Magnesium Transport and Magnesium Homeostasis

KRISZTINA M. PAPP-WALLACE* AND MICHAEL E. MAGUIRE

[SECTION EDITOR: JOHN FOSTER]

Posted September 25, 2008

Department of Pharmacology, School of Medicine, Case Western Reserve University, Cleveland, Ohio 44106-4965

*Corresponding author. Mailing address: Department of Pharmacology, School of Medicine, Case Western Reserve University, 10900 Euclid Ave., Cleveland, OH 44106-4965. Phone: (216) 368-6187, Fax: (216) 368-3395, E-mail: This e-mail address is being protected from spambots. You need JavaScript enabled to view it

This chapter reviews the properties and regulation of the Salmonella enterica serovar Typhimurium and Escherichia coli transporters that mediate Mg²⁺ influx: CorA and the Mgt P-type ATPases. In addition, potential Mg²⁺ regulation of transcription and translation, largely via the PhoPQ two-component system, is discussed.

CorA proteins are a unique class of transporters and are widespread in the Bacteria and Archaea, with rather distant but functional homologs in eukaryotes. The Mgt transporters are highly homologous to other P-type ATPases but are more closely related to the eukaryotic H⁺ and Ca²⁺ ATPases than to most prokaryotic ATPases. Together, these systems are responsible for the uptake of Mg²⁺ and the maintenance of the intracellular Mg²⁺ content. An extensive review of all prokaryotic Mg²⁺ transporters has been published recently (52).

MAGNESIUM CONTENT AND CHEMISTRY

Mg²⁺ is the most abundant divalent cation in bacterial cells. It is also very different from Ca²⁺ and from the monovalent cations Na⁺ and K⁺ in several important ways (Table 1). The radius of the hydrated form of Mg²⁺ is much larger than that of the hydrated form of any other cation, while the atomic ion is much smaller than the ionic forms of other common biological cations, Na⁺, K⁺, and Ca²⁺. Since variations in the radius correspond to variations in volume, which is equal to the radius to the third power, these differences are much clearer in comparisons of the ratios of the hydrated and ionic volumes of the cations (Table 1), which show the hydrated Mg²⁺ cation to be ~400 times larger in volume than the atomic ion and the hydrated Na⁺, K⁺, and Ca²⁺ cations to be only 4- to 25-fold larger than the respective atomic ions. A hydrated Mg²⁺ cation is approximately the size of a glucose molecule when its inner hydration shell is considered. Since, however, the transport number for Mg²⁺ (Table 1),an estimate of the average number of solvent molecules associated sufficiently tightly with an ion that they migrate through the solution as the cation diffuses, is much higher than those for other biological cations and its water exchange rate is more than 3 orders of magnitude lower, a hydrated Mg²⁺ cation diffusing in an aqueous solution is effectively far larger than all other cations. Further, Mg²⁺ is invariably hexacoordinate and strongly prefers bond angles very close to 90°, whereas, for example, Ca²⁺ can be hexa-, hepta-, octa-, or even nonacoordinate and in the hexacoordinate state exhibits a bond angle of 90° ± 40°. Bond distances for Mg²⁺ are shorter, at 1.9 to 2.1 Å, than those for Ca²⁺, which are 2.2 to 2.7 Å (26). These properties of Mg²⁺ also suggest that transport proteins or ion channels mediating Mg²⁺ flux may have unusual properties necessary to accommodate this unique cation (38, 51).

Table 1Properties of ions of common biological cations

In addition to these chemical properties, the comparatively rigid structure of the Mg²⁺ coordination sphere versus that of Ca²⁺ reflects the relative roles of these cations in cellular biochemistry. Ca²⁺, primarily a signaling molecule, must bind to many different proteins and elicit a conformational change, implying that the geometry of the Ca²⁺ binding site must be flexible. In contrast, the most common physiological role of Mg²⁺ is to bind a nucleoside triphosphate in the active site of an enzyme. The binding of Mg²⁺ to the phosphoryl groups holds them in a specific spatial conformation that is critical for catalysis. In other enzymes, Mg²⁺ appears to hold a water molecule in a specific position, either to form a specific structure or to position the water molecule for a role in catalysis (18).

The total intracellular concentration of Mg²⁺ is about 30 mM (54), although the total concentration of Mg²⁺ bound to a gram-negative bacterium (primarily to the outer membrane), such as E. coli and Salmonella serovar Typhimurium, may reach 100 mM (2). The level of intracellular free Mg²⁺ is about 0.3 to 1.0 mM, and thus, most Mg²⁺ inside a cell is bound. Forms of bound Mg²⁺ include Mg²⁺ within kinetic complexes and Mg²⁺ associated with DNA or lipids. Levels of extracellular Mg²⁺ vary depending on the environment of the bacterium. Within the host, the Mg²⁺ concentration in the stomach, in the intestine, within macrophages (phagosomes), and within serum is about 1.0 mM.

In the laboratory, Luria-Bertani broth used routinely to culture bacteria has approximately 30 to 40 μM Mg²⁺. Typical low-Mg²⁺ laboratory culture medium is a minimal medium supplemented with sugar and amino acids and contains less than 15 μM Mg²⁺. High-Mg²⁺ laboratory culture medium is typically a minimal medium supplemented with sugar and amino acids and contains >1 mM Mg²⁺.

MAGNESIUM TRANSPORTERS OF E.COLI AND SALMONELLA SEROVAR TYPHIMURIUM

corA was initially identified in E. coli by investigators in Silver’s and Kennedy’s laboratories as a locus giving moderate resistance to Co²⁺. Transport and genetic studies indicated that the locus was responsible for Mg²⁺ influx (70, 85). Nelson and Kennedy later identified a second system in E. coli which was repressible by Mg²⁺ (69). Park et al. further characterized both loci and named the inducible locus mgt (73). A detailed characterization of the systems in E. coli and Salmonella serovar Typhimurium was later performed by investigators in the Maguire laboratory (43, 44, 89, 90). E. coli has two Mg²⁺ transporters, CorA and MgtA. Salmonella serovar Typhimurium has these two systems, as well as an additional P-type ATPase, MgtB. In both species, corA and mgtA are found as single open reading frames. In Salmonella serovar Typhimurium, mgtB is found with mgtC in a two-gene operon at the 3' end of Salmonella pathogenicity island 3 (SPI-3).

CorA is constitutively expressed and is the primary Mg²⁺ transporter for both species, as indeed it is for perhaps half of all bacteria and many archaea (43, 44, 89). Hundreds of homologs of CorA are currently known from genomic sequencing. In contrast, only when extracellular and possibly intracellular Mg²⁺ levels fall significantly is the expression of mgtA and mgtB induced (20, 93, 95, 96). The transcription of mgtA and mgtB in response to low extracellular levels of Mg²⁺ is regulated by the two-component system PhoP-PhoQ (36, 59, 60). mgtA is widespread in the Enterobacteriaceae, and it has homologs in gram-positive bacteria and some extremophiles (47, 52). On the other hand, mgtB is present in only a few enterobacterial species, in accordance with its apparent acquisition via horizontal transfer (12).

Half-maximal uptake by CorA occurs at 15 μM Mg²⁺. As measured by bulk cell uptake, the V _max for CorA is >0.5 nmol of Mg²⁺ min^-1 10⁸ cells^-1 (43), but this measurement may be misleading since the protein is apparently an ion channel (see below). Cation selectivity for CorA is as follows: Mg²⁺ = Co²⁺ > Ni²⁺. Co²⁺ and Ni²⁺ have affinities of 20 and 200 μM, respectively (89). Both cations are toxic to the cell at these concentrations. Thus, the primary physiological role for CorA is Mg²⁺ influx. Mn²⁺ inhibits Mg²⁺ influx with a K_i of 30 μM but is not transported. No other cations, including Fe²⁺ (71), are transported, nor do any inhibit Mg²⁺ influx with good affinity except for Mn²⁺, which inhibits noncompetitively with a K_i of ~30 μM (89).

mgtA and mgtB expression is absent or minimal under most laboratory growth conditions unless the medium Mg²⁺ concentration is reduced below 15 μM (see below). At such low extracellular Mg²⁺ concentrations, both mgtA and the mgtCB operon are induced greatly. The affinity of MgtB for Mg²⁺ is slightly better than that of MgtA, 5 to 10 μM versus 20 to 30 μM. A V _max value is not relevant because of the regulation, but maximal uptake via MgtA or MgtB can reach 20 to 30% of that via CorA (89). Both ATPases transport only Mg²⁺ and Ni²⁺. Unlike CorA, neither can transport Co²⁺. Uptake by MgtA is inhibited by Zn²⁺, Co²⁺, and Ca²⁺ but not by Fe²⁺ or Mn²⁺, with a rank order of potency as follows: Zn²⁺ ≥ Mg²⁺ > Ni²⁺ ≈ Co²⁺ > Ca²⁺. In contrast, MgtB is inhibited by Mn²⁺ but not Zn²⁺ or Ca²⁺, with the following rank order of potency: Mg²⁺ ≈ Co²⁺ ≈ Ni²⁺ > Mn²⁺ >> Ca²⁺ or Zn²⁺ (89). Interestingly, although both MgtA and MgtB exhibit robust Mg²⁺ uptake when assayed in whole cells at 37°C, MgtB is completely inactive in cells assayed at 20°C while MgtA is still quite active. Thus, CorA, MgtA, and MgtB are highly selective for Mg²⁺ physiologically. The only other cations that interact significantly are Ni²⁺ and Co²⁺. The most straightforward reason for interaction with the latter two cations is that both have approximately the same atomic radius as Mg²⁺, about 0.65 Å. The transport properties of each are significantly different from those of Mg²⁺, especially with CorA. A graph of the rate of the whole-cell uptake of Mg²⁺, as assayed with ²⁸Mg²⁺, is linear for only a few seconds, while that of the uptake of Ni²⁺ or Co²⁺ is linear for many minutes. This pattern may reflect the channel properties of CorA, which appears to interact initially with a fully hydrated Mg²⁺ cation (see below), and suggests that CorA interacts with Ni²⁺ and Co²⁺ primarily as atomic ions rather than as hydrated cations. This scenario is also consistent with the Mn²⁺ cation’s being the only other cation that can significantly inhibit CorA. Mn²⁺ has an atomic radius of 0.75 to 0.8 Å, much closer to the 0.65-Å sizes of Mg²⁺, Ni²⁺, and Co²⁺ and smaller than those of other transition metal divalent cations.

CorA STRUCTURE

CorA is a 37-kDa protein. Salmonella serovar Typhimurium and E. coli CorA proteins differ by 8 of 316 amino acids (aa), all corresponding to conservative substitutions. Topology studies using blaM and lacZ fusions initially indicated that the Salmonella serovar Typhimurium CorA contained three transmembrane (TM) segments (87); however, subsequent data obtained using a variety of approaches showed that the CorA superfamily of proteins have only two TMs at the extreme C terminus (50, 79, 102). Thus, CorA is a two-domain protein with a small TM domain and a large soluble N-terminal domain resident in the cytosol. The latter domain is largely α-helical and can be expressed as a truncated protein with the retention of structure (101).

The crystal structure of the Thermatoga maritima CorA protein was recently determined at a resolution of 3.9 Å for the whole protein (Fig. 1A and Fig. 1B) and 1.85 Å for the soluble domain (50). Subsequently, virtually identical crystal structures of the T. maritima CorA protein at resolutions of 2.9 and 3.6 Å were presented by Eshaghi et al. (27) and Payandeh and Pai (75), respectively. Salmonella serovar Typhimurium CorA shows only about 20% sequence identity to T. maritima CorA but is virtually an exact match in terms of secondary structure predictions (50). Indeed, secondary structure predictions for CorA and its homologs and paralogs, including Mrs2 and ZntB, are all very similar, an indication that all these proteins likely have the same basic structure (53). Indeed, the coordinates for the soluble domain of the ZntB paralog from Vibrio parahaemolyticus were recently deposited in the Protein Data Bank (identification no. 3BHC). The apparent structure of the soluble domain is virtually identical to that of CorA.

Fig. 1-01The crystal structure of T.maritima CorA is shown as the complete homopentamer with the subunits color coded as follows, starting at the N terminus: α₁-α₂-α₃ helices, yellow; β₁-β₇ sheet, blue; α₄-α₅-α₆ helices (including α₅-α₆ willow helices), red; α₇ stalk helix, green fading to light green for TM1; TM2, magenta; and C-terminal basic sphincter, dark blue. The protein is oriented with the periplasm at the top. The blue arrows denoting the β-sheets show the orientation of each strand with the arrow pointing from N terminus to C terminus.

Fig. 1-02View of T.maritima CorA homopentamer from the periplasm, oriented orthogonally to the view presented in Fig. 1A and shown with the same color coding.

CorA is a funnel-shaped homopentamer (Fig. 1A and Fig. 1B). In the cytosol, the N terminus forms an α ₃-β ₇-α ₃ sandwich domain that comprises a new protein fold (Fig. 1C). This domain leads to the almost-100-Å-long α ₇, or "stalk," helix which largely forms the interior face of the funnel and continues through the membrane as TM1. TM1 ends at the periplasmic face of the membrane in the signature sequence unique to CorA proteins, YGMNF. The mutation of any of the residues in this sequence in Salmonella serovar Typhimurium abolishes transport through CorA (94). Within the homopentamer, the five TM1 helices form the cation pore (Fig. 1B). A short (9-aa) sequence in the periplasm connects TM1 and TM2. The structure of this loop is unresolved in all current T. maritima structures. TM2 lies outside the pore formed by TM1 and returns the C terminus to the cytosol. Virtually all CorA proteins have only a short (6-aa) C-terminal tail, which always contains several Arg and/or Lys residues. For example, the sequence in Salmonella serovar Typhimurium and E. coli CorA proteins is KRKNWL, while that in T. maritima CorA is KKKKWL.

Fig. 1-03Single monomer of T.maritima CorA shown in the same orientation and with the same color coding as the structure in Fig. 1A. Note that TM1 (light green) and TM2 (magenta) are not connected. The 9-aa sequence that connects the two TM segments is not resolved in any of the current crystal structures available (see the text).

In the originally determined T. maritima CorA crystal structure (50), amino acids N314 at the periplasmic end of the pore, M302 in the middle of the pore, and L294 and M291 at the cytosolic end of the pore would appear to occlude cation passage, indicating that this structure is that of a closed form. The mutation of residues of Salmonella serovar Typhimurium CorA that are near the equivalent of the L294-M291 site alters the transport properties of CorA in a manner consistent with these residues’ occluding the pore (94). Similarly, the replacement of L294 with a smaller residue generates a gain in function, most likely due to the pore’s no longer being occluded (74).

The ring of extreme positive charge at the cytosolic membrane interface created by the C terminus (20 Lys residues plus 10 more Lys residues from the extremely long stalk α-helix) creates a large positive potential field at the same level as the L294-M291 occlusion in the pore that would further inhibit the passage of Mg²⁺ through the pore (50). Thus, the combination of the positive field and the L294-M291 amino acid pair has been postulated to be the primary gate for Mg²⁺ passage through the channel. Hypothetical control of this gate is provided by the so-called willow helices (α ₅ and α ₆) that are part of the soluble domain and lie outside the funnel formed by the stalk helix and extending towards the ring of positive charge. The ends of the willow helices contain a plethora of Asp and Glu residues that may reasonably provide a negative counterpart to the ring of positive charge. Interactions between these highly charged regions are potentially critical to the mechanism of transport of CorA (50) and are likely controlled by the binding of an Mg²⁺ ion seen in the crystal structure between D253 of the stalk helix in one monomer and D89 of a more N-terminal α-helix in an adjacent monomer. Two of the crystal structures have an additional adjacent cation bound partially by a single residue, E88 (27, 75). This is likely an artifact of crystallization since the other ligands are water molecules and, therefore, the affinity of Mg²⁺ at this site would be several millimolars, well above physiological free Mg²⁺ concentrations. The homopentamer thus contains (at least) five bound Mg²⁺ ions, which are poised to control the association between the monomers (53). The replacement of either D253 or D89 with lysine creates loss-of-function mutants (74). These mutations may lead to the formation of salt bridges between monomers, keeping CorA in a closed conformation and thus resulting in the loss-of-function phenotype. Further work will be necessary to determine the roles of these various parts of the CorA structure in cation movement.

INHIBITION OF CorA

Cation hexaammines are selective inhibitors of CorA transport (48). Cobalt(III) hexaammine trichloride [Co(NH₃)₆Cl₃] and the ruthenium(II) and ruthenium(III) hexaammine analogs all inhibit ⁶³Ni²⁺ uptake through the Salmonella serovar Typhimurium and Methanococcus jannaschii CorA proteins. The cations have the same diameter (5.0 Å) as a hydrated Mg²⁺ cation. The Ni(II) hexaammine molecule, which is 6.0 Å in diameter, does not inhibit transport. Neither the MgtA or MgtB Mg²⁺ transporters nor the PhoQ Mg²⁺ sensor is inhibited by these cation hexaammines. Co(III) hexaammine does not inhibit the growth of wild-type Salmonella serovar Typhimurium but is bacteriostatic towards an mgtA mgtCB mutant strain that is dependent on corA for Mg²⁺ influx. The cation hexaammines do not enter the cell and, thus, are presumed to bind to CorA via the short periplasmic loop between TM1 and TM2 (53), although Payandeh and Pai suggest that Co(III) hexaammine can enter the external portion of the pore (75). Overall, these data suggest the hypothesis that CorA initially binds a fully hydrated Mg²⁺ cation and that the primary basis of selectivity is size, not charge.

Mg2+ EFFLUX

CorA is also essential for Mg²⁺ efflux in Salmonella serovar Typhimurium and presumably E. coli (35, 89). Under typical laboratory growth conditions, the Mg²⁺ content of the medium is about 10 to 15 μM (minimal medium) or 30 to 40 μM (Luria-Bertani medium). When cells loaded with ²⁸Mg²⁺ are diluted or are washed and resuspended in medium without added Mg²⁺, no significant Mg²⁺ efflux can be detected, even after 2 h of incubation. In contrast, if the concentration of extracellular Mg²⁺ is markedly increased, to the millimolar range, the efflux of intracellular ²⁸Mg²⁺ can be seen (89). Three additional Co²⁺ resistance genes, corB, corC, and corD, influence efflux (35), although none of these three genes map to the same locus as the corB Co²⁺ resistance locus identified in E. coli (73). The mutation of corB, corC, and corD elicits a progressive right shift of the extracellular Mg²⁺ concentration required to elicit ²⁸Mg²⁺ efflux. Since the apparent half-maximal concentration of Mg²⁺ required for influx is 15 μM and that required for a half-maximal rate of Mg²⁺ efflux is about 3 to 5 mM, such efflux via CorA cannot be due to a simple Mg²⁺-Mg²⁺ exchange process because at the extracellular Mg²⁺ concentration at which efflux can be detected, influx is already saturated (35).

MgtA

MgtA is a 95-kDa protein that belongs to the P-type ATPase superfamily (96). P-type ATPases require the ATP-dependent phosphorylation of a conserved aspartyl residue for cation binding and transport (90). Salmonella serovar Typhimurium MgtA is 91% identical to E. coli MgtA (96).

The expression of mgtA has been investigated primarily in Salmonella serovar Typhimurium. The expression of mgtA in E. coli may not be similar. When cells are grown in minimal medium containing millimolar concentrations of Mg²⁺, a decrease in the external [Mg²⁺] elicits a biphasic transcriptional response (95). A small increase in transcription occurs with an apparent half-maximal (extracellular) Mg²⁺ concentration (K _0.5) for Mg²⁺ of 0.5 to 1.0 mM, while a second, far larger increase in transcription has a K _0.5 of 0.01 mM (20, 95). mgtA regulation is complex. The two-component PhoP-PhoQ system controls the transcription of mgtA via the activation of PhoQ by low extracellular concentrations of Mg²⁺. Intracellular Mg²⁺ appears to regulate the mgtA mRNA level posttranscriptionally (see below) (20, 32). Acid shock abolishes the expression of mgtA. However, chronic exposure to acid under low-Mg²⁺ conditions induces mgtA expression, but only if CorA is present (88).

SALMONELLA SEROVAR TYPHIMURIUM MgtB AND MgtC

MgtB is also a P-type ATPase and, at 101 kDa, is slightly larger than MgtA (90). It is encoded by a gene present only in Salmonella serovar Typhimurium at the 3' end of SPI-3 as the second gene of a putative two-gene operon including mgtC. Strikingly, MgtB is as similar (with ≈50% amino acid identity) to mammalian Ca²⁺ ATPases of the sarcoplasmic reticulum and fungal P-type H⁺-ATPases as it is to MgtA. Unlike most prokaryotic P-type ATPases, which contain six TM segments, MgtB (and MgtA) has 10 TM segments (86).

MgtC is a 22.5-kDa protein encoded by the first gene of the mgtCB operon. Although homologs are widespread in prokaryotes, it lacks significant homology to any other known protein, and its function is unknown (65). MgtC is not required for MgtB function or its insertion into the membrane (95). Although MgtC has been hypothesized to be an Mg²⁺ transporter because it is required for growth in liquid medium at extracellular Mg²⁺ concentrations of <50 μM (12), subsequent data have definitively shown that it is not (65). The expression of mgtC in Xenopus laevis oocytes revealed that MgtC can constitutively activate Na⁺,K⁺-ATPase (41). The significance of this activation has not been elucidated. MgtC appears to have two independent physiological roles; it responds to low extracellular Mg²⁺ concentrations, and it is required for survival in macrophages (78). Only some MgtC proteins from other bacteria facilitate both growth in low-Mg²⁺ medium and survival in macrophages when introduced into Salmonella serovar Typhimurium. Others facilitate only growth in low-Mg²⁺ medium; interestingly, the mutation of single amino acids results in the facilitation of the macrophage survival as well (78).

Although mgtC is the first gene of the operon, and although decreased extracellular Mg²⁺ concentrations markedly increase the transcription of mgtCB via PhoP-PhoQ, only MgtB protein is made; no MgtC protein can be detected for several hours after the increase in transcription (65). Even if MgtB is rendered inactive by the mutation of the active-site aspartyl residue, no MgtC protein is made. MgtC is readily detected, however, if most or all of the mgtB gene is deleted. Another protein, MgtR, is responsible for the absence of MgtC when MgtB is present (3). MgtR is a 30-aa peptide encoded downstream of mgtB. It interacts with MgtC protein in vivo, promoting the degradation of MgtC via the FtsH AAA+ protease (3).

As that of mgtA, the expression of the mgtCB operon is biphasic with decreasing concentrations of Mg²⁺ in the growth medium (95). It is currently unknown if mgtB mRNA is regulated by Mg²⁺, like that of mgtA. However, the expression of mgtB can be induced by inorganic acid in the presence of high Mg²⁺ concentrations (8). These results indicate that potentially two forms of regulation of the mgtCB operon are present (95). The mgtCB response to acid shock is similar to that of mgtA (88). Unlike that by MgtA, temperature affects transport mediated by MgtB. In intact cells, MgtA exhibits normal temperature dependence for transport whereas MgtB is active at 37°C but completely inactive at 20°C.

REGULATION OF MAGNESIUM TRANSPORT: PhoP-PhoQ

PhoP-PhoQ is a two-component system consisting of PhoQ, the sensor/receptor histidine kinase, and PhoP, the response regulator/transcriptional activator (60). The expression of both mgtA and mgtCB in either E. coli or Salmonella serovar Typhimurium is markedly induced in a PhoPQ-dependent manner by low concentrations of Mg²⁺ in the medium (31). The activation of PhoQ is controlled by Mg²⁺, Ca²⁺, antimicrobial peptides, and acid (5, 8, 31, 61, 91). E. coli PhoP and PhoQ are 93 and 86% identical to Salmonella serovar Typhimurium PhoP and PhoQ, respectively (37, 45). However, although PhoPQ controls over 100 genes in both species, there is little overlap between the PhoPQ regulons of E. coli and Salmonella serovar Typhimurium (66). The mechanisms of gene regulation by PhoPQ and the molecular details of metal regulation by PhoQ and PhoP are complex. Disagreement exists over several aspects of regulation and biochemistry.

phoP and phoQ form an operon with two promoters in both E. coli and Salmonella serovar Typhimurium (46, 92). The expression of phoP and phoQ from promoter 2 is constitutive at a basal level and independent of the Mg²⁺ concentration. Mg²⁺ controls expression through promoter 1, by environmentally stimulating PhoQ to phosphorylate PhoP. Phosphorylated PhoP will bind promoter 1 and increase the expression of PhoP and PhoQ dramatically; thus, phoPQ transcription is autoregulated (32, 46, 92). PhoP’s autoregulation of transcription results in a surge of transcription of PhoPQ-regulated genes, followed by a decrease so that transcription eventually reaches a steady-state level. This pattern of transcription is mirrored by levels of phosphorylated PhoP; thus, phospho-PhoP is responsible for the surge (84).

Upon encountering low Mg²⁺ concentrations, Mg²⁺ dissociates from PhoQ, which then undergoes a conformational change resulting in the typical autophosphorylation of the PhoQ sensor kinase on a histidine residue (H277) (14, 17, 32, 67). PhoQ transfers the phosphate to an aspartate residue on PhoP through direct interactions with PhoP (13, 14, 67). The mechanisms by which PhoP regulates gene transcription are complex and appear to differ for different PhoPQ-regulated genes. First, in Salmonella serovar Typhimurium, PhoP typically binds to a consensus sequence of (T)G(T)TT(AA) (a PhoP box), which is slightly different from the E. coli PhoP box described by Kato et al., (T/G)GTTTA (46, 103). Second, PhoP can also bind another class of promoters which have a reverse PhoP box (104). The presence of these different boxes results in the differential regulation of transcription at different promoters by PhoP (64). Third, at some but not all PhoP-regulated sites, an additional regulatory protein is required (68). Finally, with the divergently transcribed treR and mgtA genes in E. coli, the binding of PhoP can simultaneously repress the transcription of treR and activate the transcription of mgtA (103). In addition, the presence of both PhoP and RNA polymerase at the promoter site results in an increased ability of each to bind to a promoter (103).

Most unclear at this time are the control and function of PhoP phosphorylation. All investigators agree that increasing the Mg²⁺ concentration activates an Mg²⁺-dependent phosphatase activity of PhoQ that readily dephosphorylates either PhoQ or PhoP. There is disagreement, however, on whether the autokinase activity of PhoQ is Mg²⁺ activated and on whether the phosphorylation of PhoP is required for binding to its promoters. These discrepancies are most likely due to differences in assay conditions, such as temperature, high versus low salt concentrations, pH, and the presence or absence of a His tag on PhoP. However, given the complexity of PhoP-regulated promoters, the requirements for PhoP phosphorylation, at least, may differ at different promoters.

STRUCTURE OF PhoQ

The 2.4-Å-resolution crystal structure of the Salmonella Typhimurium PhoQ periplasmic sensor domain reveals a dimer with a flat, negatively charged region parallel to the membrane surface (17). Bound metal ions bridge this highly negatively charged region and the bacterial inner membrane. Upon the loss of metal, these negatively charged regions should repel each other; this repulsion is thought to be responsible for initiating signal transduction. Several acidic residues of the PhoQ sensor domain, including E149 to E155 in E. coli and D149 to E154 in Salmonella serovar Typhimurium, are important for metal binding (17, 100). The results of reporter assays monitoring the inhibition of the expression of a psiD::lacZ transcriptional fusion in wild-type or phoQ mutant strains exposed to different concentrations of Mg²⁺ and Ca²⁺ suggest that PhoQ has distinct binding sites for Mg²⁺ and Ca²⁺, because the levels of repression by Mg²⁺ and Ca²⁺ alone are distinct and the full repression of psiD::lac expression requires the presence of both metal ions (32). Moreover, a phoQ (pho-24) mutant has differential responses to Mg²⁺ and Ca²⁺ (99). However, nuclear magnetic resonance analyses do not support these conclusions, since identical conformational changes occur in the extracellular PhoQ sensor domain upon the binding of either Ca²⁺ or Mg²⁺ (17).

A 1.6-Å-resolution closed-conformation crystal structure of the E. coli PhoQ cytoplasmic kinase domain with AMP-PNP, a nonhydrolyzable ATP analog, shows that the PhoQ cytoplasmic kinase domain has two separately folding domains: a histidine kinase domain and an ATP binding kinase domain (56). Mutational analyses suggested that three residues in the ATP binding domain, Lys392, Arg434, and Arg439, are important for catalysis. A loop forms a lid that holds AMP-PNP (ATP) into the binding pocket. Conformational changes within this loop generating interactions with the histidine kinase domain were hypothesized previously to be the mechanism of activation (56).

STRUCTURE OF PhoP

Bachhawat and Stock resolved the crystal structure of the PhoP receiver domain in E. coli with and without a phosphoryl analog, beryllofluoride (4a). The receiver domain is a dimer with twofold symmetry and is similar to other regulatory-domain family members. The active form bound to beryllofluoride and the inactive form of PhoP have similar structures. A comparison of the two forms reveals some differences in the backbone of PhoP. At high concentrations of unphosphorylated PhoP protein, PhoP is in an active conformation.

MUTANT FORMS OF PhoP AND PhoQ

Several mutant forms of PhoP and PhoQ have been characterized previously. In Salmonella serovar Typhimurium, a T48I mutant form of PhoQ (corresponding to the pho-24 mutant) is constitutively active, possibly due to either decreased phosphatase activity of PhoQ or an increased net level of phosphotransfer onto PhoP (32, 40). The T48 residue may be required for PhoQ to switch between a PhoP kinase and a PhoP phosphatase (81). Interestingly, a T48I mutant form of E. coli PhoQ is nonfunctional both in E. coli and in Salmonella serovar Typhimurium when expressed in this species (80). In contrast, the Salmonella serovar Typhimurium T48I mutant form is constitutively active in both species. A D179A or D179L mutant form of E. coli PhoQ has the same constitutively active phenotype as a T48I mutant form of Salmonella serovar Typhimurium PhoQ (63). S93N and Q203R mutant forms of PhoP in Salmonella serovar Typhimurium are constitutively active, independently of the presence of PhoQ. The combination of the S93N and Q203R mutations is synergistic (15, 39). The S93N but not the Q203R PhoP mutant form can still interact with PhoQ, and thus, the activity of the S93N mutant PhoP can still be influenced by the Mg²⁺ concentration as long as PhoQ is present.

SENSING BY PhoQ IN THE HOST

The activation of PhoPQ measured by in vitro studies versus that determined by in vivo studies are important to distinguish. PhoQ is activated by many factors, including Mg²⁺, Ca²⁺, pH, and antimicrobial peptides, in the laboratory. Martin-Orozco et al. (57) have used various ion sensors to evaluate the pH and the Mg²⁺ concentrations in the host cell and in the vacuole containing Salmonella serovar Typhimurium after its invasion. Their data suggest that Mg²⁺ concentrations do not markedly change and apparently remain too high to activate PhoQ signal transduction, even within the Salmonella-containing vacuole. Instead, it was suggested that changes in pH are more likely than Mg²⁺ to be responsible for PhoPQ-mediated gene regulation inside the host cell. While the pH dependence of regulation seen in their study is clear and likely highly physiologically relevant, it is not clear that the Mg²⁺ sensor used in the study could detect any changes in Mg²⁺ below 1 mM since its apparent affinity for Mg²⁺ is ~10 mM (72). Thus, the relative roles of pH, Mg²⁺, and even antimicrobial peptides in the regulation of PhoPQ have not yet been definitively determined.

Alternatively, the pH of the phagosome is around 5.5, and antimicrobial peptides can displace metals bound by PhoQ to activate signal transduction (6). Acid (pH 5.5) and antimicrobial peptides have been shown to activate PhoQ even when bacteria are exposed to 1.0 or 10.0 mM extracellular Mg²⁺ (6, 57, 76). Minimal medium with 1.0 mM Mg²⁺ at a neutral pH inhibits PhoQ activation under laboratory growth conditions. Moreover, acid and antimicrobial peptides have an additive effect on PhoQ activation; PhoQ activation is enhanced two- to threefold when antimicrobial peptides are used at pH 5.5 (76). Although the conformation of PhoQ is altered in the presence of acid and a pH of 5.5, PhoQ still binds Mg²⁺. H157 of PhoQ is important for the acid-mediated activation of PhoQ. Thus, acid and antimicrobial peptides seem like more relevant inducers of PhoPQ signal transduction in the host than Mg²⁺.

REGULATION OF mgtA: AN RNA SENSOR FOR Mg2+

In Salmonella serovar Typhimurium, mgtA and mgtCB expression also may be regulated at the mRNA level by intracellular Mg²⁺ (20, 93). mgtA has a 264-nucleotide 5' untranslated region (UTR) between its coding region and promoter. A low Mg²⁺ concentration in the growth medium promotes the transcription of mgtA, even if the mgtA promoter is replaced with the lac promoter. In silico analysis predicts extensive secondary structure within the mgtA 5' UTR. The 5' UTR appears to be capable of forming three stem-loops: A, B, and C. Loop C can form only when loop A and loop B do not form because of overlap. Loop C forms when the intracellular Mg²⁺ level is low, resulting in the transcription of mgtA. When the intracellular Mg²⁺ concentration rises, loop A and B form and the transcription of mgtA is terminated by an Mg²⁺-dependent riboswitch. However, Spinelli et al. indicate that both mgtA and mgtC mRNAs are rapidly degraded (93). Along with the transcription termination model proposed by Cromie et al. (20), they suggest that in the presence of high Mg²⁺ levels, the transcripts are also degraded. RNase E was found to be important for the degradation of mgtA transcripts in the presence of high Mg²⁺ concentrations. The data are interpreted to suggest that even if low concentrations of extracellular Mg²⁺ increase the PhoPQ-mediated transcription of mgtA, MgtA protein will be expressed only if intracellular Mg²⁺ concentrations are also low. Similar events appear to occur for mgtA in E. coli (20, 93). Nonetheless, reminiscent of current uncertainties over PhoPQ regulation, the physiological relevance of riboswitches and mRNA regulation by Mg²⁺ is unclear. Measurements of free intracellular Mg²⁺ in Salmonella serovar Typhimurium by using the dye Mag-Fura 2 have indicated that the free Mg²⁺ concentration is about 1 mM physiologically and does not appear to change rapidly (29). Thus, this direct Mg²⁺-mediated regulation of mRNA may not occur in free-living cells.

PhoPQ AND VIRULENCE

PhoPQ was first discovered in Salmonella serovar Typhimurium because of its role in virulence; phoP and phoQ mutants have survival deficits in macrophages and are less virulent in the mouse than wild-type strains (28, 30, 60). PhoP is also involved in macrophage cell death (23), in part by promoting the proper trafficking of the Salmonella-containing vacuole (34). PhoPQ regulates other virulence-related properties, such as resistance to host antimicrobial peptides, resistance to bile, proper biofilm formation, and the stabilization of RpoS levels (62, 77, 97, 98). Moreover, the surge in transcription caused by the surge in phospho-PhoP is necessary for virulence because a mutant constitutively expressing PhoP, which therefore lacks the surge in phosphorylation, is completely attenuated in the mouse (84). This mutant has a consensus −35 hexameric sequence replacing the PhoP box in the phoP promoter and thus is not a pho-24 mutant.

PhoPQ generally negatively regulates genes in SPI-1, as gene products repressed by PhoPQ promote epithelial cell invasion (7, 9). PhoPQ indirectly represses hilA expression; HilA is the master regulator of SPI-1 (7). The expression of components of SPI-1 was evaluated previously by reporter assays using a pho-24 mutant. SPI-1 transcription is repressed in the pho-24 mutant. This repression is not mediated through PhoPQ in response to pH (7). Recently, orgBC of SPI-1 was found to be activated directly by PhoP under PhoPQ-inducing conditions, as well as invasion-inducing conditions (1), presumably through a PhoP box lying just upstream of orgBC; these same genes are also regulated by HilA.

Controversy exists over whether PhoP and PhoQ are involved in regulating SPI-2. Investigators in some labs have shown that PhoPQ positively regulates genes in SPI-2 (11, 21, 22). Chromatin immunoprecipitation revealed that PhoP directly promotes the transcription of ssrB in macrophages by binding the ssrB promoter; SsrB-SsrA is a two-component system that regulates SPI-2 expression (11). Electrophoretic mobility shift assays and DNase I footprinting further indicated that PhoP binds to the ssrB promoter. PhoP regulates ssrA (spiR) at the posttranscriptional level through binding to the 5' UTR (11). Two-dimensional gel electrophoresis revealed that some of same proteins are altered in phoP, pho-24, and ssrB mutants (21); thus, the regulatory networks may overlap.

In a previous study, low extracellular Mg²⁺ levels induced the expression of SPI-2 components whereas high Mg²⁺ levels repressed it (22). However, transcription was measured at only one time point. By measuring the expression of SPI-2 components over time, Miao et al. (58) showed that Mg²⁺ limitation does not increase the expression of SPI-2 but that an alkaline pH represses SPI-2, as shown previously (49). An SPI-2 chaperone, SscA, is not present in a phoP mutant as measured by Western blotting after growth in a low-Mg²⁺ medium (22). SscA is not constitutively expressed in the pho-24 mutant, indicating that PhoPQ may indirectly regulate SPI-2 (22). The expression of ssaH, another SPI-2 effector, requires phoP only after the growth of the bacterium at low pH and low Mg²⁺ levels (49). Miao et al. attributed the lack of expression of SscA and ssaH to the severely limited growth of a phoP mutant in low-Mg²⁺ medium; the expression of sspH2, sscA, and ssaH is unaffected by the mutation of phoP when the bacterium is grown in medium containing higher levels of Mg²⁺ (22, 49, 58).

SpiC, an SPI-2 effector, and SsrB require PhoP for expression within macrophages (11). However, Miao et al. (58) argued that the phoP mutant and wild type should not be compared directly in vivo as they have differential growth phenotypes. Thus, the investigators measured sspH2 expression in macrophages in phoP and phoP ssrA mutant backgrounds. They found that the expression of sspH2 is not affected by phoP mutation. Further, the results of mouse competition experiments using mutations in SPI-2 genes or in phoP suggested that the PhoPQ and SsrB-SsrA regulatory networks work independently, because when both are mutated Salmonella attenuation is enhanced (10). In conclusion, PhoPQ is certainly a key player in Salmonella serovar Typhimurium pathogenesis, but its role and its regulation remain to be fully assessed.

MAGNESIUM TRANSPORT AND VIRULENCE

In addition to potential Mg²⁺ involvement in PhoPQ regulation, all three Mg²⁺ transport systems are associated with virulence as well. Salmonella serovar Typhimurium mgtA expression is induced upon entry into or invasion of J774 macrophage-like cells, RAW 264.7 macrophage-like cells, CMT93 epithelial cells, Hep-2 epithelial cells, and Henle-407 epithelial cells (42, 88). mgtCB expression is induced markedly upon entry into or invasion of mice, J774 macrophage-like cells, RAW 264.7 macrophage-like cells, MDCK epithelial cells, CMT-93 epithelial cells, Hep-2 epithelial cells, and Henle-407 epithelial cells (33, 42, 55, 88). The expression of corA within eukaryotic cells has not been measured. However, since the presence of the corA allele can significantly alter the expression of mgtA and mgtCB, the expression of corA may also be important (64).

An mgtA Salmonella serovar Typhimurium mutant does not have an invasion defect in any of the cell lines tested previously (88). The mutation of mgtA alone in the presence of wild-type alleles of corA and mgtB causes no apparent virulence defect. An mgtCB mutant strain exhibits no deficit in entry into J774 macrophage-like cells, although it is deficient for long-term survival (12, 88). After intraperitoneal injection into mice, an mgtCB mutant strain shows about a 2-log decrease in the 50% lethal dose compared to that of the wild type (12). This virulence effect has been attributed to mgtC, since a strain lacking only a functional mgtB allele is as virulent as the wild type. An mgtA mgtCB double mutant exhibits about a 3-log decrease in the 50% lethal dose compared to that of the wild type after intraperitoneal injection (12). Moreover, an mgtA mgtB double mutant is attenuated in the mouse after oral gavage (D. G. Kehres and M. E. Maguire, unpublished observations). Interestingly, the Mg²⁺ growth requirement and the invasion phenotype of mgtC mutant strains have recently been dissociated, although how each phenotype relates to putative MgtC functions is unknown (78). A corA mutant exhibits decreased invasion of Caco-2 epithelial cells (K. M. Papp-Wallace, M. Nartea, D. G. Kehres, S. Porwollik, M. McClelland, S. J. Libby, F. C. Fang, and M. E. Maguire, accepted for publication). Moreover, the lack of corA slightly decreases the survival of Salmonella within macrophages. These defects may be attributed partially to decreased expression of SPI-1 and SPI-2 in a corA mutant. A corA mutant is also attenuated in the mouse upon either oral administration or intraperitoneal injection. The regulation of transport by the CorA protein may also be important for Salmonella virulence (K. M. Papp-Wallace and M. E. Maguire, accepted for publication). In summary, Salmonella needs three functional Mg²⁺ transporters for full virulence.

POTENTIAL VIRULENCE-ASSOCIATED PHENOTYPES

Iron transport and homeostasis are associated with virulence. Iron is used by many enzymes to catalyze various redox reactions and thus is essential for life. However, too much iron is toxic, because it can catalyze the formation of harmful free radicals. Salmonella serovar Typhimurium has multiple siderophore systems for the uptake of insoluble ferric iron and FeoABC for the uptake of ferrous iron. Under most conditions, ferrous iron rapidly oxidizes to ferric iron, and thus, siderophore systems are the major pathway for iron uptake.

Salmonella serovar Typhimurium exhibits an Fe²⁺ hypersensitivity phenotype if either phoP or both mgtA and mgtCB are mutated (16). The increased Fe²⁺ sensitivity of the phoP strain is abolished by the mutation of corA. ⁶³Ni²⁺ uptake is reported to be increased in mgtA mgtCB and phoP mutants. The increased uptake is dependent on CorA and is apparently the result of an increase in the V _max without change in the K_m for Ni²⁺. Neither corA transcription nor the CorA protein level is reported to increase in the mgtA mgtCB and phoP mutants. The addition of cation hexaammines, selective CorA antagonists, also decreases the Fe²⁺ hypersensitivity phenotype of the phoP mutant. However, since the CorA protein does not transport Fe²⁺ (71), this Fe²⁺ sensitivity phenotype is apparently not a direct consequence of CorA transport. A corA mutation upregulates the transcription of one siderophore iron uptake system, enterochelin (Papp-Wallace et al., submitted). Moreover, the loss of corA results in a remodeling of the proteins present in the bacterial cell wall, which may be protective against Fe²⁺ toxicity. Thus, the abolishment of the Fe²⁺ hypersensitivity phenotype of the phoP and mgtA mgtCB mutants by the mutation of corA is most likely an indirect result and not due to CorA-mediated iron uptake.

The enzyme lactoperoxidase is produced as a defensive measure by immune cells; it oxidizes different substrates, primarily thiocyanate, by using hydrogen peroxide to generate toxic hypothiocyanite, which is antibacterial. Hypothiocyanite will oxidize sulfhydryl residues on proteins within the bacterial cytoplasmic membrane (4). These oxidizations will alter the function of these membrane proteins, which is detrimental to bacterial function. These oxidations can be either bacteriostatic or bactericidal.

Exposure to the host lactoperoxidase system induces the expression of several genes in E. coli, among them corA (82, 83). Both E. coli and Salmonella serovar Typhimurium corA mutants are hypersensitive to lactoperoxidase but not to direct hydrogen peroxide challenge or to superoxide generated by the addition of plumbagin. The addition of Mg²⁺ does not alter the response, but the addition of Ni²⁺ increases the lactoperoxidase-induced killing of the wild type but not an E. coli corA mutant, strongly suggesting that the ability of CorA to transport Ni²⁺ (or Co²⁺) mediates or influences the lactoperoxidase sensitivity. The mechanism by which lactoperoxidase exposure induces corA expression has not been investigated. However, corA is essential to protect Salmonella from the host lactoperoxidase system. Conversely, if CorA transports Ni²⁺, then corA causes Salmonella to be more susceptible to the host lactoperoxidase system. Thus, if CorA is present, but not transporting Ni²⁺, then Salmonella is protected from the host lactoperoxidase system.

SUMMARY AND UNANSWERED QUESTIONS

This chapter has reviewed what is known to date about the transport of Mg²⁺ and the regulation of Mg²⁺ transport in E. coli and Salmonella serovar Typhimurium. Both bacteria express corA, the primary Mg²⁺ transporter found throughout bacteria and archaea. They also possess Mgt P-type ATPases: MgtA in E. coli and MgtA and MgtB in Salmonella serovar Typhimurium. The expression of these mgt transporters occurs through a PhoPQ-dependent manner in response to pH, antimicrobial peptides, and low Mg²⁺ concentrations. PhoPQ is a two-component system: PhoQ is the sensor kinase, and PhoP is the transcription activator. Upon encountering a stimulus, such as low Mg²⁺ levels, PhoQ will autophosphorylate and phosphorylate PhoP; in response, around 100 genes are either activated or repressed. In addition, in Salmonella serovar Typhimurium, mgtA transcription is regulated by intracellular-Mg²⁺ control of mRNA structure. It is unknown whether mgtCB or any other PhoPQ-regulated genes also respond to intracellular [Mg²⁺] in this way.

The crystal structure of T. maritima CorA has been resolved, thus opening the door for further mutagenesis/structure/function studies. Many questions about the mechanism of transport of CorA remain, but given the recent structural information, many will most likely be answered. Are N314, M302, L294, and M291 functioning as gates to block the passage of Mg²⁺? Are there interactions between the ring of positive charge at the cytosolic membrane interface and the ring of negative charge created by the ends of the willow helices which influence the mechanism of transport? Moreover, are the five Mg²⁺ ions bound between D89 and D253 of adjacent monomers involved in the mechanism as well, since mutating D89 and D253 can close the conformation of CorA? How does efflux occur, and what role does it play physiologically? How does CorA recognize Mg²⁺ when the channel is a homopentamer and Mg²⁺ is invariably hexacoordinate? These and several more questions need to be answered in order to determine how cations get moved via CorA.

Further, the full role of CorA in general bacterial physiology remains to be determined. The mutation of corA and the inhibition of CorA via selective antagonists can abolish the Fe²⁺ sensitivity of the phoP and mgtA mgtCB mutants, even though CorA does not transport Fe²⁺. Interestingly, the V _max of CorA is dramatically increased in these mutant backgrounds: why? Moreover, exposure to lactoperoxidase induces the expression of corA, and corA mutants are hypersensitive to killing by lactoperoxidase. The mechanism and physiological relevance of these observations are yet to be determined. These phenotypes may be relevant to the attenuation of the corA mutant upon the infection of mice, but the role of corA in virulence remains to be determined. Could the regulation of CorA be the key? Mg²⁺ transport is definitely correlated with virulence, but how and why is it important given that Mg²⁺ levels in the host are sufficient for Salmonella survival?

The role of MgtC is unknown; it can constitutively activate the Na⁺, K⁺-ATPase in Xenopus oocytes, but what is the significance of this effect? Is MgtC involved in Mg²⁺ homeostasis, given that the corresponding gene occupies the upstream position in a two-gene operon with mgtB, encoding a known Mg²⁺ transporter? Why is mgtC necessary for the survival of Salmonella within macrophages? And why does MgtC have a dual role in bacteria?

Many mysteries involving PhoPQ remain as well. How does sensing by PhoQ in the host occur? Is it pH, antimicrobial peptides, or a divalent cation that triggers PhoPQ activation? There appear to be several different mechanisms by which PhoP interacts with the promoters of the genes it regulates. Conflict is present in the literature about which promoters PhoP can bind to. Moreover, is the phosphorylation of PhoP necessary for binding to its promoters? While PhoPQ generally positively regulates SPI-2 and negatively regulates SPI-1, there is overlap, and the role of this differential regulation is unknown. Do PhoPQ and SPI-2 act independently of each other? Many questions about PhoPQ, virulence, and the role of Mg²⁺ remain unanswered.