Mutations at the S1 sites of methionine aminopeptidases from Escherichia coli and Homo sapiens reveal the residues critical for substrate specificity.

Methionine aminopeptidase (MetAP) catalyzes the removal of methionine from newly synthesized polypeptides. MetAP carries out this cleavage with high precision, and Met is the only natural amino acid residue at the N terminus that is accepted, although type I and type II MetAPs use two different sets of residues to form the hydrophobic S1 site. Characteristics of the S1 binding pocket in type I MetAP were investigated by systematic mutation of each of the seven S1 residues in Escherichia coli MetAP type I (EcMetAP1) and human MetAP type I (HsMetAP1). We found that Tyr-65 and Trp-221 in EcMetAP1, as well as the corresponding residues Phe-197 and Trp-352 in HsMetAP1, were essential for the hydrolysis of a thiopeptolide substrate, Met-S-Gly-Phe. Mutation of Phe-191 to Ala in HsMetAP1 caused inactivity in contrast to the full activity of EcMetAP1(Y62A), which may suggest a subtle difference between the two type I enzymes. The more striking finding is that mutation of Cys-70 in EcMetAP1 or Cys-202 in HsMetAP1 opens up the S1 pocket. The thiopeptolides Leu-S-Gly-Phe and Phe-S-Gly-Phe, with previously unacceptable Leu or Phe as the N-terminal residue, became efficient substrates of EcMetAP1(C70A) and HsMetAP1(C202A). The relaxed specificity shown in these S1 site mutants for the N-terminal residues was confirmed by hydrolysis of peptide substrates and inhibition by reaction products. The structural features at the enzyme active site will be useful information for designing specific MetAP inhibitors for therapeutic applications.

Protein biosynthesis is initiated with either Met (in eukaryotes) or N-formylmethionine (in prokaryotes, mitochondria, and chloroplasts) (1). For a significant fraction of intracellular proteins, the N-terminal Met is removed enzymatically after the initiation of translation. Posttranslational modifications, including Met removal, acetylation, and myristoylation, have a significant impact on protein localization, activation, stability, function, and degradation (1).
Considering the importance of Met removal, it is not surprising that MetAP shows stringent specificity for the N-terminal Met (2)(3). No natural amino acid residues other than Met are accepted at this position. The formyl group in formylmethionine must be removed before Met can be cleaved by MetAP (4). MetAP from Salmonella typhimurium cleaved Met(sulfoxide)-Ala-Ser, Met(sulfone)-Ala-Ser, and Nle-Gly-Gly with efficiencies of 41, 35, and 19%, respectively, compared with the tripeptide Met-Ala-Ser, which is defined as 100% (4). Leu-Gly-Gly could not be cleaved. Acceptance of norleucine (Nle) at this S1 binding pocket has been supported by the observation of Yang et al. (5) that human type II MetAP (HsMetAP2) could hydrolyze not only Met-pNA (k cat /K m ϭ 14,000 M Ϫ1 min Ϫ1 ) but also an analog, Nle-pNA (k cat /K m ϭ 2,200 M Ϫ1 min Ϫ1 ).
MetAP enzymes also have a strong preference for a small, uncharged amino acid (Gly, Ala, Ser, Thr, Pro, Val, or Cys) as the penultimate residue, and the terminal Met is often removed if Met is followed by one of the small, uncharged amino acid residues. It has been noted that this requirement complements the N-end rule of eukaryotic protein turnover by ubiquitinization (1,6).
MetAPs are grouped into two subtypes according to sequence homology. Prokaryotes have only one MetAP, either type I (eubacteria) or type II (archaea), whereas eukaryotic cells have both type I and type II MetAPs. Structural information is available for four MetAPs, namely the type I MetAPs from Escherichia coli (EcMetAP1) (7,8) and Staphylococcus aureus (9), and the type II MetAPs from Pyrococcus furiosis (10) and human (11). Locations of the S1 and S1Ј substrate binding pockets and the residues involved in forming these pockets have been assigned according to the structures of MetAPs complexed with the ligands derived from peptide substrates. Met, a product from Met removal, and Met analogs (methionine phosphinate, norleucine phosphonate, methionine phospho- nate, and trifluoromethionine) all bind in a similar fashion at the active site of EcMetAP1, which defines the S1 site (8). The peptidic inhibitor AHHpA-Ala-Leu-Val-Phe-OMe, with the Nle analog (3R)-amino-(2S)-hydroxyheptanoic acid (AHHpA) at the N terminus, is derived from the natural product bestatin, and its complex with EcMetAP1 (7) again indicates that the S1 site is able to accommodate a side chain of Nle. It is remarkable to note that type I and type II MetAPs use two different sets of amino acid resides to form the S1 site, and both offer the same stringent specificity for the N-terminal residue for substrates. Available structural information indicates that the S1 site of EcMetAP1 is formed by Cys-59, Tyr-62, Tyr-65, Cys-70, His-79, Phe-177, and Trp-221, and the S1 site in HsMetAP2 is formed by Pro-220, Gly-222, Ile-338, His-382, Met-384, and Tyr-444 (7,8,11,12). Sequence alignment of type I MetAPs from E. coli and human (EcMetAP1 and HsMetAP1) indicates that they share 43% identity (114 of 264 residues) in the peptidase domain ( Fig. 1). Furthermore, six of the seven residues forming the S1 site in EcMetAP1 (Tyr-62, Tyr-65, Cys-70, His-79, Phe-177, and Trp-221) are either identical or conserved in HsMetAP1 (Tyr-194, Phe-197, Cys-202, His-211, Phe-308, and Trp-352).
We have successfully established a unique and efficient E. coli expression for HsMetAP1 (13). As part of our effort to map the active site of MetAP for the discovery of specific inhibitors as antibacterial, antifungal, and anticancer agents, all of the seven S1 site residues in EcMetAP1 and the corresponding residues in HsMetAP1 have been replaced by either Ala or Ser. These mutants have been evaluated for the capability of hydrolyzing Xaa-S-Gly-Phe thiopeptolide substrates (Xaa ϭ Met, Ala, Asn, Asp, Gly, Ile, Leu, Lys, Phe, Pro, Ser, Thr, Tyr, Trp, or Val) and Xaa-Ala-Ser peptide substrates (Xaa ϭ Met, Leu, Phe, or Nle). The S1 site residues that are important for substrate hydrolysis and substrate specificity have been identified.

EXPERIMENTAL PROCEDURES
Sources of Substrates-All of the Xaa-S-Gly-Phe thiopeptolide substrates, where Xaa represents Met, Ala, Asn, Asp, Gly, Ile, Leu, Lys, Phe, Pro, Ser, Thr, Tyr, Trp, or Val, were synthesized in this laboratory, and syntheses of Xaa-S-Gly-Phe will be reported elsewhere. Xaa-Ala-Ser tripeptide substrates, where Xaa is Leu, Phe, or Nle, were prepared by GL Biochem (Shanghai, China), and Met-Ala-Ser was purchased from Sigma.
Expression and Purification of Wild-type EcMetAP1 and Its S1 Site Mutants-The full-length gene of EcMetAP1 was obtained by direct PCR-amplification from an E. coli strain BL21(DE3)pLysS and cloned into pGEMEX-1 vector (Promega) as pGEMEX-1/EcMetAP1 as described previously (14). The S1 site mutants of EcMetAP1 were prepared using the QuikChange site-directed mutagenesis system (Stratagene) for generating the single point mutants C59A, C59S, Y62A, Y65A, C70A, C70S, H79A, F177A, and W221A from the plasmid pGE-MEX-1/EcMetAP1. Recombinant plasmids containing either a wildtype or mutant EcMetAP1 were verified by DNA sequencing and transformed into E. coli strain BL21(DE3)pLysS for expression. The recombinant wild-type EcMetAP1 and the S1 site mutants were all expressed as soluble protein and purified by ammonium sulfate precipitation and Q-Sepharose column separation as described previously (14). Typically, 20 mg of the purified proteins was obtained from a 1-liter culture. Each of the purified proteins appeared as a single band on an SDS-PAGE gel by Coomassie Blue staining.
Expression and Purification of Wild-type HsMetAP1 and Its S1 Site Mutants-The full-length gene of HsMetAP1 was cloned into the vector pGEX-KG (15) as an N-terminal fusion protein with glutathione Stransferase (GST) as described previously (13). To obtain the Hs-MetAP1 mutants, the plasmid pGEX-KG/HsMetAP1 was used as the template to generate plasmids for the individual HsMetAP1 mutants (P191A, Y194A, F197A, C202A, H211A, F308A, and W352A) with the QuikChange site-directed mutagenesis kit. All mutations were verified by DNA sequencing.
The plasmids were transformed into E. coli strain BL21(DE3)pLysS for expression. The GST-fusion proteins were expressed as soluble protein and purified by a GSTrap affinity column as previously described (13). The full-length enzymes were released from the fusion proteins by thrombin treatment. Typically, 10 mg of the final purified proteins was obtained from a 1-liter culture, and each of the proteins appeared as a single band on an SDS-PAGE gel by Coomassie Blue staining.
Colorimetric Assays with Thiopeptolides Xaa-S-Gly-Phe-Hydrolysis of a thiopeptolide such as Met-S-Gly-Phe generates a free thiol group, which reacts with DTNB instantly and causes an increase of absorbance at 412 nm (16). It is a sensitive and convenient assay for monitoring MetAP activity kinetically. To avoid the inactivation of EcMetAP1 by DTNB that was observed previously (16,17), the assay was modified to a two-step procedure. Hydrolysis of the thiopeptolide was allowed to proceed before DTNB was added for color development.
A typical assay was carried out at room temperature on a 96-well plate in a 100-l mixture containing 50 mM MOPS, pH 7.0, 500 M thiopeptolide, 50 M CoCl 2 , and 200 nM EcMetAP1 or 500 nM Hs-MetAP1. After 1 min, a 50-l solution containing 2 mM DTNB, 50 mM MOPS, pH 7, and 6 M guanidine hydrochloride was added to the mixture, and absorbance at 412 nm was determined immediately. The rate of background hydrolysis of the thiopeptolide was subtracted using controls without enzymes. The activity unit was defined as a micromolar product generated in 1 min by 1 M enzyme. HPLC Assays Using Tripeptides Xaa-Ala-Ser-Hydrolysis of Xaa-Ala-Ser tripeptide substrates was assayed by an HPLC method, and substrate depletion was determined by absorbance at 215 nm after the substrate and the products were separated by HPLC. Peak area was used to calculate the hydrolysis of the substrate. All assays were performed at room temperature in a 20-l reaction system containing 50 mM MOPS, pH 7, 50 M CoCl 2 , 8 mM tripeptide, and 5 M EcMetAP1 (wild-type or a mutant) for 5 min or 2 M HsMetAP1 (wild-type or a mutant) for 30 min. The reaction was quenched by the addition of 80 l of 0.1% trifluoroacetic acid, and the mixture was centrifuged at 10,000 ϫ g for 10 min. A reverse-phase C8 column (Eclipse, XDB; 5 m, 4.6 ϫ 150 mm) was used for HPLC analysis. Mobile phase A consisted of water containing 0.1% trifluoroacetic acid, and mobile phase B consisted of 1:1 water to acetonitrile plus 0.1% trifluoroacetic acid. A 10-l sample was loaded onto the column by 5% phase B at a flow rate of 1 ml/min.

Preparation of Wild-type and Mutant MetAP1
Enzymes-The S1 site residues in EcMetAP1 (Cys-59, Tyr-62, Tyr-65, Cys-70, His-79, Phe-177, and Trp-221) have been identified from available x-ray structures (Fig. 2) (7,8,12). However, there is no structural information available for HsMetAP1. Both Ec-MetAP1 and HsMetAP1 belong to the type I MetAP family and share 43% sequence identity (Fig. 1). Six of the seven S1 site residues identified in EcMetAP1 are either identical or conserved in HsMetAP1, and, therefore, it is likely that HsMetAP1 has a structure similar to that of EcMetAP1. We prepared nine single point mutants of
Wild-type and mutant EcMetAP1s were expressed in E. coli as soluble proteins and purified to homogeneity with high yield. Recombinant HsMetAP1s, both wild-type and mutant, were expressed in E. coli as GST-fusion proteins and purified with GSTrap affinity column. HsMetAP1 proteins were released from the fusion proteins by thrombin treatment. Based on SDS-PAGE analyses, the purity of wild-type and mutant Ec-MetAP1 or HsMetAP1 was Ͼ95%.
Residues Important for Hydrolysis of Met-S-Gly-Phe-Seven EcMetAP1 mutants with each S1 residue replaced by Ala were tested on the known substrate Met-S-Gly-Phe first. Three residues (Tyr-65, His-79, and Trp-221), when replaced by Ala, had huge impact on EcMetAP1 activity in hydrolyzing Met-S-Gly-Phe. EcMetAP1(Y65A) and EcMetAP1(W221A) became totally inactive, whereas EcMetAP1(H79A) had very weak activity (Fig. 3). Tyr-65 and Trp-221 are located near each other and have direct contact with the terminal methyl group of the Met side chain (Fig. 2). Their inactivity may indicate that hydrophobic interaction with the methyl group is required for binding and hydrolysis of a substrate. His-79 is in the same vicinity and close to the catalytic metal site, although it is not a metal ligand. In the case of the bestatin-based inhibitor (7), His-79 interacts with atoms in the peptide bond between the P1Ј and P2Ј residues. However, in the transition state analogs (8) this His moves toward the metal center with the potential to form hydrogen bonds with the atoms of the scissile peptide bond (P1 and P1Ј residues). Therefore, the causes for significantly reduced activity of EcMetAP1(H79A) may be more complex.
On the other hand, the other four residues (Cys-59, Tyr-62, C70A, and F177A), when replaced by Ala, showed minimal effect on the activity. EcMetAP1(C59A) and EcMetAP1(Y62A) were as active as the wild-type, and EcMetAP1(C70A) and EcMetAP1(F177A) showed only slightly reduced activity. However, a closer look revealed that both EcMetAP1(C59A) and EcMetAP1(C70A) showed increased K m , along with increased k cat , so that the overall effect on k cat /K m was minimal. All four residues are located on one side of the pocket (Fig. 2).
When corresponding mutants of HsMetAP1 were tested, we found that some results were consistent with those from Ec-MetAP1 mutants (Fig. 3). Mutation of Phe-197 (Tyr-65 in Ec-MetAP1) rendered the enzyme inactive, and mutation of Trp-352 (Trp-221 in EcMetAP1) displayed a significant reduction in activity. HsMetAP1(P191A) was even more active than the wild-type enzyme, indicating that Pro-191 (Cys-59 in Ec-MetAP1) is not essential. This explains why Cys-59 is unique to EcMetAP1 and is not conserved in HsMetAP1 and other MetAPs. HsMetAP1(C202A), like EcMetAP1(C70A), also showed both increased K m and increased k cat as compared with wild-type HsMetAP1, indicating its active participation in substrate hydrolysis.

FIG. 4. Hydrolysis of Met-S-Gly-Phe (A and D), Leu-S-Gly-Phe (B and E), or Phe-S-Gly-Phe (C and F) by wild-type and S1 site mutants of EcMetAP1 and HsMetAP1.
at the S1 site of MetAPs from two different species, which is an important feature that we are exploring for specific MetAP inhibitors.

Residues Important for Maintaining Stringent Specificity for Terminal Met as Revealed by Using Thiopeptolides Xaa-S-Gly-
Phe-It is intriguing how the stringent specificity for a terminal Met is conferred by the S1 site residues. The side chain of Met is largely hydrophobic, with a sulfur atom as a potential hydrogen bond acceptor. However, the sulfur is not involved in hydrogen bonding in the x-ray structures (8). Because Nle-Gly-Gly (4) and Nle-pNA (5) are validated substrates, the side chain of Nle can fit into the pocket as well as that of Met. It is likely that hydrophobic interaction and steric exclusion strongly influence the ability of the S1 pocket to discriminate between Met and other residues. To better probe the pocket, we prepared 14 thiopeptolides with the general structure of Xaa-S-Gly-Phe (Xaa ϭ Ala, Asn, Asp, Gly, Ile, Leu, Lys, Phe, Pro, Ser, Thr, Tyr, Trp, or Val), based on the known substrate Met-S-Gly-Phe.
None of the 14 thiopeptolides (Xaa is a natural amino acid residue but not Met) could be cleaved by wild-type EcMetAP1 or HsMetAP1, except for Thr-S-Gly-Phe, which showed barely detectable hydrolysis (7.3 M/min/M) by HsMetAP1. This result confirms the stringent substrate specificity for the Nterminal residue.
When these thiopeptolides were tested on the S1 site mutants, significant hydrolysis of some thiopeptolides was detected (Fig. 4). Leu-S-Gly-Phe was cleaved readily by EcMetAP1(C70A) and weakly by EcMetAP1(C59A) and EcMetAP1(Y62A). Cys-59 and Tyr-62 in EcMetAP1 are the same two residues that showed no adverse effect on hydrolysis of Met-S-Gly-Phe upon mutation to Ala (Figs. 3 and 4). The effect of the mutation of Cys-70 to Ala on the hydrolysis of Leu-S-Gly-Phe is interesting, because it greatly increased both K m and k cat in the hydrolysis of Met-S-Gly-Phe. EcMet-AP1(C70A) hydrolyzed Leu-S-Gly-Phe (K m , 0.35 Ϯ 0.04 mM; k cat , 20.8 Ϯ 1.0 s Ϫ1 ; k cat /K m , 59,088 M Ϫ1 s Ϫ1 ) as efficiently as wild-type EcMetAP1 hydrolyzed Met-S-Gly-Phe (K m , 0.35 Ϯ 0.01 mM; k cat , 22.5 Ϯ 0.3 s Ϫ1 ; k cat /K m , 63,757 M Ϫ1 s Ϫ1 ). The importance of this Cys in controlling specificity is even more noticeable in HsMetAP1 mutants, and HsMetAP1(C202A) (corresponding to EcMetAP1(C70A)) was the only one among the seven S1 site mutants that cleaved Leu-S-Gly-Phe. The structural difference between the side chains of Met and Leu is a branch at the ␥-carbon in Leu. Conversion of this Cys (Cys-70 in EcMetAP1 and Cys-202 in HsMetAP1) to Ala opens up a space to accommodate the methyl group at the ␥-carbon. From the results of this study, it is not known how the side chain of Leu fits actually into the pocket after the Cys to Ala mutation. The x-ray structures of EcMetAP1 complexed with a series of Met analogs (methionine phosphinate, norleucine phosphonate, methionine phosphonate, trifluoromethionine, and Met) (8), indicating that there is no significant rotation of the bonds of the side chain. Indeed, based on the structure of EcMetAP1 complexed with 1-amino-3-(methylmercapto)propylphosphonic acid (MetP), we can fit the methyl group toward this Cys without rotating any bonds of the side chain (Fig. 5). Phe is another amino acid residue with a branch at the ␥-carbon, and Phe-S-Gly-Phe was cleavable to a significant extent only by EcMetAP1(C70A) or HsMetAP1 (C202A) among the S1 site mutants (Fig. 4).
Replacement of Cys by Ala could also disrupt hydrogen bonds due to the removal of the thiol group. To reduce disruption to the S1 site, we also made two Cys to Ser mutants, namely EcMetAP1(C59S) and EcMetAP1(C70S). The side chain of Ser is considerably smaller than that of Cys (OH versus SH), and the hydroxyl group in Ser could function as a hydrogen bond donor or acceptor similar to the thiol group in Cys. The fact that EcMetAP1(C70S) cleaved both Leu-S-Gly-Phe (K m , 0.14 Ϯ 0.01 mM; k cat , 4.04 Ϯ 0.26 s Ϫ1 ; k cat /K m , 29,154 M Ϫ1 s Ϫ1 ) and Phe-S-Gly-Phe (K m , 0.24 Ϯ 0.01 mM; k cat , 4.55 Ϯ 0.22 s Ϫ1 ; k cat /K m , 19,310 M Ϫ1 s Ϫ1 ) at efficiencies similar to that for Met-S-Gly-Phe (K m , 0.68 Ϯ 0.02 mM; k cat , 30.8 Ϯ 0.6 s Ϫ1 ; k cat /K m , 45, 541 M Ϫ1 s Ϫ1 ) suggests that the space created by changing Cys to Ser, although not optimal, is already enough for the side chain of Leu or Phe of a substrate to fit into the S1 site for efficient cleavage.
Confirmation of Change of Substrate Specificity of S1 Site Mutants by Using Tripeptides Xaa-Ala-Ser-Because of the heightened reactivity intrinsic to the thiopeptolides, it becomes crucial to parallel the more convenient thiopeptolide assay with a true peptide-based assay to minimize kinetic bias. We prepared four tripeptides, namely Met-Ala-Ser, Leu-Ala-Ser, Phe-Ala-Ser, and Nle-Ala-Ser, and tested them on the four S1 site mutants that showed activity on the thiopeptolides Leu-S-Gly-Phe and Phe-S-Gly-Phe, along with wild-type enzymes (Table I).
Met-Ala-Ser has been reported as an efficient substrate (K m , 0.38 mM; k cat , 210 min Ϫ1 ; k cat /K m , 550,000 M Ϫ1 s Ϫ1 ; for Hs-MetAP2) (5). Under our assay conditions, it was hydrolyzed by wild-type EcMetAP1 and the EcMetAP1(C59A), EcMetAP1-(Y62A), and EcMetAP1(C70A) mutants, and wild-type Hs-MetAP1 and the HsMetAP1(C202A) mutant. This is consistent with the results from the thiopeptolides, which show that these mutations do not affect the cleavage of a substrate with Met as the P1 residue. It was not surprising to see that Ile-Ala-Ser was hydrolyzed by the wild-type and mutant enzymes listed in Table I, because other data already suggested that the Nle side chain is able to occupy the S1 pocket comfortably (4,5).
It caused some concerns when we did not observe significant hydrolysis of Leu-Ala-Ser by the mutants tested. The difference between Leu-Ala-Ser and Leu-S-Gly-Phe, which is an excellent substrate, is that an amide bond in the peptide is replaced by a thioester bond in the thiopeptolide, in addition to the variations at the P1Ј and P2Ј residues. This HPLC assay is certainly not a sensitive assay, but, nevertheless, it is still an adequate working tool for confirming the results obtained from the thiopeptolides. It is significant that Phe-Ala-Ser was cleaved by EcMetAP1(C70A) and HsMetAP1(C202A) at a rate similar to that of Met-Ala-Ser (Table I), whereas none of the EcMetAP1, HsMetAP1, EcMetAP1(C59A), and EcMetAP1(Y62A) enzymes showed detectable activity. This result clearly demonstrates that a mutation at Cys-70 in EcMetAP1 or Cys-202 in Hs-MetAP1 to Ala opens up the S1 site to accommodate the side chain of Phe not only for a thiopeptolide substrate but also for a peptide substrate.
Product Inhibition of Wild-type and Mutated MetAP1s-MetAP removes Met from peptide substrates, and Met as a reaction product dissociates from the enzyme. The affinity of Met to EcMetAP1 and HsMetAP1 was evaluated for its competition with Met-S-Gly-Phe by the colorimetric assay. Met binds only weakly to EcMetAP1 as indicated by its IC 50 at 28.2 mM, and it binds equally weakly to the EcMetAP1 mutants tested (Table II). Met also binds to HsMetAP1 and the mutant HsMetAP1(C202A).
If the role of Cys-70 of EcMetAP1 or Cys-202 of HsMetAP1 in maintaining the stringent specificity is critical, Leu or Phe as a product from peptide cleavage could conceivably fit into the S1 site also. Phe binds the most strongly to EcMetAP1(C70A) and relatively strongly to EcMetAP1(C70S) and has almost no binding to wild-type EcMetAP1. This order of binding affinity correlates very well with the gradual enlargement of the S1 pocket caused by changing the residue at position 70 from Cys to Ser to Ala. Although the hydrolysis of Leu-Ala-Ser cannot be detected and the affinity of Leu to the enzymes is weak, the rank order of the binding of Leu to wild-type and mutant EcMetAP1s or HsMetAP1s is in good agreement with the data from thiopeptolides and peptides. Ile cannot fit into the S1 pocket, and its inability to bind to the enzymes was expected and confirmed. DISCUSSION By removing the N-terminal Met from nascent polypeptides, MetAP plays a crucial role in protein maturation for proper localization, targeting and eventual degradation in many normal and pathological processes. MetAP has been recognized as a promising target for antibacterial, antifungal and anticancer agents (18).
In eukaryotic cells, type I and type II MetAPs are redundant, and only the deletion of both genes in yeast Saccharomyces cerevisiae is lethal (19,20). Antiangiogenic fumagillin and its analogues specifically target HsMetAP2 (the type II human enzyme) without affecting the type I isozyme HsMetAP1 (21,22), and a semisynthetic derivative, TNP-470, is in clinic trials for cancer therapy (23). Many pathogenic bacteria have only one MetAP, which is type I, and its presence is essential for cell survival. Its essential role has been demonstrated by the lethal deletion of the single MetAP gene in E. coli (24) and S. typhimurium (25). For minimal toxicity, it is desirable to design an inhibitor specific for bacterial MetAP without affecting the human type I enzyme HsMetAP1. Understanding the role of each residue at the enzyme active site in the recognition of substrates and inhibitors and the cleavage of the amide bond in the substrate is crucial in designing a MetAP inhibitor with the desired specificity for therapeutic applications.
Both type I and type II MetAPs have the same "pita bread" fold, and all of the five metal ligands are conserved (12). Marine natural products called bengamides inhibit tumor growth in vitro and in vivo, and their molecular target was recently identified as MetAP (26). LAF389, a synthetic analog, inhibited both type I and type II human MetAPs at submicromolar concentrations. However, selective MetAP inhibitors can be obtained as demonstrated by fumagillin, which covalently modifies His-231 of HsMetAP2 by opening an epoxide ring (11). Covalent modification of the conserved His in the type I enzyme EcMetAP1 (His-79) was achieved only at a much higher concentration (27).  a Rates of hydrolysis of Met-Ala-Ser by EcMetAP1 and HsMetAP1 were 248 Ϯ 9 and 116 Ϯ2 M/min/M, respectively. The detection limit was 10 M/min/M under our assay conditions. It is more challenging to discover inhibitors that discriminate among type I MetAPs, the important targets for antibacterial drug development. The residues forming the S1 site are conserved among type I MetAPs such as EcMetAP1 and Hs-MetAP1 (Fig. 1), and they shape a pocket that accommodates the side chain of Met or Nle. Side chains of other amino acids with a significant difference in size may be barred simply by steric exclusion, and their binding will not produce a productive hydrolytic event. None of the 14 thiopeptolide Xaa-S-Gly-Phe (Xaa is a natural amino acid residue other than Met) substrates was cleavable to a significant extent by either EcMetAP1 or HsMetAP1, confirming the stringent specificity for Met as the terminal residue.
Both Met and Nle have a straight side chain. There are several natural amino acids with a side chain of similar size, and from our data it is likely that Cys-70 in EcMetAP1 or Cys-202 in HsMetAP1 functions as a "gate keeper" to prevent a side chain with a branch, such as that of Leu, Ile, Val, or Phe, from fitting into this pocket. Mutation of this Cys to Ala makes it possible for side chains with a branch at the ␥-carbon to fit into the pocket, and Leu-S-Gly-Phe and Phe-S-Gly-Phe become cleavable substrates. The side chain of Val and Ile with a branch at the ␤carbon was still not able to fit into the S1 site.
Mutants C70S, W221L, and C70S/W221A of EcMetAP1 have been reported previously (28), and they had 54, 27, and 6% activity, respectively, in hydrolyzing Met-Gly-Met-Met as compared with wild-type EcMetAP1. In contrast to the inactivity of EcMetAP1(Y65A) that we observed in this study, the Y65F mutant kept the same hydrolytic activity on the peptide (28). The change from Tyr to Phe was not as dramatic as the change to Ala, which may account for this discrepancy. Our data confirm the importance of Trp-221 in the hydrolysis of a substrate with Met as the P1 residue and additionally reveal the crucial role of Cys-70 in maintaining stringent specificity at the S1 site. Our systematic mutation of all S1 site residues, combined with thiopeptolide and peptide substrates as S1 site probes, has revealed valuable information about the contribution of each residue to substrate recognition and hydrolysis.
The S1Ј site also plays an important role in substrate recognition, which is shallow and only accepts small and uncharged amino acid residues at the penultimate position of substrates. Walker and Bradshaw reported a mutagenesis study on the S1Ј site of ScMetAP1 (29), and Met-329 (Met-206 in EcMetAP1), Gln-356 (Gln-233 in EcMetAP1), and Ser-195 (Ser-72 in Ec-MetAP1) were all mutated to Ala. It is remarkable that the mutants Q356A and M329A were able to cleave N-terminal Met from substrates that have large penultimate residues, an observation similar to what we report here.
There are few inhibitors for bacterial and fungal MetAPs reported in the literature. The substrate-like inhibitor AHHpA-Ala-Leu-Val-Phe-OMe inhibited EcMetAP1 with an IC 50 of 5 M (30). We recently discovered, by using high throughput screening of a diverse compound library, a serious of nonpeptidic small molecules that inhibit EcMetAP1 specifically and show no inhibition for HsMetAP1 (13,31). We are in the process of defining their binding mode on the enzyme and are investigating how these inhibitors exploit the structural features on the enzyme for such specificity. The information gained from this and other studies of the MetAP enzyme active site will shed light on the eventual development of MetAP inhibitors as useful therapeutic agents.