Crystal Structure of the Bacillus subtilis Phosphodiesterase PhoD Reveals an Iron and Calcium-containing Active Site

Background: PhoD family enzymes liberate phosphate from organic compounds. Results: The structure of PhoD reveals an active site with one Fe3+ and two Ca2+ ions. Conclusion: PhoD represents a new class of phosphatase related to purple acid phosphatases. Significance: The requirement of PhoD for iron ions may limit bacterial phosphate acquisition in low iron environments.

PhoA family enzymes are predominantly phosphomonoesterases. The active site of the prototypical PhoA enzyme of Escherichia coli contains two Zn 2ϩ and one Mg 2ϩ ion, and the catalytic mechanism involves an enzyme-bound phosphoserine intermediate (3). PhoX enzymes are exclusively phosphomonoesterases, and their complex active site cofactor has recently been shown to contain two Fe 3ϩ ions, three Ca 2ϩ ions, and a 3 -bridging oxo group (4).
Enzymes of the PhoD family are the least well characterized of the three major classes of bacterial extra-cytoplasmic alkaline phosphatase. The archetype of the family is Bacillus subtilis PhoD. This enzyme is a phosphodiesterase but also has significant phosphomonoesterase activity (5). Metal ion reconstitution experiments show that PhoD requires Ca 2ϩ for activity (5). The physiological role of PhoD is proposed to be the release of phosphate from cell wall teichoic acids under conditions of phosphate starvation (6). A second biochemically characterized member of the PhoD family is the secreted phospholipase D (PLD) 7 from Streptomyces chromofuscus (7). PLD cleaves the phosphodiester bond that links the head group and diacyl glycerol portions of phospholipids. PLD is reported to contain one iron atom together with substoichiometric amounts of manganese. In addition, the enzyme requires Ca 2ϩ ions for significant activity and to bind to phospholipid membranes.
Sequence analysis suggests that PhoD proteins are unrelated to the PhoA or PhoX phosphatase families but instead form a distinct subgroup within a very broad structural superfamily of hydrolytic enzymes possessing binuclear metal centers at the active site (the "metallophosphatase" superfamily in the Conserved Domain Database (8)). Within this structural superfamily it has been argued that the PhoD family has greatest similarity to eukaryotic purple acid phosphatases (PAPs) (9). In PAPs the active site metal pair are a Fe 3ϩ ion and a divalent ion that can be Fe 2ϩ , Mn 2ϩ , or Zn 2ϩ depending on the enzyme (10,11). The Fe 3ϩ ion is ligated by a tyrosine ligand resulting in an intense phenolate-to-Fe(III) charge-transfer transition ( max ϭ 510 -560 nm; ⑀ ϭ ϳ3000 -4000 M Ϫ1 cm Ϫ1 ), which gives PAPs their characteristic purple color. The occurrence of a tyrosinate-ligated Fe 3ϩ ion is highly characteristic for PAPs, with the only other currently identified examples being found in the structurally unrelated diferric DNA ligase from Ferroplasma acidiphilum (12). It is, thus, striking that S. chromofuscus PLD exhibits a PAPS-like visible spectrum. Amino acid residues, including a tyrosine, that could correspond to the Fe 3ϩ ligands in PLD have been proposed on the basis of sequence conservation and mutagenesis experiments (9).
In an effort to define the active site architecture of the PhoD phosphatase family we have determined the structure of B. subtilis PhoD. Unexpectedly, the active site contains three metal ions, these being identified as a Fe 3ϩ ion and two Ca 2ϩ ions. Our structural data show that the PhoD active site is distinct from that found in PAPs but that these two enzyme families conserve the position and coordination environment of their Fe 3ϩ ions. A phosphate molecule bound to the PhoD active site gives insight into the mechanism of catalysis.

EXPERIMENTAL PROCEDURES
Cloning-The sequence coding for the mature domain of PhoD (codons 57-593) was amplified from B. subtilis strain 168 chromosomal DNA using the primers 5Ј-CCAGTGGGTCTC-AGGTGGTGCGCCTAACTAAGC-3Ј and 5Ј-CGGCGTCG-ACTTAATTCGTGATTTTTGCACG-3Ј. The resulting amplicon was digested with BsaI and SalI and cloned into the same sites in plasmid pCA597 (13) to produce plasmid pCA597-mphoD. This plasmid directs the expression of the mature domain of PhoD with a Ulp1-cleavable N-terminal Strep-SUMO tag (where SUMO is small ubiquitin-like modifier). Site-specific mutations were introduced into the phoD coding sequence in plasmid pCA597-mphoD using the QuikChange TM method (Stratagene).
Protein Purification-E. coli strain BL21 (DE3) (14) containing plasmids pCA597-mphoD and pRARE (Novagen) was grown aerobically at 37°C in 2ϫYT medium (Sigma-Aldrich) supplemented with 50 g/ml kanamycin. Protein expression was induced at an A 600 nm ϭ 0.6 with 0.5 mM isopropyl-1-thio-␤-D-galactopyranoside, and the culture grown was then grown for 15 h at 20°C. Cells were harvested by centrifugation and resuspended in resuspension buffer (20 mM Hepes NaOH, pH 7.4, 200 mM NaCl, 1 mM CaCl 2 ), with EDTA-free complete protease inhibitors (Roche Applied Science) and DNase I (Sigma-Aldrich). The cells were broken by three passages through a French press at 8000 p.s.i. The lysate was clarified by successive centrifugation at 10,000 ϫ g for 20 min at 4°C and at 45,000 ϫ g for 1 h at 4°C, and then applied to a 5-ml self-packaged Strep-Tactin column (IBA, GmbH, Göttingen, Germany). Recombinant PhoD was eluted from the column with resuspension buffer containing 2.5 mM desthiobiotin and then supplemented with 1 mg of His 6 -Ulp1 protease (13) and dialyzed overnight at 4°C against resuspension buffer. Cleaved tags and uncleaved mPhoD protein were removed by incubation with Strep-Tactin matrix, and the sample was further purified on a Superdex75 (16/60) prep grade size-exclusion column (GE Healthcare) that had been equilibrated in resuspension buffer. The mPhoD-containing fractions were pooled and concentrated using a 30-kDa cut-off Amicon PLTK Ultracel-PL membrane concentrator (Merck Millipore).
Phosphatase Activity Assays-Alkaline phosphatase activity was assayed spectrophotometrically by measuring the release of p-nitrophenol from p-nitrophenyl phosphate (pNPP) or bis-pnitrophenyl phosphate (bis-pNPP). Assays were started by the addition of 1 g/ml PhoD to 100 mM Tris-HCl, pH 8.0, 2 mM CaCl 2 , and 1 mM pNPP or bis-pNPP at 25°C. Reactions were stopped after 5 min by the addition of 1 M final concentration NaOH. p-Nitrophenol production was quantified using ⑀ 405 nm ϭ 18,000 M Ϫ1 cm Ϫ1 .
The substrate specificity of PhoD was assessed by measuring phosphate release from test substrates. Each 800-l reaction contained 100 mM Tris-HCl, pH 8.0, 2 mM CaCl 2 , 1 mM test substrate, and 1 g of PhoD. After a 5-min incubation at 25°C the reaction was terminated by the addition of 200 l of malachite green reagent (AMS Biotechnology) followed by incubation at room temperature for 30 min. Production of phosphate was obtained from absorbance at 620 nm and quantified using a phosphate standard curve.
Crystallographic Data Collection and Structure Determination-Crystallization trials were performed at 294 K by the sitting drop vapor diffusion method. The best crystals were obtained in 0.2 M NaCl, 0.1 M Na 2 HPO4/KH 2 PO4, 54% PEG 200, pH 5.6, with protein at 7.8 mg/ml, a total drop volume of 400 nl, and a 50:50 protein-to-mother liquor ratio. The crystals were cryo-protected in mother liquor containing 10% (w/v) ethylene glycol and flash-cooled in liquid nitrogen. X-ray data were collected at beamlines ID23_1 (European Synchrotron Radiation Facility (ESRF), Grenoble, France) and I03 (Diamond Light Source, Oxfordshire, UK). The data were processed in space group I4122 using Process (part of the autoPROC pipeline; Global Phasing Ltd.) (15). The multiple wavelength anomalous diffraction method was used to determine the sites of the three catalytic metals (initially assigned as three iron sites, but later refined as one iron and two calcium sites) using autoSharp (16) and to produce phases that were optimally solvent-flattened to 62% using Solomon (17). The handedness of the sites was determined by the contrast (standard deviation of solvent to standard deviation of protein) falling from 57 to 27% in one hand and from 63 to 39% in the other. Buccaneer (18) built a partial model among a number of other unlikely chains. This model was then manually manipulated in accordance with the kidney bean purple acid phosphatase structure (Protein Data Bank (PDB) 3KBP). Input of this partial model for a second round of Buccaneer building placed 524 residues of a monomer into the map. This was used as the search model for molecular replacement via MOLREP (19), placing the two translational related non-crystallographic symmetry copies at (0.5, 0.5, 0.08) of the unit cell from one another. Iterative refinement and model building were carried out with Refmac (20) and Coot (21). The quality of the model was checked using MolProbity (22), with a final MolProbity score of 1.19 (93th percentile).
The sample for proton-induced x-ray emission (micro-PIXE) analysis was purified in a calcium-free buffer composed of 20 mM Tris-HCl, pH 7.2, 200 mM potassium acetate. Measurements were taken at the Ion Beam Centre (University of Surrey).
EPR Spectroscopy-EPR spectra were collected on an X-band CW EPR spectrometer (Bruker Biospin GmbH) using an X-band Super High Sensitivity Probehead (Bruker) and equipped with a low temperature helium flow cryostat (Oxford Instruments CF935). Data were analyzed using the program EasySpin (23).

Characterization of B. subtilis PhoD-B. subtilis
PhoD was heterologously expressed in the cytoplasm of E. coli. Purified recombinant PhoD protein was a monomer in solution as judged by size-exclusion multi-angle laser light scattering (Fig.  1A). The recombinant PhoD exhibited phosphomonoesterase activity with the colorigenic substrate pNPP and exhibited phosphodiesterase activity with the colorigenic substrate bis-pNPP (Fig. 1B). However, the specific activity of the recombinant enzyme was very much lower than previously reported for the native enzyme (5). We observed that the activity of the recombinant enzyme was massively enhanced by Ca 2ϩ supplementation (Fig. 1B), and so Ca 2ϩ was subsequently included in all buffers used for the purification and analysis of PhoD. At pH 8.0 and a temperature of 25°C, the Ca 2ϩ -supplemented recombinant PhoD protein had a K m for pNPP of 50 M and a k cat of 1.2 s Ϫ1 . PhoD was able to release phosphate from compounds containing phosphoester bonds but was not able to cleave phosphorus-nitrogen or phosphorus-carbon bonds ( Fig. 2A).

B
Specific Activity (μmoles/min.mg) The inset shows a magnified part of the spectrum around g ϭ 9.6. The PhoD concentration was 10 mg/ml in 20 mM Hepes, pH 7.5, 400 mM NaCl, 2 mM calcium acetate with 20% (v/v) glycerol added. A control experiment showed that the glycerol added to the samples did not change the spectral line shape (data not shown). EPR operating conditions: microwave power, 2 milliwatts; microwave frequency, 9.425 GHz; modulation, 5.0 G at 100 kHz; temperature, 10 K. Spectra are buffer-subtracted, with negative controls taken for sodium dithionite, hydrogen peroxide, and EDTA solutions.
PhoD purified in the presence of Ca 2ϩ was purple in color with a broad absorbance band in the visible spectrum, indicating the presence of a prosthetic group (Fig. 2B). The addition of the Ca 2ϩ -selective chelator EGTA caused a shift in the visible absorbance maximum from around 540 nm to ϳ560 nm (Fig.  2B), suggesting that Ca 2ϩ is associated with the prosthetic group. Proton-induced x-ray emission spectroscopy of PhoD purified in the absence of added Ca 2ϩ ions detected iron and calcium but no cobalt, magnesium, manganese, nickel, or zinc. The presence of iron in PhoD was confirmed by EPR spectroscopy, which detected the g ϭ 4.3 and weak g ϭ 9.6 peaks characteristic of the middle and lower Kramer's doublets of an isolated high spin rhombic Fe 3ϩ ion (Fig. 2C). No further signals were elicited by oxidation with hydrogen peroxide. All EPR signals disappeared on reduction with sodium dithionite, as expected of reduction of a monomeric Fe 3ϩ ion to the EPR silent Fe 2ϩ state (Fig. 2C). The Fe 3ϩ signal was unperturbed by the addition of 50 mM EDTA (Fig. 2C), showing that the iron in PhoD is too tightly bound to be removed by chelator treatment. The presence of iron in PhoD has not previously been reported. However, iron is found in the homologous S. chromofuscus PLD enzyme (7).
Structure of PhoD-PhoD was crystallized at pH 5.6 in the presence of phosphate. The structure of PhoD was determined to a resolution of 2 Å via a multi-wavelength anomalous diffraction experiment using the iron cofactor as the source of the anomalous signal. Final R/R free values at this resolution were 17.8/20.2%. Data reduction, phasing, and crystallographic refinement statistics are shown in Table 1. Every residue in the PhoD mature domain was observed in the electron density.
The core of the PhoD structure is a sandwich of two long ␤-sheets flanked by ␣-helices of varying lengths (Figs. 3A and 4). Three active site metal ions are coordinated by residues on loops on one face of the ␤-sandwich and are covered by a long C-terminal helix. With the knowledge that PhoD contains Fe 3ϩ and Ca 2ϩ ions (above) anomalous Fourier maps at three wavelengths allowed the assignment of the three active site metal ions as one Fe 3ϩ ion and two Ca 2ϩ ions, which we label Ca A and Ca B (Figs. 3, B and C, and 5A). Substitution of individual amino acids involved in metal ion coordination at each of the three metal sites either abolished (C180A, D265A, N271A, N272A, and D436A variants) or drastically reduced (D207A and D266A variants) PhoD activity with pNPP as a substrate (Fig. 5B), showing that all three metal ions are required for the phosphomonoesterase activity of PhoD.
PhoD was crystallized in the presence of phosphate, and the structure contains a phosphate molecule bound at the active site (Fig. 3B). Phosphate is a known competitive inhibitor of PhoD (5).
PhoD Has Structural Similarity to Purple Acid Phosphatases-A search of the PDB protein structure database using the DALI server (24) revealed that PhoD has highest structural similarity to PAPs. The closest overall structural match was to red kidney bean PAP (PDB 4DHL) (25) with a DALI Z-score of 28.6 and a main chain root mean square deviation relative to PhoD of 2.47 Å. The similarity between the two protein families is illustrated in Fig. 3, A and B, where we have chosen to compare PhoD with sweet potato PAP (PDB 1XZW) (26) due to the close structural homology in the way the two proteins bind phosphate (discussed below).
At the Fe 3ϩ site the sole difference in amino acid ligand sets between PhoD and the PAPs is a cysteine-for-aspartate substitution in PhoD (Cys-180). Cysteine is an unusual ligand to a Fe 3ϩ ion. Indeed, to the best of our knowledge, the structurally unrelated alkaline phosphatase PhoX (4) is the only other protein using a thiolate group to coordinate a redox-inactive Fe 3ϩ ion. The visible absorbance bands of PhoD are likely to arise predominantly from the tyrosine ligation of the Fe 3ϩ ion, as in the PAPs (27). However, additional thiolate-to-Fe 3ϩ charge transfer bands from the cysteine ligand will also contribute to the visible spectrum (28).
The Ca A site in PhoD corresponds to the position of the second metal site in the PAPs. Nevertheless, there is little similarity between the metal ion ligands found at this site in the two types of protein (Fig. 3B), consistent with the different coordination preferences of Ca 2ϩ and the divalent metal ions found in the PAPs. Specifically, the Ca A site has irregular 8-fold coordination with all amino acid ligation being provide by Asp residues.
The Ca B site in PhoD has no analogue in the PAPs. The Ca B ion has regular octahedral ligation with Asp and Asn ligands.
All three metal ions in PhoD have bonding interactions with the phosphate group present in the active site (Figs. 3B and 5A). One oxygen atom of the phosphate ligates the iron, one oxygen atom bridges Ca A and the iron, and one oxygen atom bridges the two calcium ions. The interactions between the phosphate and the Ca A -Fe pair replicate the tripodal -2-2-geometry observed between phosphate and the two metal ions of sweet potato PAP in PDB structure 1XZW (26) (Fig. 3B). A similar mode of bridging tripodal coordination has also been seen for phosphate bound to the dimetal centers of Sporosarcina pasteurii urease (29) and an organophosphate-degrading enzyme from Agrobacterium radiobacter (30). In addition to interacting with the active site metal ions the phosphate ion in PhoD has bonding interactions with the amino acid side chains of Arg-562 and His-566 from the C-terminal capping helix (Figs. 5A and 6, A-C).
The C-terminal Helix Occludes Access to the PhoD Active Site-The C-terminal region of the B. subtilis PhoD protein forms an ␣-helix (residues 560 -577), which lies over the active site (Figs. 3A and 6). This helix is part of a C-terminal extension that is found in PhoD from B. subtilis and certain additional Bacillus species, but is absent from other bacterial PhoD proteins and from the PAPs (Fig. 6A). The C-terminal helix is held in position by a network of salt bridges, which include interactions with the phosphate ion located in the active site, and with Asp-436, which bridges the Fe and Ca A ions (Fig. 6C).
The C-terminal helix blocks access to the catalytic site. Furthermore, as judged by modeling studies with pNPP substituted for the phosphate molecule seen in the crystal structure, steric clashes with the helix would also prevent substrates binding to the active site. As a consequence, the C-terminal helix must move from the position seen in the crystal structure in order for substrates to bind to PhoD. The main chain crystallographic B-factor for the C-terminal helix is high (49.1 Å 2 ) relative to the whole protein average (28.2 Å 2 ), showing that the C-terminal helix is more mobile than the rest of the PhoD structure (Fig.  6D). This suggests that the helix is relatively weakly bound to the main body of PhoD and could plausibly be released at some stage of the catalytic cycle. Attempts to observe the predicted movement of the C-terminal helix by co-crystallization with the bulky substrate analogues phosphonoacetic acid and phosphonohexanoic in place of phosphate were unsuccessful.
The function of the PhoD C-terminal helix was investigated using variants possessing truncations of this helix. These experiments show that the whole C-terminal helix is required for catalytic activity and that even the variant with the smallest truncation tested had lost the visible absorbance bands arising from the Fe 3ϩ ion (Fig. 7). Thus the C-terminal helix is required to either insert or retain the PhoD active site metal cofactor.

DISCUSSION
The structure of B. subtilis PhoD confirms earlier suggestions (9) that PhoD family proteins have structural features in common with eukaryotic PAPs. The two protein families share the same protein fold, and both contain a tyrosinate-ligated Fe 3ϩ ion at the same position in the active site. Indeed, the sole difference in the Fe 3ϩ coordination environment between  (44). C, electron density at the PhoD active site contoured at 1 (gray) overlaid with anomalous Fourier maps contoured at 4 (purple) at the x-ray wavelengths indicated at the top of each panel. The anomalous signals for iron relative to calcium are consistent with the theoretical fЉ values for iron and calcium at these wavelengths calculated using Crossec (45), which are given under the panels.

PAPS and B. subtilis
PhoD is the substitution of one of the aspartic acid residues in the canonical PAPs ligand set by a cysteine (Cys-180). In this context it is worth noting that PhoD family proteins can be divided on sequence grounds into two subfamilies, one of which possesses the cysteine Fe 3ϩ ligand seen in B. subtilis PhoD and also S. chromofuscus PLD, whereas the second has the aspartate ligand found in the PAPS site.
Notwithstanding their similarities, the PhoD and PAPs families clearly form distinct sequence groups and differ in their active site metal complement beyond the tyrosine-ligated Fe 3ϩ ion. Although the PAPs bind an additional Fe 2ϩ , Mn 2ϩ , or Zn 2ϩ ion, B. subtilis PhoD binds two Ca 2ϩ ions. Other members of the PhoD family almost certainly bind the same two Ca 2ϩ ions because the coordinating amino acids are close to invariant within the family (the residue equivalent to B. subtilis PhoD Ca B ligand Asn-272 is sometimes conservatively substituted with an Asp, which is also a good Ca 2ϩ ligand). Although the position of the Ca A ion in PhoD is similar to that of the divalent ion in the PAPs, the ligand sets provided by the two protein families are completely distinct. Ca A in PhoD has five coordination bonds from Asp residues, whereas the divalent ion in PAPs is ligated by an invariant His 2 Asp 1 Asn 1 set (31). It

PhoD|B.subtilis F F N D Y R G Y V R C T V T P H Q W K A D Y R V M P F V T E P G A A I S T R A S F V Y Q K D Q T G L R K V S S T T I Q G G V K Q S D E V E E
PhoD|B.subtilis 5   should be noted that the earlier predictions for the non-iron ion ligands in the PhoD family are incorrect (9). A few bacteria encode proteins that conserve all PAP metal ligands (32), and these proteins can be considered to be true bacterial PAPs. By contrast, our structural analysis shows that the much more widely distributed enzymes of the PhoD family should be regarded as forming a distinct but closely related sister group to the PAPs. Because B. subtilis PhoD and S. chromofuscus PLD are thought to target specific phosphate-containing molecules in vivo (teichoic acids and phospholipids, respectively) it is possible that other PhoD family members also have specialist biological roles rather than operating as general phosphatases like members of the PhoA and PhoX families. B. subtilis PhoD has a flexible C-terminal helix overlaying the active site (Fig. 7). Although this helix is not present in most other members of the PhoD family, analogous "cap domains" are a feature of certain other enzymes in the metallophosphatase superfamily (33)(34)(35)(36). In most cases the role of these cap domains is unclear. Intriguingly, however, the cap domain of a mycobacterial metallophosphatase has recently been implicated in cell wall binding (33). An equivalent function for the PhoD C-terminal helix is an attractive hypothesis given the proposed role of PhoD in teichoic acid metabolism.
The way in which phosphate is bound to the active site of PhoD in our crystal structure is likely to mimic the substrateenzyme interaction at some point in the catalytic reaction. We can get considerable insight into the likely mechanism of PhoD catalysis by interpreting the phosphate complex in the light of the extensive body of work on the mechanism of PAPs. In PAPs the two active site metal ions are bridged by a hydroxide ion. The current view is that phosphomonoester substrates coordinate both PAP metal ions. The bridging hydroxide ion then undertakes a nucleophilic attack on the phosphorus atom in-line with the bond to the leaving group. This results in the hydroxide ion being incorporated into the phosphate product (10). A structure has been determined for sweet potato PAP in which a phosphate ion is bound in a tripodal -2-2-geometry between the two metals ions such that one of the oxygen atoms occupies the position normally taken by the bridging hydroxide ion (26). This structure has been interpreted as representing the initial product complex in which the oxygen atom of the hydroxide nucleophile is fully bonded to the substrate phosphorus atom while still maintaining bridging coordination to the metal ions. The position of the phosphate ion in the PhoD structure is highly analogous to that in the sweet potato PAPs structure (Fig. 3B), suggesting that PhoD possesses a nucleophilic hydroxide between the Fe and Ca A ions and that the PhoD structure represents the initial product complex formed by the reaction involving this ion.
PhoD is a phosphodiesterase and so must be capable of catalyzing two successive hydrolysis reactions. This process could involve repeating the reaction involving the bridging hydroxide, with the initial product rotating in the active site before the second reaction. Alternatively, the two reactions could utilize distinct nucleophiles. Comparisons with the PAPs are again potentially informative in distinguishing between these two possibilities. Although PAPs have historically been considered not to catalyze phosphodiesterase reactions due to their inability to hydrolyze bis-pNPP (37) recent work with less sterically restricted substrates has shown that PAPs do possess this capability (38). The inferred phosphodiesterase mechanism in the PAPs is nucleophilic attack by a Fe 3ϩ -coordinated terminal hydroxide group followed by a second nucleophilic attack by the bridging hydroxide. A similar mechanism is plausible for PhoD with the phosphate complex corresponding to the initial product of the second hydrolytic reaction.
A striking difference between the PhoD and PAP families (and, indeed, other members of the metallophosphatase superfamily) is the presence of a third metal ion (Ca B ) in PhoD. Because PAPs are able to catalyze phosphodiesterase reactions at a similar rate to phosphomonoester hydrolysis with only two metal ions (38) it is unlikely that the presence of CaB is specifically linked to the phosphodiesterase functionality of PhoD. Instead, Ca B is appropriately placed to stabilize the developing charge on the leaving group during nucleophilic attack by the bridging hydroxide ion. More generally, the additional metal ion may compensate for the weaker polarizing ability of Ca 2ϩ relative to the divalent metal ions found in the PAPs (Fe 2ϩ , Mn 2ϩ , Zn 2ϩ ).
The combination of Fe 3ϩ and Ca 2ϩ ions used by PhoD is unusual and to the best of our knowledge has only otherwise been seen in the structurally unrelated PhoX alkaline phosphatase (4). Remarkably, therefore, two out of the three major classes of alkaline phosphatases in bacteria use the same, rare combination of metal ions. Both PhoD and PhoX are extracytoplasmic bacterial enzymes, and so the selection of Ca 2ϩ as the divalent cation may simply reflect the high availability of Ca 2ϩ ions in most terrestrial and marine environments (39). By contrast, the bioavailability of iron is low in many environments (40) and may limit the activity of PhoD family enzymes in such locations, as has also recently been argued for PhoX (4). Export of both PhoD and PhoX family proteins uses the Tat protein translocation system, which is dedicated to the transport of folded proteins (41)(42)(43). The Tat pathway is often used by proteins that need to fold before export to bind cofactor molecules in the cytoplasm. We speculate that the limited availability of soluble iron in the external environment means that biosynthesis of PhoD and PhoX cannot rely on the acquisition of adventitious iron in the periplasm. Iron ions would, instead, be inserted into these proteins under controlled conditions in the cytoplasm with export of the folded, iron-containing, proteins taking place via the Tat apparatus. In summary, the PhoD structure reveals that members of this widely distributed family of bacterial enzymes represent a new class of Fe 3ϩ -and Ca 2ϩ -dependent phosphatase.