Site-directed mutants, at position 166, of RTEM-1 beta-lactamase that form a stable acyl-enzyme intermediate with penicillin.

Class A beta-lactamases are known to hydrolyze substrates through a Ser70-linked acyl-enzyme intermediate, although the detailed mechanism remains unknown. On the basis of the tertiary structure of the active site, the role of Glu166 of class A enzymes was investigated by replacing the residue in RTEM-1 beta-lactamase with Ala, Asp, Gln, or Asn. All the mutants, in contrast to the wild-type, accumulated a covalent complex with benzylpenicillin which corresponds to an acyl-enzyme intermediate. For the Asp mutant, the complex decayed slowly and the hydrolytic activity was slightly retained both in vivo and in vitro. In contrast, the other mutants lost the hydrolytic activity completely and their complexes were stable. These results indicate that the side-chain carboxylate of Glu166 acts as a special catalyst for deacylation. Residues for deacylation have not been identified in other acyl enzymes, such as serine proteases and class C beta-lactamases. Furthermore, the acyl-enzyme intermediates obtained are so stable that they are considered to be ideal materials for crystallographic studies for elucidating the catalytic mechanism in more detail. In addition, the mutants can more easily form inclusion bodies than the wild-type, when they are produced in a large amount, suggesting that the residue also plays an important role in proper folding of the enzyme.

in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ Recipient of a Fellowship of the Japanese Society for the Promotion of Science for Japanese Junior Scientists. To whom correspondence and reprint requests should be addressed. the residues which very likely constitute the catalytic site: LYS'~, G1u'66, and LYS*~~, in addition to Ser7' (the numbering of Ambler (1) is used). These residues are conserved in the primary structures of class A enzymes, and preliminary analyses af mutant enzymes suggested the importance of Lys73 and G1P6 (20). The next question is how these residues are involved in catalysis.
The RTEM-1 8-lactamase encoded by the bla gene contained in the E. coli plasmid, pBR322 (21), is a class A enzyme, whose structure-function relationship has been well studied by means of protein engineering approaches (22)(23)(24)(25)(26)(27). Although high resolution of the tertiary structure of the enzyme has not been reported, its primary structure is quite similar to that of the S. aureus enzyme, therefore, the properties of genetically engineered mutants of the enzyme can be interpreted through the use of the tertiary structure of the Staphylococcal enzyme and the amino acid sequence alignments already established (1,3,28). In this study, G1u'66 of RTEM-1 p-lactamase (according to the numbering of Ambler, 166; the 164th residue in the precursor of the enzyme) was replaced with Ala, Asp, Gln, or Asn by site-directed mutagenesis, and the resultant mutants were characterized to elucidate the residue's role in catalysis; facilitation of the deacylation step of the acyl-enzyme mechanism. In addition, it was also found that these mutant enzymes could more easily form inactive and insoluble inclusion bodies than the wild-type enzyme, when they were produced in a large amount, through the use of an overexpression plasmid.

RESULTS
Construction of an Expression Plasmid, pHA508, for the Wild-type and the Mutants, at Position 166, of RTEM-1 P-Lactarnase-It has already been reported that RTEM-1 Plactamase was overproduced under the control of the lpp-lac fused promoter and the lac operator from the plasmid, pJG108, in which the region coding for mature P-lactamase was inserted into an Escherichia coli secretion vector, . Although the coding region for the signal sequence of the bla gene in pJG108 was replaced with that for E. coli OmpA, the precursor produced from the plasmid was precisely processed and the mature enzyme with the original sequence was secreted into the periplasmic space in E. coli cells (32). The plasmid, pHA508, used to express the wildtype and mutant RTEM-1 p-lactamases in this study was a derivative of pJG108. Construction of the plasmids is shown Portions of this paper (including "Experimental Procedures" and Figs. 1 and 4) are presented in miniprint at the end of this paper. Miniprint is easily read with the aid of a standard magnifying glass. Full size photocopies are included in the microfilm edition of the Journal that is available from Waverly Press. in Fig. 1 in detail (see the Miniprint). In pHA508, the vector portion of pJG108 together with the lacl gene encoding the lac repressor was replaced by pUC-kan-lMa, a kanamycinresistant derivative of pUC118 (29). The mutations involving substitution of Glu'= by Ala, Asp, Gln, and Asn were produced in bacteriophage M13mp18 (33) by site-directed mutagenesis and then introduced into PIN-kan-ompAla to produce mutant pHA508s as well as wild pHA508. Through the construction of the new expression plasmid, the copy number of the plasmid was increased by replacement of the replication origin, the drug resistance was changed from chloramphenicol to kanamycin, and the lucI gene was removed. overproduction of the Wild-type and Mutant RTEM-1 p-Lactamases in E. coli MV1184 by pHA508"To express plactamase from pHA508, E. coli MV1184 (29) carrying the lac19 gene, an overproducing mutant of the lac repressor gene, was always used. Preliminary experiments showed that in LB medium supplemented with 0.1% glucose at 37 "C, MV1184 harboring pHA508 grew much slower and the turbidity of cells at full growth was much lower than in the case of MV1184 harboring PIN-kan-ompAla, a vector plasmid without the bln gene. This toxicity was due to the noninduced synthesis of @-lactamase from the high copy number plasmid.
Overproduction of /3-lactamase by a lac inducer, isopropyl p-D-thiogalactopyranoside (IPTG); from pJG108 has been found to be lethal and to lead to its aggregation in the periplasmic space (32). In the case of the mutant pHA508s, the expression was always more toxic to cells than in the case of the wild-type plasmid and the optical density at 660 nm of a full growth culture was less than 1.0 under the same conditions. In order to reduce the rate of expression and to obtain a sufficient amount of cells to analyze, cells with the plasmids were cultured in M9 medium supplemented with casamino acids and glucose (M9CA medium) at lower temperature, 30 "C. The conditions were a modification of those of Takagi et al. (38) who optimized the rate of protein synthesis for the proper folding of subtilisin E, a bacterial extracellular protease, using PIN-111-ompA as a vector. Under these conditions, the growth of cells with wild or mutant pHA508s was comparable to that of the control strain (the optical density of a full growth culture at 660 nm was more than 5.0). Surprisingly, under such noninduced conditions for the promoter, the wild-type and mutant enzymes were found to be overproduced within cells as a major protein, as judged on SDS-polyacrylamide gel electrophoresis (SDS-PAGE) (Fig. 2). Densitometric analyses showed that in each case, the mature enzyme was produced up to about 10% of the total cellular protein. In this study, for the production of P-lactamase, these conditions were always used. Similar analysis showed that in the same medium at 37°C and in LB medium with 0.1% glucose even at 30 or 37 "C, lower production per cell was obtained (data not shown). ND, not detectable due to the low activity.

Minimum Inhibitory Concentrations of Ampicillin against E. coli Producing the Wild-type or Mutant RTEM-1 0-Lacta-
colonies were grown from a single cell. As shown in Table I, the minimum inhibitory concentration values for dilution clearly showed that the substitutions at residue 166 resulted in the complete loss of hydrolytic activity, except for in the case of the Asp mutant, E166D. The Asp mutant showed slightly increased resistance to ampicillin, in comparison with the control and the other mutants, indicating that it retained a decreased level of activity. This was confirmed by the minimum inhibitory concentration value for direct spotting, which was more than 500 pg/ml. The higher values for 10' than those for lo-' dilution, especially for the E166D mutant, were due to the inactivation of ampicillin by p-lactamase present in the preculture. As for the E166Q mutant of plactamase I from Bacillus cereus, the loss of ampicillin resistance of enzyme-producing E. coli cells has already been reported (20).
Hydrolysis of Benzylpenicillin by Crude Preparations of the Wild-type and Mutant RTEM-1 p-Lactamases-By using the total cellular protein of enzyme-producing E. coli cells prepared as above, the in vitro activity of the mutant enzymes was examined, with benzylpenicillin as a substrate. Benzylpenicillin is also a good substrate for class A enzymes, therefore, it is often used as a standard substrate. As shown in Table I, the results corresponded well to those in the case of in vivo activity described above. The mutants completely lost the activity except for E166D, which showed faint but detectable activity. The specific activity of the crude enzyme of the Asp mutant was 114000 in comparison with the wild-type, although the expression levels in E. coli cells were similar (Fig. 2). These results indicated that the catalytic activity requires a carboxylate side chain at position 166. In contrast to the greatly reduced specific activity, the K,,, value for benzylpenicillin of the Asp mutant was very similar to that of the wild type (Table I), suggesting that Glu'" is not so important for the initial binding of substrates. The K,,, values for the wild-type and the mutant, 0.03 and 0.04 mM, respectively, were similar to that of the purified wild-type enzyme determined under the same conditions (15).

Covalent Complex Formation of the Mutant RTEM-1 p-
Lactamuses with /'4C]Benzylpenicillin-It is possible to determine the affected step(s) (acylation and/or deacylation) during the catalysis by the mutants by monitoring the formation of a covalent complex with benzylpenicillin. For this purpose, the method which has been used to detect penicillin-binding proteins (37) was used. Penicillin-binding proteins are bacterial target enzymes of p-lactam antibiotics which bind covalently to p-lactams through the active site serine residue and thus are inactivated. The enzyme preparations were mixed with ['4C]benzylpenicillin under conditions similar to those used for determination of the hydrolytic activity and separated by SDS-PAGE, followed by fluorography to detect [14C]benzylpenicillin-enzyme complexes.
The results are shown in Fig. 3. No complex with the wild type was detected. In contrast, all of the mutants clearly formed detectable covalent complexes, indicating that G1dS only catalyzes the deacylation step. It is reasonable that the observed complexes were acyl-enzyme intermediates formed through acylation and accumulated due to a lack of deacylation. Penicillin-binding proteins derived from the host strain A chase experiment, in which a high concentration of unlabeled benzylpenicillin was added after labeling with the radioactive compound, showed the stability of the complexes with the Gln, Asn, and Ala mutants (Fig. 3). In particular, the E166N mutant hardly decayed, at least within one h. The complex with the Asp mutant, in contrast, decayed faster than the others, corresponding well with its slight hydrolytic activity. In this case, the rate of deacylation seemed high enough for hydrolytic activity to be detected, but it had decreased to lower than that of acylation.
Inclusion Body Formation of Mutant RTEM-1 @-hctamases-Preliminary experiments showed that the conditions used for the binding of ['4C]benzylpenicillin in Fig. 3 were enough to saturate the mutant enzymes with the ligand (data not shown). The amounts of covalent complexes with the mutants observed, however, differed from each other (Fig. 3, 0 min), although the expression levels were similar (Fig. 2), suggesting that each mutant preparation contained active and inactive forms of the penicillin-binding activity. This was found to be due to different levels of inactive and insoluble inclusion bodies, which were detected only in the mutants. By means of microscopy, a few vesicles were observed inside each enzyme-producing cell and the crude enzyme preparations still contained them after sonication. In the case of the wildtype, vesicle-containing cells were rare and the preparation was clear. Fig. 4A shows the protein profiles of the soluble and insoluble fractions obtained from sonicated cells (see the Miniprint). For all the mutants, more than one-half of the enzyme protein was recovered in the insoluble fraction together with a small amount of membrane protein. Penicillinbinding activity was not detected for the insoluble proteins, being recovered completely in the soluble fraction ( Fig. 4B; see the Miniprint). Hydrolytic activity toward benzylpenicillin of the E166D mutant and the wild-type were only recovered in the soluble fraction (data not shown). In contrast, under the same culture conditions, the wild-type was produced as an active and soluble form, although a small part was recovered in the insoluble fraction (Fig. 4A). The fact that a large part of the mutant proteins was unable to fold and formed inclusion bodies suggests that G1dS also plays an important role in the normal folding of the protein. The 4,000fold reduction in the activity of the Asp mutant (Table I), therefore, was partially due to the formation of inclusion bodies. However, the accumulation of an acyl-enzyme intermediate clearly reflected a reduction in the rate of deacylation and the resulting decrease in the bat value of the mutant.

Catalytic Mechanism of Class A 8-Lmtamases-Kinetic
studies and trapping experiments on covalent substrate-enzyme or inhibitor-enzyme complexes have shown that class A p-lactamases hydrolyze substrates through a Ser70-linked acyl-enzyme intermediate. On the basis of the tertiary structure of the active site cavity around Ser70 of the Staphylococcal @-lactamase, Herzberg and Moult (17) proposed a catalytic mechanism in which the conserved residues, L Y S~~, Glu", and Lys' ", participate, in addition to acylated Ser". It is likely, on the basis of the predicted mode of substrate binding, that the enzyme-substrate complexes are largely stabilized through an electrostatic interaction between the ammonium group of LysZs4 and the C3 carboxylate of penicillins, and that the tetrahedral intermediates and associated transition states occurring during acyl-transfer to and from Ser70 are stabilized by the oxyanion hole comprised of the two main chain NH of RTEM-1 P-Lactamase at Position 166 3189 groups of Ser7' and residue 237 (not a conserved residue) (17). Very recently, Lys234 in the B. licheniformis enzyme was replaced by site-directed mutagenesis and the resultant mutants were investigated, and it was indicated that Lys234 is involved in both ground state binding and transition state binding (39). The existence of an oxyanion hole was confirmed by a kinetic study (40) and by that among 19 possible mutants in which the Gln237 of RTEM-1 p-lactamase was replaced, only the Pro mutant was inactivated completely (25). In contrast to these aspects of the catalytic mechanism, the main catalysts for acylation and deacylation have remained unclear. Herzberg and Moult (17) proposed, on the basis of the crystal structure, that Lys73 catalyzes the acylation step and Glu" the deacylation step. Madgwick and Waley (20) have already constructed site-directed mutants of p-lactamase I from B. cereus, K73R and E166Q, and have shown the reduction in their in vivo activity measured as the ampicillin resistance of E. coli cells. However, detailed analysis has not been reported.
In this study, we constructed site-directed mutants of RTEM-1 p-lactamase, as a model class A enzyme, and showed the accumulation of a stable acyl-enzyme intermediate and the resulting loss of the activity of the E166A, E166Q, and E166N mutants, and the reduced activity of the E166D mutant due to the slow decay of such an accumulated complex. These results clearly indicate that the carboxylate side chain of G1uI6'j of class A p-lactamases is not essential for the initial binding of substrates (as indicated by the similar K , value for benzylpenicillin of E166D to that of the wild-type) or acylation, but for deacylation during catalysis. The stability of the covalent complexes observed in the case of the E166A, E166Q and E166N mutants with benzylpenicillin also indicates that deacylation is unable to proceed through only the machinery involved in acylation. The importance of G1u'66 in the catalysis was supported by a kinetic study which suggested the involvement of a carboxylate residue in the catalysis (12). According to the results of this study and the structural basis for the putative position of a water molecule which attacks the carbonyl carbon of the acyl intermediate (17), deacylation is thought to be catalyzed by the carboxylate of Glu'%, which acts as a general-base catalyst. Considering that Glu'% does not participate in acylation and that the catalytic residue for acylation might be Lys73 on a structural basis, a possible catalytic mechanism is presented in Fig. 5, which is based on that of Herzberg and Moult (17). They proposed that during the proton transfer from Ser70 to the p-lactam nitrogen, protonated Lys73 provides a potential gradient which reduces the energy barrier for the transfer and polarizes the nitrogen favorably to receive the proton. In fact, the K73R mutant of P-lactamase I from B. cereus retained considerable activity (20).
Comparison of the catalytic residues of class A p-lactamases with those of other acyl-enzymes indicated that G l P 6 is a unique residue of class A p-lactamases that is specifically involved in deacylation. The fact that class A 0-lactamases are fully efficient enzymes (41) may be due to the existence of this residue. The well established catalytic mechanism of serine proteases having a catalytic triad and that of a class C P-lactamase proposed by a recent crystallographic study (42) do not involve such participation of a carboxylate residue in deacylation. The lack of such a carboxylate residue may make deacylation the rate-limiting step for class C P-lactamases, even with good substrates (8-10).
The refined tertiary structure of a class C p-lactamase (42) also showed the structural similarity between class A and C enzymes, and suggested another possible catalyst for the acylation reaction to Ser7' in class A enzymes, Ser13', according to the results of superimposing of active-sites. SerI3' of class A p-lactamases is a conserved residue and corresponds to conserved Tyr'"' of class C enzymes whose side-chain hydroxy group was proposed to catalyze both acylation and deacylation (42). We are now preparing mutants at position 130 to test the possibility that Se?"' participates in the acylation step together with LYS~~.

Acylation -
Moreover, as a result of this study, we obtained stable benzylpenicilloil-enzyme complexes as a native form. Even denatured benzylpenicilloil-enzyme could not be trapped by means of a cryoenzymological technique and quick denaturation using the B. cereus enzyme (12). This is the first demonstration of stable complexes of a class A fi-lactamase with a good substrate, not an inhibitor. The stable complexes obtained should be good materials for further structural analyses involving crystallography which will answer the question as to the catalytic residue for acylation (LYs~~ and/or Ser13') during the catalysis by class A /3-lactamases. The method is also applicable to other class A enzymes theoretically.
Inclusion Body Formation by Mutant RTEM-1 &Lackmuses--It has already been reported that the wild-type RTEM-1 @-lactamase was produced from p&2108 and localized in the periplasmic space as inclusion bodies under the control of the lpp-Zuc promoter through IPTG induction (32).

PeripIasmic inclusion
bodies of the enzyme were also produced under the control of the tat promoter with induction (43,44). Under the control of the APL promoter with induction, much higher production was obtained and the precursor of the enzyme was accumulated as inclusion bodies, probably cytoplasmic (44). These observations indicated that an excessively high rate of the protein synthesis inhibits the proper folding of RTEM-1 /3-lactamase in the maturation-secretion pathway, suggesting that inclusion body formation is related to the folding kinetics of the enzyme.
Under the conditions used in this study, the wild-type enzyme was almost normally folded into the active conformation. The fact that mutants at position 166 formed mature inclusion bodies under the same conditions strongly suggested defects of the mutants in the proper folding during secretion into the periplasm. A similar observation was previously reported for another  fl-lactamase mutant with double mutations of C77A and C123A (44). In addition, inclusion bodies of the mutants could be recovered from the insoluble fraction with considerable purity (Fig. 4A). Recovery of the activity by denaturation-renaturation of the inclusion bodies established previously for the RTEM-1 /3-lactamase (44) may be useful for purification of the mutant enzymes. Ampicillin affinity chromatography for purification of penicillin-binding proteins is also applicable to that of the 166 mutants from the soluble fraction.