Structure of Pyrimidine 5′-Nucleotidase Type 1

Eukaryotic pyrimidine 5′-nucleotidase type 1 (P5N-1) catalyzes dephosphorylation of pyrimidine 5′-mononucleotides. Deficiency of P5N-1 activity in red blood cells results in nonspherocytic hemolytic anemia. The enzyme deficiency is either familial or can be acquired through lead poisoning. We present the crystal structure of mouse P5N-1 refined to 2.35Å resolution. The mouse P5N-1 has a 92% sequence identity to its human counterpart. The structure revealed that P5N-1 adopts a fold similar to enzymes of the haloacid dehydrogenase superfamily. The active site of this enzyme is structurally highly similar to those of phosphoserine phosphatases. We propose a catalytic mechanism for P5N-1 that is also similar to that of phosphoserine phosphatases and provide experimental evidence for the mechanism in the form of structures of several reaction cycle states, including: 1) P5N-1 with bound Mg(II) at 2.25Å, 2) phosphoenzyme intermediate analog at 2.30Å, 3) product-transition complex analog at 2.35Å, and 4) product complex at 2.1Å resolution with phosphate bound in the active site. Furthermore the structure of Pb(II)-inhibited P5N-1 (at 2.35Å) revealed that Pb(II) binds within the active site in a way that compromises function of the cationic cavity, which is required for the recognition and binding of the phosphate group of nucleotides.

Two types of hP5N-1 deficiency are known: hereditary and acquired. Hereditary deficiency is associated with nonspherocytic hemolytic anemia (NHA) (2,7,10). Acquired deficiency results from inhibition of hP5N-1 by Pb(II) during lead poisoning, and its severity is related to the blood levels of Pb(II) (11)(12)(13)(14). Both NHA and severe lead poisoning (Ͼ80 g/dl Pb(II) in blood) have the same clinical manifestation: erythrocytes in blood smears of affected patients show a distinct basophilic stippling upon Wright's staining (7,14,15). At the molecular level, NHA is characterized by significantly increased levels of nucleotides in red blood cells (3-6 times), especially pyrimidine nucleotides that can account for up to 80% nucleotide content (7). Pyrimidine nucleotides are present only in minute amounts in normal erythrocytes (16). Similar although less pronounced accumulation of pyrimidine nucleotides may occur during severe lead poisoning (11). Four distinct missense mutations in hP5N-1 have been found in patients with NHA: D98V, L142P, N190S, and G241R (17)(18)(19)(20).
Here we report the x-ray structure of mouse P5N-1 (mP5N-1, UniProt code Q9D020) refined to 2.35 Å resolution and structures of several states of the mP5N-1 reaction cycle at 2.1-2.35 Å resolution that corroborate the hypothesis that the mP5N-1 catalytic mechanism is closely analogous to that of phosphoserine phosphatases. Furthermore we present a 2.35 Å structure of mP5N-1 complexed with Pb(II) and propose a mechanism for mP5N-1 inhibition by this metal. The structure of mP5N-1 represents a fold-space target (Ͻ30% sequence identity to any structure in the Protein Data Bank) and was determined under the NIGMS, National Institutes of Health Protein Structure Initiative.

EXPERIMENTAL PROCEDURES
Crystal Preparation-The gene encoding the mP5N-1 protein was cloned and selenomethionine (SeMet)-labeled protein was expressed and purified following the standard Center for Eukaryotic Structural Genomics pipeline protocol for cloning * This work was supported by NIGMS, National Institutes of Health Grants P50 GM64598 and U54 GM074901 (to J. L. Markley, principal investigator). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. To prepare complexes, crystals were soaked for 5 min to 1 h in mother liquor supplemented with a variety of additives as detailed in Table 1 and cryoprotected in the same solutions supplemented with up to 15% ethylene glycol (v/v). X-ray Data Collection and Structure Solution-X-ray diffraction data for apo-mP5N-1 were collected near the selenium K absorption edge (12,661 eV) at the Southeast Regional Collaborative Access Team 22ID beamline of Advanced Photon Source at Argonne National Laboratory. The diffraction images were integrated and scaled using HKL2000 (25). The selenium substructure of SeMet-labeled mP5N-1 crystals was determined using HYSS (26) and SHELXD (27). The protein structure was automatically phased and density-modified using autoSHARP (28) with the help of auxiliary programs from CCP4 suite (29). Iterative model building by the automatic tracing procedure of ARP/wARP (30) followed by combination of model and experimental phases by autoSHARP resulted in an initial model with ϳ86% of all possible residues traced of which 74% had side chains assigned. The structure was completed using alternate cycles of manual building in XFIT (31) and refinement in REFMAC5 (32). Two molecules of mP5N-1 were observed in the asymmetric unit of the crystal. All refinement steps were monitored using an R free value based on Ϸ5.1% of the independent reflections. The stereochemical quality of the final model was assessed using PROCHECK (33) and MOLPRO-BITY (34). X-ray diffraction data for the product complex (state VII) were collected at Northeastern Collaborative Access Team 8BM beamline of Advanced Photon Source at Argonne National Laboratory and processed using HKL2000. Data for the remaining complexes were collected with a Bruker AXS Proteum R CCD detector and Microstar rotating anode generator using copper K ␣ radiation. All x-ray data were processed and scaled with the programs SAINT and SADABS from the Proteum software suite (Bruker, Madison, WI). Structures of mP5N-1 complexes were solved by molecular replacement or rigid body refinement using apo-mP5N-1 as a model, manually adjusted with COOT (35), and refined as described for apo-mP5N-1. The figures were prepared using PyMOL (DeLano Scientific, San Carlos, CA).

RESULTS
Structure of mP5N-1-The structure of the mP5N-1 apoenzyme was determined by single wavelength anomalous diffraction and refined to a resolution of 2.35 Å. Data collection, refinement, and model statistics are summarized in Table 2. The three-dimensional structure of mP5N-1 revealed that this protein belongs to the ␣/␤-class of proteins with a three-layer sandwich architecture and Rossmann fold topology in the core domain ( Fig. 1, cyan and red). Extended loops of the fold that span residues 55-142 and 190 -211 form an auxiliary domain ( Fig. 1, yellow and green). Two of the helices of the auxiliary domain (␣4 and ␣5 in Fig. 1) form an extended "lid" located over the core domain. Two molecules of mP5N-1 present in the asymmetric unit showed only minimal differences in their main-chain conformation and were therefore refined using tight non-crystallographic symmetry restraints.
Sequence Homology Search-Analysis of the mP5N-1 sequence by profile alignment methods, including PSIBLAST (36), FFAS03 (37), and SUPERFAMILY (38), revealed that mP5N-1 is likely a member of the haloacid dehydrogenase (HAD) superfamily (SUPERFAMILY E ϭ 4.7 ϫ 10 Ϫ13 ). The HAD superfamily includes a range of phosphomutases, phosphatases, nucleotidases, and dehalogenases (39). HAD superfamily enzymes contain the signature motif DXDX(T/V). Aspartates within this motif have been implicated in the catalytic mechanism of several members of the HAD superfamily (40,41). A set of highly conserved residues of structurally characterized members of the HAD superfamily, as detected by profile-profile multiple sequence alignment by FFAS03, is also conserved in mP5N-1 and includes residues Asp 49 , Asp 51 , Thr 53 , Lys 213 , Gly 237 , Asp 238 , and Asp 242 . All these residues are involved in the formation of the active sites in HAD superfamily enzymes (39). mP5N-1 residues Asp 49 , Asp 51 , and Thr 53 reside within the signature motif DXDX(T/V). All conserved residues cluster in the mP5N-1 structure in the area that likely represents the active site ( Fig. 1, blue sticks).
Structural Homology Search-To confirm that mP5N-1 belongs to the HAD superfamily we performed a structural homology search using the DALI and VAST servers (42,43). Both DALI and VAST identified a range of structural homologs of mP5N-1, many of which were indeed established members of the HAD superfamily. The closest structural homologs of mP5N-1 identified by both servers were phosphoserine phosphatases (PSPs) from different species. Specifically the top homolog found by DALI was a PSP from Methanococcus jannaschii (MjPSP) with Z ϭ 14.7, root mean square deviation of 3.5 Å, and 17% sequence identity over 201 aligned C ␣ residues (Protein Data Bank code 1f5s (44)). The top structural homolog identified by VAST was human PSP (hPSP) with a VAST score of 20.9, root mean square deviation of 3.4 Å, and 17.5% sequence identity over 206 aligned residues (Protein Data Bank code 1l8o (45)). Fig. 2A shows a structural superposition of mP5N-1 and hPSP as determined by DALI. Although the overall fold of these proteins is clearly similar there are several notable differences. 1) mP5N-1 has a longer amino terminus (22 more residues). 2) Helices ␣4 and ␣5 in the lid domain are longer in mP5N-1 (residues 97-105). 3) Loops of mP5N-1 usually contain insertions of several residues and show significant differences in their local conformation from hPSP. Structural comparison of the active sites of PSPs and mP5N-1 revealed that they are highly similar (Fig. 2B).
Mechanism Overview-The catalytic mechanism of MjPSP and hPSP has been studied in considerable structural detail by x-ray crystallography (41, 44 -47). Both hP5N-1 and PSPs catalyze dephosphorylation of their substrates: pyrimidine 5Ј-mononucleotides and phosphoserine, respectively. The overall reaction is the intensity of an individual measurement of the reflection and ͗I(h)͘ is the mean intensity of the reflection. c R cryst ϭ ͚ h ͉͉F obs ͉ Ϫ ͉F calc ͉͉/͚ h ͉F obs ͉ where F obs and F calc are the observed and calculated structure-factor amplitudes, respectively. d R free was calculated as R cryst using ϳ5.0% of the randomly selected unique reflections that were omitted from structure refinement. e Dispersion precision indicator based on R free (57).
catalyzed by P5N-1 is shown in Fig. 3A. Because the active site of mP5N-1 is nearly identical to that of PSPs, we hypothesized that the catalytic mechanism of mP5N-1 might be analogous to that of PSPs. To provide experimental evidence for this hypothesis we set out to structurally characterize accessible states of the mP5N-1 reaction cycle. The general scheme of the proposed reaction cycle of mP5N-1, based on that of MjPSP (41), is depicted in Fig. 3B. Table 1 summarizes attempts made to prepare crystals of mP5N-1 trapped in different states of the reaction cycle. In summary, we successfully characterized the following: 1) apo-mP5N-1 (state I, reported above), 2) mP5N-1 with bound Mg(II) (state II), 3) the phosphoenzyme intermediate analog with BeF 3 Ϫ mimicking the phosphate group (state V), 4) the product-transition state analog with AlF 3 mimicking the phosphate (state VI), and 5) the product complex with Mg(II) and phosphate bound in the active site (state VII). However, we were unable to trap the nucleotide-bound states III and IV. When UMP or CMP was included during the soak procedure we ended up with the phosphate product complex (state VII) ( Table 1, lines 2 and 3). This observation suggests that mP5N-1 is catalytically active within the crystal lattice. Similarly inclusion of uridine and AlF 3 in the experiment designed to trap the substrate-transition analog complex (Table 1, lines 6 and 7; state IV) was unsuccessful even when 100 mM uridine was used and instead resulted only in the analog of the product-transition complex (state VI). Table 2 provides details of data collection and structure refinement for states II, V, VI, and VII of the mP5N-1 reaction mechanism. Snapshots of the catalytic site in structurally characterized states are shown in Fig. 4, A-E.
Mg(II) Coordination-hP5N-1 requires Mg(II) ion for activity (3,48). Our structures established that Mg(II) binds in the active site of mP5N-1 and is octahedrally coordinated by 1) the carboxyl oxygens of Asp 49 and Asp 238 , 2) the main-chain carbonyl oxygen of Asp 51 , 3) two water molecules, and 4) either an additional water (state II) or an oxygen of the phosphate group (or fluorine of phosphate analogs) (states V-VII) (Fig. 4, C-E).
Substrate Binding-The phosphate group of the substrate binds within a cationic cavity of the mP5N-1 active site. Our structures established that the phosphate binding is achieved by 1) hydrogen bonds to polar side-chain atoms of Lys 213 , Ser 164 , and in some states also Asp 51 ; 2) hydrogen bonds to main-chain amide nitrogens of Phe 50 , Asp 51 , Ala 165 , and Gly 166 ; 3) coordination to Mg(II); and 4) electrostatic interaction with Lys 213 and Mg(II). Structural details of the cationic cavity are most clearly visible in state VII (Fig. 5A). The remaining portion of the substrate is recognized and bound by unknown mP5N-1 residues. Based on our substrate docking trials (data not shown) these may possibly include Tyr 114 and Glu 96 , which has been reported to be important for binding affinity (49).
The Phosphoenzyme Intermediate-The catalytic residue, Asp 49 1 and 2b). B, the proposed catalytic mechanism for nucleotidase activity of mP5N-1. "R" represents ribonucleoside. Individual states of the reaction mechanism include apoenzyme (I), active enzyme (II), the substrate complex (III), the substrate-transition complex (IV), the phosphoenzyme intermediate (V), the product-transition complex (VI), and the product complex (VII).

phosphoenzyme intermediate analog was obtained by using BeF 3
Ϫ , which acts as a phosphate analog in proteins that are phosphorylated on aspartate residues (46,50). The intermediate with the covalently modified Asp 49 is shown in Fig. 4C.
The Product-Transition State-The covalent intermediate can either 1) undergo hydrolysis (nucleotidase activity), 2) transfer the phosphoryl group to precisely the same nucleoside (reverse reaction), 3) transfer the phosphoryl group to the same type of nucleoside (apparent reverse reaction), or 4) transfer the phosphoryl group to a different type of pyrimidine nucleoside (phosphotransferase activity) as shown in Fig. 3A. A nucleo-phile (water or nucleoside) must enter the active site and position itself to attack the phosphoenzyme at the modified Asp 49 (state V). Asp 51 is likely involved in the activation of the nucleophile as a general base. The reaction proceeds through a second transition state that leads to the formation of the second product, phosphate (state VI) or 5Ј-nucleotide. The structure of the product-transition state was obtained by using AlF 3 , which acts as the transition state analog of the phosphoryl group (51). Structural details of the catalytic site of this state are shown in Fig. 4D. The carboxyl oxygen of Asp 49 is located within 2.0 Å of the aluminum atom, and the attacking water refined to a distance of about 1.8 Å. In addition, a carboxyl oxygen of Asp 51 is located about 3.0 Å from the nucleophilic water. We hypothesize that phosphoryl transfer to a nucleoside would proceed through a similar transition state.
The Product State-Upon hydrolysis of the phosphoenzyme intermediate (or the phosphoryl group transfer onto nucleoside) the second product is formed (state VII) and leaves the active site. The structure of the product complex with Mg(II) and phosphate bound in the active site was refined to 2.1 Å resolution. The details of the active site of this complex are shown in Fig. 4E. Analysis of mP5N-1 complexes in states VI and VII confirmed that Asp 51 is the most likely candidate for the general base residue facilitating catalysis.
Productive and Non-productive States of Asp 49 -Detailed analysis of the mP5N-1 active site revealed that the most interesting changes involve the catalytic residue, Asp 49 . In summary, the side chain of Asp 49 adopts three different conformations

Pyrimidine 5-Nucleotidase Type 1 Structure
during the catalytic cycle. All conformations adopted by Asp 49 are related to rotamers of aspartate frequently found in proteins. Mg(II) binding induces the first change in the conformation of Asp 49 : the carboxyl group rotates about 70°to coordinate the metal (Fig. 5B). In addition, Asp 238 also rotates slightly (about 15°) to accommodate the metal. There is no significant change in the active site residue conformations in states V and VI compared with state II. (We expect that the same is true for states III and IV that were not observed in this work). Thus, coordination of Mg(II) by mP5N-1 precisely prepositions one of the carboxyl oxygens of Asp 49 for successful in-line attack on the phosphate group of the substrate. The second change in the conformation of Asp 49 occurs upon phosphoenzyme hydrolysis (transition from state VI to state VII). Interestingly in this conformation the carboxyl oxygen of Asp 49 , which does not coordinate Mg(II), points away from the phosphate bound in the cationic cavity (Fig. 5B). The change in rotamer is accompanied by a slight shift of the protein backbone of Asp 49 away from the metal (0.5 Å). We propose that the observed conformation of Asp 49 represents a catalytically non-productive state. The observed rotamer of Asp 49 , which occurs with low frequency (3.5%) in proteins, is stabilized by hydrogen bonds to polar atoms of side chains of Thr 53 (2.5 and 2.8 Å) and Asp 242 (3.2 Å). Both these residues are very highly conserved in the HAD superfamily of enzymes. A similar change in the rotamer state of the catalytic residue has been described for the phosphate product complex of MjPSP (Protein Data Bank code 1l7m). In that case both conformations of the catalytic residue were observed at the same time (41).
Structural Insight into Lead(II) Inhibition of mP5N-1-hP5N-1 is one of several enzymes readily inhibited by Pb(II) during lead poisoning (52,53). However, the molecular mechanism of this inhibition was previously unknown. It has been proposed that Pb(II) affects sensitive sulfhydryl groups and somehow alters hP5N-1 conformation (6). To obtain a structural insight into the mechanism of mP5N-1 inhibition by Pb(II) we derivatized mP5N-1 crystals with lead(II) acetate (Table 1, line 8). Data collection and refinement statistics for this complex are summarized in Table 2. The structure revealed that Pb(II) binds within the active site of mP5N-1 (Fig. 4F). The presence of Pb(II) within the active site was further confirmed by analysis of anomalous difference maps (10 and 7.8 peaks found in monomers A and B, respectively). However, we noted that the position of Pb(II) suggested by the anomalous difference maps was offset up to 0.6 Å toward either Asp 49 in monomer A or Asp 238 in monomer B from a position suggested by 2F o Ϫ F c difference maps. In addition, the height of the anomalous peak for Pb(II) was comparable to that of the selenium of buried SeMet 245 . The theoretical dispersion correction for forward scattering f Љ of selenium at 1.54 Å is 1.14, whereas that of lead is 8.51. These observations are therefore consistent with partial occupancy of the active site by Pb(II) and with an overlay of apo-mP5N-1 and Pb(II)-bound mP5N-1. Because of the limited resolution of the diffraction data we did not attempt to refine the alternate conformational states of mP5N-1 in this complex. We refined the structure assuming 25% occupancy for Pb(II) and estimated up to 0.6 Å error in Pb(II) location. The occupancy of Pb(II) and diffraction quality of Pb(II)⅐mP5N-1 com-plex represent an experimental compromise: higher concentration of lead(II) acetate and/or longer soaking time were poorly tolerated by crystals, which showed cracking and lowered diffraction quality (Table 1, line 9). Nevertheless we believe that the model of Pb(II)⅐mP5N-1 complex provides a good framework for understanding the molecular mechanism of mP5N-1 inhibition by Pb(II). The structure of Pb(II)⅐mP5N-1 complex established that the binding mode of Pb(II) in the active site differs from that of Mg(II) (Fig. 4, compare B-E and F). The position of Pb(II) coincides approximately with the position of one of the water molecules that coordinates Mg(II). Pb(II) is not coordinated by the backbone carbonyl of Asp 51 ; however, it is stabilized by electrostatic interaction with Asp 242 (2.8 Å). Importantly Pb(II) is clearly located away from the cationic cavity.

DISCUSSION
The mouse P5N-1 enzyme shows 92% identity and 97% similarity to the human ortholog with nonidentical residues residing mainly on the surface of mP5N-1. The structure of mP5N-1 therefore represents an excellent model for hP5N-1. We argue that all conclusions made in this work remain valid for hP5N-1.
One of the key achievements of this work is that we unequivocally established mP5N-1 as a member of the HAD superfamily of enzymes. Although profile alignment methods suggest this relationship, only a structural characterization could confirm it. We found that mP5N-1 shows significant structural homology to several established members of the HAD superfamily, especially PSPs. In addition, mP5N-1 shows a clear pattern of evolutionary conservation of residues delineating the active site (39). Finally we established that the catalytic mechanism of mP5N-1 is also closely similar to that of PSPs. The ability to assign mP5N-1 to the HAD superfamily opens an extensive body of literature related to the catalytic mechanism of PSP and other HAD superfamily enzymes.
Structural snapshots of the mP5N-1 reaction cycle clearly established that 1) Asp 49 acts as the catalytic residue and 2) Asp 51 is positioned appropriately to function as proton donor during nucleophilic attack by Asp 49 and as a general base during hydrolysis of (or phosphoryl transfer from) the phosphoenzyme intermediate. It has been observed that nucleotidase activity of hP5N-1 is 1) optimal at pH 7-7.5, 2) drops to 70% at pH 5.0, and 3) drops precipitously at basic pH and at pH 9.0 is nearly fully inhibited (48). This profile is consistent with an involvement of acidic residue(s) in the catalysis. Further evidence for the importance of Asp 49 and Asp 51 is provided in a recent mutagenesis study: both D49N and D51N mutants result in complete loss of catalytic activity (49).
The structure of the product complex provides important clues for the role of Thr 53 . This residue resides within the signature motif DXDX(T/V) of the HAD superfamily. To the best of our knowledge no hypothesis about the role of this residue has yet been formulated. We propose that Thr 53 stabilizes the non-productive conformation of the catalytic residue Asp 49 by an additional (and possibly stronger) hydrogen bond compared with the productive conformation (Fig. 5B). This stabilization leads to a lower reverse reaction rate, thus driving overall reaction toward products.
The structure of mP5N-1 provides a framework for understanding results of kinetic studies of nucleotidase and/or phosphotransferase activities of hP5N-1 (3,54). The cationic cavity is located at the interface of the core domain and the lid of mP5N-1. We argue that for steric reasons there has to be an ordered product release. The first product, a nucleoside, has to leave mP5N-1 before a nucleophile (water or another nucleo-side) can enter the active site (Fig. 3A). The second product is formed from phosphoenzyme next and can leave the active site only after that point. The structure of the phosphoenzyme analog confirmed the presence of Asp 49 -BeF 3 Ϫ in the active site and provides the first direct experimental evidence for this intermediate in the mP5N-1 catalytic cycle. Existence of this intermediate clearly suggests that nucleotidase and phosphotransferase activities are interconnected. Existence of this intermediate in the catalytic cycle of hP5N-1 was inferred from kinetic studies (3,54). The evidence for similar intermediates has been provided for several members of the HAD superfamily that catalyze dephosphorylation or phosphotransfer, including MjPSP, phosphomannomutase, P-type ATPases, and purine 5Ј-nucleotidase cN-II (40,46,55). Overall we argue that the mP5N-1 structure is consistent with an ordered bi-bi mechanism with ping-pong kinetics for the overall reaction catalyzed by this enzyme (Fig. 3A).
Lead "Poisoning" of mP5N-1-Elucidation of the mP5N-1 structure put us in the unique position to address the molecular mechanism of mP5N-1 inhibition by Pb(II) that occurs during lead poisoning. The results obtained in this study show that 1) Pb(II) binds specifically within the active site and 2) Pb(II) binds in a different position compared with Mg(II), importantly, away from the cationic cavity (Fig. 4F). We propose that mP5N-1 inhibition by Pb(II) is based on lowered affinity of the cationic cavity for the phosphate group and possibly also on improper positioning of the catalytic residue Asp 49 for in-line attack. A properly positioned metal within the active site is directly coordinated by the phosphate bound in the cationic cavity and further stabilizes the negatively charged phosphate through electrostatic interaction. Lead(II) shows much higher affinity for the active site than Mg(II) (48). It thus outcompetes Mg(II) and "poisons" the active site of mP5N-1 because simultaneous binding of Mg(II) and Pb(II) is not possible. Interestingly it has been reported that Cr(III) slightly increases hP5N-1 activity in the presence of 10 mM Mg(II) (48). We hypothesize that Cr(III) binds in the same position as Mg(II) and increases the affinity of the cationic cavity for the phosphate group due to better electrostatic stabilization. Structural Insight into the Pathologies of hP5N-1 Mutants-The structure of mP5N-1 and insight into the catalytic mechanism of mP5N-1 allows for a structural explanation of the catalytic deficiencies of hP5N-1 mutants found in patients  with NHA. Four site mutations have been described in patients suffering from NHA: D98V, L142P, N190S, and G241R (18 -20). Thermal stability and catalytic properties of all these mutants have been recently studied in vitro (56). All targeted residues are conserved in mP5N-1 and shown in Fig. 1. Our insight into the molecular pathologies of the familial mutations of P5N-1 is summarized in Table 3. The structural details relevant to understanding these pathologies are depicted in Fig. 6. It is interesting to note that none of the mutants target the catalytic site residues and that decreased catalytic efficiency of mutant proteins seems to result from secondary effects related to propagating conformational changes within the enzyme.