CHAPTER 4 – STRUCTURE OF ABC TRANSPORTERS

CHAPTER 4 – STRUCTURE OF ABC TRANSPORTERS CHAPTER 4 – STRUCTURE OF ABC TRANSPORTERS CHAPTER 4 – STRUCTURE OF ABC TRANSPORTERS CHAPTER 4 – STRUCTURE OF ABC TRANSPORTERS CHAPTER 4 – STRUCTURE OF ABC TRANSPORTERS CHAPTER 4 – STRUCTURE OF ABC TRANSPORTERS CHAPTER 4 – STRUCTURE OF ABC TRANSPORTERS CHAPTER 4 – STRUCTURE OF ABC TRANSPORTERS

I NTRODUCTION

In order to understand the mechanism by

which ABC transporters translocate solute

across cellular membranes, structural data are

essential Such data have been hard won This

is primarily because of the difficulty inherent in

overexpressing and purifying these proteins

in an active form As for many membrane

pro-teins, ABC transporters are often toxic to the

cell or misfold when overproduced, and their

vectorial active transport function is disrupted

as they are purified Perhaps most importantly,

the activity of many ABC transporters is

influ-enced by their lipid membrane environment so

ensuring that any purified protein is fully

active, and therefore properly folded, is

non-trivial There is also increasing evidence that

the transmembrane domains (TMDs) of ABC

transporters are highly flexible, a characteristic not conducive to ready crystallization

Structural data have gradually emerged from

a variety of approaches Many bacterial ABC transporters are multi-protein complexes with each of the four core domains encoded as a sep-arate polypeptide (see introductory chapter to this volume) Several of the relatively hydrophilic nucleotide-binding domains (NBDs) have been overexpressed (including one of eukaryotic ori-gin), purified and characterized at high

resolu-tion by X-ray crystallography (Table 4.1).

Although such data tell us much about interac-tions with ATP, they have had little impact on our understanding of the mechanism of trans-port This is because binding of the translocated substrate is a property of the TMDs, and trans-port requires the interaction of all four domains

Structural data for a complete transporter came

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

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

4

CHAPTER

TABLE4.1 HIGH-RESOLUTION STRUCTURES OFNBDS

RbsA E coli Ribose uptake 2.5 Armstrong et al (1998)

HisP S typhimurium Histidine uptake 1.5 Hung et al (1998)

MalK T litoralis Maltose uptake 1.9 Diederichs et al (2000)

MJ0796 M jannaschii Unknown 2.7 Yuan et al (2001)

MJ1267 M jannaschii Amino acid transport 1.6 Karpowich et al (2001)

TAP1 H sapiens Antigen presentation 2.4 Gaudet and Wiley (2001) Rad50 P furiosus Double-strand DNA repair 2.5 Hopfner et al (2000)

SMC T maritima Double-strand DNA repair 3.1 Lowe et al (2001)

initially from single-particle analyses and

electron crystallography, and low- to

medium-resolution structures are now available for three

mammalian ABC transporters: the multidrug

resistance P-glycoprotein (Pgp) to 25 Å by

trans-mission electron microscopy (TEM) of single

particles and to approximately 10 Å by electron

cryomicroscopy (ECM) of two-dimensional

(2-D) crystals (Rosenberg et al., 1997, 2001);

MRP1 to 22 Å resolution by TEM of single

particles and 2-D crystals (Rosenberg et al.,

2001); and TAP to approximately 35 Å

resolu-tion (Velarde et al., 2001) The structure of YvcC

from Bacillus subtilis has also been resolved to

25 Å by ECM The only X-ray crystallographic

data for any complete ABC transporter came

from a tour de force approach for the lipid A

transporter (MsbA) from Escherichia coli (Chang

and Roth, 2001;Chapter 7) In that study, over

twenty E coli transporters were expressed and

then tested under 96 000 crystallization

condi-tions to yield an example that crystallized and

diffracted (to a resolution of 4.5 Å)

Although not the first NBD structure to be

obtained (Stauffacher and colleagues solved

the structure of the N-terminal NBD of E coli

RbsA earlier; Armstrong et al., 1998), the HisP

structure (Hung et al., 1998) was the first

pub-lished description of an NBD at atomic

resolu-tion (1.5 Å) HisP, the NBD of the Salmonella

histidine uptake system, is a single polypeptide

domain of which two copies associate with the

TMDs (HisQ and HisM) in the intact transporter

(Kerppolla et al., 1991) The structure of HisP is

shown in Figure 4.1A The most convenient

description is that it comprises two ‘arms’

ori-ented approximately perpendicular to one

another Arm-I contains an ABC-specific ␤-sheet

subdomain (Karpowich et al., 2001), with the

Walker A motif in a typical phosphate-binding

loop conformation together with the Walker B

motif (Walker et al., 1982) The perpendicular

Arm-II, an ␣-helical subdomain, contains the

‘ABC signature’ motif The two other conserved

motifs of ABC transporter NBDs, namely the

Q-loop (Diederichs et al., 2000) and the H-loop

(Linton and Higgins, 1998), are located at the

interface of the two arms

HisP was crystallized in the presence of ATP

(Hung et al., 1998) The environment of the

ATP molecule is depicted in Figure 4.1B A

number of side-chains interact with nucleotide,

stabilizing its binding Among these are the expected interactions with the consecutive acidic amino acids at the C-terminus of the Walker B motif (D178 and E179 in HisP), and with the polar side-chains of the Walker A motif (K45, S46 and T47) The conserved polar amino acids of the Q-loop and the H-loop also make contact with ATP through water molecules bound within the ATP pocket One additional interaction is noteworthy: the adenine ring

Arm-II

Arm-I A

B

S46

H211

E179 Q100 D178

K45

Figure 4.1 The structure of HisP and the environment of nucleotide within the ATP-binding pocket A, the two perpendicular arms of HisP are displayed in cartoon format with ribbons denoting

␣-helices and arrows representing ␤-sheets The bound ATP molecule is displayed in ball-and-stick format The colours represent: yellow, Walker A motif; red, Walker B motif; blue, H-loop; magenta, Q-loop; green, ABC signature B, The ATP molecule and side-chains of residues with which it interacts are displayed in ball-and-stick format Amino acid positions are indicated To ensure clarity, the backbone of the Walker A motif has been removed These and other structural diagrams were produced using the program Molscript (Kraulis, 1991) Carbon atoms are in gray (darker in the ATP molecule), oxygen atoms are red, nitrogen blue and phosphorus pink Reproduced with permission from Kerr (2002).

stacks against the side-chain of a well-conserved

aromatic residue (Y16 in HisP) However, there

are relatively few contacts on one side of the

ATP molecule, and compared to other ATPases,

the nucleotide appears somewhat exposed

(Dreusicke et al., 1988); (Figure 4.1).

CONFORMATIONAL CHANGES IN

THENBDS

There are now structural data for NBDs in

dif-ferent conformational states (i.e

nucleotide-free, or complexed with ADP or ATP) The most

pertinent comparisons are made between

dif-ferent conformations of the same NBD

Other-wise, to assess the conformational changes

invoked by ATP binding, or by ATP hydrolysis

followed by loss of Pi, NBDs from different

ABC transporters must be compared This

lim-its any analysis to regions conserved between

NBDs

Conformational change associated with

ATP binding

Rad50 provides the best model for comparison

of nucleotide-free and ATP-bound

conforma-tions of NBDs This bacterial protein is involved

in the repair of double-strand breaks in DNA

and thus is an unusual paradigm for the

inter-action of NBDs in ABC transporters However,

Rad50 contains sequences that unequivocally

identify an ABC-transporter-like NBD, and the

structure displays the characteristic L-shaped

domain (Hopfner et al., 2000) Analysis of the

ATP-bound and ATP-free forms of Rad50

indi-cates that there is a pronounced ordering of the

Walker A, Walker B and Q-loop motifs upon

interaction with nucleotide This suggests an

induced fit of ATP This conclusion must be

tempered with some caution as it is possible

that the ATP-dependent dimerization of Rad50

might be responsible for this effect The

impli-cations of this dimer are discussed below

However, analysis of the bound and

ADP-free forms of another NBD protein (MJ1267)

supports the hypothesis of induced nucleotide

fit (see below) A second change observed in the

bound form as compared with the

ATP-free form of RAD50 is rotation of the ␣-helical

subdomain relative to the ␤-subdomain Of

course, it is difficult to assess whether rotation

of the ␣-helical subdomain is a consequence

of ATP binding or NBD:NBD dimerization

However, evidence from other NBDs suggests

that within these domains there is a degree

of conformational flexibility between the sub-domain motifs

Conformational change associated with ATP hydrolysis and release of Pi

The HisP structure (complexed with ATP) and the MJ1267 structure (complexed with Mg.ADP) have enabled comparison of NBDs

in a pre-hydrolytic and post-hydrolytic state

(Karpowich et al., 2001) These changes are

illus-trated in Figure 4.2 The ABC-specific

subdo-main (Arm-II) undergoes a 12–15° inward rotation, bringing it closer to Arm-I in the ATP-bound structure compared with the ADP-ATP-bound structure In another allosteric membrane protein (the nicotinic acetylcholine receptor)

a relatively small rotation at the presumed ligand-binding sites results in a substantial rotational movement in the channel-lining region (Unwin, 1993, 1995) Thus, although this 12–15 Å rotation may appear slight, its effects

on the TMDs could be highly significant The rotation observed in MJ1267 appears to be cen-tred on the Walker B motif, which is located close to the hinge between the two ‘arms’ The most dramatic change in orientation of residues involves the conserved glutamine of the Q-loop

(Karpowich et al., 2001) In the ATP-bound

con-formation of HisP this amino acid is located within 5 Å of the ␤-phosphate of ATP and inter-acts with it through a bound water molecule

In stark contrast, the ADP-bound conforma-tion sees this amino acid withdrawn from the nucleotide such that its closest approach is now⬎12 Å from the ␤-phosphate (Karpowich

P P

P

HisP:ATP MJ1267:ADP

Figure 4.2 The conformational changes associated with ATP hydrolysis Arm-II of the NBD undergoes a 12–15° outward rotation following hydrolysis of ATP and loss of phosphate.

The Q-loop (shown in black) is moved 7 Å further from the bound nucleotide such that it can no longer interact with the ␤-phosphate Reproduced with permission from Kerr (2002).

et al., 2001) Thus, the Q-loop is a strong

candi-date for transmitting the conformational

change associated with ATP hydrolysis to other

domains of an ABC transporter In addition,

the ABC signature motif (which is located in

Arm-II) is displaced by about 6 Å as a result of

rotation of the ␣-helical subdomain and thus

may also be involved in transmission of

con-formational change (Karpowich et al., 2001).

Similar hinge-motions have been hypothesized

in analyses of the structures of HisP (Hung

et al., 1998), TAP1 (Gaudet and Wiley, 2001)

and another Methanococcus NBD (MJ0796; Yuan

et al., 2001) Whether these changes are the

result of ATP hydrolysis or the release of

phos-phate from the post-hydrolytic intermediate

remains unanswered Pharmacological studies

indicate that conformational changes in the

drug-binding site (TMDs) of Pgp occur upon

phosphate release, rather than upon ATP

hydrolysis (Martin et al., 2001; Rosenberg et al.,

2001) The determination of NBD structures

complexed with ADP.vanadate may confirm at

which step the changes in NBD conformation

occur

Conformational change associated with

release of ADP

Conformational differences between the

ADP-bound and nucleotide-free structures of the

NBD of the branched-chain amino acid

trans-porter of Methanococcus jannaschii (MJ1267;

Karpowich et al., 2001) may be seen as a model

for the release of ADP from a post-hydrolytic

NBD Studies of other ATPases suggest that

the release of dinucleotide may often be

accom-panied by substantial conformational change

(Scheirlinckx et al., 2001; Zhou and Adams,

1997) The two predominant changes are (i) a

general destabilization of the NBD in the

absence of nucleotide and (ii) alteration in the

conformation of the H-loop The

destabiliza-tion is reflected in higher crystallographic

B-factors in several regions that provide

inter-actions with the nucleotide The B-factor is a

measure of the degree of flexibility in a region

of a structure Thus, the Walker A, Walker B,

H-loop and the ␤-strand containing the adenine

ring-interacting aromatic residue (Tyr-16 in

HisP) all exhibit higher B-factors in the absence

of nucleotide, suggesting that release of ADP

from the post-hydrolytic NBD is accompanied

by a relaxation of the domain Put another way,

it suggests an induced fit of nucleotide with

NBD (Karpowich et al., 2001) The second

structural effect is the change in conformation

of the backbone of the conserved H-loop, which displaces the side-chains of the H-loop

by as much as 12 Å This suggests that the H-loop may be involved in transmitting post-hydrolytic conformational changes in an intact transporter No other changes of comparable magnitude are observed elsewhere (except in loops that are not conserved across the ABC transporter family)

As previously stated, caution must be applied when comparing structural data on NBDs from different ABC transporters as crystal-packing forces may contribute to the conformational changes described However, it is particularly interesting that molecular dynamics simula-tions of NBDs in the presence or absence

of nucleotide demonstrate both nucleotide-dependent rotation of the ␣-helical subdomain and withdrawal of the Q-loop (Campbell and Sansom, personal communication)

INTERACTION BETWEENNBDS

The structural data described above have been obtained for isolated, monomeric NBDs (in the case of RbsA, which contains two NBDs in a sin-gle polypeptide, only the first 259 amino acids corresponding to the N-terminal NBD were

crystallized; Armstrong et al., 1998) Clearly,

our understanding of the function and dynam-ics of ABC transporters would be greatly enhanced by a description of the structural and conformational interactions between domains Both NBDs in an ABC transporter are required

for function (Azzaria et al., 1989; Gill et al.,

1992) In the alternating catalytic cycle model only one ATP molecule is hydrolyzed at a time, with the ATPase activity alternating between the

two NBDs (Hrycyna et al., 1999; Senior and

Gadsby, 1997) Although we should not over-look the possibility that the two NBDs influence each other indirectly through their cognate TMDs, the simplest explanation is that they interact directly with each other In this respect,

it is interesting that several of the monomeric NBDs have formed a crystallographic dimer (Kerr, 2002) These associations can be consid-ered as hypothetical models for the interaction

of NBDs in an intact ABC transporter

Four alternative models have been presented for NBD:NBD association and are represented

schematically in Figure 4.3 HisP forms a

back-to-back crystallographic dimer in which

the domains interact with each other through

the exposed ␤-strands that constitute the

ABC-specific ␤-sheet subdomain in Arm-I

(Figure 4.3A) The perpendicular Arm-II and the

ABC signature motif are proposed to interact

with the TMDs, HisQ and HisM (Hung et al.,

1998) The big drawback of this interface is the

very small surface area buried by dimer

forma-tion (about 1000 Å2) Detailed comparison of

dimer interfaces in proteins suggests such a

small buried surface area may be the result of

crystal-packing forces, rather than a

physiologi-cally relevant dimer (Kerr, 2002) The NBDs of

the thermophilic maltose transporter (MalK)

interlock in the crystal structure as shown in

Figure 4.3B, with close contacts between the

hinge regions of the two NBDs In particular, the

Q-loops are in close contact across the dimer

interface (4 Å), consistent with a role in

trans-mission of inter-domain conformational change

In this dimer, the TMDs would be in close

appo-sition to the ABC signature motifs (Diederichs

et al., 2000) For Rad50, the crystallographic

dimer (Figure 4.3C) shows the two monomers

interact in a head-to-tail fashion Interestingly,

nucleotide binds at the NBD:NBD interface in

Rad50 and is coordinated by interactions with

the Walker A and B motifs of one NBD, and

the ABC signature motif of the other NBD

This orientation provides extra stability for the

nucleotide and provides a possible explana-tion for domain:domain interacexplana-tion upon ATP hydrolysis through the ABC signature motifs

(Hopfner et al., 2000) A further potential

dimer-interface model is derived from the structure of

ArsA (Figure 4.3D), the ATP-hydrolytic domain

of the bacterial arsenic transporter (Zhou et al.,

2000) Although ArsA is not a member of the ABC transporter family (as it does not possess the characteristic ABC signature motif), it may be informative since the ArsA dimer is formed from

a single polypeptide with two ATP-binding sites

Thus, the association of the two domains cannot

be an artifact of crystallization conditions In this structure the two ATP-binding pockets are con-siderably closer together (about 15 Å) than in the

other three models (Zhou et al., 2000).

In attempting to assess the validity of the four dimer interface models, a number of considera-tions are necessary First, what biochemical evi-dence supports the association state? Second, are there theoretical considerations that may have an impact? Third, does the interface between the NBD and the TMD, as suggested crystallographically for MsbA (Chang and Roth, 2001), rule out any of the proposed models?

The biochemical data are diverse and consist of

attempts to measure direct interactions (e.g by

crosslinking of cysteine residues), as well as

indirect interactions (e.g by examining the effects

A

C

B

D S

S

S

P

P

P P

S

S

S

Figure 4.3 Models for NBD association In each case ‘P’ refers to the location of the phosphate-binding

loop (Walker A motif ), while ‘S’ represents the signature motif (absent from ArsA) A, HisP; B, MalK;

C, Rad50; D, ArsA Reproduced with permission from Kerr (2002).

that mutations in NBDs have on the transport

complex) Data for MalK suggest that mutation

of residues proximal to the Q-loop to cysteine

can result in dimerization of MalK, consistent

with the close apposition of these loops in the

MalK dimer model (Hunke et al., 2000) Cysteine

crosslinking data for Pgp suggest that the two

Walker A motifs may be as close as 15–20 Å apart

(Loo and Clarke, 2000; Urbatsch et al., 2001) This

appears to be consistent only with the ArsA

structure, as these motifs are more than 25 Å

apart in the Rad50, HisP and MalK dimer

mod-els However, fluorescence resonance

experi-ments on Pgp suggest that the distance between

Walker A motifs might be 30–35 Å, which would

be satisfied by models other than ArsA (Qu and

Sharom, 2001)

A considerable amount of data has been obtained for bacterial ABC transporters that is

rather more indirect, in that it pertains more

to the interaction of NBDs with the TMDs

However, the nature of this interaction could

clearly provide a considerable constraint on

potential NBD:NBD interactions For the

histi-dine and maltose transporters, the interaction

of the TMDs with the NBDs has been assessed

by mutagenesis and co-purification studies

(Liu et al., 1999; Petronilli and Ames, 1991).

Mutations on the face of HisP that is proposed

to interact with the TMDs (i.e the upper surface

of Arm-II in Figure 4.1A) diminished

co-purification with the TMDs, suggesting that

the association of the domains is disrupted

Similarly, for the maltose transporter, mutations

in the conserved region linking TM ␣-helices

4 and 5 of MalF and MalG (containing the

so-called ‘EAA’ motif) that disrupted function

could be rescued by ‘suppressor’ mutations in

the ␣-helical ABC-specific domain (Arm-II;

Mourez et al., 1997) Some residues within the

EAA motif could also be crosslinked to three

residues in the same ␣-helix of the ␣-helical

sub-domain of MalK (Hunke et al., 2000) Again this

helix is on the upper face of Arm-II as viewed in

Figure 4.1A The main caveat is that the results

can also be explained by allostery, i.e that the

mutations leading to a loosening of the HisQMP2

or MalGFK2 complexes are not necessarily at

the domain:domain interfaces but are

down-stream effects of these mutations

The interaction of nucleotide with NBD pro-teins has been cited as support for the Rad50

dimer model (Hopfner et al., 2000; Karpowich

et al., 2001) Rad50 in the absence of nucleotide

is a monomer in solution and in the crystal

However, ATP (or indeed the non-hydrolyzable

analogue AMP-PNP) leads to dimerization of Rad50 in solution and in the crystal Mutations

in the ABC signature motif disrupt ATP-dependent dimerization in Rad50 (Hopfner

et al., 2000) However, several studies have

indicated that ATP does not promote the dimerization of other NBDs (see Kerr, 2002) Furthermore, SMC, another DNA-interacting protein with an ABC-transporter-like NBD, does not show the interlocking-L arrangement that characterizes Rad50 Instead, SMC adopts a hexameric association in the crystal irrespective

of the presence of nucleotide (Lowe et al., 2001).

Recently, a structure has been obtained at 4.5 Å resolution for the entire prokaryotic lipid A transporter MsbA (Chang and Roth, 2001) This protein is a ‘half transporter’, containing a single NBD and a single TMD within the same polypeptide Although a considerable propor-tion of the NBD is not resolved in the crystal structure, the structural data do illustrate the interface between the TMD and the NBD, and so constrain putative NBD dimer interfaces (see above) The most noticeable feature of the MsbA

structure (Figure 4.4; a complete description of

the structure is given in Chapter 7of this vol-ume) is the recognition of intracellular subdo-mains (ICDs) linking the TMD and NBD In the

orientation shown (Figure 4.4) the sequences in

the NBD that interact with the ICDs are clearly visible (purple) Three conserved motifs of the NBDs which pack against the ICDs are the Walker B ␤-strand, the Q-loop and the first

␣-helix of the ␣-helical subdomain All three can

be mapped onto the lower surface of Arm-II in

HisP (Figure 4.1A) The implication of this

admittedly incomplete structure is that the MalK dimer interface (in which the Q-loops are

in very close apposition; Diederichs et al., 2000)

is not a feasible model for an NBD:NBD dimer interface in intact ABC transporters How the crosslinking data for MalK (see above), which

suggest that the upper face of the ␣-helical

sub-domain is in contact with the TMDs, can be rec-onciled is also unclear

In conclusion, the current data seem to favour the orientation of the two ATP-hydrolytic domains of Rad50 (and possibly ArsA) as a model for NBD:NBD interactions within an intact ABC transporter It is worth pointing out that the published MsbA structure consists of two molecules tilted together at the extracellu-lar face and splayed apart at the intracelluextracellu-lar side, separating their NBDs such that they do not share an interface (Chang and Roth, 2001) Thus, perhaps the NBDs do not interact or

interact only transiently during the transport

cycle Progress towards a higher-resolution

structure of mammalian ABC transporters will

be required to resolve this issue

Pgp is the best-characterized mammalian ABC

transporter and this section is focused on that

protein Data obtained for other mammalian

ABC transporters are broadly consistent and,

where appropriate, are compared and

con-trasted The MsbA structure is described in

detail elsewhere in this volume (Chapter 7): a

comparison with Pgp illustrates unresolved

questions

Structure determination requires not only pure protein, but a reasonable degree of

confi-dence that the purified protein has retained its

native fold This poses a significant problem

when working with large molecular weight membrane proteins Pgp activity is dependent

on the lipid environment of the membrane

(Callaghan et al., 1997), while the first step in

purification requires disruption of the lipid bilayer and solubilization of the protein compo-nents using detergent Not surprisingly, this destroys measurable activity of Pgp To demon-strate that the solubilized and purified protein has retained, or can regain, the native protein fold, the detergent must be replaced by lipid

so that activity can, once again, be measured

The choice of detergent is therefore crucial to the success of the purification process The deter-gent must solubilize the membrane protein without irreversible denaturation and it must be possible to replace the solubilizing detergent with lipids The non-ionic detergent,

dodecyl-␤-D-maltoside has the requisite characteristics and has proved invaluable for solubilization of Pgp from multidrug resistant Chinese hamster

ovary cells (Callaghan et al., 1997).

ALOW-RESOLUTION STRUCTURE FORPGP

A low-resolution (25 Å) structure of active Pgp (shown by drug-binding and drug-stimulated ATPase activity) was determined by TEM of single particles In this technique, multiple images of single particles with a similar orienta-tion were aligned and averaged to produce an image of higher signal:noise ratio (Rosenberg

et al., 1997) Single particle analysis was initially

carried out on reconstituted Pgp because, com-pared with solubilized protein, the lipid bilayer

confines all the particles in the z axis (i.e in the

plane of the membrane) Thus, the molecules

only exhibit rotational freedom in the x, y

dimensions, limiting the potential orientations that they can adopt Furthermore, when recon-stituted into a lipid environment, Pgp was shown to exhibit drug binding and ATPase activity similar to that of the protein in its native membranes The reconstituted protein was examined under negative-stain, such that only the stain-accessible surface of the molecule was observed More than 70% of the protein recon-stituted into the lipid bilayer adopted a single

orientation TEM of this material (Figure 4.5A)

projected an electron-dense ring of protein

12 nm in diameter with both twofold and six-fold symmetry surrounding a central chamber

of approximately 5 nm diameter The twofold symmetry in this structure was consistent with

Figure 4.4 Domain:domain interactions in MsbA.

The structure of a single MsbA molecule is

displayed in a cartoon format The ␣-helices of

the TMDs are displayed in blue, while their

intracellular extensions which comprise the ICD

are colored green The NBD is predominantly

colored yellow, but the three regions which interact

with the ICDs are colored purple These comprise

the Q-loop and Walker B motif and the first

␣-helix of Arm-II.

the presence of two homologous TMDs and the

sixfold symmetry is consistent with six pairs of

transmembrane ␣-helices, each pair linked by

an extracellular loop

To obtain images of different surfaces, the particles were examined as detergent

solu-bilized material This provided several different

views of the particles (Figure 4.5B, C and D).

One projection (Figure 4.5B) closely resembled

that of the reconstituted protein (Figure 4.5A).

Preferential labeling of this surface of

solubi-lized Pgp particles by lectin-gold shows it to be

glycosylated and therefore consistent with a

view of the extracellular surface of the TMDs

of Pgp Because the chamber accumulated the

uranyl acetate stain it is likely to be aqueous

and thus open to the extracellular milieu

Futhermore, at least for this surface, the protein

has a similar fold in the lipid bilayer and when

solubilized in detergent A second projection

(Figure 4.5C)was also circular and 10–12 nm in

diameter However, this surface, with no central

chamber and two 3 nm diameter lobes, was

dis-tinct from the extracellular view These lobes

are an appropriate size for the 200 amino acid

NBDs The third projection (Figure 4.5D) was

asymmetric in shape, with three small lobes on

one half of the particle and two larger lobes on

the other half This projection probably

repre-sents a side view, in which the two larger lobes

correspond to the two NBDs and the three

small lobes correspond to the TMDs (the

hexagonal symmetry of the TMDs when

viewed from above would be expected to

proj-ect three elproj-ectron-dense lobes when viewed

from the side)

Thus, the shape of the Pgp particle approxi-mates a short and fat cylinder, 10 nm in diame-ter and about 8 nm high The lipid bilayer is about 4 nm in thickness, suggesting that about one-half of the molecule resides within the membrane The TMDs form a chamber in the membrane, open at the extracellular face The chamber is closed at the cytoplasmic face

of the membrane and the two 3 nm lobes prob-ably correspond to the NBDs

MEDIUM-RESOLUTION STRUCTURE OF

P-GLYCOPROTEIN

Single particle analysis and negative stain limit the resolution of the data To obtain higher reso-lution data 2-D crystals of Pgp were obtained and imaged using low-dose electron cryomi-croscopy (ECM) Precipitant-induced, large, well-ordered 2-D crystals of Pgp can be grown reproducibly at the air/water interface of a droplet Projection images of frozen-hydrated, 2-D crystals displayed reflections to

approxi-mately 8 Å resolution (Rosenberg et al., 2001).

Importantly, because crystallized protein cannot

be analyzed for function, the unit cell of the crys-tals was very similar to the size and shape of the single particles, making it likely that the native protein fold had been preserved Tantalizingly, the resolution of the processed image remains around 10 Å, just outside that required for recognition of secondary structural features ECM of the holoenzyme, unlike negative stain, produced a 2-D projection of all electron

densities in the 3-D protein (Figure 4.6) The

Figure 4.5 Projection maps of Pgp Images derived from single particle alignment and averaging of

negatively stained Pgp particles, using contours and shading to delineate stain (black) and protein (white) boundaries A, averaged projection map of Pgp reconstituted into proteoliposomes B, C and D, averaged projection maps of solubilized Pgp particles Three classes of particle were observed, which differ in their orientations with respect to the electron beam B, face-on projection of the extracellular face of Pgp

C, face-on projection of the cytoplasmic face of Pgp D, side-on view of Pgp; the two NBDs are indicated Reproduced with permission from Rosenberg et al (1997).

projection structure approximates to an

ellipti-cal ring of 91 Å⫻ 60 Å with a slightly

asym-metric, central low-density region entirely

consistent with the surface views obtained by

analysis of single particles in negative stain As

expected, the protein has distinct

pseudo-twofold symmetry with several pairs of clearly

related density peaks (Figure 4.6).

The consistency with the low-resolution structure is more easily recognized in a 3-D

map of the protein, generated from negatively

stained 2-D crystals by imaging the crystal

lat-tice from different angles Although at lower

resolution, this analysis provided information

about the spatial organization of the four

domains Sections through the 3-D map at

approximately 10 Å intervals through the plane

of the membrane are shown in Figure 4.7.

Working up from what is thought to represent

the intracellular surface of the transporter, two

large densities (see filled arrows in Figure 4.7A)

are related by twofold pseudosymmetry These domains presumably reflect the NBDs and are

of an appropriate size to each accommodate

HisP (see open arrow, Figure 4.7A) In the next section through the transporter (Figure 4.7B)

the NBDs are still apparent (indicated by arrows) but there are now two extra electron densities either from a second lobe of each of the NBDs, or from intracellular loops of the TMDs (or, possibly, a combination of both) The four lobes now surround a distinct centre of low electron density indicating that the central chamber of the transporter extends deep into the plane of the membrane At this resolution it

is not possible to ascertain whether the specific residues which block the chamber come from the cytoplasmic loops of the TMDs or from the NBDs (although no evidence for such a role for

the NBDs could be found by in vivo labeling studies; Blott et al., 1999) Towards the midpoint

of the membrane, two arcuate domains form almost a complete ring of protein around the central chamber Each domain has three higher electron densities, which presumably corre-spond to pairwise clustering of the six putative transmembrane ␣-helices of each TMD (labeled

1–3 and 4–6 in Figure 4.7D) There are

notice-able ‘gaps’ between the two TMDs within the plane of the membrane, potentially permitting side-access to the chamber from the lipid phase

(arrows in Figure 4.7D) At the extracellular surface of the transporter (Figure 4.7E) the

elec-tron densities probably include contributions from both the TM ␣-helices and the extracellu-lar loops and a pronounced gap in the protein ring is evident

Comparison of Pgp with low-resolution structures of MRP1 and TAP

Structures for MRP1 and TAP have been

resolved to 22 Å resolution (Rosenberg et al., 2001) and around 35 Å (Velarde et al., 2001),

respectively MRP1, TAP and Pgp are members

of different subfamilies of ABC transporters

MRP1 has an extra TMD (TMD0) in addition

to the four core domains, and TAP is a het-erodimer of two ‘half ABC transporters’, TAP1 and TAP2 The primary sequence of the TMDs

of TAP is particularly different from those of Pgp and MRP1, and probably reflects the spe-cialized nature of TAP for antigen presentation

to MHC class I molecules Despite these differ-ences, single particle images of MRP1 and TAP

a

10 Å b

Figure 4.6 Projection map of Pgp determined by

electron cryomicroscopy at 10 Å resolution Solid

lines indicate density above the mean Twelve

major densities (A–F with their pseudosymmetric

densities A ⴕ–Fⴕ) are related by a pseudo-twofold

symmetry axis centered at the star A region of low

protein density corresponds to an aqueous chamber

within the membrane The areas circled in blue at

opposite ends of the molecule probably include

densities corresponding to the NBDs; the

three-dimensional reconstruction shows they are at

the cytoplasmic face of the membrane Reproduced

with permission from Rosenberg et al (2001).

are similar in size and shape to those for Pgp.

Each structure consists of a ring of protein

sur-rounding a large central hydrophilic chamber,

which is open at one side of the membrane and

closed at the other with two large protein lobes

In the MRP1 and TAP structures it is not

possi-ble to say whether the ‘open’ side is equivalent

to the intra- or extracellular face of the

mem-brane, although it is expected to be extracellular

by virtue of comparison with Pgp and because

the two large lobes are probably the

intracellu-lar NBDs Interestingly, analysis of TAP2 by

itself suggests that the TMD of TAP2 can form

an arcuate structure independently of TAP1,

consistent with the interpretation that each

TMD contributes half of the chamber in the

membrane (as also seen for the MsbA X-ray

structure; see below) In single particles of

MRP1, a particularly large protein density at

the outer side of the ring may represent the additional TMD0 Although MRP1 crystallizes

as a dimer, at current resolution it is not possible

to demonstrate that the crystallization dimer is

‘double barreled’ with a separate chamber per molecule, although this would seem likely given the structure of single particles of MRP1

Comparison of Pgp with the structures of MsbA and YvcC

MsbA is a ‘half transporter’ with one NBD and one TMD It is expected to function as a homo-dimer to transport lipid A across the inner

mem-brane of E coli The higher-resolution structure

(4.5 Å) for MsbA represents an important advance (Chang and Roth, 2001;Chapter 7) For the first time it is definitively demonstrated that

F E

D

1 2

3 4

5

*

6

k

Outside

Inside

Lipid bilayer

Chamber

TM TM

A B C D E h

Figure 4.7 3-D map of Pgp Panels A to E represent slices in the plane of the crystal (x, y) (parallel to the plane of the membrane), with each slice having an approximate thickness of 10 Å and progressively moving from the intracellular (A) to the extracellular (E) side of the membrane The scale bar ⴝ 28 nm In the inset of panel A, two HisP monomers (Hung et al., 1998) to the same scale are modeled onto the structure as ribbon diagrams Panel F shows two sketched views of Pgp with the TMDs in red and the NBDs in blue The upper panel shows a cross-sectional sketch of the Pgp molecule indicating the approximate plane of sections The lower panel shows a view of Pgp in the plane of the membrane as viewed from the extracellular face Reproduced with permission from Rosenberg et al (2001).