Differential Localization and Operation of Distinct Mg2+ Transporters in Apical and Basolateral Sides of Rat Liver Plasma Membrane*

Upon activation of specific cell signaling, hepatocytes rapidly accumulate or release an amount of Mg2+ equivalent to 10% of their total Mg2+ content. Although it is widely accepted that Mg2+ efflux is Na+-dependent, little is known about transporter identity and the overall regulation. Even less is known about the mechanism of cellular Mg2+uptake. Using sealed and right-sided rat liver plasma membrane vesicles representing either the basolateral (bLPM) or apical (aLPM) domain, it was possible to dissect three different Mg2+ transport mechanisms based upon specific inhibition, localization within the plasma membrane, and directionality. The bLPM possesses only one Mg2+ transporter, which is strictly Na+-dependent, bi-directional, and not inhibited by amiloride. The aLPM possesses two separate Mg2+ transporters. One, similar to that in the bLPM because it strictly depends on Na+ transport, and it can be differentiated from that of the bLPM because it is unidirectional and fully inhibited by amiloride. The second is a novel Ca2+/Mg2+ exchanger that is unidirectional and inhibited by amiloride and imipramine. Hence, the bLPM transporter may be responsible for the exchange of Mg2+ between hepatocytes and plasma, and vice versa, shown in livers upon specific metabolic stimulation, whereas the aLPM transporters can only extrude Mg2+ into the biliary tract. The dissection of these three distinct pathways and, therefore, the opportunity to study each individually will greatly facilitate further characterization of these transporters and a better understanding of Mg2+homeostasis.

Magnesium, the second most abundant cation within mammalian cells, is necessary for a variety of metabolic and cellular functions (1)(2)(3)(4)(5). Under resting conditions, cellular-free Mg 2ϩ concentration is held at 0.5-0.8 mM, well below its predicted electrochemical equilibrium, and approximately 5 mM Mg 2ϩ is complexed with ATP and other metabolites (1,5). In several tissues such as heart and liver, specific hormonal or metabolic stimulation causes, within a few minutes, a massive mobilization of cellular Mg 2ϩ (6 -9). Although cytosolic-free Mg 2ϩ undergoes minimal changes, several intracellular signals leading to a cAMP increase induce a loss of 5-10% of total cellular Mg 2ϩ (6 -9). Conversely, signals activating protein kinase C result in an accumulation of Mg 2ϩ and a consequent increase of total cellular Mg 2ϩ by 5-10% (10).
These findings underscore the operation of a very powerful Mg 2ϩ transport machinery within the plasma membranes of mammalian cells. Yet, the Mg 2ϩ transporter(s) has not been isolated, and even the basic kinetic properties of Mg 2ϩ cellular transport remain confusing and contradictory. For example, the dependence of Mg 2ϩ release on extracellular Na ϩ has been established in invertebrate and mammalian cells (11)(12)(13)(14)(15)(16) but Na ϩ -independent pathways have also been reported (1,8,9,17,18). Equally contradictory are the findings of variable Na ϩ / Mg 2ϩ stoichiometries (1,19), the role of trans-membrane potential, and the partial and variable effectiveness of inhibitors such as amiloride (for review see Refs. 1, 5, 18, and 20). In principle, these discrepancies could be accounted for by the co-existence of separate plasma membrane Mg 2ϩ transporters each operating with different mechanisms and sensitivity to inhibitors. Hence, this work was designed to dissect putatively different Mg 2ϩ transporters using an appropriate cellular model, the plasma membrane of rat hepatocytes.
In view of its size and secretory capacity, the liver should play a major role in Mg 2ϩ homeostasis. Recent evidence in the literature demonstrates that the perfused liver or isolated hepatocytes can accumulate or release very large amounts of Mg 2ϩ upon specific metabolic stimulation (7,8). Recently, this laboratory was able to successfully obtain sealed liver plasma membranes (LPM) 1 to study Mg 2ϩ transport across the hepatocyte cell membrane (21). This model provides well defined and controllable experimental conditions in which membrane transport can be quantitatively investigated in the absence of interfering intracellular signaling pathways or transport by organelles. Using sealed plasma membranes, we recently found that either extravesicular Na ϩ or Ca 2ϩ could mobilize intravesicular Mg 2ϩ (21).
A major difference in liver plasma membrane vesicles is that they release intravesicular Mg 2ϩ in the presence of extravesicular Na ϩ or Ca 2ϩ alone, without the need of metabolic stimulation (21) necessary in intact hepatocytes (7,8). Under these conditions, cellular Mg 2ϩ transport appears "uncoupled" and maximally operating in the absence of stimulatory cellular signals. One intriguing possibility is the presence in hepatocytes of a physiological "break" on Mg 2ϩ transport, which could be transiently removed as the result of activation of cellular pathways. Such a break (i.e. protein phosphorylation and dephosphorylation) appears to be absent in LPM probably because they are devoid of ATP and signaling pathways.
Those studies also provided direct evidence that extracellular Na ϩ and Ca 2ϩ elicited a larger Mg 2ϩ efflux than either cation alone and suggested the presence of distinct Na ϩ /Mg 2ϩ and Ca 2ϩ /Mg 2ϩ transporters (21). Unfortunately, both putative transporters seemed to be inhibited by the same inhibitor. Additionally, it was not clear whether the two transport mechanisms were localized within the same vesicle or in two distinct populations of vesicles.
The hepatocyte is a highly polarized cell with differential localization of enzymes or transporters in the two main membrane domains, the basolateral and apical plasma membrane. The present study has now taken advantage of the possibility of obtaining separate populations of sealed and properly oriented plasma membranes comprised mostly of either apical (aLPM) or basolateral (bLPM) domains.
The data presented here demonstrate that the Na ϩ -dependent and Ca 2ϩ -dependent Mg 2ϩ transporters are distinct in terms of localization within the plasma membrane, inhibition, and overall operation. The data indicate that two distinguishable Na ϩ -dependent Mg 2ϩ exchange mechanisms are localized to both the basolateral and apical domain, whereas a Ca 2ϩ / Mg 2ϩ exchanger operates only in the apical domain. Such a finding should be of primary importance to kinetically characterize individual Mg 2ϩ transporters and to better identify them.

EXPERIMENTAL PROCEDURES
Plasma Membrane Isolation-Total LPM were isolated as described previously (21,22). Briefly, male Harlan Sprague-Dawley rats (250 -350 g) were anesthetized by intraperitoneal injection of 50 mg of pentobarbital/kg of body weight. The abdomen was open and the liver was perfused by the portal vein with 50 ml of isolation medium: 250 mM sucrose, 5 mM K-Hepes, 1 mM EGTA (pH 7.4), at 37°C. The liver was rapidly removed, finely minced, washed twice in the isolation medium, and homogenized in 50 ml of medium using 10 passes with a loose fitting pestle followed by three passes with a tight fitting pestle. The homogenate was diluted to 6% (w/v) with the same buffer and sedimented at 1400 ϫ g for 10 min. The pellets were recovered and resuspended at 6% (final concentration) in the isolation medium by four passes with the tight fitting pestle. Percoll (Amersham Pharmacia Biotech) was added to the resuspension in the proportion of 1.4 ml of Percoll to every 10.4 ml of resuspension (21), and the LPM were isolated by centrifugation at 34,500 ϫ g for 30 min. All operations following liver removal were carried out at 4°C. The top fluffy layer, containing the plasma membrane vesicles, was collected, diluted 1:5 (v/v) with the incubation medium (250 mM sucrose, 50 mM Tris, pH 8.0), and sedimented at 34,500 ϫ g for 30 min. The resulting pellets were diluted to a final concentration of 5 mg of protein/ml in incubation medium and stored in liquid nitrogen until used. No difference in Mg 2ϩ transport was noticed using freshly isolated LPM or vesicles quickly frozen and stored in liquid nitrogen for several days.
Isolation of aLPM and bLPM from Total LPM Population-After isolation, LPM were washed with 5 volumes of 250 mM sucrose, 25 mM K-Hepes, pH 7.4 (washing and incubation buffer) and sedimented at 34,500 ϫ g for 5 min in a SS-34 Sorvall rotor to remove residual Percoll carried over from isolation. LPM vesicles were resuspended in the same buffer to a concentration of 5 mg/ml and homogenized by 75 passes with a tight fitting pestle in a Thomas C Potter (21) to fully homogenize the vesicles. LPM were then layered on a discontinuous sucrose gradient (43%, 46%, and 52% w/v from top to bottom) and sedimented at 93,000 g max for 60 min in a SW 28.1 Beckman rotor. The pellet (bLPM) and the bands at the top of the 43% and 43%/46% sucrose interface (aLPM) were carefully removed, diluted with 5 volumes of buffer, and sedimented at 34,500 ϫ g for 10 min in a SS-34 Sorvall rotor. LPM were stored in liquid nitrogen until used (within one week). All procedures were carried out at 4°C.
Plasma Membrane Purity and Orientation-The purity of the LPM vesicles was assessed using 5Ј-nucleotidase, cytochrome c oxidase, and glucose 6-phosphatase activities as markers for plasma membrane, mitochondria, and endoplasmic reticulum, respectively (see Table I). Alkaline phosphatase, Na ϩ /K ϩ -ATPase, and 5Ј-nucleotidase were used to assess the purity of bLPM and aLPM. In addition, Na ϩ /K ϩ -ATPase and 5Ј-nucleotidase were used to assess LPM orientation (see Table I).
Plasma Membrane Loading-5-ml aliquots of aLPM or bLPM vesicles were resuspended in 25 ml of incubation medium (1:5 v/v) in the presence of 20 mM MgCl 2 or, when specified, Na ϩ or Ca 2ϩ and loading was empirically determined by four passes in a Thomas C Potter with a tight fitting pestle at 4°C as described previously (21).
Mg 2ϩ Measurement-Mg 2ϩ fluxes were measured by atomic absorption spectophotometry (AAS) in a Perkin-Elmer model 3100 as the total content of Mg 2ϩ in the supernatant after fast sedimentation of the vesicles. The Mg 2ϩ -loaded LPM vesicles were resuspended in the Mg 2ϩfree incubation medium previously mentioned. An aliquot of LPM vesicles was incubated in the Mg 2ϩ -free medium, at a final concentration of 250 -350 g of protein/ml. After a 1-min equilibration, aliquots of the incubation mixture were withdrawn in duplicate at 1-2 min intervals, and the vesicles sedimented by rapid centrifugation (7000 ϫ g for 1 min) in microfuge tubes. Mg 2ϩ content was measured in the supernatants by AAS. Similar procedures were also used for Na ϩ -or Ca 2ϩ -loaded LPM.
As for the Mg 2ϩ content in LPM vesicles (pellet measurement), aliquots of the incubation medium were sedimented in microfuge tubes through an oil layer (dimethyl-phthalate:dibutyl-phthalate:dioctylphthalate (2:3:4)) to remove extravesicular Mg 2ϩ . The supernatant and oil layer were removed by vacuum suction, and the pellet was digested overnight in 500 l of 10% HNO 3 . The Mg 2ϩ content of the vesicles was measured by AAS in the acid extracts using Mg 2ϩ standards in identical acid concentration. Similar procedures were also used for Na ϩ -or Ca 2ϩ -loaded LPM.
Protein Assay-Protein was measured according to the procedure of Bradford (23), using bovine serum albumin as a standard.
Chemicals-All chemicals and assays were of the purest analytical grade (Sigma). Nitrex nylon mesh was obtained from Tetko, Inc. (Briarcliff Manor, NY). Percoll was from Amersham Pharmacia Biotech.
Statistical Analysis-Data are presented as mean Ϯ S.E. Data were first analyzed by one-way ANOVA (analysis of variance). Multiple means were then compared by the Student-Newman-Keuls method multiple comparison test performed with statistical significance designated at a p value of Յ0.05.

Purity and Orientation of bLPM and aLPM after Isolation-
The purity and enrichment of bLPM and aLPM were assessed using 5Ј-nucleotidase and alkaline phosphatase as apical markers and Na ϩ /K ϩ -ATPase as a basolateral marker (24 -26). The maximal activity (considered as 100%) was obtained by using 0.06% Triton X-100 (21). Table I shows the activities and the enrichment of bLPM and aLPM as compared with homogenate and total LPM prior to separation. As can be noted, bLPM were enriched 19-and 14-fold over homogenate and aLPM, respectively, as measured by Na ϩ /K ϩ -ATPase activity. By contrast, aLPM were enriched 4-and 15-fold over bLPM as measured by 5Ј-nucleotidase and alkaline phosphatase, respectively. The protein recovery of bLPM and aLPM from the total LPM population was 34.6% Ϯ 12% and 38.2% Ϯ 7%, respectively (or 0.67 and 0.71 mg of protein/g of liver). As described TABLE I Enzyme activity of purified basolateral and apical liver plasma membranes Alkaline phosphatase is expressed as mol of p-nitrophenyl hydrolyzed/mg of protein Ϫ1 /min Ϫ1 , whereas all the other data are expressed as mol Pi/mg of protein Ϫ1 /min Ϫ1 . All experimental protocols were performed as previously described (21). All data are reported as mean Ϯ S.E.; n ϭ 15, 8, 8, and 10 for 5Ј-nucleotidase, MgATPase, Na ϩ /K ϩ -ATPase, and alkaline phosphatase in homogenate and total LPM, respectively. n ϭ 10, 6, 6, and 10 for 5Ј-nucleotidase, MgATPase, Na ϩ /K ϩ -ATPase, and alkaline phosphatase in aLPM and bLPM, respectively. previously, correct (inside-in) orientation of LPM is greatly affected by the presence of Mg 2ϩ , which has been shown to induce correct membrane sidedness (27,28). The sidedness of LPM was determined as previously reported (21). Both the Mg 2ϩ -loaded aLPM and bLPM were approximately 80% inside-in. Dose Response of Na ϩ and Ca 2ϩ Induced Mg 2ϩ Efflux from bLPM and aLPM and the Effect of Various Inhibitors- Fig. 1A shows that bLPM loaded with 20 mM MgCl 2 and suspended in the absence of Na ϩ and Ca 2ϩ released negligible Mg 2ϩ over several min of incubation (control). The addition of increasing concentrations of extravesicular NaCl induced a dose-dependent efflux of Mg 2ϩ , which was maximal at 10 mM NaCl (138.1 Ϯ 15.1 nmol of Mg 2ϩ /mg of protein, n ϭ 4). Concentrations of NaCl larger than those reported in the figure (e.g. 25 or 50 mM) did not induce a statistically significant increase in Mg 2ϩ efflux (142.9 Ϯ 12.5 nmol of Mg 2ϩ /mg of protein, n ϭ 5, for 25 mM NaCl versus 138.1 Ϯ 15.1 nmol of Mg 2ϩ /mg of protein, n ϭ 4, for 10 mM NaCl). Irrespective of the concentration of NaCl used, Mg 2ϩ efflux at 37°C was maximal within the first min after the Na ϩ addition (Fig. 1). In contrast to what was previously observed in the total LPM population (21), the addition of 500 M CaCl 2 to bLPM elicited a negligible release of Mg 2ϩ (18.4 Ϯ 6.0 nmol of Mg 2ϩ /mg of protein as compared with 267.7 Ϯ 17.0 in the total LPM population (21).
Several agents have been reported to inhibit the Na ϩ /Mg 2ϩ transporter. Amiloride has been shown to effectively inhibit Mg 2ϩ extrusion via the putative Na ϩ /Mg 2ϩ exchanger in several mammalian cell types, including hepatocytes (5, 8, 14, and 18). Amiloride also has been shown to inhibit the Na ϩ /H ϩ exchanger as well as Na ϩ entry pathways in epithelia (29,30). In addition, Feray and Garay (13) demonstrated that tricyclic antidepressant drugs, such as imipramine, are nonspecific inhibitors of Mg 2ϩ transport, i.e. Na ϩ /Mg 2ϩ exchange, in human red cells. Interestingly, the addition of 1 mM amiloride to bLPM did not inhibit Na ϩ -induced Mg 2ϩ efflux (Fig. 1B) as compared with a 50% inhibition observed in the total LPM population (21). In fact, 1 mM amiloride enhances the Na ϩ -induced Mg 2ϩ extrusion 3 min after addition of NaCl (Fig. 1B), a phenomenon for which we have no explanation. By contrast, the addition of 200 M imipramine effectively inhibited Na ϩ -induced Mg 2ϩ mobilization up to approximately 85% (Fig. 1B). Fig. 2, A and B, illustrate the Na ϩ and Ca 2ϩ dose-dependent effect, respectively, on Mg 2ϩ efflux in aLPM. In Fig. 2A maximal Mg 2ϩ efflux is induced by much greater concentrations of NaCl than in bLPM (e.g. 25 mM induces an efflux of 184.7 Ϯ 36.7 nmol of Mg 2ϩ /mg of protein). At variance from bLPM, Ca 2ϩ could also induce Mg 2ϩ efflux in these vesicles. The maximal Mg 2ϩ efflux induced by 500 M CaCl 2 was 227.2 Ϯ 50.2 nmol of Mg 2ϩ /mg of protein, n ϭ 4 (Fig. 2B). The concentrations of Na ϩ or Ca 2ϩ able to elicit maximal Mg 2ϩ efflux from aLPM are very similar to those effective in the total LPM population (21). Also, in contrast to bLPM, 1 mM amiloride inhibited both Na ϩ -and Ca 2ϩ -induced Mg 2ϩ efflux in aLPM to a comparable extent (Fig. 2C). 200 M imipramine was equally effective at inhibiting both Mg 2ϩ extrusion mechanisms (not shown).
Partial Kinetics of Mg 2ϩ Efflux-The Mg 2ϩ efflux at 37°C shown in the previous experiments was already maximal at the first time observation point, after 2 min. Hence, it was difficult to distinguish whether the Mg 2ϩ appearing in the supernatant was the result of net efflux and/or the release of Mg 2ϩ from surface binding sites. Some kinetic resolution can be achieved by lowering the temperature of incubation. Fig. 3, A and B, show the rates of Na ϩ -or Ca 2ϩ -dependent Mg 2ϩ extrusion in the supernatant from 20 mM MgCl 2 -loaded bLPM (Fig. 3A) or aLPM (Fig. 3B) incubated in a thermostated vessel at 10°C.
Under these conditions, the efflux of Mg 2ϩ is comparable, but the rates of release are much slower and kinetically distinguishable.
Specificity of Transport-Na ϩ -dependent Mg 2ϩ extrusion in FIG. 1. Na ϩ , but not Ca 2ϩ , induces Mg 2ϩ efflux from 20 mM MgCl 2loaded basolateral LPM (A). Imipramine but not amiloride inhibits Mg 2ϩ efflux (B). Mg 2ϩ efflux is reported as the net change in Mg 2ϩ content in the supernatant with respect to that before the addition of extravesicular Na ϩ or Ca 2ϩ (Fig. 1A). Dose response for NaCl induced Mg 2ϩ efflux from 20 mM MgCl 2 -loaded bLPM vesicles incubated in the absence of extravesicular Mg 2ϩ . All experiments were performed as described under "Experimental Procedures." Vesicles (250 g/ml) were incubated at 37°C in a reaction medium containing 250 mM sucrose, 25 mM K-Hepes, pH 7.4. After a 2-min equilibration, an aliquot corresponding to 500 l of incubation medium was withdrawn in duplicate and rapidly sedimented in microfuge tubes. Mg 2ϩ content in the supernatant was measured by AAS. After withdrawal, the concentrations of NaCl or CaCl 2 indicated in the figure were added, and an aliquot of the incubation mixture was withdrawn in duplicate at the reported time points and processed as above. The figure represents the net change in Mg 2ϩ content in the supernatant with respect to that present before ion addition. In Fig. 1B, 1 mM amiloride or 200 M imipramine was added 2 min before Na ϩ addition. Data are mean Ϯ S.E. of six and eight different preparations for Na ϩ for Ca 2ϩ , respectively. ANOVA and or the Student-Newman-Keuls method was performed at all time points. *, p Ͻ 0.05 versus control and imipramine. bLPM is specific for Na ϩ . Addition of 25 mM KCl did not elicit a significant Mg 2ϩ extrusion, and the addition of 25 mM LiCl elicited an efflux of 55.7 Ϯ 29.1 nmol of Mg 2ϩ /mg of protein.
The efflux induced by Li ϩ was not statistically significant from the control but was statistically significant from 10 mM NaCl (Fig. 4A). Similarly, the addition of 25 mM LiCl or 25 mM KCl to aLPM also elicited negligible Mg 2ϩ efflux (not shown), indicat-ing that the Mg 2ϩ extrusion observed in aLPM strictly depends on extracellular Na ϩ . Compared with Na ϩ , the Ca 2ϩ requirement for Mg 2ϩ release from aLPM is far less specific, because Mg 2ϩ efflux could also be induced by the addition of equimolar concentrations of other divalent cations (Fig. 4B).
Bidirectionality of the Mg 2ϩ Exchange- Fig. 5A shows the result of a set of experiments in which bLPM loaded with 20

FIG. 2. NaCl-induced (A) or CaCl 2 -induced (B) Mg 2؉ efflux from 20 mM MgCl 2 -loaded aLPM and inhibition by amiloride (C).
Amiloride (1 mM) was added 2 min before the Ca 2ϩ or Na ϩ addition. All experiments were performed as described in the legend to Fig. 1. Data are mean Ϯ S.E. of three preparations. ANOVA and the Student-Newman-Keuls method for multiple comparisons were performed at all time points. *, p Ͻ 0.05 versus control and amiloride. mM NaCl were stimulated by the addition of varying concentrations of extravesicular Mg 2ϩ . The observation that extravesicular Mg 2ϩ can induce a Na ϩ efflux from Na ϩ -loaded bLPM demonstrates that the Na ϩ /Mg 2ϩ extrusion mechanism is bidirectional. On the other hand, when aLPM were loaded with 20 mM Na ϩ and stimulated with Mg 2ϩ , no Na ϩ efflux or Mg 2ϩ influx was observed, indicating that the Na ϩ /Mg 2ϩ transport mechanism in these vesicles is unidirectional (Fig. 5B). When similar experiments were performed in aLPM loaded with CaCl 2 , the addition of external MgCl 2 did not induce Ca 2ϩ extrusion or, consequently, Mg 2ϩ uptake. This indicates that the Ca 2ϩ /Mg 2ϩ transporter in aLPM is also unidirectional, i.e. Mg 2ϩ extrusion for Ca 2ϩ uptake (Fig. 5C). Table II compares the initial loading content under each experimental loading condition as well as the percentage of cation releasable upon thr addition of Triton X-100. Table II also reconfirms that when no movement is observed such as in Fig. 5, B or C, the absence of efflux is not due to mismatched loading conditions.

DISCUSSION
Properties of Plasma Membrane Mg 2ϩ Transport-As indicated in the Introduction, the regulation of Mg 2ϩ homeostasis in liver and other tissues is poorly understood. In response to activation of specific cell signaling, it has been shown that massive amounts of total Mg 2ϩ can be translocated within minutes in or out of hepatocytes (7,8,10). This indicated a very powerful and rapid transport machinery of Mg 2ϩ transport within the plasma membrane.
Several groups have demonstrated that the plasma membrane of different cell types possesses very active Mg 2ϩ transport mechanisms (6 -14). Data in the literature support the operation of at least two distinct Mg 2ϩ transport mechanisms in mammalian cells tentatively identified as a Na ϩ -dependent and a Na ϩ -independent exchange pathway, respectively (8,9,(12)(13)(14)(15)(16)(17). Experiments performed using a variety of perfused organs or isolated cells indicate that Mg 2ϩ transporters operate with varying stoichiometric ratios and transport properties (for review see Ref. 1, 3, 5, 18, 19, or 20) under different experimental conditions. This information makes it difficult to reconcile the available data in a comprehensive scheme.
These conflicting results could be consistent with the operation of several Mg 2ϩ transporters within the same cell, each operating under different conditions and sensitivity to inhibitors. In this work, we were successful in obtaining effective separation of LPM into two fractions enriched in aLPM or bLPM, which maintained low permeability to cations and proper sidedness. Using these preparations to study Mg 2ϩ fluxes, it was possible to observe that the plasma membrane of rat liver possesses three kinetically distinct Mg 2ϩ transport pathways. These transporters are distinguishable in terms of localization within the plasma membrane, sensitivity to inhibitors, requirement of cation for co-transport, and reversibility.
The overall conclusion is that the apical LPM possesses two distinct Mg 2ϩ transporters, one Ca 2ϩ -and one Na ϩ -dependent. In contrast, the basolateral LPM contains only one Mg 2ϩ transporter dependent upon Na ϩ . Partial evidence indicates that the Mg 2ϩ transporters activated by Na ϩ in the apical and basolateral membranes may be distinct. The kinetic properties of these transporters are outlined in Table III.
We conducted several controls as previously shown (21) to conclusively demonstrate that the Mg 2ϩ extruded in the supernatant is the result of net translocation across the membrane and that Mg 2ϩ efflux is not the result of a passive leakage from Mg 2ϩ -loaded vesicles or a release from surface binding sites after the addition of a cation or osmotic mismatch. Briefly, these experiments include the following: (a) unstimulated LPM do not release entrapped Mg 2ϩ over several min of incubation even in the presence of a gradient across the plasma membrane ( Figs. 1 and 2); (b) both Na ϩ -and Ca 2ϩ -induced Mg 2ϩ effluxes are completely prevented in the presence of inhibitors; (c) 25 mM LiCl or 25 mM KCl (Fig. 4A) fail to replace Na ϩ in mobilizing Mg 2 ϩ ; (d) at lower temperatures Mg 2ϩ release induced by Ca 2ϩ or Na ϩ is slow and kinetically resolvable; and (e) most of the Mg 2ϩ releasable by Ca 2ϩ or Na ϩ can be the same Mg 2ϩ pool released by detergent or cation ionophore (21).
Properties of Mg 2ϩ Transport in the Basolateral Membrane-Only Na ϩ can induce the release of intravesicular Mg 2ϩ from bLPM. The Na ϩ requirement is specific, because no other monovalent or divalent cation can induce Mg 2ϩ efflux. The process is fully reversible, as extravesicular Mg 2ϩ effectively mobilizes intravesicular Na ϩ . It is noteworthy that imipramine, but not amiloride, is effective at inhibiting Na ϩ -induced Mg 2ϩ extrusion in bLPM (Fig. 1B). On the contrary, amiloride inhibits the Na ϩ /Mg 2ϩ mechanism in aLPM by Ͼ85% (Fig. 2C). As the ratio of aLPM to bLPM is ϳ1 on a protein basis, the differential effectiveness of amiloride to inhibit Mg 2ϩ transport in bLPM versus aLPM could explain our previous observation that amiloride inhibits Mg 2ϩ transport in total LPM by approximately 50% (21).
Properties of Mg 2ϩ Transport in the Apical Membrane-The data from aLPM demonstrate that the apical domain of hepatocytes has both a Na ϩ /Mg 2ϩ and a Ca 2ϩ /Mg 2ϩ transport mechanism. The Na ϩ /Mg 2ϩ exchange in aLPM operates in a manner similar to the Na ϩ /Mg 2ϩ exchanger described in bLPM in terms of specificity for Na ϩ . However, three main differences are evident: (a) the dose response for extravesicular Na ϩ in aLPM is different than in bLPM in that the concentration for maximal activity is left-shifted in bLPM (Fig. 1A), (b) the transport is almost completely inhibited by amiloride (Fig. 2C), and (c) the process is not reversible because extravesicular Mg 2ϩ cannot elicit efflux of intravesicular Na ϩ (Fig. 5B). These differences can be accounted for by the presence of two separate Na ϩ /Mg 2ϩ antiporters in the apical and basolateral sides of  hepatocytes or by an identical Mg 2ϩ transporter differently regulating Na ϩ exchange. In the latter alternative, the difference in Na ϩ affinity, amiloride sensitivity, and reversibility will be confined to the different mechanism of Na ϩ influx in either plasma membrane domain. The Mg 2ϩ release mechanism dependent upon extracellular Ca 2ϩ and found in the apical membrane is a novel observation identified for the first time in mammalian cells. The data from aLPM demonstrate that the Ca 2ϩ /Mg 2ϩ transporter is strictly localized in this domain and that the transporter is nonselective as it can be activated by several divalent cations (Fig. 4B). The fact that this transporter is unidirectional suggests that in vivo it may operate only in terms of Mg 2ϩ efflux and Ca 2ϩ uptake.
Physiological Significance of Hepatocyte Mg 2ϩ Transporters-The distribution of the three Mg 2ϩ transport mechanisms within the hepatocyte is diagrammatically illustrated in Fig. 6. Based upon the limited share of knowledge of intra-and extracellular Mg 2ϩ homeostasis in liver, or for that matter, in any other tissue, any discussion of the physiological role of these different transporters in hepatocytes remains speculative. Owing to its large mass, blood flow, and active secretory capacity, the liver is positioned to play a key role in the homeostasis of metabolites and ions such as Mg 2ϩ . Hepatocytes are polarized cells in which transport between hepatocytes and plasma is regulated by the basolateral membrane, whereas that between hepatocyte and biliary tract is regulated by the apical membrane.
The basolateral membrane accounts for approximately 90% of the surface area of the hepatocyte plasma membrane. Hence, the basolateral Na ϩ -dependent Mg 2ϩ transport, described by several groups, should play the primary role in hepatocyte Mg 2ϩ homeostasis as well as in plasma Mg 2ϩ homeostasis. The novel finding of this work is that the basolateral Na ϩ /Mg 2ϩ overall exchange is reversible, whereas the Na ϩ -dependent and Ca 2ϩ -dependent Mg 2ϩ transport mechanisms in the apical membrane are not. This implies that the basolateral Na ϩ -de-pendent Mg 2ϩ transporter is most likely the effector for the observed large and rapid changes in plasma (31) and tissue (6 -10) Mg 2ϩ observed in either direction after the addition of ␣or ␤-adrenergic agonists (7-9, 32), protein kinase C activation (10), insulin (33), etc.
In contrast, the apical membrane Mg 2ϩ transporter is unidirectional and can only transport Mg 2ϩ from the hepatocytes to the bile in exchange for Ca 2ϩ or Na ϩ . Data in the literature indicate that the concentration of total Ca 2ϩ in the bile ranges from 2 to 16 mM (34). The unidirectional Ca 2ϩ /Mg 2ϩ exchange could be responsible for decreasing biliary Ca 2ϩ thereby preventing the formation of bile stones or be an additional mechanism contributing to the overall reabsorption of Ca 2ϩ (34 -36). The physiological significance of the unidirectional Na ϩ -dependent Mg 2ϩ efflux from hepatocytes through the apical membrane is equally conjectural and may operate in tandem with the described Na ϩ /bile acid uptake or for direct Mg 2ϩ extrusion (37). It will be interesting to observe whether these transporters are up-regulated in conditions of chronic alcoholism or diabetes when the tissue concentrations of Mg 2ϩ in liver and other tissues dramatically decrease (38,39).
Although the present data may provide more questions than answers as to the physiological and metabolic implications of liver Mg 2ϩ transport, they dissect an overall complex Mg 2ϩ cellular transport in individual components. Identifying three individual transporters on the basis of localization, inhibition, reversibility, and cation specificity will permit the design of experimental approaches to further investigate the distribution of these Mg 2ϩ transporters in other tissues, the more detailed mechanism of transport and the physiological significance in terms of Mg 2ϩ , and other ion and metabolic homeostasis.