ChemBioChem 2002, 3, 161 ± 165 ¹ WILEY-VCH-Verlag GmbH, 69451 Weinheim, Germany, 2002 1439-4227/02/03/02-03 $ 17.50+.50/0 161

Membrane biochemistry is a very excitingand fast-moving research area. Our under-standing of membranes and the proteinsembedded within the two-dimensionalfluid of lipids is still limited. The processesoccurring at this interface are even lessunderstood. Chang and Roth have re-cently determined the first X-ray crystalstructure of an ATP binding casette (ABC)transporter, MsbA from Escherichia coli(Figure 1; ATP� adenosine triphos-phate).[1] This structure represents a majorachievement in the field of membranebiochemistry. Biological membranes are aprotective shield against a hostile envi-ronment. However, the two-dimensional,impermeable nature of biological mem-branes[2] creates a severe and sometimeslife-threatening problem. Living organ-isms have to take up nutrients, extrudeharmful substances, and of course ex-change information. During evolutionmany transmembrane proteins haveevolved, to ensure that living organismsare able to survive and multiply.One of the most common families of

transmembrane proteins found in allthree kingdoms of life is the family ofABC transporters.[3] The substrates of thissuperfamily range from small inorganicions (such as chloride ions) to aminoacids, sugars, drugs, and even largeproteins. Despite such diversity, all mem-bers of the family of ABC transportersshare a common blueprint, which com-prises of four modules: two transmem-

brane domains (TMDs) and two nucleo-tide- or ATP-binding domains (NBDs). Inall cases, the energy released during ATPhydrolysis by the NBD is used to trans-locate the substrate. Every arrangementof these four domains is possible. Inbacteria, separate polypeptide chainscommonly make up each of the fourmodules. However, one NBD and oneTMD might be fused on a single protein,to generate a so-called half-size trans-porter. In eukaria, a single polypeptidechain generally makes up all four do-mains. This architecture corresponds tothe so-called full-size transporter; the half-sized transporter is the exception.[4]

On a functional and structural level, it isnow commonly accepted that the NBDsprovide only the energy for translocationwhile the TMDs confer substrate specific-

ity. The NBDs, which define an ABC trans-porter, contain three conserved sequencemotifs. The Walker A motif (consensussequence: GXXGXGKST, where X denotesany amino acid), the Walker B motif (con-sensus sequence: hhhhD, where h de-notes any hydrophobic amino acid),[5] andthe signature motif or ™C-loop∫ (consen-sus sequence: LSGQQR), which is specificfor ABC transporters. As a consequence,NBDs are well-conserved among thisprotein family. On the other hand, TMDsshare little sequence homology and eventhe number of �-helices seems to varyfrom one ABC transporter to the other. Asix �-helix core is the minimal require-ment for an ABC transporter, but varia-tions are generally the rule rather than theexception.[3]

The best-characterized ABC transport-ers are bacterial ones, for example, thehistidine permease[6] or maltose trans-porters.[7] In humans, the cystic fibrosistransmembrane conductance regulator(CFTR),[8] transporter associated with anti-

The First View of an ABC Transporter:The X-ray Crystal Structure of MsbA from E. coliLutz Schmitt*[a]

KEYWORDS:

ABC transporters ¥ membrane proteins ¥ multidrug resistancestructure ± activity relationships ¥ structure elucidation

[a] L. SchmittInstitute of Biochemistry, Biocenter N210Johann Wolfgang Goethe University FrankfurtMarie-Curie Strasse 960439 Frankfurt (Germany)Fax: (�69)79829-495E-mail : lschmitt@em.uni-frankfurt.de

Figure 1. Three-dimensional structure of the homodimeric MsbA. �-helices are represented in red, �-strands inblue, and coils in yellow. The figure was generated by using the MOLSCRIPT[31] and Raster3D[32] programs basedon the Protein Databank entry 1JSQ.

L. Schmitt

162 ChemBioChem 2002, 3, 161 ± 165

gen processing (TAP),[9]and multidrug re-sistance protein 1 (MDR1)[10]are the mostprominent members of the ABC trans-porter family. Point mutations in CFTR,mainly F508�, are the genetic cause ofcystic fibrosis, the most common inher-ited disease among Caucasians. MDR1causes severe problems in chemotherapy.MDR1 is overexpressed in most tumorsupon treatment with chemotherapeuticsand results in resistance against theapplied drugs. As a consequence, thedrug dose has to be increased so thateven healthy cells are affected and thechemotheraphy becomes ineffective.These two examples explain why a mo-lecular understanding of ABC transportersthat are involved in diseases, especiallytransporters in human, is more thanworthwhile. But multidrug resistance isnot only a human phenomenon. More orless every organism contains a transmem-brane protein conferring resistance tomany natural and synthetic drugs,[11] andin many cases these proteins belong tothe family of ABC transporters.[12]

Despite two decades of intensive re-search and an overwhelming body ofexperimental data, little is know about themechanisms of transport or the couplingof ATP hydrolysis and substrate trans-location; there is also a lack of structuralinformation. For example, MDR1 is able totransport more or less every hydrophobicdrug known today.[10] Examples are vincaalkaloids (e.g. , vinblastine), antibiotics(e.g. , actinomycin D), Taxol, protein-syn-thesis inhibitors (e.g. , puromycin), DNAintercalators (e.g. , ethidium bromide),toxic peptides (e.g. , valinomycin), or fluo-rescent dyes (e.g. , rhodamine). How can asingle protein achieve such diversity whilemaintaining specificity? This knowledgeis, of course, a prerequisite for the devel-opment of drugs or specific inhibitiorsthat would open up new avenues notonly in cancer therapy but also in thetreatment of infectious diseases spreadthrough bacteria, which have achievedmultidrug resistance to many commonlyused drugs. In the last three years, severalthree-dimensional structures of NBDsfrom ABC transporters have been solvedby X-ray crystallography.[13±17] However, nostructural information for the TMDs of anABC transporter was available. The onlyexceptions were the low-resolution struc-

tures of human MDR1[18, 19] and multidrugresistance related protein 1 (MRP1)[20] ob-tained from single-particle analysis andtwo-dimensional crystals.In Science,[1] Geoffrey Chen and Chris-

topher B. Roth from the Department ofMolecular Biology at The Scripps ResearchInstitute in La Jolla, USA, reported the firstthree-dimensional structure of a completeABC transporter, MsbA from E. coli (Fig-ure 1). MsbA transports lipid A. It sharesaround 30% sequence homology withhuman MDR1. However, in contrast toMDR1, which is a full-size transporter,functional MsbA is assembled from twohalf-size transporters. It might very wellserve as a structural model for many ofthe ABC transporters which give rise tothe phenomena of drug resistance.The bottleneck of modern X-ray crys-

tallography is the availability of well-ordered three-dimensional crystals. Dueto the techniques of modern molecularbiology, more or less every water-solubleprotein can be produced in quantitiessufficient for structural analysis. However,structural investigations of membraneproteins are still hampered by the limitedamounts of protein. In addition, anotherimportant parameter has to considered:the choice of the right detergent.[21, 22]

Taking these points together, even thenonexpert can understand why the struc-ture determination of a membrane pro-tein is more than an ordinary challenge.The milestone of the MsbA structure

was achieved by a tour de force. In orderto obtain X-ray suitable crystals, a tremen-dous effort was undertaken. A total of20 different MDR-mediating ABC trans-porters from 14 different organisms werecloned, overexpressed, purified, and in-vestigated for their ability to crystallize.The incredible number of 96000 crystal-lization trials was performed with around20 different detergents. At the end of thislong torture, 35 crystal forms were ob-tained. Out of these crystals, MsbA hadthe best diffraction quality. However, evenafter this impressive struggle, native crys-tals of MsbA diffracted only to 6.2 ä andshowed strong anisotropic diffraction.Nevertheless, the authors proceeded andapplied a so-called ™refinement strategy∫,which was intended to strengthen latticecontacts in order to improve diffractionquality. After another tour de force, which

included an intensive screening of deter-gents, detergent concentrations, temper-atures, and inorganic and organic com-pounds, the diffraction limit was raised to4.5 ä through OsCl3. As this was still notenough, the procedure of refinement alsohad to be modified to the needs of thisparticular structure. It is beyond the scopeof this article to describe the effortsundertaken in the work, but it finallyresulted in an electron density of MsbA,which was of sufficient quality to trace theprotein backbone chain. The quality ofthe structure determined, even in light ofthe moderate resolution, is indicated bythe R factor (27%, Rfree� 38%).The structure presented by Chang and

Roth answers many questions, but it alsoleaves many questions open and evenraises extra questions. The overall struc-ture of MsbA at 4.5 ä resolution is shownin Figure 1. The crystal structure is con-sistent with the fact that MsbA forms ahomodimer as the functional unit. Per-haps most important is the fact that theTMDs are solely composed of �-helices.This proves, beyond any doubt, that thesubstrate pathway (TMD) is in agreementwith the secondary structure predictionproposed by many other laboratories andsupported by biochemical evidence. Inthe case of MsbA, the TMDs are eachcomposed of six �-helices, which have atilt angle of 30 ± 40� with respect to thebilayer normal. This and many other de-tails of the structure agree with twodecades of experimental work and giveus confidence. The NBDs, as far as they arevisible within the experimental electrondensity, agree with the recently publishedstructures of isolated NBDs.[13±17] However,the N-terminal region (residues 341 ± 418),which includes the Walker A motif, is notvisible. The exact position and role of theNBDs are therefore under speculation.The first surprise of the structure is thepresence of a third type of domain, whichhas not been suggested before. Thisintracellular domain (ICD), which corre-sponds to residues 97 ±139 (ICD1), 193 ±252 (ICD2), and 302 ±327 (ICD3), is helicalin nature (three helices in the case of ICD1and two helices each for ICD2 and ICD3)and connects the NBD and the TMD ineach half-sized transporter. IDC1 is posi-tioned directly ™above∫ the ABC signaturemotif (LSGGQQ) of the NBD and is

X-ray Crystal Structure of MsbA

ChemBioChem 2002, 3, 161 ± 165 163

thought to transmit information betweenthe NBD and TMD (see Figure 2). Thistransmission very likely occurs throughrearrangement of the helical IDC.

Figure 2. Organization of the ICD. For simplicityonly a monomeric unit of MsbA is shown. The TMD isshown in red, the NBD in dark blue and the ICD inpurple (ICD1), light blue (ICD3), and orange (ICD2).The C-loop within the NBD is highlighted in green.Helix 2 of ICD 1 is positioned above the C-loop, whichimplys a possible pathway for signal transduction.Please note that the second helix of ICD2 is hidden inthis orientation. For further details, please see text.The figure was generated as in Figure 1 by rotation of90� in plane.

Obviously, the three-dimensional archi-tecture of the TMDs is the most excitingand surprising part of the structure. Thehelices form a cone-like structure withtwo large openings within the bilayersection of the protein (Figures 1 and 3).The part of MsbA located in the outerleaflet is closed, while the part positionedin the inner leaflet of the membrane iswidely open. The two openings facing thebilayer are around 25 ä wide and locatedsolely within the inner (cytosolic) leaflet ofthe bilayer. The base of the formedchamber, which is located at the cytosolicside of the bilayer is roughly 45 ä wide.This chamber can easily accommodatethe substrate, lipid A. The opening of thechamber within the inner leaflet of thebilayer guarantees free entry of the sub-

strate, while the closed structure withinthe outer leaflet prohibits entry or exit.Therefore, the unidirectional translocationof lipid A from one side of the bilayer tothe other is understandable. The open-ings of the chamber are defined bytransmembrane helix 2 (TM2) from onemonomer within the dimer and TM5 fromthe other. However, the whole chamber isformed from side chains of all the TMs.Another interesting point is the chargedistribution within the chamber. While thepart of the substrate binding side locatedat the inner leaflet contains a cluster ofpositively charged amino acids, the partlocated in the outer leaflet is hydrophobicin nature. As suggested in Figures 1 and 3,the chamber creates a large perturbationwithin the inner leaflet. Intuitively, onewould expect that this arrangement gen-erates a lot of stress on the membraneand that the impermeable nature of thebilayer might be endangered.Based on the structure of MsbA, which

was obtained in the absence of substrateand any nucleotide, Chang and Roth havederived a possible mechanism of sub-strate transport for MsbA and for ABCtransporters in general. The model isbased on the presented structure and abody of available biochemical data. Ineach monomer, the ICD seems to be a

™sensing unit∫, which transmits signalsfrom the TMD to the NBD and vice versaby conformational changes. Lipid A bindsto the open chamber from the innerleaflet of the bilayer. Information of thisevent is conducted through the ICDs tothe NBDs; this triggers ATP hydrolysis.Such substrate-induced stimulation ofATPase activity has also been demonstrat-ed for MDR1 (see, for example, ref. [23]) orTAP.[24] A conformational change of theNBDs upon ATP hydrolysis is proposed toinduce an interaction between bothNBDs. However, the N-terminal regionsof the NBDs are not visible in the electrondensity and their exact position andinteraction is speculative. However, sucha scenario would result in a rearrange-ment of the whole molecule. The reor-ganization of the NBDs influences theTMDs. The chamber is lined with a clusterof charges, which creates an energeticallyunfavorable situation for lipid A. Suchcharges also imply the presence of boundsolvent. However, as pointed out above,the TMDs contain an asymmetrical chargeand polarity distribution: charged andhighly polar in the lower part (chamber),while hydrophobic in the upper part.Together with the interaction of the NBDsupon ATP hydrolysis and a subsequentstructural reorganization, lipid A flips over

Figure 3. Domain organization of the homodimeric MsbA. The TMDs are given in red, the ICDs in purple, andthe NBDs in blue. The orientation of MsbA is identical to the one shown in Figure 1. The orientation andlocation of the putative bilayer is indicated with solid lines. The middle line indicates the border between theinner and outer leaflets of the membrane bilayer. The figure was generated as in Figure 1.

L. Schmitt

164 ChemBioChem 2002, 3, 161 ± 165

into the upper part, which is energeticallyfavored. Such a model is in agreementwith the observed vectorial transport oflipid A. After flipping, the substrate isproperly oriented to enter the outer leaf-let of the membrane and complete thetransport cycle. This, of course, requiresthe rearrangement of TM2 and TM5 tocreate an opening into the outer leaflet.Finally, the extrusion of lipid A sends asignal to the NBDs, probably through theICDs again, which induces ADP±ATP ex-change or spontaneous ADP release. Thisbrings the system back to the groundstate. One ATP molecule is consumed pertransported substrate during the pro-posed cycle. This agrees with data ob-tained for MDR1.[25, 26] However, the ™tilt-ing∫ movement of the TMDs is energizedby ATP hydrolysis, while the flipping oflipid A is driven by charge and polaritygradients along the TMDs. Such a tiltingwould impose a large amount of physicalstress on both leaflets of the membraneand require lipid reorganization to coun-terbalance the different space require-ments of MsbA during the transport cycle.The proposed mechanism raises anotherquestion: What is the driving force of thetransport process? A recognition step hasto take place, because MsbA transportslipid A with high specificity. After ATPhydrolysis, hydrophobic interactions drivelipid A to flip into the upper part of MsbA.At this stage and based on the proposedmodel, it is not obvious how the substrateis released into the outer leaflet of thebilayer. Are lateral, two-dimensional den-sity gradients involved or is it simply adiffusion-controlled process? Furtherstructural and biochemical investigationsare necessary to clarify this point andprove the proposed transport cycle ofMsbA.Of course, Chang and Roth do not

propose that the presented model holdsfor all ABC transporters, especially not forthose transporting hydrophilic substrates.Nevertheless, the ™tilting model∫ derivedfrom the structure might serve as ageneral scheme for MDR-mediating ABCtransporters. However, it has been shownthat LmrA, the MDR1 homologue ofLactococcus lactis, extrudes the substrateinto the extracellular medium[27] and not,like MsbA, into the outer leaflet. Evenfrom a structural point of view, contra-

dicting results exist. The low-resolutionstructure of MDR1 derived from two-dimensional crystals clearly shows a largeextracellular opening (around 25 ä) inMDR1 in the absence of substrate andnucleotide.[18] A similar observation wasmade for MRP1.[20] Additionally, two-di-mensional crystals of MDR1 in differentfunctional states of the NBDs displayedlarge conformational changes of theTMDs.[19] Apart from the extracellularopening, cross-linking studies performedwith MDR have shown that helices of theTMDs are in close proximity,[28, 29] althoughthey are far apart in the structure of MsbA.From these data, Rosenberg et al. deriveda model, in which the binding of ATP isused for substrate translocation from theinner to the outer leaflet.[19] This goes inhand with a reduced affinity of substrate.In contrast, Chang and Roth propose thathydrolysis is employed for the tilting ofthe TMDs while substrate flipping is amore or less spontaneous process. Addi-tional biochemical data indicate that theNBDs strongly interact and act in analternating fashion.[30] No evidence canbe derived from the presented structureto clarify this point.Despite the open questions and differ-

ences between the available structuraldata, it has been now demonstrated thatABC transporters can be crystallized andtheir structure solved by X-ray crystallog-raphy. The MsbA structure is only a first,very important and exciting step towardsfurther understanding the structure andfunction of ABC transporters. Of course,the dimer interface of the TMDs seen inthe structure might not be the biolog-ically relevant one. Further biochemicalanalysis is necessary to prove this, but infavor of the observed interface is the factthat Chang and Roth used only proteinthat corresponded to the dimeric state ofMsbA for the crystallization set-ups. Thelow sequence similarity of ABC transport-ers within the transmembrane regionexplains the different substrate specificitybut might also imply different transportpathways. Such a situation is not in favorof the conservation of structure andfunction in biological systems. Neverthe-less, evolutionarily related ABC transport-ers,[4] such as CFTR and MRP5 or TAP andhemolysin B, display very different sub-strate specificity. Many puzzling questions

in the field of ABC transporters still wait tobe answered. A lot of structural inves-tigations will have to be undertaken untila clear picture of the structure ± functionrelationship of ABC transporters will arise.However, the structure of Chang and Rothwill guide future biochemical and bio-physical studies that will help us tounderstand the molecular mechanismsof the extremely large family of diversemembrane transport proteins.[3]

I would like to apologize to all colleagueswhose work has not been referenced ap-propriately. I thank Carsten Horn, NilsHanekop, Stanislav Gorbulev, and Dr. Ro-bert Tampe¬ (Institute of Biochemistry, Jo-hann Wolfgang Goethe University Frank-furt) for many stimulating and excitingdiscussions about ABC transporters. Ourwork is supported by the Deutsche For-schungsgemeinschaft through the EmmyNoether program and the Graduiertenkol-leg ™Protein function on an atomic level∫(Philipps University Marburg).

[1] G. Chang, C. B. Roth, Science 2001, 293, 1793 ±1800.

[2] S. J. Singer, G. L. Nicolson, Science 1972, 175,720 ± 731.

[3] C. F. Higgins, Annu. Rev. Cell Biol. 1992, 8, 67 ±113.

[4] W. Saurin, M. Hofnung, E. Dassa, J. Mol. Evol.1999, 48, 22 ±41.

[5] J. E. Walker, M. Saraste, M. J. Runswick, N. J.Gray, EMBO J. 1982, 1, 945 ± 951.

[6] G. F.-L. Ames, Int. Rev. Cytol. 1992, 137A, 1 ± 35.[7] M. Ehrmann, R. Ehrle, E. Hofmann, W. Boos, A.

Schlosser, Mol. Microbiol. 1998, 29, 685 ±694.[8] D. C. Gadsby, A. C. Nairn, Physiol. Rev. 1999, 79,

77 ±107.[9] L. Schmitt, R. Tampe¬, ChemBioChem 2000, 1,

16 ±35.[10] M. M. Gottesman, I. Pastan, Annu. Rev. Bio-

chem. 1993, 62, 385 ± 427.[11] H. W. van Veen, W. N. Konings, Semin. Cancer

Biol. 1997, 8, 183 ± 191.[12] H. W. van Veen, W. N. Konings, Biochim. Bio-

phys. Acta 1998, 1365, 31 ± 36.[13] L. W. Hung, I. X. Y. Wang, K. Nikaido, P. Q. Liu,

G. F. L. Ames, S. H. Kim, Nature 1998, 396,703 ± 707.

[14] K. Diederichs, J. Diez, G. Greller, C. Muller, J.Breed, C. Schnell, C. Vonrhein, W. Boos, W.Welte, EMBO J. 2000, 19, 5951 ± 5961.

[15] R. Gaudet, D. C. Wiley, EMBO J. 2001, 20,4964 ± 4972.

[16] N. Karpowich, O. Martsinkevich, L. Millen, Y. R.Yuan, P. L. Dai, K. MacVey, P. J. Thomas, J. F.Hunt, Structure 2001, 9, 571 ± 586.

X-ray Crystal Structure of MsbA

ChemBioChem 2002, 3, 161 ± 165 165

[17] Y. R. Yuan, S. Blecker, O. Martsinkevich, L.Millen, P. J. Thomas, J. F. Hunt, J. Biol. Chem.2001, 276, 32313 ± 32321.

[18] M. F. Rosenberg, R. Callaghan, R. C. Ford, C. F.Higgins, J. Biol. Chem. 1997, 272, 10685 ±10694.

[19] M. F. Rosenberg, G. Velarde, R. C. Ford, C.Martin, G. Berridge, I. D. Kerr, R. Callaghan, A.Schmidlin, C. Wooding, K. J. Linton, C. F.Higgins, EMBO J. 2001, 20, 5615 ± 5625.

[20] M. F. Rosenberg, Q. Mao, A. Holzenburg, R. C.Ford, R. G. Deeley, S. P. Cole, J. Biol. Chem.2001, 276, 16076 ± 16082.

[21] R. M. Garavito, D. Picot, Methods 1990, 1, 57±69.

[22] R. M. Garavito, D. Picot, P. J. Loll, J Bioenerg.Biomembr. 1996, 28, 13 ± 27.

[23] K. Szabo, E. Welker, Bakos, M. Muller, I.Roninson, A. Varadi, B. Sarkadi, J. Biol. Chem.1998, 273, 10132 ± 10138.

[24] S. Gorbulev, R. Abele, R. Tampe¬, Proc. Natl.Acad. Sci. USA 2001, 98, 3732 ± 3737.

[25] G. D. Eytan, R. Regev, Y. G. Assaraf, J. Biol.Chem. 1996, 271, 3172 ±3178.

[26] A. B. Shapiro, V. Ling, Eur. J. Biochem. 1998,254, 189 ± 193.

[27] H. W. van Veen, A. Margolles, M. Muller, C. F.Higgins, W. N. Konings, EMBO J. 2000, 19,2503 ± 2514.

[28] T. W. Loo, D. M. Clarke, J. Biol. Chem. 1996, 271,15414 ± 15419.

[29] T. W. Loo, D. M. Clarke, J. Biol. Chem. 1996, 271,27482 ± 27487.

[30] A. E. Senior, Acta Physiol. Scand. 1998, 163,213 ± 218.

[31] P. J. Kralaukis, J. Appl. Crystallogr. 1991, 24,946 ± 950.

[32] E. A. Merritt, D. J. Bacon, Methods Enzymol.1997, 277, 505 ±524.