Mutational Analysis of VIM-2 Reveals an Essential Determinant for Metallo-β-Lactamase Stability and Folding

Abstract

Metallo-β-lactamase (MBL)-producing bacteria are emerging worldwide and represent a formidable threat to the efficacy of relevant β-lactams, including carbapenems, expanded-spectrum cephalosporins, and β-lactamase inactivator/β-lactam combinations. VIM-2 is currently the most widespread MBL and represents a primary target for MBL inhibitor research, the clinical need for which is expected to further increase in the future. Using a saturation mutagenesis approach, we probed the importance of four residues (Phe-61, Ala-64, Tyr-67, and Trp-87) located close to the VIM-2 active site and putatively relevant to the enzyme activity based on structural knowledge of the enzyme and on structure-activity relationships of the subclass B1 MBLs. The ampicillin MIC values shown by the various mutants were affected very differently depending on the randomized amino acid position. Position 64 appeared to be rather tolerant to substitution, and kinetic studies showed that the A64W mutation did not significantly affect substrate hydrolysis or binding, representing an important difference from IMP-type enzymes. Phe-61 and Tyr-67 could be replaced with several amino acids without the ampicillin MIC being significantly affected, but in contrast, Trp-87 was found to be critical for ampicillin resistance. Further kinetic and biochemical analyses of W87A and W87F variants showed that this residue is apparently important for the structure and proper folding of the enzyme but, surprisingly, not for its catalytic activity. These data support the critical role of residue 87 in the stability and folding of VIM-2 and might have strong implications for MBL inhibitor design, as this residue would represent an ideal target for interaction with small molecules.

Metallo-β-lactamases (MBLs) are bacterial zinc enzymes that are able to hydrolyze most β-lactam antibiotics, including the newer compounds such as oxyimino-cephalosporins and carbapenems, and are not inhibited by the available therapeutic β-lactamase inactivators (e.g., clavulanate and penicillanic acid sulfones) (19, 20). Although MBLs were originally found in bacterial species of limited clinical occurrence, they are now emerging as acquired resistance determinants in major gram-negative pathogens such as Enterobacteriaceae, Pseudomonas aeruginosa, and Acinetobacter spp., in which they can be responsible for extended-spectrum β-lactam resistance phenotypes (19, 20).

VIM-2, an MBL encoded by mobile genetic elements, is presently the most widespread acquired MBL determinant and has been reported in clinical isolates from Europe, Asia, South America, and the United States (20). It thus represents a very important target for MBL inhibitor research. In this context, basic information on enzyme structure, function, and mechanism is of primary interest to help in the design of efficient MBL inhibitors. In this work, we investigated the importance and role in enzyme structure and function of four potentially relevant amino acid residues identified on the basis of the recently obtained VIM-2 crystal structure and its comparison with other known MBL structures (8, 20).

MATERIALS AND METHODS

Bacterial strains, culture media, and growth conditions.

Escherichia coli XL1-Blue (Stratagene Inc., La Jolla, CA) was routinely used as the host in molecular cloning methodologies and for obtaining mutant libraries. E. coli BL21(DE3) (Novagen Inc., Madison, WI) was used for enzyme production. Bacteria were grown in Luria-Bertani (LB) medium, or in ZYP-5052 (22) supplemented with the appropriate antibiotic. Cation-adjusted Mueller-Hinton (MH) medium (agar and broth) was used to determine MICs, as recommended by the CLSI (2).

Construction and characterization of VIM-2 mutant libraries.

Saturation mutagenesis was used to create four independent libraries of VIM-2 mutants, in which codons corresponding to positions 61, 64, 67, and 87 (the BBL numbering is used throughout [7]) were randomized using the PCR-based method as previously described (15). First, an E. coli vector system to be used for the screening of mutant libraries was obtained as follows. The bla_VIM-2 gene was amplified by PCR with primers VIM-2-EXP/fwd and VIM2-EXP/rev as previously described (Table 1) (4). The PCR product was digested with EcoRI and XhoI restriction endonucleases and cloned into vector pLB-II (a derivative of vector pBC-SK [Stratagene Inc., La Jolla, CA] with an added ribosome binding site upstream the cloning site), which was previously digested with the same enzymes, to yield recombinant plasmid pLBII-VIM-2, in which expression of the MBL gene was under the transcriptional control of the P_lac promoter. Random libraries were obtained by PCR using the Expand high-fidelity PCR system (Roche Biochemicals, Mannheim, Germany). Amplification reaction mixtures contained 200 μM deoxynucleoside triphosphates (dNTPs), 125 ng of each primer (Table 1), 7 U of DNA polymerase, and 20 ng of pLBII-VIM-2 plasmid. The cycling parameters were as follows: initial denaturation at 97°C for 5 min; denaturation at 97°C for 3 min, annealing at 48°C for 1 min, and extension at 68°C for 12 min, repeated for 25 cycles; and final extension at 68°C for 20 min. The reaction products were treated with the DpnI restriction enzyme for 3 h at 37°C to specifically remove the template DNA, and the resulting mixture was transformed in E. coli XL1-Blue. For each library, approximately 100 single chloramphenicol-resistant clones were randomly selected for subsequent analysis. The sequences of mutant VIM-2 open reading frames (ORFs) were determined on plasmid DNA using primers T7 and T3. The amount of actual mutants (i.e., showing a codon different from that of the wild-type [WT] gene) in the various libraries ranged 75 to 90%. In order to probe the effect of amino acid randomization, the MICs of ampicillin were determined for each clone of the four independent libraries using the agar dilution method, as recommended by the CLSI (2). E. coli XL-1(pBC-SK) and XL-1(pLBII-VIM-2) were used as negative and positive controls, respectively. Antimicrobial susceptibility profiles were determined using the broth microdilution method, as recommended by CLSI (2).

TABLE 1.

Oligonucleotides used in this study^a

Primer	Sequence (5′ → 3′)
VIM-2-EXP/fwd	GGAATTCCATATGTTCAAACTTTTGAGT
VIM-2-EXP/rev	GGCTCGAGGATCCTGCTACTCAACGACTG
VIM-random 61(+)	CAACGCAGTCGNNNGATGGCCCAG
VIM-random 61(−)	CTGCGCCATCNNNCGACTGCCTTG
VIM-random 64(+)	GTTTGATGGCNNNGTCTACCCG
VIM-random 64(−)	CGGCTAGACNNNGCCATCAAAC
VIM-random 67(+)	GGCGCAGTCNNNCCGTCCAATG
VIM-random 67(−)	CATTGGACGGNNNGACTGCGCC
VIM-random 87(+)	GATTGATACAGCGNNNGCTGCGAAAAAC
VIM-random 87(−)	GTTTTTCGCACCNNNCGCTGTATCAATC
RT-VIM-2/fwd	TTGTCCGTGATGGTGATGAG
RT-VIM-2/rev	TGAAAGTGCGTGGAGACTGC
RT-gapA/fwd	CGACAAATATGCTGGCCAGG
RT-gapA/rev	GTAGTAGCGTGAACGGTGGT

^aThe VIM-2-EXP primers were used to clone the bla_VIM-2 ORF, which added EcoRI (italicized) and NdeI (underlined) restriction sites at the 5′ end of the gene and XhoI (italicized) and BamHI (underlined) restriction sites after the bla_VIM-2 stop codon. VIM-random primers were used to obtain libraries with mutations at selected positions. RT oligonucleotides were used in real-time PCR assays.

Production and purification of VIM-2 mutants.

Selected VIM-2 mutants (A64W, W87A, and W87F) were overproduced using a strategy similar to that for the wild-type enzyme, based on the T7 promoter expression system (4). Briefly, the NdeI-BamHI restriction fragment obtained from the relevant pLBII-VIM-2 derivative was subcloned into vector pET-9a and subsequently transformed in E. coli BL21(DE3). The mutants were produced in the autoinducing medium ZYP-5052 (22). Purification protocols were adapted from those described by Docquier et al. (4), and mutants were purified to homogeneity using two or three chromatography steps, consisting of an initial anion-exchange chromatography (Q Sepharose Fast Flow or Source 15Q; GE Healthcare) at pH 7.2 followed by a-high resolution anion exchange (MonoQ 5/50 GL column) at pH 9.2 and, when required, a gel filtration (Superdex HR 10/30 columns, GE Healthcare) using 50 mM HEPES-50 μM ZnSO₄ (pH, 7.5) buffer for elution.

Protein techniques and determination of kinetic parameters.

Total, soluble, and insoluble protein fractions from bacterial culture samples were prepared as previously described (22). SDS-PAGE was performed by the Laemmli method, using final acrylamide concentrations of 12% and 5% (wt/vol) for the separating and the stacking gels, respectively. After electrophoresis, the protein bands were stained with SimplyBlue SafeStain (Invitrogen, Carlsbad, CA). Immunological detection of the VIM-2 β-lactamase (and mutants thereof) after electrophoretic separation was carried out using standard Western blotting procedures and anti-VIM-2 antibodies elicited from rabbits, as previously described (9). β-Lactamase activities of the various samples were measured spectrophotometrically in 10 mM HEPES (pH 7.5) buffer supplemented with 50 μM ZnSO₄ using 150 μM imipenem as the substrate. Protein concentrations in solution were determined with the Bio-Rad protein assay (Bio-Rad, Richmond, CA), using serum albumin (BSA) as a standard. The authenticity of the purified VIM-2 mutants were confirmed by mass spectrometry using peptide mass fingerprint analysis (following enzymatic digestion of the protein sample with trypsin) (14). Kinetic parameters for the hydrolysis of β-lactam substrates were determined using the experimental conditions adopted for the wild-type VIM-2 (4).

Conformational stability of VIM-2 and mutants.

Chemical-induced unfolding transitions were followed at 25°C, using both circular dichroism (CD) and intrinsic fluorescence measurements, in the presence of 4 μM purified enzyme (WT VIM-2 and mutants thereof) and various concentrations of guanidinium chloride (GdmCl) (ranging from 0 to 4.5 M), as described previously (5). Samples were left to equilibrate for ca. 16 h, and the reversibility of chemical-induced denaturation was assayed by diluting (20-fold) the protein sample incubated in the presence of 6 M GdmCl (under these conditions, complete unfolding was achieved within less than 10 min) in 10 mM HEPES (pH 7.5) buffer supplemented with 50 μM ZnSO₄. Enzyme activity was measured in the same buffer using 150 μM imipenem as the substrate. GdmCl concentrations in the various samples were measured on the basis of refractive index measurements (18), using an R-5000 hand-held refractometer (Atago Co. Ltd., Tokyo, Japan).

Changes in intrinsic fluorescence emission (excitation wavelength [λ_ex] = 280 nm; emission wavelength [λ_em] = 333 and 323 nm for the WT and mutant enzymes, respectively) were monitored using a Perkin-Elmer LS50B spectrofluorimeter (Perkin-Elmer, Whaltan, MA), as described previously (12).

CD measurements were performed with a Jasco J-810 spectropolarimeter (Jasco Ltd., Great Dunmow, United Kingdom), using a protein concentration of 4 μM and a 1-mm-path-length cell, under constant nitrogen flow. Spectra were acquired at a scan speed of 20 nm/min, with a 1-nm bandwidth and a 4-ms response. Spectra were measured five times, averaged, and corrected by subtraction of the solvent spectrum obtained under the same conditions. GdmCl unfolding curves were recorded at 220 nm, using a 1-nm bandwidth. At all denaturant concentrations, at least 24 data points were acquired, averaged, and corrected for the contribution of the solvent, as described previously (5).

Equilibrium unfolding curves obtained using either intrinsic fluorescence or CD measurements were analyzed on the basis of two-state (N ⇋ U) or three-state (N ⇋ H ⇋ U) models for the unfolding transition, as described previously (12).

Thermal denaturation was monitored at 220 nm, using a protein concentration of 4 μM in 10 mM HEPES (pH 7.5) supplemented with 50 μM ZnSO₄. The temperature was increased monotonically in the range from 20°C to 70°C at a rate of 0.55°C/min. The reversibility of denaturation was assayed by following the CD signal upon cooling the sample down to 20°C at the same rate. Data were acquired using a 0.2°C interval, with a 4-s integration time and a 1-nm bandwidth, and analyzed as described previously (6).

The program Grafit 5.0.10 (Erithacus Software Ltd., Horley, United Kingdom) was used for nonlinear least-squares analysis of the data.

Real-time PCR assays.

Total RNA samples from E. coli XL-1(pLBII-VIM-2) and XL-1(pLBII-VIM-2-W87F) grown in MH broth (A₆₀₀, 1.4) were extracted in triplicate using the SV total RNA isolation kit (Promega, Madison, WI) and cDNA obtained using the ImProm II reverse transcriptase (Promega). The amount of bla_VIM-2 transcripts in the various samples was determined by real-time PCR, using the LightCycler DNA Master Hybridization Probes kit (Roche Molecular Biochemicals, Mannheim, Germany) with suitable negative and positive controls. Transcripts of gapA (a housekeeping gene encoding glyceraldehyde-3-phosphate dehydrogenase) were used as the internal reference, and relative changes in bla_VIM-2 or bla_VIM-2[W87F] expression were computed as described by Livak and Schmittgen (13).

RESULTS AND DISCUSSION

Selection of positions for saturation mutagenesis studies and analysis of mutant libraries.

On the basis of the comparison of available MBL structures, including that of VIM-2, and current knowledge on the structure-activity relationships of VIM-type enzymes (see reference 20 and references therein), four residues were identified as potentially relevant to catalysis and/or enzyme structure: (i) Ala-64, which represents the structurally equivalent residue of the mobile-flap Trp-64 residues found in IMP-type enzymes (Fig. 1) and whose role in substrate or inhibitor binding has been previously demonstrated by X-ray crystallography and mutagenesis experiments (3, 17), and (ii) residues Phe-61, Tyr-67, and Trp-87, which form a hydrophobic patch close to the active site, lying close to the second metal binding site, and whose role in substrate binding has been hypothesized (4) (Fig. 1). More recently, the importance of positions 61 and 67 for binding of a mercaptocarboxylate inhibitor was demonstrated by X-ray crystallography (24).

FIG. 1.

Cartoon representation of the VIM-2 X-ray crystal structure (PDB code 1KO3). The positions of the residues selected for saturation mutagenesis (Phe-61, Ala-64, Tyr-67, and Trp-87) and of two additional tryptophan residues (Trp-53 and Trp-242) are shown in dark gray. The active-site residues involved in metal coordination are shown as sticks and are numbered according to the standard numbering scheme for metallo-β-lactamases (7). Zn²⁺ ions are represented by spheres.

The roles of these positions were investigated using a saturation mutagenesis approach, which allowed us to investigate the effect of substitutions at each position with all possible amino acid residues. This method represents a valuable tool to investigate the roles of specific protein positions in enzyme structure and/or function (15). For each library of mutants, a random selection of approximately 100 mutants was investigated for their susceptibility to ampicillin (a substrate of the VIM-2 enzyme), and susceptibility data were correlated to the sequence of the mutated VIM-2 ORF retrieved from each individual clone.

Results for the library of mutants with mutations at position 64.

Within the library of mutants with mutations at position 64, a large proportion of clones (92%) showed no significant variation of the MIC for ampicillin (≥64 μg/ml) compared to the strain producing the wild-type (WT) VIM-2 (data not shown). Moreover, the lowest ampicillin MIC observed with clones from this mutant library was 4-fold higher than that with the negative-control strain (i.e., the E. coli host strain transformed with the empty vector) or that with the clones carrying nonsense mutations (8% of the library). These results suggested that position 64 is tolerant to substitution, and sequencing data confirmed that amino acids such as Val, Leu, Ile, Gly, Tyr, and Trp were compatible with the higher MICs (≥64 μg/ml) (data not shown). Interestingly, the A64W mutant was also found in this library. This mutant carries a residue identical to that naturally found at the corresponding position in IMP-type variants (located in the so-called mobile flap) and whose role in enzyme function has been documented by various studies (3, 15, 17). The antimicrobial susceptibility profile of the E. coli clone producing the A64W mutant did not show any significant variation from that of the strain producing the WT enzyme, suggesting that this substitution does not impair or improve enzyme activity in vivo or modify its spectrum (Table 2).

TABLE 2.

Antimicrobial susceptibility profiles of E. coli XL1-Blue(pLBII-VIM-2), producing WT VIM-2 MBL, and isogenic derivatives producing the A64W, W87A, and W87F mutants^a

Antibiotic	MIC (μg/ml)
	VIM-2	A64W	W87F	W87A	pBC-SK
Ampicillin	64	64	8	2	2
Piperacillin	8	8	0.5	0.5	0.5
Cephalothin	256	256	16	16	16
Cefoxitin	128	128	8	8	8
Cefuroxime	256	256	8	8	8
Cefotaxime	>32	>32	0.12	0.12	0.12
Ceftazidime	2	2	0.12	0.12	0.12
Cefepime	0.5	0.5	0.06	0.12	0.06
Ceftriaxone	4	4	0.12	0.12	0.12
Imipenem	1	1	0.12	0.12	0.12
Meropenem	0.5	0.5	≤0.03	≤0.03	≤0.03
Ertapenem	0.5	0.5	≤0.03	≤0.03	≤0.03
Aztreonam	0.12	0.12	0.12	0.12	0.12

^aThe strain carrying the empty vector (pBC-SK) is also shown for comparison.

Functional properties of the VIM-2 A64W mutant.

To investigate whether the A64W substitution could affect the individual kinetic parameters and, in particular, the binding of β-lactam substrates, the VIM-2 A64W mutant was purified and subjected to kinetic analysis. Peptide mass fingerprint analysis of the purified A64W mutant confirmed the authenticity of the enzyme preparation, as the observed mass of the protein fragment containing position 64 (residues 44 to 75) was in excellent agreement with the theoretical value of the peptide showing a Trp residue at this position (3392.86 and 3392.71, respectively). Steady-state kinetic parameters for the hydrolysis of representative β-lactam substrates were measured and compared to those for the WT VIM-2 and IMP-1 (bearing a Trp-64) and its W64A mutant (17) (Table 3). In good agreement with the susceptibility data, the presence of a Trp residue at position 64 in VIM-2 did not strongly affect the values of individual kinetic parameters for the hydrolysis of representative β-lactam substrates (the k_cat or K_m values did not vary more than 5-fold). This yielded overall similar k_cat/K_m values, with ampicillin (8-fold decrease) and cefuroxime (9-fold increase) being the most affected substrates (Table 3). This situation strongly differs from that previously observed with the IMP-1 enzyme, for which the presence of Trp-64 was shown to be critical to recognition of all β-lactam substrates. Indeed, the W64A substitution in IMP-1 caused a remarkable increase of K_m values for most substrates, although the magnitude of the K_m variation was also dependent on the substrate (Table 3). It has been suggested that the mobility of Trp-64 and the various conformations it could adopt upon substrate binding could contribute differently to substrate stabilization (17). In contrast, the A64W substitution in VIM-2 led to higher K_m values than with the WT enzyme for ampicillin and imipenem (i.e., an effect opposite to that observed with IMP-1), while only a moderate decrease was observed with cephalosporins and there was no effect with meropenem. Altogether, these data indicate that the role of position 64 is different in IMP-1 and VIM-2 and is likely less relevant in the latter, especially regarding the interaction with the substrate.

TABLE 3.

Steady-state kinetic parameters for the hydrolysis of β-lactam substrates of VIM-2 and its A64W mutant^a

Substrate	kcat (s−1)				Km (μM)				kcat/Km (μM−1 s−1)
	VIM-2	VIM-2 A64W	IMP-1	IMP-1 W64A	VIM-2	VIM-2 A64W	IMP-1	IMP-1 W64A	VIM-2	VIM-2 A64W	IMP-1	IMP-1 W64A
Ampicillin	125	50	200b	86b	90	300	320b	1,250b	1.4	0.17	0.63b	0.07b
Cephalothin	130	135	32c	>250c	11	6	5c	>500c	12	22	6.4c	0.31c
Cefoxitin	15	—d	9	40	13	—	3	45	1.2	—	3.0	0.9
Cefuroxime	8	15	8	—	20	4	37	—	0.4	3.7	0.22	—
Cefotaxime	70	26	13.4	98	12	7	2.8	61	5.8	3.7	4.8	1.5
Ceftazidime	3.6	0.7	8	—	72	40	44	—	0.05	0.018	0.18	—
Cefepime	>40	>50	7	—	>400	>400	11	—	0.1	0.12	0.66	—
Nitrocefin	770	270	130	700	18	10	8.2	160	43	27	14	4.3
Imipenem	34	60	68	67	9	21	25	120	3.8	2.9	2.7	0.55
Meropenem	5	6	5	—	2	2	10	—	2.5	3.0	0.5	—

^aKinetic parameters were determined using the experimental conditions described previously for wild-type VIM-2 (4). Data for IMP-1 and its W64A mutant are shown for comparison (17). Data for WT IMP-1 and cefuroxime, ceftazidime, cefepime, and meropenem are from reference 11. Standard deviations were below 10%.

^bData for benzylpenicillin.

^cData for cephaloridine.

^d—, data not available.

Results for the library of mutants with mutations at positions 61, 67, and 87.

Considering that position 64 in VIM-2 is apparently not as significant for substrate binding as it is in IMP-1, we probed the relevance of other potential substrate binding determinants that could be found close to the active site of VIM-2, and particularly in the hydrophobic pocket lying close to the zinc 2 site that was previously hypothesized to play a role in substrate recognition based on modeling and structural studies (4, 24). To this end, we investigated the importance of aromatic residues found at positions 61, 67, and 87, which might be critical for interacting with the substrate C-6 side chain, which might carry hydrophobic and/or bulky groups.

The substitutions at these three positions affected the in vitro resistance differently (Fig. 2). With mutations at position 61, 75% of clones exhibited an ampicillin MIC of ≥32 μg/ml, i.e., compatible with production of an active enzyme (Fig. 2). Sequencing results confirmed that this residue (Phe in the WT enzyme) could be replaced by hydrophobic (Ala, Val, Leu, or Ile), aromatic (Tyr or Trp), and hydrophilic (Asn, His, Lys, or Thr) residues, and also by Gly, without significantly compromising in vitro resistance (ampicillin MICs were ≥32 μg/ml for all these mutants). Some substitutions had a more negative impact on resistance, as replacement of Phe-61 with Cys, Pro, Ser, and Arg (found in 19% of the clones of this library) were associated with ampicillin MIC values ranging 8 to 16 μg/ml. Interestingly, none of the detected substitutions was nevertheless able to restore full susceptibility to ampicillin, with the exception of the mutants in which a stop codon was introduced, which showed ampicillin MIC values of ≤4 μg/ml) (Fig. 2). This behavior strongly suggests that position 61 is unlikely to be critical for enzyme function and/or structure.

FIG. 2.

Distribution of ampicillin MIC values for the libraries of mutants with mutations at positions 61 (white bars), 67 (gray bars), and 87 (black bars), constituting the hydrophobic patch in the VIM-2 active site. The nature of the amino acid obtained after sequencing the bla_VIM-2 ORF in the various clones is shown in correspondence with the measured MIC values. The amino acid corresponding to that of the WT enzyme is marked with an asterisk. The ampicillin MIC values for the WT strain ranged from 64 to 128 μg/ml. Stop codons (nonsense mutations) are also indicated (♦).

A similar situation was observed, to some extent, with the library with mutations at position 67. Apart the Y67P substitution, which restored full susceptibility to ampicillin, the other substitutions had more varied effects on ampicillin MIC values. However, although some substitutions slightly improved (ampicillin MICs, ≥128 μg/ml; 10% of the library) or impaired (ampicillin MICs, 16 to 8 μg/ml; 36% of the library) resistance (Fig. 2), several amino acids whose chemical properties are different from those of tyrosine (found in the WT) are compatible with a functional enzyme.

In contrast, drastically different results were obtained with the library with mutations at position 87 (Trp in the WT enzyme). Indeed, 69% of the randomly selected clones showed a strongly decreased ampicillin MIC (≤4 μg/ml) compared to that of the WT (Fig. 2). Moreover, sequence data revealed that (i) all mutants exhibiting MIC values of ≥32 μg/ml were associated with the presence of a Trp residue at position 87; (ii) mutants with a Phe or Tyr residue exhibited MIC values of 16 to 8 μg/ml; and (iii) all other detected substitutions (i.e., Ala, Val, Leu, Ile, Gly, Pro, Ser, Thr, Met, Asn, Gln, Lys, and Arg) were associated with MIC values of ≤4 μg/ml. Altogether, these data suggested that the Trp-87 residue is very important for enzyme structure and/or function, as its replacement by any other amino acid invariably leads to a strong decrease of ampicillin resistance in E. coli. The antimicrobial susceptibility profiles of isogenic strains producing the VIM-2 W87F and W87A mutants confirmed the dramatic decrease of MIC values for all tested β-lactam antibiotics, indicating that this behavior does not depend on the nature of the antibiotic (Table 2). To further investigate this aspect, the W87F and W87A mutants were purified and further characterized.

Functional properties of the VIM-2 W87F and W87A mutants.

Using an E. coli BL21(DE3)-based overexpression system, the VIM-2 W87F mutant could be produced and purified (although with an overall lower yield than for the WT enzyme). The authenticity of the enzyme preparation was demonstrated by peptide mass fingerprinting and mass spectrometry (observed and theoretical masses of the peptide from position 76 to 90 with a phenylalanine residue at position 87, 1577.68 and 1577.81, respectively). Surprisingly, and in contrast to the antimicrobial susceptibility data, the purified mutant was functional and showed catalytic efficiencies (k_cat/K_m) for the hydrolysis of a representative set of β-lactam substrates that were very close to those of the WT VIM-2, with variations that did not exceed a 2-fold increase or decrease (Table 4). This usually resulted from minor variations of individual kinetic parameters (k_cat and K_m), except for meropenem, for which both k_cat and K_m values increased (9- and 5-fold, respectively). Interestingly, the K_m value for imipenem was only poorly affected, likely reflecting the influence of a 1-β-methyl group or a bulkier C-2 side chain in meropenem (Table 4).

TABLE 4.

Steady-state kinetic parameters for the hydrolysis of β-lactam substrates of VIM-2 and its W87F and W87A mutants^a

Substrate	kcat (s−1)			Km (μM)			kcat/Km (μM−1 s−1)
	VIM-2	W87F	W87A	VIM-2	W87F	W87A	VIM-2	W87F	W87A
Ampicillin	125	115	165	90	56	189	1.4	2.0	0.87
Cephalothin	130	130	320	11	10	28	12	13	11
Cefuroxime	8	12	15	20	18	23	0.4	0.67	0.65
Cefotaxime	70	143	25	12	19	83	5.8	7.5	0.3
Ceftazidime	3.6	7.8	≥0.152	72	170	≥350	0.05	0.046	0.00043
Cefepime	>40	>20	>4.5	>400	>450	>350	0.1	0.048	0.012
Imipenem	34	22	119	9	6	22	3.8	3.7	5.4
Meropenem	5	45	35	2	10	18	2.5	4.5	1.9

^aStandard deviations were below 10%.

Similar results were obtained with the W87A mutant, which also showed small variations of the kinetic parameters except for the k_cat/K_m values for cefotaxime, ceftazidime, and cefepime, which were consistently lower (Table 4). For cefotaxime, and likely the other two substrates, this was due to both a k_cat decrease and a K_m increase. The apparent loss of activity toward these substrates, which present a typical bulky oxyimino group that should fit in the cavity close to Trp-87, now replaced with a shorter Ala residue, indicates that a large hydrophobic residue in that position might promote a better recognition of these substrates.

From the analysis of the kinetic data obtained with the VIM-2 W87A and W87F variants, it is clear that the replacement of the Trp-87 residue was not globally affecting the functional properties and the substrate profile of the enzyme, while it was responsible for an important decrease of the MIC values of all tested antibiotics. Thus, the discrepancy between the functionality of W87A and W87F variants and susceptibility data should rely on factors affecting enzyme structure, folding, and/or stability in vivo. In order to probe this hypothesis, the stabilities of these mutants were investigated and compared to that of the WT VIM-2, using both circular dichroism (CD) and fluorescence measurements.

Importance of Trp-87 for VIM-2 stability.

WT VIM-2 contains three tryptophan residues at positions 53, 87, and 242; two of them (residues 53 and 87) are located at the protein surface and thus are rather exposed to the solvent, and the other one (position 242) is buried in the enzyme structure (solvent-accessible surface, 0.3 Ų) (Fig. 1). The intrinsic fluorescence emission spectrum of the native enzyme shows a maximum (λ_max) at 340 nm (Fig. 3a). In comparison with that of the WT enzyme, the fluorescence spectra of the W87F and W87A mutants display reduced intensities (which is due merely to the contribution of Trp-87 in the WT enzyme) and a lower λ_max value (329 to 331 nm) (Fig. 3a). The blue shift observed in the absence of Trp-87, which is located in close proximity to the zinc binding residues, is in agreement with the rather high solvent accessibility of this residue, as indicated by its solvent-accessible surface area of ∼31 Ų (computed using the WT VIM-2 structural coordinates [PDB code 1KO3]) (8, 10).

FIG. 3.

(a) Fluorescence spectra of the native forms of wild-type VIM-2 (thick line) and its W87A mutant (continuous thin line) (the spectrum of the W87F was identical to that obtained with the W87A mutant and has been omitted for clarity); the discontinuous line shows the spectrum of the unfolded form of wild-type VIM-2 (spectra with reduced intensities but very similar λ_max values were obtained for the W87A and W87F mutants). (b) GdmCl-induced equilibrium unfolding transition of VIM-2 (empty circles) and its W87A (empty squares) and W87F (filled circles) mutants at pH 7.5 and 25°C, monitored by the change in fluorescence intensity at 333 nm (WT) and 323 nm (mutants). The data were analyzed on the basis of two-state (WT and W87A enzymes) and three-state (W87F enzyme) models, and the solid lines represent the best fits calculated using the values of the thermodynamic parameters in Table 5. Data are presented as the fraction of unfolded enzyme (f_U), as a function of GdmCl concentration. (c) Temperature-induced denaturation of VIM-2 (empty circles), the W87A mutant (empty squares), and the W87F mutant (filled circles). The unfolded fraction, obtained from the far-UV CD data at 220 nm, is plotted as a function of temperature. Apparent T_m values were computed as previously described (6).

The relative stability of WT VIM-2 and its W87F and W87A mutants was investigated by CD and fluorescence measurements in the presence of various concentrations of GdmCl (0 to 4.5 M). With the three enzymes, GdmCl-induced denaturation was shown to be reversible, since >90% of the imipenem-hydrolyzing activity (measured prior to denaturation) was recovered upon dilution of the denaturant (see Materials and Methods). The fluorescence spectra (Fig. 3a) of the unfolded forms of the three enzymes showed very similar shapes, with high λ_max values (357 to 358 nm), corresponding to fully exposed tryptophan residues. As expected from the removal of one tryptophan residue (position 87), however, both mutants showed reduced fluorescence intensities. With the three proteins, the significant reduction in fluorescence intensity measured at 333 nm (WT) and 323 nm (mutants) allowed the equilibrium transition between the native and unfolded states to be readily monitored (Fig. 3b). These experiments showed marked differences in the behaviors of the three enzymes. With the WT enzyme and the W87A mutant, the unfolding curves obtained by intrinsic fluorescence emission and CD measurements were indistinguishable and showed a single cooperative transition, compatible with a two-state process (Fig. 3b). These curves could be fitted to obtain the thermodynamic parameters for the transition (Table 5), showing a large difference between the variation of free energy for the unfolding transition for the two enzymes and thus indicating that the W87A mutation has a significant destabilizing effect (∼9 kJ/mol) (Table 5). In contrast with the other enzymes, mutant W87F exhibits a more complex unfolding transition (as monitored by both fluorescence and CD), indicating the formation of at least one intermediate species during the denaturation process (Fig. 3b). However, no satisfactory values for the thermodynamic parameters characterizing the unfolding transition of this mutant could be computed after fitting the transition curve using a three-state model.

TABLE 5.

Thermodynamic parameters for the unfolding transition of VIM-2 and its W87A and W87F mutants^a

Type of unfolding	Parameter	Value (mean ± SD) for:
		VIM-2	W87A	W87F
GdmCl induced	ΔG°(H2O) (kJ/mol)	31 ± 4	22 ± 2	NDb
	m (kJ/mol·M)	−18 ± 2	−17 ± 1	ND
	Cm (M)	1.7 ± 0.4	1.3 ± 0.2	0.9/1.8c
Thermald	Tm (°C)	54 ± 0.2	50 ± 0.1	44 ± 0.2

^aThe thermodynamic parameters of GdmCl- and temperature-induced unfolding were measured at pH 7.5 and at a temperature of 25°C and computed from the analysis of the equilibrium transitions. The ΔG°, m, and C_m values are the means of those obtained by far-UV CD and fluorescence measurements and were computed as described previously (5, 12). m is a measure of the dependency of the free energy on the denaturant concentration, and C_m is the transition midpoint, i.e., the denaturant concentration at which [N] = [u]. The apparent T_m is the temperature at the midpoint of the denaturation curve.

^bND, not determined.

^cApproximate values retrieved from the transition curves for the first and second transitions, respectively.

^dThe reversibility of thermal unfolding was estimated as the recovery of the CD signal after cooling of the sample as described in Materials and Methods and was <5% in all cases.

As an alternative to chemical-induced unfolding and in order to compare the three enzymes, thermal denaturation experiments were carried out by recording the CD signal intensity at 220 nm (Fig. 3c). Although thermal unfolding did not appear to be reversible (as <5% of the CD signal was retrieved after cooling the samples), an apparent T_m value could be computed on the assumption of an apparent two-state model for unfolding of the three enzymes (Table 5). These data provided additional confirmation that both mutants are less stable than the WT enzyme and that the W87F mutant was even less stable than the W87A mutant.

Biological consequences of the Trp-87 substitution.

Overall, the comparative analysis of the biophysical properties of WT VIM-2 and the W87A and W87F mutants indicates that replacement of Trp-87 significantly affects the stability of the enzyme in vitro and potentially also in vivo. In particular, this might affect the folding process of the mutant enzymes following protein synthesis and thus the kinetics of accumulation of the folded and active enzyme produced by the bacteria, affecting the actual level of resistance of the bacterial host to antibiotics.

To further investigate this point, β-lactamase production in isogenic E. coli strains expressing WT VIM-2 or the W87F mutant was examined. The level of bla_VIM-2 transcripts was determined by real-time PCR, and the amount of β-lactamase was estimated in crude extracts both by Western blot analysis using rabbit anti-VIM-2 antibodies and by measuring the imipenem-hydrolyzing activity (the latter allows the detection of the active and properly folded form of the enzyme). No significant differences in the growth rates of the two organisms were observed (Table 6). However, the kinetics of appearance of the imipenem-hydrolyzing activity were strikingly different in the two strains, as the activity of the strain producing the W87F variant was detectable with a consistent delay and in a much smaller amount (Table 6). Indeed, after 4.5 h of growth, the variation in the activity measured in samples obtained from the two strains was up to 60-fold (Table 6), while the amounts of the MBL gene transcripts were not significantly different (mean variation, ≥0.6 compared to the WT). A Western blot analysis carried out on the total, soluble, and insoluble protein fractions extracted from the same samples revealed, surprisingly, a roughly similar total amount of β-lactamase (data not shown) in both strains, meaning that the replacement of the Trp-87 residue does not likely affect β-lactamase synthesis. These data indicate that the W87F variant might transiently accumulate as an improperly folded protein, possibly reflecting a slower in vivo folding process than for the WT enzyme. In addition, a significant amount of metallo-β-lactamase was also detected in the insoluble fraction, showing that the replacement of Trp-87 might also affect in vivo protein solubility. Overall, residue Trp-87 seems to be critical for the stability, folding, and rate of production of the β-lactamase in the bacteria, in agreement with the lower in vitro stability of the W87F mutant compared to the WT enzyme. These data are now consistent with the susceptibility data and likely apply to all the other obtained variants which also showed full susceptibility to ampicillin (Fig. 2), since the strain producing a VIM-2 W87 variant does not seem able to produce in a timely manner a biologically active amount of β-lactamase that can protect itself from the action of the antibiotic.

TABLE 6.

Kinetics of growth and production of β-lactamases of E. coli strains XL-1(pLBII-VIM-2) and XL-1(pLBII-VIM-2- W87F), producing the wild-type VIM-2 and the W87F variant, respectively^a

Time (h)	E. coli XL-1 (pLBII-VIM-2)		E. coli XL-1 (pLBII-VIM-2-W87F)
	A600	Normalized IPM- hydrolyzing activity (nmol/min·ml)	A600	Normalized IPM-hydrolyzing activity (nmol/min·ml)
0	0.10	—b	0.10	—
0.5	0.18	—	0.18	—
1.2	0.32	—	0.32	—
1.7	0.52	30	0.51	—
2.3	0.80	39	0.77	—
2.8	1.0	110	0.98	—
3.3	1.2	170	1.2	—
3.8	1.4	240	1.3	4.7
4.3	1.7	270	1.6	4.7
5.0	2.0	330	1.8	7.9
6.0	2.3	380	2.1	9.5

^aThe cell concentration in the cultures was estimated by measuring the absorbance at 600 nm (A₆₀₀). The amount of β-lactamase was estimated as the imipenem-hydrolyzing activity in crude cell extracts normalized to the cell density (see Materials and Methods for details). These results are the means of three independent measurements, and standard deviations were below 10%.

^b—, imipenem-hydrolyzing activity not detected (≤2.7 nmol/min·ml).

Concluding remarks and significance.

MBLs represent a significant threat to antimicrobial chemotherapy and are associated with acquired multidrug resistance phenotypes and, more importantly, also with pan-resistance phenotypes exhibited by clinical isolates that belong to the most relevant pathogenic species (1, 16, 23). The development of efficient MBL inhibitors, which could help to preserve the efficacy of β-lactam antibiotics (which are by far the most successful class of anti-infective drugs), is thus becoming an issue of great practical importance. However, the diversity of MBLs represents a critical obstacle to the design of broad-spectrum inhibitors of these enzymes, and, in that perspective, further knowledge on their structure-function relationships might be valuable.

The results of this study highlighted some unique features of the VIM-2 enzyme: (i) the contribution of position 64 to the functional properties of the enzyme, and especially substrate recognition, is apparently less relevant than in IMP-type enzymes, which are the other major group of clinically relevant acquired MBLs (and also than in some other subclass B1 members, e.g., Bacteroides fragilis CcrA, which also have the “flapping” Trp in position 64) (3, 17); (ii) hydrophobic residues at positions 61 and 67, which were recognized as potentially important for IMP-1 function in a previous study (15), might be replaced with amino acids exhibiting different chemical properties without impairing in vitro resistance; and (iii) residue Trp-87 appears to be critical to enzyme structure and folding while not being determinant for substrate recognition and catalysis. The last feature represents the most important and unexpected result of this work.

The present data suggest that Trp-87 would represent an excellent target for interaction with inhibitors, considering that (i) it is rather exposed to the solvent and thus accessible to small ligands and (ii) its replacement, leading to slower accumulation of active enzyme and thus to the loss of antibacterial resistance of the organism as shown in the case of VIM-2, is very unlikely to occur in natural variants. The latter point is interesting as it gives better confidence that inhibitor resistance by means of single mutation would not easily occur in nature, although it would be fascinating to investigate the impact of the Trp-87 substitution in other hosts and/or growth conditions that more closely resemble those encountered in vivo.

It is noteworthy that the Trp-87 residue is present in all VIM-type enzymes, including VIM-1 and VIM-4, which are becoming prevalent MBLs in Enterobacteriaceae isolates in some settings (mainly Greece and Italy) (21, 23), and also in many others enzymes belonging to subclass B1 (e.g., BlaB-type [from Elizabethkingia meningoseptica] and IND-type [from Chryseobacterium indologenes] enzymes and the acquired GIM-1 enzyme) (see reference 20 and references therein). We are currently investigating the role of Trp-87 in other relevant MBLs to verify that its critical role would also apply to different enzymes, thus providing additional support for the development of broad-spectrum MBL inhibitors able to address the increasingly important clinical threat of MBL-producing pathogens.