Structural Insights into the Inhibition of the Extended-Spectrum β-Lactamase PER-2 by Avibactam

The diazabicyclooctane (DBO) avibactam (AVI) reversibly inactivates most serine-β-lactamases.

ABSTRACT

The diazabicyclooctane (DBO) avibactam (AVI) reversibly inactivates most serine-β-lactamases. Previous investigations showed that inhibition constants of AVI toward class A PER-2 are reminiscent of values observed for class C and D β-lactamases (i.e., k₂/K of ≈10³ M⁻¹ s⁻¹) but lower than other class A β-lactamases (i.e., k₂/K = 10⁴ to 10⁵ M⁻¹ s⁻¹). Herein, biochemical and structural studies were conducted with PER-2 and AVI to explore these differences. Furthermore, biochemical studies on Arg220 and Thr237 variants with AVI were conducted to gain deeper insight into the mechanism of PER-2 inactivation. The main biochemical and structural observations revealed the following: (i) both amino-acid substitutions in Arg220 and the rich hydrophobic content in the active site hinder the binding of catalytic waters and acylation, impairing AVI inhibition; (ii) movement of Ser130 upon binding of AVI favors the formation of a hydrogen bond with the sulfate group of AVI; and (iii) the Thr237Ala substitution alters the AVI inhibition constants. The acylation constant (k₂/K) of PER-2 by AVI is primarily influenced by stabilizing hydrogen bonds involving AVI and important residues such as Thr237 and Arg220. (Variants in Arg220 demonstrate a dramatic reduction in k₂/K.) We also observed that displacement of Ser130 side chain impairs AVI acylation, an observation not made in other extended-spectrum β-lactamases (ESBLs). Comparatively, relebactam combined with a β-lactam is more potent against Escherichia coli producing PER-2 variants than β-lactam–AVI combinations. Our findings provide a rationale for evaluating the utility of the currently available DBO inhibitors against unique ESBLs like PER-2 and anticipate the effectiveness of these inhibitors toward variants that may eventually be selected upon AVI usage.

INTRODUCTION

PER-1 and PER-2 are the most frequently reported members of this class A PER family of β-lactamases (1). These class A extended-spectrum β-lactamases (ESBLs) readily hydrolyze oxyimino-cephalosporins, with high catalytic efficiencies (k_cat/K_m) for both cefotaxime and ceftazidime, and are also efficiently inhibited by mechanism-based inhibitors such as clavulanate and tazobactam (2, 3).

PER-2, sharing 86% amino acid sequence identity with PER-1, is the second most frequent ESBL (after the pandemic CTX-Ms) in Argentina, accounting for nearly 10% and 5% of the oxyimino-cephalosporin resistance in Klebsiella pneumoniae and Escherichia coli, respectively (1, 4, 5).

Recently, the crystallographic structure of PER-2 at a 2.2-Å resolution revealed the presence of a unique fold in the Ω-loop of PER-2. This results in an enlarged and potentially mobile active site cavity, unlike in other class A β-lactamases, which allows for the more efficient hydrolysis of β-lactams, like the oxyimino-cephalosporins (6). Additionally, a hydrogen bond network connecting Ser70-Gln69-water-Thr237-Arg220 is observed, which is postulated to be important for the enhanced activity and inhibition of this β-lactamase (6). Generating single amino acid substitutions at positions Arg220 and, to a lesser extent, Thr237 in PER-2 decreases the potency of clavulanic acid and tazobactam, as was also reported for KPC-2 (7, 8).

Avibactam (AVI), a novel bridged diazabicyclooctane (DBO) non-β-lactam β-lactamase inhibitor (9), reversibly inactivates most Ambler class A and C β-lactamases, restoring susceptibility to certain β-lactams when used in combination. One of the most attractive features of AVI, besides its improved affinity, is that upon deacylation, recyclization occurs via ring closure (i.e., regenerating the intact AVI) and may acylate other β-lactamases (10–13). Despite these differences from other mechanism-based inhibitors like clavulanic acid and tazobactam, AVI shares some mechanistic similarities with them, as it enables acylation of Ser70 and accommodation of the carbonyl oxygen in the oxyanion hole for stabilization of the transition state (11).

An analysis of the ability of AVI to inactivate PER-2 was recently completed (14). Interestingly, the inhibition constants (i.e., k₂/K = 2 ± 0.1 × 10³ M⁻¹ s⁻¹) that were observed when testing AVI were reminiscent of values observed for class C and D β-lactamases (i.e., k₂/K range of ≈10³ M⁻¹ s⁻¹), but lower than those from class A β-lactamases (i.e., k₂/K range of 10⁴ to 10⁵ M⁻¹ s⁻¹) (9, 10, 14). Once AVI is bound to PER-2 β-lactamase, AVI forms a stable complex with PER-2, as observed via mass spectrometry (e.g., from 31,389 ± 3 amu to 31,604 ± 3 amu for 24 h). This analysis led to the hypothesis that the hydrophobic nature of the PER-2 active site was responsible in part for the slower acylation rate of AVI (14). In silico molecular modeling suggested key interactions between AVI’s sulfate and residues Arg220 and Thr237 in PER-2 (14). To gain a deeper insight into the mechanism of inactivation of PER-2 by AVI, we conducted structural and biochemical studies with wild-type PER-2 and Arg220 and Thr237 variants of PER-2.

RESULTS AND DISCUSSION

Overall structure of apo-PER-2 β-lactamase and its complex with AVI.

The structure of the apo form of PER-2 was refined at 2.69 Å (PDB 6DGU), and the covalent complex of PER-2 with AVI was refined at 2.40 Å (PDB 6D3G). The main data and refinement statistics are given in Table 1. Both refined structures consisted of four monomers (A to D) per asymmetric unit, and the electron density map is well defined along the main chains of both structures (Fig. 1). For the apo-PER-2, monomers A to D include 274, 277, 275, and 272 residues, respectively. The compositions of monomers A to D in the PER-2:AVI complex comprise 279 amino acids for monomers A and B and 281 and 275 residues for monomers C and D, respectively. The structures are solvated by 183 (apo-PER-2) and 133 (PER-2:AVI) ordered water molecules.

TABLE 1.

X-ray data collection and refinement statistics for apo-PER-2 and the PER-2:AVI complex

Crystal parameter	Value fora :
	Apo-PER-2	PER-2:AVI
PDB code	6DGU	6D3G
Data collection
No. of frames	300	1,800
Oscillation step (o)	0.5	0.1
Detector distance (mm)	75.00	280.06
Wavelength (Å)	1.5418	0.9801
Exposure per frame (s)	150	0.025
Space group	P212121	P212121
Unit cell dimensions
a, b, c (Å)	82.37, 82.58, 174.92	84.16, 85.84, 161.13
α = β = γ (°)	90	90
No. of subunits/asymmetric unit	4	4
Resolution range (Å)	41.29–2.69 (2.85–2.69)	48.17–2.40 (2.54–2.40)
No. of reflections:
Total	192,909	312,245
Unique	32,951 (4,944)	46,410 (7,289)
Redundancy	5.5 (5.8)	6.6 (6.7)
Completeness (%)	97.3 (91.4)	99.8 (98.8)
Mean I/σ⟨I⟩	9.7 (3.6)	13.7 (1.6)
Overall Wilson B factor (Å2)	27	56
Rmerge	0.164 (0.450)	0.085 (0.923)
Rmeas	0.180 (0.500)	0.092 (1.002)
CC(1/2)	0.987 (0.780)	0.999 (0.746)
Refinement
No. of reflections used in refinement	32,951	46,399
No. in Rfree test set; %	1,980; 6	2,321; 5
Rwork	0.271	0.206
Rfree	0.337	0.253
No. of:
Non-hydrogen atoms	8,616	8,706
Macromolecules	8,494	8,551
Ligands		81
Solvent	113	74
RMSDb from ideal stereochemistry
Bonds (Å)	0.010	0.013
Angles (o)	1.220	1.550
Avg B factor (Å2):
All atoms	30	64
Protein	30	64
Ligand		61
Solvent	19	50
Ramachandran plot (%):
Favored regions	92.2	96.8
Allowed regions	7.0	2.7
Outlier regions	0.8	0.5

^aStatistics for the highest-resolution shell are given in parentheses.

^bRMSD, root mean square deviation.

FIG 1.

Overall structure of the asymmetric unit of the P2₁2₁2₁ crystal of the complex at 2.4 Å (PDB code 6D3G) with its four protein monomers (A to D). The detailed structure of the AVI moieties in the four monomers and their corresponding |F_obs| − |F_calc| omit maps contoured at 2.5 σ (green mesh) are shown in the corner of each monomer. (Inset) 2|F_obs| − |F_calc| map contoured at 1 σ (gray mesh) around AVI (green backbone) and important residues (light pink backbone). Color references: oxygen, red; nitrogen, blue; sulfur, green. α helices are shown in orange, and β-sheets are shown in red.

The average root mean square deviation (RMSD) values for Cα atoms between the main chains of PER-2 and its complex with AVI is 0.58 Å, while the average RMSD values between monomers within the same structures are 0.44 Å for apo-PER-2 and 0.54 Å for the PER-2:AVI complex. Due to this observation, the Results and Discussion sections will focus on monomer A of each structure unless otherwise noted.

Inspection of the overall PER-2:AVI complex revealed that AVI is covalently bound to the Ser70-O^γ atom in the active site with a well-defined |F_obs| − |F_calc| omit map contoured at 2.5 σ for all AVI moieties (Fig. 1). AVI’s 6-member ring adopts a chair conformation with hydrogen bonds via the C-7 carbonyl of the newly formed carbamate with backbone nitrogen atoms of Ser70 and Thr237 in the oxyanion hole (Fig. 1 and 2A). Additional hydrogen-bonding interactions between the AVI amide oxygen atom and Asn132, via the AVI sulfate moiety with the γ-oxygen atom of Ser70, Ser130, Thr235, and Thr237, and with the backbone nitrogen atom of Thr237 were observed. A potential long-range hydrogen bond electrostatic interaction (3.6 Å) could also occur between the AVI sulfate group and Lys234.

FIG 2.

(A) Stereo view of the active site of PER-2 in complex with AVI (aquamarine), showing the favored hydrogen bonds (black dashed lines) involved in the stabilization of the inhibitor with conserved residues (orange) in the active site. (B) Stereo image showing the superposition of the apo-PER-2 (pink) versus the PER-2:AVI complex (orange), showing the hydrogen bonds formed upon AVI binding (green dashed lines).

Differences in the interaction of AVI with PER-2 and other class A β-lactamases.

To address the difference in acylation rates between PER-2 and other class A β-lactamases (i.e., CTX-M-15, KPC-2, and SHV-1 β-lactamases), the PER-2:AVI structure was compared to other class A β-lactamase–AVI crystal structures. After superimposing each of the four monomers of the PER-2:AVI complex and the apo form of PER-2, the side chain of Lys73 is closer to Ser130 (and more distant from Glu166), in the absence of ligand in some cases, in contrast to what is observed in other β-lactamases like CTX-M-15 (see below) (15). The difference in Lys73 conformations could potentially impact the acylation rate by AVI. However, due to the relatively high variability in the rotameric positions of the Lys73 side chain, we are unable to assess if this observation could be associated with the inhibition mechanism. After the binding of the ligand, the side chain of Lys73 is oriented toward Ser130 in the four monomers (not shown).

In all possible superpositions of PER-2:AVI versus apo-PER-2, we observed important differences that may impact activity. Upon binding of AVI, an up to 1.2-Å shift in the Ser130 side chain occurs in favor of the formation of the hydrogen bond between Ser130O^γ and the AVI sulfate moiety mentioned above (Fig. 2B).

Interestingly, the side chain of Trp105 shifts its rotameric position upon binding of AVI, probably favored through van der Waals interactions with its 6-member ring (Fig. 2B). Trp105 likely plays an important structural role in stabilization of the complex. The shift of Trp105 is accompanied by an ∼4-Å displacement of the Thr104 side chain, creating new stabilizing hydrogen bonds with His170 (Fig. 2B), which is present in the inverted Ω-loop of PER-2. As a result, a new hydrogen bond network is now formed between Lys73, Thr104, Asn132, Glu166, and His170, serving to stabilize the PER-2:AVI complex; this rearrangement is unprecedented compared to other class A β-lactamases. Other hydrophobic interactions with Gln69, Gly236, and Thr237 are possible; in addition, the benzyl moiety of Phe72 turns 90° upon AVI binding in some monomers, therefore creating an additional hydrophobic environment, reinforcing our previous hypothesis (14).

Rearrangement of residues could also involve Arg220, a key residue for which an essential role was previously attributed within the second coordination sphere of PER-2 (8). Arg220 participates in more hydrogen bond interactions when AVI is linked to Ser70: i.e., between the guanidinium moiety of the Arg220 and carbonyl groups of Gly236 and Asn245 and the side chain of Thr237 (data not shown).

As other complexes of class A β-lactamases demonstrate, AVI is covalently linked in a similar fashion as in KPC-2 (PDB code 4ZBE) and TLA-3 (PDB code 5GWA) (Fig. 3A). In concordance with TLA-3, PER-2 shares similar folds and overall tertiary structures, explaining why AVI is stabilized within the active site. In contrast, the deacylating water was not observed in our structure, probably because of lack of crystallographic resolution or, as we hypothesized above, because the more hydrophobic environment in PER-2 may prevent it from binding (14).

FIG 3.

(A) Superposed view of the covalent interaction of AVI moiety in the active site of PER-2, KPC-2 (PDB code 4ZBE), and TLA-3 (PDB code 5GWA). Color codes for amino acid residues and AVI moiety, respectively: orange and aquamarine for PER-2, light green and salmon for KPC-2, and light violet and gold for TLA-3. Deacylating water molecules are only observed in KPC-2 (violet sphere) and TLA-3 (dark orange sphere). (B) Comparison of the active sites of superposed structures of PER-2, CTX-M-15 (PDB code 4S2I), and SHV-1 (PDB code 4ZAM). Color codes (amino acids and AVI molecules, respectively): orange and aquamarine for PER-2, cyan and lime green for CTX-M-15, and pink and chocolate for SHV-1. Deacylating waters for CTX-M-15 and SHV-1 are shown as a green sphere and magenta sphere, respectively.

The catalytic site of PER-2 is significantly more hydrophobic than those of CTX-M-15 (PDB code 4S2I) and SHV-1 (PDB code 4ZAM) (Fig. 3B and 4). Previous Michaelis models have shown that hydrophobic patches near Ser70, as well as a possible displacement of catalytic water in PER-2, could contribute to the different k₂/K values, which for PER-2 (2,200 M⁻¹ s⁻¹), are 1 and 2 orders of magnitude lower than for SHV-1 and CTX-M-15, respectively (60,000 and 130,000 M⁻¹ s⁻¹) (14). The hydrophobic environment in the active site of PER-2 may play a relevant role in the lower inhibition parameters observed for AVI. However, recognizing that PER-2 is more hydrophobic than TLA-3, a displacement of the deacylating water in PER-2:AVI seems unlikely because the inhibition parameters k₂/K and k_off for both PER-2 and TLA-3 are very similar (2,200 M⁻¹ s⁻¹ and 0.0004 s⁻¹ versus 3,250 M⁻¹ s⁻¹ and 0.00096 s⁻¹, respectively). In turn, the apparent K_i (K_{i app}) for TLA-3 (1.71 μM) is 12-fold lower than that for PER-2 (16), which is also greatly increased in Arg220 variants.

FIG 4.

Molecular surface representation of PER-2, colored according to hydrophobic content (red, highly hydrophobic; white, highly hydrophilic). AVI covalently linked in the active site is shown as sticks.

Another noteworthy difference between PER-2 and other class A enzymes is the positioning of the N-6–O bond that determines the way in which AVI interacts with Ser130 (Fig. 5). In CTX-M-15, there is a closer proximity between the N-6 and C-7 atoms of AVI compared to PER-2 (2.9 versus 3.6 Å). More importantly, interaction of the AVI sulfate moiety with Ser130O^γ (2.6 Å) is more favorable compared to the AVI N-6–Ser130O^γ (3.2 Å) interaction in PER-2; in CTX-M-15 and other β-lactamases with higher values of k₂/K, association of AVI with Ser130O^γ is more favorable through its N-6 atom instead of the sulfate group (2.9 versus 3.5 Å in CTX-M-15, respectively).

FIG 5.

Detailed positioning of the N-6–O bond (red circle) of the AVI molecule in PER-2 compared to KPC-2 and CTX-M-15, which determines the way in which AVI interacts with Ser130 (see the text for details). Color codes are the same described in the legends to Fig. 2 and 3: orange and aquamarine for PER-2:AVI, light green and salmon for KPC-2:AVI, and cyan and lime green for CTX-M-15:AVI.

Finally, an important difference between PER-2 and SHV-1 is the absence of a direct link between the sulfate moiety of AVI and Arg220 in PER-2, which is fulfilled in SHV-1 via a salt bridge interaction with Arg244 (whose side chain occupies an equivalent space to Arg220) that, together with the Thr237Ala replacement in SHV-1, could force a shift in the location of AVI in the active site (11).

Although the resolution of our structures limits providing precise details concerning the mechanism of inhibition by AVI, some general observations can be made. Overall, our structures reveal that Ser70 is primed by deprotonation for nucleophilic attack on the AVI C-7 atom. It is still uncertain if this deprotonation is driven by Lys73 or Glu166 via a catalytic water that bridges both Glu166 and Ser70 (17–22). Using both high-resolution structural studies and molecular dynamics simulations, both Lys73 and Ser130 are also believed to participate in a proton shuttle resulting in the protonation of the AVI N-6 in the acylation step (15, 23). In our structures, Lys73 is observed in close proximity to Ser130; this strategic positioning could support a previously assigned role of Lys73 in the acylation mechanism. Following this model, it is also likely that Ser130 is involved in deacylation, probably activated by Lys73, as in other class A β-lactamases (24, 25).

Impact of substitutions at Arg220 and Thr237 on AVI inhibition.

To further examine the roles of Ambler positions 220 and 237, E. coli clones expressing PER-2 and select Arg220 and Thr237 variants, obtained as described before (8), were subjected to broth microdilution susceptibility testing using AVI (at 4 μg/ml) combined with either ceftazidime (CAZ), ceftaroline (TAR), or aztreonam (AZT); for comparison, the MICs of the same drugs combined with 4 μg/ml relebactam (REL) were also obtained.

Only AVI combined with TAR lowers the MIC of wild-type PER-2 expressed in E. coli Top10F′ cells; however, the MIC was 1 μg/ml (Table 2). AVI was able to restore the MIC values to the susceptible range when combined with either TAR or AZT for two and four of the Arg220 variants, respectively. In contrast, the CAZ-AVI combination was completely unable to lower MICs sufficiently. In a previous study, clinical Enterobacteriaceae isolates producing PER-2 were tested and revealed that the TAR-AVI was the most potent combination (14); these differences in the MICs from clinical isolates compared to these for E. coli recombinant clones could be explained by acknowledging that in recombinant clones, PER-2 variants are expressed from the high-copy-number plasmid pK19. However, given the results of this analysis, concerns are raised that β-lactam–AVI combinations might fail to treat an infection caused by an Enterobacteriaceae isolate producing any of these Arg220 variants of PER-2.

TABLE 2.

MIC values of different combinations containing AVI and REL for E. coli recombinant clones expressing wild-type PER-2 and derived Arg220/Thr237 mutants

E. coli Top10F′ derivative	MIC (μg ml−1)
	CAZ	CAZ- AVI	CAZ- REL	TAR	TAR- AVI	TAR- REL	AZT	AZT- AVI	AZT- REL
No plasmid	0.25	0.25	0.5	0.06	0.06	0.06	0.5	0.25	0.12
pK19	0.5	0.5	1	0.12	0.12	0.12	0.5	0.5	0.5
pK19/blaPER-2 (WT)	>4,096	128	256	256	1	0.5	>4,096	128	128
pK19/blaPER-2 Arg220Leu	128	32	4	128	2	0.25	128	4	1
pK19/blaPER-2 Arg220Ser	256	64	8	64	2	0.25	128	16	1
pK19/blaPER-2 Arg220Gly	256	32	4	128	2	0.5	256	8	1
pK19/blaPER-2 Arg220His	256	32	4	32	2	0.25	32	4	0.5
pK19/blaPER-2 Arg220Thr	128	32	2	32	1	0.12	32	2	0.5
pK19/blaPER-2 Arg220Cys	128	16	4	32	0.5	0.25	64	2	0.5
pK19/blaPER-2 Thr237Ala	512	16	4	512	0.5	0.25	>512	8	1

In comparison, combinations with REL showed that this DBO inhibitor seems to be much more potent than AVI, especially toward the E. coli clones producing mutant PER-2 (Table 2); the inhibitory potential of REL was most striking with TAR as a partner β-lactam, followed by CAZ and AZT, with MIC values decreasing up to 8 dilutions compared to the β-lactam alone. REL was recently shown to be a potent inhibitor of the KPC-2 carbapenemase, restoring the susceptibility of imipenem in Enterobacteriaceae (26). Our findings suggest that REL could serve as a better inhibitor than AVI if Arg220 variants of PER-2 eventually emerge, although the efficacy of both inhibitors is not as high as for other prevalent enzymes such as KPC-2.

Wild-type PER-2 and the Arg220 and Thr237 variants were expressed and purified as previously described (8), and the steady-state kinetic parameters were determined. As shown in Table 3, single amino acid substitutions in Arg220 resulted in up to 67-fold higher K_{i app} values for AVI compared to the wild-type enzyme. This behavior was accompanied by an up to ∼40-fold decrease in the inhibition efficiency, k₂/K. In both cases, the variants containing leucine, histidine, threonine, or cysteine residues at position 220 were the most affected. Remarkably, except for the Arg220Gly mutant, off-rate values increase between 3- and 112-fold compared to the wild-type PER-2, suggesting that the loss in affinity for AVI could be countervailed by a faster deacylation in both Arg220 and Thr237 mutants. The behavior of the Arg220Gly mutant is noteworthy in the sense that only the k₂/K parameter is strongly impaired.

TABLE 3.

Steady-state and inhibition parameters for PER-2 and the Arg220/Thr237 mutants with NCF and AVI

Kinetic parametera	PER-2 wild typeb	PER-2 R220G	PER-2 R220S	PER-2 R220H	PER-2 R220T	PER-2 R220C	PER-2 R220L	PER-2 T237A
NCF
Km (μM)	32 ± 1	310 ± 34	39 ± 2	141 ± 13	169 ± 19	76 ± 5	96 ± 12	17 ± 1
kcat (s−1)	159 ± 3	274 ± 16	92 ± 1	77 ± 3	155 ± 7	76 ± 1	71 ± 3	289 ± 3
kcat/Km (μM−1 s−1)	5 ± 0.3	0.9 ± 0.1	2.4 ± 0.1	0.55 ± 0.07	0.9 ± 0.1	1 ± 0.1	0.7 ± 0.1	17 ± 1
AVI
Ki app (μM)	20 ± 3	92 ± 24	316 ± 50	720 ± 147	509 ± 129	642 ± 161	1,343 ± 662	12 ± 2
k2/K (M−1 s−1)	2,200 ± 100	352 ± 44	246 ± 18	167 ± 21	76 ± 11	120 ± 13	56 ± 8	2,600 ± 234
koff (s−1)	0.0004	0.0005	0.012	0.045	0.0015	0.011	0.0011	0.006

^aNCF, nitrocefin; AVI, avibactam.

^bValues from reference 14.

In turn, the Thr237Ala variant was inhibited by AVI with slightly improved values compared to the wild type, albeit not at the level of SHV-1 (27). Thus, single amino acid substitutions at position 237 in PER-2 are less likely to impact inhibition by DBOs. Conversely, older β-lactamase inhibitors like clavulanate and tazobactam possess lower inhibition efficiencies for Thr237 variants compared to wild-type PER-2 (8). Although Arg220 variants showed a substantial decrease in k₂/K values, the effect on inhibition was less noticeable than those reported values for KPC-2 with single amino acid substitutions at Ser130 (28).

Importantly, the Ser130 variants of KPC-2 are slow to hydrolyze β-lactams; conversely, the PER-2 Arg220 variants maintain robust activity against β-lactams. Each residue possesses a different role in both hydrolytic as well as inhibition mechanisms.

Conclusions.

Insights into the structures of PER-2 in complex with AVI reveal distinctive interactions of AVI with residues in the active site that impact the efficiency of inhibition and partially explain the difference in acylation compared to other class A β-lactamases. As we have noted, PER-2 is not as efficiently inhibited by AVI as other class A β-lactamases like CTX-M-15, KPC-2, and SHV-1. The main biochemical and structural observations generated in this study are as follows: (i) both amino acid substitutions in Arg220 and the rich hydrophobic content in the active site of PER-2 hinder the binding of catalytic waters and acylation, greatly impairing AVI inhibition; (ii) movement of Ser130 upon binding of AVI favors the formation of a hydrogen bond with the AVI sulfate moiety at the expense of losing interactions with the AVI N6 atom; and (iii) the Thr237Ala substitution slightly improves the AVI inhibition constants.

The acylation rate, k₂/K, of PER-2 by AVI seems to be influenced by at least two factors: (i) Arg220 through stabilizing hydrogen bonds with the AVI moiety (variants in this residue demonstrate a dramatic reduction in k₂/K values) and (ii) Ser130 side-chain positioning, not observed in other class A β-lactamases, whose displacement impairs AVI acylation. The latter is supported by the observation that Ser130 is proposed as the general base to draw the proton from the N-6 atom as well as aid in rebuilding the N-6–C-7 bond in AVI recycling (24). Among β-lactamases showing higher k₂/K values (i.e., CTX-M-15 and KPC-2), the interaction of Ser130 with AVI N-6 is favored, while in PER-2, with lower k₂/K values, the preferred interaction seems to occur between Ser130 and the AVI sulfate moiety. We therefore conclude that positioning of the N-O bond of AVI moiety is critical to the inactivation mechanism in PER-2. In spite of limitations in the resolution detail of our structures, it is also likely that Lys73 participates in the proton shuttle via Ser130 to the AVI N-6 atom in the acylation steps. New higher-resolution structures are being attempted to address this.

In closing, our comparative analysis demonstrates that AVI is not an efficient inhibitor against PER-2 like other class A β-lactamases such as CTX-M-15 or KPC-2. Additionally, select variants at position Arg220 also show that the inhibitory activity of AVI could be greatly compromised if these variants are selected in vivo during treatment. The latter observation is supported by both the MIC values when combined with CAZ (the only available combination in Argentina) and biochemical analysis. Moreover, these findings are consistent with and support analyses previously performed (7). Interestingly, REL seems to be a more potent inhibitor for these variants when paired with the three partner β-lactams. The activity of REL with PER β-lactamases is under investigation.

Lastly, our findings provide a biochemical and structural basis for evaluating the utility of the currently available DBO inhibitors against unique ESBLs like PER-2 and anticipate the effectiveness of these inhibitors toward variants that we fear may be selected upon increased CAZ/AVI usage in the clinical arena. Our results provide a clearer understanding of the role of important residues (i.e., Arg220 and Thr237) in the catalytic and inhibition mechanism and explore the future impact these findings may have on clinical use.

MATERIALS AND METHODS

Bacterial strains and plasmids.

The recombinant E. coli clone harboring the pET/bla_PER-2 plasmid and recombinant E. coli clones producing different PER-2 Arg220 variants were obtained in a previous study (8).

Compounds.

Avibactam (AVI) was obtained through an investigator-initiated trial with Allergan. Nitrocefin (NCF) was purchased from Becton Dickinson and Oxoid. Relebactam (REL) was obtained from Advanced Chem Blocks. Chemical structures of AVI and REL are provided in Fig. 6.

FIG 6.

Chemical structures of avibactam (A) and relebactam (B).

Antimicrobial susceptibility test.

MICs of CAZ, TAR, and AZT alone and the combination of these drugs with 4 μg/ml AVI or REL were obtained by the broth microdilution method for recombinant E. coli Top10F′ clones expressing both the wild-type PER-2 and derived Arg220 or Thr237 mutants, following the Clinical and Laboratory Standards Institute (CLSI) guidelines (29).

Expression and purification of wild-type PER-2 and derived mutants.

Wild-type and Arg220 mutant PER-2 enzymes were expressed and purified as previously described (6, 8). Briefly, cultures of recombinant E. coli clones grown in lysogeny broth (LB) supplemented with 30 μg/ml kanamycin were induced for β-lactamase expression with 1 mM IPTG (isopropyl-β-d-thiogalactopyranoside) at 18°C overnight. After mechanical disruption of bacterial cells by sonication, the β-lactamase was purified by nickel-based affinity chromatography using a HisTrap HP affinity column (GE Healthcare Life Sciences, USA) in an ÄKTA-purifier (GE Healthcare, Uppsala, Sweden). The His tag was removed by cleavage with thrombin. The purity of PER-2 was assessed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE).

Steady-state and inhibition kinetics.

Steady-state kinetic parameters were determined using a T80 UV-visible (UV-Vis) spectrophotometer (PG Instruments, Ltd., United Kingdom). Each assay was performed in triplicate in 1× phosphate-buffered saline (PBS) at pH 7.4 at room temperature (22°C). The steady-state kinetic parameters K_m and V_max for nitrocefin and the inhibition parameters for AVI were obtained as described previously (14, 30), with nonlinear least-squares fitting of the data (Henri Michaelis-Menten equation) using GraphPad Prism 5.03 for Windows (GraphPad Software, San Diego, CA, USA):

$$v=\frac{V_{\max }\times [S]}{K_m+[S]}$$ (1)

The interaction of PER-2 with AVI was assumed to follow previous models proposed for other class A β-lactamases, excluding KPC-2 (9). Determination of apparent K_i (K_{i app}) was achieved using a direct competition assay under steady-state conditions with nitrocefin (NCF) as the reporter substrate and measuring initial velocities. The velocity (V₀) obtained after mixing corresponds to equation 2:

$$v_0=\frac{V_{\max }\times [S]}{K_m\times (1+[I]/K_{iapp}+[S])}$$ (2)

Data were linearized by plotting inverse initial steady-state velocities (1/v₀) against inhibitor concentration, [I]. K_{i app} was determined by dividing the value for the y-intercept by the slope of the line and corrected to account for the concentration and affinity of NCF for the β-lactamase according to equation 3:

$$K_{iapp(corrected)}=\frac{K_{iapp(observed)}}{(1+[S]/K_{mnitrocefin})}$$ (3)

To determine k₂/K, progress curves were obtained by incubating PER-2 with increasing concentrations of AVI and maintaining NCF as the reporter substrate. Progress curves were subsequently fit to equation 4 to obtain k_obs values, by nonlinear least-squares fitting of the data, and k₂/K was determined from equation 5:

$$A=v_f\times t+(v_0-v_f)\times (1-e^{-k_{{\text{obs}}^t}})/k_{\text{obs}}+A_0$$ (4)

$$k_{\text{obs}}=k_{-2}+(k_2/K)\times \{[I]/[1+([S]/K_{m\text{nitrocefin}})]\}$$ (5)

For equation 4, A is absorbance at λ = 482 nm, v_f is final velocity, t is time, v₀ is initial velocity, and A₀ is initial absorbance at λ = 482 nm. For equation 5, [I] is concentration of AVI and [S] is concentration of NCF. The data were plotted as k_obs versus [I], and the k₂/K value was obtained by correcting the value for the slope of the line for the concentration and affinity of NCF (equation 6):

$$k_2/K_{(\text{corrected})}=k_2/K_{(\text{observed})}\times [([S]/K_{m\text{nitrocefin}})+1]$$ (6)

The k_off value was determined by incubating the wild-type and mutant PER-2 β-lactamases with an AVI concentration that resulted in complete inhibition. A 1 μM concentration of enzyme was preincubated with 120 μM AVI. Samples were serially diluted (1:2,000), and hydrolysis of 100 μM NCF was measured. Progress curves were fit to a single exponential decay equation as previously described (9, 14).

Crystallization of the apoenzyme form of PER-2 and the PER-2:avibactam complex.

Crystals of apo PER-2 were grown at 20°C by the hanging drop vapor diffusion method with 20% (wt/vol) polyethylene glycol (PEG) 6000 plus 0.05 M imidazole (pH 8.0). The AVI adduct was obtained by soaking apo-PER-2 crystals in crystallization solution supplemented with 26 mM AVI in the mother liquor for 4 days.

Data collection, phasing, model building, and refinement.

X-ray diffraction data for the apo enzyme were collected at 100 K on a Bruker D8 QUEST microfocus diffractometer equipped with a Photon 100 CMOS detector. Crystals of the PER-2:AVI complex were subjected to diffraction studies at 100 K at the Proxima 2A beamline at the Soleil synchrotron, Saint Aubin, France, using a Dectris EIGER 9M detector. Integration and scaling were performed using XDS (31). Structure resolution was achieved by molecular replacement through Phaser (32), using the previously solved structure of PER-2 (PDB code 4D2O) as the starting model (6). Refinement and model building were performed with Phenix 1.12-2829 (33), and Coot 0.8.6.1 (Turtle Bay) (34), respectively. Models were visualized with PyMol 1.7.0.3 (35).

Accession number(s).

The coordinates and structure factors of the apo-PER-2 β-lactamase and the AVI:PER-2 complex structures have been deposited in the Protein Data Bank under accession codes 6DGU and 6D3G, respectively.