Crystal structure of the bacterial acetate transporter SatP reveals that it forms a hexameric channel

Acetate is found ubiquitously in the natural environment and can be used as an exogenous carbon source by bacteria, fungi, and mammalian cells. A representative member of the acetate uptake transporter (AceTr) family named SatP (also yaaH) has been preliminarily identified as a succinate–acetate/proton symporter in Escherichia coli. However, the molecular mechanism of acetate uptake by SatP still remains elusive. Here, we report the crystal structure of SatP from E. coli at 2.8 Å resolution, determined with a molecular replacement approach using a previously developed predicted model algorithm, which revealed a hexameric UreI-like channel structure. Structural analysis identified six transmembrane (TM) helices surrounding the central channel pore in each protomer and three conserved hydrophobic residues, FLY, located in the middle of the TM region for pore constriction. According to single-channel conductance recordings, performed with purified SatP reconstituted into lipid bilayer, three conserved polar residues in the TM1 facing to the periplasmic side are closely associated with acetate translocation activity. These analyses provide critical insights into the mechanism of acetate translocation in bacteria and a first glimpse of a structure of an AceTr family transporter.

Acetate is found ubiquitously in the natural environment and can be used as an exogenous carbon source by bacteria, fungi, and mammalian cells. A representative member of the acetate uptake transporter (AceTr) family named SatP (also yaaH) has been preliminarily identified as a succinate-acetate/proton symporter in Escherichia coli. However, the molecular mechanism of acetate uptake by SatP still remains elusive. Here, we report the crystal structure of SatP from E. coli at 2.8 Å resolution, determined with a molecular replacement approach using a previously developed predicted model algorithm, which revealed a hexameric UreI-like channel structure. Structural analysis identified six transmembrane (TM) helices surrounding the central channel pore in each protomer and three conserved hydrophobic residues, FLY, located in the middle of the TM region for pore constriction. According to single-channel conductance recordings, performed with purified SatP reconstituted into lipid bilayer, three conserved polar residues in the TM1 facing to the periplasmic side are closely associated with acetate translocation activity. These analyses provide critical insights into the mechanism of acetate translocation in bacteria and a first glimpse of a structure of an AceTr family transporter.
Acetate is a common anion in all organisms. In microorganisms, acetate is involved in the regulation of multiple functional processes, including the formation of biofilms, responses to stress, and modulation of pathophysiological responses (1). Furthermore, acetate can be used as a single carbon source through the conversion of acetic acid to acetyl-CoA (1). In mammals, acetate can also be used in lipid synthesis and energy production (2). This is especially true for several types of cancer cells that rapidly consume acetate to facilitate growth and proliferation (3)(4)(5).
Acetate is a weak carboxylic acid that partially dissociates in aqueous solution and establishes an equilibrium between uncharged and anionic forms. Therefore, the translocation processes of acetate can be divided into two major paths: simple diffusion and facilitated translocation (6). Currently, there are several kinds of transporters, or channels, that have been identified to assist the translocation of acetate and other monocarboxylates. In mammals, the monocarboxylate transporter (MCT) is responsible for the transport of acetate (7,8). Furthermore, the specific mitochondrial pyruvate carrier (MPC) formed by MPC1 and MPC2, which is conserved from yeast to mammals, also has a similar function of pyruvate transport (9). Ady2, a member of the GPR1/FUN34/ YaaH family, and Jen1 mediate the permeation of monocarboxylates in fungi (6,10,11). In Escherichia coli, FocA is a representative member of the formate-nitrate transporter family (TC 2.A.44), which has the conserved function of transporting of short-chain carboxylates in parasites (12). In addition, acetate permease (ActP) (13) and succinate-acetate transporter (SatP) (14) belong to the sodium solute symporter (TC 2.A.21) and acetate uptake transporter family (AceTr, TC 2.A.96), respectively.
Monocarboxylate transport has already been identified for decades. However, the structure and transport mechanisms of monocarboxylate transporters are complicated, except the formate transporter FocA, which reveals a pentameric, aquaporinlike channel (15). The SatP in E. coli (ecSatP) 4 has been predicted to consist of six transmembrane (TM) helices (14). Additionally, its sequence is not similar to the known structures of membrane proteins, which strongly indicates that its fold structure is novel. The unclear translocational mechanism of acetate and novel fold structure of the AceTr family may be highly conserved from bacteria to fungi based on their sequence similarity. Thus, we performed a structural and functional study of ecSatP to elucidate its function at the molecular level.

Characterization of purified ecSatP
We successfully purified the full-length ecSatP protein with a monodispersed behavior in aqueous solution using various  1 These authors contributed equally to this work. 2 To whom correspondence may be addressed. E-mail: geng.jia@scu.edu.cn. 3 To whom correspondence may be addressed. E-mail: dengd@scu.edu.cn. detergents (Fig. S1A). Surprisingly, ecSatP, with a molecular mass of 20 kDa, was eluted at ϳ12 ml from size-exclusion chromatography, which suggested that ecSatP maintains more than 158 kDa within detergent micelles and indicated that ecSatP stabilizes at an oligomeric state (Fig. S1A). Consequently, we carried out an in vitro cross-link assay using purified ecSatP, and the result strongly implied that ecSatP was present in a hexameric state (Fig. S1B).
Moreover, to confirm the functionality of the oligomeric ecSatP in detergent micelles, purified ecSatP was reconstituted into lipid bilayer and measured using a single-channel conductance assay (Fig. 1A). We generated a planar bilayer membrane with a 150-m diameter aperture separating the cis-compartment and trans-compartment. After the addition of the ecSatP protein into the cis-compartment, a stepwise current increase was observed as shown in Fig. 1B, indicating the incorporation of individual ecSatP protein into the lipid bilayer membrane. The control experiments without ecSatP protein did not show any current increase. The conductance data provided a peak value of 1.2 Ϯ 0.34 nS (n ϭ 25) (Fig. 1C). In addition, the I-V curves of ecSatP in symmetric NaOAc or NaCl buffer both indicated that this protein functions as a bidirectional channel (Fig.  1D). When applying potential less than Ϫ100 mV or greater than ϩ100 mV, we observed three-step gating of this protein (Fig. S2).
To evaluate the ion selectivity of ecSatP, another ion conductance assay in an asymmetric 200/20 mM (cis-/trans-) acetate buffer was conducted. A reverse potential was applied to eliminate the current generated by acetate translocation as ecSatP was incorporated. When other kinds of anion were applied to the cis-solution, the current was recorded over time.
The histogram clearly showed that ecSatP was strongly selective for chloride, butyrate, and succinate in the presence of acetate (Fig. 1E). The order of the relative anion permeability of SatP was as follows: current change (I) chloride Ͼ I butyrate Ͼ I succinate Ͼ I lactate Ͼ I glycine Ͼ I formate Ͼ I pyruvate Ͼ I alanine (Fig.  1E). Although this in vitro substrate transport assay, performed with single-channel recording, did not replicate the exact physiological conditions of ecSatP, it provided evidence that ecSatP had a preferred translocation selectivity for carboxylates and chloride.

Overall structure of ecSatP
To understand the translocation mechanism of SatP, fulllength ecSatP was crystallized in the space group P3 2 . Through additive screening and continuous dehydration, the native diffraction data set was collected at ϳ3 Å. Large amounts of heavy atom-derived crystals were generated to solve the phase problem. However, none of them resulted in a data set having desirable anomalous signals. Finally, the structure was determined through molecular replacement using the predicted model of ecSatP from Baker's group (16) and refined to 2.8 Å resolution (Table 1). In each asymmetric unit, there are six molecules that form a compact cylinder with a diameter of ϳ95 Å and a height of ϳ45 Å ( Fig. 2A). The outer surface of the cylinder is highly hydrophobic, which is consistent with a TM domain in lipid bilayer, whereas the periplasmic and cytoplasmic sides of the cylinder have negative and positive electrostatic potential, respectively (Fig. 2B). There is a hole in the center of the cylin-der that is plugged by a discontinuous weak electron density (Fig. S3A). Based on the lipids or detergent molecules found in other oligomeric membrane proteins (15,17), the extra density indicates that lipids from the plasma membrane or detergents from the purification procedures occupied the central pore. The interface between two monomers is primarily constructed by TM1 and TM2 from monomer 1 and TM3 and TM4 from the nearby monomer 2 (Fig. 2C). The extensive hydrophobic interactions between the TM domains stabilize the oligomer (Fig. 2C). Surprisingly, the ConSurf evolutionary conservation analysis showed that none of the residues located laterally from TM1/2/3/4 shared reasonable similarity (Fig. S4).
Recently, another group resolved the structure of Citrobacter koseri SatP (ckSatP). Consistent with the high percentage of sequence conservation, the two SatP proteins exhibited a nearly identical conformation, with a root mean square deviation (RMSD) of 0.49 Å over 990 C␣ atoms (hexamer) and an RMSD of 0.22 Å over 161 C␣ atoms (monomer) (Fig. S5A). The comparison of the resolved structure and the predicted model of ecSatP also reveals limited differences, with an RMSD of 2.54 Å over 998 C␣ atoms (hexamer) and an RMSD of 1.79 Å over 159 C␣ atoms (monomer) (Fig. S5B). Additionally, there is a correlation between ecSatP and another hexameric channel, UreI. Although ecSatP shares no significant sequence similarity with the urea channel UreI, the superposition of ecSatP and UreI revealed a quite similar hexameric ring, with an RMSD of 2.98 Å over 866 C␣ atoms (Fig. S5C). Additionally, the protomer of ecSatP could be well superposed with the protomer of UreI with an RMSD of 2.61 Å over 149 C␣ atoms (Fig. S5C). The superposition of ecSatP and these two bacterial channels could strongly imply its channel-like structure. Nevertheless, the predicted cytoplasmic locations of the N and C termini of SatP indicated a reverse topology compared with UreI ( Fig. S6A). To confirm the topology of ecSatP, a PEGylation assay was carried out (Fig. S6B). The engineered protein containing a single cysteine in the C-terminal tail could only be PEGylated by mPEG-Mal-5K after sonication, which suggested the cytoplasmic localization of C terminus as well as the N terminus (Fig. S6B). In contrast, the N terminus of UreI was identified as having a periplasmic location in previous studies (17,18). Therefore, ecSatP shows the same fold structure but reverse topology compared with UreI, which reveals the evolutionary conservation of the UreI-like channel structure. This structure was also consistent with the results of the single-channel conductance assay. Thus, ecSatP reveals an oligomeric channel-like feature.

The protomer of ecSatP is a functional unit
The superposition of all ecSatP protomers showed that each protomer exhibits a nearly identical conformation (Fig. 3A). In each protomer, six TM helices form a tight bundle with a successive pattern (Fig. 3B). There are two residues (Leu-131/Ala-164) located in the channel of protomer that are involved in enhancing the transport ability of lactate (14) (Fig. 3B). This indicates that the protomer is a functional unit, which is similar to the urea channel UreI (17). The ConSurf evolutionary conservation also showed that the residues pointed into the channel of monomer share a greater sequence identity (Fig. S4). The channel-lining residues in the axial channel of each protomer

Structure of bacterial acetate channel SatP
exhibit an amphipathic feature, containing both polar and nonpolar amino acids that are highly conserved in the AceTr family ( Fig. 3C and Fig. S4). The highly conserved channel suggests a common translocational mechanism within members of the AceTr family and is consistent with the chemical properties of short-chain carboxylic substrates. Unexpectedly, the cytoplasmic side of the channel is occupied by electron density with a short tail, and the detergent molecule (NG) fit well. The hydrophobic tail of the NG molecule plugs into the cytoplasmic side of the central channel, and it undergoes extensive van der Waals interactions with the side chains of hydrophobic residues (Fig. S3B).

The constrictive site of ecSatP
We calculated the channel radius of a protomer using HOLE (19) (Fig. 4A). Obviously, there is only one constriction site in the central portion of the channel that narrows to ϳ0.8 Å. The overall structure of the protomer forms an hourglass-like shape, and the constrictive site is created by side chains from three hydrophobic residues (Phe-17, Tyr-72, and Leu-131) (Fig.  4B). Therefore, ecSatP is stabilized in the closed state under the crystallization condition. The open cavities on both the periplasmic and cytoplasmic sides of ecSatP distinguish it from the classic transporter and present a novel channel feature. Meanwhile, the constrictive residues "FLY" are almost identical to those of other transporters in the AceTr transporter family (Fig. S4), suggesting that members have highly conserved con-strictive gates. The mutation of L131V, or the corresponding residue in the yeast homolog Ady2, promoted a change of substrate specificity (14). Our structural analysis suggests that the three hydrophobic residues not only form the constrictive site but also contribute to the substrate selectivity.

Putative substrate-binding site of ecSatP
The specific recognition of substrate is very important for translocation events by transporters and channels. During translocation, the water-soluble substrates bind to the specific polar or charged residues in the cavity of the transporters/channels. Although we failed to resolve the structure of ecSatP in the complex with substrate, we predicted the substrate binding site of ecSatP based on resolved apo-structure. In the periplasmic cavity of the channel, there are three conserved polar residues (Thr-21, Asn-25, and Asn-28) in the TM1 that are presumably involved in substrate recognition (Fig. 5A). To verify those key residues, a single-channel transport assay was performed (Fig.  5, B and C). In addition, we constructed several mutants of ecSatP with T21A, N25A, and N28A. These mutants, excluding N28A, formed hexamers during protein expression and purification, and they could be incorporated into the lipid bilayer to form conducting channels as well. As observed in our experiment, ecSatP maintained a constantly open state in NaCl solutions, and the translocation of acetate blocked ecSatP transiently, which generated reversible blockade signals. The number of signals was positively correlated with the acetate transport capability. Blockade frequency indicates the frequency of reversible blockade signals observed upon the addition of acetate. Under the same conditions, the event frequency changes were ϩ0.51/s, ϩ0.35/s, and Ϫ0.04/s for WT ecSatP, T21A, and N25A, respectively (n ϭ 3 for each construct) (Fig. 5C). These results confirmed that the WT ecSatP had the greatest sodium acetate transport activity among all of the constructs under the experimental conditions. Furthermore, this observation suggested that these three residues play important roles in translocation and have a central role in substrate recognition. We also considered the involvement of the polar residues Gln-50 and Trp-76, both of which are close to the constrictive site, in translocational events. Considering the behavior of ecSatP with W76A, but not with Q50A/N/E, only W76A was used in the channel conductance recording experiment. As shown in Fig. 5 (B and C), W76A did not result in an obvious increase in the translocation frequency (Ϫ0.01/s).

Discussion
Our structural investigation provides a first glance of transporters in the AceTr family. The structure of ecSatP was resolved by molecular replacement using the predicted structure as the searching model. At present, few structures have been determined using this approach. As more structures are added to the Protein Data Bank, structural predictions will become more powerful and facilitate protein design as well.
Previous studies have indicated that the members belonging to this family are all transporters. However, the single-channel conductance assay of ecSatP showed that the rate of ionic translocation was ϳ10 7 ions/s with a conductance of 1.2 Ϯ 0.34 nS. The structure of ecSatP also presented the channel-like feature,

Structure of bacterial acetate channel SatP
which suggested that ecSatP is an acetate channel, not a transporter. Considering the highly conserved sequence and the similar fold structures of CLC channels and transporters (20 -22), we cannot simply infer that other members in the AceTr family are channels. Meanwhile, the interface between monomers of ecSatP was not highly conserved. Whether proteins belonging to the AceTr family have different oligomerization states needs further investigation. Even so, the details of acetate translocation remain perplexing. Additional structural investigations and molecular dynamic simulations should be carried out to explore the translocational mechanism. How does a protomer transition from the closed state to the open state? The sequence analysis indicated that the conserved residues (GGXXQ) in the TM2 are involved (Fig. S4). The double glycine motif may be a helical blocker that forms a kink (23). A discontinuous TM helix is a common transporter feature, and the local conformational change in this area may also play an important role in determining the transport activity (24). Therefore, we hypothesize that the conserved double glycine motif is involved in the translocational activity of this protein. Overall, this study increased our understanding of the translocation mechanism of acetate.
During the single-channel conductance assay, we observed three-step gating of ecSatP (Fig. S2). Six levels of conductance

Structure of bacterial acetate channel SatP
were recorded with the SatP from C. koseri (25). Considering ecSatP has a hexameric channel-like and the protomer serves as the functional unit, we propose that the six protomers do conduct simultaneously. Occasionally, we observed double or triple substrate translocational events in a single channel, which had a dwell time and blockade that resembled those of single translocational events. This suggested that the monomer ecSatP could function independently and that multiple translocational events could occur through one hexameric ecSatP. More biochemical and computational studies are needed to elucidate the details of the translocational mechanism.
Recently, the structure of the succinate-acetate transporter from ckSatP, which was identified as a unidirectional acetate channel, was reported by Liao's group (25). Although ecSatP shares a high sequence identity with ckSatP, the former presents bidirectional acetate permeation features under our experimental condition, and this is different from the ckSatP. However, both investigations confirmed the chloride translocation by SatP. Whether SatP plays an important role in chloride homeostasis in vivo requires more investigation.

Cloning, expression, and purification of ecSatP
The complementary DNA of the WT SatP was cloned from the genome of the K-12 strain of E. coli into a modified pET15b vector (Novagen), with an additional DrICE-cutting site (DEVDA) inserted after the original thrombin-cutting site. All mutations and truncations were generated using a PCR-based strategy.
The transformed E. coli C43 (DE3) cells were cultured in LB medium (BD Biosciences) at 37°C to reach an A 600 of 1.0 for further induction. Cells were induced with 0.2 mM isopropyl ␤-D-thiogalactoside (Amresco) at 22°C overnight and harvested with lysis buffer containing 25 mM Tris, pH 8.0, and 150 mM NaCl.
For protein purification, cells were lysed with sonication, followed by centrifugation (20,000 ϫ g) for 10 min at 4°C to remove the debris. The supernatant was collected and supplemented with 1% (w/v) n-dodecyl-␤-D-maltoside (DDM; Anatrace) and 1 mM phenylmethylsulfonyl fluoride to solubilize the membrane for 1.5 h at 4°C. The insoluble portion was removed by an additional ultracentrifugation step (150,000 ϫ g) for 30 min at 4°C. The remaining soluble portion was loaded on nickel-nitrilotriacetic acid resin (Qiagen) and rinsed with wash buffer containing 25 mM HEPES, pH 7.0, 150 mM NaCl, 30 mM imidazole, and 0.05% (w/v) DDM. The protein was eluted with wash buffer plus 270 mM imidazole and concentrated to 20 mg/ml. The further size-exclusion chromatography (Superdex 200 Increase 10/300, GE Healthcare) was carried out after preequilibration with 25 mM HEPES, pH 7.0, 150 mM NaCl, and 0.05% (w/v) DDM. For crystallization, the N-terminal His 6 tag of the target protein was removed using DrICE, and many kinds of buffers and detergents were tested during the purification procedure. For the ion conductance assay in sodium acetate, the buffer was changed to 25 mM HEPES, pH 7.0, 150 mM NaAc, 0.05% (w/v) DDM during gel filtration.

Crystallization and data collection
We performed extensive crystallization trials on WT ecSatP proteins that were purified with different buffers and detergents. Crystals were grown at 18°C using the hanging-drop vapor-diffusion method. Finally, the protein in 25 mM Tris (pH 8.0), 150 mM NaCl, and 0.4% (w/v) ␤-NG formed the hexagonshaped crystal that appeared within 2 days in a buffer containing 0.05 M glycine (pH 9.8), 27% (v/v) PEG300, and 0.1 M NaCl. It required 4 days for the crystal to reach a full size of 150 ϫ 150 ϫ 40 m. However, most of the crystals behaved poorly beyond 4 Å at Shanghai Synchrotron Radiation Facility beamline BL19U1, with only a few reaching the 3.5-4 Å resolution. Further optimization was carried out using detergent and additive screening kits (Hampton Research). The dropwise addition of 0.15 mM n-nonyl-␤-D-thioglucoside worked well and produced higher-quality diffracted crystals to a 3.0 -3.2 Å resolution. Finally, dehydration trials were carried out. Fully grown crystals were treated by adding PEG300 into the well buffer to a final concentration of 40% (v/v) over 24 h. Then the crystals were harvested and flash-frozen in liquid nitrogen.
The final data sets were collected at Shanghai Synchrotron Radiation Facility (SSRF) beamline BL19U1 and processed with the HKL3000 package (26). Further processing was carried out using programs from the CCP4 suite (27). Each data set was from a single crystal.

Structure determination and refinement
Briefly, the predicted model from Baker's group (17) was applied to PHASER (28) for molecular replacement as the search model. The model was first modified by CHAINSAW to generate polyalanine peptides and then rebuilt in the program COOT (29) and refined with PHENIX (30). The sequence dock-

Structure of bacterial acetate channel SatP
ing was undoubtedly aided by the SatP sequence itself. Data collection and structural refinement statistics are summarized in Table 1. The atomic coordinates have been deposited in the Protein Data Bank (accession code 5ZUG).

PEGylation assay
To identify the topology of SatP, we introduced an S187C mutation to the Cys-free construct and used mPEG-Mal-5k (Sigma) to label the protein. Briefly, 150 ml of transformed E. coli was cultured as described previously. Cells were divided into three equal portions and centrifuged at 2500 ϫ g, followed by resuspension using 1 ml of the reaction buffer containing 20 mM HEPES, pH 7.0, 150 mM NaCl, and 10% (v/v) glycerol. One fraction was sonicated as a positive control. The reaction was initiated by adding 10 mM mPEG-Mal-5k to the mixture. The negative control was supplemented with an equal amount of

Structure of bacterial acetate channel SatP
buffer. After incubating for 1 h at room temperature, 20 mM ␤-mercaptoethanol was added to stop the reaction. Unbroken cells were also sonicated, and all treatments were purified following the same protocol as above. Purified protein samples were applied to Western blotting using anti-His antibody for analysis.

Cross-linking assay
In total, 50 l of WT SatP at a concentration of 5 mg/ml was supplemented with 0 -2 mM EGS-sulfo (Thermo) diluted from a 10 mM stock solution and incubated for 1 h at 18°C. The reactions were terminated using 70 mM Tris, pH 8.8, for 30 min. Samples were analyzed using SDS-PAGE.

Incorporation of SatP into planar bilayer lipid membrane and electrophysiological measurement
A bilayer of 1,2-diphytanoylphosphatidylcholine was formed on an aperture (150 m) in a Teflon septum (25-m thick) that divided the chamber into cis-and trans-compartments (BCH-13A, from Warner, catalog no. 6406451), The methodology used to form the bilayer was as described previously (31). The experiments were performed under a series of symmetrical or asymmetrical buffer conditions with a 2.0-ml solution composed of different concentrations of NaCl or NaOAc at 22°C. Then 0.8 l of SatP protein was added to the cis-compartment containing 1 ml of the buffer. The substrate binding and conductance measurement experiments were conducted in 500 mM NaCl, 25 mM MES (pH 5.5) due to the high transport activity under acidic conditions (14). The directionality experiment of SatP was performed in 200 mM NaCl or 200 mM NaOAc buffered with 10 mM HEPES (pH 7.0). The cis-compartment was connected to ground, and the potential was applied to the trans-compartment. After the insertion of a single SatP protein channel or more, the increasing current would analogously show discrete, increasing steps. The final concentration of the SatP protein was 2-10 ng/ml. Currents were recorded with a patch clamp amplifier (10USB, HEKA, Lambrecht, Germany). They were low-pass-filtered with a Bessel filter at 5 kHz and sampled at 100 kHz by a computer equipped with Patchmaster version 2.65.
A plot of the current-voltage curve was obtained at different voltages for the conductance measurements. As a comparison, the conductance measurements were also performed under a ramp voltage, with the slope representing its single-pore conductance at 1.2 nS.

Ion conductance assay under asymmetric buffer conditions
For the ion conductance assay, sodium formate, sodium acetate, sodium succinate, glycine, sodium pyruvate, sodium lactate, sodium butyrate, alanine, and sodium chloride were tested. The substrates used in this trial were purchased from Sigma-Aldrich. For the transport assay buffer, 200 and 20 mM sodium acetate buffered with 10 mM HEPES (pH 7.0) were used. After the incorporation of ecSatP into the bilayer lipid at asymmetric 200/20 mM sodium acetate (cis-/trans-) solution, a reverse potential was applied to eliminate the acetate current, and then 20 l of different 1 M substrate stocks were individually added to the cis-compartment. The current changes (I x ) generated by the addition of different substrates were analyzed. If the addition of one type of ion induces a current change, then it demonstrates that SatP is permeable to this type of ion. The current change generated by the addition of different substrates is positively correlated with their transport capability.