First Penicillin-Binding Protein Occupancy Patterns of β-Lactams and β-Lactamase Inhibitors in Klebsiella pneumoniae

ABSTRACT

Penicillin-binding proteins (PBPs) are the high-affinity target sites of all β-lactam antibiotics in bacteria. It is well known that each β-lactam covalently binds to and thereby inactivates different PBPs with various affinities. Despite β-lactams serving as the cornerstone of our therapeutic armamentarium against Klebsiella pneumoniae, PBP binding data are missing for this pathogen. We aimed to generate the first PBP binding data on 13 chemically diverse and clinically relevant β-lactams and β-lactamase inhibitors in K. pneumoniae. PBP binding was determined using isolated membrane fractions from K. pneumoniae strains ATCC 43816 and ATCC 13883. Binding reactions were conducted using β-lactam concentrations from 0.0075 to 256 mg/liter (or 128 mg/liter). After β-lactam exposure, unbound PBPs were labeled by Bocillin FL. Binding affinities (50% inhibitory concentrations [IC₅₀]) were reported as the β-lactam concentrations that half-maximally inhibited Bocillin FL binding. PBP occupancy patterns by β-lactams were consistent across both strains. Carbapenems bound to all PBPs, with PBP2 and PBP4 as the highest-affinity targets (IC₅₀, <0.0075 mg/liter). Preferential PBP2 binding was observed by mecillinam (amdinocillin; IC₅₀, <0.0075 mg/liter) and avibactam (IC₅₀, 2 mg/liter). Aztreonam showed high affinity for PBP3 (IC₅₀, 0.06 to 0.12 mg/liter). Ceftazidime bound PBP3 at low concentrations (IC₅₀, 0.06 to 0.25 mg/liter) and PBP1a/b at higher concentrations (4 mg/liter), whereas cefepime bound PBPs 1 to 4 at more even concentrations (IC₅₀, 0.015 to 2 mg/liter). These PBP binding data on a comprehensive set of 13 clinically relevant β-lactams and β-lactamase inhibitors in K. pneumoniae enable, for the first time, the rational design and optimization of double β-lactam and β-lactam–β-lactamase inhibitor combinations.

INTRODUCTION

The emergence of carbapenem-resistant (CR) Klebsiella pneumoniae is an urgent, global threat to human health (1–3). One of the common resistance mechanisms in K. pneumoniae is the acquisition of carbapenemase enzymes (4). While these enzymes rapidly hydrolyze carbapenems, K. pneumoniae isolates often additionally harbor an extended-spectrum β-lactamase (ESBL). Together, these two types of β-lactamases can confer high-level resistance to virtually all β-lactam antibiotics.

Two strategies have been employed to combat such β-lactamase-producing strains. The first approach involves the development and use of a β-lactamase inhibitor (BLI) which protects the β-lactam from β-lactamase-related degradation. This approach has gained significant interest over recent years, as shown by the development of avibactam (5), vaborbactam (6), relebactam (7), ETX2514 (8), and other β-lactamase inhibitors (9–11).

The second strategy involves the use of double β-lactam combination therapy, which can inactivate an optimal set of penicillin-binding proteins (PBPs). Ideally, the two β-lactams combined should not be extensively hydrolyzed by the same type of β-lactamase. Inactivation of multiple PBPs has been shown to yield synergistic bacterial killing, as observed for combinations of mecillinam (amdinocillin) with other β-lactams against Enterobacteriaceae (12, 13). All β-lactams are well known to covalently bind to and thereby inactivate different PBPs as their primary mechanism of action (14, 15). However, the target affinities for different PBPs can vary substantially between β-lactams. The PBPs are highly conserved enzymes and have been extensively studied (16, 17). While the PBP binding affinities are known for essentially all β-lactams in Escherichia coli and Pseudomonas aeruginosa (18–24), we are not aware of any published data on PBP binding affinities in K. pneumoniae. Consequently, the scientific basis to combine two β-lactams to inactivate a larger set of PBPs is lacking for K. pneumoniae.

For several decades, β-lactams have commonly been used in monotherapy to treat infections by susceptible K. pneumoniae isolates. Interestingly, double β-lactam combination therapies have been evaluated in several clinical trials in the 1970s and 1980s (25–29). These empirical combinations lacked, however, mechanistic support from PBP binding data and have never been optimized. Despite this, some of these clinical trials showed promising results for nonoptimized double β-lactam therapy compared to those for β-lactam-plus-aminoglycoside combinations against K. pneumoniae (25–29). In addition to their promising efficacy, mechanistically optimized, double β-lactam combinations are appealing, since β-lactams are very safe and can be used in patients of all ages.

This study aimed to create the mechanistic basis for double β-lactam antibiotic combination therapies in K. pneumoniae. We generated the first PBP binding data for a comprehensive set of 13 chemically diverse and clinically relevant β-lactam antibiotics and β-lactamase inhibitors. These included four carbapenems, two cephalosporins, three penicillins, one monobactam, and three potentially PBP-binding β-lactamase inhibitors.

(Part of this research was presented at the 28th European Society of Clinical Microbiology and Infectious Diseases, 21 to 24 April 2018, Madrid, Spain.)

RESULTS

Mass spectrometry analysis of labeled bands confirmed the PBP protein sequences (Fig. 1) for the first time in K. pneumoniae. The highest expression was observed for PBP5/6 and PBP1a/1b based on the Bocillin-labeled SDS-PAGE images for the no-drug controls (Fig. 2). The expression of PBPs 2, 3, and 4 was 24- to 133-fold lower than that of PBP5/6. The expression of PBPs during late log growth phase was relatively consistent across both strains.

FIG 1.

Validated PBP bands based on mass spectrometry analysis for strains ATCC 43816 and ATCC 13883. Bands of PBPs labeled by Bocillin FL were separated on SDS-PAGE gels, and then labeled bands were excised and identified via mass spectrometry analysis of membrane proteins.

FIG 2.

Relative expression of different PBPs in K. pneumoniae strains ATCC 43816 and ATCC 13883. Labeled PBPs were separated on SDS-PAGE gels and quantified using the ImageQuantTL software. Data are the average ± SD of results from triplicates for the PBP expression relative to that of PBP5/6.

Binding data (50% inhibitory concentrations [IC₅₀]) for 13 different β-lactams and β-lactamase inhibitors (Table 1) were generated in duplicate for two wild-type K. pneumoniae strains (ATCC 43816 and ATCC 13883). Binding data associated with 75% (IC₇₅) and 90% inhibition (IC₉₀) of Bocillin FL binding are provided in the supplemental material. These ATCC strains lack an AmpC and a KPC β-lactamase; however, they produce a β-lactamase belonging to the SHV-24 cluster and additional β-lactamase superfamily proteins with negligible or unknown spectra of activity (30, 31). We showed via the nitrocefin slide assay that their β-lactamase activity was nondetectable even at a very high bacterial density (>10⁹ CFU/ml; results not shown); nitrocefin is a rapidly hydrolyzed universal substrate of β-lactamases.

TABLE 1.

PBP occupancy patterns of β-lactam antibiotics and BLIs in K. pneumoniae

Strain and PBP	IC50 of the indicated drug (mg/liter)a
	BIA	DOR	IPM	MEM	ATM	PIP	CAZ	FEP	MEC	AMOX	AVI	SUL	TAZ
ATCC 43816
1a/b	0.5	0.5	0.25	2	>256	16	4	1	>256	32	>256	>256	>256
2	<0.0075	<0.0075	<0.0075	<0.0075	16	2	256	0.12	<0.0075	0.5	2	16	16
3	64	1	8	0.12	0.06	0.06	0.25	0.06	128	8	>256	128	128
4	0.015	<0.0075	<0.0075	0.015	32	64	>256	2	>256	0.25	128	>256	>256
5/6	16	8	1	8	128	64	>256	>128	>256	64	>256	>256	>256
MIC	0.06	0.06	0.25	0.06	0.06	8	0.25	0.06	0.125	32	16	64	256
ATCC 13883
1a/b	0.5	1	0.25	0.5	64	8	4	0.5	>256	8	>256	>256	>256
2	<0.0075	<0.0075	<0.0075	<0.0075	16	1	128	0.12	<0.0075	0.5	2	16	16
3	4	0.5	0.5	0.06	0.12	0.015	0.06	0.015	>256	2	>256	128	128
4	<0.0075	<0.0075	<0.0075	<0.0075	8	4	>256	1	128	0.03	32	32	32
5/6	4	4	0.25	2	>256	32	>256	>256	>256	32	>256	>256	>256
MIC	1	0.125	1	0.06	0.25	>32	0.5	0.125	>32	>32	>256	64	256

^aConcentration of β-lactam that inhibits 50% of Bocillin FL compared to that of a control containing no drug. BIA, biapenem; DOR, doripenem; IPM, imipenem; MEM, meropenem; AMOX, amoxicillin; ATM, aztreonam; MEC, mecillinam (amdinocillin); PIP, piperacillin; SUL, sulbactam; TAZ, tazobactam; AVI, avibactam; CAZ, ceftazidime; FEP, cefepime.

In both strains, all carbapenems had very low IC₅₀ for PBP2 and PBP4 and also bound PBP1a/1b, PBP3, and PBP5/6 at higher concentrations (Table 1; Fig. 3 and 4). The IC₅₀ for PBP2 and PBP4 were usually below the lowest tested carbapenem concentration (0.0075 mg/liter). Carbapenems bound PBP3 and PBP5/6 from ATCC 13883 at lower concentrations than those needed for strain ATCC 43816. Imipenem was the carbapenem with the lowest IC₅₀ for the low-molecular-weight PBP5/6. The MICs for the four carbapenems ranged from 0.06 mg/liter to 1 mg/liter in the two ATCC strains; these MICs were higher than the IC₅₀ for PBP2 and PBP4, suggesting that binding and inactivation of PBP2 and PBP4 at very low carbapenem concentrations was insufficient to cause considerable bacterial killing and to affect the MIC. The carbapenem MICs tended to be lower than the IC₅₀ for PBP1a/1b and PBP4.

FIG 3.

Binding patterns of β-lactams (see Table 1, footnote a, for abbreviations) for K. pneumoniae PBPs from strain ATCC 43816. The indicated β-lactam concentrations were used during the 30-min binding reaction before Bocillin FL was added. Labeled PBPs were separated by SDS-PAGE and detected using a fluorimager. A large range of β-lactam concentrations was captured using two different gels (left, 0.007 to 1 mg/liter; right, 2 to 128 mg/liter) for each drug.

FIG 4.

Binding patterns of β-lactams for K. pneumoniae PBPs from strain ATCC 13883 (as described in the legend to Fig. 3).

Aztreonam, piperacillin, ceftazidime, cefepime, and meropenem showed a low IC₅₀ for PBP3 (Table 1; Fig. 3 and 4), whereas the three other carbapenems (imipenem, biapenem, and doripenem) bound PBP3 only at higher concentrations (especially for ATCC 43816). Noncarbapenem β-lactams and β-lactamase inhibitors showed minimal or no binding of PBP5/6. The MIC was comparable to the IC₅₀ for PBP3 for meropenem, aztreonam, piperacillin, ceftazidime, and cefepime; however, the MIC was at least 128-fold higher than the IC₅₀ for PBP3 for piperacillin.

Ceftazidime had a 2- to >64-fold lower IC₅₀ for PBP1a/1b than that of aztreonam and piperacillin. Preferred binding of one or two PBPs was observed for aztreonam (PBP3), piperacillin (PBP3), and ceftazidime (PBP3 and PBP1a/1b). In contrast, cefepime bound PBPs 1 to 4 at more even concentrations (Table 1).

Mecillinam and amoxicillin both bound PBP2, and amoxicillin also bound PBP4. While the PBP IC₅₀ patterns were consistent between the two strains for mecillinam and amoxicillin, the mecillinam MIC was substantially higher for strain ATCC 13883 than for strain ATCC 43816. This suggested that additional factors (such as altered efflux, outer membrane permeability, or both) contributed to the observed differences in the mecillinam MIC between the two strains. While all three β-lactamase inhibitors bound PBP2, avibactam had an 8-fold lower IC₅₀ for PBP2 than that of sulbactam and tazobactam (Table 1; Fig. 5). The MICs of the β-lactamase inhibitors were 4- to >128-fold higher than their IC₅₀s for PBP2; this large range was likely caused by differences in efflux or outer membrane permeability between the three β-lactamase inhibitors and between the two ATCC strains.

FIG 5.

Binding patterns of β-lactamase inhibitors for K. pneumoniae PBPs from strains ATCC 43816 and ATCC 13883 (as described in the legend to Fig. 3).

Our principal component analysis of the log-transformed PBP IC₅₀ showed that the first two eigenvectors explained 86.5% of the total variance in the PBP IC₅₀ for the 13 compounds (Fig. 6). Compounds could be broadly categorized into two clusters. The first contained β-lactams which primarily targeted PBP3 but differed substantially in the IC₅₀ for their secondary targets (i.e., PBP1a/1b for ceftazidime, PBPs 2 and 4 for piperacillin and aztreonam, and PBPs 2, 1a/1b, and 4 for cefepime).

FIG 6.

Principal component analysis of the log-transformed PBP IC₅₀ data for our 13 tested compounds in two K. pneumoniae strains. The plot shows the clustering of compounds according to their positions on the first and second eigenvector. These two vectors explained 86.5% of the total variance. Compounds were grouped into two general clusters. The first cluster contained β-lactams that primarily targeted PBP3 but differed in their secondary targets. The second cluster was comprised of compounds that primarily targeted PBP2 or both PBPs 2 and 4. Of note, among compounds in the second cluster, the carbapenems, amoxicillin, and mecillinam had substantially lower IC₅₀ for their primary targets than the β-lactamase inhibitors. Symbols for β-lactamase inhibitors are smaller due to their much higher IC₅₀ for their primary PBP target relative to the primary PBP target IC₅₀ of the β-lactams.

The second cluster was comprised of compounds which primarily targeted PBP2 with or without PBP4 (Fig. 6). The upper half of the second cluster, shown in Fig. 6, included compounds that additionally bound PBP3 (tazobactam, sulbactam, meropenem, and amoxicillin), although often at substantially higher concentrations than those needed to bind PBP2. The lower half of the second cluster included avibactam as well as doripenem, biapenem, and imipenem, which had substantially higher IC₅₀ for PBP3 than that of meropenem. While the three β-lactamase inhibitors were part of the second cluster due to PBP2 being their primary target, their IC₅₀ were orders of magnitude higher than those for the primary PBP targets of the carbapenems, mecillinam and amoxicillin.

DISCUSSION

This study presents the first PBP binding data for K. pneumoniae using 13 chemically diverse and clinically relevant β-lactams and β-lactamase inhibitors. Furthermore, this is the first report of PBP protein sequences validated by mass spectrometry in K. pneumoniae. Overall, the PBP IC₅₀ and occupancy patterns (i.e., the IC₅₀ for different PBPs for each drug) were consistent across the two K. pneumoniae strains for each compound. The IC₅₀ for biapenem, mecillinam, and avibactam were comparable in the two strains; however, the MIC of these three compounds was 16- to >256-fold higher in strain ATCC 13883 than in strain ATCC 43816. This suggested that other factors (including efflux and outer membrane permeability) contributed to the observed MIC differences.

In general, the PBP occupancy patterns in K. pneumoniae were comparable, but not identical, to those published for E. coli and P. aeruginosa (19, 20, 32). Potentially important differences should be noted. Considerably lower concentrations are sufficient for carbapenems to bind PBP3, for aztreonam or piperacillin to bind PBPs 1a and 1b, and for ceftazidime to bind PBP1a in P. aeruginosa than the concentrations required by the same β-lactams to bind the respective PBP(s) in K. pneumoniae. In contrast, cefepime has a much higher IC₅₀ for PBP2 in P. aeruginosa than in K. pneumoniae. Future studies are required to assess the impact of these PBP IC₅₀ differences.

Based on our results in two ATCC strains, the IC₅₀ for PBP3 was comparable to the MIC for meropenem, aztreonam, ceftazidime, and cefepime, but this did not hold true for piperacillin. There was no obvious correlation between the MICs and IC₅₀ for PBP2 by mecillinam, amoxicillin, or the β-lactamase inhibitors. For carbapenems, the MICs were higher than the IC₅₀ for PBPs 2 and 4 but usually lower than the IC₅₀ for PBP1a/1b. This suggested that binding of PBPs 2 and 4 at low carbapenem concentrations did not correlate with the MIC. In-depth future studies on a larger number of strains are warranted to better correlate the MICs with the PBP IC₅₀ patterns in K. pneumoniae.

Importantly, the MIC is a summary measure after 18 to 24 h of bacterial growth, killing, and potential resistance emergence (33). Thus, multiple factors, including efflux pump(s) and outer membrane permeability, contribute to the MICs. Systematic studies on the time course of bacterial killing and bacterial morphology in response to different PBP occupancy patterns are required to link PBP IC₅₀ to bacterial killing and resistance prevention. Our PBP binding studies were performed on isolated membranes; this efficient approach has been applied by virtually all prior studies in this area. Informed by these PBP binding data from isolated membranes, the PBP binding of selected β-lactams and double β-lactam combinations should be tested in intact bacteria in future studies. Likewise, data on the rates of outer membrane penetration will be highly valuable.

The principal component analysis clustered the tested compounds according to their primary and secondary targets (Fig. 6). The first cluster contained β-lactams which primarily targeted PBP3. The second cluster was comprised of compounds that primarily targeted PBP2 or both PBPs 2 and 4. Interestingly, amoxicillin was part of the second cluster and had a PBP occupancy pattern that was substantially different from that of piperacillin. This was likely caused by the acylureido group in piperacillin, which is lacking in amoxicillin and may provide insights for future structural studies on the binding of PBPs 2, 3, and 4 in K. pneumoniae. To maximize the set of bound PBPs, combining a β-lactam from the first cluster with a carbapenem from the second cluster appears to be a rational choice. This is in agreement with earlier in vitro studies which suggested synergy for combinations of mecillinam (targeting PBP2; also called amdinocillin) with other β-lactams in Enterobacteriaceae (13).

Different complexes are involved in the synthesis and remodeling of peptidoglycan (i.e., the cell wall), including the elongasome and divisome complexes. Both PBPs 1a and 2 are part of the elongasome; likewise, PBPs 1b and 3 both contribute to the divisome in E. coli (34). Inactivation of different sets of PBPs via double β-lactam combinations or gene deletion can achieve synergistic bacterial killing, as shown in E. coli (12, 13).

For K. pneumoniae, however, the composition of the elongasome and divisome complexes is unknown. While this work presents the first PBP IC₅₀ in K. pneumoniae, there is a near-complete dearth of PBP IC₅₀ in other Enterobacteriaceae for virtually all β-lactams (35). While a small number of prior studies showed promising in vitro results in Enterobacteriaceae (13), systematic time course studies are required to design and identify PBP occupancy patterns that achieve synergistic bacterial killing and minimize resistance. The present study enables the design and rational optimization of such double β-lactam combinations in K. pneumoniae.

For these double β-lactam combinations, the relative expression of PBPs may be important (Fig. 2). Highly expressed PBPs such as PBP5/6 and to a lesser extent PBP1a/1b may act as a “sponge” in the periplasmic space. A carbapenem molecule, for example, which covalently binds PBP5/6, is then no longer available to bind PBPs 2, 3, and 4. The expression of PBP5/6 (and PBP1a/1b) is considerably more dominant in K. pneumoniae than in E. coli (36). It is a strength of our employed isolated membrane assay that PBP binding is studied in the presence of all PBPs at their natural relative abundance (Fig. 2). There is a competition of β-lactam molecules binding to different PBPs both in this isolated membrane assay and in the periplasmic space of intact bacteria; for the latter, this competition is more intense, since β-lactams have to penetrate the outer membrane. The next generation of mechanistic models (37) will have to account for the relative PBP expression, as well as the rates of outer membrane penetration and efflux, to establish mass balance equations for β-lactams in the periplasm.

Several randomized clinical trials in febrile granulocytopenic patients have compared double β-lactam to β-lactam-plus-aminoglycoside combinations (25–29) and found promising efficacy for double β-lactam therapy. However, P. aeruginosa showed emergence of β-lactam resistance during moxalactam-plus-piperacillin therapy in four patients (i.e., two in each study) (28, 38). Emergence of resistance by P. aeruginosa was further found in one of three patients receiving cefoperazone plus piperacillin (39) and in three patients treated with moxalactam plus amikacin (38).

It is critical to note that the IC₅₀ for PBP4 in P. aeruginosa is approximately 0.4 mg/liter for moxalactam (40, 41) and 0.5 mg/liter for cefoperazone (42). Thus, both of these β-lactams extensively bind PBP4 at clinical concentrations. We showed more recently (22) that inactivation of PBP4 rapidly and extensively upregulates the AmpC β-lactamase in P. aeruginosa. As this confers high-level resistance to moxalactam, piperacillin, and cefoperazone, it is now clear why these empirical double β-lactam combinations (28, 38, 39) were highly nonoptimal and led to resistance emergence by P. aeruginosa.

Importantly, K. pneumoniae has no native inducible β-lactamase; the only known exception is the horizontally transmitted DHA-1 plasmid-mediated inducible β-lactamase (carrying the regulator ampR gene), which is extremely rare (43, 44). A mutation(s) in AmpD presents another mechanism which leads to stable upregulation of the AmpC β-lactamase in Serratia spp., P. aeruginosa, indole-positive Proteus spp., Citrobacter spp., and Enterobacter cloacae (45). However, this mechanism is not present in K. pneumoniae.

The development of newer β-lactams over the last 30 years enables us to combine two β-lactams which are not inactivated by the same type of β-lactamase. To further enhance double β-lactam combinations, a β-lactamase inhibitor can be added to inhibit β-lactamase activity and further contribute to PBP2 binding (Table 1). Overall, the mechanistic insights on PBP binding and β-lactamases available today enable us to rationally design and optimize highly promising two- and three-drug combination dosing strategies to combat multidrug-resistant K. pneumoniae. These strategies will likely be superior to the empirical, nonoptimized double β-lactam combinations which showed promising efficacy against K. pneumoniae in the 1970s and 1980s (25–29).

To achieve synergistic bacterial killing throughout most of day by a double β-lactam combination, it is important to ensure that both β-lactams are present at the primary infection site at sufficient concentrations; thus, the concentration-time profiles of both β-lactams should be matched. The relatively slow dissociation rate of a bound β-lactam from a PBP facilitates optimization of the time course of PBP occupancy patterns.

In summary, this study provides the first data on PBP occupancy patterns in K. pneumoniae for 13 chemically diverse, clinically relevant β-lactams and β-lactamase inhibitors. The PBP IC₅₀ were consistent across both ATCC strains. We identified two clusters of compounds based on their PBP occupancy patterns. While the PBP IC₅₀ in K. pneumoniae were in general comparable to those reported for P. aeruginosa and E. coli, potentially important differences in PBP IC₅₀ between pathogens were noted. Future studies are warranted to assess the target site penetration of β-lactams in K. pneumoniae and to systematically link PBP occupancy patterns to synergistic bacterial killing and resistance prevention. The mechanistic insights from these future studies will further enhance the foundation for rationally optimizing innovative two- and three-drug combination dosing strategies to combat multidrug-resistant K. pneumoniae.

MATERIALS AND METHODS

PBP binding assays and susceptibility testing.

PBP binding by β-lactam antibiotics was determined using a binding assay with Bocillin FL as the fluorogenic probe. We studied, in duplicate, two K. pneumoniae strains (ATCC 43816 and ATCC 13883) which have minimal β-lactamase activity. These two widely accessible ATCC strains were studied to enhance the robustness of our PBP IC₅₀ data sets and provide some information on strain-to-strain variability. Both of these ATCC strains express the outer membrane porins OmpK35 and OmpK36 at a wild-type level (46, 47). To assess β-lactamase activity, we grew both ATCC K. pneumoniae strains to >9 log₁₀ CFU/ml. Ten microliters of this bacterial suspension were dispensed on a nitrocefin slide (BD DrySlide nitrocefin) and compared against sterile broth. Nitrocefin hydrolysis would manifest as a red color on the slide and was monitored via a Protocol plate reader (Microbiology International, Frederick, MD).

Bacteria were lysed by sonication, and membrane fractions were isolated using ultracentrifugation (48). Protein concentration was determined from the isolated membranes using the Bradford protein assay (49). Binding reactions were conducted for 13 chemically diverse β-lactams or β-lactamase inhibitors using 20 μg of isolated protein. Biapenem was purchased from TRC Canada (Toronto, Ontario, Canada); doripenem and avibactam were from MedChem Express (Monmouth Junction, NJ); imipenem and meropenem were from AK Scientific (Union City, CA); sulbactam was from TCI America (Portland, OR); tazobactam, amoxicillin, piperacillin, aztreonam, cefepime, and cefsulodin were from Chem-Impex International, Inc. (Wood Dale, IL); and mecillinam was from Molekula (Newcastle Upon Tyne, United Kingdom).

The antibiotics at concentrations from 0.0075 to 128 or 256 mg/liter were incubated with the membranes at 35°C for 30 min before the addition of Bocillin FL. The resulting membranes were then analyzed by SDS-PAGE and scanned directly by a Typhoon 9410 fluorimager. Protein gel bands were quantified using an ImageQuantTL 8.1 from GE Health Care Life Sciences (Pittsburg, PA, USA). Binding affinities (IC₅₀) were reported as the β-lactam concentrations that half-maximally inhibited Bocillin FL binding. Additionally, the β-lactam concentrations which inhibited Bocillin FL binding by 75% (IC₇₅) and 90% (IC₉₀) were obtained and reported. Postimaging, PBP labeled gel fragments were excised and analyzed via mass spectrometry to validate the PBP sequences. Broth microdilution MICs were determined according to standard CLSI methods in cation-adjusted Mueller-Hinton broth at least in duplicate (50).

Principal component analysis.

We performed a principal component analysis for the 13 tested compounds using the PBP IC₅₀ data from both K. pneumoniae strains. The IC₅₀ were log-transformed, and compounds clustered according to their positions on the first two eigenvectors, which explained most of the variance in PBP IC₅₀. This analysis was performed in XLSTAT (version 19.02).

Mass spectrometry analysis.

Nano-liquid chromatography–tandem mass spectrometry (nano-LC/MS/MS) was performed on a Thermo Scientific Q Exactive HF Orbitrap mass spectrometer. This instrument was equipped with an EASY Spray nanospray source (Thermo Scientific) and operated in positive ion mode. An EASY Spray PepMAP column from Thermo Scientific was used for chromatographic separations (C₁₈, 75-μm inside diameter, 25-cm length, and 3-μm, 100-Å pore size).

The scan sequence of the mass spectrometer was based on the TopFive method; the analysis was programmed for a full scan recorded from 350 to 2,000 Da and an MS/MS scan to generate product ion spectra. This method determines the amino acid sequence in consecutive instrument scans of the 10 most abundant peaks in the spectrum. Sequence information from the MS/MS data was processed by converting the RAW files into a merged file (MGF) via the msConvert software (ProteoWizard). The resulting MGF files were searched using the Mascot Daemon by Matrix Science (Boston, MA) software (version 2.4.0); the database was searched against the full Swiss-Prot database (version 2017_06; containing 554,860 sequences and 198,649,153 residues).

The mass accuracy was set to 20 ppm for the precursor ions and to 0.1 Da for the fragments. Considered variable modifications were methionine oxidation and deamidation of asparagine-glutamine (NQ). A fixed modification for carbamidomethyl cysteine was considered. Two missed cleavages for the enzyme were permitted. A decoy database was searched to determine the false discovery rate (FDR). Peptides were filtered according to the FDR. Protein identifications were checked manually, and proteins with a Mascot significance threshold P of <0.05 with a minimum of two unique peptides from one protein having a b-ion or y-ion sequence tag of five residues or better were accepted. This approach allowed us to conclusively identify the excised PBPs from the SDS-PAGE gel bands.