The multidrug-resistant iCCA cell lines MT-CHC01R1.5 (with acquired resistance to GEM) and 82.3 (with primary resistance to GEM), isolated and stabilized at our Institution (13, 14, 26), were maintained in KnockOut DMEM/F-12 (Thermo Fisher Scientific, Waltham, MA, USA) supplemented with 10% fetal bovine serum (FBS), Hepes buffer, 100 U/ml penicillin, and 100 μg/ml streptomycin (P/S) (all from Sigma-Aldrich, St. Louis, MO, USA).

GEM hydrochloride (Sandoz, Novartis Division, Siena, Italy) and PTX (Sigma-Aldrich) were dissolved in water for injection and aliquoted in working solutions. For the in vivo experiments, Nab-PTX (Abraxane, Celgene, Milan, Italy) was used as PTX formulation.

Cell viability was evaluated by plating 2,500 cells per well into 96-well plates in optimal culture medium. After 24 h, cells were treated with escalating doses of PTX (from 0.9375 to 15 ng/ml). Viability was quantified after 72 h with the Cell Titer-Glo^® commercial kit (Promega, Milan, Italy), and luminescence was quantified with a GloMax microplate reader (GloMax-Multi Detection System, Promega). The IC50 of the drug was calculated using the CalcuSyn software (Biosoft, Cambridge, UK) based on the Chou-Talalay method.

For the cell cycle analysis, 2 × 10⁶ cells were plated into 10-cm plates. After 24 h, cells were treated with GEM (1.5 µM), PTX (15 ng/ml), or their combination; after 48 h of treatment, cells were fixed with ice-cold 70% ethanol. Cells were then incubated at 37°C for 15 min in the dark in PBS containing 10 μg/ml of propidium iodide (PI) (Thermo Fisher Scientific), 0.1% (v/v) Triton X-100, and 100 μg/ml RNase. Samples were then acquired with a CyAn ADP™ flow cytometer (Beckman Coulter, Brea, CA, USA) and data analyzed with the Summit software v4.3 (Beckman Coulter). The percentage of cells in the various phases of the cell cycle (G0/G1, S, G2/M) was calculated. Three independent experiments were performed.

For the apoptosis analysis, 1 × 10⁵ cells were seeded per well of 6-well plates. Cells were treated after 24 h with GEM (1.5 µM), PTX (15 ng/ml), or their combination. The quantification of apoptotic cells was performed after 48 h with the Annexin V-allophycocyanin (APC)/PI staining assay. Briefly, supernatants were recovered and cells were preincubated for 30 min at 4°C in the dark with a commercial binding buffer (Bender MedSystems, Wien, Austria). Next, cells were suspended in binding buffer containing Annexin V-APC (Bender MedSystems) and PI 50 µg/ml (Thermo Fisher Scientific) and incubated for 1 h at 4°C in the dark. Samples (30,000 events/experimental point) were acquired with a CyAn ADP™ flow cytometer (Beckman Coulter). Cell death was expressed as the percent average of dead cells (early and late apoptotic) obtained in three independent experiments.

Cells were subjected to lysis with lysis buffer (Cell Signaling, Danvers, MA, USA) and centrifuged at 20,000xg for 30 min at 4°C. Twenty μg of proteins was electrophoresed on Mini-PROTEAN TGX Precast Gels (4%–20%) and transferred on nitrocellulose Midi membranes using Trans-Blot Turbo (Bio-Rad, Hercules, CA, USA). Non-specific sites were blocked with 5% non-fat dry milk (Bio-Rad Laboratories, Munchen, Germany), and then the blots were stained using standard procedures with Cleaved Caspase 3 (CC3) and cleaved poly(ADP-ribose) polymerase (cPARP) primary (Cell Signaling) antibodies. Finally, signals were revealed by a chemiluminescence reagent (Euroclone, Milan, Italy). Densitometric analysis was performed with Fiji ImageJ (27) using the Analyze/Gel function. Briefly, after selecting each lane with a rectangular selector, densitometric histograms were derived with the Plot Lanes function and the peaks were measured excluding the side curves that originated from background signals.

For the sphere formation assay, 2.5 × 10⁵ cells were seeded per well into 24-well ultra-low attachment plates in serum-free DMEM-F12 (Sigma) with 1× B27 Supplement (Thermo Fisher Scientific), 200 ng/ml human EGF (PeproTech, London, UK), 0.4% bovine serum albumin (BSA) (Sigma), 4 μg/ml insulin (Sigma), and P/S (Life Technologies). Cells were treated with GEM (1.5 µM) or PTX (15 ng/ml). After 7 days, spheres with a diameter >50 µm were counted under an optical microscope (DMIL, Leica Microsystems GmbH). Three fields per well were acquired, and three independent experiments were performed.

NOD-SCID (Non-Obese Diabetic-Severe Combined Immunodeficient) female mice (4–6 weeks old) (Charles River Laboratory, Wilmington, MA, USA) were maintained under sterile conditions in micro-isolator cages at the animal facilities of Candiolo Cancer Institute, FPO-IRCCS. All animal procedures were approved by the Institutional Ethical Committee for Animal Experimentation (Fondazione Piemontese per la Ricerca sul Cancro) and by the Italian Ministry of Health (Ethic code: 177/2015-PR; 178/2015-PR and 106/2021-PR). In three independent experiments, MT-CHC01R1.5 cells resuspended in 50% (v/v) growth factor-reduced basement membrane matrix (Matrigel, BD Bioscience) were injected subcutaneously into the right flank of mice. When tumors reached a volume of 80–100 mm³ (approximately 3 weeks after injection), animals were randomized into four arms of treatment (6 mice/arm) that were treated twice a week with vehicle only (NT), GEM (25 mg/kg), Nab-PTX (10 mg/kg), or their combination. Tumor growth was measured with a caliper twice a week until the end of the experiment. Tumor volumes were calculated using the formula V = (A × B ²)/2 (V = tumor volume, A = largest diameter, B = smallest diameter). In addition, at the end of the experiments, tumors were explanted and weighted. Glucose uptake was analyzed in live animals using IVIS Lumina II for 2D acquisitions and IVIS Spectrum CT for 3D acquisition (Perkin Elmer Waltham, MA, USA) after tail vein injection of XenoLightRediJect 2-DeoxyGlucosone (DG)-750 (Caliper Life Sciences, Waltham, MA, USA). Images were analyzed with the Living Image Software (PerkinElmer, Caliper Life Sciences). To calculate glucose uptake by tumor xenografts, the ratio between total fluorescence emitted and total number of tumor mass voxels was calculated.

To evaluate the presence of necrosis, tumors were formalin-fixed and paraffin-embedded (FFPE) and 3-μm-thick sections were stained with hematoxylin and eosin (H&E). The Polygon Selection tool of ImageJ software was used to measure necroti areas. Five fields of each slide were analyzed. The proliferation marker Ki67, the apoptosis marker CC3, and the endothelial cell marker CD146 were investigated by immunohistochemistry (IHC). Briefly, 5-μm FFPE tissue sections were deparaffinized, rehydrated, and permeabilized, followed by antigen retrieval and saturation. Next, sections were decorated with the following primary antibodies: anti-Ki67 (DAKO, Carpinteria, CA, USA), anti-CC3 (Cell Signaling), and anti-CD146 (Abcam, Cambridge, UK). HRP-conjugated secondary antibodies (Anti-Mouse or Anti-Rabbit, DAKO) were applied, and the signal was detected with a diaminobenzidine tetrahydrochloride (DAB)-based system (Liquid DAB chromogen system, DAKO). Slides were counterstained with hematoxylin. Proliferation rates were calculated as the percentage of Ki67-positive cells out of the total number of cells in five fields of each slide. Apoptosis was expressed as the absolute value of CC3-positive cells in the total area of each slide. Tumor vasculature was evaluated by CD146 immunostaining; the result was represented as percentage of CD146-positive area in five fields of each slide. For this purpose, the Color Deconvolution plugin (ImageJ) was used.

Experiments were analyzed using mean ± standard error of mean (SEM) or standard deviation (SD), as indicated in the figure legends. Two-way ANOVA was used to analyze cell and tumor growth, as well as the differences in terms of drug response. One-way ANOVA (multiple comparisons) was used for the other statistical analysis. *p < 0.05; **p < 0.01; ***p < 0.001; and ****p < 0.0001 were considered statistically significant.

We first quantified the cytotoxic effect of PTX on two multidrug-resistant iCCA cell lines, namely, MT-CHC01R1.5 (acquired resistance to GEM) and 82.3 (primary resistance to GEM) (13, 14). Cells were incubated for 72 h with escalating doses of the drug, followed by a cell viability assay ( Figure 1 ). The derived IC50 values were calculated as 4.81 ± 1.98 and 10.55 ± 0.77 ng/ml for MT-CHC01R1.5 and 82.3 cells, respectively.

We next analyzed the cell cycle after treatment with GEM (1.5 μM), PTX (15 ng/ml), or GEM (1.5 μM) + PTX (15 ng/ml) for 48 h. As shown in Figure 2 and Supplementary Table 1 in both cell lines, as expected due to the previously characterized resistance, treatment with GEM did not induce any change in the cell cycle and was therefore considered as an additional negative control. In MT-CHC01R1.5 cells, treatment with PTX induced a decrease of cells in the G0/G1 phase compared to both NT and GEM and an increase of cells in the S phase compared to both NT and GEM. Treatment with GEM + PTX induced a drastic decrease of cells in the G0/G1 phase compared to all other experimental points, an increase of cells in the S phase compared to both NT and GEM, and a massive increase of cells in the G2/M phase compared to all other experimental points. In 82.3 cells, treatment with PTX induced a substantial decrease of cells in the G0/G1 phase compared to both NT and GEM, an increase of cells in the S phase compared to both NT and GEM, and a considerable increase of cells in the G2/M phase compared to both NT and GEM. Treatment with GEM + PTX was not different from GEM alone.

We further characterized drug efficacy in terms of cell death induced by 48 h of incubation with GEM (1.5 μM), PTX (15 ng/ml), or GEM (1.5 μM) + PTX (15 ng/ml). After treatment, cells were stained with APC-labeled Annexin V (a marker of early apoptosis) and PI (a marker of late apoptosis and necrosis), followed by flow cytometry. Early (positivity to Annexin V) and late apoptosis (positivity to both PI and Annexin V) were induced by PTX alone in both cell lines. The addition of gemcitabine did not increase the apoptotic fraction on MT-CH01R1.5 cells, while in 82.3, GEM acts as a PTX antagonist ( Figure 3 ).

A Western blot analysis of two apoptotic markers, namely, CC3 and cPARP confirmed the dissimilar behavior of the two cell models ( Figure S1 ). In the case of MT-CH01R1.5 cells, this analysis fully mirrored the flow cytometry results: both CC3 and cPARP were upregulated by PTX alone or in combination. For 82.3 cells, the results were discordant: CC3 was downregulated by GEM-containing treatments, while cPARP was upregulated by PTX-containing treatments (necrosis, positivity to PI) was slightly induced by PTX alone or in combination in MT-CHC01R1.5 cells but not in 82.3 cells.

Together, these data suggest that PTX in vitro, alone, exerted a cytotoxic effect in both cell lines. In 82.3 cells, the cytostatic effect of GEM prevailed in the drug combination, resulting as an antagonist of PTX.

We tested the ability of cells to grow in a non-adherent, serum-free environment, an assay commonly used to quantify cancer cell subpopulations with stemness properties, which are usually more refractory to chemotherapy (28, 29). The effect of PTX on the stem cell compartment was assessed on MT-CHC01R1.5 and 82.3 cells as their capacity to form spheroids (cholangiospheres) when grown in suspension and in the presence of GEM (1.5 μM) or PTX (15 ng/ml). Cholangiospheres were imaged and quantified under a light microscope after 7 days of treatment ( Figure 4A ). In 82.3 cells, but not in MT-CHC01R1.5, GEM was capable of reducing the numbers of presumed cancer stem cells. In both cell lines, PTX reduced the number of cholangiospheres compared to the controls NT or GEM ( Figures 4B, C ).

These results suggest that PTX may overcome drug resistance by affecting a subpopulation of cells with stemness features.

Having demonstrated the feasibility of PTX-based preclinical applications in multidrug-resistant iCCA cell lines, we set up the corresponding in vivo experiment. Following a recent study that demonstrated the efficacy of Nab-PTX over PTX in animal models (30), we used Nab-PTX, either as a single agent or in combination with GEM, for our in vivo studies with MT-CHC01R1.5 xenografts. Three weeks after injection, when tumor volumes reached 80–100 mm³, mice were randomized into four arms of treatment: (i) drug vehicle (saline solution) administered intraperitoneally (i.p.) (NT), (ii) GEM (i.p., 25 mg/kg), (iii) Nab-PTX (i.p., 10 mg/kg), and (iv) drug combination. GEM and Nab-PTX were administered every 3 days ( Figure 5A ). Tumors were monitored and measured twice a week. After a total of 23 days of treatments, animals were euthanized and tumor masses explanted ( Supplementary Figure 1 ). Figures 5B, C show that the combination of GEM and Nab-PTX delayed tumor growth, in terms of both tumor weight, which was reduced in the drug combination arm compared with NT (p < 0.001) and single-agent GEM (p < 0.05) ( Figure 5B ), and tumor volume, lower in the drug combination arm compared with NT (p < 0.0001), Nab-PTX (p < 0.0001), and GEM (p < 0.01) ( Figure 5C ). The response in terms of tumor growth was also evaluated by 3D (in vivo) and 2D (ex vivo) imaging upon injection of the fluorescent glucose derivative 2GD 750. In live animals, the combination of GEM and Nab-PTX reduced glucose uptake compared to NT (p < 0.05) and GEM (p < 0.01), as shown in Figure 6 , supporting (i) impaired cell viability and (ii) reduced tumor mass. The same results were obtained when glucose uptake was analyzed ex vivo on explanted tumors ( Supplementary Figures 2A, B ).

To dissect the antitumor activity of GEM and Nab-PTX association, we characterized tumor morphology in the different treatment arms. H&E staining of xenograft sections ( Figure 7 ) showed an increase in necrotic areas in GEM + Nab-PTX-treated tumors compared to both NT (p < 0.01) and GEM (p < 0.01). IHC analysis revealed a lower proliferation rate ( Figures 7A, C ), as quantified by Ki67 staining, in the drug combination arm in comparison with both NT (p < 0.001) and GEM (p < 0.05) and in the Nab-PTX monotherapy arm compared to NT (p < 0.01). The results of CC3 staining ( Figures 7A, D ) were consistent with those obtained in vitro, with an increase of apoptosis in the combination arm with respect to GEM (p < 0.01). Lastly, we evaluated tumor vasculature by staining for the endothelial cell marker CD146 ( Figures 7A, E ); results indicated a significant role of both Nab-PTX (p < 0.05) and Nab-PTX + GEM (p < 0.01) in impairing tumor blood vessels compared to GEM.

Together, these results demonstrate the efficacy of a Nab-PTX-including drug combination in reducing tumor growth in a xenograft model of multidrug-resistant iCCA in which GEM or PTX monotherapy is ineffective.