New structure–activity relationships of chalcone inhibitors of breast cancer resistance protein: polyspecificity toward inhibition and critical substitutions against cytotoxicity

Abstract

Adenosine triphosphate-binding cassette subfamily G member 2 (ABCG2) plays a major role in cancer cell multidrug resistance, which contributes to low eifficacy of chemotherapy. Chalcones were recently found to be potent and specific inhibitors, but unfortunately display a significant cytotoxicity. A cellular screening against ABCG2-mediated mitoxantrone efflux was performed here by flow cytometry on 54 chalcone derivatives from three different series with a wide panel of substituents. The identified leads, with submicromolar IC₅₀ (half maximal inhibitory concentration) values, showed that the previously identified 2′-OH-4′,6′-dimethoxyphenyl, as A-ring, could be efficiently replaced by a 2′-naphthyl group, or a 3′,4′-methylenedioxyphenyl with lower affinity. Such a structural variability indicates 3polyspecificity of the multidrug transporter for inhibitors. At least two methoxyl groups were necessary on B-ring for optimal inhibition, but substitution at positions 3, 4, and 5 induced cytotoxicity. The presence of a large O-benzyl substituent at position 4 and a 2′-naphthyl as A-ring markedly decreased the cytotoxicity, giving a high therapeutic ratio, which constitutes a critical requirement for future in-vivo assays in animal models.

Introduction

Cancer cells display a strong ability to acquire resistance to antitumor drugs, termed multidrug resistance (MDR), which constitutes a critical hurdle to cancer therapy. While several mechanisms can induce cellular resistance to chemotherapeutics, the energy-dependent drug efflux, mediated by adenosine triphosphate-binding cassette (ABC) transporters, is recognized to play a major role. Overexpression of multidrug ABC transporters in tumors alters anticancer drug efficacy by significantly reducing their intracellular accumulation. ABC subfamily G member 2 (ABCG2), also called ABCP for its abundance in placenta,1 BCRP (breast cancer resistance protein),2 or MXR (mitoxantrone resistance protein),3 was the most recently discovered of the three main multidrug ABC transporters overexpressed in cancer cells, after ABC subfamily C member 1 (ABCC1)/MDR protein 1 (MRP1)4 and ABCB1/MDR1/P-glycoprotein.5 ABCG2 was shown to confer resistance to a wide variety of anticancer agents,3 revealing a characteristic polyspecificity toward transport drug substrates. Its clinical relevance was demonstrated in both adult and childhood acute myeloid leukemia,6,7 and it was clearly involved in drug bioavailability at various protection barriers.8,9 Overcoming MDR against anticancer agents, which is of importance for future clinical treatments, might be achieved through effective inhibitors of multidrug ABC transporters. ABCG2 inhibitors were therefore investigated in a number of studies, leading to identification of a few potent compounds.10,11 Different types of inhibitors were characterized: firstly, nonselective inhibitors already known as ABCB1 inhibitors and then considered as dual inhibitors, such as GF120918/elacridar12 and XR9576/tariquidar.13 A few inhibitors appeared to be selective: fumitremorgin C, from Aspergillus fumigatus,14 which was however highly neurotoxic and of which synthetic derivatives, such as Ko143, were prepared.15 Derivatives of XR9576/tariquidar16 and GF120918/elacridar17 were also ABCG2-specific, but displayed cytotoxicity and a limited in-vivo bioavailability.18 A flavonoid-binding site was identified for chrysin and other flavones,19–21 while methoxy-trans-stilbenes appeared less toxic.22 Lower-affinity rotenoids,23 acridone derivatives,17 and chromones24 appeared to bind to the same site, as for a number of flavones, and benzo-pyrane/furane derivatives, of which three-dimensional quantitative structure-activity relationship analyses allowed to establish a molecular model.25,26 Recent studies with chalcones and derivatives (Figure 1) reported a role of substituents, essentially OH, methoxyl (OMe), and Cl, on both A- and B-rings,27,28 but the results were not always easily interpretable since both rings were simultaneously substituted. In addition, we showed that OMe groups on phenyl B-ring, especially at position 4, were quite detrimental towards cytotoxicity.29 Our present aim was therefore to use unsubsti-tuted A-ring for studying the role of a large panel of B-ring substitutions, on both inhibition potency and cytotoxicity. We show here, with a number of chalcone derivatives, that large unsubstituted groups such as 2′-naphthyl or 3′,4′-methylenedioxy-phenyl efficiently replaced the phenyl A-ring, supporting polyspecificity of ABCG2 for inhibitors. At least two OMe were required on the phenyl B-ring for optimal inhibition, although inducing a significant cytotoxicity, whereas most other substituents were not, or only weakly, efficient toward inhibition. Using both a large O-benzyl substituent at critical B-ring position 4 and a 2′-naphthyl group as A-ring additively decreased cytotoxicity and gave a high therapeutic ratio (TR), making chalcone 45 a good candidate for future in-vivo trials in mice models.

Figure 1.

Core structure of generic chalcones (A) and substituted derivatives studied in our previous work (B–K).29

Abbreviations: OMe, methoxyl; Me, methyl.

Materials and methods

Synthesis, purification, and physicochemical analyses of chalcones

All reagents used were of analytical grade, and purchased from Merck KGaA, Darmstadt, Germany or Sigma-Aldrich (St Louis, MO, USA), except for 2,4,6-trimethoxy-acetophenone, xanthoxylin [2-hydroxy-4,6-dimethoxyaceto-phenone], 2-hydroxy-3-bromo-4,6-dimethoxyacetophenone, and 3-methoxy-4-(phenylmetho-xy)-benzaldehyde, which were prepared according to Figure 2, as previously described.30–34 The chalcones were prepared by magnetic stirring of the acetophenone derivative (1.2 mmol), methanol (20 mL), KOH 50% w/v (5 mL), and the corresponding aldehyde (1.2 mmol), at room temperature for 24 hours. Distilled water and hydrochloric acid (10%) were added to the reaction for total precipitation of the products. The compounds were then obtained by vacuum filtration and later recrystallized from dichloromethane and hexane. All chalcones were previously described in the literature.30,32,35–50 The purified compounds were obtained in yields ranging from 40% to 99%. Melting points were determined with a Microquímica MGAPF-301 apparatus (Monte Mor, São Paulo, Brazil). Infrared spectra were recorded with an Abb Bomen FTLA 2000 spectrometer (Zurich, Switzerland) on KBr disks. Nuclear magnetic resonance (¹H and ¹³C) measurements were recorded on a Varian Inc. Oxford AS-400 (Palo Alto, CA, United States) (400 MHz) spectrometer, using tetramethylsilane as an internal standard. Elemental analysis was carried out with a CE Instruments CHNS EA 1110 (Hindley, United Kingdom). Percentages of C and H were in agreement with the product formula (within ± 0.4% of theoretical values to C).

Figure 2.

Synthesis of chalcones.

Notes: a = KOH 50% w/v, CH₃OH, room temperature, 24 hours. In classical chalcones, both A and B are phenyl groups.

Compounds for biological assays

Mitoxantrone, calcein-AM, and 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-tetrazolium bromide (MTT) were purchased from Sigma-Aldrich. All other reagents were commercial products with the highest available purity grade. All chalcone derivatives were dissolved in DMSO (dimethylsulfoxide) and then diluted in Dulbecco’s Modified Eagle’s Medium (DMEM) high glucose medium. The stock solution was stored at −20°C and warmed to 25°C just before use.

Cell cultures

The human fibroblast HEK293 cell lines transfected with either ABCG2 (HEK293-ABCG2) or the empty vector (HEK293-pcDNA3.1 cells) were obtained as previously described.21 Flp-In-293, an isogenic HEK293 cell line, was co-transfected using Lipofectamine™ (Invitrogen, Carlsbad CA, USA) with either the empty vector pcDNAR/FRT or the pcDNA/FRT-ABCC1 vector, in combination with the Flp recombinase vector pOG44, resulting in targeted integration of the expression vector to the same locus in each cell. HEK293 cell lines transfected with ABCB1 (P-glycoprotein) were kindly provided by Dr SE Bates (National Cancer Institute [NCI] at the National Institutes of Health [NIH], Bethesda, MD, USA). All cells were maintained in DMEM high glucose, supplemented with 10% fetal bovine serum, 1% penicillin/streptomycin, and supplemented in some cases with either 0.75 mg/mL G418 (for HEK293-pcDNA3.1 and HEK293-ABCG2), or 2 mg/mL G418 (for HEK293-ABCB1), or 200 μg/mL hygromycin B (for Flp-In-293-pcDNA5/FRT and Flp-In-293-ABCC1).

ABCG2- and ABCB1-mediated mitoxantrone transport

HEK293 cells were seeded at a density of 1 × 10⁵ cells/well into 24-well culture plates. After 48 hours incubation, the cells were exposed to 5 μM mitoxantrone (HEK293-ABCG2 and HEK293-MDR1 cells) for 30 minutes at 37°C, in the presence or absence of compounds at various concentrations. After cell washing with phosphate buffer saline, the cells were trypsinized. The intracellular drug fluorescence was monitored by flow cytometry with a FACS Calibur cytometer (BD Biosciences, San Jose, CA, USA). At least 10,000 events were collected for which the maximal fluorescence (100%) was the difference between geometric mean fluorescence of cells incubated with 5 μM GF120918 and without inhibitor.23 For ABCB1-mediated mitoxantrone transport, the cells transfected with the empty vector were used as a control.

MRP1-mediated calcein transport

HEK293 cells transfected with either ABCC1 or the empty vector were exposed to 0.2 μM calcein-AM and analyzed by flow cytometry as described above. The maximal fluorescence (100%) was the difference between geometric mean fluorescence of control cells (HEK293-pcDNA3.1) and MRP1-transfected cells, incubated with substrate but without inhibitor.

Cytotoxicity assays

HEK293 cells transfected by either pcDNA3.1-ABCG2 (resistant cells) or the empty vector (control sensitive cells) were seeded into 96-well culture plates at a 1 × 10⁴ cells/well density. After overnight incubation, the cells were treated with various concentrations of compounds for 72 hours at 37°C under 5% CO₂. Cell viability was evaluated with an MTT colorimetric assay51 Control experiments were performed with DMEM high glucose containing 0.1% of DMSO (v/v). The results from at least three replicates were expressed as percentage of viable cells versus control cells, taken as 100%. The curves were fitted with the Sigma Plot™ (Systat Software Inc, San Jose, CA, USA) software.

Statistical analysis

Each experiment was performed at least in triplicate. The data are presented as mean ± standard deviation.

Results and discussion

New structure–activity relationships (SARS) among inhibitory chalcones

A total of 54 chalcone derivatives were investigated here, belonging to three different series listed in Tables 1–3: (1) the first series (Table 1), with 23 derivatives, containing a 3 ′,4′-methylenedioxy-phenyl unit as A-ring; (2) the second series (Table 2), with 22 derivatives, containing a 2′-naphthyl group as A-ring; and (3) the third series (Table 3), with nine derivatives, containing a 1-naphthyl group or other substituents as B-ring.

Table 1.

Inhibition of ABCG2-mediated mitoxantrone efflux by chalcones 1–23

First series
Compound	Ring B	% inhibition at 5 μM
1		15.3 ± 6.7
2		26.5 ± 7.1
3		40.9 ± 1.0
4		12.8 ± 7.5
5		13.2 ± 5.2
6		28.8 ± 11.2
7		13.7 ± 7.5
8		24.7 ± 3.9
9		73.0 ± 10.1
10		15.1 ± 4.6
11		31.8 ± 3.8
12		8.3 ± 8.1
13		57.2 ± 8.3
14		33.3 ± 4.0
15		18.7 ± 2.7
16		25.4 ± 2.8
17		59.1 ± 11.2
18		29.8 ± 3.0
19		30.2 ± 6.4
20		18.2 ± 5.2
21		85.7 ± 6.7
22		79.5 ± 5.7
23		73.1 ± 4.5

Notes: The preparation of all compounds is described in the Materials and methods section. Each chalcone was assayed at 5 μM for its ability to inhibit mitoxantrone efflux from ABCG2-transfected HEK293 cells by flow cytometry as described. The relative inhibition was calculated by reference to 5 μM GF120918 producing 100% inhibition.

Abbreviations: OBut, buthoxyl; Me, methyl; OMe, methoxyl.

Table 3.

Inhibition by chalcones 46–54

Third series
Compound	Ring A	Ring B	% inhibition at 5 μM
46			88.2 ± 5.6
47			73.0 ± 2.6
48			49.2 ± 7.2
49			57.0 ± 6.7
50			82.3 ± 11.1
51			21.3 ± 9.0
52			60.9 ± 6.7
53			63.7 ± 12.7
54			55.3 ± 8.9

Note: The conditions were the same as in Table 1.

Abbreviation: OMe, methoxyl.

Table 2.

Inhibition by chalcones 24–45

Second series
Compound	Ring B	% inhibition at 5 μM
24		39.7 ± 6.4
25		16.7 ± 6.8
26		9.2 ± 1.7
27		2.0 ± 1.4
28		42.9 ± 1.2
29		22.5 ± 8.8
30		23.8 ± 13.1
31		11.0 ± 2.8
32		49.7 ± 1.5
33		96.9 ± 7.4
34		17.4 ± 8.1
35		33.7 ± 8.7
36		14.6 ± 6.9
37		29.0 ± 11.0
38		6.7 ± 1.3
39		24.1 ± 6.9
40		1.2 ± 2.9
41		31.8 ± 5.3
42		7.5 ± 1.9
43		23.2 ± 7.7
44		42.1 ± 8.8
45		81.5 ± 4.4

Note: The conditions were the same as in Table 1.

Abbreviations: OBut, buthoxyl; Me, methyl; OMe, methoxyl.

New structure–activity relationships were identified by comparing the inhibitory activity of each compound at 5 μM against ABCG2-mediated mitoxantrone efflux, measured by flow cytometry and using 5 μM GF120918 as a reference for complete inhibition (Tables 1–3). The following substitutions were identified to positively contribute to inhibition.

1. A 3′,4′-methylenedioxy-phenyl unit, replacing A-ring in compounds of the first series (Table 1), was as efficient in 9 as the active 2′-OH-4′,6′-diOMe-phenyl (A-ring) substituent previously identified in chalcone i (Figure 1).29

2. A 2′-naphthyl group replacing A-ring in compounds of the second series (Table 2) was also quite efficient in 33, as indolyl in chalcones j and k (Figure 1).29 Such a polyspecificity toward A-ring chalcones is consistent with that observed for different classes of flavonoidic and flavonoid-like inhibitors, assumed to bind to the same site or to overlapping sites, such as hydrophobic flavones,19–21 rotenoids,23 acridones,17 tariquidar derivatives,16 methoxy trans-stilbenes,22 and other chalcones.27–29 Polyspecificity of inhibitory sites is also consistent with the well known polyspecificity of ABCG252 and other multidrug ABC transporters toward large panels of transport substrates. It is worthwhile mentioning that the gain-of-function R482T/G mutation in ABCG2, which was found to extend the transport substrate panel to rhodamine 123 and anthracyclins,53 also allowed inhibitors such as GF120918 and imatinib to be transported.52

3. In both series, at least two OMe (in 9, 22, 23, and 33) were required for optimal inhibition when comparing with compounds with only one OMe (8, 19, 32) or without any OMe (1–7, 10–18, 20, 24–31, 34–43). The only exception was 44, which suggested a negative contribution of OMe at position 6 versus position 5 in 33. One of the two OMe could be efficiently replaced by O-benzyl at position 4 (in 21 and 45). Interestingly, an O-benzyl substituent was also shown to positively contribute to ABCG2 inhibition in chromones.24

4. In contrast, all other tested substituents on B-ring were poorly efficient, such as Cl (in 1–4, 17, and 26–29), Br (in 7 and 38), F (in 15, 16, 36, and 37), NO₂ (in 5, 6, 24, and 25), methyl (Me) (in 10, 18, 31, and 43), N(Me)₂ (in 11 and 34) and buthoxyl (in 14 and 35), similarly as using a naphthyl group to replace B-ring (in 12, 13, 40, and 41).

5. Comparable positive contributions were observed from the third series of chalcones (Table 3): the requirement for at least two OMe (in 46, 50, 52, 53, and 54), or one OMe and one O-benzyl (in 47), while the low efficiency of 51 was most likely related to the charged carboxyl preventing diffusion through the plasma membrane and therefore the compound uptake.

Selectivity for ABCG2 and cytotoxicity

The 54 chalcones described in this work were then investigated for their ability to inhibit mitoxantrone efflux by ABCB1/MDR1/P-glycoprotein, and calcein efflux by ABCC1/MRP1. As shown in Figure 3, all chalcones were essentially specific for ABCG2, since no significant inhibition was produced in most cases. A low inhibition of ABCB1 activity was however observed at 10 μM with two strong ABCG2 inhibitors, 9 and 33, both containing three OMe groups on the B-ring. Three other compounds slightly inhibited ABCC1-mediated drug efflux, by reference to MK571: 34 and 11 containing a 4-N(Me)₂ group on B-ring, and 48 containing a 2′-OH on A-ring. The N(Me)₂ substituent present in 11 and 34 might allow these chalcones to be transported, as observed for ABCB1 transport substrates,54 which could explain lower specificity. In contrast, the three potent ABCG2 inhibitors containing a large 4-O-benzyl (ie, 21, 45, and 47) were highly selective and did not inhibit ABCB1 nor ABCC1.

Figure 3.

Inhibition by chalcones of P-glycoprotein (ABCB1)- and MRP1 (ABCC1)-drug efflux activity. All chalcones from the three series were checked individually at 10 μM by flow cytometry for their ability to inhibit either ABCB1-mediated mitoxantrone efflux (grey bars) or ABCC1-mediated calcein efflux (black bars). The control of complete inhibition was performed with either 5 μM GF120919 or 30 μM MK571, respectively.

Abbreviations: ABC, adenosine triphosphate-binding cassette; CTRL, control.

The most potent compounds, inhibiting by more than 70% at 5 μM, were further analyzed for both their IC₅₀ (half maximal inhibitory concentration) values by flow cytometry (Table 4), and cytotoxicity by an MTT cell survival test (Figure 4A and B). The different compounds greatly varied in cytotoxicity, with 9, containing a 3′,4′-methylenedioxy-phenyl as A-ring and a 3,4,5-triOMe substituted B-ring, being the most cytotoxic with an IG₅₀ (concentration producing 50% growth inhibition) value of 4.4 μM, giving a low TR (TR = IG₅₀/IC₅₀) of 7, similarly to 22 with a 2,5-diOMe substitution (TR = 8). A rather high cytotoxicity was also observed for 23, with 3,4-diOMe substitution (IG₅₀ = 12.9 μM, TR = 24), and 46, with a 3,4-methylenedioxy-phenyl instead of B-ring, which confirmed the cytotoxic role of OMe at positions 3, 4, and 5 of phenyl B-ring previously observed when A-ring was an OH- and OMe-substituted phenyl.29 Therefore, the 3′,4′-methylenedioxy-phenyl as A-ring behaved similarly to the above substituted phenyl, but with a 3–4-fold lower affinity. By contrast, the addition of O-benzyl at position 4 was positive in 21, versus 23, by increasing threefold the TR (70 versus 24), as due to both increased affinity for inhibition and decreased cytotoxicity (Figure 4A). The effect was even more marked in 47 with a 4′-Br-phenyl as A-ring (TR = 99). Quite interestingly, the presence of a naphthyl group not only increased inhibition affinity but also lowered the cytotoxic effect of OMe substituents, especially when present as A-ring in 33 (IG₅₀ = 34.9 μM and TR = 152) versus B-ring in 50 (IG₅₀ = 20.6 μM and TR = 94) (Figure 4B and Table 4). This naphthyl positive effect on cytotoxicity appeared additive to that induced by 4-O-benzyl in 45 (IG₅₀ > 100 μM, TR > 435). Such decreases in cytotoxicity might be possibly due to the prevention of interactions with unknown cellular target(s) inducing cell death.

Table 4.

TR values of potent chalcone inhibitors

Compound	A-ring substituents	B-ring substituents	IG50 (μM)	IC50 (μM)	TR
9	methylenedioxy-phenyl	3,4.5-triOMe-phenyl	4.4 ± 0.5	0.64 ± 0.16	7
22	methylenedioxy-phenyl	2,5-diOMe-phenyl	8.1 ± 1.3	1.07 ± 0.20	8
46	2′,4′-diOMe-phenyl	methylenedioxy-phenyl	11.7 ± 1.8	0.64 ± 0.12	18
23	methylenedioxy-phenyl	3,4-diOMe-phenyl	12.9 ± 1.6	0.54 ± 0.06	24
21	methylenedioxy-phenyl	3-OMe,4-OBz-phenyl	17.5 ± 1.2	0.25 ± 0.12	70
47	4′-Br-phenyl	3-OMe,4-OBz-phenyl	28.7 ± 5.4	0.29 ± 0.04	99
50	2′-Oh, 4′,6′-diOMe-phenyl	1-naphthyl	20.6 ± 0.8	0.22 ± 0.07	94
33	2-naphthyl	2,4,5-triOMe-phenyl	34.9 ± 2.5	0.23 ± 0.06	152
45	2-naphthyl	3-OMe,4-OBz-phenyl	>100	0.23 ± 0.07	>435

Notes: The cytotoxicity IG₅₀ values were evaluated from Figure 4A and B. The IC₅₀ values were determined with increasing inhibitor concentrations up to 10 μM by flow cytometry as in Tables 1–3.

Abbreviations: IC₅₀, half maximal inhibitory concentration; IG₅₀, concentration producing 50% growth inhibition; OBz, O-benzyl; OMe, methoxyl; TR, therapeutic ratio.

Figure 4.

Intrinsic cytotoxicity on HEK293 cells and sensitization of BCRP (ABCG2)-transfected cells to mitoxantrone. The survival upon 72-hour culture of control sensitive cells in the presence of increasing concentrations of the indicated chalcones was quantified by MTT assays, as described in the Materials and methods section. Compared intrinsic cytotoxicities are shown in (A) for 21 (black triangles), 22 (black circles), 23 (white circles), and 47 (white triangles), and in (B) for 9 (black circles), 33 (white triangles), 45 (black squares), 46 (white circles), and 50 (black triangles). The sensitization to mitoxantrone, in (C) of the growth of ABCG2-transfected resistant cells (black circles) by 1 μM 50 (black triangles) reached a similar level as control sensitive cells (white circles).

Abbreviations: ABC, adenosine triphosphate-binding cassette; MTT, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-tetrazolium bromide; HEK293, human embryonic kidney HEK293.

The potent inhibition of lead compounds against ABCG2-mediated mitoxantrone efflux in flow cytometry experiments was confirmed by their ability to chemosensitize cell growth of resistant cells (IG₅₀ = 82.8 ± 21.7 nM) to the cytotoxic drug, as illustrated in Figure 4C for 50: a 1 μM noncytotoxic concentration of the chalcone efficiently chemosensitized cell growth to mitoxantrone, giving an IG₅₀ value of 17.5 ± 2.2 nM, similar to that of sensitive control cells (14.5 ± 3.6 nM). A comparable efficiency was observed with 45 (IG₅₀ = 16.4 ± 3.1 nM), whereas 22 only produced a partial sensitization (26.7 ± 11.4 nM) (data not shown) in agreement to its 4–5-fold lower affinity for inhibition (see Table 4). The fact that the same low concentrations of 50 and 45 chemosensitized cell growth in MTT assays as efficiently as they inhibited drug efflux in flow cytometry suggests that these chalcones were not mainly metabolized nor transported during the 72-hour timescale of the experiments. Chalcone 45, which is a potent and selective inhibitor with low toxicity, therefore constitutes a good candidate for future in-vivo assays in animal models.

Conclusion

Original data about ABCG2 inhibitory sites and molecular mechanism have been obtained in the present work, thanks to a wide variety of chalcones and derivatives. Firstly, the flavonoid-specific inhibitory site appeared to be polyspecific since it could accommodate diverse large substituents instead of the original phenyl A-ring, with 2′-naphthyl and 3′,4′-methylenedioxy-phenyl behaving quite efficiently toward inhibition. Secondly, substitution at position 4 of the phenyl B-ring was quite critical toward both inhibition and cytotoxicity, with a major role played by an O-benzyl group. The present data of cytotoxicity obtained from in-vitro screening on culture cells constitute an interesting first step that requires to be further developed for the identified lead compounds, such as chalcone 45, toward in-vivo pharmacokinetics and pharmacodynamics in mice.