Penicillin Binding Protein 7/8 Is a Potential Drug Target in Carbapenem-Resistant Acinetobacter baumannii

ABSTRACT

Limited therapeutic options dictate the need for new classes of antimicrobials active against carbapenem-resistant Acinetobacter baumannii. Presented data confirm and extend penicillin binding protein 7/8 (PBP 7/8) as a high-value target in the CR A. baumannii strain HUMC1. PBP 7/8 was essential for optimal growth/survival of HUMC1 in ex vivo human ascites and in a rat subcutaneous abscess model; in a mouse pneumonia model, the absence of PBP 7/8 decreased lethality 11-fold. The loss of PBP 7/8 resulted in increased permeability, sensitivity to complement, and lysozyme-mediated bactericidal activity. These changes did not appear to be due to alterations in the cellular fatty acid composition or capsule production. However, a decrease in lipid A and an increase in coccoidal cells and cell aggregation were noted. The compromise of the stringent permeability barrier in the PBP 7/8 mutant was reflected by an increased susceptibility to several antimicrobials. Importantly, expression of ampC was not significantly affected by the loss of PBP 7/8 and serial passage of the mutant strain in human ascites over 7 days did not yield revertants possessing a wild-type phenotype. In summary, these data and other features support PBP 7/8 as a high-value drug target for extensively drug-resistant and CR A. baumannii. Our results guide next-stage studies; the determination that the inactivation of PBP 7/8 results in an increased sensitivity to lysozyme enables the design of a high-throughput screening assay to identify small molecule compounds that can specifically inhibit PBP 7/8 activity.

INTRODUCTION

The continual increase of antimicrobial resistance in Acinetobacter baumannii is an impending clinical crisis (1, 2). Extensively drug-resistant (XDR) and carbapenem-resistant (CR) A. baumannii are the “poster-child” for the urgent threat of antibiotic resistance to public health. The CDC reported a 78% increase in CR A. baumannii infection in 2019–2020, more than any other pathogen of concern (3). A large proportion of A. baumannii strains are already resistant to nearly all antibiotic classes, including aminoglycosides, carbapenems, cephalosporins, and β-lactam/β-lactamase inhibitor (4–7). An estimated 24 to 41% excess mortality is caused by multidrug-resistant A. baumannii, far greater than what was observed for Staphylococcus aureus (15%), Escherichia coli (5%), Klebsiella pneumoniae (2%), and Pseudomonas aeruginosa (4%) (2, 8). The Centers for Disease Control and the WHO have characterized A. baumannii as the top priority (Priority 1:Critical, #1) among all pathogens (9, 10).

Antimicrobial agents active against XDR A. baumannii are limited. Unfortunately, the antimicrobials ceftolozane-tazobactam, ceftazidime-avibactam, meropenem-vaborbactam, imipenem-cilastatin-relebactam, fosfomycin, and plazomicin are poorly active against XDR and pan drug-resistant A. baumannii (11) and increasing resistance has been reported for the last resort antimicrobials tigecycline, polymyxins, and cefiderocol (6, 12). Further, there are virtually no antibiotics in late development to adequately combat XDR and CR A. baumannii (4, 9, 13). Thus, there is an urgent need to develop new strategies to combat XDR A. baumannii (14–16).

Low-molecular-weight (LMW) penicillin binding proteins (PBP) may prove to be A. baumannii’s Achilles’ heel. Sequence analysis has delineated eight putative PBPs in A. baumannii (17); four high-molecular-weight (HMW; _AbPBPs 1a, 1b, 2, and 3), three LMW (_AbPBPs 5/6, 6b, and 7/8), and a monofunctional enzyme (MtgA) were identified. Enzymes that mediate peptidoglycan (PG) synthesis are important targets for β-lactam antibiotics. In E. coli, one HMW class A and one class B are requisite for bacterial survival, whereas the loss of all LMW class C enzymes, including PBP 7, is not lethal (18, 19). In E. coli, PBP 7 has a single functional domain with putative periplasmic d-alanyl-d-alanine endopeptidase activity (20, 21), whereas in A. baumannii PBP 7 possesses both dd-endopeptidase and dd-carboxypeptidase activity (22); PBP 8 is an OmpT-mediated degradation product of PBP 7 (19). It was hypothesized that PBP 7 modifies peptidoglycan to prevent autolysis in nonreplicating cells (23). In Salmonella, PBP 7 expression is induced in carbon-starved medium (starvation-stress response) and is postulated to play a role in cell wall remodeling that enhances survival (24), and in A. baumannii, PBP 7 has been shown to contribute to tolerance to meropenem (22). An apparent redundancy in function of LMW enzymes was hypothesized to be responsible for PBP 7/8 not being essential (25, 26). As a result, to date, the focus in antimicrobial development has been on HMW PBPs (27). However, in the study that established PBP 7 in E. coli was not essential, it was stated “the low-molecular weight PBPs may not be essential for laboratory growth of E. coli but might, instead, affect bacterial viability or physiology under conditions not yet tested” (18). Consistent with this speculation, we identified the LMW PBP 7/8 of A. baumannii as an essential gene ex vivo in human ascites and serum, and in vivo, but not when cultured in rich laboratory medium (28). These data have created a paradigm shift making PBP 7/8 a potential therapeutic target. Since our initial observation, other studies have further supported that A. baumannii PBP 7/8 is required for growth/survival in the Galleria mellonella insect infection model (29) and human serum (30). Critical advantages of targeting PBP 7/8 include known druggability (23, 31, 32), high conservation and 100% prevalence across A. baumannii strains (17), lack of human homologs, accessibility (33), and essentiality (28, 34).

In this report, we demonstrate that PBP 7/8 is essential in the clinically relevant, CR, XDR A. baumannii strain HUMC1 under clinically relevant ex vivo and in vivo conditions. The loss of PBP 7/8 on A. baumannii’s susceptibility to antimicrobials and insights into the mechanism(s) by which PBP 7/8 is requisite for A. baumannii’s survival ex vivo and in vivo are described herein. These data support that PBP 7/8 is a high-value target for the development of novel therapeutics.

RESULTS

The loss of PBP 7/8 decreases the survival of the extensively drug-resistant strain HUMC1 in various clinical niches.

Both HUMC1 and HUMC1ΔpbpG were able to grow in laboratory medium (lysogeny broth [LB]) and minimal media plus Casamino Acids [MM-CA]), but HUMC1ΔpbpG growth was not equivalent (P = 0.008 in LB and P = 0.002 in MM-CA) (Fig. 1A). Quantitative growth curves confirmed that the XDR strain HUMC1ΔpbpG demonstrated significantly less survival in ascites compared to HUMC1 (P < 0.0001) (Fig. 1B). Polar effects from allelic deletion of pbp7/8 are unlikely given that the direction of transcription for a putative threonine synthase, encoded by the open reading frame 3′ to pbpG, is in the opposite direction. Nonetheless, to confirm this, quantitative growth curves were performed in ascites with the complemented construct HUMC1ΔpbpG/pNLAC1[pbpG] (PBP7/8 mutant derivative of HUMC1 containing cloned pbpG), which demonstrated similar growth to HUMC1, thereby confirming that inactivation of PBP7/8 was indeed responsible for decreased growth/survival in ascites.

FIG 1.

Growth of HUMC1 (wt) and HUMC1ΔpbpG (PBP 7/8 negative) in minimal medium, rich laboratory medium, and human ascites and in the rat soft tissue and mouse pneumonia infection models. (A) Growth of HUMC1 (wt) and HUMC1ΔpbpG (PBP 7/8 negative) as assessed by measurement of CFU (data are mean ± SD) in minimal medium plus Casamino Acids (MM-CA) (3 biological and 2 to 4 technical replicates for each strain) and rich laboratory medium (lysogeny broth, [LB]) (3 biological and 2 technical replicates for each strain) in vitro). (B) Growth/survival of HUMC1 (wt), HUMC1ΔpbpG (PBP 7/8 negative), and HUMC1ΔpbpG/pNLAC[pbpG] (PBP 7/8 positive) as assessed by measurement of CFU (data are means ± SD) in human ascites ex vivo (6 biologic and 3 technical replicates for each strain). (C) Growth/survival of HUMC1 and HUMC1ΔpbpG as assessed by measurement of CFU (data are means ± SD) in the rat subcutaneous abscess model (3 biological and 2 technical replicates for each strain). The horizontal black line designates the level of detection. (D) Percent mortality at day 5 of C57BL/6 mice after intrapulmonary challenge with HUMC1 and HUMC1ΔpbpG. LD₅₀, 50% lethal dose.Closed symbol, n = 4; open symbol n = 8. For panel A: *, P < 0.05 (two-tailed unpaired t tests for the area under the curve), HUMC1 compared to HUMC1ΔpbpG. For panels B and C: *, P < 0.05 (two-tailed unpaired t tests for the area under the curve), HUMC1, and HUMC1ΔpbpG/pNLAC[pbpG] compared to HUMC1ΔpbpG. For panel D: LD₅₀ was calculated using an exact logistic regression model.

To confirm that PBP7/8 was a factor that contributed to Acinetobacter infection, we performed an in vivo validation of our ex vivo findings. We compared the survival of HUMC1 and its isogenic derivative HUMC1ΔpbpG in a rat subcutaneous abscess model. A major advantage of this infection model is that multiple samplings can be performed over time in each animal, making it time and cost efficient for the initial assessment of strains in vivo. Further, it is clinically relevant given that A. baumannii has been increasingly recognized as a cause of a variety of soft tissue infections (35, 36). Interestingly, both strains demonstrated decreased growth/survival over time; however, compared to HUMC1, HUMC1ΔpbpG decreased growth/survival was significantly greater (P < 0.0001) (Fig. 1C). These data demonstrate that PBP7/8 is important for the survival of HUMC1 in soft tissue infection.

A. baumannii is an important and often lethal cause of pneumonia (4). To assess the role of PBP7/8 for this infection, C57BL/6 mice underwent pulmonary challenge with various inocula of HUMC1 or HUMC1ΔpbpG and were monitored over 5 days for an in extremis state or death. After pulmonary challenge, the odds of death were 11.0-fold greater when infected with HUMC1 compared to HUMC1ΔpbpG (P = 0.0001), after adjusting for challenge inoculum (Fig. 1D; Table S4 in the supplemental material). These data demonstrate that PBP7/8 is important for the virulence of HUMC1 for pneumonia.

The loss of PBP 7/8 decreases the growth/survival of HUMC1 in human serum ex vivo and increases its susceptibility to lysozyme.

The innate host defense response is instrumental in determining whether extracellular bacterial pathogens such as A. baumannii are successfully cleared or establish an infection (4). Therefore, the bactericidal-mediated activity of complement against HUMC1, HUMC1ΔpbpG, and HUMC1ΔpbpG/pNLAC1[pbpG] was assessed to gain insights into the mechanism by which the loss of PBP7/8 decreases growth/survival in ascites and virulence in vivo. Compared to HUMC1, HUMC1ΔpbpG demonstrated a significant decrease in survival in 50% human serum (P < 0.0001) (Fig. 2A). The complemented derivative HUMC1ΔpbpG/pNLAC1[pbpG] demonstrated restored growth/survival but not to the level of HUMC1. Interestingly, heating of the serum to 56°C for 30 min (which inactivates complement-mediated bactericidal activity) resulted in a smaller, but still, significant loss in viability of HUMC1ΔpbpG. We hypothesized that this is due to residual lysozyme activity, which is relatively stable under these conditions (37).

FIG 2.

Growth/survival of HUMC1 (wt), HUMC1ΔpbpG (PBP 7/8 negative), and HUMC1ΔpbpG/pNLAC[pbpG] (PBP 7/8 positive) in human serum and minimal medium plus Casamino Acids (MM-CA) with or without lysozyme. (A) Growth/survival of HUMC1 (wt), HUMC1ΔpbpG (PBP 7/8 negative), and HUMC1ΔpbpG/pNLAC[pbpG] (PBP 7/8 positive) as assessed by measurement of CFU (data are means ± SD) in 50% human serum (black symbols) or 50% human serum heated at 56°C for 30 min (Δ56° serum) (5 to 8 biological and 2 to 3 technical replicates for each strain). *, P < 0.05 (two-tailed unpaired t tests for the area under the curve), HUMC1 plus serum and compared to HUMC1ΔpbpG and HUMC1ΔpbpG/pNLAC[pbpG] plus serum or Δ56° serum. (B) Growth/survival of HUMC1, HUMC1ΔpbpG, and HUMC1ΔpbpG/pNLAC[pbpG] as assessed by measurement of CFU (data are means ± SD) in MM-CA with (black symbols) or without lysozyme (500 μg/mL) (2 to 4 biological and 2 to 3 technical replicates for each strain). *, P < 0.05 (two-tailed unpaired t tests for the area under the curve), HUMC1 plus lysozyme and compared to HUMC1ΔpbpG plus lysozyme.

The sensitivity to lysozyme of HUMC1, HUMC1ΔpbpG, and HUMC1ΔpbpG/pNLAC1[pbpG] was assessed in minimal medium (MM-CA). Compared to HUMC1, HUMC1ΔpbpG demonstrated a significant decrease in survival in the presence of a physiologic concentration of lysozyme (500 μg/mL) (P < 0.0001) (Fig. 2B) (38). Growth/survival was restored to wild-type levels with the complemented derivative HUMC1ΔpbpG/pNLAC1[pbpG].

The loss of PBP 7/8 compromises the permeability barrier of A. baumannii.

The inability to produce PBP 7/8 results in increased susceptibility to complement and lysozyme, which suggests the loss of PBP 7/8 increases permeability (28). Therefore, permeability was directly assessed in HUMC1, HUMC1ΔpbpG, and HUMC1ΔpbpG/pNLAC1[pbpG] in MM-CA and LB media via a fluorescent-based assay, using the probe N-phenyl-l-naphthylamine (NPN), to quantify outer membrane permeability (Fig. 3; Table S5). Compared to HUMC1, the uptake of NPN was greater for HUMC1ΔpbpG in MM-CA (98,595 ± 20,104 versus 33,915 ± 9,148, P < 0.0001) and in LB (54,094 ± 9,492 versus 29,072 ± 5,606, P < 0.0001). In the complemented strain HUMC1ΔpbpG/pNLAC1[pbpG], compared to wild-type parent HUMC1, this phenotype was fully complemented in LB (31,207 ± 3,230 versus 29,072 ± 5,606, P = 0.52) and partially complemented in MM-CA (52,593 ± 8,210 versus 33,915 ± 9,148, P = 0.009), thereby establishing the involvement of PBP 7/8 in maintaining the permeability barrier. These findings were replicated in the antimicrobial-sensitive strain AB307-0294 (Fig. S1; Table S5); in fact, the increase in permeability in AB307ΔpbpG was more pronounced. These data demonstrate that the loss of PBP7/8 compromised the stringent permeability barrier in A. baumannii.

FIG 3.

Permeability of HUMC1 (wt), HUMC1ΔpbpG (PBP 7/8 negative), and HUMC1ΔpbpG/pNLAC[pbpG] (PBP 7/8 positive) in minimal and rich laboratory medium. Permeability, measured by uptake of the fluorescent compound 1-N-phenylnaphthylamine (NPN), of HUMC1, HUMC1ΔpbpG, and HUMC1ΔpbpG/pNLAC[pbpG]. (A) Growth in the rich laboratory medium lysogeny broth (LB) (3 to 4 biological and 3 to 4 technical replicates for each strain and condition). (B) Growth in minimal medium plus Casamino Acids (MM-CA). Bacteria were resuspended in HEPES buffer (5 mmol/L). *, P < 0.05 (one-way ANOVA).

The loss of PBP 7/8 impacts lipid A/lipooligosaccharide and the structure of HUMC1.

In an attempt to gain mechanistic insights, we first investigated whether the altered permeability barrier due to the loss of PBP 7/8 was mediated by modulation of capsule production as assessed by silica density. These data demonstrated that the average density of cells was similar across the three strains; however, HUMC1ΔpbpG displayed greater variation, and the observed band from HUMC1ΔpbpG was less compact compared to its wild-type parent HUMC1 and its complemented derivative HUMC1ΔpbpG/pNLAC1[pbpG] (Fig. S2).

Next, the relative abundance of lipid A, and by extension lipooligosaccharide (LOS), was ascertained by silver staining and densitometric quantitation of proteinase K-digested total cell lysates that were separated via SDS-PAGE (Fig. 4). The total cellular lipids were quantitated in parallel for comparative purposes. The total cellular lipid pool was extracted via chloroform:methanol solubilization, spotted onto silica plates along glycerophospholipid standards, and detected using copper staining and heat treatment; this allowed for subsequent densitometric analyses (Fig. 4). HUMC1ΔpbpG was found to have significantly lower levels of lipid A/LOS (40.0% decrease), whereas the total lipid content decreased to a lesser degree (11.6% decrease), when including the same biomass, as defined by measurements of optical density at 600 nm. These data support that the loss of PBP 7/8 results in diminished lipid A/LOS abundance.

FIG 4.

Quantitation of lipid A/lipooligosaccharides and total cellular lipids. HUMC1 (wt), HUMC1ΔpbpG (PBP 7/8 negative), and HUMC1ΔpbpG/pNLAC[pbpG] (PBP 7/8 positive) were grown to mid-log phase in LB medium and divided for either lipid A/lipooligosaccharide (LOS) (y axis) or total cellular lipid (x axis) analyses. The data (mean values of biological triplicates) for HUMC1ΔpbpG (circle) and the complemented strain HUMC1ΔpbpG/pNLAC[pbpG] (triangle) were corrected against those observed for HUMC1 wild-type cells (diamond). The error bars show the ± standard error of the means for the two parameters. Statistical analyses were performed using a one-way ANOVA where the HUMC1ΔpbpG was found to have significantly lower levels of lipid A/LOS (****, P < 0.0001) and total lipids (*, P < 0.05).

Finally, to further interrogate for changes resulting from the loss of PBP 7/8, the fluorescence of SynaptoGreen C4 in HUMC1, HUMC1ΔpbpG, and HUMC1ΔpbpG/pNLAC1[pbpG] was assessed by flow cytometry. Fluorescence of SynaptoGreen C4-stained HUMC1ΔpbpG (28,193 ± 1,605 [SD]) was significantly greater than HUMC1 (8,999 ± 200 [SD]) and HUMC1ΔpbpG/pNLAC1[pbpG] (9,556 ± 508 [SD]) (P < 0.0001). Since styryl probes, such as SynaptoGreen C4, can function as an indicator of phospholipids abundance in the outer membrane (39), we also interrogated possible implications of PBP 7/8 loss on the cell’s fatty acid composition. However, chromatography-mass spectrometry of HUMC1, HUMC1ΔpbpG, and HUMC1ΔpbpG/pNLAC1[pbpG] did not identify any major changes in total cellular fatty acid composition (Table 1). In addition to the quantitative difference in SynaptoGreen C4 fluorescence in HUMC1ΔpbpG, the forward and side scatter profiles also differed compared to HUMC1 and HUMC1ΔpbpG/pNLAC1[pbpG] (Fig. 5). This is indicative of changes in cell morphology. Hence, these strains were examined by fluorescent and differential interference contrast (DIC) microscopy (Fig. 5), which showed that loss of PBP7/8 resulted in the formation of coccoidal cells and a greater tendency for cells to aggregate. The changes in the cell morphology are the likely cause for changes in SynaptoGreen C4 fluorescence as determined by flow cytometry, as the microscopy did not reveal major changes in the fluorescent intensity of the cellular membrane(s) of the PBP 7/8 mutant compared to wild type. Collectively, these data demonstrate that the loss of PBP7/8 compromised the stringent permeability barrier in A. baumannii, and is likely to be due to, at least in part, to changes in the outer membrane composition mediated by lipid A/LOS abundance.

TABLE 1.

Total cellular fatty acid composition

Species	HUMC1	HUMC1ΔpbpG (% abundance ± SD)	HUMC1ΔpbpG/pNLAC1[pbpG]
12:0	2.6 ± 0.25	2.9 ± 0.04	3.1 ± 0.76
14:0	0.8 ± 0.06	0.8 ± 0.002	0.7 ± 0.06
16:0	36.6 ± 0.40	36.5 ± 0.20	36.1 ± 1.06
16:1	32.7 ± 0.02	32.2 ± 0.31	31.4 ± 0.65
18:1	27.1 ± 0.56	27.4 ± 0.097	28.8 ± 1.02

FIG 5.

Microscopy and flow cytometry evaluation of HUMC1 (wt), HUMC1ΔpbpG (PBP 7/8 negative), and HUMC1ΔpbpG/pNLAC[pbpG] (PBP 7/8 positive). Both differential interference contrast (DIC) and fluorescent (SynaptoGreen C4) microscopy were performed on the wild-type, mutant, and complemented strains. The forward (FSC) and side (SSC) scatterplots from flow cytometric (FC) analyses are aligned with the microscopy images of the respective strains. The dotted line indicates the FSC-SSC distribution of HUMC1 wild-type cells in each of the FC plots.

The loss of PBP 7/8 in A. baumannii variably affects its susceptibility to various antimicrobials.

To assess the potential clinical consequences of the loss of PBP 7/8, susceptibility to a variety of antimicrobials, which included amikacin, azithromycin, aztreonam, polymyxin E, meropenem, tigecycline, and vancomycin, was assessed in the XDR A. baumannii strains HUMC1 and HUMC1ΔpbpG (Fig. 6). To account for intrinsic growth differences observed between HUMC and HUMC1ΔpbpG (Fig. 1A), a comparative mutant susceptibility index (MSI) was determined. MSI values <1 indicate an increased antimicrobial susceptibility for HUMC1ΔpbpG. The loss of PBP 7/8 significantly increased the susceptibility of HUMC1ΔpbpG to all antimicrobials assessed at selected concentrations. However, a significant increase in activity within a clinically achievable serum concentration was only observed for meropenem, polymyxin E, and tigecycline (40–42). These data are presented as MICs in Fig. S3. These data support that inhibition/inactivation of PBP7/8 holds the potential for enhancing the activity of multiple antimicrobial classes.

FIG 6.

Antimicrobial susceptibility of HUMC1 (wt) and HUMC1ΔpbpG (PBP 7/8 negative). The activity of the various antimicrobials listed was assessed against HUMC1 and HUMC1ΔpbpG as defined by growth measured by optical density (Å₆₀₀) after 20 h at 37°C in MM-CA. Each symbol represents a biologic replicate. Four to 6 biological replicates and 4 technical replicates were performed for each of the various antimicrobials, except 10 biological replicates were performed for vancomycin. The mutant susceptibility index (MSI) is the ratio of (mutant Å_600/wild-type Å₆₀₀) with antimicrobial/(mutant Å₆₀₀/wild-type Å₆₀₀) without antimicrobial. MSI values <1 indicate an increased antimicrobial susceptibility for HUMC1ΔpbpG. NS, P > 0.05; *, P ≦ 0.05; **, P ≦ 0.01; ***, P ≦ 0.001; ****, P ≦ 0.0001 (one-sample t and Wilcoxon tests).

(i) The loss of PBP 7/8 has an uncertain effect on biofilm formation.

The treatment of bacterial biofilms, which commonly occurs with prosthetic device infections, can be challenging. HUMC1, HUMC1ΔpbpG, and HUMC1ΔpbpG/pNLAC1[pbpG] were grown in either LB of MM-CA to assess for the effect of PBP 7/8 on biofilm formation (Table 2). When grown in LB or MM-CA, biofilm formation was significantly decreased compared to HUMC1. This decrease was restored with complementation in MM-CA medium but not in LB medium. However, both the planktonic and biofilm CFU enumerated for HUMC1ΔpbpG was significantly decreased in both LB and MM-CA medium compared to HUMC1 under these assay conditions. Therefore, it is unclear whether the diminished ability of HUMC1ΔpbpG to form biofilms is due to diminished growth versus biofilm-forming properties. However, importantly, the loss of PBP 7/8 does not result in an increase in biofilm formation, which would be a liability.

TABLE 2.

Biofilm formation by HUMC1, HUMC1ΔpbpG, and HUMC1ΔpbpG/pNLAC[pbpG] when grown in LB and MM-CA^a

Biofilm production
Strain	Å570 in LB	P value, compared to HUMC1	Å570 in MM	P value, compared to HUMC1
HUMC1	0.986 ± 0.151		1.136 ± 0.319
HUMC1Δpbp7/8	0.732 ± 0.137	<0.0001****	0.686 ± 0.222	<0.0001****
HUMC1Δpbp7/8/pNLAC[pbp7/8]	0.778 ± 0.194	0.0002***	1.033 ± 0.159	0.3470, ns
Media only	0.090 ± 0.040	Not compared	0.087 ± 0.043	Not compared
Biofilm (CFU/mL)
	LB	P value, compared to HUMC1	MM	P value, compared to HUMC1
HUMC1	5.76e7 ± 2.07e7		8.94e7 ± 1.42e7
HUMC1Δpbp7/8	1.77e7 ± 1.56e7	0.0001***	1.88e7 ± 0.52e7	<0.0001****
HUMC1Δpbp7/8/pNLAC[pbp7/8]	4.29e7 ± 1.45e7	0.1464, ns	6.20e7 ± 2.80e7	0.0077**
Planktonic (CFU/mL)
	LB	P value, compared to HUMC1	MM	P value, compared to HUMC1
HUMC1	2.02e9 ± 0.61e9		1.40e9 ± 0.278e9
HUMC1Δpbp7/8	1.46e8 ± 0.67e8	<0.0001****	3.47e8 ± 0.96e8	<0.0001****
HUMC1Δpbp7/8/pNLAC[pbp7/8]	1.85e9 ± 0.26e9	0.5480, ns	9.81e8 ± 1.91e8	0.0004***

^aData are presented as the means ± SD. ***, P ≦  0.001; ****, P ≦  0.0001, P values were calculated using a one-way ANOVA comparing HUMC1 to HUMC1ΔpbpG or HUMC1ΔpbpG/pNLAC[pbpG].

The loss of PBP 7/8 does not increase the expression of ampC.

The loss of other peptidoglycan endopeptidases has increased the abundance of ampC β-lactamase (43), which is undesirable since this would increase resistance to several classes of antimicrobials. To assess whether this occurred with the loss of PBP 7/8, quantitative PCR (qPCR) was performed for HUMC1 and HUMC1ΔpbpG when grown in MM-CA and LB. Two primer pairs (ampC1 and ampC2) were used for ampC detection in independent assays and 16S RNA served as the reference gene. 16S RNA abundance was similar for both strains and growth conditions as expected (Table 3). When grown in LB medium, the quantification cycle (C_q) for HUMC1 and HUMC1ΔpbpG was similar (Table 2). In MM-CA the C_q for HUMC1ΔpbpG demonstrated a small, but significant, statistical increase in C_q (P = 0.03) (Table 3). However, if this small increase was biologically significant, it would reflect a decrease in ampC expression with the loss of PBP 7/8, not an increase, which was the concern. These data were then used to calculate the normalized expression (ΔΔC_q) relative to a control with the Bio-Rad CFX Maestro 2.3 qPCR analysis software (file version 5.3.022.1030). The relative quantities of ampC expression in HUMC1ΔpbpG were normalized to ampC expression in HUMC1, which was set to a value of 1. The relative normalized expression for HUMC1ΔpbpG grown in LB was 1.099 and 0.7699 when the ampC1 and ampC2 primer pairs were used, respectively, and 0.9725 and 0.6368 when grown in MM-CA. Taken together these data demonstrate that the loss of PBP 7/8 does not result in increased expression of ampC but instead a slight decrease in its expression.

TABLE 3.

ampC and 16S rRNA C_q of HUMC1 and HUMC1ΔpbpG grown to mid-log phase in LB and MM-CAA^a

Strain/Assay	LB	MM	P value (LB, MM)b
HUMC1
ampC1	20.02 ± 0.2531	19.97 ± 0.1879
ampC2	19.40 ± 0.2025	19.35 ± 0.2027
ampC1 + 2	19.71 ± 0.3899	19.66 ± 0.3721
HUMC1ΔpbpG
ampC1	19.95 ± 0.2643	20.37 ± 0.2349	0.4375, 0.0312
ampC2	19.45 ± 0.1587	20.07 ± 0.2481	0.7500, 0.0312
ampC1 + 2	19.74 ± 0.3881	20.22 ± 2786	0.7510, 0.0005
HUMC1 16S rRNA	9.343 ± 0.4933	9.156 ± 0.3770
HUMC1ΔpbpG
16S rRNA	9.189 ± 0.2086	9.296 ± 0.1700	0.3394, 0.3394

^aData are presented as the means ± SD.

^bP values were calculated using the Wilcoxon matched-pairs signed rank test comparing HUMC1ΔpbpG to HUMC1.

Serial passage of HUMC1ΔpbpG in ex vivo ascites has a minor impact on lysozyme resistance.

The ability to develop compensatory mutations at a high frequency without a significant effect on biological fitness decreases the value of an antimicrobial target. Therefore, the development of such mutations in HUMC1ΔpbpG was assessed following serial passage in either 10% or 90% ascites over 7 days. Colonies of HUMC1ΔpbpG that survived serial passage in human ascites over 7 days were first assessed for growth in MM-CA plus lysozyme, a surrogate for compensation for the loss of PBP 7/8 activity (Fig. 7; Table 4). The proportion of isolates with enhanced growth increased over time and was greater in 90% ascites compared to 10% ascites. When passaged in 10% ascites, 0%, 11%, 10%, and 16% for isolates from days 1, 3, 5, and 7, respectively, demonstrated increased growth in MM-CA plus lysozyme compared to HUMC1ΔpbpG. For isolates passaged in 90% ascites, 0%, 30%, 40%, and 50% of isolates from days 1, 3, 5, and 7, respectively, demonstrated increased growth in MM-CA plus lysozyme compared to HUMC1ΔpbpG. However, importantly, no isolate reverted to the HUMC1 wild-type growth phenotype.

FIG 7.

Growth of HUMC1ΔpbpG (PBP 7/8 negative) ascites passaged isolates in minimal medium plus Casamino Acids (MM-CA) with lysozyme. HUMC1ΔpbpG was serially passaged in either 10% or 90% human ascites over 7 days. A number of surviving isolates were collected after each passage and subsequently assessed for growth in MM-CA plus lysozyme. (A, C, E, and G) Growth as measured by Å₆₀₀ of isolates obtained after serial passage in 10% ascites from days 1, 3, 5, and 7 days, respectively. (B, D, F, and H) Growth as measured by Å₆₀₀ of isolates obtained after serial passage in 90% ascites from days 1, 3, 5, and 7 days, respectively. The number of isolates assessed from each day is listed in each panel. For comparison, the growth of nonpassaged HUMC1 and HUMC1ΔpbpG was performed in parallel. The black line depicts the 20-h growth of HUMC1ΔpbpG. Two to 6 biological replicates and 2 to 3 technical replicates were performed for each isolate. The number of serially passaged HUMC1ΔpbpG isolates that demonstrated increased growth compared to their nonserially passaged parent is listed in Table 3.

TABLE 4.

Assessment for the development of compensatory mutations that restored of PBP 7/8 activity via passage in human ascites

Condition/serial passage day	Titer postpassage	No. colonies saved for subsequent screening in MM-CA plus lysozyme	Colonies with growth > than HUMC1Δpbp7/8 in MM-CA plus 3 mg/mL lysozyme (%)
10% ascites
1	6.35 × 108	10	0 (0%)
3	1.30 × 108	9	1 (11%)
5	5.00 × 108	10	1 (10%)
7	3.50 × 108	32	5 (16%)
90% ascites
1	6.95 × 108	10	0 (0%)
3	6.05 × 108	10	3 (30%)
5	1.27 × 108	10	4 (40%)
7	2.85 × 107	28	14 (50%)

Seven ascite-passaged HUMC1ΔpbpG isolates that demonstrated the largest increase in growth in MM-CA plus lysozyme were assessed for growth in MM-CA with no added lysozyme. Interestingly, all serially passaged HUMC1ΔpbpG isolates demonstrated improved growth in MM-CA compared to their parent HUMC1ΔpbpG, but less than wild-type HUMC1 (Fig. S4). These data demonstrate that a minority of serial ascites passaged HUMC1ΔpbpG isolates acquired compensatory mutations that enhanced growth in MM-CA in the presence and absence of lysozyme but did not revert to the wild-type growth phenotype.

DISCUSSION

There is a dire need for the development of a new class of antimicrobials active against XDR and CR A. baumannii. Herein, we describe studies on a clinically relevant strain, which further support PBP 7/8 as a potential drug target for such difficult to treat isolates. Initial studies demonstrated that PBP 7/8 was essential ex vivo and in vivo in a drug-sensitive isolate of A. baumannii (28). In this report, we extend that observation to the XDR CR strain HUMC1 (Fig. 1; Table S4). Further, new mechanistic insights on the consequences of PBP 7/8 inactivation were elucidated. Importantly, the loss of PBP 7/8 led to increased sensitivity to complement and lysozyme-mediated antibacterial activity (Fig. 2). In addition, the loss of PBP 7/8 decreased the permeability barrier in A. baumannii (Fig. 3), which is a fundamental mechanism for A. baumannii’s intrinsic antimicrobial resistance (44). Further, HUMC1ΔpbpG demonstrated increased sensitivity to several classes of antimicrobials (Fig. 5), which for some classes of antimicrobials may be due to this altered permeability barrier. These physiological changes observed in the mutant are supported by physical data that demonstrated a decrease in lipid A/LOS abundance and the formation of coccoidal cells, with a tendency to aggregate following the loss of PBP 7/8 (Fig. 4 and 5). Finally, potential liabilities were explored. Disruption of pbpG does not induce the expression of the β-lactamase AmpC (Table 3). Further, despite passage of HUMC1ΔpbpG in human ascites for 7 days, none of the isolates recovered and assessed reverted to a wild-type phenotype, suggesting that A. baumannii does not possess an endogenous bypass pathway. Finally, the loss of PBP 7/8 does not appear to increase biofilm formation (Table 2). Taken together, these data support continued studies of PBP 7/8 as a novel drug target for XDR and CR A. baumannii.

These data are consistent with prior studies. Geisinger et al. (45) also demonstrated that the loss of PBP 7/8 increased the activity of various antimicrobials, albeit some differences were strain dependent suggesting strain-to-strain differences in PBP 7/8-independent factors that modulate antimicrobial activity. In addition, Geisinger et al. (45) demonstrated that the loss of PBP 7/8 results in a reduction in LOS and an increased sensitivity to sodium dodecyl sulfate, suggesting that PBP 7/8 is needed for optimal integrity of the outer membrane. Islam et al. (22) demonstrated that the loss of PBP 7 increased outer membrane permeability as demonstrated by increased uptake of ethidium bromide and decreased outer membrane integrity as demonstrated by increased release of outer membrane vesicles. Islam et al. (22) also demonstrated that the loss of PBP 7 affected the cell morphology and observed increased clumping.

These data suggest several potential mechanisms by which PBP 7/8 enables the survival of A. baumannii ex vivo and in vivo. Lysozyme is a glycoside hydrolase that cleaves 1,4-β-linkages between N-acetylmuramic acid and N-acetyl-d-glucosamine residues in the peptidoglycan backbone. PBP 7/8 is not known to directly affect that site; however, modifications have been shown to be protective against lysozyme’s activity, primarily in Gram-positive bacteria (46). Lysozyme can also kill bacteria via pore formation in bacterial membranes, as can complement (47). These observations have generated several hypotheses. (i) Acinetobacter’s permeability to lysozyme and complement is increased by alterations in the outer membrane, which in turn enhances killing via pore formation. (ii) The lysozyme peptidoglycan cleavage site is directly protected by PBP 7/8 activity. (iii) The loss of PBP 7/8 has resulted in a structural change in the peptidoglycan thereby enabling lysozyme access and/or increased activity. (iv) The loss of PBP 7/8 modulates the activity of lysozyme inhibitors (48). Since the loss of PBP 7/8 affected lysozyme, complement, and selected antimicrobial-mediated bactericidal activity, an effect not solely specific to lysozyme activity would appear to be operational. Lipid A/LOS and morphologic studies support that the loss of PBP 7/8 does indeed result in structural changes, and it is tempting to speculate that these observed changes have resulted in a modulation of the permeability barrier. Initial studies did not demonstrate a change in total cellular fatty acid composition; however, a functional alteration cannot be excluded. Additional next steps include studies to assess whether the loss of PBP 7/8 affects outer membrane proteins or peptidoglycan. Interestingly, Islam et al. (22) demonstrated that in the ATCC strain 17978 the loss of PBP 7/8 resulted in a modification of peptidoglycan. Additional studies are required to fully delineate the mechanism(s) responsible for the phenotypic changes observed for HUMC1ΔpbpG.

The determination that HUMC1ΔpbpG demonstrated an increased sensitivity to lysozyme has several important implications. First, any direct effects of PBP 7/8 inactivation on the bacterial cell’s viability will be augmented by lysozyme-mediated bactericidal activity. Lysozyme is present at high concentrations in numerous body fluids, including blood, urine, and airway secretions at concentrations as high as 1 to 3 mg/mL (48, 49). In addition, neutrophils that migrate to the site of infection also secrete lysozyme, thereby further increasing local concentrations at the site of infection. We have also demonstrated that the loss of PBP 7/8 increases complement-mediated bactericidal activity. Therefore, inactivation of PBP 7/8 would render A. baumannii susceptible to at least two of the critical arms of the host’s innate defense system. Second, the increased sensitivity of HUMC1ΔpbpG to lysozyme can form the basis for a high-throughput drug screen. The sensitivity of HUMC1 in the presence of lysozyme may reflect inactivation of PBP 7/8. Off-target or nonspecific inhibitors will be identified, in part, by the inhibition of HUMC1 growth in the absence of lysozyme and compounds that inhibit the growth of HUMC1ΔpbpG.

PBP 7/8 possesses multiple features desirable in an antimicrobial target. Most importantly, it is essential ex vivo and in vivo (28, 29, 34). It is highly conserved in A. baumannii (17). At the protein level, 100% identity was observed for the sequence from 1630 A. baumannii strains in the public domain. Another critical feature is that the ability of PBP 7/8 to bind penicillins and carbapenems establishes that it is “druggable” (23, 31, 32). Small molecule entry into Gram-negative bacilli (GNB) and subsequent efflux have been limiting factors in antimicrobial development (33). PBP 7/8’s periplasmic location partially mitigates this liability. Presented data support that the inactivation of PBP 7/8 increases permeability and susceptibility to other antimicrobials, which in turn may enable synergistic activity with combination therapy. PBP 7/8 does not have any human homologs. Although this characteristic is not requisite for drug development, it does make the process easier and is predicted to decrease the likelihood of toxicity. Finally, inactivation of PBP 7/8 may lyse nonreplicating bacteria. Antimicrobial binding to PBP 7 in Salmonella is associated with lysis of nonreplicating cells, although other factors may be involved (23), and in A. baumannii, PBP 7 contributes to tolerance (22); this trait, if generalizable, could afford advantages in the treatment of certain infections (e.g., biofilms, abscesses) and help mitigate the development of resistance.

In this report, we also addressed some potential liabilities. The loss of PBP 7/8 could result in increased AmpC production, a nondesirable consequence since this would increase resistance to several classes of antimicrobials. Although this has not been demonstrated for PBP 7/8 in other Gram-negative bacilli, the loss of PBP 4 (another endopeptidase) can result in this phenotype (43). Although PBP 4 is not present in A. baumannii, it was important to establish that RNA abundance for ampC was similar in HUMC1 and HUMC1ΔpbpG (Fig. 6). This observation was further supported by the MIC for HUMC1 and HUMC1ΔpbpG for aztreonam being similar (64 μg/mL and 32 μg/mL, respectively, Fig. S4). Another potential liability would be the rapid development of compensatory mutations. A minority of isolates identified by serial passage in both 10% and 90% human ascites over 7 days developed increased resistance to lysozyme, but none achieved wild-type levels. Interestingly, partial compensatory mutants were more commonly observed in 90% ascites compared to 10% ascites. Further, the increased survival of these isolates when assessed in MM plus/minus lysozyme was shown to be due to a combination of both increased resistance to lysozyme as well as improved growth in MM relative to HUMC1ΔpbpG. Whether the isolates with the compensatory mutation have altered biofitness with altered growth in vivo remains to be seen. Finally, the loss of PBP 7/8 does not appear to increase biofilm formation, thereby obviating the concern that the management of infections complicated by biofilm formation will not be more challenging.

In summary, data reported support PBP 7/8 as a high-value drug target for XDR and CR A. baumannii, for which new antimicrobials for needed. PBP 7/8 has many features desirable for a drug target. Additional studies are needed to identify the mechanism/structural changes by which the loss of PBP 7/8 disrupts A. baumannii’s permeability barrier. Importantly, the determination that the inactivation of PBP 7/8 results in an increased sensitivity to lysozyme enables the design of a high-throughput screening assay for next-stage studies to identify small molecule compounds that can specifically inhibit PBP 7/8 activity.

MATERIALS AND METHODS

Bacterial strains, media, and antibiotics.

A. baumannii strain HUMC1 (blood and lung isolate, bla OXA-51-like, bla OXA-23-like, bla PER, bla GES, aac, IS-Aba1) (50) has a K4 capsular serotype (51) and was isolated from a patient in Los Angeles, CA (accession number NZ_LQRQ01000007.1). HUMC1ΔpbpG (pbpG encodes PBP7/8) is an isogenic derivative of HUMC1 generated by allelic exchange as described previously (52, 53), and bases 127 to 882 (756 bp) of the HUCM1 pbpG open reading frame were replaced with the hygromycin cassette (see Tables S1A and 2A and B for primer and sequence details). The complemented derivative HUMC1ΔpbpG/pNLAC1[pbpG], which was used to confirm that the observed phenotypic differences between HUMC1 and HUMC1ΔpbpG were due to PBP 7/8, was generated by cloning pbpG, including 188 bp upstream and 143 bp downstream of the pbpG open reading frame, into pNLAC with subsequent transformation of pNLAC1[pbpG] into HUMC1ΔpbpG. Phase-variable control of various phenotypes in A. baumannii strain AB5075, regulated by the ompR-envZ two-component system, has been reported (54). Phase variants are recognized by translucent appearance when viewed under indirect light and demonstrate decreased osmotic tolerance and virulence (55). HUMC1 was assessed for phase variation by assessing colonies incubated at 37°C for 24 h on LB agar. Ten plates for each strain were examined; plates with more than 100 colonies per plate were analyzed for phase variants under oblique light. The high-frequency phase variation (i.e., 10⁻¹ to 10⁻²) observed in other strains (55) was not observed within the examined samples. Phase variation in HUMC1, if present, occurs at low frequency and is unlikely to have significantly influenced our findings. Strains were maintained at −80°C in glycerol-LB (1:1; vol:vol). HUMC1ΔpbpG was grown in the presence of 500 μg/mL of hygromycin, and HUMC1ΔpbpG/pNLAC1[pbpG] was grown in the presence of 500 μg/mL of hygromycin and 20 μg/mL of tetracycline.

The procedures for obtaining human ascites fluid and serum were reviewed and approved by the Western New York Veterans Administration Institutional Review Board; informed consent for ascites fluid was waived because it was collected from deidentified patients who were undergoing therapeutic paracentesis for symptoms due to abdominal distension. These individuals were not being treated with antimicrobials and were not infected with human immunodeficiency virus, hepatitis B virus, or hepatitis C virus. The ascites fluid was cultured to confirm sterility, divided into aliquots, and stored at −80°C. BBL Mueller-Hinton II cation adjusted (CAMH) broth consisted of 22 g/L (Becton, Dickinson). Lysogeny broth (LB) consisted of 5 g yeast extract, 10 g tryptone, and 10 g NaCl. Minimal medium plus Casamino Acids (MM-CA) consisted of 200 mL of solution A (2.0 g (NH4)₂SO₄, 6.0 g Na₂HPO₄, 3.0 g KH₂PO₄, 3.0 g NaCl, 0.011 g Na₂SO₄), 800 mL of solution B (0.2 g MgCl₂, 0.0132 g CaCl₂-2H₂O, 0.0005 g FeCl₃-7H₂O, 2.9241 g citrate [trisodium salt dehydrate]), and 3 g of Casamino Acids.

Growth in human ascites fluid and serum.

Growth experiments for HUMC1, HUMC1ΔpbpG, and HUMC1ΔpbpG/pNLAC1[pbpG] in human ascites and serum were performed as previously described (56).

Animal infection models.

Animal studies were reviewed and approved by the Veterans Administration and the University at Buffalo-SUNY Institutional Animal Care Committees and were carried out in strict accordance with the recommendations in the guidelines delineated in the NIH “Guide for the Care and Use of Laboratory Animals” (revised 1985) and the “Ethics of Animal Experimentation Statement” (Canadian Council on Animal Care, July, 1980) and monitored by the Institutional Animal Care and Use Committee. All efforts were made to minimize suffering. Veterinary care for the animals was supplied by the staff of the Veterans Administration Animal Facility under the direction of a fully licensed veterinarian.

(i) Rat subcutaneous abscess model. Briefly, a subcutaneous space was created through the injection of 30 mL of air into the back of anesthetized Long-Evans rats (200 to 225 g), followed by the injection of 1 mL of 1% croton oil in a filter-sterilized vegetable oil vehicle. The space was allowed to mature into an encapsulated, fluid-filled (8 to 12 mL) “pouch” over 6 to 8 days. Neutrophils would have migrated into the abscess in response to appropriate chemotactic signals. The abscess’ subcutaneous location enabled multiple injections and samplings to be performed over time. HUMC1 or HUMC1ΔpbpG was injected alone into the abscess of an anesthetized animal, resulting in an estimated starting abscess concentration of 1 × 10⁷ CFU/mL. Within 1 min after the bacteria were injected into the abscess, 0.5 mL of abscess fluid was removed to measure the actual starting bacterial titer. Fluid aliquots (0.5 mL) were subsequently obtained from anesthetized animals 4, 24, 48, and 96 h after the initial bacterial challenge, and bacterial titers were enumerated.

(ii) Mouse pneumonia model. Pulmonary challenge was performed as described previously (57) with the following modifications. In brief, mice were treated with cobra venom factor (Sigma-Aldrich; cat. no. 233552; 4 units/mouse in 100 μL of 1× phosphate-buffered saline [PBS] approximately 18 h before bacterial challenge) to deplete complement components. Pulmonary challenge was achieved by suspending anesthetized C57BL/6 mice by their incisors, the tongue was pulled horizontally to occlude the esophagus, the nares were manually occluded, 50 μL of the bacterial inoculum in PBS pH 7.4 and 9.6% porcine mucin was deposited in the posterior oro-pharynx, and tongue traction and nares occlusion was maintained until the mouse took three to five breaths, which resulted in aspiration of the inoculum. Multiple challenge inocula were assessed as described in the relevant section. Animals were followed for 5 days with an in extremis state or death used as the study endpoint.

Lysozyme assay.

HUMC1, HUMC1ΔpbpG, and HUMC1ΔpbpG/pNLAC1[pbpG] were grown overnight in LB broth with appropriate antibiotics, washed in PBS and diluted in MM-CA ± lysozyme (500 μg/mL) to achieve a starting titer of approximately 1 to 2 × 10⁵ CFU/mL. Bacteria were incubated at 37°C, shaking at 300 rpm. Aliquots were removed at 0, 2, 4, 6, and 24 h, and bacteria were enumerated via 10-fold serial dilutions. In some experiments, growth was monitored by measuring the Å₆₀₀.

Permeability assay.

The assay was performed as described (58) using the fluorescent compound NPN with the following modifications. HUMC1, HUMC1ΔpbpG, and HUMC1ΔpbpG/pNLAC1[pbpG] were grown overnight in either LB or MM-CA media with the appropriate antibiotics, diluted into fresh medium the next day, grown to logarithmic phase, and resuspended in HEPES buffer (5 mmol/L) so that 100 μL contained approximately 1 × 10⁷ CFU; bacterial titers were confirmed by serial dilution. The assay was performed in a Corning polystyrene 96-well tissue culture-treated black/clear flat-bottomed microplate (number 3603). Each well contained 200 μL; control wells consisted of 200 μL of HEPES or 150 μL of HEPES and 50 μL of NPN (40 μM/L) or 100 μL of HEPES and 100 μL of bacteria, and experimental wells consisted of 50 μL of HEPES, 50 μL of NPN, and 100 μL of bacteria. Fluorescence (excitation: 355 nm; emission: 405 nm) was measured using a Bioteck SYNERGY-H1 fluorescence spectrophotometer 30 s after the addition of bacteria and reported as relative fluorescence units.

Antimicrobial susceptibility testing.

HUMC1 and HUMC1ΔpbpG were grown for 18 h in LB. Bacteria were diluted 1:100 in fresh LB medium, and titers were confirmed by performing serial 10-fold dilutions and enumeration on LB plates plus/minus appropriate antibiotics. Antimicrobial activity was determined by the measurement of optical density (Å₆₀₀) in 96-well microtiter plates containing various concentrations of antimicrobials in 100 μL of MM-CA, which more closely reflects in vivo conditions than LB, to which 100 μL MM-CA containing bacteria was added, with a final bacterial concentration of approximately 5 × 10⁵ CFU/mL of the strain being assessed. The activity of amikacin, azithromycin, aztreonam, polymyxin E, meropenem, tigecycline, and vancomycin was assessed. Control wells contained medium only or bacteria without antimicrobials. Plates were incubated for 20 h at 37°C with double-orbital shaking in an Epoch 2, Biotek, spectrophotometer with Å₆₀₀ measured in each well every 15 min. Comparisons between HUMC1 and its isogenic mutant derivative HUMC1ΔpbpG when tested against antimicrobials were performed in parallel to control for intertest variability. To account for intrinsic growth differences between HUMC1 and HUMC1ΔpbpG a comparative MSI was determined. MSI is the ratio of (mutant Å_600/wild-type Å₆₀₀) with antimicrobial/(mutant Å₆₀₀/wild-type Å₆₀₀) without antimicrobial. MSI values <1 indicate an increased antimicrobial susceptibility for HUMC1ΔpbpG.

Microscan susceptibility testing was also performed on HUMC1 and HUMC1ΔpbpG.

Capsule production analyses.

Semiquantitative analysis of bacterial capsule in HUMC1 and its mutant derivatives was ascertained via a previously established protocol, with minor modifications (59). Two milliliters of overnight culture was harvested by centrifugation at 7,000 × g for 2 min. The cell pellet was resuspended in 2 mL of PBS, of which 1,125 μL of the suspension was mixed with 375 μL of LUDOX colloidal silica (30% [weight] suspension in H₂O). The final mixture was centrifuged at 12,000 × g for 30 min and imaged immediately after (iPhone 11, Apple Inc.) in front of a black background.

Fatty acid analyses.

Bacterial lipid extraction and fatty acid profiling were performed as described previously (60). Overnight cultures of HUMC1, HUMC1ΔpbpG, and HUMC1ΔpbpG/pNLAC1[pbpG] were diluted to an Å₆₀₀ of 0.05 in fresh LB media (20 mL) and grown to mid-log phase (Å₆₀₀ = 0.7). Cells were harvested by centrifugation at 7,000 × g for 10 min and washed with PBS, and processed pellets were then resuspended in 50 μL of 1.5% NaCl solution. For lipid extraction, 1 mL of chloroform:methanol (2:1; vol/vol) was added to the cell suspension, mixed vigorously for 2 min, and then incubated at room temperature for 10 min. Following the addition of 200 μL 1.5% NaCl, the suspension was mixed vigorously for 1 min and centrifuged at 6,000 × g for phase separation. The lower phase was recovered and concentrated via nitrogen evaporation. All samples were stored at −20°C before gas chromatography/mass spectrometry analysis.

To generate fatty acid methyl esters (FAMEs), concentrated lipid samples were resuspended in 1:1 (vol/vol) chloroform and trimethylsulfonium hydroxide and subsequently analyzed using an Agilent 7890A GC system with a 30 m Agilent DB-FastFAME column (Agilent Technologies). Mass spectrometry was completed using a coupled Agilent 5975C MSD system (Agilent Technologies). FAME species were differentiated and determined by comparing them to the FAME mix c4-24 standard (Sigma-Aldrich). The data were analyzed using the Agilent MassHunter Qualitative Navigator software (Agilent Technologies).

Lipid A/LOS analysis.

Overnight cultures of HUMC1, HUMC1ΔpbpG, and HUMC1ΔpbpG/pNLAC1[pbpG] were diluted to an Å₆₀₀ of 0.05 in fresh LB media (20 mL) and grown to mid-log phase (Å₆₀₀ = 0.7). Following this, 15 mL of cells were harvested by centrifugation 7,000 × g for 10 min, resuspended in 2× lysis buffer, and heated at 100°C for 10 min. Samples containing cells were left to incubate with 0.5 mg·mL⁻¹ proteinase K (Sigma-Aldrich) at 56°C. Silver staining of A. baumannii LOS and lipid extracts was performed using an established method (61). Densitometry analyses of lipid A bands were performed using Image Lab v6.1 software (Bio-Rad).

Relative quantitation of A. baumannii total lipids.

For analysis of total bacterial lipid extracts, concentrated lipids samples harvested from 5 mL mid-log cultures of HUMC1, HUMC1ΔpbpG, and HUMC1ΔpbpG/pNLAC1[pbpG] were resuspended in 50 μL of methanol, of which 5 μL of each suspension was spotted onto Silica gel 60 plates (Supelco, Sigma-Aldrich). Detection of lipids was performed by spraying the silica plates with 100 mg·mL⁻¹ copper sulfate containing 8% phosphoric acid followed by charring at 180°C for 5 min. Densitometry of lipid spots was performed using Image Lab v6.1 software (Bio-Rad).

Cell morphology studies.

(i) Flow cytometry. Overnight cultures of HUMC1, HUMC1ΔpbpG, and HUMC1ΔpbpG/pNLAC1[pbpG] were diluted to an Å₆₀₀ of 0.05 in fresh LB media (20 mL) and grown to mid-log phase (Å₆₀₀ = 0.7). One milliliter of the mid-log cultures was harvested by centrifugation at full speed for 5 min and resuspended in 1 mL of PBS. One hundred microliters of cell suspension was incubated with 100 μL of PBS containing SynaptoGreen C4 (Sigma) at a final concentration of 5 μM, for 15 min at room temperature. The stained cells were diluted with 800 μL PBS before analysis using an Attun NxT Acoustic Focusing Cytometer (Thermo Fisher) at an excitation wavelength of 488 nm (blue laser). Bacterial cells were identified based on the forward scatter and side scatter (small particle filter) profiles and comparison to untreated PBS controls. A total of 100,000 bacterial cells were analyzed, which were collected at approximately 6,000 events/second. Fluorescence of SynaptoGreen C4 was measured in the Attune BL2 detection channel with an emission filter of 576 ± 24 nm. The data were processed using Attune Cytometric Software.

(ii) Microscopy. Overnight cultures of HUMC1, HUMC1ΔpbpG, and HUMC1ΔpbpG/pNLAC1[pbpG] were diluted to an Å₆₀₀ of 0.05 in fresh LB media (20 mL) and grown to mid-log phase (Å₆₀₀ = 0.7). Five hundred microliters of the mid-log cultures were labeled directly with 5 μM SynaptoGreen C4 (Sigma) for 5 min at room temperature. Following washing in PBS, cells were analyzed at ×1,000 magnification (UPLXAPO 100× objective, Olympus) on a BX-53 (Olympus) with DIC microscopy, or cells were excited using a U-LGPS light source (Olympus) and U-FGFP filterset (Olympus) to examine SynaptoGreen C4 fluorescence. Images were taken using a DP74 camera (Olympus) and processed with CellSense Dimensions (Olympus).

Biofilm assay.

This assay was performed as described with minor modifications (62). Overnight cultures of HUMC1, HUMC1ΔpbpG, and HUMC1ΔpbpG/pNLAC1[pbpG] in either LB of MM-CA were diluted so that a well in a flat-bottomed, polystyrene 96-well microtiter plate with a low evaporation lid (costar number 3370, Corning Inc.) contained 100 μL of the same media with approximately 1 × 10⁶ CFU. Control wells contained media alone. The plate was incubated for 24 h at 37°C in a partially sealed plastic bag containing a moistened paper towel. The supernatant was carefully aspirated, and bacterial CFU were enumerated. The residual biofilm was washed twice with 125 μL of 1× PBS. Dedicated wells were used to determine either biofilm CFU or biofilm formation. To measure biofilm CFU 100 μL of 1× PBS was added, the well was scraped, and bacterial CFU were enumerated. To measure biofilm formation, 125 μL of cold 95% ethanol was added and a 15-min incubation at room temperature. Next, the ethanol was aspirated and the wells were air dried for 60 min. One hundred twenty-five microliters of 0.1% crystal violet were added to each well, after a 60-minute incubation at room temperature the crystal violet was removed, the stained biofilm was washed once with 125 μL of water to remove any residual stain, and 125 μL of 95% ethanol was then added to each well. The plate was incubated at room temperature for 15 min on a plate shaker (200 rpm), the wells were scraped, and the Å₅₇₀ was measured.

qPCR assay.

Minimum Information for Publication of Quantitative Real-Time PCR Experiments (MIQE) guidelines were followed to ensure optimum RNA quality, primer efficiency, and qPCR parameters (63).

(i) Growth conditions for RNA harvest and DNase treatment. Three individual cultures of either HUMC1 or HUMC1ΔpbpG were grown in LB broth overnight at 37°C. Two hundred fifty microliters of each overnight culture were used to inoculate 20 mL of either LB or MM-CA. The cultures were grown at 37°C, 250 rpm and harvested at mid-log phase (Å₆₀₀ ~ 0.7), and CFU were determined (Table S3A). Two milliliters of culture were centrifuged at 8,000 × g for 5 min at room temperature. The pellet was resuspended in 300 μL of TE buffer (10 mM Tris, 1 mM EDTA, pH 8.0) containing 0.53% SDS and 0.17 mg/mL of proteinase K and incubated at 65°C for 15 min, followed by 5 min on ice. Thirty-six microliters of 5 M sodium chloride was added and each sample was shaken vigorously for 10 s before centrifugation at 11,500 × g for 10 min at 4°C. Supernatants were transferred to new 1.5-mL tubes, 500 μL of isopropanol was added, and the tubes were inverted until the precipitated total nucleic acids were visible. The samples were again centrifuged at 11,500 × g for 10 min at 4°C, the supernatants were removed, the pellets were air dried for 3 min, and resuspended in 50 μL of water.

(ii) DNase treatment of nucleic acid samples. Five microliters each of 10× DNase I buffer and DNase I (Sigma Cat. No. AMPD1) were added to each sample followed by a room temperature incubation for 15 min. Forty microliters of water and 140 μL of TE buffer were added. After tapping the tubes to mix, 36 μL of 5 M sodium chloride were added and each sample was shaken vigorously for 10 s before incubating on ice for 5 min. After centrifugation at 11,500 × g for 10 min at 4°C, the supernatants were transferred to new 1.5-mL tubes and the RNA was precipitated with 500 μL of isopropanol. After centrifugation at 11,500 × g for 10 min at 4°C, the supernatants were removed, and the pellets were air dried for 3 min and resuspended in 50 μL of water. After the DNase treatment was repeated a second time, the pellets were washed twice with 500 μL of 70% ethanol and resuspended in 100 μL of 10 mM Tris, pH 8.5.

(iii) RNA integrity, gene sequences, primer design, qPCR assay specificity, annealing temperature optimization, primer efficiency, and reverse transcription. Methods for RNA integrity determination, primer design, qPCR assay specificity, annealing temperature optimization, and reverse transcription are described in the supplemental material (Supplemental Methods and Tables S1B, S2C and D, and S3B).

(iv) qPCR protocol. qPCR was performed as described previously (64) with the following modifications: 5 μL cDNA (0.5 ng), 10 μL SYBR green, 1 μL primer pair (500 nM each), and 4 μL water. PCR protocol: 98.0°C for 2 min (95.0°C for 15 s, 60°C for 30 s) × 40 cycles, followed by melt curves.

Assessment for the development of mutations that compensated for the loss of PBP 7/8 activity ex vivo.

Serial passage of HUMC1ΔpbpG in 10% or 90% ascites was performed over 7 days as follows. HUMC1ΔpbpG was grown overnight in LB medium plus hygromycin (500 μg/mL), washed in PBS pH 7.4, and concentrated 10-fold to achieve a titer of 1.4 × 10¹⁰ CFU/mL. One hundred microliters of concentrated bacteria were added to either 900 μL of ascites (final concentration 90% ascites) or to 100 μL of ascites and 800 μL of MM-CA (final concentration 10% ascites). Bacteria for each condition were grown at 37°C (300 rpm) for 24 h, after which one aliquot was removed to enumerate the bacterial CFU and 100 μL was added to either 900 μL of fresh ascites or 100 μL of fresh ascites and fresh 800 μL of MM-CA. This process was repeated daily for 7 days. Colonies were saved after each passage to assess for sensitivity to lysozyme (3 mg/mL), which served as a surrogate for compensation for the loss of PBP 7/8 activity. The lysozyme assay was performed as described above; growth at 37°C was assessed by measuring by Å₆₀₀ in a Biotek spectrophotometer (double-orbital shaking (speed 237 rpm [4 mm], slow pace) every 15 min for 20 h. Ten, 9, 10, and 32 isolates and 10, 10, 10, and 28 isolates recovered from days 1, 3, 5, and 7 after passage in 10% and 90% ascites, respectively, were individually tested.

Statistical analyses.

To normalize in vitro, ex vivo, and in vivo growth/survival data (Fig. 1 and 2), log₁₀-transformed values were utilized, and the area under each curve was calculated. Data were assessed for normal distribution; an unpaired, two-tailed unpaired t test was used for normally distributed data, while a Mann-Whitney test was used if data were not normally distributed (Prism 9 for MacIntosh; GraphPad Software Inc.). Permeability assays (Fig. 3), quantitation of lipid A/LOS (Fig. 4), and biofilm formation (Table 2) were assessed by a one-way ANOVA. For antimicrobial susceptibility studies (Fig. 6), one-sample t and Wilcoxon tests were used, which compare the mean of the sample with a hypothetical mean, which we set to 1 since a mean of 1 is indicative of equal susceptibility of the strains. For pulmonary challenge studies (Fig. 1D; Table S4), day 5 in extremis state or death was modeled by infection type (pulmonary, intraperitoneal) as a function of log₁₀(confidence interval [CI]) level and strain using exact logistic regression. Nonsignificant strain by log₁₀(CI) level interactions were removed from the model. Exact tests about the odds ratio = 1 in comparing strains were performed at alpha = 0.05 (two-sided) adjusted for log₁₀(CI) level. Analysis of qPCR data used the Wilcoxon matched-pairs signed rank test (Table 3). P values of 0.05/n (n = the number of comparisons) were considered statistically significant based on the Bonferroni correction for multiple comparisons.

Data and materials availability.

Data are available in the main text or the supplemental materials, and primary data are available at request from the authors.