8/13/2019 2857.full

1/6

Three-dimensional Structure of P-glycoproteinTHE TRANSMEMBRANE REGIONS ADOPT AN ASYMMETRIC CONFIGURATION INTHE NUCLEOTIDE-BOUND STATE*

Received for publication, September 8, 2004, and in revised form, October 12, 200Published, JBC Papers in Press, October 13, 2004, DOI 10.1074/jbc.M410296200

Mark F. Rosenberg, Richard Callaghan, Szabolcs Modok, Christopher F. Higgins ,and Robert C. Ford From the Department of Biomolecular Sciences, University of Manchester Institute of Science and Technology, P. O. Box 88, Manchester M60 1QD, United Kingdom, Department of Clinical Laboratory Sciences, University of Oxford, John Radcliffe Hospital, Oxford OX3 9DU, United Kingdom, and MRC Clinical Sciences Centre, Faculty of Medicine, Imperial College, Hammersmith Hospital Campus, Du Cane Road, London, W12 0NN, United Kingdom

Multidrug resistance of cancer cells and pathogens isa serious clinical problem. A major factor contributing to drug resistance in cancer is the over-expression of P-glycoprotein, a plasma membrane ATP-binding cas-sette (ABC) drug efflux pump. Three-dimensional struc-tural data with a resolution limit of 8 have been

obtained from two-dimensional crystals of P-glycopro-tein trapped in the nucleotide-bound state. Each of thetwo transmembrane domains of P-glycoprotein consistsof six long -helical segments. Five of the -helices fromeach transmembrane domain are related by a pseudo-2-fold symmetry, whereas the sixth breaks the symmetry.The two -helices positioned closest to the (pseudo-)symmetry axis at the center of the molecule appear to bekinked. A large loop of density at the extracellular sur-face of the transporter is likely to correspond to theglycosylated first extracellular loop, whereas two glob-ular densities at the cytoplasmic side correspond to thehydrophilic, nucleotide-binding domains. This is thefirst three-dimensional structure for an intact eukary-otic ABCtransporter.Comparison with the structures of

two prokaryotic ABC transporters suggests significantdifferences in the packing of the transmembrane -hel-ices within this protein family.

The ATP-binding cassette (ABC)1 family of membrane pro-tein transporters is associated with many genetic disorders aswell as the resistance of pathogenic bacteria to antibiotics andcancer cells to chemotherapy. The minimal functional unit of an ABC transporter typically consists of four domains, twohydrophilic nucleotide-binding domains (NBDs) and two trans-membrane domains (TMDs), each consisting of several (fre-quently six) putative membrane-spanning -helices. X-raycrystallography-derived structures of two bacterial ABC trans-porters (MsbA and BtuCD) have recently been published (13).The structure of the hydrophilic energy-transducing NBD

dimer of BtuCD is similar to the structures determined forisolated NBDs (see Ref. 4 for a review and Refs. 59 for exam-ples). No significant structural homology between the TMDs of the two transporters was observed. Indeed BtuCD has 20 mem-brane-spanning -helices, whereas MsbA has 12.

Eukaryotic ABC transporters have been even more refrac-

tory to structural analysis than bacterial ABC transporters.Electron microscopy coupled with image analysis has revealedlow resolution data for single particles (10) and two-dimen-sional crystals (11, 12, 15) of P-glycoprotein (P-gp). These datademonstrated large scale conformational changes upon nucle-otide binding (11, 12), reflecting coupling between ATP hydrol-ysis and drug binding. Low resolution structures from singleparticles of a peptide transporter, TAP (13), a yeast multidrugtransporter, Pdr5p (14), and another mammalian multidrugtransporter, MRP-1 (16), have also been reported. Most re-cently, the low resolution three-dimensional structure of thecystic fibrosis transmembrane conductance regulator protein(CFTR) was reported (17) and was demonstrated to show astrong structural homology to P-gp including conformational

plasticity. In this study we present the three-dimensionalstructure for P-gp at 8 resolution obtained by cryo-electroncrystallography of two-dimensional crystals. These structuraldata are the highest resolution for any eukaryotic ABC trans-porter and the first to show the location and packing of thetransmembrane -helices. This structure was obtained in thepresence of bound nucleotide, enabling a new conformationalstate of an ABC transporter to be assessed.

MATERIALS AND METHODSChinese hamster P-gp was purified using dodecyl maltoside and

two-dimensional crystals produced by vapor diffusion at the air/waterinterface of hanging droplets containing protein, buffer, precipitant,and a non-hydrolysable ATP analogue (AMP-PNP) as described previ-ously (11, 12, 18, 19). After 17 h, the surface of the droplet was probedwith a carbon-coated electron-microscope grid. Grids were embedded in2% (w/v) glucose and transferred at liquid-nitrogen temperatures into aGatan cryotransfer specimen holder and observed at 100 K in aPhilips CM200 FEG electron microscope operated at 200 kV accelera-tion voltage. Images were recorded with a 1- or 2-s exposure time underlow dose conditions at a magnification of 38,000 on Kodak SO-163 filmand developed for 12 min with full strength Kodak D19 developer.Micrographs were screened by optical diffraction. Areas of 20482048pixels were digitized using a Zeiss SCAI scanner with a 7-m step sizeor a UMAX Powerlook 3000 scanner with an 8-m step size. Imageswere processed using Medical Research Council image-processing pro-grams (20). After three cycles of lattice unbending and then correctionfor contrast transfer function, amplitude and phase data were merged(see Table I). Images were recorded over a broad tilt range and over alarge range of defocus values ( 200 nm to 1500 nm). Projection mapsand three-dimensional density maps were generated using the CCP4

* This work was supported by Cancer Research UK, a EuropeanMolecular Biology Organization short-term fellowship (to M. F. R.), andthe Medical Research Council, UK. The costs of publication of thisarticle were defrayed in part by the payment of page charges. Thisarticle must therefore be hereby marked advertisement in accordancewith 18 U.S.C. Section 1734 solely to indicate this fact.

To whom correspondence should be addressed: Dept. of Biomolecu-lar Sciences, University of Manchester Institute of Science and Tech-nology, P. O. Box 88, Manchester M60 1QD, UK. Tel.: 44-1612004187;Fax: 44-1612360409; E-mail: r.ford@umist.ac.uk.

1 The abbreviations used are: ABC, ATP-binding cassette; NBD, nu-cleotide-binding domain; TMD, transmembrane domain; P-gp, P-glyco-protein; ICD, intracytoplasmic domain.

THE JOURNAL OF BIOLOGICAL CHEMISTRY Vol. 280, No. 4, Issue of January 28, pp. 28572862, 2005 2005 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in U.S.A.

This paper is available on line at http://www.jbc.org 2857

8/13/2019 2857.full

2/6

software programs (21) and visualized using the XtalView softwarepackage (22). A map calculated with a low temperature factor of 40 2was used initially for identification of the molecular envelope andapproximate location of the NBDs. A map calculated with a highertemperature factor of 200 2 was then employed for a more preciseplacement of -helices using manual alignment to continuous cylindri-cal densities and using the xfit routines within the XtalView package(22). The NBDs were subsequently fitted manually to the three-dimen-sional map using either the MJ0796 (homodimer with ATP bound) (9) orthe BtuD homodimer (2) as models.

RESULTS

Two-dimensional crystals of Chinese hamster P-gp in thenucleotide-bound state were grown, and data were collected bycryo-electron microscopy and analyzed by crystallographic im-age processing (Table I). The crystals were well ordered withstrong reflections to 8 resolution for both untilted and highlytilted crystals (Fig. 1), whereas weaker reflections could bedetected to 6 resolution. The flatness of the crystals was goodas determined by checking for any systematic deviation of thelattice parameters for several 2048 2048 pixel boxes overlarge crystalline areas.

The final resolution limit of the data was judged to be 6 in the direction parallel to the crystal plane and 12 perpen-dicular to the crystal plane. These judgments were based onphase errors, figure of merit, and percentage of completenessover increasing resolution ranges as detailed in Table I to-gether with a visual assessment of the degree of scatter of phases along continuous lattice lines in the z* direction inreciprocal space (Fig. 2). The three-dimensional sampling of reciprocal space is 70 80% complete within the tilt range em-ployed to about 11 resolution and falls off to about 50%completeness to 8.5 resolution (Table I). However, samplingof data is reduced considerably beyond this point; between 8.5and 7.2 resolution only 11% of expected structure factorsare measured with a figure of merit 0.1, indicating the cur-rent resolution limit of these three-dimensional data. A reso-lution cutoff of 6 was used for display of the projection map inFig. 3 B, and 8 resolution for display of the three-dimensionalmap in Fig. 4.

Fig. 3 shows projection maps obtained from the unstainedtwo-dimensional crystals of P-gp observed under cryo-condi-tions. Fig. 3 A displays a low resolution map with a resolutioncutoff at 20 to allow comparison with previous low resolutionstudies of negatively stained P-gp (1012). It shows a unit cellconsisting of a single molecule (black circle ) with a clover leafshape similar to that obtained from negatively stained crystalsof P-gp in the nucleotide-bound state (1112). A similarlyshaped projection map has also been observed with negatively

stained two-dimensional crystals of CFTR with nucleotide

bound (17). Fig. 3 B shows the P-gp map with a high resolutioncut off at the 6 resolution limit of these data. Several strongelliptical or round peaks of density of 10 diameter can beseen, consistent with densities expected for transmembrane

-helices.The three-dimensional density map of P-gp is displayed in

Fig. 4. Two orthogonal views from a direction perpendicular tothe long molecular axis are shown in A and B, respectively. Views from a direction perpendicular to the crystal plane areshown in CE . Each view corresponds to a section of density

30 thick. The blue three-dimensional netting drawn at athreshold of 1.5 times the standard deviation above the meandensity level indicates strong protein density. The approximateenvelope of one P-gp molecule within the density map is delin-eated by a dashed white line . The P-gp molecules in thesecrystals were packed so that their long molecular axes subtendan angle of about 35o to the normal to the crystal plane (thedirection of the crystal plane is indicated by the double-headedarrows in A and B). The molecule is 160 long and 45 65 wide, dimensions that are similar to those obtained for P-gpand for three other ABC transporters (Pdr5p, YvcC, and CFTR)obtained by low resolution structural studies (1012, 14, 17,23).

Twelve cylindrical densities were observed in the centralregion of the 8 -resolution three-dimensional map, each

5075 long and 10 in diameter (Fig. 4, A and B). Thesefeatures were clear at high thresholds equivalent to two timesstandard deviation (2 ) above the mean density in the map andcould be modeled as -helical segments within the transmem-brane domains (13, 2426). The length of most of these cylin-drical densities was 1520 greater than that required to spanthe lipid bilayer. Hydrophilic intracytoplasmic loops of theTMDs (the so-called intracytoplasmic domains or ICDs) were

observed as helical extensions of the membrane-spanning-helices in MsbA (1, 3), which interact directly with the NBDs.In contrast, BtuCD does not contain -helical ICDs, and thecontacts between transmembrane domains and NBDs are me-diated by structured loops (2). The lengths of the cylindricaldensities observed in P-gp suggest that it has ICD helicalextensions similar to MsbA (1, 3).

The long axes of the putative -helices were approximatelyparallel to the long axis of the molecule and formed a 70 -thick belt across its center with a roughly elliptical profile of 60 40 (Fig. 4C). Only one putative -helix was significantlytilted away from the long molecular axis. This elongated den-sity was found to cross from one half of the molecule to contactthe other half (see Fig. 5 B, helix indicated by arrowhead ). The

12 cylindrical densities of sufficient length to span the mem-

T ABLE I Summary of the crystallographic image processing data

Two-dimensional plane group p1Lattice constants a 67.4 0.8 , b 68.2 1.2 , 124.4 2.1Thickness used for map 200 Number of imagesa 93In-plane resolution limit 6 Maximum tilt anglea 55.0Number of merged reflectionsb 6210Number of interpolated structure factors to 8 resolution 1146

Mean phase errorb

, mean figure of merit (FOM) (27), and %completenessc over the given resolution ranges (20020 ) 18.8 0.920 97(20011 ) 34.2 0.779 78(2008.5 ) 38.0 0.734 48(2007.2 ) 39.3 0.721 40Resolution range Phase error FOM % Completeness

a Distribution of images with given tilt angles 12 (025), 14 (2535), 29 (3545), and 38 (4560).b IQ (signal: noise measure (27)) 17 reflections were included; standard error of mean phase is shown (27).c % Completeness of the data in a given resolution range was calculated by dividing the number of experimentally measured structure factors

with figure of merit 0.1 by the theoretical maximum number of structure factors available in this range.

Three-dimensional Structure of P-glycoprotein2858

8/13/2019 2857.full

3/6

brane are seen to be segregated into two clusters of six, pre-sumably corresponding to the two TMDs, when viewed alongthe long axis of the protein (dashed line in D). Their packing isdiscussed in more detail below.

The crystal surface in contact with the electron microscopegrid support film has previously been assigned to the extracel-lular, glycosylated face of the P-gp molecule (1012). This side

of the crystal (Fig. 4C and the uppermost region in A and B)

contained densities which are partly folded back over the top of the molecule ( A, arrowhead ), probably corresponding to thesingle large extracellular loop of P-gp and can be modeled astwo -helices (Fig. 4C). It seems likely that glycosylation alsocontributes to this density.

The region furthest away from the support film in the three-dimensional crystals corresponding to the cytoplasmic side of the molecule (1012) (lowermost in Fig. 4, A and B), consists of two large heart-shaped domains related by a (pseudo-)2-foldrotational symmetry and has the 2-fold axis running approxi-mately parallel to the long molecular axis. These densitiescorrespond to the two homologous NBDs (2, 3, 49). The

-strands and short -helices of the NBDs will be poorly re-solved in an 8 resolution map, and their identification willalso be impeded by the fading of density toward the edges of themap because of the missing cone problem associated with two-

dimensional crystallography (20, 27). The MJ0796 NBD struc-

FIG. 1. Quality of the two-dimensional crystals. Computed Fourier transforms of images obtained for crystals of P-gp tilted to 53 (right ) anduntilted (left ). The dashed line indicates the tilt axis. Resolution in is marked by the concentric circles . The size of the boxes and the numberswithin designate the signal:noise ratio of the reflection with larger boxes and lower numbers indicating higher signal:noise (27).

FIG. 2. Quality of the three-dimensional data along the direc-tion perpendicular to the crystal plane. Selected lattice lines forthe two-dimensional crystals showing the observed amplitudes (ampl. )and phases ( points ) and the fitting of the data to a continuous transform(solidline ). The interpolated structure factors sampled at regular pointsalong the lattice line are shown with their estimated error (bars ). Notethat in this flat graphical representation of the phase data, 180 isequivalent to 180, hence the anomalously large phase fluctuationsfrom 180 to 180 are in fact smooth transitions in phase.

FIG. 3. Projection data (contour maps). Projection maps calcu-lated from untilted P-gp crystals with AMP-PNP bound. Themap on theleft ( A) shows the projection data curtailed at a resolution of 20 forcomparison with previous data. The map on the right ( B) shows thesame data to a resolution of 6 . The corresponding areas in the twomaps are identified by the circle . The scale bar represents 5 nm.

Three-dimensional Structure of P-glycoprotein 2859

8/13/2019 2857.full

4/6

ture with ATP bound (9) and the BtuD NBD structure (notshown) (3) can both be modeled onto the P-gp density map in arobust manner and in almost identical positions, suggestingthat the arrangement of the NBDs with respect to the TMDs inP-gp is not radically different from either of them at thisresolution. The model for the NBDs shown in Figs. 4 and 5 isbased on the MJ0796 dimer in the ATP-bound state (9).

DISCUSSIONWe have generated the first three-dimensional structure for

P-gp at 8 resolution, and its interpretation is summarized inFig. 5. The overall space-filling volume is shown in A, whereas BD show the transmembrane region of the transporter with

-helices represented by spiral ribbons. In cross-section, themolecule has an overall clover leaf shape ( B, dashed line )similar to the low resolution envelope (Fig. 3) determined pre- viously by other methods (11, 12). Twelve transmembrane

-helices of sufficient length to span the membrane are seen intwo blocks of six exhibiting a pseudo-2-fold symmetry as ex-pected given the sequence homology between the two TMDs of the transporter. Pseudo-symmetry-related -helices are shownin the same color, whereas -helices with no apparent pseudo-

symmetry-related partner are colored gray . The pseudo-sym-

metry-related halves of the TMDs are compared in Fig. 5, Cand D after rotation of one half by 180 around the pseudo-2-fold axis. Five -helices (dashed enclosure ) from each TMD areclearly repeated in each of the two halves of the transporter.The sixth pair of -helices ( BD , yellow helices ) that make upthe 12 transmembrane -helices show a less symmetrical rela-tionship. One is tilted so that it crosses over to contact the otherTMD, whereas its partner aligns roughly parallel to the otherlong -helical segments (C and D , arrows ). Four further densi-ties, fitted as short -helices, are unrelated by the symmetry of the transporter ( A and B, gray helices ) and are probably notmembrane-spanning. Two of these densities ( A, top ) likely cor-respond to the major extracellular loop in P-gp, whereas one ( B,asterisk ) is tilted well away from the long axis of the moleculeand is too short to cross the membrane. The final density ( Aand B, X ) lies at the interface between adjacent unit cells,contacting the TMD region of one unit cell and the extracellularsurface of P-gp of the adjacent unit cell. There are therefore twopossible locations of this density (as indicated by X , A) whichcannot be distinguished as full tracing of the polypeptide chainis unfeasible at the current resolution limits.

The deviation from true 2-fold symmetry in the TMD region

FIG. 4. Three-dimensional data (density map). The three-dimensional density map for P-gp in the AMP-PNP-bound state displayed with aresolution cutoff of 8 at a threshold of 1.5 above the mean density level and a temperature factor of 200 2 (blue netting ). Fitting of -helicesand NBDs is shown by the yellow C- traces. A and B show side views of the map (the yellow scale bar represents 20 ). B is related to A by a 90orotation around the vertical axis as indicated at the top of each panel. Molecules are packed in the crystal with the longest axis of the protein tiltedat an angle of 35o from the perpendicular with respect to the crystal plane direction (indicated by the double-headed arrow in each panel). Thelocations of the TMDs, ICDs, and NBDs are indicated by the arrows and dashed lines . The putative extracellular loop ( ECL ) is indicated at the topof the panels. CE show sections through the map corresponding to the positions of the pink arrows in B. The viewing direction is perpendicularto the crystal plane from the intracellular side. A roughly elliptical profile is observed in D and E , with two clusters of densities (segregated by thered dashed lines ). The yellow scale bar in C indicates 20 (and also applies to DE ).

Three-dimensional Structure of P-glycoprotein2860

8/13/2019 2857.full

5/6

is likely to be a consequence of conformational changes inducedby nucleotide binding (11, 12). Low resolution (20 ) maps of P-gp in the nucleotide-free configuration have a more pro-nounced 2-fold (pseudo)-symmetry and display a cavity in themembrane (11, 12). Similar conformational states have beenobserved in a related ABC transporter, CFTR (17), reinforcingthe original conclusions for P-gp. The present 8 -resolutionstructure agrees well with the (lower resolution) structure of P-gp in negatively stained crystals grown in the presence of nucleotide. Cryo-electron microscopy studies of P-gp in thenucleotide-free state are now needed to describe with greaterprecision the nature and magnitude of the conformationalchanges induced by nucleotide binding. For the time being, acomparison of 20 - and 8 -resolution structures for the nu-cleotide-free and -bound states, respectively, concurs with theoriginal interpretation (11) that the conformational changeopens out a central cavity in the TMD region of the P-gpstructure. These and other studies have suggested that theconformational changes mediate transport of substances bythese transporters (2, 3, 11, 12, 2832; see Ref. 33 for a review).

P-gp has been the subject of extensive cross-linking andsite-directed mutagenesis studies (24, 26, 29, 34, 35). Thesestudies frequently place transmembrane -helices 6 and 12(TM6 and TM12) close to each other, especially at the cytoplas-mic side of the membrane, and these helices have been impli-cated in drug binding (36, 37). The two -helices at the centerof the protein (Fig. 5 B, green helices ) are therefore good candi-dates for TM6 and TM12. These are closer at the bottom (cy-toplasmic side) where they are probably in van der Waalscontact in agreement with cross-linking data (24). These heli-ces also display a kink or discontinuity in the density about

halfway through the membrane. Interestingly, mutations in a

similar position in TM6 and TM12 cause significant loss of activity in the transporter (38). Recent biochemical data haveshown that upon ATP binding (AMP-PNP) TM6 tilts with re-spect to other helices (39) consistent with the prediction thatthe tilted helix at the center of the structure (Fig. 5, greenhelices ) is indeed TM6. TM5 and TM8 have also been shown bycross-linking to be near to each other, as have -helices TM2and TM11 (34, 35). It is not, however, possible at this resolutionto predict which of the -helices in the P-gp structure reflectthe TM5/TM8 and TM2/TM11 pairs.

As expected from the sequence homology between P-gp andthe bacterial ABC transporter MsbA, the packing of the trans-membrane helices observed here for P-gp is not inconsistentwith that reported for the MsbA monomer (1, 3, 24, 29). Differ-ences are likely to be a consequence of the fact that the P-gpstructure has bound nucleotide, known to cause significantconformational changes in the transmembrane domains (11,12). However, the packing of the two halves of the molecule inP-gp with respect to each other is very different from thetapering or teepee arrangement reported for the MsbA ho-modimer (1, 3). Thus, the interface between the two TMDs inP-gp is much more extensive than in the Escherichia coli MsbA homodimer. Molecular dynamics simulations of E. coli MsbA (29) suggest that this small dimer interface would not conferthermodynamic stability, whereas the current P-gp structureappears to be thermodynamically stable in such moleculardynamics simulations.2 Indeed, the overall shape of P-gp issimilar to the BtuCD homodimer (2) where the interface be-tween the monomers is extensive and the long axes of each

2

J. Campbell and M. S. P. Sansom, personal communication.

FIG. 5. Interpretation of the three-dimensional map. Space-filling model of P-gp in the nucleotide-bound form based on cryo-electronmicroscopy data and with all residues modeled as alanine. A shows a side view of the protein with the NBDs (violet ) at the bottom with a top viewbelow. The 12 putative membrane-spanning -helices have been colored in pairs to indicate the two halves of the transporter. A pseudo-symmetryrelationship is seen. Four additional gray-colored helices do not show an obvious symmetry relationship; one (*) is intracellular, tilted, and too shortto cross a membrane; another (shown in the side view by X ) is ambiguous in its location (see text); the other two are at the extracellular side of the protein (the dashed lines indicate the putative bounds of a 4.5-nm-thick lipid bilayer). Scale bar 5 nm. B shows a view along the long axisof the molecule from the extracellular side with the two densities on the extracellular surface and the NBDs removed for clarity. Spiral ribbonsindicate helical densities in the structure. The pseudo-2-fold axis is indicated by the rotation symbol . Dotted lines indicate discontinuities or kinksin two central helices (dark green ) that wrap around the pseudo-symmetry axis. C and D show the two pseudo-symmetry related halves of thetransporter after the rotation of one half by 180 around the pseudo-2-fold axis. The dashed enclosure highlights the most strongly symmetry-related helices. Five of the six helices in each TMD show strong pseudo-symmetry; the sixth pair (colored yellow ) do not.

Three-dimensional Structure of P-glycoprotein 2861

8/13/2019 2857.full

6/6

monomer line up side by side. We also find that the P-gp NBDsin the AMP-PNP bound state are arranged similarly to those inthe BtuD dimer (2) and the MJ0796 dimer with ATP bound (9)but are very different from the NBDs reported for E. coli MsbA (1). The NBDs in the Vibrio cholera MsbA structure (3) are incontact and therefore are closer to the arrangement in the P-gpmap, but they still display major differences in their structurewith respect to the other isolated NBD structures (4, 59) andwith respect to the BtuD dimer and are therefore unsuitable formodeling the P-gp structure. In summary, the structural datafor P-gp support the suggestion (24) that the MsbA homodimerstructure does not reflect the native domain interfaces for thissubfamily of ABC transporters.

The structure described herein provides a further insightinto the packing of transmembrane -helices in ABC transport-ers and shows a novel asymmetric conformational state (nucle-otide-bound) that will be important for developing molecularmodels of P-gp and ABC transporter mechanisms of action (24,29). This represents the first three-dimensional structure for aeukaryotic ABC transporter and the fourth structure of any ABC transporter at a resolution sufficient to resolve the pack-ing of the transmembrane -helices (13). In P-gp/MsbA andBtuCD, members of two distinct subfamilies of ABC transport-

ers, the packing of the transmembrane -helices is very differ-ent. It remains to be seen whether ABC transporters willdisplay a wide range of folds for the TMDs or whether somestructural consensus will emerge.

Acknowledgments We thank Alice Rothnie, Janet Storm,Per Bullough, Jeff Campbell, Jeremy Derrick, Alhaji Bukar Kamis,Ian Kerr, Werner Kuhlbrandt, Janet Vonck, Kenny Linton, andSteve Prince for useful discussions and comments on the manuscript.We thank Per Bullough, Werner Kuhlbrandt, and Marin Van Heel forfacilitating access to microscopes at the University of Sheffield, MPIFrankfurt, and Imperial College London, and Derryk Mills, Peiyi Wang,Ed Morris, and Shoaxia Chen for assistance with microscopy.

REFERENCES1. Chang, G., and Roth, C. B. (2001) Science 293, 179318002. Locher, K. P., Lee, A. T., and Rees, D. C. (2002) Science 296, 109110983. Chang, G. (2003) J. Mol. Biol. 330, 4194304. Kerr, I. D. (2002) Biochim. Biophys. Acta 1561, 47645. Hung, L. W., Wang, I. X., Nikaido, K., Liu, P. Q., Ames, G. F., and Kim, S. H.

(1998) Nature 396 , 7037076. Diederichs, K., Diez, J., Greller,G., Muller, C., Breed, J., Schnell, C., Vonrhein,

C., Boos, W., and Welte, W. (2000) EMBO J. 19, 59515961

7. Hopfner, K.-P., Karchner, A., Shin, D. S., Craig. L., Arthur, L. M., Carney,J. P., and Trainer, J. A. (2000) Cell 101, 789800

8. Gaudet, R., and Wiley, D. C. (2001) EMBO J. 20, 4964 49729. Smith, P. C., Karpowich, N., Millen, L., Moody, J. E., Rosen, J., Thomas, P. J.,

and Hunt, J. F. (2002) Mol. Cell 10, 13914910. Rosenberg, M. F., Callaghan, R., Ford, R. C., and Higgins, C. F. (1997) J. Biol.

Chem. 272 , 106851069411. Rosenberg, M. F., Velarde, G. S., Ford, R. C., Martin, C., Berridge, G., Kerr,

I. D., Callaghan, R., Schmidlin, A., Wooding, C., Linton, K. J., and Higgins,C. F. (2001) EMBO J. 20, 56155625

12. Rosenberg, M. F., Kamis, A. B., Callaghan, R., Higgins, C. F., and Ford, R. C.(2003) J. Biol. Chem. 278, 8294 8299

13. Velarde, G. S., Ford, R. C., Rosenberg, M. F., and Powis, S. J. (2001) J. Biol.Chem. 276, 4605446063

14. Ferreira-Pereira, A., Marco, S., Decottignies, A., Nader, J., Goffeau, A., andRigaud, J.-L. (2003) J. Biol. Chem. 278, 1199511999

15. Lee, J.-Y., Urbatsch, I. L., Senior, A. E., and Wilkens, S. (2002) J. Biol. Chem.277, 4012540131

16. Rosenberg, M. F., Mao, Q., Holzenburg, A., Ford, R. C., Deeley, R. G., and ColeS. P. (2001) J. Biol. Chem. 276, 1607616082

17. Rosenberg, M. F., Kamis, A. B., Aleksandrov, L. A., Ford, R. C., and Riordan,J. R. (2004) J. Biol. Chem. 279, 3905139057

18. Cyrklaff, M., Auer, M., Kuhlbrandt, W., and Scarborough, G. A. (1995) EMBO J. 14, 18541857

19. Callaghan, R., Berridge, G., Ferry, D. R., and Higgins, C. F. (1997) Biochim. Biophys. Acta 1328, 109124

20. Crowther, R. A., Henderson, R., and Smith, J. M. (1996) J. Struct. Biol. 116,916

21. Bailey, S. (1994) Acta Crystallogr. Sect. D Biol. Crystallogr. 50, 76076322. McRee, D. E. (1992) J. Mol. Graphics , 10, 444723. Chami, M., Steinfels, E., Orelle, C., Jault, J.-M., Di Pietro, A., Rigaud, J.-L.,

and Marco, S. (2002) J. Mol. Biol. 315, 1075108524. Stenham, D. R., Campbell, J. D., Sansom, M. S. P., Higgins, C. F., Kerr, I. D.,

and Linton, K. J. (2003) FASEB J. 17, 2287228925. Kast, C., Canfield, V., Levenson, R., and Gros, P. (1995) Biochemistry 34,

4402441126. Loo, T. W., and Clarke, D. M. (1995) J. Biol. Chem. 270, 84384827. Glaeser, R. M., and Downing, K. H. (1992) Ultramicroscopy 47, 25626528. Wang, G. C., Pincheira, R., Zhang, M., and Zhang, J. T. (1997) Biochem. J. 328,

89790429. Campbell, J. D., Biggin, P. C., Baaden, M., and Sansom, M. S. P. (2003)

Biochemistry 42, 3666367330. Druley, T. E., Stein, W. D., and Roninson, I. B. (2001) Biochemistry 40,

4312432231. Julien, M., and Gros, P. (2000) Biochemistry 39, 4559 456832. Gabriel, M. P., Storm, J., Taylor, A. M., Linton, K. J., Kerr, I. D., and Cal-

laghan, R. (2003) Biochemistry 42, 7780778933. Higgins, C. F., and Linton, K. J. (2004) Nat. Struct. Mol. Biol. 11, 91892634. Loo, T. W., Bartlett, C., and Clarke, D. M. (2004) J. Biol. Chem. 279,

7692769735. Loo, T. W., Bartlett, C., and Clarke, D. M. (2004) J. Biol. Chem. 279,

182321823836. Loo, T. W., Bartlett, C., and Clarke, D. M. (2003) J. Biol. Chem. 278,

397063971037. Loo, T. W., and Clarke, D. M. (1997) J. Biol. Chem. 272, 319453194838. Loo, T. W., and Clarke, D. M. (2001) J. Biol. Chem. 276, 149721497939. Rothnie, A., Storm, J., Campbell, J., Linton, K. J., Kerr, I. D., and Callaghan,

R. (2004) J. Biol. Chem. 279, 3491334921

Three-dimensional Structure of P-glycoprotein2862