Crystal Structure of a PhoU Protein Homologue

PhoU proteins are known to play a role in the regulation of phosphate uptake. In Thermotoga maritima, two PhoU homologues have been identified bioinformatically. Here we report the crystal structure of one of the PhoU homologues at 2.0 Å resolution. The structure of the PhoU protein homologue contains a highly symmetric new structural fold composed of two repeats of a three-helix bundle. The structure unexpectedly revealed a trinuclear and a tetranuclear iron cluster that were found to be bound on the surface. Each of the two multinuclear iron clusters is coordinated by a conserved E(D)XXXD motif pair. Our structure reveals a new class of metalloprotein containing multinuclear iron clusters. The possible functional implication based on the structure are discussed.

PhoU proteins are known to play a role in the regulation of phosphate uptake. In Thermotoga maritima, two PhoU homologues have been identified bioinformatically. Here we report the crystal structure of one of the PhoU homologues at 2.0 Å resolution. The structure of the PhoU protein homologue contains a highly symmetric new structural fold composed of two repeats of a three-helix bundle. The structure unexpectedly revealed a trinuclear and a tetranuclear iron cluster that were found to be bound on the surface. Each of the two multinuclear iron clusters is coordinated by a conserved E(D)XXXD motif pair. Our structure reveals a new class of metalloprotein containing multinuclear iron clusters. The possible functional implication based on the structure are discussed.
Inorganic phosphate (P i ) uptake is of fundamental importance in the cell physiology of bacteria because P i is required as a nutrient. Escherichia coli has developed a P i acquisition system that allows the assimilation of P i via a variety of systems. Two distinct systems for the uptake of P i have been described: the low affinity phosphate inorganic transporter and the high affinity phosphate-specific transporter (PstSCAB) (1,2). When the preferred P i source is in excess, it is taken up by the phosphate inorganic transporter. When the extracellular P i concentration is less than ϳ4 M, the synthesis of the high affinity transporter is induced, and P i is taken up by PstSCAB. The PstSCAB transporter belongs to the superfamily of ATPbinding cassette transporters and is encoded by the pst operon (3,4). This operon contains five genes that are transcribed counterclockwise in the following order: pstS, pstC, pstA, pstB, and phoU. PstS is a periplasmic P i -binding protein, and PstC and PstA are integral membrane proteins that mediate the translocation of P i through the inner membrane. PstB is an ATPase that energizes the transport. The Pst operon is part of the phosphate (PHO) 1 regulon that consists of 31 genes arranged in eight different operons in E. coli (5,6). The genes and operons of the PHO regulon are co-regulated by a two-compo-nent system composed of the regulatory proteins PhoB and PhoR. When the concentration of P i in the medium falls below ϳ4 M, the sensor protein PhoR phosphorylates PhoB, and the phosphorylated PhoB binds to the PHO boxes in the control region of Pst, recruiting the 70 subunit of the RNA polymerase and initiating transcription. When the P i concentration in the medium is in excess, the PHO regulon is repressed. Repression of the PHO regulon requires not only an excess concentration of extracellular P i but also the intact PstSCAB transporter and the PhoU protein.
The PhoU gene encodes a polypeptide of molecular mass ϳ27,000 Da. Although it is located in the pst operon, the encoded PhoU protein does not seem to participate in P i transport (4). Besides its role as a repressor of the PHO regulon, PhoU was also reported to be involved in intracellular P i metabolism (presumably related to the synthesis of ATP) (7).
Currently, there is a total of 44 entries of annotated PhoU protein homology sequences in the Protein Information Resource data base (www.pir.georgetown.edu). Only the E. coli PhoU protein was reported to be purified as a protein aggregate, and no structures for this protein family are available (8). Two PhoU homologues have been identified in Thermotoga maritima, and they are Tm1260 and Tm1734 (www.tigr.org). The Tm1260 gene is located in a gene cluster similar to the PhoU gene in E. coli, and Tm1734 is located in a gene cluster containing several hypothetical genes. The detailed molecular function for both PhoU homologues in T. maritima is unknown. However, these two PhoU homologues are highly sequencerelated (25% identity and 54% homology). As part of the Berkeley Structural Genomics Project (www.strgen.org), we purified and solved the crystal structure of one of the two PhoU homologues (Tm1734) from T. maritima (Tm_PhoU2) at 2.0 Å resolution. This structure reveals a new class of metalloproteins containing multinuclear iron clusters.

EXPERIMENTAL PROCEDURES
Cloning, Protein Expression, and Purification-The Tm_PhoU2 (gi number 4982311) gene was amplified by PCR using T. maritima genomic DNA template and primers designed for ligation-independent cloning (9). The amplified PCR product was prepared for vector insertion by purification, quantitation, and treatment with T4 DNA polymerase (New England Biolabs, Beverly, MA) in the presence of 1 mM dTTP. The prepared insert was annealed into the ligation-independent cloning expression vector pB3, a derivative of pET21a (Novagen, Madison, WI) that expresses the cloned gene with an Nterminal His 6 -tagged tobacco etch virus cleavage sequence, and transformed into chemical competent DH5␣ to obtain fusion clones. The seleno-L-methionine (Se-Met) isoform of Tm_PhoU2 protein was expressed in E. coli B834 (DE3). Cell paste was prepared with Studier's autoinduction method. 2 For protein purification, cell paste was resuspended in buffer A containing 50 mM HEPES, pH 7.0, 100 mM NaCl, supplemented with Roche protease inhibitors (Roche Applied Science). Cells were opened with a Microfluidizer (Microfluidics, Newton, MA). After centrifugation, the cleared cell lysate was loaded onto a 5-ml HiTrap MC column (Amersham Biosciences). The column was first washed with 10 column volumes of buffer A with 10 mM ␤-mercaptoethanol (BME) followed by 10 column volumes of buffer B containing 50 mM HEPES, pH 7.5, 1 M NaCl, and 10 mM BME. The target protein was eluted with buffer C containing 50 mM HEPES, pH 7.5, 100 mM NaCl, 10 mM BME using an imidazole gradient from 40 to 300 mM in 20 column volumes. Target protein was eluted at ϳ200 mM imidazole concentration. The peak fractions were pooled together and diluted 10-fold with buffer A. The diluted protein solution was then loaded onto a 5-ml HiTrapQ anion exchange column (Amersham Biosciences), and eluted with a linear NaCl gradient from 0 to 0.5 M NaCl in 20 column volumes. Protein came off at ϳ200 mM NaCl concentration. The peak fractions were then concentrated to 10 mg/ml in 20 mM HEPES, pH 7.0, 300 mM NaCl, 10 mM BME for crystallization. Inductively Coupled Plasma Mass Spectrometry (ICP-MS)-Purified Tm_PhoU2 protein sample was sent for ICP-MS experiments for trace metal analysis. The ICP-MS experiment was conducted at the University of Georgia (www.cal.uga.edu) using a Thermo Jarrell-Ash Enviro 36 inductively coupled argon plasm. Prior to analysis, the purified protein was placed in a buffer containing 20 mM Tris-HCl, pH 7.0, 1 mM dithiothreitol. The control buffer contained no protein but 20 mM Tris-HCl, pH 7.0, 1 mM dithiothreitol. The ICP-MS results showed that the iron and nickel concentration in Tm_PhoU2 protein is 7.1-and 11.3-fold higher than in the control buffer, respectively.
Crystallization and Structure Determination-Screening for crystallization conditions was done by the sparse matrix screening method (10). Optimum conditions were obtained by mixing 1 l of protein solution (10 mg/ml in 20 mM HEPES, pH 7.0, 300 mM NaCl, 10 mM BME) with 1 l of reservoir solution (100 mM HEPES, pH 7.5, 200 mM CaCl 2 , and 25% polyethylene glycol 400) and equilibrating over 1 ml of reservoir solution. Large crystals were obtained within 2-3 days at room temperature.
For data collection, crystals of Tm_PhoU2 were transferred to the reservoir solution containing protein buffer (20 mM HEPES, pH 7.0, 300 mM NaCl, 10 mM BME) plus 10% glycerol before they were directly frozen in liquid nitrogen. The multiple-wavelength anomalous dispersion data for Tm_PhoU2 Se-Met crystals were collected at Advanced Light Source Beamline 5.0.2 (Lawrence Berkeley National Laboratory, Berkeley, CA) at 100 K with a Quantum 315 charge-coupled device detector and processed with HKL 2000 (11). The crystals belong to the space group P3 1 21 or P3 2 21 with approximate unit cell dimensions of a ϭ b ϭ 90.677 Å, c ϭ 45.255 Å. The crystals contain one monomer/ asymmetric unit corresponding to a solvent content of 44.7%. Data statistics are summarized in Table I. The data for iron peak wavelength and nickel peak wavelength were also collected at Advanced Light Source Beamline 5.0.2.
The initial selenium atoms were determined using the HySS program (12). A total of 14 sites in the asymmetric unit were identified using the multiple wavelength anomalous dispersion data at 3.0 Å resolution, even though the protein monomer has only 12 seleniums. A calculation of the electron density map confirmed P3 1 21 to be the correct space group. Phasing and phase extension to 2.0 Å resolution by density modification were done with the crystallography NMR software program (13). The resultant electron density map was readily interpretable. The initial model was built with program O (14). The initial model building confirmed that the 14 sites found by HySS include 11 ordered selenium sites and three strong metal sites. Further experiments revealed that the three metal sites belong to two multinuclear iron clusters. Model refinement was carried out with the REFMAC5 program (15). The TLS method was utilized for structural refinement. At the final stage of refinement, iron clusters were built into the structural model, and water molecules were added. The final model of the Tm_PhoU2 protein was refined to R free ϭ 25.5% and R-factor ϭ 21.7% at 2.0 Å resolution and contained residues 2-226, seven iron ions, one nickel ion, one calcium ion, two acetate ions, and 78 water molecules. The final refinement statistics are given in Table I.

RESULTS
The crystal structure of Tm_PhoU2 was determined by multiple-wavelength anomalous dispersion using Se-Met protein.
The protein crystallized in space group P3 1 21 and contained one monomer in the asymmetric unit. The final model of the Tm_PhoU2 protein was refined to R free ϭ 25.5% and R-factor ϭ 21.7% at 2.0 Å resolution and contained residues 2-226, seven iron ions, one nickel ion, one calcium ion, two acetate ions, and 78 water molecules. The N-terminal methionine and nine Cterminal residues were not included in this model because of poor electron density. PROCHECK was used to assess the stereochemistry of this model (16). It shows 95.7% of the residues in the most favored regions of theplot, and no outliers were present. The r.m.s. deviations of bond lengths and bond angles from the standard geometry were 0.012Å and 1.17°, respectively (Table I).
Overall Structure-The Tm_PhoU2 monomer is mainly an ␣-helical structure with 81.7% of the amino acids in ␣-helical conformations, 19.6% of the amino acids in turns, and 3.65% of the amino acids in ␤-strand conformations (Fig. 1A). The overall structure of Tm_PhoU2 is composed of a six-helix bundle and a short two-stranded anti-parallel ␤-sheet attached at the C terminus. The six-helix bundle is composed of two three-helix bundles connected by a 10-residue loop. The N-terminal threehelix bundle is composed of H1A, H2A, and H3A, and the C-terminal three-helix bundle is composed of H1B, H2B, and H3B. The interface between these two three-helix bundles is mediated through H1A-H1B and H3A-H3B and contains mainly hydrophobic residues. The two three-helix bundles are structural repeats. The superposition of the two structural repeats of the three-helix bundle results in an r.m.s. deviation of 1.45 Å over 79 equivalent C␣ atoms (Fig. 1B). The major difference between these two repeats is that H1A is 12 amino acids longer than H1B. The other two ␣-helices in the structural repeats can be well superimposed. The superposition of these two repeats also revealed that the protein sequences in these two repeats are related. In fact, the two iron clusters identified in this structure are found at the equivalent positions of these two structural repeats and coordinated by the same E(D)XXXD motif repeats (discussed below).
Tm_PhoU2 Structure Represents a New Structural Fold-The DALI program was used for a structural homology search (17). No match was found for the complete model of the Tm_PhoU2 protein. A search with the whole model led to structural homologues to the first or second repeat. Not surprisingly, a search with the two structural repeats of the threehelix bundle led to many similar structural homologues. The structures with the highest Z-scores are the STAT (signal transducers and activators of transcription) protein fragment (Protein Database accession number 1UUR, Z-score 11.0, r.m.s. deviation 2.8 Å over 103 C␣ atoms with 7% sequence identity), the ␣-spectrin fragment mutant (Protein Database accession code 1CUN, Z-score 10.8, r.m.s. deviation 2.2 Å over 105 C␣ atoms with 10% sequence identity), and the heat shock cognate 71-kDa fragment (hsc70) bag-family (Protein Database accession code 1HX1, Z-score 10.7, r.m.s. deviation 2.2 Å over 97 C␣ atoms with 11% sequence identity). These proteins have very different sequences and molecular functions. They do not appear to be remotely related to the PhoU protein family. To the best of our knowledge, the structural fold composed of two repeats of a three-helix bundle for Tm_PhoU2 is unique and represents a novel structural fold in the protein data bank.
Two Multinuclear Metal Clusters-Metal binding of the Tm_PhoU2 protein was first suggested by the Se-Met phased ͉F obs Ϫ F calc ͉ difference density map and subsequently verified by ICP-MS results (see the "Experimental Procedures"). ICP-MS results showed that Tm_PhoU2 protein contains significant amounts of iron and nickel ions. Two diffraction data sets were collected subsequently at the iron peak wavelength (1.7433 Å) and the nickel peak wavelength (1.4879 Å) with the Se-Met crystals (Table I). Calculating the anomalous difference Fourier maps revealed two multinuclear metal clusters bound to the protein surface: a tetranuclear iron cluster, which contains three iron ions and one nickel ion bound to the N-terminal structural repeat, and a trinuclear iron cluster containing three iron ions bound to the C-terminal structural repeat (Fig.  1A). Both clusters are bound on the same side of the protein surface defined by highly acidic surface patches (Fig. 1C). Superposition of the N-and C-terminal structural repeats bound with multinuclear metal clusters revealed that the two iron clusters are bound at the same position of the two structural repeats (Fig. 1B). Because no iron ions were included in the purification and crystallization steps, these metal clusters had to be obtained during cell growth.
The assignment of iron or nickel in these two metal clusters was based on the anomalous difference Fourier maps calculated using 2.5 Å resolution data collected at 1.7433 Å (iron peak wavelength) and 1.4879 Å (nickel peak wavelength) (Fig.  2, A and B). Iron is a strong anomalous scatterer at both wavelengths (f Љ ϳ3.99 at 1.7433 Å and ϳ2.99 at 1. 4879 Å). The contribution of nickel is weak at 1.7433 Å (f Љ ϳ0.64) but strong for the wavelength of 1.4879 Å (f Љ ϳ3.92). The identification of nickel ion at the N-terminal iron cluster is based on the stronger anomalous difference peak at 1.4879 Å rather than that at 1.7433 Å, when both maps are contoured at the same level (Fig. 2B).
Trinuclear Iron Cluster-The trinuclear iron cluster is bound on the protein surface of the C-terminal repeat, coordinated by residues from H2B and H3B (Figs. 1C and 2C). All three iron ions (Fe-6, -5, and -7) in this cluster are well ordered and form a nearly isosceles triangle with a distance between Fe-5 and -6, Fe-5 and -7, and Fe-6 and -7 of 3.82 Å, 3.69 Å, and 6.68 Å, respectively. In the trinuclear cluster, Fe-5 can be described as a core ion and is partially buried inside the protein surface between H3B and H2B, coordinated by the terminal carboxylates of four conserved acidic residues: Asp 152 and Asp 156 from H2B and Glu 191 and Asp 195 from H3B. Fe-6 and -7 can be described as adjacent ions with Fe-6 bound on top of H3B, coordinated to Glu 191 and Asp 195 , and Fe-7 bound on top of H2B, coordinated to Asp 152 and Asp 156 . Toward the solvent region, the trinuclear iron cluster is internally capped by a modeled acetate ion with two oxygens coordinated by all three iron ions. The acetate ion was tentatively modeled based on the ͉F obs Ϫ F calc ͉ difference density map and the geometry of the trinuclear iron cluster. Although this difference density was modeled as an acetate ion, no acetate ion was used during purification and crystallization steps, and we are not sure about the actual identity of this difference density. A similar example for this modeling is the crystal structure of a ribonucleotide reductase R2 protein complexed with an oxo-centered trinuclear iron cluster, in which an acetate ion was found to coordinate two of the irons in the oxo-centered trinuclear iron cluster (18). Besides the modeled acetate group and protein residues, seven water molecules were also found to be involved in the coordination of the trinuclear iron cluster, with one water binding to Fe-5 and three waters each binding to Fe-6 and -7. The overall chemical environment around the trinuclear iron cluster is nearly symmetric. Fe-5 has a pentagonal bipyramidal coordination to seven oxygens. There are five equatorial oxygens: one from a well ordered internal water, two from the acetate oxygens, and two from the carboxyl oxygens of Asp 152 and Glu 191 . The two axial oxygens for Fe-5 are from the carboxylates of Asp 156 and Asp 195 . Among the five equatorial oxygens, four (except the internal water) are bridging oxygens between core iron Fe-5 and the two adjacent ions Fe-6 and -7. Fe-6 also has a pentagonal bipyramidal coordination to seven oxygens, in which the five equatorial oxygens are two water molecules, two bridging oxygens to Fe-5 (one from the acetate oxygen, another from the carboxyl oxygen of Glu 191 ), and another carboxyl oxygen from Glu 191 . The two axial oxygens are one water molecule and one carboxyl oxygen of Asp 195 . Fe-7 has a slightly distorted octahedral coordination to six oxygens, which are three water molecules, two bridging oxygens to Fe-5 (one from acetate oxygen, and another from the carboxyl oxygen of Asp 152 ), and one carboxyl oxygen from Asp 156 . Unlike Glu 191 , where both carboxyl oxygens are coordinated to Fe-6, Asp 152 has only one carboxyl oxygen coordinating to Fe-7. The other carboxyl oxygen of Asp 152 forms a hydrogen bond with the internal water that coordinates Fe-5, possibly stabilizing the internal water.
Tretranuclear Metal Cluster-The tetranuclear iron cluster is bound on the protein surface of the N-terminal repeat coordinated by residues from H2A and H3A (Figs. 1C and 2D). The tetranuclear cluster includes four metal ions: nickel, Fe-2, -3, and -4, in which nickel and Fe-2 and -3 form a similar triangle as the Fe-6 -5-7 trinuclear cluster. Here, Fe-2 represents the core ion, whereas nickel and Fe-3 represent the adjacent ions similar to Fe-7 and -6, respectively. An acetate ion was also modeled into a similar position as that of the C-terminal cluster. This acetate ion is less well defined, indicated by its high B factors (average ϳ70 Å Ϫ2 ). Again, we are not sure about the actual identity of the difference density, which we modeled as an acetate ion. The nickel ion has a higher B factor compared with other iron ions (68 versus 30 -47 Å Ϫ2 for iron ions), indicating that this metal ion is more mobile than other iron ions. Furthermore, it is possible that the nickel ion in this structure could be a result of an ion exchange between nickel and iron during protein purification because a nickel affinity column was used for purification.
The coordination for this tetranuclear iron cluster involves seven water molecules and eight protein residues including Glu 49 , Glu 50 , Asp 53 , and Glu 56 from H2A and Glu 88 , Glu 91 , Asn 92 , and Asp 95 from H3A. Fe-2 is partially buried inside the protein surface and is coordinated by four conserved residues: Glu 49 and Asp 53 from H2A and Glu 91 and Asp 95 from H3A (Figs. 1C and 2D). Overall, Fe-2 has an octahedral coordination to six oxygens: one from an internal water molecule, one from the acetate oxygen, and four from the carboxyl oxygens of Glu 49 , Asp 53 , Glu 91 , and Asp 95 . Here, Glu 49 and Glu 91 provide the bridging oxygens. Fe-3 has a distorted pentagonal bipyramidal coordination to seven oxygens: two from water molecules, one from the acetate oxygen, and four others from the terminal oxygens of Glu 91 , Asp 95 , and Asn 92 . Nickel has coordination to six oxygens: two from water molecules, two from carboxylates of Glu 49 , and two others from carboxylates of Asp 53 and Glu 50 . Because the nickel has a higher B factor, this coordination of the nickel ion shown is tentative. The extended iron ion, Fe-4, is coordinated by both H2A and H3A. Fe-4 has a slightly distorted octahedral coordination to six oxygens, three of which are from water molecules and three others from the carboxyl oxygens of Glu 56 , Glu 88 , and Glu 91 . Here, one water molecule and another carboxyl oxygen of Glu 91 act as bridging oxygens between Fe-3 and -4. E(D)XXXD Motif-A conserved motif was identified in the PhoU family by identifying the protein residues that coordinate the N-and C-terminal iron clusters in the Tm_PhoU2 structure. A total of four repeats of the E(D)XXXD motif can be identified in the PhoU family (Fig. 3). For the first N-terminal repeat, this motif can be extended to EXXXDXXE(D). For the second and the fourth repeat, the motif is restricted to EXXXD, and for the third repeat, the motif is restricted to DXXXD. The four motif repeats can be grouped into two motif pairs. The two N-terminal motif repeats form the N-terminal motif pair and constitute the N-terminal tetranuclear iron cluster binding site, whereas the two C-terminal motif repeats form the Cterminal motif pair and constitute the C-terminal trinuclear iron cluster binding site. Both sites are located at the equiva- lent position of the two structural repeats, and conserved carboxyl residues from each motif pair form a highly acidic surface patch. In the coordination between each of the two E(D)XXXD motif pairs and their bound multinuclear iron cluster, all four conserved carboxyl residues in each motif pair are involved in coordination to the core iron ion. The two N-terminal conserved residues (Glu or Asp) provide the bridging oxygens and symmetrically bind to the core iron, whereas the two conserved C-terminal residues (Asp) provide each of its carboxyl oxygens coordinating the core iron ion and an adjacent iron.
Dimeric Structure-It is not clear whether the Tm_PhoU2 protein functions as a monomer or as an oligomer. Gel-filtration chromatography indicates a protein size between a dimer and a monomer (ϳ40 kDa). In the crystal, two Tm_PhoU2 monomers form a dimer through the crystallographic dyad, resulting in a 12-helix bundle (Fig. 4A). The four helices bound with clusters in one monomer (H2A, H3A, H2B, and H3B) are parallel with their counterparts in another monomer and form Residues that are involved in the coordination of the two iron clusters are shown as ball-and-stick models. The identification of iron and nickel is based on the ICP-MS results and anomalous difference Fourier maps. The anomalous difference Fourier maps were calculated at 2.5 Å resolution using the refined Tm_PhoU2 model phases at 1.7433 Å (iron peak wavelength, shown as magenta on the left side for both A and B) and 1.4879 Å (nickel peak wavelength, shown as dark blue on the right side for both A and B). Both maps are contoured at 4.4 . C and D, stereo view of the coordination between the N-and C-terminal iron clusters and the Tm_PhoU2 protein residues. Acetate ions (Act) and residues that coordinate the two iron clusters are shown as ball-and-stick models. Iron ions are shown as spheres in green, and nickel is in pink. Water molecules are shown as spheres in red. C, view of the C-terminal trinuclear iron cluster. D, view of the Nterminal tetranuclear iron cluster. the dimer interface. There are a total of 64 interactions between the two monomers, including 30 hydrophobic interactions. A total of 5995 Å 2 surface area is buried between these two monomers, corresponding to 46% of the overall surface of the monomer. A crystal structure of the apo-PhoU protein from Aquifex aeolicus, recently solved in our laboratory in two different crystal forms, also shows a very similar dimer interface. 3 The striking feature about the Tm_PhoU2 dimer is that a central pore is observed at the dimer interface (Fig. 4B). At both ends of this pore, a tetranuclear iron cluster from one monomer and a trinuclear iron cluster from another monomer are close to each other and bind to the open acidic protein surface formed by conserved residues that may be functionally important. DISCUSSION The crystal structure of the Tm_PhoU2 protein revealed unexpected results. The Se-Met phased ͉F obs Ϫ F calc ͉ difference density map indicates that Tm_PhoU2 binds two metal clusters on the protein surface. The metal content was subsequently aeolicus, and Archaeoglobus fulgidus are aligned together using ClustalW (27). Tm1734 and Tm1260 are the two PhoU homologues from T. maritima. The invariant residues are labeled with asterisks, highly conserved residues are labeled with double dots, and generally conserved residues are labeled with a single dot. The secondary structure of Tm_PhoU2 is labeled on top of the sequences (arrows indicate ␤-strands, squares indicate ␣-helices, lines indicate loops, and dashes indicate disordered regions). The sequence motifs involved in iron cluster coordination are boxed. Residues that are conserved in the motif are labeled in red.

FIG. 4.
Dimer structure of Tm_PhoU2. A, ribbon diagram of the Tm_PhoU2 dimer. One monomer is shown in cyan with its metals shown in green, and another monomer is shown in purple with its metals shown in yellow. B, electrostatic surface presentation of the Tm_PhoU2 dimer. The left view is looking from one side of the central pore, and the right view is looking from the opposite side of the central pore. The central pore and the iron clusters are indicated. confirmed by the ICP-MS results and the anomalous difference density. The octahedral or pentagonal bipyramidal coordination for each individual iron ion in the two clusters is consistent with previously defined coordination geometry for the iron ion (18 -20). The fact that a multinuclear iron cluster is coordinated by an EXXXD motif pair has never been described previously. However, an oxo-centered trinuclear iron cluster coordinated by protein-derived carboxylates has been structurally characterized in the ribonucleotide reductase R2 protein, although these carboxyl residues are not conserved (18). In ferritins, the formation of an iron core is crucial for iron storage, detoxification, and mobilization throughout all three kingdoms. All ferritins contain several conserved carboxylate residues that are clustered in a patch on the inside surface of the ferritin sphere. Mutation studies have been performed on some of these residues and resulted in diminished core formation. The carboxylates are thus believed to make up nucleation sites for the mineral core (21)(22)(23)(24)(25). Although we cannot completely rule out the possibility that the iron ion could be another metal, the evidence presented here strongly supports that the two multinuclear metal clusters in Tm_PhoU2 protein are multinuclear iron clusters. The nickel ion is likely to be the result of an ion exchange between iron and nickel during protein purification.
So far, structures reported for a trinuclear iron cluster or multinuclear iron cluster-binding proteins are only for proteins involved in iron transport and iron metabolism. This includes ferric ion-binding protein, a protein involved in iron uptake from the transferrin superfamily, which binds trinuclear oxoiron clusters in an open cleft with a conserved dityrosyl ironbinding motif (19 -20), and the DpsA protein (Dps, DNA-protecting protein during starvation), a Dps-like ferritin family protein, which binds to iron clusters in a manner similar to that found in diiron-carboxyl oxygen-activating proteins (26). The Tm_PhoU2 structure defines a new class of metalloproteins that bind multinuclear iron clusters by its conserved E(D)XXXD motif pair in a manner not observed previously.
An intriguing feature about the Tm_PhoU2 structure is the highly symmetric nature in the protein sequence motifs, the structural fold, and the coordination between the protein and the multinuclear iron clusters. Such a symmetric coordination environment is only observed in small organic structures and is rarely found in protein structures. The Tm_PhoU2 structure bound by two multinuclear iron clusters gives the first example of how a protein uses its structural fold and conserved motifs to create a nearly symmetric coordination environment for multinuclear iron clusters.
The biological function of these carboxylate-coordinated multinuclear iron clusters bound to the Tm_PhoU2 protein is unknown. However, the structure of the Tm_PhoU2 protein suggests that the iron cluster binding for PhoU protein is likely to be conserved and functionally related. We speculate that the PhoU protein might use its conserved E(D)XXXD motif pair to recruit iron clusters, which may in turn act as a cofactor involved in unknown PhoU protein-related functions, such as P i metabolism (7).