Molecular Basis of AmpC β-Lactamase Induction by Avibactam in Pseudomonas aeruginosa: PBP Occupancy, Live Cell Binding Dynamics and Impact on Resistant Clinical Isolates Harboring PDC-X Variants

Avibactam belongs to the new class of diazabicyclooctane β-lactamase inhibitors. Its inhibitory spectrum includes class A, C and D enzymes, including P. aeruginosa AmpC. Nonetheless, recent reports have revealed strain-dependent avibactam AmpC induction. In the present work, we wanted to assess the mechanistic basis underlying AmpC induction and determine if derepressed PDC-X mutated enzymes from ceftazidime/avibactam-resistant clinical isolates were further inducible. We determined avibactam concentrations that half-maximally inhibited (IC50) bocillin FL binding. Inducer β-lactams were also studied as comparators. Live cells’ time-course penicillin-binding proteins (PBPs) occupancy of avibactam was studied. To assess the ampC induction capacity of avibactam and comparators, qRT-PCR was performed in wild-type PAO1, PBP4, triple PBP4, 5/6 and 7 knockout derivatives and two ceftazidime/avibactam-susceptible/resistant XDR clinical isolates belonging to the epidemic high-risk clone ST175. PBP4 inhibition was observed for avibactam and β-lactam comparators. Induction capacity was consistently correlated with PBP4 binding affinity. Outer membrane permeability-limited PBP4 binding was observed in the live cells’ assay. As expected, imipenem and cefoxitin showed strong induction in PAO1, especially for carbapenem; avibactam induction was conversely weaker. Overall, the inducer effect was less remarkable in ampC-derepressed mutants and nonetheless absent upon avibactam exposure in the clinical isolates harboring mutated AmpC variants and their parental strains.


Introduction
The worldwide emergence of multidrug-resistant Gram-negative Pseudomonas aeruginosa isolates has caused a substantial public health problem, which is exacerbated by few therapeutic options remaining [1]. Owing to their extensively proven efficacy and safety, β-lactams have been the drug of choice to treat infections caused by P. aeruginosa [2]. All β-lactams bind to and inactivate multiple penicillin-binding proteins (PBPs) with different affinities. PBPs play different roles in peptidoglycan biosynthesis. Roughly, high-molecular-weight PBPs catalyze peptidoglycan polymerization and the cross-linking of glycan strands (transglycosylase (ponA and mrcB; PBP1a and 1b) and DD-transpeptidase (PBP1a, 1b, 2 and 3)), cell wall elongation (pbpA; PBP2) and septum formation (fstI; PBP3). Low-molecular-weight PBPs catalyze DDendopeptidase (dac and pbpG; PBP4 and 7) and DD-carboxypeptidase (dacC; PBP5/6) activities, necessary for the correct peptidoglycan incorporation and cell growth. The bactericidal activity of β-lactams requires the saturation of different PBP receptors, causing the inhibition of the aforementioned biological processes. However, intrinsic and acquired β-lactam resistance is one of the striking features of this microorganism. The chromosomal cephalosporinase AmpC (or Pseudomonas-derived cephalosporinase 1 (PDC-1)) represents one of the most menacing weapons among the extensive P. aeruginosa enzymatic warfare [3]. AmpC is an inducible broad-spectrum Ambler class C β-lactamase capable of hydrolyzing most β-lactams, excluding carbapenems.
In the face of the continuous upscaling of β-lactamase-driven multidrug resistance (MDR) in Gram-negatives, new β-lactam-β-lactamase inhibitor (BLI) treatment options have become available. However, AmpC inhibition is out of the spectrum of most BLIs. Novel combinations such as ceftazidime/avibactam, ceftolozane/tazobactam, meropenem/ vaborbactam and cefepime/zidebactam have based their activity against P. aeruginosa upon different strategies [7,8]. The avibactam β-lactamase inhibitor spectrum includes class C enzymes, which protect the partner β-lactam from AmpC-related hydrolysis [9]. While tazobactam does not have a major impact on the activity of ceftolozane against P. aeruginosa, the latter retains its activity against extremely high ampC (PDC-1) expression in addition to its improved PBP-binding profile [10,11]. Zidebactam, in spite of its fair inhibition capability towards class C β-lactamases, focuses its activity on β-lactam-enhancer properties derived from targeting PBP2 [12,13]. Finally, vaborbactam, a cyclic boronic-based narrow-spectrum BLI, does not offer additional protection against AmpC hyperproduction; however, it seems to enhance meropenem activity against P. aeruginosa [7].
Furthermore, clinically relevant BLIs such as clavulanic acid, tazobactam and sulbactam have been shown to induce ampC expression at clinically relevant concentrations [18,19]. Livermore et al. described a similar effect for the non-β-lactam BLI avibactam, showing significant induction at higher than clinically relevant concentrations (32 µg/mL) and profound strain-to-strain variability. However, according to their work, such induction capacity would become clinically significant only if two conditions were met for the AmpC enzyme: loss of inhibition by avibactam while retaining inducibility [20].
Under these circumstances, knowing whether a novel BLI would be able to induce AmpC expression in avibactam-insensitive fifth-generation cephalosporin-resistant strains would be of great interest. On that premise, we examined the basis for the underlying AmpC induction in the presence of clinically relevant avibactam concentrations (≈50% of C max after 500 mg standard dosage) [21] and the behavior of mutated AmpC enzymes (PDC-221 and -223) [15] upon induction in CAZ/AVI-resistant P. aeruginosa clinical isolates.

Molecular Basis of Avibactam AmpC Induction
To investigate whether the induction of ampC expression in the wild-type P. aeruginosa PAO1 strain was correlated to PBP4 inhibition, the fluorogenic bocillin FL PBP-binding assay was used. The PBP-binding IC 50 values were determined by incubating PBP-containing membrane preparations in growing concentrations (0.25-32 µg/mL) of avibactam and two typical PBP4-binding AmpC inducers (cefoxitin and imipenem) and afterwards labeling with bocillin FL to determine the β-lactam concentrations that half-maximally inhibited (IC 50 ) bocillin FL binding (Table 1). Cefoxitin and imipenem, as previously described, showed low PBP4 IC 50 values, the lowest observed for imipenem (0.1 and 1.5 µg/mL, respectively) [11,22]. Imipenem bound to all PBPs with the highest efficiency with concentrations that ranged from 0.05 to 0.3 µg/mL, whereas cefoxitin binding values for non-PBP4 enzymes were >250-fold higher (range: 7.8-14.6 µg/mL), and no PBP1b binding was detected. Avibactam bound to PBP1b, 2, 4 and 5/6 (IC 50 = 3.6, 4.2, 3.1 and 2.2 µg/mL, respectively). However, in contrast to previous works, no PBP1a and 3 binding was detected under our experimental conditions in the studied strain ( Figure 1) [23]. Besides showing greater PBP4 affinity than previously reported values, avibactam effective concentration was the highest observed from all the three compounds tested (3.1 µg/mL) [23]. Consequently, the differential PBP affinities could be attributed to strain-to-strain variability [20]. enzymes (PDC-221 and -223) [15] upon induction in CAZ/AVI-resistant P. aeruginosa clinical isolates.

Molecular Basis of Avibactam AmpC Induction
To investigate whether the induction of ampC expression in the wild-type P. aeruginosa PAO1 strain was correlated to PBP4 inhibition, the fluorogenic bocillin FL PBP-binding assay was used. The PBP-binding IC50 values were determined by incubating PBP-containing membrane preparations in growing concentrations (0.25-32 µg/mL) of avibactam and two typical PBP4-binding AmpC inducers (cefoxitin and imipenem) and afterwards labeling with bocillin FL to determine the β-lactam concentrations that half-maximally inhibited (IC50) bocillin FL binding (Table 1). Cefoxitin and imipenem, as previously described, showed low PBP4 IC50 values, the lowest observed for imipenem (0.1 and 1.5 µg/mL, respectively) [11,22]. Imipenem bound to all PBPs with the highest efficiency with concentrations that ranged from 0.05 to 0.3 µg/mL, whereas cefoxitin binding values for non-PBP4 enzymes were >250-fold higher (range: 7.8-14.6 µg/mL), and no PBP1b binding was detected. Avibactam bound to PBP1b, 2, 4 and 5/6 (IC50 = 3.6, 4.2, 3.1 and 2.2 µg/mL, respectively). However, in contrast to previous works, no PBP1a and 3 binding was detected under our experimental conditions in the studied strain ( Figure 1) [23]. Besides showing greater PBP4 affinity than previously reported values, avibactam effective concentration was the highest observed from all the three compounds tested (3.1 µg/mL) [23]. Consequently, the differential PBP affinities could be attributed to strain-to-strain variability [20].

Avibactam Target Binding in Live P. aeruginosa Cells
To gather further information regarding the PBP4-related ampC induction by avibactam, the time-course of PBP-binding in live P. aeruginosa cells was determined as previously described with several modifications [24]. We used intact PAO1 cells in the late exponential growth phase with an initial inoculum of ≈4.0 × 10 7 CFU/mL. To infer the outer membrane permeability-limited target binding, a matching experiment with isolated PBP-containing membranes (from previously lysed cells) was similarly performed in the presence of 4 µg/mL avibactam (AVI) (fixed concentration used for minimum inhibitory concentrations and pharmacokinetic model studies). Figure 2 shows the intact cells' (Figure 2a) and isolated membranes' (Figure 2b) binding affinities. No significant PBP1a, 3 and 5/6 binding was observed for the whole-cell assay, whereas 20% of PBP1b and 2 fractions and up to 50% of PBP4 were bound by avibactam after 60 min of incubation. Previously reported data showed that avibactam uptake was indeed limited by the outer membrane permeability barrier, and its activity could be further enhanced upon permeabilization in Gram-negatives, including P. aeruginosa [25]. Unsurprisingly, isolated membranes assay determined a more extensive inhibition of all the PBPs but PBP1a and 3, binding up to 68 and 82% of PBP4 and 5/6 total fractions respectively. As previously suggested by other authors, we confirmed PBP4 binding by avibactam in P. aeruginosa, in both whole cells and isolated membranes [20,26]. Figure 1. Illustrative example of a PBP-binding IC50 SDS-polyacrylamide gel. P. aeruginosa PAO1 cultures were grown to the midexponential phase (7.6 log10 CFU/mL), and membranes were isolated by ultracentrifugation. Isolated PBP-containing membrane preparations (0.5 mg/mL) were incubated in the presence of increasing concentrations of avibactam and afterwards were labeled with 25 µM bocillin FL. Labeled PBPs were separated by SDS-PAGE and detected using a Fluorimager. a Penicillin-binding proteins identified in P. aeruginosa PAO1. b AVI, avibactam, range of concentrations tested: 0.25-32 µg/mL.

Avibactam Target Binding in Live P. aeruginosa Cells
To gather further information regarding the PBP4-related ampC induction by avibactam, the time-course of PBP-binding in live P. aeruginosa cells was determined as previously described with several modifications [24]. We used intact PAO1 cells in the late exponential growth phase with an initial inoculum of ≈4.0 × 10 7 CFU/mL. To infer the outer membrane permeability-limited target binding, a matching experiment with isolated PBP-containing membranes (from previously lysed cells) was similarly performed in the presence of 4 µg/mL avibactam (AVI) (fixed concentration used for minimum inhibitory concentrations and pharmacokinetic model studies). Figure 2 shows the intact cells' (Figure 2a) and isolated membranes' (Figure 2b) binding affinities. No significant PBP1a, 3 and 5/6 binding was observed for the whole-cell assay, whereas 20% of PBP1b and 2 fractions and up to 50% of PBP4 were bound by avibactam after 60 min of incubation. Previously reported data showed that avibactam uptake was indeed limited by the outer membrane permeability barrier, and its activity could be further enhanced upon permeabilization in Gram-negatives, including P. aeruginosa [25]. Unsurprisingly, isolated membranes assay determined a more extensive inhibition of all the PBPs but PBP1a and 3, binding up to 68 and 82% of PBP4 and 5/6 total fractions respectively. As previously suggested by other authors, we confirmed PBP4 binding by avibactam in P. aeruginosa, in both whole cells and isolated membranes [20,26].

qRT-PCR-Based ampC Induction Assays
To investigate whether the binding of avibactam to PBP4 in live cells was correlated with ampC induction, qRT-PCR was performed to determine the time-course expression of β-lactamase mRNA (AVI 16 µg/mL; t = 0, 30 and 180 min). Typical AmpC inducers, imipenem and cefoxitin, were used as comparators (IPM 8 µg/mL, FOX 64 µg/mL). To assess the potential clinical implications of ampC induction in the clinical worst-case scenario, two clinical isolates belonging to the high-risk clone ST175, harboring AmpC structural modifications in the Ω loop or adjacent residues and their parental wild-type

qRT-PCR-Based ampC Induction Assays
To investigate whether the binding of avibactam to PBP4 in live cells was correlated with ampC induction, qRT-PCR was performed to determine the time-course expression of β-lactamase mRNA (AVI 16 µg/mL; t = 0, 30 and 180 min). Typical AmpC inducers, imipenem and cefoxitin, were used as comparators (IPM 8 µg/mL, FOX 64 µg/mL). To assess the potential clinical implications of ampC induction in the clinical worst-case scenario, two clinical isolates belonging to the high-risk clone ST175, harboring AmpC structural modifications in the Ω loop or adjacent residues and their parental wild-type AmpC isolates, were also studied. Described AmpC mutations were previously shown to reduce susceptibility to avibactam inhibition and enhance ceftolozane and ceftazidime catalytic efficiencies simultaneously [15,17]. Table 2 shows the basal ampC mRNA expression levels for each of the strains studied relative to wild-type strain PAO1. As previously demonstrated by our group and other independent researchers, PBP4 inactivation (PA∆dacB) elicited significant ampC hyperexpression; the highest relative expression values were observed at the 180 min determination (1100 ± 243). Both initial ST175 isolates, 101-E5 and 109-E9 (wild-type AmpC), and their CAZ/AVI post-therapy resistant isolates, 103-H8 and 110-G8, showed an even more extensive ampC relative expression (2400 ± 1100, 1700 ± 382, 5400 ± 2900 and 3500 ± 1600, respectively) consistent with the ampR-G154R-resulting enhanced ampC expression observed in nearly all of the ST175 isolates [15,27,28]. Inactivation of the three low-molecular-weight (LMW) PBPs (PBP4, 5/6 and 7) in PA∆dacBdacCpbpG elicited the most remarkable cephalosporinase constitutive overexpression (58,000 ± 26,000) after 180 min.   Figure 3. The values represent each strain ampC relative expression normalized to its expression under basal conditions at t = 0 min. The most notable cephalosporinase induction was observed for the wild-type strain PAO1 upon imipenem incubation. The maximum values (31,200 ± 17,000) were observed by the end of the assay (180 min). As previously reported in works from our group and others, cefoxitin induction was less significant compared to imipenem, achieving maximum expression after 30 min incubation [11,20]. On the other hand, avibactam showed the lowest ampC induction capacity, however still significant (according to the >10 times threshold proposed by several authors) after 180 min incubation (60.6 ± 25.2) [20,29]. As opposed to observations in other species, avibactam is capable of inducing P. aeruginosa chromosomal cephalosporinase AmpC [20,26], although not as strongly as classical inducers such as imipenem or cefoxitin. Experimental induction capacity was consistently correlated with observed PBP4 inhibition values for each of the studied drugs (  According to the premise that higher basal ampC expression levels (constitutive overexpression) determine lesser inducible endpoints, the PBP4 (dacB) mutant showed a much lower ampC inducibility compared to its parental strain PAO1 for both imipenem and cefoxitin for all the observations, 107-and 32-fold lower (289 ± 138 and 77.6 ± 35.7) for the knockout strain at 180 min, respectively [6]. Surprisingly, avibactam caused a slightly higher ampC expression increase in dacB mutant, especially towards the end of the assay (180 min). This observation could be attributed to the concomitant inhibition of LMW PBP5/6 and 7 (dacC, pbpG) beyond PBP4 inactivation [4,22]. In fact, regardless of a remarkably higher relative basal ampC expression compared to PAO1 (58,000 ± 26,000; t = According to the premise that higher basal ampC expression levels (constitutive overexpression) determine lesser inducible endpoints, the PBP4 (dacB) mutant showed a much lower ampC inducibility compared to its parental strain PAO1 for both imipenem and cefoxitin for all the observations, 107-and 32-fold lower (289 ± 138 and 77.6 ± 35.7) for the knockout strain at 180 min, respectively [6]. Surprisingly, avibactam caused a slightly higher ampC expression increase in dacB mutant, especially towards the end of the assay (180 min). This observation could be attributed to the concomitant inhibition of LMW PBP5/6 and 7 (dacC, pbpG) beyond PBP4 inactivation [4,22]. In fact, regardless of a remarkably higher relative basal ampC expression compared to PAO1 (58,000 ± 26,000; t = 180 min), triple PA∆dacBdacCpbpG mutant showed no significant avibactam induction (9.1 ± 4.3) (Figure 3).
In addition to the lower induction capacity thoroughly observed for the AmpCderepressed mutants, the lower imipenem β-lactamase induction effect in ST175 clinical isolates (101-E5, 109-E9, 103-H8 and 110-G8) could be further explained by the observed OprD mutations [15]; that is, restricting imipenem periplasmic uptake would reduce the number of molecules capable of binding to the low-molecular-weight PBPs, which, in turn, cause the β-lactamase expression induction [4,22]. Moreover, AmpC structural mutations have been shown to reduce imipenem resistance and presumably enzyme's hydrolytic capacity towards this compound [30].
In summary, the results presented in this work are in accordance with previous studies suggesting a link between avibactam ampC induction and PBP4 inhibition. Furthermore, we demonstrated "in vivo" PBP4-derived induction process and outer membrane permeability-limited avibactam target site penetration and binding. This may explain the added synergistic effects of CAZ/AVI when coadministered with an outer membrane permeabilizer [25]. What is more important is that, even when the foretold conditions for the clinical significance of avibactam induction are met (i.e., mutated AmpC with loss of affinity for avibactam while retaining inducibility) [20], ampC variants studied in this work (PDC-221 and -223) were not reactive to avibactam induction. In terms of bacterial physiology and fitness, it seems reasonable not to retain full inducibility once the derepressed enzyme expression has already been selected [6]. Understanding the dynamics of mutational ampC regulation and resistance to antimicrobials and β-lactamase inhibitors will definitely help the rational development of novel compounds and combinations that are less prone to select mutational resistance.

Bacterial Strains and Antibiotics
The wild-type reference strain P. aeruginosa PAO1 [31] was used for PBP-binding, IC 50 determination and quantification of the AmpC induction assay. P. aeruginosa knockout strains PA∆dacB (PBP4) [6] and PA∆dacBdacCpbpG (PBP4, 5/6 and 7) [22], and two previously characterized [15] pairs of TOL/TAZ susceptible/resistant XDR clinical isolates belonging to the epidemic high-risk clone ST175, were also used for the quantification of AmpC induction.
3.2. Intact Cells' Time-Course of PBP-Binding Assay P. aeruginosa PAO1 cultures grown to the midexponential phase (7.6 log 10 CFU/mL) were preincubated in cation-adjusted Mueller-Hinton broth (CAMHB) at 37 • C (180 rpm) for 20 min. Afterwards, 4 µg/mL of avibactam (AVI) was added, and the bacterial cultures were incubated (37 • C, 180 rpm). Control and treatment samples were taken after 15, 30 and 60 min incubation, kept in ice and centrifuged (3220× g for 10 min at 4 • C). The bacterial pellets were washed in 30 mL phosphate-buffered saline (PBS) (pH 7.5) four times, resuspended in 10 mL of PBS and sonicated by using a Digital Sonifier Unit model S-450D (Branson Ultrasonics Corp., Danbury, CT, USA) at 40% amplitude for three 2 min bursts (while immersed in an ice bath) and centrifuged at 3220× g for 15 min at 4 • C. Membranes containing the PBPs were isolated by ultracentrifugation at 150,000× g for 30 min at 4 • C using an Optima L100XP Ultra centrifuge (Beckman Coulter, Inc., Palo Alto, CA, USA) and were resuspended in 60 µL of PBS. Total protein content was measured using the Quick Start Bradford protein assay kit with bovine serum albumin as a standard (Bio-Rad Laboratories, Hercules, CA, USA), according to the manufacturer's instructions. To determine the PBP fraction unbound, membranes containing PBPs (20 µL, at 0.5 mg/mL) were labeled with a 25 µM concentration of the fluorescent penicillin bocillin FL [32]. Labeled PBPs were denatured and separated through 4-15% SDS-polyacrylamide gel electrophoresis (Bio-Rad Laboratories, Hercules, CA, USA). Labeled PBPs were visualized using a Typhoon FLA 9500 biomolecular imager (GE Healthcare Bio-Sciences AB, Uppsala, Sweden) (excitation at 488 nm and emission at 530 nm), and binding to different PBPs was determined from at least three independent experiments using ImageQuant TL software v8.1.0.0 (GE Healthcare Bio-Sciences AB, Uppsala, Sweden).
In parallel to the aforementioned assay, we performed analogous experiments in which isolated PBP-containing membrane preparations instead of intact cells were preexposed to avibactam. Briefly, membrane preparations containing PBPs were obtained by following previously described protocols [8] and were incubated for 15, 30 and 60 min at 37 • C in the presence of 4 µg/mL avibactam and were afterwards labeled with bocillin FL (25 µM). Determination of the PBP fraction unbound was performed as described in the above section.

Determination of PBP-Binding Affinity (IC 50 )
PBP-binding affinity was determined using previously described methods [8]. Briefly, P. aeruginosa PAO1 cultures grown to the midexponential phase (7.6 log 10 CFU/mL) were incubated in CAMHB at 37 • C and 180 rpm, washed and centrifuged (3220× g for 10 min at 4 • C). Bacterial pellets were washed in 30 mL of PBS buffer (pH 7.5), resuspended in 10 mL of PBS and sonicated at 40% amplitude for three 2 min bursts (while immersed in an ice bath) and centrifuged at 3220× g for 15 min at 4 • C. PBP-containing membranes were isolated by ultracentrifugation at 150,000× g for 30 min at 4 • C. Total protein concentrations were measured through the Bradford method.