RNA was extracted and purified from surgically resected tumor, four primary CCA cell lines (CCA-UK5, CCA-UK6, CCA-UK7, and CCA-UK9) and patient-matched, histologically tumor-free resected margin as a control. One microgram of total RNA was reverse transcribed using 500 ng Oligo-d(T) and 250 ng random primers and M-MLV reverse transcriptase (Takara Shuzo Co., Tokyo, Japan). One microliter of the total 20 μL cDNA was added to 1 μL Taqman Probe (EGFR, Hs01076092; HER2, Hs01001580; HER3, Hs00176538; and HER4, Hs00955522), and 8 μL ddH₂O with 10 μL droplet digital polymerase chain reaction (ddPCR) super mix for probes (Bio-Rad, Hemel Hempstead, UK). Droplets were generated by adding 20 to 70 μL Droplet generation oil (Bio-Rad, Hemel Hempstead, UK) using a QX100 droplet generator (Bio-Rad, Hemel Hempstead, UK). Amplification was carried out using standard Taqman protocols (10-minute activation step, 95°C, 40 cycles of 30 seconds denaturation at 95°C, 1 minute annealing and elongation 60°C and final extension 98°C 10 minutes using a PCR thermal cycler (Bio-Rad, Hemel Hempstead, UK). Samples were analyzed using QX100 droplet reader (Bio-Rad, Hemel Hempstead, UK).

Immunohistochemistry was used to examine expression of EGFR family receptors in CCA patient tissues from tumours with different etiologies (OV related in the Thai group and nonOV in the UK group). Formalin-fixed-paraffin-embedded (FFPE) tumor samples from thirty patients were retrieved from the Department of Pathology, Ramathibodi Hospital, Mahidol University and Department of Pathology, Rajavithi Hospital, Bangkok, Thailand, and thirty patients from the Department of Pathology, Nottingham Universities Hospital Trust, for immunohistochemistry assessment. The tissues were histologically confirmed as mass forming CCA by the two pathologists of each institute (NL, CS and AZ, AM). OV-related CCA in Thai patients was diagnosed by the patient fulfilling one of these five criteria: (1) a history of previous positive stool examination for OV or its eggs; (2) identification of OV or its eggs in stool or bile; (3) demonstration of a typical bead-like cholangiogram; (4) demonstration of dilated peripheral small intrahepatic bile ducts, compatible with biliary parasitic disease on sonography, computed tomography or magnetic resonance image; and (5) histopathological evidence of OV in the hepatic resected specimen. Each case was tested using the following primary antibodies: anti-EGFR, anti-HER2, anti-HER3, and anti-HER4 antibodies. Serial 5-μm-thick sections of FFPE tissue were cut for a standard method for immunohistochemical analysis. The immunoreactivity was scored based on membranous and/or cytoplasmic staining [11] compared to positive control, as follows: no staining or faint staining in < 10% of tumor cells; 1+, weak perceptible membranous staining in ≥ 10% of tumor cells; 2+, moderate complete membranous staining in ≥ 10% of tumor cells; 3+, strong complete membranous staining in ≥ 10% of tumor cells [12,13]. Reference positive control tissue was normal placenta (for EGFR), receptor positive breast cancer for HER2 and HER3 and normal liver for HER4.

Cell lines used in this study included human CCA cell lines established from CCA tissue of Thai patients, HuCCA-1, KKU-M213, KKU-100, and KKU-M055 (obtained from the Japanese Cell Research Bank (HuCCA-1 [JCRB1657], KKU-M213 [JCRB1557], KKU-100 [JCRB1568], and KKU-M055 [JCRB1551]), and a normal cholangiocyte cell line, MMNK-1 [14,15]. All cell lines were cultured in RPMI supplemented with 10% fetal bovine serum and 2 mM L-glutamine (Sigma-Aldrich, Dorset, UK) in 5% CO₂ atmosphere at 37°C. KKU-M213 cells harbor mutations of KRAS (p.G13C), TP53 (p.V31I), ERBB2 (p.N125D), and KKU-100 cells harbor mutations of KRAS (p.G12D) and TP53 (p.P72R). There is no mutation data for HuCCA-1 and KKU-M055. Mutation profile of CCA cell lines (the DepMap portal website; https://depmap.org/portal/cell_line/ACH-001538?tab=mutation; DepMap ID no. ACH-001538).

ErbB protein expression was examined in CCA cell lines. Protein extracted using NP-40 lysis buffer containing 1% (vol/vol) Triton X (Calbiochem, Darmstadt, Germany), 1× protease inhibitor (Roche, Mannheim, Germany), 50 mM NaF (Sigma-Aldrich, St. Louis, MO), 2 mM sodium orthovanadate (Sigma-Aldrich, Dorset, UK) and phenylmethylsulphonyl fluoride (Sigma-Aldrich) and quantified using Pierce BCA Assay Kit (Thermo Scientific, Basingstoke, UK). Twenty-five micrograms of protein was separated via sodium dodecyl sulfate polyacrylamide gel electrophoresis on 10% gels followed by wet transfer of the proteins to a nitrocellulose membrane using Bio-Rad transfer equipment (Bio-Rad, Hercules, CA), and transfer efficiency was confirmed by Ponceau S staining. The membranes were blocked in 5% BSA for 1.5 hours. Blots were then incubated with primary antibodies: anti-EGFR (Abcam, Cambridge, UK), anti-HER2, anti-HER3 (Cell Signaling Technology Inc., Beverly, MA), anti-HER4 (Abcam), and anti–α-tubulin antibodies (Cell Signaling Technology Inc.) at 4°C, overnight. The blots were then incubated with horseradish peroxidase–conjugated anti-rabbit or anti-mouse IgG secondary antibodies (Cell Signaling Technology Inc.) for 2 hours at room temperature and imaged using an Alliance Q9 mini Chemiluminescent gel imager (UVITEC, Cambridge, UK). The relative intensity of the protein band was quantified by ImageJ (from NIH website by Scion Corporation, Frederick, MD) and Image Studio (LI-COR).

Chemotherapeutic drugs used in this study included gemcitabine, cisplatin, and 5-fluorouracil (5-FU). The specific ErbB inhibitors included afatinib, lapatinib, and erlotinib, all were purchased from Selleckchem (Houston, TX). The drugs were dissolved in cell culture grade dimethyl sulfoxide (DMSO) (AppliChem, Barcelona, Spain). Frozen DMSO stocks were prepared together with the drug to be used as vehicle. For cell lines, all drug treatments were performed at 60%–80% cell confluence.

Fresh surgical material from tumor resections at Nottingham University NHS Trust, was immediately placed into tissue transfer media (Dulbecco’s modified Eagle’s medium with 10% fetal bovine serum, 1% 0.2 M L-glutamate, 1% penicillin/streptomycin/amphotericin B, 0.1% 1 mg/mL hydrocortisone, 0.75% 1 mg/mL insulin) at 4°C, and processed within 4–6 hours. The majority of tissue was finely minced for immediate use and further portions were FFPE for immunohistochemistry, or stored in RNAlater (Ambion, Austin, TX) for subsequent analysis.

In vitro tumor cell growth was established from five patients, four of which we went on to investigate drug sensitivities (CCA-UK5, CCA-UK6, CCA-UK7, and CCA-UK9), and expanded with a layer of supporting feeder layer cells according to the method of Liu et al. [16] and as previously described by Saunders et al. [17]. A small amount of finely minced tumor tissue was enzymatically disaggregated by type II collagenase (100 U/mL, Invitrogen, Burlington, Canada) and dispase (2.4 U/mL, Invitrogen) in Hank’s balanced salt solution without calcium or magnesium (Sigma-Aldrich, Dorset, UK) at 37°C under constant rotation. Cells were removed at hourly intervals until the tumor was completely disaggregated. Cells were allowed to settle on the feeder layer, and expanded. If fibroblasts (assumed cancer associated fibroblasts [CAFs]) were seen amongst the epithelial population, they were harvested using differential trypsinisation, and subsequently grown separately. Cell number and viability were determined using trypan blue exclusion.

The CCA cells were harvested and employed for close-to-patient three-dimensional tumor growth assays. Briefly, cells were resuspended in 9 mg/mL ice-cold Cultrex basement membrane extract (Trevigen, Gaithersburg, MD), diluted in modified RPMI-1640 containing phenol red-free and 6 mmol/L D-Glucose at pH 6.8) (Life Technologies, Inc., Rockville, MD). CCA cells (confirmed by human-specific antibodies) were suspended in basal membrane extract in the presence or absence of the CAFs or human bone-marrow-derived mesenchymal stem cells (MSCs, Sciencell, UK). Cells were then plated into low adherent, black-walled, clear bottom 384-well plates at 6,250 tumor cells±CAFs/MSCs. Drugs at 0.1, 1, 10, and 100 μM were added in six replicates on day 3. Drugs used in combination were premixed and serially diluted together before adding to the assay. Drug exposure was for 96 hours before final endpoint readings. An AlamarBlue assay (Invitrogen); 10% (v/v) was added for 1 hour at 37°C to monitor cell growth daily using fluorescent plate reader (FLUOstar Omega, BMG Labtech). Drug sensitivity was calculated as a percentage of a matched untreated control and IC₅₀ were determined using GraphPad Prism 8 (GraphPad Software, Inc., La Jolla, CA), nonlinear curve fit of y=100/(1+ 10^(Log1C50-X)×HillSlope).

The proliferative effect of specific ErbB tyrosine kinase inhibitors (afatinib, lapatinib, and erlotinib) (Selleckchem) on human CCA cells was determined by MTT assay. In brief, CCA cells (HuCCA-1, KKU-M213, KKU-100, and KKU-M055) were grown overnight in 96-well plates at density of 5×10³ cells/well, then incubated with different concentrations of specific ErbB tyrosine kinase inhibitors (0.1, 1, and 10 μM) for 48 hours. Cells (HuCCA-1 and KKU-M213) were also incubated with the chemotherapeutic drugs: gemcitabine, cisplatin, 5-FU (Selleckchem) at 0.1, 1, and 10 μM for 48 hours. Cells without drug treatment were used as a control group and 0.1% DMSO treated cells were used as a vehicle control group. After incubation, 100 μL of MTT solution (0.5 mg/mL, Sigma-Aldrich, St. Louis, MO) was added to each well and incubated for a further 4 hours at 37°C in the dark. Subsequently, 100 μL of DMSO (Merck, Darmstadt, Germany) was added to each well and absorbance of the sample was measured at OD490 nm by a Versamax microplate reader using SoftMax Pro 4.8 analysis software (Molecular Devices, San Jose, CA). The IC₅₀ value was calculated based on the nonlinear regression curve fit method by GraphPad Prism 6.0 software (GraphPad Software, Inc.). The greatest best-fit curve was used for calculation and R² value was more than 0.8.

The drug combination study was performed in HuCCA-1 and KKU-M213 cells by using a fixed concentration of gemcitabine (0.1 μM) combined with cisplatin, 5-FU, or with specific ErbB inhibitors (afatinib, lapatinib, and erlotinib) at 0.1, 1, and 10 μM. Interactions between the different drugs were evaluated using the combination index (CI) as described by Chou [18]. The CI for each fraction-affected value (Fa-CI) representing the percentage of proliferation inhibited by a drug. A plot of CI values and Fa-CI for each drug-pair was generated using the CompuSyn software (ComboSyn, Inc., Paramus, NJ; available for free download from http://www.combosyn.com. A CI interpreted as: CI value < 1.0, synergistic; 1.0, additive and > 1.0, antagonistic.

Cell cycle distribution was analyzed using flow cytometry. HuCCA-1 and KKU-M213 cells were cultured and starved with serum-free media overnight. Cells were treated with a single treatment of gemcitabine (0.1 μM) or afatinib (0.1 μM) for 24 hours. Concurrent treatment with gemcitabine and afatinib (G+A) and sequential treatments of gemcitabine (6 hours) followed with afatinib (18 hours) (G→A) and an inverted order with the same interval (A→G) were also performed. Cells were harvested and prepared for flow cytometry by incubated with propidium iodide (Sigma-Aldrich, St. Louis, MO). DNA content of cells was determined using a flow cytometer (BD FACSCanto model, BD Biosciences). The protein samples were applied to western blotting and incubated with the following primary antibodies: anti-cyclin D1, anti-cyclin E (Santa Cruz Biotechnology, Inc., Santa Cruz, CA), and anti α-tubulin antibodies (Cell Signaling Technology Inc.).

Experiments were performed in triplicate. Data are presented as means±standard error of mean and statistically analyzed using GraphPad Prism program ver. 6 (GraphPad Software, Inc.) by one-way ANOVA followed by Tukey’s multiple comparison tests and within-treatment group comparisons were performed using paired t test (2-tailed). Difference with p-values less than 0.05 were considered statistically significant.

mRNA expression of ErbB receptors by ddPCR revealed high expression of EGFR, HER2 HER3, and HER4 (Fig. 1A) in tumor tissues compared with tissue resection margin, and in tumor cells isolated from primary tissues compared with normal cells isolated from the margin (Fig. 1B). Immunohistochemistry of all tissue sections from both OV positive patients from Thailand and presumed OV negative patients from the UK patients (compared due to different etiologies) were positive for all ErbB family receptors; however, they showed different intensities (Table 1, Fig. 1). Immunoreactive staining intensity was scored relative to that of a reference positive control. Compared with histologically normal tissues taken from the resection margin of the CCA all ErbB receptors had expression that was more intense in tumor cells from CCA patients from both Thailand and the UK (Fig. 1C). All CCA cases showed positive staining for EGFR with most showing moderate or strong staining (Table 1). CCA tissues from both Thailand and the UK were weakly to moderately stained for HER2 (Table 1). HER3 was weakly stained in both Thailand and the UK CCA tissues. HER4 was strongly stained in Thailand and the UK tumors. Notably, these findings highlighted that EGFR, HER2, and HER4 are highly expressed in CCA tumors irrespective of whether they come from - UK patients not exposed to OV, or from fluke-induced Thai cancers. We confirmed that the high expression of ErbB was preserved in immortalized CCA cell lines (S1A Fig.). ErbB expression was investigated in 4 CCA cell lines derived from patients with CCA associated with fluke infection including KKU-M213, HuCCA-1, KKU-100, and KKU-M05 and a cholangiocyte cell line, MMNK-1. Immunoblotting showed that KKU-M055 cells expressed low levels of all the ErbB receptors while KKU-M213, HuCCA-1, and KKU-100 cells highly expressed EGFR, HER2, and HER3 proteins (S1B Fig.). The relative expression levels of EGFR, HER2, and HER3 proteins were found to be HuCCA-1 > KKU-M213 > KKU-100 > KKU-M055 while for HER4 this was HuCCA-1 > KKU-M213 > KKU-M055 and KKU-M100. We also determined mRNA expression of ErbB at transcription level and found that the mRNA transcript of EGFR and its family members, HER2, HER3, and HER4 in these CCA cell lines were not consistent with the protein expression, while all ErbB receptors were equally expressed in MMNK-1 cells (S1D Fig.).

We then investigated the effects of ErbB inhibitors on the growth of CCA cell lines. All three inhibitors (afatinib, lapatinib, and erlotinib) were cytotoxic in a dose-dependent manner (Fig. 2A–C) with IC₅₀ values ranging from 0.31–23.65 μM (Table 2). The HuCCA-1 cells were the most sensitive to all drug treatments: afatinib (IC₅₀=0.73 μM), lapatinib (IC₅₀=0.32 μM), and erlotinib (IC₅₀=4.52 μM). KKU-M213 exhibited the least sensitivity to all drugs: afatinib (IC₅₀=4.31 μM), erlotinib (IC₅₀=8.06 μM), and lapatinib (IC₅₀=11.14 μM). There was no apparent correlation between the response to the treatments of ErbB inhibitors (IC₅₀ values) and the expression of ErbB proteins (S2 Fig.). The most sensitive HuCCA-1 cells and the least sensitive KKU-M213 cells were selected for further experiments.

To determine whether ErbB inhibitors could in addition to existing therapies we examined the effect of ErbB2 inhibitors in the presence of effective cytotoxic agents. We first confirmed the sensitivity of HuCCA-1 and KKU-M213 cells to chemotherapeutic drugs (gemcitabine, cisplatin, and 5-FU). This confirmed that HuCCA-1 cells were more sensitive to gemcitabine than (IC₅₀=0.76 μM) 5-FU (IC₅₀=4.64 μM), or cisplatin (18 μM) (S3A Fig.) while KKU-M213 showed less sensitivity to all three drugs: gemcitabine (IC₅₀=14.73 μM), 5-FU (IC₅₀=15.68 μM) and cisplatin (16 μM) (S3B Fig.). We further determined the interaction effect of combining gemcitabine with cisplatin or gemcitabine with 5-FU. There was an antagonistic effect of adding cisplatin to gemcitabine in HuCCA-1 (S3C Fig.), and KKU-M213 cells (S3D Fig.) with the computed CI values > 1 (S3E and S3F Fig.), whereas, gemcitabine (0.1 μM) combined with 5-FU (0.1, 1, and 10 μM) showed a significant additive effect on growth inhibition in both cell lines in a dose-dependent manner (S3C and S3D Fig.), with the computed CI values < 1 (S3E and S3F Fig.).

As gemcitabine seemed the most likely to provide additional benefit, we therefore determined the effect of gemcitabine combination with tyrosine kinase ErbB inhibitors (afatinib, lapatinib, and erlotinib) at a fixed ratio (0.1: 0.1, 1, 10 μM) and the efficacy of combination treatment was evaluated in the least and most sensitive cell lines (KKU-M213 and HuCCA-1). In HuCCA-1 cells, a synergistic effect was observed when afatinib (0.1, 1, and 10 μM), lapatinib (10 μM), or erlotinib (10 μM) were combined with 0.1 μM gemcitabine compared to gemcitabine alone (Fig. 2D). In KKU-M213, a synergistic effect was observed when afatinib (0.1, 1, and 10 μM) or erlotinib (1 and 10 μM) was combined with 0.1 μM gemcitabine while gemcitabine combined with lapatinib showed no beneficial effect (Fig. 2E). Combined treatment of gemcitabine with either afatinib or erlotinib showed a CI value toward synergism in both HuCCA-1 (S4A Fig.) and KKU-M213 cells (S4B Fig.). In contrast, the combined treatment of gemcitabine and lapatinib mostly produced antagonistic effect (very high CI values) (S4A and S4B Fig.). These results indicated that the growth inhibitory effect of gemcitabine was enhanced by the addition of afatinib and erlotinib but not lapatinib.

We chose afatinib to further investigate a synergistic effect of the combined treatments using three different approaches: (1) concurrent treatment of gemcitabine (0.1 μM) with afatinib for 24 hours (G+A), (2) sequential treatment of gemcitabine (6 hours) followed with afatinib for 18 hours (G→A) and (3) inverted treatment with afatinib then gemcitabine (A→G) with the same interval (Fig. 3). Cells were also treated with gemcitabine alone (0.1 μM) or afatinib alone (0.1, 1, or 10 μM) for comparison. The results demonstrated that a single treatment of gemcitabine for 24 hours at 0.1 μM slightly decreased cells viability while afatinib at 1 and 10 μM significantly decreased viability of both HuCCA1 (Fig. 3A) and KKU-M213 cells (Fig. 3B). For HuCCA1 cells, only a concurrent treatment of 0.1 μM gemcitabine and 0.1 μM afatinib showed a beneficial additive effect compared to their respective dose of afatinib; however, growth inhibition was small. Increasing the concentration of afatinib did not enhance the additive effect. Sequential treatment (G→A) showed an additive effect to 0.1 μM afatinib alone. While an additive effect for inverted treatment (A→G) was observed using afatinib at 0.1 and 1 μM. Moreover, the sequential treatment using 0.1 μM afatinib with gemcitabine in both G→A and A→G showed similar levels of growth inhibition (Fig. 3A). For KKU-M213 cells, all concurrent treatments of gemcitabine and afatinib showed no additive effect to afatinib. Additive effect was followed with 0.1 μM afatinib (G→A) and the reverted order (A→G) at afatinib 0.1 and 1 μM (Fig. 3B).

We performed cell cycle analysis to determine whether the result of the synergistic effect of gemcitabine combined with afatinib resulted from cell cycle arrest or apoptosis. The cell cycle distribution of HuCCA-1 and KKU-M213 cells after single drug treatment, concurrent or sequential treatments of gemcitabine (0.1 μM) and afatinib (0.1 μM) for 24 hours are shown in Fig. 4. Treatment of HuCCA-1 cells with gemcitabine alone increased cells accumulated in the subG1 (apoptotic/dead cells) and S phases consistent with inhibition of DNA synthesis, and this was associated with decreased cell viability (Fig. 4A) while with afatinib treatment alone, cells accumulated in subG1 phase. Concurrent treatment of gemcitabine and afatinib (G+A) increased cell accumulation in S phase. Both sequential treatments of G→A and A→G induced accumulation in the subG1 and S phases, accompanied by decreased cells in G2/M phase compared to control (Fig. 4A).

For KKU-M213 cells, either gemcitabine or afatinib single treatment showed no significant effect on cell cycle distribution compared to untreated control cells. Concurrent treatment of gemcitabine and afatinib significantly increased cells in subG1 population associated with a decrease in G0/G1 and S phases. Sequential treatments of either G→A or A→G significantly increased cells accumulation in S phase (Fig. 4B). These results support the hypothesis that concurrent treatment forced the majority of cells to undergo cell death while sequential treatment resulted in S-phase arrest.

We next determined the expression of cyclin D1 and cyclin E cell cycle regulators. Western blot analysis showed that treatment with gemcitabine had no effect on cyclin D1 but upregulated cyclin E in HuCCA-1 cells (Fig. 4C–E). Afatinib single treatment slightly decreased levels of cyclin D1 in both HuCCA-1 (Fig. 4C and D) and KKU-M213 cells (Fig. 4F and G). The combination treatments of gemcitabine and afatinib (G+A, G→A, A→G) significantly decreased expression of cyclin D1 (Fig. 4C, D, F, and G) but not cyclin E compared to control and gemcitabine alone in both cells (Fig. 4C, E, F, and H).

We next assessed the response to ErbB inhibitors in primary CCA cells isolated from patients, in a 3D setting, in the presence of either CAFs or MSCs (Fig. 5A). Cells were grown in 3D 96 well plates, and then treated after growth and cell survival measured (Fig. 5B). While all four cell lines tested showed a sensitivity to gemcitabine/cisplatin treatment that was below the mean peak serum concentration (mPSC) for patients undergoing treatment with these agents, the addition of CAFs or MSCs increased the resistance to these treatments, and in two cases to above the mPSC, indicating that while cells from these two tumors might appear to be sensitive on their own, one would predict that they would be insensitive to treatment clinically (Fig. 5C). When cells were treated with erlotinib, they appeared to be borderline sensitive or insensitive by themselves, and addition of CAFs or MSCs made them much less sensitive (i.e., resistant) to erlotinib treatment (Fig. 5D). In contrast, all four cell types were sensitive (IC₅₀ less than or equal to the mPSC) for Afatinib, and three of the four were highly sensitive (IC₅₀ less than 80% of the mPSC) in the presence of CAFs, and all four were highly sensitive in the presence of MSCs (Fig. 5E). These results suggest that not only are primary CCA cells sensitive to afatinib, they are potentially more so in a tumor relevant setting (in the presence of other stromal cells).