Magnolol derivatives as specific and noncytotoxic inhibitors of breast cancer resistance protein (BCRP/ABCG2)

Abstract

The breast cancer resistance protein (BCRP/ABCG2) transporter mediates the efflux of numerous antineoplastic drugs, playing a central role in multidrug resistance related to cancer. The absence of successful clinical trials using specific ABCG2 inhibitors reveals the urge to identify new compounds to attend this critical demand. In this work, a series of 13 magnolol derivatives was tested as ABCG2 inhibitors. Only two compounds, derivatives 10 and 11, showed partial and complete ABCG2 inhibitory effect, respectively. This inhibition was selective toward ABCG2, since none of the 13 compounds inhibited neither P-glycoprotein nor MRP1. Both inhibitors (10 and 11) were not transported by ABCG2 and demonstrated a low cytotoxic profile even at high concentrations (up to 100 μM). 11 emerged as the most promising compound of the series, considering the ratio between cytotoxicity (IG₅₀) and ABCG2 inhibition potency (IC₅₀), showing a therapeutic ratio (TR) higher than observed for 10 (10.5 versus 1.6, respectively). This derivative showed a substrate-independent and a mixed type of inhibition. The effect of compound 11 on the ABCG2 ATPase activity and thermostability revealed allosteric protein changes. This compound did not affect the expression levels of ABCG2 and increased the binding of the conformational-sensitive antibody 5D3. A docking study showed that 11 did not share the same binding site with ABCG2 substrate mitoxantrone. Finally, 11 could revert the chemoresistance to SN-38 mediated by ABCG2.

Graphical Abstract

1. INTRODUCTION

Multidrug resistance (MDR) related to cancer has been described as a major challenge in oncological therapy [1]. Among several biological mechanisms of MDR, the overexpression of ATP-binding cassette (ABC) transporters is considered the leading cause of MDR in neoplastic cells [2]. These overexpressed ABC transporters increase the efflux of chemotherapeutic drugs, decreasing their intracellular accumulation to subclinical concentrations, and leading the protocols to their failure [3]. ABC transporters are described as polyspecific due to their ability to export, in eukaryotic cells [4], a wide range of drugs with unrelated chemical structures and cellular targets [5]. The human genome encodes 48 ABC proteins [6] and among them, three ABC transporters are closely involved with the MDR phenotype: glycoprotein-P (P-gp or MDR1), multidrug resistance protein 1 (MRP1) and breast cancer resistance protein (BCRP or ABCG2) [7].

The use of inhibitors of ABC transporters is the most common strategy to overcome MDR. The first identified ABC transporter related to MDR was P-gp [8]. Nevertheless, clinical studies of P-gp inhibitors still fail to improve the chemotherapeutic efficacy of oncological protocols [9–12]. One of the diverse reasons for the reported clinical failure is that many drugs transported by P-gp are also transported by ABCG2 and other ABC transporters [13]. Since P-gp and ABCG2 are both overexpressed in several cancers, and despite recent advances, there are very few potent, non-toxic ABCG2 inhibitors available for clinical studies, demonstrating that the development of new ABCG2 inhibitors is an urgent necessity. During the last few years, we have identified certain classes of compounds, naturally occurring or not, as potent ABCG2 inhibitors, including chromones [14,15], stilbenes [16], chalcones [17], indeno[1,2-b]indoles [18–21] and porphyrins [22].

Indeed, natural products are a great source of chemical scaffolds for new drugs and represent a powerful starting point for ABCG2 inhibitors. Flavonoids [23], botryllamides [24], and curcumin [25] are among the natural compounds identified as ABCG2 inhibitors. Among the most widely researched polyphenols, two dimeric neolignans, magnolol (1) and its isomer honokiol (2), isolated from the bark of Magnolia spp and constituted by a bisphenolic core bearing two allyl side chains, have shown numerous and increasing research data [26].

Numerous preclinical studies have verified an antitumoral activity exerted by 1 against different types of cancer, such as lung, prostate, breast, gallbladder, colon, skin and liver [27–32]. In vitro and, subsequently, in vivo studies have shown that magnolol has a satisfactory safety profile [33,34], contributes to the reduction of tumor growth, induces apoptosis and inhibits invasion, migration and metastasis [32,35,36].

Some bisphenol neolignans inspired by magnolol (1), and synthesized through enzymatic dimerization and regioselective ortho-hydroxylation, showed inhibitory activities against tankyrase-2 [37] and yeasts α-glucosidase enzymes [38]. The former plays an important role in cancer, whereas the latter is an important target for development of new antidiabetic drugs. Other studies on bisphenol neolignans inspired by honokiol pointed out their antioxidant [39], antiproliferative [40] and anti-obesity activity as lipase inhibitor [41].

Additional studies suggest that magnolol, honokiol and 4-O-methylhonokiol may be promising agents in combating MDR through down-regulation of P-gp (ABCB1) expression levels [42]. More recently, the two isomers were described as promising inhibitors of ABCG2 activity and down-regulators of expression levels [43]. In this study, we evaluated the inhibitory potential of magnolol derivatives (new and previously described structures) toward the main MDR-related ABC transporters and the mechanism of ABCG2 inhibition of the most promising compounds was characterized.

2. MATERIAL AND METHODS

2.1. GENERAL EXPERIMENTAL PROCEDURE

NMR spectra were run on a Varian Unity Inova spectrometer operating at 500 MHz (¹H) and 125 MHz (¹³C). Chemical shifts (δ) were indirectly referred to TMS using residual solvent signals (δ 3.31 for CD₃OD, 7.26 for CDCl₃, 2.05 for (CD₃)₂CO). High-resolution mass spectra were acquired with a Q Exactive Orbitrap mass spectrometer (Thermo Fisher Scientific, Bremen, Germany) equipped with an ESI ion source operating in negative ion mode. Compound 13 (1 × 10⁻⁵ Min 50:50 MeOH:H₂O + 1% formic acid) was directly infused in the mass spectrometer. A survey scan was performed from m/z 150 to 1000 at 140 K resolution.

HPLC analyses were carried out using an Agilent Series (Milan, Italy) G1354A pump and an Agilent UV G1315D as a diode array detector. An Agilent Series 1100 G1313A autosampler was used for sample injection. Samples were purified by semi-preparative HPLC-UV onto a RP-18 (Phenomenex luna, 10um, 250 × 10 mm) with a gradient of CH₃CN (solvent b) in water (solvent a) at 3 ml/min: t_0min b = 90%, t_1min b = 100%, t_7min b = 100%, t_9min b = 90%.

Thin-layer chromatography (TLC) was conducted using pre-coated silica gel F254 plates (Merck Millipore, Milan, Italy). Each reaction mixture was observed under UV light at wavelengths of 254 and 366 nm or through staining with a cerium sulfate solution.

All chemicals were of reagent grade and were used without further purification. Horseradish peroxidase (type I), tyrosol, and homovanillyl alcohol were purchased from Merck (Milan, Italy). Candida antarctica lipase (Chirazyme, L-2, c.-f. C2, lyo) was purchased from Roche (Monza, MB, Italy). Magnolol and eugenol were purchased from TCI Europe N.V. (Zwijndrecht, Belgium). IBX was prepared according to the literature [44].

2.2. CHEMISTRY

The syntheses of compounds 3 – 10 were achieved as previously reported in Pulvirenti et al. and Baschieri et al. [38,45]. Compound 3 was obtained by regioselective ortho-hydroxylation of magnolol (1) with IBX (2-iodox-ybenzoic acid). Compounds 4 and 6 were obtained by enzymatic dimerization of eugenol and tyrosol in the presence of horseradish peroxidase type I (HRP). Compound 5 was obtained from 4 after a reaction with IBX. Compound 7 was obtained from tyrosol after enzymatic acetylation with Candida antarctica lipase (CAL) followed by enzymatic dimerization in the presence of HRP. Compound 8 was obtained from homovanillyl alcohol after enzymatic acetylation with CAL and subsequent dimerization with HRP. Compounds 9 and 10 were obtained from 8 after IBX-mediated oxidative demethylation. The monomethylated and dimethylated magnolol derivatives 11 (Scheme 1) and 12, were obtained from 1 by a reaction with methyl iodide. Compound 14 was obtained from 1 after catalytic hydrogenation. Compound 15 was obtained from 11 after catalytic hydrogenation as well. The NMR spectroscopic data of compounds 3 – 10 were compared with those previously reported [38]. Also the spectroscopic characterization of compounds 11, 12, 14 and 15 were in agreement with previously published data [46–48]. All the reactions are summarized in Supplementary Material (Scheme S1-6 and Fig. S1-3).

Scheme 1.

Synthesis of 13. a) dry acetone, K₂CO₃ (2 equiv), MeI (2 equiv), 56 °C, 4h. b) 11 (0.2 M in MeOH), IBX (1.2 equiv), 0 °C, 30 min; Na₂S₂O₄, rt, 5 min.

2.2.1. SYNTHESIS OF 5,5’-diallyl-2’-methoxy-[1,1’-biphenyl]-2,3-diol (13)

Compound 11 (171.6 mg, 0.61 mmol) was dissolved in MeOH (3.0 mL, 0.2 M), then IBX (205.2 mg, 0.73 mmol) was added (See Supplementary material). The solution was stirred at 0 °C until the disappearance of the substrate (30 min). Then, a Na₂S₂O₄ solution (83.5 mg, 0.81 mmol in 8.9 mL of H₂O) was added, and the solution was stirred for 5 min at rt. The crude reaction was concentrated in vacuo, and the residue was solubilized with ethyl acetate (20 mL) and partitioned with a saturated NaHCO₃ solution (10 mL). The aqueous phase was extracted with ethyl acetate (2 × 10 mL). The organic phases were washed with a saturated NaCl solution and dried over Na2SO4. After filtration, the solvent was evaporated under vacuum. The flash chromatography with DIOL Silica-gel, eluted with n-hexane: CH₂Cl₂ (30:70 → 0:100) and n-hexane:CH₂Cl₂: (100 → 97:3) afforded the neolignan 13 with 86% purity. This fraction was further purified by semi-preparative HPLC-UV onto a RP-18 (Phenomenex Luna C18, 10um, 250 × 10 mm) with a gradient of CH₃CN (solvent B) in water (solvent A) at 3 mL/min: t_0min B = 90%, t_1min B = 100%, t_7min B = 100%, t_9min B = 90%.

The desired compound was recovered with 11.6 % of yield (22.6 mg): Rf (TLC) 0.46 (100% CH₂Cl₂). ¹H-NMR (500 MHz,CDCl₃): δ = 7.21–7.19 (m, 2 H, H-6’ and H-4’), 6.99 (d, J = 8.5 Hz, 1H, H-3’), 6.82 (d, J = 2.0 Hz, 1 H, H-4), 6.70 (s, 1 H, OH), 6.65 (d, J = 2.0 Hz, 1 H, H-6), 6.00–5.95 (m, 2 H, H-8/8’), 5.85 (s, 1 H, OH), 5.12–5.04 (m, 4 H, CH₂-9/9’), 3.91 (s, 3 H, OMe), 3.31 (d, J = 7.0 Hz, 2 H, CH₂-7’), 3.34 (d, J = 7.0 Hz, 2 H, CH₂-7). ¹³C-NMR (125 MHz): δ = 153.5 (C, C-2’), 146.3 (C, C-3), 139.0 (C, C-2), 137.7 (CH, C-8), 137.4 (CH, C-8’), 134.4 (C, C-5’), 133.8 (C, C-5), 132.5 (CH, C-6’), 129.3 (CH, C-4’), 127.4 (C, C-1), 126.8 (C, C-1’), 121.9 (CH, C-6), 116.0 (CH, C-3’), 115.8 (CH₂, C-9), 114.4 (CH₂, C-9’), 112.3 (CH, C-4), 56.9 (CH₃, OMe), 39.9 (CH₂, C-7), 39.4 (CH₂, C-7’). HRESIMS m/z 295.1368 [M - H]⁻, calcd for C₁₉H₁₉O₃ m/z 295.1334.

2.3. CELL LINES

The human fibroblast HEK293 cell line was stably transfected to overexpress ABCG2 (HEK293-ABCG2) as previously described [49]. The murine fibroblast NIH3T3 was stably transfected to overexpress P-gp (NIH3T3-ABCB1) as previously described [50]. BHK21 cell line was stably transfected to overexpress MRP1 (BHK21-ABCC1) as previously described [51]. Both parental and stably transfected cell lines were kindly provided by Dr. Attilio Di Pietro (IBCP, Lyon, France). All cell lines were cultivated in High Glucose modified Dulbecco’s Modified Eagle Medium, supplemented with 10% fetal bovine serum, 1% penicillin/streptomycin and maintained at 37 °C and 5% CO₂ under controlled humidity. The stably transfected cells were additionally treated with 0.75 mg mL⁻¹ of G418 (HEK293-ABCG2), 60 ng mL⁻¹ of colchicine (NIH3T3-ABCB1) or 0.1 mg mL⁻¹ methotrexate (BHK21-ABCC1). The cells were cultivated until reaching 80–90% of confluence and then used for experimentation.

2.4. INHIBITION ASSAY

The transport inhibition assay was carried out with cells overexpressing the ABC transporters: ABCG2 (HEK293-ABCG2), P-gp (NIH3T3-ABCB1) and MRP1 (BHK21-ABCC1). Cells were seeded at the density of 1.5×10⁵ cells/well in 24 well culture plates and incubated for 48 h at 37 °C under 5% CO₂. Cells were treated with compounds 3 – 12 (10 or 50 μM) and substrate (5 μM of mitoxantrone, 5 μM rhodamine 123 and 0.15 μM of calcein-AM for ABCG2, P-gp and MRP1, respectively). Stably transfected cells treated with the reference inhibitors Ko143 (0.5 μM), GF120918 (0.5 μM) and verapamil (30 μM) were used as positive inhibition controls for ABCG2, P-gp and MRP1 respectively. The negative control consisted of stably transfected cells exposed only to their respective substrates. Cells in all conditions (negative/positive controls and tested conditions) were incubated for 30 min at 37 °C and 5% CO₂. After treatment, cells were washed with PBS (400 μl), detached with trypsin (50 μl) and resuspended in ice-cold PBS (300 μl). The intracellular fluorescence was monitored with a FACS Calibur cytometer (Becton Dickinson) using Red Laser/FL4 channel for mitoxantrone and Blue Laser/FL1 channel for rhodamine 123 and calcein-AM, with at least 10,000 events acquired inside the gate (geometrical delimitation of higher density and most homogeneous cells within a dot plot). The maximal fluorescence (assumed as 100%) was determined by the median value of intracellular accumulation of substrates in stably transfected cells incubated with their specific reference inhibitors (or using their respective parental cell lines). A minimum ratio of 2 between the maximum and minimum fluorescence values obtained by the positive and negative inhibition controls, respectively, was necessary to validate each experiment. The percentage of transport inhibition was calculated by the following equation:

$$\mathrm{\%}\mathit{inhibition}=(C-S)/(I-S)\times 100$$

where “C” corresponds to the intracellular fluorescence of cells in the presence of the tested compounds and their specific substrate, “S” corresponds to the intracellular fluorescence of cells in the presence of the substrate alone, and “I” corresponds to the intracellular fluorescence of cells in the presence of both their specific substrate and reference inhibitor. The IC₅₀ curves of ABCG2 inhibition were obtained with increasing concentrations of magnolol derivatives (0.39 – 50 μM) using mitoxantrone as fluorescent substrate.

2.5. CONFOCAL MICROSCOPY

HEK293-ABCG2 cells were seeded at a density of 1×10⁵ cells/well in 24-well culture plates containing coverslips for microscopy and incubated for 48 h at 37 °C under 5% CO₂. Cells that adhered to the coverslips were treated with 50 μM of inhibitor and 1 μM of the ABCG2 substrate Hoechst 33342 for 30 min at 37°C and 5% CO₂. After incubation, the coverslips were removed from the plate and placed on slides for microscopy. The slides were then read in a confocal microscopy Nikon A1R MP + (NIKON, Tokyo, Japan) using an oil-immersed 40X objective (with the numerical aperture of 1.15). A laser of 405 nm was used for excitation, and the fluorescence emission was recorded using a bandpass filter of 425–475 nm. The software Nis Elements 4.20 (NIKON, Tokyo, Japan) was used for the visualization of the images.

2.6. CELL VIABILITY ASSAYS

Cell viability was evaluated with a 3-(4,5-dimethylthiazol-2-yl)-2,5-iphenyltetrazolium bromide (MTT) colorimetric assay [52]. HEK293 and HEK293-ABCG2 cells were seeded into 96-well culture plates at 1.5×10⁴ cells/well density. After overnight incubation, the cells were treated with various concentrations of magnolol derivatives (0.19 – 100 μM) for 72 h at 37 °C under 5% CO₂. For reversion tests (sensitization), cells were simultaneously treated with a saturation concentration of 11 and 0.05 μM of SN-38 and incubated under the same conditions. After the treatment, the culture medium was removed and 100 μL of a 0.5 mg mL⁻¹ of MTT solution was added. Cells were then incubated for 4 h at 37°C under 5% CO₂. The formazan crystals were dissolved with a solution of DMSO/ethanol (1:1) and the absorbance was measured at 570 nm using a Multiscan FC microplate reader (Thermo Scientific). The results were expressed as percentage of viable cells versus control cells (0.1% DMSO, v/v) taken as 100%.

2.7. mRNA EXPRESSION LEVELS BY qPCR

Total RNA was obtained of HEK293-ABCG2 cells from cultures with approximately 90% of confluence after treatment for 72 h with 11 (50 μM). The total RNA isolation was performed using TRIzol (Invitrogen) protocol according to the manufacturer’s instructions. A NanoDrop^™ Spectrophotometer was used to quantify RNA concentration and the integrity was evaluated by 1% agarose gel electrophoresis and, subsequently, RNA was stored at −80 °C. Two micrograms of total RNA were reverse transcribed using High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems) according to the manufacturer’s instructions, and the resulting cDNA was stored at −20 °C. Using cDNAs as the template, quantitative real-time PCR (qPCR) was performed using the SYBR Green PCR Master Mix (Applied Biosystems) in a 7500^™ Real-Time PCR Detection System (Applied Biosystems). A dissociation cycle was performed after each run to check for non-specific amplification or contamination. The mRNA expression levels were normalized using the geNorm 3.4 software, and the corresponding housekeeping gene expression levels. Sets of specific primers were designed using Primer designing tool - NCBI and validated through BLAST and BLAT, and their respective sequences are shown in Table 1. Relative expression levels were estimated using the method described by Pfaffl [53].

TABLE 1 -.

NUCLEOTIDE SEQUENCES OF PRIMERS USED FOR qPCR.

Gene	NCBI reference	Sequence	Reference
PPIA	NM_021130.5	F- TAAAGCATACGGGTCCTGGC R- TGCCATCCAACCACTCAGTC	[54–58]
RPS13	NM_001017.3	F- CGTCCCCACTTGGTTGAAGT R- TGAATCTCTCAGGATTACACCGA	[55,58,59]
HPRT1	NM_000194.3	F- CAGGGATTTGAATCATGTTTGTGT R- ACTCCAGATGTTTCCAAACTCAAC	[54,58,60,61]
ABCG2	NM_001257386.2	F - ATGGTCTGTTGGTCAATCTCAC R – TTATGCTGCAAAGCCGTAAATCC	[58]

2.8. ATPase ASSAY

The ATPase activity was determined as previously described [62]. High-Five insect cell total membranes overexpressing ABCG2 were used at a concentration of 5 μg protein/tube in a final volume of 100 μL. The membranes were incubated in assay buffer with the following composition: 50 mM Tris–HCl, pH 6.8, 150 mM N-methyl-D-glucamine (NMDG)-Cl, 5 mM sodium azide, 1 mM EGTA, 1 mM ouabain, 2 mM DTT, and 10 mM MgCl₂, in the presence or absence of sodium orthovanadate (0.3 mM). The protein-buffer mix was treated with the inhibitor at increasing concentrations (0 – 50 μM) and incubated for 20 min at 37 °C in the presence of ATP (5 mM). The reaction was stopped with the addition of 100 μL of 5% SDS, 400 μL of Pi solution (sulfuric acid 36.2 N, water, ammonium molybdate and antimony potassium tartrate) and 200 μL of 1% ascorbic acid. After 10 min, the absorbance was measured at 880 nm using Ultrospec 3100 pro spectrophotometer (Amersham Biosciences, UK).

2.9. CONFORMATIONAL ANTIBODY BINDING (5D3)

HEK293-ABCG2 cells were seeded at a density of 2×10⁵ cells/well in 24 well culture plates and incubated for 48 h at 37 °C under 5% CO₂. After incubation the cells were washed with PBS, detached by trypsin, collected with 300 μL of PBS and centrifuged (1000×g for 3 min). The cell pellet was resuspended in 100 μL of PBS containing 4 μL of a BSA solution (1 mg/mL). Cells were treated with inhibitor at 50 μM for 10 min at 37 °C. The 5D3 primary antibody (Purified mouse anti-human CD338, BD Pharmigen - 1:100) was added and incubated at 37 °C for 30 min. After incubation, centrifugation was performed at 1000×g for 3 min and the cell pellet was resuspended in 100 μL of PBS containing the secondary antibody conjugated with phycoerythrin (PE Goat anti-mouse IgG, Abcam - 1:200) and incubated at 37 °C for 30 min. After incubation, centrifugation was performed at 1000×g for 3 min and the pellet was resuspended in 300 μL of ice-cold PBS. The flow cytometry analysis was performed as previously described, with a FACS Calibur cytometer (Becton Dickinson) using the blue laser (488 nm) and FL-2 filter.

2.10. THERMOSTABILITY ASSAY

The thermal stability assay was performed as previously described [63]. Total membrane prepared from High Five insect cells overexpressing ABCG2 were incubated (3 μg protein/tube) with assay buffer with the following composition: 50 mM Tris–HCl, pH 6.8, 150 mM N-methyl-d-glucamine (NMDG)-Cl, 5 mM sodium azide, 1 mM ouabain and 2 mM DTT, in the presence or absence of 0.3 mM orthovanadate at a final volume of 50 μL. To evaluate the effect of 11 over thermal stability, each sample was prepared with 12.5 mM MgCl₂ or 6.25 mM ATP and incubated with a temperature range from 37 to 71 °C for 10 min using a thermocycler C1000 Touch (Bio-Rad, Hercules, CA). After incubation, 10 μL of 25 mM ATP or 50 mM MgCl₂ (5 and 10 mM final concentration, respectively) was added and incubated at 37 °C/20 min to allow ATP hydrolysis. The reaction was stopped with the addition of 50 μL of Pi solution containing 1% ammonium molibdate, 2.5 N H₂SO₄ and 0.014% potassium-antimony tartrate. To evaluate the absorbance, the samples were transferred to a 96 well plate (50 μL/well) and added 150 μL of 0.33% sodium ascorbate solution. The absorbance was measured after 15 min of incubation at room temperature using the microplate reader Spectramax iD3 (Molecular Devices, San Jose, CA). The sensitive activity to vanadate (Vi) was calculated as the difference between the activity in the absence of Vi and the activity in the presence of Vi to each temperature.

2.11. DOCKING AND ADMET STUDIES

Molecular docking studies were performed using the ABCG2 structure deposited in the Protein Data Bank under the ID code 6VXI [64]. The inhibitors were structurally optimized by applying the Mopac 2016 program (MOPAC2016, James J. P. Stewart, Stewart Computational Chemistry, Colorado Springs, CO, USA, http://OpenMOPAC.net (2016) by using the PM7 method [65]. The structures for ABCG2 and inhibitors were prepared for use with AutoDock Vina [66] using the prepare_ligand.py script from AutoDock Tools (v.1.5.6). The protein structure had unsolved side chains completed and their protonation state estimated through Maestro PrepWizard. For molecular docking, an area of 29.41 × 30.33 × 32.55 Å and coordinates X= 0.201, Y= −0.921 and Z= −2.044 was delimited. Two initial strategies were used: first, docking on the apoprotein and then docking on the structure containing a redocked mitoxantrone molecule. After, the resulting structure with 11 was used to dock a second molecule of 11, to simulate possible interactions between ABCG2 and this compound at high concentrations. The exhaustiveness parameter was set to 75 and a maximum number of 9 binding modes was used. Docked models of the inhibitors, binding sites and amino acid interactions were visualized with ABCG2 using PYMOL (Molecular Graphics System, Version 1.3, Schrödinger, LLC) and PLIP web server [67], the latter being used to determine the intermolecular interactions between the molecules and the side chains of the transporter. Physicochemical, pharmacokinetic (absorption, distribution, metabolism and excretion), and toxicity properties of 11 and the reference inhibitor Ko143 were calculated using the ADMETlab 2.0 webtool [68], by inserting the molecules SMILES strings into the screening tab.

3. RESULTS

3.1. CHEMISTRY

A total of 13 magnolol derivatives were included in this study. Compounds 3 - 10 were resynthesized according to a chemo-enzymatic procedure previously reported by Pulvirenti et al. [38]. The detailed reaction conditions are summarized in Supplementary Material. The synthesis of 13 is reported for the first time herein (Scheme 1). This derivative was achieved by semi-synthesis into two steps: methylation of magnolol afforded the monomethyl derivative (11) with 89.4%, then, a selective hydroxylation carried out with IBX gave the expected compound 13 with 11.6% yield. The compound was characterized by HRMS spectrometry and by ¹H and ¹³C NMR spectroscopy, thus confirming the obtained compound has the expected OH functionality in ortho position to OH.

3.2. IDENTIFICATION OF 11 AS A SELECTIVE AND NONCYTOTOXIC ABCG2 INHIBITOR

A total of 13 magnolol derivatives (Table 2) were evaluated as potential inhibitors of ABCG2 transporter by examining their effect on mitoxantrone ABCG2-mediated efflux.

TABLE 2 -.

SCREENING OF MAGNOLOL DERIVATIVES ON ABC TRANSPORTERS (ABCG2, P-GP AND MRP1)

	Inhibition (% ± SD)
	Compound	ABCG2 (BCRP)		ABCB1 (P-gp)		ABCC1 (MRP1)
		10 μM	50 μM	10 μM	50 μM	10 μM	50 μM
	3	2.5 ± 3.6	6.9 ± 3.8	−0.3 ± 2.0	−0.3 ± 2.8	3.3 ± 2.8	6.2 ± 1.3
	4	6.1 ± 7.9	12.8 ± 13.2	−0.8 ± 2.0	0.0 ± 2.4	2.7 ± 4.5	2.2 ± 1.5
	5	0.6 ± 2.0	1.0 ± 0.5	0.4 ± 1.9	−1.0 ± 2.0	3.5 ± 4.0	2.2 ± 3.0
	6	0.9 ± 2.7	3.6 ± 2.8	−0.3 ± 2.0	−0.5 ± 1.5	1.7 ± 2.0	3.2 ± 3.5
	7	10.1 ± 2.7	21.9 ± 3.4	0.0 ± 1.2	−0.1 ± 1.6	3.8 ± 2.8	9.0 ± 5.8
	8	7.0 ± 2.1	21.5 ± 4.7	0.3 ± 0.6	−0.4 ± 1.3	4.3 ± 3.4	9.9 ± 5.2
	9	0.1 ± 1.8	2.7 ± 0.5	0.0 ± 2.3	−0.3 ± 1.6	2.5 ± 2.0	10.0 ± 6.8
	10	21.8 ± 12.0	53.1 ± 8.5	−0.6 ± 1.4	−0.4 ± 1.6	1.0 ± 1.2	6.9 ± 3.2
	11	62.0 ± 11.9	105.4 ± 13.8	−0.7 ± 1.2	2.8 ± 3.8	4.9 ± 7.0	11.5 ± 5.8
	12	5.9 ± 10.9	32.8 ± 5.5	−1.1 ± 1.5	−0.5 ± 2.1	2.6 ± 1.8	9.1 ± 3.3
	13	4.0 ± 4.4	19.5 ± 4.4	1.1 ± 4.6	−1.2 ± 1.8	2.9 ± 4.3	4.7 ± 6.4
	14	1.8 ± 1.7	19.5 ± 4.5	0.5 ± 0.9	−6.4 ± 0.3	5.6 ± 5.5	16.0 ± 10.7
	15	1.1 ± 5.4	20.4 ± 1.7	0.2 ± 0.8	2.0 ± 1.9	4.3 ± 3.9	5.8 ± 4.1

These bisphenolic neolignans were mainly obtained starting through enzymatic dimerization of simple natural phenols (eugenol, tyrosol and homovanillic alcohol) as well as through chemical modification of magnolol. The products differ from magnolol scaffold for various structural details: ortho insertion of hydroxyl (as in compound 3) and/or methoxy (4, 5) groups; simple side chain modifications through insertion of hydroxyl (6) or acetoxyl group (7); combination of acetylated side chain with the addition of methoxy (8), hydroxyl (9) or both (10) groups. In compounds 11 – 13, the magnolol scaffold was decorated with hydroxyl or methoxy groups; 14 and 15 are the hydrogenated analogues of 1 and 11.

Stably transfected cells overexpressing ABCG2 (HEK293-ABCG2) were incubated with these compounds at concentrations of 10 and 50 μM for 30 min. The intracellular fluorescence of mitoxantrone was monitored by flow cytometry. One magnolol derivative (10) showed a partial inhibition (~50% of inhibition) at the higher concentration tested. Only compound 11 showed a total inhibition (~100% of inhibition) of mitoxantrone ABCG2-mediated efflux (Table 2).

Additionally, the selectivity toward ABCG2 was evaluated by comparing the effect of these magnolol derivatives on P-gp and MRP1 activity using rhodamine 123 and calcein-AM as fluorescent substrates, respectively. Stably transfected cells overexpressing P-gp (NIH3T3-ABCB1) and MRP1 (BHK21-ABCC1) were used. As shown in Table 2, none of these compounds inhibited neither P-gp nor MRP1 activities. These results suggest that magnolol derivatives are a class of compounds that exclusively affects the activity of ABCG2 transporter. Once the chemical structure of the two most potent compounds (10 and 11) are very different, the data of table 2 are not enough to perform a correct structure-activity relationship study with this class of compounds.

HEK293-ABCG2, NIH3T3-ABCB1 and BHK21-ABCC1 cells were exposed to mitoxantrone (5 μM), rhodamine 123 (5 μM) and calcein-AM (0.15 μM) respectively, and magnolol derivatives (10 and 50 μM). Ko143 (0.5 μM), GF120918 (0,5 μM) and verapamil (30 μM) were used as positive controls, which produce 100% inhibition, for ABCG2, P-gp and MRP1 respectively. Intracellular accumulation of the fluorescent substrates was monitored by flow cytometry using FL4-H (mitoxantrone) and FL1-H (rhodamine 123 and calcein-AM) channels. Results were expressed as percent of inhibition related to positive controls. The data represents the mean ± SD of, at least, three independent experiments.

The effect of these two most promising magnolol derivatives on ABCG2 was studied by different approaches. HEK293-ABCG2 cells were submitted to a short-term (30 min) treatment using increasing concentrations of 10 and 11 to determine the IC₅₀ values (concentration that produce the half-maximal inhibition). Compounds 10 and 11 showed IC₅₀ values of ABCG2 inhibition of 22.6 (Fig. 1A) and 9.5 μM (Fig. 1B), respectively. In addition, these data revealed that 10 is a partial ABCG2 inhibitor, in contrast to 11, that is a complete ABCG2 inhibitor, such as the reference inhibitor Ko143 (Fig. 1C).

Figure 1.

Inhibition curves of magnolol derivatives and histogram overlays. IC₅₀ curves of ABCG2 inhibition for 10 (A) and 11 (B). HEK293-ABCG2 cells were exposed to mitoxantrone (5 μM) and magnolol derivatives (10 and 11 at 0.039 – 50 μM). Ko143 (0.5 μM) was used as positive control (100% inhibition). Results were expressed as percent of inhibition related to the positive control. (C) Histograms of ABCG2 inhibition: 10, 11 and Ko143.

The cytotoxicity effect of drugs in parental and cells overexpressing the target ABC transporter can provide valuable information about a possible ABCG2-mediated cross-resistance or collateral sensitivity [69]. HEK293 and HEK293-ABCG2 cells were treated for 72 hours with increasing concentrations of 10 and 11 (Fig. 2). Compound 10 was cytotoxic at high concentrations (≥ 50 μM). The similar cytotoxicity pattern for both cell lines suggests an absence of cross-resistance, which means that 10 is not recognized as an ABCG2 substrate (Fig. 2A). In addition, 11 did not showed a cytotoxic effect even at 100 μM, the highest concentration tested (Fig. 2B).

Figure 2 -. Cell viability assay and therapeutic ratio representation.

Cytotoxic profile of the two best compounds, 10 (A) and 11 (B). HEK293 and HEK293-ABCG2 cells were submitted at a long-term treatment (72 hours) with increasing concentrations of magnolol derivatives (0.19 – 100 μM). (C) Inhibition potency and cytotoxicity of the compounds with their respective therapeutic ratios.

Considering the two most important parameters for identification of new ABCG2 inhibitors, potency of inhibition and intrinsic cell cytotoxicity, the therapeutic ratio (TR) values were calculated as the ratio between the IG₅₀ value (concentration that reduces 50% of the cell viability) and IC₅₀ value of inhibition. As shown in figure 2C, 11 was the best compound, showing a TR almost ten times greater than 10. Together, these data highlight that 11 is the most promising compound from this series.

3.3. STUDY OF 11 ON ABCG2 USING MEMBRANE-BASED APPROACHES

Substrates and inhibitors of ABC transporters commonly exert opposite effects on the ATPase activity, stimulating and inhibiting, respectively [70,71]. The effect of 11 on the ABCG2 ATPase activity was studied using total membranes from High-Five insect cells overexpressing the ABCG2 transporter. As shown in Fig. 3A, this magnolol derivative triggered a dual effect, stimulating (~60%) the ATPase activity of ABCG2 at low concentrations and inhibiting (~30%) at the higher concentration (50 μM).

Figure 3 -.

Effects of 11 on ABCG2 ATPase activity and protein thermostability. (A) Effect of 11 at increasing concentrations (0.001 – 50 μM) on basal ABCG2 ATPase activity. Thermostability assay in the absence (B) and the presence (C) of 11. The data represent mean ± SD of three independent experiments performed in duplicate.

Additionally, the effect of 11 on the thermostability of the protein was studied in membranes of insect cells overexpressing ABCG2. The thermostability of a protein is often altered due to conformational changes induced by a ligand [72]. To study the binding of 11 on ABCG2 and possible allosteric conformational changes, a thermostability assay in the presence and absence of ATP was performed. Interestingly, 11 decreased the IT₅₀ values compared to the control (DMSO) (Fig. 3B and C), suggesting that the binding of this magnolol derivative destabilizes the protein structure.

3.4. STUDY OF 11 ON ABCG2 USING CELL-BASED APPROACHES

To additionally characterize the mechanism of inhibition caused by 11, the functional inhibition was studied using other substrates of ABCG2. The intracellular fluorescence of Hoechst 33342 was analyzed by confocal microscopy. The results obtained by compound 11 were compared with the effect caused by the reference inhibitor Ko143. A diffuse fluorescent pattern was observed in the absence of inhibitors (Fig. 4A). Whereas, similarly to Ko143, 11 promoted an increase in the intracellular levels of Hoechst 33342 (Fig. 4A), confirming that 11 is not a substrate-specific inhibitor.

Figure 4 –

Mechanism of ABCG2 inhibition caused by compound 11. (A) Intracellular accumulation of Hoechst 33342 (1 μM) in HEK293-ABCG2 cells by confocal microscopy. Ko143 (0.5 μM) was used as control and 11 was tested at 50 μM. (B) Effect of 11 on mRNA expression levels. (C and D) Histograms from the conformational 5D3 antibody shift assay. Overlay of untreated control and cells treated with Ko143 2 μM (C) and 11 50 μM (D). (E) Lineweaver-Burk plot using different concentrations of 11 and mitoxantrone as substrate.

To further explore the mechanism of inhibition, the effect of 11 on mRNA ABCG2 expression levels was studied by qPCR. As shown in figure 4B, 11 did not significantly affect the transcriptional levels of ABCG2, suggesting that this compound is not an ABCG2 expression modulator, but a functional ABCG2 inhibitor. In addition, the effect of 11 on protein conformation was investigated by the 5D3 shift assay. This assay consists in the use of the conformational antibody 5D3, which recognizes an epitope in the extracellular loop of ABCG2. As shown in Fig. 4C and D, the magnolol derivative triggered an increase of the 5D3 binding similar to the reference inhibitor Ko143, suggesting that the interaction of 11 on ABCG2 caused important conformational changes in the extracellular region of the transporter.

The absence of ABCG2-mediated transport of 11 suggests that this magnolol derivative does not share the same binding region or site with ABCG2 substrates. To pursue this hypothesis, the type of inhibition was investigated using increasing concentrations of both mitoxantrone and 11. The data showed an increasing of Vmax and Km while inhibitor concentration increased, suggesting a mixed type of inhibition, which can be visualized by the Lineweaver-Burk plot (Fig. 4E).

3.5. STUDY OF 11 USING in silico APPROACHES

A molecular docking study was performed to better understand the binding site of 10 and 11 on ABCG2. As shown in figure 5A, similarly to the reference inhibitor Ko143, 11 binds in the transmembrane region of ABCG2 (apo structure). To compare the binding of 10 and 11 on ABCG2, the inactive analogue of 11, that corresponds to compound 12, was also used. The difference between these two analogues corresponds to a methoxy group substitution (12) instead of a hydroxyl group (11), which leads to opposite inhibitory effects. The analysis using the ABCG2 apo structure revealed an overlap of the binding site for all three molecules (Fig. 5B), with a binding affinity of −8.5 kcal/mol for 10 and −9.7 kcal/mol for both 11-12. All molecules bound preferentially to the central pocket of drug binding site (drug binding cavity), and the pair of analogues 11-12 additionally presented identical binding conformation (Fig. 5B).

Figure 5 -. Docking of Ko143, magnolol derivatives and mitoxantrone on ABCG2.

(A) Binding site of Ko143 and 11 (spheres representations in purple and orange, respectively) and (B) sequential docking analysis of 10 (in yellow), 11 (in orange) and 12 (in green) on ABCG2 structure (PDB: 6VXI) without mitoxantrone (MTX). (C) Redocked MTX interactions with ABCG2. (D) Binding site comparison of mitoxantrone molecules (original MTX from 6VXI in light blue and redocked MTX in dark blue). (E) Sequential docking analysis and interactions of 10, 11 and 12 (in yellow, orange, and green, respectively) with ABCG2 structure containing MTX (in dark blue). 11 (F) and 12 (G) interactions with ABCG2 in the presence of MTX. (H) Sequential docking and interactions involving two molecules of 11. ABCG2 chains are light gray (A chain) and dark gray (B chain), hydrophobic interactions (C, F, G and H) displayed as dashed gray lines. Amino acids in B, D and E for spatial reference only. Figures were generated using PYMOL (Molecular Graphics System, Version 1.3, Schrödinger, LLC) and PLIP web server [67].

In addition, the docking analysis revealed that all molecules were stabilized through pi-stacking interactions between the F439 residues of the A and B chains and the aromatic rings of the biphenolic core (Fig. 5B), as observed for mitoxantrone alone (Fig. 5C). A docking analysis was also performed with two other small molecules that inhibit ABCG2, chalcone 27 [17] and stilbene 9 [16], which showed binding affinities of −8.2 and −9.2 kcal/mol, respectively (Fig. S1B and C). These compounds also presented pi-stacking interactions between their aromatic rings and F439 residues. Additionally, chalcone 27 binding was also supported by two H-bonds with N436A and T452B, such as Ko143 (Fig. S4A and C). In summary, an overlap in the ABCG2 binding site was observed for all these inhibitors (Fig. S4D).

Interestingly, the docking analysis using the ABCG2 structure containing the redocked substrate mitoxantrone, which presents the same binding region as the original molecule (Fig. 5D), revealed a shift of 10, 11 and 12 binding sites (Fig. 5E). This shift was different for 11 (Fig. 5F) when compared with 12 (Fig. 5G), which should implicate in a decreased binding affinity value (−7.1 vs −6.4 kcal/mol, respectively). This different binding mode in presence of the substrate could be related to the lack of a H-bond donor in 12, which is present on 11 through the hydroxyl group and a H-bond with Q398A sidechain (Fig. 5F). In addition, no specific intermolecular interactions other than hydrophobic contacts were identified in the docking of the structure in the presence of mitoxantrone. Curiously, the 11-11 sequential docking resulted in a second molecule binding in a similar conformation and identical energy (−7.1 kcal/mol) (Fig. 5H). Finally, the different binding sites of 11 and mitoxantrone (Fig. 5E) support the mixed type of inhibition caused by this inhibitor (Fig. 4E). Although we can assume that 11 occupies a different site of mitoxantrone, it should induce conformational changes on protein, which affects Km value of substrate.

Further, an in silico analysis to investigate physicochemical and ADMET (Absorption, Distribution, Metabolism, Excretion and Toxicity) features of 11 and Ko143 was also performed. The main parameters of interest regarding those properties are summarized in table 3.

TABLE 3 –

PHYSICOCHEMICAL AND ADMET CALCULATION

Compound	Physicochemical		Absorption		Distribution
	MW	LogP	F20%	F30%	PPB	Fu
11	280.15	5.169	0.744	0.983	100.859%	0.668%
Ko143	469.26	3.768	0.012	0.034	95.423%	1.903%
	Metabolism		Excretion		Toxicity
	iCYP1A2	iCYP2C19	CL	T ½	DILI	RAO
11	0.968	0.930	10.798	0.298	0.078	0.042
Ko143	0.045	0.453	9.540	0.709	0.905	0.886

Physicochemical and ADMET properties of 11 and Ko143. MW = Molecular Weight; F20% and F30% represent the probability (0–1) of oral bioavailability lower than 20 and 30%, respectively; PPB = Plasma Protein Binding; Fu = Fraction unbounded; iCYP1A2 and iCYP2C19 = probability (0–1) to inhibit CYP1A2 and CYP2C19. CL = clearance (ml/min/kg); T ½ = probability (0–1) of having a half-life lower than 3 hours; DILI = probability (0–1) of inducing liver damage; RAO (Rat Oral Acute Toxicity) = probability (0–1) of being highly toxic.

3.6. SENSITIZATION STUDY OF 11 TO ANTICANCER DRUGS

To verify the potential utilization of magnolol derivatives as ABCG2 inhibitors in further pre-clinical studies, an in vitro chemosensitization assay using cells overexpressing ABCG2 was performed after 72 h of co-treatment using 11 and SN-38 (active metabolite of irinotecan which is transported by ABCG2).

As shown in Fig. 7, transfected cells (HEK293-ABCG2) are resistant to the treatment of SN-38 when compared to HEK293 cells. HEK293-ABCG2 cells exposed to the co-treatment (11 + SN-38) showed a significant decrease in the cell viability when compared to the treatment with SN-38 alone. Additionally, HEK293-ABCG2 cells co-treated with 11 + SN-38 or Ko143 (reference inhibitor) + SN-38 showed the same percentage of cell viability as the observed for HEK293 cells exposed to SN-38 alone. These findings demonstrate that 11 was able to completely revert the MDR phenotype mediated by ABCG2.

Figure 7 – Chemosensitization assay using of stably transfected cells overexpressing ABCG2.

Cell viability after 72 h of treatment. HEK293-ABCG2 and HEK293 cells treated with SN-38 alone and HEK293-ABCG2 cells in co-treatment with SN-38 and 11 at saturating concentration or SN-38 and Ko143 at 0.5 μM (positive control). Data represents mean ± SD of at least three independent experiments performed in triplicate. Groups were statistically compared using one-way ANOVA (Sidak’s multiple comparisons test). NS = non-significant; ***p<0.001 and ****p<0.0001.

4. DISCUSSION

Magnolol (1) core structure and its structural isomer honokiol (2) have been recently associated with a down-regulation of ABCG2 expression levels and ABCG2 functional inhibition [43]. Here, we screened 13 magnolol derivatives to verify their potential as specific ABCG2 inhibitors (Table 2). Unfortunately, the nature and the number of the substituents in the magnolol scaffold from this series of compounds did not allow us to establish a precise structure-activity relationship study. The best magnolol derivative was 11, which is structurally similar to the precursor scaffold. The single difference between 11 and the magnolol core structure was the replacement of the hydroxyl group of magnolol by a methoxy group.

In general, the presence of methoxy groups increases the ABCG2 inhibition effect, as reported by stilbenes [16,17]. Despite the increase of the ABCG2 inhibition observed by the replacement of a hydroxyl group with a methoxy group (9 versus 10), the presence of a second methoxy group decreased the inhibition effect (10 versus 8). The same negative effect of the replacement of a hydroxyl group by a methoxy group was also observed by 11 versus 12 (Table 2). This negative effect on ABCG2 inhibition was observed when replacing the two allyl groups (11) with two propyl groups (15), and the replacement of a methoxy group with a hydroxyl group did not produce any effect (15 versus 14). Interestingly, the presence of a second hydroxyl group (13) in the structure abrogates the inhibition effect. These data suggest that a single chemical change in the studied compounds can markedly modify their biological effects. These findings are not commonly observed for other classes of ABCG2 inhibitors, such as chalcones [17], stilbenes [16], indeno[1,2-b]indoles [73], and chromones [15], where a single substitution commonly shows a mild impact in the potential of inhibition.

Only 11 and 10 showed more than 50% of ABCG2 inhibition (Table 2), and 11 showed a lower IC₅₀ value (Fig. 1) and a non-cytotoxic profile (Fig. 2). Compound 11 showed an IC₅₀ value of 9.5 μM when mitoxantrone was used as substrate. In addition, 11 should be considered a complete inhibitor since it produced a saturation effect of inhibition at 100%, similar to the reference inhibitor Ko143 (Fig. 1C). For ABCG2, the maximal inhibition effect is an important characteristic to be considered during the screening of new inhibitors [74]. Another desirable feature is the selectivity toward one single ABC transporter [75,76]. The selectivity toward ABCG2 can reduce collateral effects due the interference on homeostasis processes and the pharmacokinetics of other drugs [77]. In this study, all magnolol derivatives were selective toward ABCG2, showing no inhibition on P-gp (Table 2).

A second important feature of ABC transporter inhibitors is the absence of transport mediated by these proteins [74]. Interestingly, magnolol and honokiol have been previously described as substrates of ABCG2 [43]. Here, a cell viability assay investigated the hypothesis of a transport mediated by ABCG2. As shown in Figure 2B, 11 was not cytotoxic even when using high concentrations (10-fold the IC₅₀ value). However, 10 showed a similar cytotoxic effect in both parental and cells overexpressing ABCG2. These data suggest that this magnolol derivative is not recognized as a substrate of ABCG2 (Fig. 2A). Substrates of ABCG2 constitute a good scaffold resource for the development of new inhibitors, as described by stilbenes and porphyrins [16,22,74]. A third important feature of inhibitors is the therapeutic ratio, which consists of the ratio between the potency of inhibition (IC₅₀ value) and cytotoxicity (IG₅₀ value). In this study, 11 showed a lower IC₅₀ value and an estimated cytotoxic effect > 100 μM, thus, 11 showed a TR at least 6-fold higher than 10. The TR observed for 11 was similar to the porphyrin 4B, recently described as a new selective ABCG2 inhibitor [22], but very lower than the observed for chromone 6g (MBL-II-141) [15] and Ko143 [78,79] (TR of ~ 2000 and ~2600, respectively). However, despite being more potent (IC₅₀ = 0.01 μM) [79], Ko143 is more cytotoxic than 10 and 11 (IG₅₀ = 18 – 34 μM) [80].

To better understand the mechanism of inhibition, membrane-based approaches were used to evaluate the effect of 11 on ATPase activity and thermostability of ABCG2. The efflux of substrates mediated by ABC transporters through the cellular membrane depends on the binding and hydrolysis of ATP [2] and, in general, functional inhibitors (compounds that inhibit the transport activity) also inhibit the ATPase activity, such as Ko143, a reference ABCG2 inhibitor [81]. However, differently of P-gp, it is not uncommon for ABCG2 that functional inhibitors stimulate the ATPase activity, such as stilbenes [16], tetrahydroquinoline/4,5-dihydroisoxazole molecular hybrids [82], indeno[1,2-b]indoles [20], Ulixertinib [83] and Rociletinib [84]. Here, we observed a rare and complex effect triggered by 11, a dual behavior: stimulation at low concentrations and inhibition at high concentrations (Fig. 3A). Despite rare, this effect already was described for nilotinib and gefitinib [85,86]. The thermostability assay showed a decrease in protein stability in the presence of 11 and ATP, since the IT₅₀ decreased by approximately 10 °C. This result was similar to the observed for porphyrin 4B [22] and confirmed structural protein changes induced by 11 binding (Fig. 3C).

Cell-based approaches were also employed to better understand the inhibition mechanism of 11. Confocal microscopy allowed us to verify that this magnolol derivative was able to inhibit the ABCG2-mediated efflux of Hoechst 33342, similarly to Ko143 (Fig. 4A), and confirming that the ABCG2 inhibition caused by 11 is independent of the substrate. A kinetic experiment was conducted to better understand the type of inhibition caused by 11. Using increasing concentrations of the inhibitor and mitoxantrone as substrate, 11 showed a mixed type of inhibition (Fig. 4E). The same type of inhibition was recently observed for porphyrin 4B [22]. Another way to study the interaction of ligands on ABCG2 is by the 5D3 shift assay [74]. Generally, the binding of the conformation-sensitive antibody 5D3 on the extracellular loop of ABCG2 is increased by inhibitors, and not affected or even decreased by substrates [87–89]. 11 showed the same behavior that functional inhibitors (Fig. 4C and D), confirming conformational changes due to 11 binding.

It was found that 11 binds to the transmembrane region that interacts with ABCG2 physiological substrate uric acid (Q398, N436, L539, T542 and V546) during translocation [90], and other ABCG2 inhibitor stravatinib (Q398, N436, V401) [91]. The docking analysis revealed that 11 shares the same ABCG2 binding site with the two others ABCG2 inhibitors that have a similar size: chalcone 27 [17] and stilbene 9 [16]. In addition, chalcone 27 and stilbene 9 showed a similar binding conformation as 11, suggesting that these inhibitors are stabilized inside the central pocket of ABCG2 mainly by pi-stacking interactions with F439A/B (Fig. S4). In the case of 12, the inactive analogue of 11, a significant smaller affinity was determined by the docking algorithm. It is possible that the presence of the hydroxyl in 11, acting as a hydrogen bond donor with Q398 and other amino acid residues of the protein, has a relevant role anchoring the compound and triggering inhibition. Although it is not possible to exclude the ability of 12 to interact through water-mediated hydrogen bonds with its methoxy groups, the stabilization promoted is possibly inferior to that present in 11, being insufficient for the molecule to be able to bind firmly to the transporter and sterically prevent the constriction of the transmembrane helices of the central cavity, characteristic of the progression from the catalytic cycle to the substrate transport step [92]. The docking analysis of 10 suggests that a neighboring methoxy group could promote a more stable H-bond with the protein side chains, possibly by decreasing the acidity of the hydrogen donor. This observation partially explains the inhibitory effect of 8 and 10 in contrast to 9. Additionally, the docking of two 11 molecules indicated that the molecules should preferentially bind to the F439 (substrate) region, due to the better binding affinity, with a second molecule able to bind to the Q398 region, when the first is occupied. This model fits the results obtained from the ATPase assay, with the first 11 molecule stimulating ATP hydrolysis (substrate site). As the concentration of 11 increases, the second site is filled, effectively preventing the necessary conformational steps for NBDs dimerization and hydrolytic activity.

The physicochemical and ADMET calculations estimated that 11 presented reduced toxicity in comparison to Ko143, showing a much lower probability of causing hepatotoxicity. However, the high LogP value of 11 might imposes potential difficulties in terms of the water solubility. In addition, the lower oral bioavailability and the high percentage of plasma protein binding suggests a difficult tissue penetration, which may be compensated by a slower excretion rate. Together, these data suggest that a drug delivery system is required for preclinical studies with 11.

Finally, the co-treatment of cells with 11 and SN-38 successfully chemosensitized cells overexpressing ABCG2 to the anticancer drug, confirming that 11 is an effective inhibitor to overcome MDR in cancer mediated by ABCG2. In summary, the magnolol derivative 11 is a promising complete ABCG2 inhibitor, that differently of the magnolol core structure does not affect the expression levels and is not transported by ABCG2 [43]. 11 is selective toward ABCG2 and trigger allosteric effects, binding in a different site that substrates. Finally, the absence of cytotoxicity observed for 11 and its ability to revert MDR phenotype make this inhibitor attractive for future pre-clinical studies.

HIGHLIGHTS.

• 13 magnolol derivatives were tested as ABCG2 inhibitors.

• The monomethoxylated magnolol derivative 11 showed complete ABCG2 inhibition.

• Magnolol derivative 11 was not transported, noncytotoxic, and showed a specific ABCG2 inhibition.