Sub-lines with acquired drug resistance to docetaxel (C4-2B^R and LNCaP^R, respectively) were generated as previously described (Lima et al., 2021). C4-2B^R and LNCaP^R showed docetaxel IC₅₀ values of 99.47–100.50 nmol/L and 49.50–50.65 nmol/L, respectively, compared to the C4-2B and LNCaP parental cells that had IC₅₀ values of 1.00–1.40 nmol/L, and 0.78–1.06 nmol/L, respectively (Lima et al., 2021). DU145 parental and docetaxel-resistant DU145^R, and PC-3 parental and docetaxel-resistant PC-3^R prostate cancer cells (Ellegaard et al., 2013), were kindly provided by Marja Jaättellä (University of Copenhagen, Denmark). All PCa cells were cultured and maintained in RPMI-1640 medium + glutaMAX™-I (Gibco, Invitrogen, Carlsbad, CA, United States) supplemented with 10% fetal bovine serum (FBS). MDA-MB-231 parental and DTX-resistant MDA-MB-231^R, and MCF-7 parental and DTX-resistant MCF-7^R breast cancer cells (Hansen et al., 2016), were kindly provided by Jan Stenvang. MDA-MB-231 and MCF-7 cells were cultured and maintained in Dulbecco’s modified Eagle’s medium (DMEM) including L-glutamine, supplemented with 10% and 1% fetal bovine serum (FBS), respectively. The clear cell renal cell carcinoma line RCC-FG2, and breast ductal carcinoma lines HCC1419, HCC1569, HCC202, and BT483 (ATCC, United States) were cultured and maintained in RPMI-1640 medium supplemented with 2 mM L-glutamine and 10% fetal bovine serum. All cell lines were tested for mycoplasma contamination using Mycoplasma PCR detection kit (GATC-Biotech, Köln, Germany). The FDA-approved drug library (SCREEN-WELL FDA-approved drug library V2) was from Enzo Life Sciences (Farmingdale, NY, United States). Itraconazole was acquired from Sigma-Aldrich (Sporanox; Merck Life Science A/S, Denmark).

Cytotoxicity was performed using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-tetrazolium bromide (MTT) assay as previously described (Carmichael et al., 1987). Briefly, cells were plated at a density of 8,000 cells/well in triplicates in 96-well cell culture plate and allowed to grow in drug free medium over 24 h prior to treatment. After 48 or 72 h of treatment, drugs were removed and 0.5 mg/ml MTT (Sigma-Aldrich) was added to each well. Following incubation for three hours, 20% sodium dodecyl sulphate (SDS) in 0.02 M hydrochloric acid (HCL) was added to each well to dissolve the formed formazan crystals overnight. Absorbance of formazan was measured in a microplate spectrophotometer (PowerWaveX, Bio-Tek Instruments, INC.) at 570 nm and the background absorbance of MTT was measured at 670 nm. Cell viability was expressed in percent relative to untreated control cells.

Whole-cells were harvested and lysed using lysate buffer M-PER Mammalian Protein Extraction Reagent (Thermo Scientific) supplemented with Pierce Protease and Phosphatase Inhibitor Mini tables (Thermo Scientific). Cell lysates were centrifuged at 14,000 g for 10min at 4°C and supernatants collected. Total amount of protein was assessed by the Pierce™ BCA Protein Assay Kit (Thermo Scientific), according to manufacturer’s instructions. The Novex^® NuPAGE^® MES SDS Running Buffer (Thermo Fisher Scientific) was used for separation of proteins according to manufacturer’s instructions. Samples were loaded onto precast 10- or 15- well 4–12% Bis-Tris Gel gels (Novex^® NuPAGE^®, Invitrogen). Proteins were blotted onto a nitrocellulose membrane (iBlot^®2 NC, Invitrogen) using iBlot^®2 gel transfer device. Blots were blocked for 1 h in washing buffer (PBS+0.1% Tween 20) containing 5% non-fat dry milk and incubated overnight with appropriate primary antibody diluted in blocking reagent: Anti-MDR1 (Abcam, Cambridge, United Kingdom); Purified Mouse Anti- β-actin (BD Transduction Laboratories™, NJ, United States). After washed for 3 × 10 min in TBS-T, membranes were incubated with horseradish peroxidase-conjugated secondary antibody (Mouse/Rabbit) for 1 h at RT. Membranes were further washed 3 × 10 min in TBS-T and developed using Clarity Western ECL substrate (Bio-Rad) detection reagent. Proteins bands were detected with UVP BioSpectrum Imaging System (UVP, CA, United States).

The set of TMA blocks analysed by immunohistochemistry comprised samples from a consecutive series of men (n = 336), with clinically localized prostate cancer who underwent radical prostatectomy with curative intent from 1 January 2002 until 31 December 2005 at the Department of Urology, Rigshospitalet, Copenhagen, Denmark. Detailed construction of the TMA blocks and clinicopathological parameters of the samples has been previously reported (Kristensen et al., 2018). Briefly, two 1-mm cores from each of a representative malignant and non-malignant tissue areas were mounted on a total of 44 TMA blocks. The study was approved by the Danish National Committee on Health Research Ethics for the Capital Region (J.nr. H-6-2014-111).

Freshly cut 2.5 µm sections of each of 44 TMA blocks were deparaffinized, rehydrated and exposed to thermal-induced antigen retrieval to unmask epitopes. Sections were boiled for 10 min in Envision Flex Target Retrieval Solution, high pH (Dako, Glostrup, Denmark) diluted 1:50 in miliQ H₂O, before being incubated for 1 h with primary MDR1 antibody (Abcam #ab170904) diluted 1:1,000 in antibody diluent with background reducing components (Dako). Then sections were washed twice in TBS + 0.5% Triton X-100 and incubated for 20 min with High Definition Polymer Detector (AH diagnostics). Colorimetric signals were detected using DAB. Sections were developed with EnvisionTM FLEX DAB + Chromogen (Dako) diluted in EnvisionTM FLEX Substrate Buffer to visualize the primary antibody. Sections were counterstained with Mayer’s hematoxylin (Hounisen, Skanderborg, Denmark) and mounted with Pertex xylene-based mounting media (Hounisen). Positive controls for MDR1 staining consisted of tissue cores from normal liver and FFPE-embedded docetaxel resistant cells (LNCaP^R and C4-2B^R, respectively).

The rhodamine 123 (Rho123; Sigma-Aldrich) efflux assay was used to assess functional activity of ABCB1 in the presence or absence of ITZ. Briefly, cells (5 × 10⁵) were incubated at 37°C in the dark for 30 min in DMEM medium supplemented with 5% FCS and containing 0.5 μmol/L Rho123. Cells were then washed extensively with ice-cold PBS to completely remove extracellular Rho123, and subsequently incubated in the presence or absence of ITZ 2.5 μmol/L in DMEM at 37°C for 90 min. As a control, Valspodar (PSC-833; 1 μmol/L), a selective P-glycoprotein inhibitor, was used instead of ITZ. Cells were then washed with ice-cold PBS and immediately analyzed for intracellular Rho123 fluorescence on a FACSort flow cytometer. Mean fluorescence intensity (MFI) was calculated for each analysis. Five thousand events were analyzed.

To examine the cellular behavior of DTX in the presence of ITZ, we performed in vivo imaging of drug resistant C4-2B^R and LNCaP^R cells and the corresponding parental lines C4-2B and LNCaP. For imaging, 9–12,000 cells were seeded into black, 190 μm clear bottom 96-well microplates (Screenstar, Greiner). Two days later cells were treated, as indicated, with ITZ or vehicle together with 0.5 μM nuclear violet (AAT Bioquest) and 0.1 μM DTX^DeepRed (Tubulin Tracker DeepRed, Invitrogen), which consists of DTX conjugated with a bright and stable far-red fluorescent dye. One hour later cells were set up for live imaging using ImageXpress Micro confocal high-throughput microscope (Molecular Devices) and scanned and imaged every two hours, until the final 7-h timepoint using widefield mode and a ×40 objective, in 5% CO₂ atmosphere and 37°C. Nuclear violet was imaged with a DAPI filter (Ex. 377/54nm, Em. 447/60) and DTX^DeepRed with a Cy5 filter (Ex. 631/28nm, Em. 692/40 nm). The images were analyzed using MetaXpress software (Molecular Devices) and the cell scoring program using the violet nucleus as an indicator of a cell to calculate the % of cells with far-red dye incorporated to their tubulin network. Results presented are representative of three independent experiments.

The recently determined three-dimensional structures of ABCB1 from Cyanidioschyzon merolae (CmABCB1, PDB entries 6A6M and 6A6N) were retrieved from the Protein Data Bank and used as model for the human ABCB1 (Kodan et al., 2019). After water molecules and ligands were deleted, the Protein Preparation Wizard in Maestro version 11.1 was used to prepare the protein structures for docking by adding missing side-chains, assigning proper atom types and bond orders, and optimizing the hydrogen-bonding network pH = 7.0 (Sastry et al., 2013). A restrained minimization of the proteins was performed with convergence of heavy atoms to RMSD = 0.30 Å using the OPLS3e force field (Roos et al., 2019). The structure of ITZ, the (2R,4S,2′R)-stereoisomer, was retrieved from PubChem (entry CID 55283). ITZ is a weak base with pKa values of the triazole and piperazine moieties predicted by the MarvinSketch version 15.4.20 program from ChemAxon (https://www.chemaxon.com) to be 2.2 and 3.9, respectively. Possible tautomers and protonation states were generated at pH 7.0 ± 2.0. The different stereoisomers were generated in Maestro and subjected to a short energy minimization. The GOLD (Genetic Optimisation for Ligand Docking) program version 5.6 was used for docking of ITZ to both the homodimeric structures and the monomeric structure (A-chain) of ABCB1 (Jones et al., 1997). Binding sites were defined by a 30 Å sphere centered at a residue in the binding cavity (Glu530 in 6A6M and Tyr 358 in 6A6N). Ligands were docked with the slow genetic algorithm and using ChemScore as scoring function (Verdonk et al., 2003). Ten poses were sampled and analyzed for each docking. The Prime program implemented as part of the Schrodinger software system version 2019-4 (Schrödinger, LLC, New York, NY, 2014, http://www.schrodinger.com) was used for determination of the MM/GBSA free energies of binding (Li et al., 2011; Genheden and Ryde, 2015). The program yields free energies of binding based on the actually provided as well as on relaxed target and ligand structures (Table 2, ΔG Bind and ΔG Bind(NS), respectively).

Comparison of dose-response curves was performed by non-linear regression fitting of curves with a built-in model [log (inhibitor) vs. response] using GraphPad Prism software (GraphPad Prism 9, GraphPad Software, United States). For each dataset we then used ANOVA to test for significance in differences in best-fit values in IC₅₀ of the dataset, with IC50 values are the same for all data sets as null hypothesis. Significance level was set to 5%.

Although up-regulation of ABCB1 expression is expected to occur following exposure of cancer cells to DTX, thus underpinning acquisition of resistance to chemotherapy, high levels of expression of ABCB1 may also contribute to intrinsic resistance to DTX in chemo-naïve tumors. Then, if that was the case one could use expression of ABCB1 as a predictive biomarker. Because cancer samples are an admixture of many different cell types, and there were conflicting reports of ABCB1 predictive value depending on whether studies were done at the mRNA or protein level (Vergier et al., 1993), we performed an immunohistochemistry (IHC) study of ABCB1 expression in a large cohort of chemo-naïve prostate cancer tissue samples. In this way we could assess the specific expression of ABCB1 protein (P-gp) in cancer cells, as well as the relative expression of the membrane-bound functional drug transporter. Prostatectomies from 336 chemo-naïve castration resistant prostate cancer (CRPC) patients (Kristensen et al., 2018) (Table 1), were used to assess the potential of ABCB1 expression as a predictive biomarker for DTX resistance in prostate cancer (Figure 1). We found that apart from endothelial cells (yellow arrows; Figures 1D,F), PCa samples were negative for ABCB1 expression (white arrows; illustrated in Figures 1E,F). As expected, the positive controls, FFPE embedded LNCaP^R and C4-2B^R cells, and normal liver tissue (black arrows; Figures 1A–C, respectively), showed ABCB1 staining. FFPE embedded LNCaP and C4-2B cells showed no immunoreactivity for ABCB1 (Figures 1A,B, left panel, respectively). Regardless of the functional importance of ABCB1 in MDR in PCa, our data suggested that ABCB1 may not be useful as a predictive biomarker of intrinsic DTX resistance in chemo-naïve CRPC patients.

Currently, drug development strategies targeting ABCB1-mediated drug resistance in cancer mostly combine ABCB1 inhibitors with substrate drugs to re-establish drug sensitivity in the drug-resistant cells. We used our DTX-resistant cell models to screen a library of 786 FDA-approved drugs (Screen-Well FDA-approved drug library V2) for compounds (drug X, 10 µM) that could kill the DTX-resistant C4-2B^R cells in combination with 0.01 μM DTX. We found that one drug, itraconazole (ITZ), showed little cytotoxicity of its own, with only a modest effect observed at higher concentrations (>2.5 μM; Figure 2A), however, ITZ was able to efficiently kill DTX-resistant cells when combined with DTX. Given that ITZ has a favorable toxicity profile and can be used for long-term maintenance treatment (Persat et al., 1992), should the combination of ITZ with low doses of DTX efficiently overcame drug resistance it may be of clinical value. To assess the effect of ITZ on DTX resistance, LNCaP^R and C4-2B^R cells were treated with various concentrations of DTX combined with ITZ for 72 h, and cell viability was assessed by MTT (Figures 2B,C). Addition of ITZ showed a dose-dependent reversal of sensitivity to DTX in LNCaP^R (Figure 2B, p < 0.01) and C4-2B^R cells (Figure 2C, p < 0.0001), respectively.

Unlike other azole antifungal drugs, such as fluconazole and ketoconazole, which are known to block gonadal and adrenal steroidogenesis (Patel et al., 2018; Munkboel et al., 2019), ITZ is not expected to interfere with androgen synthesis. To determine if the effect of ITZ on drug resistance was independent of androgen receptor (AR) activity, we investigated if ITZ could reverse DTX resistance in AR-negative ABCB1-overexpressing prostate cancer cell lines. We treated two previously described AR-negative DTX-resistant cell lines, PC-3^R and DU145^R, with DTX in combination with ITZ for 72 h (Figures 2D,E, respectively). Cell viability was assessed by MTT and, as shown in Figure 2, AR signaling was not a determining factor as ITZ could also revert DTX resistance in these cells in a dose-dependent manner (Figures 2D,E; p < 0.0001 and p < 0.0001, respectively).

ITZ has been shown to revert daunorubicin resistance in murine leukemia P388/ADR multidrug resistant cells, as well as adriamycin resistance in K562 cells (K562/ADR) and HL60 human leukemia cells (HL60/ADR) (Gupta et al., 1991), suggesting that ITZ may reverse multidrug resistance in various settings. DTX is used not only for treatment of metastatic PCa, but also in other cancers such as breast cancer. We used the DTX-resistant breast cancer cell lines MDA-MB-231^R and MCF-7^R, previously established in our group (Hansen et al., 2016), to evaluate the activity potential of ITZ in breast cancer cells. The MDA-MB-231^R and MCF-7^R sublines have been previously characterized, and ABCB1 upregulation was identified as a major alteration associated with docetaxel resistance in these cells (Hansen et al., 2016). As shown in Figure 3, ITZ was able to reverse DTX resistance in MDA-MB-231^R cells (Figure 3B, p < 0.05) but failed to do so on MCF-7^R cells (Figure 3A), respectively.

To determine whether ITZ was able to overcome DTX resistance unrelated to ABCB1 upregulation, we analyzed the DTX sensitivity data from the Genomic of Drug Sensitivity in Cancer project platform (GDSC1 dataset) to identify cell lines with substantial intrinsic DTX resistance (IC₅₀ > 0.1 µM; Figure 3C). We found five lines that had no detectable expression of ABCB1 (Figure 3C, lower right subpanel). The five cell lines are shown ranked by their sensitivity and respective IC₅₀ (Figure 3C, table inset). As previously, all five cell lines were treated with DTX alone (5–25 nM), ITZ alone and a combination of both, for 72 h. As shown in Figure 3D through Figure 3H, BT-483 (Figure 3D), HCC202 (Figure 3F), HCC1569 (Figure 3G), and HCC1419 (Figure 3H) cell lines did not demonstrate significant differences in cell viability when exposed to DTX or ITZ, alone or in combination, respectively at the concentrations tested. The RCC-FG2 cell line showed a borderline significant effect of ITZ in combination with DTX (p = 0.0475) (Figure 3E). Taken together, our data suggested that ITZ might only surpass DTX resistance associated with ABCB1 upregulation.

Although ITZ has been reported to be an ABCB1 inhibitor at clinically relevant doses (Wang et al., 2002), the exact mechanism(s) by which it can overcome multidrug resistance remains unclear. ITZ is reportedly a potent inhibitor of the hedgehog (Hh) signaling pathway (Kim et al., 2010; Inoue et al., 2016), and the Hh pathway effector transcriptional factor GLI1 regulates expression of ABCB1 (Chen et al., 2014). To determine whether ITZ modulated expression of ABCB1 in the DTX-resistant PCa cells, we examined the expression of ABCB1 in ITZ treated cells alone or in combination with DTX. We found no effect of ITZ on ABCB1 expression in any of the cell lines tested (illustrated for C4-2B^R in Figure 4A). We then evaluated whether ITZ ability to reverse docetaxel resistance stems from inhibition of ABCB1 efflux activity. We analysed the efflux activity of ABCB1 using the cationic fluorescent dye Rhodamine 123 (Rho123) as a tracer. Rho123 is a substrate of ABCB1 and its cellular accumulation provides an useful proxy for evaluating ABCB1-mediated drug-efflux activity (Ludescher et al., 1992). We found that ITZ had no significant effect on Rho123 efflux from LNCaP^R and C4-2B^R cells (Figures 4B,C). By contrast, the ABCB1 inhibitor valspodar (PSC-833) was able to efficiently block efflux of Rho123 from LNCaP^R and C4-2B^R cells (Figures 4B,C, green line), respectively. Since it was possible that ITZ affected only some very specific ABCB1 substrates, we examined the levels of intracellular DTX in LNCaP^R and C4-2B^R cells, in the absence or presence of various concentrations of ITZ (Figures 4C–G). We performed single-cell in vivo imaging of drug resistant C4-2B^R and LNCaP^R cells using DTX conjugated with a bright and stable far-red fluorescent dye (DTX^DeepRed) in the absence or presence of various concentrations of ITZ. As can be seen in Figure 4, in the absence of ITZ DTX^DeepRed accumulates in the LNCaP parental cell line (Figure 4D) but is efficiently extruded from the DTX resistant LNCaP^R line (Figure 4E). ITZ was able to block efflux of DTX^DeepRed in a dose-dependent manner in LNCaP^R cells (Figure 4G), restoring intracellular concentrations of DTX^DeepRed to levels comparable, albeit lower, to those of LNCaP cells (Figure 4F). The same effect was observed in C4-2B^R cells (data not shown), indicating that ITZ is an ABCB1 inhibitor.

Docking of ITZ to the inward-facing form of ABCB1 yielded a series of very similar poses, all indicating that ITZ binds at the trans-membrane part of the transporter between the two monomers (Figure 5A). The dioxolane end of ITZ is sandwiched between helices H6 and H6* with the dichloro-substituted benzene and triazole rings pointing towards H5 and H1*. Side-chains from these helices (H5: Met351, Ile354 and Tyr358; H6: Phe383, Phe384, Ile387 and Met391; H1*: Leu135, Phe138 and Phe142; H6*: Phe384, Ile387, Leu388 and Met391) form a hydrophobic cavity encircling this part of the ITZ structure. Five and eight of the 11 residues interacting with ITZ are identical and similar, respectively, to the corresponding residues in the human ABCB1. The other end of the ITZ molecule with the 1,2,4-triazole-3-one moiety is poking in between H4* and H5*. The only polar interaction we observe is a hydrogen bond between Ser350 on H5* and the carbonyl group in the 1,2,4-triazole-3-one moiety. For comparison, we have also docked ITZ to the inward-facing form of the ABCB1 homodimer and to the ABCB1 monomer to compare the energetics of the different binding scenario. For both proteins ITZ binds close to the ATP-binding sites (data not shown). GOLD binding energies (GOLD ΔG, Table 2) show that binding to the homodimers is more favorable than binding to the monomer, and that binding to the outward-facing form is superior to binding to the outward-facing form. The ranking is further supported by more rigorously determination of the free energy of binding determined by the Schrödinger implementation of the MM/GBSA method (MM/GBSA ΔG, Table 2). ITZ contains three chiral centers and, accordingly, may exist as eight stereoisomers (Shi et al., 2010). The commercial ITZ formulation is a mixture of the four cis-stereoisomers, which recently have been shown not only to be more potent than the trans-stereoisomers, but also to display different hepatotoxicity and antiangiogenic activity (Shi et al., 2010; Shim et al., 2016). Our original studies were done on the (2R,4S,2′R)-isomer retrieved from PubChem, but subsequently we have docked all eight stereoisomers to the inward-facing form of the transporter. For all eight stereoisomers we find docking poses displaying binding modes like the one reported for the (2R,4S,2′R)-isomer. Generally, we find that the cis-isomers have better GOLD scores than the trans-isomers by an average on 4 kcal/mol. Thus, our docking studies indicate that ITZ binds tightly deep in between the trans-membrane spanning alpha-helices of the inward-facing form of ABCB1 and thereby can inhibit the transport of substrates such as DTX as illustrated in Figure 5B.