INTRODUCTION

The most common mechanism of bacterial resistance to -lactam antibiotics is the production of -lactamase, a bacterial enzyme that catalyzes the hydrolysis of -lactams. -Lactamases are grouped into 4 classes (A, B, C, and D) based on primary sequence. TEM-1 -lactamase, a class A serine hydrolase, is the most prevalent -lactamase found in Gram-negative bacteria. TEM-1 -lactamase is capable of hydrolyzing both penicillins and cephalosporins. However, it cannot hydrolyze the recently developed extended-spectrum antibiotics, such as ceftazidime and aztreonam. Extended-spectrum antibiotics were developed in part to combat Gram-negative bacteria that had developed resistance to existing penicillins and first and second generation cephalosporins through the expression of -lactamases such as TEM-1. However, within a few years after the introduction of the extended-spectrum antibiotics, clinical isolates were discovered that were capable of hydrolyzing these new antibiotics (1). These enzymes, termed extended-spectrum -lactamases, contain one to four amino acid substitutions near the active site and are derived either from the TEM or SHV -lactamases (2). These substitutions occur at residues 104, 164, and 237-240, both as individual and as combinatorial mutations (numbering of amino acids according to Ambler et al. (26)). Specifically, the substitutions of a lysine for glutamate at position 104 (E104K), a serine for arginine at position 164 (R164S), a serine for a glycine at position 238 (G238S), and a lysine for a glutamate at position 240 (E240K), either alone or in combination, have been shown to provide increased catalytic activity toward the extended-spectrum cephalosporin ceftazidime (3, 4, 5, 6). Understanding how amino acid substitutions alter the substrate specificity of -lactamase for a given antibiotic may aid in the design of new antibiotics that avoid inactivation by extended spectrum -lactamases. For example, by knowing if amino acid substitutions in extended spectrum -lactamases act through enhanced binding of a specific part of an antibiotic side chain it may be possible to modify that part of the antibiotic to avoid inactivation. In addition, knowledge of how mutations alter the substrate specificity of -lactamase may be of direct empiric value for combination therapy. If the set of substitutions in -lactamase that result in increased activity are different for two antibiotics then the combination of those antibiotics in therapy may avoid the development of resistance.

Previously, random replacement mutagenesis has been used to randomize the nucleotide sequence of three contiguous codons in the bla_TEM-1 gene to create libraries that encode all possible amino acid combinations for the target region (6, 7, 8, 9, 10). In the random library covering residues 238-241, -lactamase mutants were selected and characterized having 100-fold higher activity than the wild-type level toward ceftazidime (6). That study determined that the G238S and E240K substitutions were the primary cause for ceftazidime resistance while substitutions at position 241 provided only a minor improvement.

While residue 241 is on the outer edge of the active site, crystallographic studies of TEM -lactamase have shown residue 237 to be part of the hydrogen-bond network that stabilizes initial binding of the substrate in the active site (11, 12). Since residues 237-240 are the only contiguous residues in which substitutions have been shown to greatly increase the hydrolysis spectrum of TEM -lactamase (4, 5, 6, 8, 13), and residue 237 seems to have a greater importance than residue 241 in hydrolysis, the region 237-240 is more relevant for the study of enzyme-substrate interactions in extended-spectrum -lactamases. In this report, random replacement mutagenesis has been used to create the random library L237-240,¹ which contains all possible amino acid combinations for residues 237-240. This library was used to select mutants with high levels of activity toward the cephalosporin ceftazidime and the monobactam aztreonam to study the importance of this region in determining substrate specificity. Ceftazidime and aztreonam contain an identical aminothiazole-oxime side chain and so amino acid substitutions that affect ceftazidime hydrolysis may exhibit a similar effect on aztreonam hydrolysis. In fact, analysis of kinetic parameters of the E104K, R164S, G238S, and E240K enzymes indicates that the catalytic efficiency of each of these enzymes is increased for both ceftazidime and aztreonam (4, 5). In this study, the majority of mutants had a serine for glycine substitution at position 238 and a lysine or arginine for glutamate substitution at position 240, which reinforces the importance of the G238S and E240K/R substitutions in the hydrolysis of antibiotics with an aminothiazole-oxime side chain (3, 4, 5, 6). All of the mutants selected for ceftazidime hydrolysis had the wild-type alanine conserved at position 237, but 17 out of 18 mutants selected for aztreonam hydrolysis had an A237G substitution. The role of the individual substitutions was determined by introducing the substitutions individually and in combination. Both the A237G single and G238S:E240K double substitution improve catalytic efficiency for both ceftazidime and aztreonam. However, when the A237G substitution is added to the double mutant, catalytic efficiency is decreased 12-fold for ceftazidime but is increased 3-fold for aztreonam. Thus, the effect of the A237G substitution on ceftazidime hydrolysis is strongly dependent on whether it resides in the wild-type or G238S:E240K enzyme.

EXPERIMENTAL PROCEDURES

All enzymes were purchased from New England Biolabs, except T7 DNA polymerase, which was purchased from U. S. Biochemical Corp. Ceftazidime was provided by Glaxo (Greenford, United Kingdom) and aztreonam was provided by Bristol-Meyers Squibb, Inc. Nitrocefin and antibiotic paper discs containing either ceftazidime or aztreonam were purchased from Becton Dickinson (Cockeysville, MD). G-75 Sephadex was obtained from Pharmacia Biotech Inc. (Piscataway, NJ).

E. coli BW313 [Hfr lysA(61-62), dut1, ung1, thi1, relA1 spoT1] was used for the propagation of plasmid DNA prior to mutagenesis (14). E. coli ES1301 [lacZ53, mutS201::Tn5, thyA36, rha5, metB1, deoC,IN (rrnD-rrnE)] was used for the introduction of mutagenized DNA (15). E. coli XL1-Blue [recA1, endA1, gyrA96, thi-1, hsdR17, supE44, relA1, lac, [F::Tn10(Tet^r)proAB, lacI^q (lacZ)M15] was used for the determination of antibiotic susceptibility and the preparation of single stranded DNA (16).

Plasmid pBG66 is the parent plasmid of all random library constructions (9). The pBG66 plasmid is a 4.8-kilobase derivative of pBR322 and pBR325 that contains the wild-type bla_TEM-1 gene and the cat gene, which encodes for chloramphenicol acetyltransferase. The plasmid also contains the ColE1 and f1 origins of DNA replication.

Oligonucleotide primers used for mutagenesis and DNA sequencing were synthesized by Genentech, Inc., at the PAN facility at Stanford University Medical School, and at Genosys Biotechnologies, Inc. Random replacement mutagenesis was done as described previously (17).

Briefly, the strategy behind the modified mutagenesis was to first insert a unique XhoI restriction site into a location within the bla gene, which has been targeted for mutagenesis. The XhoI recognition sequence is flanked by two 12 base arms which are complementary to the sequence adjacent to the site targeted for mutagenesis. The restriction site was positioned at or near the middle of the three codons to be randomized, and the second base of the middle codon was deleted to create a frameshift mutation. A XhoI restriction site was previously inserted at codon 238 (9). The frameshift mutation, resulting from this insertion, renders the bla gene non-functional. Subsequent randomization is achieved by replacing the unique restriction site with a 9-base randomized DNA sequence. An oligonucleotide was designed to replace the 9-base window (including the XhoI site) with random sequence, 5-NNS NNS NNS-3 (where N indicates an equal probability of any base, and S indicates an equal probability of either C or G). This insured all amino acids would be sampled in the window. Two 14-base complementary arms flank the random sequence. The calculations involved in the determination of the percent of randomization have been described previously (9). Library DNA was electroporated into E. coli XL1-B cells for further screening.

Single-stranded plasmid DNA was prepared for sequencing as described (18). DNA sequencing was performed using the dideoxy chain termination method (19). Oligonucleotides were designed to prime synthesis from specific sites within the bla_TEM-1 gene.

Selections were done using the disc diffusion method. The TEM-1 -lactamase or the L237-240 was introduced into E. coli XL1-B and was grown in 2 × YT media plus 12.5 µg/ml chloramphenicol to stationary phase with aeration at 37 °C. A 1:20 dilution of this culture was made with Mueller-Hinton media, and a sterile cotton swab was used to plate the dilution on Mueller-Hinton agar plates. Antibiotic paper discs containing either 30 µg of ceftazidime or aztreonam were applied to the surface of the plates which were then incubated overnight at 37 °C. The zone of inhibition for wild-type -lactamase was used as a control. Any colonies from L237-240 that grew near or within the wild-type zone of inhibition were considered to have greater than or equal to TEM-1 activity.

Minimum inhibitory concentrations (MICs) were determined by broth microdilution. 1 × 10⁴ E. coli XL1-B cells containing the selected mutant -lactamase were inoculated into microtiter wells containing 100 µl of LB media having 2-fold dilutions of the antibiotic being tested. The ranges of antibiotic concentrations tested for ceftazidime and aztreonam were 0.03-64 µg/ml. The microtiter plates were incubated at 37 °C for 18-24 h. The plates were examined visually, and the lowest concentration of antibiotic that inhibited visual growth was recorded as the MIC. A 2-fold difference in MIC values was determined to be insignificant.

The E240K and G238S:E240K mutants were previously constructed by cassette mutagenesis (6). The A237G mutant was originally isolated from the L235-237 random library in a selection for ampicillin resistance (9).

-Lactamase Purification

TEM-1 -lactamase and the mutant -lactamases were purified to >90% homogeneity. The wild-type TEM-1, E240K, and G238S:E240K enzymes were previously cloned into an expression vector under the control of the tac promoter (6). E. coli XL1-Blue cells containing the expression vector were grown in LB media (18) containing 25 µg/ml kanamycin to early log phase. At an OD₆₀₀ of ~0.5, 0.1 mM isopropyl-1-thio-D-galactopyranoside was added and the culture was incubated an additional 3 h until it reached late log phase. In contrast, the A237G and A237G:G238S:E240K enzymes were purified from E. coli XL1-B cells containing the pBG66 plasmid and were expressed from the natural -lactamase promoter. These strains were grown in culture for 9 h to reach late log phase. From this point, all enzymes were purified by the method that follows. -Lactamase and other periplasmic proteins were first isolated by an osmotic shock procedure (20). The solution obtained by osmotic shock was adjusted to a final concentrate of 100 mM NaAc, pH 7.5, 800 mM NaCl (buffer A). The protein solution was concentrated to a 5-ml volume with an Amicon Centriprep-10 concentrator. The concentrated protein solution was then applied to a 1-ml HiTrap zinc chelating column (Pharmacia). -Lactamase bound strongly to the column while other periplasmic proteins eluted out with buffer A. -Lactamase was eluted using a linear gradient of buffer B (100 mM NaAc, pH 4.0, 800 mM NaCl). Fractions containing -lactamase activity were identified by nitrocefin hydrolysis and SDS-polyacrylamide gel electrophoresis. The -lactamase fractions were concentrated with a Centriprep-10 concentrator. The -lactamase was further purified by Sephadex G-75 gel filtration chromatography using a 25 mM sodium phosphate buffer, pH 7.0. The purity of the enzymes was verified by SDS-polyacrylamide gel electrophoresis. All preparations were >90% pure. The purified enzymes were stored at 4 °C until the determination of kinetic parameters.

The kinetics of TEM-1 -lactamase and the selected mutants were determined with ampicillin, cephaloridine, ceftazidime, and aztreonam. The extinction coefficients used were: ampicillin, 235 nm,  = 900 M¹ cm¹ (6); cephaloridine, 260 nm,  = 10,200 M¹ cm¹ (21); ceftazidime, 260 nm,  = 8,660 M¹ cm¹ (22); aztreonam, 318 nm,  = 660 M¹ cm¹ (22). The hydrolysis of the substrates was monitored spectrophotometrically at 30 °C in 0.05 M phosphate buffer, pH 7.0, on a Beckman model DU 640 spectrophotometer using a 0.1- or 1.0-cm pathlength cuvette as needed. Kinetic parameters V_max and K_m were determined by: 1) initial velocity kinetic analysis by fitting to the equation of a rectangular hyperbola using unweighted nonlinear least squares (23), or 2) using complete progress curves with starting concentrations at least 3 × K_m. The progress curves were evaluated by determining the instantaneous rate at a minimum of 50 points in the reaction by taking the derivative of the substrate concentration at those points in the reaction course. This data was fitted to the Michaelis-Menton equation using unweighted nonlinear least squares (23). To insure that substrate inhibition was not occurring, initial rates were determined at 6 to 8 substrate concentrations chosen to bracket the K_m value determined by the progress curve. In the case of aztreonam, the K_m was measured as the K_i with cephaloridine as the reporter substrate. The enzyme and inhibitor were preincubated for 5 min at 30 °C before the substrate was added. At least five concentrations of reporter substrate bracketing the K_m were tested with at least three concentrations of inhibitor. Enzyme concentration was at least 10-fold less than inhibitor concentrations and at least 100-fold less than reporter substrate concentrations. K_m/V_max values were then plotted versus inhibitor concentration to obtain K_i. These experiments were done at least in triplicate and analyzed using the EnzymeKinetics v1.42 program from Trinity Software. The k_cat was obtained directly from the linear decrease in A₃₁₈ with 100 µM aztreonam and 69 nM enzyme. For the enzymes where k_cat and K_m could not be determined independently the k_cat/K_m values were estimated using the equation  = (k_cat/K_m)[E][S], where [S]  K_m (24). The values reported for the wild-type enzyme for ceftazidime are based on velocity measurements at 10, 15, and 20 µM substrate, and for the E240K enzyme on velocity measurements at 25 and 50 µM substrate. Both were done using the 1.0-cm pathlength cuvette. The values reported for the A237G and A237G:G238S:E240K enzymes are based on velocity measurements at 100, 250, and 500 µM substrate using the 0.1-cm pathlength cuvette. The value reported for wild-type enzyme for aztreonam is based on velocity measurements at 250 µM substrate using both the 0.1- and 1.0-cm pathlength cuvette.

The 1.8-Å crystallographic structure of E. coli TEM-1 -lactamase (Protein Data Bank, Brookhaven National Laboratory, entry 1BTL) was used for the model of the -lactam binding site (12). Crystallographic data from Sowek et al. (5) was used to construct the structure of aztreonam. The 2.5-Å crystal structure of aztreonam bound to the class C -lactamase from Citrobacter freundii (25) describes the acyl-enzyme complex as having a 2.7-Å hydrogen bond between the lactam carbonyl oxygen and the main chain amide group of Ser-64 (Ser-70 in the TEM-1 enzyme) and a 2.9-Å hydrogen bond between the lactam carbonyl oxygen and the main chain amide group of Ser-318 (Ala-237 in TEM-1 enzyme). The 1.7-Å crystal structure of benzylpenicllin with a E166N TEM-1 mutant (11) describes the acyl-enzyme complex as having a 2.7-Å hydrogen bond between the lactam carbonyl oxygen and the Ser-70 main chain amide group and a 3.0-Å hydrogen bond between the lactam carbonyl oxygen and the Ala-237 main chain amide group. The aztreonam molecule was manually docked into the active site of the TEM-1 enzyme using Quanta, version 3.3, from Molecular Simulations, Inc. (Waltham, MA). The substrate was fitted to position the -lactam carbonyl oxygen in the oxyanion hole within ±0.5 Å of the distances mentioned above. The model was adjusted manually to relieve obvious steric problems with the enzyme. No attempt was made to alter the conformation of the aztreonam molecule or to optimize the fit by energy minimization.

RESULTS

Naturally occurring amino acid substitutions at positions 237, 238, and 240 have been described in several extended-spectrum -lactamases. For example, TEM-3, TEM-4, TEM-8, TEM-14, TEM-15, and TEM-19 contain the G238S substitution, TEM-10 and TEM-24 contain the E240K substitution, and TEM-5 contains the A237T:E240K double substitution (2). Note that residues 238 and 240 are adjacent to each other in the TEM sequences. The absence of residue 239 is due to the ABL numbering system for Class A -lactamases (26). In previous experiments with the random library 238-241, -lactamase mutants were selected that had 100-fold greater resistance than wild-type toward ceftazidime (6). It was found that the multiple substitution of G238S:E240K was responsible for the large increase in catalytic efficiency for ceftazidime.

In order to better study the role of substitutions at positions important for extended-spectrum -lactamase activity, the random library 237-240 was constructed. Selection experiments were done using the extended-spectrum cephalosporin ceftazidime and the monobactam aztreonam (Fig. 1). The similarity of these substrates should provide further information about the ability of substitutions in this region to not only increase catalytic efficiency but also to differentiate between similar substrates. The L237-240 library was transformed into E. coli XL1-B cells to isolate mutants with high level activity for ceftazidime or aztreonam. The transformants were selected using a disc diffusion method with discs containing either 30 µg of ceftazidime or 30 µg of aztreonam. E. coli cells containing wild-type TEM-1 -lactamase or the L237-240 library were spread on Mueller-Hinton agar plates and the antibiotic discs were applied to the surface of the plates. The antibiotic diffuses outward from the disc forming a circular concentration gradient on the agar plate. After an overnight incubation at 37 °C, a zone of inhibited bacterial growth will be visible around each antibiotic disc. The radius of the zone of inhibition of wild-type enzyme for each antibiotic is a relative indicator of the wild-type level of activity and was used as a control for these experiments. Only mutant transformants with greatly enhanced activity toward the antibiotic being tested could survive at a much closer radius to the antibiotic disc than the cells containing the wild-type enzyme. The transformants growing within the wild-type zone of inhibition radius on the random library plates were selected for further study. The disc diffusion selection was performed twice for each antibiotic.

Structures of -lactam antibiotics used in this study.

The DNA sequences of 20 mutants selected for ceftazidime hydrolytic activity (10 from each selection) and 18 mutants selected for aztreonam hydrolytic activity (10 from the first selection and 8 from the second) were determined to identify the substitutions that resulted in increased -lactamase activity toward these antibiotics. For the ceftazidime selection (Fig. 2A), alanine was conserved at position 237 for all the mutants, suggesting that alanine is the optimal amino acid at this position for ceftazidime hydrolysis. At position 238, 15 of 20 mutants had a serine while 5 of 20 had an asparagine residue. At position 240, all 20 mutants had either a lysine or arginine. The predominance of serine at 238 and lysine at 240 is in agreement with previous studies showing the importance of these substitutions for ceftazidime hydrolysis (3, 4, 6). For the aztreonam selection (Fig. 2B), glycine was observed at position 237 in 17 of the 18 mutants sequenced, suggesting that glycine is the optimal amino acid for aztreonam hydrolysis. At position 238, 12 of 18 mutants had G238S, 5 of 18 had G238T, and 1 of 18 had G238N. At position 240, lysine or arginine was observed in 15 of 18 mutants selected for aztreonam hydrolysis while the remaining mutants contained either glutamine or threonine. The predominance of serine at 238 and of lysine or arginine at 240 in the aztreonam selection matches the profile of the ceftazidime selection. This result suggests the substitutions at position 238 and 240 act via the aminothiazole-oxime side group that is common to both antibiotics. In contrast, the sequence requirements at position 237 for maximal aztreonam hydrolysis appear to be different from those required for maximal ceftazidime hydrolysis.

To better understand the extent to which the mutations conferred increased hydrolytic activity, the MICs were determined for aztreonam and ceftazidime for all the selected mutants (Fig. 2). The MICs for ceftazidime were 60- to 100-fold greater than wild-type for mutants with the G238S and E240K/R substitutions. However, the G238S:E240K/R mutants that contained the additional substitution of A237G had MICs only 30-fold greater than wild-type. Thus, the addition of the A237G substitution to the G238S:E240K double mutant decreased the activity of the enzyme toward ceftazidime.

The MIC's for aztreonam were 250-fold higher than wild-type for the G238S:E240K double mutant. However, in contrast to the ceftazidime MIC's, the addition of the A237G substitution to the double mutant resulted in a further increase in the aztreonam MIC. These results suggest that the A237G substitution allows the enzyme to discriminate between aztreonam and ceftazidime.

In order to further understand the effects of the substitutions described above on catalytic activity, the A237G, E240K, G238S:E240K, and A237G:G238S:E240K -lactamases were purified for kinetic analysis (``Experimental Procedures''). The A237G, E240K, and G238S:E240K mutants were constructed previously (6, 9). The kinetic parameters for ampicillin, cephaloridine, ceftazidime, and aztreonam hydrolysis were determined for the wild-type TEM-1, A237G, E240K, G238S:E240K, and A237G:G238S:E240K enzymes. The structures of these antibiotics are shown in Fig. 1.

The kinetics of ampicillin and cephaloridine hydrolysis (Tables I and II) were determined to compare the effect of the substitutions on the hydrolysis of a penicillin and a non-extended-spectrum cephalosporin with that of ceftazidime and aztreonam. The effect of the A237G substitution on the hydrolysis of ampicillin and cephaloridine followed a similar pattern in that for both substrates the K_m value was increased relative to the wild-type enzyme. The double mutant G238S:E240K exhibits a large decrease in both k_cat and K_m values. This has been reported previously and it was shown that the majority of this effect is due to G238S substitution (6). The K_m values determined for the G238S:E240K enzyme in this study are somewhat lower than has been described previously (6). However, all data reported here was done with the same batches of enzyme and antibiotic substrate to facilitate a direct comparison of values. The addition of the A237G substitution to the G238S:E240K enzyme increases the K_m value of the triple mutant for both ampicillin and cephaloridine relative to the G238S:E240K double mutant. Therefore, both alone and in combination with the G238S:E240K substitutions, the A237G substitution increases the K_m for ampicillin and cephaloridine hydrolysis.

Table I. Kinetic parameters of wild-type (wt) and mutant -lactamases for ampicillin Units are s1 for kcat, µM for Km, and M1 s1 for kcat/Km.  -Lactamases kcat Km kcat/Km Relativea wt 237-AGE-240 1625  ± 74 63  ± 9 2.58  × 107 1.00 A237G 1530  ± 94 100  ± 13 1.53  × 107 0.59 E240K 851  ± 96 72  ± 19 1.18  × 107 0.46 G238S:E240K 8.3  ± 0.5 3.7  ± 1.1 2.24  × 106 0.09 A237G:G238S:E240K 24  ± 1 28  ± 2 8.57  × 105 0.03 a  Ratio of kcat/Km of the mutant relative to the wild-type.

Table II. Kinetic parameters of wild-type (wt) and mutant -lactamases for cephaloridine Units are s1 for kcat, µM for Km, and M1 s1 for kcat/Km.  -Lactamases kcat Km kcat/Km Relativea wt 237-AGE-240 1070  ± 124 761  ± 44 1.41  × 106 1.00 A237G 849  ± 38 1635  ± 94 5.19  × 105 0.37 E240K 281  ± 22 669  ± 95 4.20  × 105 0.30 G238S:E240K 9.7  ± 0.4 16.1  ± 2.0 6.02  × 105 0.43 A237G:G238S:E240K 10.3  ± 0.5 28.6  ± 3.0 3.60  × 105 0.26 a  Ratio of kcat/Km of the mutant relative to the wild-type.

As seen in Table III, all of the mutant enzymes had increased catalytic efficiency for ceftazidime. With ceftazidime as substrate, a high K_m value (>1000 µM) prevented the individual values of k_cat and K_m from being determined for all but the G238S:E240K enzyme. The k_cat/K_m values were estimated by examining the reaction rate at [S]  K_m (24). As described previously (6), the effects of the G238S and E240K substitutions on ceftazidime hydrolysis were that G238S causes a significant decrease in K_m, allowing the k_cat and K_m of G238S:E240K to be measured, and E240K causes an increase in k_cat and a small decrease in K_m (6). Accordingly, the double mutant G238S:E240K provides a 1294-fold increased catalytic efficiency (k_cat/K_m) over the wild-type enzyme. Interestingly, the A237G substitution alone provides a 13-fold increase in catalytic efficiency for ceftazidime, but the triple mutant A237G:G238S:E240K provides only a 109-fold increase, which is 12-fold less than the double mutant. The results suggest that the A237G substitution, while beneficial on its own, detracts from the contributions of the G238S:E240K double mutant. Because the K_m of the G238S:E240K double mutant is 178 µM and the K_m of the A237G:G238S:E240K triple mutant is too high to measure (>1000 µM), it can be concluded that the addition of the A237G substitution to the double mutant results in at least a 5-fold increase in K_m. Note that an increase in K_m is also observed with ampicillin and cephaloridine as substrates (Tables I and II). However, because of the inability to measure K_m in the A237G single mutant, it is unclear whether the increase in catalytic efficiency of this enzyme is due to decreased K_m or increased k_cat or both.

Table III. Kinetic parameters of wild-type (wt) and mutant -lactamases for ceftazidime Units are s1 for kcat, µM for Km, and M1 s1 for kcat/Km.  -Lactamases kcat Km kcat/Km Relativea wt 237-AGE-240 NMb NM 5.51  × 101 1 A237G NM NM 6.94  × 102 13 E240K NM NM 5.64  × 102 10 G238S:E240K 12.7 ± 0.4  178 ± 10 7.13  × 104 1294 A237G:G238S:E240K NM NM 6.02  × 103 109 a  Ratio of kcat/Km of the mutant relative to the wild-type. b  NM, not measurable.

With aztreonam as the substrate (Table IV), each substitution lowered the K_m value such that k_cat and K_m could be determined for each enzyme except wild-type. The effect of the A237G and E240K substitutions was to lower the K_m to a level that we could measure. An even larger reduction, relative to that of A237G and E240K, was observed for the K_m values of G238S:E240K and A237G:G238S:E240K, with reductions of an additional 200-600-fold. The addition of the G238S substitution to the E240K enzyme had the effect of further reducing both K_m and k_cat. In contrast to the kinetic patterns observed with ceftazidime, the triple mutant A237G:G238S:E240K had the highest catalytic efficiency for aztreonam, with a 3437-fold increase over the wild-type enzyme. The reason for this is that the addition of the A237G substitution to the G238S:E240K enzyme has the effect of lowering K_m and thus providing a 3-fold increase in catalytic efficiency over the G238S:E240K double mutant. It is interesting that the A237G substitution resulted in a lower K_m value when aztreonam was the substrate versus a higher K_m value for ampicillin, cephaloridine, and ceftazidime.

Table IV. Kinetic parameters of wild-type (wt) and mutant -lactamases for aztreonam Units are s1 for kcat, µM for Km, and M1 s1 for kcat/Km.  -Lactamases kcat Km kcat/Km Relativea wt 237-AGE-240 NMb NM 2.22  × 102 1 A237G 0.44  ± 0.07 103  ± 43 4.13  × 103 19 E240K 0.76  ± 0.04 229  ± 23 3.32  × 103 15 G238S:E240K 0.135c 0.452d 2.99  × 105 1347 A237G:G238S:E240K 0.330c 0.384d 7.63  × 105 3437 a  Ratio of kcat/Km of the mutant relative to the wild-type. b  NM, not measurable. c  kcat was obtained directly from the linear decrease of A318 with 100 µM aztreonam. d  Km was measured as Ki with cephaloridine as the reporter substrate.

A comparison of the substrate specificity, as measured by the catalytic efficiency, of wild-type TEM-1 -lactamase and each of the mutant enzymes is shown in Fig. 3. It is apparent from the data that the wild-type enzyme has very high catalytic efficiency for ampicillin and cephaloridine and poor catalytic efficiency for ceftazidime and aztreonam. The double mutant G238S:E240K, although, has moderately high catalytic efficiency for all substrates. With the addition of A237G, however, the triple mutant exhibits a further increase in catalytic efficiency for aztreonam hydrolysis while showing reduced efficiency for ampicillin, cephaloridine, and ceftazidime hydrolysis. Thus, the double substitution of G238S:E240K provides the wild-type enzyme with the ability to hydrolyze antibiotics with an aminothiazole-oxime side group, and the further addition of A237G allows the enzyme to differentiate between a monobactam and a cephalosporin with the aminothiazole-oxime side group.

Comparison of the substrate specificity of wild-type TEM-1 -lactamase and the mutant -lactamases.

DISCUSSION

The goal of this study was to identify substitutions in the 237-240 region of TEM-1 -lactamase that alter specificity and enhance catalytic efficiency toward different antibiotic substrates, such as the cephalosporin ceftazidime and the monobactam aztreonam. In order to accomplish this goal, random replacement mutagenesis was used to completely randomize residues 237-240 to form a random library that contains all the possible amino acid combinations for that region. Then, mutants were selected that had greater than the wild-type level of activity toward the given antibiotic. It was found that the preferred sequence for ceftazidime hydrolysis was a conserved alanine at position 237, a serine or asparagine at position 238, and a lysine or arginine at position 240. The preferred sequence for aztreonam hydrolysis involved more variability and was found to have a glycine at position 237, a serine, threonine, or asparagine at position 238, and a lysine, arginine, or threonine at position 240. Previous studies with TEM-1 and SHV-type -lactamases have shown the importance of the G238S:E240K double mutant for ceftazidime hydrolysis (3, 6). For TEM-1 -lactamase, the G238S:E240K double substitution has been shown to increase k_cat while decreasing K_m (6). Characterization of these enzymes for aztreonam hydrolysis indicate that the increases in k_cat/K_m are mostly due to very large reductions in K_m.

To better understand the role of the A237G substitution, we determined the contributions of the A237G, E240K, G238S:E240K, and A237G:G238S:E240K substitutions in ampicillin, cephaloridine, ceftazidime, and aztreonam hydrolysis. Characterization of these enzymes for ampicillin and cephaloridine hydrolysis indicate that the A237G substitution results in an increased K_m value which causes a slight decrease in the overall catalytic efficiency for these substrates (Tables I and II). However, characterization of these enzymes for ceftazidime hydrolysis indicate that, although A237G alone provides an increase in catalytic efficiency, when it is added to the G238S:E240K double mutant a reduction in catalytic efficiency occurs and therefore the A237G substitution detracts from the contributions of the other substitutions.

In contrast to the hydrolysis of ceftazidime, the A237G substitution is beneficial for aztreonam hydrolysis both individually and in combination with the G238S:E240K double mutant. The k_cat/K_m of the A237G mutant is increased 19-fold more than the wild-type enzyme (Table IV). In addition, the k_cat/K_m of the G238S:E240K double mutant for aztreonam hydrolysis is increased 3-fold when the A237G substitution is added (Table IV). Whereas optimal ceftazidime hydrolysis is achieved by the double mutant G238S:E240K, optimal aztreonam hydrolysis is achieved by the triple mutant A237G:G238S:E240K.

The contributions of the substitutions to the overall catalytic efficiency can be calculated for each substrate from the data in Tables I through IV. This was done using the following equation (24):

(Eq. 1)

which calculates the difference in binding energy between the wild-type and mutant enzymes in going from free enzyme and substrate to the transition state. With ceftazidime as the substrate, the A237G:G238S:E240K mutant decreases the free energy barrier by 2.8 kcal/mol. The A237G and G238S:E240K substitutions decrease the free energy barrier by 1.5 and 4.3 kcal/mol, respectively. The free energy changes of these two mutants can be related to the triple mutant A237G:G238S:E240K by the following equation (27, 28):

(Eq. 2)

The G₁ term represents a change in the interaction energy provided by the individual mutants to the triple mutant. If the contributions of the substitutions display simple additivity, the value of G₁ will be near zero. The value of G₁ for the triple mutant A237G:G238S:E240K is +3.0 kcal/mol. This value indicates complex additivity and that the contribution of the A237G substitution to transition state stabilization is lost when the substitution is added to the G238S:E240K double mutant. In fact, the A237G substitution antagonizes the transition state stabilization provided by the G238S:E240K substitutions.

Applying the same analysis to data from Table IV with aztreonam as substrate, G = 4.9 kcal/mol for the triple mutant A237G:G238S:E240K, while the values for A237G and G238S:E240K were 1.8 and 4.3 kcal/mol, respectively. The value of G₁ was +1.2 kcal/mol for the triple mutant. Although this value is non-zero, interaction energies from 1.0 to 1.5 kcal/mol are generally considered additive because of errors involved in calculating the free energy effect in the multiple mutant from the component mutants (28, 29). Thus, the addition of A237G to G238S:E240K exhibits simple additivity. The G for the A237G:G238S:E240K enzyme with ampicillin as substrate was +2.1 kcal/mol with a G₁ term of +0.3 kcal/mol, while the A237G:G238S:E240K G for cephaloridine was +0.8 kcal/mol with a G₁ term of 0.3 kcal/mol. Thus, for aztreonam the contribution of the A237G substitution to transition state stabilization is retained when the substitution is added to the G238S:E240K double mutant. Also, the destabilizing effect of the substitutions on ampicillin and cephaloridine hydrolysis displays simple additivity, in contrast to the non-additive destabilizing effect seen with ceftazidime.

A review of site-directed mutagenesis studies of tyrosyl-tRNA synthetase and substilisin BPN has shown that complex additivity, or non-additivity, in transition state stabilization is observed in two instances (28). First, when the mutation sites are very close together a change in the interaction energy between the sites often occurs. This is thought to be due to either a disruption in direct contacts by the residues or an indirect change in electrostatic interactions or structural perturbations of the residues. Second, when a mutation site is involved in a cooperative interaction in the enzyme mechanism a change in the rate-limiting step may occur. Thus, the complex additivity shown by the TEM-1 -lactamase A237G and G238S:E240K mutants for ceftazidime hydrolysis is consistent with the above mentioned observations on the additivity relationships of adjacent residues (28). However, these same mutants display simple additivity for ampicillin, cephaloridine, and aztreonam hydrolysis. There are several possible explanations for the non-additivity observed with ceftazidime but not the other antibiotics. For example, the A237G substitution may introduce a conformational change in the G238S:E240K enzyme structure or the solvent structure around the active-site that antagonizes the contributions of the individual substitutions. However, the change in enzyme or solvent structure would not affect the hydrolysis of the other antibiotics. An alternative explanation is that ceftazidime binding may introduce a conformational change in the G238S:E240K enzyme that does not occur in the A237G:G238S:E240K triple mutant. This model more readily explains why non-additivity is observed only with ceftazidime. Support for this model comes from x-ray crystallographic studies of a DD-peptidase with cephalothin or cefotaxime in the active site (30). It was found that a conformational change in an active-site threonine residue (T301) occurs in the cefotaxime-enzyme complex but not the cephalothin complex (30). Finally, the ceftazidime substrate may undergo a conformational change upon binding the G238S:E240K enzyme that is impeded by the addition of the A237G substitution.

In the case of aztreonam hydrolysis, the A237G enzyme exhibits a large reduction in the K_m value and when the A237G substitution is added to the G238S:E240K enzyme, the K_m of this enzyme is further reduced. In contrast, the A237G substitution results in an increased K_m for ampicillin and cephaloridine hydrolysis both alone and in combination with the G238S:E240K substitutions. These opposite effects may be due to the presence of the single lactam ring in the monobactam compared to the fused ring system in the other antibiotics. The structure of an acyl-enzyme intermediate of TEM-1 -lactamase with penicillin G (PenG) has been solved (11). This structure shows that the PenG carboxylate group at position C-3 of the thiazolidine ring forms strong hydrogen bonds to both the Arg-244 and Ser-235 side chains. An important difference between monobactams such as aztreonam and the penicillins and cephalosporins is that the monobactams have a sulfonic acid group bonded directly to the nitrogen of the lactam ring in place of the functionally equivalent carboxylate group (Fig. 1). The sulfonic acid is therefore in a different position relative to the carbonyl-carbon of the lactam ring than is the carboxylate. This may result in weaker interactions between the sulfonic acid oxygens and the side chains of Ser-235 and Arg-244. A model of aztreonam in the active site of the wild-type TEM-1 -lactamase is shown in Fig. 4. The A237G substitution may act by altering the conformational flexibility of the S3 -strand on which both Ser-235 and Ala-237 are found. This could reposition the side chain of Ser-235 so that it could form a stronger hydrogen bond with the sulfonic acid group. Alternatively, the removal of the methyl group at position 237 in the A237G mutant may allow the side chain of Arg-244 to rotate to a position closer to the sulfonic acid group and form a stronger interaction. Either of these possibilities might improve the binding of the substrate and account for the reduced K_m value. A change in enzyme mechanism also must be considered. Acylation has been determined as the rate-limiting step of TEM-1 -lactamase for the extended-spectrum cephalosporin cefotaxime, and the G238S substitution was shown to improve the acylation rate, although not enough to change the rate-limiting step (31). The large reduction in K_m for aztreonam could be explained by a possible change in rate-limiting step from acylation to deacylation. However, a detailed interpretation of the effect of the A237G substitution awaits further structural and kinetic data on the enzyme.

Stereoview of the model of the binding site of TEM-1 -lactamase with aztreonam.