Biochemical Insights into Imipenem Collateral Susceptibility Driven by ampC Mutations Conferring Ceftolozane/Tazobactam Resistance in Pseudomonas aeruginosa

ABSTRACT

Several Pseudomonas aeruginosa AmpC mutants have emerged that exhibit enhanced activity against ceftazidime and ceftolozane, while also evading inhibition by avibactam. Interestingly, P. aeruginosa strains harboring these AmpC mutations fortuitously exhibit enhanced carbapenem susceptibility. This acquired susceptibility was investigated by comparing the degradation of imipenem by wild-type and cephalosporin-resistant AmpC. We show that cephalosporin-resistant AmpC enzymes lose their efficacy for hydrolyzing imipenem and suggest that this may be due to their increased flexibility and dynamics relative to the wild type.

INTRODUCTION

The opportunistic pathogen Pseudomonas aeruginosa often resists β-lactam action through the expression of AmpC (or Pseudomonas-derived cephalosporinase [PDC]), a broad-spectrum class C β-lactamase (1–3). Enhanced antibiotic potency is therefore ensured through the concurrent administration of β-lactamase inhibitors (BLIs) (4). Combinations such as ceftolozane-tazobactam and ceftazidime-avibactam provide effective therapies against multidrug-resistant (MDR) or extensively drug-resistant (XDR) P. aeruginosa infections, partially alleviating the urgent clinical need for new agents (5–7). However, the emergence of resistance toward these antibiotic-plus-BLI treatments is of particular concern (8).

Previously, we described AmpC-mediated ceftolozane-tazobactam resistance mechanisms, under both in vitro (9) and treatment (10) conditions. The AmpC mutations underlying these mechanisms include substitutions or deletions in (i) Ω-loop residues (11, 12) or (ii) residues in the Ω-loop vicinity interacting with it (11–15); we quantified their effects on the antibiotic ring-opening efficacy and BLI binding strength (13). Interestingly, cephalosporin-resistant P. aeruginosa strains harboring these mutations also exhibit increased susceptibility to imipenem (9, 10); however, the biochemical basis for this enhanced susceptibility remains unclear. To gain insight into imipenem susceptibility, we quantified the reduction in its degradation efficiency by wild-type (WT) AmpC and the above cephalosporin/BLI variants of the enzyme. Based on these data, we provide insights into the molecular basis for why these ampC mutations exhibit opposite effects on imipenem versus cephalosporin hydrolysis.

Cephalosporin-resistant AmpC enzymes (WT, T96I, G183D, E247K, and ΔG229-E247 [a deletion from position 229 to 247]) were expressed and purified as previously described (13). Antibiotic hydrolysis was measured using an Infinite 200 PRO plate reader (Tecan Trading AG, Switzerland) operating in absorbance mode. Nitrocefin hydrolysis was carried out as before to ensure consistency with our previous work on these variants (13). Imipenem hydrolysis was monitored at 280 nm by challenging 0.5 μM AmpC enzyme solution with imipenem concentrations ranging from 0 to 400 μM for an hour. The reaction was carried out at 25°C in far-UV-transparent plates (Greiner, Monroe, NC, USA) in pH 7 buffer (0.1 M Na₂HPO₄ + 0.01% [wt/wt] bovine serum albumin [BSA]). The initial rate of product (P) formation was determined from the absorbance A by linear regression, using the following relationship:

$${\left(\frac{d[P]}{\mathit{dt}}\right)}_{t=0}=\frac{1}{\Delta {\epsilon }_{280}}{\left(\frac{\mathit{dA}}{\mathit{dt}}\right)}_{t=0}$$

where $\Delta {\epsilon }_{280}=-4,900{\text{M}}^{-1}{\text{cm}}^{-1}$ is the change in extinction coefficient (16). All enzyme runs were performed in triplicate and analyzed using SigmaPlot (Systat Software, Inc., CA, USA). All chemicals were obtained from Sigma-Aldrich.

Table 1 lists the enzyme kinetic parameters associated with the degradation of imipenem, nitrocefin, ceftolozane, and ceftazidime for all studied variants. Nitrocefin hydrolysis was measured as a control to ensure that all mutant enzymes are active. Clearly, cephalosporin-resistant AmpC variants hydrolyze ceftolozane and ceftazidime better than wild-type enzyme ($\frac{k_{\text{cat}}}{K_M}$ values jump by more than an order of magnitude). In contrast, these same variants degrade imipenem much less efficiently than wild-type enzyme ($\frac{k_{\text{cat}}}{K_M}$ values drop to less than 10%). These contradictory mutational effects upon the β-lactam ring opening are best understood if the molecular mechanisms of antibiotic degradation are carefully considered.

TABLE 1.

Michaelis-Menten kinetic parameters of WT and mutant AmpC enzymes for nitrocefin, imipenem, ceftazidime, and ceftolozane^a

Enzyme	Parameterb	Mean ± SD for:
		Ceftolozane	Ceftazidime	Nitrocefin	Imipenem
WT	kcat (s−1)	(1.0 ± 0.1) × 10−2	(4.9 ± 0.6) × 10−3	50 ± 8	(4.57 ± 0.01) × 10−3
	Km (μM)	2,100 ± 300	350 ± 80	90 ± 30	37 ± 5
	kcat/Km (μM−1 s−1)	(5 ± 1) × 10−6	(1.4 ± 0.4) × 10−5	(5 ± 2) × 10−1	(1.2 ± 0.2) × 10−4
	Fold change	1	1	1	1
T96I	kcat (s−1)	NDc	ND	1.4 ± 0.2	(1.2 ± 0.2) × 10−3
	Km (μM)	ND	ND	40 ± 10	61 ± 40
	kcat/Km (μM−1 s−1)	(1.99 ± 0.04) × 10−4	(2.27 ± 0.04) × 10−4	(4 ± 1) × 10−2	(2 ± 1) × 10−5
	Fold change	40 ± 2	17 ± 3	0.07 ± 0.02	0.16 ± 0.08
G183D	kcat (s−1)	(1.8 ± 0.2) × 10−1	ND	0.8 ± 0.1	(1.8 ± 0.2) × 10−3
	Km (μM)	1,200 ± 200	ND	24 ± 11	140 ± 40
	kcat/Km (μM−1 s−1)	(1.5 ± 0.3) × 10−4	(1.41 ± 0.02) × 10−4	(3 ± 2) × 10−2	(1.3 ± 0.4) × 10−5
	Fold change	31 ± 5	10.3 ± 0.1	0.06 ± 0.04	0.11 ± 0.03
E247K	kcat (s−1)	1.5 ± 0.3	(6.6 ± 0.4) × 10−1	0.9 ± 0.04	(2.61 ± 0.18) × 10−3
	Km (μM)	3,600 ± 700	1,020 ± 90	25 ± 4	140 ± 20
	kcat/Km (μM−1 s−1)	(4 ± 1) × 10−4	(6.5 ± 0.7) × 10−4	(36 ± 6) × 10−3	(1.9 ± 0.4) × 10−5
	Fold change	80 ± 20	47 ± 5	0.07 ± 0.01	0.16 ± 0.03
ΔG229-E247	kcat (s−1)	ND	ND	0.5 ± 0.05	(3.5 ± 0.1) × 10−3
	Km (μM)	ND	ND	50 ± 20	320 ± 20
	kcat/Km (μM−1 s−1)	(1.22 ± 0.02) × 10−4	(1.38 ± 0.02) × 10−4	(10 ± 4) × 10−3	(1.1 ± 0.1) × 10−5
	Fold change	25 ± 1	10.1 ± 0.1	0.019 ± 0.008	0.09 ± 0.01

^aThe kinetic parameters for nitrocefin and imipenem were obtained from the experimental data collected for this study (see Fig. S1 and S2 in the supplemental material); the parameters for cephalosporin were obtained from reference 7. All experiments were performed in triplicate. Reported uncertainties were obtained from regression fits.

^bFold change, ratio of the mutant to WT AmpC substrate specificity constants (k_cat/K_m).

^cND, not determined. For these enzymes, the K_m values for ceftolozane and ceftazidime are significantly larger than 5 mM, and the Michaelis-Menten plots are linear not hyperbolic, precluding the determination of k_cat and K_m as independent parameters; the slope of the linear Michaelis-Menten plot measured at substrate concentrations close to zero, however, will be equal to $\frac{k_{\text{cat}}}{K_M}[\text{total enzyme concentration}]$, allowing k_cat/K_m to be determined.

β-Lactamase enzymes (E) degrade cephalosporins (C) through the following mechanism (17):

$$E+C{}_{\underset{k_{\text{off}}}{\leftarrow }}^{\stackrel{k_{\text{on}}}{\to }}\mathit{EC}\stackrel{k_{\text{ring\hspace{0.17em}opening}}}{\to }E-C\stackrel{k_{\text{hydrolysis}}}{\to }E+\text{product}$$ (1)

The third step of this mechanism is postulated (13, 17) to be the rate-determining step for ceftolozane and ceftazidime hydrolysis, and structural data suggest that cephalosporins bind AmpC in catalytically competent conformations (15)—the carbonyl oxygen lies in the oxyanion hole (formed by the backbone amides of S64 and A318), allowing T150 and water to activate it (Fig. 1A).

FIG 1.

Active site structures of AmpC with ceftazidime (PDB accession number 1IEL) (A) and imipenem (1LL5) (B). Residues are labeled with numbers corresponding to E. coli AmpC; numbers in parentheses correspond to Pseudomonas aeruginosa PAO1. α-Helix residues (91 to 106) is shown in yellow, α-helix residues (154 to 165) appears in green, and Ω-loop residues (229 to 247) is indicated in magenta.

Unfortunately, we were unable to crystalize the P. aeruginosa AmpC mutant proteins in this study. However, the structures of wild-type P. aeruginosa and Escherichia coli AmpC proteins are very similar (the root mean square deviation [RMSD] of alpha carbon is ~1.49 Å) (18, 19). Based on the X-ray crystal structures of E. coli AmpC mutant enzymes solved by Thomas et al. (20), we speculate that the effects of the amino acid changes (T96I, G183D, E247K, and ΔG229-E247) upon the structure of P. aeruginosa mutant proteins are as follows (see Fig. S3 in the supplemental material). In wild-type AmpC, the threonine (T96) hydroxyl can form a hydrogen bond with the backbone carbonyl of E247, which stabilizes the Ω-loop. Replacing threonine with isoleucine will prevent this hydrogen bond from forming. Similarly, the E247K replacement precludes the formation of a side chain-backbone hydrogen bond. On the other hand, the G183D replacement exchanges a small hydrogen atom for a bulky negatively charged group that can clash with other nearby moieties, such as E247. These three replacements may reduce favorable interactions between the Ω-loop and the (91 to 106) and (154 to 165) helices, thereby increasing the Ω-loop dynamics. In fact, Thomas et al. saw both loop structure perturbation and a reduction in protein stability as characterized by the melting point when these mutants were introduced (20). Based upon Fig. S3, the deletion mutation of the Ω-loop (ΔG229-E247) may also affect the folding free energy of the (91 to 106) and (154 to 165) helices.

Indeed, we measured the melting temperature of the P. aeruginosa AmpC mutant enzymes (T96I, G183D, E247K, and ΔG229-E247) in our previous work (13), and all the proteins exhibited a reduction in melting point relative to the wild type. This reduction in protein stability causes an increase in the enzyme dynamics (13, 21), allowing water molecules to access the protein interior. We postulate that this increase in dynamics and water access is the likely reason why the magnitude of k_hydrolysis for ceftolozane and ceftazidime is greater for the mutant enzymes than for wild-type enzyme (13).

Conversely, the crystal structure showed that the thermodynamically stable imipenem-AmpC complex has the imipenem carbonyl oxygen O7 outside the oxyanion hole (Fig. 1B) (14, 22). The activation of the carbonyl oxygen is required for β-lactamase catalysis, which led Maveyraud et al. to suggest (22) the existence of a catalytically active “canonical” acyl-enzyme complex structure having the carbonyl oxygen in optimal geometry. This canonical structure is a high-energy conformer because of the unique torsional strain it requires—namely, the strain that the 6α-1R-hydroxyethyl substituent applies to the acyl-serine bond (22). Therefore, we suggest the following mechanism of imipenem (I) ring opening to be more realistic than the scheme in equation 1:

$$E+I\stackrel{K_S}{\iff }\mathit{EI}\stackrel{K_{\text{ring\hspace{0.17em}opening}}}{\iff }E-I_{\text{canonical}}\stackrel{K_{\text{stable}}}{\iff }E-I_{\text{stable}}$$

$$E-I_{\text{canonical}}\stackrel{k_{\text{hydrolysis}}}{\to }E+\text{product}$$ (2)

In this mechanism, the hydrolysis step is still the slow rate-determining step, allowing all prior steps to be represented as equilibria. After substrate binding, the first necessary step for catalysis is the activation of its carbonyl oxygen through the displacement of water from the oxyanion hole and S64 acylation (22). Next, once the complex is formed, it equilibrates between two conformations (22): (i) the canonical form, in which the ester carbonyl is held in the oxyanion hole by two hydrogen bonds, and (ii) the “lower energy form,” similar to the crystal structure. Finally, hydrolysis of the acyl bond occurs, which requires the complex to be in the canonical form prior to deacylation (22).

We express the experimentally determined enzyme specificity constant of imipenem hydrolysis in terms of the microscopic equilibrium/rate constants seen in the scheme in equation 2 (see the supplemental material):

$$\frac{k_{\text{cat}}}{K_m}=\frac{k_{\text{hydrolysis}}}{K_S{K}_{\text{ring\hspace{0.17em}opening}}}$$ (3)

Equation 3 allows us to rationalize the observed the loss of imipenem degradation activity seen in AmpC mutants. Because these mutations presumably increase protein flexibility/dynamics, all enzyme species depicted in the scheme in equation 2 (except E − I_canonical, which has to be “locked” in a catalytically competent state) gain additional degrees of freedom. This increases the entropic penalty required for accessing the E − I_canonical conformation as represented by the constant K_{ring opening} (23). The net effect of this penalty is an increase in the apparent activation barrier, which leads to an increase in K_m and a decrease in the measured $\frac{k_{\text{cat}}}{K_m}$ values for imipenem seen in the mutant AmpCs.

In conclusion, here, we compared the imipenem degradation efficiency of wild-type AmpC with mutants of the enzyme that evade avibactam and have enhanced activity toward cephalosporins (13). We showed that in contrast to cephalosporins, these mutants exhibit significantly smaller imipenem substrate specificity than the wild type. We rationalized this by considering the structural requirements of the catalytic cycle and the effects these point mutations have upon the AmpC dynamics. Cephalosporin resistance-conferring AmpC mutations cause a decrease in protein thermal stability (13), which increases the frequency of dynamic fluctuations (entropy) (21). We suggest that in contrast to cephalosporin degradation, this entropy gain lowers the imipenem degradation efficacy of AmpC enzymes. This efficiency loss leads to higher steady-state concentrations of imipenem inside the bacterium (13), increasing the susceptibility of cephalosporin-resistant P. aeruginosa strains toward imipenem.