Yeast Methionine Aminopeptidase I

In eukaryotes, two isozymes (I and II) of methionine aminopeptidase (MetAP) catalyze the removal of the initiator methionine if the penultimate residue has a small radius of gyration (glycine, alanine, serine, threonine, proline, valine, and cysteine). Using site-directed mutagenesis, recombinant yeast MetAP I derivatives that are able to cleave N-terminal methionine from substrates that have larger penultimate residues have been expressed. A Met to Ala change at 329 (Met206 in Escherichia coli enzyme) produces an average catalytic efficiency 1.5-fold higher than the native enzyme on normal substrates and cleaves substrates containing penultimate asparagine, glutamine, isoleucine, leucine, methionine, and phenylalanine. Interestingly, the native enzyme also has significant activity with the asparagine peptide not previously identified as a substrate. Mutation of Gln356 (Gln233 inE. coli MetAP) to alanine results in a catalytic efficiency about one-third that of native with normal substrates but which can cleave methionine from substrates with penultimate histidine, asparagine, glutamine, leucine, methionine, phenylalanine, and tryptophan. Mutation of Ser195 to alanine had no effect on substrate specificity. None of the altered enzymes produced cleaved substrates with a fully charged residue (lysine, arginine, aspartic acid, or glutamic acid) or tyrosine in the penultimate position.

Proteins synthesized in eukaryotic cells undergo two common types of co-/post-translational modifications at their N termini: initiator methionine cleavage and N ␣ -acetylation. These reactions are catalyzed by two classes of enzymes, the methionine aminopeptidases (MetAP) 1 and N ␣ -acetyltransferases (1). In combination, they produce four distinct types of N termini: those with and without initiator methionine, and those with and without N ␣ -acetylation. The penultimate residue adjacent to the initiator methionine is the principal factor that determines enzyme specificity and hence which of these four types of N termini the mature protein will possess. Proteins that have signal peptides removed during translocation steps do not retain these modifications. As predicted from mutant cytochrome sequences (2), the seven penultimate amino acid residues with the smallest side chain radii of gyration (glycine, alanine, serine, threonine, proline, valine, and cysteine) direct MetAP to cleave the initiator methionine. In ad-dition, N-terminal glycine, alanine, serine, and threonine residues are usually N ␣ -acetylated. This modification also occurs on the retained methionine of proteins with penultimate aspartic acid, glutamic acid, and asparagine (3)(4)(5). Prokaryotes have an initiator methionine cleavage pattern identical to that of the eukaryotes (5, 6); however, they N ␣ -acetylate very few proteins and apparently have individual acetyltransferases for each substrate (1).
There are two major types of MetAPs that have been identified with this substrate cleavage pattern and both are expressed in eukaryotes. In addition, the type I enzymes are found in Eubacteria, while the type II enzymes are found in Archaea (8). Comparison of the Escherichia coli structure (9) with that of Pyrococcus furiosis (10) reveals that despite their low overall sequence similarity, the type I and type II enzymes possess a very similar fold in the catalytic domain. The most significant difference between these enzymes is a large helical domain insertion on the surface of the protein characteristic of the type II isozymes. The eukaryotic MetAP isozymes are differentiated from their prokaryotic counterparts by an additional N-terminal domain. The eukaryotic type I MetAP has two putative zinc finger motifs in this ϳ12-kDa region (11,12), and the eukaryotic type II enzyme has a highly charged N terminus with alternating polyacidic and polybasic stretches in a similarly sized segment (8). Although it has not yet been demonstrated, these N-terminal extensions may be involved in the association of eukaryotic MetAP isozymes with intracellular structures/organelles such as the ribosome.
Historically, it has been reported that both types of MetAP are Co 2ϩ -dependent metalloproteases, having two metal ions per catalytic unit (13)(14)(15). However, recent experiments have determined that Saccharomyces cerevisiae MetAP I containing Zn 2ϩ in place of Co 2ϩ has substantially higher activity under in vivo conditions than the Co 2ϩ -substituted enzyme, albeit zinc ions are inhibitory at higher (nonphysiological) concentrations (16). Furthermore, unlike the Zn 2ϩ enzyme, the Co 2ϩ enzyme is inactivated by glutathione, which is present in high concentrations in the cytosol, further supporting the view that yeast MetAP I is a Zn 2ϩ -metalloprotease in situ. However, in reconstituted preparations, the Zn 2ϩ and Co 2ϩ -MetAP I preparations act essentially identically.
Deletion of the MetAP gene in prokaryotes is a lethal event (17); in yeast, both the type I and type II genes must be disrupted for lethality, indicating some redundancy in function, at least in that simple eukaryote (18). However, the specific inactivation of MetAP II by the antiangiogenic compounds, fumagillin and ovalicin, indicates some uniqueness in function (19,20). The selectivity of fumagillin for the type II enzymes appears to be a matter of dose-response, since it has recently been shown that E. coli and yeast MetAP I can be inactivated by these reagents (21). Since antiangiogenic compounds have excellent potential in the treatment of cancer, the ability of fumagillin and ovalicin to differentiate between the type I and type II isozymes is of major interest currently.
The substrate specificity of the MetAP isozymes suggests a high degree of selectivity for methionine in the S 1 site 2 with an S 1 Ј site that primarily limits the side chain length of the substrate to Ͻ3.68 Å (6). Although the SЈ 1 pocket is not yet defined in molecular terms, the structure of the E. coli MetAP (9) is a useful guide to predict the catalytic site residues that define the cavity. Fig. 1 depicts the residues that presumably form the penultimate residue-binding site, of which Gln 356 and Met 329 are two such candidates.
In this study, data are presented for mutant forms of S. cerevisiae MetAP I, in which these two residues individually and together have been converted to alanine. A third site, Ser 195 , that is conserved in all of the MetAP isozymes was also mutated to alanine to determine if it has a significant catalytic function. Detailed kinetic analysis has established that the methionine and glutamine-substituted enzymes have an expanded substrate profile, while the serine substitution was without effect on specificity. In addition, evidence is presented that the current MetAP specificity profile, as determined from in situ measurements, although generally correct, should be modified to indicate that the midsized penultimate residues, such as threonine and valine, may not always allow initiator methionine processing and that asparagine in a limited number of cases may direct processing.

EXPERIMENTAL PROCEDURES
Cloning of Yeast MetAP I-MetAP I was cloned from S. cerevisiae total genomic DNA, based on the sequence of Chang et al. (11), using polymerase chain reaction with the following oligonucleotides: ACAG-AATTCAGCACTGCAACTACAACAGTT and ACAGAATTCCTATTTA-ATTCTCTGTCTTGG (Genset, San Diego, CA). Polymerase chain reaction products were restricted with EcoRI, purified on a 1.2% agarose gel, and subcloned into pBluescript II (Stratagene, La Jolla, CA). Several clones were then sequenced using the Sequenase 2.0 protocol (Amersham Pharmacia Biotech). Polymerase chain reaction-generated errors were removed by exchanging restriction fragments among the sequenced clones. The final product was resequenced to verify the accuracy of the corrections and subcloned into the pGEX-5X-1 expression vector (Amersham Pharmacia Biotech) using the EcoRI restriction site.
Expression and Purification of Yeast MetAP I-S. cerevisiae MetAP I fused in frame to glutathione S-transferase was expressed in E. coli BL21 cells (Novagen Inc., Madison, WI) as described previously (22). Briefly, E. coli cells harboring the glutathione S-transferase-MetAP I plasmid was cultured in expression media (Luria-Bertani medium, 2% glucose, 50 g/ml ampicillin) and induced using 0.1 mM isopropyl thio-␤-D-galactoside (Fisher). Cells were harvested and resuspended in phosphate-buffered saline (140 mM NaCl, 10 mM Na 2 HPO 4 , 1.8 mM KH 2 PO 4 , 2.7 mM KCl) and ruptured using a French press. The lysate was then absorbed onto GSH-Sepharose (Amersham Pharmacia Biotech) and then cleaved with factor Xa (Amersham Pharmacia Biotech) to release the glutathione S-transferase domain. The liberated MetAP I was eluted and purified on a Source S FPLC column (Amersham Pharmacia Biotech). Fractions were then flash frozen in liquid nitrogen and stored at Ϫ70°C until needed. The molecular weight of the product was verified using matrix-assisted laser desorption/ionization mass spectroscopy (data not shown).
Determination of MetAP I Initial Velocities and Kinetic Constants-Twenty peptides based on the sequence MXSHRWDW (where X represents each of the 20 normally coded amino acids) were purchased from Quality Controlled Biochemicals (Hopkinton, MA) and purified on a 10-mm Vydac C 4 column (The Separations Group, Hesperia, CA) using an acetonitrile/water gradient with 0.1% trifluoroacetic acid. The purified peptides were assessed using mass spectroscopy and reverse phase high pressure liquid chromatography on a 4.6-mm Vydac C 18 column. Initial velocities were determined as described previously (16), except that assays used 50 M substrate, 50 nM enzyme, and various time intervals (15 s to 30 min), as required by the individual substrates. Kinetic constants were also determined as described previously (16), except various time scales were used as required for each peptide substrate. Due to an interfering reaction of Co 2ϩ and the MCSHRWDW peptide, kinetic constants for this peptide could not be determined using the Co 2ϩ -substituted enzyme. However, the Zn 2ϩ -substituted enzyme prepared as described previously (16) was able to cleave this peptide (data not shown).

RESULTS
The evaluation of the kinetic behavior of the native and mutant forms of recombinant S. cerevisiae MetAP I can be divided into two major sections based on normal and nonnormal MetAP substrates, defined as follows. Normal substrates are those that possess penultimate residues reported to specify N-terminal methionine cleavage (glycine, alanine, serine, threonine, proline, valine, and cysteine), and nonnormal substrates are those that possess penultimate residues generally considered to prevent methionine removal (aspartic acid, glutamic acid, phenylalanine, histidine, isoleucine, lysine, leucine, methionine, asparagine, glutamine, arginine, tryptophan, and tyrosine) (2)(3)(4)(5).
Kinetic Parameters of Yeast MetAP I with Normal Substrates-As seen in Fig highest turnover rates on substrates with penultimate alanine or serine. The S195A mutant has a similar substrate preference (Fig. 2B), but its average turnover is only about two-thirds that of the native enzyme (Table I). In contrast, with the exception of the valine substrate, the M329A mutation produces turnover rates very similar to those of the native enzyme ( Fig. 2C and Table I). The MetAP derivatives with Q356A and M329A/Q356A mutations have almost identical substrate preference patterns (Fig. 2, D and E, respectively) but with turnover rates that are only about one-third that of the native enzyme (Table I). Additionally, the k cat values for the threonine and valine substrates, whose side chains are somewhat larger, are exceptionally depressed, being less than 12 and 5% of the native enzyme, respectively.
For native MetAP I, substrates among the cleaved subset with larger penultimate residues (threonine, proline, and valine) have higher K m values than the smaller residues (glycine, alanine, and serine) (Fig. 3A). All mutant enzymes show a similar K m profile (Fig. 3, B-E) and have average K m values comparable with the native enzyme (Table I). However, the enzyme with the M329A/Q356A double substitution has a low K m for the threonine substrate (27% of native), mimicking the S195A enzyme (34% of native) much more than the enzyme with the Q356A substitution (149% of native) that it otherwise resembles (Fig. 3, E, B, and D, respectively).
All five enzymes have similar substrate specificity profiles with respect to catalytic efficiency (k cat /K m ); the three smallest penultimate residues (glycine, alanine, and serine) having the highest values in every case (Fig. 4, A-E). This is due to a combination of both higher k cat and lower K m values for these substrates. However, as shown in Table I, the S195A and M329A mutant enzymes have noticeably improved catalytic efficiencies (131 and 150% of native), while the mutant enzymes with Q356A and M329A/Q356A substitutions have substantially lower catalytic efficiencies (37 and 31% of native, respectively).
Velocity of Methionine Hydrolysis with Nonnormal Substrates-In agreement with previous reports (2)(3)(4)(5), native MetAP I does not cleave N-terminal methionine from the nonnormal subset of peptides with the notable exception of the asparagine-containing substrate, which was cleaved at a low but significant rate (Fig. 5A). The S195A enzyme behaves almost identically to the native enzyme on the nonnormal substrates, including showing a significant activity with the asparagine peptide (Fig. 5B). The enzymes with M329A, Q356A, and M329A/Q356A mutations, however, are very effective at cleaving many of these nonnormal peptides showing significant activity toward all but substrates with fully charged penultimate residues (aspartic acid, glutamic acid, lysine, and arginine) and with tyrosine (Fig. 5, C-E). The M329A enzyme is more effective with the asparagine, leucine, isoleucine, and phenylalanine substrates than either of the enzymes with the Q356A or M329A/Q356A mutations, while the latter are better at cleaving histidine, glutamine, and tryptophan-containing peptides than the former. Consistent with the activities on the normal substrates, the enzyme with the Q356A substitution behaves almost identically to the one with the M329A/Q356A double mutation. Both of these enzymes cleave the MMSHR-WDW substrate with velocities that are comparable with the wild type enzyme acting on normal substrates and far exceed the velocity of other nonnormal substrates with any of the enzymes in this study.
Kinetic Constants of MetAP I for Nonnormal Substrates-As seen in Fig. 6, A-C, native MetAP I has measurable activity with the nonnormal asparagine peptide. However, its k cat is 37-fold lower than the slowest normal substrate (valine); combined with a 2.8-fold lower K m , its catalytic efficiency is over 13-fold lower than for the valine peptide. The enzyme with the S195A mutation behaves almost identically to the native enzyme having similar kinetic constants for the asparagine peptide. However, the enzymes with the M329A, Q356A, and M329A/Q356A mutations are much more effective than either the native or the S195A mutant enzymes on the asparagine substrate, having k cat values that are 8 -20-fold higher. The improvement in catalytic efficiency on this substrate (7-12fold) is principally manifested through higher k cat constants, since the K m values do not vary much between the enzymes.
As expected, native MetAP I has no measurable activity toward the nonnormal substrate MMSHRWDW, and the enzyme with the S195A mutation has only very low activity (Fig.  7). In contrast, enzymes with M329A, Q356A, and M329A/ Q356A mutations display substantial activity, with k cat values (400 -525 min Ϫ1 ) as good as or better than three of the six normal MetAP substrates measured with the native enzyme (Fig. 4E). Importantly, these represent minimal k cat values, since the enzyme produced two products during the reaction (MSHRWDW and SHRWDW). These mutants have K m values ranging from 25 to 65 M that are comparable with those of the native enzyme with normal substrates (23-254 M) (Fig. 3A).

FIG. 2. Turnover numbers for S. cerevisiae MetAP I on normal substrates.
A, native; B, S195A; C, M329A; D, Q356A; and E, M329A/Q356A. Assays were conducted as described under "Experimental Procedures" using substrates of the family MXSHRWDW, where the penultimate (X) amino acid is indicated below each bar. Catalytic constants were determined using Linweaver-Burk analysis. Three k cat values were determined for each sample, and the error bars were calculated as ϮS.E.
The combination of these constants yields catalytic efficiencies ranging from 8.5 to 15.7 M Ϫ1 min Ϫ1 , which is superior to the native enzyme with the normal substrates (1.3-7.6 M Ϫ1 min Ϫ1 ) with the larger threonine, proline, and valine residues in the penultimate position. DISCUSSION An octapeptide substrate family with all 20 possible penultimate residues was used to detail the in vitro specificity of yeast recombinant MetAP I. Consistent with previous in situ studies of initiator methionine cleavage patterns, yeast MetAP I is capable of cleaving N-terminal methionine from substrates that have penultimate residues with small side chain radii of gyration (glycine, alanine, serine, threonine, proline, valine, and cysteine) in vitro (2)(3)(4)(5). The enzyme cleaves methionine from substrates with penultimate serine and alanine much more efficiently than the other permissive substrates, due to a combination of higher k cat values and lower K m values for the smaller substrates. It is perhaps not surprising, therefore, that Met-Ala and Met-Ser N termini dominate the sequences found in the yeast genome, particularly among the metabolic or "housekeeping" enzymes that characterize the cytoplasmic protein population. Removal of initiator methionines from these proteins would return a significant amount of energetically expensive, free methionine to the amino acid pool, allowing other functions dependent on methionine, including new protein synthesis, to occur.
Although the k cat values measured are comparable with previous studies, the K m values determined for normal substrates range from 23 to 253 M, which is orders of magnitude lower than previous studies (2680 -6560 M) (12,23). This may be due to the use of longer, octapeptide substrates, compared with the three-and four-residue substrates used by others, possibly indicating that residues well downstream of the penultimate residue may contribute to substrate binding to the enzyme. ; and E, M329A/Q356A. Assays were conducted as described under "Experimental Procedures" using substrates of the family MXSHRWDW, where the penultimate (X) amino acid is indicated below each bar. Initial velocities were determined by linear regression analysis. Three initial velocities were determined for each sample, and the error bars were calculated as Ϯ S.E. Note that the initial velocities for the MMSHRWDW peptide represent the lower limit of the velocity, since two products are formed during the reaction (MSHRWDW and SHRWDW). This may primarily operate through backbone interactions on the substrate and thus be largely sequence-independent.
It was also found that MetAP I can cleave substrates with penultimate asparagine, indicating that the established specificity pattern for co-/post-translational modification may have a greater degree of flexibility than originally detected (2,7). Analysis of cytosolic proteins for whom the N terminus has been directly sequenced has shown that, although proteins with penultimate glycine, alanine, serine, cysteine, and proline are completely processed, threonine and valine result in retention of the initiator methionine residue 15 and 60% of the time, respectively (24). An additional study using recombinant methionyl-tRNA synthetase with 20 different penultimate residues, has shown that midsized penultimate residues such as asparagine, aspartic acid, leucine, and isoleucine are partially processed in E. coli (25). Taken together, this may indicate that midsize penultimate residues such as threonine, valine, and asparagine may only specify methionine removal part of the time, possibly dependent on other downstream determinants.
In this study, the E. coli MetAP structure (9) was instrumental in developing a mutagenesis strategy for the S. cerevisiae MetAP I enzyme although their catalytic domains are only 39% identical. However, sequence alignments with 12 other type I MetAPs also aided in developing the mutagenesis strategy. Residues in the active site cavity that were highly conserved and appeared to define cavity perimeters were considered good candidates for site-directed mutagenesis to alter the substrate specificity of the enzyme. Since the M329A and Q356A mutations allow yeast MetAP I to cleave six and seven of the 13 nonnormal substrates, respectively, it is clear that these residues have a significant role in determining if a substrate has the correct size penultimate residue. However, Met 329 does not appear to have any other significant role in catalysis, since the M329A enzyme is at least as effective as the native enzyme in catalyzing N-terminal methionine removal from normal substrates. In contrast, Gln 356 may serve a dual role with respect to substrate specificity and involvement in the normal catalytic reaction, since conversion of this residue to alanine results in a . Assays were conducted as described under "Experimental Procedures." Catalytic constants were determined using Linweaver-Burk analysis. Three constants were determined for each sample, and the error bars were calculated as ϮS.E. Note that the turnover numbers for the MMSHRWDW peptide represent the lower limit of the k cat , since two products are made during the reaction (MSHRWDW and SHRWDW). loss of about two-thirds of the turnover rate. This may be due to an involvement of Gln 356 in positioning of the substrate in the active site for optimum catalytic efficiency.
There are several features that distinguish the M329A and Q356A mutants. The Q356A mutant is able to cleave substrates with the very large tryptophan residues, while the M329A enzyme cannot. Although M329A and Q356A are positioned side by side in the enzyme, forming a common wall (Fig.  1), the replacement of Gln 356 with the smaller side chain allows a large tryptophan to fit into the penultimate residue-binding site. In addition, the Q356A mutant cleaves histidine-containing substrates 30-fold more efficiently than the M329A mutant. It is possible that histidine also fits into the pocket formed by the missing glutamine, allowing reformation of the glutamine hydrogen bonding network. Consistent with this hypothesis, the M329A enzyme is much more efficient with phenylalanine and isoleucine substrates than is the Q356A mutant. These hydrophobic residues, which are about the same size as the missing methionine, may fill the cavity created by the M329A mutation. Gln 356 has an important role in this hydrophobic interaction, since the M329A/Q356A mutant is not nearly as effective at cleaving the phenylalanine and isoleucine peptides as the M329A single mutant. This effect may be due to a hydrophobic interaction between the adjacent Met 329 side chain (or the hydrophobic substrate in the M329A mutant) and the ␤and ␥-carbons of Gln 356 . Not surprisingly, Met 329 does not play a significant role in the Gln 356 hydrogen bonding network, since the double mutant is as effective as the Gln 356 single mutant at cleaving the histidine substrate.
It is clear from these studies that size of the penultimate residue is not the only determinant of N-terminal methionine cleavage, since all enzymes tested are highly selective against residues expected to be fully charged at the pH of the reaction mixture. None of the five MetAP I enzymes tested has significant activity toward aspartic acid, glutamic acid, lysine, and arginine although aspartic acid and glutamic acid are not exceptionally large residues. Although the reason for this is uncertain, it may simply be due to charge repulsion. The shorter aspartic acid and glutamic acid residues may be repulsed by Glu 327 on the wall of the pocket, and the longer lysine and arginine side chains may be repelled by His 308 at the bottom of the pocket. Although mutation of Glu 327 would likely be very detrimental to the enzyme due to its involvement in metal coordination, conversion of His 308 to alanine could prove instructive in this regard.
The information gained from these site-directed mutagenesis studies has allowed a reexamination of the structure of E. coli MetAP I to identify other residues that may be involved in  (Table II). Tyr 291 forms a wall in this cavity and is conserved as tyrosine or phenylalanine in all 13 type I MetAP sequences known; all type II MetAPs have a conservative substitution of leucine in this position. His 308 is at the bottom of this pocket and may define the maximum length of the penultimate residue that can be bound and, therefore, may be the residue responsible for discrimination between phenylalanine and tyrosine in the M329A mutant. His 308 is conserved as either asparagine or histidine in all type I enzymes and as asparagine in all type II MetAPs. Glu 327 serves a dual function as one of the metal coordinating residues as well as forming a wall in the putative penultimate residue binding pocket; Glu 327 is completely conserved in all MetAPs. An absolutely conserved Gly 293 is also part of the wall of this pocket. The absence of a side chain on Gly 293 may be required to keep the cavity open, making it an excellent candidate for creating a MetAP mutant with a more limited specificity. The backbone of two residues, Cys 292 and Pro 328 , also complete this pocket. The Pro 328 residue is completely conserved in the type I enzymes and as proline or threonine in five and three of the type II enzymes, respectively. The Cys 292 is less conserved, with 7 of 13 residues being cysteine in the type I enzymes with no particular pattern for this residue in the type II enzymes. Since the side chain of Cys 292 is solvent-accessible, absolute conservation of this residue is not critical for maintaining the cavity shape. Alignment of the type I E. coli and type II P. furiosis enzymes reveals that all of these putative penultimate binding site residues occupy the same spatial geometry maintaining a similar cavity shape although these enzymes are highly divergent (16% identity) (data not shown).
The use of sequence alignments in conjunction with an enzyme crystal structure has proven to be a powerful tool in determining enzyme structure-function relationships. Two out of the three point mutants studied produced novel activities for the enzyme. One of the substrates, not normally cleaved by the native enzyme, was hydrolyzed with kinetic constants superior to three of the six normal substrates for MetAP I. Furthermore, one of these mutants (M329A) retained at least as much activity for the normal substrates as the native enzyme.
MetAP has also been studied for its role in the production of recombinant proteins, since incorrect processing of the N terminus can produce proteins that are inactive or immunogenic (40,41). Overexpression of recombinant proteins can overload the ability of the host cell to process initiator methionine resi-  (26), Bacillus subtilis (27), E. coli (13), Hemophilus influenzae (28), Helicobacter pylori (29), Mycobacterium leprae (30), Mycobacterium tuberculosis (31), Mycoplasma gallisepticum, (DDBJ/GenBank ™ /EBI entry g2766524), Mycoplasma genitalium (32), S. cerevisiae (11), Homo sapiens (DDBJ/GenBank ™ /EBI entry g57731), Arabidopsis thaliana (DDBJ/GenBank ™ /EBI entry g2583129), and M. pneumoniae (33)) and eight type II MetAPs (Sulfolobus solfataricus (34), H. sapiens (8), Methanobacterium thermoautotrophicum (35), Methanococcus jannaschii (36), P. furiosis (DDBJ entry g2382623), Rattus norvegicus (37), S. cerevisiae (18), and Archaeoglobus fulgidus (38)). Alignments were calculated using the Clustal method with a PAM 250 residue weight table (39). Residues mutated in this study are denoted with an asterisk. dues, resulting in a product that has a mixture of different N termini (7). In addition, cytosolic expression of a secreted protein can result in the presence of an initiator methionine that would not normally be found in the mature product. The mutant forms of MetAP presented here may prove useful as reagents for the in vitro or in vivo processing of recombinant proteins. The mature forms of enzymes with cleavable signal sequences often have N-terminal residues with large side chains. For cytoplasmic expression systems, the signal sequence can be removed during the cloning process, but all protein synthesis must be initiated with methionine (or Nformyl methionine); therefore, if the recombinant protein has a large penultimate residue, the methionine will be retained. The expanded specificity mutants of MetAP I could be used in these cases to remove these artificially retained initiator methionines. It may also be possible to create host cells that have MetAPs with contracted substrate specificity to allow protection of N termini from unwanted post-translational modifications. The initiator methionine could be retained in this system to prevent N-terminal post-translational modifications, and later in vivo treatment with the native MetAP could remove this protective methionine.