Is ATP binding responsible for initiating drug translocation
by the multidrug transporter ABCG2?
Christopher A. McDevitt
1
, Emily Crowley
1
, Gemma Hobbs
1
, Kate J. Starr
2
, Ian D. Kerr
2
and Richard
Callaghan
1
1 Nuffield Department of Clinical Laboratory Sciences, John Radcliffe Hospital, University of Oxford, UK
2 Centre for Biochemistry and Cell Biology, School of Biomedical Sciences, University of Nottingham, UK
Resistance to chemotherapy presents a continuing and
significant obstacle in the treatment of both solid
tumours and haematological malignancies. One of the
most prevalent primary cellular defence mechanisms
against chemotherapeutic agents is the membrane-
bound transporter [1]. The defining feature of these
transporters is their ability to interact with a broad
range of structurally unrelated compounds, a property
that has led them to be described as ‘multidrug trans-
porters’ [2–4]. The resistant phenotype is conferred by
the reduction in cytoplasmic concentrations of chemo-
therapeutic drugs to levels below that required for
cytotoxicity. Resistance to chemotherapy has been
attributed to the expression of three ‘multidrug trans-

2008, accepted 27 June 2008)
doi:10.1111/j.1742-4658.2008.06578.x
ABCG2 confers resistance to cancer cells by mediating the ATP-dependent
outward efflux of chemotherapeutic compounds. Recent studies have indi-
cated that the protein contains a number of interconnected drug binding
sites. The present investigation examines the coupling of drug binding to
ATP hydrolysis. Initial drug binding to the protein requires a high-affinity
interaction with the drug binding site, followed by transition and reorien-
tation to the low-affinity state to enable dissociation at the extracellular
face. [
3
H]Daunomycin binding to the ABCG2
R482G
isoform was examined
in the nucleotide-bound and post-hydrolytic conformations. Binding of
[
3
H]daunomycin was displaced by ATP analogues, indicating transition to
a low-affinity conformation prior to hydrolysis. The low-affinity state was
observed to be retained immediately post-hydrolysis. Therefore, the dissoci-
ation of phosphate and ⁄ or ADP is likely to be responsible for resetting
of the transporter. The data indicate that, like ABCB1 and ABCC1,
the ‘power stroke’ for translocation in ABCG2
R482G
is the binding of
nucleotide.
Abbreviations
ABC, ATP binding cassette; ATP-c-S, adenosine 5¢-[c-thio]-triphosphate; NBD, nucleotide binding domain; TMD, transmembrane domain;
TNP-ATP, 2¢,3¢-O-(2,4,6-trinitrophenyl) adenosine 5¢-triphosphate.
4354 FEBS Journal 275 (2008) 4354–4362 ª 2008 The Authors Journal compilation ª 2008 FEBS

, like other multidrug transporters, con-
tains more than one drug binding site. In addition, the
binding sites are linked by both negative and positive
heterotropic allostery. In a departure from the drug–
protein interactions with ABCB1, the R482G isoform
also contains multiple sites of interaction for a single
drug (daunomycin), which can manifest as homotropic
allostery [22]. The latter has been observed for the
bacterial half-transporter LmrA, but not for any
eukaryotic ABC protein [23].
The translocation of drugs across the plasma mem-
brane requires that the drug binding event(s) in TMD
is intrinsically coupled to the catalytic cycle within
NBDs. The best evidence for an interaction between
the two domains is the ability of numerous substrates
and modulators of ABCG2 (and the R482G isoform)
to stimulate the rate of ATP hydrolysis [21,24,25],
albeit to a lesser degree than that commonly encoun-
tered with ABCB1. The translocation event requires
that the drug binding sites switch from the initial
high-affinity, inward-facing configuration to an
outward-facing, low-affinity configuration to facilitate
dissociation [26]. Originally, the impetus for the switch
in binding site affinity and orientation was thought to
be the energy produced by nucleotide hydrolysis. In
the case of ABCB1, this was revised through the obser-
vations that nucleotide binding in the absence of
hydrolysis could cause the conformational alteration
(reviewed in [27,28]). The low-affinity conformation of
drug binding sites in ABC multidrug efflux pumps is

)1
, which was significantly reduced
following the addition of a large molar excess of doxoru-
bicin (30 lm). The remaining [
3
H]daunomycin associ-
ated with the membranes corresponded to nonspecific
binding at sites other than the ABCG2
R482G
protein.
This fraction corresponded to 37 ± 8 pmolÆmg
)1
, and
therefore the specific binding component in the mem-
branes was 70 pmolÆmg
)1
. The dissociation constant for
[
3
H]daunomycin binding to ABCG2
R482G
has previ-
ously been estimated as 98 nm [22], and all subsequent
binding assays in this study were conducted with 300–
350 nm of the radioligand. There was no detectable dis-
placement of [
3
H]daunomycin binding to membranes
that did not express ABCG2
R482G

32
P]azido-ATP was
used to characterize the interaction of nucleotides with
the transporter. As shown in Fig. 2A, [a
32
P]azido-ATP
binds to ABCG2
R482G
in a dose-dependent manner.
Unfortunately, commercial preparations of the
photo-active nucleotide do not attain sufficiently high
concentrations to enable complete saturation of binding.
However, the binding isotherm in Fig. 2A provides an
estimate of the binding affinity for [a
32
P]azido-ATP as
K
D
= 201 ± 80 lm. This affinity is similar to the value
obtained for ATP binding to ABCB1 [29]. The ability of
nucleotides to displace binding is shown in Fig. 2B, with
values normalized to the amount bound in the absence
of added nucleotide. Neither ADP nor AMP altered the
photolabelling of [a
32
P]azido-ATP bound, whereas the
ATP analogues adenosine 5¢-[c-thio]-triphosphate
(ATP-c-S) and 2¢,3¢-O-(2,4,6-trinitrophenyl) adenosine
Fig. 1. Heterologous displacement of [
3

nucleotide. The only exception was the ATP analogue TNP-ATP,
which was used at a concentration of 0.6 m
M. The amount of
[
3
H]daunomycin bound to the membranes in the absence of nucleo-
tide was assigned a value of unity, and all other data were
expressed as a fraction of this. Values correspond to the mean ±
SEM of three independent membrane preparations.
Fig. 2. The binding of nucleotides and analogues to ABCG2
R482G
.
(A) Purified ABCG2
R482G
(0.25 lg) was photolabelled with
[a
32
P]azido-ATP (3–300 lM) as described in Materials and methods.
Labelled protein was visualized and quantified by autoradiography
of SDS-PAGE analysis. The amount of bound protein was plotted
as a function of nucleotide concentration, and the data were fitted
with the Langmuir binding isotherm using nonlinear least-squares
regression. (B) Photoaffinity labelling of purified ABCG2
R482G
(0.25 lg) was undertaken using a fixed concentration (30 lM)of
[a
32
P]azido-ATP in the presence or absence of ADP (10 mM), AMP
(10 m
M), ATP-c-S (10 mM) or TNP-ATP (1 mM). The intensity of

ADP produced a marginal decrease to 80 ± 5%
(n =4, P > 0.05). The addition of ATP produced a
statistically significant decrease (n =4, P < 0.05) in
the amount of [
3
H]daunomycin bound to a value of
59 ± 9%. The nonhydrolysable nucleotide, ATP-c-S,
produced an even greater decrease to 59 ± 4%
(n =8, P < 0.05). Despite the use of a considerably
lower concentration (0.6 mm), the fluorescent and
slowly hydrolysable analogue TNP-ATP reduced the
binding to 35 ± 4% (n =6,P < 0.05).
Binding of [
3
H]daunomycin to ABCG2
R482G
in a
pre-hydrolysis configuration
ATP, and its nonhydrolysable analogues ATP-c-S and
TNP-ATP, reduced the degree of [
3
H]daunomycin
binding to ABCG2
R482G
, thus warranting further
examination of the effect of these nucleotide
analogues. Figure 3 shows the effects of a range of
ATP-c-S concentrations on the interaction of [
3
H]dau-

tive allosteric mechanism. The addition of either nucle-
otide analogue will effectively trap the protein in a
conformation closely resembling the pre-hydrolytic
state. The decrease in capacity for drug binding reflects
a lower affinity interaction between [
3
H]daunomycin
and the protein immediately prior to ATP hydrolysis.
Fig. 3. Heterologous displacement of [
3
H]daunomycin binding to
ABCG2
R482G
by the nonhydrolysable nucleotide ATP-c-S. The effect
of the nonhydrolysable ATP analogue ATP-c-S (100 l
M to 20 mM)
on [
3
H]daunomycin (300 nM) binding to ABCG2
R482G
was examined
using High-5 cell membranes (20 lg). Incubations were undertaken
at 20 °C for a period of 120 min, and the membrane-bound radioli-
gand was harvested by vacuum filtration through a manifold. The
general dose–response relationship was fitted to the data
(mean ± SEM, n ‡ 3) using nonlinear least-squares regression.
Fig. 4. Heterologous displacement of [
3
H]daunomycin binding to
ABCG2

switched to a low-affinity configuration on nucleotide
binding, what is the consequence of ATP hydrolysis
for the drug binding sites? To address this issue,
ABCG2
R482G
was trapped immediately post-hydroly-
sis using sodium orthovanadate [21]. The metal oxo-
anion vanadate serves as a transition state mimic,
exploiting its chemical similarity to phosphate. Thus,
ATP and vanadate generate an ADP-vanadate struc-
ture mimicking the transition state for the hydrolysis
of the c-phosphate of ATP [30]. Figure 5 demon-
strates the effect of pre-incubation of ABCG2
R482G
-
containing membranes with 100 lm NaVO
3
and a
series of ATP concentrations. The data show a 90%
decrease in the amount of [
3
H]daunomycin bound to
the membranes, indicating that the capacity for sub-
strate interaction is considerably reduced. The
potency for vanadate trapping to reduce [
3
H]dauno-
mycin binding to ABCG2
R482G
was 21.3 ± 3.3 mm

daunomycin by transporting it out of the cytoplasm
[31]. A previous study has indicated that there are two
allosterically coupled binding sites for daunomycin,
although it is unclear whether the coupling is between
the two monomers in a transporter, or between distinct
dimeric units [22]. Measurement of [
3
H]daunomycin
binding provides a useful insight into the pharmacology
of the ABCG2
R482G
isoform, as the binding site is in
communication with those for different drug substrates.
The initial nucleotide screen revealed that several
nucleotide species were capable of modulating drug
binding, thereby reaffirming the interdomain communi-
cation reported for ABCG2. However, despite using
relatively high concentrations, neither AMP nor ADP
was capable of altering the drug–ABCG2 interaction.
In the case of the monophosphate AMP, this was
entirely expected as this nucleotide plays no role in the
catalytic process of ABCG2, and was therefore a
control for specificity of the interaction. The lack of
effect of the diphosphate nucleotide indicates that,
following inorganic phosphate release, the ADP-bound
ABCG2
R482G
isoform adopts a conformation capable
of supporting the binding of [
3

orthovanadate (100 l
M) and a series of ATP concentrations (100 lM
to 300 mM)at37°C for 30 min. The vanadate-trapped protein was
then incubated with [
3
H]daunomycin (300 nM) for 120 min at 20 °C.
Bound and free radioligand were separated using a rapid filtration
assay, and the former was detected using liquid scintillation count-
ing. Values correspond to the mean ± SEM of at least three inde-
pendent membrane preparations, and the dose–response curve
was fitted using nonlinear least-squares regression.
The power stroke in ABCG2 C. A. McDevitt et al.
4358 FEBS Journal 275 (2008) 4354–4362 ª 2008 The Authors Journal compilation ª 2008 FEBS
of sufficient ATP-c-S or TNP-ATP. Thus, the ATP-
loaded conformation of ABCG2
R482G
([E]Æ[ATP],
where ‘E’ refers to ABCG2
R482G
) facilitates a negative
heterotropic allosteric effect of NBDs on TMDs. This
finding with ABCG2
R482G
is entirely consistent with
the observation for the interaction of [
3
H]vinblastine
with ABCB1 and for [
3
H]estrone-sulfate binding to

[
3
H]daunomycin binding conformation. By inference,
therefore, the step in the catalytic cycle corresponding
to the release of inorganic phosphate ([P
i
]) is likely to
correspond to the restoration of a high-affinity
conformation for the transporter, which is supported
by the restoration of high-affinity binding in the ADP-
bound conformation. That this step of the catalytic
cycle is associated with the greatest free energy change
also makes it ideal for the mediation of drug binding
site reorientation, although the binding data cannot
unequivocally inform on the orientation of the sites,
only their affinity for interaction with drugs.
The data presented here suggest that the ABC-
G2
R482G
isoform undergoes the following sequence of
conformational transitions:
½E
H
$½E
L
Á½ATP$½E
L
Á½ADP½P
i


scintillation fluid was obtained from Beckman
Coulter (High Wycombe, UK). Doxorubicin, sodium ortho-
vanadate, ATP, ADP, AMP, ATP-c-S and TNP-ATP were
purchased from Sigma (Poole, UK). GF ⁄ F filters were
purchased from VWR International (Lutterworth, UK).
Insect Xpress medium was obtained from Cambrex (Read-
ing, UK) and Ex-cell 405 medium from JRH Biosciences
(Andover, UK).
Insect cell culture and membrane preparation
The Trichoplusia ni (High-5) cell line was routinely used for
the expression of ABCG2
R482G
and maintained in shaking
suspension cultures, as described previously [22]. High-5
cells at a density of approximately 3 · 10
6
cellsÆmL
)1
were
infected with recombinant baculovirus (approximately
1 · 10
8
plaque-forming unitsÆmL
)1
) at a multiplicity of
infection of five. After 1 h of incubation with virus, the
cells were diluted to a density of 1.5 · 10
6
cellsÆmL
)1

other drugs were incubated in hypotonic binding buffer,
comprising 50 mm Tris ⁄ HCl, pH 7.4. Hypotonic buffer was
used in binding assays to ensure no intraliposomal drug
accumulation. Nonspecific binding to the filters or the lipid
component of membranes was defined as the amount of
[
3
H]daunomycin bound in the presence of a large molar
excess (30 lm) of doxorubicin. Drugs were added from con-
centrated stocks in dimethylsulfoxide and the solvent con-
centration was maintained at < 1% (v ⁄ v). Actual
concentrations of [
3
H]daunomycin added to the tubes were
determined by liquid scintillation counting. Unbound ligand
was separated from bound ligand through porous glass-
fibre filters (GF ⁄ F) using rapid vacuum filtration on a
48-well manifold. The GF ⁄ F filters were pre-soaked in
wash buffer supplemented with 0.1% (w ⁄ v) BSA for
10 min. Samples on the filters were rinsed twice with 10 mL
of ice-cold wash buffer (50 mm Tris ⁄ HCl, pH 7.4, 20 mm
MgSO
4
). [
3
H]Daunomycin bound to the filters was mea-
sured by liquid scintillation counting using Ready Protein
+
scintillation fluid.
Heterologous displacement assays used ABCG2

the binding assay, according to a previously published
procedure [21].
The amount of [
3
H]daunomycin bound at each concen-
tration of heterologous drug or nucleotide was expressed as
a fraction of that obtained with radiolabel alone. The
fraction bound was plotted as a function of added drug
concentration, and nonlinear regression of the general
dose–response relation (Eqn 1) was used to ascertain the
potency (IC
50
) and degree of displacement (F
D
).
Binding of [a
32
P]azido-ATP to purified
ABCG2
R482G
ABCG2
R482G
was purified using immobilized metal affinity,
anion exchange and gel filtration chromatography; full and
extensive details have been described previously [22]. Binding
of nucleotide to ABCG2
R482G
was determined using photo-
affinity labelling with [a
32

max
À B
min
Þ
1 þ 10
½ðlogIC
50
ÀLÞ
n

ð1Þ
where B is the maximal [
3
H]daunomycin binding, B
max
is
the maximal binding, B
min
is the minimum binding, IC
50
is
the concentration of drug that leads to half-maximal bind-
ing of radiolabel (nm), n is the Hill slope factor and L is
log
10
[ligand concentration (m)]. The binding capacities are
expressed as a fraction of the total obtained in the absence
of drug or nucleotide.
Equation (1) was fitted to the displacement data by non-
linear least-squares regression using the graphpad prism

Scheper RJ & Taylor CG (2001) Breast cancer resis-
tance protein expression and resistance to daunorubicin
in blast cells from patients with acute myeloid leukae-
mia. Br J Haematol 115, 257–262.
7 Steinbach D, Sell W, Voigt A, Hermann J, Zintl F &
Sauerbrey A (2002) BCRP gene expression is associated
with a poor response to remission induction therapy in
childhood acute myeloid leukemia. Leukemia 16, 1443–
1447.
8 van den Heuvel-Eibrink MM, Sonneveld P & Pieters R
(2000) The prognostic significance of membrane trans-
port-associated multidrug resistance (MDR) proteins in
leukemia. Int J Clin Pharmacol Ther 38, 94–110.
9 Allikmets R, Schriml LM, Hutchinson A, Romano-
Spica V & Dean M (1998) A human placenta-specific
ATP-binding cassette gene (ABCP) on chromosome
4q22 that is involved in multidrug resistance. Cancer
Res 58, 5337–5339.
10 Doyle LA, Yang W, Abruzzo LV, Krogmann T, Gao
Y, Rishi AK & Ross DD (1998) A multidrug resistance
transporter from human MCF-7 breast cancer cells.
Proc Natl Acad Sci USA 95, 15665–15670.
11 Miyake K, Mickley L, Litman T, Zhan Z, Robey R,
Cristensen B, Brangi M, Greenberger L, Dean M, Fojo T
et al. (1999) Molecular cloning of cDNAs which are
highly overexpressed in mitoxantrone-resistant cells:
demonstration of homology to ABC transport genes.
Cancer Res 59, 8–13.
12 McDevitt CA, Collins RF, Conway M, Modok S,
Storm J, Kerr ID, Ford RC & Callaghan R (2006) Puri-

R482 on methotrexate transport. Cancer Res 63 , 4048–
4054.
19 Robey RW, Honjo Y, van de Laar A, Miyake K,
Regis JT, Litman T & Bates SE (2001) A functional assay
for detection of the mitoxantrone resistance protein,
MXR (ABCG2). Biochim Biophys Acta 1512, 171–182.
20 Mitomo H, Kato R, Ito A, Kasamatsu S, Ikegami Y,
Kii I, Kudo A, Kobatake E, Sumino Y & Ishikawa T
(2003) A functional study on polymorphism of the
ATP-binding cassette transporter ABCG2: critical role
of arginine-482 in methotrexate transport. Biochem J
373, 767–774.
21 Ozvegy C, Varadi A & Sarkadi B (2002) Characteriza-
tion of drug transport, ATP hydrolysis, and nucleotide
trapping by the human ABCG2 multidrug transporter.
Modulation of substrate specificity by a point mutation.
J Biol Chem 277, 47980–47990.
22 Clark R, Kerr ID & Callaghan R (2006) Multiple drug
binding sites on the R482G isoform of the ABCG2
transporter. Br J Pharmacol 149, 506–515.
23 van Veen HW, Margolles A, Muller M, Higgins CF &
Konings WN (2000) The homodimeric ATP-binding
cassette transporter LmrA mediates multidrug transport
by an alternating two-site (two-cylinder engine) mecha-
nism. EMBO J 19, 2503–2514.
24 Ozvegy C, Litman T, Szakacs G, Nagy Z, Bates S,
Varadi A & Sarkadi B (2001) Functional characteriza-
tion of the human multidrug transporter, ABCG2,
expressed in insect cells. Biochem Biophys Res Commun
285, 111–117.

tein. Biochemistry 40, 15733–15742.
33 Rothnie A, Callaghan R, Deeley RG & Cole SP (2006)
Role of GSH in estrone sulphate binding and transloca-
tion by the multidrug resistance protein 1
(MRP1 ⁄ ABCC1). J Biol Chem 281, 13906–13914.
34 Taylor AM, Storm J, Soceneantu L, Linton KJ, Gabriel
M, Martin C, Woodhouse J, Blott E, Higgins CF &
Callaghan R (2001) Detailed characterization of cyste-
ine-less P-glycoprotein reveals subtle pharmacological
differences in function from wild-type protein. Br J
Pharmacol 134, 1609–1618.
35 Martin C, Berridge G, Mistry P, Higgins C, Charlton P
& Callaghan R (1999) The molecular interaction of the
high affinity reversal agent XR9576 with P-glycoprotein.
Br J Pharmacol 128, 403–411.
36 Goodno CC (1982) Myosin active site trapping with
vanadate ion. Methods Enzymol 85, 116–123.
The power stroke in ABCG2 C. A. McDevitt et al.
4362 FEBS Journal 275 (2008) 4354–4362 ª 2008 The Authors Journal compilation ª 2008 FEBS