Structural basis of catalysis by monometalated methionine aminopeptidase

Abstract

Methionine aminopeptidase (MetAP) removes the amino-terminal methionine residue from newly synthesized proteins, and it is a target for the development of antibacterial and anticancer agents. Available x-ray structures of MetAP, as well as other metalloaminopeptidases, show an active site containing two adjacent divalent metal ions bridged by a water molecule or hydroxide ion. The predominance of dimetalated structures leads naturally to proposed mechanisms of catalysis involving both metal ions. However, kinetic studies indicate that in many cases, only a single metal ion is required for full activity. By limiting the amount of metal ion present during crystal growth, we have now obtained a crystal structure for a complex of Escherichia coli MetAP with norleucine phosphonate, a transition-state analog, and only a single Mn(II) ion bound at the active site in the position designated M1, and three related structures of the same complex that show the transition from the mono-Mn(II) form to the di-Mn(II) form. An unliganded structure was also solved. In view of the full kinetic competence of the monometalated MetAP, the much weaker binding constant for occupancy of the M2 site compared with the M1 site, and the newly determined structures, we propose a revised mechanism of peptide bond hydrolysis by E. coli MetAP. We also suggest that the crystallization of dimetalated forms of metallohydrolases may, in some cases, be a misleading experimental artifact, and caution must be taken when structures are generated to aid in elucidation of reaction mechanisms or to support structure-aided drug design efforts.

All newly synthesized proteins have an amino-terminal methionine residue corresponding to the start codon AUG. In a significant number of cases, this initiating methionine residue is removed, either co- or posttranslationally, by the enzyme methionine aminopeptidase (MetAP) (1). In bacteria, this enzyme is the product of a single gene, and it is absolutely essential for bacterial survival, as demonstrated by gene deletion experiments in Escherichia coli (2) and Salmonella typhimurium (3). The single but critically essential MetAP enzyme of bacteria thus stands out as an attractive target for the design of antibacterial agents (4). Eukaryotes, on the other hand, have two distinct MetAPs, types I and II, arising from different genes (5). The human type II MetAP is a target of the antiangiogenic compounds fumagillin, ovalicin, and TNP-470 (6–8). Bengamides inhibit both types of MetAP (9) and cause inhibition of the growth of several human tumor cell lines in vitro at low-nanomolar concentrations. Therefore, human MetAPs may also serve as targets for the development of new anticancer agents. Some small molecules that inhibit MetAPs potently in vitro are known, but they lack potent antibacterial (10–12) or antiangiogenic (13) activities. One reason for this failure may be that they do not penetrate the bacterial or mammalian cells to reach their intended target, or perhaps they are efficiently transported back out of the cells.

It has long been known that MetAPs can be activated in vitro by a number of different divalent metal ions, including Mn(II), Fe(II), Co(II), Ni(II), and Zn(II) (14–16), and we have shown that inhibitors of the Co(II) form may or may not inhibit MetAP in other metalloforms (14). It is not known which of these ions may be present in the MetAP under physiological conditions. Thus, another reason for the apparent lack of antibacterial and antiangiogenic activities of some potent in vitro MetAP inhibitors may be a mismatch between the metal ion used for activity measurement in vitro and the metal ion used to activate the apoenzyme inside cells. The former has most often been Co(II), whereas Mn(II) and Fe(II) have been suggested for the latter (15, 17). Our laboratory has recently discovered several classes of in vitro inhibitor for E. coli MetAP that inhibit different metalloforms of the enzyme with remarkable selectivity (18); they may become powerful tools to define the intrinsic metal used by MetAP under physiological conditions.

MetAP is a member of the metalloaminopeptidase family. X-ray structural analyses of numerous enzymes in this family, both with and without small ligands bound at the active site, show that all possess a dinuclear metal site (dimetalated) comprising conserved amino acid residues that furnish imidazole and carboxylate donors (19). These analyses have led to proposed reaction mechanisms involving both metal ions (i.e., M1 and M2) (20, 21). A water molecule (or hydroxide ion) that bridges M1 and M2 serves as a nucleophile to attack the carbonyl group of the scissile peptide bond. These metal ions are suggested to bind and activate the water for deprotonation and assist the nucleophilic attack, whereas the metal ion M2 also binds the free unprotonated amino group of the methionine residue (P1) of the substrate for a productive bound conformation. In some recent structures, a third divalent metal ion is also present, serving as a bridge between a small molecule ligand and active-site groups on the enzyme (11, 22–24). Despite the great number of x-ray structures of these enzymes that show two or even three divalent metal ions bound in the active-site region, metal titration studies indicate that some of these metalloaminopeptidases can function fully with only a single metal ion bound. Included in this group are the Aeromonas aminopeptidase (25), porcine kidney leucine aminopeptidase (26), human cytosolic (27) and porcine membrane-bound (28) aminopeptidase P, and the MetAPs from E. coli and Pyrococcus furiosus (29, 30). The apparent binding constants for divalent metal ions by apo-MetAPs are in the low- or submicromolar range for the first equivalent of metal, but they are often much higher for the second equivalent (29, 30).

From a drug-design viewpoint, x-ray crystal structures of target enzymes often provide extremely valuable guidance for optimizing, by structural modification, the activity of lead compounds discovered by screening or designed as transition-state analogs of natural substrates. Although there are many crystal structures of MetAP enzymes with two Co(II) ions bound (21) and a few with two Mn(II) ions bound (18, 22, 23, 31), there are no crystal structures of any metalloaminopeptidase in a monometalated form. Because proteins are usually crystallized from concentrated solutions containing excess amounts of small ligands and cofactors such as metal ions, it is not surprising that both of the available metal-binding sites would be occupied in solution before crystal formation. Thus, although dimetalated MetAP enzymes may be catalytically active, they may not be the most physiologically relevant form, at least for MetAP and possibly for other metallopeptidases, and they may not provide the best information for purposes of structure-aided drug design. Here, we present a structure of a mono-Mn(II) form of E. coli MetAP with a transition-state analog bound at the active site. We also present structures of the same complex showing the transition from the mono-Mn(II) form to the di-Mn(II) form, and we show how various groups on the enzyme and the ligand move as the second metal site fills.

Results and Discussion

Activation of E. coli MetAP Requires only One Equivalent of Co(II) or Mn(II).

Although the available x-ray structures of MetAP are exclusively of the di- or trimetalated forms, kinetic evidence indicates that MetAP can be fully functional as a monometalated enzyme (29, 30). To verify this conclusion, we titrated the apoenzyme of E. coli MetAP with increasing amounts of Co(II) or Mn(II) while monitoring enzymatic activity with a kinetic fluorescence assay. It is clear that only a single equivalent of Co(II) or Mn(II) is required to activate the enzyme fully (Fig. 1). Furthermore, because the binding of M2 in E. coli MetAP is much weaker [K_d 2.5 mM for Co(II)] than the binding of M1 [K_d 0.3, 6, and 0.2 μM for Co(II), Mn(II), and Fe(II), respectively] (29, 30), the dimetalated form is relatively unlikely to exist inside cells.

Fig. 1.

Activation of E. coli MetAP by Co(II) (open circles with error bars and solid line) and Mn(II) (filled squares and dashed line) as monitored by hydrolysis of the fluorogenic substrate Met-7-amido-4-methylcoumarin (AMC). The apoenzyme concentrations [20 μM or 75 μM for Co(II) or Mn(II), respectively] are 66.7 and 12.5 times higher than the reported K_d values of 0.3 μM and 6 μM (29, 30) to ensure high metal occupancies at the equivalence point of the titration. The enzyme shows essentially maximum activity when the metal/apoenzyme ratio reaches 1:1. The curve shapes are consistent with the tight binding behavior of the M1 site.

Crystallization of Mono- and Dimetalated MetAPs.

To obtain crystals of the monometalated form of E. coli MetAP, we used conditions that strictly limited the Mn(II)/apoenzyme ratio and provided a transition-state analog inhibitor of the enzyme, norleucine phosphonate (NleP), as an auxiliary ligand. We anticipated that the α-amino group of this ligand would be able to occupy or at least obstruct the M2 site, further favoring the formation of crystals in the monometalated form. Thus, hanging drops containing MnCl₂, apoenzyme (7.5–15 mg/ml), and NleP in the mole ratio 0.25:1:9 were set up and maintained at 21–23°C. Similar experiments were conducted with the same excess of NleP (or absence of NleP for the unliganded structure) but varying the metal/apoenzyme ratio from 0.5:1 to 2:1. Crystals were generated that consistently produced high-quality diffraction data for structural solution to resolutions of 1.6–2.0 Å (Table 1, which is published as supporting information on the PNAS web site).

Mono- and Dimetalated Structures of MetAP Complexed with NleP.

From the crystals described above, we solved a series of x-ray structures that show a progression of associated structural changes around the active site (Fig. 2A–D). In particular, although the M1 site was always filled with Mn(II) before the M2 site, occupancy of the M2 site varied from virtually empty to essentially full in accordance with changes in the Mn(II) concentration during crystallization.

Fig. 2.

X-ray structures. (A–D) Binding modes of NleP at the active site of E. coli MetAP in four related but independently obtained x-ray structures. Shown here are stereoviews of partial structures in the mono-Mn(II) form (A), the di-Mn(II)-form (D), and structures (B and C) in between. An x-ray structure of unliganded E. coli MetAP is also shown in E. The metal/apoprotein concentration ratios during crystallization were 0.25:1 (A), 0.5:1 (B), 1:1 (C), 2:1 (D), and 0.5:1 (E). Two molecules of enzyme, A and B, are present in each asymmetric unit. For molecule A, the metal site occupancies (in percent, M1/M2) are 40/0 (A), 81/29 (B), 90/46 (C), 100/99 (D), and 71/28 (E). Those for molecule B are 47/0 (A), 83/33 (B), 91/45 (C), 98/96 (D), and 46/31 (E). The structures shown are for molecule A. The color scheme is as follows: gray, carbon (protein residues); yellow, carbon (inhibitor); blue, nitrogen; red, oxygen; and orange, phosphorus. Mn(II) ions are shown as green spheres, and water molecules are shown as a red sphere. F_obs − F_calc omit maps (inhibitor and metal ions were not included in the model) are shown superimposed on the refined structures as blue meshes contoured at 3.5 σ.

As we reported in ref. 18, the structure of the di-Mn(II) form (Fig. 2D) is very similar to that of the di-Co(II) form of the enzyme (32) [Protein Data Bank (PDB) ID code 1C27]. Alignment of the two structures over 1,748 main-chain and side-chain atoms gives an rms deviation of 0.33 Å, whereas the rms deviation is reduced to 0.15 Å if only the residues within 5 Å of the two metal ions (87 atoms) are aligned. In the mono-Mn(II) form (Fig. 2A), however, several larger structural changes are readily apparent in the vicinity of the active site. For example, the carboxylate group of residue Asp-97 rotates, whereas that of residue Asp-108 swings in the direction of Asp-97; both become closer to the amino nitrogen of the NleP ligand. In the monometalated structure, the distances from this nitrogen to the carboxylate oxygens of Asp-97 and Asp-108 are 2.74 and 3.70 Å, and 2.95 and 3.22 Å, respectively, whereas in the dimetalated structure, the corresponding distances are 3.39 and 3.40 Å (Asp-97), and 3.68 and 3.37 Å (Asp-108). The smaller distances in the monometalated structure suggest the development of strong charge–charge interactions as well as hydrogen bonding between the amino group (which is probably protonated) and the carboxylate groups (which are probably deprotonated). Another major conformational change involves the oxygens of the phosphonate moiety of NleP. In the di-Mn(II) form (Fig. 2D), one of the oxygens is firmly anchored to and bridges between the two Mn(II) ions, but in the monometalated form (Fig. 2A), the tripod of phosphonate oxygens has rotated around the P–C^α axis by nearly 30°. Accompanying this rotation is a tipping of the plane of the imidazole ring of His-178 by ≈10° (i.e., rotation around the C^β–C^γ axis). Despite this tipping, the intermolecular hydrogen bond between the second phosphonate oxygen and the imidazole nitrogen is maintained. Similarly, the imidazole ring of His-79 adjusts its position with a rotation to maintain the hydrogen bond between the imidazole nitrogen and the third oxygen of the phosphonate. Finally, M1 changes from hexacoordinate in the dimetalated structure to pentacoordinate in the monometalated structure.

Whereas l-methionine is a weak product inhibitor, NleP and the bestatin-based peptide (3R)-amino-(2S)-hydroxyheptanoic acid (AHHpA)-Ala-Leu-Val-Phe-OMe (Fig. 3A) are both potent transition-state analog inhibitors of MetAP. NleP mimics the transition-state and tetrahedral intermediate, with two of its phosphonate oxygens imitating the gem-diol grouping and the remaining oxygen taking the place of the amide nitrogen. A comparison of the structures with NleP or AHHpA-Ala-Leu-Val-Phe-OMe complexed in the mono- and dimetalated forms of MetAP is displayed in Fig. 3B. Overlaying the structure of mono-Mn(II) MetAP and di-Mn(II) MetAP complexed with NleP shows that the ligand side chains virtually superimpose. In the complex of di-Co(II) MetAP with AHHpA-Ala-Leu-Val-Phe-OMe (20) (PDB ID code 3MAT), the butyl side chain of the AHHpA moiety occupies the same region of space, but it does not superimpose because the configuration of the β-amino carbon is opposite that of NleP and Met. Nevertheless, the β-amino group of AHHpA-Ala-Leu-Val-Phe-OMe acts as a ligand toward M2, whereas the α-hydroxyl group bridges M1 and M2. Presumably, these interactions contribute greatly to transition-state mimicry by this ligand and hence to its potency as an inhibitor. Another noticeable difference is that His-79 moves considerably closer to one of the phosphonate oxygens in both mono- and dimetalated structures. Movement of His-79 to this position makes it possible to deliver a proton to the amide nitrogen of the scissile bond.

Fig. 3.

Chemical and x-ray structures. (A) Chemical structures of l-methionine (Met), NleP, and the peptide inhibitor AHHpA-Ala-Leu-Val-Phe-OMe. (B) Stereoview of overlay of E. coli MetAP structures of the mono-Mn(II) form with NleP (red), the di-Mn(II) form with NleP (green), and the di-Co(II) form with AHHpA-Ala-Leu-Val-Phe-OMe (yellow; only the AHHpA-Ala-Leu moiety shown here was observed in the crystal structure).

Structure of an Unliganded MetAP.

A key structural feature highly relevant to catalysis by MetAP, as well as by other metalloaminopeptidases, is a coordinated water molecule (or hydroxide ion), seen bridging the two metal ions in numerous crystal structures of these enzymes. This water is proposed to function as an active-site nucleophile that attacks the carbonyl group of the scissile peptide bond (19, 21). Because the monometalated enzymes are also highly active catalytically, a question arises about the existence and location of this coordinated water. We therefore also sought, and by limiting the concentration of Mn(II), obtained, crystals of the unliganded enzyme in which the M1 site is substantially filled, whereas the M2 site has much reduced occupancy (Fig. 2E). A single water molecule is clearly seen coordinated to M1, but this water is, on average, now much closer to M1 and farther from the mostly unoccupied M2 site. The distances to M1 and M2 are 3.74 and 4.04 Å, respectively, in molecule A and 2.16 and 3.16 Å, respectively, in molecule B [two protein molecules (A and B) are in one asymmetric unit of the crystal]. In the absence of a substrate-analog ligand, M1 is now clearly pentacoordinate, which is in agreement with findings from EPR and extended x-ray absorption fine-structure studies in solution (29, 30, 33, 34).

Catalysis by a Monometalated MetAP.

Based on the liganded and unliganded structures reported here and the structure of the previously reported dimetalated enzyme complexed with AHHpA-Ala-Leu-Val-Phe-OMe, the binding of a typical peptide substrate (Met-Ala-Leu) and its tetrahedral intermediate for hydrolysis by the monometalated E. coli MetAP can be modeled as shown in Fig. 4A and B, respectively. The coordinates of the protein active-site residues are those of the liganded monometalated structure. The coordinates for Met of the tripeptide substrate are based on those of NleP in the monometalated structure, whereas those for Ala and Leu are adopted from AHHpA-Ala-Leu-Val-Phe-OMe in the dimetalated structure. The conformation of Met at the S1 site is well supported by several available structures in the di-Co(II) form of MetAP (32). However, the structure of E. coli MetAP complexed with AHHpA-Ala-Leu-Val-Phe-OMe is the only one that provides the binding mode for the substrate residues at the prime S1′ and S2′ sites of MetAP. The bound conformation shown for the prime-side residues is supported by the x-ray structures of bovine lens leucine aminopeptidase complexed with amastatin (PDB ID code 1BLL) (35) and that of aminopeptidase from Aeromonas proteolytica complexed with bestatin (PDB ID code 1TXR) (36).

Fig. 4.

Proposed models and catalytic mechanism. Proposed models of tripeptide substrate Met-Ala-Leu (A) and its tetrahedral intermediate (B) bound to the active site. A schematic drawing of the proposed catalytic mechanism of monometalated MetAP is shown in C.

To form a Michaelis complex with the monometalated enzyme, the peptide substrate likely approaches M1 with the scissile peptide bond in a trans conformation (Fig. 4A). In the absence of M2, the negatively charged side chains of Asp-97 and Asp-108 are moved closer to the positively charged amino terminus of the peptide by the developing charge–charge interaction. The imidazole ring of His-79 moves toward, and hydrogen bonds to the scissile amide nitrogen, while the imidazole of His-178 forms a hydrogen bond to the oxygen of the scissile carbonyl group. Interaction of this oxygen with M1 increases the coordination number of M1 from five to six. With the assistance of Glu-204 as a general base, the metal-coordinated water acts as a nucleophile and attacks the re-face of the carbonyl group to generate a tetrahedral intermediate (Fig. 4C). Interactions between the tetrahedral intermediate and active-side residues are depicted in Fig. 4B.

In the catalytically active monometalated enzyme, Asp-97 is no longer needed for metal ion coordination, yet the D97A mutant has only 4% of the activity of the wild-type enzyme (37), and mutation of its counterpart Asp-219 in Saccharomyces cerevisiae type I MetAP to the neutral Asn reduces activity by 1,000-fold (38). Thus, it is apparent that the major role of Asp-97 is to provide negative charge density, along with Asp-108, to interact with the protonated amino-terminal amino group and orient the peptide substrate for productive binding. The movement of His-79 toward the amide nitrogen allows it to stabilize both the Michaelis complex and the tetrahedral intermediate of hydrolysis and to deliver a proton to the nitrogen to facilitate the breakdown of the tetrahedral intermediate. The importance of this role is consistent with the substantially reduced activity of the H79A mutant (39). It is interesting to note that during the catalysis, coordination of M1 changes from pentacoordinate to hexacoordinate after binding of the substrate. Both pentacoordinate and hexacoordinate metal environments have been seen in MetAP structures reported here and earlier with different inhibitors bound (18, 20, 32). Further support for the scissile carbonyl oxygen binding to M1 at the same time as a water molecule comes from a study on P. furiosus MetAP by EPR, which showed that both nucleophile and substrate bind to a catalytic Fe(II) center in that enzyme (40).

This proposed mechanism of catalysis by monometalated MetAP modifies the mechanisms proposed in refs. 20 and 41 for the dimetalated enzymes in several key respects. First, M2 is no longer required (or present). Instead of the unprotonated amino terminus of the substrate coordinating to M2, charge–charge interactions involving Asp-97 and Asp-108 with the protonated amino terminus position and bind the substrate. Kinetic measurements suggested that the terminal amino group is extensively protonated at the assay pH (7.0) and that the protonated form binds to the enzyme more tightly (41). It might be expected that the positively charged amino terminus would move toward the void generated by the absence of M2, but instead, the side chains of Asp-97 and Asp-108 shift toward the amino group. The carboxyl group of Glu-204 is positioned to act as a general base to deprotonate the water molecule coordinated to M1, allowing it to act as a nucleophile and add to the scissile carbonyl. In contrast to earlier suggestions, Glu-204 is not the hydrogen donor to the departing amino group because it is too far away to form a hydrogen bond to the amide nitrogen. His-79 shows some mobility in different x-ray structures that allows it to reach, through hydrogen bonding, either the amide nitrogen or the carbonyl oxygen of the P1′ Ala residue of the substrate. However, it does not form a hydrogen bond with the oxygen (20) or merely hydrogen bond to the nitrogen (41); instead, it likely participates actively in catalysis by donating a proton to the departing amino group through a charge relay involving the imidazole ring of His-79, a possibility that has been recognized earlier (21). His-79 is also the residue that is covalently modified by fumagillin, and in this case, the N^ε2 nitrogen of the imidazole of His-79 serves as a nucleophile toward the oxirane moiety of fumagillin (8). A similarly positioned histidine in arginase has been suggested as a proton donor (42), and the apparent pK_a of His-79 at 7.9, as determined by Copik et al. (41), supports its function as a proton donor in the catalysis.

Potential Problems in Structures of Dimetalated Metallohydrolases.

MetAP and other metalloaminopeptidases belong to the superfamily of dimetalated (or dinuclear/binuclear) metallohydrolases that includes not only peptidases but also phosphohydrolases such as Ser/Thr phosphatases and nucleases (43). In addition to the metalloaminopeptidases that are catalytically active as monometalated enzymes, several metallophosphatases, such as the Klenow fragment (43, 44) and ribonuclease H (45), also require only one metal ion for activity. Although some of the metallohydrolases may still use two metal ions, catalysis with a single metal ion is certainly a plausible mechanism based on the available kinetic data and the structural information reported here. It is not uncommon to see large differences in affinity between the two metal sites. For example, both clostridial aminopeptidase and aminopeptidase P bind only one Co(II) per enzyme molecule at dissociation constants of 52 and 15 μM, respectively (46). The dissociation constants for the binding of Co(II) to the first and second metal sites of β-lactamase are estimated to be 0.14 and 2.52 mM (47), and DNA polymerase I has a K_d of 2.5 μM for its tighter site and a K_d of 600 μM for its weaker site (48). Nevertheless, the structures most often reported for these and other enzymes are dimetalated. Given the nonphysiologically high concentrations of both protein and metal that are generally used for growing crystals (very much higher than those that exist in cells), the crystallization of dimetalated forms of the enzymes may, at least in some cases, be a misleading experimental artifact. Such consideration must be taken into account when structures are generated to aid in elucidation of reaction mechanisms or to support structure-aided drug design efforts.

Materials and Methods

Preparation and Quantitation of E. coli MetAP.

The recombinant E. coli MetAP was purified as an apoenzyme (14). The protein concentration was determined by the Bradford method with a kit from Bio-Rad using BSA as the protein standard. To ensure the accuracy of the protein concentration, we also calculated it by titration of the apoenzyme with 5,5′-dithiobis(2-nitrobenzoic acid) (DTNB), based on the known seven cysteine residues in the purified apoenzyme. In the titration, the absorbance at 412 nm was determined when the apoenzyme (≈20 μM) was mixed with 2 mM DTNB in a solution containing 6 M guanidine·HCl and 50 mM Mops (pH 7.0), and the amount of protein was derived from a standard curve obtained with known amounts of homocysteine. The Bradford method and the DTNB titration gave similar numbers for protein concentration.

Enzyme Activity Determination.

The fluorogenic substrate Met-AMC was purchased from Bachem. In the metal-titration experiments, the enzyme activities were monitored with a fluorescence assay by using Met-AMC (14, 49) on a 384-well microplate. Briefly, an 80-μl assay mixture in each well contained 50 mM Mops (pH 7.0), 200 μM Met-AMC, 20 μM apoenzyme for CoCl₂ or 75 μM for MnCl₂, and increasing amounts of metal ions (CoCl₂ or MnCl₂). The formation of AMC was monitored continuously for 10 min at room temperature by fluorescence (λ_ex 360 nm, λ_em 460 nm) on a SpectraMax Gemini plate reader (Molecular Devices).

Crystallization Conditions.

Both NleP and leucine phosphonate (LeuP) were purchased as racemic mixtures from Fisher. Initial crystallization conditions were determined by using Crystal Screen and Index HT kits in 96-well sitting-drop plates (Hampton Research, Aliso Viejo, CA) at room temperature. Final crystals of the E. coli MetAP·NleP complexes and the unliganded E. coli MetAP were obtained independently by the hanging-drop vapor-diffusion method at 21–23°C. For the E. coli MetAP·NleP complexes, NleP (75 mM in H₂O) was added to concentrated apoenzyme (7.5–15 mg/ml, 0.25–0.5 mM) in 10 mM Mops (pH 7.0) at an NleP/apoenyzme concentration ratio of 9:1. Hanging drops contained 3 μl of the protein solution mixed with 3 μl of reservoir solution. The reservoir solution consisted of 10–17% polyethylene glycol 20000, 0.1 M Mes (pH 6.5), and various amounts of MnCl₂ for metal/apoenzyme concentration ratios of 0.25:1, 0.5:1, 1:1, and 2:1. The crystal of the unliganded enzyme was obtained under the same condition but in the presence of LeuP instead of NleP (the LeuP/apoenzyme concentration ratio was 160:1). The metal/apoenzyme ratio was 0.5:1 during crystallization of the unliganded enzyme.

Data Collection and Structural Refinement.

The personnel at the Protein Structure Laboratory at the University of Kansas assisted in collecting the reflection data. Data were collected on an R-Axis IV imaging plate detector (Rigaku, Tokyo) with a rotating anode generator operated at 50 kV and 100 mA. Images were recorded over 180° in 0.5° increments at 100 K. The raw reflection data were processed by using mosflm and merged and scaled by using scala in CCP4i (50). Cell analysis of the data indicated two molecules of the enzyme per asymmetric unit for all crystals examined. The coordinates of our previously solved structure of E. coli MetAP (PDB ID code 1XNZ), with ligand, metal ions, and water molecules removed, were used as the search model for molecular replacement by using molrep in CCP4i. Crystallographic refinement was performed with cns software (51). The refinement was monitored with free R factor throughout the whole refinement process, with 5% of the total number of reflections set aside. Initial refinement started with simulated annealing at a temperature of 4,000 K and a 25 K drop in temperature per cycle. The models were refined with iterative cycles of individual B factor refinement, positional refinement, and manual model building by using wincoot (52). The Mn(II) atoms were not included in the initial refinement procedure to reduce the model bias in phases, and they were then added to the model to the center of the peak in the Mn(II)-omitted F_obs − F_calc electron density map. The coordinate file for ligand NleP, along with topology and parameter files, was obtained from the HIC-Up database (53). The ligand was added by using wincoot when the electron densities shown in 2F_obs − F_calc and F_obs − F_calc maps for their placement were unequivocal. These electron density maps were examined with different contour levels. The 2F_obs − F_calc map was contoured at 1.0σ, whereas the F_obs − F_calc map was contoured at 3.0σ and −3.0σ. Model building and crystallographic refinement were performed iteratively. The occupancies of metal ions were refined (described in more detail below) after the addition of the ligand and water molecules. The initial crystallographic R factors were ≈33%, and the free R factor continued to drop as the refinement progressed. The resultant 2F_obs − F_calc maps showed clear electron densities for most of the atoms except for a few side chains at the molecular surface. The final models for all of the structures were analyzed by using the program procheck (54), and all of them showed that 99.6% of residues were in allowed Ramachandran plot ranges. Two protein molecules in the asymmetric unit were refined independently. Overlay of the two molecules showed no significant differences except for a few surface side chains. Although racemic NleP was used in the crystallization of the complexes, only l-NleP was seen in the complex structures. For the unliganded structure, although LeuP was present during crystallization, no electron density for LeuP in the refined structure was observed. The atomic coordinates and structure factors for the liganded and unliganded structures have been deposited in the Protein Data Bank (PDB ID codes 2GTX, 2GU4, 2GU5, 2GU6, and 2GU7). All drawings for protein structures in the figures were generated by using pymol (DeLano Scientific, South San Francisco, CA). Statistic parameters in data collection and structural refinement are shown in Table 1.

Calculation of Metal Occupancies.

The occupancies of the metal ions were refined by using cns. At the final stages of structure refinement, B factors of the atoms of residues Asp-97, Asp-108, His-171, Glu-204, and Glu-235, which are residues directly ligating to the metal ions, were averaged, and the averaged B factor value was used as the B factor value for the metal ions for their occupancy refinement. Then, with the refined occupancies of the metal ions, individual B factor refinement for each individual metal ion was carried out. The occupancy for each metal ion, in molecule A and molecule B in the asymmetric unit, was refined independently. The occupancy refinement and the B factor refinement were performed iteratively for several cycles until they converged to consistent values. The B factors from the refinement and the corresponding occupancies are listed in Table 2, which is published as supporting information on the PNAS web site.

Abbreviations

AHHpA

(3R)-amino-(2S)-hydroxyheptanoic acid

AMC

7-amino-4-methylcoumarin

DTNB

5,5′-dithiobis(2-nitrobenzoic acid)

LeuP

leucine phosphonate

MetAP

methionine aminopeptidase

NleP

norleucine phosphonate

PDB

Protein Data Bank.