SLC41A1 Is a Novel Mammalian Mg2+ Carrier*

The molecular biology of mammalian magnesium transporters and their interrelations in cellular magnesium homeostasis are largely unknown. Recently, the mouse SLC41A1 protein was suggested to be a candidate magnesium transporter with channel-like properties when overexpressed in Xenopus laevis oocytes. Here, we demonstrate that human SLC41A1 overexpressed in HEK293 cells forms protein complexes and locates to the plasma membrane without, however, giving rise to any detectable magnesium currents during whole cell patch clamp experiments. Nevertheless, in a strain of Salmonella enterica exhibiting disruption of all three distinct magnesium transport systems (CorA, MgtA, and MgtB), overexpression of human SLC41A1 functionally substitutes these transporters and restores the growth of the mutant bacteria at magnesium concentrations otherwise non-permissive for growth. Thus, we have identified human SLC41A1 as being a bona fide magnesium transporter. Most importantly, overexpressed SLC41A1 provide HEK293 cells with an increased magnesium efflux capacity. With outwardly directed Mg2+ gradients, a SLC41A1-dependent reduction of the free intracellular magnesium concentration accompanied by a significant net decrease of the total cellular magnesium concentration could be observed in such cells. SLC41A1 activity is temperature-sensitive but not sensitive to the only known magnesium channel blocker, cobalt(III) hexaammine. Taken together, these data functionally identify SLC41A1 as a mammalian carrier mediating magnesium efflux.

The molecular biology of mammalian magnesium transporters and their interrelations in cellular magnesium homeostasis are largely unknown. Recently, the mouse SLC41A1 protein was suggested to be a candidate magnesium transporter with channel-like properties when overexpressed in Xenopus laevis oocytes. Here, we demonstrate that human SLC41A1 overexpressed in HEK293 cells forms protein complexes and locates to the plasma membrane without, however, giving rise to any detectable magnesium currents during whole cell patch clamp experiments. Nevertheless, in a strain of Salmonella enterica exhibiting disruption of all three distinct magnesium transport systems (CorA, MgtA, and MgtB), overexpression of human SLC41A1 functionally substitutes these transporters and restores the growth of the mutant bacteria at magnesium concentrations otherwise non-permissive for growth. Thus, we have identified human SLC41A1 as being a bona fide magnesium transporter. Most importantly, overexpressed SLC41A1 provide HEK293 cells with an increased magnesium efflux capacity. With outwardly directed Mg 2؉ gradients, a SLC41A1-dependent reduction of the free intracellular magnesium concentration accompanied by a significant net decrease of the total cellular magnesium concentration could be observed in such cells. SLC41A1 activity is temperature-sensitive but not sensitive to the only known magnesium channel blocker, cobalt(III) hexaammine. Taken together, these data functionally identify SLC41A1 as a mammalian carrier mediating magnesium efflux.
Intracellular magnesium, especially its ionized fraction (Mg 2ϩ ), plays a critical role in enzyme activation, making the ion essential for numerous metabolic processes (1). Mg 2ϩ is an important co-factor in a number of other physiological func-tions, including the synthesis of biomacromolecules, secretion of hormones, and modulation of ion channel activity (2,3). It is therefore not surprising that an abnormal Mg 2ϩ homeostasis is associated with several disease conditions, such as cardiovascular diseases, essential hypertension, diabetes mellitus, and metabolic syndrome (4 -6). However, a better understanding of cellular Mg 2ϩ transport mechanisms and regulation is needed to elucidate the exact role of Mg 2ϩ in these disease processes; at present, this is hampered by limited knowledge of the molecular fundament of the mammalian Mg 2ϩ transport network. Despite extensive evidence for the existence of various regulated Mg 2ϩ transport proteins (7)(8)(9)(10), only two plasma-membrane localized proteins have been identified at the molecular level, namely, TRPM6 and TRPM7, which are ion channels of the melastatin-related transient receptor potential family, and MRS2, a channel located in the inner mitochondrial membrane (11)(12)(13). Thus, the recent description of novel putative Mg 2ϩ transporters, such as the A1 and A2 members of the solute carrier family 41 (SLC41) (14 -17), the ancient conserved domain protein subtype 2 (18,19), a protein termed magnesium transporter 1 (MagT1) (20) and the protein NIPA1 (21), have significantly expanded the field of research into cellular Mg 2ϩ transport systems.
The eukaryotic proteins SLC41A2 and SLC41A3, together with the protein SLC41A1, form a novel and unique family among the SLC superfamily, which contains 44 families of proteins involved in the transport of various inorganic and organic solutes (Ref. 22; HUGO data base). SLC41A1 was first identified and bioinformatically described by Wabakken et al. (14). Human SLC41A1 (hSLC41A1) has been mapped to chromosome 1q31-32 and encodes a protein consisting of 513 amino acids with a predicted molecular mass of 56 kDa (14). In humans and mice, the 5-kb long SLC41A1 transcripts have been found in most tissues (notably in heart, muscle, testis, thyroid gland, and kidney) (14,15). Homologues of the hSLC41A1 have also been identified in worms and insects.
A role of SLC41A1 in Mg 2ϩ cellular transport suggests itself because of its partial sequence homology to the bacterial Mg 2ϩ transporter MgtE (14,23,24). Experiments show that feeding mice on a low Mg 2ϩ diet causes increased expression of SLC41A1 in the kidney, colon, and heart (15). Moreover, analysis of published sequences has predicted SLC41A1 to be an integral cell membrane protein possessing 10 transmembrane domains. However, the only direct experimental evidence for SLC41A1 being an Mg 2ϩ transporter has been reported by Goytain and Quamme (15). By using a two-electrode-voltage clamp (TEV), 6 the authors suggest that heterologous expression of mouse SLC41A1 (mSLC41A1) in Xenopus laevis oocytes induces large inward currents carried by Mg 2ϩ .
In this study, we have identified SLC41A1 as an eukaryotic Mg 2ϩ carrier with the ability to form protein complexes. We show that SLC41A1 mediates a slow temperature-sensitive transport of Mg 2ϩ and, importantly, that it is able to substitute genetically distant bacterial Mg 2ϩ transporters CorA, MgtA, and MgtB at a functional level in Salmonella. Overall, our data suggest that SLC41A1 is an Mg 2ϩ carrier playing a significant role in transmembrane Mg 2ϩ transport and, by extrapolation, in cellular Mg 2ϩ homeostasis.
LB medium containing 10 mmol liter Ϫ1 MgCl 2 was used to culture the MM281 strain. The solid and liquid N-minimal media for complementation tests were prepared according to Nelson and Kennedy (26), except that 0.5 mmol liter Ϫ1 Na 2 SO 4 was used instead of 0.5 mmol liter Ϫ1 K 2 SO 4 . In addition, the media were supplemented with 0.1% casamino acids (Difco BD) and thiamine (2 mg liter Ϫ1 , Sigma). Overnight cultures grown in LB medium (37°C, provided with Mg 2ϩ if necessary) were washed with 0.7% saline, adjusted to an A 600 of 0.1 and diluted as indicated in Fig. 4. Serial dilutions were spotted onto N-minimal medium plates containing 10 mmol liter Ϫ1 , 100 mol liter Ϫ1 , or 10 mol liter Ϫ1 MgCl 2 . Spotted bacteria were cultivated for 36 h. To establish growth curves, overnight cultures grown in LB medium were washed with 0.7% saline, adjusted to an A 600 of 0.1, and inoculated into liquid N-minimal media containing 10 mmol liter Ϫ1 , 100 mol liter Ϫ1 , or 10 mol liter Ϫ1 MgCl 2 .

Immunoprecipitation and Western Blot Analysis
Total proteins were extracted from 250 ml of the bacterial culture (ϪIPTG or ϩIPTG, as indicated) using trichloroacetic acid/acetone. Proteins of the membrane fraction were isolated using the ProteoExtract TM Partial Bacterial Proteome Extraction Kit (Calbiochem, La Jolla, CA). His-tagged hSLC41A1 was immunoprecipitated from the membrane protein fraction with a His 6 tag antibody (GenWay Biotech, San Diego, CA). Protein samples were separated by SDS-PAGE utilizing 12.5% polyacrylamide gels, blotted, and labeled with His 6 tag antibody and goat anti-mouse (GAM)-HRP (Molecular Probes, Eugene, OR) or GAM--HRP (SBA, Birmingham, AL) antibodies. Antibody binding was visualized using the Chemilmager TM 5500 (Alpha Innotech) or AGFA Cronex 5 medical x-ray films developed with the Curix 60 (AGFA).

Determination of Total Magnesium in Salmonella by ICP-Mass Spectroscopy (ICP-MS)
Cultures of strains MM1927, MM281, and MM281-pUC18-hSLC41A1, grown (24 h) in N-minimal medium supplemented with 2 or 10 mmol liter Ϫ1 Mg 2ϩ , were washed 3 times with 0.7% saline and diluted to a bacterial density of 3 ϫ 10 8 bacteria ml Ϫ1 . Diluted bacterial suspensions (1 ml each) were centrifuged. Dried bacterial pellets were resuspended in 0.3 ml of 1 N HNO 3 and 0.7 ml of 1-bromododecane (purum-purum, Roth Karlsruhe Germany). Samples were centrifuged and the upper water fractions were used to determine total magnesium content (ICP-MS ELAN 6100, PerkinElmer Life Sciences). The organic fractions were used to determine protein content.

Determination of Free Intracellular Mg 2؉ in Salmonella by mag-fura 2 FF-Spectrofluorometry
Experimental procedures and data analyses were conducted according to Froschauer et al. (27) except the mag-fura 2 AM loading facilitator Pluronic F-127 was used at a final concentration of 5 mol liter Ϫ1 and the mag-fura 2 AM loading period was 30 min. Measurements were performed with LS-55 spectrofluorometer, operated by FL WinLab software version 4.0 (both products of Perkin-Elmer) at 37°C, in 3-ml cuvettes containing bacterial suspension (2 ml, 3 ϫ 10 8 bacteria ml Ϫ1 ).

Growth Media and Culture Conditions
HEK293-(FLAG-SLC41A1)-Full-length hSLC41A1 cDNA was cloned into a modified version of the pCDNA4/TO vector (Invitrogen) with an N-terminal FLAG tag. The FLAG-hSLC41A1 cDNA in pCDNA4/T0 was electroporated into HEK293 cells previously transfected with the pCDNA6/TR 6  construct for Tet-repressor expression. Cells were placed under zeocin selection; zeocin-resistant clones were screened for tet-inducible expression of the FLAG-tagged hSLC41A1 protein.
The same samples were immunoprecipitated by M2 anti-FLAG (Sigma) or isotype control, resolved by 10% SDS-PAGE, and transferred to polyvinylidene difluoride membranes. The membrane was immunoblotted with M2 anti-FLAG (Sigma) and GAM--HRP (SBA, Birmingham, AL). Membranes were developed by enhanced chemical luminescence (ECL) (Amersham Biosciences).

Blue-native Polyacrylamide Gel Electrophoretic (BN-PAGE) Separation and Two-dimensional SDS-PAGE
Enriched native membrane proteins were isolated from ϩtet (15 h) HEK293-(SLC41A1) cells by use of the ProteoExtract TM M-PEK. Native protein samples were mixed with SDS and incubated for 10 min in a thermomixer at 37°C with moderate shaking before being separated on the BN-polyacrylamide gel gradient (4 Ͼ 12%) according to the protocol of Swamy et al. (28). Proteins forming complexes with SLC41A1 were resolved by two-dimensional 10% SDS-PAGE and stained with Silver Stain Plus (Bio-Rad). The two-dimensional gels running in parallel with those used for silver staining were blotted and immunodecorated with M2 anti-FLAG and GAM HRP-linked antibodies and FLAG-SLC41A1 was visualized by a Chemilmager TM 5500 (Alpha Innotech). Protein marker Native Mark TM was purchased from Invitrogen.

Confocal Microscopy
5 ϫ 10 5 HEK293-(FLAG-SLC41A1) cells were plated on 12-mm glass, gelatin (2%)-coated coverslips and cultured for 24 h. Thereafter, FLAG-hSLC-41A1 overexpression was induced with tetracycline (15 h). Then, labeling of ϩtet and Ϫtet cells with Alexa Fluor-594 wheat germ agglutinin (2 g ml Ϫ1 , 10 min at 4°C) purchased from Invitrogen was performed. After rinsing with phosphate-buffered saline (PBS), cells were fixed in 100% methanol (10 min at Ϫ20°C). All following steps were carried out at room temperature. Cells were rinsed with PBS, blocked for 1 h in PBS containing 0.5% fish skin gelatin (Sigma), and then rinsed with PBS containing 0.02% fish skin gelatin. Subsequently, they were incubated for 45 min each with the primary M2 anti-FLAG antibody (1 mg ml Ϫ1 ) and with the secondary GAM antibody (0.4 mg ml Ϫ1 , Invitrogen) labeled with Alexa Fluor-488. Processed samples were coated with 5 l of vectashield (Vector Laboratories, Burlingame, CA) and digital images were acquired using a confocal microscope Zeiss LSM 510 META (Zeiss Jena Germany). Colocalization correlation analysis was performed using the Zeiss LSM 510 Image Browser (Zeiss).

Determination of Free Intracellular Mg 2؉ in ؉tet and ؊tet HEK293-(SLC41A1) Cells by mag-fura 2 FF-Spectrofluorometry
The Ϫtet and ϩtet HEK293-(SLC41A1) cells were rinsed twice with ice-cold, completely divalent-free PBS, detached by use of Hytase (Perbio Science, Bonn, Germany), centrifuged, washed twice in PBS, and finally re-suspended in completely Ca 2ϩ -and Mg 2ϩ -free Hanks balanced solution (CMF-HBS, pH 7.4, PAN Biotech). Loading of cells with 7.5 mol liter Ϫ1 magfura 2 AM (Molecular Probes) was performed for 25 min at 37°C in the presence of pluronic acid. After being washed in CMF-HBS, cells were incubated for a further 30 min to allow for complete de-esterification of the fluorescence probe, washed twice in CMF-HBS to remove extracellular mag-fura 2, and stored in CMF-HBS complemented with 10 mmol liter Ϫ1 HEPES and 5 mmol liter Ϫ1 glucose (CMF-HBSϩ) until used for measurements of free intracellular [Mg 2ϩ ] ([Mg 2ϩ ] i ). Measurements were made at 37°C (or as indicated under "Results") in 3-ml cuvettes containing cell suspension (2 ml, CMF-HBSϩ with a cytocrit of 10%) under stirring after the cells had been washed twice in CMF-HBSϩ. In experiments with inside-directed Mg 2ϩ gradients, MgCl 2 was added to give final concentrations of 2, 5, or 10 mmol liter Ϫ1 (30 to 40 s prior to start of the measurements). In control measurements, no Mg 2ϩ was added but, instead, 2, 5, or 10 mmol liter Ϫ1 Ca 2ϩ was present in the measuring solution. [Mg 2ϩ ] i was determined by measuring the fluorescence of the probe-loaded cells in a spectrofluorometer (LS50-B, PerkinElmer Life Sciences) by using the fast filter accessory, which allowed fluorescence to be measured at 20-ms intervals with excitation at 340 and 380 nm, and emission at 515 nm. [Mg 2ϩ ] i values were calculated from the 340/380-nm ratio according to the formula of Grynkiewicz et al. (29) using FL WinLab version 4.0 (PerkinElmer Life Sciences). A dissociation constant of 1.5 mmol liter Ϫ1 for the mag-fura 2-Mg 2ϩ complex was used for calculations; minimum (R min ) and maximum (R max ) ratios were determined at the end of each experiment by using digitonin. R max was found by the addition of 25 mmol liter Ϫ1 MgCl 2 in the absence of Ca 2ϩ , whereas R min was obtained by addition of 50 mmol liter Ϫ1 EDTA, pH 7.2, to remove all Mg 2ϩ from the solution. For data evaluation, 10-s data sets each were averaged at the beginning of the measurement and then always after 50 s. The final [Mg 2ϩ ] i was determined as the mean [Mg 2ϩ ] i of the last 10 s of the measurement. Thus, for the calculation of any given [Mg 2ϩ ] i , 500 data points were used. If not otherwise stated, data are presented as mean Ϯ S.E.

Determination of Free Intracellular Ca 2؉ in ؉tet and ؊tet HEK293-(SLC41A1) Cells by fura 2 FF-Spectrofluorometry
The general procedure was the same as that described for the determination of [Mg 2ϩ ] i with the following exceptions. Cells were loaded with 10 mol liter Ϫ1 fura 2 AM. The R max for fura 2 was obtained in solutions with 2 mM Ca 2ϩ and the R min by the addition of 20 mmol liter Ϫ1 EGTA, pH 8.0; a dissociation constant of 224 nmol liter Ϫ1 was used for the fura 2-Ca 2ϩ complex.

Determination of the Total Mg in ؊tet and ؉tet (15 h) HEK293-(FLAG-SLC41A1) by Atomic Mass Spectroscopy
The Ϫtet and ϩtet (15 h) HEK293-(FLAG-SLC41A1) cells were grown to ϳ80% confluence, washed twice with serumfree, Mg 2ϩ /Ca 2ϩ -free HEK293 experimental medium (PAN Biotech), detached by 0.25% trypsin-EDTA buffer, and resuspended in HEK293 medium to give a final cell count of 6 ϫ 10 6 cells ml Ϫ1 . The viability of the cells was determined using trypan blue exclusion. Diluted cells were held in the synthetic HEK293 medium for 60 min prior to the addition of Mg 2ϩ to give a final [Mg 2ϩ ] e of 10 mmol liter Ϫ1 . Subsequently, the cells were incubated in the presence of Mg 2ϩ at 37°C in 5% CO 2 atmosphere for 20 or 180 min. After incubation, they were washed three times with Mg 2ϩ /Ca 2ϩ -free PBS and dried pellets were mixed with 0.3 ml of 1 N HNO 3 and 0.7 ml of 1-bromododecane (purum-purum). Samples were centrifuged and the upper water fractions were used to determine total magnesium contents (flame AM Spectrometer M Series, Thermo Scientific). Protein contents were determined in the organic fractions.

Statistics
All statistical calculations were performed by using Sigma-Stat (Jandel Scientific). Significance was determined by Student's t test; p Ͻ 0.05 was considered to be significant.

Inhibitors
DIDS and cobalt(III) hexaammine (CoHex) were obtained from Sigma. H 2 DIDS was purchased from Molecular Probes.

RESULTS
To assess the basic molecular characteristics of SLC41A1 and its role in cellular Mg 2ϩ transport, we took advantage of the well established tetracycline-controlled expression system in the HEK293 cell line. Several zeocin-resistant clones were tested; clone 17 was selected for this study, because of the high level of overexpression and the lack of molecular leakiness ( Fig. 1, B and C).
Cell Topography of Recombinant FLAG-hSLC41A1-Computational analyses predicted SLC41A1 to be an integral cell membrane protein with 10 putative transmembrane domains and possibly both N and C termini located intracellularly ( Fig.  1A) (14, 15) (PSORT II and WOLF PSORT II Prediction). To test whether overexpressed FLAG-hSLC41A1 was targeted to the plasma membrane of the HEK293 cells, we designed several experiments comprising confocal immunolocalization and Western blot analysis of the membrane protein fraction isolated from non-induced (Ϫtet) and tet-induced (ϩtet) HEK293-(FLAG-hSLC41A1).
As shown in Fig. 1B, the recombinant FLAG-tagged SLC41A1 protein was specifically detected in the plasma membrane of ϩtet (15 h) HEK293-(FLAG-hSLC41A1) cells investigated by confocal microscopy. This was confirmed by colocalization of the green fluorescent signal of immunolabeled hSLC41A1 (M2 anti-FLAG: GAM Alexa 488) with the red fluorescent signal of wheat germ agglutinin conjugated to Alexa 594 (Fig. 1B). The latter is known to recognize sialic acid and N-acetylglucosaminyl sugar residues predominantly found on the plasma membrane. Colocalization correlation analysis revealed a 59.3 Ϯ 1.6% overlap of red and green pixels. In contrast, no FLAG-hSLC41A1-specific fluorescence was found in Ϫtet cells (Fig. 1B). Fig. 1C shows data obtained by Western blot analysis of membrane protein fractions and non-membrane protein fractions from Ϫtet and ϩtet (18 h) cells. The 56-kDa band corresponding to FLAG-hSLC41A1 was predominantly detected in the membrane fraction with lower abundance in the non-membrane fraction. Western blot analysis of immunoprecipitated FLAG-hSLC41A1 from membrane and non-membrane protein fraction lysates revealed the same results (Fig.  1C). FLAG-hSLC41A1-specific band was not detected in samples prepared from Ϫtet cells. Taken together, these data demonstrate the plasma membrane localization of FLAG-hSLC41A1 when overexpressed in HEK293 cells (Fig. 1A).
Complex Forming Ability of hSLC41A1-Various solute transporters have been shown to form stable or transient protein complexes, which are necessary for them to be functional (31,32). To test whether hSLC41A1 formed such complexes with other proteins, we performed BN-PAGE with native proteins isolated from ϩtet (15 h) HEK293-(SLC41A1) cells. FLAG-hSLC41A1-containing complexes were immunodetected with M2 anti-FLAG and goat anti-mouse HRP-linked antibodies. We identified two complexes (C1 and C2; Fig. 2A) with molecular masses lying between 720 and 1236 kDa (720 kDa Ͻ C1, C2 Ͻ 1236 kDa). Next, the hSLC41A1 complexes were gradually degraded by adding SDS in a stepwise manner to give concentrations from 0.05 to 1%. Upon addition of 0.1% SDS, we were able to detect the break-down products of C1 and C2: 480 kDa Ͻ C3 Ͻ 720 kDa; 242 kDa Ͻ C4 Ͻ 480 kDa and M ϳ56 kDa, the latter corresponding to the molecular mass of the SLC41A1 monomer ( Fig. 2A). A successive increase of SDS strengthened the signal of C4 and M and, as expected, weakened the signal of C1 and C2. The two-dimensional SDS-PAGE separation of the C1 and C2 complexes followed by silver staining revealed heterogeneous compositions of C1 and C2 complexes (data not shown). Because of the limited resolution of the high molecular mass protein complexes (750 kDa Ͻ Ͻ Cx) in the first native dimension, the presence of SLC41A1 in C1 and C2 complexes was confirmed by SLC41A1 immunodecoration after two-dimensional SDS-PAGE (Fig. 2B).
Effect of hSLC41A1 Overexpression on Growth and Mg 2ϩ Content of Mg 2ϩ -deficient Salmonella Strain MM281-The hSLC41A1 gene shares sequence similarity with the bacterial gene mgtE (14,15). Gene mgtE has been identified in various bacteria (23,24), but not in Salmonella sp. Based on its ability to restore growth of the Mg 2ϩ -deficient strain MM281 of S. enterica, Smith and colleagues (24) have proposed the direct involvement of MgtE in Mg 2ϩ transport. Strain MM281 exhibits disruption of genes corA, mgtA, and mgtB, the three major Mg 2ϩ influx systems of Salmonella. Compared with normal strains that can grow at [Mg 2ϩ ] e of 10 -100 mol liter Ϫ1 , this

SLC41A1, A Novel Mg 2؉ Carrier
strain requires [Mg 2ϩ ] e from 10 to 100 mmol liter Ϫ1 to proliferate (24,28). We tested the ability of hSLC41A1 to complement the Mg 2ϩ -dependent growth-deficient phenotype of strain MM281 by transforming it with plasmids pUC18-hSLC41A1 or pUC18-(empty).
The expression of His-hSLC41A1 after addition of IPTG (0.02 to 0.05 mmol liter Ϫ1 ) was confirmed by Western blot analysis of the total protein isolate as well as of the immunoprecipitated His-hSLC41A1 from the bacterial membrane protein fraction (Fig. 3). Growth curves were established within 24 h for strains MM281-pUC18-(empty), MM281-pUC18-hSLC41A1, and MM1927 in media containing 10 mol liter Ϫ1 , 100 mol liter Ϫ1 , or 10 mmol liter Ϫ1 Mg 2ϩ . The growth maxima of strains MM1927 and MM281-pUC18-hSLC41A1 were almost identical at [Mg 2ϩ ] e of 10 mmol liter Ϫ1 , whereas the growth maximum of strain MM281-pUC18-(empty) was ϳ33% lower in comparison with the growth maximum of strain MM1927 (Fig. 4A). The growth maximum of strain MM281-pUC18-hSLC41A1 reached 43% of the growth maximum of strain MM1927 when cultivated at an [Mg 2ϩ ] e of 100 mol liter Ϫ1 (Fig. 4B) and 32.5% when cultivated at an [Mg 2ϩ ] e of 10 mol liter Ϫ1 (Fig. 4C). Strain MM281-pUC18-(empty) did not grow in media supplemented with an [Mg 2ϩ ] e of 10 or 100 mol liter Ϫ1 . As shown in Fig. 4, images of the plated serial dilutions obtained after 24 h of incubation at 37°C clearly corresponded to the respective sets of the growth curves.
Furthermore, we measured the [Mg 2ϩ ] i of bacteria from strains MM1927, MM281-pUC18-(empty), and MM281-pUC18-hSLC41A1 by using mag-fura 2 fast filter spectroscopy (27). Mg 2ϩ -starved bacteria were incubated in 0.9% saline containing 0 or 10 mmol liter Ϫ1 Mg 2ϩ and the [Mg 2ϩ ] i was determined over 20 min. The results are summarized in Fig. 4D The mag-fura 2 data are in agreement with our results obtained by using ICP-MS. With this technique, the relative   Patch Clamp Characterization of hSLC41A1-Using TEV, Goytain and Quamme (15) observed large Mg 2ϩ currents associated with mouse SLC41A1 when overexpressed in X. laevis oocytes. Therefore, we expected Mg 2ϩ carried currents to appear after hSLC41A1 overexpression in HEK293 cells. To characterize such currents, patch clamp experiments in the whole cell configuration with ϩtet (15-18 h) and non-induced HEK293-(SLC41A1) cells were performed. Repetitive voltage ramps that spanned Ϫ100 to ϩ100 mV over 50 ms were delivered every 2 s from a holding potential of 0 mV. Inward currents were assessed at Ϫ80 mV and outward currents at ϩ80 mV. An inwardly directed Mg 2ϩ concentration gradient was created by perfusion of cells with Mg 2ϩ -free internal saline (K ϩ -Glubased, if not stated otherwise), whereas the external solution contained 2 mmol liter Ϫ1 Mg 2ϩ . Under these experimental conditions, development of a small but identifiable current at negative membrane potentials (Ϫ100 to 0 mV) would be predicted in SLC41A1 overexpressing cells that would not be seen in non-induced cells. This current would be expected to have a more positive reversal potential (E rev ) and would be carried by Mg 2ϩ . Instead, SLC41A1 overexpressing cells developed a large outwardly rectifying conductance (Fig. 5A). This current was fully activated within 200 s of the experiment and its currentvoltage (I-V) relationship (Fig. 5B) revealed a highly nonlinear current with a reversal potential of around Ϫ35 mV. The development of the SLC41A1-induced current could be prevented in the presence of 1 mmol liter Ϫ1 intracellular Mg 2ϩ (Fig. 5, C and D).
To test whether the SLC41A1-induced conductance could support Mg 2ϩ influx, cells were initially bathed in the standard external solution containing 1 mmol liter Ϫ1 Ca 2ϩ and 2 mmol liter Ϫ1 Mg 2ϩ . At 200 s, when the SLC41A1-induced conductance had reached its full amplitude, an isotonic solution of 115 mmol liter Ϫ1 Mg 2ϩ was applied for 60 s via a buffer pipette (Fig.  5E). This had no significant effect on either inward or outward currents, and the shape of the I-V relationship extracted at the end of the application was also not affected compared with the control (data not shown). In conclusion these unexpected results clearly show that the SLC41A1-induced conductance did not give rise to an Mg 2ϩ influx but exhibited typical characteristics of a chloride conductance.
Therefore, further experiments were set out to confirm the latter. To this end, we allowed the current to develop fully before applying an external solution supplemented with 100 mol liter Ϫ1 of the Cl Ϫ channel inhibitor DIDS. This resulted in a fast and almost complete block of the current (Fig. 5, F and  G). In control experiments with ϩtet (15-18 h) HEK293-(TRPM7) cells, the application of 100 mol liter Ϫ1 DIDS had no effect on TRPM7 current (data not shown). In the next set of experiments we used Mg 2ϩ -free KCl-based instead of K ϩ -Glu-based internal saline. Under these conditions we observed: 1) an inward current that could not be seen when K ϩ -Glu buffer was used for perfusion of ϩtet HEK293-(SLC41A1) cells (data not shown) and 2) a depolarizing shift of the E rev as predicted for Cl Ϫ conductance by the Nernst equation (data not shown). At 300 s, a low Cl Ϫ solution was applied for 100 s via a buffer pipette. As expected, this resulted in a strong reduction of the outward current during application, whereas the inward current remained the same (data not shown). The application of low Cl Ϫ solution also evoked a further depolarizing shift of the E rev (data not shown). These data in conjunction with the DIDS sensitivity of the current clearly confirm the involvement of Cl Ϫ channels.
Because some Cl Ϫ channels are activated by protein phosphorylation (33, 34) we wished to determine whether the SLC41A1-induced conductance would also be activated. To this end, we perfused both Ϫtet and ϩtet cells with a Mg 2ϩ -free intracellular solution supplemented with 1 mmol liter Ϫ1 ATP␥S, a non-hydrolysable substrate for ATPases. In Ϫtet cells, ATP␥S gave rise to a Cl Ϫ conductance that was identical to the conductance and I-V curves seen in ϩtet cells in the absence of this substrate (Fig. 5H). Moreover, ATP␥S did not cause recruitment of any additional currents in SLC41A1-overexpressing cells (Fig. 5, A versus H and I) and the ATP␥S-induced currents developed in an identical manner even in the complete absence of intracellular and extracellular Mg 2ϩ (data not shown). We wondered whether suppression of the SLC41A1-induced Cl Ϫ conductance would reveal any Mg 2ϩ influx that might have been masked by the large currents that develop in ϩtet cells. However, upon suppression of the Cl Ϫ currents by supplementing the extracellular solution with 100 mol liter Ϫ1 DIDS and superfusing the cells with an isotonic Mg 2ϩ solution, no further Mg 2ϩ influx could be detected (Fig. 5J).
It is known that two molecules of tetracycline can chelate one Mg 2ϩ (30). To exclude any tetracycline effects on our measurements, wild type (WT) HEK293 cells grown for 15 h in tetracycline-containing medium (1 g/ml) were perfused with Mg 2ϩfree internal saline and examined in whole cell mode patch clamp experiments. As predicted, no conductance similar to that measured in SLC41A1 overexpressing HEK293 cells was found in WT cells grown in ϩtet medium (data not shown).
Functional Characterization of hSLC41A1 in HEK293 Cells by Use of mag-fura 2-Because of the sequence homology of SLC41A1 to the bacterial Mg 2ϩ transporter MgtE, we wondered whether this protein might be involved in Mg 2ϩ transport functioning as a carrier protein rather than an ion channel mechanism. We therefore set out to measure intracellular Mg 2ϩ concentrations by using a mag-fura 2-based ratiometric assay. HEK293 cells bearing FLAG-tagged SLC41A1 were induced for 5, 10, or 15 h with tetracycline and, afterward, the The incubation of ϩtet HEK293-(SLC41A1) cells in completely Mg 2ϩ -free medium always led to a significant decrease of their [Mg 2ϩ ] i compared with that of Ϫtet HEK293-(SLC41A1) cells (Fig. 6, A and B, (Fig. 6B). Such a process was never seen in Ϫtet HEK293-(SLC41A1) cells or wild type HEK293 cells, which showed a negligible (56 Ϯ 7 mol liter Ϫ1 ) [Mg 2ϩ ] i increase under these conditions. These surprising results point to an increased efflux capacity of HEK293 cells overexpressing SLC41A1.
Compared with the zero-Mg 2ϩ conditions, higher [Mg 2ϩ ] i values were observed in both ϩtet and Ϫtet cells if they were incubated in Mg 2ϩ -containing medium (Fig. 6A and Table 1). However, from 10 h and more after induction, ϩtet HEK293-(SLC41A1) cells had a significantly higher [Mg 2ϩ ] i at the end of the measuring period than Ϫtet cells at all [Mg 2ϩ ] e used ( Table  1). In contrast, no [Mg 2ϩ ] i increase was observable in the presence of transmembrane Ca 2ϩ gradients favoring calcium influx (Fig. 6A) Table 1 showing that the [Mg 2ϩ ] i of Ϫtet or ϩtet wt HEK293 cells was not different from that measured in Ϫtet HEK293-(SLC41A1) cells.
Because the patch clamp data revealed an inhibition of the SLC41A1-related Cl Ϫ conductance in ϩtet HEK293-(SLC41A1) cells treated with 100 mol liter Ϫ1 DIDS (Fig. 5, F and G), we investigated whether this inhibitor also influenced their [Mg 2ϩ ] i . As shown in Fig. 6C (Fig. 6D). After correction for this linear component, a [Mg 2ϩ ] i elevation was still observable in ϩtet HEK293-(SLC41A1) cells (Fig. 6E). This remaining component was assumed to result mainly from SLC41A1 overexpression and its extent was dependent on [Mg 2ϩ ] e and on the duration of tet-induction (Fig. 6E) (Fig. 7A).
Next, we wished to determine whether the observed [Mg 2ϩ ] i changes were accompanied by net changes of [Mg] t (measured by atomic mass spectroscopy). In these experiments, all cells were prestarved in Mg 2ϩ -free medium for 60 min (this time was adequate for mag-fura 2 AM loading and activation in the fast filter spectroscopy measurements described above) and then incubated in the presence of 10 mmol liter Ϫ1 Mg 2ϩ over a time period of 20 min. As shown in Fig.  7B, the [Mg] t of Ϫtet HEK293-(SLC41A1) cells was not influenced by incubation in Mg 2ϩ -free or high-Mg 2ϩ (10 mmol liter Ϫ1 ) medium. However, when ϩtet (5 h) HEK293-(SLC41A1) cells were kept in Mg 2ϩfree medium, their [Mg] t decreased by 25.6% compared with that of Ϫtet cells incubated under the same conditions. Again, such results are only explainable by an increased SLC41A1-mediated Mg 2ϩ efflux from these cells. When Mg 2ϩ (10 mmol liter Ϫ1 ) was present during the 20-min incubation time, the [Mg] t of these cells increased by    (Fig. 8A). In Ϫtet cells incubated in 0 mmol liter Ϫ1 Mg 2ϩ medium, CoHex had no significant effect on the [Mg 2ϩ ] i but the inhibitor led to a significant 30% reduction of [Mg 2ϩ ] i in 10 mmol liter Ϫ1 Mg 2ϩ medium (Fig. 8A). In contrast, the SLC41A1-related [Mg 2ϩ ] i change was not influenced by CoHex (Fig. 8A). These data confirm the existence of a CoHex-blockable Mg 2ϩ influx mechanism(s) not identical to SLC41A1 in HEK293 cells. A likely candidate for such a transport mechanism is the TRPM7 ion channel, which is endogenously expressed in this cell type (36). To study the effect of CoHex on TRPM7 current development, we performed patch clamp experiments in the whole cell configuration mode with ϩtet (14 -20 h) HEK293-(TRPM7) cells (12). CoHex was applied 60 s after the start of the experiment when TRPM7 currents were fully developed. CoHex at 1 mmol liter Ϫ1 reversibly blocked inward TRPM7 currents (relevant to divalent cations, mainly Mg 2ϩ conductance) by 51.3 Ϯ 1.8%, whereas outward TRPM7 currents (relevant to monovalent ion con-  ductance) remained almost unaffected (12.3 Ϯ 0.6% inhibition) in the presence of 2 mmol liter Ϫ1 [Mg 2ϩ ] e (Fig. 8, B-E).

DISCUSSION
At present, our understanding of the molecular identity and cellular functions of SLC41A1 is limited. The sequential similarity between SLC41A1 and the putative bacterial Mg 2ϩ transporter MgtE (14) and the up-regulation of SLC41A1 expression in response to a low Mg 2ϩ diet (15) lead to the hypothesis that SLC41A1 is involved in Mg 2ϩ homeostasis and/or Mg 2ϩ transport in cells of higher eukaryotes. This hypothesis is supported by our data showing the functional substitution of CorA, MgtA, and MgtB Mg 2ϩ transporters by hSLC41A1 in the Salmonella strain MM281. Moreover, the results described here provide experimental evidence that SLC41A1, the first molecularly characterized Mg 2ϩ carrier in eukaryotes, probably mediates Mg 2ϩ efflux. The basis for this conclusion is 4-fold: 1) overexpression of SLC41A1 in HEK293 cells does not induce detectable Mg 2ϩ -carried currents, 2) in Mg 2ϩ -free media, SLC41A1 overexpression leads to a significant reduction of [Mg 2ϩ ] i and [Mg] t , 3) the intensity of the Mg 2ϩ loss depends on the induction time and thus on the number of SLC41A1 molecules in the cell membrane, and 4) SLC41A1-related [Mg 2ϩ ] i changes are temperature-sensitive but not influenced by the Mg 2ϩ channel blocker CoHex.
hSLC41A1 Functionally Complements Disruption of the CorA-MgtA-MgtB Transport System in S. enterica sv. typhymurium-The Mg 2ϩ -dependent growth-deficient Salmonella strain MM281 represents, with certain limitations, a simple model for testing the ability of the candidate Mg 2ϩ transporters to restore its growth and thus to identify the direct involvement of these transporters in Mg 2ϩ transport (24,37,38). SLC41A1 has only been identified in the genomes of eukaryotes (14,15), however, due to its distant sequential ancestry with the bacterial MgtE, we reasoned that it might be able to complement the growth-deficient phenotype of the MM281 strain. MgtE can mediate Mg 2ϩ uptake in bacteria but lacks homology to the other known bacterial Mg 2ϩ transporters as it does not possess the typical F/YGMN motif, which is characteristic for members of the CorA-Mrs2-Alr1 superfamily of Mg 2ϩ transporters (13). Nevertheless, as our data show, hSLC41A1, when overexpressed from pUC18-hSLC41A1 in the MM281 strain, partly restores the growth of this triple disruption of Salmonella in low Mg 2ϩ media. However, the growthpromoting effect of SLC41A1 is less than that of Mrs2 (13). The latter is present in the mitochondria of the eukaryotes and represents a distant homologue of the bacterial Mg 2ϩ channel CorA. Functional complementation by SLC41A1 corresponds well to our data obtained by ICP-MS demonstrating a significant increase of the total magnesium concentration in the MM281 strain transformed with pUC18-hSLC41A1 in comparison with the [Mg] t in the MM281 strain transformed with pUC18-(empty). The ability of hSLC41A1 to complement the Mg 2ϩ -linked growth-deficient phenotype of Salmonella strain MM281 identifies hSLC41A1 as being a bona fide Mg 2ϩ transporter.
hSLC41A1 Probably Forms Hetero-oligomeric Complexes in a Mammalian Expression System-Taking into account that hSLC41A1 maintains its functionality when expressed in Salmonella and that the Salmonella genome lacks mgtE, hSLC41A1 probably works as a monomer and/or a homo-oligomer in this expression system. However, various solute transporters have been shown to form stable or transient protein complexes to become functional in their native systems (31,32). This is in agreement with our findings establishing that SLC41A1 forms protein complexes of "high" molecular mass (ϳ1000 kDa) when overexpressed in HEK293 cells. In addition, our two-dimensional PAGE data indicate the presence of distinct proteins in the observed SLC41A1 complexes, further suggesting the hetero-oligomeric character of SLC41A1 complexes in the mammalian system. SLC41A2 and SLC41A3 are possible candidates for being binding partners in such complexes. This hypothesis is indirectly supported by our recent observation that all three genes are being overexpressed simultaneously in response to extracellular Mg 2ϩ starvation in lymphocytes. 7 Even so, protein(s) other than SLC41A2 or SLC41A3 (e.g. protein components of the cytoskeleton, other ion transporters, and/or enzymes) must be integrated in SLC41A1-containing complexes to reach the observed molecular masses between 720 and 1236 kDa. Future studies investigating SLC41A1-binding partners and the composition of the SLC41A1 complexes in response to specific physiological conditions will clarify this.
hSLC41A1 Overexpression Does Not Induce Measurable Mg 2ϩ Currents, but Allows Mg 2ϩ Efflux and Is Associated with an Endogenous Cl Ϫ Conductance-Overexpression of mSLC41A1 in X. laevis oocytes has been shown to induce large Mg 2ϩ -carried currents, although various other divalent cations are also transported (15). Using TEV, Goytain and Quamme (15) determined the following SLC41A1-specific permeation profile: Mg 2ϩ Ն Sr 2ϩ Ն Fe 2ϩ Ն Ba 2ϩ Ն Cu 2ϩ Ն Zn 2ϩ Ն Co 2ϩ Ͼ Cd 2ϩ . However, because of the lack of a control for the intracellular ion milieu, TEV does not allow the establishment of a true permeation profile. Nevertheless, these data suggest that SLC41A1 is an unspecific divalent cation channel. In contrast, the currents induced by SLC41A1 overexpression in our ϩtet HEK293-(SLC41A1) cells have been identified as endogenous Cl Ϫ currents, recruited by depletion of intracellular Mg 2ϩ and blockable by the broad-spectrum Cl Ϫ transport antagonist DIDS. These currents are not affected by changing the driving force for Mg 2ϩ across the plasma membrane. In accordance with our data, SLC41A2, another member of the SLC41 transporter family, has also been reported to mediate large Mg 2ϩ currents when expressed in X. laevis oocytes (16) but induces no significant currents after expression in TRPM7-deficient DT40 cells (17 (15), an increased Mg 2ϩ influx capacity of ϩtet HEK293-(SLC41A1) was expected. For this reason, our experiments were originally designed to support such SLC41A1-related Mg 2ϩ uptake by performing all preparation and storage procedures before the actual measurements in Mg 2ϩ -free solutions. It is very likely that ϩtet HEK293-(SLC41A1) cells already lose relatively high amounts of intracellular Mg 2ϩ during this time period due to increased magnesium efflux compared with wild type cells. This assumption is supported by the very low initial [Mg 2ϩ ] i levels (ϳ0.2 mmol liter Ϫ1 ) measured in ϩtet HEK293-(SLC41A1) cells incubated in Mg 2ϩ -free medium. HEK293 cells express the constitutively active channel TRPM7, which has been shown to mediate Mg 2ϩ uptake in various cell types (11,12,39). Thus, TRPM7 background activity mainly explains the [Mg 2ϩ ] i increase seen in Ϫtet and ϩtet HEK293-(SLC41A1) cells in the presence of an inwardly directed Mg 2ϩ gradient. However, a higher efflux capacity after hSLC41A1 overexpression in conjunction with lower initial [Mg 2ϩ ] i levels may result in a stronger TRPM7-mediated influx component in ϩtet HEK293-(SLC41A1) cells. After correction for this component, an apparent "Mg 2ϩ uptake" still persists resulting in an additional increase of [Mg 2ϩ ] i and significantly higher end point [Mg 2ϩ ] i levels compared with non-induced control cells. At least at the high [Mg 2ϩ ] e of 10 mmol liter Ϫ1 , this is accompanied by a net increase of [Mg] t . Although we cannot preclude from the presented results that SLC41A1 can also mediate Mg 2ϩ influx in the presence of strong inside-directed Mg 2ϩ gradients, our data suggest a [Mg 2ϩ ] e -dependent depression of the SLC41A1-related efflux as the underlying mechanism. Nevertheless, the [Mg 2ϩ ] i increase levels off at about 1 mmol liter Ϫ1 , far below the electrochemical equilibrium for Mg 2ϩ under our experimental conditions. This could be attributable to a negative feedback regulation of TRPM7mediated Mg 2ϩ transport or the existence of another unknown Mg 2ϩ efflux mechanism, such as the Na ϩ /Mg 2ϩ exchanger in HEK293 cells (10).
At a functional level, a DIDS-sensitive anion-linked Mg 2ϩ efflux system has been described in ventricular heart muscle cells (8). Interestingly, abundant levels of the SLC41A1 transcript has been found in the heart (14) and, together with our data, this makes the protein a good candidate for being the proposed efflux pathway. The failure of H 2 -DIDS to change [Mg 2ϩ ] i in our study does not exclude this possibility because it could result from complete inhibition of SLC41A1-related Mg 2ϩ transport by the unphysiologically high extracellular [Mg 2ϩ ] of 10 mmol liter Ϫ1 used in our experiments. Low affinity (K m for [Mg 2ϩ ] e about 2 to 6 mmol liter Ϫ1 ), slow and anionlinked (mostly HCO 3 Ϫ and Cl Ϫ ) Mg 2ϩ transporters also have been functionally described in the basolateral membrane of enterocytes (40,41), erythrocytes (42), and ruminal epithelial cells (43).
In some studies (42), Na ϩ -independent Mg 2ϩ efflux was accompanied by channel-mediated and, therefore, separate Cl Ϫ efflux. This corresponds to our data showing that SLC41A1-related DIDS-blockable Cl Ϫ conductance and [Mg 2ϩ ] i changes in ϩtet HEK293-(SLC41A1) cells are not directly linked. Rather, as described in other studies, endogenous Cl Ϫ channels are activated simply by the reduction of intracellular Mg 2ϩ , a condition that would also favor Mg 2ϩ transport by TRPM7. An investigation of the functional role of the observed Cl Ϫ conductance was beyond the scope of this study. However, the free intracellular [Mg 2ϩ ] is known to be an important regulator of various ion channels, e.g. K ϩ and Na ϩ channels, with very different functions depending on the cell type. Activation of SLC41A1-related Mg 2ϩ efflux by at present unknown mechanisms can thus play a special role in such processes.
CoHex is the only known Mg 2ϩ channel inhibitor showing significant blocking effects on Mg 2ϩ transport conducted by the bacterial CorA and the mitochondrial Mrs2 channels (13,35,44). Here, we demonstrate that CoHex significantly (approximately 50%) and reversibly inhibits the Mg 2ϩ conductance of the TRPM7 ion channels while leaving SLC41A1-mediated [Mg 2ϩ ] i change unaffected. Hence, CoHex may prove to interfere with channel-based Mg 2ϩ transport mechanisms but not carrier-based mechanisms, increasing the possibilities of identifying distinct Mg 2ϩ transport mechanisms in various cell systems. Moreover these results give some indication that SLCA1 functions as an Mg 2ϩ carrier rather than as a channel. An additional feature functionally pointing to a carrier mechanism is the temperature sensitivity of the SLCA1-related Mg 2ϩ change. Wolf et al. (45) have found a similar 80% reduction of Na ϩ /Mg 2ϩ exchanger activity after a temperature reduction from 37 to 15-18°C, although the same temperature change has no significant effect on Mg 2ϩ uptake by the mitochondrial Mg 2ϩ channel Mrs2 (13).
Goytain and Quamme (15) observed Mg 2ϩ currents after overexpression of mouse SLC41A1 in Xenopus oocytes. This, in contrast to our data, points to a channel-like behavior of mouse SLC41A1. Some possible explanation for these diverse results should be given here. One explanation is the simple assumption that, during evolution, the hSLC41A1 Mg 2ϩ carrier evolved from the mouse SLC41A1 ion channel. SLC41A1 from mouse and human are sequentially almost identical (92% identity and 92% similarity, BlastP version 2.2.9; mSLC41A1 protein sequence Q8BJA3/NCBI was blasted against hSLC41A1 protein sequence NP776253/NCBI); thus, on the basis of "structures predetermine functions," they could transport Mg 2ϩ in a similar manner. However, this assumption can be easily refuted by considering that certain point mutation(s) can alter not only the ion specificity of the transporter(s) but also the mechanism(s) of the ion transport itself (46 -48).
Another explanation for the above mentioned difference might be that interactions between SLC41A1 and its binding partners keep the protein functioning as a Mg 2ϩ carrier in mammalian cells, whereas when it is overexpressed in Xenopus oocytes, Salmonella, or any other non-mammalian expression system, the quantitative and/or qualitative lack of such binding partners result in SLC41A1 functioning as an ion channel. This hypothesis is also supported by the finding that Mg 2ϩ accumulation observed after the overexpression of hSLC41A1 in Salmonella occurs rapidly and resembles the kinetics of Mg 2ϩ transport conducted via the CorA channel (27). 8 Although we favor this explanation over the first, further experimental investigation will be necessary to describe its molecular basis. In conclusion, our results show that hSLC41A1 represents a functionally active Mg 2ϩ carrier mediating Mg 2ϩ efflux in mammalian cell systems.