ABC (ATP-binding cassette) transporters are primary active membrane proteins that translocate solutes (allocrites) across lipid bilayers. The prototypical ABC transporter consists of four domains: two cytoplasmic NBDs (nucleotide-binding domains) and two TMDs (transmembrane domains). The NBDs, whose primary sequence is highly conserved throughout the superfamily, bind and hydrolyse ATP to power the transport cycle. The TMDs, whose primary sequence and protein fold can be quite disparate, form the translocation pathway across the membrane and generally (but not always) determine allocrite specificity. Structure determination of ABC proteins initially took advantage of the relative ease of expression and crystallization of the hydrophilic bacterial NBDs in isolation from the transporter complex, and revealed detailed information on the structural fold of these domains, the amino acids involved in the binding and hydrolysis of nucleotide, and the head-to-tail arrangement of the NBD–NBD dimer interface. More recently, several intact transporters have been crystallized and three types have, so far, been characterized: type I and II ABC importers, and ABC exporters. All three are present in prokaryotes, but only the ABC exporters appear to be present in eukaryotes. Their structural determination has provided insight into the mechanisms of energy and signal transduction between the NBDs and TMDs (i.e. between the ATP- and allocrite-binding sites) and, for some, the nature of the allocrite-binding site(s) within the TMDs. In this chapter, we focus primarily on the ABC exporters and describe the structural, biochemical and biophysical evidence for and against the controversial bellows-like mechanism proposed for allocrite efflux.
ABC (ATP-binding cassette) proteins form one of the largest and most diverse protein superfamilies, whose members can be found in all prokaryotic and eukaryotic organisms studied to date. The vast majority of ABC proteins are transporters that translocate allocrites across biological membranes. (ATP is the true substrate of ABC transporters. The term ‘allocrite’ was coined by Blight and Holland  to describe the ‘transport substrate’, the compound that is translocated, but not changed chemically.) A typical ABC transporter consists of four domains, which together form the minimum functional unit necessary for transport and comprise two TMDs (transmembrane domains), containing multiple MSHs (membrane-spanning α-helices) that form the translocation pathway across the membrane, and two NBDs (nucleotide-binding domains) that utilize the energy released by ATP binding and hydrolysis to power allocrite transport.
Domain organization: complexes compared with multidomain proteins
ABC transporters vary in their modular composition and can consist of any domain arrangement between a complex of four separate single-domain polypeptides to a single four-domain polypeptide. For example, the histidine permease from Salmonella enterica serotype Typhimurium has a core complex of four polypeptides with two NBDs and two TMDs, whereas in the Escherichia coli ribose transporter RbsA, the two NBDs are fused in a single polypeptide and form a complex with two separate TMD polypeptides. Alternatively, the transporter may comprise two half-transporters with one NBD and one TMD per polypeptide. In some cases, these half-transporters homodimerize, as observed for the bacterial multidrug transporter LmrA, and in others they heterodimerize, as is the case for the mammalian transporter associated with antigen presentation TAP1/TAP2 (described in Chapter 13). There are also examples of four-domain proteins, such as the mammalian multidrug-resistance pump, ABCB1 [also known as P-glycoprotein/MDR1 (multidrug resistance 1)]. ABCB1 is believed to have arisen from a half-transporter by a gene-duplication event, resulting in the domain order TMD–NBD–TMD–NBD.
In some ABC transporters, the core domain architecture is augmented by additional domains which may add functionality. For example, prokaryotic ABC importers require a specific SBP (solute-binding protein), present in the periplasm of Gram-negative bacteria, to bind allocrite with high affinity and deliver it to the transporter complex. Maltose is imported by such a mechanism in E. coli, and the function of the transporter complex MalFGK2 with its SBP MalE is described in detail in Chapter 5. The human CFTR (cystic fibrosis transmembrane conductance regulator) (ABCC7) also has an extra domain, the regulatory (‘R’) domain, separating the two halves of the transporter. Phosphorylation of the R domain serves to regulate Cl− channel activity of CFTR . Other ABC transporters of the mammalian ABCC subfamily, such as MRP1 (multidrug resistance-associated protein 1) (ABCC1) and SUR (sulfonylurea receptor) (ABCC8/9), have an additional N-terminal TMD (TMD0), which is often smaller than the core TMDs. Whereas this additional TMD has been shown to modulate the activity of SUR , its deletion from ABCC1 has little effect on function, although the linker region between TMD0 and TMD1 is important for activity .
The NBDs of different ABC transporters share a high degree of amino acid sequence identity, primarily in seven conserved motifs, reflecting commonality of the ATP-powered mechanism [5,6]. In contrast, the TMDs are often very dissimilar, and this is mirrored in the wide range of allocrites that can be transported by different members of the superfamily. Most individual ABC transporters are highly specific, recognizing a very limited number of allocrites, but some, such as ABCB1, are unusually promiscuous and transport many different allocrites . The determinants of specificity compared with promiscuity are still not clearly understood and are not simply the properties of the TMDs to which the NBDs are associated (for example, ABCB4, the very close relative of ABCB1, does not transport drugs, but flops the lipid phosphatidylcholine into the bile duct ). However, the combination of detailed biochemical and biophysical analyses, allied with the structural data that are becoming available, is beginning to shed light on their molecular mechanism(s).
History of structure determination in the ABC field
Crystal structures of isolated NBDs
Determining the functional dynamic mechanism of any protein is greatly aided by knowledge of its molecular structure. The acquisition of such structural information is particularly difficult for membrane proteins because of technical limitations in expression and purification, and their inherent amphipathicity (often requiring the use of detergent/lipid mixtures to maintain fold and function during purification) impairs functional, biophysical and structural studies. However, the modular nature of ABC transporters offered one solution: the NBDs could be expressed independently as soluble domains. Thus the earliest crystal structures, which remain the highest-resolution structures we have for this superfamily, are of isolated NBDs. The first, published in 1998 by Hung et al. , was the NBD of the histidine permease, HisP, from Salmonella Typhimurium. This 1.5 Å (1 Å=0.1 nm) structure, co-crystallized with ATP, provided detail of the interaction between the substrate and many of the conserved NBD motifs, but shed little light on the dimeric nature of the domains. Subsequently, the crystal structures of isolated NBDs from Rad50 (a two-NBD polypeptide involved in DNA repair), the maltose importer (MalK), haemolysin exporter (HlyB) and several transporters from archaea (often of unknown function) have revealed the true nature of the NBD–NBD dimer, in which the two monomers juxtapose with head-to-tail arrangement, sandwiching two molecules of nucleotide at the interface [10–17].
The NBD dimer and ATP binding
Each NBD has an L-shaped topology, with two subdomains (Figure 1). The F1-type ‘core’ subdomain contains six of the conserved motifs: the Walker A and B motifs, together with the A-, Q-, D- and H-loops. The α-helical subdomain contains the ABC signature motif, the presence of which is diagnostic for members of the ABC protein superfamily. When the NBDs are arranged in their head-to-tail dimer, these motifs are bought into close proximity at the interface between the two NBDs. Mutant NBDs with changes to key motifs that prevented ATP hydrolysis, allowed co-crystallization of MJ0796 NBD dimers with ATP in what appears to be an otherwise hydrolytic conformation . This was followed by the publication of crystal structures of mutant HlyB and wild-type MalK NBD dimers in the presence of ATP [11,17]. In these crystals, two molecules of ATP are observed at the NBD–NBD interface with all motifs involved in co-ordination, except for the D-loop (which seems to form a direct hydrogen bond between the NBDs). Each nucleotide-binding pocket is a composite site with a single nucleotide co-ordinated between the core subdomain of one NBD and the α-helical subdomain of the apposed NBD.
X-ray structures of full-size ABC transporters
The first crystal structure for a complete ABC transporter (with both NBDs and TMDs), the E. coli vitamin B12 importer, BtuCD, was solved in 2002 . This was followed by the first structure for a complete ABC exporter, Sav1866, from Staphylococcus aureus in 2006 . Since then, there has been a relative flurry of complete ABC transporter crystal structures, including the putative metal importer HI1470/71 from Haemophilus influenzae , the molybdate/tungstate importer ModBC from Archaeoglobus fulgidus  and Methanosarcina acetivorans , the maltose importer MalFGK2 from E. coli , the lipid floppase MsbA (an exporter) from several sources (Salmonella Typhimurium, E. coli and Vibrio cholerae) , the methionine importer MetNI from E. coli , and the ABCB1 homologue, Abcb1a, from Mus musculus . The different structural architecture of the TMDs indicates that there are at least three groups of ABC transporter (more are likely to be revealed as structure determination continues): type I ABC importers (ModBC, MalFGK2 and MetNI), type II ABC importers (BtuCD and HI1470/71) and ABC exporters (Sav1866, MsbA and Abcb1a) (Figure 2).
The TMDs of the type I ABC importers share a conserved core of five MSHs per TMD, as found in MetI (Figure 2C), although some have additional helices (MalF and MalG have six and eight MSHs respectively; Figure 2B). The two type II importers, BtuCD (Figure 2D) and HI1470/71 (Figure 2E), both have ten densely packed MSHs per TMD, arranged with MSH2 central to each domain and surrounded by the other helices. Together, the two TMDs of the type II importers form a 20-helix complex within the membrane, featuring an apparent gated central channel at the interface. The solute importers are described in detail by Amy Davidson in Chapter 5, so here we concentrate on the final group, the ABC exporters.
In ABC exporters, each TMD contains six MSHs, and the two most complete crystal structures (i.e. with most residues resolved and at the highest resolution), Sav1866 (Figure 2F) and Abcb1a (Figure 2H), are solved in distinct conformations (notwithstanding the controversy over the Abcb1a structure; see below). The architecture of the exporters is distinctly different in several key features compared to the importers. First, the MSHs of ABC exporters extend much further into the cytosol (approximately 25 Å) to contact the NBDs. Secondly, the MSHs cross the membrane in bundles of six helices, and the composition of these bundles is not limited to the six MSHs of each TMD. This is the phenomenon described as ‘domain swapping’ by Dawson and Locher , when they first described the structure of ADP-bound Sav1866. In Sav1866, MSH1 and MSH2 from one TMD cross the membrane in close association with MSH3, MSH4, MSH5 and MSH6 from the second TMD (in a reciprocal arrangement, the remaining six MSHs of the transporter also cross the membrane in a closely associated bundle). Intriguingly, in the nucleotide-free structure of Abcb1a, which may describe a different conformation adopted by the transporter during the transport cycle (see below), there is also domain swapping, but the composition of the two helix bundles is different (MSH4 and MSH5 of TMD1 now closely associate with MSH1, MSH2, MSH3 and MSH6 of TMD2, and so on) . One important consequence of this domain swapping is that the MSHs are not perpendicular to the plane of the membrane and this allows the two intracellular loops of each domain to interface with different NBDs (Figure 3). This is a marked departure from both types of ABC importer where each TMD only interfaces with a single NBD through a single intracellular loop from each TMD.
As mentioned above, the NBDs of ABC transporters are highly conserved among members of the family, whereas primary sequence similarity between the TMDs is low. In spite of this, the NBDs and TMDs must be mechanistically coupled and conduits must exist to allow energy transduction between the domains. The likely conduits, the coupling helices, formed at the ends of the intracellular loops of the TMDs, share little or no primary sequence similarity, but do share secondary structure elements that are conserved throughout ABC transporters [19,21,23,24].
The coupling helices are short α-helices, preceded and followed by tight turns that are located at the apices of the intracellular loops. They interact with grooves on the surface of the NBDs, to allow the bidirectional transmission of mechanical energy. For example, structural rearrangements caused by ATP binding and hydrolysis at the NBDs must be transduced into allocrite translocation from the TMDs, and likewise, the energy released by binding of allocrite to the TMDs must be transmitted to the NBDs to enable control over the ATP catalytic cycle. In ABC exporters, there are two coupling helices per TMD: coupling helix 1 is formed by the intracellular loop between MSH2 and MSH3, and coupling helix 2 is formed by the intracellular loop between MSH4 and MSH5. Coupling helix 1 contacts the directly apposed NBD in close proximity to the Walker A motif (i.e. coupling helix 1 in TMD1 contacts NBD1, whereas the equivalent helix in TMD2 contacts NBD2). In contrast, coupling helix 2 interfaces with a groove between the two subdomains of the diametrically apposed NBD (i.e. coupling helix 2 of TMD1 contacts NBD2, and the equivalent helix of TMD2 contacts NBD1), demonstrated biochemically by cysteine cross-linking using ABCB1 . The groove in the NBD, at the juncture of the subdomains, is formed by the Q-loop which is therefore positioned such that it could control allosteric coupling between the two subdomains of the NBD (and therefore both nucleotide-binding pockets), as well as the allosteric interactions between the NBDs and the allocrite-binding sites of the TMDs. Bacterial ABC importers have only a single coupling helix which, even before structural data, had been identified as containing a primary sequence motif, the EAA box, conserved within the TMDs of a subset of importers , and had been shown to be important for function. Coupling helix 2 in ABC exporters is structurally equivalent to the single coupling helix found in the importers, which likewise interacts with residues in and around the Q-loop of the NBD subunit, although in importers this is the NBD directly apposed to the respective TMD, rather than diametrically apposed. Crucially, each TMD in the ABC exporters therefore interacts with both NBDs, whereas in the importers, each TMD only interfaces with a single NBD.
Is the Abcb1a structure correct?
Sav1866 and Abcb1a are homologous proteins. Sav1866 crystallized with its NBDs in a close dimeric arrangement with ADP in both nucleotide-binding pockets, and a cavity formed by the TMDs that is open towards what would be the extracellular environment. The overall shape of Sav1866 describes a regular ‘V’. On the other hand, the crystal form of Abcb1a describes an inverted ‘V’ with the NBDs completely detached from one another, such that their only possible interaction is indirect via the TMDs. To move from the Abcb1a conformation to the Sav1866 conformation would be similar to the closure of a bellows, and is a possible mechanism of action for the ABC exporters (the proposed bellows-like mechanism is described in more detail below). However, the ‘inverted V’ conformation of Abcb1a was unexpected and remains controversial, so it is worth asking whether such a conformation is physiologically relevant or is merely an artefact of crystallization? The major problems with the observed conformation can be condensed in the knowledge that the NBDs of ABC transporters are known to interact (and should interface directly to bind nucleotide), and that too much of the surface area of the protein is exposed (which is likely to be why, in molecular dynamics simulations, it is unstable).
It is also true that protein crystals are produced under non-physiological conditions that may promote artefactual protein–protein crystal lattice interactions (the most substantial crystal contacts are between the NBDs of different Abcb1a molecules that are not in the same plane and these cannot be physiologically relevant). It is therefore important to test the validity of models based upon crystallographic data using biochemical and biophysical techniques that can be performed on protein under conditions in which it is shown to have functional activity, and it is well worth asking whether such data are available for ABC proteins.
There are five lines of evidence that lend some credence to the Abcb1a model.
(i) The protein, when solubilized from the crystal, retains drug-stimulated ATPase activity. However, it is possible that the protein can snap back into a physiologically relevant conformation once it is released from its crystal lattice, so the best that can be said is that the protein in the crystal lattice is not irreversibly inactivated.
(ii) The protein co-crystallizes with inhibitors bound in, essentially, the same conformation (see below).
(iii) Electron microscopic analysis of purified hamster Abcb1 and cysteine-free human ABCB1 provides evidence of gross nucleotide-dependent conformational changes in the transporter (while these data involve a different crystal lattice, two-dimensional rather than three-dimensional, and analysis by different techniques, the same caveat exists as for point i) [29,30].
(iv) Electron microscopic analysis of a member of the mammalian ABCC family (an exporter of the products of drug metabolism) in the apo state also shows evidence of a large separation between the NBDs .
(v) EPR (electron paramagnetic resonance) studies on the lipid A exporter MsbA, in a lipid environment, using functional protein, provides evidence (one data point) of the distance between two amino acids that is consistent with the protein adopting a similar conformation in the apo state . (EPR is a biophysical technique that has many applications within protein biochemistry. We refer to experiments where EPR has been used to make distance measurements between specific points within the lipid transporter, MsbA. This involved ‘spin-labelling’ several pairs of amino acids along the MSHs and NBDs. Spin-labels are molecules that contain unpaired electrons, which will align their spin-state either parallel or antiparallel to an applied magnetic field. The energetic difference between these states is, essentially, what is measured using an EPR spectrometer. When brought into close proximity, the unpaired electrons in two spin-labels exhibit a dipole–dipole interaction and this has a measurable effect on the energetic difference between spin states. The strength of this interaction is inversely proportional to the distance between the two spin-labels and so can be used to infer the distance between the amino acids in the protein.)
The last point, while controversial, is perhaps the most compelling to date. Crystal structures of the lipid A exporter MsbA have been solved, but some (the apo state) are even more extreme than that observed for the apo state of Abcb1a. Apo MsbA crystallizes with a large gap of ~50 Å between the NBDs, but the resolution is poor (5.3 Å) and the protein is modelled in a state which would also require significant rotation of the NBDs, with respect to the TMDs, to form the nucleotide-binding pockets, a state widely considered to be non-physiological. Nevertheless, the unusual conformation prompted testing by a significantly less invasive technique. Using EPR, it was possible to record distance measurements between several pairs of paramagnetic groups specifically introduced at positions along the length of MsbA. The results suggest that, before binding nucleotide, the MsbA TMDs and intracellular loops are angled away from each other, forming a large cavity in the membrane that extends into the aqueous phase (Figure 4). These findings would be consistent with both the tweezers-like mechanism proposed for the type I ABC importers described in Chapter 5 and the bellows-like mechanism suggested for the ABC exporters. However, one cysteine residue was introduced towards the base of the NBD to provide a pair of cysteine residues in the MsbA homodimer and a single data point that could be used to discern between a tweezers-like and bellows-like mechanism of action (in the tweezers-like model, the NBDs remain in direct contact throughout the transport cycle via a regulatory domain that acts as a hinge at the base of the NBD–NBD dimer). The data recorded were more consistent with a bellows-like mechanism as the distance between the spin-labelled cysteine residues was much larger in the apo form than a form trapped by nucleotide in the closed NBD–NBD dimer conformation. Measuring the distance between the α-carbons of the equivalent residues in the models of Abcb1a and Sav1866 suggests a closure of 12 Å during the transport cycle, but the empirical measurements made in MsbA by EPR involved a change of 28 Å. This 28 Å is the best evidence that we have for a bellows-like mechanism for the exporters, but the discrepancy between the observed and expected results is difficult to explain and may even suggest that the Abcb1a crystal conformation does not represent the full extent of separation of the NBDs that can be achieved. The implied veracity of the bellows-like mechanism must be viewed with some caution, given the technical difficulty of the experiments involved and only a single data point able to discriminate between the two mechanistic models proposed. Nevertheless, the EPR study serves to illustrate the potential power of this technique to test structural and mechanistic models. More data are certainly required before the scientific community is persuaded by a bellows-like mechanism for the ABC exporters but, at present, it is the simplest interpretation of the available empiric data. The scientific community far from universally accepts a bellows-like mechanism for the ABC exporters. For example, Jones and George  have modelled the movement of NBDs in silico (albeit in the absence of TMDs and lipid) and argue cogently for a constant contact mechanism in which the NBDs remain in direct contact throughout the transport cycle, oscillating around the vertical axis to allow alternating access to the two nucleotide-binding pockets. This model certainly has merit, but may be more likely to describe the mechanism of a subset of ABC transporters, such as CFTR, which have one degenerate nucleotide-binding pocket, to which ATP may remain bound while several rounds of hydrolysis occur in the other pocket.
The basis of the polyspecificity of a multidrug exporter
Abcb1a was crystallized in the apo state, and also co-crystallized with inhibitors, providing an insight into the promiscuous nature of the transporter. Both co-crystallized forms adopt the ‘inverted V’ conformation of the apo state. The inhibitors, stereoisomers of a cyclic hexapeptide, bind within the cavity formed by the TMDs. This cavity, the suggested drug-binding pocket, formed within the lipid bilayer by Abcb1a has been estimated to describe a volume of more than 6000 Å3, theoretically big enough to simultaneously accommodate multiple allocrites . Indeed, one of the Abcb1a crystal forms shows two molecules within the binding pocket of the protein. Interestingly, the second inhibitor-bound structure shows a single molecule bound at a distinct, but overlapping, position within the binding pocket. This is consistent with earlier biochemical and pharmacological data suggesting that Abcb1a is capable of binding multiple compounds, that combinations of drugs can produce non-competitive inhibition, and that the addition of one compound can stimulate transport of another [34–36]. The emerging model for allocrite recognition by multidrug transporters such as ABCB1 is that polyspecificity arises from relatively non-specific electrostatic and van der Waals interactions between the planar amphipathic drugs and the binding pocket of the protein [37,38], in contrast with the lock-and-key mechanism of less promiscuous enzymes. ABCB1 has been shown to bind to drugs directly from the inner leaflet of the membrane . Weak electrostatic interactions are particularly favourable within the lipid environment of the plasma membrane owing to the absence of water. This is also reflected in the observation that drugs able to accept hydrogen bonds are more likely to be allocrites of ABCB1 .
The fact that ABCB1 resides in the plasma membrane and is able to bind allocrites directly from the lipid phase means that the composition and organization of the lipid bilayer will influence the behaviour of the transporter. Mammalian cells are capable of synthesizing thousands of different membrane lipids and the lipid composition of different cell types can vary greatly . The lipids in the plasma membrane (and the membrane surrounding most organelles) are also arranged asymmetrically between the two leaflets, and this will inevitably influence both the partitioning and rate of diffusion of hydrophobic drugs within each leaflet, and also the rate of flip–flop between leaflets. This greatly complicates the pharmacokinetic characterization of ABCB1. Reconstitution of purified ABCB1 into proteoliposomes with the lipids arranged asymmetrically has never been achieved (and may not be possible), and there are few tools available that allow the lipid composition of cellular systems to be manipulated. However, the cholesterol content of cellular and in vitro systems is more amenable to manipulation. Several groups have reported that cholesterol depletion can reduce transporter activity [42–44], whereas addition of cholesterol increases apparent ABCB1 activity [44,45]. However, the mechanism by which cholesterol exerts this effect remains the subject of debate within the field. It has been suggested that the requirement for cholesterol is related to drug recognition by ABCB1, in that small drugs may be unable to efficiently stimulate the conduits linking the drug-binding cavity to the ATP-binding pockets unless cholesterol also occupies some of the volume . Other researchers, however, argue that the enhancement in apparent ABCB1 function by cholesterol is an indirect result of changes in the partitioning of hydrophobic drugs within the cholesterol-rich or cholesterol-poor membrane [47,48]. These data serve to illustrate that membrane proteins in general and ABC transporters in particular need to be studied in the context of the lipid bilayer and the nature of the bilayer, whether in vitro, ex vivo or in vivo, can have a profound effect on the observations made.
The bellows-like model of ABC exporter function
If we assume the veracity of both the Abcb1a and Sav1866 structural models (and the available biochemical and, albeit limited, biophysical data suggest that we should), then the transport cycle should involve a switch between the two forms. At its heart is an ATP switch that describes how the ATP-catalytic cycle of the NBDs and drug transport cycles of the TMDs are coupled mechanistically [49,50] (Figure 5). The model can be summarized in four steps, and as both of these transporters are drug pumps, it is described in terms of drug binding, although it is expected that the transport cycle would be similar with only minor variance needed to explain the idiosyncrasies of other transporters with different allocrites.
Step I: drug binding to an inward-facing high-affinity drug-binding pocket initiates the transport cycle. Drugs are likely to access ABCB1 directly from the inner leaflet of the membrane. In the absence of drug, the TMDs are likely to negatively regulate the NBDs (this has been demonstrated in the relationship between the NBD of the maltose importer, MalK, and its cognate TMDs, MalF and MalG ), thus reducing the likelihood of futile ATP hydrolysis. It is possible that the ‘inverted V’ conformation of the apo form of Abcb1a may be required to fully open the drug-binding pocket to accept large drug molecule(s) (or smaller drugs plus cholesterol), and that, to achieve this, the exporters have evolved the second coupling helix per TMD to remain in direct contact with both NBDs at all times during the transport cycle. The binding of drug(s) to the TMDs is postulated to induce conformational changes that are conveyed to the NBDs via the coupling helices to promote closure of the space between the NBDs.
Step II: the switch to the ‘closed NBD dimer’ generates an outward-facing low-affinity drug-binding pocket. In the presence of ATP, structural data indicate that ABC transporters adopt a closed NBD dimer conformation with two molecules of ATP sandwiched at the interface. Martin et al.  showed that binding of a non-hydrolysable ATP analogue to ABCB1 lowered the affinity for vinblastine although the capacity remained the same, strongly suggesting that binding of the nucleotide is sufficient for drug transport. Studies on the bacterial exporter LmrA demonstrated that the high-affinity site faced the inner leaflet and the low-affinity site faced the extracellular surface . The mechanistic details of NBD closure remain obscure (particularly if the apo state resembles the Abcb1a model, as this would demand a very large conformational change to adopt the conformation described by Sav1866). Other important details are also lacking; for example, it is not known whether ATP binds, in vivo, to the individual NBDs in the apo state with the conformation of the TMDs preventing NBD dimerization, or whether stimulation of the coupling helices promotes ATP binding to the NBDs and the allosteric coupling of the two nucleotide-binding pockets (first suggested by Smith et al. , and subsequently demonstrated experimentally by Zaitseva et al. , as one role of the Q-loop when describing the dimerization of isolated NBDs in solution).
Step III: ATP hydrolysis initiates dissolution of the closed NBD–NBD dimer. In solution, bacterial NBDs dimerize transiently to hydrolyse ATP. The closed NBD dimer conformation of the full transporter may therefore represent an autocatalytic intermediate in the transport cycle, although it has also been hypothesized that drug release might trigger a conformational change in the TMDs which, when transmitted to the NBDs, leads to ATP hydrolysis . ABCB1 requires both NBDs to be catalytic but ATP hydrolysis appears to be non-simultaneous at the two nucleotide-binding pockets, suggesting, but not proving, that two ATPs are hydrolysed per transport cycle. Some ABC exporters such as ABCB2/3 (TAP1/2) and the ion channel CFTR probably only hydrolyse a single ATP in each cycle [54,55]. The simplest hypothesis is that one or two ATPs are hydrolysed in a transporter depending on the amount of energy required to destabilize the particular closed NBD–NBD dimer. However, other theories have been postulated, including ATP hydrolysis alternating between the two nucleotide-binding pockets to control different stages of the transport cycle. A more detailed discussion of the catalytic cycle of ABC transporters is given in Chapter 4.
Step IV: Pi and then ADP are released, restoring the transporter to its basal state. Following ATP hydrolysis, Pi dissociates from the NBD dimer first, as demonstrated by the ability of vanadate to trap ADP in the pocket and arrest the transporter in a post-hydrolytic state. ADP release then follows, but exactly how this happens again remains contentious and may be different for ABC transporter subtypes. For example, the non-transporting ABC protein MutS (involved in DNA repair) appears to retain at least one molecule of ADP until DNA binding triggers conformational change and the replacement of ADP with ATP .
Structural data, interpreted in the light of biochemical and biophysical data, have been instrumental in the formulation of mechanistic models of how ABC transporters move their allocrites across lipid bilayers. The current structural data provide evidence that appears to fulfil the predictions of Oleg Jardetzky  that a transporter should cycle between two states (for an exporter, an inward-facing state with high affinity, and an outward-facing state with low affinity). These states have been observed in ABC transporters (both in crystal form and in the test tube) and data are available to couple the states to the biochemistry of the ATP catalytic cycle. The bellows-like model, which incorporates an ATP switch that controls the transition between the open and closed conformations of the NBDs in an allocrite- and ATP-dependent manner, is attractive, perhaps because of its relative simplicity. It also has the flexibility to account for the idiosyncrasies of different ABC transporters while maintaining a common general mechanism.
• The majority of ABC protein family members are integral membrane proteins that act as primary active transporters of solutes (allocrites) across biological membranes.
• Structure determination suggests that there are at least three different types of ABC transporter: Type I and II ABC importers, and ABC exporters.
• The typical ABC transporter contains four domains: two TMDs and two NBDs. Additional domains may be present and can have a regulatory role.
• The primary sequence of the NBDs is highly conserved and defines the ABC transporter family.
• The TMDs share little sequence identity across the whole family and can differ in the number and length of their MSH.
• The two NBDs are aligned in a head-to-tail arrangement to generate two composite nucleotide-binding pockets.
• Each TMD in an ABC importer couples to a single NBD through a coupling helix formed at the apex of an intracellular loop.
• Each TMD in an ABC exporter forms two coupling helices and makes contact with both NBDs.
• ABC exporters may function via a bellows-like mechanism and cycle between the conformations described by the structures of the drug transporters Abcb1a and Sav1866
• In the bellows-like model, allocrite binds to a high-affinity inward facing site on the TMDs and induces the NBDs to dimerize. The NBDs come together around two molecules of ATP. This ‘ATP switch’ induces a conformational change in the TMD allocrite-binding site, which is reoriented to face outward and now be of low affinity. Allocrite is released and ATP hydrolysis initiates resetting of exporter.
- © The Authors Journal compilation © 2011 Biochemical Society