The structure of MgtE in the absence of magnesium provides new insights into channel gating

MgtE is a Mg2+ channel conserved in organisms ranging from prokaryotes to eukaryotes, including humans, and plays an important role in Mg2+ homeostasis. The previously determined MgtE structures in the Mg2+-bound, closed-state, and structure-based functional analyses of MgtE revealed that the binding of Mg2+ ions to the MgtE cytoplasmic domain induces channel inactivation to maintain Mg2+ homeostasis. There are no structures of the transmembrane (TM) domain for MgtE in Mg2+-free conditions, and the pore-opening mechanism has thus remained unclear. Here, we determined the cryo-electron microscopy (cryo-EM) structure of the MgtE-Fab complex in the absence of Mg2+ ions. The Mg2+-free MgtE TM domain structure and its comparison with the Mg2+-bound, closed-state structure, together with functional analyses, showed the Mg2+-dependent pore opening of MgtE on the cytoplasmic side and revealed the kink motions of the TM2 and TM5 helices at the glycine residues, which are important for channel activity. Overall, our work provides structure-based mechanistic insights into the channel gating of MgtE.


Introduction
Mg 2+ ions are biologically essential elements involved in various physiological processes, including the catalytic action of numerous enzymes, the utilization and synthesis of ATP, and  attempted to generate Fab antibodies for MgtE to enable determination of the cryo-EM structure of the MgtE-Fab complex by increasing its total molecular weight. After mouse immunization and generation of multiple monoclonal antibodies for MgtE, fluorescence detection size exclusion chromatography (FSEC) analysis showed that one of the Fab fragments, Fab705, bound to full-length MgtE with 0 mM Mg 2+ , as indicated by the apparent peak shift (S1A Fig), but the binding did not occur at a high concentration of Mg 2+ (20 mM) (S1B Fig). Consistent with this, isothermal titration calorimetry (ITC) analysis showed that Fab705 bound to full-length MgtE with 0 mM MgCl 2 , with a K d of 216.1 ± 52.7 nM, whereas no interaction was detected between Fab705 and MgtE at a high concentration of Mg 2 + (20 mM) (Fig 1D and 1E, S1 Table). We then performed patch clamp analysis of full-length MgtE using Escherichia coli giant spheroplasts at 0 mM MgCl 2 in the presence and absence of Fab705 and observed little change in the open probabilities upon addition of Fab705 into the bath solution (Fig 1F and 1H). Furthermore, in the patch clamp recording, when we tested a high concentration of MgCl 2 (20 mM MgCl 2 ) in the bath solution after the addition of Fab705, we did not observe channel opening, as similarly observed with a high concentration of MgCl 2 in the absence of Fab705 (Fig 1G). These results demonstrate that Fab705 binds to Mg 2+ -free MgtE but not to MgtE with high Mg 2+ concentrations and neither positively nor negatively modulate channel opening at 0 mM Mg 2+ . These functional properties of Fab705 are desirable for the determination of the MgtE structure under Mg 2+ -free conditions because it does not affect channel gating under Mg 2+ -free conditions, particularly in an inhibitory manner. Thus, the utilization of Fab705 enables the gating motions of MgtE to be captured by cryo-EM.

Cryo-EM of the MgtE-Fab complex in Mg 2+ -free conditions
To determine the structure of MgtE in the absence of Mg 2+ or other divalent cations, the fulllength MgtE-Fab705 complex was reconstituted in PMAL-C8 amphipol and subjected to cryo-EM single-particle analysis (Fig 2, S2 and S3 Figs). A 3D reconstruction of the cryo-EM images was performed to reach an overall resolution of 3.7 Å, with C2 symmetry imposed ( Fig  2C), and the cryo-EM density of the TM domain was clear enough to build an atomic model (Fig 2D-2F, S4 Fig). An additional cryo-EM density map of the TM domain was calculated at an approximately 3.1 Å resolution by signal subtraction and additional 3D refinement (S5 Fig).
Notably, we did not detect the cryo-EM density of the cytoplasmic domain of MgtE. SDS-PAGE analysis showed the presence of the cytoplasmic domain of MgtE in the purified sample for cryo-EM (S2 Fig), and thus, the absence of the MgtE cytoplasmic domain in the cryo-EM density was probably because of the high structural flexibility of the Mg 2+ -free cytoplasmic domain [30,31]. The highly flexible nature of the Mg 2+ -free MgtE cytoplasmic domain in our cryo-EM structure is consistent with previous studies conducted on MgtE by highspeed atomic force microscopy (HS-AFM) [32] and molecular dynamics (MD) simulations [31]. In particular, the HS-AFM results directly visualized the continuous shaking motions of MgtE cytoplasmic domains superimposed on the CBS domains and viewed from the cytoplasmic side (B) and parallel to the membrane (C). The black arrows and red bars denote the rotation of the N domain (B) and plug helices (C). The N and CBS domains, plug helix of the Mg 2+ -free state, are colored blue, green, and yellow, respectively, whereas the Mg 2+ -bound state structure is colored gray. (D, E) Representative ITC profiles of MgtE with Fab705 in the presence of 0 mM MgCl 2 (D) and 20 mM MgCl 2 (E). (F, G) Single-channel recording of MgtE in the inside-out configuration from E. coli giant spheroplasts at −60 mV. Numbers within and outside trapezoids indicate MgCl 2 concentrations in pipettes and bath solutions, respectively. Representative current traces from triple-knockout E. coli spheroplasts expressing MgtE in the presence (lower) or absence (upper) of 2 μM Fab705. (H) The open probabilities for MgtE with (+) and without (−) 2 μM Fab705 (n = 6 for each condition, p = 0.804). Error bars represent the standard error of the mean. The individual numerical values that underlie the summary data displayed in this figure can be found in S1 Data. CAU : Abbreviationlistshaveb BS, cystathionine-beta-synthase; ITC, isothermal titration calorimetry; TM, transmembrane. https://doi.org/10.1371/journal.pbio.3001231.g001 the MgtE cytoplasmic domains under Mg 2+ -free conditions. The flexibility of the MgtE cytoplasmic domain in Mg 2+ -free conditions is necessary for unlocking the closed conformation of the MgtE TM domain for channel opening [14].
Nevertheless, the details of the structural changes in the TM domain following the release of Mg 2+ ions from the cytoplasmic domain were previously unclear, and for the first time, we have successfully visualized the conformation of the TM domain under Mg 2+ -free conditions. MgtE forms a symmetric dimer, and each subunit is composed of 5 TM helices (Fig 3). The Fab705 molecules are bound to the cytoplasmic side of the MgtE TM domain, and this binding is mediated mainly by electrostatic interactions and hydrogen bonds, primarily involving the residues in the TM2 to TM3 linker and the TM4 to TM5 linker (Fig 3A, S6A Fig). Among these residues involved in Fab binding, Arg345 interacts with Glu217 in the CBS domain, and Asp418 binds to Mg 2+ ions to bridge the TM domain and plug helices in the full-length MgtE structure in the Mg 2+   Notably, in FSEC analysis, the addition of 20 mM Mg 2+ into the sample of the MgtE-Fab complex sample, which was preformed before the addition of Mg 2+ , disrupted complex formation, where we did not see a Fab-dependent FSEC peak shift (S1C Fig), consistent with the patch clamp recording at a high concentration of MgCl 2 (20 mM MgCl 2 ) in the bath solution (Fig 1G). These results indicate that after Fab binding, a high concentration of Mg 2+ ions can still close the MgtE channel by disrupting the MgtE-Fab complex.
We then compared our structure with the previously determined Mg 2+ -bound, closed-state MgtE structure (PDB ID: 2ZY9) and performed further functional analysis to gain structural insight into MgtE channel opening.

Ion-conducting pore
The MgtE structure possesses an ion-conducting pore at the center of the dimer in the Mg 2+bound, closed state (Fig 4). The MgtE structure in Mg 2+ -free conditions has a wider pore on the cytoplasmic side than that in the previously determined structure in the presence of Mg 2+ ions (PDB ID: 2ZY9) (Fig 4C). Notably, at the cytoplasmic end of the pore, the side chains of Asn424 of both subunits face each other in the Mg 2+ -bound closed structure (PDB ID: 2ZY9) (Fig 4B), whereas in our new Mg 2+ -free structure, the side chains of Asn424 are turned away from each other (Fig 4A) to form a wider cytoplasmic pore, suggesting that this region might form the cytoplasmic gate. In addition, Asn332 is in close contact with Asn424, seemingly contributing to the formation of the cytoplasmic gate (Fig 4B).

PLOS BIOLOGY
To test this hypothesis, we generated MgtE mutants possessing an alanine substitution at Asn332 or Asn424, which may disrupt the cytoplasmic gate (N332A and N424A). We conducted a Ni 2+ sensitivity assay in E. coli using the N332A-and N424A-MgtE mutants. The coordination chemistry required for Ni 2+ is highly similar to that of Mg 2+ , and MgtE and other Mg 2+ channels typically transport Ni 2+ [7]. We employed Ni 2+ , not Mg 2+ , for the

PLOS BIOLOGY
sensitivity assay in E. coli because Ni 2+ is much more toxic to E. coli cells than Mg 2+ : Therefore, Ni 2+ is more suitable to evaluate the change in the sensitivities in E. coli. Increased sensitivity to Ni 2+ indicates up-regulation of Ni 2+ uptake, and the change in the sensitivities to Ni 2+ was already previously employed to characterize MgtE and its mutants in the past [17]. E. coli cells expressing the N332A and N424A mutants of MgtE exhibited increased Ni 2+ sensitivity ( Fig  4D) compared to cells expressing wild-type MgtE, suggesting that the disruption of the cytoplasmic gate led to increased Ni 2+ uptake. In contrast to the opening of the cytoplasmic side of the pore, the pore remains closed around the Ala317 residues on the periplasmic side, where the pore radius is approximately 1 Å (Fig 4C) in both MgtE structures in the presence and absence of Mg 2+ ions, indicating that the current structure would remain in a nonconductive conformation on the periplasmic side.

Cytoplasmic pore-opening motions
Consistent with the wider opening of the ion-conducting pore on the cytoplasmic side, the TM helices, including the pore-forming TM2 and TM5 helices, are located in exterior positions with respect to the center of the ion-conducting pore under Mg 2+ -free conditions (Fig 5A and 5B, S1 Video). Furthermore, the TM3 and TM4 helices have also moved away from the center, as they interact with both the TM2 and TM5 helices (Fig 5B). For instance, compared to the structure of the Mg 2+ -bound state, Thr344 at TM2, in the Mg 2+ -free state, Asp348 at TM3, Leu415 at TM4, and Ala420 at TM5 are rotated by approximately 8˚, 5˚, 8˚, and approximately 14˚, respectively, clockwise from the central axis along the pore and have moved away from the central axis by approximately 3 Å, approximately 3 Å, 3 Å, and approximately 1 Å, respectively.
To verify the Mg 2+ -dependent structural changes in the TM helices on the cytoplasmic side, we generated MgtE mutants possessing cysteine substitutions at Leu421 and Thr336, where the Cα distances between Leu421 (A) and Thr336 (B) in the presence and absence of Mg 2+ ions were 5.2 and 14.0 Å, respectively (Fig 5C). Using Cu 2+ phenanthroline as a catalyst, crosslinking experiments were performed with these mutants (Fig 5D and 5E). Wild-type MgtE does not contain any endogenous cysteine residues, and the addition of MgCl 2 to MgtE would shift the equilibrium from the Mg 2+ -free form to the Mg 2+ -bound form, according to the previously reported limited proteolysis assay of MgtE [14]. The double cysteine mutant T336C/ L421C exhibited a strong band corresponding to the MgtE dimer in the presence of Mg 2+ at a concentration gradient (0.01 to 10 mM) and Cu 2+ phenanthroline (Fig 5D and 5E), consistent with the Mg 2+ -dependent structural changes observed in our cryo-EM structure.
To further support the current conformation of the MgtE TM domain structure, we performed MD simulations starting from the Mg 2+  The results further support the cytoplasmic pore opening of MgtE under Mg 2+ -free conditions, as we observed in the cryo-EM structure. At the same time, we also do not exclude the possibility that we might see some structural changes if we run the MD simulation for a much longer duration.

Kink motions of the TM helices for channel gating
Detailed observation and comparison of the MgtE structures further revealed the kink motions of TM2 at Gly325 and Gly328 by approximately 10˚and of TM5 at Gly435 by approximately

PLOS BIOLOGY
10˚ (Fig 6A and 6B). Since the TM2 and TM5 helices form the ion-conducting pore, these kink motions seemingly enable the TM2 and TM5 helices to expand the pore on the cytoplasmic side.
To examine the functional importance of these glycine residues, we generated a series of alanine-substituted mutants of MgtE (G325A, G328A, and G435A) and performed a growth complementation assay using a Mg 2+ -auxotrophic E. coli strain lacking the major genes (ΔmgtA ΔcorA ΔyhiD) encoding Mg 2+ transporters that survived only when the medium was supplemented with a sufficiently high concentration of Mg 2+ [14]. The exogenously expressed MgtE could reconstitute the Mg 2+ transport activity of this strain, and the strain transformed with the wild-type MgtE gene could survive in Mg 2+ -free LB medium [14]. While the expression of wild-type MgtE rescued Mg 2+ -auxotrophic growth, the expression of the mutations at these glycine residues almost lost Mg 2+ -auxotrophic growth complementation activity ( Fig  6C). In particular, mutation of the strictly conserved Gly328 completely eliminated the growth complementation activity (Fig 6C). These results indicated that these glycine residues at the kinks of the TM2 and TM5 helices are important for the channel activity of MgtE.

Discussion
In this work, we determined the cryo-EM structure of MgtE in Mg 2+ -free conditions. The utilization of antibodies has been successfully applied to the single-particle analysis of membrane proteins [33][34][35][36][37]. Notably, in our MgtE-Fab complex structure, the ordered region of MgtE (MgtE TM domain dimer) has a mAU : PleasenotethatasperPLOSstyle; molecularweightisunitless; dono olecular weight of only 39 kDa, further demonstrating antibody-based techniques to be powerful tools for the determination of membrane protein structures by cryo-EM. Notably, Fab705 employed for cryo-EM does not have inhibitory effects on the channel opening under Mg 2+ -free conditions (Fig 1F), which would be desirable for determining the MgtE structure under Mg 2+ -free conditions (S9 Fig) because Fabs with inhibitory effects on channel gating may hinder structural changes in the TM domain, which would not

PLOS BIOLOGY
be suitable to gain structural insights into the channel gating. Based on this structure, we further conducted structure-based functional analysis, and our new findings together with our previous research on MgtE provided structural insights into the gating mechanism of MgtE, as discussed below.
First, under high Mg 2+ concentrations, the binding of intracellular Mg 2+ to the cytoplasmic domain of MgtE, including the plug helices, stabilizes the tightly packed, closed conformation of the protein, leading to pore closure in the TM domain through extensive interactions between the plug helices and the TM domain, mainly via helices TM2 and TM5 on the cytoplasmic side [30] (Figs 1A and 7A).
In contrast, under Mg 2+ -free conditions, the MgtE cytoplasmic domain, including the plug helices, was totally disordered and could not be built into the cryo-EM map, indicating the highly flexible nature of the MgtE cytoplasmic domain under Mg 2+ -free conditions (Fig 7B). In such disordered conditions, the plug helices of MgtE cannot tightly interact with the TM domain to lock the channel in the closed state anymore, unlike the structure in the Mg 2 + -bound state (Figs 1A and 7A). In other words, under Mg 2+ -free conditions, the high flexibility of the cytoplasmic domain leads to the loss of the interactions between the plug helices and the TM domains, unlocking the TM domain conformation from the closed state. Highly flexible domain motions in the MgtE cytoplasmic domain were also observed by HS-AFM [32], in the crystal structure of the Mg 2+ -free cytoplasmic domain structure [30], and via MD simulations of MgtE [31]. In particular, in a previous report on HS-AFM [32], Haruyama and colleagues observed a jagged topography of MgtE in the absence of Mg 2+ under HS-AFM from

PLOS BIOLOGY
the end-up orientation (cytoplasmic view), indicating that the cytoplasmic domains are constantly fluctuating. Furthermore, they observed that in the side-on orientation (side view), the cytoplasmic regions of MgtE were highly flexible and ambiguous in the absence of Mg 2+ , and successive images demonstrated the association and disassembly of the N and CBS domains in solution. Such disassembly of the N and CBS domains is consistent with the previous structure of the MgtE cytoplasmic domain under Mg 2+ -free conditions (Figs 1B and 7B). Intriguingly, similar types of regulatory mechanisms, in which the cytoplasmic domain undergoes a transition from a rigid conformation to a highly flexible or disordered state upon the release of ligands from the cytoplasmic domain, have been observed in other channels/transporters such as the CLC Cl − channel and the OpuA ABC transporter, which also possesses a CBS domain [38,39].
Following the unlocking of the TM domain due to the high flexibility of the Mg 2+ -free cytoplasmic domain, we observed the structural differences of the TM domains between our cryo-EM structure and the previously determined Mg 2+ -bound, inactivated structure (Fig 5). In our cryo-EM structure, the ion-conducting pore is wide open on the cytoplasmic side but still closed on the periplasmic side (Fig 4), indicating that the pore is still nonconductive. According to previous patch clamp recording results, MgtE is in equilibrium between the open and closed states under low-Mg 2+ conditions rather than in a constitutively open state [14]. Therefore, our cryo-EM structure of MgtE may represent a preopening state of MgtE under low-Mg 2+ conditions, where the pore is partially opened on the cytoplasmic side due to loss of the tight interactions with the plug helices (Fig 7B). Intriguingly, after the 3-μs MD simulation of the Mg 2+ -free, MgtE TM domain structure was performed by adding Mg 2+ ion to the Mg 2+ binding site in the pore; the entrance region of the periplasmic side between the TM1 and TM2 helices became partially opened, with Cα distances between Glu311 (A) and Glu311 (B) of approximately 12 Å (S8E-S8G Fig), whereas the periplasmic gate remained nonconductive, with Cα distances between Ala317 (A) and Ala317 (B) of approximately 6 Å (S8D Fig). To fully understand the gating cycle of the MgtE channel, a structure in a fully opening state is still required. Furthermore, whereas the previous results are most consistent with the bacterial MgtE acting as an ion channel [14,16,17], since some of the SLC41 proteins, the eukaryotic counterparts of the bacterial MgtE, are Na + -driven Mg 2+ transporters [24], the understanding of their transport mechanism would require structural analysis of eukaryotic SLC41 proteins.
In our structure, the structural changes in the TM domain occur mainly at the TM2 and TM5 helices (Figs 5 and 7B), and the glycine residues in the TM2 and TM5 helices enable the kink motions of the TM2 and TM5 helices for cytoplasmic pore opening (Figs 6 and 7B). We verified these structural changes by biochemical cross-linking experiments and genetic assays (Figs 4D, 5D, 5E and 6C). Notably, none of these functional experiments to verify the structural changes were performed in the presence of Fab705, suggesting that these structural changes in our MgtE-Fab complex structure are unlikely to be nonphysiological artifacts due to the use of Fab for structure determination.
Taken together, our structural and functional analyses provide mechanistic insight into the channel-opening mechanisms of MgtE, particularly cytoplasmic pore opening.

Preparation of the MgtE-Fab complex
The full-length T. thermophilus MgtE gene was subcloned into a pET28a-derived vector containing an N-terminal hexahistidine tag and a human rhinovirus 3C (HRV3C) protease cleavage site, as described previously [30]. The protein was overexpressed in E. coli C41 (DE3) cells in LB medium containing 30 μg/ml kanamycin at 37˚C by induction at an OD 600 of approximately 0.5 with 0.5 mM isopropyl-β-D-thiogalactoside (IPTG) for 16 hours at 18˚C. The cells were subsequently harvested by centrifugation (6,000 × g, 15 minutes) and resuspended in buffer H [50 mM HEPES (pH 7.0), 150 mM NaCl, 0.5 mM phenylmethanesulfonyl fluoride (PMSF)]. All purification steps were performed at 4˚C. The cells were then disrupted with a microfluidizer, and the supernatants were collected after centrifugation (20,000 × g, 30 minutes) and ultracentrifugation (200,000 × g, 1 hour). Next, the membrane fraction was solubilized with buffer S [50 mM HEPES (pH 7.0), 150 mM NaCl, 2% (w/v) n-dodecyl-beta-d-maltopyranoside (DDM) (Anatrace, United States of America) and 0.5 mM PMSF] for 2 hours. The solubilized fraction was loaded onto a Ni-NTA column pre-equilibrated with buffer A [50 mM HEPES (pH 7.0), 150 mM NaCl, and 0.05% (w/v) DDM] containing 20 mM imidazole and mixed for 1 hour. The column was then washed with buffer A containing 50 mM imidazole, and the protein was eluted with buffer A containing 300 mM imidazole. The N-terminal hexahistidine tag was cleaved by HRV3C protease against dialysis buffer (buffer A containing 20 mM imidazole) overnight, and the cleaved protein was reloaded onto a Ni-NTA column pre-equilibrated with dialysis buffer.
The flow-through fraction from the column was collected and concentrated using an Amicon Ultra 50K filter (Merck Millipore, USA) and then loaded onto a Superdex 200 10/300 size exclusion chromatography (SEC) column (GE Healthcare, USA) in buffer B [20 mM HEPES (pH 7.0), and 150 mM NaCl] containing 0.05% (w/v) DDM. The main peak fractions from the column were collected and mixed with PMAL-C8 amphipol (Anatrace) at a ratio of 1:5 (w:w) overnight. The sample was mixed with Bio-Beads SM-2 (Bio-Rad, USA) and rotated gently for 4 hours. The Bio-Beads were subsequently removed by a disposable Poly-Prep column, and the eluents were concentrated using an Amicon Ultra 50K filter and loaded onto a Superdex 200 10/300 SEC column in buffer B. The SEC fractions corresponding to the amphipol-reconstituted MgtE were collected, mixed with the Fab antibody fragment Fab705 at a ratio of 1:2 (w:w) for 1 hour, and applied to a Superdex 200 10/300 SEC column in buffer B. The SEC fractions corresponding to the MgtE-Fab complex were collected and concentrated to approximately 1 mg/ml with an Amicon Ultra 50K filter for single-particle analysis using cryo-EM.

EM data acquisition and analysis
A total of 4.0 μl of the MgtE-Fab complex in PMAL-C8 was applied to a glow-discharged holey carbon-film grid (Zhongkexinghua BioTech, GiG-A31213, 300M-Au-R1.2/1.3, China) blotted with a Vitrobot (FEI, USA) system using a 1.0-second blotting time with 100% humidity at 9˚C and plunge-frozen in liquid ethane. Cryo-EM images were collected on a Titan Krios (FEI) electron microscope operated at an acceleration voltage of 300 kV. The specimen stage temperature was maintained at 80 K. Images were recorded with a K2 Summit direct electron detector camera (GatanAU : PleasenotethatasperPLOSstyle; donotuseInc:; Ltd:; etc:exceptasappropriate , USA) set to super-resolution mode with a pixel size of 0.41 Å (a physical pixel size of 0.82 Å) and a defocus ranging from −1.5 μm to −2.3 μm. The dose rate was 8 es -1 , and each movie was 6-second long, dose-fractioned into 40 frames with an exposure of 1.75 e -Å -2 for each frame. The cryo-EM data are summarized in S2 Table. Image processing A total of 5,610 movies of the MgtE-Fab complex were motion-corrected and binned with MotionCor2 [40] with 5 × 5 patches, producing summed and dose-weighted micrographs with a pixel size of 0.82 Å. Contrast transfer function (CTF) parameters were estimated by CTFFIND 4.1 [41]. Particle picking and further image processing were performed using RELION 3.0 [42]. A total of 1,860,461 particles were autopicked and extracted with a box size of 256 × 256 pixels. After several rounds of 2D classification, the re-extracted particles were classified into 12 classes for 3D classification without any symmetry imposed. Refinement and postprocessing using RELION produced a final map at a 3.7 Å resolution from 168,911 particles with C2 symmetry imposed. The final 3D reconstruction was calculated from 2 × 2 binned images (0.82 Å). The final resolution was estimated using the Fourier shell correlation (FSC) = 0.143 criterion on the corrected FSC curves, in which the influence of the mask was removed. The local resolution was estimated using RELION. The workflow for image processing and for the 3D reconstruction and angular distribution plot are shown in S3 Fig.

Model building
The initial model of Mg 2+ -free MgtE-Fab was manually built starting from the MgtE TM domain structure and the homology model of Fab705 generated by SWISS-MODEL [43]. Manual model building was performed using Coot [44]. Real-space refinement was performed using PHENIX software [45]. The final atomic model of MgtE includes residues 271 to 448 for chains A and B. All figures showing structures were generated using PyMol (https://pymol. org/). The sequence alignment figure was generated using Clustal Omega [46] and ESPript 3.0 [47]. The MgtE pore shown in

Preparation of antibody fragments
Mouse monoclonal antibodies against MgtE were raised by an established method [49]. Briefly, a proteoliposome antigen was prepared by reconstituting purified, functional MgtE at high density in phospholipid vesicles consisting of a 10:1 mixture of egg phosphatidylcholine (Avanti Polar Lipids, USA) and the adjuvant lipid A (Sigma-Aldrich, USA) to facilitate the immune response. BALB/c mice immunization was performed with the proteoliposome antigen using 3 injections at 2-week intervals. Antibody-producing hybridoma cell lines were generated using a conventional fusion protocol [50]. Hybridoma clones producing antibodies recognizing conformational epitopes in MgtE were selected by an enzyme-linked immunosorbent assay on immobilized phospholipid vesicles containing purified MgtE (liposome ELISA), allowing positive selection of the antibodies that recognized the native conformation of MgtE. Additional screening for reduced antibody binding to SDS-denatured MgtE was used for negative selection against linear epitope-recognizing antibodies. Stable complex formation between MgtE and each antibody clone was checked using FSEC. From these screens, we isolated 4 MgtE-specific monoclonal antibodies: YN0705, YN0710, YN0721, and YN0736. Whole IgG molecules, collected from the large-scale culture supernatant of monoclonal hybridomas and purified using protein G affinity chromatography, were digested with papain, and Fab fragments were isolated using a HiLoad 16/600 Superdex 200 gel filtration column followed by protein A affinity chromatography. The cDNAs encoding the light and heavy chains of Fab YN0705 were cloned from hybridoma cells using rapid amplification of 5 0 complementary DNA ends (5 0 -RACE) and sequenced.

Isothermal titration calorimetry
The binding affinities of the Fab antibody fragment Fab705 for full-length MgtE were measured using MicroCal iTC200 (GE Healthcare). The full-length MgtE proteins were purified by a similar procedure to that described above but with SEC buffer [buffer B containing 0.05% (w/v) DDM] containing either 0 or 20 mM MgCl 2 . Fab was also used for gel filtration with the same SEC buffer containing either 0 or 20 mM MgCl 2 . The peak fractions of MgtE were collected and diluted to approximately 2.5 μM (monomer) with SEC buffer, while Fab was diluted to 25 μM. Fab was injected into the cell containing the full-length MgtE 20 times (0.5 μl for injection 1, 2 μl for injections 2 to 20) at 25˚C, with 120-second intervals between injections.
The background data were measured by injecting the SEC buffer into a cell containing the fulllength MgtE. The results were analyzed by Origin8 software (MicroCal). The experiments were repeated 3 times for each measurement, and similar results were achieved. All ITC profiles are shown in S10 Fig. The protein concentration was estimated using the absorbance at 280 nm.

Chemical cross-linking
The wild-type and mutant MgtE proteins were purified as described above, except that dithiothreitol (DTT) was added at a final concentration of 20 mM to the MgtE mutant proteins before removal by SEC, and then diluted to 0.5 mg/ml with buffer B containing 0.05% (w/v) DDM. MgtE proteins (4.0 μl) were mixed with 0.5 μl of MgCl 2 at appropriate concentrations and incubated for 30 minutes at 4˚C. Then, 0.5 μl of 10 mM Cu 2+ bis-1,10-phenanthroline was added at a 1:3 molar ratio, and the samples were incubated for another 30 minutes at 4˚C. The reaction mixtures were loaded with nonreduced loading buffer and analyzed immediately by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). The relative intensity of the bands was analyzed by ImageJ software and fitted to a sigmoidal equation. The experiments were repeated 6 times. All SDS-PAGE results are shown in S11 Fig.

Molecular dynamics simulations
As described previously [51,52], DESMOND [53] was used to perform MD analysis with a constant number of particles, constant pressure (1 bar) and temperature (300 K) (NPT), and periodic boundary conditions (PBCs) by using the Nose-Hoover chain thermostat and Martyna-Tobias-Klein barostat methods. The MgtE TM domain dimer was embedded in a phosphoryl-oleoyl phosphatidylcholine (POPC) bilayer. The structure of MgtE/POPC was solvated in an orthorhombic box of SPC water molecules. All-atom OPLS_2005 force fields for proteins, ions, lipids, and simple point charge (SPC) waters were used in all simulations [54,55]. Before performing simulations, a default relaxation protocol in DESMOND was employed: (1) NVT ensemble with Brownian dynamics for 100 ps at 10 K with small time step and solute heavy atom restrained; (2) NVT ensemble using Berendsen thermostat for 12 ps at 10 K with small time step and solute heavy atom restrained; (3) NPT ensemble using a Berendsen thermostat and barostat for 12 ps at 10 K and 1 atm with solute heavy atom restrained; (4) NPT ensemble using a Berendsen for 12 ps at 300 K and 1 atm with solute heavy atom restrained; and (5) NPT ensemble using a Berendsen for 24 ps at 300 K and 1 atm with no restraints. After the relaxation, MD simulation was performed for 200 ns. The integration time step used was 2 fs, and coordinate trajectories were saved every 200 ps. A DELL T7910 graphics workstation (with multiple NVIDA Tesla K40C GPUs) was used to run the MD simulations, and a 12-CPU CORE DELL T7500 graphics workstation was used to perform the preparation, analysis, and visualization. The Simulation Event Analysis module in DESMOND was used to analyze MD trajectories.

Electrophysiological recordings
Patch clamp recording of MgtE using E. coli giant spheroplasts was performed in the insideout configuration, as described previously [14,56]. In brief, spheroplasts were placed in a bath solution containing 210 mM N-methyl-D-glucamine, 90 mM MgCl 2 , 300 mM D-glucose, and 5 mM HEPES (pH 7.2). After gigaseal formation with a borosilicate glass pipette (Harvard Apparatus, Kent, United Kingdom) with 6 to 8 mOhm resistance, the bath medium was changed to bath solution containing 300 mM N-methyl-D-glucamine, 300 mM D-glucose, and 5 mM HEPES (pH 7.2) by perfusion using a Rainer perfusion pump and a custom-made perfusion system. Bath medium containing 2 μM Fab705 was added by perfusion. The pipette solution contained 210 mM N-methyl-D-glucamine, 90 mM MgCl 2 , 300 mM sucrose, and 5 mM HEPES (pH 7.2). Currents were measured with an Axopatch 200B amplifier and recorded with a Digidata 1440B A/D converter under the control of pCLAMP software (Molecular DevicesAU : PleaseprovidethemanufacturerlocationforMolecularDevicesinthesentenceCurrentswereme ). Currents were filtered at 2 kHz and recorded at 5 kHz. Additional reporting details regarding the electrophysiology experiments are shown in S12 Fig.

E. coli genetics
As described previously [14], the Mg 2+ -auxotrophic E. coli strain (BW25113 ΔmgtA ΔcorA ΔyhiD DE3) was transformed with MgtE or MgtE mutant plasmids and grown on LB medium plates containing 50 μg/ml kanamycin and 100 mM MgSO 4 . Each transformant was inoculated in liquid LB medium containing 50 μg/ml kanamycin and 100 mM MgSO 4 and incubated at 37˚C overnight with shaking. The liquid samples were diluted 1:100 with the same medium and grown until the OD600 reached 0.6. The sample cultures were then serially diluted 10-fold in LB medium containing 50 μg/ml kanamycin, spotted onto assay plates, and incubated at 37˚C overnight, as indicated in Fig 6C. In the Ni 2+ sensitivity assay, the isogenic magnesium-prototrophic wild-type E. coli strain (W3110 DE3) was transformed with the plasmid, and transformants were obtained on LB (+50 μg/ml kanamycin) plates. The liquid samples were diluted 1:100 with the same medium and grown until the OD600 reached 0.6. The sample cultures were then serially diluted 10-fold and spotted onto assay plates, and the plates were incubated at 37˚C overnight, as indicated in Fig 4D. The expression of MgtE and its mutants was confirmed by western blotting (S13 and S14 Figs).