Fine-tuning of Substrate Affinity Leads to Alternative Roles of Mycobacterium tuberculosis Fe2+-ATPases*

Little is known about iron efflux transporters within bacterial systems. Recently, the participation of Bacillus subtilis PfeT, a P1B4-ATPase, in cytoplasmic Fe2+ efflux has been proposed. We report here the distinct roles of mycobacterial P1B4-ATPases in the homeostasis of Co2+ and Fe2+. Mutation of Mycobacterium smegmatis ctpJ affects the homeostasis of both ions. Alternatively, an M. tuberculosis ctpJ mutant is more sensitive to Co2+ than Fe2+, whereas mutation of the homologous M. tuberculosis ctpD leads to Fe2+ sensitivity but no alterations in Co2+ homeostasis. In vitro, the three enzymes are activated by both Fe2+ and Co2+ and bind 1 eq of either ion at their transport site. However, equilibrium binding affinities and activity kinetics show that M. tuberculosis CtpD has higher affinity for Fe2+ and twice the Fe2+-stimulated activity than the CtpJs. These parameters are paralleled by a lower activation and affinity for Co2+. Analysis of Fe2+ and Co2+ binding to CtpD by x-ray absorption spectroscopy shows that both ions are five- to six-coordinate, constrained within oxygen/nitrogen environments with similar geometries. Mutagenesis studies suggest the involvement of invariant Ser, His, and Glu residues in metal coordination. Interestingly, replacement of the conserved Cys at the metal binding pocket leads to a large reduction in Fe2+ but not Co2+ binding affinity. We propose that CtpJ ATPases participate in the control of steady state Fe2+ levels. CtpD, required for M. tuberculosis virulence, is a high affinity Fe2+ transporter involved in the rapid response to iron dyshomeostasis generated upon redox stress.

Iron is an essential micronutrient required for numerous biological processes as it is used as a prosthetic group by several different enzymes (1,2). However, in excess, it can be toxic due to its participation in Fenton chemistry and potential mismetallation in non-iron-containing metalloproteins. In this context, damage of iron-sulfur centers and mononuclear iron enzymes produced by various redox stresses are particular con-tributors to iron dyshomeostasis and consequent toxicity (3)(4)(5)(6). Characterization of bacterial Fe 2ϩ homeostasis has mainly been focused in mechanisms of uptake (by divalent metal, siderophore, and heme transporters), transcriptional regulation (by Fur and IdeR systems), and Fe 2ϩ sequestration (by bacterioferritin and Dps proteins) (2,(7)(8)(9). Nevertheless, studies have suggested that cation diffusion facilitators and iron-citrate transporters participate in Fe 2ϩ efflux (10 -12). We recently observed that Bacillus subtilis PfeT, a P 1B4 -ATPase, confers Fe 2ϩ tolerance (13). PfeT is expressed under the control of PerR in response to peroxide exposure (14). Initial biochemical characterization showed that Fe 2ϩ activates isolated PfeT ATPase, leading to a higher V max than generated by Co 2ϩ , which is the proposed substrate of P 1B4 -ATPases (13,(15)(16)(17). Interestingly, phenotypic analysis of Listeria monocytogenes lacking the P 1B4 -ATPase FrvA showed a role of this ATPase in resistance to heme toxicity (18). These observations suggest a significant role of this subfamily of P-type ATPases in Fe 2ϩ homeostasis (13,14). P 1B4 -ATPases present in prokaryotes and plant chloroplasts are part of the large family of P-type ATPases (15,19,20). P-type ATPases are polytopic membrane proteins that transport a variety of ions using the energy provided by ATP hydrolysis (21)(22)(23). The P 1B subgroup includes proteins responsible for the efflux of cytoplasmic transition metals including Cu ϩ , Zn 2ϩ , Co 2ϩ , and Ni 2ϩ (19,22,23). The specificity of their transmembrane metal binding sites (TM-MBSs) 2 is determined by invariant amino acid sequences in their last three transmembrane segments (TMs) (17, 19, 24 -26). However, activation by non-cognate substrates has been reported for most P 1B -ATPase subgroups (22,27). In particular, activation of P 1B4 -ATPases by Co 2ϩ , Ni 2ϩ , Ca 2ϩ , Cu ϩ , Zn 2ϩ , and Cd 2ϩ has been proposed (15-17, 28 -30). We previously reported in vivo and in vitro functional studies directed at understanding the metal selectivity and consequent physiological roles of mycobacterial P 1B4 -ATPases (15,16). The presence of one or two P 1B4 -ATPasecoding genes in mycobacterial species enabled comparative studies of Mycobacterium smegmatis CtpJ and Mycobacterium * This work was supported by National Institutes of Health Grant DK068139 (to T. L. S.). The authors declare that they have no conflicts of interest with the contents of this article. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
tuberculosis CtpJ and CtpD. In vitro, MsCtpJ and MtCtpJ display a higher activation by Co 2ϩ and Ni 2ϩ compared with Zn 2ϩ , although equilibrium binding affinities show K D values for Zn 2ϩ Ͻ Co 2ϩ ϭ Ni 2ϩ (15,16). In vivo, ctpJ expression is induced by Co 2ϩ , whereas mutant strains show accumulation and sensitivity to the metal. On the contrary, the expression of the homologous MtctpD is not induced by Co 2ϩ but rather by redox stress. Mutation of MtctpD does not lead to Co 2ϩ sensitivity or higher intracellular levels of this metal. Nevertheless, MtCtpD ATPase activity is partially activated by Co 2ϩ . Surprisingly, MtCtpD but not MtCtpJ is required for M. tuberculosis virulence.
Previous studies have not explored the activation of mycobacterial P 1B4 -ATPases by Fe 2ϩ . Could a differential activation by Co 2ϩ /Fe 2ϩ explain the presence of paralogous genes in M. tuberculosis? Why is MtCtpD but not MtCtpJ required for virulence? To address these questions, we examined the activation of M. smegmatis and M. tuberculosis P 1B4 -ATPases by Fe 2ϩ and their participation in Fe 2ϩ homeostasis and stress response. In addition, we explored the molecular basis of the different Fe 2ϩ and Co 2ϩ -ATPase activities by determining the coordination of these metals during transport by MtCtpD.
Iron and Hemin Sensitivity Tests-Liquid LIMM cultures of M. smegmatis mc2155, M. tuberculosis H37Rv, mutant, and complemented strains were inoculated at 0.05 A 600 from late exponential phase cultures and supplemented with the desired concentration of FeCl 3 or hemin (Sigma). A hemin stock solution was prepared at 25 mg/ml in 1.4 M NaOH. Cells were incubated for 16 h (M. smegmatis) or 5 days (M. tuberculosis), and A 600 was measured. To avoid hemin interference in A 600 readings, cells grown in hemin-containing medium were collected, washed twice with LIMM, and suspended in the original LIMM volume, and A 600 was measured.
Streptonigrin Sensitivity Tests-M. smegmatis mc2155, M. tuberculosis H37Rv, mutant, and complemented strains grown in LIMM to midlog phase were diluted to 0.05 A 600 in LIMM. The cultures were supplemented with 1 g ml Ϫ1 streptonigrin (STN) and 10 M FeCl 3 as indicated in the figures. Cells were incubated for 16 h (M. smegmatis) or 5 days (M. tuberculosis), and A 600 was measured.
Metal Accumulation Assays-Liquid LIMM cultures in midexponential phase (A 600 ϳ 1.0) were supplemented with increasing concentrations of FeCl 3 and incubated for 4 (M. smegmatis) or 8 h (M. tuberculosis). After this incubation, cells were harvested and washed with 5 mM EDTA and 0.9% NaCl. Aliquots were taken for protein determinations (32). Pellets were acid-digested with 0.5 ml of NO 3 H for 1 h at 80°C and then overnight at 20°C. Digestions were concluded by adding 1 ⁄ 8 volume of 30% (v/v) H 2 O 2 followed by a 1:5 dilution with water. Metal contents in digested samples were measured by AAS.
Fe 2ϩ Binding to Proteins-Metal binding to isolated enzymes was measured as described previously (15,36). Five micromolar His-less enzyme was incubated for 1 min at 4°C in 25 mM HEPES-NaOH, pH 7.5, 150 mM NaCl, 1 mM tris(2-carboxyethyl)phosphine (TCEP), and 20 M either FeSO 4 or CoCl 2 . Excess metal was removed by washing in a 30-kDa-cutoff Centricon filtration device. Protein samples were acid-digested as described above, and metal concentrations were measured using AAS.
Metal binding affinities were determined using the divalent metal-binding chromophore mag-fura-2 (Invitrogen) (15,36). Five micromolar His-less protein and 10 M mag-fura-2 were titrated with 1 mM Fe 2ϩ or Co 2ϩ in Chelex-treated buffer (25 mM HEPES-NaOH, pH 7.5, 150 mM NaCl, and 1 mM TCEP where is the molar ratio of metal bound to protein and n is the apparent stoichiometry (37). Reported errors for K D and n are asymptotic standard errors provided by the fitting software KaleidaGraph (Synergy, Reading, PA).
ATPase Assays-ATPase assays were performed as described (15,24,34). The assay mixture contained 50 mM Tris, pH 7.4, 50 mM NaCl, 3 mM MgCl 2 , 3 mM ATP, 0.01% asolectin, 0.01% dodecyl ␤-D-maltoside, 2.5 mM TCEP, 20 g/ml purified protein, and freshly prepared transition metal ions at the desired concentrations. Fe 3ϩ was added as FeCl 3 , Cu 2ϩ was added as CuSO 4 , and in both cases TCEP was not included in the assay medium. Cu ϩ was obtained by including TCEP with CuSO 4 salt. Fe 2ϩ and Zn 2ϩ were included in the assay medium as the sulfate salts, whereas Co 2ϩ , Ni 2ϩ , and Mn 2ϩ were included as their chloride salts. ATPase activity was stopped after a 20-min incubation at 37°C, and released P i was determined (38). ATPase activity measured in the absence of transition metals was subtracted from plotted values. Curves of ATPase activity metal dependence were fit to The reported standard errors for V max and K 1/2 are asymptotic standard errors reported by the fitting software KaleidaGraph.
X-ray Absorption Spectroscopy (XAS)-XAS samples were loaded in a Coy anaerobic chamber at a 1:1 metal:protein molar ratio (1.2 mM Fe 2ϩ or 0.43 mM Co 2ϩ ). Sample was injected into a Kapton-wrapped Lucite cell, flash frozen, and stored in liquid nitrogen. XAS data were collected at the Stanford Synchrotron Radiation Lightsource on beamline 7-3 equipped with a Si(220) double crystal monochromator with a harmonic rejection mirror. Fluorescence spectra were collected using a 30-element germanium solid-state detector (Canberra, Meriden, CT). During data collection, the continuous flow liquid helium cryostat (Oxford Instruments, Concord, MA) was stabilized at 10 K. Iron and cobalt data were collected using a 6-m magnesium or a 3-m iron filter, respectively, placed between the cryostat and the detector to reduce unas-sociated scattering. Iron and cobalt foil spectra were collected simultaneously with protein data for direct energy calibration of the data. The first inflection points for iron and cobalt were set at 7111.3 and 7709.5 eV, respectively. Iron XAS spectra were recorded using 5-eV steps in the pre-edge regions (6900 -7094 eV), 0.25-eV steps in the edge regions (7095-7135 eV), and 0.05-Å Ϫ1 increments in the extended x-ray absorption fine structure (EXAFS) region (to k ϭ 13.5 Å Ϫ1 ), integrating from 1 to 20 s in a k 3 -weighted manner. Cobalt XAS spectra were recorded using 5-eV steps in the pre-edge regions (7542-7702 eV), 0.25-eV steps in the edge regions (7702-7780 eV), and 0.05-Å Ϫ1 increments in the extended EXAFS region (to k ϭ 13.5 Å Ϫ1 ), integrating from 1 to 25 s in a k 3 -weighted manner. A total of eight scans were taken for each sample, and these were then averaged.
XAS spectra were processed and analyzed using the EXAF-SPAK program suite for Macintosh OSX (39). A Gaussian function was used in the pre-edge region, and a three-region cubic spline was used in the EXAFS region. EXAFS data were converted to k space using E 0 values of 7130 and 7745 eV for iron and cobalt, respectively. Spectra were simulated using single and multiple scattering amplitude and phase functions generated using the Feff v8 software integrated within EXAFSPAK. Single scattering models were calculated for oxygen, nitrogen, and carbon to simulate possible iron-or cobalt-ligand environment. Calibrated scale factors and model E 0 values were not allowed to vary during fitting; the scale factor for iron was 0.95, and that for cobalt was 0.98. Iron data were fit out to a k value of 13.5 Å Ϫ1 . Calibration from Fe 2ϩ and Fe 3ϩ model compounds was used for determination of E 0 and scale factor parameters for iron. E 0 values for Fe-O and Fe-C were set at Ϫ10 eV. Cobalt data were fit out to a k value of 13.5 Å Ϫ1 . Calibration from Co 2ϩ model compounds was used for determination of E 0 and scale factor parameters. E 0 values for Co-O and Co-C were set at Ϫ11.3 eV. EXAFS spectra were simulated using both filtered and unfiltered data; however, simulation results are presented only for fits to unfiltered (raw) data. Simulation protocols and criteria for determining the best fit have been described elsewhere (25).

Results
Mycobacterial P 1B4 -ATPases Confer Fe 2ϩ Tolerance-We previously reported the activation of mycobacterial P 1B4 -ATPases by Co 2ϩ (15,16). However, M. tuberculosis CtpJ and CtpD appear to have distinct roles in Co 2ϩ tolerance and cellular response to redox stress (16). We recently observed that B. subtilis PfeT, a P 1B4 -ATPase in the PerR regulon, transports and confers tolerance to Fe 2ϩ in addition to Co 2ϩ (13). Similarly, the L. monocytogenes P 1B4 -ATPase, FrvA, confers resistance to heme toxicity (18). Thus, we hypothesized that selective activation by Fe 2ϩ might explain the presence of the ctpJ and ctpD paralogs in the M. tuberculosis genome. To test this idea, the capability of mycobacterial P 1B4 -ATPases to confer tolerance to Fe 2ϩ was assessed. Fig. 1 shows the growth of Ms⌬ctpJ, Mt⌬ctpD, Mt⌬ctpJ, and Mt⌬ctpD:⌬ctpJ double mutant strains in LIMM supplemented with different concentrations of FeCl 3 or hemin. The Ms⌬ctpJ strain showed a growth defect at high FeCl 3 or hemin as compared with M. smegmatis WT (Fig. 1, A and B). The complemented Ms⌬ctpJ strain, carrying the plasmid pMV306 harboring the full-length MsctpJ gene under the regulation of its native promoter, showed similar growth as the M. smegmatis WT. A comparable deficiency was observed in the Mt⌬ctpJ strain grown in the presence of 1 mM FeCl 3 (Fig. 1D). However, this mutant strain was not affected by the presence of hemin in the medium. On the contrary, mutation of the MtctpD gene significantly affected the growth in both FeCl 3 -and hemin-supplemented medium. Interestingly, the Mt⌬ctpD:⌬ctpJ double mutant strain showed a behavior identical to that of the Mt⌬ctpD strain. In all cases, complemented strains showed the growth phenotype of M. tuberculosis WT (Fig. 1, D and E).
The data indicate that MsCtpJ and MtCtpD confer iron tolerance when cells are exposed to relatively high metal levels.
Exploring their role at lower iron levels, the sensitivity to STN was tested. STN is a quinone antibiotic whose activity is correlated with intracellular iron availability (40). The Ms⌬ctpJ, Mt⌬ctpD, Mt⌬ctpJ, and Mt⌬ctpD:⌬ctpJ mutant strains displayed a significantly increased STN sensitivity in LIMM supplemented with 1 g ml Ϫ1 STN and only 10 M FeCl 3 (Fig. 1, C  and F). In contrast to their distinct tolerance to high Fe 3ϩ in the medium, there was no significant difference in the sensitivity of Mt⌬ctpD and Mt⌬ctpJ strains to STN-Fe 2ϩ .
These results suggest that to different extents mycobacterial P 1B4 -ATPases contribute to Fe 2ϩ homeostasis by driving this metal efflux. To further explore this hypothesis, Ms⌬ctpJ, Mt⌬ctpD, Mt⌬ctpJ, and Mt⌬ctpD:⌬ctpJ mutant strains were challenged with sublethal concentrations of FeCl 3 , and the resulting cellular Fe 2ϩ levels were determined. Consistent with the iron sensitivity phenotypes (Fig. 1), Fe 2ϩ accumulation was observed in mutant strains (Fig. 2, A and B). Iron levels in the Ms⌬ctpJ strain were approximately 5 times higher than those in M. smegmatis WT (Fig. 2A). The partial recovery observed in the Ms⌬ctpJ mutant strain complemented with MsctpJ appears to be associated with lower levels of transcript (35% of WT; not shown). Reinforcing the predominant role of MtCtpD in Fe 2ϩ homeostasis, a significant increase of Fe 2ϩ content was observed in the Mt⌬ctpD strain, whereas 50% smaller changes were observed in the Mt⌬ctpJ strain. Similar Fe 2ϩ accumulation was observed in the Mt⌬ctpD and Mt⌬ctpD:⌬ctpJ double mutant strains (Fig.  2B). These results suggest that although mycobacterial CtpJs are involved in controlling Co 2ϩ levels they also participate in Fe 2ϩ efflux, particularly when they are the only P 1B4 -ATPase in the organism as in M. smegmatis. In contrast, MtCtpD appears to play a dominant role in maintaining the cytoplasmic Fe 2ϩ level in this organism.
Distinct Biochemical Properties of Mycobacterial P 1B4 -ATPase-P-ATPases couple the transmembrane transport of their substrate to ATP hydrolysis following the Albers-Post E1/E2-like mechanism (23). Consequently, the metal dependence of ATPase activity provides a starting point to analyze substrate selectivity. Previous reports showed that MsCtpJ, MtCtpD, and MtCtpJ are differently activated by Co 2ϩ , Ni 2ϩ , and to a lesser extent Zn 2ϩ (15,16). The activation of mycobacterial P 1B4 -ATPases by Fe 2ϩ/3ϩ was tested using purified proteins stabilized in lipid/detergent micelles. All three proteins were strongly activated by Fe 2ϩ and only minimally by Fe 3ϩ (Fig. 3). For comparison, activation by Co 2ϩ , Ni 2ϩ , and Zn 2ϩ at 0.1 and 1 mM concentrations is shown. MtCtpD Fe 2ϩ -dependent activity was ϳ2-fold higher than those observed in MtCtpJ and MsCtpJ (Table 1 and Fig. 3D) and quite similar to that of B. subtilis PfeT (3.25 Ϯ 0.21 mol/mg/h) (13). MtCtpD also showed significant activation at 1 mM Zn 2ϩ (Fig. 3B). Zn 2ϩ binding to P 1B4 -ATPases as well as Zn 2ϩ transport has been reported (15,30). The K 1/2 for Fe 2ϩ activation of the mycobacterial enzymes confirmed a tendency observed in B. subtilis PfeT: the larger activation by Fe 2ϩ is associated with a K 1/2 much larger than that of Co 2ϩ (Table 1). However, the observed K 1/2 values do not describe the selectivity to the enzymes. These parameters result from the k on /k off of the metals binding the cytoplasmic facing transmembrane sites and the k on /k off for the release/backward binding of the metal to the periplasmic facing sites (41). As shown below, equilibrium binding determinations of K D better report the relative selectivity for the activating metals.
The described Fe 2ϩ -ATPase activities require the binding of the transported substrate to the TM-MBS. The stoichiometry of this interaction was verified by measuring Fe 2ϩ binding to MsCtpJ, MtCtpD, and MtCtpJ in non-turnover conditions lacking ATP ( Table 1). The His 6 -less enzymes were incubated with excess Fe 2ϩ , unbound metal was removed by filtration, and bound metal was quantified by AAS. As expected, the proteins bind Fe 2ϩ in a 1:1 molar ratio. Discarding the possibility of nonspecific interactions, metal binding was largely abolished in the presence of 1.5 mM vanadate (not shown). This binding stoichiometry is similar to that previously observed for Zn 2ϩ , Ni 2ϩ , and Co 2ϩ binding to P 1B4 -ATPases (15,17). Notably, although the TM-MBSs of these enzymes appear to accommodate divalent cations, no significant binding of Fe 3ϩ was observed (not shown).
Mycobacterial P 1B4 -ATPase affinities for Fe 2ϩ and Co 2ϩ under equilibrium conditions were determined by titration of isolated enzymes in the presence of the fluorescence indicator mag-fura-2 (15,42). In these experiments, mag-fura-2 forms 1:1 indicator-metal complexes of known K D . The concentration of free indicator can be spectrophotometrically monitored, and the free metal and metal-protein complex levels can be calculated. The enzyme-metal K D and the apparent stoichiometry of the interactions were obtained by fitting mag-fura-2 A 366 versus free metal concentration curves (Table 1). MsCtpJ and MtCtpJ showed a similar K D for Fe 2ϩ . These were also comparable with those previously reported for Co 2ϩ (included in Table 1 for comparison). Notably, MtCtpD has ϳ3-fold higher affinity for Fe 2ϩ compared with Co 2ϩ . Moreover, the affinity of MtCtpD is 20 times higher for Fe 2ϩ (lower K D ) and 5 times higher for Co 2ϩ when compared with those observed in the CtpJ enzymes. The relative preference of MtCtpD for Fe 2ϩ when compared with MtCtpJ further supports a dominant role of MtCtpD in Fe 2ϩ tolerance (Fig. 1).
Distinct Co 2ϩ and Fe 2ϩ Coordination by MtCtpD-Full appreciation of the different enzymatic activities and metal selectivity observed in P 1B4 -ATPases requires understanding of the structural basis of these phenomena. P 1B4 -ATPases share a number of invariant residues in the transmembrane region proposed to participate in metal coordination (19, 24 -26). A sixcoordinate Co 2ϩ species by the Sulfitobacter sp. P 1B4 -ATPase has been postulated with participation of a Ser in the conserved SPC in the fourth TM and invariant His, Glu, and Thr in the sixth TM (17). Surprisingly, this coordination did not include the invariant Cys located in the fourth transmembrane segment of all P 1B -ATPases.
Considering the results shown above and the possible distinct Fe 2ϩ /Co 2ϩ coordination, the binding environment of Fe 2ϩ and Co 2ϩ in MtCtpD was analyzed by XAS. The x-ray absorption near edge spectroscopy (XANES) portion of the XAS spectrum is element-specific and local bonding-sensitive; therefore it is useful for reporting the oxidation and coordination states of metals bound to the enzyme. The spectra of Fe 2ϩloaded protein were compared with Fe 2ϩ and Fe 3ϩ model systems (Fig. 4A). The first inflection point energy for protein-bound iron occurs at 7127.6 eV, consistent with a 50%/50% Fe 2ϩ /Fe 3ϩ redox state mixture. Although all spectra were closely screened for photoreduction, iron oxidation during protein concentration postmetal loading may have led to this observation. Pre-edge features observed in the XANES of iron-MtCtpD are characteristic of 1s 3 3d electronic transitions. These pre-edge features are consistent with pseudosymmetric six-coordinate iron-ligand systems (43). Cobalt-MtCtpD XANES pre-edge features and the general edge structure are consistent with Co 2ϩ bound to protein systems in a six-coordinate ligand environment as observed previously (Fig. 4B) (17,  44). The EXAFS region of an XAS spectrum provides high res-   (Fig. 3D). b Values were previously reported (15,16). olution distances for ligand coordination environments of metals bound to metalloproteins. Fourier transform of the EXAFS provides a pseudoradial distribution function of ligand environments surrounding the metal. EXAFS analysis was used to compare differences in coordination of iron and cobalt to MtCtpD (Fig. 5). The resulting coordination number, ligand identity, bond lengths, and statistical fitting parameters are described in Table 2. The iron-MtCtpD spectrum was best simulated by a ligand environment containing five oxygen/nitrogen ligands at two sets of coordinating distances. Long range iron-ligand scattering was best fit with four Fe-C ligands. The general features in the iron EXAFS and bond lengths obtained from the simulations suggest that iron is most likely coordinated by six oxygen/nitrogen-based side chain ligands from amino acids and water molecules. The cobalt-MtCtpD EXAFS served as a comparison for the iron bound to MtCtpD. The best fit simulation contained two unique Co-O/N environments ( Table 2). The Co-O/N bond lengths, compared with crystallographically characterized model compounds in the Cambridge Structural Database, are again consistent with five-to six-coordinate Co-(O/N) 6 compounds (45). This is similar to what has been observed in the Sulfitobacter sp. P 1B4 -ATPase (17). The spectroscopy analysis points to a distinct coordination for Co 2ϩ and Fe 2ϩ . However, the spectroscopy is not able to reveal alternative ligands. Conserved residues of MtCtpD possibly involved in metal coordination are Ser-316 and Cys-318 in TM4 and His-642, Glu-643, Gly-644, Ser-645, and Thr-646 in TM6. Seeking a more detailed understanding of the Fe 2ϩ and Co 2ϩ coordination at the TM-MBS, residues likely involved in the metal coordination were exchanged by site-directed mutagenesis, and the resulting proteins were functionally characterized. The single mutants E643D and T646S (included as conservative control modifications) and G644A and S645A did not alter the Co 2ϩ -or Fe 2ϩ -ATPase activities of MtCtpD (Fig.  6). In agreement with previous reports, mutation of S316A, H642A, E643A, or T646A led to significant loss of Co 2ϩ ATPase activity (17). The single substitutions S316A, E643A, and T646A equally affected the Co 2ϩ -and Fe 2ϩ -ATPase activ-ities. However, the H642A mutation largely abolished Co 2ϩ stimulation (6% of WT activity) while preserving significant Fe 2ϩ sensitivity (33%) at saturating metal concentrations. In contrast, the C318A mutation had diminished Fe 2ϩ activation (18%) while retaining 39% of the Co 2ϩ -ATPase activity. The differential effects of H642A and C318A mutations on the ATPase activity point toward a plausible mechanism. It appears that CtpD differentiates Co 2ϩ and Fe 2ϩ as substrates perhaps via alternative coordination despite binding these ions with quite similar affinities.
Considering that the ATPase activities might be affected by the removal of a ligating group or by the inability to undergo structural changes required for transport, the capability of the C318A and H642A mutant proteins to bind Co 2ϩ and Fe 2ϩ was tested. In comparison, the C318A mutant showed significantly lowered MtCtpD affinity for Fe 2ϩ , but it did not change the binding affinity for Co 2ϩ , suggesting an important role of this conserved Cys in Fe 2ϩ binding ( Fig. 7 and Table 3). The critical role of Cys-318 was further confirmed by the determination of the metal binding by AAS after incubation of the C318A mutant protein with metals at concentrations 10 times over the observed K D . The C318A mutant protein was able to bind 1.05 Ϯ 0.14 Co 2ϩ but only 0.32 Ϯ 0.06 Fe 2ϩ . Finally, the H642A mutant had no detectable effect on Fe 2ϩ or Co 2ϩ binding affinities compared with WT (Table 3), raising the possibility that His-642 has a role other than the direct participation in the TM-MBS.

Discussion
The substrates and consequent functional roles of bacterial and eukaryotic P 1B4 -ATPases have remained elusive as their capabilities to transport different transition metals have been reported (15-17, 28 -30). This is relevant as some of these transporters are required for bacterial virulence (16,18) and critical for metal homeostasis in chloroplasts (29,46). An extra layer of complexity is present in bacterial systems containing homologous non-redundant P 1B4 -ATPases (16). From a different perspective, defining P 1B4 -ATPases substrates is significant to understanding the coordination of transition metals by transport proteins as well as likely additional selectivity mechanisms acting in vivo. Here we describe the roles of mycobacterial P 14B -ATPases in Fe 2ϩ homeostasis, the kinetics of transport, and the structural elements that in part determine the selectivity of these enzymes. These results indicate that mycobacterial P 14B -ATPases are Fe 2ϩ /Co 2ϩ -ATPases; however, the various isoforms show differential participation in the homeostasis of these ions.

Mycobacterial P 1B4 -ATPases Participates in Both Fe 2ϩ and
Co 2ϩ Homeostasis-We observed that mycobacterial CtpJ proteins contribute to the homeostasis of Fe 2ϩ and Co 2ϩ to different extents. M. smegmatis has a single P 1B4 -ATPase, MsCtpJ. Expression of the coding gene is induced by Co 2ϩ and partially by the superoxide generator paraquat but not by H 2 O 2 (15). Deletion of MsctpJ leads to lower tolerance to Co 2ϩ , Fe 2ϩ , and hemin as well as increments in intracellular Co 2ϩ and Fe 2ϩ  levels ( Fig. 1) (15). The M. tuberculosis genome encodes two P 1B4 -ATPases. MtCtpJ expression, like MsCtpJ, is induced by Co 2ϩ and to a lesser extent by redox stressors and Fe 2ϩ (16). The Mt⌬ctpJ strain accumulates higher levels of both Co 2ϩ and Fe 2ϩ (Fig. 1) (16), again in a fashion similar to that of the Ms⌬ctpJ strain. These characteristics appear similar to those observed for PfteT, the single P 1B4 -ATPase present in B. subtilis (13). The comparable functions suggested by the observed phenotypes correlate with the analogous biochemistry of MsCtpJ, MtCtpJ, and B. subtilis PfteT. These three ATPases transport Fe 2ϩ and Co 2ϩ with surprisingly similar V max and K 1/2 for activation. Moreover, both CtpJs bind Fe 2ϩ and Co 2ϩ with micromolar affinities (equilibrium binding determinations have not been performed for PfeT). These affinities explain the observed capability of these enzymes to influence the cellular response to STN when extracellular iron is maintained at just 10 M. More importantly, 2-3 M Fe 2ϩ affinities appear consistent with reported free Fe 2ϩ levels in the 1-10 M region (4). In fact, the iron-sensing transcriptional regulators Fur and IdeR have 9 M K D for Fe 2ϩ (47,48), indicating that these regulators are sensitive to the same concentration of Fe 2ϩ as the P 1B4 -ATPases. Consequently, efflux CtpJ ATPases and influx transporter regulators are likely to coordinately respond to changes in metal levels not only under Fe 2ϩ stress conditions but also under normal conditions. Homologous P 1B4 -ATPases Present in Mycobacterial Genomes Have Distinct Roles-M. tuberculosis, as in other mycobacteria, has an additional P 1B4 -ATPase, CtpD. Notably, MtCtpD, but not MtCtpJ, is required for bacterial virulence. What unique function does CtpD provide? The Mt⌬ctpD strain is more sensitive to iron stress and accumulates higher levels of this metal than Mt⌬ctpJ. The phenotypic differences between the Mt⌬ctpD and Mt⌬ctpJ strains should have a molecular basis either in the biochemistry of these enzymes or the iron pool that they transport. Notably, the phenotype of the Mt⌬ctpD:⌬ctpJ double mutant strain is similar to that of the Mt⌬ctpD cells, suggesting that CtpD and CtpJ use the same iron pool as a substrate, and this can be controlled by MtCtpD alone. Alternatively, the molecular activities of MtCtpD and MtCtpJ appear distinct. MtCtpD has significantly higher Fe 2ϩ -ATPase activity. Moreover, if the relative activation induced by Fe 2ϩ /Co 2ϩ is considered, MtCtpD shows Fe 2ϩ activity 12 times larger than that generated by Co 2ϩ . On the contrary, MtCtpJ shows higher activation by Co 2ϩ than MtCtpD and only a 1.5 Fe 2ϩ -ATPase: Co 2ϩ -ATPase ratio. Although these relative activities approximately correlate with the observed phenotypes, the higher affinity of MtCtpD for Fe 2ϩ (K D of 0.1 M) appears to confer its dominant role in Fe 2ϩ homeostasis. This K D is 1 order of magnitude smaller than that reported for Fe 2ϩsensing transcriptional regulators of influx systems (47,48). Distinct from CtpJ, CtpD is not induced by divalent metals but by redox stressors, such as the nitric oxide generator nitroprusside and the respiratory poison cyanide (16). In fact, the region upstream of ctpD contains the TTG XXXXTTCXXG operator sequence for the redox-sensing MtFurA regulator (49). Considering the release of Fe 2ϩ from iron-sulfur and mononuclear iron-containing proteins upon redox stress, it can be hypothesized that CtpD constitutes an early response to Fe 2ϩ dyshomeostasis that is independent of efflux (CtpJ), storage (bacterioferritin), and regulators (IdeR) that respond to higher free Fe 2ϩ levels.
The Coordination of Fe 2ϩ by P 1B4 -ATPases Likely Requires the Invariant Cys in the Fourth TM-Metal selectivity is central to the physiological roles of P 1B -ATPases. In early studies, invariant Cys in the sixth TM (fourth TM in P 1B4 -ATPases) were instrumental in defining P 1B -ATPases. Detection of other conserve residues in the transmembrane region led to the identification of P 1B -ATPases subgroups (19). The participation of these signature residues in the binding sites of P 1B1 Cu ϩ -ATPases and P 1B2 Zn 2ϩ -ATPases was later established (25,26). Then it was relevant to establish the metal coordination in P 1B4 -ATPases. Previous studies proposed that P 1B4 -ATPases coordinate Co 2ϩ with a Ser in the conserved SPC in the fourth TM and invariant His, Glu, and Thr in the sixth TM of these proteins (17). Surprisingly, no participation of the archetypical Cys in the fourth TM in Co 2ϩ coordination was observed. However, a different coordination of Fe 2ϩ by MtCtpD might explain its distinct biochemistry, i.e. higher affinity for Fe 2ϩ and Co 2ϩ and higher activity in the presence of Fe 2ϩ . We studied the coordination of Fe 2ϩ and Co 2ϩ while bound to MtCtpD TM-MBS by XAS and functionally analyzed variants carrying mutations in putative coordinating groups. XAS data indicate that both Fe 2ϩ and Co 2ϩ are coordinated by five to six oxygen/nitrogen ligands in a manner similar to that described previously for the Sulfitobacter sp. P 1B4 -ATPase. That is, the spectroscopy does not show the participation of sulfur atoms from the invariant Cys in the fourth TM as a metal ligand.
Notably, mutagenesis studies showed an alternative portrait of MtCtpD TM-MBS. As shown in the case the Sulfitobacter sp. P 1B4 -ATPase, we observed that mutation of S316A, H642A, E643A, and T646A led to an almost complete inhibition of Co 2ϩ activation, whereas replacement C318A retains significant (39%) Co 2ϩ -ATPase activity. A different pattern is observed, however, for the effects of these mutations on the Fe 2ϩ -ATPase. In this case, H642A retains some activity, whereas C318A causes a larger decrease in the activation by Fe 2ϩ . Although these differences are not dramatic, they suggest a putative differential involvement of these residues. Determination of the equilibrium binding affinities provided a more detailed view. Surprisingly, mutation H642A did not affect the metal binding to MtCtpD, suggesting that the reduced V max of this mutant is associated with alterations in rate-limiting conformational steps rather than ion coordination. Keep in mind that metal release is the rate-limiting step in P-type ATPases (21,50). More remarkably, replacement of C318A leads to a large reduction in the affinity for Fe 2ϩ without affecting Co 2ϩ binding. This datum in itself does not show a role of the conserved Cys in coordinating metals but suggests a direct effect, perhaps steric or through the second coordination sphere, in determining the affinity for Fe 2ϩ . In this case, the conservation of this Cys in the CPS signature sequence appears to be a logical consequence of the need to maintain a high binding affinity for Fe 2ϩ .
In summary, our observations suggest that mycobacterial P 1B4 -ATPases play a central role in Fe 2ϩ homeostasis. CtpD in particular, likely regulated by FurA, constitutes part of the cellular response to redox-induced damage of iron centers.  Values obtained by fitting curves resulting from competitive metal binding with mag-fura-2 (Fig. 7). Values are the mean Ϯ S.D. (n ϭ 3).