Insights into β-Lactamases from Burkholderia Species, Two Phylogenetically Related yet Distinct Resistance Determinants*

Background: Resistance to β-lactams in Burkholderia is mediated by different β-lactamases (e.g. PenA and PenI). Results: PenA from B. multivorans is a carbapenemase, and PenI from B. pseudomallei is an extended-spectrum enzyme. Conclusion: Subtle changes within the active site of β-lactamases result in major phenotypic changes. Significance: Future antibiotic design must consider the distinctive phenotypes of PenA and PenI β-lactamases. Burkholderia cepacia complex and Burkholderia pseudomallei are opportunistic human pathogens. Resistance to β-lactams among Burkholderia spp. is attributable to expression of β-lactamases (e.g. PenA in B. cepacia complex and PenI in B. pseudomallei). Phylogenetic comparisons reveal that PenA and PenI are highly related. However, the analyses presented here reveal that PenA is an inhibitor-resistant carbapenemase, most similar to KPC-2 (the most clinically significant serine carbapenemase), whereas PenI is an extended spectrum β-lactamase. PenA hydrolyzes β-lactams with kcat values ranging from 0.38 ± 0.04 to 460 ± 46 s−1 and possesses high kcat/kinact values of 2000, 1500, and 75 for β-lactamase inhibitors. PenI demonstrates the highest kcat value for cefotaxime of 9.0 ± 0.9 s−1. Crystal structure determination of PenA and PenI reveals important differences that aid in understanding their contrasting phenotypes. Changes in the positioning of conserved catalytic residues (e.g. Lys-73, Ser-130, and Tyr-105) as well as altered anchoring and decreased occupancy of the deacylation water explain the lower kcat values of PenI. The crystal structure of PenA with imipenem docked into the active site suggests why this carbapenem is hydrolyzed and the important role of Arg-220, which was functionally confirmed by mutagenesis and biochemical characterization. Conversely, the conformation of Tyr-105 hindered docking of imipenem into the active site of PenI. The structural and biochemical analyses of PenA and PenI provide key insights into the hydrolytic mechanisms of β-lactamases, which can lead to the rational design of novel agents against these pathogens.

Currently, ␤-lactam antibiotics are recommended as a treatment option for infections due to B. cepacia complex and B. pseudomallei (27,28). However, resistance to ␤-lactams is present, and the mechanistic basis for resistance is correlated to the expression of ␤-lactamases (29 -32) as well as decreased expression of penicillin-binding proteins in B. pseudomallei (33). B. multivorans and B. pseudomallei possess related ␤-lactamases, PenA and PenI, respectively, that share 71% similarity and 62% identity (34).
At present, ␤-lactamases (EC 3.5.2.6) are classified into four different classes based on structural similarities (A-D) (35). Class A, C, and D ␤-lactamases use a serine (e.g. Ser-70 in class A enzymes; Ambler amino acid-numbering scheme (35)), as the nucleophile, and class B ␤-lactamases are metalloenzymes using one or two Zn 2ϩ ions for catalysis. PenA and PenI are class A ␤-lactamases. In general, class A ␤-lactamases hydrolyze ␤-lactams through the reaction mechanism represented in Fig. 1. Here, we use biochemical assays, crystallography, and molecular modeling to dissect the hydrolytic mechanisms of PenA and PenI. Understanding how ␤-lactamases interact with their target ligand, ␤-lactams, is critical for devising strategies to circumvent them and treat infections by these drug-resistant and virulent pathogens.

EXPERIMENTAL PROCEDURES
Protein Analysis-Protein sequences were analyzed by BlastP and Bl2Seq from the National Center for Biotechnology Information website (www.ncbi.nlm.nih.gov). Multiple sequence alignments were generated using Clustal 2.0.12 and BOXSHADE version 3.31C from the Mobyle@Pasteur website. Phylogenetic trees were constructed by submitting amino acid sequences and using the "one-click" mode of the phylogenetic analysis tool developed by Information Génomique et Structurale and funded by Réseau National des Génopoles (36 -42).
Bacterial Strains and Plasmids-The bla PenA FIGURE 1. Proposed reaction mechanism for a cephalosporin (red) with a class A serine ␤-lactamase (black). The carbonyl of ␤-lactam sits in the oxyanion hole (electrophilic center) formed by the backbone nitrogens of Ser-70 and Xaa-237. The carboxylate forms a salt bridge to a positively charged amino acid, Arg-220 with PenA and PenI. Either Lys-73 directly removes a proton from Ser-70 or Glu-166 removes a proton from a water molecule (blue), which activates the nucleophilic Ser-70 to initiate acylation. Unprotonated Ser-70 attacks the ␤-lactam bond forming an acyl-enzyme. A proton shuttle starting with the ␤-lactam and Ser-130 to Lys-73 followed by Glu-166 results in an unprotonated Glu-166. Unprotonated Glu-166 removes a proton from a water molecule (i.e. DW), which attacks the acyl-enzyme complex resulting in water being added across the bond and deacylation of the inactivated ␤-lactam (69,80).
ATCC 17616. The bla PenI gene was a kind gift from Dr. Herbert P. Schweizer (Colorado State University, Fort Collins, CO). The full-length bla PenA and bla PenI genes were cloned into the pET24a(ϩ) plasmid using NdeI and BamHI restriction sites. Both genes were subsequently subcloned from the pET24a(ϩ) plasmid into pBC SK(ϩ) phagemid, including the ribosomebinding site of pET24a(ϩ) (pBC SK(ϩ) lacks a ribosome-binding site) by using SacII and BamHI restriction sites for bla PenA and SacI and BamHI restriction sites for bla PenI . The bla PenA gene minus the first 81 nucleotides, which encode the signal peptide for PenA, was cloned into the pGEX-6p-2 using the BamHI and XhoI restriction sites with a glutathione S-transferase gene at the 5Ј end of the gene. The bla PenI gene minus the first 90 nucleotides, which encode its signal peptide, was cloned into the pET24a(ϩ) plasmid using the NdeI and BamHI restriction sites. bla PenA and bla PenI genes cloned into pBC SK(ϩ) phagemid were expressed in Escherichia coli DH10B cells for susceptibility testing. The bla PenA gene (⌬1-81) cloned into the pGEX-6p-2 and the bla PenI gene (⌬1-90) cloned into the pET24a(ϩ) plasmid were expressed in E. coli Origami 2 DE3 cells for protein purification. The QuikChange XL site-directed mutagenesis kit (Stratagene) was used to perform site-directed mutagenesis of pBC SK(ϩ) bla PenA to generate the R220G substitution in PenA using the manufacturer's protocol.
Antibiotic Susceptibility-Mueller-Hinton agar dilution MICs, 2 according to the Clinical and Laboratory Standards Institute (43,44), were used to phenotypically characterize E. coli DH10B pBC SK(ϩ) with bla PenA and bla PenI , as described previously (45). Etest MIC assays were conducted using the manufacturer's instructions. Colonies, directly resuspended into sterile water equivalent to a 0.5 McFarland standard, were used to inoculate Mueller-Hinton agar plates. The IMI Etests (Biomerieux Diagnostics) were placed on each plate; the bacteria were grown at 37°C for 18 h, and MICs were read. Chemical structures of compounds tested are presented in Fig. 2, and the sources of these compounds were described previously (46).
␤-Lactamase Purification-The E. coli Origami 2 DE3 cells producing the PenA and PenI ␤-lactamases were grown and lysed as described previously (47). PenA with a cleavable N-terminal GST tag was purified from crude extracts using GSTrap FF column as described previously (48). Crude extracts from PenI pellets were used for preparative isoelectric focusing using a pH 3.5-10 gradient, as described previously (48,49).
Purity of the fractions from both preparations was determined by SDS-PAGE. Gels were stained with Coomassie Brilliant Blue R-250. Protein concentrations were determined by measuring absorbance at 280 nm and using the protein's extinction coefficient (⌬⑀, 32,555 M Ϫ1 cm Ϫ1 for PenA and ⌬⑀, 21,555 M Ϫ1 cm Ϫ1 for PenI at 280 nm), which were obtained using the ExPASy ProtParam tool. To maintain full activity, PenA and PenI were stored in 10 mM PBS with 25% glycerol at Ϫ20°C.
Kinetics-Steady-state kinetic parameters were determined using an Agilent 8453 diode array spectrophotometer (Santa Clara, CA), as described previously (45,50). Using Enzfitter (Biosoft Corp., Ferguson, MO), a nonlinear least square fit of the data (Henri-Michaelis-Menten equation) determined the kinetic parameters, V max and K m as shown in Equation 1, The ␤-lactamase inhibitors (CLAV, SUL, and TAZO) inactivate ␤-lactamases by interfering with the catalytic mechanism. For class A enzymes, CLAV, SUL, and TAZO are recognized as ␤-lactams. The apparent K m (K m, app ) value of the inhibitors for the enzyme was determined using a direct competition assay under steady-state conditions, as described previously (45,50).
Partition ratios (k cat /k inact ) at 15 min for PenA and PenI were obtained by incubating the enzymes with increasing concentrations of inhibitor at room temperature (45,50). The ratio of inhibitor to enzyme (I/E) necessary to inhibit the hydrolysis of nitrocefin by greater than 99% was determined.
Crystallization and Structure Refinement-PenA and PenI were crystallized by the vapor diffusion method using a 250-l reservoir with a 4-l hanging drop (2 l reservoir solution ϩ 2 l protein solution). For PenA, the well solution contained 25% polyethylene glycol of 8 kilodaltons (PEG8K), 0.2 M sodium chloride, 0.1 M sodium phosphate/citric acid at pH 4.2 and 15 mg/ml of PenA in 2 mM HEPES, pH 7.5. Two successful conditions were obtained for PenI, one at pH 7.5 and the other at pH 9.5. One well solution contained 25% PEG8K, 0.1 M ammonium sulfate, 0.1 M HEPES, pH 7.5, and 10 mg/ml protein in 2 mM HEPES, pH 7.5, and the second contained 25% PEG8K, 0.1 M N-cyclohexyl-2-aminoethanesulfonic acid, pH 9.5 (reservoir), and 10 mg/ml protein in 2 mM HEPES, pH 7.5. Crystals appeared in 1-2 weeks reaching sizes of 0.4 -0.8 mm. The crystals were cryoprotected by dipping them into a reservoir solution containing 20% glycerol, flush-cooled, and kept at 100 K with a nitrogen gas stream.
The 0.5°oscillation images were collected on a Q270 CCD detector with synchrotron radiation ( ϭ 1.00 Å at beamline NE3A of the Advanced Ring of the Photon Factory, Tsukuba, Japan). The HKL2000 program was used to index and scale x-ray intensities (51). For the PenI structure, the ␤ was close to 90°(89.996°as the HKL2000 output) in the C2 space group. Other bravais-lattice candidates, which were suggested by HKL2000, were also examined. Only in the case of the C Centered Monoclinic was a reasonable R merge value (0.04 versus Ͼ0.4) obtained.
Molecular replacement for PenI structure obtained at pH 9.5 was completed at 2.0 Å using CTX-M-9 ␤-lactamase (PDB entry 2P74) as a search model with the program Phaser with Phenix Auto-MR wizard (52). For molecular replacement of the other two structures, the working model of PenI at pH 9.5 structure was used as a search model. With the highest resolution data, anisotropic B-factors with SIMU restraint 0.025 were introduced. At this step, R factor and R free values were reduced from 0.176/0.205 to 0.136/0.170, from 0.201/0.218 to 0.159/ 0.186, and from 0.201/0.242 to 0.141/0.189 for PenI at pH 9.5, PenI at pH 7.5, and PenA, respectively. The validity of these steps in the refinement was verified with PARVATI, a program that analyzes the modeled anisotropy (53). Riding hydrogen atoms were later added. R factor and R free values were reduced from 0.126/0.167 to 0.115/0.154, from 0.144/0.171 to 0.134/ 0.159, and from 0.143/0.187 to 0.132/0.174 for PenI at pH 9.5, PenI at pH 7.5, and PenA, respectively.
The occupancies of the DWs of PenI structures were refined, because the electron densities in the 2F o Ϫ F c map were poor compared with the PenA structure. The occupancies of the amino acid residues with poor density with the 1.0 2F o Ϫ F c map were set as 0.6. The atoms with no electron density in the map were not included in the final model.
The results showed clear initial solutions and a reasonable molecular arrangement of each of the ␤-lactamases in an asymmetric unit. The model was then subjected to numerous cycles of conjugate gradient refinement and stepping to add higher resolution data with the program SHELX-97 followed by manual model building with the program COOT (54,55). See Table  1 for data collection and refinement statistics. The coordinates were deposited in the Protein Data Bank (PenA, PDB code 3W4Q; PenI, pH 7.5, PDB code 3W4P; and PenI, pH 9.5, PDB code 3W4O).
Molecular Modeling-The atomic coordinates of PenA and PenI, from the crystal structures reported here were used to construct Michaelis-Menten and acyl-enzyme complexes with IMI as described previously using Discovery Studio 3.1 (Accelrys, Inc. San Diego) molecular modeling software (45). IMI was constructed using Fragment Builder tools and was minimized using a Standard Dynamics Cascade protocol of Discovery Studio 3.1. IMI was automatically docked into the active site of PenA using the CDOCKER module of Discovery Studio 3.1 (56). This protocol uses a CHARMm-based molecular dynamics scheme to dock ligands into a receptor-binding site. For IMI docking into the PenI active site, LibDock algorithm was employed (57)(58)(59). LibDock is a high throughput algorithm for docking, which uses a set of pre-generated ligand conformations to dock ligands into a receptor-active site. From a maximum of 1000 pre-calculated IMI conformations, 100 hits were retained and scored (with a minimum LibDock score of 100). The best conformations were automatically aligned to polar and apolar active site hot spots, and the best scoring poses were reported. At this step, the hydrogen atoms were not maintained. To further optimize the docked poses (i.e. add hydrogens and prevent the clashes between the receptor and ligand), a CHARMm minimization step was used. For this step the Smart Minimization algorithm was employed (i.e. 1000 steps of steepest descent with a root mean square gradient tolerance of 3 Å, followed by Conjugate Gradient minimization, with a rootmean square deviation (r.m.s.d.)-minimization gradient of 0.001 Å). For the final minimization of the IMI conformations into the active site of PenI, a r.m.s.d. cutoff of 1 Å was chosen. The resulting conformations of both PenA and PenI complexes were analyzed; the most favorable positioning of the IMI was chosen, and the complexes between the enzyme and inhibitor created, as described previously (60).
To check the stability of the complexes, a 6-ps Molecular Dynamics Simulation (MDS) was conducted for all enzyme-IMI complexes, as described previously (46,50). During the heating/cooling, equilibration and production stages of MDS, a temperature of 300 K and a constant pressure were maintained. The long range electrostatics were treated with Particle Mesh Ewald and explicit solvation with Periodic Boundary Condition. The MDS for PenA-IMI complexes was run without any constraints. The first 4000 steps (4 ps) of MDS (heating/cooling and equilibration) of PenI with IMI were run using distance constraints to tether the carbonyl oxygen of IMI toward the oxyanion hole, which is made up of the amide backbone nitrogens of Ser-70 and Thr-237. For the production step of MDS, the system was free without any constraints.
Imipenem Hydrolysis Assay-Periplasmic extracts were prepared from E. coli DH10B pBC SK(ϩ), E. coli DH10B pBC SK(ϩ) bla PenA , and E. coli DH10B pBC SK(ϩ) bla PenA carrying the R220G substitution as described previously (47). Total protein concentrations were measured for periplasmic extracts using protein assay according to the manufacturer's instructions (Bio-Rad). Imipenem hydrolysis was monitored at 297 nm for 20 min using 100 M imipenem and 12 g of protein from the periplasmic extracts.

RESULTS AND DISCUSSION
Divergent Phenotypes, PenA Is a Carbapenemase and PenI Is an ESBL-One of the most significant threats in clinical medicine is resistance to carbapenems, as they are the "last line" of therapy for many types of infections (61,62). Thus, the identification of class A ␤-lactamases, like PenA, with carbapenemase activity is highly alarming (63)(64)(65)(66)(67). Similarly, our analysis of PenI reveals it possesses ESBL properties. Thus, Burkholderia spp. harbor intrinsic ␤-lactamases that confer multidrug resistance phenotypes.
In vivo susceptibility testing reveals that E. coli expressing bla PenA demonstrate resistance to all four classes of ␤-lactam antibiotics (i.e. AMP, THIN and TAX, AZT, and IMI) as well as all currently available ␤-lactam-␤-lactamase inhibitor combinations (i.e. AMP-CLAV, AMP-SUL, and PIP-TAZO) ( Table 2 and Fig. 2). On the contrary, E. coli-producing bla PenI shows lower levels of resistance to all tested compounds compared

Hydrolytic Mechanisms of Two ␤-Lactamases from Burkholderia
with cells carrying bla PenA , and it is only resistant to AMP, THIN, and to a lesser extent, TAZ, TAX, and AZT (Table 2). Steady-state kinetic analyses parallel susceptibility testing as PenA can hydrolyze AMP, THIN, TAX, AZT, and IMI with k cat /K m values of 3.6 Ϯ 0.4 s Ϫ1 , 3.1 Ϯ 0.3 s Ϫ1 , 0.48 Ϯ 0.05 s Ϫ1 , 0.16 Ϯ 0.02 s Ϫ1 , and 0.05 Ϯ 0.01 s Ϫ1 , respectively (Table 3). PenI demonstrates lower catalytic efficiencies for all ␤-lactams compared with PenA, but the highest k cat value (i.e. 9.0 Ϯ 0.9 s Ϫ1 ) is for TAX (Table 3). Interestingly, PenA and PenI display distinct substrate spectra.
The ␤-lactamase inhibitors CLAV, SUL, and TAZO were developed to circumvent the production of ␤-lactamases. Both PenA and PenI exhibit K m, app values in the low micromolar range (i.e. 0.5 Ϯ 0.1 to 4.1 Ϯ 0.5 M) to the ␤-lactamase inhibitors CLAV, SUL, and TAZO (Table 4). However, the partition ratios (k cat /k inact ) tested at 15 min differ greatly between the two enzymes. PenA possesses much larger k cat /k inact values of 2000, 1500, and 50 for CLAV, SUL, and TAZO, respectively, whereas the values for PenI are 5, 5, and 1.
The ability of PenA to hydrolyze inhibitors as well as all classes of ␤-lactams (i.e. penicillin, cephalosporins, monobac-tams, and carbapenems) is worrisome; currently, ␤-lactamase inhibitors are not available that can inhibit this enzyme.

Crystallography Reveals Key Differences between PenA and PenI That Explain Phenotype and Catalytic Properties-The
PenA crystal structure was obtained at a 1.2 Å resolution at pH 4.2, and the PenI structures were obtained at 1.05 Å resolution at pH 7.5 and at 1.18 Å resolution at pH 9.5. The PenA ␤-lactamase crystal was in the space group C2 with three molecules per asymmetric unit ( Table 1). The PenI ␤-lactamases at pH 7.5 and pH 9.5 were both crystallized in space group P2 1 with one molecule in the asymmetric unit (Table 1).
Despite differences in amino acid sequence, PenA and PenI, pH 7.5, have very similar overall three-dimensional structures (i.e. r.m.s.d. ϭ 0.59 Å for C␣ atoms between positions 32 and 289). To identify differences that may contribute to the contrasting kinetic properties of PenA and PenI, the 70 (Fig. 4). 70 SXXK 73 Motif-The 70 SXXK 73 motif contains the nucleophilic Ser-70 as well as Lys-73, which can serve as a general base to initiate ␤-lactam acylation and is also involved in the proton shuttle during ␤-lactam hydrolysis (Fig. 1). Fig. 5A shows that two conformations of Lys-73 are found in the structure of PenI obtained at pH 9.5. An overall decreased hydrolytic profile is observed in PenI compared with PenA, and the positioning of Lys-73 may be an important contributing factor. In conformation 1, which is identical to the conformation observed in PenA (Fig. 5A, cyan), the N of Lys-73 can form hydrogen bonds with the O␦1 of Asn-132 (distance 2.6 Å) and the O␥ of Ser-70 (distance 2.5 Å), and in conformation 2, the N of Lys-73 is within hydrogen bonding distance of the O⑀1 of Glu-166 (distance 2.6 Å) and the O␥ of Ser-70 (distance 2.9 Å). Not only are Lys-73 and Ser-70 important for ␤-lactam hydrolysis, but Glu-166 can   also serve as a general base important for acylation and deacylation of ␤-lactams (Fig. 1). As a consequence, both the interactions observed between Lys-73, Glu-166, and Ser-70 in conformation 2, as well as the ability of Lys-73 to have two conformations, may decrease substrate hydrolysis due to inhibiting the proper function of these residues (e.g. decreased ability of Lys-73 to abstract a proton from Ser-70, altered proton shuttling, and changes in the pK a of Glu-166). Unlike PenI, PenA possesses a Phe at position 72. The S72F substitution was implicated in promoting CLAV resistance in PenI by increasing the K i of the enzyme for CLAV (31,68). This hydrophobic amino acid may contribute to the CLAV resistance phenotype found in PenA by altering the overall geometry of the oxyanion hole.

130
SDN 132 Loop-The 130 SDN 132 loop contains Ser-130, which participates in the proton shuttle for ␤-lactam hydrolysis (Fig. 1) and is critical for inactivation of a ␤-lactamase by a ␤-lactamase inhibitor (69). Also present in the loop is Asn-132, which stabilizes the Michaelis complex. Two conformations of Ser-130 are found in PenI at pH 7.5. In conformation 1, the O␥ of Ser-130 is within hydrogen bonding distance of the N of Lys-73 (distance 3.0 Å), which is similar to the orientation observed in PenA (Fig. 5A, cyan). In conformation 2, the O␥ of Ser-130 can form hydrogen bonds with the N of Lys-234 (distance 2.7 Å); and an additional hydrogen bond can be observed between the O␥ of Ser-130 and a sulfate molecule (SO42010, distance 2.6 Å) trapped in the active site of PenI. Similar to Lys-73, flexibility of Ser-130 may impede hydrolysis of ␤-lactams by PenI, further supporting the lower -catalytic efficiencies for ␤-lactams observed with PenI compared with PenA (70). As with Lys-73, the decreased ability to shuttle protons by Ser-130 affects hydrolysis rates.

234
KTG 236 Motif and B3 ␤-Strand-The hydrogen bonding network between residues 236, 237, 220, and 276 was previously demonstrated to be critical for binding and hydrolysis of ␤-lactams and ␤-lactamase inhibitors by the carbapenemase, KPC-2 (46,47). In both PenA and PenI, the O␥1 of Thr-237 and the O of Gly-236 can form a hydrogen bonding network with the N2 and the N1 of Arg-220, respectively, in addition the N⑀ of Arg-220 is within hydrogen bonding distance of the O␦1 Asp-276.
102-110 Loop-The importance of 105 in ␤-lactam binding was previously shown in KPC-2 (50), and as a result attention was directed to the 102-110 loop. The crystal structure shows that the O of Tyr-105 of PenA can form a hydrogen bond with the hydroxyl side chain of Tyr-129 (distance 2.8 Å) and possesses a t rotamer orientation (trans, t80°); conversely, the Tyr-105 of PenI is in the m rotamer orientation (minus, m Ϫ85°/m Ϫ30°) (Fig. 5A) (71). Typically, the m rotamer is only observed in crystal structures of ␤-lactamases bound by an inhibitor and is hypothesized to stabilize hydrophobic ligand side chains (71). Molecular dynamics studies revealed that Tyr-105 is a dynamic residue, and for optimal ␤-lactam binding the t rotamer is nec- essary (71). The decreased catalytic activity of PenI compared with PenA may also be the result of the orientation of Tyr-105 obstructing the active site thus preventing the ␤-lactam from binding.
⍀ Loop and Deacylation Water-The ⍀ loop includes residues 164 -179 and contains two critical active-site residues, Glu-166 and Asn-170; both residues anchor the DW molecule required for ␤-lactam hydrolysis (Fig. 1). Hydrogen bonding interactions between DW molecules and the three structures are different. In the PenA structure, the DW molecule forms hydrogen bonds with the O⑀2 of Glu-166, the N␦2 of Asn-170, the N of Ser-70, and O␥ of Ser-70. Compared with PenA ( Fig.  5A), in PenI at pH 7.5 and pH 9.5, the side chains of Asn-170 and Glu-166 are rotated 180°, and thus the hydrogen bonding interactions with the DW are different (Fig. 5A). Specifically, in the pH 7.5 structure, the DW is not anchored appropriately to Asn-170; this suggests that the DW may be impaired in promoting deacylation of the ␤-lactam. In addition, the rotation of Glu-166 changes the hydrogen bonding partner from the O⑀2 of Glu-166 to the O⑀1 of Glu-166 with the DW. In the PenI structure at pH 9.5, the DW possesses additional hydrogen bonds to the N␦2 and O␦1 of Asn-170, which may decrease the rate of deacylation. In addition to the alterations in hydrogen bonding pattern of the DW in the three structures, the DW molecule is at a lower occupancy in PenI about 25% (21.7% at pH 7.5 and 27.4% at pH 9.5) compared with PenA, which is at 100% (Fig. 5, B and C). The lower DW occupancy may further explain why PenI demonstrates lower catalytic activity compared with PenA.
In class A ␤-lactamases, substitutions within the ⍀ loop can confer cephalosporin resistance (72,73). A substitution from Pro-167 to Thr results in TAZ resistance in the class A ESBL, CTX-M-23 (74). PenA possesses Thr-167; conversely, PenI has Pro-167. PenA demonstrates a TAZ resistance level one dilution higher than PenI.
Hydrogen bonding interactions between the ⍀ loop and B3 ␤-strand have been shown to decrease the flexibility of the B3 ␤-strand and narrow the spectrum of the ␤-lactamase (75). Specifically, Arg-241, which is near the B3 ␤-strand in the SHV-1 class A ␤-lactamase, was shown to be important for stabilizing the ⍀ loop; however, in PenA and PenI, a tyrosine is present at position 241. This change may allow the B3 ␤-strand to be flexible and thus be displaced from the ⍀ loop, increase the distance between the two, and allow larger cephalosporins (e.g. TAX) to be hydrolyzed.
PenA Shares Significant Structural Similarities with KPC-2-PenA was compared with KPC-2 as this ␤-lactamase is the most prevalent and clinically important class A carbapenemase (76). r.m.s.d. for PenA to KPC-2 is 1.0 Å. KPC-2 is similar to PenA in the five major motifs and demonstrates similar differences from PenI as observed in the PenA-PenI comparison. A major difference is the lack of a disulfide bond in PenA and PenI that is present between Cys-69 and Cys-238 in KPC-2 as well as other class A carbapenemases. Notably, KPC-2 contains the 237, 220, and 276 hydrogen bonding network, which is necessary for binding and hydrolysis of ␤-lactams as well as CLAV resistance (46). The occupancy of DW in KPC-2 is at 100% and is anchored between Glu-166 and Asn-170. The phenotype of PenA, being so comparable to that of KPC-2, may be attributed to their similarities.
Molecular Modeling of PenA and PenI with IMI, Insights into Carbapenemase Activity-Carbapenems are the last line of therapy for many types of infections (61,62). PenA hydrolyzes carbapenems with a k cat of 0.38 Ϯ 0.4 s Ϫ1 , although carbapenem hydrolysis by PenI was not detectable. In addition, E. coli DH10B carrying bla PenA is resistant to imipenem (MIC 4 mg/li- Water molecules (not shown) are present near Arg-220; thus IMI is in a favorable orientation for tautomerization of its C 2 -C 3 double bond for generation of the more stable ⌬ 1 isoform. ter), whereas E. coli DH10B-producing bla PenI is susceptible. As a result, PenA and PenI were modeled with the carbapenem, IMI, to gain insight into why PenA is a carbapenemase and PenI is not. Michaelis-Menten complexes of IMI with PenA reveal that IMI is primed for acylation with PenA (Fig. 6A). The carbonyl of IMI is oriented toward the backbone amides of Ser-70 and Thr-237, which is the oxyanion hole or electrophilic center of ␤-lactamases. In contrast, the positioning of Tyr-105 hindered the docking of IMI into the active site of PenI. Docked complexes were eventually achieved, and the carbonyl of IMI is positioned toward the oxyanion hole (Fig. 6B).
To gain further insights into the mechanism, the acyl-enzyme complexes of PenA and PenI with IMI were constructed with the ⌬ 2 isoform of IMI. Upon acylation, carbapenems can form two isoforms (e.g. ⌬ 2 3 ⌬ 1 ) due to tautomerization of the pyrroline-double bond from C 2 -C 3 to C 3 -C 4 (Fig. 2). In TEM-1, a class A enzyme, deacylation of the carbapenem proceeds more rapidly with the ⌬ 2 isoform; the ⌬ 1 isoform deacylates slowly (77,78). An arginine residue at 244 in the TEM-1 ␤-lactamase coordinates a water molecule that is the source of the proton for this tautomerization (79); the Arg-244 equivalent in PenA and PenI is Arg-220.
In the molecular representation of PenA with the ⌬ 2 isoform of IMI, the carbonyl of IMI is docked in the oxyanion hole (2.7 Å from the N of Ser-70 and 2.9 Å from the N of Thr-237) (Fig. 6C). The O4 of the IMI carboxylate forms a hydrogen bonding interaction with the N1 of Arg-220 (distance 2.9 Å); no waters are found near (within 4 Å) Arg-220 to promote tautomerization. Thus, the IMI is oriented in a position for deacylation.
To confirm this model, we substituted Arg-220 to Gly; this resulted in a loss of IMI resistance by PenA and a significant decrease in catalytic activity when measured enzymatically (Fig. 7,  A and B). Therefore, Arg-220 is critical for interacting with IMI.
In contrast, in the model of PenI with the ⌬ 2 isoform of IMI, the carbonyl is not in an oxyanion hole (Fig. 6D). The Arg-220 can form a hydrogen bond with the O4 of the IMI carboxylate, and water molecules are within hydrogen bonding distance of Arg-220, which may influence the tautomerization of IMI in PenI to the unhydrolyzable and stable ⌬ 1 isoform.
The carbapenemase phenotype of the PenA enzymes compared with PenI may also be attributed to the positioning of Tyr-105, thus allowing IMI to enter the active site. In addition, the ⌬ 2 isoform of IMI forms interactions that are favorable for deacylation; the carbonyl of IMI remains in the oxyanion hole during the 6-ps molecular dynamics simulation, unlike with PenI. Finally, the water molecules necessary for tautomerization of the IMI to the more stable ⌬ 1 isoform are not near Arg-220 in PenA but are present in PenI.
Conclusions-B. multivorans is an opportunistic nosocomial pathogen primarily infecting patients with cystic fibrosis, and B. pseudomallei is the causative agent of melioidosis and possesses the potential to be used as a bioweapon. Both organisms are resistant to ␤-lactam antibiotics, and this phenotype is linked to the expression of a class A ␤-lactamase, PenA in B. multivorans, and PenI in B. pseudomallei.
Here, the crystal structures of two phylogenetically similar yet phenotypically distinct ␤-lactamases from Burkholderia were analyzed to dissect their different mechanisms. Kinetic and crystallographic analysis of both enzymes revealed key differences that may directly lead to their unique phenotypes. Alterations in the positioning of important catalytic residues (e.g. Lys-73, Ser-130, and Tyr-105) may directly contribute to PenI's lower catalytic efficiencies and k cat /k inact values toward ␤-lactams and ␤-lactamase inhibitors, respectively, compared with PenA. In addition, altered anchoring of and decreased occupancy of the DW molecule is most likely another consideration of importance. On a different note, resistance to ␤lactamase inhibitors by PenA may be the result of the presence of Phe at 72, as well as the topology of the active site; PenA demonstrates strong similarity to KPC-2, an inhibitor-resistant carbapenemase.
Distressingly, these studies highlight the tremendous hurdles that appear before the medicinal chemist and clinician facing the challenge of designing drugs to treat multidrug resistance Gram-negative pathogens. For the former, designing novel ␤-lactams that can be effective against a pathogen harboring two different ␤-lactamases that possess contrasting properties is a complex endeavor. To the latter, the correct combination of drugs that will affect a durable cure must be chosen.