Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Introduction

ATP-binding cassette (ABC) proteins serve many functions, including the transport of nutrients into the cell, transport of compounds across organellar membranes, the secretion of proteins, antigen (peptide) presentation, cell volume regulation, regulation of protein synthesis, detoxification, and antibiotic resistance. The vast majority of ABC proteins are part of complexes that mediate transport of molecules across cellular or organellar membranes. A smaller group of ABC proteins is associated with soluble (supra) molecular complexes and involved in DNA repair, recombination, chromosome condensation and segregation, and translation elongation. Regardless of whether the ABC proteins are found in membrane transport or soluble (supra) molecular complexes, they provide a power stroke in which chemical energy is converted into work (e.g., for a translocation or dislocation event).

ABC transporters are subdivided into importers and exporters. Both importers and exporters consist of two transmembrane domains (TMDs) and two cytoplasmic nucleotide-binding domains (NBDs), which power the transport through hydrolysis of ATP. The TMDs and NBDs together form the translocator. In this chapter we focus on ABC importers, which on structural and mechanistic grounds are subdivided into Type I, II, and III importers (Fig. 1) (Ter Beek et al. 2014). The mechanism of transport of Type I and II ABC importers involves the binding and release of substrate from a dedicated extracytoplasmic substrate-binding protein (SBP) and alternating access of the substrate-binding pocket in the translocator domain. Type III importers, also named energy-coupling factor (ECF) transporters, capture ligands via so-called S-components, which are small integral membrane proteins that associate with a transmembrane coupling protein and two NBDs to form a full transporter; the different types of modularity are presented in Box I. The mechanism of transport of Type III importers may involve substrate translocation by toppling of the S-component rather than alternating access of the translocator domain. The current status of the mechanisms of transport by Type I, II, and III importers is presented in separate sections below, but we first introduce the substrate-binding proteins associated with Type I and II importers and a recently proposed classification of these important accessory proteins.

Fig. 1
figure 1

ABC superfamily of transporters: Type I, II, and III importers and exporters. The TMDs are colored orange and yellow, the NBDs light and dark blue, and additional domains which often have a regulatory function, are colored green. The SBPs are in purple. From Ter Beek et al. (2014)

Full size image

Box I: Classification within Type III ABC importers

Like the Type I and Type II ABC importers, the Type III importers consist of two identical or highly similar cytoplasmic NBDs (EcfA and EcfA’) and two TMDs. One of the TMDs is called the S-component and functions as the substrate-binding protein. The other three subunits form the eponymous ECF module and fuel transport by ATP binding and hydrolysis. Based on the chromosomal location of the genes encoding all four subunits, the ECF transporters are subdivided into group I and group II. In group I transporters, the genes encoding all four subunits are present in one operon and under the control of a single promotor; the gene products form a dedicated transporter. In group II transporters, the genes for the ECF module are clustered together, while genes encoding S-components for different substrates are scattered around the genome. These S-components are all able to interact with the same ECF module. This modularity is analogous to the promiscuity of a subset of Type I importers that can interact with different SBPs or have multiple different SBDs fused to their TMD (Fig. 2).

Fig. 2
figure 2

Schematic representation of group I and group II ECF transporters, showing both the chromosomal location of the genes as well as the architecture of the transporters. After Slotboom (2014)

Full size image

Structural Features of Substrate-Binding Proteins

Substrate-binding proteins were first discovered in the periplasm of Escherichia coli (Berger and Heppel 1974), a Gram-negative bacterium, and are still often referred to as periplasmic binding proteins. A recent proteomics study showed that up to 80 % of the proteins of the periplasm of E. coli can be SBPs depending on the growth conditions, with the number of most SBPs increasing with decreasing growth rate (Heinemann et al., unpublished). Thus, cells express more SBPs when nutrient conditions become harsh. In a few cases a limited number of related SBPs can associate with the same translocator (Davidson et al. 2008), enabling the system to import multiple distinct substrates. The first SBP crystal structure, the l -arabinose-binding protein (ABP), was solved in 1974 (Quiocho et al. 1974).

The soluble SBPs of Type I importers are present in large excess over the translocator complexes in the membrane, allowing efficient capture of substrates and initiation of the translocation reaction (Heinemann et al., unpublished). In contrast, the SBPs of Type II importers appear to be stoichiometric with the translocators. In Gram-positive bacteria and archaea, i.e., microorganisms lacking an outer membrane and periplasm, SBPs are exposed on the cell surface and attached to the cytoplasmic membrane via a lipid anchor or a transmembrane peptide (observed in archaea only), or they can be fused to the TMDs. The latter has been observed in hetero- and homodimeric TMDs, and thus results in one or two SBDs per functional complex (van der Heide and Poolman 2002; Schuurman-Wolters and Poolman, unpublished). In some cases, two or even three SBDs fused in tandem are linked to the TMDs and these systems have a total of four or six extracytoplasmic substrate-binding sites (Fig. 3). Systems with SBDs fused to the TMDs can also be found in Gram-negative bacteria but less frequently compared to Gram-positive bacteria (van der Heide and Poolman 2002). The linking of SBDs to the membrane or the fusion of multiple SBDs to the TMD increases the effective concentration of the substrate-binding sites near the translocator and increases the efficiency of transport (Schuurman-Wolters and Poolman, unpublished). The relatively short (10–20 amino acid) flexible linkers allow the SBDs to probe a small volume around the translocator, leading to an effective concentration of the SBDs in the millimolar range. The affinity of the fused SBDs for the translocator is not known, but the dissociation constants might be millimolar or higher. In case of soluble or lipid-anchored SBPs, the affinity for the translocator is ~0.1 mM when the proteins are in the closed-liganded state (Prossnitz et al. 1989; Dean et al. 1992; Doeven et al. 2004). These low affinities may necessitate a high concentration of SBPs in the periplasm [the maltose-binding protein reaches a concentration of ~1 mM; Manson et al. (1985)] or surface-tethering of the proteins. Given that the periplasm is highly crowded and diffusion is slow (diffusion coefficients for proteins of the size of SBPs in the periplasm are even lower than in the cytoplasm; van den Berg et al., unpublished) and synthesizing a large excess of SBPs is costly, it is perhaps surprising that the covalent linking of SBDs is not more widespread or even universal for ABC importers. The mechanism of substrate binding of periplasmic, membrane-anchored, and TMD-fused SBDs is similar and so is the mode of action of the corresponding ABC transporters (vide infra).

Fig. 3
figure 3

Types of Type I ABC importers. The TMDs are colored gray and the associated SBPs or covalently linked SBDs are shown in different shades of gray (from white to dark gray); the NBDs are colored orange and two molecules of ATP are indicated by bars. After van der Heide and Poolman (2002)

Full size image

SBPs consist of two lobes connected by a linker or hinge (domain). The two lobes close and engulf the ligand upon substrate binding (mode of substrate capture akin that of a Venus’s Flytrap) (Quiocho and Ledvina 1996). Structures are available of SBPs in the open-unliganded (O), closed-unliganded (C), open-liganded (OL), and closed-liganded (CL) forms; the latter conformation is thought to productively interact with the translocator complex. Mutational and structural analyses indicate that each lobe of the SBP binds to one of the TMDs (Hollenstein et al. 2007a; Davidson et al. 2008), and conformational changes in the NBDs upon binding of ATP are transmitted via the TMDs to the SBP. Thus, ATP is involved in the opening of the SBPs and substrate transfer to the translocator.

SBPs associated with ABC transporters are related to similar domains present in a wide variety of translocation and signal transduction systems in both prokaryotic and eukaryotic organisms, including tripartite ATP-independent periplasmic (TRAP) transporters, two-component regulatory systems, guanylate cyclase–atrial natriuretic peptide receptors, G-protein coupled receptors (GPCRs), and ligand-gated ion channels (Berntsson et al. 2010). In addition, SBDs are part of prokaryotic DNA-binding proteins involved in gene regulation. SBPs are very diverse in sequence, and phylogenetic analyses based on multiple sequence alignments do not yield stable alignments (the sequence identity of the proteins is often < 20 %). However, the structures of SBPs are remarkably similar, which has been used to cluster the SBPs based on structural similarity instead of sequence similarity (Berntsson et al. 2010).

Pairwise structural alignments of more than 100 SBPs have been used to produce a structural distance tree. The SBPs were found to group into six defined clusters (A–F), three of which (clusters A, D, and F) were further subdivided. The analyses have shown that the proteins within the six clusters can be discriminated on the basis of the linker (hinge) region that connects the two protein lobes (Fig. 4): the generic secondary structure is (β)4/5(α) n -hinge-(β)4/5(α) n (Fukami-Kobayashi et al. 1999). There is little or no correlation between the structural clustering and functional classification, that is, SBPs with very different substrate specificities are present in each cluster. Furthermore, in the individual clusters, the proteins are not necessarily homologous, as judged from the absence of significant sequence similarity. Clusters B and F contain the most diverse SBPs, associated with various types of transport and signal transduction systems; SBPs in cluster A are unique to type II ABC importers; clusters C and D contain SBPs of Type I ABC importers; and SBPs in cluster E are found in TRAP transporters. Specific information on the six clusters is presented in Box II.

Fig. 4
figure 4

The different clusters of SBPs are shown with their distinct structural features colored in orange. a Cluster A contains SBPs having a single connection between the two domains in the form of a rigid helix. b Cluster B contains SBPs with three interconnecting segments between the two domains. c Cluster C contains SBPs that have an extra domain and are significantly larger in size when compared with the others. d Cluster D contains SBPs with two relatively short hinges. e Cluster E contains SBPs associated with TRAP transporters which all contain a large helix functioning as the hinge region. f Cluster F contains SBPs with two hinges similarly like those of cluster D, however, these hinges have almost double the length creating more flexibility inside the SBP. Please note that Clusters A, D, and F can further be subdivided based on the substrate of the SBP (see text). The proteins used to illustrate the features in Cluster A–F are BtuF (PDB code: 1N2Z), RBP (PDB code: 1DRJ), OppA (PDB code: 3DRF), ModA (PDB code: 1ONR), UehA (PDB code: 3FXB), and HisJ (PDB code: 1HSL), respectively. Figure was taken from Berntsson et al. (2010)

Full size image

Box II: Structural features distinguishing substrate-binding proteins

Cluster A: The cluster A SBPs associate with Type II ABC importers. The distinguishing characteristic of these SBPs is an α-helix that serves as a hinge between the two domains. The rigidity of this helix is reflected in the small movement of both domains upon substrate binding. All of the SBPs in cluster A play a role in metal binding, either directly or as metal chelates.

Cluster B: Cluster B consists of SBPs that bind carbohydrates, branched chain amino acids, natriuretic peptides, and autoinducer-2 (AI-2). The SBPs in cluster B interact with type I ABC importers, two-component histidine-sensory complexes, and guanylate cyclase-atrial natriuretic peptide receptors. The hinge of the SBPs in cluster B is built of three distinct regions connecting the lobes. Homologous to the proteins in cluster B are the lac-repressor type transcription factors.

Cluster C: The cluster C SBPs interact with Type I ABC importers and bind diverse ligands including di- and oligopeptides, arginine, nickel ions, and cellobiose. They all have an extra domain, which in oligopeptide transporters extends the binding cavity to accommodate very large ligands (peptides of 20 residues or more).

Cluster D: The discernible feature of these proteins is that their hinge region consists of two short stretches, 4–5 amino acids long. This large group of SBPs binds a large variety of substrates: carbohydrates, putrescine, thiamine, tetrahedral oxyanion as well as ferric or ferrous iron.

Cluster E: These SBPs are part of the TRAP transporter (tripartite ATP-independent periplasmic transporter) family. TRAP transporters use an electrochemical ion gradient to fuel the uphill translocation of substrates. The remarkable feature of TRAP SBPs is a large single ß-strand that is part of the two five-stranded ß-sheets of both lobes. All TRAP-dependent SBPs structurally characterized have a conserved strand order, typical of class II SBPs (β2β1β3βnβ4), and an additional ß-strand connecting both domains. A second distinguishing feature is a long helix that spans both domains, although in some structures it is interrupted by a kink. The known substrates are ectoine, pyroglutamic acid, lactate, 2-keto acids, and sialic acid.

Cluster F: The distinguishing feature of the cluster F proteins is a hinge consisting of two segments connecting the two lobes. The linker stretches of cluster F proteins are significantly longer (8–10 amino acids) than the hinges of 4–5 amino acids observed in SBPs of cluster D. Possibly the longer linker provides more flexibility between the open and closed conformation. Cluster F SBPs bind a large variety of substrates ranging from trigonal planar anions (nitrate, bicarbonate) to amino acids and compatible solutes such as glycine betaine.

Mechanism of Ligand Binding

SBPs have been captured in crystals in different conformations: open (O) and closed (C), either with or without ligand (L). In solution, the equilibrium state of the SBPs is toward the open (O) conformation in the absence of ligand and toward the closed conformation in the presence of ligand (CL). The crystal structures provide crucial snapshots of different conformational states of the SBPs, but they do not reveal the dynamics of ligand binding. Different models for ligand binding are possible. In the induced fit mechanism, the initial interaction between protein and ligand (OL) is relatively weak but triggers the closing of the protein and formation of CL (Fig. 5). In the conformational selection mechanism, the protein samples closed states (partially closed (PC) or fully closed (C) in Fig. 5) without the involvement of a ligand. Ligand binding stabilizes the PC or C-form and thereby drives the equilibrium to the closed form CL. A combination of both mechanisms is also possible, but nowadays it is thought that conformational selection is the dominant mechanism of ligand binding by enzymes and receptors (Vogt and Di Cera 2013; Clore 2014).

Fig. 5
figure 5

Ligand-binding mechanisms. Conformational transitions in SBDs according to the induced fit or the conformational selection binding mechanism, as explained in the main text. V closing, rate of closing of the binding site; τ c , life-time of the closed state; and [L], ligand concentration

Full size image

The two mechanisms make specific predictions about the kinetics of ligand binding, i.e., the closing rate increases with ligand concentration in induced fit but not in conformational selection (Fig. 5). Moreover, in the conformational selection mechanism, the life-time of the closed state increases with ligand concentration. To discriminate between induced fit and conformational selection is generally not possible in ensemble measurements because differences in conformations are averaged out. Single-molecule observations with high time resolution allow the two mechanisms or hybrid forms to be discriminated. Recent single-molecule Förster resonance energy transfer (smFRET) studies of the maltose-binding protein (MalE) (Kim et al. 2013; Seo et al. 2014) and the two SBDs of the amino acid ABC importer GlnPQ (Gouridis et al. 2015) have shown that these proteins bind their ligands by a previously undocumented type of induced fit mechanism: the SBDs intrinsically transit from O to C in addition to being triggered to close by the ligand. Remarkably, as shown for both SBDs from GlnPQ, the life-time of C is the same as that of CL when high-affinity ligands (with sub-micromolar dissociation constants) are bound. In contrast, the life-time of CL is dramatically shortened when low-affinity ligands are bound. This observation indicates that the ligands do not “stabilize” the closed conformation, which is often thought on the basis of the multiple interactions between protein and ligand in crystal structures.

Box III. The GlnPQ transporter

The GlnPQ transporter from Lactococcus lactis and other Gram-positive bacteria has homodimeric TMDs with two SBDs fused in tandem to each of them (Fig. 3, 6th cartoon). The N-terminal SBD (SBD1) is preceded by a signal sequence that is cleaved off following translocation of the SBDs across the membrane (Schuurman-Wolters and Poolman 2005).

Thus, in GlnPQ four SBDs compete for interaction with the translocator to deliver the bound ligand. The close proximity of multiple SBDs, each of which can adopt different conformations (O, C, OL, or CL) and close intrinsically, affects transport (Gouridis et al. 2015). GlnPQ imports asparagine solely via SBD1 (K D  = 0.2 μM), and glutamine via both SBD1 (K D  ~ 100 μM) and SBD2 (K D  = 1 μM), but with affinities that differ by two orders of magnitude (Fulyani et al. 2013). The transport of asparagine has been determined in the absence of glutamine (SBD2 ~ 98 % in O, ~2 % in C), in the presence of relatively low concentrations of glutamine that do not significantly compete with Asn for binding to SBD1 (SBD2 > 95 % in CL), and in the absence or presence of glutamine and a mutation (D417F) that prevents SBD2 from binding ligand and, more importantly, prevents closing (SBD2 ~ 100 in O) (Gouridis et al. 2015). The remarkable observation was made that the D417F mutation increases the rate of asparagine import by a factor of 3, showing that 2 % of C already has a strong inhibitory effect on transport via SBD1. In fact, the inhibition of asparagine transport via SBD1 by 2 % of SBD2 in the C state is comparable to the inhibition by 95 % in the CL.

These experiments indicate that the TMDs interact with the closed state of the SBD, but the interaction is shorter-lasting for CL (leading to a productive translocation cycle) than for C (leading to a long-lived blocked intermediate state). This work illustrates the importance of understanding the mechanism of ligand binding to substrate-binding proteins in order to begin to understand the full translocation cycle. Although not so easily probed in ABC importers with periplasmic or lipid-anchored SBPs, similar kinetics of ligand binding and competition between proteins in different conformation states could play a role in these systems.

Structural Features of ABC Importers

Type I Importers

Crystal structures of Type I importers for four different substrates are available: the molybdate/tungstate transporter ModB2C2 from Archaeoglobus fulgides (Hollenstein et al. 2007b) and Methanosarcina acetivorans (Gerber et al. 2008), the maltose transporter MalFGK2 from E. coli (Oldham et al. 2007, 2013; Khare et al. 2009; Oldham and Chen 2011a, b; Chen et al. 2013), the methionine transporter MetN2I2 from E. coli (Kadaba et al. 2008; Johnson et al. 2012), and the transporter for basic amino acids Art(QN)2 from Thermoanaerobacter tengcongensis (Yu et al. 2015). MalFGK2 is by far the most studied transporter. The different crystal structures are listed in Table 1.

Table 1 The different crystal structures available of ABC importers
Full size table

The crystal structure of MalFGK2 in the inward-facing apo-state is likely representing the resting state of the transport cycle (Khare et al. 2009). The two TMDs of MalFGK2, MalF and MalG, are not identical. They consist of eight and six transmembrane helices, respectively, from which TM3-8 of MalF are related to TM1-6 of MalG. These helices form the core region of the TMDs with pseudo-twofold symmetry. In the inward-facing state, the interface between MalF and MalG creates a translocation cavity, which is accessible from the cytoplasm. On the periplasmic end, this cavity is closed by a periplasmic, hydrophobic gate that is composed of four loops, each at the kink of a transmembrane helix (TM5 and TM7 from MalF, TM3 and TM5 from MalG). Both MalF and MalG have a coupling helix, which is located in loop L6 between TM6 and TM7 of MalF and in loop L4 between TM4 and TM5 of MalG. Both coupling helices dock into a groove on the surface of each of the two NBDs (MalK subunits). These grooves are lined by two helices from the helical subdomain of MalK, the helix following the Walker A motif and residues from the Q-loop. The interactions between the coupling helices and the MalK subunits transduce conformational changes upon ATP binding and hydrolysis to MalF and MalG and allow alternate access of the binding pocket in the TMDs.

Within the MalF–MalG dimer, an occluded maltose-binding pocket is formed about halfway the membrane. In the structures showing maltose-bound MalFGK2 in a pre-translocation state (Oldham and Chen 2011a), this pocket is closed at the periplasmic side by a hydrophobic gate. On the cytoplasmic side, the pocket is closed by a network of van der Waals interactions. The binding pocket is lined by residues from MalF only, and the binding mode resembles that of MBP in terms of aromatic stacking of binding site residues with the sugar rings, as well as a hydrogen-bond network involved in sugar recognition.

A crystal structure of nucleotide-free MalFGK2–MBP in the pre-translocation state with maltoheptaose bound has shed light on the selectivity of sugar transport (Oldham et al. 2013). MalFGK2 imports linear malto-oligosaccharides of a length from two to seven glycosyl units, linked through α-1,4 glycosidic bonds, with an unmodified reducing end. In the crystal structure, four glycosyl units from the reducing end of maltoheptaose are bound in the groove between the N-terminal and C-terminal lobe of MBP. A conserved glutamine in the periplasmic L5 loop of MalG is also inserted into this groove and forms hydrogen bonds with the first glycosyl unit at the reducing end. There is no clear electron density found for the glycosyl units at the non-reducing end of maltoheptaose. In a different AMP–PNP bound, pre-translocation state structure with an outward-facing conformation, three glucosyl units from the non-reducing end of maltoheptaose are bound in the MalF-binding site (Oldham et al. 2013). Aromatic stacking interactions and five direct protein–sugar hydrogen bonds within this binding site indicate specificity for α-1,4 linkage of the glycosyl units. The other glucosyl units of maltoheptaose are not visible, probably because they do not have specific interactions with the transporter. Overall, it seems that the size-exclusion limit of transport by MalFGK2 is determined by the size of the cavity in the occluded state, which is about 2400 Å, just large enough to fit a maltoheptaose molecule.

Five crystal structures of the transporter for basic amino acids Art(NQ)2 from T. tengcongensis in the apo-, arginine-, histidine-, ATP-, and ATP plus arginine-bound state all show an inward-facing conformation, with the nucleotide-binding domains (ArtNs) in a semi-open state (Yu et al. 2015). The transmembrane domains (ArtQs) contain five transmembrane segments, which correspond to TM4-8 of MalF and TM2-6 of MalG, and form a homodimer with twofold symmetry. Together, the ArtQs form a highly negatively charged tunnel reaching from the periplasmic side to the cytoplasmic side of the membrane, which allows positively charged amino acids to pass through. Halfway across the predicted membrane, each ArtQ subunit has a highly negatively charged substrate-binding pocket, lined by highly conserved residues that are involved in specific binding of both arginine and histidine. Remarkably, two amino acids are present in the ArtQ dimer, which raises questions on the transport mechanism as one amino acid at a time is transferred from the substrate-binding protein ArtI to Art(NQ)2. A crystal structure of ArtI in the arginine bound state shows only one binding site. We therefore consider it more likely that the two amino acids have accessed the binding pocket when the protein was trapped in the inward-facing conformation, and that the structures do not reflect a true translocation intermediate.

Capturing MalFGK2 in an ATP-bound state was achieved by mutating the catalytic glutamate in the ATP hydrolysis site to a glutamine (E159Q) (Oldham et al. 2007). In the outward-facing conformation, the translocation cavity is now open to the bound MBP, which is in a substrate-free state (Oldham et al. 2007; Oldham and Chen 2011a, b). On the cytoplasmic side, the translocation cavity is closed by a tightly packed helix bundle of TM6 and TM7 from MalF and TM4 and TM5 from MalG. In the pre-translocation, occluded state, only four residues from these TMs are forming the gate. In the closed, nucleotide-bound MalK dimer, two ATP or AMP-PNP molecules are sandwiched at the dimeric interface, where they interact with residues from the Walker A and Walker B motifs of one MalK subunit and residues from the LSGGQ signature sequence of the other subunit. Compared to the pre-translocation, occluded state, the helical domains of the NBDs have rotated another 15°, thereby breaking the intersubunit hydrogen-bond network of the MalK dimer. Overall, when comparing the inward-facing and outward-facing conformations, the structures of the core helices TM4-7 in MalF and TM2-5 in MalG are maintained as rigid bodies during the transport cycle. The other helices, TM2, TM3, and TM8 of MalF and TM1 and TM6 of MalG, move together with the core helices of the other TMD subunit. The helical domains of the NBDs, as well as the coupling helices, rotate over an angle of about 30° upon closure of the MalK dimer, which allows the conformational change in the MalFG dimer.

MalFGK2 has been crystallized in complex with the MBP, maltose, and ADP together with phosphate analogs VO4 −, AlF4 − , or BeF3 −, which has provided insight into the mechanism of ATP hydrolysis within the MalK dimer (Oldham and Chen 2011b). The overall structures show the same outward-facing conformation as described above, suggesting that ATP hydrolysis does not force the transporter in a different conformation. Even the conserved residues within the ATP hydrolysis site are superimposable on all four structures. The hydrolysis of ATP proceeds by the attack of the γ-phosphate of ATP by a water molecule, which is thought to transit a pentacovalent intermediate state. The VO4 − and AlF4 − bound structures show a trigonal bipyramidal and octahedral geometry, respectively, which represent the transition state of the hydrolysis. Three of the four oxygen atoms of VO4 − or the fluorides of AlF4 − lie in the equatorial plane, while the fourth oxygen of VO4 − or the attacking water molecule in case of the AlF4 − bound structure and the oxygen connecting the β-phosphate group are found at the axial positions. This transition state supports the catalysis-by-a-general-base model, where the attacking water molecule is activated by the conserved glutamate (E159) by polarizing the water molecule. The conserved histidine (H192) in the H-loop positions the γ-phosphate group, the attacking water molecule, and the conserved glutamate, while the conserved glutamine (Q82) in the Q-loop is involved in coordination of the Mg2+ ion.

In the final step before returning to the resting state, the SBP is released from the transporter and the inorganic phosphate and ADP are released from the NBDs. Intermediate states of this step are observed in the crystal structure of the molybdate/tungstate transporter ModB2C2 from A. fulgides, which is in the inward-facing conformation in complex with its substrate-binding protein ModA (Hollenstein et al. 2007a, b), and two crystal structures of ADP-bound MetN2I2 in the inward-facing conformation without the substrate-binding protein MetQ bound (Johnson et al. 2012). The structure of the ModB2C2-A complex is similar to that of MalFGK2 in the resting state, with the NBDs (ModC) in an open, nucleotide-free conformation and the TMDs (ModB) in an inward-facing conformation. ModC adopts a post-hydrolysis conformation and two molecules of inorganic phosphate are bound at the position of the β-phosphate groups of ATP in the ATP-bound structure of MalFGK-MBP (Oldham et al. 2007), suggesting that binding of a phosphate group at this position is probably the strongest compared to the position of the α- or γ-phosphate group.

Two pairs of crystal structures of the methionine transporter MetN2I2 from E. coli are available (Kadaba et al. 2008; Johnson et al. 2012), with the NBDs (MetN) spaced differently between the nucleotide-free and the ADP-bound structures. Similar to the structure of MalFGK2 in the resting state, the MetN dimer is held together by dimerization of the C-terminal domains. Each TMD contains five transmembrane segments, which correspond to TM4-8 of MalF and TM2-6 of MalG. The region of highest sequence and structure similarity is found at the region of the coupling helices and TM3 and TM4 located on either side of them. Similar to the maltose transporter, a network of salt bridges between the coupling helices of MetI and the helical subdomain of MetN relay the conformational changes between the TMD and NBD. The gate that closes the translocation cavity at the periplasmic side of MetN2I2 is structurally conserved and found in the ModB2C2 transporter.

Type II Importers

The Type II importers facilitate the uptake of metal chelates including vitamin B12, heme, and oxanions (Klein and Lewinson 2011). For this group of ABC importers, seven crystal structures of complete transporters are available (Table 1). These structures correspond to five different states, which represent the major steps of the transport cycle. Five crystal structures are of the vitamin B12 transporter BtuCD from E. coli (Locher et al. 2002; Hvorup et al. 2007; Korkhov et al. 2012a, b, 2014), which is the most studied Type II importer. The two other crystal structures are of the molybdate/tungstate transporter MolB2C2A from Haemophilus influenzae (Pinkett et al. 2007) and the heme transporter HmuUV from Yersinia pestis (Woo et al. 2012).

In 2002, the vitamin B12 transporter BtuCD from E. coli was crystallized in an outward-facing conformation, forming a cavity accessible from the periplasmic side (Locher et al. 2002). No substrate or nucleotides are bound and this outward-facing conformation probably represents the resting state of the transport cycle. The BtuC subunit consists of ten transmembrane helices, of which TM2-5 and TM7-10 are related by a pseudo-twofold rotation axis. Within the BtuC dimer, there is a twofold symmetry axis running down the translocation pathway. This translocation pathway is lined by residues from TM5, TM5a (small transmembrane helix following TM5), and TM10 of each BtuC, and the loops preceding TM3 and TM8. In this outward-facing conformation, the translocation cavity stretches out two-thirds into the predicted membrane and is closed by cytoplasmic gate I.

Given the millimolar ATP concentration in the cell, the resting state of the transporter will be short lived. The BtuDs will quickly bind ATP and the transporter will convert into an outward-facing, nucleotide-bound state as represented by the recently solved crystal structure of BtuCD in complex with two molecules of the non-hydrolyzable ATP analog AMP-PNP and two Mg2+ ions (Korkhov et al. 2014). Like in the resting state, the BtuC dimer is in an outward-facing conformation with a translocation cavity open to the periplasmic side but closed at the cytoplasmic side by a second cytoplasmic gate, known as cytoplasmic gate II. The BtuDs form a closed dimer with the AMP-PNP molecules and Mg2+ ions bound at the ATP hydrolysis site; in this state the two coupling helices of the BtuCs are about 9.5 Å closer to each other compared to the resting state. Apparently, the binding of AMP-PNP promotes closure of the BtuD dimer, which is coupled to opening of cytoplasmic gate I and closure of cytoplasmic gate II in BtuC.

In the next step of the transport cycle, the substrate-binding protein BtuF binds vitamin B12 and subsequently docks onto the periplasmic side of the BtuC dimer. A study with BtuCD reconstituted in proteoliposomes has shown that ATP binding promotes docking of BtuF onto the BtuCD complex, rather than simulating BtuF release to scavenge vitamin B12 (Korkhov et al. 2014). Upon docking of BtuF onto the BtuCD complex, the N-terminal and C-terminal lobe of BtuF are spread apart while making interactions with periplasmic loops of BtuC. TM5a of BtuC sticks into the vitamin B12 binding site of BtuF, and this distortion forces vitamin B12 to move into the translocation cavity. Rearrangements within the BtuC dimer will then result in a substrate-bound, occluded state, which has been obtained in the presence of AMP-PNP (Korkhov et al. 2012b). In this state, the BtuD dimer closely resembles the previous state, but TM5 and TM5a of both BtuC subunits have now closed the translocation cavity from the periplasmic side, while cytoplasmic gate II is still closed. Although vitamin B12 is not present in this crystal structure, the cavity is wide enough to host the molecule with only minor steric hindrance. The interior of the cavity does not resemble the substrate-binding pocket of BtuF, rather it forms a low-affinity chamber.

After hydrolysis of ATP, the cytoplasmic gate II in the BtuD dimer will open to release inorganic phosphate and ADP. The substrate may then be squeezed out by a peristaltic-like movement. The state after substrate release is visualized by the crystal structure of the molybdate/tungstate transporter MolB2C2 from Haemophilus influenzae (Pinkett et al. 2007). The transporter was crystallized in a nucleotide-free, inward-facing conformation in the absence of the substrate-binding protein MolA. The arrangement of the transmembrane helices of MolB is similar to that of BtuC. In MolB2C2 only the TM5a of each subunit contributes to the formation of the periplasmic gate. EPR and selective crosslinking studies show that this gate offers a narrower opening upon nucleotide binding compared to the periplasmic gate in BtuCD, which is in agreement with the much smaller substrates molybdate or tungstate as compared to vitamin B12 (Rice et al. 2013).

To return to the outward-facing resting state, BtuF should dissociate and cytoplasmic gate I should close while the periplasmic gate needs to open up. A post-translation, intermediate state was crystallized in two different asymmetric conformations (Hvorup et al. 2007; Korkhov et al. 2012a). While the two non-identical lobes of BtuF interact with the periplasmic loops of the BtuC subunits, a slight asymmetry is introduced in the loops, resulting in an asymmetric arrangement of the transmembrane helices. Especially the orientation of TM3-5a is thought to control on which side of the membrane the translocation cavity is opened. The closure of the periplasmic gate and cytoplasmic gate I ensures that no continuous channel is formed, while switching from the inward-facing to the outward-facing conformation. Once the transporter reaches the outward-facing conformation by opening the periplasmic gate, BtuF will be released and the transporter is back in the resting state.

Type III Importers

Compared to the Type I and Type II importers, the structural information on the Type III importers/ECF transporters is limited. To date, crystal structures of four different S-components (Zhang et al. 2010; Erkens et al. 2011; Berntsson et al. 2012; Yu et al. 2014), two NBD dimers (Chai et al. 2013; Karpowich and Wang 2013), and three complete transporters are available (Xu et al. 2013; Wang et al. 2013; Zhang et al. 2014) (Table 1). The four substrate-bound S-components are RibU for riboflavin from Staphylococcus aureus (Zhang et al. 2010), ThiT for thiamin and BioY for biotin from Lactococcus lactis (Erkens et al. 2011; Berntsson et al. 2012), and NikM2 for Ni2+ for Thermoanaerobacter tengcongensis (Yu et al. 2014).

S-components. RibU, ThiT, and BioY associate with group II ECF transporters and despite a sequence identity of only 14–16 %, these S-components share a common structural fold of a six transmembrane helical bundle. The N-terminal parts consisting of TM1-3 are structurally the most conserved and were predicted to be involved in the interaction with the ECF module. TM1 contains a conserved AxxxA motif, which is important for complex formation and thiamin transport by ECF ThiT from L. lactis (Erkens et al. 2011) and biotin transport by the group I transporter BioMNY from Rhodobacter capsulatus (Finkenwirth et al. 2015). TM4-6 in the C-terminal parts of S-components are structurally more different and are the main contributors to the substrate-binding site. This combination of a structurally conserved part and a highly divergent part explains how different S-components can interact with one and the same ECF module and maintain a high specificity for chemically diverse substrates (dissociation constants in the subnanomolar to nanomolar range). The substrate-binding pocket is located on the predicted extracytoplasmic side of the S-component and closed by the L1 loop connecting TM1 and TM2, which lies over the pocket in a lid-like manner. EPR studies on thiamin-specific ThiT from L. lactis have shown that the main difference between thiamin-bound and substrate-free ThiT is the conformation of this L1 loop, covering the substrate-binding pocket in the liganded state, while exposing it to the environment in the unliganded state (Majsnerowska et al. 2013).

NikM2 is the S-component of the NikM2N2OQ transporter from T. tengcongensis, which is a metal transporter of the group I ECF transporters (Yu et al. 2014). Compared to the three vitamin-specific S-components, NikM2 has an additional N-terminal α-helix. The highly conserved extracellular N-terminus inserts into the center of the S-component and occupies a space that corresponds to the substrate-binding pocket in the vitamin-specific S-components. Within this space, Ni2+ is bound in a square-planar geometry by four nitrogen atoms. The nickel-binding residues are stabilized and oriented by a network of hydrogen bonds.

Full-length transporters. The structures of the complete Type III transporters all show the same conformation, in which the EcfAA’ heterodimer adopts an open conformation and the substrate-free S-components lie in the membrane with the helical axis almost perpendicular to the plane of the membrane (Fig. 6). These transporters have the same ECF module in complex with three different S-components: FolT for folate (Xu et al. 2013), PanT for pantothenate (Zhang et al. 2014), and HmpT; the latter is thought to bind pyridoxine (and should therefore be named PdxU) (Wang et al. 2013). The NBDs have the same structural fold as in Type I and Type II importers, including the RecA-like domain, the helical domain, and a C-terminal domain. All the conserved features (Walker A, Walker B, LSGGQ signature motif, D-, H-, and Q-loop) are present. A C-terminal extra domain provides the basis for dimerization (Chai et al. 2013; Karpowich and Wang 2013). Like the NBDs of the Type I and Type II importers, both dimers have two grooves at the interface of the RecA-like domain and the helical domain within one ATPase subunit, which can form the anchor points of eventual coupling helices. These acidic and highly conserved grooves are followed at the end of the Q-loop by a short 310 helix, which seems to be conserved among the ECF transporters. This Q-helix has a conserved xPD/ExQφ (φ is a hydrophobic residue) motif, in which the conserved acidic residue and the conserved glutamine point to opposite sides of the helix. The latter points into the groove and is therefore thought to play a role in the interaction with transmembrane domains. The structures of the complete transporter also reveal a L-shaped EcfT consisting of eight α-helices, of which five are transmembrane helices (TM1-4 and TM8). Helix 5 lies at the cytoplasmic side of EcfT almost perpendicular to the transmembrane helices and helices 6 and 7 form two cytoplasmic coupling helices, which stick with their C-terminal ends into the grooves at the surface of EcfA’ and EcfA, respectively. Together these coupling helices form a X-shaped bundle, which is stabilized by extensive hydrophobic interactions and hydrogen bonds between the conserved Glu174 in helix 6 and the conserved Ser211 and Arg215 in helix 7. The C-terminal end of these helices is formed by a conserved XRX motif, in which the first X is usually an alanine and the second a glycine or serine (Neubauer et al. 2009). This signature motif participates in the interaction with the NBDs, and the interaction network is probably used to transfer conformational changes upon ATP binding and hydrolysis in the EcfAA’ dimer to the transmembrane domains of the transporter.

Fig. 6
figure 6

Crystal structure of a Type III ABC importer. The crystal structure of ECF FolT is shown in the secondary-structure representation from the plane of the membrane from two sides, related by rotation of 90°. The two NBDs are colored in brown, EcfT is colored in gray and the S-component is colored from the N-terminus in blue to the C-terminus in red. In the upper-right corner, a schematic representation showing all subunits in the same color code is given to indicate the position of the membrane. The figure was generated with PyMoL, using structure PBD ID: 4HUQ

Full size image

In the transmembrane part of the transporter, the S-components bind in the cleft of the L-shaped EcfT, being separated from the EcfAA’ dimer only by the coupling helices. Not TM1-3 which were predicted to interact with the ECF module, but TM1, TM2, and the C-terminal end of TM6 of all three S-components show a common surface, through which they interact with the coupling helices. The first coupling helix of EcfT mainly interacts with TM1 of the S-components via hydrophobic interactions and the AxxxA motif. The second coupling helix lies in a hydrophobic groove formed by TM1, TM2, and the C-terminal end of TM6 of the S-components. A similar groove is found in RibU, ThiT, and BioY. TM1, TM2, and the C-terminal end of TM6 contain several small hydrophobic residues that form an extensive hydrophobic interaction network with small hydrophobic residues in the coupling helix. This interaction network is important for the formation and stabilization of the whole transporter (Zhang et al. 2014). A second interaction site is formed by hydrophobic residues in the L3 and L5 loops of the S-components and hydrophobic residues in TM1, TM3, and TM4 in EcfT.

Mechanism of Transport

Type I Importers

The maltose transporter MalFGK2E from E. coli is the best-studied Type I importer (Davidson et al. 2008; Oldham et al. 2007). In fact, a lot of what we know of full-length Type I ABC transporters comes from crystal structures of MalFGK2-MBP, elucidated by the Davidson and Chen labs. Currently, structures of the maltose transporter in the inward-facing apo-state without the maltose-binding protein (MBP) (SV in Fig. 7; Khare et al. 2009; Chen et al. 2013); in the outward-facing state with maltose, nucleotides, and MBP bound (SIV in Fig. 7; Oldham et al. 2007; Oldham and Chen 2011b); and a pre-translocation state between the inward-facing and outward-facing state, with maltose and MBP bound and in the presence (SIII → SIV in Fig. 7; Oldham and Chen 2011a) and absence of nucleotides (SIII → SIV in Fig. 7; Oldham et al. 2013) are available. However, there is not yet a crystal structure of the nucleotide-bound inward-facing conformation (SI–SIII in Fig. 7).

Fig. 7
figure 7

Model of substrate translocation via a Type I ABC importer. After Gouridis et al. (2015)

Full size image

In growing cells, the levels of ATP (typically millimolar) are well above the KD for binding of ATP to the NBDs and this primes the translocator for import (Fig. 7, SV → SI). In SI the translocator is in the nucleotide-bound inward-facing conformation. Upon substrate capture via induced fit, the SBP changes conformation from open to closed (SII) and, subsequently, docks onto the translocator (SIII). Through the binding of SBP in the CL state to the TMDs and Mg–ATP bound in the NBDs, the so-called pre-translocation state is formed (SIII), that is, the NBDs come closer together, forming a tipping for the access of the binding site in the TMDs. The binding energy of both ATP and SBP is used to transit the translocator from an inward to an outward-facing conformation (SIII → SIV), which allows the transfer of the substrate from the SBP to the TMD (SIV). Subsequent hydrolysis of ATP and release of ADP and inorganic phosphate (Pi) complete the reaction cycle and release the substrate on the trans side of the membrane (SIV → SV).

In another model of a Type I importer (Bao and Duong 2013), based on biochemical studies of the maltose transporter, the open state of the SBD interacts with the TMD in the outward-facing conformation, which then allows the binding of maltose. Upon subsequent hydrolysis of ATP, the transporter converts to the inward-facing conformation and releases the maltose into the cytoplasm. This model finds little support in the work on other Type I ABC importers, and the initiation of translocation through binding of SBP in the open state is controversial. In fact, the model of the Duong lab is based on studies in nanodiscs in which the coupling between maltose-MBP binding and ATP hydrolysis is relatively poor compared what others measure (Alvarez et al. 2010), and MalFGK2-MBP may not have been in a fully native conformation.

An outstanding question is whether substrate recognition (selectivity) by ABC importers is counted for solely by the SBP or whether the TMD also contributes. For the maltose system there is data (Ferenci 1980; Ferenci et al. 1986; Treptow and Shuman 1985) suggesting that part of the selectivity originates from the TMD. For instance, mutants that function independent of the maltose-binding protein still transport no sugars other than maltodextrins (Treptow and Shuman 1985). In a recent structural study by the Chen lab (Oldham et al. 2013), it was shown that the transmembrane subunit MalF binds three glucosyl units from the non-reducing end of the sugar, suggesting that both MBP and the TMD contribute to the selectivity of the transporter. In accordance with these observations, the final step in translocation (Fig. 7, SIV → SV) by GlnPQ is faster for asparagine than for glutamine (Gouridis et al. 2015). This implies that the two amino acids interact differently with the TMD and that additional selectivity to transport is provided beyond the step of substrate binding to the SBDs.

Another point of discussion is the stoichiometry of the transport reaction, which is technically challenging to determine. Many ABC transporters display poor coupling between ATP hydrolysis and translocation when purified and reconstituted in lipid vesicles, that is, significant amounts of ATP are hydrolyzed even in the absence of substrate. Another difficulty is that the substrate and ATP need to be present on opposite sides of the membrane, and the affinity of the TMD for the SBP is low (Prossnitz et al. 1989; Dean et al. 1992; Doeven et al. 2004). The Type I ABC importer OpuA transports glycine betaine and is gated by ionic strength. Like GlnPQ it has the SBD covalently linked (Fig. 3), which facilitates the membrane reconstitution and the ability to perform functional assays. Importantly, below threshold levels of ionic strength, there is hardly any ATP hydrolysis provided the lipid composition is “physiological” (>25 mol% of anionic lipids, 40–50 % non-bilayer lipids) (van der Heide et al. 2001; Biemans-Oldehinkel et al. 2006a). Under these conditions and upon activation of the transporter, the experimental ATP/glycine betaine ratio was 2.0 ± 0.5 (Patzlaff et al. 2003), suggesting a mechanistic stoichiometry of 2. At present, this is probably the best estimate of the ATP/substrate ratio for any ABC transporter.

A subset of ABC importers has additional domains fused to the NBDs, which serve to control the activity of the transporter. MalFGK2-MBP from E. coli and OpuA from Lactococcus lactis are the best-studied systems in terms of regulation of transport. The C-terminal domain of MalK of the maltose transporter is involved in carbon catabolite repression (Böhm et al. 2002), which determines the hierarchy of sugar utilization and is also known as an inducer exclusion mechanism. The C-terminal domain of MalK can interact with IIAglc, a component of the bacterial phosphoenolpyruvate-dependent sugar phosphotransferase system (PTS). IIAglc can be in the unphosphorylated and phosphorylated states, depending on the availability of PTS sugars (Nelson and Postma 1984; Dean et al. 1991). In the unphosphorylated form, i.e., when a PTS sugar such as glucose is present, IIAglc binds to MalK and thereby prevents uptake of maltose, the inducer of the maltose operon. It has been recently shown that IIAglc stabilizes the inward-facing conformation of the maltose transporter by wedging between the NBD of one MalK and the regulatory domain of the opposite MalK (Chen et al. 2013). Herewith, the closure of the NBDs and thus the formation of the outward-facing conformation are prevented, which are key steps for ATP hydrolysis and maltose translocation. The inhibition is relieved when glucose is exhausted and IIAglc becomes phosphorylated.

The activity of OpuA increases with the osmolality of the medium, which is signaled to the protein as an increase in the cytoplasmic electrolyte concentration (gating by ionic strength). Two cystathionine-β-synthase (CBS) domains fused in tandem (CBS module) to the C-terminus of the NBD and an anionic membrane surface are central to the gating of the transporter by ionic strength (Biemans-Oldehinkel et al. 2006b; Karasawa et al. 2013). CBS2 is involved in electrolyte sensing, whereas the CBS1 domain merely serves as linker between the nucleotide-binding domain (NBD) and the CBS2 domain (Karasawa et al. 2011). In one model the interaction of the CBS module with an anionic membrane surface locks the transporter in an inactive conformation. Increasing the intracellular ionic strength beyond threshold levels screens the electrostatic interactions of oppositely charged surfaces and activates the transporter. Alternatively, the ionic gating of OpuA could involve two like-charged surfaces, such as the anionic membrane and anionic protein residues, in which case a high ionic strength would promote their interaction. The verdict is not yet out but it is evident that screened electrostatic forces and an anionic membrane surface are intrinsic to the gating mechanism of OpuA.

For a number of Type I ABC transporters, it has been observed that trans-accumulated substrate inhibits the translocation. In Listeria monocytogenes and Lactobacillus plantarum, the transport of glycine betaine and carnitine is inhibited by pre-accumulated (trans) substrate at 0.1 M or above (Verheul et al. 1997; Glaasker et al. 1998). Upon raising the medium osmolality, the corresponding ABC transporters are rapidly activated through a diminished trans-inhibition. Once cellular osmostatic conditions are restored, the transporters are switched off. This so-called trans-inhibition serves as a control mechanism to prevent the accumulation of the compatible solutes glycine betaine and carnitine to too high levels and thereby prevents the turgor pressure from becoming too high. Although the physiological effect is the same, trans-inhibition is mechanistically different from the ionic strength gating observed in OpuA. The ABC transporter OpuA is not inhibited by trans-substrate up to at least 400 mM (Patzlaff et al. 2003). For two Type I ABC transporters the structural basis for trans-inhibition has been elucidated, that are the methionine transporter MetN2I2 from E. coli and the molybdate transporter ModB2C2 from Methanosarcina acetivorans. In both cases, the NBD subunits have a C-terminal regulatory domain albeit with distinct folds (Gerber et al. 2008; Johnson et al. 2012). The MetN2I2 system is inhibited when cytoplasmic levels of methionine reach threshold values (Kadner 1977), and the binding of methionine to the C-terminal domain stabilizes the transporter in the inward-facing conformation, similar to what IIAglc does to the maltose transporter. The ModB2C2 system is inhibited by the binding of molybdate or tungstate to its C-terminal regulatory domain.

In summary: In recent years, we have obtained enormous insight into the energetics and mechanism of transport and selectivity for ligands, as well as the gating of a subset of Type I importers. The regulation of ABC transporters is typically affected through one or more additional domains, linked to the NBD, which allow the transporter to be fully switched off under certain nutrient or osmotic conditions, even if substrate and ATP are available.

Type II Importers

Type I and Type II importers differ in their architecture of the SBP and TMD, but also mechanistically these transporters appear to operate differently (Davidson et al. 2008; Rees et al. 2009; Lewinson et al. 2010; Joseph et al. 2011). The mechanism of transport by a Type II importer presented in Fig. 8 is based on crystal structures of each of the states (Korkhov et al. 2014). The vitamin B12 transporter BtuCDF from E. coli is the best-studied Type II importer (Locher et al. 2002), and most of the states are based on this transporter; state SIII is based on the structure of MolB2C2 from Haemophilus influenzae (Pinkett et al. 2007). The structural analysis suggests two cytoplasmic and one periplasmic gate(s), depicted as vertical (red) and horizontal (gray) bars, respectively (Fig. 8). SI represents the nucleotide-bound, outward-facing conformation. Ligand-bound SBP docks onto the translocator and this results in an occluded state (SII), following the diffusion of the ligand from the SBP to the TMD. Next, ATP is hydrolyzed and inorganic phosphate and ADP are released, which opens the cytoplasmic gate II and the substrate diffuses into the cytoplasm (SIII). The SIII → SIV transition is only possible after the substrate has been released, as in the case of BtuCDF the cavity in the TMD has become too small to accommodate vitamin B12. Korkhov et al. (2014) speculate that SIV may prevent the back-reaction or non-specific solute transport through the translocation pathway. Next, the SBP dissociates from the translocator and the TMDs rearrange to form an outward-facing conformation (SV). Given the high intracellular concentrations of ATP, SV is rapidly converted into SI; the ATP binding closes the cytoplasmic gate II and the system is reset for another translocation cycle.

Fig. 8
figure 8

Model of substrate translocation via Type II ABC importer. After Khorkhov et al. (2014)

Full size image

The BtuCDF system has a very high apparent ATP/substrate stoichiometry, i.e., of the order of 50 (Borths et al. 2005), which may reflect a high rate of ATP hydrolysis of SI or, as Korkhov et al. (2014) propose, SIV to SI-like transitions. This results in futile ATPase activity and a high apparent ATP/substrate stoichiometry; the mechanistic stoichiometry for full translocation is most likely 2. Contrary to the maltose transporter, there are no specific interactions between the substrate (vitamin B12) and the TMD (BtuC subunits). The substrate pocket is very hydrophobic and forms a “Teflon-like” wall to prevent interactions with the hydrophilic substrate (Korkhov et al. 2012b). The NBDs of Type II importers discovered so far do not have extra domains and gating mechanisms (osmotic, catabolite, or product control) are not known.

In summary: Although Type I and Type II importers both employ extracytoplasmic SBPs or SBDs and operate by alternate access of the binding pockets in the TMDs, the mechanism of transport is very different. For instance, in the maltose transporter (Type I) the liganded SBP interacts with inward-facing TMD, whereas in the vitamin B12 transporter (Type II) the liganded SBP interacts with outward-facing TMD. Accordingly, the subsequent steps are different for Type I and Type II importers.

Type III Importers

Apart from what has been revealed by the three crystal structures of the complete ECF FolT, ECF PantT, and ECF HmptT transporters from L. brevis, not much is known about the transport mechanism of the Type III importers. Unlike the Type I and Type II importers, the Type III importers contain a transmembrane domain that binds the substrate specifically without the help of a soluble SBP. At least 21 different S-component families are known, which are able to bind a wide range of small, chemically different substrates (Rodionov et al. 2006, 2009). Molecular dynamic simulation studies on the S-component ThiT for thiamin from L. lactis have shown that the six hydrophobic α-helices of the protein zig-zag through the lipid bilayer, with the connecting loops being located alternatively on the cytoplasmic and the extracellular side of it (Majsnerowska et al. 2013). In the four crystal structures of the substrate-bound S-components available, the substrate-binding site is located near the extracellular surface of the proteins (Zhang et al. 2010; Erkens et al. 2011; Berntsson et al. 2012; Yu et al. 2014). The site is formed by residues in the three C-terminal helices of the S-components together with residues in the L1 and L3 loops. These residues create specific, high-affinity binding sites with Kd values in the low nanomolar range (Erkens and Slotboom 2010; Zhang et al. 2010; Berntsson et al. 2012). The structure of the binding site of ThiT has been shown to be very robust, suggesting that a substantial input of energy, presumably in the form of ATP hydrolysis, is needed to accomplish conformational changes in the binding site to release its substrate (Swier et al. 2015). The N-terminal helices of the S-components are structurally more similar and provide interaction sites to interact with the EcfT domain. In the substrate-bound S-components, the substrate-binding site is closed by the L1 loop. However, in the structures of the complete transporters the L1 loop has moved away, opening the binding site to the cytoplasm (Wang et al. 2013; Xu et al. 2013; Zhang et al. 2014). This is in agreement with EPR studies performed on ThiT, which showed that the major difference between substrate-bound and substrate-free ThiT is a conformational change of the L1 loop (Majsnerowska et al. 2013).

In the structure of the complete transporters, the S-component adopts a highly unusual orientation, lying almost parallel in the membrane like it has toppled over from the orientation of the solitary S-components (SI in Fig. 9) (Wang et al. 2013; Xu et al. 2013; Zhang et al. 2014). The S-component interacts with EcfT, but not with the NBDs, via two interaction sites with the interactions between TM1, TM2, and the C-terminal end of TM6 of the S-components and the coupling helices of EcfT forming the largest interaction interface. The two coupling helices create a X-shaped structure, of which the C-terminal ends stick into the grooves between the RecA-like domain and the helical domain at the surface of the EcfAA’ dimer. Via an interaction network of electrostatic interactions and hydrogen bonds, these coupling helices are thought to transfer the conformational changes upon ATP binding and hydrolysis to the S-component. Like the NBDs in other ABC transporters, the EcfAA’ dimer will undergo a tweezer-like closure upon ATP binding (SII in Fig. 9) (Finkenwirth et al. 2015). This will probably push the C-terminal ends of the coupling helices together, creating a scissor-like conformation. The interactions between the S-component and EcfT are mainly hydrophobic (van der Waals), which allows the subunits to slide along each other. ATP binding might allow the S-component to rotate back to the up-right conformation in the membrane in order to capture its substrate on the extracytoplasmic side, while maintaining the interaction with EcfT (Finkenwirth et al. 2015). It is also possible that in this orientation, the interaction between the substrate-free S-component and EcfT is weakened and the two will dissociate, allowing another substrate-bound S-component to bind (SIII in Fig. 9) (Karpowich et al. 2015). Early research has shown that within the group II ECF transporters multiple S-components can compete for the same ECF module and that the affinity of the S-component for the ECF module is dependent on the substrate concentration, with substrate-bound S-components binding stronger to EcfT than S-components in the absence of their substrate (Henderson et al. 1979). ATP hydrolysis and subsequent release of the hydrolysis products will then allow the coupling helices to move back to the open-scissor conformation and allow the S-component to topple over (SIV in Fig. 9). The energy generated by ATP hydrolysis could be used to lower the binding affinity for the substrate, so it can be released into the cytoplasm. This mechanism would be new form of the moving-carrier mechanism, in which the S-component moves as a rigid body like in the elevator mechanism, but topples over to expose the substrate binding site to the cytoplasm, rather than making a translational movement as in the elevator mechanism.

Fig. 9
figure 9

Model of substrate translocation via a Type III ABC importer

Full size image

A 1:1:1:1 stoichiometry was found in the crystal structures of the complete transporters and this stoichiometry has been confirmed by light scattering experiments performed on the group II ECF transporters from L. lactis (ter Beek et al. 2011). On the other hand, a 1:1:2:2 (A:A’:T:S) stoichiometry was suggested for the group II riboflavin transporter from Thermotoga maritima (Karpowich and Wang 2013) and oligomeric EcfT and S-components were suggested for the group I biotin transporter from Rhodobacter capsulatus (Finkenwirth et al. 2010; Neubauer et al. 2011; Kirsch et al. 2012). The former result is difficult to reconcile with the crystal structures, but in the latter case it is possible that two or more complexes oligomerized in the membrane, and that the basic complex has 1:1:1:1 stoichiometry. Besides the debate on stoichiometry, rare solitary S-components have been found that seem to function independent of the ECF module, i.e., they are found in organisms that do not have the genes for the ECF module. An example is the biotin-specific S-component BioY, found in almost all Chlamydia strains, diverse proteobacterial and cyanobacterial species (Rodionov et al. 2009; Fisher et al. 2012; Finkerwirth et al. 2013). Furthermore, BioY and the cobalt-specific S-component CbiMN from Rhodobacter capsulatus may facilitate substrate entry into the cell in the absence of the ECF module (Hebbeln et al. 2007; Siche et al. 2010). Maybe these S-components topple over spontaneously in the membrane, a transition that would be energetically unfavourable but such rare events may suffice to allow the cell to take up small amounts of micronutrients. The S-components could also oligomerize and subsequently assist each other in the toppling in the membrane, or they could be assisted by unknown protein(s).

In conclusion: Type I and Type II ABC importers employ extramembranous substrate-binding proteins to capture their substrate and deliver the molecule to the translocator. Type III ABC importers use an integral membrane protein (S-component) for the initial binding of substrate; transport is thought to take place by reorientation of the S-component in the membrane. Whereas full structures of Type I and Type II ABC importers with/without SBPs are available, the translocation mechanism by Type III importers awaits further biochemical and structural analysis.