The drug/metabolite transporter (DMT) superfamily is a large group of membrane transporters ubiquitously found in eukaryotes, bacteria and archaea, and includes exporters for a remarkably wide range of substrates, such as toxic compounds and metabolites1. YddG is a bacterial DMT protein that expels aromatic amino acids and exogenous toxic compounds, thereby contributing to cellular homeostasis2,3. Here we present structural and functional analyses of YddG. Using liposome-based analyses, we show that Escherichia coli and Starkeya novella YddG export various amino acids. The crystal structure of S. novella YddG at 2.4 Å resolution reveals a new membrane transporter topology, with ten transmembrane segments in an outward-facing state. The overall structure is basket-shaped, with a large substrate-binding cavity at the centre of the molecule, and is composed of inverted structural repeats related by two-fold pseudo-symmetry. On the basis of this intramolecular symmetry, we propose a structural model for the inward-facing state and a mechanism of the conformational change for substrate transport, which we confirmed by biochemical analyses. These findings provide a structural basis for the mechanism of transport of DMT superfamily proteins.


An important physiological function is the expulsion of various compounds from cells to the extracellular space, which is essential for cellular homeostasis. This process involves specific membrane transporters that export their substrates across the cellular membrane. An example is the exporters of toxic compounds, which are crucial for the growth of microorganisms in the presence of antibiotics and antiseptics4,5. These drug exporters cause the emergence of multi-drug-resistant strains, which are a major obstacle to the effective treatment of bacterial infections6,7. Furthermore, the exporters of metabolites, such as amino acids and sugars, are important for maintaining their appropriate concentrations in the cytosol8,9,10. These metabolite transporters have crucial roles in multicellular organisms to direct metabolites to their appropriate locations, including tissues and cellular compartments.

The DMT superfamily is a large group of membrane transporters, comprising more than 32 families1. Numerous members of this superfamily are involved in the export of a wide range of substrates, including drugs and metabolites, and DMT proteins are ubiquitously distributed in eukaryotes, bacteria and archaea. For example, nucleotide sugar transporter family proteins export nucleotide–sugar conjugates (such as UDP–galactose and CMP–sialate) to the Golgi apparatus and endoplasmic reticulum of eukaryotic cells to supply building blocks for the sugar chains of glycoproteins, glycolipids and polysaccharides11,12. Many DMT proteins are predicted to contain ten transmembrane segments with a five-transmembrane unit internal repeat, which was probably formed by gene duplication1,13,14,15. Despite the importance of the DMT superfamily proteins, their structural mechanism of drug/metabolite transport has remained elusive.

YddG, a bacterial inner-membrane protein belonging to the DMT superfamily, exports drugs and metabolites. YddG from Salmonella enterica sv. Typhimurium is involved in the efflux of the di-cationic herbicide methyl viologen, and is postulated to be important for the efflux of multiple toxic compounds2. E. coli YddG (EcYddG) exports aromatic amino acids, and is essential for alleviating the growth inhibition caused by their excessive cytosolic accumulation3. To explore the substrate specificity of YddG further, we performed an in vitro functional analysis using YddG-reconstituted proteoliposomes (Fig. 1a). The results showed that EcYddG transports various amino acids, including threonine, methionine, lysine and glutamic acid (Fig. 1b), suggesting the broad substrate specificity of EcYddG. Although we could not examine the transport activity of aromatic amino acids in the in vitro assay system, because of their low solubility, YddG probably transports aromatic amino acids as well, based on a previous genetic analysis3. We also confirmed the in vivo amino acid export activity of EcYddG by a metabolomics analysis (Extended Data Fig. 1). We identified YddG from S. novella (SnYddG, 287 amino acids, 28% sequence identity with EcYddG) as a suitable candidate for structural studies, by the fluorescence-based screening method16,17. The transport activity of SnYddG was confirmed using the in vitro assay system (Fig. 1), which showed that SnYddG is also an amino acid transporter with broad substrate specificity.

Figure 1: Functional characterization of YddG.
Figure 1

a, b, Liposome-based analyses of EcYddG and SnYddG transport activities. a, Time-dependent uptake of [14C]threonine into empty, EcYddG- and SnYddG-containing liposomes. b, Uptake of 14C-labelled amino acids into empty, EcYddG- and SnYddG-containing liposomes, after 30 min. c.p.m., counts per minute. Error bars, s.d.; n = 3. **P < 0.01; ***P < 0.001 compared with empty liposomes (Student’s t-test).

We determined the crystal structure of SnYddG by the single-isomorphous replacement with anomalous scattering method, using Hg-derivatized crystals, and refined it at 2.4 Å resolution (Extended Data Table 1, Extended Data Fig. 2a, b). The asymmetric unit of the crystalline lattice contains six SnYddG molecules with nearly identical structures, and they are superimposable with root mean square deviations (r.m.s.d.) of less than 0.9 Å. Thus, we hereafter focus on molecule B in the asymmetric unit, as the quality of its electron density is the best among these molecules. The overall structure of YddG is basket-shaped, with a deep cavity facing the extracellular solvent (Fig. 2a, Extended Data Fig. 2c). As expected from previous informatics and biochemical analyses18, YddG comprises 10 α-helical transmembrane segments, with its N and C termini located on the intracellular side. The topology is composed of four pairs of two consecutive transmembrane segments forming two-helix hairpins; that is, transmembrane (TM) 1–TM2, TM3–TM4, TM6–TM7 and TM8–TM9, which are arranged alternately to surround the central cavity (Fig. 2b). Namely, the transmembrane segments in the N-half (TM1–TM5) and the C-half (TM6–TM10) surround the central cavity in anticlockwise and clockwise manners, respectively, as viewed from the periplasmic side (Fig. 2a, b). TM5 and TM10 form a four-helix bundle together with TM4 and TM9, which seals one side of the central cavity. TM4 and TM9 are respectively interrupted by short loops, with sequences that are well conserved among the YddG proteins from other species (Extended Data Fig. 3a), and thereby form the short helical segments, TM4a, TM4b, TM9a and TM9b. The N and C halves of SnYddG share weak sequence similarity (Extended Data Fig. 3b). Accordingly, the structures of these two halves are related by two-fold pseudo-symmetry with an axis running parallel to the membrane, and superimpose well with an r.m.s.d. of 2.7 Å for 90 Cα atoms (Fig. 2c). Notably, the resulting topology of YddG is unique and completely different from those of membrane transporters with known structures.

Figure 2: Overall structure of S. novella YddG.
Figure 2

a, Overall structure in cylinder representations, viewed from the plane of the membrane (left) and the periplasmic side (right). b, Topology diagram of SnYddG, coloured as in a. The primary structural organization of the transmembrane segments and the N and C halves is illustrated at the bottom. c, Structural superimposition of the N and C halves of SnYddG, viewed from two different directions. In the right panel, the transmembrane segments that do not overlap in the superimposition are indicated by arrows. All molecular graphics were illustrated with CueMol (http://www.cuemol.org/).

Previous bioinformatics analyses suggested that the small multidrug resistance (SMR) family is the progenitor of the DMT proteins1,13,14,15. E. coli EmrE is the best-characterized member of the SMR family19,20. The crystal structure of EmrE at 3.8 Å resolution revealed its dimeric architecture, with four-transmembrane segment protomers21. Although there is no detectable sequence similarity and their transmembrane topologies are different, the superimposition of the transmembrane helices of YddG and the EmrE dimer (PDB accession 3B5D) revealed good structural alignment (r.m.s.d. of 2.9 Å over 127 Cα atoms) (Extended Data Fig. 4a–c). The superimposition suggests the possible evolutional relationship between the four-transmembrane SMR and other ten-transmembrane DMT proteins (see Supplementary Information).

The central cavity of YddG deeply penetrates the inner leaflet of the membrane, and is formed by six transmembrane segments: TM1, TM3, TM4, TM6, TM8 and TM9 (Fig. 3a). Notably, TM1, TM4, TM6 and TM9 contain several residues conserved among YddGs from other species (Extended Data Fig. 3a). At the centre of the molecule, the strictly conserved Trp residues, Trp17 (TM1), Trp101 (TM4a), and Trp163 (TM6), form the bottom of the cavity (Fig. 3a). The wall of the central cavity is created by the conserved hydrophobic residues, Leu20 (TM1), Phe40 (TM2), and Phe225 (TM8) (Fig. 3a). Notably, a large density blob is observed in the central cavity (Fig. 3a). The shape of this density fits well with the monoolein molecule used in the LCP crystallization, suggesting that monoolein is bound to this site. This density peak interacts with the conserved residues, including Trp17, Tyr78, Trp101 and Trp163. Thus, we proposed that this cavity functions as a substrate-binding pocket, where these conserved hydrophobic residues bind the hydrophobic groups of the substrates. Moreover, Tyr78 (TM3), Tyr82 (TM3), and Tyr99(TM4) are located in the central cavity (Fig. 3a), and may provide both hydrophobic and hydrophilic environments for substrate binding. In addition, several hydrophilic residues, including His79 (TM3), Ser244 (TM9) and Ser251 (TM9), are also present in the central cavity (Fig. 3a), and may provide binding sites for the hydrophilic groups of the substrates.

Figure 3: Central cavity and intracellular gate.
Figure 3

a, The structure of the central cavity. In the right panel, the monoolein molecule bound to the cavity and its mFoDFc omit map, contoured at 3σ, are shown. b, c, Liposome-based analyses of SnYddG mutants of the central cavity, using threonine (b) and methionine (c) as substrates. The substrate uptake activities were plotted as percentages of the wild-type transport activity. Error bars, s.d.; n = 3. df, The structure of the intracellular gate. The mFo − DFc maps, calculated by omitting the key residues for the interactions and contoured at 3σ, are shown.

To explore the functional importance of these conserved residues for the substrate recognition and transport activity, we measured the transport activities of SnYddG mutants by a liposome-based assay, using 14C-labelled threonine and methionine (Fig. 3b, c). The structural integrity of the mutant proteins was verified by gel-filtration chromatography at the final purification step (Extended Data Fig. 5a). The activity was normalized to the amount of protein reconstituted into the proteoliposomes (Extended Data Fig. 5b). The results showed that the His79Ala mutant abolished the transport activities for both threonine and methionine, thus revealing the critical role of this hydrophilic residue. Furthermore, the Trp101Ala and Trp163Ala mutants exhibited decreased transport activities for threonine, but not for methionine, suggesting the importance of these aromatic residues for specific types of substrates. In contrast, the Tyr78Ala mutation showed moderate effects on the transport activities of both threonine and methionine, suggesting that Tyr78 is involved in, but not crucial for, the recognition and/or transport of these substrates. The Tyr82Ala mutation enhanced methionine transport, and slightly reduced threonine transport. This mutation could increase the size of the substrate-binding site in the inward-facing state, which may facilitate the transport of large substrates. Taken together, our results strongly suggest that the central cavity functions as the binding site for a wide range of YddG substrates.

The intracellular side of the central cavity is closed by the intracellular gate, which is formed by the interactions among the side-by-side helices of TM4b and TM9a, and the intracellular tips of the TM6–TM7 and TM8–TM9a hairpins (Fig. 2a). Just beneath the central cavity, the conserved Trp228 (TM8) and Met232 (TM8) residues form the hydrophobic core of the intracellular gate (Fig. 3d). Trp228 hydrophobically interacts with the conserved Trp17 and Trp163, which form the bottom of the central cavity. The NE1 atom of the Trp228 side chain hydrogen bonds with the Ser167 (TM6) side chain. Furthermore, Met232 forms the hydrophobic core with Ile105 (TM4b), Val168 (TM6) and Val237 (TM9a), which are weakly conserved as hydrophobic residues among the YddGs from other species. In the vicinity of this hydrophobic core, the main-chain carbonyl group of Phe225 and the side chains of His70, Tyr166, Ser170 and Asp229 form a hydrogen-bonding network (Fig. 3e). Water molecules are captured by this hydrogen-bonding network, suggesting that the interactions in the intracellular gate are impermanent and can dissociate during the transport cycle. Furthermore, Arg171 (TM6) and Lys233 (TM8) form hydrogen bonds with the main-chain carbonyl groups in the TM8–TM9a and TM6–TM7 loops, respectively, which seal the intracellular side of the intracellular gate (Fig. 3f). Together, these tight interactions separate the central cavity from the intracellular space.

While the present crystal structure of YddG represents the outward-facing state in the alternating-access mechanism, the structural and sequence similarities between the N and C halves (Fig. 2c and Extended Data Fig. 3b) allowed us to generate a feasible structural model for the inward-facing state, as in other secondary transporters with inverted structural repeats22,23,24 (Fig. 4a, b). In this model of the inward-facing state, the intracellular gate interactions observed in the outward-facing structure are completely dissociated, thus opening the pathway directed towards the intracellular side (Fig. 4a). In contrast, the extracellular gate is formed by the interactions among the TM1–TM2 and TM3–TM4a hairpins, and TM4a and TM9b without any obvious steric clashes, and thus the substrate-binding site is occluded from the extracellular side. In the crystal structure of the outward-facing state, TM9a contains the Gly241 residue, which enables the tight side-by-side interaction between TM4b and TM9a (Fig. 4a). TM4a also contains the Gly95 residue, which could enable a similar side-by-side interaction between TM4a and TM9b in the inward-facing structure (Fig. 4b). Moreover, hydrophobic packing interactions are probably formed between the extracellular tips of the TM1–TM2 and TM3–TM4a hairpins, to create the extracellular gate. The side chain of Leu86 (TM3) may be surrounded by hydrophobic residues, including Ala21 (TM1) and Tyr82 (TM3), and Tyr82 may form the top of the substrate-binding site in the inward-open state (Fig. 4b). The results of the Cys-crosslinking experiment, as well as an evolutionary covariation analysis of YddG homologues, provide strong support for the extracellular gate formation and our inward-facing model structure (Extended Data Figs 6 and 7; see Supplementary Information for further discussion).

Figure 4: The outward–inward conformational change of SnYddG.
Figure 4

a, Crystal structure of the outward-facing state (left) and the modelled structure of the inward-facing state (right), viewed from the cytoplasmic side. The residues involved in the intracellular gate are shown in stick models. b, The same structures as in a, viewed from the periplasmic side. The residues possibly involved in the extracellular gate are shown in stick models. c, Proposed transport mechanism, illustrating the bending motion of the transmembrane segments during the outward–inward conformational change. The molecular envelopes are indicated by dotted curves.

A comparison between the inward-facing model and outward-facing crystal structures provides further insights into the structural changes that occur during the transport cycle (Fig. 4c and Supplementary Video 1). The structures of TM3 and TM4 suggest the bending and straightening of the extracellular halves of these transmembrane segments, with the region around Gly71–Gly77 in TM3 (Extended Data Fig. 8a), as well as the intra-membrane loop in TM4, serving as hinges. The bending and straightening of the TM3–TM4 hairpin may further involve the tilting and upright motions of TM6, which collectively close and open the extracellular entrance of the central cavity (Fig. 4c). Similar structural changes may occur in TM8, TM9 and TM1, which are related by the intramolecular pseudo-symmetry to TM3, TM4 and TM6. Along with the tilting and upright motions of TM1, the hinge motions in the TM8–TM9 hairpin occur around Gly217–Gly222 in TM8 (Extended Data Fig. 8b) and the intra-membrane loop in TM9. These structural changes collectively close and open the intracellular entrance of the central cavity (Fig. 4c). The results from the molecular dynamics simulations also supported this structural change mechanism (Extended Data Fig. 9; see Supplementary Information for further discussion).

In summary, we determined the crystal structure of SnYddG at 2.4 Å resolution, which revealed the novel membrane transporter topology, with 10-transmembrane segments. The structural and complementary functional analyses suggested that YddG operates by a unique type of alternating-access mechanism, which is completely different from those of other known transporters. Our results provide further insight into the common transport mechanism shared among the DMT superfamily members, including the SMR transporters.


Cloning and expression of YddG

The S. novella yddG gene (gi:502932551) was cloned from S. novella genomic DNA (Strain: JCM 20403) into a plasmid derived from the pET expression vector, which includes a C-terminal (His)8-tag and a tobacco etch virus (TEV) protease cleavage site. The SnYddG protein was overexpressed in E. coli Rosetta2 (DE3) strain cells, grown in LB medium containing ampicillin (50 μg ml−1). When the culture reached an absorbance at 600 nm of ~0.5, the cells were induced with 0.5 mM isopropyl β-thiogalactopyranoside (IPTG) for 2 h at 37 °C. The E. coli yddG gene (gi:152031741) was cloned from E. coli K-12 genomic DNA (strain: JCM 20135) into a plasmid derived from the expression vector pCGFP-BC16, which includes a C-terminal green fluorescent protein (GFP), a (His)8-tag and a TEV protease cleavage site. The EcYddG protein was overexpressed in E. coli C41(DE3)ΔacrB cells, grown in LB medium containing ampicillin (50 μg ml−1). When the culture reached an absorbance at 600 nm of ~0.5, the cells were induced with 0.5 mM IPTG for 18 h at 20 °C.

Purification and crystallization of SnYddG

The SnYddG protein for crystallization were purified according to the following procedure at 4 °C. The cells were pelleted by centrifugation at 4,500g, and were disrupted by a Microfluidizer (Microfluidics). After centrifugation (12,000g), the supernatant was ultra-centrifuged (200,000g), and the membrane fraction was collected. The proteins were solubilized from the membrane fraction with 50 mM HEPES (pH 7.0), containing 300 mM NaCl, 20 mM imidazole, 1 mM phenylmethylsulfonyl fluoride, 1.2% (w/v) DDM, 0.24% (w/v) cholesteryl hemisuccinate (CHS), and were purified by the following three chromatography steps. The insoluble material was removed by ultracentrifugation (Beckman Type 70 Ti rotor, 150,000g, 30 min), and the supernatant was mixed with Ni-NTA resin (QIAGEN). The (His)8-tag was cleaved by TEV protease at 4 °C overnight, and the proteins were re-chromatographed on a Ni-NTA column. The (His)8-tag-cleaved protein was further purified by gel-filtration chromatography (Superdex 200 Increase 10/300 GL, GE Healthcare) in 20 mM HEPES (pH 7.0), containing 150 mM NaCl, 0.03% (w/v) DDM and 0.006% (w/v) CHS.

For crystallization, the purified protein was concentrated to approximately 15 mg ml−1, using an Amicon Ultra 50K filter (Millipore). SnYddG was mixed with liquefied monoolein (Sigma) in a 2:3 protein to lipid ratio (w/v), using the twin-syringe mixing method. For the sandwich-drop crystallization, aliquots of the protein-LCP mixture were dispensed onto 96-well glass plates and overlaid with the precipitant solution, using a Gryphon LCP (Art Robbins Instruments, LLC). Initial crystallization conditions were searched, using screening kits including MemMeso, MemGold I and II, and MemStart/MemSys (Molecular Dimensions). The initial hits were optimized by changing the concentration of each component, as well as additive screening, using the hanging-drop crystallization method. For the hanging-drop crystallization, the protein-LCP drops were manually spotted onto siliconized glass coverslips and overlaid with the precipitant solutions, and then the coverslips were placed upside down onto 24-well plates and sealed with each well containing 300 μl of reservoir solution. We finally found that the addition of (NH4)2SO4 to the precipitant solution markedly improved the size of the crystals. The native crystals were grown in hanging-drop plates at 20 °C, with 50 nl protein-LCP drops overlaid with 800 or 1,600 nl precipitant solution, which consisted of 32–34% PEG550MME, 100 mM Na-citrate (pH 4.5), 100 mM (NH4)2HPO4, and 100 mM (NH4)2SO4. The heavy atom-derivatized crystals were prepared by the soaking method. After the native crystals were grown on hanging-drop plates to the full size, the overlaid crystallization solution was replaced with 2,400 nl of the solution supplemented with a slightly higher concentration of PEG550MME and 1 mM CH3HgCl. The crystals were incubated at 20 °C for 3 h. All of the crystals were flash-cooled in liquid nitrogen for data collection, using the reservoir solution as a cryoprotectant.

Data collection and structure determination of SnYddG

All diffraction data sets were collected at the station BL32XU at SPring-8. Data sets were processed with the program XDS25 and the CCP4 suite26. The data processing statistics are summarized in Extended Data Table 1. The structure was determined by the single isomorphous replacement with anomalous scattering method, using the native and CH3HgCl-soaked SnYddG crystals. Twenty-four Hg atom sites were identified with the program SHELXD27. The initial phases were calculated with the program SHARP28. The resulting phases were improved by solvent flattening with the program SOLOMON29 and six-fold non-crystallographic symmetry averaging with the program DM30. The initial model was built into the map, using the program COOT31. The model was subsequently improved through alternating cycles of manual building with COOT and refinement with the program PHENIX32. The structural refinement statistics are summarized in Extended Data Table 1. Molecular graphics were illustrated with CueMol (http://www.cuemol.org/).

Metabolomics analysis

To construct the assay strain W3110 (DE3) ΔyddG::Km containing the IPTG-inducible T7 polymerase expression unit (λDE3), the yddG gene knockout allele was transferred to the destination E. coli strain W3110 (DE3)33 by P1 phage transduction from the systematic E. coli knockout library strain, with kanamycin resistance as the selection marker34. The knockout allele was confirmed by a PCR analysis, using the DNA primers 5′-ATAGCGGTAGAAAAACGCACCA-3′ and 5′-TGAGATATAAGGTGAATTACTGGTATTTG-3′. E. coli strains W3110 (DE3) and W3110 (DE3) ΔyddG::Km cells, cultivated on LB plates, were inoculated into 5 ml M9 medium, containing 0.5% glucose, and shaken at 37 °C. When the optical density reached 0.5, IPTG was added to the medium (final concentration 1 mM), and subsequently, the cells were cultivated for approximately 24 h. After the cells were removed by centrifugation and filtration, the supernatants were analysed by capillary electrophoresis–mass spectrometry (CE–MS) at Human Metabolome Technologies Inc.

Liposome-based assay

The SnYddG protein for liposome assay was purified by the same procedure as those for the crystallization, but 1.2% (w/v) DDM was used for solubilization and 0.25% (w/v) n-decyl-β-d-maltopyranoside (DM) was included in every step after solubilization to gel filtration. The EcYddG protein for liposome assay was also purified by the same procedure as the SnYddG protein. The purified SnYddG and EcYddG proteins were reconstituted into liposomes by the following procedure. An E. coli polar lipid extract (Avanti) was dissolved in chloroform and dried into a thin film. This film was then resuspended to a final concentration of 20 mg ml−1 in 10 mM HEPES buffer (pH 7.0) containing 100 mM NaCl, and sonicated for 2 min to obtain the liposome solution. The purified proteins were added to the liposome solution at a lipid to protein ratio of 50:1 (w/w). The protein–liposome mixtures were incubated at 4 °C for 30 min, and then ultra-centrifuged (200,000g) at 4 °C for 3 h to remove the detergent. The proteoliposomes were re-suspended to a final concentration of 20 mg ml−1 and stored at −80 °C. Protein-free liposomes were prepared by a similar procedure, except that the protein solution was replaced with the buffer used for the final purification step. The liposomes were sonicated immediately before the measurements to prepare uniformly-sized liposomes. The time-dependent [14C]Thr (175 mCi mmol−1; Moravek Biochemicals) uptake assay was initiated by mixing the liposome solution (45 μl) with an equal volume of the extraliposomal solution, consisting of 10 mM HEPES (pH 7.0), 100 mM NaCl, 100 μM amino acid, and 2% (v/v) [14C]amino acid. After the reaction at 37 °C, the aliquot (20 μl) of the reaction mixture were isolated by Sephadex G-50 (GE Healthcare) gel filtration, and the radioactivity of the incorporated [14C]amino acid was measured by liquid scintillation counting. The Met-, Glu- and Lys-uptake assays were also performed with a similar condition to that of Thr-uptake assay, using [14C]Met, [14C]Glu and [14C]Lys (55 mCi mmol−1, 210 mCi mmol−1 and 288 mCi mmol−1, respectively; American Radiolabelled Chemicals). For mutational analyses, mutations were introduced by a PCR-based method. The mutant proteins were expressed, purified, and reconstituted into liposomes, and the transport activities were measured by a similar procedure to that for the wild type. The assays were initiated by mixing the liposome solution (25 μl) with an equal volume of the extraliposomal solution, and after the 30-min reaction at 37 °C, the aliquot (20 μl) were isolated for the subsequent radioactivity measurement. The reconstitution rates of the wild-type and mutant proteins were determined by fractionating the proteoliposome samples on an SDS–PAGE gel, and quantifying the amount of SnYddG protein by an LAS-3000 image analyser. All assays were repeated three times. Error bars represent s.d.

Cysteine cross-link analysis of SnYddG

The Cys-free mutant of SnYddG (C159A/C185A/195A/C271A) was constructed using a PCR-based method. The double-Cys mutants (A21C/P91C, A21C/A92C, T24C/A85C, T24C/L86C, A138C/A266C) were also constructed using a PCR-based method, based on the Cys-free mutant. The mutant proteins for the cross-link analysis were prepared by the same procedure as those for the crystallization, but 1.2% (w/v) DDM was used for solubilization and 0.03% (w/v) DDM was included in every step after solubilization to gel filtration. The purified mutant proteins were incubated with 10 mM tris(2-carboxyethyl)phosphine (TCEP) or 1 mM copper phenanthroline [Cu(phen)3] at 37 °C for 30 min, followed by trichloroacetic acid precipitation. The pellets were dissolved in SDS–PAGE sample buffer, containing 1% SDS and 20 μM tetramethylrhodamine maleimide (TMRM), incubated at 37 °C for 90 min, and then analysed by SDS–PAGE. The fluorescence of the TMRM-modified proteins was visualized with an LAS-3000 image analyser.

For quantification of intramolecular disulphide formation by SnYddG double-Cys mutants, they were subjected to reductive or non-reductive carboxymethylation. For reductive carboxymethylation, the protein (~2 μg) was dissolved in 20 μl of 1% dithiothreitol in 6 M guanidine hydrochloride, 1 M Tris-HCl (pH 8.5) and 10 mM EDTA, and was heated at 80 °C for 30 min. After cooling, alkylation was performed by the addition of 2 μl of a 25% iodoacetic acid solution in 1 N NaOH and an incubation at room temperature for 30 min in the dark. For non-reductive carboxymethylation, the protein was dissolved in 2.5% iodoacetic acid in 6 M guanidine HCl, 1 M Tris-HCl (pH 8.5) and 10 mM EDTA. The reaction mixtures were desalted with a Sephadex G-25 syringe (1 ml), and pooled fractions of the carboxymethylated protein were dried and hydrolysed in 6 N HCl vapour at 110 °C for 20 h. The acid hydrolysate was derivatized with 6-aminoquinolyl-N-hydroxysuccinimidyl carbamate, and was quantified as described previously35 (Extended Data Fig. 6b).

Co-evolutionary analysis

The co-evolutionary analysis of YddG homologues and other DMT proteins was performed using the EVcoupling36 web interface (http://evfold.org/evfold-web/evfold.do). The default parameters of the web interface were employed for the calculations, except that the E-value threshold for generating the sequence alignment was changed to −30. The resulting number of YddG and DMT homologue sequences used for the calculation was 59,114. The subsequent 3D-structure prediction of SnYddG was performed by the program EVfold_membrane37. The default parameters of the web interface were employed for the calculation, except that the ‘membrane protein’ option was turned on and the numbers of flanking upstream and downstream residues in the secondary structure prediction were changed to 0.

Molecular dynamics simulation

The atomic coordinates of the crystal structure of SnYddG (molecule B) were used for the simulation. The disordered region (Ala138–Gly144) was modelled by adding the corresponding coordinates in molecule E. All of the water molecules observed in the crystal structure were kept. The missing hydrogen atoms were built with the program VMD38. A periodic boundary system, including explicit solvent and a phosphoryloleoylphosphatidylethanolamine (POPE) lipid bilayer39, was prepared. The net charge of the simulation system was neutralized through the addition of 150 mM NaCl. The simulation system was 96 × 96 × 96 Å, and contained 80,530 atoms. The molecular topologies and parameters from the Charmm36 force field39 were used for the protein, lipid and water molecules.

Molecular dynamics simulations were performed with the program NAMD 2.10 (ref. 40). The systems were first energy minimized for 1,000 steps with fixed positions of the non-hydrogen atoms, and then for another 1,000 steps with 10 kcal mol−1 restraints for the non-hydrogen atoms, except for the lipid molecules within 5.0 Å from the proteins. Next, equilibrations were performed for 0.01 ns under NVT conditions, with 10 kcal mol−1 restraints for the heavy atoms of the protein. Finally, equilibrations were performed for 0.5 ns under NPT conditions with the 1.0 kcal mol−1 restraints. In the equilibration and production processes, the pressure and temperature were set to 1.0 atm and 310 K, respectively. Constant temperature was maintained by using Langevin dynamics. Constant pressure was maintained by using the Langevin piston Nosé–Hoover method41. Long-range electrostatic interactions were calculated by using the particle mesh Ewald method42. The production run of the equilibrium simulation was performed for 500 ns, starting from the crystal structure. The outward-to-occluded simulation run was also performed for 500 ns, starting from the crystal structure, with a harmonic distance restraint (force constant = 10.0 kcal mol−1 Å−2) between centres of mass of the Cα atoms of TM4a (Pro91–Ala98) and TM9b (Ala246–Leu253). The equilibrium distance of the harmonic restraint was gradually decreased from 15 Å to 9 Å during the 500-ns run. Next, the occluded-to-inward simulation run was performed, starting from the final snapshot of the outward-to-occluded simulation. A similar harmonic distance restraint was applied between centres of mass of the Cα atoms of TM4b (Trp101–Phe108) and TM9a (Val237–Ser244), with the equilibrium distance gradually increased from 9 Å to 15 Å.

Statistical analyses

The statistical significance of differences in mean values was calculated using unpaired, two-tailed Student’s t-test. No statistical methods were used to predetermine sample size.


Primary accessions

Protein Data Bank

Data deposits

The atomic coordinates and structure factors for SnYddG have been deposited in the Protein Data Bank (PDB) under accession number 5I20.


  1. 1.

    , & The drug/metabolite transporter superfamily. Eur. J. Biochem. 268, 3620–3639 (2001)

  2. 2.

    et al. The Salmonella enterica sv. Typhimurium smvA, yddG and ompD (porin) genes are required for the efficient efflux of methyl viologen. Mol. Microbiol. 46, 687–698 (2002)

  3. 3.

    et al. YddG from Escherichia coli promotes export of aromatic amino acids. FEMS Microbiol. Lett. 275, 312–318 (2007)

  4. 4.

    Prevention of drug access to bacterial targets: permeability barriers and active efflux. Science 264, 382–388 (1994)

  5. 5.

    , & Proton-dependent multidrug efflux systems. Microbiol. Rev. 60, 575–608 (1996)

  6. 6.

    Efflux-mediated multiresistance in Gram-negative bacteria. Clin. Microbiol. Infect. 10, 12–26 (2004)

  7. 7.

    Multidrug-resistance efflux pumps - not just for resistance. Nature Rev. Microbiol. 4, 629–636 (2006)

  8. 8.

    Systems and mechanisms of amino acid uptake and excretion in prokaryotes. Arch. Microbiol. 162, 1–13 (1994)

  9. 9.

    , , & Identification and characterization of the new gene rhtA involved in threonine and homoserine efflux in Escherichia coli. Res. Microbiol. 154, 123–135 (2003)

  10. 10.

    et al. Expression control and specificity of the basic amino acid exporter LysE of Corynebacterium glutamicum. Microbiology 147, 1765–1774 (2001)

  11. 11.

    & Molecular physiology and pathology of the nucleotide sugar transporter family (SLC35). Pflugers Arch. 447, 768–775 (2004)

  12. 12.

    Roles of the nucleotide sugar transporters (SLC35 family) in health and disease. Mol. Aspects Med. 34, 590–600 (2013)

  13. 13.

    , & Evolution of antiparallel two-domain membrane proteins: tracing multiple gene duplication events in the DUF606 family. J. Mol. Biol. 378, 596–606 (2008)

  14. 14.

    , , , & Functional specialization in nucleotide sugar transporters occurred through differentiation of the gene cluster EamA (DUF6) before the radiation of Viridiplantae. BMC Evol. Biol. 11, 123 (2011)

  15. 15.

    & Diversity and evolution of the small multidrug resistance protein family. BMC Evol. Biol. 9, 140 (2009)

  16. 16.

    & Fluorescence-detection size-exclusion chromatography for precrystallization screening of integral membrane proteins. Structure 14, 673–681 (2006)

  17. 17.

    , & A fluorescence-detection size-exclusion chromatography-based thermostability assay for membrane protein precrystallization screening. Structure 20, 1293–1299 (2012)

  18. 18.

    et al. Membrane topology analysis of the Escherichia coli aromatic amino acid efflux protein YddG. J. Mol. Microbiol. Biotechnol. 19, 189–197 (2010)

  19. 19.

    , , & Emulating membrane protein evolution by rational design. Science 315, 1282–1284 (2007)

  20. 20.

    EmrE, a model for studying evolution and mechanism of ion-coupled transporters. Biochim. Biophys. Acta 1794, 748–762 (2009)

  21. 21.

    et al. X-ray structure of EmrE supports dual topology model. Proc. Natl Acad. Sci. USA 104, 18999–19004 (2007)

  22. 22.

    Structural biology. (Pseudo-)symmetrical transport. Science 339, 399–401 (2013)

  23. 23.

    et al. Structural insight into the ion-exchange mechanism of the sodium/calcium exchanger. Science 335, 686–690 (2012)

  24. 24.

    , , & Structure and mechanism of a bacterial sodium-dependent dicarboxylate transporter. Nature 491, 622–626 (2012)

  25. 25.

    XDS. Acta Crystallogr. D 66, 125–132 (2010)

  26. 26.

    et al. Overview of the CCP4 suite and current developments. Acta Crystallogr. D 67, 235–242 (2011)

  27. 27.

    & Substructure solution with SHELXD. Acta Crystallogr. D 58, 1772–1779 (2002)

  28. 28.

    & Maximum-likelihood heavy-atom parameter refinement for multiple isomorphous replacement and multiwavelength anomalous diffraction methods. Methods Enzymol. 276, 472–494 (1997)

  29. 29.

    & Methods used in the structure determination of bovine mitochondrial F1 ATPase. Acta Crystallogr. D 52, 30–42 (1996)

  30. 30.

    & Improvement of macromolecular electron-density maps by the simultaneous application of real and reciprocal space constraints. Acta Crystallogr. D 49, 148–157 (1993)

  31. 31.

    , , & Features and development of Coot. Acta Crystallogr. D 66, 486–501 (2010)

  32. 32.

    et al. PHENIX: Building new software for automated crystallographic structure determination. Acta Crystallogr. D 58, 1948–1954 (2002)

  33. 33.

    et al. Mg2+-dependent gating of bacterial MgtE channel underlies Mg2+ homeostasis. EMBO J. 28, 3602–3612 (2009)

  34. 34.

    et al. Construction of Escherichia coli K-12 in-frame, single-gene knockout mutants: the Keio collection. Mol. Syst. Biol. 2, 2006.0008 (2006)

  35. 35.

    & Examination of an absolute quantity of less than a hundred nanograms of proteins by amino acid analysis. Anal. Bioanal. Chem. 405, 8073–8081 (2013)

  36. 36.

    et al. Protein 3D structure computed from evolutionary sequence variation. PLoS ONE 6, e28766 (2011)

  37. 37.

    et al. Three-dimensional structures of membrane proteins from genomic sequencing. Cell 149, 1607–1621 (2012)

  38. 38.

    , & VMD: visual molecular dynamics. J. Mol. Graph. 14, 33–8, 27–8 (1996)

  39. 39.

    et al. CHARMM: the biomolecular simulation program. J. Comput. Chem. 30, 1545–1614 (2009)

  40. 40.

    et al. Scalable molecular dynamics with NAMD. J. Comput. Chem. 26, 1781–1802 (2005)

  41. 41.

    , , & Constant pressure molecular dynamics simulation: The Langevin piston method. J. Chem. Phys. 103, 4613 (1995)

  42. 42.

    , & Particle mesh Ewald: An Nlog(N) method for Ewald sums in large systems. J. Chem. Phys. 98, 10089 (1993)

  43. 43.

    & Secondary-structure matching (SSM), a new tool for fast protein structure alignment in three dimensions. Acta Crystallogr. D 60, 2256–2268 (2004)

Download references


We thank H. Nishimasu and M. Hattori for comments on the manuscript; T. Tsukazaki and D. Drew for discussion; Y. Lee and A. Kurabayashi for technical assistance; the RIKEN BioResource Center for providing Starkeya novella genomic DNA; and the beam-line scientists at BL41XU and BL32XU of SPring-8 for assistance with data collection. The diffraction experiments were performed at SPring-8 BL41XU and BL32XU (proposals 2014A1091, 2014A1061, 2014A1093, 2014A1116 and 2014B1194). This work was supported by grants from the Platform for Drug Discovery, Informatics and Structural Life Science by the Ministry of Education, Culture, Sports, Science and Technology (MEXT), JSPS KAKENHI (grants 24227004, 25291011 and 26711003), the FIRST program, and a Grant-in-Aid for JSPS Fellows. Partial calculations were performed on the HOKUSAI GreatWave supercomputer system at RIKEN and the NIG supercomputer at ROIS National Institute of Genetics.

Author information


  1. Department of Biological Sciences, Graduate School of Science, The University of Tokyo, 2-11-16 Yayoi, Bunkyo-ku, Tokyo 113-0032, Japan

    • Hirotoshi Tsuchiya
    • , Shintaro Doki
    • , Mizuki Takemoto
    • , Tatsuya Ikuta
    • , Takashi Higuchi
    • , Tomohiro Nishizawa
    • , Ryuichiro Ishitani
    •  & Osamu Nureki
  2. Research Institute for Bioscience Products & Fine Chemicals, Ajinomoto Co., Inc., 1-1 Suzuki-cho, Kawasaki-ku, Kawasaki 210-8681, Japan

    • Keita Fukui
  3. Institute for Innovation, Ajinomoto Co., Inc., 1-1 Suzuki-cho, Kawasaki-ku, Kawasaki 210-8681, Japan.

    • Yoshihiro Usuda
    •  & Eri Tabuchi
  4. Department of Bioengineering, School of Engineering, The University of Tokyo, 7-3-1 Hongo, Bunkyo-Ku, Tokyo 113-8656, Japan

    • Satoru Nagatoishi
    •  & Kouhei Tsumoto
  5. Department of Medical Genome Sciences, Graduate School of Frontier Sciences, The University of Tokyo, Chiba 277-8562, Japan

    • Koichi Ito
  6. Biomolecular Characterization Unit, RIKEN Center for Sustainable Resource Science, Wako, Saitama 351-0198, Japan

    • Naoshi Dohmae
  7. Theoretical Molecular Science Laboratory, RIKEN, Wako, Saitama 351-0198, Japan

    • Ryuichiro Ishitani


  1. Search for Hirotoshi Tsuchiya in:

  2. Search for Shintaro Doki in:

  3. Search for Mizuki Takemoto in:

  4. Search for Tatsuya Ikuta in:

  5. Search for Takashi Higuchi in:

  6. Search for Keita Fukui in:

  7. Search for Yoshihiro Usuda in:

  8. Search for Eri Tabuchi in:

  9. Search for Satoru Nagatoishi in:

  10. Search for Kouhei Tsumoto in:

  11. Search for Tomohiro Nishizawa in:

  12. Search for Koichi Ito in:

  13. Search for Naoshi Dohmae in:

  14. Search for Ryuichiro Ishitani in:

  15. Search for Osamu Nureki in:


H.T., S.D., T.H., and E.T. constructed the expression systems. H.T. expressed, purified, and crystallized SnYddG, collected the diffraction data, determined the structure, and performed liposome-based assays. K.T. and S.N. assisted with the liposome-based assays. M.T., T.N. and R.I. assisted with the structure determinations. K.I. constructed the E. coli knockout strain. K.F. and Y.U. performed the metabolomics analysis. M.T. and T.I. performed the molecular dynamics simulations. N.D. performed disulphide-bond quantification. H.T., R.I. and O.N. wrote the manuscript. R.I. and O.N. directed and supervised all of the research.

Competing interests

The authors declare no competing financial interests.

Corresponding authors

Correspondence to Ryuichiro Ishitani or Osamu Nureki.

Extended data

Supplementary information

PDF files

  1. 1.

    Supplementary Information

    This file contains a Supplementary Discussion and Supplementary References.


  1. 1.

    The outward-inward conformational change of SnYddG

    The video sequence showing the conformational changes from the outward- to inward-facing state of SnYddG, viewed from four different directions. The structure is shown with rainbow coloring, with the N- and C-termini colored blue and red, respectively. The residues involved in the intracellular and extracellular gates are shown in stick models.

About this article

Publication history






Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.