Research ArticleSTRUCTURAL BIOLOGY

Structural basis for the Mg2+ recognition and regulation of the CorC Mg2+ transporter

See allHide authors and affiliations

Science Advances  10 Feb 2021:
Vol. 7, no. 7, eabe6140
DOI: 10.1126/sciadv.abe6140

Abstract

The CNNM/CorC family proteins are Mg2+ transporters that are widely distributed in all domains of life. In bacteria, CorC has been implicated in the survival of pathogenic microorganisms. In humans, CNNM proteins are involved in various biological events, such as body absorption/reabsorption of Mg2+ and genetic disorders. Here, we determined the crystal structure of the Mg2+-bound CorC TM domain dimer. Each protomer has a single Mg2+ binding site with a fully dehydrated Mg2+ ion. The residues at the Mg2+ binding site are strictly conserved in both human CNNM2 and CNNM4, and many of these residues are associated with genetic diseases. Furthermore, we determined the structures of the CorC cytoplasmic region containing its regulatory ATP-binding domain. A combination of structural and functional analyses not only revealed the potential interface between the TM and cytoplasmic domains but also showed that ATP binding is important for the Mg2+ export activity of CorC.

INTRODUCTION

The magnesium ion (Mg2+) is the most abundant biological divalent cation and is essential for all living organisms (1). Mg2+ plays a vital role in various physiological processes, such as maintaining genome stability and acting as enzymatic cofactors for adenosine 5′-triphosphate (ATP) hydrolysis and DNA replication (2, 3). In humans, Mg2+ is involved in the maintenance of heart rhythm, blood pressure, neuromuscular conduction, bone integrity, and glucose metabolism (4). Accordingly, cellular imbalance of Mg2+ homeostasis is associated with multiple diseases (4), and Mg2+ transport proteins play a primary role in homeostasis (5, 6).

The CNNM (cyclin M)/CorC family of Mg2+ transporters is widely distributed in all domains of life, from prokaryotes to eukaryotes, including humans (7). CorC, a prokaryotic member of the CNNM/CorC family of proteins, is implicated in Mg2+ transport in several microbial organisms (812). In Staphylococcus aureus, a pathogenic bacteria, CorC proteins confer resistance to high Mg2+ concentrations in their infected host environment, enhancing their pathogenicity (10, 12). In Lactobacillus plantarum, an interaction between CorC and an antimicrobial peptide of plantaricin EF is associated with divalent cation stress (13). Furthermore, CorC expression is up-regulated in the L22* strain of Bacillus subtilis to increase Mg2+ flux, which is associated with resilience to ribosome-targeting antibiotics in the L22* strain (11).

In mammals, including humans, CNNM proteins (CNNM1 to CNNM4) are widely expressed throughout the body (14). Among them, CNNM2 and CNNM4 mediate Mg2+ efflux, which is responsible for the absorption/reabsorption of Mg2+ from the intestines and kidney, respectively (1517). Mutations of CNNM2 and CNNM4 are responsible for human congenital diseases, dominant hypomagnesemia (17) and Jalili syndrome (18, 19), respectively. Knockout mouse studies confirmed the occurrence of some disease-related phenotypes (15, 16) and revealed the importance of CNNM2 in blood pressure regulation (16). Furthermore, abnormal functions of CNNM3 and CNNM4 have been reported to cause tumor progression (2022). Accordingly, the CorC and CNNM proteins are potential targets for the development of previously unknown antibiotics and cancer therapeutics, respectively.

Both CorC and CNNM proteins share a DUF21 transmembrane (TM) domain and cystathionine β-synthase (CBS) domain (8, 10, 14, 23). The DUF21 domain is responsible for the TM transport of Mg2+, and many disease-associated mutations of CNNMs are localized to the DUF21 domain (18, 19, 24). The CBS domain is an evolutionally conserved domain found in a wide range of proteins and typically has binding activity to adenosine and its analogs, such as ATP and adenosine 5′-monophosphate, to regulate the functions of the associated enzymes and transporters (2527). Mutations in the CBS domain of MpfA, a CorC ortholog from S. aureus, reportedly affect its transporter functions (10). ATP binding to the CBS domain of CNNM2 and CNNM4 is required for their Mg2+ efflux activities (28, 29). Furthermore, there have been multiple structures containing the CBS domain of the CNNM/CorC family proteins reported to date (2936). Despite the physiological importance of the CNNM/CorC family, little is known about the Mg2+ transport mechanism of the CNNM/CorC family proteins, mainly because the structure of the CNNM/CorC proteins containing the DUF21 TM domain has not yet been reported.

In this work, we determined the high-resolution structure of the CorC TM domain in complex with Mg2+ ions, illuminating the Mg2+ recognition mechanism, which highly contrasts with that of the known Mg2+ channel structures. Notably, all residues at the Mg2+ binding site are strictly conserved in both human CNNM2 and CNNM4 Mg2+ exporters, and many of the corresponding residues in human CNNM2 and CNNM4 were associated with congenital diseases, providing a structural interpretation of these disease-associated mutations. Further structural analyses of the CorC cytoplasmic domain containing the ATP-binding site and associated functional analyses not only revealed the potential interface between the TM and cytoplasmic domains but also showed that ATP binding is important for the Mg2+ export activity of CorC. Overall, our work provides structural insights into the Mg2+ recognition and regulation of the CorC Mg2+ transporter.

RESULTS

Structural determination and functional characterization

We screened the expression of over 140 CNNM/CorC homologs by fluorescence-detection size exclusion chromatography (FSEC; table S1) (37). In the green fluorescent protein (GFP)–based FSEC method, we expressed targeted membrane proteins fused to the GFP tag, and whole-cell extracts were solubilized with detergents and subjected to size exclusion chromatography attached to a fluorescence detector to analyze the expression level and monodispersity of targeted membrane proteins (37). Among the screened CNNM/CorC orthologs, we identified the CorC protein from Thermus parvatiensis (TpCorC) as a suitable candidate for structural studies, with a high level of expression and sharp peak profile (Fig. 1A). The domain organization of TpCorC consists of a DUF21 TM domain, a CBS domain, and a CorC/HlyC domain, which is typical of CorC family proteins (Fig. 1B). Among these domains, the DUF21 and CBS domains are also shared by CNNM family proteins and have many disease-associated mutations (1719, 24), indicating the possible functional importance of these core regions in CNNM/CorC family proteins.

Fig. 1 Expression and functional characterization of TpCorC.

(A) FSEC profile of the C-terminal GFPuv-tagged TpCorC (residues 23 to 441) on a Superdex 200 Increase 10/300 GL column (GE Healthcare) (excitation, 395 nm; emission, 507 nm). (B) Representative domain organization of the CorC and CNNM proteins. (C) Cell surface expression of TpCorC. (D and E) Mg2+ export assay. (D) Representative time course of mean relative fluorescent intensities. Mg2+ was depleted at the time point indicated with an arrowhead. (E) Relative fluorescence intensities after Mg2+ depletion (at 5 min, means ± SEM; empty, n = 30; wild type (WT), n = 30; WT (−Na+), n = 20; TM-CBS, n = 20; and V101A, n = 20). (F) Thermostability assay of the WT and TpCorC TM domain mutants by GFP-based FSEC-TS.

TpCorC shares approximately 30% sequence identity with the S. aureus CorC orthologs MpfA and MpfB, which are known Mg2+ export CorC proteins (fig. S1) (10, 12). Thus, we tested the Mg2+ export activity of TpCorC using the Mg2+ efflux assay in human embryonic kidney (HEK) 293 cells (Fig. 1, C to E), which have been used to characterize the Mg2+ export activity of human CNNM2 and CNNM4 (28). To facilitate the cell surface expression of TpCorC, we fused the membrane targeting sequence from human CNNM4 (residues 1 to 178) at the N terminus (14), established a HEK293 cell line stably expressing TpCorC with the targeting sequence, and confirmed the cell surface expression (Fig. 1C). Using this cell line, we performed imaging analyses using Magnesium Green, a fluorescent indicator dye for Mg2+. The intensity of the fluorescent signal in the cells expressing TpCorC decreased after the removal of Mg2+ ions from the bath solution, which was similarly observed with CNNM2 and CNNM4 Mg2+ exporters (28), whereas we observed little change in the fluorescent signal in HEK293 cells transfected with empty vector (Fig. 1, D and E). We then tested the Na+ dependency of the Mg2+ export activity of TpCorC. The depletion of Na+ from the bath solution induced the loss of Mg2+ export activity, suggesting that the Na+ gradient could be a potential driving force for Mg2+ export from TpCorC, as is the case with human CNNM4 (Fig. 1, D and E) (15). The Mg2+ export activity of TpCorC suggests the functional similarity of TpCorC to previously characterized Mg2+-exporting CNNM/CorC family proteins such as MpfA/B and CNNM2/4 (10, 12, 15, 28).

Truncation of the cytoplasmic domain yielded TpCorC crystals of the TM domain (residues 26 to 182) in detergent micelles, poorly diffracting to ~7 Å resolution. Thus, to improve the diffraction quality of TpCorC crystals, we performed alanine scanning of the TpCorC TM domain by an FSEC-based thermostability (FSEC-TS) assay (38) and identified two mutants (V101A and V101A/G115A) with increased melting temperatures (Tm values) of 8.4° and 8.9°C, respectively (Fig. 1F). Mutation at Val101 did not affect Mg2+ export activity (Fig. 1, D and E), whereas truncation of the entire cytoplasmic domain induced a loss of localization at the cell surface in HEK293 cells (Fig. 1C). We applied the lipidic cubic phase (LCP) method, which has been successful for membrane protein crystallization (39), for the crystallization of the TpCorC TM domain constructs (wild type and V101A), yielding crystals diffracting to 3.2 and 2.0 Å resolutions, respectively (table S2). Notably, these crystals appeared specifically in the presence of Mg2+ ions, whereas we also tested cocrystallization with other divalent cations such as Ca2+, Co2+, Ni2+, and Mn2+. The initial phase was determined by the single-wavelength anomalous diffraction (SAD) method with the selenomethionine (SeMet)–labeled TpCorC V101A/G115A mutant protein and was further refined to 2.0 Å resolution using the native datasets of the TpCorC TM domain constructs (wild type and V101A) (fig. S2, A and B, and table S2). The two TpCorC TM domain structures (wild type and V101A) are essentially identical with a root mean square deviation (RMSD) of 0.2 Å. Thus, we mainly describe the TpCorC TM domain V101A structure unless otherwise noted, as it was determined at higher resolution. The location of Val101 in the TM domain structure and a possible explanation for the thermostabilization effect of the V101A mutation are described in fig. S3.

TM domain architecture

The crystallographic asymmetric unit of the TpCorC TM domain contains one protomer, which forms a dimer, related by crystallographic twofold symmetry (Fig. 2 and fig. S2, A and B). Each TpCorC protomer is composed of three TM helices with three cytoplasmic helices, termed cytoplasmic helix 1 (CH1) and CH2, located between the TM1 and TM2 helices, and the belt helix following the TM3 helix (Fig. 2, D and E). The N- and C-terminal ends of TpCorC extend into the periplasm and cytoplasm, respectively, because the C-terminal region of TpCorC has a cytoplasmic region including the CBS domain (Fig. 1B).

Fig. 2 Architecture of the TM domain.

(A to C) Cartoon representations of the TpCorC TM domain dimer in the Mg2+-bound form, viewed parallel to the membrane (A), from the extracellular side (B), and from the intracellular side (C). (D) Each protomer of the TM domain is rainbow colored from the N terminus to the C terminus (TM1, blue; CH1 and CH2, cyan; TM2, green; TM3, yellow; belt helix, red). (E) Schematic topology of the TpCorC TM domain protomer. The cytoplasmic domains (CBS and CorC domains) are also shown with dashed lines, which are not included in the TM domain structure. (F) Structural deviation from the TpCorC TM domain structure during 300-ns MD simulations with and without Mg2+ ions. (G) The root mean square fluctuation (RMSF) from the TpCorC TM domain structure with and without Mg2+ ions is shown as a function of residue number.

From the cytoplasmic view, there are close interactions within the TM1, TM2, and TM3 helices in each protomer but not within the TM helices of the neighboring subunit (Fig. 2C). In contrast, from the periplasmic view, the TM2 and TM3 helices interact closely with their counterparts in the neighboring subunit (Fig. 2B). Consequently, the TpCorC transporter structure seemingly adopts an inward-opening conformation, where the cytoplasmic side is solvent accessible but the periplasmic side is closed (Fig. 3A).

Fig. 3 Mg2+ binding site of TpCorC.

(A) Cross-sectional view of the surface model of the TpCorC TM domain dimer with the close-up view of the Mg2+ binding site. The potential displayed represents the range from −20 (red) to 20 (blue) kT/e. Amino acid residues involved in Mg2+ ions are shown in stick representation. Mg2+ ions are shown as spheres. Dashed lines indicate hydrogen bonds. (B) Cell surface expression of TpCorC Mg2+ binding site mutants. (C) Mg2+ export assay. Bar graph: Relative fluorescence intensities after Mg2+ depletion (at 5 min, means ± SEM; empty, n = 10; WT, n = 10; S47A, n = 15; G129A, n = 12; and E130A, n = 10).

The belt helix after the TM3 helix has three unique features (Fig. 2 and fig. S4). First, it is nearly parallel to the plane of the membrane (Fig. 2A). Second, it is amphipathic so that the hydrophobic side of the belt helix is embedded in the membrane bilayer, whereas the hydrophilic side faces the solvent (Fig. 2A and fig. S4, A and B). Last, the belt helix is quite long (approximately 35 residues) and highly curved to interact with the TM1 and TM2 helices within the protomer and with the TM3 helix in the neighboring subunit (Fig. 2, B and C). Molecular dynamics (MD) simulations of the TpCorC TM domain dimer embedded in a 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine membrane showed that the overall conformation of the TpCorC TM domain structure, including the unique belt helix, was mostly stable throughout the simulations (Fig. 2, F and G). The structural deviations from the original crystal structure were within 2.5 Å during 300-ns MD simulations, and the root mean square fluctuation values as a function of residue number were also within 2.5 Å after the 300-ns MD simulations. Overall, the protein fold of TpCorC is distinct from those of any previously reported membrane protein structures that are likely to be conserved among CNNM/CorC family proteins.

Mg2+ ion binding site

The TpCorC TM domain dimer structure revealed a strong residual electron density peak surrounded by the side chain or main chain atoms of five amino acid residues of each protomer within the TM domain (fig. S5). We obtained crystals of the TpCorC TM domain by cocrystallization specifically in the presence of Mg2+ ions. In addition to Mg2+ ions, our crystallization conditions include Zn2+ ions as divalent cations. Therefore, these two types of divalent cations, Mg2+ and Zn2+, would be potential origins for the strong electron density in our structure.

To clarify the origin of the strong electron density in the TM domain structure, we initially attempted to use Co2+, Ni2+, and Mn2+, heavy divalent cations, which can occupy Mg2+ binding sites based on their octahedral coordination geometry, similar to that of Mg2+ ions (40, 41). However, neither cocrystallization nor soaking with these heavy atoms yielded well-diffracting crystals. We also collected the diffraction data at the wavelength for the Zn2+ anomalous peak to verify the possibility of Zn2+ as an origin of the strong electron density.

The anomalous difference Fourier map at the wavelength for the Zn2+ anomalous peak did not show any anomalous peaks at the corresponding position (fig. S5E), whereas it clearly showed anomalous peaks for Zn2+ ions at the crystal-packing interface between neighboring dimers (fig. S5F). Furthermore, the coordination of the surrounding side chain or main chain atoms with the residual electron density peak is highly consistent with that of the fully hydrated Mg2+ ion, adopting an octahedral geometry (Fig. 3A) (42) instead of the typical tetrahedral coordination geometry of Zn2+ ions (42). Overall, on the basis of the abovementioned reasons, we interpret the density as Mg2+ ions.

Mg2+ ions are directly coordinated to the side chains of Ser43, Ser47, Asn90, and Glu130 and the main chain carbonyl oxygen atoms of Ser43 and Gly129; thus, Mg2+ ions are fully dehydrated in the TpCorC TM domain structure. The bonding distances between the Mg2+ ion and the coordinated atoms of TpCorC are 2.1 to 2.2 Å, which is consistent with the canonical distance of 2.1 Å between the Mg2+ ion and oxygen atom in an aqueous environment (42). All five of these residues are similarly or highly conserved among CorC and CNNM proteins (fig. S1). Notably, all five amino acid residues involved in Mg2+ binding are strictly conserved in both human CNNM2 and CNNM4 (fig. S1). All TM helices within the protomer participate in Mg2+ binding, but there is no intersubunit bridging involved in Mg2+ coordination. To test the functional importance of the residues involved in Mg2+ binding, we then generated alanine-substituted mutants of TpCorC (S43A, S47A, N90A, G129A, and E130A), established HEK293 cell lines stably expressing these mutants, and performed Mg2+ efflux assays (Fig. 3, B and C).

While S43A and N90A were not expressed at the cell surface, all other mutants were properly expressed at the membrane surface (Fig. 3B). The fluorescence intensity from cells expressing the wild-type protein decreased after the depletion of Mg2+ ions, but all Mg2+ binding site mutants expressed at the cell surface exhibited a smaller change in fluorescence signal (Fig. 3C). These results indicated the important role of these residues in Mg2+ transport by TpCorC.

Mg2+-dependent conformational equilibrium

The current TpCorC TM domain structure seemingly adopts an inward-facing conformation (Fig. 3A) that seems to be stabilized by the binding of Mg2+ ions, which bridge all the TM helices in the protomer (Fig. 3A). Therefore, we hypothesized that Mg2+-bound TpCorC preferentially adopts an inward-facing conformation. This hypothesis predicts that the removal of Mg2+ from TpCorC may disrupt the stable formation of the inward-facing conformation.

To test our hypothesis, we established biochemical cross-linking experiments for TpCorC (Fig. 4). First, we generated the TpCorC Cys-less mutant C282A and its mutant with cysteine substitutions at Thr106 (T106C/C282A), where the Cβ distance between the Thr106 residues of the two subunits is 7.1 Å, which is close enough for chemical cross-linking via Cys residues (Fig. 4A). Using Cu2+ phenanthroline as an oxidizing reagent, we conducted a cross-linking experiment on the wild type, the Cys-less C282A mutant, and the T106C/C282A mutant of TpCorC (Fig. 4B). The cysteine mutant T106C exhibited a strong band for the TpCorC dimer in the presence of Cu2+ phenanthroline, verifying our TM domain dimer structure, whereas the Cys-less mutant showed only the band for the TpCorC monomer (Fig. 4B). In addition, FSEC analysis showed that the wild type, the C282A mutant, and the T106C/C282A mutant treated with Cu2+ phenanthroline eluted in a similar position, indicating that Cu2+ phenanthroline treatment did not affect the oligomeric state of all constructs (Fig. 4C).

Fig. 4 Mg2+-dependent conformational equilibrium.

(A) Cartoon representations of the TpCorC TM domain dimer, viewed parallel to the membrane (left) and from the extracellular side (right). The Thr106 residues are shown in stick representation. Dotted lines indicate the Cβ distances between the Thr106 residues between subunits. (B) Chemical cross-linking experiments of the wild-type, Cys-less mutant (C282A) and intersubunit cross-linking mutant (T106C/C282A) of TpCorC. (C) FSEC profiles of the TpCorC WT, C282A, and T106C/C282A mutant on a Superdex 200 Increase 10/300 GL column (excitation, 280 nm; emission, 325 nm). (D) Chemical cross-linking experiments of the intersubunit cross-linking mutant (T106C/C282A) and its mutants. The symbols below the SDS–polyacrylamide gel electrophoresis (SDS-PAGE) gel indicate the experimental conditions (Mg2+ +, 10 mM MgCl2; Mg2+ −, 0.2 mM EDTA; Na+ +, 150 mM NaCl; Na+ −, 150 mM KCl).

As we hypothesized, the removal of Mg2+ from the T106C/C282A mutant with EDTA in the presence of Na+ caused the loss of the cross-linked dimer (Fig. 4D), indicating disruption of the inward-facing conformation of the transporter. In contrast, there was a strong band for the TpCorC dimer in the presence of Mg2+ (Fig. 4D). The combination of Mg2+ removal with EDTA and the replacement of Na+ with K+ resulted in bands for both the TpCorC monomer and dimer (Fig. 4D), potentially suggesting a conformational equilibrium in the absence of both Mg2+ and Na+.

We then introduced mutations to the Mg2+ binding site of TpCorC (N90A and E130A) into the TpCorC T106C/C282A mutant. Notably, each mutant exhibited both bands for the TpCorC monomer and dimer and was not sensitive to either the addition of Mg2+ or EDTA (Fig. 4D). These results suggest that the Mg2+ binding site is crucial for the stable formation of the inward-facing conformation and suggest that adding Na+ in the absence of Mg2+ may further destabilize the inward-facing conformation.

Mutational analysis of putative Na+ binding site

Our Mg2+ efflux assay for TpCorC in HEK293 cells suggested Na+ dependency of the transport activity (Fig. 1, D and E). Upon close inspection of the TpCorC TM domain structure, we speculated that the dimer interface, located near the Mg2+ binding site, might be a potential candidate for Na+ binding sites (Fig. 5A) because this region includes multiple Asn residues (Asn91 and Asn94), which are often involved in monovalent cation binding in transport proteins (4346). Among them, Asn94 is conserved in both CNNM2 and CNNM4 (fig. S1), and CNNM4 is shown to have Na+-dependent transport activity (15). Furthermore, the existence of a Na+ binding site proximal to the Mg2+ binding site would be consistent with Na+-coupled Mg2+ transport, as often observed in other Na+-driven secondary active transporter structures (4346).

Fig. 5 A potential residue implicated in Na+ binding.

(A) Close-up views of the TpCorC intersubunit interface near the Mg2+ binding site. The overall structure is also shown. Amino acid residues are shown in stick representation, and Mg2+ ions are shown as green spheres. (B) Chemical cross-linking experiments of the intersubunit cross-linking mutant (T106C/C282A) and its mutants. The symbols below the SDS-PAGE gel indicate the experimental conditions (Mg2+ +, 10 mM MgCl2; Mg2+ −, 0.2 mM EDTA; Na+ +, 150 mM NaCl; and Na+ −, 150 mM KCl). (C) Cell surface expression of TpCorC and CNNM mutants. (D and E) Mg2+ export assay of TpCorC mutants (D) and CNNM mutants (E). Bar graph: Relative fluorescence intensities after Mg2+ depletion (at 5 min, means ± SEM, n = 10).

To test our notion, we introduced alanine-substituted mutations (N91A and N94A) into the T106C/C282A mutant of TpCorC and performed chemical cross-linking experiments (Fig. 5B). Whereas the addition of Na+ to the T106C/C282A mutant in the absence of Mg2+ caused the loss of the cross-linked dimer, the N94A/T106C/C282A mutant exhibited bands for both the TpCorC monomer and dimer upon the addition of Na+ in the absence of Mg2+ (Fig. 5B), suggesting that the Asn94 residue may be involved in Na+ sensitivity of TpCorC. In contrast, mutation of the nonconserved Asn91 had little effect on Na+ sensitivity (Fig. 5B).

In the N94A/T106C/C282A mutant, Mg2+ had a weaker effect on the formation of the cross-linking dimer (Fig. 5B). This may be because Asn94 is indirectly involved in the formation of the Mg2+ binding site by forming a hydrogen bond with Ser43, which is directly involved in Mg2+ binding (Fig. 5A). In addition, the fractions of the chemical cross-linking dimer were reduced in the Na+-free N91A/T106C/C282A mutant in the presence and absence of Mg2+ compared to those in the T106C/C282A mutant. The reason is unknown, and further characterizations would be necessary but are beyond the scope of our study.

Furthermore, the Na+-nonsensitive mutation at Asn94 and the corresponding mutation in Na+-driven CNNM4 reduced the Mg2+ transport activities of both constructs (Fig. 5, C to E). Overall, these results suggest that Asn94 is involved in Na+ sensitivity in TpCorC and is important in its Mg2+ transport activity.

Mg2+ binding site and disease-associated mutations of human CNNM2/4

To gain mechanistic insights and to interpret the structure-function relationship of the Mg2+ binding site in human CNNM proteins, we performed structure-based mutational analysis of human CNNM2 and CNNM4 by the Mg2+ export assay in HEK293 cells (Fig. 6). We chose to characterize CNNM2 and CNNM4 among the CNNM transporters for the following three reasons: First, all of the amino acid residues involved in the Mg2+ binding site of TpCorC (Ser43, Ser47, Asn90, Gly129, and Glu130) are strictly conserved in both human CNNM2 (Ser269, Ser273, Asn323, Gly356, and Glu357) and CNNM4 (Ser196, Ser200, Asn250, Gly283, and Glu284) (Fig. 6A). Second, both CNNM2 and CNNM4 also function as Mg2+ exporters (15, 28), as observed with TpCorC in our study. Last, on the basis of the sequence alignment, we noticed that many known disease-associated mutations in CNNM2 and CNNM4 are located at the Mg2+ binding site (Fig. 6A) (18, 24).

Fig. 6 Mg2+ binding sites of human CNNM2 and CNNM4.

(A) Sequence alignment of TpCorC and human CNNM2 and CNNM4 at the Mg2+ binding site. Purple and green circles indicate disease-associated mutation sites and Mg2+ binding sites, respectively. (B and C) Close-up view of the Mg2+ binding sites in human CNNM2 (B) and CNNM4 (C) based on the homology model. Amino acid residues involved with Mg2+ ions are shown in stick representation. Mg2+ ions are shown as spheres. Dashed lines indicate hydrogen bonds. (D) Cell surface expression of CNNM Mg2+ binding site mutants. (E and F) Mg2+ export assay. Bar graph: Relative fluorescence intensities after Mg2+ depletion (at 5 min, means ± SEM, n = 10).

On the basis of the sequence similarity between TpCorC and human CNNM2/4, we constructed homology models of the TM domain of CNNM2 and CNNM4 using our TpCorC TM domain structure. Because all amino acid residues in the Mg2+ binding site are strictly conserved, the Mg2+ binding pockets and their coordination to Mg2+ ions in the homology models are almost identical to those of TpCorC (Figs. 3A and 6, B and C).

We first generated alanine-substituted mutants of human CNNM2 at (Ser269, Ser273, Asn323, Gly356, and Glu357) and near (Gly322 and Pro360) the Mg2+ binding site (Fig. 6A). Pro360 is strictly conserved among the CNNM/CorC family proteins, and Gly322 is also highly conserved in the family (fig. S1). Among the seven mutants at or near the Mg2+ binding site, the S269A and N323A mutants were not properly expressed at the cell surface, and S273A was not expressed at all (Fig. 6D and fig. S6). While the expression of wild-type CNNM2 induced a clear decrease in the fluorescence intensity after the depletion of Mg2+, G356A and E357A, both Mg2+ binding site mutants that were properly expressed at the cell surface, produced little change in the fluorescence signal compared to the level of the control (Fig. 6E). The mutation at the strictly conserved Pro360 located near the Mg2+ binding site also abolished the Mg2+ export activity of CNNM2 (Fig. 6E), whereas G322A maintained Mg2+ export activity (Fig. 6E).

Because we were not able to measure the Mg2+ export activity of three of the Mg2+ binding site mutants of CNNM2 (S269A, S273A, and N323A) because of technical problems with cell surface expression, we then generated three mutants of CNNM4 at the corresponding positions (S196P, S200Y, and N250A). For S196P and S200Y, we chose Pro and Tyr residues, respectively, because S196P and S200Y are reportedly associated with Jalili syndrome (18). All three mutants were successfully expressed at the cell membrane surface and exhibited the loss of Mg2+ export activity (Fig. 6, D and F).

Overall, our results showed that these Mg2+binding site mutants (CNNM2: G356A and E357A; CNNM4: S196P, S200Y, and N250A) abolished Mg2+ export activity (Fig. 6, E and F), indicating that both CNNM2 and CNNM4 would likely recognize Mg2+ ions in essentially the same manner as TpCorC (Figs. 3A and 6, B and C). Since many disease-associated mutations of CNNM2 and CNNM4 are located in the Mg2+ binding site, these results indicate structure-function relationships between the Mg2+ binding site and disease-associated mutations of CNNM2 and CNNM4.

CBS domain structure and TM-CBS interface

We further determined a nearly complete structure of the TpCorC CBS domain, including the region between the TM and CBS domains (residues 184 to 352) (fig. S2C). Thus, the structures that we obtained (TM, residues 28 to 182; CBS, residues 184 to 352) almost completely covered the entire region of the TM and CBS domains (Fig. 7, A and B), allowing us not only to obtain the structure of the CBS domain but also to gain insight into the proximal region between the TM and CBS domains.

Fig. 7 CBS domain structure and potential TM-CBS interface.

(A) Cartoon representation of the TpCorC TM domain dimer viewed parallel to the membrane. (B) Cartoon representation of the TpCorC CBS domain dimer. (C) Model of the TpCorC TM-CBS structure viewed parallel to the membrane. The coloring scheme of one subunit is the same as in fig. S1. The other subunit is colored gray. (D) Chemical cross-linking experiments on the cross-linking mutant (K60C/E190C) at the TM-CBS domain interface. (E) FSEC profiles of the TpCorC cross-linking mutant K60C/E190C on a Superdex 200 Increase 10/300 GL column (GE Healthcare) (excitation, 280 nm; emission, 325 nm). TCEP, tris(2-carboxyethyl)phosphine. (F) Structural deviation from the TpCorC TM-CBS domain structure model during a 300-ns MD simulation. (G) Cα distances between Lys60 (chain A) and Glu190 (chain B) during a 300-ns MD simulation.

The linker loop region between the TM and CBS domains consists of 11 amino acid residues (Gly178-Ser188), and with this loop length constrained, the loop connection between the C-terminal end of the TM domain structure and the N-terminal end of the CBS domain structure yielded a nearly unique TM-CBS interface, as shown in Fig. 7C. In this model, CH3 of the CBS domain interacts with CH1 of the TM domain in the neighboring subunit (Fig. 7C).

To verify this model, we generated the double cysteine mutant of the TpCorC Cys-less mutant (C282A), having cysteine substitutions at Lys60 in CH1 and Glu190 in CH3, where the Cα distance between Lys60 in one subunit and Glu190 in the other subunit is 9.4 Å. Consistent with our model, the K60C/E190C mutant of TpCorC showed a strong band for the TpCorC dimer in the presence of Cu2+ phenanthroline (Fig. 7D). FSEC analysis showed that the Cys-less mutant and the K60C/E190C double mutant treated with Cu2+ phenanthroline eluted in similar positions, indicating that Cu2+ phenanthroline–dependent dimer formation likely occurs within two protomers in a single TpCorC dimer and is not caused by nonspecific disulfide bond formation between the two adjacent TpCorC dimers (Fig. 7E). Furthermore, MD simulations of the TM-CBS model showed that the overall structure and TM-CBS interface (plotted by Cα distances between Lys60 in one subunit and Glu190 in the other subunit) are mostly stable throughout the simulation (Fig. 7, F and G), further verifying our TM-CBS structure model.

Nevertheless, because we have tested only one disulfide pair in the chemical cross-linking experiments, the constraint is insufficient to verify our proposed model as the unique one. Thus, further biophysical and structural analyses will be required to draw a firm conclusion.

ATP-binding site in the CBS domain

To gain insights into the ATP binding and ATP-dependent modulation of TpCorC, we also determined the crystal structure of the TpCorC CBS domain in complex with ATP (Fig. 8A and fig. S2, D and E). ATP molecules are located at the interface between two tandem CBS repeats in each subunit, with their phosphate groups facing each other (Fig. 8A).

Fig. 8 ATP-binding site.

(A) Close-up views of the ATP-binding site in the TpCorC CBS domain. ATP and the residues involved in ATP binding are shown in stick representation. (B) Isothermal titration calorimetry (ITC) data on the TpCorC CBS domain and its mutants with ATP. The raw ITC data and profiles are shown. Measurements were repeated twice, and similar results were obtained. [WT, n = 0.19 ± 0.01, ΔΗ° = −37.9 ± 0.6 (kcal/mol), −TΔS° = 29.3 (kcal/mol), and ΔG° = −8.6 ± 0.6 (kcal/mol); T336I, n = 0.92 ± 0.01, ΔΗ° = −16.9 ± 0.3, −TΔS° = 10, and ΔG° = −6.9 ± 0.3; Y255A, n = 1.08 ± 0.01, ΔΗ° = −24.2 ± 0.3, −TΔS° = 16.6, and ΔG° = −7.6 ± 0.3]. (C) The ATP hydrolysis assay of TpCorC was performed with a time course of 0 to 90 min (means ± SEM, n = 6). (D) Cell surface expression of TpCorC ATP-binding site mutants. (E) Mg2+ export assay. Bar graph: Relative fluorescence intensities after Mg2+ depletion (at 5 min, means ± SEM; empty, n = 20; WT, n = 17; and Y255A/T336I, n = 20). (F) Superposition of the ATP-bound CBS domain structure (cyan) onto the apo structure (red). Both subunits in the dimer are shown, and the neighboring subunit of the dimer is colored gray. Both the overall structure and close-up view of the region near the exterior of the ATP-binding site are shown. Black arrows indicate the structural changes between two conformations. OD, optical density.

The adenine base of ATP is recognized by three hydrogen bonds with the main chain amino and carbonyl groups of Val235 and the main chain carbonyl group of Arg257 and by a stacking interaction with Tyr255 (Fig. 8A). The ribose group forms two hydrogen bonds with the side chain of Asp339 (Fig. 8A), whereas the phosphate groups of ATP interact with the side chain and main chain amino group of Ser256 and the side chains of Arg257 and Thr336 (Fig. 8A). Furthermore, the conformation of the phosphate groups is seemingly stabilized by Mg2+ ions, which are further bridged by Glu338 via a water molecule (Fig. 8A). Notably, most of these residues at the ATP-binding site are highly conserved among the CNNM/CorC family proteins (fig. S1).

To test the ATP-binding mechanism, we generated three ATP-binding site mutants of TpCorC, Y255A (adenine ring), T336I (ribose), and Y255A/T336I and performed an ATP-binding assay using isothermal titration calorimetry (ITC). The disease-associated CNNM2 mutant T568I, corresponding to the T336I mutant of TpCorC, reportedly lost both ATP-binding activity and Mg2+ export activity (17, 28). The Y255A mutant and T336I mutant of the TpCorC CBS domain exhibited dissociation constant (Kd) values of 2.8 and 9.4 μM for ATP, whereas the wild type showed a Kd of 0.46 μM for ATP, indicating that these mutations weakened the affinity of ATP for TpCorC (Fig. 8B). The Y255A/T336I double mutant lost ATP-binding activity (Fig. 8B). Notably, in addition to ATP binding, we performed an ATP hydrolysis assay with TpCorC (Fig. 8C) and did not detect hydrolysis activity compared to apyrase, an ATP diphosphohydrolase (Fig. 8C).

We then performed an Mg2+ export assay in HEK293 cells with the Y255A/T366I mutant, which lacks ATP-binding ability (Fig. 8, D and E). Consistent with the loss of affinity for ATP, the Y255A/T336I double mutant showed low Mg2+ export activity (Fig. 8, D and E). Given the high affinity of the TpCorC CBS domain for ATP (~500 nM) and the typical concentration of cytoplasmic ATP at millimolar level (47), ATP may constitutively bind to CorC as a regulatory cofactor in vivo rather than functioning in the binding and release cycles. Overall, our results not only verify the ATP-binding mechanism of CorC but also show that ATP binding to the CBS domain is important for Mg2+ transport by CorC.

DISCUSSION

In this work, we determined the crystal structure of the TpCorC TM domain in complex with Mg2+ ions (Fig. 2), unveiling the novel fold of the DUF21 TM domain conserved in the CNNM/CorC family. Notably, Mg2+ ions are fully dehydrated by the surrounding amino acid residues in the structure (Fig. 3), and these residues are important for the Mg2+ efflux activities of TpCorC and human CNNM2 and CNNM4 (Figs. 3 and 6). This Mg2+ recognition contrasts with that of the known Mg2+ channel structures, such as the MgtE and CorA Mg2+ channels (48, 49). For instance, in the MgtE structure, Mg2+ ions are fully hydrated with all six water molecules in the first hydration shell (49). The dehydration energy for Mg2+ ions is much higher than that of monovalent cations and other biological divalent cations (50), and ion channels typically accomplish ion transport much faster than secondary active transporters. Therefore, this difference in Mg2+ recognition may account for the differences in transport kinetics between ion channels (MgtE and CorA) and transporters (CNNM/CorC).

Another unique feature in the TpCorC TM domain structure is the amphipathic belt helix (Fig. 2A and fig. S4). Since the belt helix is directly linked to the following regulatory CBS domain, it is tempting to speculate that the belt helix in the CorC TM domain might play an important role in transport regulation. Furthermore, the belt helix interacts closely with all three TM helices and may therefore affect the conformation of the TM domain (Fig. 2, B and C). Supporting our notion, mutation at the conserved Gly178 of TpCorC in the belt helix reduced Mg2+ efflux activity (figs. S1 and S4, C to E), whereas the P392A mutant of CNNM2, targeting its belt helix and corresponding to the P161A mutant of TpCorC, exhibited faster Mg2+ efflux than the wild type (figs. S1 and S4, C, D, F, and G).

Here, we describe our speculation on the transport mechanism of CorC, including how Mg2+ enters the binding pocket for dehydration and how it is released upon Na+ ion binding (fig. S7). First, the Mg2+ binding site is essentially solvent accessible in the current inward-facing conformation, but the bound Mg2+ ion is embedded in the binding pocket (Fig. 3A). The highly conserved Gly and Pro residues are located near or among the residues for Mg2+ binding (Gly89 in TM2 and Gly129 and Pro133 in TM3) (fig. S1), which may provide flexibility for the structural changes of the protein associated with Mg2+ binding in the transport cycle. Consistent with this notion, the 300-ns MD simulations of the TpCorC TM domain dimer with and without a bound Mg2+ ion showed local structural differences at the Mg2+ binding site, including the conserved residue Gly129 (fig. S7A). Whereas the overall conformation was essentially consistent (Fig. 2, F and G), we might see further structural changes by running the simulations for a much longer duration. In the absence of Mg2+, the side chains of Ser47 and Glu130 and the main chain carbonyl group of Gly129 face out of the Mg2+ binding pocket (fig. S7A). Alternatively, the side chains of Ser47 and Glu130 are ligated to the side chain of Lys134, which is conserved among CorC proteins (fig. S1). In the Mg2+-free inward-facing state, these local structural differences may make the binding pocket more solvent accessible (fig. S7A). In addition, since Glu130 might face outward from the binding pocket in the Mg2+-free inward-facing state (fig. S7A), Glu130 would be positioned to interact more easily with Mg2+ ion in the cytoplasm (fig. S7A). Thus, it is tempting to speculate that in the Mg2+ binding process, Glu130 may first interact with the fully hydrated Mg2+ and then partially dehydrate the Mg2+ ion (fig. S7B). Such interaction may cancel the interaction between Glu130 and the positively charged Lys134, and the following conformational change of Glu130 might push Mg2+ deeper into the Mg2+ binding pocket, facilitating further dehydration by the other residues by mimicking the hexacoordination of Mg2+ (fig. S7B). In this way, CorC might facilitate the dehydration of Mg2+ (fig. S7, A and B), whereas the dehydration of Mg2+ is usually energetically unfavorable and very slow, as we mentioned above. On the other hand, the details of the structural changes from the inward-facing conformation to the outward-facing conformation are not yet clear, but the chemical cross-linking experiment suggested that the addition of Na+ seems to destabilize the inward-facing state (Fig. 4D). Therefore, it is possible that the addition of Na+ may shift the structural equilibrium toward an outward-facing state (fig. S7, C to E), and such a structural change may promote the release of Mg2+ ion to the periplasmic side (fig. S7E). However, further structural and functional analyses are required for future work to fully understand the transport cycle.

Our results show that ATP binding to the CorC CBS domain is important for its Mg2+ transport activity (Fig. 8). A comparison of the apo and ATP-bound structures of the CBS domain suggests that in the ATP-bound structure, the helix region exterior of the ATP-binding site moves slightly outward from the pocket, mainly via its contact with the ribose moiety of ATP (Fig. 8F). Such structural changes in the CBS domain might be important for the regulation of transport activity through interaction with the TM-CBS interface region. Overall, our study presents a structural framework for understanding the Mg2+ recognition and regulation of the CNNM/CorC family, whose eukaryotic and bacterial physiological functions are implicated in various diseases.

METHODS

Expression and purification

The expression screening of CNNM/CorC homologs was performed by the GFP-based FSEC method (37). Alanine scanning was carried out by the FSEC-TS assay (38). The TM domain (residues 26 to 182) of the CorC gene from TpCorC and its mutants were synthesized (GENEWIZ, China); subcloned into a pET vector containing a human rhinovirus (HRV) 3C protease cleavage site, GFPuv, and an octahistidine tag at the C terminus; and transformed into the Escherichia coli Rosetta (DE3) strain. Cells were cultured in LB medium supplemented with ampicillin (50 μg/ml) at 37°C and then induced by 0.5 mM isopropyl-β-d-thiogalactopyranoside (IPTG) at an optical density at 600 nm (OD600) of 0.6. The cells were additionally cultured at 18°C for 16 hours, harvested by centrifugation (5000g, 15 min), and disrupted in buffer A [150 mM NaCl and 50 mM tris (pH 8.0) with 1 mM phenylmethanesulfonyl fluoride].

All purification steps were performed at 4°C. The cell lysate was centrifuged (20,000g, 30 min), and the membrane fractions in the supernatant were isolated by ultracentrifugation (180,000g, 1 hour). The membrane pellet was homogenized and solubilized in buffer A containing 1% n-dodecyl-β-d-maltopyranoside (DDM) (Anatrace, USA) and 0.2% cholesteryl hemisuccinate (CHS) (Sigma-Aldrich, USA). Then, the insoluble material was removed by ultracentrifugation (180,000g, 1 hour), and the supernatant was mixed with TALON resin (Takara, Japan). After binding for 2 hours, the resin was washed with buffer B [150 mM NaCl, 50 mM tris (pH 8.0), 0.05% DDM, and 0.01% CHS] containing 10 mM imidazole and eluted with buffer B containing 300 mM imidazole. The eluate was mixed with a His-tagged HRV 3C protease to cleave the GFPuv-octahistidine tag and dialyzed against buffer B overnight. The sample was reapplied to TALON resin equilibrated with buffer B, and the flowthrough fractions were concentrated using an Amicon Ultra 50K filter (Merck Millipore, USA) and applied to a Superdex 200 10/300 size exclusion chromatography column (GE Healthcare, USA) in buffer C [150 mM NaCl, 20 mM Hepes (pH 7.5), 0.05% DDM, and 0.01% CHS]. The peak fractions were concentrated to 10 mg/ml for crystallization.

The V101A mutant of the TpCorC TM domain was expressed and purified as described above. For cocrystallization of the TpCorC TM domain V101A mutant with CsCl, 150 mM NaCl in buffer C was replaced with 150 mM CsCl for size exclusion chromatography. To generate SeMet-substituted proteins, the TpCorC TM domain mutant of L38M/V101A/G115A/L173M was overexpressed in the E. coli B834 (DE3) strain, cultured in LeMaster medium supplemented with SeMet (50 mg/liter), and purified as described above.

All CBS domain constructs of TpCorC (residues 183 to 361, Y255A, and residues 202 to 361) were subcloned into the pET vector containing an HRV 3C protease cleavage site and an octahistidine tag and transformed into the E. coli Rosetta (DE3) strain. Cells were cultured in LB medium supplemented with ampicillin (50 μg/ml; residues 183 to 361, Y255A; residues 202 to 361) at 37°C and then induced by 0.5 mM IPTG at an OD600 of 0.6. The cells were further cultured at 37°C for 3 hours, harvested by centrifugation (5000g, 15 min), and disrupted in buffer A. The cell lysate was centrifuged (20,000g, 1 hour), and the supernatant was mixed with TALON resin (Takara, Japan). After binding for 1 hour, the resin was washed with buffer A containing 10 mM imidazole and eluted with buffer A containing 300 mM imidazole. The eluate was mixed with a His-tagged HRV 3C protease to cleave the octahistidine tag and dialyzed against buffer A overnight. The sample was reapplied to TALON resin equilibrated with buffer A, and the flowthrough fractions were concentrated using an Amicon Ultra 30K filter (Merck Millipore, USA) and applied to a Superdex 75 Increase 10/300 size exclusion chromatography column (GE Healthcare, USA) in buffer D [100 mM NaCl and 20 mM Hepes (pH 7.5)]. The peak fractions were concentrated to 10 mg/ml for crystallization. The CBS domain construct of TpCorC (residues 202 to 361) was mixed with ATP and MgCl2, at final concentrations of 1.5 and 10 mM, respectively, before crystallization.

Crystallization

Purified CorC TM domain proteins were mixed with MgCl2 at a final concentration of 50 mM, incubated for 30 min, and then mixed with monoolein (Nu-Chek, USA) at a ratio of 2:3 (w/w) with a coupled syringe mixer to generate LCP. The robotic LCP crystallization trial was performed by dispensing 50 nl of LCP drops onto 96-well sandwich plates and overlaid with 700 nl of reservoir solution using a Gryphon LCP crystallization robot (Art Robbins Instruments, USA). All LCP crystals were obtained in a solution containing 10 mM ZnCl2, 100 mM sodium acetate (pH 4.0), and 40% polyethylene glycol 200 (PEG 200) at 18°C. Crystals typically appeared in 3 days and grew to their maximum size within 1 week. Crystals were then harvested in reservoir solution supplemented with 40% PEG 200 and 50 mM MgCl2 and flash-frozen with liquid nitrogen for x-ray diffraction experiments.

For crystallization by the vapor diffusion method, 1 μl of the TpCorC CBS domain proteins (residues 183 to 361, Y255A, and residues 202 to 361) was mixed with 1 μl of reservoir solution [0.4 M ammonium thiocyanate, 0.1 M sodium acetate (pH 4.5), and 15% PEG 4000 and 0.1 M CaCl2, 0.1 M Hepes (pH 7.5), and 5% PEG 8000] and stored at 18°C. Before flash freezing in liquid nitrogen, crystals were harvested with the corresponding cryoprotectant solutions: for the crystals of the TpCorC CBS domain protein (residues 183 to 361, Y255A), 30% glycerol, 0.4 M ammonium thiocyanate, 0.1 M sodium acetate (pH 4.5), and 15% PEG 4000 and for the crystals of the TpCorC CBS domain protein (residues 202 to 361), 30% glycerol, 0.1 M CaCl2, 0.1 M Hepes (pH 7.5), and 5% PEG 8000.

Data collection and structure determination

X-ray diffraction datasets were collected at the SPring-8 beamlines of BL32XU and BL41XU with the automated data collection system ZOO (51) and processed with KAMO (52) for automatic data processing using XDS (53). For the SeMet-substituted TM domain of TpCorC, selenium sites were identified from the SAD data with SHELXD (54, 55). The initial phase was calculated and improved with SHELXE (54, 55), followed by automatic model building. The model was then further manually rebuilt with Coot (56) and refined with PHENIX (57). The structures of the wild type and the V101A mutant of the TpCorC TM domain were determined by molecular replacement with Phaser using the initial model of the SeMet-substituted TM domain of TpCorC (58) and further refined through multiple rounds of manual model building by Coot (56) and refinement by PHENIX (57).

The structure of the TpCorC CBS domain constructs of TpCorC (residues 183 to 361, Y255A, and residues 202 to 361) were initially determined by molecular replacement with Phaser using the homology model from SWISS-MODEL (59). The structure was further manually rebuilt using Coot (56) and refined by PHENIX (57).

The Ramachandran plots were calculated with MolProbity (60). Data collection and refinement statistics are summarized in table S2. All figures showing structures were generated using PyMOL (https://pymol.org/).

Isothermal titration calorimetry

All measurements were performed using a MicroCal iTC200 (GE Healthcare, USA) at 25°C. The TpCorC CBS domain and its mutants were purified by a similar method to that described above, and buffer E [100 mM KCl, 5 mM MgCl2, and 20 mM Hepes (pH 7.5)] was used for size exclusion chromatography. The ATP solutions were prepared by adding ATP to buffer E for size exclusion chromatography at a final concentration of 0.5 mM. A total of 250 μl of TpCorC CBS domain proteins (50 to 100 μM) was applied to the thermally equilibrated ITC cell, and the ligand syringe was filled with 40 μl of ATP solution. The molar ratio of proteins to ligands was 1:10. The ligands were injected 20 times (0.5 μl for injection 1 and 2 μl for injections 2 to 20), with 120-s intervals between injections. The background data obtained from buffer E (for ATP) were subtracted before data analysis. The data were analyzed with Microcal Origin software. Measurements were performed at least twice, and similar results were obtained.

Biochemical cross-linking

The wild-type, Cys-less mutant C282A and its mutants with cysteine substitutions of TpCorC (residues 23 to 441) were purified by a similar method to that described above, except that buffer F [150 mM NaCl, 20 mM Hepes (pH 7.5), and 0.03% DDM] and buffer G [150 mM KCl, 20 mM Hepes (pH 7.5), and 0.03% DDM] were used for size exclusion chromatography. A total of 4.0 μl of 20 μM TpCorC protein was mixed with 0.5 μl of 2 mM EDTA, 100 mM MgCl2, or Milli-Q water at the respective concentrations and then incubated for 1 hour at 4°C. Then, 0.5 μl of the reaction solution (10 mM Cu2+ bis-1,10-phenanthroline in a 1:3 molar ratio), Milli-Q water, or 20 mM tris(2-carboxyethyl)phosphine solution was added, followed by incubation for 30 min at 4°C. The samples were analyzed by nonreducing SDS–polyacrylamide gel electrophoresis and FSEC. Experiments were performed at least twice, and similar results were obtained.

ATP hydrolysis activity assay

The ATP hydrolysis assay of TpCorC and its mutant was performed by the malachite green method (61). Briefly, malachite green dye solution was freshly prepared on the day of the experiment by mixing 0.045% malachite green, 4.2% ammonium molybdate, and 1% Triton X-100 at a volume ratio of 36:12:1. Then, 1 μM TpCorC or 1 nM apyrase (Sigma-Aldrich, USA) was prepared in buffer H [20 mM Hepes (pH 7.5), 150 mM NaCl, and 0.05% DDM]. ATP was dissolved in buffer H containing 4 mM MgCl2 at a final concentration of 4 mM. To initiate the reaction, the ATP solution was added to the protein solution at an equal volume to obtain a final reaction mixture consisting of 20 mM Hepes (pH 7.5), 2 mM MgCl2, 150 mM NaCl, 0.05% DDM, and 2 mM ATP. The samples were then incubated at room temperature, and aliquots were taken at multiple time points (5, 30, 60, 90, and 120 min). Fifty microliters of each aliquot taken at each time point was mixed with 850 μl of the malachite green dye solution and 100 μl of 34% citric acid, and the absorbance was measured at 660 nm. The absorbance standard curve for inorganic phosphate was established with standard H3PO4 solutions.

MD simulations of the TpCorC TM and TM-CBS domains

MD simulations were carried out using Desmond (62). The CorC TM domain structure and TM-CBS structure model were embedded into a pre-equilibrated 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine lipid bilayer membrane. The protein/lipid system was placed at the center of a 10 × 10 × 10 box and then solvated with the simple point charge water model. The system was neutralized by adding sodium ions as counterions. NaCl (150 mM) was added to mimic the physiological conditions. The OPLS-2005 force field was used (63, 64). All simulations were carried out for 300 ns under an NPT (isothermal–isobaric) ensemble using a K40c Nvidia graphics processing unit, with the temperature kept at 300 K using a Nose-Hoover chain thermostat and the pressure at 1 atm using a Martyna-Tobias-Klein barostat. The long-range electrostatic interactions were calculated using the particle mesh Ewald method. The Coulomb interactions were analyzed with a cutoff of 9.0 Å. The reversible reference system propagator algorithm integrator was used with a time step of 2 fs, and coordinate trajectories were saved every 200 ps during the sampling process. The trajectory analysis was carried out using the Simulation Event Analysis and the Simulation Interactions Diagram modules of Desmond.

Expression constructs for Mg2+ export experiments

The cDNAs for human CNNM4 and mouse CNNM2 were generated in previous studies (15). The cDNAs for TpCorC containing the membrane targeting sequence from human CNNM4 (residues 1 to 178) were synthesized (GENEWIZ, China). Amino acid–substituted mutants were generated by GENEWIZ. Each cDNA was inserted into pCMV tag 4A (Agilent) for expression in HEK293 cells.

Cell culture and transfection

HEK293 cells were cultured in Dulbecco’s modified Eagle’s medium (Nissui, Japan) supplemented with 10% fetal bovine serum and antibiotics. Expression plasmids were transfected using Lipofectamine 2000 (Invitrogen, USA) according to the manufacturer’s instructions. For the analyses of CNNM2 and CNNM4, cells were directly subjected to immunofluorescence microscopy or the Mg2+ export assay. For the analyses of TpCorC, HEK293-derived cell lines stably expressing TpCorC mutants were established by selecting the transfected cells with growth medium supplemented with G418 (800 ng/ml) and 40 mM Mg2+ (Mg2+ was added to avoid potential decreases in intracellular Mg2+ level and subsequent growth arrest by the expressed proteins).

Mg2+ export assay

CNNM2- or CNNM4-transfected HEK293 cells or HEK293-derived cell lines stably expressing TpCorC were cultured in growth medium supplemented with 40 mM Mg2+ until use. The expression of each construct of TpCorC and CNNM proteins was confirmed by Western blotting (fig. S6). The Mg2+ export assay with Magnesium Green was performed as described in a previous study (15). The data are shown as either line graphs indicating the time course of mean relative fluorescent intensities or bar graphs of mean relative fluorescent intensities after Mg2+ depletion (at 5 min). Relative fluorescent intensities were calculated as the ratio of fluorescent intensity at a desired time point to that at time zero. After imaging analyses, cells were fixed with phosphate-buffered saline containing 3.7% formaldehyde and subjected to immunofluorescence microscopy to confirm protein expression. Representative line graphs are shown in the respective figures or in fig. S8.

Statistical analyses

Statistical analyses were performed using GraphPad Prism 6 software (GraphPad Software), and data are presented as means ± SEM. The P values were obtained by one-way analysis of variance (ANOVA) with Holm-Sidak post hoc tests.

Homology models of the TM domain of CNNM2 and CNNM4

The homology models of the TM domain of CNNM2 and CNNM4 were constructed by MODELLER using the TpCorC TM domain structure as a template (65).

SUPPLEMENTARY MATERIALS

Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/7/7/eabe6140/DC1

https://creativecommons.org/licenses/by-nc/4.0/

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.

REFERENCES AND NOTES

Acknowledgments: We thank F. Cai, Z. Zhang, and K. Hirata for technical support and C.-H. Lee and H. Kato for critical comments on the manuscript. We also thank the staff from BL32XU and BL41XU beamlines at SPring-8 and from BL17U1 at Shanghai Synchrotron Radiation Facility (SSRF) for assistance during data collection. The diffraction experiments were performed at SPring-8 BL32XU and BL41XU (proposal nos. 2017A2523, 2018A2507, and 2019A2514) and at SSRF BL17U1 (proposal no. 2018-SSRF-PT-004257). Funding: This work was supported by funding provided by the Ministry of Science and Technology of China (National Key R&D Program of China: 2016YFA0502800) to M.H. and by funding provided by the National Natural Science Foundation of China (31850410466 and 32071234), the Opening Project of State Key Laboratory of Genetic Engineering (SKLGE-1813), the Innovative Research Team of High-Level Local University in Shanghai, and a key laboratory program of the Education Commission of Shanghai Municipality (ZDSYS14005). This work was also supported by funding provided by the Japan Society for the Promotion of Science (JP26111007, JP17H04041, and JP20H03515) to H.M. and (JP20K07312, JP20H05508, and JP17K19396) Y.F. Author contributions: Y.H. and F.J. expressed and purified CorC and its mutants for the structural and functional studies and determined the structures with assistance from M.S., Y.Z., and M.H. Y.H. performed the ITC and biochemical cross-linking experiments. Y.F. performed the Mg2+ export assay. Z.X. and W.Z. performed the computational experiments at the early stage of the project. J.W. and Y.Y. performed MD simulations. Y.H., F.J., Y.F., H.M., and M.H. wrote the manuscript. H.M. and M.H. supervised the research. All authors discussed the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. The atomic coordinates and structural factors for the structures of the wild-type CorC TM domain, its V101A mutant, and the CorC CBS domain in the apo and ATP-bound forms have been deposited in the Protein Data Bank under the accession codes 7CFG, 7CFF, 7CFH, and 7CFI, respectively. Additional data related to this paper may be requested from the authors.

Stay Connected to Science Advances

Navigate This Article