Asymmetry in the Structure of the ABC Transporter-Binding Protein Complex BtuCD-BtuF

See allHide authors and affiliations

Science  07 Sep 2007:
Vol. 317, Issue 5843, pp. 1387-1390
DOI: 10.1126/science.1145950


BtuCD is an adenosine triphosphate–binding cassette (ABC) transporter that translocates vitamin B12 from the periplasmic binding protein BtuF into the cytoplasm of Escherichia coli. The 2.6 angstrom crystal structure of a complex BtuCD-F reveals substantial conformational changes as compared with the previously reported structures of BtuCD and BtuF. The lobes of BtuF are spread apart, and B12 is displaced from the binding pocket. The transmembrane BtuC subunits reveal two distinct conformations, and the translocation pathway is closed to both sides of the membrane. Electron paramagnetic resonance spectra of spin-labeled cysteine mutants reconstituted in proteoliposomes are consistent with the conformation of BtuCD-F that was observed in the crystal structure. A comparison with BtuCD and the homologous HI1470/71 protein suggests that the structure of BtuCD-F may reflect a posttranslocation intermediate.

Adenosine triphosphate (ATP)–binding cassette (ABC) transporters are integral membrane proteins that use the energy gained from hydrolyzing ATP to drive the transport of diverse substrates across cellular membranes (1). In bacteria, binding protein–dependent ABC importers facilitate the uptake of essential nutrients from the environment (2). One such protein is the Escherichia coli vitamin B12 transporter BtuCD (3, 4) that is related in sequence and mechanism to iron-siderophore transporters associated with the virulence of certain pathogenic bacteria (5, 6). Like other ABC transporters, BtuCD consists of two transmembrane domains (TMDs) (i.e., BtuC subunits) that form a pathway for the substrate and two cytoplasmic nucleotide-binding domains (NBDs) (BtuD subunits) that bind and hydrolyze ATP. The fold of the NBDs and their arrangement in ABC transporters are conserved, whereas the architectures of the TMDs are not (2). For example, BtuCD has 20 transmembrane (TM) helices, distinct from the 12 TM helices of the multidrug exporter Sav1866 (7) or the molybdate/tungstate importer ModBC (8). The cognate periplasmic binding protein of BtuCD is BtuF, which captures B12 and feeds it to the external side of the transporter (9).

A detailed understanding of transport phenomena requires direct visualization of the transporters in different conformations and at high resolution. At the present time, no ABC transporter has been visualized in more than one state, illustrating the challenges involved. Here we present the structure of a BtuCD-F complex (stoichiometry BtuC2D2F), which has captured a conformation that may reflect a posttranslocation intermediate. We compared this structure with those previously determined of BtuCD (10) and BtuF (11, 12) and with that of the homologous metalchelate transporter HI1470/71 (13). We found substantial structural changes that may be relevant for formulating a transport mechanism.

To generate BtuCD-F, we exploited an earlier finding (11, 14) and added B12-bound BtuF to detergent-lyzed E. coli cells overexpressing BtuCD (15). From this mixture, we purified a BtuCD-F complex that is colorless and therefore devoid of bound B12. We used a mutant of BtuCD with all surface-exposed cysteines replaced by serines (“cys-less”), which offered the advantage that it was also suitable for spin labeling and electron paramagnetic resonance (EPR) spectroscopy. Cys-less BtuCD has unchanged adenosine triphosphatase (ATPase) activity as compared to native BtuCD and also shows a characteristic, nearly twofold stimulation of ATP hydrolysis rates upon the addition of B12-bound BtuF (ATP hydrolysis rates of 492 ± 17 nmol mg–1 min–1 for BtuCD and 784 ± 8 nmol mg–1 min–1 for BtuCD-F at room temperature).

The crystal structure of the BtuCD-F complex (Fig. 1) revealed that BtuF is bound to the periplasmic face of BtuCD. Although the diffraction data extended to 2.6 Å (table S1), high-quality electron density was visible for only the BtuC and BtuD subunits, whereas in the region of BtuF, the electron density was not as good (fig. S1A). This probably reflects the flexibility of BtuF when bound to the transporter or the absence of lattice contacts involving BtuF. The experimental electron density nevertheless allowed for unambiguous tracing of BtuF, using the previously determined high-resolution structure as a template.

Fig. 1.

Structure of the BtuCD-F complex in ribbon representation. (A) Front view, illustrating the arrangement of the five protein subunits. The horizontal lines indicate the approximate boundaries of the membrane. N and C denote the amino and carboxyl termini, respectively. (B) View from the periplasmic side. Regions of the BtuC subunits in contact with BtuF are teal. (C) Similar to (B) but for clarity, BtuF is not shown. The TM helices are numbered, and the regions in contact with BtuF are teal. The arginine residue conserved in metal-chelate–type ABC importers (R56) is shown as sticks and indicated. (D) BtuC subunits viewed from the cytoplasm; helices are numbered. The pronounced asymmetry of TM5 and, to a lesser degree, TM3 and TM4 is shown. (E) BtuD subunits and the BtuC helices 6a and 6b (previously labeled L helices) viewed from the membrane. Helix 6b is the coupling helix, a feature also present in other ABC-transporter structures.

In the structure of BtuCD-F, the binding protein is devoid of substrate, which is remarkable given the high affinity (∼15 nM) of isolated BtuF for B12 (9). The loss of B12 from the binding pocket correlates with a substantial opening or spreading of the two lobes (N lobe and C lobe) of BtuF and with the insertion of periplasmic BtuC loops into the B12 binding pocket. When the N lobe of the structure of BtuF in isolation is superimposed onto that of the complex, the change in the C lobe can be described as a pivoting by ∼8° around a hinge approximately located at the C-terminal end of the “backbone helix” of BtuF (Fig. 2). As a consequence, the regions of the C lobe in contact with BtuCD are shifted outward by some 4 Å, away from the B12 binding site. Lobe opening has previously been suspected to be important for substrate delivery of other binding protein–dependent ABC importers (16), but no sizeable lobe opening was observed in the crystal structure of apo-BtuF (12), suggesting that binding to BtuCD is critical for this event.

Fig. 2.

Conformational changes in BtuF. A stereo figure of the structure of isolated, B12-bound BtuF (11, 12) is shown in light blue, and that of BtuF, devoid of B12 and in complex with BtuCD, is shown in red. The view is from the side, and the superposition is based on the N lobe of BtuF (left lobe in the present view). As a result, only the C lobe reveals substantial changes, and for clarity, the N lobe of BtuCD-bound BtuF is not shown. The arrow denotes the C-terminal end of the backbone helix (top of the figure) of BtuF.

The contact interface between BtuF and BtuC2 involves the two nonidentical lobes of BtuF (Fig. 1B). Each lobe contacts primarily one BtuC subunit, thus inducing slight asymmetry in the periplasmic loops of the two BtuC subunits. The interface involves various BtuC loops, including parts of helix 5a (Fig. 1C). There are several charged residues participating in the interface, which is consistent with observations seen in other ABC importers (8, 17, 18). In particular, it was predicted that arginine residues from BtuC, conserved in metal-chelate type ABC importers, may interact with conserved glutamate residues on the surface of BtuF (11). Our BtuCD-F structure confirms this prediction and reveals salt bridges between Arg56 residues from both BtuC subunits and the glutamate residues Glu74 and Glu202 from BtuF. These salt bridges appear to be important for proper interaction in vivo, because the deletion of the analogous glutamate residues in the ferrichrome binding protein FhuD abolished transport (18). We also found that mutating Glu74 of E. coli BtuF into an alanine residue prevents the efficient formation of the BtuCD-F complex in vitro. The periplasmic loops between TM5 and helix 5a from both BtuC subunits reach into the B12 binding site, partially occupying the space of previously bound B12. Combined with the spreading of the lobes of BtuF, the insertion of periplasmic BtuC loops is probably responsible for triggering B12 displacement from the binding pocket.

The BtuD subunits (NBDs) reveal a nucleotidefree “open” conformation with a gap between the P loops (phosphate-binding loops) and the LSGGQ (Leu-Ser-Gly-Gly-Gln) motifs of opposing NBDs. It has been noted earlier (8) that there are substantial structural differences between the nucleotidefree states of the full transporters ModBC, HI1470/71, and BtuCD and the NBD dimer from the maltose transporter MalK2 (19). This is in contrast to the highly similar conformations of the ATP-bound states, as revealed by the structures of the multidrug transporter Sav1866 and various ATP-bound NBD dimers (7, 1922). There are no substantial structural changes of the BtuD subunits in BtuCD-F when compared to those observed in BtuCD, suggesting that for substrate-induced ATPase stimulation, only changes in the flexibilities of the NBDs (but no distinct structural changes) may be necessary.

The BtuC subunits are thought to provide a central translocation pathway for B12. Whereas twofold rotational symmetry related these subunits in the original structure of BtuCD (10), BtuCD-F reveals striking asymmetry that is primarily evident in the helices TM3, TM4, TM5, and 5a but to a smaller extent also in other helices. The observed asymmetry is relatively moderate at the interface with BtuF (Fig. 1, B and C) but is very substantial at the cytoplasmic side of the membrane (Fig. 1D). Helices TM3 to 5a form a subset of TM helices whose orientations appear to control to which side of the membrane the translocation pathway is exposed. In the HI1470/71 structure, this subset was found to be distinct in conformation as compared with BtuCD, causing HI1471 to adopt an inward-facing conformation, whereas BtuCD was outward-facing. In BtuCD-F, helices TM3 to 5a of one BtuC subunit (yellow in Figs. 1 and 3; interacting mostly with the C lobe of BtuF) are similar in conformation to BtuCD, whereas the same helices of the other BtuC subunit (blue in Figs. 1 and 3; interacting mostly with the N lobe of BtuF) are similar to HI1471 (Fig. 3). This is reflected in the crossing angles of TM5 in the three crystal structures (table S2). As a consequence of the asymmetric conformations of helices TM3 to 5a, the central cavity in BtuCD-F is accessible to neither side of the membrane and appears too small to harbor a B12 molecule.

Fig. 3.

Comparison of the TMD conformations of BtuCD (red), BtuCD-F (yellow and blue), and HI1470/71 (teal). For clarity, only helices TM5 and 5a from both TMDs are shown after the superposition of the NBDs and all TM helices except TM3, TM4, TM5, and 5a. For a stereo figure of the superimposed structures, see fig. S2.

A structure-based alignment of BtuC and HI1471 (fig. S3) indicates conserved hydrophobic residues in TM5 and helix 5a. These include Leu146 and Leu147 at the cytoplasmic side of TM5 and Leu172 and Met176 in helix 5a. In BtuCD-F, these residues shield the central cavity from the cytoplasm and the periplasm. Analogous hydrophobic residues are also present in the amino acid sequences of other metal-chelate ABC importers such as the E. coli ferrichrome importer FhuB, suggesting a common role in gating.

Because the asymmetrically occluded conformation of BtuCD-F was notable and unexpected, we sought to confirm its relevance in lipidic membranes. Hence, we introduced cysteine side chains at strategically placed positions in BtuC, in the loops preceding and following the key transmembrane helix TM5 (residues Ser141 and Thr168, respectively). Because of the stoichiometry, each mutation introduced two cysteines in the assembled transporter. We modified the engineered cysteines with spin labels and recorded continuous-wave (CW) EPR spectra of BtuCD and BtuCD-F after reconstitution in proteoliposomes, which mimic the native environment of membrane proteins. CW EPR spectra reflect the local dynamics of spin labels, which is influenced by the environment of the protein (23). At the same time, the spectra can reveal spin-spin coupling if the distance between two spin labels becomes short enough (<20 Å) (24). Our spectra (Fig. 4) show considerable differences between BtuCD and BtuCD-F. At the periplasmic side of the membrane, the spin labels at position 168 reveal substantial mobility in BtuCD, with two clearly defined dynamic components. The addition of B12-bound BtuF appears to restrict the overall dynamics, and coupling between spin labels becomes evident in the spectrum, indicating a shortened distance between the labels. The spectra are consistent with the structures of BtuCD and BtuCD-F. In the latter, the distance between Cα positions of the residues 168 from the BtuC subunits decreased to ∼14 Å from >25 Å in BtuCD. In addition, the external loops between TM5 and helix 5a extensively interact with BtuF, which probably accounts for the decreased mobility of the labels in BtuCD-F.

Fig. 4.

EPR studies of BtuCD and BtuCD-F. (A) Strategically placed reporter labels 141 and 168 in the BtuC subunits are indicated by green and red spheres, respectively. They are located in the cytoplasmic loop preceding TM5 (position 141) and in the periplasmic loop between TM5 and helix 5a (position 168). (B) Normalized CW EPR spectra of BtuCD and BtuCD-F (black and red, respectively) after spin labeling at position 168 and reconstitution in proteoliposomes. Mobile (m) and immobile (i) components of the spectra are shown, and the region of the spectrum revealing spin-spin coupling is indicated with an arrow. The scale bar represents 15 gauss. (C) Similar to (B) but with spin labels at position 141. The spectra of BtuCD and BtuCD-F are in black and green, respectively.

At the cytoplasmic side of the membrane (in the loop between TM4 and TM5), the differences between the spectra are even more dramatic. In BtuCD, the labels at position 141 are highly immobile (almost at the rigid limit), consistent with their location at the center of the transporter in an outward-facing conformation. The addition of B12-bound BtuF and the formation of BtuCD-F lead to two distinct components in the spectrum, one indicating an immobile label, the other a mobile label. Although we cannot rule out the possibility that the two components are the result of a dynamic exchange of labels between two environments, the BtuCD-F structure offers a simpler interpretation: The asymmetry between the BtuC subunits probably causes the two labels to face distinctly different environments (Fig. 1D). Whereas residue 141 of one BtuC subunit (yellow in Fig. 1D) is located at the center of the complex, where the flexibility of the spin label is probably restricted, that of the other BtuC subunit (blue in Fig. 1D) is at the periphery, where the label may experience a much higher conformational flexibility. This interpretation predicts equal contributions of the mobile and immobile components in the EPR spectra, which agrees with a quantitative analysis of the spectra (fig. S4). The EPR spectra are thus consistent with the conformations of the TM5 helices that were observed in the crystal structures of BtuCD and BtuCD-F (Fig. 3), and we conclude that both structures probably reflect conformations that are relevant in native membranes.

On the basis of biochemical data and the structures of full ABC transporters determined in recent years, a conserved coupling mechanism for ABC transporters has recently been proposed that suggests that binding of ATP to the NBDs promotes an outward-facing conformation of the TMDs, whereas a release of the hydrolysis products promotes an inward-facing conformation (8, 25). ABC importers would thus acquire substrates from their binding proteins in the ATP-bound state and release the substrates to the cytoplasm upon dissociation of the ATP hydrolysis products (2). This basic two-state scenario is in agreement with the structures of Sav1866, HI1470/71, and ModBC, which all visualize such states, whereas those of BtuCD and BtuCD-F reveal intermediate conformations. Although there may be differences in the detailed transport mechanism of BtuCD when compared with that of the maltose transporter MalFGK or the molybdate/tungstate transporter ModBC, BtuCD may nevertheless largely follow the common coupling mechanism of ABC transporters.

The BtuCD-F complex is stable in the absence of ATP, which is different from MalFGK or ModBC, where ATP and vanadate are required to generate stable complexes with the binding proteins (8, 26). Even though the interaction of BtuCD with BtuF is sufficient to release B12 from the high-affinity binding site, the transport of B12 across the membrane requires ATP both in vivo and in vitro. During a productive transport cycle, B12 is probably fed into an outward-facing conformation such as that observed in the structure of BtuCD and later released into the cytoplasm from an inward-facing conformation similar to that revealed by HI1470/71. The BtuCD-F structure presented here is a conformation that is, with respect to its TMDs, an intermediate between those of BtuCD and HI1470/71. This suggests that during the conversion from the inward-to the outward-facing conformation, the BtuC subunits may not alter their conformations simultaneously.

Unlike BtuCD, many other ABC transporters have pairs of TMDs with distinct amino acid sequences, which may have mechanistic consequences. The structure of BtuCD-F provides an opportunity to study conformational asymmetry at high resolution, which could prove useful for the mechanistic understanding of intrinsically asymmetric ABC transporters.

Supporting Online Material

Materials and Methods

Figs. S1 to S4

Tables S1 and S2


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article