Modulation of the function of the multidrug resistance–linked ATP-binding cassette transporter ABCG2 by the cancer chemopreventive agent curcumin

Curcumin (curcumin I), demethoxycurcumin (curcumin II), and bisdemethoxycurcumin (curcumin III) are the major forms of curcuminoids found in the turmeric powder, which exhibit anticancer, antioxidant, and anti-inflammatory activities. In this study, we evaluated the ability of purified curcuminoids to modulate the function of either the wild-type 482R or the mutant 482T ABCG2 transporter stably expressed in HEK293 cells and drug-selected MCF-7 FLV1000 and MCF-7 AdVp3000 cells. Curcuminoids inhibited the transport of mitoxantrone and pheophorbide a from ABCG2-expressing cells. However, both cytotoxicity and [3H]curcumin I accumulation assays showed that curcuminoids are not transported by ABCG2. Nontoxic concentration of curcumin I, II, and III sensitized the ABCG2-expressing cells to mitoxantrone, topotecan, SN-38, and doxorubicin. This reversal was not due to reduced expression because ABCG2 protein levels were unaltered by treatment with 10 μmol/L curcuminoids for 72 hours. Curcumin I, II, and III stimulated (2.4- to 3.3-fold) ABCG2-mediated ATP hydrolysis and the IC50s were in the range of 7.5 to 18 nmol/L, suggesting a high affinity of curcuminoids for ABCG2. Curcuminoids also inhibited the photolabeling of ABCG2 with [125I]iodoarylazidoprazosin and [3H]azidopine as well as the transport of these two substrates in ABCG2-expressing cells. Curcuminoids did not inhibit the binding of [α-32P]8-azidoATP to ABCG2, suggesting that they do not interact with the ATP-binding site of the transporter. Collectively, these data show that, among curcuminoids, curcumin I is the most potent modulator of ABCG2 and thus should be considered as a treatment to increase the efficacy of conventional chemotherapeutic drugs. [Mol Cancer Ther 2006;5(8):1995–2006]


Introduction
Development of multidrug resistance (MDR) is a problem in cancer chemotherapy that limits the effectiveness of anticancer drugs (1,2). The overexpression of ATP-binding cassette (ABC) transporters, such as ABCB1 (P-glycoprotein) and ABCC1 (MRP1), have been shown to confer resistance to chemotherapeutic agents by exporting drugs from cells in an ATP-dependent manner. ABCG2 (also called mitoxantrone resistance-associated protein, breast cancer resistance protein, and placental ABC transporter) is a half-transporter of the ABCG subfamily of ABC transporters, the overexpression of which also plays a major role in the development of the MDR phenotype of malignant cells (3 -6). It confers resistance to various anticancer agents, such as doxorubicin, mitoxantrone, topotecan, and SN-38 (7 -10). One of the possible mechanisms by which the development of MDR can be circumvented during chemotherapy is by blocking the functions of these ABC transporters. Tsuruo et al. (11) showed the ability of the calcium channel blocker, verapamil, to reverse ABCB1mediated MDR. Since then, many other reversal agents, including trifluoperazine, amiodarone, reserpine, phenothiazines, cyclosporine A, FK506, and a cyclosporine derivative (PSC-833; reviewed in refs. 1, 2), have been introduced and explored. A mycotoxin fumitremorgin C from the fungus Aspergillus fumigatus was found to be a potent and specific inhibitor of ABCG2 and in vitro reversed MDR mediated by this transporter (12). However, the neurotoxic effects of fumitremorgin C and its structural analogues precluded its use as a modulator for in vivo studies (13 -15).
Turmeric powder has been used for centuries as a spice, coloring, and therapeutic agent (16 -22). The active constituents of turmeric are three major curcuminoids: curcumin or diferuloylmethane (curcumin I), demethoxycurcumin (curcumin II), and bisdemethoxycurcumin (curcumin III; refs. 23,24). We reported earlier that curcumin mixture and purified curcumin I, II, and III could reverse the drug resistance in cells expressing ABCB1 and ABCC1 by inhibiting the functions of these transporters (25 -27). It was also shown that curcumin I, which is a major constituent (70 -75%) of curcumin mixture, among the purified curcuminoids, was the most potent inhibitor of the activity of both the transporters (25 -27).
It has been reported that the amino acid at position 482 has a crucial role in the substrate and inhibitor specificity of ABCG2 and that mutants R482-T/G exhibit altered drug resistance profiles and substrate specificity of ABCG2 (28 -30). Therefore, we investigated the modulating effects of curcuminoids on ABCG2 activity in both wild-type R482-expressing HEK293 cells and mutant 482T-expressing MCF-7 AdVp3000 cells. The drug-selected MCF-7 FLV1000 and MCF-7 AdVp3000, which overexpressed the wild-type R482 and mutant 482T ABCG2, respectively, were chosen because the protein was overexpressed in these breast cancer cell lines under its own promoter at higher levels (31) for biochemical characterization. Curcuminoids inhibited the efflux of ABCG2 substrates, such as mitoxantrone and pheophorbide a, and the presence of nontoxic concentrations of curcuminoids increased the sensitivity of ABCG2-expressing cells to anticancer drugs. In addition, [ 3 H]curcumin I transport assays show that the curcuminoids are not transported by ABCG2. At biochemical levels, curcuminoids stimulated ATPase activity of ABCG2 at very low concentrations (7 -18 nmol/L), and fumitremorgin C, which is an inhibitor of ABCG2, abolished the stimulatory effect of curcumin, indicating that curcuminoids have an effect on ABCG2-specific ATP hydrolysis. Curcuminoids also inhibited both the photolabeling of ABCG2 with two photoaffinity analogs, [ 125 I]iodoarylazidoprazosin (IAAP) and [ 3 H]azidopine and the transport of these agents.
Curcuminoids are natural compounds that are consumed daily in Southeast Asian countries. They interact with the transporter with very high affinity and inhibit ABCG2mediated drug resistance. Taken together, our previous work with ABCB1 and ABCC1 (25 -27) and this study with ABCG2 suggest that curcumin I is a very effective modulator, which should be considered as a potential compound for development of reversal agents designed to overcome MDR-mediated by these three major ABC drug transporters. Cell Lines and Culture Conditions HEK293 cells stably transfected with either empty pcDNA3.1 vector (pcDNA3.1-HEK293) or pcDNA3.1 containing ABCG2 coding either arginine 482 (482R-HEK293) or threonine 482 (482T-HEK293) were maintained in Eagle's MEM supplemented with 10% FCS, penicillin, streptomycin, and 2 mg/mL G418 (29). MCF-7 FLV1000 (482R) and MCF-7 AdVp3000 (482T) cells overexpressing ABCG2 were cultured in RPMI with 10% fetal bovine serum with 1 Ag/mL flavopiridol or 3 Ag/mL doxorubicin and 5 Ag/mL verapamil, respectively (28,32). These cell lines were provided by Dr. Susan Bates (National Cancer Institute/NIH, Bethesda, MD). The cells were maintained in drug-free medium for 3 to 6 days before conducting the experiments to prevent possible effects or interactions of these drugs with curcuminoids. In crude membranes of control HEK293 and MCF-7 and ABCG2-expressing HEK293, MCF-7 FLV1000, and MCF-7 AdVp3000 cells, both ABCB1 and ABCC1 were not detected by Western blot analysis (data not shown), consistent with previous reports (29,32,33).

Materials and Methods
Purification of Curcuminoids from Turmeric Powder Turmeric rhizomes were dried and blended to a powder form. The powder was extracted with 95% ethanol for 24 hours. The ethanolic extract was filtered through Whatman filter paper and ethanol was removed by a rotary evaporator. Curcumin I, II, and III were further purified from the crude ethanolic extract as described earlier (26).
Isolation of Crude Membranes from ABCG2-Expressing HEK293 or MCF-7 Cells Crude membranes from ABCG2-transfected HEK293 or MCF-7 cells were prepared as described elsewhere (26,34). The protein content was estimated using the Amido Black B dye-binding assay as described earlier (35).
Fluorescent Drug Accumulation Assay by Fluorescence-Activated Cell Sorting Accumulation assays with mitoxantrone or pheophorbide a (5 Amol/L) in ABCG2-transfected HEK293 cells or MCF-7, MCF-7 FLV1000, and MCF-7 AdVp3000 cells were done using fluorescence-activated cell sorting as described previously (31). Fumitremorgin C (10 Amol/L) was used as an inhibitor of ABCG2 in flow cytometry assays. For all samples, 10,000 events were counted and the analysis was done with CellQuest software (Becton Dickinson Immunocytometry Systems, San Jose, CA). The mean fluorescence intensity was calculated using the histogram stat program in CellQuest software.
Cytotoxicity Assay Cytotoxicity of curcuminoids was determined by MTT assay as described previously (26). The IC 50 was calculated from linear regression analysis of the linear portion of the growth curves (26). The cytotoxicity assays were also carried out using 2-(2-methoxy-4-nitrophenyl)-3-(4-nitrophenyl)-5-(2,4-disulfophenyl)-2H-tetrazolium, monosodium salt (Dojindo Molecular Technologies, Gaithersburg, MD), which produces a water-soluble formazan dye on bioreduction to monitor the viability of cells, and similar results were obtained; for this reason, only data obtained with MTT dye are given. In both assays, curcuminoids were incubated with CCK8 or MTT dye in the absence of cells and these values were subtracted from samples with cells.
ATPase Assay Crude membranes prepared from 482R-HEK293 and pcDNA3.1-HEK293 cells (100 Ag protein/mL) were incubated with varying concentrations of curcuminoids, prazosin, 10 Amol/L fumitremorgin C, or indicated concentration of curcumin I plus 10 Amol/L fumitremorgin C in the presence and absence of beryllium fluoride (BeF x ; 0.2 mmol/L BeSO 4 , 2.5 mmol/L NaF) in ATPase assay buffer [0.05 mmol/L KCl, 5 mmol/L NaN 3 , 2 mmol/L EGTA, 10 mmol/L MgCl 2 , 1 mmol/L DTT, 50 mmol/L MOPS (pH 7.5)] for 5 minutes at 37jC. The reaction was started by the addition of 5 mmol/L ATP and incubated for 20 minutes at 37jC. SDS solution (0.1 mL of 5% SDS) was added to terminate the reaction and the amount of inorganic phosphate released was quantified by modification of the sensitive colorimetric reaction as described previously (34). The specific activity of the transporter was recorded as BeF x -sensitive ATPase activity (34).

Photoaffinity Labeling of ABCG2 with [ 125 I]IAAP and [ 3 H]Azidopine
Crude membranes (1 mg protein/mL) from MCF-7 AdVp3000 cells were incubated with various concentrations of curcuminoids for 10 minutes at 21jC to 23jC in 50 mmol/L Tris-HCl (pH 7.5). [ 125 I]IAAP (2,200 Ci/mmol) or 0.5 Amol/L [ 3 H]azidopine (60 Ci/mmol) at 3 to 5 nmol/L was added and incubated for an additional 5 minutes under subdued light. The samples were illuminated with a UV lamp (365 nm) for 10 minutes at room temperature. The labeled ABCG2 was immunoprecipitated by adding 800 AL radioimmunoprecipitation assay buffer containing 1% aprotinin followed by addition of 10 Ag BXP-21 antibody. Protein A-Sepharose beads (100 AL) were added and further incubated at 4jC for 16 hours. The protein A-Sepharose beads were pelleted at 4jC and washed with radioimmunoprecipitation assay buffer containing 1% aprotinin. SDS-PAGE sample buffer (50 AL) was added and incubated at 37jC for 1 hour. The samples were separated on a 7% Tris-acetate gel at constant voltage. The [ 3 H]azidopine gel was incubated with Fluoro-Hance (Research Products) for 30 to 45 minutes and both gels were dried under vacuum and exposed to X-ray film for 12 to 24 hours at À80jC. The incorporation of [ 125 I]IAAP or [ 3 H]azidopine into the ABCG2 band was quantified by estimating the radioactivity incorporated into the band using the STORM 860 PhosphorImager System (Molecular Dynamics, Sunnyvale, CA) and the software ImageQuant as described (36).

I]IAAP, [ 3 H]Azidopine, and [ 3 H]Curcumin I Accumulation Assays
MCF-7, MCF-7 AdVp3000, or MCF-7 FLV1000 cells (0.25 Â 10 6 per well in 24-well plates) were grown in monolayer at 37jC. The assay was initiated by incubating cells with 3 to 5 nmol/L [ 125 I]IAAP (2,200 Ci/mmol) or 25 nmol/L [ 3 H]azidopine (60 Ci/mmol) or 200 nmol/L [ 3 H]curcumin I (10 Ci/mmol). Curcumin mixture or fumitremorgin C at 10 Amol/L was added to [ 125 I]IAAP or [ 3 H]azidopine wells and incubated at 32jC for 5 to 60 minutes under subdued light (to avoid photo-crosslinking). After incubation, cells were washed with PBS and lysed by incubation with 0.3 mL/well trypsin/EDTA at 37jC for 30 minutes. The cell lysates were transferred to scintillation vials containing 15 mL Bio-Safe II scintillation fluid and the radioactivity was measured in a scintillation counter. Cells washed with PBS immediately after addition of the assay mix were used as the 0-minute time point. The value for accumulated [ 125 I]IAAP, [ 3 H]azidopine, or [ 3 H]curcumin I at 0 minute (within 3 -5 seconds) was subtracted from a given time as nonspecific binding of these compounds to the cells. The accumulation of labeled compounds was expressed as pmol/10 6 cells. These assays were done at 32jC to slow the rate of efflux from these cells for studying the accumulation/efflux in a time-dependent manner.
Curcumin I and Mitoxantrone Accumulation Assay MCF-7 (control), MCF-7 AdVp3000, and MCF-7 FLV1000 cells (2.5 Â 10 6 per well in 24-well plates) were grown in monolayer. The assay was initiated by incubating the cells with medium containing 10 Amol/L curcumin I or 5 Amol/L mitoxantrone at 32jC in a CO 2 incubator for 60 minutes. After incubation, the cells were washed with cold PBS. The accumulated curcumin I or mitoxantrone in the cells was extracted by adding 95% ethanol. The extract was centrifuged at 3,000 rpm to pellet the cell debris and the supernatant was dried and then resuspended in 500 AL ethanol for the mitoxantrone assay or 25 AL for the curcumin I assay. The ethanolic curcumin extract (10 AL) was spotted on Silica Gel 60 F 254 -coated TLC plates and separated using the solvent system (CHCl 3 /ethanol/acetic acid, 94:5:1; ref. 26). Alternatively, the fluorescence intensity of 200 AL mitoxantrone or 5 AL curcumin ethanolic extract diluted with 195 AL ethanol was read at 695 and 540 nm, respectively (excitation wavelength, 620 nm for mitoxantrone and 420 nm for curcumin I).
Photo-Cross-Linking of ABCG2 with [A-32 P]8-AzidoATP The crude membranes from MCF-7 AdVp3000 cells were incubated with 25 Amol/L curcumin mixture, I, II, and III or 10 mmol/L ATP for 10 minutes at 4jC in ATPase assay buffer (without DTT) and 10 Amol/L [a-32 P]8-azidoATP (10 ACi/nmol) was added under subdued light and incucbated for an additional 5 minutes at 4jC. The samples were then illuminated with a UV lamp (365 nm) on ice for 10 minutes and separated on SDS-7% Tris-acetate polyacrylamide gel at constant voltage. The gel was dried and exposed to X-ray film for 12 to 24 hours at À80jC.

Western Blot Analysis
Crude membrane protein (0.2 -2 Ag/lane) was subjected to electrophoresis and transferred to nitrocellulose membranes as described previously (28). Blots were probed with mouse monoclonal ABCG2-specific BXP-21 antibodies at 1:500 dilution. Horseradish peroxidase -conjugated goat anti-mouse IgG at 1:10,000 dilution was used as the secondary antibody and the signals were detected using enhanced chemiluminescence kit (GE Healthcare, Piscataway, NJ). For detection of ABCB1 and ABCC1, C219 and MRP1 antibodies, respectively, were used as described previously (33,37).
Statistical Analysis Data are the mean F SD from at least three independent experiments. Differences between the means were analyzed by one-way ANOVA. Results were considered to be statistically significant when P < 0.05.

Curcuminoids Inhibit the Efflux of Mitoxantrone and Pheophorbide a from Both Wild-type and R4 82T Mutant ABCG2-Expressing Cells
To investigate the effect of curcuminoids on the accumulation of ABCG2 substrates, mitoxantrone and pheophorbide a accumulation assays (31) were done in wild-type 482R-HEK293 cells and mutant 482T ABCG2-expressing MCF-7 AdVp3000 cells. The pcDNA3.1-HEK293, 482R-HEK293, MCF-7 and MCF-7 AdVp3000 cells were incubated with 5 Amol/L mitoxantrone or pheophorbide a in combination with increasing concentrations of curcumin mixture, curcumin I, II, and III, or 10 Amol/L fumitremorgin C, a specific inhibitor of ABCG2, and incubated at 37jC in the dark for 45 minutes. The accumulation of the fluorescent substrate in the cells was measured as described in Materials and Methods by flow cytometry. The presence of curcuminoids increased the accumulation of mitoxantrone and pheophorbide a in wild-type (R482) ABCG2-expressing cells in a concentration-dependent manner. Representative results at 5 Amol/L curcuminoids are shown in histograms (Fig. 1). Control (pcDNA3.1transfected) cells showed higher accumulation of both mitoxantrone and pheophorbide a and curcuminoids (up to 20 Amol/L) did not have any effect (data not shown). It should be noted that the curcumin mixture and curcumin I (at 5 Amol/L) were more potent in inhibiting the accumulation of both fluorescent substrates than curcumin II and III (compare brown, blue, purple, and light blue lines in Fig. 1). Curcuminoids also inhibited the efflux of mitoxantrone and pheophorbide a from the mutant 482T ABCG2-expressing MCF-7 AdVp3000 cells in a concentration-dependent manner. Fig. 1 (top two right and bottom two right) shows representative histograms at 5 Amol/L curcuminoids. Curcumin I and the curcumin mixture were also more effective in inhibiting the transport in mutant 482T ABCG2-expressing cells than curcumin II and III, whereas the highest concentration of curcuminoids tested (20 Amol/L) had no effect on the control MCF-7 cells (data not shown). The effect of curcuminoids on efflux of these two fluorescent substrates was determined in terms of mean fluorescence intensity and these results are given in Supplementary Fig. S1. 3 The IC 50 s derived from data in Supplementary Fig. S1 are given in Table 1. These data clearly show that curcumin I and curcumin mixture were more efficient in inhibiting the ABCG2-mediated efflux of the substrates than curcumin II and III. These findings suggest that curcumin I, which is the major constituent of the curcumin mixture, is the form that most effectively inhibits this transporter.

Curcumin I and [ 3 H]Curcumin I Accumulation in ABCG2-Expressing Cells
To confirm whether curcumin is transported by ABCG2, curcumin I accumulation was measured in ABCG2-over-  (Fig. 2A). Similar results were also observed in MCF-7 AdVp3000 cells (data not shown). These data suggest that curcumin I is not transported by ABCG2. An accumulation assay with unlabeled 10 Amol/L curcumin I (concentration used for inhibition of substrate accumulation) was also carried out to make sure that curcumin I was not a transport substrate 3 Supplementary material for this article is available at Molecular Cancer Therapeutics Online (http://mct.aacrjournals.org/). of ABCG2 at higher concentrations. The accumulated curcumin I in MCF-7, MCF-7 FLV1000, and MCF-7 AdVp3000 cells after 60 minutes was extracted with ethanol and the concentration of accumulated curcumin I was determined by measuring the fluorescence of the samples. When mitoxantrone, a known substrate (29) was used, as shown in Fig. 2B, the MCF-7 cells showed 4-to 4.5-fold higher accumulation of mitoxantrone than the MCF-7 AdVp3000 or MCF-7 FLV1000 cells. On the other hand, there was no significant difference in the fluorescence of the curcumin I extracted from the MCF-7 control, MCF-7 AdVp3000, and MCF-7 FLV1000 cells (Fig. 2B). The curcumin samples were also separated on TLC plates and the spot corresponding to curcumin I was visualized (Fig. 2C), which showed no significant differences in the intensity of the yellow spot corresponding to curcumin I between the drug-sensitive MCF-7 and the drug-resistant MCF-7 FLV1000 or MCFAdrVp3000 cells. These results show that curcumin I is not transported by ABCG2-expressing cells; furthermore, it is clear that curcumin I is not metabolized 'to a detectable level during 1-hour incubation in control or ABCG2-expressing MCF-7 cells.

Reversal of Drug Resistance by Curcuminoids
As it is known that mitoxantrone (29), topotecan, SN-38, and doxorubicin (39,40) are substrates of ABCG2, the ability of curcuminoids to augment the cytotoxic effect of The difference in the mean fluorescence intensity in the presence of 10 Amol/L fumitremorgin C or curcuminoids and in the absence of any addition was taken as control (100%). Mean F SD (n = 3) derived from data given in Supplementary Fig. S1 (available at http://mct.aacrjournals.org/). these drugs in ABCG2-expressing drug-resistant cells was evaluated. A nontoxic concentration of curcuminoids (5 Amol/L; <5% cell killing at this concentration) was used along with varying concentrations of mitoxantrone, topotecan, and SN-38 in cytotoxicity assays. The relative resistance was calculated based on IC 50 s in the absence and presence of the curcuminoids. The results shown in Table 3 show that curcuminoids were able to increase the sensitivity of the drug-resistant cells to these drugs, ranging from 3-to 8-fold. A nontoxic concentrations of fumitremorgin C (5 Amol/L) and mitoxantrone (0.5 nmol/L) of with f95% cell survival was used as a positive control in the reversal of mitoxantrone and doxorubicin cytotoxicity, respectively, in the mutant 482T ABCG2-expressing cells.

Expression of ABCG2 Is Not Affected by 3-Day Treatment with Curcuminoids in Culture
To confirm that the reversal of drug resistance by curcuminoids was a direct effect on the function of the transporter and not an effect due to reduced expression of ABCG2, the MCF-7 AdVp3000 cells were grown in the presence of nontoxic concentrations (10 Amol/L) of the curcumin mixture and curcumin I, II, and III for 3 days. Crude membrane proteins (10 Ag) from these cells were separated on a 7.5% SDS-PAGE gel and the protein level of ABCG2 was analyzed by Western blotting using BXP-21, the anti-ABCG2 antibody (41). No difference in ABCG2 protein level was observed in the control (DMSO treated) and curcumin-treated cells (Supplementary Fig. S2A). 3 Similarly, the curcumin mixture-treated (10 Amol/L) MCF-7 AdVp3000 cells did not show any significant difference in their ability to efflux mitoxantrone from the control untreated cells (Supplementary Figure S2B -E). 3 Effect of Curcuminoids on ATP Hydrolysis by ABCG2 The ATPase activity of ABCG2 is inhibited by orthovanadate (42) as well as by BeF x . 4 The BeF x -sensitive basal and the substrate-stimulated ATPase activity was determined in crude membranes isolated from wild-type (482R-HEK293) cells. As shown in Fig. 3A, prazosin stimulated the ATPase activity of ABCG2 in a concentration-dependent manner to 3-to 3.2-fold of the basal level with a concentration of 2.4 Amol/L required for 50% stimulation (Supplementary Table S1), 3 which was consistent with previous reports (42,43). The presence of the curcumin mixture and curcumin I, II, and III stimulated BeF xsensitive ABCG2 ATPase activity, with a concentration of 7.52, 7.47, 15.06, and 18.05 nmol/L required for 50% stimulation, respectively ([ Fig. 3B]; Supplementary Table S1), 3 whereas there was no effect of curcuminoids on ATP hydrolysis in membranes prepared from pcDNA3.1-HEK293 cells (data not shown). It is important to note that basal ABCG2 ATPase activity is in the range of 20 to 25 nmol P i /min/mg protein in HEK293 membranes and the curcuminoids stimulate this activity at significantly lower (7.4 -18 nmol/L) concentration compared with prazosin or mitoxantrone (Supplementary Table S1). 3 The ABCG2-specific inhibitor fumitremorgin C inhibited the ATPase activity in a concentration-dependent manner and the IC 50 for inhibition was 1.2 to 1.4 Amol/L (Fig. 3C). Furthermore, fumitremorgin C at 10 Amol/L inhibited curcumin I -stimulated ATPase activity in crude membranes of R482 wild-type ABCG2-expressing cells (Fig. 3D). These results show that curcuminoids stimulate ABCG2specific ATPase activity.

Photoaffinity Labeling of ABCG2 with [ 125 I]IAAP and [ 3 H]Azidopine
The results described above suggest that curcuminoids might interact with the drug substrate-binding sites of ABCG2. Therefore, the binding of the two photoaffinity substrate analogues [ 125 I]IAAP and [ 3 H]azidopine to ABCG2 was studied. The crude membranes from MCF-7 AdVp3000 (482T) cells were used to study the photolabeling of ABCG2, as there was no difference in the binding of these agents to the wild-type and R482-G/T mutant ABCG2 expressed in HEK293 cells (data not shown). The crude membranes from MCF-7 ( Fig. 4A and B, lane 1) and MCF-7 AdVp3000 (Fig. 4A and B, lanes 2 -6) cells were incubated with 10 Amol/L of the indicated curcuminoids for 10 minutes at 21jC to 23jC and 3 to 6 nmol/L [ 125 I]IAAP (Fig. 4A)   was most effective in inhibiting this labeling ( Fig. 4A and B,  lane 4). Additional experiments showed that the inhibition of [ 125 I]IAAP binding to ABCG2 was concentration dependent, with IC 50 s of 0.81, 0.54, 3.2, and >25 Amol/L for the curcumin mixture and curcumin I, II, and III, respectively (data not shown).

Curcuminoids Inhibit the Transport of [ 125 I]IAAP and [ 3 H]Azidopine
An accumulation assay in intact cells was done to test whether the transport of [ 125 I]IAAP and [ 3 H]azidopine was also inhibited by curcuminoids. The MCF-7 or MCF-7 AdVp3000 cells were incubated with 3 to 5 nmol/L [ 125 I]IAAP or 25 nmol/L [ 3 H]azidopine in the absence (control) or presence of 10 Amol/L curcumin mixture or fumitremorgin C at 32jC for 60 minutes. The radioactivity (drug accumulation) associated with the cells was determined as described in Materials and Methods. As shown in Fig. 4C, after 60 minutes, the MCF-7 cells accumulated 4-fold more [ 125 I]IAAP than MCF-7 AdVp3000 cells. The presence of 10 Amol/L curcumin mixture or fumitremorgin C completely inhibited the efflux mediated by ABCG2 and [ 125 I]IAAP accumulation reached the same level in MCF-7 AdVp3000 cells as MCF-7 cells. Similar results were also observed with [ 3 H]azidopine accumulation. Whereas the control MCF-7 cells accumulated 14 times more [ 3 H]azidopine than the MCF-7 AdVp3000 cells, the presence of 10 Amol/L curcumin mixture or fumitremorgin C completely inhibited the efflux mediated by ABCG2 in MCF-7 AdVp3000 cells (Fig. 4D). As shown in Fig. 2C, during 60-minute incubation, curcumin I is not metabolized, indicating that the curcumin I itself interferes with the binding of the transport substrate. Taken together, the transport and the photolabeling assay clearly showed that curcuminoids inhibit the binding of the substrates to the transporter and prevent efflux of [ 125 I]IAAP and [ 3 H]azizidopine, thus increasing their concentrations inside the cells.
Effect of Curcuminoids on the Binding of [A-32 P]8-AzidoATP to ABCG2 We reported earlier that that compounds, such as disulfiram, interact with both the ATP-binding site and the substrate-binding site of ABCB1 (36). Therefore, the effect of curcuminoids on the binding of [a-32 P]8-azidoATP to ABCG2 was investigated. The crude membranes from MCF-7 AdVp3000 cells were incubated with 25 Amol/L of the curcuminoids for 10 minutes at 4jC in ATPase assay buffer (without DTT), and 10 Amol/L [a-32 P]8-azidoATP (10 ACi/nmol) was added. The samples were then photo-cross-linked on ice and processed as described in Materials and Methods. We observed that the presence of curcuminoids did not have any detectable effect on the binding of the [a-32 P]8-azidoATP to ABCG2 ( Supplementary Fig. S3), 3 which suggests that they interact with the substrate-binding site(s) and do not interfere with binding of ATP to the nucleotide-binding domain. Therefore, the stimulation of ATPase activity of   (Fig. 3B) by curcuminoids is most likely due to their interaction at the transport substrate site(s) or at allosteric site(s), which affect interaction of substrates with transport substrate-binding site(s).

Discussion
We have shown previously that curcuminoids are inhibitors of the ABC transporters, ABCB1 (25,26) and ABCC1 (27), which are known to play a major role in the development of MDR in cancer cells. In this study, purified curcuminoids were tested for their interaction with ABCG2. The curcuminoids inhibited the transport of the substrates from ABCG2expressing cells ( Fig. 1; Table 1) and the maximal inhibition was comparable with 10 Amol/L fumitremorgin C, a potent inhibitor of ABCG2 (12). The cytotoxicity data suggested that the curcuminoids themselves may not be transported by ABCG2 (Table 2). This was further confirmed by curcumin I and [ 3 H]curcumin I accumulation assays, which showed no difference in the accumulation of this curcuminoid in the control and ABCG2-expressing cells (Fig. 2).
The regulation of ABCG2 at the gene level is being studied and recent work has shown functional hormone (estrogen) and hypoxia response elements in the promoter (44 -47) and curcuminoids have been shown to affect the function of several transcription factors (reviewed in ref. 20). For this reason, we tested the effect of curcuminoids on the expression of ABCG2. No detectable difference in the expression of ABCG2 was observed in MCF-7 AdVp3000 cells grown in the presence of 10 Amol/L curcuminoids for 72 hours compared with the untreated cells ( Supplementary  Fig. S2), 3 indicating that curcuminoids do not affect the protein levels of the transporter and the reversal of resistance is due to their effect on the function of ABCG2.
A possible mechanism of ABCG2 inhibition, based on the above findings, was the direct interaction of curcumin with ABCG2. Therefore, it was investigated by two different approaches, an ATPase assay and photoaffinity labeling with transport substrates. It was observed that the curcuminoids stimulate ATP hydrolysis mediated by ABCG2 at very low concentrations (Fig. 3B), and fumitremorgin C, which is an inhibitor of ABCG2 ATPase (Fig. 3C), abolished the stimulatory effect of curcumin I (Fig. 3D). However, the transport data ( Fig. 2) Fig. 4A and B) and observed that the curcuminoids inhibited this photolabeling. Additionally, we used an intact cell assay also to determine the effect of curcuminoids on the transport of [ 125 I]IAAP and [ 3 H]azidopine and found that similar to its effect on the photolabeling with these two analogues in the crude membranes it also completely blocked the transport of these two substrates ( Fig. 4C and D). At present, it is not clear whether curcuminoids bind to same transport substrate site(s) or at allosteric site(s), which affect interaction of substrates with transport substrate-binding site(s). Supplementary Table S2 3 compares various variables assessing the interaction of curcumin I with the three ABC transporters (ABCB1, ABCC1, and ABCG2). The degree of stimulation of substrate-stimulated ATP hydrolysis by this curcuminoid is maximal for ABCG2 (3.3-fold) compared with ABCB1 and ABCC1. In addition, the concentration of curcumin I required to obtain maximum stimulation was 5 nmol/L for ABCG2, whereas it was 500 nmol/L for ABCB1, suggesting that curcumin I exhibits higher affinity for ABCG2 than the other two ABC transporters. Similarly, the lower IC 50 for ABCG2 (0.54 Amol/L) compared with ABCB1 (5.8 Amol/L) for inhibition of IAAP labeling suggests that curcumin I is a potent inhibitor of the IAAP binding to ABCG2 than ABCB1. Allen et al. reported that Ko143, a synthetic compound, is a potent inhibitor of ABCG2 (13). Both Ko143 and GF120918 compounds represent a new generation of promising inhibitors (13). The IC 50 of Ko143 and GF120918 was reported to be 19 and 18 Amol/L for MCF-7 cells (13), respectively, whereas curcuminoids have a cytotoxicity value ranging from 25 to 50 Amol/L for the same cells (Table 2). This indicates that curcuminoids are less toxic than these compounds and may safely be used as ABCG2 inhibitors.
Several experimental and clinical studies indicate that curcumin exhibits antioxidant, anti-inflammatory, and anticancer properties (20,23,49). It has also been reported that the use of commercially available curcumin supplements (2 -4 g/d) or recommended therapeutic oral dose of 6 to 8 g/d results in micromolar concentrations of Figure 3. Effect of curcuminoids and fumitremorgin C on BeF x -sensitive ATPase activity in crude membranes from 482R-HEK293 cells: Crude membrane protein (100 Ag protein/mL) from 482R-HEK293 cells was incubated at 37jC with varying concentrations of (A) prazosin; (B) curcumin mixture (n), curcumin I (E), curcumin II (!), or curcumin III (y); (C) fumitremorgin C (n); and (D) curcumin I (n) alone or 10 Amol/L fumitremorgin C + curcumin I (E) in the presence and absence of BeF x (0.2 mmol/L beryllium sulfate, 2.5 mmol/L NaF) in ATPase assay buffer for 10 min. The reaction was started by the addition of 5 mmol/L ATP and was stopped by the addition of 0.1 mL of 5% SDS solution. The amount of inorganic phosphate released and the BeF xsensitive ATPase activity of ABCG2 was determined as described in Materials and Methods. Points, mean of three independent experiments done in duplicates; bars, SD. curcuminoids in the gastrointestinal tract (50 -55). In addition, oral intake of 8 g/d curcumin resulted in f2 Amol/L concentration in serum (56). A recent report detected low levels of curcumin after 8 g of oral dose, but it could be because of the type of formulation used, which had 66% less curcumin than the pure curcumin powder (57). Therefore, curcuminoid intake can be expected to result in concentrations high enough to increase the levels of the anticancer drugs in the cells through its inhibitory effect on ABCG2. A similar approach has been reported to increase the oral bioavailability of topotecan by coadministration of the ABCG2 and ABCB1 inhibitor GF120918 (58). Additionally, simultaneous administration or intake of curcuminoids with ABCG2 substrates and/or antitumor agents might result in the alteration of their pharmacokinetics. This also might increase the toxicity or bioavailability of the different drugs and nutrients.
This work and our previous studies have shown that curcumin I modulates effectively the function of ABCG2, ABCB1, and ABCC1 (25 -27). This is not surprising, as these transporters exhibit overlapping substrate specificity (6,59). In summary, these findings suggest that curcuminoids may play an important role in clinical therapy and the simultaneous administration of recommended therapeutic doses of curcuminoids with anticancer drugs would probably result in an increased bioavailability of the drugs inside the cells. Thus, these curcuminoids (or curcumin mixture) might be used as broad-spectrum modulators offering clinical benefits to improve the effectiveness of chemotherapy in cancer patients.