Crystal structure of a bacterial homologue of the bile acid sodium symporter ASBT.

High cholesterol levels greatly increase the risk of cardiovascular disease. About 50 per cent of cholesterol is eliminated from the body by its conversion into bile acids. However, bile acids released from the bile duct are constantly recycled, being reabsorbed in the intestine by the apical sodium-dependent bile acid transporter (ASBT, also known as SLC10A2). It has been shown in animal models that plasma cholesterol levels are considerably lowered by specific inhibitors of ASBT, and ASBT is thus a target for hypercholesterolaemia drugs. Here we report the crystal structure of a bacterial homologue of ASBT from Neisseria meningitidis (ASBT(NM)) at 2.2 Å. ASBT(NM) contains two inverted structural repeats of five transmembrane helices. A core domain of six helices harbours two sodium ions, and the remaining four helices pack in a row to form a flat, 'panel'-like domain. Overall, the architecture of the protein is remarkably similar to the sodium/proton antiporter NhaA, despite having no detectable sequence homology. The ASBT(NM) structure was captured with the substrate taurocholate present, bound between the core and panel domains in a large, inward-facing, hydrophobic cavity. Residues near this cavity have been shown to affect the binding of specific inhibitors of human ASBT. The position of the taurocholate molecule, together with the molecular architecture, suggests the rudiments of a possible transport mechanism.

ASBT/IBAT is a SLC10 (Sodium bile acid co-transporter family) member that moves bile acids across the apical membrane of the ileum into the portal blood vein[5,6]. ASBT utilizes the sodium ion gradient to drive the uphill transport of bile acids across membranes, with a stoichiometry of two sodium ions per substrate reported[7]. Mutations in the human ASBT gene cause a condition of primary bile acid malabsorption[8]. ASBT is a pharmaceutical target for drugs aimed at lowering cholesterol and several ASBT inhibitors have been developed that are effective in animal models[1,2]. As some drugs are poorly absorbed in the intestine or need to be targeted to the liver, ASBT and its close liver paralogue NTCP have also received attention as pro-drug carriers, capable of transporting various compounds coupled to bile acid, e.g. HMG-CoA reductase inhibitors, the anti-viral drug acyclovir, nucleotides and cytostatic drugs[9].

ASBTNM from Neisseria meningitidis, with 26% identity and 54% similarity to human ASBT was identified by fluorescent-based screening methods[10,11] as a suitable candidate for structural studies (Supplementary Fig. 1 and Fig. 2). Residues known to be functionally important in mammalian ASBT and other SLC10 members[12] are well conserved in ASBTNM (Supplementary Fig. 1). Bile acid transport by ASBTNM was confirmed in whole-cells by the sodium-dependent uptake of [3H]-taurocholate (Fig. 1a). The observed Km for [3H]-taurocholate is in the low μM range ~50μM (Fig. 1b), which is similar to that measured for rat and human ASBT[7,13,14]. The ASBT inhibitors cyclosporin A15, bromosulfophthalein[15] and the drug Fluvastatin[16], are also competitors for ASBTNM-mediated [3H]-taurocholate transport (Fig. 1c). Thus, ASBTNM is a valid model of mammalian bile acid transporters. The ASBTNM structure was solved by single wavelength anomalous scattering and refined at a resolution of 2.2Å (Supplementary Tables 1 and 2, see Methods).

Fig. 1

Sodium-dependent transport of bile acid by ASBTNM

a, Time-dependent uptake of [3H]-taurocholate after expression of ASBTNM in E. coli as monitored in buffer containing 137 mM sodium (filled circles) or <1 mM sodium (non-filled circles) b, Michaelis-Menten transport kinetics of ASBTNM-mediated [3H]-taurocholate uptake. The Specific uptake (filled circles) was calculated by subtracting the internalization measured from control cells lacking the transporter (non-filled squares) from the total uptake (non-filled circles), as detailed in Methods. c, ASBTNM-mediated [3H]-taurocholate uptake after 5 min in the presence of 150 μM of taurocholate, cyclosporin A, fluvastatin or bromosulfophthalein (black-filled bars) measured as a percentage of the uptake without their addition (non-filled bar). d, ASBTNM-mediated [3H]-taurocholate uptake after 5 min for wild-type (non-filled bar) and single alanine point mutants (filled-bars): Q77A, E260A, N265A and N295A. The uptake for the mutants is displayed as a percentage of the wild type activity. The expression and detergent-solubilised folded-state of all mutants was similar to wild-type protein, Supplementary Fig. 2a. In all experiments errors bars, s.e.m.; n = 3.

ASBTNM has cytoplasmic N- and C- termini, is comprised of 10 transmembrane helices (TMs) that are linked by short loops, and has overall dimensions of approximately 45 × 30 × 30Å (Figs. 2a and b and Supplementary Fig. 3). TMs 1 to 5 and TMs 6 to 10 are topologically similar but oppositely orientated in the plane of the membrane. The r.m.s.d. (root mean square deviation) after superposition of the two topology-inverted repeats is 3.7Å (Supplementary Fig. 4a and b, and see Methods). Each repeating unit is made of an N-terminal V-motif (TMs 1-2, 6-7) and a Core motif of 3 helices (TMs 3-5, 8-10) (Fig. 2, Supplementary Fig. 3 and 4). If the V and Core-motifs are superposed separately, the r.m.s.d. is lower, 2.6Å and 2.8Å respectively (Supplementary Fig. 4c). The Core motifs from each repeat form the “Core” domain, whereas the two V-motifs create a “Panel” like domain (Fig. 2b). TMs 4 and 9 in the Core domain are broken in the middle (discontinuous), and form helical hairpins with kinked TMs 5 and 10, respectively. At the point where TMs 4 and 9 are broken by well-conserved peptide motifs, they cross over (Fig. 2, Supplementary Fig. 5 and 6). On the intracellular side a wide crevice separates the Core from the Panel domain (Fig. 3a). The cavity extends over halfway through the protein. The extracellular side of the cavity is tightly closed by TMs 1, 2, 4b, 7, 9b and 10. Previously, two topology models of ASBT were proposed with 7 or 9 TMs respectively[17,18]. As TM1 is not conserved in ASBT the structure is broadly consistent with the 9-TM model (Supplementary Fig. 5). TMs 4 and 9 were annotated as extracellular loops in the 7-TM topology model, but were correctly identified in the 9-TM model.

Fig. 2

ASBTNM structure

a, Ribbon representation of ASBTNM as viewed in the plane of the membrane. TMs 1 to 10 have been coloured from red at the N-terminus to blue at the C-terminus and the position of the membrane is depicted in grey. The pink circles indicate sodium sites, Na1/Na2, and the wine-red stick model the substrate taurocholate. b, ASBTNM structure as viewed from the intracellular side as a ribbon representation (left) and as a simplified cartoon (right): sodium ions (pink spheres), taurocholate stick model (wine red).

Fig. 3

ASBTNM structure is inward-facing and contains bound sodium and bile acid

a, Surface representation showing the location of the taurocholate-bound intracellular cavity as a section through the protein. b, The sodium binding sites in ASBTNM. Na1 is octahedrally coordinated by Ser114 and Asn115 on TM4b, Thr132, and Ser128 on TM5 and Glu260 on TM9a. The square pyramidal arrangement of the Na2 ligands is made up of Glu260, Val261, Met263 and Gln264 on TM9, and Gln77 on TM3. c, The intracellular cavity in ASBTNM. Residues lining the cavity and near to the taurocholate are shown. The figures have been coloured as in Fig. 2. A 150-fold difference in inhibition of the mouse and human forms of ASBT by benzothiazepines[4] has been assigned to sequence differences corresponding to Ser291 at the bottom of the cavity. Supplementary Figure 10 shows a stereo version of b and c.

Discontinuous TMs are a common motif in secondary active transporters[3,19,20]. However, the sodium-proton antiporter NhaA is the only other known example where these helices cross as observed in ASBTNM (Supplementary Fig. 6). Indeed, ASBTNM has a similar structure to NhaA, and they superpose with an r.m.s.d. of 2.9Å over 202 Cα atoms (Supplementary Fig. 7a, see Methods). The similarity is more striking when the Core and Panel domains are superposed separately (Supplementary Fig. 7b). This unexpected finding further emphasizes the remarkable plasticity of transporters to utilize a common scaffold to translocate different substrates[20].

In ASBT and NTCP two sodium ions are translocated per bile acid molecule[7,21]. In the highly conserved Core domain of ASBTNM (Supplementary Fig. 8), we have identified two sodium-binding sites (Na1 and Na2) based on the coordination and bond distances (2.0-2.5Å) (Fig. 3b, Supplementary Fig. 9a and 10a, see Methods). Na1, is located approximately 10Å from the cytoplasmic surface between TMs 4b and 5, but also interacts with the carboxylate moiety of Glu260 on TM9a, (Fig. 3b and Supplementary Fig. 10a). The Na2 site is located 8Å from Na1, near the centre at the crossover points of TMs 4a-4b and 9a-9b. Four backbone carbonyl-oxygen atoms coordinate Na2, including Glu260 on TM9a, and the side chains of Gln264 on TM9a and Gln77 on TM3. The residues for which the side-chains interact with the two sodium ions are completely conserved in ASBT and NTCP (Supplementary Figs. 5 and 8). The equivalent glutamate residue to Glu260 is essential for activity in ASBT and NTCP[13,22]. In ASBTNM its replacement with alanine significantly affects transport, as does the mutation of Gln77 to alanine (Fig. 1d and Supplementary Fig. 2a). Thus, it appears that both sodium ions are required for efficient transport. Mechanistically sodium at the Na2 site is almost certainly important to neutralize the partial negative dipole of TM9a, and by doing so, stabilize the interaction with TM4a. Neutralization of the helix dipoles seems a conserved feature for this fold. In NhaA the corresponding TM is thought to be neutralized by the positive charge of Lys300, which is essential for transport[3,23].

The substrate-binding cavity is open to the cytoplasm and is approximately 6 × 12 × 14Å with a solvent accessible volume of 550Å3 (Fig. 3a and see Methods). As the N-terminal half of TM1 is profoundly bent outwards it is more open to one side. The cavity is much bigger than taurocholate, perhaps reflecting the large variety of compounds that are recognized by ASBT[9,12,16] (Fig. 3a and c). It is predominantly hydrophobic but near the bottom there are a number of polar residues and water molecules (Fig. 3c and Supplementary Fig. 10b). As judged from high B-factors, taurocholate appears weakly bound (Supplementary Table 2 and Supplementary Fig. 9b). Consistent with this observation there is only one direct hydrogen bond between ASBTNM and taurocholate, from Asn295 on TM10 to the 7α hydroxyl group. The mutation of Asn295 to alanine causes a dramatic reduction in taurocholate transport (Fig. 1d and Supplementary Fig. 2a). Water molecules bridge the 7α hydroxyl with His294 and the 3α hydroxyl with Asn265, located at the crossover region of TM9. Thr112 is also in the vicinity of the 3α group but cannot be unambiguously placed. The 12α hydroxyl group does not have any apparent hydrogen-bonding partner. The taurine moiety binds between TM1 and TM10. Interaction of the taurocholate with residues in TM10 is in agreement with biochemical data, which have proposed that the last helix in ASBT plays a dominant role in the translocation process[24]. The location of Asn265 between the TM4b and 9b dipoles suggests that it may play a role in the mechanism. The importance of this residue has been inferred from mutagenesis studies on NTCP[22]. In ASBTNM, if it is replaced by alanine, transporter activity is reduced by ~80% (Fig. 1d and Supplementary Fig 2a). Though there are clear similarities in the binding sites between ASBTNM and ASBT there are also sequence differences (Supplementary Fig. 5). Such differences may affect substrate specificity.

For transport to take place the protein must switch between outward and inward facing states[25]. The architecture of ASBTNM provides a clue to understanding how this might occur. The sodium ions are located in the Core domain close to the crossover points of the discontinuous helices and occluded from the bulk solvent. In NhaA sodium binding causes a rearrangement of these helices[26,27]. In ASBTNM similar rearrangements in the Core domain are therefore likely. Since NhaA only translocates ions[26] these TM movements might be sufficient for transport. However, because ASBTNM transports much larger substrates, structural movements in more than the Core domain are needed. For the sodium-coupled transporter LeuT, Forrest et al used the internal asymmetry of the repeating motifs to predict global movements from a single structure[28]; which have been substantiated by crystallographic studies[29]. In an analogous manner to LeuT, an outward-facing model of ASBTNM was generated by superimposing TMs 1-5 on TMs 6-10 and vice versa (Fig. 4a and see Methods). Comparing the inward-facing ASBTNM structure with the outward-facing model, the largest difference is the position of the Panel relative to the Core domain (Fig 4c). A route through the protein between these domains is in agreement with experimental data, that suggest that the last helix of ASBT and TM9 of NhaA line the transport pathway[3,24,26,30]. Interestingly, the NhaA domain equivalent to the Panel is placed between that of the outward-facing and inward-facing ASBTNM states (Fig. 4b). This may either be because NhaA translocates a much smaller substrate, or it could represent another conformation of the transporter, likely an occluded state.

Fig. 4

Putative mechanism for ASBTNM transport

a, Superposition of ASBTNM (red Panel, blue Core) and the outward-facing model as described in the text (light grey). The superposition has been optimized on the Core domains. Loops have been removed for clarity. In the image on the right the Panel of the model has been rotated 25° relative to the Core domain, around the axis shown in the left image, to superimpose the Panels. Significant kinks in the helices are represented as breaks. The area of the cavity is depicted by a salmon trapezoid. b, NhaA shown in the same view as ASBTNM in a. The Core domain is shown in light blue and the Panel in brown. The two additional TMs and β-strands that are not present in ASBTNM are shown in grey. The position that sodium is thought to bind[3] is shown with a black ring. c, Schematic of the proposed mechanism that illustrates the movement of the Panel against the Core domain to transport sodium and bile acid.

In summary, we propose that sodium binding controls the conformation of the Core domain of ASBTNM, which, in turn, drives the movement of the Panel domain. This large conformational change of the Panel relative to the Core domain is required to alter the accessibility to the substrate-binding pocket. The ASBTNM structure should provide important new avenues for designing inhibitors against ASBT with the goal to treat hypercholesterolemia.

Methods Summary

ASBTNM was cloned into a cleavable GFP-His8 fusion vector pWaldoGFPe[10]. The fusion protein was expressed in E. coli, solubilised in 1% dodecyl-β-D-maltopyranoside (DDM) and purified to homogeneity. Prior to crystallisation, untagged ASBTNM was exchanged into 0.06% n-dodecyl-N,N-dimthylamine-N-oxide (LDAO) by size-exclusion chromatography. Crystals were grown in the presence of 10 mM taurocholate by the vapour diffusion method. Data were collected on beamlines I02 and I03 at the Diamond Light Source, dehydration of the crystals being necessary to collect high-resolution data. The protein was derivatised by short soaking a surface engineered cysteine mutant (ASBTNM_1) with 1 mM mercury acetate. The structure of ASBTNM_1 was solved by Hg-SAD and subsequently refined against data collected from ASBTNM at a resolution of 2.2Å. The cell-based bile acid uptake assay for ASBTNM was modified from that previously described[6]