The More the Better—Investigation of Polymethoxylated N-Carboranyl Quinazolines as Novel Hybrid Breast Cancer Resistance Protein Inhibitors

The ineffectiveness and failing of chemotherapeutic treatments are often associated with multidrug resistance (MDR). MDR is primarily linked to the overexpression of ATP-binding cassette (ABC) transporter proteins in cancer cells. ABCG2 (ATP-binding cassette subfamily G member 2, also known as the breast cancer resistance protein (BCRP)) mediates MDR by an increased drug efflux from the cancer cells. Therefore, the inhibition of ABCG2 activity during chemotherapy ought to improve the efficacy of the administered anti-cancer agents by reversing MDR or by enhancing the agents’ pharmacokinetic properties. Significant efforts have been made to develop novel, powerful, selective, and non-toxic inhibitors of BCRP. However, thus far the clinical relevance of BCRP-selective MDR-reversal has been unsuccessful, due to either adverse drug reactions or significant toxicities in vivo. We here report a facile access towards carboranyl quinazoline-based inhibitors of ABCG2. We determined the influence of different methoxy-substitution patterns on the 2-phenylquinazoline scaffold in combination with the beneficial properties of an incorporated inorganic carborane moiety. A series of eight compounds was synthesized and their inhibitory effect on the ABCG2-mediated Hoechst transport was evaluated. Molecular docking studies were performed to better understand the structure-protein interactions of the novel inhibitors, exhibiting putative binding modes within the inner binding site. Further, the most potent, non-toxic compounds were investigated for their potential to reverse ABCG2-mediated mitoxantrone (MXN) resistance. Of these five evaluated compounds, N-(closo-1,7-dicarbadodecaboran(12)-9-yl)-6,7-dimethoxy-2-(3,4,5-trimethoxyphenyl)-quinazolin-4-amine (DMQCd) exhibited the strongest inhibitory effect towards ABCG2 in the lower nanomolar ranges. Additionally, DMQCd was able to reverse BCRP-mediated MDR, making it a promising candidate for further research on hybrid inorganic-organic compounds.


Introduction
The development and improvement of different chemotherapeutics for cancer treatment have been the main research topic of numerous groups worldwide. Due to the occurrence of multidrug resistance (MDR), the efficacy of chemotherapy is often decreased [1][2][3][4]. MDR was proven to exist either as an intrinsic property of some cancer cells, or, more address hydrophobic binding sites within the target proteins. Further, utilizing the inorganic nature of the hybrid structures appears to be beneficial for the improvement of the drugs' metabolic stability [32].  [24]; (C) the first reported carborane-functionalized ABCG2 inhibitors [33]. Köhler et al. previously reported the implementation of a carborane cluster as a functional group in BCRP inhibitors [33]. However, the two tested compounds (see Figure 1C) exhibited lower activity than their organic counterparts. Nevertheless, combining the favorable properties of a carborane cluster (hydrophobicity and 3D appearance) with a suitable organic scaffold could be useful for the interaction with the ABCG2 protein, leading to specific inhibition and, thus, a successful reversal of multidrug resistance.
Herein, we report the synthesis of eight N-carboranyl quinazoline derivatives, based on either a quinazoline or 6,7-dimethoxyquinazoline scaffold. Various methoxy substitution patterns of the quinazoline scaffold were studied to determine the substituents' effect on the inhibition efficacy towards BCRP. Furthermore, in vitro and in silico experiments were conducted to test the potential of these molecules as BCRP inhibitors. In silico studies serve as an important starting point in understanding the interactions of the N-carboranyl quinazoline derivatives with the binding site of the ABCG2 protein. The experiments revealed these compounds to be prominent ABCG2 inhibitors with good levels of cytotoxicity showing signs of MDR reversal, therefore making them promising candidates for further research on hybrid inorganic-organic compounds.

Synthetic Procedures and Analytical Data
All solvents were degassed, dried, and purified with the solvent purification system SPS-800 by MBraun and stored over activated 4 Å molecular sieves. All reactions were carried out under a nitrogen or argon atmosphere using standard Schlenk techniques and anhydrous, degassed solvents, unless otherwise stated. The deuterated solvent (CDCl3) was dried over P2O5, vacuum-transferred, and degassed by freeze-pump-thaw cycling. SPhos Pd G4 and closo-9-bromo-1,7-dicarbadodecaborane (12) (CbBr) were prepared according to the literature [34,35]. All other starting materials and reagents are commercially available and were used without further purification. The NMR spectra were recorded on  [24]; (C) the first reported carborane-functionalized ABCG2 inhibitors [33]. Köhler et al. previously reported the implementation of a carborane cluster as a functional group in BCRP inhibitors [33]. However, the two tested compounds (see Figure 1C) exhibited lower activity than their organic counterparts. Nevertheless, combining the favorable properties of a carborane cluster (hydrophobicity and 3D appearance) with a suitable organic scaffold could be useful for the interaction with the ABCG2 protein, leading to specific inhibition and, thus, a successful reversal of multidrug resistance.
Herein, we report the synthesis of eight N-carboranyl quinazoline derivatives, based on either a quinazoline or 6,7-dimethoxyquinazoline scaffold. Various methoxy substitution patterns of the quinazoline scaffold were studied to determine the substituents' effect on the inhibition efficacy towards BCRP. Furthermore, in vitro and in silico experiments were conducted to test the potential of these molecules as BCRP inhibitors. In silico studies serve as an important starting point in understanding the interactions of the Ncarboranyl quinazoline derivatives with the binding site of the ABCG2 protein. The experiments revealed these compounds to be prominent ABCG2 inhibitors with good levels of cytotoxicity showing signs of MDR reversal, therefore making them promising candidates for further research on hybrid inorganic-organic compounds.

Synthetic Procedures and Analytical Data
All solvents were degassed, dried, and purified with the solvent purification system SPS-800 by MBraun and stored over activated 4 Å molecular sieves. All reactions were carried out under a nitrogen or argon atmosphere using standard Schlenk techniques and anhydrous, degassed solvents, unless otherwise stated. The deuterated solvent (CDCl 3 ) was dried over P 2 O 5 , vacuum-transferred, and degassed by freeze-pump-thaw cycling. SPhos Pd G4 and closo-9-bromo-1,7-dicarbadodecaborane(12) (CbBr) were prepared according to the literature [34,35]. All other starting materials and reagents are commercially available and were used without further purification. The NMR spectra were recorded on a BRUKER Avance III HD 400 MHz NMR spectrometer at 25 • C ( 1 H 400.13 MHz, 11 B 128.38 MHz, 13 C 100.63 MHz, two-dimensional ( 1 H-1 H COSY, 1 H-13 C HSQC, 1 H-13 C HMBC)). The chemical shifts δ are reported in parts per million (ppm). Tetramethylsilane (TMS) or solvent residual peaks were used as the internal reference in 1 H and 13 C NMR spectra, and all other nuclei spectra were referenced to TMS using the Ξ scale [36]. The numbering scheme for 1 H and 13 C NMR signals is presented for each compound in the SI. Electrospray ionization mass spectrometry was carried out with an ESI-qTOF Impact II by Bruker Daltonics GmbH in the positive mode. The IR spectra were obtained with an FT-IR spectrometer Nicolet iS5 (ATR, transmission, Thermo Scientific, Waltham, MA, USA) scanning between 4000-400 cm −1 with a KBr beam splitter (only selected frequencies given). Elemental analyses (C, H, and N) were performed with a Heraeus VARIO EL microanalyzer. The melting points were determined in glass capillaries using a Gallenkamp apparatus and are uncorrected. The column chromatography was performed using a Biotage Isolera ONE and KP-SIL columns.
Detailed chemical synthesis procedures and chemical analysis results of all compounds (Qa-d, DMQa-d, QCa-d, and DMQCa-d) are described in Appendix A and the Supplementary Materials.

Determination of Cell Viability by WST-1 Cell Proliferation Assay
MDCKII-hABCG2 cells and their parental MDCKII-WT cells were seeded in 96-well plates (TPP, Trasadingen, Switzerland) in a density of 2 × 10 4 cells/mL and 3 × 10 4 cells/mL, respectively. After 48 h, the cells were incubated with increasing concentrations of up to 50 µM of the novel synthesized compounds, with 0.1% Triton X-100 as the positive control and solvent (0.1% DMSO) as the negative control for 48 h. Afterwards, the substance-specific cytotoxicity was determined by the water-soluble tetrazolium-1 (4-[3-(4-iodophenyl)-2-(4-nitro-phenyl)-2H-5-tetrazolio]-1,3-benzene sulfonate; WST-1) assay as previously described [37]. The cell viability was determined by microplate reader at 450 nm (Tecan Sunrise, Crailsheim, Germany). The data of three independent experiments (N = 3) were tested for normality with the Shapiro-Wilk test. The significant differences compared to solvent control were analyzed by one-way ANOVA with the Holm-Šidák post hoc test using SigmaPlot 14.5 (Systec Software, San Jose, CA, USA). The data were normalized against the vehicle control (0.1% DMSO) and were expressed as mean ± SEM.

Determination of ABCG2 Interaction with Hoechst 33342 Accumulation Assay
The Hoechst 33342 accumulation assay was used to detect an interaction of the investigated compounds with the human ABCG2 transporter. The assay protocol was adapted from that published by Wassermann et al. [38]. The MDCKII-hABCG2 (2 × 10 4 cells/mL) and MDCKII-WT (3 × 10 4 cells/mL) cells were seeded in 96-well plates and cultured for 72 h. Afterwards, the sub-confluent monolayers were treated with 0.5 µM and 1.0 µM of the investigational compounds or its respective solvent control (0.1% DMSO) for 4 h. The known ABCG2 inhibitor Ko143 (1 µM) was added as positive control. After 5 and 20 min, the cells were washed twice with ice-cold PBS buffer solution and lyzed with 100 µL/well 0.1% (v/v) SDS/PBS. The intracellular Hoechst 33342 amount was detected by spectrofluorometry (360 nm excitation/465 nm emission wavelengths, Tecan Infinite F200 Pro, Crailsheim, Germany). The examined fluorescence subtracted by the background of non-treated cells was correlated to the protein amount quantified by the bicinchoninic acid assay (BCA, Thermo Scientific, Rockford, IL, USA) following the manufacturer's instruc-tions. Relative fluorescence units (RFU) were calculated as the relation of fluorescence/mg protein of MDCKII-hABCG2 cells to MDCKII-WT cells, normalized to the solvent control set as 1 and expressed as mean ± SEM. The data from the Hoechst 3342 accumulation assay (five independent experiments N = 5) were tested for normality with the Shapiro-Wilk test. Significant differences compared to the solvent control were analyzed by one-way ANOVA with the Holm-Šidák post hoc test using SigmaPlot 14.5 (Systec Software, San Jose, CA, USA). An inhibition of human ABCG2 efflux transport activity is defined as the significant difference to the solvent control (0.1% DMSO).

Determination of Autofluorescence
The autofluorescence of all compounds was investigated with a modified Hoechst 33342 accumulation assay. The MDCKII-hABCG2 and MDCKII-WT cells were seeded and treated as previously described for the Hoechst accumulation assay. Afterwards, the cells were washed twice with ice-cold PBS buffer, lyzed with 100 µL/well 0.1% (v/v) SDS/PBS, and, subsequently, the intracellular fluorescence was measured by spectrofluorometer (360 nm excitation/465 nm emission wavelengths, Tecan Infinite F200 Pro, Crailsheim, Germany). The protein amounts were quantified by BCA assay. For this approach, the Hoechst 33342 dye was not added. The background of non-treated cells was subtracted from the measured intracellular fluorescence and was correlated to the protein amount for MDCKII-hABCG2 or MDCKII-WT cells, respectively. All data were expressed as mean ± SEM and tested for normality by the Shapiro Wilk test using SigmaPlot 14.5. An autofluorescence was defined as the significant increase in total intracellular relative fluorescence unit (RFU) in comparison to the solvent-treated control and was determined by one-way ANOVA with Holm-Šidák post hoc test using SigmaPlot 14.5. As the negative control, the ABCG2-Inhibitor Ko143 (1 µM) was added.

Reversal of Multidrug Resistance
The MDCKII cells were seeded as described in the WST-1 assay section, and treated with increasing concentrations of mitoxantrone (0.01 µM up to 50 µM) or solvent (0.1% DMSO) for 48 h. In order to detect a reversal of the ABCG2-mediated chemoresistance, novel ABCG2 interacting compounds (QCc, DMQCa, DMQCb, DMQCc, and DMQCd) were added to mitoxantrone in 1.0 µM for 48 h. Afterwards, the WST-1 assay was performed as described [37]. The left-shift factor was calculated from IC 50 minus SEM from the mitoxantrone alone treatment by IC 50 plus SEM from combined use of mitoxantrone and carborane-based derivatives (QCc, DMQCa, DMQCb, DMQCc, and DMQCd). Data were tested for normality by the Shapiro Wilk test and significant differences between the mitoxantrone treatment groups were determined by two-way ANOVA with the Holm-Šidák post hoc test using SigmaPlot 14.5. The IC 50 values were defined as 50% reduced cell viability and calculated with SigmaPlot 14.5 by nonlinear regression. All data were normalized against the vehicle control and were expressed as mean ± SEM, calculated from at least three independent experiments.

Molecular Docking/Computational Methods
The structures were built with open-source molecular builder Avogadro 1.1.1 [39] and geometries were further optimized with ORCA 4.2.1, using the PBEh-3c method [40]. The electrostatic potential derived CHELPG charges were obtained from the ORCA-internal orca_chelpg program. The cryo-EM crystal structures of the whole human ABCG2 transporter protein (PDB code 5NJ3) and the MXN-ABCG2 co-crystallized protein-substrate adduct (PDB code 6VXI) were obtained from the Protein Data Bank (http://www.rcsb.org (accessed on 26 April 2021)) [41][42][43]. The UCSF Chimera package was used to remove all ligands, non-standard residues, and water molecules from the crystal structures [44]. The AutoDock Tools 1.5.6 (ADT) was used to prepare the structures for docking [45]. The protein and ligand structures were constructed for docking using a previously described process [46]. To comply to the AutoDock atom types, the missing hydrogen atoms were added to the protein structures, all polar hydrogen atoms were preserved, and all non-polar hydrogen atoms were merged. All histidine, lysine, and arginine residues were set to their protonated state, whereas all the aspartic acid and glutamic acid residues were deprotonated. For the protein, Gasteiger charges were applied to each atom. The above-described optimized geometries were used for docking. All torsion angles inside the carborane clusters were adjusted to non-rotatable, and the non-polar hydrogen atoms of the ligand were combined to match the AutoDock atom types. The docking was performed with AutoDock 4.2.6 according to a previously reported protocol (100 docking runs) [45,46]. The docking grid box dimension was set to 120 × 120 × 120 grid points along the x-, y-, and z-axes, respectively. The grid spacing of 0.375 Å was utilized. The grid center coordinates were 124.747, 124.778, and 143.202 (x, y, and z assignments, respectively). In order to group the final docked conformations, a tolerance of 1.5 Å root-mean-square deviations (RMSD) was used. The ADT 1.5.6 was used for setting up the docking protocols and result analysis. The AutoDock free energy scoring function had a standard error of 2-3 kcal/mol [45]. The UCSF ChimeraX software was used to visualize and render all molecular docking figures [44].

Biological Investigations
The eight target compounds (QCa-d and DMQCa-d) were evaluated for their cytotoxicity and ability to inhibit the ABCG2 transporter using the Madin-Darby canine kidney cells (MDCKII wild-type and MDCKII human ABCG2-transfected). In addition, a cell proliferation assay was used to investigate whether the most potent and promising ABCG2 inhibitors were able to reverse ABCG2-mediated mitoxantrone (MXN) resistance in the MDCKII-hABCG2 cells.

Determination of Cytotoxicity in MDCKII-hABCG2 and MDCKII Wild-Type Cells
The cytotoxicity of the eight N-carboranyl quinazoline amines towards human ABCG2transfected MDCKII cells and their parental MDCKII-WT cells were studied using a WST-1 proliferation assay. Triton X-100 and solvent (0.1% DMSO) were used as the positive and negative control, respectively. Substance-specific IC 50 values (defined as 50% reduced cell viability) could only be calculated for unsubstituted (QCa, Figure S56A, Supplementary Materials) and mono-substituted (QCb, Figure S56B (Table 1). The MDCKII-hABCG2 cells treated with the highest concentrations of 50 µM led to 30% cell viability, thus the IC 50 values could only be given as higher than 10 µM. Due to limited solubility, only stock solutions of 50 mM (DMQCb: 25 mM) were used and, thus, no further IC 50 values could be determined for the target compounds. A decreased cell viability of about 40% was detected for QCc ( Figure S56C, Supplementary Materials) in MDCKII-hABCG2 cells and 30% for DMQCb ( Figure S57B) in MDCKII-WT cells. All other compounds did not significantly affect the viability of MDCKII cells up to 50 µM ( Figures S56 and S57, Supplementary Materials). The comparison of the toxicities between wild-type and transfected cells revealed no significant difference ( Figures S56 and S57, Supplementary Materials), suggesting that the potential inhibitors themselves have a low toxicity. The results indicated that cell viability improves with an increasing degree of substitution. Especially, a higher level of substitution at the 2-phenyl moiety appears to be beneficial ( Figure S56

Biological Investigations
The eight target compounds (QCa-d and DMQCa-d) were evaluated for their cytotoxicity and ability to inhibit the ABCG2 transporter using the Madin-Darby canine kidney cells (MDCKII wild-type and MDCKII human ABCG2-transfected). In addition, a cell proliferation assay was used to investigate whether the most potent and promising ABCG2 inhibitors were able to reverse ABCG2-mediated mitoxantrone (MXN) resistance in the MDCKII-hABCG2 cells.

Determination of Cytotoxicity in MDCKII-hABCG2 and MDCKII Wild-Type Cells
The cytotoxicity of the eight N-carboranyl quinazoline amines towards human ABCG2-transfected MDCKII cells and their parental MDCKII-WT cells were studied using a WST-1 proliferation assay. Triton X-100 and solvent (0.1% DMSO) were used as the positive and negative control, respectively. Substance-specific IC50 values (defined as 50% reduced cell viability) could only be calculated for unsubstituted (QCa, Figure S56A, Supplementary Material) and mono-substituted (QCb, Figure S56B  The Hoechst 33342 accumulation assay was used to evaluate the efficacy of compounds QCa-d and DMQCa-d towards human ABCG2. Since the fluorescent dye Hoechst 33342 (2 -(4-ethoxyphenyl)-5-(4-methyl-1-piperazinyl)-2,5 -bi-1H-benzimidazol) is known to be a substrate of ABCG2, it is actively transported out of the cell. The inhibition of the ABCG2 transporter by the investigated compounds leads to an accumulation of the dye and, thus, increased intracellular fluorescence. The commercially available BCRP inhibitor Ko143 [(3S,6S,12aS)-1,2,3,4,6,7,12,12a-octahydro-9-methoxy-6-(2-methylpropyl)-1,4-dioxopyrazino[1 ,2 :1,6]pyrido [3,4-b]indole-3-propanoic acid 1,1-dimethylethyl ester] was used as reference. The carboranyl quinazoline derivatives were applied in concentrations of 0.5 and 1.0 µM to determine the inhibition of the human ABCG2 transporter in low-and sub-micromolar ranges (Figure 2). These concentrations were selected due to our previous study showing that concentrations of 0.5 µM Ko143 exhibited lower intracellular Hoechst 33342 amounts than 1 µM. To screen whether the investigated compounds are able to inhibit in nanomolar ranges, 0.5 µM of each selected compound was chosen for a first screening approach. While the positive control Ko143 doubled the intracellular dye amount, the parental compound QCa and the methoxy-substituted derivatives QCb and QCd did not affect the human ABCG2-mediated Hoechst efflux ( Figure 2A). Thus, among the unsubstituted quinazoline-scaffold structures (QCa-d), the 2-(3,4-dimethoxyphenyl) quinazoline derivative (QCc) was the only compound exhibiting inhibition within the tested concentration range. The presence of a third methoxy group at the phenyl moiety (QCd) unexpectedly diminished the efficacy towards the human ABCG2 in the Hoechst 33342 accumulation assay (Figure 2A). Juvale et al. investigated similar scaffolds and observed a beneficial effect of the 3,4-dimethoxyphenyl moiety in the 2-position for their compounds [24]. In the second group of compounds (DMQCa-d, Figure 2B), methoxy groups were introduced in the 6-and 7-position of the quinazoline moiety. For DMQCa, an improved inhibitory potency at 1.0 µM was observed, compared to its parental compound QCa. Interestingly, compound DMQCd, bearing five methoxy groups, was by far superior to all other tested compounds, showing the strongest inhibition at 0.5 µM with a comparable level as the reference Ko143. Unexpectedly, DMQCb and DMQCc caused only marginal increases in Hoechst 33342 accumulation ( Figure 2B). However, high background values were observed, indicating a potential autofluorescence of DMQCb and DMQCc, which may have disturbed the fluorescence measurements, leading to an underestimation of ABCG2 interaction. Therefore, this phenomenon was further investigated.

Determination of Autofluorescence
As indicated by the obtained data of the Hoechst accumulation assay, DMQCb and DMQCc seemed to exhibit autofluorescence. Therefore, all investigated compounds were tested for autofluorescence. To detect intracellular autofluorescence, the Hoechst assay was performed in a modified protocol without the Hoechst 33342 dye. After 4 h treatment,

Determination of Autofluorescence
As indicated by the obtained data of the Hoechst accumulation assay, DMQCb and DMQCc seemed to exhibit autofluorescence. Therefore, all investigated compounds were tested for autofluorescence. To detect intracellular autofluorescence, the Hoechst assay was performed in a modified protocol without the Hoechst 33342 dye. After 4 h treatment, the medium was removed, the cells were washed twice, and the buffer solution was entirely removed before the lysis buffer was added. In contrast to the Hoechst 33342 accumulation assay, the Hoechst 33342 dye was not added and, thus, the detected fluorescence is solely intracellular, and an influence of other compounds can be excluded by comparison with the background and the solvent control. As shown in Figure S58 (Supplementary Materials), QCb, QCd, DMQCa, DMQCb, DMQCc, and DMQCd showed a dose-dependent significant increase in the total intracellular fluorescence in comparison with the solvent control. In addition, a treatment with 1 µM of QCc enhanced the intracellular fluorescence in a significant manner. The reference Ko143 emitted no fluorescence itself. This is in line with the data obtained after 5 min in the Hoechst accumulation assay, indicating an increased fluorescence for MDCKII-hABCG2 cells especially treated with DMQCb and DMQCc. The autofluorescence enhanced the background significantly, preventing the determination of hABCG2-mediated Hoechst accumulation. These results proved that QCb, QCc, QCd, DMQCa, DMQCb, DMQCc, or DMQCd are accumulating in MDCKII cells. However, further studies are needed to validate whether these compounds are ABCG2 inhibitors.

Reversal of ABCG2-Mediated MDR
The expression of ABCG2 efflux transporters in several tumors, such as human melanoma and colon cancer, caused chemoresistance and poor 5-year survival rates [4,48,49]. Therefore, it was of interest to investigate whether the quinazoline derivatives reverse the hABCG2mediated mitoxantrone (MXN) resistance in MDCKII cells. The hABCG2 inhibitors identified in the Hoechst assay (QCc, DMQCa, DMQCd, Figure 3A,B,E) and the compounds exhibiting the highest level of autofluorescence (DMQCb, DMQCc, Figure 3C,D) were selected for testing. MXN, as a known substrate for the human ABCG2 transporter, is considerably stronger excreted in the hABCG2-overexpressing MDCKII cell lines, compared with their parental cell line (MDCKII-WT). We have shown that the hABCG2-expressing cell line (MDCKII-hABCG2) exhibited a significantly higher MXN-specific IC 50 value (2.649 ± 0.594 µM) than observed for MDCKII-WT (0.519 ± 0.042 µM). These results underline the hABCG2-mediated decreased toxicity of MXN, which is reversed by 1 µM of the reference compound Ko143 (0.295 ± 0.063 µM). As presented in Figure 3, the combined treatment of MXN with the hABCG2-inhibiting compounds (1 µM) significantly left-shifted the IC 50 value ( Table 2). The strongest left shift (11.6-fold) of MXN-specific IC 50 values was obtained for DMQCd ( Figure 3D). In comparison to the IC 50 value of the combined treatment with Ko143 (1 µM; IC 50 : 0.295 ± 0.063 µM; left-shift factor: 5.7), the addition of DMQCd ( Figure 3E) further reduced the IC 50 value to half the value (0.144 ± 0.033 µM). This is in agreement with the data from the Hoechst 33342 accumulation assay, with DMQCd exhibiting the strongest ABCG2 inhibition among all tested compounds ( Figure 2). Consequently, QCc and DMQa-d were able to reverse MXN resistance by inhibition of the human ABCG2 transporter in MDCKII cells.

In Silico Investigations-Molecular Docking Simulations and Investigation of Mode of Inhibition
Structure-based molecular docking, even though a rather 'rough' approach, represents a rational and useful in silico method to better understand binding modes and interactions between a designed compound and its biological target. As several carboranyl quinazoline derivatives showed cell assay-based inhibition for the ABCG2 transporter protein, we carried out docking simulations in comparison with Hoechst 33342 (in accordance with the literature [50][51][52][53]) and mitoxantrone (MXN, substrate of ABCG2) on the cryo-electron microscopy (cryo-EM) structures of human ABCG2 (PDB code 5NJ3) and human ABCG2 bound to MXN (PDB code 6VXI) [42,43]. Docking simulations of Hoechst and MXN were performed to assess the binding behavior of the carboranyl quinazoline derivatives. Binding of the molecules in the same manner at the binding site could thus indicate competitive inhibition between the inhibitor and substrate. The binding free energy values of all carboranyl derivatives DMQCa-d are given in

In Silico Investigations-Molecular Docking Simulations and Investigation of Mode of Inhibition
Structure-based molecular docking, even though a rather 'rough' approach, repre-  The predicted binding modalities of the carboranyl quinazoline-based structures showed similar orientations and interactions towards the transporter protein. Exemplarily, the best scoring binding pose of compound DMQCd (−8.1 kcal/mol) is shown in Figure 4A,B, as DMQCd exhibited the most promising results in the biological investigations. The predicted simulations were found foremost in the slit-like gap between the ABCG2 monomers and bound mostly by hydrophobic interaction-surrounded by hydrophobic amino acids, including Leu-405, Phe-432, Thr-542, Ile-543, and Val-546. Furthermore, π-π stacking was observed in a sandwich-like manner for the aromatic rings between the phenyl moiety of Phe-439 from the opposing monomers ( Figure 4C). A detailed 2D interaction diagram of DMQCd and the protein side chains is presented in the Supplementary Materials ( Figure S59). This is in agreement with the cryo-EM structures of the small molecule inhibitors in complex with ABCG2 [54]. In total, 98% of the top 200 docking poses of DMQCd discovered from molecular docking experiments using the AutoDock 4.2.6 were occupying the inner binding site, generally referred to as inhibitor binding site S1 [26,45]. In general, all quinazoline derivatives exhibited S1-occupying docking poses in >95% of the top 100 docking poses. However, depending on the substitution pattern of the quinazoline scaffold, we observed deviations in their orientations. The carborane moiety of the dimethoxyquinazoline scaffolds (DMQCa-d) protruded into the inner cavity, suggesting a 'blocking'-like fashion of the molecule inside the binding site, whereas derivatives with an unsubstituted quinazoline scaffold (QCa-d) showed an upward facing (into the extracellular space) orientation of the three-dimensional moiety.
In general, all quinazoline derivatives exhibited S1-occupying docking poses in >95% of the top 100 docking poses. However, depending on the substitution pattern of the quinazoline scaffold, we observed deviations in their orientations. The carborane moiety of the dimethoxyquinazoline scaffolds (DMQCa-d) protruded into the inner cavity, suggesting a 'blocking'-like fashion of the molecule inside the binding site, whereas derivatives with an unsubstituted quinazoline scaffold (QCa-d) showed an upward facing (into the extracellular space) orientation of the three-dimensional moiety.  Additionally, the most abundant predicted docking modes of all examined structures overlaid with the top-ranked poses of MXN in the alleged binding cavity of ABCG2 (Figure 4D). This further agrees with the co-crystallized structure of MXN bound to BCRP (PDB code 6VXI) and the literature [55]. The obtained data indicate that the inhibition of the transporter protein is possibly caused by the competition of the inhibitor and substrate.

Conclusions
The ability of carboranyl 2-phenylquinazoline derivatives to inhibit the human ABCG2 transporter protein and thus reversing ABCG2-mediated multidrug resistance was studied. The 2-phenylquinazoline scaffold has been reported before as the main structure for strong inhibitors. However, despite the ability for strong inhibition, inconsistencies often persist among the inhibitor's intrinsic toxicity and metabolic stability. Therefore, the inorganic three-dimensional carborane moiety was introduced as a non-toxic and metabolically stable pharmacophore. Furthermore, the influence of different methoxy-substitution patterns on the 2-phenylquinazoline scaffold on the inhibitory potency towards the human ABCG2 transporter was investigated.
We discovered five potent compounds (QCc, DMQCa, DMQCb, DMQCc, and DMQCd), exhibiting low cytotoxicity in wild-type and human ABCG2-transfected MDCKII cells, and inhibiting the human ABCG2 transporter at 1 µM. In addition, QCc and DMQCd inhibited the human ABCG2 transporter in Hoechst 33342 accumulation assays in sub-micromolar concentrations (0.5 µM). Among the examined compounds, DMQCd, bearing the highest number of methoxy groups of the tested quinazoline deriv- Additionally, the most abundant predicted docking modes of all examined structures overlaid with the top-ranked poses of MXN in the alleged binding cavity of ABCG2 ( Figure 4D). This further agrees with the co-crystallized structure of MXN bound to BCRP (PDB code 6VXI) and the literature [55]. The obtained data indicate that the inhibition of the transporter protein is possibly caused by the competition of the inhibitor and substrate.

Conclusions
The ability of carboranyl 2-phenylquinazoline derivatives to inhibit the human ABCG2 transporter protein and thus reversing ABCG2-mediated multidrug resistance was studied. The 2-phenylquinazoline scaffold has been reported before as the main structure for strong inhibitors. However, despite the ability for strong inhibition, inconsistencies often persist among the inhibitor's intrinsic toxicity and metabolic stability. Therefore, the inorganic three-dimensional carborane moiety was introduced as a non-toxic and metabolically stable pharmacophore. Furthermore, the influence of different methoxy-substitution patterns on the 2-phenylquinazoline scaffold on the inhibitory potency towards the human ABCG2 transporter was investigated.
We discovered five potent compounds (QCc, DMQCa, DMQCb, DMQCc, and DMQCd), exhibiting low cytotoxicity in wild-type and human ABCG2-transfected MDCKII cells, and inhibiting the human ABCG2 transporter at 1 µM. In addition, QCc and DMQCd inhibited the human ABCG2 transporter in Hoechst 33342 accumulation assays in submicromolar concentrations (0.5 µM). Among the examined compounds, DMQCd, bearing the highest number of methoxy groups of the tested quinazoline derivatives, showed the strongest inhibition of the transporter protein in both tested concentrations comparable to the reference Ko143 (1 µM). DMQCb and DMQCc showed no significant inhibition of the human ABCG2 transporter in a Hoechst 33342 accumulation assay, which was explained and confirmed by autofluorescence of both compounds. Therefore, both derivatives were also included in further assessments. As three possible binding sites within the BCRP are known, molecular docking studies were carried out on the cryo-EM structure of ABCG2 in comparison with the Hoechst 33342 and the antineoplastic agent mitoxantrone, in order to understand the mode of inhibition. The computational methods revealed a strong binding of the carboranyl quinazoline derivatives within the inner, slit-like cavity S1 of the protein. The putative binding modes of the examined structures furthermore overlapped with the binding poses of MXN. This suggests a competitive inhibition behavior between the inhibitor and the substrate. Eventually, the most potent compounds were investigated for reversal of ABCG2-mediated MDR in a WST-1-based cytotoxicity assay by co-administration of the chemotherapeutic agent MXN. All five investigated compounds reversed the MXN resistance by inhibiting human ABCG2 efflux activity. Prominently, the compounds DMQCc and DMQCd showed by far the strongest ABCG2 inhibition with very low cytotoxicity and, strikingly, the ability to reverse ABCG2-mediated resistance in combination with the ABCG2 substrate and antineoplastic agent MXN. DMQCd was able to reverse MDR, causing an 11.6-fold left shift of the MXN-specific IC 50 value, which was even stronger than detected for the reference inhibitor Ko143.
In summary, combining the chemical, computational, and biological methods shown here provides a promising approach for further development of BCRP inhibitors. We have demonstrated that the incorporation of carboranes as a three-dimensional pharmacophore has beneficial effects on the targeted inhibition and reversion of ABCG2-mediated MDR, making these carboranyl quinazoline derivatives promising candidates for further studies.  Acknowledgments: Financial support from the Graduate School Leipzig School of Natural Sciences-Building with Molecules and Nano-objects (BuildMoNa) is gratefully acknowledged.

Conflicts of Interest:
The authors declare no conflict of interest.

Appendix A
A. General procedure for preparation of quinazolin-4-amines In an oven-dried single-neck flask, the sodium hydride (1.2 eq., 60% dispersion in mineral oil) was dissolved in dry THF (5 mL/mmol). The corresponding 2-aminobenzonitrile (1.0 eq.) was dissolved in dry THF (2 mL/mmol) and added dropwise at 0 • C. After warming to room temperature, the nitrile (2.0 eq.) was added dropwise or in small portions. The reaction mixture was heated under reflux for 15−24 h, or until complete conversion (TLC-monitored). After cooling to room temperature, the distilled water (5 mL/mmol) and conc. HCl aq (1.0 eq.) were added, and the organic solvents were removed under reduced pressure. Water (10 mL/mmol) and EtOAc (25 mL/mmol) were added, and the pH was adjusted to 7-8 using 1.5 M NaOH solution. After phase separation, the aqueous phase was further extracted with EtOAc (2 × 30 mL/mmol). The combined organic phases were washed with saturated brine and dried over MgSO 4 . The crude product was purified by column chromatography, using n-hexane/EtOAc as eluents on silica gel.