P-type ATPase magnesium transporter MgtA acts as a dimer

Abstract

Magnesium (Mg²⁺) uptake systems are present in all domains of life, consistent with the vital role of this ion. P-type ATPase Mg²⁺ importers are required for bacterial growth when Mg²⁺ is limiting or during pathogenesis. However, insights into their mechanisms of action are missing. Here we solved the cryo-EM structure of the Mg²⁺ transporter MgtA from Escherichia coli. We obtained high-resolution structures of both homodimeric (2.9 Å) and monomeric (3.6 Å) forms. The dimer structure is formed by multiple contacts between residues in adjacent soluble N and P subdomains. Our structures revealed an ion, assigned as Mg²⁺, in the transmembrane segment. Moreover, we detected two cytoplasmic ion-binding sites and determined the structure of the N-terminal tail. Sequence conservation, mutagenesis and ATPase assays indicate dimerization, the ion-binding sites and the N-terminal tail facilitate cation transport or serve regulatory roles.

Extended Data Fig 1. Multisequence alignment of MgtA and MgtB illustrating conserved structural and functional features.

Sequence alignment of E. coli MgtA (EcMgtA), S. enterica serovar Typhimurium MgtA (SeMgtA), and S. enterica serovar Typhimurium MgtB (SeMgtB) generated using Clustal Omega. Domains are colored and named according to Fig. 1. Both the canonical P-type ATPase nomenclature and more detailed description of the fold of the domain with boundaries are given. The soluble A domain is split into two regions: a and b. The b segment of the A domain is comprised of a Double Stranded beta-Helix fold (DSβH). The soluble P subdomain is a noncontiguous segment comprised of two regions a and b that house the key catalytic residues required for phosphorylation. D373 is phosphorylated in EcMgtA. The N or CAP subdomain is a contiguous sequence that binds the nucleotide and aids in catalysis. The P and N subdomains comprise the haloacid dehalogenase (HAD) domain. The TM-spanning alpha-helical regions, as determined by a residue’s alpha-carbon position residing, on average, within the hydrocarbon bilayer interior of the molecular dynamics (MD) simulations (+/− 15 Angstroms from the bilayer midplane) are denoted by TM1–10. Gray circles denote residues present at the dimer interface, red circles denote residues near ATP, green circles denote residues near Mg²⁺ in the transmembrane domain, and black circles denote residues near Mg²⁺ in the cytoplasmic domain in our dimeric cryo-EM structures, as annotated in Fig. 2–4. Black dashed boxes indicate sequences conserved across all P-type ATPases as shown in the logos in Extended Data Fig. 2b. Red dashed boxes indicate sequences specific to Mg²⁺ importers as shown in the logos in Extended Data Fig. 7b.

Extended Data Fig 2. Tree of P-type ATPases and amino acids related to catalysis and structural architecture that are highly conserved across the P-type ATPase family.

a, In the tree, the clades are labeled and colored as per their known or predicted transport substrates. The P-type ATPase subclass is provided in brackets next to the aforestated label. All the major clades with IQtree bootstrap sport of 90% or higher are marked with a filled circle. Branches of special representatives, namely MgtA, MgtB, SERCA and SPCA1 are separately labeled and indicated with bold branches. The tree is based on a multiple sequence alignment of representatives of the major clades of P-type ATPase, which is in figshare. b, Sequence logos showing conservation of amino acid residues involved in ATP hydrolysis and structural architecture conserved among entire family of P-type ATPases (indicated by black dashed boxes in Extended Data Fig. 1). Letters represent amino acid abbreviations and height represents the probability of conservation in the P-type ATPase family. As in Extended Data Fig. 1, gray circles denote residues present at the dimer interface, red circles denote residues involved in ATP hydrolysis, green circles denote residues near Mg²⁺ in the transmembrane domain, and black circles denote residues near Mg²⁺ in the cytoplasmic domain in our dimeric structures, as annotated in Fig. 2–4. Gray, red, green and black arrows, respectively, indicate residues located at the dimer interface, involved in ATP hydrolysis, surrounding the transmembrane or cytoplasmic Mg²⁺, which were mutated in subsequent experiments.

Extended Data Fig 3. Purification of MgtA and ATPase activity of purified protein.

a, Solubilization of membranes expressing MgtA or MgtAS with detergents of varying strengths show differences in native protein interactions when analyzed by Native-PAGE and Western blot analysis. Membranes from cells overexpressing MgtA or MgtAS were solubilized with the detergents LMNG (L), DDM (D), or GDN (G) prior to Native-PAGE and Western blot analysis using α-FLAG antibodies against tagged MgtS (left two panels; middle panel is a longer exposure of the left panel) or α-His₆ antibodies against tagged MgtA (right panel). b, Purification of MgtA results in two distinct MW protein complexes. SEC profile (left) of MgtA used to solve the dimer and monomer structures of MgtA in Fig. 1. Fractions were analyzed by SDS-PAGE (middle) and Native-PAGE gels (right). V indicates the void volume of the SEC column. Fractions that were used to solve the structure are referred to as pooled. c, Dimeric MgtA can be separated from monomeric MgtA. SEC profile (left) of MgtA used to solve the nucleotide bound dimer structures of MgtA in Fig. 3 and Supplementary Fig. 3 and 5. Fractions were analyzed by SDS-PAGE (middle) and Native-PAGE gels (right). Fraction 18 which possessed predominantly dimer species was used for structural analysis. d, SDS-PAGE gel showing purity of purified WT and D373N mutant proteins. e, ATPase activity of WT MgtA in GDN in the presence of 0.5 or 5 mM MgCl₂ and with and without the addition of cardiolipin. f, ATPase activity of WT and D373N mutant MgtA in GDN in different concentrations of MgCl₂ between 0.5 to 20 mM. For e and f, protein was purified in 0.5 mM MgCl₂ and assays were supplemented with the indicated concentrations of MgCl₂. Assays were performed with 6.25 μg MgtA and 2 mM ATP at 37°C. Free phosphate was quantified using molybdate/malachite green-based assays (see methods). Data are shown for duplicate experiments with mean. Separate preparations of WT MgtA were assayed for e and f.

Extended Data Fig 4. Quality of cryo-EM dimer map for key structural features.

Cryo-EM map and atomic model of the TM segments (1–10), soluble domains (A, P, N), N-terminus, and Mg²⁺ ions colored as in Fig. 1. Atomic model of each structural element is shown in stick representation, the atoms are colored by heteroatom within the cryo-EM map and the corresponding map features are represented in gray mesh.

Extended Data Fig 5. Experimental support of transmembrane borders and lipid distribution.

a, Side views of cryo-EM maps at different thresholds showing extra densities in orange corresponding to the detergent micelle, detergent molecules, or potential co-purified lipids near the transmembrane region. b, Side view of MgtA simulated in a native lipid environment displayed in surface representation colored by electrostatic potential (UCSF Chimera coloring varies from red [−10 kcal/mol/e] to blue [+10 kcal/mol/e] with distance-dependent dielectric constant 4, distance from surface 1.4). Phospholipids corresponding to phosphatidylethanolamine (PE), phosphatidylglycerol (PG) and cardiolipin are colored tan, yellow, and green, respectively. c, Arginine and lysine residues that interact with anionic lipids during at least 20% of the simulation, with cutoff enclosing the first peak of the radial distribution function. d, Side view of MgtA with arginine (R), lysine (K), and tryptophan (W) residues near the lipid membrane borders highlighted in blue spheres. e, Voronoi decomposition of lipid centers-of-geometry in the cytoplasmic leaflet initially (initial) and at the end of the simulation (final). Yellow and green are anionic PG and cardiolipin lipids, respectively. At right, average enrichment or depletion of lipids based on solvation shell as assigned by Voronoi decomposition averaged over the trajectory. Data are shown for independent replicates (n=10) from the same simulation with mean and standard deviation (error bars) for each solvation shell.

Extended Data Fig 6. More splayed conformation of monomeric MgtA structure relative to the dimer with major changes in TM1–4, N and A domains.

a, Dimeric structure with two subunits in two different gray tones and Mg²⁺ ions as green spheres. b, Monomeric structure colored as in Fig. 1. c-d, To visualize structural changes the colored monomer and a single dimer subunit in gray were superimposed (c) and RMSD was calculated (d). The monomer structure is shown colored by RMSD indicating the largest differences when comparing the two structures by high RMSD values in red. e-f, Structural differences between the dimer (gray) and monomer (purple) for TM1–2 (e) and TM3–4 (f). The monomer colored according to RMSD is on the right with densities in mesh indicating the fit. g, Monomer density features and fitted model of TM5 and TM7 near ion-binding site A in the middle of the membrane which is similar to the dimer structures. h-j, Differences in position in the overall conformation between the soluble domains of the dimer (gray) and monomer (colored).

Extended Data Fig. 7. Stability of Mg²⁺ binding in simulation and conservation of key Mg²⁺-binding residues.

a, Adaptive-biasing MD simulations perturbing Mg²⁺ from the three binding sites and K⁺ from site A. Simulated barriers are similar for Mg²⁺ at each site at 13–15 kcal/mol, with K⁺ much weaker. Data are for independent simulations (n=3) with mean (colored solid line) and standard deviation (shaded band) indicated. b, Sequence logos displaying conservation of amino acids in different members of the P-type ATPase transporters. Letters represent amino acid abbreviations; the height of each letter represents the relative probability of conservation among members of the P-type ATPase family. Logos correspond to the Mg²⁺ TM binding site A to illustrate residues conserved in the MgtA clade (top) compared to the Ca²⁺ clade (bottom). The sequence logos are also highlighted in the reduced multisequence alignment in Extended Data Fig. 1. Green circles denote residues near Mg²⁺ in the transmembrane domain.

Extended Data Fig. 8. Water accessibility in the TM domain of MgtA.

Figures are arranged such that rows correspond with overlapping 1.5 nanometer thick cuts spaced every 1 nanometer. The left column is the side view from the simulation, while the middle column is from the top-down starting on the periplasmic side of the transporter. For the simulation, waters (including hydrogens) are shown in sphere representation. At right is the corresponding top-down view of the MgtA dimer from cryo-EM, with resolved water molecules shown as red spheres. The protein ribbon model is shown opaque through the cut of the simulation, while outside of the cut waters are not shown, and the protein is transparent. The transmembrane Mg²⁺ as well as glutamic and aspartic acid residues 331 and 780 (respectively) are shown in stick representation.

Extended Data Fig. 9. Extended N-terminus cannot be predicted by AlphaFold, conceals negative patches but does not affect ATPase activity.

a, Alignment showing extended N-terminus present in MgtA, H⁺-pumps, H⁺/K⁺-ATPases and Na⁺/K⁺-ATPases but not Ca²⁺-pump SERCA. The sequence alignment of E. coli MgtA (EcMgtA), Plant Pma2, Fungal Pma1, ATP4A (H⁺/K⁺ transporter) from pig and human, ATP1A1 (Na⁺/K⁺ transporter) from pig and human, and SERCA (ATP2A1) from rabbit and human was generated using Clustal Omega. Domains are colored and named according to Fig. 1. A more complete alignment can be found in figshare. b, AlphaFold models of Fungal Pma1, SeMgtB and EcMgtA. AlphaFold’s low confidence score for the N-terminal tails predicts them to be disordered. c, RMSD calculated between the AlphaFold model and a single subunit of the dimeric MgtA structure to visualize differences. d, Front and side view of the surface representation of the electrostatic potential displayed for WT and the deletion of the N-terminus (Δ1–36). Side view is rotated 90°. e, Native-PAGE gel of SEC fractions 1–6 following Superose 6 fractionation of Δ1–31 in 5 mM MgCl₂. Bolded fractions 2 and 5 were used in the ATPase assay. f, ATPase activity of dimer and monomer fractions with either 5 mM or 20 mM MgCl₂. Assays were performed with 6.25 μg MgtA and 2 mM ATP at 37°C and free phosphate was quantified using molybdate/malachite green-based assays (see methods). Data are shown for duplicate experiments with mean. g, Levels of MgtA are reduced by mutations of the N-terminus, to either eliminate the N-terminus or mutate residues conserved in the Mgt family and suggested to play a role in membrane sensing. Cells were grown uninduced (- IPTG) overnight at 37°C in LB supplemented with 100 mM MgSO₄ and normalized in lysis buffer prior to western blot analysis with polyclonal anti-MgtA antibodies. Non-specific bands, also observed with just the vector control, are detected with the native antibody for the low levels of MgtA assayed here. Ponceau S-stained membrane (right) serves as a loading control.

Extended Data Fig. 10. Predicted and documented interactions of small proteins with P-type ATPase proteins.

a, Dimer cryo-EM structure of E. coli MgtA from this study with E. coli MgtS binding predicted by AlphaFold Multimer beta. b, Models of E. coli MgtA monomer and E. coli MgtS and S. enterica MgtA and S. enterica MgtS, MgtU and MgtR predicted by AlphaFold Multimer beta. c, Selected structures of indicated P-type ATPases solved with small α-helical proteins. P-type ATPases are in gray with small protein in red (refs).

Fig. 1. Cryo-EM of Mg²⁺ transporter MgtA reveals a high-resolution dimer and a monomeric structure.

a, Schematic representation of MgtA/B based on P-type ATPase structural homology predicting ten conserved TM helices colored in purple (1–10), the actuator (A) domain in orchid, the phosphorylation (P) subdomain in light blue, the nucleotide (N) binding or CAP subdomain in light green, and the predicted unstructured N-terminal tail in orange. The A domain is split into two regions a/b. The b segment of the A domain is comprised of a Double Stranded beta-Helix fold (DSβH). The soluble P subdomain is a noncontiguous segment comprised of two regions a/b that house the key catalytic residues required for phosphorylation. The N subdomain binds the nucleotide and aids in catalysis. The P and N subdomains together comprise the Haloacid dehalogenase (HAD) domain. Created with BioRender.com. b, Post-Albers reaction scheme indicating the typical cytoplasmic side-open (E1), ATP-bound cytoplasmic side-open (E1.ATP), periplasmic side-open (E2P) and ion-occluded (E2) states. c, Representative micrograph of purified Escherichia coli MgtA. Representative image of the n > 17,000 total micrographs. d-e, Representative final 2D class averages for the ~200 kDa dimer (d) and the ~100 kDa monomer (e), respectively, with a box size of 384 pixels (approximately 319 Å). f-g, Cryo-EM reconstruction of the MgtA monomer (f, also see Supplementary Fig. 2 and 3c–d, Supplementary Video 2) and of the dimer (g, also see Supplementary Fig. 2 and 3a–b, Supplementary Video 1). Cryo-EM maps in f-g are colored as in a. A transparent gray map at a much lower threshold indicates the detergent micelle and the map features for the more flexible loops and the A domain in the monomer map.

Fig. 2. The dimer interface is formed by both hydrophobic and polar interactions.

a, Side view of the overall dimer structure with the left monomer in gray and the right monomer colored as in Fig. 1. A close-up view of the extensive dimer interface between the two N and P subdomains with sidechain residues displaying charged interactions across the dimer interface is shown below. Rotation of the structure highlights hydrophobic interactions at the dimer interface. Molecular dynamic simulations (see Supplementary Videos 4 and 5) show consistent interactions across the dimer interface between K382-E582 (64%), Q380-Q380 (87%), and K548-E549 (88%) shown in bold. b, Native-PAGE gel of SEC fractions 1–6 following Superose 6 Increase 10/300 GL. Bolded fractions 2 and 4 for 0.5 and 5 mM MgCl₂ were used in ATPase assays. c, ATPase activity of dimer and monomer fractions purified in low and high magnesium conditions. Assays were performed with 6.25 μg of MgtA and 2 mM ATP at 37°C and phosphate released was quantified using molybdate/malachite green-based assays. Data are shown for duplicate experiments with mean. d, Co-purification of two differentially tagged MgtA derivatives, one tagged with His6 and the other tagged with the larger SPA tag as shown in the top schematic reveals copurification of the two proteins (right elution panel). The experiment was performed 2x with consistent results. Proteins were visualized by Western blot analysis using MgtA antibodies.

Fig. 3. The nucleotide binding pocket of MgtA is accessible in the dimeric state.

a, Side view of the MgtA dimer with the left monomer in gray and the right monomer colored as in Fig. 1, highlighting the ATP molecule represented in spheres located in between the soluble A domain and P and N subdomains and the dephosphorylation TGES loop in yellow. b, A close-up view of the ATP-binding site highlighting residues in close proximity to the Mg-ATP molecule which is shown in a ball and stick representation and the cryo-EM map in gray mesh. Residues from the TGES loop involved in dephosphorylation and located in the A domain are colored yellow. N415 is omitted for a better view of the site. c, Distances from ATP to amino acids, water molecules and Mg²⁺ ion in the nucleotide binding pocket of MgtA from E. coli, determined using LIGPLOT+. d, The D373N mutant is unable to completement a Mg²⁺-auxotrophic E. coli strain indicating this mutant transporter does not import Mg²⁺ ions. The D373 residue gets phosphorylated upon ATP hydrolysis. The E215A mutant is only partially able to complement upon overexpression. The E215 residue is part of the TGES loop involved in dephosphorylation. Overnight cultures were serial diluted and spotted onto LB agar plates supplemented with high (100 mM) or low (1 mM) MgSO₄ with (+) and without (−) 0.1 mM IPTG for induction and grown at 37°C (see replicate and protein levels at figshare).

Fig. 4. Mg²⁺ binds sites in the transmembrane domain and between the cytosolic A domain and N and P subdomains.

a, Side view of the dimeric structure with left monomer in gray and right monomer colored as in Fig. 1 and close-up views of the resolved Mg²⁺ ions (green), nearby residues, and proximal resolved water molecules (red spheres). Black dotted lines indicate residue distances to Mg²⁺. b, Structural comparison of selected ion-bound P-type ATPases. One subunit of the dimeric EcMgtA transporter is shown with corresponding regions of the SERCA1 Ca²⁺ transporter (PDB 2ZBD) and ATP1A1 Na⁺/K⁺ transporter with Rb⁺ as K⁺ congeners (PDB 3B8E) or Na⁺ (PDB 4HQJ). Residues of EcMgtA, SERCA1 and ATP1A1 predicted to be involved in ion binding are colored pink, the Mg²⁺ and Ca²⁺ ions are green and the Rb⁺ and Na⁺ ions are purple. The residues mutated in MgtA are labeled in bold font. Those residues in SERCA1 and ATP1A1 that are part of the CBS and correspond to the MgtA E331, N709 and D738 are also labeled in bold font. For the ATP1A1 Na⁺-bound structure, a Na⁺ ion coordinated by D926, is found in a location similar to the Mg²⁺ observed in MgtA. c, Mutants with E331A (TM4) or D780A (TM7) are unable to complement a Mg²⁺-auxotrophic E. coli strain. d, The D441A (site C) mutant is only partially able to complement the Mg²⁺-auxotrophic E. coli strain upon overexpression. Complementation assays in c and d were carried out as for Fig. 3d (see replicate and protein levels at figshare). e, Native-PAGE gel of WT and mutants as described in Fig. 2b. Bolded fractions 2 and 4 for WT and D441A and fractions 1 and 5 for D780A were used for ATPase activity. The experiment was performed 2x with similar results. f, ATPase activity of dimer and monomer fractions of WT and mutant MgtA purified in 5 mM MgCl₂. Assays were performed as described in Fig. 2c. Data are shown for duplicate experiments and mean. Data for WT MgtA are the same as in Fig. 2b–c.

Fig. 5. The extended N-terminus forms multi-domain electrostatic interactions between the A, P and TM domains.

a, Front and side view of the EcMgtA dimer structure colored as in Fig. 1 with the N-terminus highlighted in a thicker orange loop. b, Close-up views of sites I-IV which mark salt bridges formed between the N-terminus and the various domains of MgtA. All residues are represented in stick and colored by heteroatom.

Fig. 6. Summary of MgtA structural insights.

a, Post-Albers cycle with proposed states of an MgtA dimer model based on known states from other P-type ATPases. Given that MgtA transports Mg²⁺ from the periplasm into the cytoplasm, a counter ion may be needed to complete the transport cycle. These states associate with a cytoplasmic ion-dependent autophosphorylation and periplasmic ion-dependent dephosphorylation, respectively. In the E1 state, P-type ATPases adopt a cytoplasmic side-open conformation where the ion-binding sites are open to the cytoplasm, allowing the cytoplasmic ion to access the transmembrane binding sites. Upon Mg-ATP and ion-binding, P-type ATPases undergo autophosphorylation at the conserved aspartate residue in the P subdomain, thereby switching to an ion-occluded E1•P-ADP state where the ion-binding sites are not accessible to the cytoplasm. Upon ADP release, the P-type ATPases switch to an E2P state and the ion-binding sites open to the periplasmic side for ion release and subsequent periplasmic ion binding. The binding of the periplasmic ion triggers dephosphorylation and drives the conformational change of the P-type ATPase to an ion-occluded state (E2•P_i). The P-type ATPase subsequently switches to the E2 state after the release of the phosphate, and finally back to an E1 state releasing the ion into the cytoplasm allowing for the transport cycle to begin again. b, Overview of putative regulatory features of MgtA including the dimer interface, cytoplasmic Mg²⁺-binding sites, N-terminal domain, lipid environment, and small protein binding. Green circles represent Mg²⁺ ions. Color scheme is the same as in Fig. 1.