CHAPTER 5 – SUBSTRATE BINDING SITES IN ABC TRANSPORTERS

CHAPTER 5 – SUBSTRATE BINDING SITES IN ABC TRANSPORTERS CHAPTER 5 – SUBSTRATE BINDING SITES IN ABC TRANSPORTERS CHAPTER 5 – SUBSTRATE BINDING SITES IN ABC TRANSPORTERS CHAPTER 5 – SUBSTRATE BINDING SITES IN ABC TRANSPORTERS CHAPTER 5 – SUBSTRATE BINDING SITES IN ABC TRANSPORTERS CHAPTER 5 – SUBSTRATE BINDING SITES IN ABC TRANSPORTERS CHAPTER 5 – SUBSTRATE BINDING SITES IN ABC TRANSPORTERS CHAPTER 5 – SUBSTRATE BINDING SITES IN ABC TRANSPORTERS CHAPTER 5 – SUBSTRATE BINDING SITES IN ABC TRANSPORTERS

G ENERAL

As is evident from this volume, the ABC family

of membrane transporters comprises a

fascinat-ingly diverse range of proteins, which mediate a

variety of different types of transport processes

A particularly distinguishing feature of this

family is that the substrates recognized by ABC

proteins appear to know no chemical, physical

or functional boundaries Perhaps this is not

sur-prising given that gaining membership to this

family is purely based on structural features

rather than the nature of substrate translocated

With a few apparent exceptions (e.g CFTR,

SUR), all ABC proteins are active transporters

that move substrates against their concentration

gradients The ‘engine room’ of all ABC proteins

comprises the two nucleotide-binding domains

(NBDs), whose catalytic activity drives the

trans-port process, and the NBDs share a high level of

sequence similarity and common overall

mech-anism Does this mean that the substrate

inter-actions and translocation processes in these

proteins, primarily involving the

transmem-brane domains, also display common themes,

irrespective of the array of physiological

processes in which ABC transporters areinvolved?

Another issue addressed in this chapterconcerns the number of substrate-binding sitespresent in ABC transporters Many ABC trans-porters interact with a single substrate (or class

of substrates) and this is particularly evidentwith the bacterial import pumps, which areoften associated with dedicated substrate ‘cap-ture and delivery’ proteins These transporters

provide a marked contrast to the so-called drug pumps, which interact with a myriad of

multi-compounds ABC transporters may contain asingle all-encompassing substrate-binding site,

or multiple sites with highly selective substratespecificity There is a paucity of information onwhere substrate-binding sites are located onABC proteins If ABC proteins have more thanone site, do these sites interact and, if so, what isthe nature of the communication between them?

The aim of this chapter will be to summarizeour current insights into the physicochemicalaspects of substrate interactions with the differ-ent types of ABC proteins mentioned above

The information available will be used to ulate on possible common molecular mecha-nisms of substrate translocation amongst ABCtransporters

spec-ABC Proteins: From Bacteria to Man ISBN 0-12-352551-9

Copyright 2003 Elsevier Science Ltd All rights of reproduction in any form reserved

P ROPERTIES OF

ABC PROTEINS MEDIATE A VARIETY OF

DIFFERENT TRANSPORT PROCESSES

The variety of substrates handled by different

ABC transporters is enormous As shown in

Figure 5.1, the substrates transported include

the majority of organic and inorganic chemical

classes found in cells: amino acids, sugars,

inor-ganic ions, lipids, polysaccharides, peptides and

even proteins, in addition to compounds that

are foreign to the organism itself Hence, the

ABC transporter family is very different from

other transport protein families, which are

char-acterized by, or even named after, the

com-pound(s) translocated (van Winkle, 1999).

Whereas most eukaryotic ABC transportersappear to mediate substrate efflux only, prokary-

otic members are divided into import and export

proteins (Higgins, 1992) The bacterial importers

usually interact with accessory proteins (e.g

periplasmic binding proteins) that bind and

deliver substrate to the translocation machinery

The high-resolution structures for periplasmic

binding proteins shown to interact witholigopeptides (OppA), histidine (HisJ) andribose (RBP) demonstrate the presence of bind-ing sites with high specificity for the transported

substrate (Mowbray and Cole, 1992; Tame et al., 1994; Yao et al., 1994) On the other hand, most

prokaryotic export proteins (e.g those involved

in antibiotic extrusion) mediate highly selectivetransport that is independent of substrate-binding proteins Furthermore, several prokary-otic and eukaryotic ABC transporters display anapparent ‘broad selectivity’ for substrates andare known as multidrug pumps Well-knownexamples are P-glycoprotein (Pgp) and MRP1,overexpression of which are major causes ofresistance of human tumors to chemotherapy

(Cole and Deeley, 1998; Gottesman et al., 1995; Lum et al., 1993), and LmrA, a bacterial homo-

logue of Pgp which mediates transport ofamphiphilic and toxic compounds, and of clini-

cally relevant antibiotics (Margolles et al., 1999; Putman et al., 2000; van Veen et al., 1996, 1998).

MRP1, which has a broad specificity for gated drugs, is the only ABC protein demon-strated thus far to mediate symport, by virtue ofits ability to co-transport drugs and glutathione(Cole and Deeley, 1998)

conju-The heterogeneity of ABC transporters hashindered the elucidation of the molecular basis

of substrate recognition by these proteins, and the subsequent steps involved in thetranslocation process Many ABC transporters

CoA

H2C O CO

O CO CH x

O

O

P O CH2 O

COOCH2COOCH2N

N N OH

H

N R HO

CH2O HO

NH2

O O

OH

SH H H

Ions

Bile acids Steroids Cholesterol

Multidrugs Multidrug

conjugates

Peptides Hormones

Fatty acids Phospholipids

Eukaryotic substrates

Figure 5.1 Illustration of the various general chemical species that are recognized and transported by

eukaryotic ABC proteins.

are intimately involved in disease states and

the elucidation of the molecular details of their

substrate-binding site(s) would prove

invalu-able in designing drugs to target specific

pro-teins in the clinical setting

SPECIFICABC PROTEIN–SUBSTRATE

INTERACTIONS

Specific characteristics required of transport

substrates for recognition by ABC transporters

have been delineated most thoroughly for the

eukaryotic TAP transporters These proteins

mediate the transport of peptides across the

endoplasmic reticulum to facilitate their

load-ing on the MHC class I complex

(Lankat-Buttgereit and Tampe, 1999; Uebel and Tampe,

1999) Peptides containing 9–16 amino acids

are the preferred substrates, although lengths

of up to 40 residues are possible, with reduced

efficiency (Koopmann et al., 1996) The

involve-ment of TAP transporters in cellular antigen

processing and the corresponding variability

in peptide substrates suggests a low-specificity

binding site However, an elegant

investiga-tion using a combinatorial peptide library has

revealed marked selectivity (Uebel et al., 1997).

The transporter exhibits preference for

hydro-phobic and positively charged amino acids on

the C-terminus of peptide substrates, whilst

aspartate or glycine residues are not tolerated

(Uebel et al., 1997) Positions 1 and 3 at the

N-terminus of peptides also greatly affect

bind-ing to TAP, although a strict pharmacophoric

preference is not obvious The interaction of

positions 1, 3 and 9 of a nonapeptide substrate

with the TAP-binding site appears to involve

contributions from the peptide backbone and the

side-chains In contrast, positions 4–8 provide

almost no determinants for substrate–protein

interaction Together these physicochemical

characteristics of peptide–TAP interaction

sug-gest that the peptide ‘docks’ at two sites by

virtue of its N- and C-terminal residues, whilst

the central amino acids span a cavity within the

transporter structure

THE CONCEPT OF MULTIDRUG PUMPS

Unfortunately, no such directed investigations

have been possible for the ABC proteins able

to transport a variety of compounds that share

no discernible structural similarities Initially,

these ‘multidrug transporters’ were thought

to contravene a central dogma of substraterecognition by proteins; namely the ‘lock–keyhypothesis’ postulated by Emil Fischer in 1894

This hypothesis, or its adaptation to an

‘induced fit model’ (Koshland, 1987),

ade-quately describes the interactions of enzymes ortransporters with hydrophilic substrates Suchsystems are characterized by highly specificinteractions between protein side-chains and

the substrate It is inconceivable that all the

compounds recognized by multidrug pumpsare able to invoke such specific interactionswith a single protein Most of the compoundsrecognized by multidrug pumps are hydropho-bic or amphiphilic organic molecules Theirinteraction with these pumps is perhaps gov-erned by a different set of ‘rules’ than thoseobserved for hydrophilic agents Consequently,

it was suggested that the recognition site inmultidrug pumps might be a simple non-specific hydrophobic core or pocket (Gottesmanand Pastan, 1993) This notion may now be dis-counted owing to the quite large observed dif-ferences in the relative affinities of compoundsfor interaction with multidrug ABC trans-

porters such as Pgp (Martin et al., 1999; Sharom

et al., 1999), LmrA (van Veen et al., 1998), and PDR5 (Rogers et al., 2001) Therefore, the multi-

drug pumps should not be considered as specific transporters, but rather as transportersdisplaying polyspecific recognition of sub-strates, not necessarily different from TAP (seeabove) or the OppA oligopeptide-binding pro-tein, where three-dimensional (3-D) structure

non-has been solved (Tame et al., 1994)

Unfor-tunately, multidrug pumps, by virtue of thischaracteristic, often provide a general pathwayfor mediating drug resistance, which is notrestricted to a single drug, in a variety of clinical

settings (Lum et al., 1993; Nikaido, 1994).

SPECIFIC CHARACTERISTICS OF SUBSTRATES TRANSPORTED BY MULTIDRUG TRANSPORTERS

The promiscuity with which Pgp recognizes strates has sparked many investigations into theidentification of potent blockers of the transportprocess in tumor cells The reader is directed totwo reviews that provide a guide to the manydifferent clinically relevant compounds identi-

sub-fied with Pgp inhibitory actions (Lum et al., 1993;

Sikic, 1997) However, these studies have notprovided significant information on the over-all chemical features that characterize Pgp

substrates To provide such information, several

groups embarked on manipulation of the

phys-icochemical properties of known Pgp

sub-strates in an effort to elucidate key molecular

constituents for recognition by the protein

(Chiba et al., 1996; Ford et al., 1990; Horton

et al., 1993; Lawrence et al., 2001; Pearce et al.,

1989; Tang-Wai et al., 1993; Toffoli et al., 1995).

Figure 5.2 shows the general structures of the

compounds used Unfortunately, these

investi-gations failed to elucidate precise and conserved

pharmacophoric elements for Pgp substrates

However, they did highlight some key physical

requirements Hydrophobicity is a key element

and planar aromatic groups contribute

signifi-cantly to this property A basic nitrogen atom is

frequently observed and a tertiary amino moiety

is associated with the ability of compounds to

display high-affinity interaction with the protein

Compounds for which the nitrogen is located

within non-aromatic rings display the greatest

potency to interact with, and bind to, Pgp

Hydrogen bonds play major roles in the actions of many biological molecules and may

inter-impart a high specificity by virtue of their

requirement for directionality (Fersht, 1998).This prompted a screen of compounds that hadpreviously been examined for their interactionwith Pgp, to determine the percentage of con-stituent groups capable of mediating hydrogenbonds (Seelig, 1998) This exhaustive screenestablished that compounds known to interactwith Pgp display noticeable clustering of elec-tron donor groups that are vital to create hydro-gen bonds Moreover, these clusters displaycharacteristic relative spatial arrangements oftheir electron donor groups On this basis, theauthor suggested that two distinct spatial pat-terns of electron donor groups are required for acompound to interact with Pgp Electron donorgroups with spatial separations of 2.5⫾ 0.3 Åare classified as type I units Type II units on theother hand have a separation of 4.6⫾ 0.6 Åbetween two donor groups, or the outer twogroups in a series of three Interestingly, muta-tion of residues 939 and 941 in Pgp to non-hydrogen-bonding side-chain groups impairsinteraction of the protein with substrates (Kajiji

et al., 1993) These residues are located in

trans-membrane segment (TM) 11, which has been

R1

R2 R1 R3

OH O

CH3S

R3 R2

R1

R4 R6

R1

R2 O N N

strongly implicated in drug binding (Loo and

Clarke, 1999a) More recently, the screen of

hydrogen-bonding groups was used to identify

or correlate to the type of drug interaction with

Pgp (Seelig and Landwojtowicz, 2000) There was

a positive correlation between the propensities of

a drug to form hydrogen bonds and inhibition of

Pgp function Compounds such as cyclosporin A,

which may form an extensive network of

hydro-gen bonds, are potent and long-lasting inhibitors

owing to the resultant low dissociation rate from

the protein (Seelig and Landwojtowicz, 2000) It

is thought that such compounds will be poorly

transported by Pgp MRP1 substrates share

chemical properties with Pgp substrates, and

often contain at least one electrically neutral type

I unit together with one negatively charged type I

unit or two electrically neutral type I units (Seelig

et al., 2000) Compounds with cationic type I

units, which are good substrates for Pgp, are not

transported by MRP1

In summary, the large number of exhaustivestudies employing chemical modifications of

substrates or correlations with biophysical

prop-erties have given guidelines for physicochemical

properties required for interaction of molecules

with Pgp Unfortunately, however, they have not

produced a clear idea of what constitutes a strate of a multidrug pump such as Pgp

sub-CAN A DISTANT STRANGER PROVIDE THE CLUE TO UNRAVELING REQUISITE FEATURES OF SUBSTRATE INTERACTION?

The concept of multidrug pumps is by no meansrestricted to ABC proteins Secondary trans-porter families such as the major facilitatorsuperfamily (MFS), the small multidrug resis-tance family (SMR), the resistance-nodulation-cell division (RND) family, and the multidrugand toxic compounds extrusion (MATE) familyall contain multidrug pumps (for reviews seeHiggins, 1992; Marger and Saier, 1993; Paulsen

et al., 1996; Putman et al., 2000; Saier et al., 1994).

There are over a hundred known multidrugpumps and their distribution is widespread,with examples in mammalian cells, lowereukaryotes, eubacteria and archaea Interest-ingly, there is a small selection of compoundsthat appear to be ‘universal’ substrates for unre-lated multidrug pumps, such as human Pgp,

Lactococcus lactis LmrA, Escherichia coli MdfA

and Bacillus subtilis Bmr (Figure 5.3).These

Eukaryotes Eubacteria Archae

ABC MFS SMR RND

Tetraphenylphosphonium

(TPP)

Ethidium bromide Rhodamine 6G

Figure 5.3 Rhodamine 6G, ethidium bromide and tetraphenylphosphonium are transported by multidrug

pumps from a wide distribution of organisms and belonging to many different transporter families.

compounds were included in a screen of

poten-tial Pgp substrates (Seelig, 1998), and display

the characteristic physicochemical features and

spatial arrangements of electron donor groups

necessary for interaction

Investigations into drug resistance in B subtilis

have provided a new avenue to understanding

substrate interaction with multidrug pumps

in the ABC family The expression of the Bmr

transporter is regulated by the BmrR

tran-scription factor, which is activated by binding

of aromatic cationic substrates for the Bmr

transporter (Markham et al., 1997) Elucidating

the crystal structure of this transcriptional

acti-vator to 2.8 Å resolution in the presence of

sub-strate has provided considerable insight into

the molecular basis of multidrug recognition

(Zheleznova et al., 1999) The authors argue

that two main factors are involved in this

polyspecific interaction Firstly, the binding of

substrates has a minimal hydrogen-bonding

component, but relies instead on contributions

from Van der Waals forces, stacking

interac-tions, electrostatic forces and the hydrophobic

effect to reduce water contacts (Zheleznova

et al., 1999) Secondly, the drug-binding pocket

is capable of producing structural

rearrange-ments to accommodate the substrate in a

man-ner analogous to the ‘induced-fit’ model The

processes underlying drug–protein interactions

deviate from the rigid molecular specificity

associated with dedicated unisubstrate

trans-porters, and may contribute to the ability of

multidrug ABC transporters such as Pgp, MRP,

LmrA and others to confound chemotherapy in

a variety of clinical settings Clearly, it is vital

that the location of binding sites within

multi-drug pumps are elucidated in order that we

may inhibit the actions of these proteins

In order to mediate the translocation of

sub-strates across biological membranes, ABC

trans-porters must contain domains and regions that

interact with the transported substrate Theperiplasmic binding proteins are essential indetermining substrate specificity in soluteuptake systems of Gram-negative bacteria.However, mutant bacterial strains without substrate-binding proteins still exhibit specificuptake of maltose and histidine via their respec-tive ABC transporters (Petronilli and Ames,1991; Treptow and Shuman, 1985) Furthermore,mutations that alter the selectivity of the histi-

dine transporter in Salmonella typhimurium from

L-histidine to L-histidinol were found to localize

as amino acid deletions in the membrane-bound

HisM protein (Payne et al., 1985) The ability of the maltose transporter in E coli to transport p-nitrophenyl-␣-maltoside was shown to bedependent on mutations in the transmembrane

domains (TMDs) of the transporter (Reyes et al.,

1986) These early investigations of ABC port systems clearly demonstrate that TMDs areinvolved in substrate recognition and transloca-tion, even for those transporters with a periplas-mic binding protein

trans-For several ABC transporters that do not have

an extracellular substrate-binding protein, theTMDs have also been demonstrated to mediatesubstrate recognition For example, the replace-ment of charged residues lining the transmem-brane pore of the chloride channel CFTRchanges its ion selectivity profile (Anderson

et al., 1991) The SUR1 and SUR2 receptor

pro-teins confer different responsiveness of theirassociated Kir6.1 protein to the channel openerglibenclamide The differential effects of the sul-fonylurea receptor proteins SUR1 and SUR2have also been related to differences in the pri-mary structures of the TMDs in these proteins

(Babenko et al., 2000; Morbach et al., 1993).

Investigations using chimeric TAP1/2 porters in which the TAP1 TMD was replaced

trans-by the TAP2 TMD, and vice versa, have revealedthat both TMDs are essential for high-affinitypeptide binding In the absence of either, sub-strate binding to the TAP complex is impaired,indicating specific roles for each TMD in sub-

strate recognition (Arora et al., 2001) A more

detailed analysis of residues in TAP2 controllingpeptide binding/recognition utilized transpor-ters composed of the ‘a’ and ‘u’ allele-encodedrat TAP2 proteins These TAP2 proteins displaydifferent substrate specificities Chimeras of therespective proteins demonstrated that criticalresidues in determining substrate specificity arelocated in putative cytoplasmic loops in theTMD, close to the plasma membrane (Momburg

et al., 1996).

The role of TMDs in substrate specificity hasbeen most extensively investigated for Pgp.

Evidence is accumulating that the drug–protein

interactions, which determine the binding

spec-ificity of Pgp, are organized within the TMDs

of the proteins Independent photoaffinity

labeling and epitope mapping studies of Pgp

involving the 1,4-dihydropyridine derivative

azidopine (Bruggemann et al., 1992),

iodoaryl-azidoprazosin (Greenberger, 1993; Isenberg

et al., 2001), iodoaryl-azidoforskolin (Busche

et al., 1989), the 3⬘- and 7⬘-benzophenone

ana-logues of taxol (Wu et al., 1998), and the

dauno-mycin derivative iododauno-mycin (Demmer et al.,

1997) have identified the same two major

photo-binding regions within the TMDs The sites

encompass transmembrane segments (TMS) 5

and 6 in the N-terminal half and TMS 11 and

12 in the C-terminal half Moreover, a deletion

mutant of Pgp consisting of the TMDs in the

absence of the NBDs retained the ability to

inter-act with drugs (Loo and Clarke, 1999b).

Mutational analyses have also proved useful

in attempting to pinpoint specific regions

within the TMDs involved in substrate binding

An informative approach was to generate a

chimera of Pgp (encoded by the MDR gene)

with the human MDR3 gene product

MDR3-Pgp is a phosphatidylcholine transporter with a

low affinity for multiple multidrug substrates

and is present in the canalicular membrane of

hepatocytes Pgp (MDR1) and MDR3 share

about 80% sequence identity at the amino acid

sequence level In chimeras, the replacements

limited to TMS 12 severely impaired Pgp

(MDR1)-mediated transport of actinomycin D,

vincristine and doxorubicin, but not colchicine,

suggesting the importance of TMS 12 in the

specificity to certain drugs (Zhang et al., 1995b).

Mutating residues that are not conserved

amongst the MDR1- and MDR3-Pgp isoforms

from different species provided evidence of

fur-ther involvement of TMS 12 in conferring

speci-ficity The results indicated that non-conserved

residues within the amino-terminal half of TMS

12 determine the relative rates of transport for

a variety of different substrates (Hafkemeyer

et al., 1998).

The literature is replete with investigationsthat have mutated residues throughout the pro-

tein in an attempt to localize the drug-binding

sites, and these are well summarized in a

pub-lished review (Ambudkar et al., 1999) Drug

specificity appears to be particularly sensitive

to mutations in TMS 5, 6, 11 and 12, but many

other mutations that affect substrate specificity

are scattered throughout the polypeptide Theinvestigations have usually determined theeffect of mutations on (i) the ability of Pgp toconfer drug resistance to whole cells, (ii) alter-ations in steady-state cellular accumulation ofPgp substrates, or (iii) modifications in drug-stimulated ATP hydrolysis by Pgp These stud-ies are difficult to interpret as simple changes

in drug–Pgp interaction, since altered activitymay be manifest by a number of possible fac-tors including drug binding, communicationbetween TMDs and NBDs, or conformationalchanges involved in the translocation step

Furthermore, altered drug recognition andbinding may be due to changes in protein–druginteractions within binding sites, or due to con-formational changes in binding sites related tolong-range perturbations in the global structure

of the protein Recently, however, cysteine ning mutagenesis of all the predicted TMSs ofPgp (1 through 6, and 7 through 12), combinedwith thiol modification using the thiol-reactivesubstrates dibromobimane and methanethio-sulfonateverapamil, demonstrated that residues

scan-in TMS 4, 5, 6, 10, 11 and 12 directly scan-interact

with drug molecules (Loo and Clarke, 1999b,

2000, 2001) A model was proposed wherein all

of these segments contribute to a large domainconsisting of multiple recognition elements fortransported substrates

Although the NBDs in Ppg are known tointeract with non-transported modulators thatcompete with nucleotides for binding (e.g

flavenoids) (Conseil et al., 1998), at present

there is no evidence that the NBDs in ABCtransporters play a direct role in determiningsubstrate specificity Experiments with chime-ric multidrug resistance genes argue against

such a role (Buschman and Gros, 1991) Mouse

Mdr1 confers resistance to drugs whereasmouse Mdr2 acts as a phosphatidylcholinetransporter A chimeric protein in which theNBDs of Mdr2 replaced those in Mdr1 stilltransported drugs, while the replacement ofTMDs of Mdr1 by those in Mdr2 abolisheddrug transport

In summary, the TMDs of ABC transportersclearly mediate substrate-binding events(although in bacterial uptake systems with aperiplasmic binding protein, the initial event isbinding by the PBP) in both uptake and effluxpumps, and TMSs involved in this process havebeen identified for a number of transporters

However, we still require significant effort toelucidate the precise molecular components ofdrug-binding sites

NUMBER OF SUBSTRATE-BINDING SITES

Two regions of Pgp photolabeled by drugs

may contribute to a single binding surface,

which is able to interact with different drugs

(Bruggemann et al., 1992) Alternatively, the

labeled regions may each represent separate

sites, giving two distinct drug-binding sites per

Pgp monomer (Dey et al., 1997) A number of

observations suggest that Pgp and related ABC

multidrug transporters possess two or more

drug-binding sites:

(a) Kinetic functional assays: Using steady-state

drug accumulation assays in whole cells,

com-petitive and non-comcom-petitive and cooperative

interactions have been detected between

dif-ferent transport substrates (Ayesh et al., 1996).

Non-competitive interactions demonstrate the

presence of multiple transport-competent

drug-binding sites in Pgp and cooperative effects

suggest that these sites may interact with more

than one ligand Rhodamine 123 and Hoechst

33342 (Shapiro and Ling, 1997), colchicine and

synthetic hydrophobic peptides (Sharom et al.,

1996), and colchicine and tetramethylrhosamine

(Lu et al., 2001) stimulated transport of each

other using isolated membranes or purified

Pgp preparations Shapiro and Ling suggested

that this effect in Pgp may arise from positively

cooperative interactions between two

transport-competent drug-binding sites, denoted the R

site and H site The non-competitive

interac-tions between drugs in their ability to stimulate

ATP hydrolysis also provides strong evidence

for multiple binding sites on the protein

(Orlowski et al., 1996; Pascaud et al., 1998).

Positively cooperative interactions between

transported substrates have also been observed

for other ABC transporters For bacterial LmrA,

vinblastine and Hoechst 33342 stimulated the

transport of each other, pointing to the presence

of two transport-competent drug-binding sites

in the transporter with overlapping drug

speci-ficities (van Veen et al., 2000) The stimulation of

the transport of non-conjugated drugs by

glu-tathione, and vice versa, in MRP1 and MRP2

(Evers et al., 2000; Loe et al., 2000a, 2000b) also

suggests the presence of at least two positively

cooperative transport-competent

substrate-binding sites in the proteins, one for glutathione

and the other for unconjugated drugs (Evers

et al., 2000).

(b) Conformational change assays: The quenching

by vinblastine and verapamil of the fluorescence

of intrinsic tryptophan residues and maleimidyl-anilino]naphthalene 6-sulfonic acid(MIANS)-labeled Pgp has been shown to exhibit

2-[4⬘-a biph2-[4⬘-asic profile (Liu et 2-[4⬘-al., 2000; Sh2-[4⬘-arom et 2-[4⬘-al.,

1999) This was suggested to provide evidence

of multiple sites of interaction for vinblastine.However, for the MIANS-labeled protein, thisobservation may also be explained by a differen-tial sensitivity of the two labeled NBDs toallosteric effects caused by the drug Biphasicdisplacement of [3H]-verapamil binding by vinblastine to Pgp was also observed by using

a radioligand-binding assay (Doppenschmitt

et al., 1999) Studies on the Pgp

conformation-sensitive monoclonal antibody UIC2 suggestedthat the stoichiometric binding of two vinblas-tine molecules per Pgp was required to confor-mationally increase the accessibility of the UIC2

epitope (Druley et al., 2001).

(c) Photoaffinity labeling approaches: The tor cis(Z)-flupentixol increased the affinity of

modula-iodoarylazidoprazosin for the C-terminal half

of Pgp (C site) without changing the affinity for

the N-terminal half (N site) (Dey et al., 1997)

In addition, iodoarylprasozin binding to thesesites was differentially inhibited by both vin-blastine and cyclosporin A

(d) Direct measurements of binding: Equilibrium

or kinetic radioligand-binding assays provide a

direct insight into drug–protein interaction The

increased dissociation rate of [3H]-vinblastinefrom Pgp and bacterial LmrA by various modulators provided the first solid pharmaco-logical proof for the existence of multiple drug-

binding sites on the proteins (Ferry et al., 1992, 2000; Malkhandi et al., 1994; Martin et al., 1997, 1999; van Veen et al., 1998) More recently, the

kinetic data for Pgp were combined with Schildanalyses of drug–drug interactions at equilib-rium to demonstrate that Pgp contains at leastfour distinct sites involved in drug binding

(Martin et al., 2000a) For LmrA, vinblastine

equilibrium binding experiments provide dence for the presence of two vinblastine-binding sites in the homodimeric transporter

evi-(van Veen et al., 2000).

Collectively, these data show that ABC drug transporters must contain more than onedrug-interaction site The drug-interaction sitescould represent two (or more) physically andspatially distinct binding sites or, alternatively,

multi-be present in a single flexible binding regionwithin the transporters

SUBSTRATE-BINDING SITES

We still know very little about the structural

ele-ments in multidrug transport proteins that

dic-tate drug specificity However, crystal structures

at 2.7 Å resolution, in the absence and presence

of drug, obtained for the polyspecific

transcrip-tion regulator BmrR from B subtilis provide

some insight (Zheleznova et al., 1999) One

of the universal multidrug pump ligands,

tetra-phenylphosphonium (Figure 5.3), appears to

penetrate into the hydrophobic core of BmrR,

where it forms van der Waals and stacking

inter-actions with hydrophobic and aromatic residues,

and makes an ion pair interaction with a buried

glutamic residue Glu134 (Vazquez-Laslop et al.,

1999; Zheleznova et al., 1999) The Bmr

trans-porter also binds and translocates

tetraphenyl-phosphonium (Neyfakh et al., 1991) Similar

drug–protein interactions are likely to occur in

the transcription regulator and the Bmr

porter and, by inference, other multidrug

trans-porters (Zheleznova et al., 2000) Indeed, the

presence of acidic residues within TMSs of

sec-ondary multidrug transporters MdfA and EmrE

in E coli (Edgar and Bibi, 1999; Muth and

Schuldiner, 2000; Yerushalmi and Schuldiner,

2000), QacA in Staphylococcus aureus (Paulsen

et al., 1996) and LmrP in L lactis (van Veen,

2001) was shown to be related to the cation

speci-ficity of these transporters More recently, similar

observations have been made for mammalian

ABC multidrug transporters A glutamate

residue in human MRP (Glu1089) in predicted

TMS 14 appears to be critical for the ability of the

protein to confer resistance to cationic drugs,

such as anthracyclines, whereas this residue is

not critical for its ability to transport endogenous

glutathione conjugates and glucuronides (e.g.,

leukotriene C4 and 17-␤-estradiol 17-␤-D

-glucuronide) (Zhang et al., 2001).

Similar to the role of acidic residues in determining the specificity for cationic drugs,

basic residues can play a role in the specificity

for anionic drugs For example, the specificity

of human MRP2 for glutathione gates (leukotriene C4 and 2,4-dinitrophenyl-

conju-S-glutathione) is related to the presence of a

lysine at position 325 and an arginine at tion 586, in TMS 6 and 11, respectively (Ito

posi-et al., 2001) For the breast cancer resistance

protein (BCRP), different versions of the BCRPcDNA have been obtained from cell lines thathad been selected for resistance by drug expo-sure and display distinctly different MDR phenotypes These versions of BCRP containdifferent amino acid substitutions at position

482 (arginine, glycine or threonine) ingly, BCRP protein containing an arginine atposition 482 is unable to transport (cationic)rhodamine 123, whereas BCRP with a neutralresidue at this position (glycine or threonine)does transport this substrate These data sup-port a role for charged residues in determin-

Surpris-ing the drug specificity of BCRP (Litman et al.,

2001).

Although acidic and basic residues in TMSs

of ABC multidrug transporters appear to play arole in the selectivity towards charged drugs,Pgp, LmrA and others do not possess chargedresidues in their TMDs, and yet, do transportcationic amphipathic drugs (see above) Hence,alternative mechanism(s) for cation selectivitymust exist Interestingly, it has been shown thatcations can bind to the ␲ face of the aromaticring structures of tyrosine, phenylalanine and

tryptophan residues (Dougherty, 1996) Since,

in the hydrophobic environment of the pholipid bilayer, this binding can be as strong

phos-as the electrostatic interactions between ionpairs, aromatic residues may determine cationselectivity in ABC multidrug transporters Thisnotion is supported by the observation thatsite-directed substitution of aromatic residues

in the protein does affect the specificity orpotency of drug interaction with Pgp (Kwan

et al., 2000; Ueda et al., 1997) Analysis of the

transmembrane ␣-helices in Pgp and LmrA hasrevealed that aromatic residues and polaramino acid residues with hydrogen donor side-chains are often clustered together on one side

of a helix, with amino acid residues with hydrogen-bonding side-chains on the other

non-side (Seelig and Landwojtowicz, 2000; van

Veen, 2001) Hence, the TMSs could be orientedwith their non-interactive residues facing thehydrophobic phospholipid bilayer, and theirinteractive residues facing a translocation pore Within this pore, ␲ electrons may enablecation binding and may even provide a ‘slide

guide’ system for cationic drugs, whereas

hydro-gen bonds and stacking interactions facilitate

the interaction with electroneutral moieties

within the amphipathic drug molecules (Seelig,

1998)

HOW DO SUBSTRATES ACCESS

THE BINDING SITES?

The binding sites for drugs on multidrug ABC

transporters appear to exist in two

conforma-tional states that display high- or low-affinity

binding for their specific ligand These binding

sites exist in an equilibrium that, in the case of

Pgp, may be switched between affinities by

allosteric communication from either (i)

bind-ing of drug at an alternate site (Martin et al.,

2000a), or (ii) events during the hydrolytic cycle

in the NBDs (Martin et al., 2000b; Sauna and

Ambudkar, 2001) The orientation and location

of the high- and low-affinity conformations of

drug-binding sites remain elusive to date for

any multidrug transporter The conventional

view of substrates gaining access to multidrug

ABC transporters via the aqueous phase seems

a bit naive given the highly hydrophobic nature

of many of the compounds In the case of Pgp

and LmrA (Chapter 12), several lines of

evi-dence support the proposal that drugs access

their binding sites via the lipid bilayer For

example, fluorescent compounds such as

dox-orubicin enable the membrane-localized probe

5-[125I]-iodonapthalene-1-azide to label Pgp via

direct energy transfer between the compounds

This labeling can only occur over a short range

and within the bilayer (Raviv et al., 1990).

Further proof has been obtained from

investiga-tions into the transport of the acetoxy-methyl

ester of calcein (calcein-AM) and BCECF

(BCECF-AM) in cells These compounds are

rapidly metabolized to the highly fluorescent

fluorescein derivatives by cytoplasmic esterases

However, only the parent non-fluorescent

ace-toxy-methyl esters are substrates for efflux by

Pgp and LmrA Cells expressing Pgp or LmrA

exhibit measurable fluorescence only following

inhibition of these multidrug pumps, indicating

that these proteins actively extrude the

acetoxy-methyl esters before they can reach the

cyto-plasm (i.e from within the bilayer) (Bolhuis

et al., 1996; Homolya et al., 1993) The transport

of substrates from the lipid bilayer may also be

relevant for other ABC transporters with

hydro-phobic substrates The human MDR3-encoded

Pgp transports phosphatidylcholine from the

cytoplasmic leaflet of the bile canalicular brane of hepatocytes into the bile (Ruetz and

mem-Gros, 1994; Smit et al., 1993) In addition, the

E coli ␣-hemolysin transporter HlyB appears

to bind the signal sequence of ␣-hemolysinwhen the signal sequence forms an amphiphilichelix that binds to the cytoplasmic leaflet of the

plasma membrane (Sheps et al., 1995; Zhang

et al., 1995a) (but see also Chapter 11).Whilst our understanding of binding-siteproperties is growing, we still have little knowl-edge regarding the molecular consequences ofsubstrate interaction on the local protein struc-ture of ABC transporters

The movement of molecules against their centration gradients by ABC proteins requires

con-a series of coordincon-ated events con-as outlined in

Figure 5.4 The ‘driven substrate’ (e.g drug) and

the ‘driving substrate’ (ATP) will interact withthe transporter in a mutually dependent fashion

to prevent ATP hydrolysis in the absence oftranslocation (Jencks, 1980; Krupka, 1993).Understanding how these two processes arecoupled is fundamental to elucidating themechanism by which ABC proteins translocate substrates The mechanism of translocation iscomposed of many discrete stages leading totwo major events: (i) reorientation of a substrate-binding site across the membrane and (ii) analteration in the binding site from high to lowaffinity for transported substrate Transport ofcompounds against a concentration gradientwill be driven by the Gibbs energy change of theoverall process The individual steps play vitalroles to ensure a reasonable rate of turnover that

is free of ‘bottlenecks’ (Jencks, 1980) Energyproduced by the binding and dissociation oftransported molecules, ATP and its metabolites,and the energy produced by the hydrolysis ofATP will be used to ensure adequate turnoverrates and efficient coupling (Jencks, 1980;Krupka, 1993)

Substrate binding and nucleotide hydrolysisevents are spatially distinct in ABC proteins and

therefore a series of communication pathways

will be required to ensure efficient coupling

ABC PROTEINS: COUPLED OR NOT?

Despite intensive research efforts, the

mecha-nism of coupling in ABC transporters remains

elusive An obvious and seemingly

straight-forward question is whether the NBDs require

a stimulus to hydrolyze ATP When associated

with their compatriot membrane proteins, the

NBDs of all ABC transporters are capable of

hydrolyzing ATP When isolated from their

membrane-bound domains the situation is not

as clear The NBDs of the maltose transporter

(MalK) and the histidine permease (HisP)

dis-play high levels of ATP hydrolysis (0.5–1.0␮mol

ATP min⫺1mg⫺1) when expressed separately

from the membrane domains of the whole

trans-porter (Morbach et al., 1993; Nikaido et al., 1997).

When HisP and MalK are associated with the

membrane-bound subunits of their respective

transporters, the ATPase activity is inhibited

and only reaches the high levels observed for

isolated domains when the transported

sub-strate is present (Davidson and Nikaido, 1991;

Liu and Ames, 1998) In contrast with these

bac-terial importers, isolated NBDs of the eukaryotic

proteins Pgp (Dayan et al., 1996), MRP (Kern

et al., 2000) and CFTR (Ko and Pedersen, 1995)

only display low activities of approximately0.05␮mol min⫺1mg⫺1 The activities of thesedomains within full-length human isoforms of

Pgp (Loo and Clarke, 1995; Ramachandra et al., 1998), MRP (Chang et al., 1997) and TAP (Gorbulev et al., 2001) have been reported to be

greater than 1␮mol min⫺1mg⫺1 These ities between the behavior of isolated NBDs andthat found when they are associated with theTMDs indicate a significant degree of functionalinteraction between the two types of domains

dispar-The opposing influence of TMDs on the activity

of NBDs observed between the prokaryotic andeukaryotic members suggests that distinctmechanisms of coupling may occur

However, the role of substrate binding anddissociation in regulating ATP hydrolysis atNBDs is of universal importance within theABC transporter family Does this provide anyinsight into a possible coupling mechanism?

Table 5.1 shows the degree to which severaltransported agents are able to stimulate ATPhydrolysis by their target ABC transporters

The maltose, histidine and TAP transporters,which display transport with high specificity,

N

N H

D

N H

Figure 5.4 Schematic presentation of the steps involved in the translocation of drug across a plasma

membrane by Pgp Step 1 is the initial interaction of drug with the drug-binding site (DBS) in the TMD

via the lipid bilayer Step 2 represents the signal to stimulate ATP hydrolysis in the nucleotide-binding

domain (NBD) Step 3 is the signal to initiate conformational changes in the TMD during a catalytic cycle.

Step 4 is the translocation and release of drug across the membrane.

do not display appreciable basal or intrinsic

ATPase activity in the absence of substrate

(Davidson et al., 1992; Gorbulev et al., 2001;

P.-Q Liu et al., 1999) For these ‘dedicated’

transporters it appears that the process of ATP

hydrolysis is tightly coupled to substrate

bind-ing, and therefore transport Interestingly, the

polyspecific or multidrug efflux pumps Pgp,

LmrA and MRP display a high basal ATPase

activity with only modest amounts of

stimula-tion by substrate (Callaghan et al., 1997; Chang

et al., 1997; Ramachandra et al., 1998; Shapiro

and Ling, 1994) The compounds used to

stim-ulate Pgp, LmrA and MRP ATPase activities

(see Table 5.1) vary markedly in their affinities

to bind to the proteins (50 nM to 50␮M),

yet the degree to which they stimulate ATP

hydrolysis is similar (1.5–4 fold) This lack of

correlation between maximal ATPase activity

(reflecting transport rate) and substrate

affin-ity is a characteristic feature of passive

trans-port processes Clearly, Pgp, LmrA and MRP

are active transporters and therefore this

characteristic has been used to argue for a

par-tially uncoupled transport mechanism (Krupka,

1999) An alternative explanation for the highbasal ATPase activity of Pgp and other multi-drug transporters may be that, even in theabsence of added drugs, these systems encoun-ter endogenous substrates (e.g lipids) in the biological membranes in which they are embed-ded (Ferte, 2000)

The catalytic cycle and drug-binding eventsmust be intertwined to some degree in order toproduce the vectorial transport by single sub-strate specific and multidrug ABC proteins.The subsequent sections explore how these twoprocesses interact and which events during thetransport cycle are critical to the coupling

WHAT DRIVES TRANSLOCATION:

SUBSTRATE BINDING OR

ATP HYDROLYSIS?

Investigations with the histidine transporter of

S typhimurium have provided significant insight

into the interaction between substrate bindingand ATP hydrolysis As indicated above, themembrane segments of the transporter (HisQM)regulate the intrinsic ATP hydrolytic capability

of HisP (Liu et al., 1997) and a significant

increase in ATPase activity is observed in thepresence of liganded substrate-binding protein(HisJ) How is this signal transmitted? The HisJprotein undergoes significant and, importantly,substrate-dependent conformational changes

upon binding histidine (Wolf et al., 1996) It

was therefore concluded that the conformation

of HisJ produced by ligand binding provides the driving force to stimulate ATP hydrolysisand initiate transport through HisQM How-ever, a subsequent investigation from the same laboratory demonstrated (i) no direct correlationbetween the affinities of different carbohydrates

to bind to HisJ and their translocation rates and(ii) a poor correlation between translocationrates and substrate-induced stimulation of

ATPase activity (C.E Liu et al., 1999) These

findings at first appeared difficult to reconcilewith a coupled vectorial ATP-dependent trans-location process (Jencks, 1980) A more recentpublication has demonstrated that binding of

ATP to the HisQMP2 complex, prior to

associa-tion of liganded HisJ, initiates quaternary tural changes within the complex (P.Q Liu

struc-et al., 1999) These changes in association of subunits in turn facilitate the ability of liganded

HisJ to stimulate further ATP hydrolysis,

TABLE5.1 STIMULATION OFATP

HYDROLYSIS BY SUBSTRATES FOR TRANSPORT IN PROKARYOTIC AND

Transporter and Fold stimulating agent stimulation

Human Pgp (Ramachandra et al., 1998)

Nicardipine 3.6 Vinblastine 2.0

Human TAP (Gorbulev et al., 2001)

(Davidson et al., 1992)

Liganded MalE ⬎300