Flavonoid Structure-Activity Studies Identify 6-Prenylchrysin and Tectochrysin as Potent and Specific Inhibitors of Breast Cancer Resistance Protein ABCG2

Intrinsic or acquired drug resistance in tumors is a main reason for the failure of anticancer chemotherapy. The multidrug resistance (MDR) cell phenotype has classically referred to tumors that are refractory to chemically unrelated cytotoxic agents, as due to their active outward transport. Several ATP-binding cassette (ABC) transporters are involved in MDR, including P-glycoprotein (MDR1/ABCB1; ref. 1), the multidrug resistance protein (MRP1/ABCC1; ref. 2), and the more recently discovered breast cancer resistance protein (ABCG2/BCRP/MXR/ABCP; refs. 3–5). These transporters are located in the plasma membrane and extrude a variety of drugs by using the energy of ATP hydrolysis. ABCG2 is a “half transporter” that homodimerizes to form an active transporter (6), but it has also been detected as tetramers or higher forms of oligomerization (7). The spectrum of ABCG2-effluxed anticancer drugs includes mitoxantrone, camptothecin derivatives, methotrexate, and anthracyclines (8). The precise substrate profile depends on a hotspot mutation at position 482 such that methotrexate is only transported by wild-type ABCG2 (R482) and anthracyclines and rhodamine 123 only by mutant ABCG2 (R482T or R482G), whereas Hoechst33342 and mitoxantrone are transported by all variants (9). Although single-nucleotide polymorphisms have been found at several residues in ABCG2, only arginine has been identified at position 482 in clinical isolates (10). Although the substrate spectrum is highly overlapping with P-glycoprotein, this is not the case for inhibitors, which may explain the failure of attempts to circumvent clinical MDR with P-glycoprotein inhibitors (11). The wide tissue distribution of ABCG2 also supports its role in controlling the permeability of important pharmacologic barriers (8). Therefore, ABCG2 inhibitors are of high interest, not only as reversal agents of clinical drug resistance, but also as agents that might improve the pharmacokinetics of substrate chemotherapeutic drugs by enhancing their oral bioavailability, plasma half-lives, and brain and fetal penetration. Recently, we and others have exhaustively studied natural and hemisynthetic flavonoids as modulators of P-glycoprotein (12–15), MRP1 (16–18), and related parasite (19, 20) and yeast (21) multidrug transporters. Flavonoids constitute a group of interesting polyphenolic compounds with a wide distribution in plants (22). As ABCG2 has been described to confer resistance to flavopiridol (23), a flavonoid-like cell cycle inhibitor currently in clinical trials, we analyzed the ability of flavonoids to modulate the transporter activity and found a potent inhibitory activity for some flavonol derivatives (24). Several publications then appeared, describing the interaction of a number of natural flavonoids inducing either inhibition of drug efflux in drug-selected ABCG2-overexpressing cells (25, 26) or reversal of multidrug resistance in transfected cells (27), but no correlation between flavonoid structure and activity was evident. Our aim was to establish structure-activity relationships for the inhibition of mitoxantrone efflux by wild-type ABCG2-transfected cells by screening a number of structurally related derivatives. The present results showed that the hydrophobic flavone 6-prenylchrysin was a potent inhibitor, 15-fold more efficient than chrysin. Interestingly, the R482T mutation limited this prenylation-dependent effect, whereas the inhibitory potency of tectochrysin was even increased, especially towards rhodamine transport. Both 6-prenylchrysin and tectochrysin were able to sensitize ABCG2-transfected cells to mitoxantrone, with low intrinsic cytotoxicity and no significant interaction with either P-glycoprotein or MRP1. These compounds therefore might be considered as potent and specific ABCG2 modulators with potential clinical interest.

Chemical compounds. Mitoxantrone and rhodamine 6G were purchased from Sigma (L'isle d'Abeau, Chesnes, France), and nonprenylated flavonoids from Extrasynthèse (Genay, France). Prenylated derivatives of chrysin were obtained as previously described (15, 28). GF120918 was provided by GlaxoSmithKline (Madrid, Spain).

Cell cultures. The human fibroblast HEK-293 lines transfected with either wild-type or mutant ABCG2 or the empty vector were obtained as described (29). The human breast cancer cell lines MDA-MB-231 stably transfected by either pcDNA3 or pcDNA3-ABCG2 encoding for the T482 mutant transporter (4), were kindly provided by Dr. Douglas Ross (University of Maryland School of Medicine, Baltimore, MD). As this ABCG2-transfected cell line showed a heterogeneous profile of rhodamine 6G efflux by flow cytometry, cell line subcloning was done by the limited dilution method. The NIH-3T3 drug-sensitive parental cell line and the NIH-3T3 MDR-G185 cell line transfected with human MDR1-G185 (30) were provided by Dr. I. Pastan (National Cancer Institute, NIH, Bethesda, MD). The 2008 human ovarian carcinoma cell line transfected with the plasmid pCMV-Neo and the 2008-MRP1 cell line transfected with the same plasmid but containing the cDNA of human MRP1 (31) were provided by Dr. P. Borst (Division of Molecular Biology, The Netherlands Cancer Institute, Amsterdam, the Netherlands). All cells were cultured in DMEM glutamax II (Invitrogen, Cergy-Pontoise, France) supplemented with 10% fetal bovine serum, 1% penicillin/streptomycin, and drug supplemented in some cases with either 500 μg/mL G418 (MDA-MB-231 and HEK-293) or 60 ng/mL of colchicine (NIH-3T3 MDR-G185).

Flow cytometry. HEK-293 cells were exposed to 5 μmol/L mitoxantrone for 30 minutes at 37°C in the presence or absence of various concentrations of flavonoids added as DMSO solutions (0.5% final concentration), washed in PBS, and incubated in substrate-free medium with the same flavonoid concentration for 1 hour. When indicated, HEK-293 or MDA-MB-231 cells were exposed to either 0.5 μg/mL rhodamine 6G or 2 μmol/L Hoechst 33342. Intracellular drug fluorescence was monitored with a FACscan flow cytometer (Becton Dickinson, Mountain View, CA) for mitoxantrone and rhodamine 6G, or a FACSVantage SE Option Diva apparatus equipped with an UV laser for Hoechst 33342. The maximal fluorescence (100%) was the difference between mean fluorescence of control cells and either R482 or T482 ABCG2-transfected cells, incubated with substrate but without inhibitor. The highest flavonoid concentration tested did not significantly modify the fluorescence of control cells. Cells without drug were included as an autofluorescence control.

P-glycoprotein and MRP1 transport activities. The flavones tectochrysin and 6-prenylchrysin were tested for their ability to interfere with P-glycoprotein– and MRP1-mediated drug efflux using a microplate assay, based on the screening method described by Sarver et al. (32) with minor modifications. Briefly, the day before the experiment, parental drug-sensitive cells and MDR1- or MRP1-transfected cells were plated at a density of 2 × 10⁴ cells per well in 96-well plates (NUNC, Roskilde, Denmark). After 24 hours, when cells were forming a 90% confluent monolayer in all the wells, the original medium was removed and the cells were incubated at 37°C in the presence of increasing concentrations of flavone in the cell culture medium, with the following P-glycoprotein and MRP1 fluorescent substrates and incubation times: either 10 μmol/L daunomycin, 3 μmol/L rhodamine 6G, or 3 μmol/L bodipy-vinblastine for 2 hours, or 1 μmol/L calcein-AM for 20 minutes. Values of P-glycoprotein substrate accumulation in the presence of flavone were normalized to those obtained in the presence of 3 μmol/L GF120918, which were considered as 100% inhibition of MDR1-mediated transport. On the other hand, the flavone-induced accumulation of MRP1 substrates was normalized to the accumulation values observed in the parental counterpart of MRP1-expressing cells (100% drug accumulation). After the incubation step in the presence of the appropriate fluorescent probes, the cell monolayers were washed twice with cell culture medium to remove any trace of extracellular fluorescence. Finally, 50 μmol/L of lysis buffer [20 mmol/L Tris-HCl (pH 7.4) and 0.2% SDS] were added to each well and incubated at 4°C overnight in the dark. The accumulated intracellular fluorescence due to each fluorophore was measured in a microplate spectrofluorometer “SpectraMax Gemini EM” (Molecular Devices, Sunnyvale, CA) using the SoftmaxPro 4.3 software.

Generation of recombinant baculoviruses and membrane preparation. The pcDNA3 plasmid containing T482 ABCG2 cDNA, provided by Dr. D. Ross, was used for subcloning into the pTriEx-4-Neo plasmid (Novagen,VWR, Fontenay-sous-Bois, France). Recombinant baculoviruses, carrying human T482 ABCG2 cDNA, were generated with Bacvector transfection kits (Novagen), according to manufacturer instructions. Briefly, Sf9 cells were cotransfected with BacVector-3000 Triple Cut Virus DNA and either pTriEx or pTriEx-ABCG2(T482) to obtain recombinant baculoviruses. Sf9 cells (10⁶/mL) were infected with recombinant viruses at an infection multiplicity of 20 for 1 hour. After a 72-hour culture at 27°C, cells were harvested by centrifugation at 500 × g for 10 minutes and the membranes were prepared as described (33). Sf9 membranes containing overexpressed wild-type R482 ABCG2 (reference SB-MXR-Sf9-ATPase) were purchased from Solvo Biotechnology (Szeged, Hungary).

ATPase activity assay. Membranes from Sf9 insect cells (2.5 μg protein) were incubated for 90 minutes at 37°C in 50 mmol/L HEPES/KOH (pH 8) with 5 mmol/L ATP, 5 mmol/L MgCl₂, 500 μmol/L EGTA, 5 mmol/L sodium azide, 1 mmol/L ouabain, and an ATP-regenerating system (4 mmol/L phosphoenolpyruvate and 60 μg/mL pyruvate kinase) in 50 μL, final volume. All effectors (mitoxantrone, flavones, and GF120918) were added as DMSO solutions (1% final concentration). The reactions were stopped by adding 1 mL of 20 mmol/L H₂SO₄, and samples were immediately transferred on ice. P_i release was measured by a colorimetric method (34).

Cytotoxicity assay. Exponentially growing HEK293-R482 and HEK293-T482 cells, respectively transfected with wild-type and mutant ABCG2 vectors, and the corresponding control cells were trypsinized and plated (2 × 10³ cells per 200 μL per well) in 96-well microplates and allowed to attach for 24 hours at 37°C under 5% CO₂. A 200-μL sample of drug solution, diluted in medium with 0.5% DMSO, final concentration, was added and cells were incubated for 48 hours at 37°C under 5% CO₂. MDA-MB-231-(T482)ABCG2 cells and respective control cells were used as 10³ cells per 200 μL per well and incubated for 96 hours. Cytotoxicity was evaluated with a 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide colorimetric assay (35).

Stastistical analysis. Sets of data were compared with a Student's t test. Differences were considered either statistically significant, *P < 0.05, **P < 0.01, ***P < 0.001, or not significant, NS. All statistical tests were done using the Sigma Stat software for Windows.

Rational screening of inhibitory flavonoids on ABCG2-mediated drug efflux in transfected cells. HEK-293 cells transfected with wild-type (R482) ABCG2 actively effluxed mitoxantrone and Hoechst 33342, whereas mutant (T482) ABCG2–transfected cells (either HEK-293 or MDA-MB-231 lines) also showed efflux of rhodamine 6G (data not shown). Mean fluorescence analysis indicated a 7- and 12-fold higher intracellular mitoxantrone accumulation in control HEK-293 cells compared with the wild-type and mutant ABCG2-transfected cells, respectively.

Comparison of representatives of different classes of flavonoids, each containing three OH groups (two at positions 5 and 7, cf. Fig. 1A), indicated that all of them were able to inhibit wild-type (R482) ABCG2-mediated mitoxantrone efflux (***P), with varying efficiencies (Fig. 1B). The flavone apigenin, with OH at position 4′, was more efficient (IC₅₀ = 16 μmol/L) than the flavonol galangin, with OH at position 3 (19 μmol/L, *P) and the isoflavone genistein, with ring B branched to ring C at position 3 instead of 2 (24 μmol/L, **P). The least efficient compound was the flavanone naringenin, with its reduced 2,3-bond (37 μmol/L, ***P). A similar preference for flavones was observed for mutant (T482) ABCG2 when rhodamine 6G was the substrate with MDA-MB-231 transfected cells (data not shown).

Structure-activity relationships identifying the flavones 6-prenylchrysin and tectochrysin as potent and specific inhibitors. Because the flavone apigenin seemed the most effective inhibitor of ABCG2, a series of flavones were screened to evaluate the effects produced by OH number and position and hydrophobic substituents (Fig. 2A). IC₅₀ values were determined from plots of percent mitoxantrone accumulation as a function of inhibitor concentration, as illustrated in Fig. 3. Several OH groups exhibited clearly negative effects on the inhibition of drug efflux: (i) at position 3 in 3-OH flavone compared with unsubstituted flavone (IC₅₀ = 8.1 ± 1.9 versus 2.8 ± 0.6 μmol/L, **P), consistent with the slightly lower efficiency of galangin compared with apigenin in Fig. 1B; (ii) at position 7 in 7-OH flavone compared with flavone (7.1 ± 0.3 versus 2.8 ± 0.6 μmol/L, ***P); and (iii) at position 4′ in apigenin compared with chrysin (16 versus 4.6 μmol/L, **P). In contrast, OH at position 5 produced a positive effect in chrysin compared with 7-OH flavone (4.6 ± 0.5 versus 7.1 ± 0.3 μmol/L, **P).

C-isoprenylation at either position 6 or 8 markedly increased the inhibition potency (***P), the highest effect being produced by prenyl (=3,3-dimethylallyl, cf. Fig. 2A) at position 6, with a 15-fold increased affinity comparatively to chrysin (0.29 ± 0.06 versus 4.6 ± 0.5 μmol/L). The inhibitory potency of 6-prenylchrysin was equivalent to GF120918. Slightly lower effects were produced by either the 1,1-dimethylallyl isomer substituent at the same position (**P), or prenyl at position 8 (*P). The effect did not depend only on hydrophobicity because more hydrophobic 6,8-digeranyl and 6-farnesyl derivatives were much less efficient (**P).

The above structure-activity relationships are summarized in Fig. 2B, highlighting both positive and negative substituents for inhibition. Apparently as potent as well-known GF120918, the best flavone identified in this work was 6-prenylchrysin, with an OH at position 5 and a prenyl at position 6 on ring A, a 2,3-double bond on ring C and no OH on ring B. Mutant (T482) ABCG2 exhibited the same preference for the class of flavonoid (i.e., flavones) and similar dependence on OH position (negative at positions 3 and 7 but positive at position 5, **P). However, the main difference was the effect of prenylation: in the case of 6-prenylchrysin, it was 12-fold lower for T482 than R482 (3.6 ± 1.9 versus 0.29 ± 0.06 μmol/L, respectively; **P, cf. Fig. 3A); similar differences were observed for the other prenylated derivatives. Interestingly, a similar distinction was observed for the hydrophobic conventional inhibitor GF120918 (6.9 ± 2.6 versus 0.31 ± 0.14 μmol/L, for T482 and R482, respectively, **P), as illustrated in Fig. 3C. In contrast, the interaction with tectochrysin was even slightly increased (Fig. 3B), which made it one of the best flavone inhibitors for T482 ABCG2 (IC₅₀ = 1.9 ± 0.3 μmol/L). Notably, the inhibitory potency of tectochrysin was further increased when using rhodamine 6G as the substrate (IC₅₀ = 0.32 ± 0.04 μmol/L, data not shown, **P).

The specificity of 6-prenylchrysin and tectochrysin for ABCG2 relatively to P-glycoprotein and MRP1 was analyzed by assaying the ability of increasing concentrations of these compounds to inhibit the efflux of daunomycin, calcein-AM, and bodipy-vinblastine which are substrates of both transporters, in addition to rhodamine 6G that is selectively transported by P-glycoprotein. Table 1 shows that tectochrysin, up to 200 μmol/L, and 6-prenylchysin, up to 25 μmol/L, exhibited a limited effect against both P-glycoprotein and MRP1 transport activities in MDR1- and MRP1-transfected cells respectively.

Effects of 6-prenylchrysin, tectochrysin, and GF120918 on the ATPase activity of wild-type (R482) and mutant (T482) ABCG2. Modification of the ATPase activity of ABC transporters by a compound is a well-established way to prove its direct interaction. We therefore investigated the effects produced by flavones on ATP hydrolysis in insect cell membranes enriched with human R482 or T482 ABCG2. The functionality of the recombinant transporter was previously controlled by flow cytometry, with rhodamine 6G in the case of mutant ABCG2, on insect cells collected after 48 hours of infection of which drug-efflux activity was fully sensitive to GF120918 (data not shown). The ATPase activity of mutant ABCG2-enriched membranes from insect cells was 21.7 ± 1.2 nmol ATP hydrolyzed × min⁻¹ × mg protein⁻¹; it was inhibited by vanadate with high affinity (IC₅₀ around 4 μmol/L) and conversely stimulated up to 43 ± 2% at 20 μmol/L mitoxantrone (Fig. 4). Interestingly, GF120918 at 5 μmol/L, as well as the different chrysin derivatives (chrysin, tectochrysin, and 6-prenylchrysin), produced a similar stimulation as mitoxantrone. In sharp contrast, the ATPase activity of wild-type ABCG2 was insensitive to both mitoxantrone and chrysin derivatives, whereas GF120918 produced a strong inhibition. As a control, the glycosylated flavonoid rutin, which did not alter rhodamine 6G efflux by mutant ABCG2-transfected cells, did not modify the ATPase activity of the corresponding mutant membranes (data not shown).

6-Prenyl-chrysin and tectochrysin as apparently nontransported modulators that potently revert resistance to mitoxantrone. Because chrysin and derivatives stimulated the ATPase activity of mutant ABCG2 in the same way as the substrate mitoxantrone, the possibility that these compounds might be transported was investigated by determining the ABCG2-induced cross-resistance to flavones. Tectochrysin and 6-prenylchrysin exhibited a relatively low cytotoxicity in both wild-type and mutant ABCG2-transfected cells (Fig. 5A), with respective IC₅₀ values of 15.2 ± 1.9 and 16.2 ± 2.3 μmol/L for tectochrysin (b) and 11.2 ± 1.3 μmol/L and 9.0 ± 1.6 μmol/L for 6-prenylchrysin (a). Interestingly, cytotoxicity was not significantly different in control cells (15.9 ± 2.1 and 8.5 ± 1.2 μmol/L, respectively, NS), indicating that ABCG2-transfected cells were not cross-resistant to either tectochrysin or 6-prenylchrysin. No cross-resistance was observed for GF120918 in the MDA-MB-231 cells (NS), either, where the cytotoxicity was higher with IC₅₀ values of 2.9 ± 0.9 μmol/L for mutant ABCG2-transfected cells and 4.5 ± 1.3 μmol/L for control cells (data not shown).

The ABCG2-transfected cells exhibited a relative resistance to mitoxantrone of 10.3-fold for wild-type R482 and 30-fold for mutant T482, by reference to control cells (transfected with empty vector), with respective IC₅₀ values of 1.03 ± 0.54 and 3.0 ± 0.5 μmol/L, compared with 0.10 ± 0.07 μmol/L (Table 2). In the case of wild-type ABCG2-transfected cells, mitoxantrone resistance was almost or completely reverted by 0.5 μmol/L 6-prenylchrysin, 2 μmol/L tectochrysin, or 1 μmol/L GF120918. In contrast, the most efficient inhibitor in the case of mutant ABCG2-transfected cells was tectochrysin, which fully reverted mitoxantrone resistance at 1 μmol/L, with an already significant effect at 0.2 μmol/L (data not shown), whereas 6-prenylchrysin and GF120918 were less efficient in these mutant cells, requiring at least a 2 μmol/L concentration to reach complete reversion. Figure 5B summarizes the reversing effect produced by either 6-prenylchrysin (a and d), tectochrysin (b and e), or GF120918 (c and f) at 1 μmol/L on both wild-type R482 (a-c) and mutant T482 (d-f) ABCG2-transfected cells. The enhanced mitoxantrone resistance in the cells transfected with T482 is compatible with the previously noted gain-of-function in mitoxantrone transport in these cells (29). It seems that 6-prenylchrysin produced a more complete reversion in wild-type than in mutant ABCG2, in contrast to tectochrysin that was more efficient against mutant ABCG2. The differential effect between wild-type and mutant was also observed with GF120918. In contrast, the above concentrations of 6-prenylchrysin and tectochrysin did not sensitize control cells to mitoxantrone. Previous results with mutant ABCG2-transfected MDA-MB-231 cells showed that unsubstituted chrysin also completely reverted mitoxantrone resistance at 5 μmol/L (data not shown).

The main original result of this article is that two hydrophobic flavones, 6-prenylchrysin and tectochrysin, are new, potent, and specific inhibitors of wild-type and mutant ABCG2 drug efflux activity and therefore constitute promising candidates for future clinical trials.

6-Prenylchrysin and tectochrysin identified as potent inhibitors of ABCG2-mediated drug efflux through screening of flavonoids. Rational screening was facilitated by the use of ABCG2-transfected cell lines, with a high drug efflux capacity (monitored by 7- and 12-fold higher mitoxantrone accumulations in control cells compared with R482 and T482 ABCG2-transfected cells, respectively), which correlates the resistance index values of 10.3 and 30 for the two cell lines. This confirms that ABCG2 is indeed the single main difference between the two types of cells, which contrasts with highly resistant lines (with resistance index up to 3,900-35,800 but much lower relative drug efflux capacity) selected from successive treatments with increasing drug concentrations. These stepwise selections likely produce a number of side modifications in addition to ABCG2 overexpression, leading to questions about the accuracy of controls. This was stressed in several studies with either ABCG2 (4, 29, 36) or P-glycoprotein (37), where additional modifications of glutathione transferase and topoisomerase II activities were reported. This might explain, at least partly, the difficulties of Zhang et al. to draw clear-cut structure-activity relationships and the high concentration of flavonoids required for complete chemosensitization (26). An additional mechanism leading to resistance to flavopiridol was indeed evidenced in ABCG2-expressing drug-selected cells, relative to resistance in transfected ones (29). In contrast, the differential effects observed here on both ABCG2-transfected cells and suited control cells (transfected with empty vector) allowed a rational two-step flavonoid screening, first identifying flavones as the favored class over flavonols, isoflavones, and flavanones, and second discriminating negative and positive roles of hydroxyl and hydrophobic substituents on drug efflux inhibition. The best inhibitors of ABCG2 drug transport activity were found to be 6-prenylchrysin and tectochrysin. A positive effect was associated with a hydroxyl at position 5 of ring A, possibly related to conjugation with the vicinal ketone on adjacent ring C. On the contrary, a negative effect of hydroxyl at position 3 was clearly observed here, which contrasts with the potential positive role proposed by Imai et al. (27). This discrepancy might be due to the difficulty inherent in drawing quantitative structure-activity relationships from cytotoxicity experiments (27) compared with flow cytometry analysis as done here. Hydrophobic substituents seemed to increase the binding affinity at both positions 6 and 7, in agreement with the negative effects of hydrophilic hydroxyls on inhibition. Because negative effects were also produced at position 4′, one may wonder about the possibility of hydrophobic substituents on ring B to further increase the inhibitory activity. The relatively simple structure of tectochrysin, a natural compound found in various plants, and its commercial availability readily allows the investigation of a series of new derivatives, such as the 6-prenyl one, to identify more potent inhibitors.

Prenylation is shown here to increase the binding affinity compared with the large chlorophenylmethylpiperidinyl substituent present in flavopiridol, which bound with lower affinity based on the high concentrations required to inhibit mitoxantrone efflux (23). The structure-activity relationships established here for ABCG2 inhibition are clearly different from those reported, on one hand, for P-glycoprotein (14) with a positive effect of 3-OH in contrast to the negative effect on ABCG2; and a much stronger effect of prenylation at position 8 over position 6 (15, 28) in contrast to the preferred position 6 in ABCG2. On the other hand, only flavonols such as quercetin (17) and dehydrosilybin (18) were able to revert multidrug resistance in MRP1 whereas prenylated derivatives were inefficient. Interestingly, a similar preference for 6-prenylation was found for the yeast Pdr5p transporter (21), which belongs to the same G class as ABCG2, suggesting that this 6-prenylation dependence of inhibition might represent a specificity for the ABCG multidrug transporters.

6-Prenylchrysin and tectochrysin behave differently from GF120918. In contrast to the strong inhibition of wild-type ABCG2 ATPase activity shown here for GF120918, which is consistent with the effects produced by both fumitremorgin C (6, 38) and Ko143 (39), 6-prenylchrysin and tectochrysin exhibited no apparent effect. This suggested a different inhibition mechanism of ABCG2-mediated drug-transport activity for chrysin derivatives by comparison to GF120918, fumitremorgin C, and Ko143. These inhibitors either bind to distinct sites on ABCG2 or induce differential effects on the coupling between ATP hydrolysis and drug transport. A combination of these different classes of ABCG2 inhibitors might therefore lead to synergistic inhibitory effects, a possibility that we are currently analyzing. On the contrary, GF120918 and flavone derivatives increased the ATPase activity of mutant T482 ABCG2, similar to the effect of the substrate mitoxantrone, which further contrasts with Ko143, that strongly inhibited in all cases (39). This indicates that the various classes of inhibitors obey overlapping but different inhibition mechanisms, which is consistent with the existence of a “large polyspecific drug-binding site” as characterized for P-glycoprotein and MRP, as well as for other non-ABC multidrug bacterial efflux transporters (40). Concerning ABCG2, this concept was early suggested by the effects of the R482 mutation on the differential panels of transport substrates; this work shows that it also applies to specific inhibitors, as previously observed for novobiocin (29) but not for tariquidar (41). The R482 residue might, directly or indirectly, play a role in the coupling between energy supply and drug transport, and this effect would be altered upon mutation. A comparable activation or inhibition of ATP hydrolysis by modulators has been often reported for P-glycoprotein (42) and MRP1 (17, 43).

The lack of cross-resistance to tectochrysin and 6-prenylchrysin in wild-type and mutant ABCG2-transfected cells, compared with control cells, favors an apparent absence of transport. The situation was not as clear for flavopiridol that exhibited cross-resistance in highly resistant drug-selected cells (8, 23) but not in transfected ones (29); flavopiridol was found to accumulate in transfected Xenopus laevis oocytes in the presence of fumitremorgin C (44). A low transport of genistein was also found in polarized cells (27). In contrast, the present hydrophobic derivatives, 6-prenylchrysin and tectochrysin, exhibit a much stronger affinity than flavopiridol and genistein and, like GF120918, seem not transported; they therefore might bind to a modulatory site strongly interfering with binding of transported drugs.

6-Prenylchrysin and tectochrysin as promising reversal agents for potential clinical trials. 6-Prenylchrysin and tectochrysin seem specific for ABCG2 among the three types of multidrug ABC transporters, because they do not significantly interfere with either P-glycoprotein or MRP1 at much higher concentrations than those that completely inhibit ABCG2 transport activity. These results are consistent with the quite different flavonoid structure-activity relationships determined for the three transporters as discussed above. 6-Prenylchrysin and tectochrysin specificity for ABCG2 contrasts with GF120918 which is at least 40-fold more efficient on P-glycoprotein (45, 46). The fact that these two chrysin hydrophobic derivatives recognize both wild-type and mutant ABCG2 is quite important in view of a potential clinical application. 6-Prenylchrysin seems here the most powerful derivative against the wild-type transporter. Because ABCG2 is susceptible to a hotspot mutation at position 482, as observed in in vitro drug selection of cultured cells, it cannot be completely excluded that similar mutations might occur in vivo upon extensive chemotherapeutic treatments; in this case, a compound such as tectochrysin would be quite valuable.

Another advantage of 6-prenylchrysin and tectochrysin over GF120918 for ABCG2 inhibition is their 3- to 5-fold lower intrinsic cytotoxicity in our transfected cell lines, whereas a similar or even higher GF120918 cytotoxicity has been reported in drug-selected cell lines (45, 46). In addition, tectochrysin is a main component in a number of nutritive supplements and therefore is expected to be well tolerated. 6-Prenylchrysin and tectochrysin displayed a similarly high efficiency as GF120918 to chemosensitize transfected-cell growth to mitoxantrone, with completion at 0.5 to 1.0 μmol/L. These flavone derivatives were therefore more than one order of magnitude more potent to revert multidrug resistance, mediated by either wild-type or mutant ABCG2, than cytotoxic. Such characteristics of 6-prenylchrysin and tectochrysin make them good candidates for future clinical trials, especially because the same submicromolar concentrations were at least as efficient in vivo to sensitize cell growth to mitoxantrone as in vitro to inhibit drug efflux; this suggested the absence of marked modification and inactivation for both chrysin derivatives.

Because the two derivatives have been proposed here to inhibit ABCG2 in a different way than GF120918, fumitremorgin C, and Ko143 (46), it would be interesting to investigate pluri-modulation with the aim to get additive or even synergistic modulatory effects.

Grant support: Centre National de la Recherche Scientifique, the Université Claude Bernard de Lyon grant UMR 5086, French-Spanish Picasso grant HF2002-0140 (interministerial agreements between A. Di Pietro and F. Gamarro), the Association pour la Recherche sur le Cancer grant ARC 3519, the Comité du Rhône de la Ligue Nationale contre le Cancer, the Spanish grant SAF-2003-04200-CO2-01 (F. Gamarro), Ligue contre le Cancer de la Creuse doctoral fellowship (A. Ahmed-Belkacem), Ligue contre le Cancer du Rhône doctoral fellowship (A. Pozza), and European Community Marie Curie fellowship grant HMPF-CF-2001-01244 (J.M. Pérez-Victoria).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Dr. Douglas D. Ross for providing the MDA-MB-231-(T482)ABCG2 and control cell lines, Dr. Ira Pastan for the NIH-3T3 MDR-G185 and control cell lines, Dr. Piet Borst for the 2008-MRP1 clone 8 and control cell lines, Robert Robey for the HEK-293-ABCG2 (R482 and T482) and pcDNA-transfected control cell lines, and GlaxoSmithKline (Madrid, Spain) for the kind gift of GF120918.