research Bank (JCRB1557 (KKU-213) and JCRB1568 (KKU-100)). HuCCA-1 (29) and RmCCA-1 (30) were kindly provided by the originators Professor Stitaya Sirisinha and Associate Professor Rutaiwan Tohtong, respectively. HuCCA-1 and RmCCA-1 were grown in Ham's F12 medium (Thermo Fisher IBR21041025) supplemented with 10% FBS. All cell lines were Mycoplasma tested every 3 months using EZ-PCR Mycoplasma Test Kit (Biological Industries). Primary BECs were isolated and grown as previously described (31). Stable cell lines were generated by transfecting PRH shRNA in pRS (Origene, TR312464), Notch3 shRNA in pLKO (Sigma-Aldrich, TRCN0000363316), CDH1 cDNA in pCDNA3 (hE-cadherin-pcDNA3 was a gift from Dr. Barry Gumbiner, Seattle Children's Hospital, Seattle, WA; Addgene plasmid #45769), or EGFP-PRH-myc in pEGFP-C1. Transient PRH knockdowns were performed using 100 nmol/L each of 4 HHEX-targeted siRNAs (Qiagen, 1027416). pEGFP-C1-PRH-Myc was generated by insertion of the human PRH cDNA between EcoRI and KpnI sites in pEGFP-C1. PCR was used to generate a PstI-KpnI fragment that replaced PRH coding sequence and placed an in-frame Myc tag followed by a translation stop codon at the end of the PRH coding sequence. This creates a double-tagged protein. Stable transfectants were selected for vector integration using either puromycin (pRS, pLKO) or G418 (pcDNA3, pEGFP-C1). For transient overexpression, PRH was overexpressed either using an adenoviral construct at multiplicity of infection (MOI) 50, or using pMUG1-myc-PRH (32). For transient Rb knockdown, we used 100 nmol/L of Rb-targeted siRNAs (Qiagen, 1027416).

EdU incorporation

Cells were plated at 104 cells per well in 96-well plates, and EdU incorporation was measured after 24 hours. For inhibitor experiments, inhibitors were added at the time of plating. Click-It EdU Microplate kit (Thermo Fisher, C10214) was used according to the manufacturer's instructions with the exception of replacement of the fluorescent kit substrate with the colorimetric peroxidase substrate o-Phenylenediamine (Sigma-Aldrich, P9187) to improve signal:noise. Cell lines were incubated with EdU for 2 hours and primary cells for 4 hours.

Immunohistochemistry

A formalin-fixed paraffin-embedded CCA tissue microarray (TMA) was purchased from Abcam (ab178201). Antigen retrieval was performed using 10 mmol/L citrate and 0.05% tween-20, pH 6.0, in a microwave on full power (800 W) for 30 minutes. The array was stained using an in-house polyclonal mouse anti-PRH antibody and the Vector ImmPRESS kit (Vector labs, MP-7402), and imaged on a Zeiss Axio Scan.Z1 microscope.

Immunocytochemistry

Cells were seeded onto poly-l-lysine–coated glass coverslips and left to adhere for 24 hours. Cells were washed with PBS and fixed using 4% formaldehyde in PBS for 10 minutes. Cells were permeabilized using 0.1% Triton X-100 in PBS for 10 minutes and then blocked in 5% BSA + 20% serum from the secondary antibody host species in PBS. Primary antibodies used were E-cadherin (Cell Signaling Technology, 24E10) and β-catenin (Cell Signaling Technology, D10A8). Secondary antibody was Texas Red conjugated anti-rabbit IgG (Sigma-Aldrich, SAB3700873). Cells were mounted using Prolong Gold with DAPI (Thermo Fisher, P36931) and imaged using a Zeiss LSM 780 confocal microscope.

Western blotting

The following primary antibodies were used for Western blotting: PRH (in-house mouse polyclonal; ref. 33), E-cadherin (Cell Signaling Technology, 24E10), lamin A/C (Santa Cruz Biotechnology, H-110), myc tag (Cell Signaling Technology, 9B11), Notch3 (Cell Signaling Technology, D11B8), vimentin (Cell Signaling Technology, D21H3), cyclin D2 (Cell Signaling Technology, D52F9), Rb (Cell Signaling Technology, 4H1), and phospho-Rb (Ser807/811; Cell Signaling Technology, 9308).

Quantitative RT-PCR

RNA was extracted using a Bioline Isolate II kit according to the manufacturer's instructions. Note that 1 μg of total RNA was used for reverse transcription (Quantitect Reverse Transcription Kit; Qiagen, 205311). qRT-PCR was performed in a Rotor-Gene Q cycler (Qiagen) using Quantitect SYBR Green PCR Kit (Qiagen, 204143). Genes of interest were normalized to β-actin expression using primer efficiency-normalized relative quantification, with primer efficiencies calculated from standard curves generated from cDNA dilutions. All primers are listed in Supplementary Table S1.

Mouse xenografts

All animal experiments and procedures were approved by the UK Home Office in accordance with the Animals (Scientific Procedures) Act 1986, and the Guide for the Care and Use of Laboratory Animals was followed. Note that 106 CCLP-1 PRH knockdown or scrambled control cells were resuspended in 100 μL Matrigel and injected s.c. into the flank of male CD-1 nude mice. Tumors were measured with a caliper every 3 days until the tumor reached 12 mm in diameter. Tumor volume was estimated using the following formula: [(length + width)/2]*length*width.

RNA sequencing

RNA was isolated using a Bioline Isolate II kit according to the manufacturer's instructions. Total polyadenylated RNA was purified and adapter ligated using Illumina TruSeq RNA Library Prep Kit according to the manufacturer's instructions. This was followed by Illumina sequencing of 75 bp paired end reads (minimum 2 × 35 million reads/sample). Each experimental condition was run in biological duplicate. Reads were quality-trimmed using TrimGalore!, aligned using the gapped read mapper TopHat, and differential expression analysis was performed using DESeq2. Gene Ontology (GO) analysis was performed using the PANTHER webserver (http://www.pantherdb.org), and Gene Set Enrichment Analysis (GSEA) for Hallmark gene sets was performed using the Broad Institute GSEA webserver (http://software.broadinstitute.org/gsea/msigdb). Raw and processed RNA sequencing (RNA-seq) data have been submitted to NCBI GEO database Accession no. GSE124429.

Chromatin immunoprecipitation sequencing

CCLP1 cells were infected with a recombinant adenovirus to express myc-PRH. Chromatin immunoprecipitation (ChIP) of myc-PRH (using the same antibody as for myc-tag Western blots) was carried out as previously described (34) with 1.5 × 107 CCLP1 cells infected with Ad-myc-PRH or empty adenovirus at MOI 50 for 48 hours. Sequencing libraries were prepared using NEBNext Ultra II DNA Library preparation kit, followed by Illumina sequencing of 100 bp reads (minimum 20 million reads/sample). Reads were quality-trimmed using TrimGalore!, aligned using Bowtie2, and peaks were called using MACS2. Chromatin prepared from empty adenovirus-infected cells and subjected to myc tag immunoprecipitation was used as background in sequencing and peak calling. De novo motif analysis of ChIP peaks was performed using HOMER. Raw and processed ChIP-seq data have been submitted to NCBI GEO database Accession no. GSE124430.

TCF/LEF reporters

TCF/LEF-dependent transcriptional activity was measured using the TOPFlash firefly luciferase reporter system, in which the firefly luciferase gene is downstream of several TCF/LEF consensus-binding sites; the same construct with scrambled TCF/LEF sites (FOPflash) is used to determine the TCF/LEF-independent activity of the promoter. Firefly luciferase activity was normalized to Renilla luciferase activity from a constitutively active promoter (pRL, Promega, E2261). For overexpression experiments, reporter constructs were cotransfected with either myc-PRH, myc-PRH N187A DNA-binding deficient mutant (32), or 3xFLAG-Notch3-ICD (hNICD3(3xFLAG)-pCDF1-MCS2-EF1-copGFP), a gift from Dr. Brenda Lilly (Abigail Wexner Research Institute, Columbus, OH), Addgene plasmid #40640) expression vectors.

Soft-agar colony formation

Colony formation assays and subsequent imaging were done as described by Borowicz and colleagues (35). Note that 103 cells were plated per well, and colonies were imaged after 10 days.

Transwell migration

Cells were starved overnight in serum-free medium with 0.2% BSA and 1 mmol/L hydroxyurea. Twenty-four–well ThinCert inserts (Greiner bio-one) were placed in the wells of a 24-well plate. Six hundred microliter DMEM with 10% serum was added to each. Two hundred microliter of serum-free medium with 0.2% BSA containing 2 × 105 cells was added to each insert. After 72 hours, medium was replaced with 450 μL of 8 μmol/L calcein-AM in DMEM with 10% serum. After 45-minute incubation, inserts were transferred to a fresh 24-well plate containing 500 μL prewarmed Trypsin-EDTA in each well. Two hundred microliter of the Trypsin-EDTA cell suspension was transferred to a black flat bottom 96-well plate. The fluorescence signal was read in a fluorescence plate reader at an excitation wavelength of 485 nm and an emission wavelength of 520 nm.

Matrigel invasion

For invasion assays, insert transwells with 8.0 μm polycarbonate membrane (Corning Costar) were coated with 50 μL of a 1:10 mixture of Matrigel (BD Biosciences) in serum-free medium. Cells were infected with Ad-empty or Ad-PRH (MOI 50) and 24 hours after infection cells incubated with 1 mmol/L hydroxyurea. Forty-eight hours after infection, 104 cells were plated per well in serum-free media containing 1 mmol/L hydroxyurea (200 μL) and left to invade towards complete medium (500 μL) in the bottom chamber for 24 hours. Cells remaining on underside of insert after swabbing with a cotton swab were fixed in methanol for 5 minutes and stained with 0.1%w/v crystal violet in 12% glutaraldehyde in water for 5 minutes and counted by microscopy.

PRH is highly expressed in CCA

Analysis of the CCA dataset (n = 45) in The Cancer Genome Atlas (TCGA) revealed elevated expression of PRH mRNA in CCA samples compared with all TCGA samples, with a mean log2-fold change of 3.2±0.9 (Supplementary Fig. S1A), and high PRH expression is limited to a small number of tumor types (Supplementary Fig. S1B). In the majority of CCA samples, elevated PRH mRNA expression is in the absence of gene amplification, and in several samples, the transcript is elevated despite a single allele deletion. No coding mutations in the HHEX gene encoding PRH were detected in any CCA samples. To examine PRH protein staining intensity in CCA samples, we performed IHC on a TMA containing 2 tumor cores and a noninvolved border core from 42 patients with CCA (Abcam). Representative IHC images of carcinomas and patient-matched border bile ducts are shown in Fig. 1A. Of the 42 noninvolved border cores, 26 had at least one bile duct. In a paired analysis of these 26 samples compared with their matched carcinoma, we found increased PRH staining intensity in 20 of 26 (P = 0.00005), and in an unpaired analysis comparing all 42 carcinomas with the 26 noninvolved bile ducts, we found a 20% increase in median PRH expression (P = 0.0004, Fig. 1B). We did not find any differences in PRH staining intensity between different grades and tumor–node–metastasis stages of carcinoma, although this may reflect lack of statistical power as the majority of our samples were from grade II tumors. We next examined PRH protein expression in four CCA cell lines compared with immortalized AKN-1 BECs and two independently isolated primary human BEC cultures derived from livers with alcoholic liver disease and steatotic liver disease (NASH). All four CCA cell lines showed increased PRH protein expression (Fig. 1C), and we conclude that PRH mRNA and protein is highly expressed in CCA compared with primary BECs.

Figure 1.

PRH is overexpressed in CCA. A, Representative immunostaining of PRH in a CCA TMA. Arrowheads, border tissue biliary epithelium. B, Quantification of PRH staining intensity in noninvolved bile ducts compared with CCA. C, Western blot showing PRH expression in four human CCA cell lines, two human primary BEC isolates, and an immortalized cholangiocyte cell line. D, Growth of CCLP1 control (34±17 mm3) and PRH knockdown tumors (176±27 mm2) in nude mouse xenografts; n = 9 per group; P = 0.0006 at day 25. E, Proportion of colony-initiating cells in soft agar for PRH knockdown (3.4±0.4%) and control (7.6±0.5%) CCLP1 cells; n = 3; P = 0.01. F, Final cross-sectional area of colonies in soft agar for PRH knockdown (2,063±103 μm2) and control (4,783±53 μm2) CCLP1 cells; n = 3; P = 0.002. G and H, As in E and F for AKN1 cells. I, Colony formation in soft agar of primary BECs infected with Ad myc-PRH or empty adenovirus. *, P < 0.05.

Figure 1.

PRH is overexpressed in CCA. A, Representative immunostaining of PRH in a CCA TMA. Arrowheads, border tissue biliary epithelium. B, Quantification of PRH staining intensity in noninvolved bile ducts compared with CCA. C, Western blot showing PRH expression in four human CCA cell lines, two human primary BEC isolates, and an immortalized cholangiocyte cell line. D, Growth of CCLP1 control (34±17 mm3) and PRH knockdown tumors (176±27 mm2) in nude mouse xenografts; n = 9 per group; P = 0.0006 at day 25. E, Proportion of colony-initiating cells in soft agar for PRH knockdown (3.4±0.4%) and control (7.6±0.5%) CCLP1 cells; n = 3; P = 0.01. F, Final cross-sectional area of colonies in soft agar for PRH knockdown (2,063±103 μm2) and control (4,783±53 μm2) CCLP1 cells; n = 3; P = 0.002. G and H, As in E and F for AKN1 cells. I, Colony formation in soft agar of primary BECs infected with Ad myc-PRH or empty adenovirus. *, P < 0.05.

Close modal

PRH promotes tumor growth by CCLP1 cells in nude mice

To determine whether elevated PRH is important in CCA, we generated stable PRH knockdown (KD) in CCLP1 CCA cells. Western blotting and quantitative RT-PCR demonstrated effective knockdown of PRH (Supplementary Fig. S1C and S1D). One million CCLP1 control and PRH knockdown cells were then injected s.c. into nude mice, and tumor size measured over 25 days and at the termination of the experiment. Only 3 of 9 mice injected with PRH knockdown cells produced tumors compared with 9 of 9 controls (P = 0.003), and knockdown of PRH expression significantly reduced tumor growth (Fig. 1D). PRH knockdown also reduced the proportion of CCLP1 cells able to form colonies in soft agar (Fig. 1E) and the average colony cross-sectional area (Fig. 1F). To test whether overexpression of PRH in BECs would be sufficient to recapitulate the phenotype observed in CCLP1 cells, AKN1-immortalized BECs were transfected with a double-tagged GFP-PRH-myc expression vector and selected to generate a stable GFP-PRH-myc–expressing cell line. In addition, we overexpressed PRH in primary BECs using an adenovirus-expressing myc-tagged PRH. Colony formation in soft agar was increased in an AKN1 cell population overexpressing PRH compared with controls (Fig. 1G and H) and strikingly, when expressing myc-PRH multiple independently isolated primary BEC populations from different donors were able to form colonies in soft agar whereas controls were not (Fig. 1I). We conclude that elevated PRH expression promotes anchorage-independent growth of primary and immortalized BECs in vitro and that reduction of PRH levels inhibits CCLP1 tumor growth in a xenograft model.

PRH drives CCA cell proliferation

To examine why depletion of PRH in CCA cells decreases tumor growth in nude mice, we examined the effect of PRH depletion and PRH overexpression on cell proliferation in culture. Knockdown of PRH in CCLP1 cells reduced cell growth, with the doubling time increasing from 28±3 to 40±4 hours (Supplementary Fig. S1E). To determine whether this was due to reduced proliferation or increased cell death, we measured proliferation by EdU incorporation and apoptosis by caspase-3 enzymatic activity assay. EdU incorporation was reduced to 53±7% of control with PRH shRNA knockdown (Fig. 2A). There was a small decrease in caspase-3 activity in PRH knockdown cells compared with controls (Supplementary Fig. S1F). We conclude that the reduction in cell number on PRH knockdown was predominantly a result of decreased cell proliferation. To determine whether this was a CCLP1-specific effect, or an off-target effect of the PRH shRNA, we also measured EdU incorporation in KKU-M055 and KKU-M213 CCA cells 72 hours after transfection with a PRH siRNA with a different target sequence to the shRNA. EdU incorporation was reduced in both cell types (Supplementary Fig. S1G) ruling out cell line–specific effects and off-target effects of the shRNA. To confirm that PRH promotes the proliferation of CCA cells, we overexpressed PRH in CCLP1 and CCSW1 CCA cells using a recombinant adenovirus encoding myc-PRH (Supplementary Fig. S2A). Overexpression of myc-PRH in both cell lines increased EdU incorporation (Fig. 2B) and increased growth rate, with doubling times decreased from 30±3 to 22±1 hours and 29±4 to 19±3 hours, respectively (Supplementary Fig. S2B). Caspase-3 activity was also robustly decreased (Supplementary Fig. S2C). Proliferation measured by EdU incorporation was increased by PRH overexpression in both primary BECs and AKN1 cells (Fig. 2C). We conclude that PRH overexpression promotes the proliferation of primary BECs and CCA cell lines and reduces basal levels of apoptosis.

Figure 2.

Effects of PRH manipulation on CCA cell biology. A, Proliferation of CCLP1 cells stably transfected with PRH shRNA or scrambled control; n = 3; P = 0.03. B, Proliferation of CCLP1 and CCSW1 cells infected with Ad myc-PRH or empty virus control; n = 3; P = 0.03 (CCLP1); P = 0.02 (CCSW1). C, Proliferation of AKN1 and primary BECs overexpressing GFP-PRH-myc (stable) or myc-PRH (transient, 48 hours); n = 3; P = 0.003 (AKN1); P = 0.006 (BEC). D, Morphology of CCLP1 cells stably transfected with PRH shRNA or scrambled control. E, Western blot showing increased expression of E-cadherin protein and decreased expression of vimentin protein after PRH knockdown. Lamin A/C was used as loading control. F, Migration of CCLP1 PRH knockdown cells through transwell filters in a 10% serum gradient; n = 3; P = 0.03. G, Invasion of CCLP1 cells through Matrigel; n = 4; P = 0.03. H, Western blotting for myc-PRH and EMT-associated proteins E-cadherin and vimentin in AKN1 cells and primary BECs. I and J, As in F and G for AKN1 cells and primary BECs. K, Morphology of equal numbers of AKN1 cells stably transfected with plasmids expressing GFP (control) or a GFP-PRH-Myc–tagged fusion protein (GFP-PRH-Myc). Scale bars, 50 μm in length. *, P < 0.05.

Figure 2.

Effects of PRH manipulation on CCA cell biology. A, Proliferation of CCLP1 cells stably transfected with PRH shRNA or scrambled control; n = 3; P = 0.03. B, Proliferation of CCLP1 and CCSW1 cells infected with Ad myc-PRH or empty virus control; n = 3; P = 0.03 (CCLP1); P = 0.02 (CCSW1). C, Proliferation of AKN1 and primary BECs overexpressing GFP-PRH-myc (stable) or myc-PRH (transient, 48 hours); n = 3; P = 0.003 (AKN1); P = 0.006 (BEC). D, Morphology of CCLP1 cells stably transfected with PRH shRNA or scrambled control. E, Western blot showing increased expression of E-cadherin protein and decreased expression of vimentin protein after PRH knockdown. Lamin A/C was used as loading control. F, Migration of CCLP1 PRH knockdown cells through transwell filters in a 10% serum gradient; n = 3; P = 0.03. G, Invasion of CCLP1 cells through Matrigel; n = 4; P = 0.03. H, Western blotting for myc-PRH and EMT-associated proteins E-cadherin and vimentin in AKN1 cells and primary BECs. I and J, As in F and G for AKN1 cells and primary BECs. K, Morphology of equal numbers of AKN1 cells stably transfected with plasmids expressing GFP (control) or a GFP-PRH-Myc–tagged fusion protein (GFP-PRH-Myc). Scale bars, 50 μm in length. *, P < 0.05.

Close modal

PRH maintains the mesenchymal phenotype of CCA cells

Following PRH knockdown in culture, we noticed a marked change in morphology of CCLP1 cells from an elongated mesenchymal-like morphology to an epithelial-like morphology (Fig. 2D). Unlike normal BECs, CCLP1 cells do not express the epithelial cell adhesion molecule E-cadherin. However, E-cadherin expression was restored upon PRH knockdown (Fig. 2E), and the mesenchymal marker protein vimentin was strongly decreased in these cells (Fig. 2E, lower). As changes in cell morphology and E-cadherin expression are associated with mesenchymal to epithelial transition (MET), we measured the migration and invasion of PRH-depleted CCLP1 cells in response to a serum gradient. Proliferation was inhibited by hydroxyurea treatment in both experiments. PRH knockdown reduced the number of migrated cells (Fig. 2F) and reduced the number of invaded cells (Fig. 2G). Overexpression of PRH in the presence of hydroxyurea increased the invasion of both CCLP1 and CCSW tumor cell lines (Supplementary Fig. S2D). Overexpression of PRH in both AKN1 and primary human BECs reduced E-cadherin expression and increased expression of vimentin (Fig. 2H). In addition, PRH overexpression increased cell migration and matrix invasion by both primary BECs and AKN1 cells (Fig. 2I and J). As might be expected based on these results, PRH expression in AKN1 cells resulted in a change from an epithelial-like cell morphology to a more mesenchymal cell morphology (Fig. 2K). We therefore examined changes in the expression of EMT-related genes in AKN1 cells overexpressing PRH using qRT-PCR. The epithelial and mesenchymal marker genes CDH1 and VIM encoding E-cadherin and vimentin were downregulated and upregulated, respectively, following PRH expression (Supplementary Fig. S2E). In addition, the genes encoding EMT transcription factors (ZEB1 and TWIST1) were upregulated in AKN1 cells expressing PRH. Thus, PRH overexpression in primary BECs and in an immortalized BEC cell line suppresses the epithelial phenotype and promotes a migratory mesenchymal phenotype.

PRH regulates pathways associated with Wnt signaling and with EMT

To understand how PRH promotes the proliferation of CCA cells and maintains a mesenchymal phenotype, we performed RNA-seq with total polyadenylated RNA isolated from CCLP1 cells with stably knocked down or transiently overexpressed PRH. We found 189 downregulated and 430 upregulated genes in PRH knockdown cells, and 889 downregulated and 1,410 upregulated genes in PRH overexpressing cells when compared with their respective controls (Fig. 3A). To validate the RNA-seq data, we performed qRT-PCR for several genes identified as differentially expressed, all of which were in agreement with the RNA-seq (Supplementary Fig. S3A). GSEA using the Molecular Signatures Database (MSigDB) Hallmark gene sets (36) showed that both PRH knockdown and PRH overexpression led to the differential expression of genes within the same pathways, many of which have been shown to be aberrantly activated in CCA including EMT, Wnt/β-catenin, TGFβ, IL6/JAK/STAT3, and estrogen signaling (Fig. 3B and C). GO term enrichment analysis was in broad agreement with GSEA analysis and suggested enrichment of genes associated with epithelial and mesenchymal differentiation as well as genes associated with both canonical and noncanonical Wnt signaling (Supplementary Fig. S3B and S3C). We also noted changes in the expression of multiple transcription factors and markers associated with biliary differentiation and Notch signaling (Fig. 3D). These results are suggestive of alteration of the transcriptional network underlying biliary differentiation upon PRH knockdown.

Figure 3.

RNA-seq of PRH knockdown and overexpressing CCLP1 cells. A, Number of DEGs detected in PRH KD and overexpression (OE) experiments, and overlap of these gene sets. B and C, GSEA using Hallmark gene sets for PRH KD and OE DEGs. Red bars represent Hallmark sets that are enriched in both PRH KD and OE DEG lists. D, Log2-fold change of gene expression from PRH OE and KD RNA-seq for BEC-related genes. TF, transcription factor.

Figure 3.

RNA-seq of PRH knockdown and overexpressing CCLP1 cells. A, Number of DEGs detected in PRH KD and overexpression (OE) experiments, and overlap of these gene sets. B and C, GSEA using Hallmark gene sets for PRH KD and OE DEGs. Red bars represent Hallmark sets that are enriched in both PRH KD and OE DEG lists. D, Log2-fold change of gene expression from PRH OE and KD RNA-seq for BEC-related genes. TF, transcription factor.

Close modal

In addition, GO analysis identified enrichment of genes associated with the control of cell proliferation in both the knockdown and overexpression experiments, including genes encoding cyclin D2 (CCND2), and the p27 (CDKN1B) and p15 (CDKN2B) cyclin-dependent kinase (CDK) inhibitors (Supplementary Fig. S3B and S3C), and GSEA analysis showed enrichment of the G2M gene set, which contains cell-cycle genes (Fig. 3C). Of particular interest in the GSEA data was the finding that Notch and PI3K/Akt pathway genes were enriched in the PRH overexpression gene set (Fig. 3C) as NOTCH3 oncogenic activity in CCA models is at least partly due to noncanonical NOTCH signaling via PI3K/Akt (10). Furthermore, the NOTCH3 gene was one of the 53 genes that were differentially regulated in both PRH overexpression and knockdown data sets, and both qRT-PCR and Western blotting showed reduced NOTCH3 gene expression and Notch3 ICD protein expression in PRH knockdown cells (Figs. 3D and 4A, respectively).

Figure 4.

Notch3 expression is regulated by PRH. A, Western blot showing Notch3 ICD protein expression in PRH knockdown CCLP1 cells. B,NOTCH3 gene expression in CCLP1 and CCSW1 cells infected with Ad myc-PRH, Ad myc-PRH DNA-binding deficient N187A mutant, or empty virus control. C, Western blot for CCLP1 samples in B. D, Notch3 ICD rotein expression correlates with PRH protein expression in four human CCA cell lines, two human primary BEC isolates, and an immortalized BEC line. E, Western blot showing elevated expression of Notch3 ICD upon PRH overexpression in both AKN1 cells and primary BECs. F, Western blot showing increased expression of E-cadherin and reduced expression of PRH proteins after Notch3 knockdown. Lamin A/C was used as loading control. G, Proliferation of CCLP1 cells stably transfected with Notch3 shRNA or scrambled control; n = 3; P = 0.04. H, Morphology of CCLP1 cells stably transfected with Notch3 shRNA or scrambled control. I, DEGs detected in Notch3 KD compared with PRH KD experiments. Hypergeometric test, P = 10−229 for upregulated genes and P = 10−31 for downregulated genes. J, Hallmark GSEA of genes differentially expressed after Notch3 KD. Red bars, gene sets that are also enriched after PRH knockdown. *, P < 0.05.

Figure 4.

Notch3 expression is regulated by PRH. A, Western blot showing Notch3 ICD protein expression in PRH knockdown CCLP1 cells. B,NOTCH3 gene expression in CCLP1 and CCSW1 cells infected with Ad myc-PRH, Ad myc-PRH DNA-binding deficient N187A mutant, or empty virus control. C, Western blot for CCLP1 samples in B. D, Notch3 ICD rotein expression correlates with PRH protein expression in four human CCA cell lines, two human primary BEC isolates, and an immortalized BEC line. E, Western blot showing elevated expression of Notch3 ICD upon PRH overexpression in both AKN1 cells and primary BECs. F, Western blot showing increased expression of E-cadherin and reduced expression of PRH proteins after Notch3 knockdown. Lamin A/C was used as loading control. G, Proliferation of CCLP1 cells stably transfected with Notch3 shRNA or scrambled control; n = 3; P = 0.04. H, Morphology of CCLP1 cells stably transfected with Notch3 shRNA or scrambled control. I, DEGs detected in Notch3 KD compared with PRH KD experiments. Hypergeometric test, P = 10−229 for upregulated genes and P = 10−31 for downregulated genes. J, Hallmark GSEA of genes differentially expressed after Notch3 KD. Red bars, gene sets that are also enriched after PRH knockdown. *, P < 0.05.

Close modal

A PRH-Notch3–positive feedback loop promotes cell proliferation and EMT

To determine whether NOTCH3 expression was increased by PRH in other CCA cell lines, we overexpressed myc-PRH and a DNA-binding–deficient myc-PRH N187A mutant in CCLP1 and CCSW1 cells. Wild-type PRH but not the DNA-binding–deficient mutant increased NOTCH3 mRNA (Fig. 4B) and protein levels in these cells (Fig. 4C). Furthermore, we found a striking correlation between PRH expression and Notch3 expression in cell lines and primary BECs (Fig. 4D). In addition, HHEX and NOTCH3 gene expression positively correlate in the TCGA CCA RNA-seq dataset as do HHEX and HES1 gene expression (Supplementary Fig. S3D and S3E), suggesting that PRH regulates NOTCH3 and HES1 in primary tumors. Finally, overexpression of PRH in both primary BECs and AKN1 cells led to expression of the Notch3 protein (Fig. 4E).

To determine whether the effects of PRH knockdown are recapitulated by depletion of Notch3, we generated Notch3 knockdown CCLP1 cells by integrating a Notch3 shRNA plasmid (Fig. 4F). Strikingly, the proliferative and morphologic phenotype of PRH knockdown cells (Fig. 2A and D) was reproduced by Notch3 knockdown (Fig. 4G and H). Moreover, knockdown of Notch3 resulted in a reduction of HHEX mRNA (Supplementary Fig. S3F) and PRH protein (Fig. 4F). These data suggest that PRH and Notch3 could form a positive transcriptional feedback loop where each regulates the other.

To better understand the transcriptome changes underlying the common phenotype between Notch3 and PRH knockdown cell lines, we performed RNA-seq with Notch3 knockdown and scrambled shRNA cell lines and compared differentially expressed genes (DEG) in PRH knockdown cells with those in Notch3 knockdown cells. These experiments clearly showed that both factors act in the same pathways as the DEGs after PRH depletion were a subset of those differentially expressed after Notch3 knockdown (see Venn diagrams, Fig. 4I). In agreement with this finding, the gene sets enriched in the Notch3 knockdown DEGs were largely the same sets enriched in PRH knockdown (red bars, Fig. 4J), with a small number of Notch3-specific sets (black bars) including PI3K/Akt signaling and Notch signaling. In agreement with the GSEA, Western blotting experiments showed that both PRH knockdown and Notch3 knockdown reduced phosphorylation of Akt at both T308 and S473 (both of which are known to increase kinase activity, T308 by 100-fold and S473 by a further 10-fold for full kinase activity; ref. 37; Supplementary Fig. S3G).

Notch3-dependent and Notch3-independent PRH target genes

To separate PRH-regulated genes into Notch3-dependent (Notch correlated) and Notch3-independent (PRH correlated) subgroups, we overexpressed PRH in Notch3 knockdown CCLP1 cells. Figure 5A shows qRT-PCR analyses examining representative PRH-correlated and Notch3-correlated genes. Expression of the PRH regulated EMT-associated genes CDH1 and VIM, as well as CCND2 was regulated by Notch3 rather than by PRH alone, and this result was reproduced at protein level (Fig. 5B). EdU incorporation experiments after overexpression of PRH in Notch3 knockdown cells suggested that PRH was unable to drive proliferation when it was decoupled from activation of Notch3 expression (Fig. 5C). Interestingly, we noted a reduced level of PRH mRNA and protein expression from the adenoviral PRH construct in the Notch3 knockdown cells compared with the control (Fig. 5A and B), suggesting that Notch3 may regulate PRH expression at the level of transcript stability. We found the same effects on proliferation and expression of genes and proteins shown in Fig. 5A and B when the viral MOI was increased in the Notch3 knockdown cells to give equal PRH protein expression, and we therefore proceeded to perform RNA-seq in the presence of equal MOIs. Across all four sample combinations, we found 5,397 DEGs that were correlated with either PRH or Notch3 expression. Of these, 4,356 were Notch3-correlated (2,110 positively correlated, i.e., activated downstream of Notch3, and 2,246 negatively correlated, i.e., repressed downstream of Notch3), and 1,041 were PRH-correlated (604 positively and 437 negatively). We found strong enrichment of c-myc target genes when we performed GSEA analysis on the 1,041 PRH-correlated genes (Fig. 5D, bottom plot). In addition to c-myc target genes, we found enrichment of a variety of gene sets (including Wnt/β-catenin signaling) that are also enriched in the Notch3-correlated gene set, suggesting that PRH and Notch3 regulate different genes in the same pathways (Fig. 5D). Finally, we present summary heat maps with hierarchical clustering analysis of all RNA-seq samples showing PRH dependence, Notch dependence or codependence of genes (Supplementary Fig. S4A), and heat maps with hierarchical clustering analysis of RNA-seq samples for selected pathways (Supplementary Fig. S4B).

Figure 5.

Notch3- and PRH-correlated gene sets. A, qRT-PCR analysis of genes from CCLP1 cells overexpressing PRH in the presence or absence of Notch3 shRNA identifying PRH and Notch3-correlated expression signatures. B, Western blot analysis of EMT proteins E-cadherin and vimentin and cyclin D2 in CCLP1 cells overexpressing myc-PRH in the presence of absence of Notch3 shRNA. C, Proliferation of CCLP1 cells overexpressing myc-PRH in the presence or absence of Notch3 shRNA. *, P < 0.05 after Bonferroni correction, compared with nontargeting shRNA/empty virus control. #, no statistically significant difference in the comparison indicated. D, Hallmark GSEA of Notch3-correlated and PRH-correlated gene sets identified from analysis of RNA-seq data. Red bars, gene sets enriched in both PRH- and Notch3-correlated sets. E, Overlap of genes with PRH-binding sites determined by ChIP-seq and genes that are differentially expressed after PRH overexpression determined by RNA-seq in CCLP1 cells. F, Comparison of the primary motif underlying PRH ChIP-seq peaks identified using HOMER with the PRH SELEX motif (derived from ref. 38). G, RNA-seq and ChIP-seq tracks of putative direct PRH target DKK1. Red tracks, myc-PRH overexpression. H, RNA-seq and ChIP-seq tracks of putative direct PRH target WNT11. Red tracks, myc-PRH overexpression.

Figure 5.

Notch3- and PRH-correlated gene sets. A, qRT-PCR analysis of genes from CCLP1 cells overexpressing PRH in the presence or absence of Notch3 shRNA identifying PRH and Notch3-correlated expression signatures. B, Western blot analysis of EMT proteins E-cadherin and vimentin and cyclin D2 in CCLP1 cells overexpressing myc-PRH in the presence of absence of Notch3 shRNA. C, Proliferation of CCLP1 cells overexpressing myc-PRH in the presence or absence of Notch3 shRNA. *, P < 0.05 after Bonferroni correction, compared with nontargeting shRNA/empty virus control. #, no statistically significant difference in the comparison indicated. D, Hallmark GSEA of Notch3-correlated and PRH-correlated gene sets identified from analysis of RNA-seq data. Red bars, gene sets enriched in both PRH- and Notch3-correlated sets. E, Overlap of genes with PRH-binding sites determined by ChIP-seq and genes that are differentially expressed after PRH overexpression determined by RNA-seq in CCLP1 cells. F, Comparison of the primary motif underlying PRH ChIP-seq peaks identified using HOMER with the PRH SELEX motif (derived from ref. 38). G, RNA-seq and ChIP-seq tracks of putative direct PRH target DKK1. Red tracks, myc-PRH overexpression. H, RNA-seq and ChIP-seq tracks of putative direct PRH target WNT11. Red tracks, myc-PRH overexpression.

Close modal

Identification of PRH-binding sites in CCA cells

To identify putative direct targets of PRH, we determined the genome-wide binding sites of PRH in CCLP1 cells using ChIP-seq with a myc-tag antibody and chromatin prepared from myc-PRH or empty adenovirus-infected CCLP1 cells and assigned ChIP peaks to the nearest transcription start site within 100 kb. Comparison of these putative direct PRH target genes with the DEGs from the myc-PRH overexpression RNA-seq experiment suggests that of the 1,410 upregulated genes, only 143 (10.1%) were direct targets, whereas of the 889 downregulated genes, 397 (44.7%) were direct PRH targets (see Venn diagram, Fig. 5E). These data produced the first consensus-binding site identified for PRH in cells (Fig. 5F); the most strongly enriched motif (64.2% of peaks, P = 10−96) underlying PRH ChIP-seq peaks contains a core ATTA motif characteristic of homeodomain transcription factor–binding sites (Fig. 5F) and is in good agreement with the avian PRH-binding site determined in vitro by SELEX (38). Variants of this sequence have been identified upstream of several PRH-regulated genes in other cell types (39). Analysis of the putative directly regulated genes using GSEA revealed enrichment of genes associated with apoptosis, proliferation (mitotic spindle and G2–M checkpoint), EMT, and signaling pathways including IL6, IL2, p53, and TNFα (Supplementary Fig. S5A). Multiple genes associated with EMT including genes involved in cell adhesion or cell migration were present in these groups; however, these targets do not include the SNAI/TWIST/ZEB families of EMT transcription factors that regulate CDH1 or CDH1 itself. PRH did not directly bind near NOTCH3 and thus PRH may regulate Notch pathway genes by indirect means. However, CDKN1B and CDKN2B genes (Supplementary Fig. S5B and S5C) and a number of genes associated with Wnt signaling including DKK1, WNT11 (Fig. 5G and H), TCF7L1, and WNT16 had nearby PRH ChIP peaks.

Notch3-dependent and Notch3-independent effects on Wnt signaling

To determine whether PRH regulates Wnt signaling in these cells, we made use of the TOPFlash TCF/LEF luciferase Wnt-signaling reporter system (40). Transient transfection of plasmids encoding myc-PRH into both CCLP1 and CCSW1 cells significantly increased TOPFlash activity compared with the empty vector control (Fig. 6A) or after overexpression of the DNA-binding–deficient N187A PRH mutant (Fig. 6A, inset). Transfection of FLAG-Notch3-ICD did not robustly alter the activity of the TOPFlash reporter compared with empty vector (Fig. 6A) despite robust protein expression (Supplementary Fig. S3H) and changes in the expression of Notch3 target genes (HEY2 and CCND2; Supplementary Fig. S3I). This suggests that genes directly downstream of PRH rather than genes regulated by Notch3 are crucial for controlling the output of the canonical Wnt signaling pathway in these cells.

Figure 6.

Regulation of Wnt signaling. A, TOPFlash TCF/LEF reporter activity in CCLP1 and CCSW1 cells expressing myc-PRH, DNA-binding–deficient N187A mutant of myc-PRH and Flag-Notch3-ICD. Inset, Western blot showing expression of myc-PRH constructs. B, Representative immunofluorescence micrographs of CCLP1 PRH knockdown cells stained for E-cadherin and β-catenin. C, Western blot of subcellular fractions of CCLP1 PRH knockdown cells. D, TOPFlash reporter activity in PRH knockdown CCLP1 cells. E, TOPFlash reporter activity in control CCLP1 cells (pcDNA empty) and CCLP1 cells overexpressing E-cadherin (CDH1) in the presence and absence of myc-PRH expression. Inset, Western blot showing successful overexpression of E-cadherin. *, P < 0.05; **, P < 0.01; #, no statistically significant difference.

Figure 6.

Regulation of Wnt signaling. A, TOPFlash TCF/LEF reporter activity in CCLP1 and CCSW1 cells expressing myc-PRH, DNA-binding–deficient N187A mutant of myc-PRH and Flag-Notch3-ICD. Inset, Western blot showing expression of myc-PRH constructs. B, Representative immunofluorescence micrographs of CCLP1 PRH knockdown cells stained for E-cadherin and β-catenin. C, Western blot of subcellular fractions of CCLP1 PRH knockdown cells. D, TOPFlash reporter activity in PRH knockdown CCLP1 cells. E, TOPFlash reporter activity in control CCLP1 cells (pcDNA empty) and CCLP1 cells overexpressing E-cadherin (CDH1) in the presence and absence of myc-PRH expression. Inset, Western blot showing successful overexpression of E-cadherin. *, P < 0.05; **, P < 0.01; #, no statistically significant difference.

Close modal

β-Catenin facilitates cross-talk between canonical Wnt signaling and cell–cell adhesion by acting as a coactivator for TCF/LEF transcription factors and a structural component of the adherens junctions complex that includes E-cadherin (41). Although there was no change of β-catenin gene expression after PRH overexpression or knockdown, we observed a large increase in E-cadherin expression after PRH knockdown in CCLP1 cells (Fig. 2E) and decreased E-cadherin expression after PRH overexpression in BECs (Fig. 2H). PRH knockdown reduced β-catenin nuclear localization compared with the control as measured by immunofluorescence microscopy (Fig. 6B) and densitometry of Western blots for β-catenin following subcellular fractionation (Fig. 6C). TCF/LEF reporter activity in PRH knockdown CCLP1 cells was also decreased (Fig. 6D). To determine whether the increased expression of E-cadherin could explain the reduction in TCF/LEF transcriptional activity in PRH KD cells, we restored E-cadherin protein expression independently of EMT transcription factors in CCLP1 cells by generating a stable cell line expressing the CDH1 gene under the control of the CMV promoter. Expression of E-cadherin decreased TOPFlash reporter activity (CDH1 empty TOP compared with pcDNA empty TOP in Fig. 6E), suggesting that the increase in E-cadherin seen on PRH depletion (and consequent decrease in β-catenin nuclear localization) is likely to be responsible for decreased Wnt signaling. Moreover, E-cadherin overexpression significantly reduced the ability of PRH to increase TOPFlash reporter activity (Fig. 6E). Taken together, these data indicate that PRH directly promotes aberrant expression of Wnt pathway genes, leading to activation of Wnt-responsive transcription independently of Notch3. In addition, repression of CDH1 downstream of PRH via Notch3 amplifies the aberrant PRH-dependent Wnt signal. Thus, PRH and Notch3 exert regulation over Wnt signaling at multiple levels.

PRH overexpression and sensitivity to chemotherapeutic drugs

As PRH regulated several components of the Notch signaling pathway including γ-secretase components, we investigated whether PRH could modulate the inhibition of cell proliferation by γ-secretase inhibition. Treatment with the γ-secretase inhibitor DAPT had a greater inhibitory effect on the proliferation of CCLP1 cells than it did on the proliferation of CCLP1 cells overexpressing PRH (Fig. 7A), indicating that PRH overexpression increases resistance. To summarize these data, we calculated a log ratio of sensitivity between CCLP1 and PRH overexpressing CCLP1, as log2((PPRH,drug/PPRH,vehicle)/(Pempty,drug/Pempty,vehicle)) where P is the relative proliferative rate measured by EdU incorporation (Fig. 7B). The log ratio is negative if PRH expression increases sensitivity to a drug, and positive if it increases resistance. PRH also drives canonical Wnt signaling in CCA cell lines, and we wondered whether the antiproliferative effect of Wnt pathway inhibition previously reported in CCA cell lines (7) would also be modulated by the PRH expression level. Treatment with the β-catenin/CBP interaction inhibitor ICG-001 had a greater inhibitory effect on the proliferation of CCLP1 cells than it did on the proliferation CCLP1 cells overexpressing PRH (Fig. 7A and B and dose–response curve in Supplementary Fig. S6A). These data indicate that high PRH levels induce resistance to canonical Wnt pathway inhibition by ICG-001. It has recently been shown in a mouse model of acute myeloid leukemia that repression of Cdkn2a by PRH is dependent on recruitment of polycomb repressive complex 2 (PRC2; ref. 23). We therefore tested whether inhibition of EZH2, the catalytic subunit of PRC2 (using UNC1999), would block proliferation of CCLP1 cells in a PRH-dependent fashion. However, although EZH2 inhibition reduced cell proliferation, the effect was PRH-independent (Fig. 7A and B).

Figure 7.

Altered sensitivity to therapeutics. A, Proliferation measured by EdU incorporation in CCLP1 cells infected with Ad empty or Ad myc-PRH and treated with various compounds used at their LD50 as shown. B, Log sensitivity ratio to compounds from A. Drugs with a log ratio >0 are less effective after PRH overexpression and vice versa. *, a log ratio significantly different (P < 0.05) from 0; #, no significant difference. C, Proliferation measured by EdU incorporation in a panel of CCA cell lines infected with Ad empty or Ad myc-PRH and treated with palbociclib at their LD50. CCLP1, CCSW, and KKU-213 (LD50 = 100 nmol/L) and KKU-100 (LD50 = 150 nmol/L) as in A and presented as log sensitivity ratio as in B. *, P < 0.05; **, P < 0.01. D, Three independent control and Rb knockdown CCLP1 cell populations were treated with palbociclib at 100 nmol/L, and cell proliferation was measured using EdU incorporation assays. Inset, Western blot showing successful knockdown of Rb. *, P < 0.05; ns, no statistically significant difference. E, Schematic of pathways affected by PRH in CCA cells and compounds targeting these pathways.

Figure 7.

Altered sensitivity to therapeutics. A, Proliferation measured by EdU incorporation in CCLP1 cells infected with Ad empty or Ad myc-PRH and treated with various compounds used at their LD50 as shown. B, Log sensitivity ratio to compounds from A. Drugs with a log ratio >0 are less effective after PRH overexpression and vice versa. *, a log ratio significantly different (P < 0.05) from 0; #, no significant difference. C, Proliferation measured by EdU incorporation in a panel of CCA cell lines infected with Ad empty or Ad myc-PRH and treated with palbociclib at their LD50. CCLP1, CCSW, and KKU-213 (LD50 = 100 nmol/L) and KKU-100 (LD50 = 150 nmol/L) as in A and presented as log sensitivity ratio as in B. *, P < 0.05; **, P < 0.01. D, Three independent control and Rb knockdown CCLP1 cell populations were treated with palbociclib at 100 nmol/L, and cell proliferation was measured using EdU incorporation assays. Inset, Western blot showing successful knockdown of Rb. *, P < 0.05; ns, no statistically significant difference. E, Schematic of pathways affected by PRH in CCA cells and compounds targeting these pathways.

Close modal

Because exogenous PRH represses CDKN1B and CDKN2B and activates CCND2 indirectly via Notch3, we next examined whether inhibition of CDK4/6 using palbociclib would block proliferation in CCLP1 cells and whether sensitivity to this drug is modulated by PRH expression level. EdU incorporation experiments showed that high levels of PRH sensitized CCLP1 cells to the antiproliferative effects of palbociclib (Fig. 7A and B and dose–response curve in Supplementary Fig. S6B). We conclude that PRH overexpression drives CDK4/6 activity via both increased expression of cyclin D2 (via Notch3) and by direct repression of p15 and p27 expression. To determine whether increased PRH levels result in increased sensitivity to palbociclib in other CCA cell lines, we overexpressed the protein in a panel of cell lines in the presence and absence of palbociclib and measured EdU incorporation. Figure 7C shows the log ratio of sensitivity between control cells and PRH overexpressing cells in each case. In all of the cell lines tested, increased PRH levels result in increased sensitivity to palbociclib treatment.

The effect of palbociclib on cell proliferation is thought to be dependent on the presence and inhibition of phosphorylation of the Rb tumor-suppressor protein (42). We therefore examined Rb expression in the CCA cell lines and BECs used in this study. Interestingly, Rb protein expression was very low or not detectable in primary BECs or immortalized BECs, although it is present in CCA cell lines (Supplementary Fig. S7A). Moreover, treatment of CCA cell lines with palbociclib at their LD50 reduced the levels of phosphorylated Rb in each case (Supplementary Fig. S7B). To determine whether the effects of palbociclib on the proliferation of CCLP1 cells requires the presence of Rb, we knocked down Rb using siRNA. Although the proliferation of control cells was inhibited by palbociclib treatment, Rb knockdown CCLP1 cells are far less sensitive to the effects of this drug (Fig. 7D). In addition, palbociclib has no effect on the proliferation of PRH knockdown CCLP1 cells (Fig. 7D).

Figure 7E summarizes the transcriptome and phenotypic changes that are dependent on PRH expression and the associated altered sensitivities to palbociclib and other chemotherapeutics. We conclude that the clinical efficacy of various chemotherapeutic strategies is likely to depend on PRH expression level, and that patient stratification on the basis of PRH expression could improve the clinical usefulness of several compounds that have recently been suggested as potential novel CCA treatments including palbociclib.

Previous work in a variety of animal and in vitro models has suggested that hyperactivation of both Wnt signaling (7) and Notch signaling (10) are crucial events in carcinogenesis and progression of CCA. Here, we show that both of these events have a common origin in the dysregulation of PRH. We reveal that the PRH protein is elevated in human CCA cell lines and primary tumors relative to primary BECs, and we use a xenograft model and colony formation assays with primary human BECs to demonstrate the importance of PRH in CCA. We show that PRH depletion in CCA cells decreases cell proliferation and inhibits cell migration and cell invasion and brings about changes in gene expression consistent with MET. The ability of PRH to influence tumor cell behavior is not confined to increasing cell proliferation and maintaining a mesenchymal phenotype in CCA cell lines. Indeed, overexpression of PRH increases cell proliferation, migration, invasion, and anchorage-independent growth of primary human BECs and AKN1 cells. Moreover, transient elevated PRH expression in primary BECs induces changes in gene expression characteristic of EMT. Collectively, these experiments show that PRH is a novel oncoprotein in at least a large proportion of CCA tumors, and they suggest that PRH dysregulation underlies both CCA and the biliary pathologies that precede CCA.

PRH regulates NOTCH3 in CCA

We have shown that PRH is a regulator of NOTCH3 gene expression. However, ChIP-seq experiments did not show binding of PRH near NOTCH3 or near to any of the other PRH-regulated genes in the Notch pathway, such as NOTCH1, JAG1, and JAG2. Although the regulation of NOTCH3 appears to be indirect, it is one of only 53 genes that are differentially expressed upon both overexpression and knockdown of PRH. We also show that NOTCH3 knockdown decreases PRH expression and that this creates a positive feedback loop. One consequence of activation of a Notch3-PRH–positive reinforcement loop in CCA and in primary BECs is that a small perturbation to the activity or expression of either factor could lead to amplification of gene expression changes that ultimately give rise to the phenotypic alterations associated with CCA. Because we observed a strong correlation between HHEX and NOTCH3 gene expression in the whole TCGA dataset, which contains samples with a variety of driver mutations, we propose that a positive feedback loop between Notch3 and PRH is initiated independently of the underlying mutational landscape through common changes in the tumor microenvironment and the intracellular signaling milieu. We also see correlation of HHEX and HES1mRNA expression, which may occur because of coregulation of HES1 by PRH and Notch3. The coactivation of multiple genes in the pathway such as the gamma secretase activator PSEN2 by Notch3 and PRH likely also leads to the increase in Notch3 ICD observed following PRH expression.

Previous studies showed that Notch3 promotes PI3K/Akt signaling in CCLP1 cells in a noncanonical manner (10). Here, we present the first genome-wide identification of Notch3-regulated genes in CCA, and we use GSEA to show that multiple PI3K/Akt/mTOR pathway genes are regulated by Notch3. Interestingly, the most strongly enriched pathway in the Notch3-correlated gene set is “cholesterol homeostasis,” and “bile acid metabolism” is also enriched. Cholesterol-derived conjugated bile acids can drive CCA cell proliferation, and cholestasis is a known risk factor for CCA (43, 44), suggesting that part of the effect of Notch3 in CCA could be due to dysregulation of bile acid synthesis and metabolism.

PRH regulates Wnt signaling

PRH overexpression in CCLP1 cells resulted in the differential expression of several genes involved in Wnt signaling including WNT11, WNT16, TCF7L1, and the endogenous LRP6 inhibitor DKK1, and PRH ChIP-seq showed binding of PRH at these loci. These results suggest that PRH directly regulates several key genes in the Wnt signaling pathway. PRH also indirectly promotes Wnt signaling in CCLP1 cells through Notch3-mediated repression of CDH1. We infer that the dominant effect of Notch3 on Wnt signaling is to increase the available pool of β-catenin and thus amplify the effects of PRH on Wnt signaling. In addition, our PRH overexpression ChIP-seq and RNA-seq data reveal that many EMT-associated genes are direct targets of PRH, including FAP, DST, and ITGAV. Thus, we conclude that PRH and Notch3 collaborate to drive EMT and Wnt signaling; PRH directly regulates genes that affect both pathways and indirectly regulates additional genes in the same pathways via Notch3. Macrophages in the tumor microenvironment have been proposed to provide a source of Wnt ligands to drive the dysregulated Wnt signaling observed in CCA (7). We suggest that independently of exogenous Wnts, the Notch3-PRH loop may also drive aberrant autocrine Wnt signaling in CCA.

PRH plays a complex role in liver development; PRH-null mice are embryonic lethal (45) with multiple defects including defective liver development, decreased proliferation, and migration of hepatic progenitors (46, 47). Conditional deletion of PRH (FoxA3-Cre) results in liver hypoplasia and loss of extrahepatic ducts, whereas a later conditional deletion (Alfp-Cre) results in viable mice with cystic ducts and decreased differentiated intrahepatic bile ducts (48). Here, we show that PRH protein expression is turned off in mature human bile ducts and that its reexpression in bile duct epithelial cells promotes cell proliferation and cell invasion. One possibility is that the aberrant expression of PRH in differentiated adult bile duct cells mimics the role that PRH plays in promoting tissue growth in early organogenesis as the PRH partners that are required for PRH-dependent bile duct differentiation are likely limiting.

PRH and response to chemotherapeutics

We have shown that PRH influences resistance to canonical Wnt pathway inhibition by ICG-001 (an inhibitor of β-catenin acetylation by CBP). As ICG-001 is reported to be effective at reducing tumor growth in a mouse model of CCA carcinogenesis (7) and is in clinical trial for various solid tumors (49), it could become a compound of interest in the development of novel CCA treatments. Our data suggest that in this case, patient stratification on the basis of PRH expression may be useful to optimize the clinical benefits of this drug or its future derivatives. Inhibition of Notch signaling by targeting the γ-secretase complex is an emerging chemotherapeutic strategy for a variety of cancers (71 clinical trials undertaken as of 2018; ref. 50). Our data suggest that as well as driving aberrant Notch signaling, PRH also determines resistance to Notch inhibition, at least by nontransition state analog γ-secretase inhibitors such as DAPT. Our Notch3 knockdown data suggest that direct targeting of Notch3 (e.g., by a blocking antibody or by a small molecule that specifically interrupts interactions between Notch3-ICD and components of the Notch transcriptional complex) could be more effective than γ-secretase inhibition. In addition, the CDK4/6 inhibitor palbociclib strongly inhibited the proliferation of CCA cells in the presence of exogenous PRH, whereas it had significantly less effect on control cells at the same concentration and PRH knockdown cells were resistant to palbociclib treatment. The effects of palbociclib on CCA cells are thought to be dependent on Rb expression, and it is interesting to note that primary BECs and immortalized BECs express much lower levels of Rb than CCA cell lines. Moreover, the effects of palbociclib on CCLP1 cells are largely lost in the absence of Rb and are abolished when PRH is knocked down. These data suggest that CCA cells with high PRH are likely to be highly sensitive to palbociclib treatment and that this sensitivity is driven by the presence of PRH in these cells, but also requires Rb. These data also indicate that at least part of the mechanism by which PRH drives cell proliferation is likely to be through hyperactivation of CDK4/6. We show that PRH activates cyclin D2 expression via Notch3 and directly represses CDKN2B (p27) and CDKN1B (p15). Moreover, these data suggest that palbociclib or other CDK4/6-specific inhibitors could be especially effective in the treatment of CCA with high PRH expression levels.

In conclusion, we propose that monitoring PRH and Notch3 levels in patients with high CCA risk biliary pathologies, either directly by biopsy or indirectly by detection of the protein products of a set of PRH/Notch3 transcriptional targets (such as cell surface proteins or secreted proteins) in bile or serum, may be a useful diagnostic tool to help predict the development of CCA in at-risk groups, such as patients with primary sclerosing cholangitis in the West and liver fluke–infected patients in southeast Asia. In addition, monitoring PRH expression in patients with CCA may be a useful tool for guiding the choice of chemotherapeutic strategy.

No potential conflicts of interest were disclosed.

Conception and design: P. Kitchen, J. Lertsuwan, J. Satayavivad, S. Oltean, S. Afford, K. Gaston, P.-S. Jayaraman

Development of methodology: P. Kitchen, K.Y. Lee, K. Gaston

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): P. Kitchen, K.Y. Lee, D. Clark, N. Lau, J. Satayavivad, S. Oltean, K. Gaston

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): P. Kitchen, K.Y. Lee, D. Clark, A. Sawasdichai, K. Gaston, P.-S. Jayaraman

Writing, review, and/or revision of the manuscript: P. Kitchen, J. Lertsuwan, J. Satayavivad, S. Oltean, S. Afford, K. Gaston, P.-S. Jayaraman

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Sawasdichai

Study supervision: S. Afford, P.-S. Jayaraman

We would like to thank L. Wallace for technical assistance isolating primary BECs, Dr. Anetta Ptasinska for help with the ChIP protocol, and Gagandeep Batth for Akt Western blotting experiments. We thank Professor Constanze Bonifer for comments on the article and for providing palbociclib and Dr. John Halsall for UNC1999. We also thank Professor Stefan Roberts (University of Bristol) for critical reading of this article and useful discussions. This work was funded by an MRC Newton Fund Project Grant (MR/N012615/1), Thailand Research Fund grant (TRF; DBG5980005), and National Institute of Health Birmingham Biomedical Research Centre (NIH BBRC).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Bridgewater
J
,
Galle
PR
,
Khan
SA
,
Llovet
JM
,
Park
JW
,
Patel
T
, et al
Guidelines for the diagnosis and management of intrahepatic cholangiocarcinoma
.
J Hepatol
2014
;
60
:
1268
89
.
2.
Landskron
G
,
De laFuente
M
,
Thuwajit
P
,
Thuwajit
C
,
Hermoso
MA
. 
Chronic inflammation and cytokines in the tumor microenvironment
.
J Immunol Res
2014
;
2014
:
149185
.
3.
Ye
X
,
Weinberg
RA
. 
Epithelial-mesenchymal plasticity: a central regulator of cancer progression
.
Trends Cell Biol
2015
;
25
:
675
86
.
4.
Takebe
N
,
Warren
RQ
,
Ivy
SP
. 
Breast cancer growth and metastasis: interplay between cancer stem cells, embryonic signaling pathways and epithelial-to-mesenchymal transition
.
Breast Cancer Res
2011
;
13
:
211
.
5.
Polyak
K
,
Weinberg
RA
. 
Transitions between epithelial and mesenchymal states: acquisition of malignant and stem cell traits
.
Nat Rev Cancer
2009
;
9
:
265
73
.
6.
Cardinale
V
,
Renzi
A
,
Carpino
G
,
Torrice
A
,
Bragazzi
MC
,
Giuliante
F
, et al
Profiles of cancer stem cell subpopulations in cholangiocarcinomas
.
Am J Pathol
2015
;
185
:
1724
39
.
7.
Boulter
L
,
Guest
RV
,
Kendall
TJ
,
Wilson
DH
,
Wojtacha
D
,
Robson
AJ
, et al
WNT signaling drives cholangiocarcinoma growth and can be pharmacologically inhibited
.
J Clin Invest
2015
;
125
:
1269
85
.
8.
Guest
RV
,
Boulter
L
,
Dwyer
BJ
,
Forbes
SJ
. 
Understanding liver regeneration to bring new insights to the mechanisms driving cholangiocarcinoma
.
NPJ Regen Med
2017
;
2
:
13
.
9.
Loilome
W
,
Bungkanjana
P
,
Techasen
A
,
Namwat
N
,
Yongvanit
P
,
Puapairoj
A
, et al
Activated macrophages promote Wnt/beta-catenin signaling in cholangiocarcinoma cells
.
Tumour Biol
2014
;
35
:
5357
67
.
10.
Guest
RV
,
Boulter
L
,
Dwyer
BJ
,
Kendall
TJ
,
Man
TY
,
Minnis-Lyons
SE
, et al
Notch3 drives development and progression of cholangiocarcinoma
.
Proc Natl Acad Sci U S A
2016
;
113
:
12250
5
.
11.
Geisler
F
,
Strazzabosco
M
. 
Emerging roles of Notch signaling in liver disease
.
Hepatology
2015
;
61
:
382
92
.
12.
Boulter
L
,
Govaere
O
,
Bird
TG
,
Radulescu
S
,
Ramachandran
P
,
Pellicoro
A
, et al
Macrophage-derived Wnt opposes Notch signaling to specify hepatic progenitor cell fate in chronic liver disease
.
Nat Med
2012
;
18
:
572
9
.
13.
Gaston
K
,
Tsitsilianos
MA
,
Wadey
K
,
Jayaraman
PS
. 
Misregulation of the proline rich homeodomain (PRH/HHEX) protein in cancer cells and its consequences for tumour growth and invasion
.
Cell Biosci
2016
;
6
:
12
.
14.
Zong
Y
,
Stanger
BZ
. 
Molecular mechanisms of bile duct development
.
Int J Biochem Cell Biol
2011
;
43
:
257
64
.
15.
Bhave
VS
,
Mars
W
,
Donthamsetty
S
,
Zhang
X
,
Tan
L
,
Luo
J
, et al
Regulation of liver growth by glypican 3, CD81, hedgehog, and Hhex
.
Am J Pathol
2013
;
183
:
153
9
.
16.
Su
J
,
You
P
,
Zhao
JP
,
Zhang
SL
,
Song
SH
,
Fu
ZR
, et al
A potential role for the homeoprotein Hhex in hepatocellular carcinoma progression
.
Med Oncol
2012
;
29
:
1059
67
.
17.
Marfil
V
,
Blazquez
M
,
Serrano
F
,
Castell
JV
,
Bort
R
. 
Growth-promoting and tumourigenic activity of c-Myc is suppressed by Hhex
.
Oncogene
2015
;
34
:
3011
22
.
18.
Kershaw
RM
,
Siddiqui
YH
,
Roberts
D
,
Jayaraman
PS
,
Gaston
K
. 
PRH/HHex inhibits the migration of breast and prostate epithelial cells through direct transcriptional regulation of Endoglin
.
Oncogene
2014
;
33
:
5592
600
.
19.
Siddiqui
YH
,
Kershaw
RM
,
Humphreys
EH
,
Assis Junior
EM
,
Chaudhri
S
,
Jayaraman
PS
, et al
CK2 abrogates the inhibitory effects of PRH/HHEX on prostate cancer cell migration and invasion and acts through PRH to control cell proliferation
.
Oncogenesis
2017
;
6
:
e293
.
20.
Kershaw
RM
,
Roberts
D
,
Wragg
J
,
Shaaban
AM
,
Humphreys
E
,
Halsall
J
, et al
Proline-rich homeodomain protein (PRH/HHEX) is a suppressor of breast tumour growth
.
Oncogenesis
2017
;
6
:
e346
.
21.
Smith
S
,
Tripathi
R
,
Goodings
C
,
Cleveland
S
,
Mathias
E
,
Hardaway
JA
, et al
LIM domain only-2 (LMO2) induces T-cell leukemia by two distinct pathways
.
PLoS One
2014
;
9
.
22.
McCormack
MP
,
Young
LF
,
Vasudevan
S
,
de Graaf
CA
,
Codrington
R
,
Rabbitts
TH
, et al
The Lmo2 oncogene initiates leukemia in mice by inducing thymocyte self-renewal
.
Science
2010
;
327
:
879
83
.
23.
Shields
BJ
,
Jackson
JT
,
Metcalf
D
,
Shi
W
,
Huang
Q
,
Garnham
AL
, et al
Acute myeloid leukemia requires Hhex to enable PRC2-mediated epigenetic repression of Cdkn2a
.
Genes Dev
2016
;
30
:
78
91
.
24.
Zamparini
AL
,
Watts
T
,
Gardner
CE
,
Tomlinson
SR
,
Johnston
GI
,
Brickman
JM
. 
Hex acts with beta-catenin to regulate anteroposterior patterning via a Groucho-related co-repressor and Nodal
.
Development
2006
;
133
:
3709
22
.
25.
Marfil
V
,
Moya
M
,
Pierreux
CE
,
Castell
JV
,
Lemaigre
FP
,
Real
FX
, et al
Interaction between Hhex and SOX13 modulates Wnt/TCF activity
.
J Biol Chem
2010
;
285
:
5726
37
.
26.
Nussler
AK
,
Vergani
G
,
Gollin
SM
,
Dorko
K
,
Morris
SM
 Jr.
,
Demetris
AJ
, et al
Isolation and characterization of a human hepatic epithelial-like cell line (AKN-1) from a normal liver
.
In Vitro Cell Dev Biol Anim
1999
;
35
:
190
7
.
27.
Shimizu
Y
,
Demetris
AJ
,
Gollin
SM
,
Storto
PD
,
Bedford
HM
,
Altarac
S
, et al
Two new human cholangiocarcinoma cell lines and their cytogenetics and responses to growth factors, hormones, cytokines or immunologic effector cells
.
Int J Cancer
1992
;
52
:
252
60
.
28.
Sripa
B
,
Leungwattanawanit
S
,
Nitta
T
,
Wongkham
C
,
Bhudhisawasdi
V
,
Puapairoj
A
, et al
Establishment and characterization of an opisthorchiasis-associated cholangiocarcinoma cell line (KKU-100)
.
World J Gastroenterol
2005
;
11
:
3392
7
.
29.
Sirisinha
S
,
Tengchaisri
T
,
Boonpucknavig
S
,
Prempracha
N
,
Ratanarapee
S
,
Pausawasdi
A
. 
Establishment and characterization of a cholangiocarcinoma cell line from a Thai patient with intrahepatic bile duct cancer
.
Asian Pac J Allergy Immunol
1991
;
9
:
153
7
.
30.
Rattanasinganchan
P
,
Leelawat
K
,
Treepongkaruna
SA
,
Tocharoentanaphol
C
,
Subwongcharoen
S
,
Suthiphongchai
T
, et al
Establishment and characterization of a cholangiocarcinoma cell line (RMCCA-1) from a Thai patient
.
World J Gastroenterol
2006
;
12
:
6500
6
.
31.
Lleo
A
,
Zhang
W
,
McDonald
WH
,
Seeley
EH
,
Leung
PS
,
Coppel
RL
, et al
Shotgun proteomics: identification of unique protein profiles of apoptotic bodies from biliary epithelial cells
.
Hepatology
2014
;
60
:
1314
23
.
32.
Desjobert
C
,
Noy
P
,
Swingler
T
,
Williams
H
,
Gaston
K
,
Jayaraman
PS
. 
The PRH/Hex repressor protein causes nuclear retention of Groucho/TLE co-repressors
.
Biochem J
2009
;
417
:
121
32
.
33.
Jayaraman
P
,
Frampton
J
,
Goodwin
G
. 
The homeodomain protein PRH influences the differentiation of haematopoietic cells
.
Leuk Res
2000
;
24
:
1023
31
.
34.
Noy
P
,
Sawasdichai
A
,
Jayaraman
PS
,
Gaston
K
. 
Protein kinase CK2 inactivates PRH/Hhex using multiple mechanisms to de-repress VEGF-signalling genes and promote cell survival
.
Nucleic Acids Res
2012
;
40
:
9008
20
.
35.
Borowicz
S
,
Van Scoyk
M
,
Avasarala
S
,
Karuppusamy Rathinam
MK
,
Tauler
J
,
Bikkavilli
RK
, et al
The soft agar colony formation assay
.
J Vis Exp
2014
:
e51998
.
36.
Liberzon
A
,
Birger
C
,
Thorvaldsdottir
H
,
Ghandi
M
,
Mesirov
JP
,
Tamayo
P
. 
The molecular signatures database (MSigDB) hallmark gene set collection
.
Cell Syst
2015
;
1
:
417
25
.
37.
Manning
BD
,
Toker
A
. 
AKT/PKB signaling: navigating the network
.
Cell
2017
;
169
:
381
405
.
38.
Crompton
MR
,
Bartlett
TJ
,
MacGregor
AD
,
Manfioletti
G
,
Buratti
E
,
Giancotti
V
, et al
Identification of a novel vertebrate homeobox gene expressed in haematopoietic cells
.
Nucleic Acids Res
1992
;
20
:
5661
7
.
39.
Williams
H
,
Jayaraman
PS
,
Gaston
K
. 
DNA wrapping and distortion by an oligomeric homeodomain protein
.
J Mol Biol
2008
;
383
:
10
23
.
40.
Korinek
V
,
Barker
N
,
Morin
PJ
,
van Wichen
D
,
de Weger
R
,
Kinzler
KW
, et al
Constitutive transcriptional activation by a beta-catenin-Tcf complex in APC-/- colon carcinoma
.
Science
1997
;
275
:
1784
7
.
41.
Bienz
M
. 
beta-Catenin: a pivot between cell adhesion and Wnt signalling
.
Curr Biol
2005
;
15
:
R64
7
.
42.
Knudsen
ES
,
Hutcheson
J
,
Vail
P
,
Witkiewicz
AK
. 
Biological specificity of CDK4/6 inhibitors: dose response relationship, in vivo signaling, and composite response signature
.
Oncotarget
2017
;
8
:
43678
91
.
43.
Liu
R
,
Zhao
R
,
Zhou
X
,
Liang
X
,
Campbell
DJ
,
Zhang
X
, et al
Conjugated bile acids promote cholangiocarcinoma cell invasive growth through activation of sphingosine 1-phosphate receptor 2
.
Hepatology
2014
;
60
:
908
18
.
44.
Lozano
E
,
Sanchez-Vicente
L
,
Monte
MJ
,
Herraez
E
,
Briz
O
,
Banales
JM
, et al
Cocarcinogenic effects of intrahepatic bile acid accumulation in cholangiocarcinoma development
.
Mol Cancer Res
2014
;
12
:
91
100
.
45.
Hallaq
H
,
Pinter
E
,
Enciso
J
,
McGrath
J
,
Zeiss
C
,
Brueckner
M
, et al
A null mutation of Hhex results in abnormal cardiac development, defective vasculogenesis and elevated Vegfa levels
.
Development
2004
;
131
:
5197
209
.
46.
Bort
R
,
Martinez-Barbera
JP
,
Beddington
RS
,
Zaret
KS
. 
Hex homeobox gene-dependent tissue positioning is required for organogenesis of the ventral pancreas
.
Development
2004
;
131
:
797
806
.
47.
Bort
R
,
Signore
M
,
Tremblay
K
,
Martinez Barbera
JP
,
Zaret
KS
. 
Hex homeobox gene controls the transition of the endoderm to a pseudostratified, cell emergent epithelium for liver bud development
.
Dev Biol
2006
;
290
:
44
56
.
48.
Hunter
MP
,
Wilson
CM
,
Jiang
X
,
Cong
R
,
Vasavada
H
,
Kaestner
KH
, et al
The homeobox gene Hhex is essential for proper hepatoblast differentiation and bile duct morphogenesis
.
Dev Biol
2007
;
308
:
355
67
.
49.
Lyou
Y
,
Habowski
AN
,
Chen
GT
,
Waterman
ML
. 
Inhibition of nuclear Wnt signalling: challenges of an elusive target for cancer therapy
.
Br J Pharmacol
2017
;
174
:
4589
99
.
50.
Tamagnone
L
,
Zacchigna
S
,
Rehman
M
. 
Taming the notch transcriptional regulator for cancer therapy
.
Molecules
2018
;
23
.