FUNCTION-DEPENDENT CONFORMATIONAL CHANGES OF THE ABCG2 MULTIDRUG TRANSPORTER MODIFY ITS INTERACTION WITH A MONOCLONAL ANTIBODY ON THE CELL SURFACE

multidrug ABC transporters: ATP-Binding Cassette transporters; AMP-PNP: Adenosine 5'-( γ -imido)triphosphate; DFP: diisopropyl-fluorophosphate; FP: flavopiridol; GAM-PE: goat anti mouse phycoerythrin conjugated secondary mAb: monoclonal MDR1: human multidrug resistance protein (P-glycoprotein, ABCB1); MRP1: human multidrug resistance protein 1, ABCC1; MX: mitoxantrone; PFA: paraformaldehyde; Sf9 cells: Spodoptera frugiperda ovarian cells; V i sodium-orthovanadate. ABSTRACT The human ABCG2 protein is an important primary active transporter for hydrophobic compounds in several cell types, and its overexpression causes multidrug resistance in tumors. A monoclonal antibody (5D3) recognizes this protein on the cell surface. In ABCG2-expressing cells 5D3 antibody showed a saturable labeling and inhibited ABCG2 transport and ATPase function. However, at low antibody concentrations 5D3 binding to intact cells depended on the actual conformation of the ABCG2 protein. ATP depletion, or the addition of the ABCG2-inhibitor Ko143, significantly increased, while the vanadate-induced arrest of ABCG2 strongly decreased 5D3 binding. The binding of the 5D3 antibody to a non-functional ABCG2 catalytic center mutant (K86M) in intact cells was not affected by the addition of vanadate, while still increased by Ko143. In isolated membrane fragments the ligand modulation of 5D3 binding to ABCG2 could be analyzed in detail. In this case 5D3 binding was maximum in the presence of ATP, ADP or Ko143, while the non-hydrolysable ATP analog, AMP-PNP, and nucleotide trapping by vanadate, decreased antibody binding. In membranes, expressing the ABCG2-K86M mutant, both ATP, ADP and AMP-PNP decreased, while Ko143 increased 5D3 binding. Based on these data we suggest that the 5D3 antibody can be used as a sensitive tool to reveal intramolecular changes, reflecting ATP binding, the formation of a catalytic intermediate, or substrate inhibition within the transport cycle of the ABCG2 protein. phosphate from ATP with a colorimetric reaction (11). When the effect of antibody binding was investigated, membranes were preincubated with anti-ABCG2 5D3 monoclonal antibody (eBioscience) or mouse IgG2b (isotype control, SIGMA) in 20 or 160 µ g/ mg membrane concentration for 30 minutes at 37 o C and then washed twice in ice-cold buffer (40 mM MOPS-Tris pH 7.0, 50 mM KCl, 2 mM dithiotreitol and 0.5 mM EDTA) prior to the ATPase activity measurement. The figures represent the mean values of at least three independent experiments with duplicates. both ATP 5D3 binding the ABCG2 protein. These effects were similar both in the wild-type ABCG2 and the K86M mutant variant (not documented in detail). These data indicate that the binding of ADP, ATP or ABCG2 (causing low 5D3 reactivity) occurs even in the absence of Mg 2+ , but no further steps of the catalytic cycle are performed.


INTRODUCTION
(Rockville, MD, USA). The construction of the ABCG2 retroviral vectors and cell transduction methods were described in detail in (30). Transduced cells in some cases were selected by stepwise increases in mitoxantrone or flavopiridol concentrations or single-cell cloned for the desired level of protein expression. Sf9 cells expressing the ABCG2 protein or its K86M variant were prepared as described previously (31). In the present study we used the K86M variant introduced into the wild type (R482) ABCG2, by cloning the Not1-Spe1 fragment of pAcUW21-L/K86M-R482G (31) into the corresponding site of the pAcUW21-L/R482 vector.

Immunodetection of ABCG2
For immunoblotting washed cells were suspended in the presence of 2 mM DFP in 2 × Laemmli buffer and sonicated for 3 × 5 seconds at 4 o C. Sf9 membranes were also suspended in Laemmli buffer. The proteins separated on 7.5 % SDS-polyacrylamide gels were electroblotted onto PVDF membranes, and immuno-detection was performed by using the monoclonal antibody BXP-21 (500 × dilution), and a HRP-conjugated goat antimouse IgG (5,000 × dilution, Jackson Immunoresearch). Enhanced chemiluminescence (ECL) technique was applied to detect HRP activity on the blots. When the labeling was carried out with PFA-prefixed cells, the cells were incubated in 200 µl of PBS (phosphate buffered saline) solution containing 1% paraformaldehyde for 10 minutes at 37°C before the above mentioned labeling procedure.
For obtaining PFA-fixed and permeabilized cells, the cells were incubated in 200 µl PBS solution, containing 4% paraformaldehyde and 0.05% Triton-X 100, for 10 minutes at 37°C. The same 0.05% Triton-X 100 was present during all steps of the labeling procedure. When labeling was carried out in the presence of modifying agents (5 µM Ko143, 10 mM Na-orthovanadate, 50 µM flavopiridol or 5 µM mitoxantrone), the cells were preincubated with these agents for 10 minutes at 37°C before labeling, and the agents were present during antibody labeling. When applicable, ATP depletion of the cells was carried out before the labeling procedure by washing the cells twice in sugar-free Hank's medium and 30 minutes incubation at 37°C in Hank's medium containing 50 mM 2-deoxy-D-glucose and 15 mM sodium azide. During cell labeling and washing the media contained the same ATP-depleting agents.
Isolated membrane fragments from Sf9 cells (45 µg) were labeled with 1 µg/ml 5D3 (or mouse IgG2b as isotype control) in 100 µl final volume of assay mix (40 mM MOPS-Tris pH 7.0, 5 mM Na-azide, 50 mM KCl, 2 mM DTT and 500 µM EGTA-Tris pH 7.0) for 30 minutes at 37 o C. The membranes were then washed with 500 µl assay mix and pelleted at 10,000 g for 4 minutes. The pellet was suspended in assay mix, containing 1 µg/ml GAM-PE, and incubated at 37 o C for 30 minutes. The membranes were then washed and centrifuged (10,000 g for 4 minutes). Finally, the pellet was suspended in 200 µl assay mix and the fluorescence was detected in a fluorescence plate reader (Fluoroskan II, Labsystems) at 485 nm (excitation)/ 590 nm (emission). When the effects of different agents were investigated the membranes were preincubated in assay mix containing 2 mM Na-orthovanadate, 1 µM Ko143, 10 mM MgAMP, MgADP, MgAMP-PNP, MgATP or 10 mM AMP, ADP, AMP-PNP, ATP + 2 mM EDTA or the combination of these agents (as described in the Figure Legends) for 5 minutes at 37 o C prior to the addition of the 5D3 antibody. The relative level of 5D3 binding was calculated as follows: (F X -F IT )/(F 0 -F IT )*100. F X : fluorescence measured in the presence of 5D3 and the investigated compound, F IT : fluorescence measured in the presence of mouse IgG2b (isotype control), F 0 : fluorescence measured in the presence of 5D3 alone.

Cellular mitoxantrone uptake
The drug extrusion function of ABCG2 in intact cells was evaluated by the mitoxantrone (MX) uptake assay of Robey et al. (32) as modified by (30). After 5D3 labeling at 37 o C for 30 minutes and washing (as described for immuno-labeling), the cells were suspended in phenol red-free Hank's balanced salt solution containing 5 µM MX or 5 µM MX + 5 µM Ko143 (in some experiments 10 mM Na-orthovanadate, or 50 µM flavopiridol) and incubated at 37 o C for 30 minutes. After washing, MX fluorescence was analyzed by flow cytometry (FACSCalibur, Becton Dickinson) at 635 nm excitation and 661/16 nm emission wavelengths (FL4). Dead cells were excluded based on propidium iodide (5 µg/ ml) staining.

ATPase activity measurement
Sf9 membranes containing human ABCG2, MDR1 or ABCG2-K86M were harvested and their membranes were isolated and stored at -80 ºC according to (34,35). ATPase activity was measured as described previously, by determining the liberation of inorganic by guest on March 23, 2020 http://www.jbc.org/ Downloaded from phosphate from ATP with a colorimetric reaction (11). When the effect of antibody binding was investigated, membranes were preincubated with anti-ABCG2 5D3 monoclonal antibody (eBioscience) or mouse IgG2b (isotype control, SIGMA) in 20 or

Antibody detection of ABCG2
For the immuno-detection of the human ABCG2 protein in various cell types we used two monoclonal antibodies. The BXP-21 antibody was generated against an N-terminal intracellular epitope (aa. 271-396 -see (28)), while mAb 5D3 was produced by immunizing mice with intact mouse fibroblasts expressing the human ABCG2 protein (17).
As documented earlier, BXP-21 recognizes the ABCG2 protein both in immunoblots and in permeabilized cells (28). In contrast, the 5D3 antibody could be used to recognize human ABCG2 on the surface of intact cells (17), but not on immunoblots (see below).  antibody. It should be noted that, in accordance with previous results, we did not find any immuno-reactivity of the 5D3 antibody with ABCG2 on immunoblots. We found that in the parental PLBs the 5D3 antibody showed no immunoreactivity, even if the cells were fixed by PFA, or fixed and permeabilized by PFA+Triton X-100 treatment (Fig. 2, Panel A). When parental PLB cells were labeled with the BXP-21 antibody (Fig. 2, Panel B), there was some background labeling observed, as compared to the isotype control. However, in these parental cells BXP-21 labeling did not increase upon treatment with PFA or PFA+Triton X-100.
As shown in Figure 2, Panel D, in the case of the ABCG2-expressing PLB cells, there was no reaction with the BXP-21 mAb, unless the cells were both fixed and Tritonpermeabilized. In this latter case a significant, ABCG2-dependent labeling of the cells by BXP-21 was found. In contrast, the 5D3 antibody showed a well visible immunoreactivity with the native ABCG2-expressing PLBs ( Figure 2, Panel C). This reactivity was increased by PFA fixation, while a further permeabilization with Triton X-100 had no effect on 5D3 binding.
It has to be noted that a similar shift in 5D3 reactivity was found upon PFA fixation, and independent of membrane permeabilization, in all ABCG2 expressing cell types studied, including Sf9 insect cells (not shown here). The 5D3 labeling in this latter cell line indicates that the level or even the absence of N-glycosylation does not influence the interaction of 5D3 antibody with ABCG2.

Inhibition of ABCG2 function by the 5D3 antibody
The data presented in Fig. 2 were obtained with relatively low concentrations of the 5D3 antibody (0.2 µg/ 10 6 cells). By increasing the antibody concentration up to 10 µg/10 6 cells, a saturable level of ABCG2 labeling could be achieved, which was not significantly modified by PFA fixation ( Figure 3A).
In order to investigate the effect of 5D3 on the ABCG2 function, we preincubated the PLB-ABCG2 cells with the 5D3 antibody (40 µg/10 6 cells) and then measured Hoechst 33342 dye extrusion. As shown in Fig. 3B, at high 5D3 concentrations (40 µg/10 6 cells), a significant (p = 0.002), about 65% inhibition of dye transport was observed. In contrast, 5D3 did not inhibit the Hoechst dye transport measured in MDR1-expressing PLBs. In addition, the anti-MDR1 inhibitory monoclonal antibody, UIC2 inhibited Hoechst 33342 extrusion in the MDR1-expressing cells, while did not modify the transport activity in the PLB-ABCG2 cells (not shown).
In order to further explore the ABCG2 inhibitory potential and selectivity of the 5D3 antibody, we have performed direct ABCG2-ATPase measurements in isolated Sf9 cell membranes (Fig. 3C). In these experiments we preincubated the isolated membranes for 30 min at 37 o C with two different 5D3 concentrations (20 µg and 160 µg 5D3/ mg membrane protein, respectively) in the absence of ATP, to assure maximum 5D3 labeling of ABCG2 (see below). We found that the application of the lower, 20 µg/mg membrane 5D3 concentration, although at least 20 times greater than that used in the whole-cell experiments, did not significantly affect the ABCG2-ATPase (p = 0.1). However, when the ATPase activity was measured after labeling with 160 µg 5D3/mg membrane protein, a significant (p= 0.007), about 30% decrease in the vanadate-sensitive ATPase activity of ABCG2 was observed. No inhibition was seen in the presence of similar concentrations of an isotype control antibody. There was no effect of 5D3 antibody on the ATPase activity of MDR1 or ABCG2-K86M membranes. All these data indicate that the 5D3 antibody, when applied in high concentrations, specifically inhibits the transport and ATPase function of the ABCG2 protein.

Effects of ABCG2 inhibitors on 5D3 reactivity and mitoxantrone transport by ABCG2 in intact cells
In the following experiments we have studied the effects of a specific ABCG2 inhibitor, Ko143 (36) and the general ABC transporter inhibitor, Na-orthovanadate (V i ) on the binding of 5D3 antibody in intact cells by flow cytometry. The 5D3 labeling conditions were as described for Fig. 2, that is relatively low antibody concentrations were applied. In the same cells we have also measured mitoxantrone (MX) accumulation, by using a different fluorescence detection channel (see Experimental procedures). In the ABCG2-expressing PLBs we found a low, but measurable 5D3 reactivity (Fig. 4. Panel C), which was greatly increased by Ko143, while slightly reduced by the addition of Na-orthovanadate. In the parallel MX uptake experiments (Panel D), in the ABCG2expressing PLBs MX accumulation was reduced, as compared to that found in the parental cells. ABCG2 inhibition by both Ko143 and V i significantly increased intracellular MX level, similar to that seen in cells not expressing ABCG2. Cell labeling with 5D3 at these low antibody concentrations did not cause any change in MX uptake.
According to these results, both Ko143 and V i blocked the ABCG2 transporter function, but Ko143 increased, while V i rather decreased 5D3 binding on the cell surface. On the other hand, 5D3 labeling at this lower antibody concentrations did not inhibit MX transport activity of ABCG2 (see below).
When we analyzed 5D3 binding and MX uptake in other ABCG2 expressing mammalian cell types, we found a similar modulation of 5D3 binding and MX transport by these inhibitors. The data presented in Fig. 4, Panel G, document that in ABCG2-transduced HEK-293T cells 5D3 binding was decreased by V i treatment and increased by Ko143. As shown in Panel H, MX transport in these cells was inhibited by both inhibitors (interestingly, vanadate preincubation could not block MX extrusion in all HEK cells, a variable population of transporting cells was still observed in these experiments). Parental HEK cells did not show a significant ABCG2 expression or MX transport activity (Panels E and F).
We obtained essentially similar data in the MCF-7/MX cells and the PLBs expressing the gain-of function R482G mutant of ABCG2 (not shown). As a summary, the addition of Ko143 and V i treatment blocked ABCG2 function in all these cell types, and Ko143 significantly increased, while Na-orthovanadate decreased 5D3 binding to the ABCG2 protein.

Effect of ATP depletion and transported substrates on 5D3 reactivity and mitoxantrone transport by ABCG2 in intact cells
In the following experiments we have studied the effect of ATP depletion and various transported substrates on 5D3 binding and MX extrusion by ABCG2 in intact PLB cells.
For achieving an efficient ATP depletion of the ABCG2-expressing PLBs, we used a 30 min pretreatment at 37 o C, with a combination of Na-azide and 2-deoxy-D-glucose (see Experimental procedures). As documented earlier in many hematopoietic cell lines, this treatment reduces the ATP level below 5% of the original levels and results in the accumulation of both ADP and AMP in the cells.
As shown in Fig. 5, this ATP depletion strongly inhibited the ABCG2 transport function, that is eliminated the ABCG2-dependent MX extrusion in these cells (Panel B).
Interestingly, ATP depletion significantly increased 5D3 binding, thus transforming the ABCG2 protein in a conformation optimal for 5D3 labeling (Panel A).
We have examined the effects of various agents on 5D3 binding, which were demonstrated transported substrates of the ABCG2 protein. The co-incubation of the ABCG2 cells with mitoxantrone (2-5 µM) did not influence 5D3 labeling (see Figure 5 Panel C). We also found no appreciable effect on 5D3 binding by the addition of other substrates, prazosin (10-50 µM), or ZD1839 (0.1-1 µM) (not shown) (5,32,33).
Flavopiridol (FP), another transported substrate of ABCG2 (37) in low (1-5 µM) concentrations had no effect on 5D3 antibody labeling, while in concentrations above 50 µM this agent significantly increased 5D3 labeling and interfered with MX extrusion ( Figure 5 Panels B and C). This is in line with the ABCG2-ATPase measurements, where high flavopiridol concentrations were inhibitory, thus could act similarly to Ko143 (data not shown in detail).

Effects of substrates, inhibitors and ATP depletion on 5D3 reactivity in the mutant, non-functional K86M-ABCG2, expressed in intact cells
In the next set of experiments we studied intact mammalian cells expressing a nonfunctional mutant (K86M) variant of ABCG2. This mutation in the highly conserved Walker A motif does not affect ATP binding by ABCG2, but impairs its drug transport and ATPase activity, as well as the formation of a vanadate-induced trapped nucleotide (31).
As shown in Fig. 6, Panels A and B, this K86M-ABCG2 had no MX extrusion function, but showed a well measurable 5D3 binding on the cell surface.
In these studies we found that the 5D3 binding of the K86M mutant ABCG2 was significantly increased by PFA fixation, ATP-depletion or Ko143 treatment. Still, the relative increase in 5D3 binding due to these effects was much smaller than in the case of the wt ABCG2, and 5D3 binding was unaffected by pretreatment with Na-orthovanadate

Effects of nucleotides and transport inhibitors on 5D3 reactivity of ABCG2 in isolated membrane fragments
In the following experiments we examined the effects of various nucleotides and transport inhibitors on 5D3 binding by human ABCG2 and its mutant (K86M) variant in isolated insect cell membrane fragments. In these membrane preparations ABCG2 expression reaches a high level (up to 5% of the membrane proteins), in a fully active form, as reflected by the ABCG2-ATPase activity (11,31). A large fraction of the isolated membrane fragments are accessible both from the cytoplasmic and the external cell surface, as tested by the trypsin sensitivity of open fragments (38) and simultaneous staining of the membrane fragments with two antibodies (pAb 405 and mAb 5D3), that recognize an intracellular (5), and an extracellular epitope of ABCG2, respectively (not shown here). Therefore this assay system allows a direct estimation of the effects of cytoplasmic ligands on the cell surface interaction of ABCG2 with the 5D3 antibody.
As shown in Fig. 7, Panel A, 5D3 binding to isolated Sf9 cell membranes, containing the human ABCG2 protein, reached a high level, significantly exceeding that seen in the control, MDR1-containing membranes, or the labeling obtained with an isotype control antibody. MgATP+vanadate. An interesting finding was in these experiments, that if Ko143 was added after a preincubation with MgAMP-PNP, the reduction in 5D3 binding by this nucleotide could not be reversed by Ko143 (data not shown).
These data indicate that in the case of a functional ABCG2, 5D3 labeling has a relatively high level either in a nucleotide-free, or in a nucleotide-liganded, flexible state of the transporter. However, when the transport cycle is blocked by a non-hydrolysable ATP analog, or by the inhibition of ATP hydrolysis by Na-orthovanadate, a strong reduction in 5D3 binding occurs. Arresting the ABCG2 transport cycle by Ko143, however, produces a high 5D3 binding, and this effect is not reversed the nucleotides and/or vanadate. Still, a low 5D3 binding conformation first fixed by MgAMP-PNP, cannot be changed to a high binding form by a later addition of Ko143. In experiments not documented here in detail, we have performed 5D3 binding to ABCG2 in isolated Sf9 membranes at 4 o C, in order to investigate labeling at nonhydrolytic conditions. We found that 5D3 binding at 4 o C was somewhat reduced (75 ± 1.4% of that measured at 37 o C), and the addition of nucleotides or inhibitors (Ko143 or V i ) did not cause a measurable change in 5D3 binding.
We have also investigated 5D3 binding to ABCG2 in isolated Sf9 membranes upon the addition of AMP, ADP, AMP-PNP and ATP, but in the absence of Mg 2+ ions (that is in the presence of excess EDTA), at 37 o C. Interestingly, we found that in the absence of Mg 2+ , both ADP, AMP-PNP and ATP (but not AMP) significantly decreased 5D3 binding to the ABCG2 protein. These effects were similar both in the wild-type ABCG2 and the K86M mutant variant (not documented in detail). These data indicate that the binding of ADP, ATP or AMP-PNP to ABCG2 (causing low 5D3 reactivity) occurs even in the absence of Mg 2+ , but no further steps of the catalytic cycle are performed.

DISCUSSION
In the present experiments we have studied the interaction of the 5D3 monoclonal antibody, prepared against a cell surface epitope of human ABCG2, with this multidrug transporter both in intact cells and in isolated membranes. We found that in intact cells 5D3 recognition of the ABCG2 protein occurred at an external epitope. The specific antibody binding was significantly increased by fixation of the intact cells by paraformaldehyde (PFA), but this interaction did not require membrane permeabilization In accordance with data in the literature regarding 5D3 effect on ABCG2-induced drug resistance (22), we found that the 5D3 antibody significantly inhibited both the dye transport and the ATPase activity of the ABCG2 protein (Figure 3, Panels B and C). Still, the inhibition of the transport or ATPase activity of ABCG2 found here was incomplete even at very high 5D3 concentrations (see Figures 3B and 3C). This finding is most probably due to the steric and mechanical constrains in such antibody-transporter interactions. A similarly selective, but only partial functional inhibition has been reported for several anti-MDR1 antibodies, e.g. MRK16 or UIC2, reacting with cell surface epitopes of the MDR1 multidrug transporter (23,24).
In this study we found that at low 5D3 concentrations the actual conformation of the ABCG2 protein significantly modified 5D3 binding to the extracellular epitope. In intact cells ABCG2 interaction with 5D3 was greatly increased by the inhibition of ABCG2 function with a specific, high affinity inhibitor, Ko143 (see Figure 4, Panels C and G), or by cellular ATP depletion. (Fig. 5, Panel A). Similarly, an increase in 5D3 reactivity was observed in the presence of high, inhibitory concentrations of a drug substrate of ABCG2, flavopiridol ( Figure 5C) (37).
In contrast, a reduction in 5D3 binding was observed when the cells were preincubated with Na-orthovanadate, a transition-state inhibitor of ABC transporters, including ABCG2 (31,(39)(40)(41). In this case, within the nucleotide binding domain of the protein, vanadate anions replace phosphate after ATP hydrolysis, and the transport cycle of ABCG2 is arrested in a transition-state. This can be experimentally followed by measuring the According to these data, 5D3 interaction with ABCG2 in intact cells depends on the actual conformation within the transport cycle of this multidrug resistance protein. 5D3 binding is relatively low in the case of the actively functioning protein or in its stabilized transition state. In contrast, 5D3 binding is greatly increased when ABCG2 conformation is stabilized in other specific conformations (by Ko143 or ATP depletion). In unpublished experiments we found that Ko143 inhibition of ABCG2 was reversible by repeated washings. Also, ABCG2-ATPase inhibition achieved by low (10 nM) Ko143 concentration could be removed by the addition of increasing concentrations of transported substrates, e.g. prazosin. These results indicate that Ko143 probably inhibits ABCG2 by interacting with its substrate binding site.
When examining the binding of the 5D3 antibody in intact cells to a non-functional ABCG2 catalytic center mutant (K86M-ABCG2), we found that 5D3 binding to this mutant protein was also efficient. In the case of this mutant ABCG2, 5D3 binding was not affected by the addition of transported substrates or vanadate, while it was increased by ATP-depletion or by the addition of Ko143 ( Figure 6). These data are in line with the impaired catalytic cycle and transition state forming ability of this mutant, with unchanged ATP binding (31), and probably with conserved drug/inhibitor binding properties.
In order to further explore the mechanistic details of the ABCG2 catalytic cycle, we have performed a detailed analysis of 5D3 binding to ABCG2 in isolated membrane fragments, accessible from both sides of the membrane (see Fig. 7, Panel B). It is important to note that the experiments carried out with isolated membranes exclude the possibility that the changes in 5D3-ABCG2 interactions might be due to variable cell surface expression of the multidrug resistance protein in intact cells. They also allow to study the interaction of non cell-permeating ligands with cytoplasmic domains of the transporter.
In these experiments we observed that the non-hydrolysable ATP analog, AMP-PNP, strongly reduced 5D3 binding to ABCG2. MgAMP, MgADP, or MgATP had no major effect, but MgATP+Na-orthovanadate induced a major decrease in 5D3 binding.
Preincubation with the inhibitor molecule, Ko143 maximized 5D3 binding under all conditions.
In the K86M-ABCG2 variant, the addition of MgATP, MgADP and MgAMP-PNP, all caused a major reduction of 5D3 binding, which was not further modulated by Naorthovanadate. These results coincide with the conserved ATP binding, but impaired catalytic intermediate formation by this mutant protein. In the case of this non-functional mutant we still found an increased 5D3 binding upon preincubation with Ko143, even if MgATP, MgADP or MgAMP-PNP were added thereafter to the media (see Fig. 7, Panel C). These results suggest a preserved substrate/inhibitor binding site in this mutant protein.
Interestingly, in Sf9 membranes containing either wild-type ABCG2 or its K86M mutant, in the absence of Mg 2+ (that is in the presence of EDTA), both ATP, ADP and AMP-PNP caused a decrease in 5D3 binding (not shown in detail). These data may indicate that at these high nucleotide concentrations (10 mM), ABCG2 binds nucleotides even in the absence of Mg 2+ , although further ATP hydrolysis is absent. When trying to investigate the possible effects of transported substrates (e.g. prazosin, flavopiridol, or mitoxantrone) on 5D3 binding by ABCG2 in isolated Sf9 cell membrane fragments, we could not detect any major changes evoked by relevant substrate concentrations. This is similar to the lack on ATPase stimulation by substrates in this system, and most probably due to the presence of endogenous substrates of ABCG2 in the Sf9 membranes (31).
These data collectively indicate that the binding of the 5D3 monoclonal antibody closely reflects the changes in the drug-and ATP binding, as well as the catalytic state of the ABCG2 transporter. This is most probably due to the variable appearance of a conformational epitope within the ABCG2 protein on the cell surface. This study is the first demonstration of such a conformation-sensitivity of an antibody binding to the ABCG2 protein, although a conformation dependent binding of some extracellular antibodies, e.g. MRK16 or UIC2, to another multidrug transporter, the MDR1 protein, has already been documented (23,24). The determination of the actual epitope structure involved in 5D3 binding should require a detailed molecular mapping of potentially cellsurface domains of ABCG2.
As a summary, the various steps within the catalytic cycle of the ABCG2 multidrug resistance transporter could be visualized through changes in 5D3 binding. A low level 5D3 binding was observed when the non-hydrolysable ATP analog, MgAMP-PNP, or the addition of ATP or ADP without Mg 2+ ions stabilized the protein in a pre-hydrolytic state (42,43). The formation of a catalytic intermediate, reflected by nucleotide trapping in the presence of vanadate anions (40,41), also coincided with a low 5D3 reactivity of ABCG2.
In contrast, transport inhibition by Ko143 or by high concentrations of flavopiridol, as well as by ATP depletion, stabilized the protein in a conformation with high 5D3 binding capacity.
Based on these data we suggest that the 5D3 reactive form of ABCG2 is a stabilized "substrate off-site" conformation of the transporter. It has to be noted that the ABCG2 protein is an ABC half-transporter, and its function requires homo-dimerization (6)(7)(8)11).
Conformational changes detected through a complex extracellular epitope of a membrane protein can be due to a function-dependent rearrangement of the transmembrane helices, triggering the movements of the extracellular loops, or to the surface-exposure of membrane-embedded short segments (for MDR1 see (44)). ABCG2 acts as a homodimer, and one additional possible explanation for the conformational changes described in the present study is the function-dependent re-orientation of the monomers within the dimer, or facilitation of the dimer formation. However, further experiments are needed to elucidate the dependence of 5D3 binding on the molecular interactions between the dimerizing ABCG2 molecules.
Based on this study we suggest that the 5D3 antibody can be used to reveal major intramolecular changes in the ABCG2 protein during its catalytic/transport cycle.
Examining 5D3 binding to various mutant, polymorphic, or stabilized forms of ABCG2 may further help structure-function relationship studies. Moreover, based on the present data, optimum conditions can be selected for the investigation of ABCG2 expression and function by 5D3 binding in intact cell preparations, thus employing this antibody for a sensitive clinical laboratory detection of ABCG2 expression and function.