Structural basis of sodium-dependent bile salt uptake into the liver

The liver takes up bile salts from blood to generate bile, enabling absorption of lipophilic nutrients and excretion of metabolites and drugs1. Human Na+–taurocholate co-transporting polypeptide (NTCP) is the main bile salt uptake system in liver. NTCP is also the cellular entry receptor of human hepatitis B and D viruses2,3 (HBV/HDV), and has emerged as an important target for antiviral drugs4. However, the molecular mechanisms underlying NTCP transport and viral receptor functions remain incompletely understood. Here we present cryo-electron microscopy structures of human NTCP in complexes with nanobodies, revealing key conformations of its transport cycle. NTCP undergoes a conformational transition opening a wide transmembrane pore that serves as the transport pathway for bile salts, and exposes key determinant residues for HBV/HDV binding to the outside of the cell. A nanobody that stabilizes pore closure and inward-facing states impairs recognition of the HBV/HDV receptor-binding domain preS1, demonstrating binding selectivity of the viruses for open-to-outside over inward-facing conformations of the NTCP transport cycle. These results provide molecular insights into NTCP ‘gated-pore’ transport and HBV/HDV receptor recognition mechanisms, and are expected to help with development of liver disease therapies targeting NTCP.

Bile salts are essential molecules for absorption of lipophilic nutrients and vitamins (vitamin A, D, E and K) in the small intestine, as well as for maintenance of endocrine and cholesterol homeostasis and excretion of toxins 1 . The vast majority-more than 90%-of the body's bile salts pool is recycled daily, shuttling between intestine and liver, where bile salts are used to aid nutrient absorption and generate bile, respectively. Human members of the solute carrier 10 (SLC10) protein family are key bile salt transporters for the maintenance of enterohepatic circulation 5,6 : NTCP 7 (also known as SLC10A1) is mainly expressed in the hepatocyte basolateral membrane, and constitutes the main active transport route of bile salts into the liver from blood, whereas apical sodium-dependent bile acid transporter 8 (ASBT (also known as SLA10A2)) is expressed in ileum enterocytes and takes up bile salts from the intestinal lumen. Both transporters are important pharmacological targets, as they can be used to facilitate oral absorption 9,10 (ASBT) and liver uptake 11,12 (NTCP) of drugs conjugated to bile salts, and are involved in the action mechanism (ASBT) 13 and pharmacokinetics (NTCP) 14,15 of cholesterol-lowering therapies. Moreover, NTCP downregulation in mouse models is associated with increased cholesterol and phospholipid excretion 16 , as well as decreased weight gain with a high-fat diet 17 . Notably, NTCP has a fundamental role in liver pathology, as the human cellular entry receptor for HBV/HDV 2,3 . Chronic HBV infection is a major cause of hepatocellular carcinoma and liver cirrhosis, and affects around 250 million people globally 18,19 . The viruses use the myristoylated and unstructured N-terminal domain in the large envelope protein-the preS1 domain (myr-preS1)-to recognize and bind human NTCP [20][21][22] , explaining viral hepatotropism and the narrow range of animal hosts. Consistently, myristoylated peptides encompassing the residues 2-48 of myr-preS1 (myr-preS1 48 ) act as potent inhibitors of HBV/HDV entry into cells [23][24][25][26] .
Structural insights into the transport mechanism of NTCP and ASBT have come from early X-ray crystal structures of prokaryotic homologues that revealed a ten-transmembrane-helix topology, arranged into core and panel domains 27,28 . The homologues follow an alternating-access transport mechanism, in which relative movements of the two domains provide alternating access to substrate-and sodium-binding sites on opposite sides of the membrane.
Here we set out to study the structural basis of human NTCP function using cryo-electron microscopy (cryo-EM) in combination with conformation-specific nanobodies to reveal key conformational transitions of the NTCP transport cycle. Article detergent solution. To minimize the number of consensus exchanges and maximize stability, we determined the contribution of single exchanges, and retained only those that increased stability, yielding a final construct that shares approximately 98% amino acid identity with wild-type NTCP (Extended Data Fig. 1) and enables purification of monodisperse material in milligram amounts. We refer to this construct as NTCP EM . NTCP EM showed robust Na + -dependent uptake of the fluorescent substrate analogue tauro-nor-THCA-24-DBD (4.5 ± 1.3-fold increase in the sodium-over the choline-based condition), similar to that of wild-type NTCP (10.2 ± 3.9), whereas control cells expressing the unrelated Na + -dependent neurotransmitter transporter EAAT1 lacked bile salt uptake (1.6 ± 0.1 sodium-dependent increase) (Fig. 1a). These results show that the transport mechanism is conserved in NTCP EM .
Second, to provide molecular features on the NTCP EM surface for cryo-EM analysis, we generated and selected nanobodies that potently bind NTCP EM . Nanobody (Nb)87 and Nb91 inhibit Na + -induced fluorescent-substrate uptake by cells expressing NTCP EM with half-maximal inhibitory concentrations (IC 50 ) of approximately 180 and 34 nM (Fig. 1b), respectively, showing that they recognize NTCP EM from the extracellular side, and suggesting that they stabilize conformational intermediates of the transport cycle. During cryo-EM sample optimization, we screened NTCP EM complexes with these nanobodies and megabody scaffolds that result in an additional 85 kDa of folded domains 30 in both detergent solutions, as well as reconstituted in nanodiscs. This yielded final cryo-EM maps of NTCP EM -Nb87 in nanodiscs, and NTCP EM -megabody (Mb)91 in detergent at overall resolutions of 3.7 and 3.3 Å, respectively, enabling structure determination (Fig. 1c

NTCP architecture
NTCP EM adopts an SLC10 fold with two structurally distinct domainscore and panel (Fig. 2a, b)-and contains nine transmembrane α-helices (TM1-9) with an unstructured N terminus on the extracellular side. The transmembrane helices are connected by short loops, as well as extracellular α-helices (ECH) and intracellular α-helices (ICH) lying nearly parallel to the membrane. The panel domain is formed by TM1, TM5 and TM6, and has lost pseudo-internal symmetry compared with its equivalent in SLC10 prokaryotic homologues, owing to the evolutionary loss of one transmembrane helix. The NTCP EM core domain is formed by packing of two helix bundles, TM2-4 and TM7-9, which are related by pseudo-two-fold symmetry (Cα root mean squared deviation (r.m.s.d.) ≈ 5 Å). TM3 and TM8 unwind close to the middle of the membrane, and pack against each other to form a characteristic X-shaped structure that displays highly conserved polar residue motifs among vertebrate SLC10 bile salt transporters (Extended Data Figs. 5, 6).   Most of the reported residues important for binding of sodium and substrate map to the core domain. Sodium-binding sites 1 (Na1; including S105, N106, T123 and E257 sidechains) and 2 (Na2; including Q68 and Q261), which were first observed in crystallographic studies of prokaryotic SLC10 homologues 27 , are structurally conserved in NTCP EM (Fig. 2c). Structural conservation and NTCP mutagenesis 31,32 strongly suggest that the two sodium ions that are thermodynamically coupled to bile salt transport 33,34 bind to these sites. Beyond Na1 and Na2, mutations at residues in the X motif 27 (equivalent to N262) or in close proximity 32 (Q293) impaired transport function, suggesting a role in substrate binding. Consistently, the NTCP-inactivating mutation 35 S267F, which is associated with hypercholanaemia and vitamin D deficiency in humans 36 lays just above the X motif, and A64T 37 is close to the sodium-binding sites.

NTCP inward-facing state
In complex with Nb87, NTCP EM adopts an inward-facing state with core and panel domains tightly packing against each other on the extracellular side of the membrane (Fig. 3a, b). On the intracellular side, the domains separate, uncovering an amphiphilic large cavity (molecular surface volume > 1,500 Å 3 ) that opens to the cytoplasm, as well as laterally to the hydrophobic core of the membrane through a crevice between TM6 and TM9. On the other side of the transporter, TM1 and TM5 pack against the core domain, occluding the cavity from the membrane. Na1 and Na2 face this cavity and localize behind the conserved X motif. In addition, the cavity is lined by several conserved polar residues from the core domain (including N103, N262, Q264 and Q293), some of which have been shown to be important for transport; hydrophobic residues are mostly located in the panel domain. Amino acid conservation, mutagenesis studies and the large volume of the cavity suggest that it is part of the substrate pathway on the cytoplasmic side. Consistently, a molecule of taurocholate has been reported to bind to the equivalent region in the structure of a prokaryotic homologue 27 .

Transition to the open-pore conformation
In complex with Mb91, NTCP EM shows a marked conformational change compared with the inward-facing state ( Fig. 3a, b, Supplementary Video 1). Core and panel domains rotate around 20º and move approximately 5 Å towards opposite sides of the membrane as nearly rigid bodies. These movements are facilitated by conserved glycine and proline residues that act as hinges in the connecting loops, as well as in the ICH and ECH (Extended Data Fig. 5). As a consequence, the two domains separate from each other on both extracellular and cytoplasmic sides, and open a wide pore through the transporter, exposing Na + -binding sites and X motif residues simultaneously to opposite Article sides of the membrane. This is an unexpected conformational transition, as active transporters typically alternate exposure of their ligand binding sites to the extracellular and intracellular milieus, and adopt intermediate states with substrates occluded within the protein 38,39 .
We discuss a plausible NTCP transport mechanism including an open-pore state below. The surface lining the pore is amphiphilic, and most polar residues in this surface come from the core domain, including conserved sidechains in the X motif. Human NTCP mutations S267F and S199R-both of which are associated with hypercholanaemia 35,36,40 -also map to this surface, on opposite sides of the membrane. The pore has a minimum diameter of approximately 5 Å, and contains a large volume (2,400 Å 3 ), with its long axis oriented at an angle of about 45º to the membrane plane. It displays wide openings on both extracellular and intracellular sides to bulk solutions, as well as hydrophobic membrane leaflets. The amino acid conservation, the architecture and the amphiphilic nature of the pore strongly suggest that it is the pathway for translocation of a wide range of amphiphilic bulky substrates transported by NTCP, including bile salts 7,41 , sulfated steroids 42,43 and statins 14,15 . Consistently, in the cryo-EM map of NTCP EM -Mb91, we observed extra density that partially occupies the pore on the cytoplasmic side, wedged in the crevice between TM6 and TM9 and in close proximity to conserved residues in the X motifs (Fig. 3a, Extended Data Fig. 7). Our cryo-EM sample included both Na + and substrate taurocholate, and the density probably corresponds to a substrate molecule bound to NTCP EM . However, the lack of molecular features in the density precluded unambiguous determination of the bound molecule.
It is worth noting that the NTCP EM -Mb91 complex structure was determined from samples in detergent solutions, raising the possibility that detergent molecules could have facilitated the open-pore state. To shed light on this question, we determined the cryo-EM structure of NTCP EM -Nb91 complex reconstituted in nanodiscs. Despite the limited   Fig. 8), demonstrating that NTCP EM adopts an open-pore state in a lipid bilayer, which therefore represents a functional state of the transport cycle. We also observed similar extra density localized to the pore in the nanodisc-reconstituted NTCP EM -Nb91 complex (Extended Data Fig. 7), further supporting the idea that the density corresponds to a substrate molecule, rather than detergent bound to the transporter.

Nb87 impairs myr-preS1 binding
The conformational changes associated with NTCP EM pore opening have implications for the HBV/HDV receptor-recognition mechanism. A reported critical region for myr-preS1 binding and viral infection 2 (NTCP residues K157-L165) maps to the extracellular half of TM5 in the panel domain, and localizes far (more than 20 Å) from both Nb87-and Nb91-binding interfaces on the surface of the core domain (Fig. 4a).
Notably, there is a conformational change around TM5 when comparing NTCP EM inward-facing and open-pore states (Fig. 4a,  To test this hypothesis, we optimized a fluorescence-based myr-preS1 binding assay in cells using a purified myr-preS1 48 lipopeptide fused to GFP (myr-preS1 48 -GFP). Indeed, myr-preS1 48 -GFP labelling of cells expressing wild-type NTCP or NTCP EM was greatly decreased in the presence of Nb87, but was not affected by Nb91 (Fig. 4c). Nb87-and Nb91-overlapping epitopes on the surface of the core domain distant from HBV/HDV binding determinants strongly indicate that the inhibitory effect of Nb87 on myr-preS1 48 -GFP binding is not owing to direct steric hindrance, but rather to stabilization of the inward-facing state that allosterically buries myr-preS1 binding determinants in the protein core. Overall, structural and functional results indicate that myr-preS1 binds preferentially to the open-pore state and interacts with exposed residues lining the pore at the interface between core and panel domains.

Discussion
Our structural and functional analyses of NTCP EM in complexes with conformation-specific nanobodies reveal key molecular aspects of NTCP transport and HBV/HDV receptor-recognition mechanisms.
The NTCP EM open-pore structure is apparently at odds with the alternating-access transport mechanisms observed in most solute carrier families 38,39 , including prokaryotic homologues of SLC10 27,28 , which involve occluded substrate-bound intermediates of the transport cycle, raising the question of how to reconcile an open-pore intermediate state with thermodynamically active transport. Our structures suggest a plausible mechanism in which the pore is transiently open in the presence of substrate (and thermodynamically coupled Na + ) and closes upon release of ligands into the cytoplasm in the inward-facing state (Fig. 4d). The presence of an additional cryo-EM density in the pore, probably representing a bile salt molecule bound to the transporter, supports this type of mechanism. Moreover, sodium ions would contribute to gate the pore, avoiding bile salt permeation down its electrochemical gradient and preferentially enabling bile salt binding at high extracellular sodium concentrations from the outside (under physiological ionic gradients), for instance by inducing outward-facing states that resemble those observed in prokaryotic homologues of SLC10 28 . Consistent with this line of thinking, early ion transport theories considered active carriers as pores whose gates are controlled by the energy source 44,45 , challenging the classical distinction between channels and pumps 46 . To our knowledge, the NTCP open-pore state is the first structural demonstration of an active transporter displaying a wide-open-pore transport pathway for a bulky solute. Detailed knowledge of how the NTCP pore is gated on the extracellular side will require further structural and biophysical work.
The NTCP EM open-pore structure further shows that HBV/HDVbinding determinants line the pore within the membrane plane, accessible to the outside, and overlap with the substrate transport pathway. By sharp contrast, the inward-facing state shows tight packing of core and panel domains on the extracellular side burying virus-binding determinant residues within the protein core, and consistently, Nb87 antagonizes myr-preS1 binding. These results converge to suggest that myr-preS1 interacts with residues in the pore and, hence, that HBV/HDV selectively recognize NTCP conformations with an open-to-outside substrate pathway, while binding to inward-facing states is impaired (Fig. 4d). Such a mechanism explains the reported inhibitory effect of myr-preS1 binding on bile salt transport 32 , as bound myr-preS1 would stabilize open-to-outside states and preclude isomerization to inward-facing ones and the antagonism between myr-preS1 and substrate binding 32 , as both ligands would interact with overlapping binding sites within the pore.
The inhibitory effect of Nb87 on myr-preS1 binding reveals the therapeutic potential of molecules that stabilize NTCP inward-facing state(s), as allosteric inhibitors of viral cell entry. Such molecules could constitute alternative and/or synergistic therapeutic tools to existing lipopeptides that mimic high-affinity myr-preS1 binding 23,47 , as well as neutralizing antibodies against HBV 48,49 .

Online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-022-04723-z.

Thermostable NTCP constructs
Consensus amino acids were calculated using JALVIEW 50 and reported criteria 29 from sequences of representative NTCP vertebrate orthologues (Extended Data Fig. 1), aligned using Muscle 51 . Consensus amino acid exchanges were simultaneously introduced into wild-type NTCP sequence background with N-glycosylation mutations N5T and N11T, improving protein stability. Deletions of N-terminal residue E2, and the unstructured C terminus (residues T329-A349) in the consensus non-glycosylated construct further improved homogeneity of the sample, yielding the so-called NTCP CO .
In general, the consensus approach generates protein samples with overall improved stability, but it is expected that by simultaneously introducing all consensus mutations, some destabilizing exchanges are included. To minimize the latter, we probed thermal stability of single-point NTCP CO mutants, in which we reverted consensus amino acids to the wild-type residues, using fluorescence-detection size-exclusion chromatography 52 (SEC). Removal of destabilizing consensus exchanges in NTCP CO , yielded a consensus design, NTCP EM , which is nearly identical to wild-type NTCP (approximately 98% identity) (Extended Data Figs. 1, 6), while preserving Na + -dependent bile salt transport as well as myr-preS1 recognition mechanisms.

Protein expression and purification
cDNAs encoding NTCP constructs were synthesized (GenScript) and subcloned into a pcDNA3.1(+) vector encompassing a C-terminal Pre-Scission site, followed by GFP, and two Strep-tags in tandem for affinity purification. Protein expression was done in HEK 293F cells (Thermo Fisher; cells were not authenticated or tested for mycoplasma contamination) by transient transfection, as described 53 with small variations. In brief, cells grown in FreeStyle 293 medium (Thermo Scientific) were transfected with linear 25K polyethyleneimine (PEI) (Polysciences) at a cell density of 2.5 × 10 6 cells per ml using 3 µg ml −1 DNA. Valproic acid (VPA) was added to the culture at a final concentration of 2.2 mM 6-12 h after transfection and cells were grown for additional 48 h before collection.
NTCP EM complexes with nanobodies and megabodies, respectively, were formed by mixing purified protein samples at 1:1.2 (transporter:nanobody, or megabody) molar ratio, and incubated for 2h at 4 °C. Excess nanobody or megabody was removed by SEC using SEC buffer. MSP1D1 nanodisc-scaffold protein was expressed and purified using published protocols 54 . Reconstitution was done by mixing purified NTCP EM -Nb and NTCP EM -Mb complexes, respectively, with MSP1D1 and liver total lipid extract (Avanti Polar Lipids) at 0.1:1:15 molar ratio, and incubated with methanol-activated biobeads for 2 h. Biobeads were exchanged once, and the mixture was further incubated overnight. Nanodisc-reconstituted sample was purified in a Superdex 200 increase column (GE Healthcare Life Sciences) in buffer containing 20 mM HEPES pH 7.4, 100 mM NaCl, and 0.2 mM sodium taurocholate. Samples were concentrated as described above, and immediately used for cryo-EM grid preparation.

Nanobody generation, expression and purification
Nanobodies against NTCP CO were generated using published protocols 55 . In brief, one llama (Lama glama) was six times immunized with a total 0.9 mg of NTCP CO reconstituted in proteoliposomes. Four days after the final boost, blood was taken from the llama to isolate peripheral blood lymphocytes. RNA was purified from these lymphocytes and reverse transcribed by PCR to obtain the cDNA of the open reading frames coding for the nanobodies. The resulting library was cloned into the phage display vector pMESy4 bearing a C-terminal His 6 tag and a CaptureSelect sequence tag (Glu-Pro-Glu-Ala). Different nanobody families, as defined by the difference in the CDR3, were selected by biopanning. For this, NTCP CO reconstituted in proteoliposomes was solid phase coated directly on plates. NTCP CO specific phage were recovered by limited trypsinization, and after two rounds of selection, periplasmic extracts were made and analysed using ELISA screens. Nb87 and Nb91 were expressed in Escherichia coli for subsequent purification from the bacterial periplasm. After Ni-NTA (Sigma) affinity purification, nanobodies were further purified by SEC in buffer: 10 mM HEPES pH 7.4, and 110 mM NaCl.
Nb91 was enlarged by fusion to the circular permutated glucosidase of E. coli K12 (YgjK, 86 kDa) to build the megabody referred to as Mb91. Mb91 was generated and purified using previously described protocols 30 .

Fluorescent substrate analogue transport assay
Sodium-dependent substrate uptake was measured in HEK 293F cells transfected with 2 µg µl −1 cDNA using the above-mentioned protocol with small modifications. Forty-eight h after transfection, around 1 million cells were pelleted, washed, and resuspended in 500 µl of transport buffer (110 mM NaCl, 4 mM KCl, 1 mM MgSO 4 , 1 mM CaCl 2 , 45 mM mannitol, 5 mM glucose and 10 mM HEPES pH 7.4), or control buffer in which NaCl was substituted with choline chloride (ChCl). To probe the effect of nanobodies on bile salt transport, cells were incubated with nanobodies for 1.5 h, followed by addition of the fluorescent substrate analogue tauro-nor-THCA-24-DBD 56,57 (tebu-bio) to a final concentration of 10 µM for 30 min at 37 °C. Excess fluorescent analogue was removed by centrifugation (13,000g for 30 s), and 1 wash with the above-mentioned control buffer. Then, cells were resuspended and lysed using Pierce IP lysis buffer (Thermo Fisher). Finally, lysates were centrifuged (13,000g for 10 min), and transferred to black 96-well flat-bottom plates (Grenier), and quantified by fluorescence in a micro-plate reader (CLARIOstar-Plus) using excitation at 454 nm and emission of 570 nm. Three biologically independent experiments were quantified in triplicate samples. Nb titrations data were fitted in Prism 8.0.1 (GraphPad) to the following dose-respond curve: Where Y min corresponds to fraction of transport at saturating Nb concentrations, IC 50 is the half-maximal inhibitory concentration, and x is log[Nb].
Myr-preS1 purification and binding assay cDNA encoding the N-terminal myristoylated consensus residues 2-48 of human HBV myr-preS1 domain (myr-GTNLSVPNPLGFFPDHQL DPAFRANSNNPDWDFNPNKDHWPEANKVG) was synthesized (GenScript) and subcloned into a pcDNA3.1(+) vector encompassing a C-terminal GFP, and poly Histidine-tag (namely, myr-PreS1 48 -GFP Myr-preS1 48 -GFP binding to NTCP constructs was assayed in HEK 293F cells, grown and transfected with 1µg ml −1 DNA using the protocol described above. Forty-eight hours after transfection, cells were washed with pre-warmed PBS, and ~1 million cells were pelleted and resuspended in 1 ml of PBS. To probe the effect of nanobodies, cells expressing NTCP constructs were pre-incubated with 10 µM nanobodies for 1.5 h. They were then labelled with 10 nM (wild-type NTCP) or 50 nM (NTCP EM ) purified myr-preS1 48 -GFP for 30 min. Excess fluorescent-probe was removed by centrifugation (13,000g for 30 s), and one wash with PBS. Cells were then re-suspended in PBS and GFP fluorescence was recorded in a micro-plate reader (CLARIOstar-Plus) using excitation at 470 nm and emission at 508 nm.

Electron microscopy sample preparation and data acquisition
Purified NTCP EM -Nb or -Mb complexes were applied to glow-discharged Au 300 mesh Quantifoil R1.2/ 1.3. Typically, 4 µl of sample at 3-4 mg ml −1 was applied to the grids, and the Vitrobot chamber was maintained at 100% humidity and 4 °C. Grids were screened in 200 kV Talos Arctica microscope (ThermoFisher) at the IECB cryo-EM imaging facility. Final data collection was performed in 300 kV Titan Krios microscope (ThermoFisher) at EMBL-Heidelberg Cryo-Electron Microscopy Service Platform, equipped with K3 direct electron detector (Gatan). Final images were recorded with SerialEM 58 at a pixel size of 0.504 Å. Dose rate was 15-20 e − pixel s −1 .
Cryo-EM data processing, model building and structure analysis All datasets were processed with cryoSPARC v2 and v3 59 . Movies were gain corrected, and aligned using in-built patch-motion correction routine. Contrast transfer function (CTF) parameters were estimated using the in-built patch-CTF routine in cryoSPARC. Low-quality images were discarded manually upon visual inspection.
For the NTCP EM -Mb91 complex, 5,796,802 particles were templatepicked from 21,390 micrographs, and selected through several rounds of 2D, as well as 3D ab initio classifications. Particles from 3D ab initio classes displaying interpretable density for transmembrane helices were pooled, and used for homogenous refinement (Extended Data Fig. 2). Cryo-EM density corresponding to both detergent micelle and megabody scaffold were masked out, and particles were further subjected to local refinement using a fulcrum that localized to center of NTCP EM transmembrane region. Focused refinement yielded a final map at an overall resolution of 3.3 Å, based on the gold-standard 0.143 Fourier shell correlation (FSC) cut-off.
For the NTCP EM -Nb87 complex, 6,535,687 particles were template-picked from 21,792 micrographs, and classified through several rounds of 2D and 3D ab initio classifications (Extended Data Fig. 3). Around 220,000 selected particles were further classified by heterogenous refinement, yielding a final set of 61,053 particles that were processed by non-uniform refinement 60 . Further focused refinement excluding nanodisc scaffold yielded a final map at an overall resolution of 3.7 Å, based on the gold-standard 0.143 FSC cut-off. Maps were visualized using UCSF Chimera 61 and ChimeraX 62 .
The cryo-EM map of the NTCP EM -Mb91 complex showed clear density for most sidechains in the transmembrane helices, although TM1 and TM6 in the panel domain displayed fewer molecular features, and was used to build an atomic model of NTCP EM using Coot 63,64 . Secondary structure predictions using Psipred 65 and bacterial homologue structure (Protein Data Bank ID 3ZUY) were used to help initial sequence assignment. Initial Nb models were created with I-TASSER 66 , and then fit as rigid bodies into the density, followed by manual building and modification in Coot 63,64 . The inward-facing conformation in the NTCP EM -Nb87 complex was built by fitting core and panel domains from NTCP EM -Mb91 structure as separate rigid bodies into the density, followed by manual modification in Coot. All atomic models were refined using PHENIX 67 .
Structural analyses were carried out as follows: protein cavity calculations with CASTp 3.0 68 , pore calculations MOLEonline 2.5 69 , proteinprotein interfaces with PISA 70 , and amino acid conservation surface mapping with ConSurf 71 .

Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this paper.

Data availability
Structural models of NTCP EM -Nb87 and NTCP EM -Mb91 complexes have been deposited in the Protein Data Bank (PDB) with accession codes 7PQG and 7PQQ, respectively, and the corresponding cryo-EM maps were deposited in the Electron Microscopy Data Bank (EMDB) under accession numbers EMD-13593 and EMD-13596. Materials are available upon reasonable request and signing of a Material Transfer Agreement.