Together, these experiments demonstrated that glo-bal conformational changes occur in the protein and are relayed from the NBDs to the TMDs as a conse-quence of nucleotide binding and dr
Trang 1energy provision and drug binding in ABCB1
Emily Crowley1, Megan L O’Mara2, Ian D Kerr3and Richard Callaghan1
1 Nuffield Department of Clinical Laboratory Sciences, John Radcliffe Hospital, University of Oxford, UK
2 Molecular Dynamics Group, School of Chemistry and Molecular Biosciences, University of Queensland, Brisbane, Australia
3 School of Biomedical Sciences, University of Nottingham, Queen’s Medical Centre, UK
Introduction
ABCB1 (P-glycoprotein) is a member of the
ATP-bind-ing cassette (ABC) family of membrane transporters,
and is located in the plasma membrane of cells The
transporter is localized to a number of tissues
associ-ated with absorptive, secretory or barrier roles [1–3],
and its primary function is therefore to provide a defensive mechanism against xenobiotics ABCB1 pro-vides this defence by acting as a multidrug efflux pump The expression pattern in physiological tissues enables ABCB1 to play a prominent role in shaping
Keywords
ABC transporter; bioenergetic coupling; drug
resistance; efflux pumps; P-glycoprotein
Correspondence
R Callaghan, Nuffield Department of Clinical
Laboratory Sciences, John Radcliffe
Hospital, University of Oxford, Oxford, OX3
9DU, UK
Fax: +44 1865 221 834
Tel: +44 1865 221 110
E-mail: richard.callaghan@ndcls.ox.ac.uk
(Received 5 May 2010, revised 2 July 2010,
accepted 27 July 2010)
doi:10.1111/j.1742-4658.2010.07789.x
Describing the molecular details of the multidrug efflux process of ABCB1,
in particular the interdomain communication associated with bioenergetic coupling, continues to prove difficult A number of investigations to date have implicated transmembrane helix 12 (TM12) in mediating communica-tion between the transmembrane domains and nucleotide-binding domains (NBDs) of ABCB1 The present investigation further addressed the role of TM12 in ABCB1 by characterizing its topography during the multidrug efflux process with the use of cysteine-directed mutagenesis Cysteines were introduced at various positions along TM12 and assessed for their ability
to covalently bind thiol-reactive fluorescent probes with differing physio-chemical properties By analysing each isoform in the basal, ATP-bound and posthydrolytic states, it was possible to determine how the local envi-ronment of TM12 alters during the catalytic cycle Labelling with hydro-phobic CM and zwitterionic BM was extensive throughout the helix in the basal, prehydrolytic and posthydrolytic states, suggesting that TM12 is in a predominantly hydrophobic environment Overall, the carboxy region (intracellular half) of TM12 appeared to be more responsive to changes in the catalytic state of the protein than the amino region (extracellular half) Thus, the carboxy region of TM12 is suggested to be responsive to nucleo-tide binding and hydrolysis at the NBDs and therefore directly involved in interdomain communication This data can be reconciled with an atomic-scale model of human ABCB1 Taken together, these results indicate that TM12 plays a key role in the progression of the ATP hydrolytic cycle in ABCB1 and, in particular, in coordinating conformational changes between the NBDs and transmembrane domains
Abbreviations
ABC, ATP-binding cassette; AMP-PNP, 5¢-adenylylimidodiphosphate; BM, BODIPY maleimide; CM, coumarin maleimide; FM, fluorescein maleimide; L ext, maximum extent of labelling; NBD, nucleotide-binding domain; TMD, transmembrane domain; TM6, transmembrane helix 6; TM12, transmembrane helix 12.
Trang 2the pharmacokinetic profile (adsorption, distribution,
metabolism and excretion) of many commonly used
medications [4] Unfortunately, cancer cells overexpress
ABCB1 in order to evade the toxic effects of
antican-cer drugs, a phenomenon known as multidrug
resis-tance The extraordinary range of compounds
recognized by ABCB1 (over 200 known drugs) makes
it a powerful mediator of resistance against
chemother-apeutic intervention in a number of cancer types The
ability to recognize such an array of compounds
remains a biological enigma, thereby making the
devel-opment of inhibitors that may restore the efficacy of
chemotherapy in cancer treatment a difficult task
The functional unit of ABCB1 consists of two
trans-membrane domains (TMDs), each comprising six
membrane-spanning helices, and two
nucleotide-bind-ing domains (NBDs) [5] Much is understood
regard-ing NBD function, owregard-ing to the high sequence
homology between members of the ABC transporter
family Furthermore, several crystal structures of ABC
transporters have been solved in the presence and
absence of nucleotides, improving our understanding
of the mechanism of transport [6–9] However, much
remains unclear about structure–function relationships
of the TMDs of multidrug resistance pumps, including
the location of the drug-binding sites and the
molecu-lar mechanism underlying drug translocation The
recent 4–4.3 A˚ resolution crystal structure of
full-length ABCB1 has provided a location for the binding
of a purpose-built peptide inhibitor [6] However, more
pharmacological information is required to evaluate
this inhibitor and how its binding relates to more
established substrates or modulators [10–12]
Another unresolved issue pertaining to ABCB1
func-tion is the molecular detail of the process of coupling
between the NBDs and TMDs The most striking
evi-dence for the presence of coupling between the two
domains is the ability of transported drugs to stimulate
the basal rate of ATP hydrolysis by ABCB1 [13,14] It
is well established that drug binding occurs in the
TMDs, and stimulation of hydrolysis therefore
requires long-distance communication with the
cyto-solic NBDs This is supported by evidence that
muta-tions of numerous residues within the TMDs are
capable of disrupting the stimulation of ATP
hydroly-sis [15–19] Moreover, drug translocation and ATP
hydrolysis must be coordinated for active efflux This
requires interdomain communication in both the TMD
to NBD and NBD to TMD directions [20] The latter
route has also been demonstrated; for example,
bind-ing of the nucleotide analogues ATPcS or
5¢-adenylyli-midodiphosphate (AMP-PNP) to the NBDs of ABCB1
was shown to significantly decrease the binding of the
UIC2 antibody, which recognizes a conformation-sen-sitive epitope in the TMD [21,22] Furthermore, the cryo-electron microscopy structure of ABCB1 showed that in the presence of the AMP-PNP the architecture
of the TMDs is significantly rearranged [23,24] Together, these experiments demonstrated that glo-bal conformational changes occur in the protein and are relayed from the NBDs to the TMDs as a conse-quence of nucleotide binding and drug binding, respec-tively, thereby enabling active drug efflux by ABCB1 However, we are yet to understand exactly how these conformational changes are relayed between the TMD and NBD, and how they enable drug translocation Transmembrane helix 6 (TM6) and transmembrane helix 12 (TM12) are likely candidates to effect coupling, given their direct links to the two NBDs of ABCB1
We have previously constructed, and analysed, a ser-ies of TM6s with single cysteine mutations, and demon-strated that this helix plays a prominent role in the coupling process in ABCB1 [25–28] A number of muta-tions in TM6 caused alteramuta-tions in drug-stimulated ATP hydrolysis, irrespective of whether they contributed to drug binding Moreover, several residues in TM6 were demonstrated to undergo topographical alterations dur-ing conformational changes of ABCB1 In a recent study, we demonstrated, using a similar approach, that the mutation of several residues within TM12 also influ-ences the communication between the TMDs and NBDs [15] The present article describes the conforma-tional changes adopted by TM12 in response to events occurring in the NBDs The data indicate that nucleo-tide binding and hydrolysis at the NBDs causes confor-mational changes that are transmitted through TM12
Results
Three thiol-reactive fluorescent probes were used to assess the relative accessibility of selected residues in TM12 that had been mutated to cysteine The probes possess distinct physicochemical properties and have been shown to partition to hydrophilic or hydrophobic environments [25] By assessment of the ability of each probe to label residues in TM12, a topographical map
of the helix can be generated Furthermore, trapping the protein at distinct stages of the catalytic cycle will reflect how the environment of individual residues in TM12 changes as ABCB1 switches conformational states
The maximum extent of labelling of TM12 residues in ABCB1
The maximum extent of labelling (Lext) of selected TM12 mutant isoforms was initially investigated with
Trang 3reconstituted protein in the basal (nucleotide-free)
state Following incubation with the fluorescent probe,
the proteins were resolved by SDS⁄ PAGE, and the
covalent binding of the probe was detected under UV
light Figure 1 (lower panel) shows a representative
labelling reaction, in this case a time course for the
V988C isoform with coumarin maleimide (CM) The
gel in the upper panel of Fig 1 shows the same gel but
stained with PageBlue to demonstrate purity of the
samples and to enable normalization of labelling for
protein loading Labelling was time dependent during
the 300 min incubation, and the extents of labelling
were quantified in comparison with that found with
cysteine-less ABCB1 and the G324C isoform The
G324C mutant was assigned as the positive control
and given a value of 100%, as this residue is located
on an external loop and is freely accessible to each of
the probes used [25,28] Furthermore, the complete
labelling of the G324C mutant with the zwitterionic and hydrophilic probes BODIPY maleimide (BM) and fluorescein maleimide (FM), respectively, demonstrated that the protein was not preferentially oriented in one direction within the proteoliposomes Consequently, labelling of TM12 isoforms was determined as a percentage of G324C labelling, as outlined in Experi-mental procedures Additionally, labelling of the cyste-ine-less ABCB1 isoform was also examined as a negative control Any nonspecific association of the three probes with cysteine-less ABCB1 was subtracted from the specific labelling intensity observed with the single-cysteine-containing isoforms Obtaining full labelling and its accurate quantitation are difficult to achieve in practice, resulting in occasional instances where values for the Lext of single-cysteine isoforms are apparently > 100% The approach does, however, provide strong predictions of relative labelling propen-sity, reflecting accessibility of the specified residue Labelling of each isoform was analysed by densitom-etry and plotted as a function of time, as shown for the M986C isoform for the three probes in Fig 2A Nonlinear regression of the exponential reaction curve estimated that the maximum extent of labelling for the representative curve of the M986C isoform in the basal state was 78% for CM (t1⁄ 2= 8 min), 59% for BM (t1⁄ 2= 4 min) and 23% for FM (t1⁄ 2= 45 min) Clearly, this mutant isoform was avidly labelled with the hydrophobic (CM) and zwitterionic (BM) probes,
on the basis of the extent and rapid half-life of the interactions In contrast, the hydrophilic FM displayed only partial labelling, with a considerably longer half-time for the reaction Similar analysis was undertaken for each of the TM12 single-cysteine mutants (using multiple protein preparations) in the basal (i.e nucleo-tide-free) state; the extent and time course of labelling are shown in Table 1
All of the mutant isoforms examined were capable
of interacting with CM, which has a high octa-nol⁄ water partition coefficient, indicating a preference for hydrophobic regions The central region of TM12, from V982C to M986C, displayed the highest extent of labelling, with Lextvalues of 75–100% The C-terminal stretch (V988C–F994C) was also capable of interacting with CM, albeit with lower values of Lext, in the range 50–60% The lowest labelling observed in the selection
of TM12 mutant isoforms was at L976C, with an Lext
of 38 ± 5% Lower labelling presumably reflects the location of the residues at the membrane–water inter-face or significant local steric hindrance The half-lives for the interaction of CM with ABCB1 ranged from
18 to 30 min, but did not reveal further details con-cerning the accessibility of the residues Of the three
150 kDa
100 kDa
150 kDa
100 kDa
Fig 1 Detection of CM labelling of the V988C isoform SDS ⁄ PAGE
analysis of the V988C isoform incubated in the presence of CM for
0–300 min The reaction was stopped at various time points by the
addition of dithiothreitol Upper panel: the gel protein was visualized
with PageBlue staining to indicate sample purity and to enable
load-ing correction Lower panel: the samples were resolved by
SDS ⁄ PAGE and the protein was visualized with the BioDocIt
sys-tem, using a UV light source Molecular mass markers are shown
on the left Lane assignments are: (i) 300 min; (ii) 120 min; (iii)
60 min; (iv) 30 min; (v) 10 min; and (vi) 0 min Lane (vii) contains
the G324C isoform, which has been assigned a 100% value for
labelling with BM.
Trang 4probes used in this investigation, the lipophilic CM
has the lowest molecular volume, and the interaction
of all but one residue at > 50% suggests that the helix
is in a hydrophobic environment
BM also displays a high octanol⁄ water partition
coefficient, and is therefore likely to reveal
hydropho-bic regions of ABCB1 However, unlike CM, this
probe contains a delocalized charge and is zwitterionic
in nature Presumably, it assumes a more polarized
ori-entation to accommodate this ampiphilicity Like CM,
each of the TM12 residues examined was able to
undergo covalent modification with BM (Table 1),
which also suggests that the helix is located in a
hydro-phobic environment A similar stretch of TM12
(namely V982C–V988C) displayed the greatest
propen-sity to be labelled with BM, with only isoform M986C
being not completely labelled by the probe Either side
of this central region was labelled with BM, but to
only a partial extent Unlike the case for CM, there
was considerable variation in the half-lives of labelling with BM of the TM12 mutant isoforms The rate of labelling (i.e t1⁄ 2) was divided into fast (L986C– G992C, average t1 ⁄ 2 8 min) and slow (L976C– G984C, average t1 ⁄ 2 25 min) kinetics between the carboxy-half and the amino-half, respectively So, although the helix is in a predominantly hydrophobic region, there were some differences in topography detected by the amphiphilic BM This may suggest that the carboxy region (i.e cytosolic) lies at an interface with a more hydrophilic domain of ABCB1, as this region displayed more rapid labelling kinetics This hypothesis is supported by the fact that F994C, which
is proximal to the membrane surface, has a consider-ably greater Lext (111 ± 35%) for BM than the near neighbours examined An alternative explanation for the two distinct kinetic divisions is that the amino region is closely packed with another helix of ABCB1 that imparts steric restrictions on the kinetics of label-ling in TM12
The final probe used to examine the topography of TM12 was the large hydrophilic FM; the extents and time courses of interactions are shown in Table 1 The data on extent of labelling data are in broad agreement with the information provided by BM and CM Only one residue displayed avid labelling with FM, namely F994C (Lextof 129 ± 24%), and this is at the extreme carboxy-end of TM12, in proximity to the aqueous environment The proximally located S992C was also able to interact with FM, although to only a partial degree The central and amino regions of TM12 dis-played low labelling with the hydrophilic probe How-ever, two residues (G984C and M986C) in the central region of TM12 did display labelling above back-ground, albeit with Lextvalues of approximately 20% This may reflect that these two residues, although in a hydrophobic local environment, are in the vicinity of a more hydrophilic region of ABCB1 The rapid kinetics
of labelling of M986C with both BM and FM would also support this local increase in hydrophilicity It is also worth noting that the extent of labelling is affected by numerous factors, including steric effects and local chemistry These may have differential effects
on the kinetic parameters for certain residues
Do conformational transitions alter the labelling
of residues in TM12 of ABCB1?
During the drug translocation process, ABCB1 adopts
a number of conformational states As drug transloca-tion is coupled to ATP hydrolysis, the conformatransloca-tional transitions will be driven by events at the NBDs If TM12 is involved in the coupling process between the
Fig 2 Analysis of probe labelling of mutant TM12 isoforms of
ABCB1 For each of the mutant isoforms, densitometric analysis
was used to quantify the UV images and values of labelling at each
time point These were then expressed as a percentage of the
maximal extent of G324C labelling The degree of labelling (% of
G324C level) was plotted as a function of time (min) and fitted with
an exponential reaction curve, using nonlinear least squares
regres-sion (A) Representative data for labelling of the M986C isoform
with CM ( ), FM (d) and BM (s) (B) Representative data for
label-ling of the F994C isoform with FM in the basal (d), AMP-PNP (s)
and vanadate-trapped ( ) conformational states.
Trang 5TMDs and NBDs, then it will presumably undergo
multiple topographical transitions during the catalytic
cycle The previous section outlined the overall
topog-raphy of TM12, by examining the accessibility of
introduced cysteines to maleimide-containing probes
The next phase of investigation involved trapping
ABCB1 mutant isoforms at distinct conformational
stages (e.g nucleotide-bound and immediately
posthy-drolysis) and reassessing the accessibility to maleimide
probes The data thereby identified the dynamic
changes produced during transition between various
stages of the catalytic cycle
The data in Fig 2B show a representative time
course for labelling of the F994C mutant isoform with
FM in the basal, nucleotide-bound and
vanadate-trapped conformations The nucleotide-bound
(prehy-drolytic) conformation was achieved by incubation of
the mutant isoforms with the nonhydrolysable ATP
analogue AMP-PNP, as previously described [24] The
posthydrolytic (but pre-ADP or phosphate release)
stage was produced by the vanadate-trapping
proce-dure [24] In the basal state, the protein was fully
labelled with FM (Lext of 105%); however, Lext was
reduced to 47% upon binding of AMP-PNP, and
fur-ther reduced to 14% following vanadate trapping
Accessibility data, as shown in Fig 2B, were
obtained (using multiple protein preparations) for each
mutant isoform in the three conformations
(nucleotide-free, AMP-PNP-bound and vanadate-trapped)
Experi-ments were carried out as described in the previous
section, and the Lextand t1⁄ 2parameters were obtained
from the labelling time course profiles To simplify
analysis, a qualitative representation has been adopted
(Table 2)
Conformational changes – amino region of TM12
As shown in Table 2, the amino region of TM12 (L976C–V982C) was not associated with large altera-tions in topography In particular, accessibility of the two residues to FM was negligible in the basal state, and this did not change for the nucleotide-bound and posthydrolytic states There were, however, some sub-tle changes in accessibility of the two more hydropho-bic probes For example, L976C became less accessible
to BM, but more accessible to CM, following a shift from the basal to the nucleotide-bound conformation
As ABCB1 shifted to the posthydrolytic conformation, the extent of BM labelling returned to the basal level, whereas CM accessibility was retained A980C shifted
to a low level of BM accessibility following nucleotide binding by ABCB1, and again, an opposite shift was seen for CM The subsequent transition to a vanadate-trapped state resulted in the highest possible extent of labelling for BM, but with no alteration for CM Over-all, nucleotide binding shifts the amino region to a dis-tinctly hydrophobic environment, such that labelling with the zwitterionic BM is, in fact, reduced Given that BM is ampiphilic, this would suggest a shift from
a possible interfacial region to a buried hydrophobic one Furthermore, the progression to the
posthydrolyt-ic state restored the topographposthydrolyt-ical features seen under basal conditions In complete contrast, V982C did not undergo any alterations of probe accessibility during transition to the nucleotide-bound and posthydrolytic conformational states This was the only residue exam-ined in TM12 that retaexam-ined an unaltered topography between the states despite the conformational changes within the TMDs induced by the NBDs
Table 1 Propensity for and rate of labelling of ABCB1 with thiol-reactive probes The propensity for labelling of the TM12 mutant isoforms was determined for the thiol-reactive probes CM, BM and FM The reaction was stopped by the addition of dithiothreitol, and proteins were resolved by SDS ⁄ PAGE Densitometric analysis was used to determine the amount of labelling for each ABCB1 isoform The extent (L ext ) and half-life (t1⁄ 2 ) of labelling were determined by nonlinear regression of the exponential reaction curve The Lextfor labelling is expressed
as the fraction of specific labelling of single-cysteine isoforms over the specific labelling of the G324C positive control Values represent the means ± standard errors of the mean from at least four independent protein preparations –, no labelling; ND, values where the extent of labelling was too low to accurately assign a value for t1⁄ 2
Mutant
Trang 6Conformational changes – central region
Two of the residues examined in the central region
(G984C and M986C) of TM12 have been shown to
accommodate partial labelling with FM, suggestive of
aqueous accessibility in the basal state At M986C, the
extent of labelling with the hydrophilic probe was
increased following the addition of nonhydrolysable
nucleotide This was accompanied by a moderate
increase in labelling with the zwitterionic BM, but with
a reduction in accessibility with the hydrophobic CM
This pattern of change suggests a shift towards a more
polar environment for this central residue This
appeared to be a transient shift in microenvironment,
as the posthydrolytic state adopted a topography
similar to that in the basal configuration G984C underwent a broadly similar shift in topography as M986C, although the degree of alteration was some-what less striking
Conformational changes – proximal to the central region
The region immediately proximal to the centre of TM12 (V988C–G989C) showed avid labelling by both
of the lipophilic probes (BM and CM) in the basal configurations, and there were no significant altera-tions in accessibility upon progression of the catalytic cycle Labelling of V988C and G989C with the hydro-philic FM was negligible, regardless of the conforma-tional state The refractoriness of labelling to conformational change is clearly demonstrated by G989C In particular, this residue displayed the lowest overall accessibility to covalent modification, regardless
of the conformational state At no stage of the cata-lytic cycle was either CM or BM able to fully label G989C, which was the only residue to exhibit this property Similarly, no interaction between the hydro-philic FM and G989C was observed The variation in physicochemical properties of the three probes suggests that the inherently low labelling at any stage of the catalytic cycle was unlikely to result from the local sol-vent environment A more likely explanation is steric hindrance to labelling by neighbouring residues or heli-ces in the TMD The labelling properties of V988C– G989C suggest that this region of TM12 undergoes minimal conformational transition
Conformational changes – carboxy region Considerably greater changes in accessibility to probes were observed at the extreme carboxy region of TM12, suggesting a more prominent role in mediating confor-mational transitions In the basal state, none of the probes could effect complete labelling of the S992C isoform However, progression to the nucleotide-bound state resulted in a universal increase in accessibility of the residue to covalent modification by all three probes Further progression to the posthydrolytic state caused a reversion in accessibility in comparison to that seen in the basal state The uniform changes in accessibility to three probes with distinct chemical properties suggest that the adoption of the nucleotide-bound state relieves the steric hindrance to labelling found in the basal conformation, and that this is restored as the catalytic cycle continues
F994C displays the highest accessibility of any resi-due in the basal conformation of ABCB1, which may
Table 2 Relative accessibilities of TM12 residues Accessibilities
of cysteines to FM, BM and CM were determined at distinct
stages of the catalytic cycle for each ABCB1 isoform The extent of
labelling was compared with that of the cysteine-less ABCB1
iso-form Basal refers to the nucleotide-free state, whereas the
AMP-PNP and Vi-trapped states mimic prehydrolytic and posthydrolytic
states of the protein, respectively +++, complete labelling
(L ext > 75%); ++, partial labelling (L ext = 50–75%); +, weak labelling
(Lext< 50%); ), labelling below the amount observed for
cysteine-less ABCB1 All values were determined as described in Table 1
and obtained from four independent protein preparations.
ABCB1 isoform
Catalytic
Trang 7reflect localization at the membrane–solute interface.
There was no alteration in the extent of labelling by
BM in any conformational state examined In contrast,
there was a dramatic reduction in labelling by the
hydrophilic FM as the protein progressed to the
nucle-otide-bound and posthydrolytic states This was
accompanied by a concomitant increase in accessibility
to the hydrophobic CM Clearly, F994C undergoes
considerable changes in accessibility, suggestive of
a move from a relatively hydrophilic region to a
more lipophilic one as ABCB1 binds and hydrolyses
nucleotide
Discussion
TM12 has previously been demonstrated to play an
integral role in coupling between the drug binding and
translocation process (TMD), with the hydrolysis of
nucleotide (NBD) [15,29] Moreover, perturbation of
TM12 altered not only drug-stimulated ATP
hydroly-sis, but also the inherent (basal) hydrolytic activity
The latter demonstrates that activity of the NBDs,
even in the absence of substrate, is subject to some
degree of control or modulation by the TMDs of
ABCB1 TM6 in the amino-half of ABCB1 has often
been regarded as a mirror image of TM12, but, from a
purely functional perspective, cysteine introduction
within TM12 generated considerably greater functional
consequences for ABCB1 than corresponding
muta-tions in TM6 The present study investigated whether
the ‘mirror image’ relationship holds true, particularly
with respect to the topographical changes in TM12
throughout the catalytic cycle
The topographical changes were examined by
intro-ducing cysteines at distinct positions in TM12 and
assessing their accessibility to covalent modification
with thiol-reactive probes In order to determine how
changes in the extent and rate of labelling reflect
con-formational changes in TM12, we used molecular
models of ABCB1 [30] in the basal and ATP-bound
states as the basis for in silico characterization
Homol-ogy modelling has previously been used to characterize
the effects of mutations in TM12⁄ TM6 on the overall
function of ABCB1 and to interpret the changes in
labelling accessibility that occur in TM6 [27] This
approach provided a mechanistic explanation for the
role of TM6 in the translocation mechanism of
ABCB1, and was reproduced in the present
investiga-tion for TM12
The changes in the accessibility to probes of mutated
residues within TM12 showed both increases and
decreases in the propensity for labelling throughout
the catalytic states, suggesting that TM12 undergoes
conformational alterations, or is subjected to changes
in its local environment There were two major obser-vations to be drawn from studying the topography of TM12 in the homology model First, the midregion of the helix, i.e V982–G984, was rigid with respect to the intracellular and extracellular sections of the helix in the basal and ATP-bound states of the model, and that this section of TM12 appeared to act as an anchor around which the rest of the helix moved Second, the model shows that the intracellular part of TM12 also contributes residues (between Met986 and Ser992) to the band of hydrophilic residues that line the central aqueous pore in ABCB1 (Fig 3) Both of these obser-vations can be rationalized with the molecular models for ABCB1
The homology models predict that both V982C and G984C, located within the centre of the helix, experi-ence little change in molecular environment upon ATP binding, which is in agreement with the biochemical data TM12 is predicted by homology modelling to rotate by approximately a quarter of a turn following ATP binding, which is also in agreement with the bio-chemical data This rotation is accompanied by a dis-placement towards Tyr953 (TM11), the nearest neighbour of Val982 in the closed-state model Despite this motion, Val982 does not form a close contact with Tyr953, and the local environment is therefore unchanged and does not impact on the accessibility of the residue to the fluorescent probes In support of this, no change in labelling was observed In addition, the position of Gly984 does not change between the closed and open states of the model, and would not result in a change in the polarity of the environment This rigidity is clearly reflected in the labelling experi-ment, which demonstrated little change in residue accessibility among the catalytic states
A hydrophilic band of residues in the TMD lines the central cavity of ABCB1 (Fig 3) and presumably con-tributes to the solvent accessibility of the residues in this region M986C and S992C (Fig 3) on TM12 straddle the boundaries of this hydrophilic band, and also face directly into the presumed translocation pore These two residues were readily labelled by the fluores-cent probes, and displayed differences in accessibility between the conformational states examined It has been suggested that conformational transitions may alter the nature of the residues lining the translocation pore [10,31], e.g from hydrophobic to hydrophilic This type of switch may be responsible for the cycling
of affinity of ABCB1 for drug substrates during the translocation process [24] Such observations have been made in both the ABCB1 homology models [30] and the low-resolution crystal structure of ABCB1 [6]
Trang 8Surprisingly, although FM labels G984C, the
homol-ogy model suggests that this residue faces into the lipid
bilayer However, G984C is not in a very densely
packed region, and it may be possible for FM to gain
access to the residue via the translocation pore In
addition, the loss of labelling of G984C with FM
fol-lowing progression to a vanadate-trapped state
sug-gests that labelling is not optimal and therefore is very
sensitive to even minor environmental changes S992C
and F994C are believed to be located at the boundary
of the membrane Indeed, Ser992 faces into the
trans-location pore near the entrance and is highly solvent
exposed Consequently, both residues are accessible to
labelling by FM Moreover, Phe994 is located within
the prominent kink in TM12, which was first identified
by the homology model of ABCB1 and subsequently
confirmed in the crystal structure [6,30]
It is conceivable that this kink may facilitate (or
dampen) transmission of movement initiated by events
in the NBDs to conformational changes in TM12 For
example, upon ATP binding, the NBDs will form a
dimer to enable hydrolysis of nucleotide The resultant
hydrolytic cleavage of ATP will result in
disengage-ment of the dimer because of the considerable
repul-sion between ADP and Pi TM12 is directly linked to
NBD2, and is therefore ideally placed to transmit these
conformational changes The communication would
extend in both directions, and the central region of TM12 would act as a stationary element about which the conformational changes occur Similarly, the bind-ing of substrates is thought to stimulate ATP hydroly-sis by facilitating conformational changes associated with NBD dimer assembly This might occur through communication between the drug-binding site(s) and TM12 In fact, mutations in TM12 were demonstrated
to affect transport or ATPase activity [32,33], in partic-ular, the stimulation of ATP hydrolysis by vinblastine and nicardipine [15] These two compounds are known
to interact at pharmacologically distinct (allosterically linked) sites in ABCB1 [34], and this supports the notion of TM12 acting as a key conduit Moreover, the observation that mutations in TM12 could alter stable ATP binding by the NBDs further supports the tight coupling imparted by TM12 on the process of ATP hydrolysis Further biochemical and structural studies will reveal the exact contribution of individual residues in TM12 to drug binding and the role of the TM12 anchor region identified here in allosteric communication
A previous investigation has also demonstrated that, upon ATP binding, the extracellular faces of the two helices can form a zero-length cross-link, indicating a close approach [35] This close approach of the helices
is relaxed following progression of ATP hydrolysis
Fig 3 Molecular modelling of the TMDs in ABCB1 Representations of the TMDs of ABCB1 obtained from molecular modelling are shown, with the NBDs removed for clarity (A) The TMD of ABCB1 predicted to represent the basal (nucleotide-free) conformation (B) The TMD of ABCB1 predicted to occur in the nucleotide-bound conformation of the protein The two TMDs of ABCB1 are shown with helices from TMD1 (N-terminal) in grey and those from TMD2 (C-terminal) in black The TMDs display a hydrophilic band of residues (cyan) that lines the central cavity, and these are shown in the ‘space-fill’ representation Relative to TM12, the hydrophilic band is located at a depth that corre-sponds to the region bounded by residues Met986 and Ser992, which are depicted in purple (C) The TMD helices (cylinders) neighbouring,
or in the vicinity of, TM12 (ribbon) The helices are shown in the nucleotide-free (bold) or bound (pastel) conformations: orange, TM9; gold, TM10; red, TM11 All other helices have been removed from the diagram to aid clarity The diagram also demonstrates (comparison of bold and pastel representations) that TM12 undergoes relatively little motion in switching between these conformations The structures are shown in the panel as viewed from the translocation pore; the relative environments of V982C (cyan) and G984C (blue) are unaltered by nucleotide binding The nearest neighbouring residue, Tyr953, is shown in red space-fill representation.
Trang 9Moreover, there is a large amount of evidence
demon-strating that TM6 and TM12 are intimately involved
in numerous aspects of the molecular mechanism of
ABCB1 The present investigation focused on TM12,
and it is clear that the helix does undergo
conforma-tional changes, with the centre of the helix being rigid
and motion being amplified at the extracellular and
intracellular ends of the helix
Experimental procedures
Materials
Octyl-b-d-glucoside, C219 antibody and Ni2+
–nitrilotriace-tic acid His Bind Superflow resin were obtained from
Merck Chemicals (Nottingham, UK) Dimethylsulfoxide,
Na2ATP, AMP-PNP, sodium orthovanadate and
choles-terol were purchased from Sigma Aldrich (Poole, UK)
Crude Escherichia coli lipid extract was obtained from
Avanti Polar Lipids (Alabaster, USA) Insect-Xpress
medium was purchased from Lonza (Wokingham, UK) and
Excell 405 from SAFC Biosciences (Andover, UK) CM,
FM and BM were purchased from Molecular Probes
(Leiden, The Netherlands)
Site-directed mutagenesis of TM12 in
ABCB1 – introduction of cysteines
Mutants were constructed with QuikChange or Altered
Sites II mutagenesis systems with a pAlter-MCHS or
pFast-Bac1-MCHS template The MCHS cDNA encodes an
ABCB1 isoform devoid of cysteines with a C-terminal His6
tag and numerous strategically inserted restriction enzyme
sites Full details of the construction of mutant ABCB1
isoforms have been given in previous publications [25,36]
Expression, purification and reconstitution of
ABCB1
Recombinant baculovirus was generated using the
Bac-to-Bac baculovirus expression system, as previously described
[25,36] and according to the manufacturer’s instructions
(Invitrogen) Trichoplusia ni (High-five) cells were infected
with recombinant baculovirus at a multiplicity of infection
of 5, and harvested 72 h postinfection by centrifugation
(2000 g, 10 min) For comparative analysis of protein
expression, 2· 106cells were resuspended in NaCl⁄ Pi
sup-plemented with 2% (w⁄ v) SDS, and proteins were resolved
by SDS⁄ PAGE ABCB1 was detected with the C219
anti-body following immunoblotting [37]
For large-scale expression of ABCB1 isoforms, 1.5· 109
T ni (High-five) cells were infected, and cell membranes
were isolated by nitrogen cavitation and density gradient
ultracentrifugation and stored at )80 C for up to 1 year
[25,36] ABCB1 isoforms were purified by immobilized metal affinity chromatography (Ni2+–nitrilotriacetic acid resin), and reconstituted by the detergent adsorption tech-nique [25,36] Confirmation of reconstitution was per-formed by examining the relative migration of lipid and protein through sucrose density (0–30% w⁄ v) gradients Protein concentration following reconstitution was deter-mined with an adapted Lowry colorimetric assay with BSA
as standard (DC-Brad Protein Assay; BioRad) [38]
Fluorescent labelling of single-cysteine isoforms
of ABCB1
The topography of TM12 was assessed by following the labelling kinetics of each single-cysteine mutant isoform with three fluorescent thiol-reactive probes The probes display distinct physicochemical properties, with variations
in charge, size and hydrophobicity [25]; for example, CM is hydrophobic, FM is hydrophilic and BM is zwitterionic Purified, reconstituted ABCB1 isoforms (2 lg) were incu-bated with 10 lm CM, BM or FM for 0, 10, 30, 60, 120 and 300 min in the dark at 20C The ligand was added from concentrated stocks in dimethylsulfoxide, and the final solvent concentration was maintained at < 0.05% (v⁄ v)
A 100-fold molar excess of probe to protein was used to facilitate labelling and prevent significant depletion of the probes The reaction was stopped by the addition of
100 lm dithiothreitol, which binds avidly to unreacted maleimide probe, and subsequently placed on ice The pro-tein was diluted 1 : 1 with buffer (50 mm Tris⁄ HCl, pH 7.4,
150 mm NH4Cl, 5 mm MgSO4, 0.02% NaN3) to reduce glycerol content, and centrifuged for 30 min at 125 000 g and 4C to remove unbound probe The pellet was washed and then resuspended in 20 lL Laemmli sample buffer, and proteins were resolved by 7.5% (v⁄ v) SDS ⁄ PAGE Nonspe-cific association of the fluorescent probe with the protein and lipid membrane was determined using a cysteine-free ABCB1 isoform The G324C mutation, located on a freely accessible extracellular loop, has previously been demon-strated to be freely accessible to each maleimide probe [28]; labelling of the isoform containing this mutation was there-fore assigned the value of 100% after 300 min Both the cysteine-less and G324C isoforms were incubated with
10 lm probe for 300 min and treated identically to the other isoforms The extent of labelling for each single-cyste-ine mutant was therefore determsingle-cyste-ined by comparison with G324C In order to calculate the specificity of labelling for each single-cysteine mutant, the background or nonspecific labelling of the cysteine-less isoform was subtracted The propensity for labelling was calculated with the following equation:
L¼ Liso Lcys
L324C Lcys
100
Trang 10where L is extent of labelling (%), Liso is the extent
of isoform labelling, Lcys is labelling of the cysteine-less
isoform, and L324Cis labelling of the G324C isoform
The extent of fluorescence labelling for ABCB1 mutant
isoforms was also determined in the nucleotide-bound
state by trapping with AMP-PNP The ABCB1
nucleo-tide-bound conformation was generated by the addition of
AMP-PNP (2 mm), followed by a 20 min incubation at
20C Trapping of ABCB1 in the posthydrolytic state was
achieved by the addition of 300 lm orthovanadate (Vi)
and 2 mm ATP, followed by a 30 min incubation at 37C
in order to generate the ADPÆVitransition state
intermedi-ate [22,24,25] Fluorescence labelling was subsequently
car-ried out as detailed in the preceding paragraph
The extent of labelling was determined by examining the
gel using the BioDocIT Imaging System (UVP), with a UV
light source of wavelength 302 nm and a CCD camera The
gel was subsequently stained with PageBlue to validate
equivalent protein loading Densitometric analysis (scion
image) was used to quantify the extent of labelling The
maximum extent of labelling (Lmax) and half-time of
label-ling (t1 ⁄ 2) were determined by nonlinear regression of the
exponential reaction curve (graphpad prism 4.0) to plots
of labelling as a function of time:
L¼ Lmax 1 ekt
where L is the percentage of labelling, Lmax is the
maxi-mum extent of labelling (%), k is the observed rate
con-stant for labelling (min)1), and t is time (min) The
labelling rate constant was converted to half-time of
label-ling according to the following relationship:
t1=2¼ Ln2=k
Statistical analysis
All data manipulations and statistical analyses were
per-formed using graphpad prism 4.0 Comparison of datasets
for each isoform was performed with Student’s t-test or
ANOVA (where n > 3), applying Dunnett’s test, where
sig-nificance was determined by a P-value < 0.05 Values
reported correspond to means ± standard errors of the
mean obtained from at least four independent preparations
of ABCB1
Homology modelling
A homology model of a nucleotide-free, open-state human
ABCB1 was developed from the open-state mouse
P-glyco-protein crystal structure (3G5U.pdb), using the swissmodel
homology modelling server [39], with the aim of producing
an open-state homology model of human ABCB1 that
would complement the previous closed-state Sav1866-based
ABCB1 model [30] The sequence identity between human
ABCB1 and mouse P-glycoprotein is 86%, giving a very
high degree of confidence to the sequence alignment of the resulting model To verify that the residue threading of this open-state ABCB1 model corresponds to the previously developed closed-state ABCB1 model [30], the sequence alignments were cross-referenced to ensure that there was positional correspondence of the residues in both conforma-tions The series of single-point mutations to cysteine were performed at positions 976, 978, 980, 988, 989 and 990 in the open-state ABCB1 homology model, to give a set of six single-point mutation open-state ABCB1 models These models were developed with the method described in Storm
et al [28]; they provide an alternative conformation to the set of closed-state ABCB1 point mutations developed in Crowley et al [15], and allow a comparison of the local environment of each residue in both the open and closed conformation of ABCB1
Acknowledgements
E Crowley was generously supported by a Cancer Research UK Studentship (C362⁄ A5502) awarded to
I D Kerr and R Callaghan M L O’Mara is supported
by a University of Queensland Post-doctoral Fellowship
References
1 Cordon-Cardo C, O’Brien JP, Boccia J, Casals D, Bertino JR & Melamed MR (1990) Expression of the multidrug resistance gene product (P-glycoprotein) in human normal and tumor tissues J Histochem Cytochem 38, 1277–1287
2 Cordon-Cardo C, O’Brien JP, Casals D, Rittman-Gra-uer L, Biedler JL, Melamed MR & Bertino JR (1989) Multidrug-resistance gene (P-glycoprotein) is expressed
by endothelial cells at blood–brain barrier sites Proc Natl Acad Sci USA 86, 695–698
3 Leslie EM, Deeley RG & Cole SPC (2005) Multidrug resistance proteins: role of P-glycoprotein, MRP1, MRP2, and BCRP (ABCG2) in tissue defense Toxicol Appl Pharmacol 204, 216–237
4 Glavinas H, Krajcsi P, Cserepes J & Sarkadi B (2004) The role of ABC transporters in drug resistance, metab-olism and toxicity Curr Drug Deliv 1, 27–42
5 Loo TW & Clarke DM (1996) The minimum functional unit of human P-glycoprotein appears to be a mono-mer J Biol Chem 271, 27488–27492
6 Aller SG, Yu J, Ward A, Weng Y, Chittaboina S, Zhuo
R, Harrell PM, Trinh YT, Zhang Q, Urbatsch IL et al (2009) Structure of P-glycoprotein reveals a molecular basis for poly-specific drug binding Science 323, 1718– 1722
7 Dawson RJ & Locher KP (2006) Structure of a bacterial multidrug ABC transporter Nature 443, 180– 185