Multiple resistance factors collectively promote inoculum-dependent dynamic survival during antimicrobial peptide exposure in Enterobacter cloacae

Abstract

Antimicrobial peptides (AMPs) are a promising tool with which to fight rising antibiotic resistance. However, pathogenic bacteria are equipped with several AMP defense mechanisms, whose contributions to AMP resistance are often poorly defined. Here, we evaluate the genetic determinants of resistance to an insect AMP, cecropin B, in the opportunistic pathogen Enterobacter cloacae. Single-cell analysis of E. cloacae’s response to cecropin revealed marked heterogeneity in cell survival, phenotypically reminiscent of heteroresistance (the ability of a subpopulation to grow in the presence of supra-MIC concentration of antimicrobial). The magnitude of this response was highly dependent on initial E. cloacae inoculum. We identified 3 genetic factors which collectively contribute to E. cloacae resistance in response to the AMP cecropin: The PhoPQ-two-component system, OmpT-mediated proteolytic cleavage of cecropin, and Rcs-mediated membrane stress response. Altogether, this evidence suggests that multiple, independent mechanisms contribute to AMP resistance in E. cloacae.

Introduction

Bacterial pathogens develop resistance to antibiotics at a remarkable rate, complicating our ability to fight life-threatening bacterial infections. This includes the evolution of multi-drug resistant (MDR) infections for which no currently available antibiotic is effective [1]. This is a particular problem with many Gram-negative pathogens, such as prominent Enterobacterales. Enterobacter cloacae, for example, is an opportunistic, rod-shaped Gram-negative pathogen, which can cause respiratory, soft tissue, and bloodstream infections [2]. While it is a member of our typical gut microflora, MDR E. cloacae have been designated a major public health risk by the CDC (the genus Enterobacter is the second “E” in the CDC’s infamous ESKAPE high priority pathogen list), particularly in immunocompromised patients [3]. Similar risks exist for other Enterobacterales, such as Escherichia coli [4] and Klebsiella pneumoniae [4, 5]. It is therefore essential to explore and develop alternative antimicrobials capable of eradicating MDR infections.

Antimicrobial peptides (AMPs) are short, often cationic peptides produced by nearly all domains of life, making up a key component of innate immunity [6–9]. This includes humans, which produce over 100 different AMPs to defend epithelial surfaces and to aid phagocytes in bacterial killing. AMPs also comprise a portion of our initial systemic response to infection [10]. Thousands of structurally unique AMPs have been isolated from other mammals, amphibians, reptiles (e.g., crocodylians), fish, birds, arachnids (tarantulas, araneomorph spiders, and scorpions), and mollusks [11–13]. Invertebrates lack adaptive immunity and rely heavily on AMPs to ward off invading microbes [14, 15].

Most positively-charged AMPs bind to negatively-charged bacterial membranes [6]. Binding may occur in an orderly, structured manner to create pores in the membrane. This may disrupt the permeability barrier, dissipate ion gradients, or promote entry of other AMPs with intracellular targets, which will result in the death of the bacterial cell [16]. In other cases, AMPs will coat the cell surface. This causes a detergent-like effect, which scrambles the integrity of the cell envelope and results in lysis [17]. Importantly, AMPs tend to kill bacteria more quickly than antibiotics [18] with a more narrow range of concentrations eliciting responses from the bacterial cell [19]. In principle, this means fewer opportunities for a population of bacteria to develop resistance mechanisms to AMPs, at least through mutation.

Due to their natural abundance, diversity, and potent antibacterial activity, AMPs are of high interest as potential alternatives to classical antibiotics [20]. One such example of AMPs being explored for therapeutic purposes are cecropins, a family of linear α-helical AMPs isolated from insects [21, 22]. Cecropins display potent antimicrobial activity against several species of Gram-positive and Gram-negative bacteria, but predominantly target Gram-negative LPS [23–26]. Dozens of other AMPs are currently in clinical development [27]. Therefore, it is vital to understand the mechanisms by which pathogens can resist killing by AMPs. This may enable the more rational design of AMPs for future therapeutic use.

The main characterized AMP resistance mechanisms are target (membrane) modification, AMP degradation, efflux and exclusion of AMPs. One of the most well-understood mechanisms for AMP resistance is modification of the outer membrane (OM) [28, 29]. For example, addition of amino sugars [30], phosphoethanolamine [31], or other amine-containing residues, or removal of phosphate residues [32], prevents the electrostatic binding of AMPs to normally negatively charged lipopolysaccharide (LPS), the primary component of the bacterial OM [33]. The rigidity of the OM likely also impacts AMP susceptibility; several members of Enterobacterales attach additional acyl chains to their lipid A in response to AMP exposure, modifying the packing and fluidity of LPS molecules [34]. In addition to membrane modification, many bacteria produce proteases capable of degrading AMPs. This includes secreted proteases [35], membrane-bound enzymes [36], or even AMP importers for intracellular cleavage [37, 38]. The susceptibility of an AMP to proteolytic cleavage is highly dependent on its structure, and linear AMPs may be more prone to degradation than an AMP with multiple disulfide bonds [39]. Additionally, biofilms can provide an additional layer of protection, hampering the ability of AMPs to reach the bacterial cell envelope. Exopolysaccharides and capsular polysaccharides (CPS) have been shown to confer up to 1000-fold higher AMP resistance in capsulated vs non-capsulated strains of many bacteria [40, 41]. In Enterobacterales, production of CPS is regulated by the envelope stress response system known as the Rcs phosphorelay [42].

Complicating the study of these resistance mechanisms is so-called “heteroresistance”. In a heteroresistant population of bacteria, a subset of cells within the genetically identical population displays higher resistance to an antibiotic or antimicrobial peptide [43–46]. Distinct from outright resistance, heteroresistance is often not stable and is instead a result of varying temporal levels of gene expression or unstable genetic alterations [47]. As heteroresistance can complicate our ability to diagnose and treat bacterial infections, it is important to investigate the nuanced mechanisms that drive this seemingly stochastic phenotype. Of note, the reason for the stochastic nature of heteroresistance (i.e., the underlying cause of population heterogeneity) is poorly understood.

In this study, we used the opportunistic pathogen Enterobacter cloacae and the insect AMP cecropin B (cecB) as a model to study mechanisms responsible for AMP resistance in Enterobacterales. We observed inoculum-dependent killing of E. cloacae by cecB, with cells at high inoculum conditions ultimately displaying higher resistance to killing after an initial die-off than cells at low inoculum. Single-cell analysis revealed this growth was due to initial heteroresistance within the population, uncovering the underlying cause of the inoculum effect against AMPs. We then identified the PhoPQ two-component system, the OM protease OmpT, and Rcs signaling as collective contributors to this form of heteroresistance.

Results

Enterobacter cloacae is resistant to cecB in an inoculum-dependent manner

It has been shown previously that cecropin B (cecB) is effective at killing Gram-negative bacteria at low MICs [48], but previous experiments have been mostly conducted with well-studied model organisms like E. coli. As part of an ongoing investigation into AMP resistance mechanisms in opportunistic pathogens, we assessed the killing dynamics of Enterobacter cloacae exposed to cecB. To this end, we treated a high density of cells (10-fold diluted subculture from overnight, about 10⁸ CFU/mL) with cecB in a growth curve assay at 2x MIC (Fig 1A). Despite the cecB concentration above the MIC, E. cloacae eventually grew after a lag phase of 2 hours. This phenotype, however, was highly dependent on initial cell density. At a more dilute, 100-fold subculture, E. cloacae exhibited a longer, but variable apparent lag phase lasting approximately 10 +/− 5 hours (>6 biological replicates; one example shown in Fig. 1B), followed by exponential growth (Fig 1B). In contrast, the laboratory E. coli strain DH5α (a strain commonly used in AMP studies [49]) failed to grow in the presence of cecB, despite its MIC being in essential agreement with the E. cloacae MIC. We next determined viability in cecB-treated cultures. Viable cell counts in the 1:10 dilution dropped up to 10-fold upon exposure to 2x MIC cecB (Fig. S1A), before recovering, consistent with our OD₆₀₀ measurements. In contrast, upon 1:100 back-dilution, viable cell count dropped ~1000 fold after exposure to cecB before recovering to control levels over a 24-hour period (Fig. 1C). This was also reflected in MIC measurements where, consistent with previous studies [50–52], we observed a marked inoculum effect, both in LB medium and the standard clinical microbiology medium MHB (Fig. 1D). In the following, “fold-MIC” always refers to the 1000-fold dilution condition in LB. Taken together, our data show that E. cloacae exhibits inoculum-dependent killing by an AMP, and a complex dynamic of killing vs. regrowth.

Fig 1. E. cloacae displays inoculum-dependent heteroresistance to cecB.

Overnight cultures were diluted (A) 10-fold or (B) 100-fold into fresh LB containing 20 μg/mL cecB, and cells were grown at 37°C for 24 hours. OD₆₀₀ measurements were taken every 5 minutes. Error bars represent standard deviation (n=3). (C) Cultures were prepared as in (B), but samples were taken at indicated timepoints to quantify colony forming units (CFU) per mL. Error bars represent standard deviation (n=3). (D) Inoculum effect in different growth media. MIC assays were conducted at the indicated dilution of an overnight culture of E. cloacae.

Delayed growth is due to initial heteroresistance within a population

To dissect the intriguing observation of delayed growth at high dilution further, we first asked whether regrowth reflected stable genetic suppressor mutations. To this end, we sampled cultures of E. cloacae that had been exposed to 2x MIC at the endpoint (overnight post-exposure regrowth) and repeated the growth curve experiment at a 100-fold back dilution with another dose of 2x MIC cecropin (Fig S2A). Time to growth was remarkably variable, even across technical replicates of the same biological sample, ranging from 8 to 18 hours. More importantly, however, none of the re-exposed cultures grew immediately upon renewed exposure to cecB, excluding stable resistance as an explanation for the regrowth. Instead, the delayed apparent cecB resistance is transient and stochastic. These phenotypes are reminiscent of “heteroresistance”, high and transient resistance of a small subpopulation against clinically used AMPs such as the polymyxins [45, 53], and also reminiscent of the phenomenologically similar “dynamic persistence” in response to antibiotics, where apparent population stability is mediated by an equilibrium of growth and death of bacterial subpopulations [54].

To better understand the dynamics of E. cloacae growth in cecB within a population, we turned to single-cell analysis conducted via phase-contrast microscopy. Cells were grown to mid-exponential phase, then placed on 0.8% agarose LB pads containing cecB and incubated at 37°C while images were acquired. At 2x MIC, 39% of cells immediately began dividing, while 61% immediately lysed (Fig 2A and 2B). For comparison, 97% of cells grew when no drug was present. At 1x MIC, 89% of cells immediately divided, and at 20x MIC, 0% of cells divided. Thus, E. cloacae populations contain a phenotypically resistant subpopulation of cells that can survive lower cecB concentrations but are killed at higher concentrations. Note that this is different from classical heteroresistance to the therapeutic AMP colistin, where a subpopulation is transiently resistant to even high concentrations of the antimicrobial [53, 55]. To evaluate the concentration-dependence more quantitatively, we next conducted lysis experiments by growing E. cloacae to mid-exponential phase followed by addition of increasing concentrations of cecB. We observed increasingly delayed growth with increasing cecB concentrations, with near complete lack of recovery at the highest concentrations (Fig. S2B). Taken together, these data strongly suggest that regrowth observed in our killing experiments is caused by initial cecB heteroresistance/dynamic persistence (cycles of growth and death) but not primarily by hypertolerant persister cells that are refractory to antimicrobial killing independent of compound concentration [56–58]).

Fig 2. E. cloacae exhibits population heteroresistance to cecropin B.

(A) Cells were grown to mid-exponential phase in liquid medium and then placed on 0.8% agarose pads containing the indicated amounts of cecropin and incubated at 37°C. Representative, cropped frames from a time series are shown. In the 20 μg/mL condition, red arrows indicate examples of lysing cells. (B) Percentage of cells that divided during the time lapse, grouped by drug concentration (0, 10, 20, 200 μg/mL). Cells are binned by birth frame, i.e., the frame of the time lapse during which the cell was first identified by SuperSegger [82]. Blue represents cells that underwent a successful division event throughout the course of the timelapse, red indicates cells which did not divide. Missing bars indicate that the initial population of cells was unable to divide.

We then asked whether regrowth might be caused by a transient physiological adaptation to the AMP, e.g. via induced OM modifications. If so, we would predict that the emerging growing population should be more refractory to subsequent exposure. To test this, we grew E. cloacae to mid-exponential phase and exposed the population to 20 µg/mL cecB (2x MIC) at OD₆₀₀ = 0.5. As in previous experiments, the culture density decreased at first (indicating lysis), followed by regrowth. We then exposed the culture to the same concentration of cecB again after reaching OD₆₀₀ = 0.5. Both lysis (Fig. S2C) and survival after 1 hour (Fig. S2D) were similar at both exposures, suggesting that cells surviving the first cecB exposure were not primed to survive second exposure.

PhoPQ partially contributes to cecB resistance

To investigate the molecular mechanisms underlying heteroresistance in E. cloacae, we took a targeted genetic approach. The PhoPQ two-component system is well-characterized as a regulator of AMP resistance mechanisms in Enterobacterales and is essential for colistin heteroresistance [53]. In short, the sensor kinase PhoQ autophosphorylates in response to AMPs and other stimuli (Mg²⁺ deficiency, osmotic stress, pH) [59], resulting in phosphorylation of the response regulator PhoP. Phosphorylated PhoP then upregulates expression of many genes, including those encoding enzymes associated with outer membrane modifications (primarily L-amino arabinose in E. cloacae) that promote AMP resistance [53]. To test the role of PhoPQ in cecB heteroresistance, we constructed a ∆phoPQ mutant and tested its resistance to cecB. The ∆phoPQ cecB MIC was in essential agreement (2-fold lower) with WT. We then tested the ∆phoPQ mutant in a growth curve assay at 2x WT MIC. At a 10-fold back dilution, the ∆phoPQ mutant population exhibited delayed growth (preceded by partial lysis) compared to WT (Fig. 3A), taking 3 hours to enter exponential growth. At a 100-fold dilution, this difference was exacerbated to 8 hours after WT entered exponential growth (Fig. 3B). Complementation of phoPQ via plasmid-based overexpression restored resistance to WT levels (Fig 3A and 3B). Viable cell counts of ∆phoPQ (when diluted 100-fold into 2x WT MIC cecB) indicated more severe killing than WT initially, but the ability to recover by 24 hr was unaffected (Fig S1A), consistent with OD₆₀₀ measurements (Fig. 3B). Interestingly, we observed a marked condition-dependence of survival dynamics in the mutant. While the ∆phoPQ mutant grew at 2x WT MIC after just a few hours delay in a growth curve assay in liquid culture, exposure to 1x WT MIC resulted in complete lack of growth on an agarose pad, with some cells becoming slightly phase light (indicative of cell envelope integrity failures), while others stayed apparently intact, but did not grow (Fig. 3C). In either liquid culture or agarose pads, however, the ∆phoPQ mutant was more cecB sensitive than the WT. These results are consistent with the well-established view that PhoPQ acts as an important AMP resistance mechanism.

Fig 3. PhoPQ partially contributes to cecB heteroresistance.

Overnight cultures were diluted (A) 10-fold or (B) 100-fold into fresh LB containing 20 μg/mL cecB, and cells were grown at 37°C for 24 hours. For the complement strain, expression of phoPQ from pMMB was induced with 200 μM IPTG. OD₆₀₀ measurements were taken every 5 minutes. Error bars represent standard deviation (n=3). (C) Mid-exponential phase cells were placed on 0.8% agarose pads containing the indicated amounts of cecropin and incubated at 37°C. Representative, cropped time frames are shown. Red arrow indicates phase-light cell, white arrow indicates a cell that stays intact.

The Rcs phosphorelay partially contributes to AMP resistance

Since we still observed heteroresistance (albeit at a reduced level) in the ∆phoPQ mutant, we explored additional putative AMP resistance factors as the underlying cause of heteroresistance. The Rcs (regulator of capsule synthesis) system is a cell envelope stress response system present across Enterobacterales. The system comprises the RcsFCDB phosphorelay, which responds to cell envelope stress and effects its response through the transcriptional regulator RcsB and, as one of its main functions, upregulates capsule synthesis [60]. AMPs induce the Rcs system [61], and capsule has been implicated in AMP resistance in other bacteria [41], though the extent to which Rcs contributes to AMP resistance is largely unknown. We therefore constructed a ∆rcsB mutant and conducted time-dependent killing experiments with cecB. At 2x WT MIC, the ∆rcsB mutant exhibited reduced heteroresistance compared to WT, undergoing initial lysis, followed by growth after 5 hours +/− 2 (10-fold dilution), or 20 hours +/− 4 (100-fold dilution), respectively (Fig 4A and 4B). Once again, viable cell counts were consistent with this observation (Fig S1B). In timelapse microscopy, ∆rcsB was completely unable to grow at 2x WT MIC, a concentration at which 39% of WT cells grow. At 1x WT MIC, 94% of ∆rcsB cells grew (Fig 4C). Taken together, our data suggest that the Rcs phosphorelay contributes substantially to cecB heteroresistance in E. cloacae.

Fig 4. The Rcs phosphorelay partially contributes to heteroresistance.

Overnight cultures were diluted (A) 10-fold or (B) 100-fold into fresh LB containing 20 μg/mL cecB, and cells were grown at 37°C for 24 hours. OD₆₀₀ measurements were taken every 5 minutes. Bars at each point represent standard deviation (n=3). (C) Mid-log cells were placed on 0.8% agarose pads containing indicated amounts of cecropin and incubated at 37°C. Representative, cropped frames from a timelapse movie are shown.

OmpT protease contributes to cecB resistance via proteolytic cleavage

We then turned our attention to proteases, which are well-characterized as a mechanism of resistance against AMPs [62, 63]. We first analyzed the E. cloacae genome for homologs of proteases that have been implicated in AMP degradation. We identified homologs of Serratia marcescens prtS (73% AA identity), E. coli ompT (55% AA identity) (Fig. S3), and two homologs of the E. coli Sap system (38% and 36% AA identity). The Sap system, which imports AMPs for cytoplasmic degradation, has been implicated in AMP resistance for various Gram-negative pathogens [37, 64]. PrtS is a zinc metalloprotease that cleaves cecropin A [65], while OmpT is a well-characterized E. coli outer membrane protease, which, in addition to housekeeping functions [66], degrades the human AMP LL-37 (cathelicidin) and thereby promotes intestinal colonization [67].

To test the impact of these putative proteases on E. cloacae heteroresistance to cecB, deletion mutants were generated. Of the three single mutants, only deletion of ompT caused increased susceptibility to cecB in liquid culture at 10- and 100-fold dilutions at 2x WT MIC (Fig. 5A and S4). The ΔompT mutant underwent lysis before regrowth at 6 hours (+/− 3) at a 10-fold dilution, or 13 hours (+/− 4) at a 100-fold dilution (Fig. 5B). Overexpression of ompT in trans not only complemented the phenotype but resulted in increased resistance compared to WT, suggesting an important role for OmpT in resistance to cecB (Fig 5A and 5B). In single cell analysis at 1x WT MIC, 40% of ∆ompT mutant cells grew while the remaining cells in the population lysed. Notably, descendants of cells that initially displayed resistance frequently lysed at later timepoints (Fig. 5C), consistent with the notion that cecropin heteroresistance is a transient state and not a defined, homogeneous phenotype across a population.

Fig 5. OmpT contributes to heteroresistance via proteolytic cleavage of cecB.

Overnight cultures were diluted (A) 10-fold or (B) 100-fold into fresh LB containing 20 μg/mL cecB, and cells were grown at 37°C for 24 hours. Where applicable, expression of ompT from pBAD was induced with 0.1% arabinose. OD₆₀₀ measurements were taken every 5 minutes. Error bars represent standard deviation (n=3). (C) Mid-exponential cells were placed on 0.8% agarose pads containing the indicated amounts of cecropin and incubated at 37°C. Representative, cropped frames from a timelapse movie are shown. Red arrow indicates a cell which initially divided, but later succumbed to cecropin and lysed. (D) Effects of OmpT active site mutations were tested using conditions in (A).

We next aimed to disrupt the proteolytic function of OmpT to ask if the predicted active site was required for OmpT-mediated cecB heteroresistance. Four amino acid residues, a His-Asp and Asp-Asp couple, have been implicated in the OmpT active site in E. coli [68]. These residues are conserved in E. cloacae OmpT (D104, D106, D226, and H228). Due to the close proximity of residues, two amino acid substitution mutant strains were created: OmpT^D104A/D106A and OmpT^D226A/H228A. Disruption of either AA couple abrogated the ability of plasmid-borne OmpT to rescue cecB sensitivity of the ∆ompT mutant (Fig 5D), suggesting OmpT proteolytic activity is important for E. cloacae heteroresistance.

We were intrigued by the involvement of OmpT, since E. coli DH5α encodes an OmpT homolog (50% AA identity) but was not resistant against cecB. We thus tested whether E. cloacae ompT expression conferred cecB resistance to DH5α. OmpT_Ecl overexpression was sufficient to promote resistance to cecropin in DH5α, which otherwise is completely susceptible to 2x WT MIC levels of cecropin (Fig. S5). This suggests that E. cloacae’s native OmpT cleaves cecB while E. coli’s shows reduced or no activity against cecB, at least at native expression levels.

PhoPQ, Rcs, and OmpT are collectively required for cecB resistance

Individual deletions of phoPQ, rcsB, or ompT resulted in a significant increase in susceptibility to cecB; however, all mutants still exhibited (albeit, reduced) apparent heteroresistance around WT MIC concentrations. Notably, repassaging the heteroresistant mutant populations into fresh cecB resulted in sensitivity comparable to the initial exposure in all mutants, confirming that the populations do not harbor stable genetic mutations and suggesting that cecropin may have been depleted from the media (Fig. S6). To test the additive effect of these deletions, double and triple knockout strains were constructed, starting with a ∆phoPQ background. Upon exposure to 2x WT MIC at a 10-fold or 100-fold dilution, the ∆phoPQ ∆ompT double mutant had a more pronounced growth delay than either single mutant (Fig. S7A), suggesting these mechanisms independently promote cecropin heteroresistance. Similarly, a ∆phoPQ ∆rcsB double mutant had a more pronounced growth delay than either ∆phoPQ or ∆rcsB mutants (Fig. S7A). It should be noted that some replicates of ∆phoPQ ∆rcsB failed to grow at all in the presence of cecB, but this was not frequently observed (Fig. S7B).

In contrast, deletion of PhoPQ, OmpT, and RcsB (∆phoPQ ∆ompT ∆rcsB, “Δ3”) resulted in near complete eradication by cecropin at both 10-fold and 100-fold back dilution (Fig. 6A and 6B), with no full regrowth at least up to 24 hours. In almost all experiments, the triple mutant was completed killed by the AMP (Fig. 6C), though we did note a small surviving fraction in one experiment (perhaps indicating that some rare AMP persisters exist [69]). In addition, single cell analysis revealed that Δ3 was completely lysed by cecropin at 1x WT MIC (Fig. 6C and 6D), though at this lower concentration, we observed initial attempts at a few divisions on the agarose pad before lysis. Despite this dramatic phenotype, however, we still observed an inoculum effect in MIC measurements for ∆3 – while the absolute values for MICs were lower than for WT, the dynamic relationship between inoculum size and MIC was preserved. In aggregate, our data show that all three AMP resistance mechanisms collectively contribute to cecB resistance in E. cloacae and inhibition of all 3 AMP resistance mechanisms could lead to almost total elimination of heteroresistance.

Fig 6. PhoPQ, Rcs, and OmpT are collectively required for heteroresistance.

Overnight cultures were diluted (A) 10-fold into fresh LB containing 20 μg/mL cecB, and cells were grown at 37°C for 24 hours. OD₆₀₀ measurements were taken every 5 minutes. Bars at each point represent standard deviation (n=3). (B) Cultures were diluted 100fold into fresh medium with or without cecropin; samples were taken at indicated timepoints to quantify colony forming units (CFU) per mL. Error bars represent standard deviation (n=3). (C) Mid-exponential phase cells were placed on 0.8% agarose pads containing indicated amounts of cecropin and incubated at 37°C. Representative, cropped time frames are shown. (D) Percentage of cells that divided during the time lapse, grouped by drug concentration (0, 10, 20, 200 μg/mL). Cells are binned by birth frame, i.e., the frame of the time lapse during which the cell was first identified by SuperSegger [82]. Blue represents cells that underwent a successful division event throughout the course of the timelapse, red indicates cells which did not divide. Missing bars indicate that the initial population of cells was unable to divide.

Discussion

As pathogens continue to develop resistance against antibiotics at an alarming rate, AMPs have come into focus as potentially safe and effective therapeutics with which to treat MDR infections [70–73]. With many AMPs undergoing clinical trials, it is imperative to determine how bacteria sense and respond to AMP treatment, potentially enabling the development of pre-emptive measures against inevitable resistance development. While classical antibiotic resistance can often be pinned to a single genetic factor (e.g., target modification [74], drug inactivation [75] and efflux [76]), we demonstrate that AMP resistance can be far more nuanced. In the case of cecropin, we show that multiple resistance strategies are deployed by Enterobacter cloacae – outer membrane modifications, protease production, and membrane stress response collectively maximize E. cloacae’s survival in the presence of cecB (Fig 7A). Thus, E. cloacae is not entirely dependent on any one of these factors to survive cecropin exposure, and, in liquid culture, presence of just one was sufficient to promote resistance. Since cecropin mimics the mechanism of action (LPS disruption) of many human AMPs, it is likely that the resistance mechanisms studied here have broad implications for pathogenesis.

Fig 7. Models of Enterobacter cloacae cecropin B heteroresistance.

(A) Mechanisms of AMP heteroresistance discussed in this study. 1) OM stress induces the Rcs phosphorelay, which phosphorylates the response regulator RcsB. Among other systems, RcsB upregulates capsule production. 2) AMPs and envelope stress induces the PhoPQ signaling system, which upregulates transcription of enzymes which modify the outer membrane. 3) OmpT, bound in the outer membrane, may be able to cleave AMPs before they act on the cell surface. (B) Model of E. cloacae growth vs concentration of cecB over time. Black line represents hypothetical population density (indicated by OD₆₀₀), solid red line represents the hypothetical concentration of cecropin in the sample, and the dotted red line represents the hypothetical concentration of cecropin necessary for lysis. In this model, the population must bring the amount of cecropin below the effective killing concentration before all cells in the population are lysed. Factors like initial starting concentration or protease production may modify the rate at which bacteria can reduce cecropin concentration, while other factors like PhoPQ and RcsB may raise the concentration of cecropin necessary for lysis.

We also observed heteroresistance to cecropin. Resistance was unstable, with populations rapidly reverting to initial cecB sensitivity upon additional doses of drug. One potential explanation for lack of stable cecropin resistance is the formation of unknown tandem gene amplifications, which are intrinsically unstable and thus transient [77]. More likely, the phenotype we observed could be due to stochastic variance of gene expression in cells, even across a genetically identical population. For example, if 50% of cells in the population are expressing an initial “optimal” level of OmpT or PhoPQ-regulated genes, they will survive and divide while the remaining cells die. Without antimicrobial pressure, heterogeneity of gene expression may return to the population. This is supported by the fact we observed individual cells in populations of E. cloacae displaying seemingly stochastic and divergent outcomes. Regardless of the mechanism, this striking population heterogeneity speaks to the importance of single-cell analysis when investigating bacterial susceptibility to AMPs.

It is also intriguing that even a 10-fold difference in initial inoculum resulted in drastically different outcomes at the population level. This has been demonstrated previously with the conclusion that, as a universal property of AMPs, a certain AMP/cell ratio must be reached for killing to occur [52]. This is particularly important for AMPs compared to classical antibiotics, given that AMPs have a very low range of intermediate efficacy [7]. In other words, AMPs will have little effect on a bacterial cell up until reaching an AMP/cell ratio where complete killing occurs. Given this narrow range, it is plausible that 10 times more cells in a population affords a tremendous growth advantage to E. cloacae treated with cecropin. The dynamics of this ratio become more complex when considering an actively growing cell. For instance, as AMPs are binding to a cell and nearing the critical killing threshold, what if the cell completes its division? Suddenly, the AMP/cell ratio has been halved (at least in a simple model; heterogeneous distribution of surface-bound AMPs among daughter cells is also conceivable) – maximal AMP killing may thus be a function of both AMP concentration and bacterial division rate. In this way, the binding kinetics of a particular AMP may be vital in its ability to kill a population of bacteria. Further, dead cells within a population can play a protective role to growing and dividing cells; the membranes of lysed cells can absorb AMPs which could otherwise act on growing bacteria [78, 79]. We propose that each of the genetic factors we describe in this study help E. cloacae to raise the AMP/cell ratio necessary to lyse a cell. In WT, for instance, although some cells will initially die, the vast majority of cells in the population will not be bound by enough cecB to lyse, due to degradation of some surface-bound cecB (via OmpT), reduction of target binding (through PhoPQ-mediated OM modifications), and/or delay in binding due to (putatively) capsule production (induced by Rcs). However, even at higher AMP concentrations, all it takes is for a few cells to win the initial race of division vs AMP binding for the population to eventually recover, if the AMP concentration sinks to growth-permissive levels in the meantime. Sequestration by dead cells or slow degradation by another protease therefore likely contributes to allowing regrowth to full density after 24 hours. We propose a model wherein the eventual survival of an E. cloacae population is dependent on the ability of the population to lower the cecB concentration to beneath its effective killing threshold (Fig. 7B). In the case of our Δ3 mutant, the population lyses completely before cecropin levels have been sufficiently lowered. This model would also explain why the resulting population is sensitive to cecropin if exposed again, because no individual cell in the population exhibits outright resistance. This model is also consistent with our observation that the ∆3 mutant still exhibits an inoculum effect, albeit at lower AMP concentration.

AMPs are incredibly diverse in structure, mode of action, and efficacy against human pathogens [11]. Specific AMP resistance mechanisms may therefore not be equivalent across bacteria. For instance, DH5α encodes its own homologue of OmpT, which appeared to be completely ineffective in combating cecB compared to E. cloacae’s OmpT (at least at native expression levels). Indeed, the surprising diversity of OmpT-mediated AMPs degradation, even within the same species, has been noted before in different strains of E. coli [67].

Altogether, this work demonstrates how multiple genetic factors can independently contribute to AMP heteroresistance in an opportunistic pathogen. This supports the notion that transient, stochastic heteroresistance may have an impact on infection outcomes. Further study will be necessary to parse out the fine mechanisms which govern the ever-evolving battle between host immunity and pathogenic invaders.

Methods

Bacterial strains and growth

All strains, plasmids, and primers used in this study are listed in Table 1 and 2. We used the Enterobacter cloacae type strain ATCC13047 and its derivatives for all experiments. All strains were initially grown from freezer stocks on solid Luria-Bertani (LB) agar at 37°C. Isolated colonies were used to inoculate LB at 37°C. Where applicable, kanamycin was used at 50 μg/mL, chloramphenicol at 100 μg/mL, and cecropin B (MedChemExpress, Monmouth Junction, NJ) at 20 μg/mL unless otherwise noted. For complementation experiments, 0.2% arabinose or 200 μM Isopropyl-ß-D-thiogalactopyranoside (IPTG) was used to induce plasmid-borne genes.

Mutant construction

The ompT, rcsB, sapA, and prtS mutants were constructed using the suicide vector pTOX5 as described in [80]. ~500 bp upstream and downstream flanking homology regions were amplified from ATCC13047 using primers ANM28/29 and ANM30/31 for ompT, ANM78/79 and ANM80/81 for rcsB, ANM16/17 and ANM18/19 for sapA, and ANM42/43 and ANM44/45 for prtS. For a given gene, flanking homology regions were then cloned into pTOX5 digested with EcoRV using isothermal assembly [81]. Successful constructs were then transformed into E. coli donor strain MFDλpir and subsequently conjugated into ATCC13047. Recombinants were selected on LB plates containing 1% glucose and 100 µg/mL chloramphenicol. Upon single colony purification, colonies were directly streaked out on an M9 minimal medium plate containing 0.2% casamino acids and 1% rhamnose, followed by incubation at 37°C for 24 hours. Mutants were then tested with flanking primers for each gene (ompT, ANM26/27; rcsB, ANM86/87; sapA, ANM14/15; and prtS, ANM32/33).

Amino acids substitutions in ompT were generated using the Q5^® Site-Directed Mutagenesis Kit (New England Biolabs, Ipswich, MA). Briefly, primers containing the desired mutations were designed (D104A-D106A, ANM146/147; D226A-H228A, ANM148/149) to anneal to ompT cloned onto the pBAD33 vector and amplify the entire construct. Kinase, Ligase, and DpnI (KLD) reaction mix was then used to phosphorylate the ends of the PCR products and ligate them. The resulting mutant ompT constructs were sequenced using pBAD flanking primers (TC34/35) and electroporated into ΔompT.

Growth curve experiments

Growth curve experiments were conducted using 100-well honeycomb plates in a Bioscreen C growth curve analyzer (Growth Curves USA, Piscataway NJ). Overnight cultures grown at 37°C (shaking 200 rpm) were diluted 10- or 100-fold into fresh LB medium containing cecropin B (20 µg/mL, 2x MIC) and transferred to honeycomb plates (200 µL volume/culture). Cultures were then grown at 37°C OD₆₀₀ was measured automatically by the plate reader every 10 minutes for 24 hours.

Cecropin killing experiments

Overnight cultures were grown at 37°C shaking (200 rpm). For the experiment, cultures were then diluted 10 or 100-fold into fresh LB medium containing cecropin B (20 µg/mL, 2x MIC) in 1.5 mL microfuge tubes. Subcultures were incubated in a 37°C standing incubator. At indicated timepoints, a 50 µL aliquot was removed from the sample and serially diluted to determine CFU/mL.

MIC assays

Cultures were grown overnight at 37°C shaking, then diluted 1000-fold into fresh LB. Subcultures were grown for 1 hour at 37°C shaking before being diluted 1000-fold again into fresh LB to create a “seed culture”. One-hundred microliters of seed culture was subsequently diluted 2-fold into a 96-well plate containing cecropin B concentrations ranging 0.13 – 64 μg/mL. Reported values are medians of 4 technical replicates.

Microscopy

Cells were harvested in mid-exponential phase and spotted on a 0.8% agarose pad containing LB and indicated concentrations of cecropin B. Cells were kept at 37°C while being imaged on a Leica Dmi8 inverted microscope. Images were acquired every 10 minutes. Time lapse images were analyzed using SuperSegger [82] and custom R scripts. Errors were excluded from the analysis by manual curation.

Supplementary Material

Supplement 1

Figure S1. PhoPQ, RcsB, and OmpT each promote E. cloacae resistance to cecropin. Overnight cultures of WT and (A) ΔphoPQ (B) ΔrcsB or (C) ΔompT were diluted 10-fold into fresh LB containing 20 μg/mL cecB, and cells were grown at 37°C for 24 hours. Samples were taken at select timepoints to quantify colony forming units (CFU) per mL. Error bars represent standard deviation (n=3). In (C), individual replicates of ΔompT +cecB are shown due to high variance at the 3-hour timepoint.

Figure S2. (A) WT cells from three independent wells in Fig. 1A were sampled after 24 hours of growth in the presence of cecropin, grown overnight in LB, then re-exposed to 20 μg/ml cecB after 100fold dilution. Each of the 3 biological replicates is represented by a different color (black, red, blue), and each color contains 18 technical replicates. (B) Concentration-dependent growth and lysis. Overnight cultures were diluted 100fold into fresh medium containing the indicated concentration of cecB, growth (OD₆₀₀) was measured in a plate reader. (C) Overnight cultures were diluted 100fold into fresh medium, grown to OD = 0.5, then cec B (2 x MIC) was added. Following regrowth to the same OD, cecB was again added at the same concentration (D) Cultures were treated as described in (C), but plated for CFU/mL after 1 hour of exposure to cecB.

Figure S3. The structure of E. cloacae OmpT is conserved with E. coli OmpT.

The crystal structure of E. coli OmpT (blue), overlaid with an AlphaFold prediction of the homologous OmpT we identified in E. cloacae (red). Pairwise amino acid sequence alignment revealed 50% identity.

Figure S4. Homologues of the Sap system and protease PrtS do not promote cecropin resistance in E. cloacae.

Overnight cultures were diluted 10-fold into fresh LB containing 20 μg/mL cecB, and cells were grown at 37°C for 24 hours. Error bars represent standard deviation (n=3).

Figure S5. E. cloacae OmpT confers cecropin resistance to DH5α.

Overnight cultures were diluted 10-fold into fresh LB containing 20 μg/mL cecB, and cells were grown at 37°C for 24 hours. Error bars represent standard deviation (n=3).

Figure S6. Heteroresistance in mutant populations is unstable.

ΔphoPQ, ΔrcsB, and ΔompT were grown at 37°C in the presence of 20 μg/mL cecropin. After 24 hours, surviving cells were passaged ON in LB, then diluted 100-fold and re-exposed to 20 μg/mL cecropin. Cells were grown at 37°C for 24 more hours. Error bars represent standard deviation (n=3).

Figure S7. Deletion of ompT or rcsB in a ΔphoPQ background exacerbates cecropin sensitivity.

Overnight cultures were diluted 10-fold into fresh LB containing 20 μg/mL cecB, and cells were grown at 37°C for 24 hours. Error bars represent standard deviation (n=3). (A) Representative growth curve of ΔphoPQΔrcsB and ΔphoPQΔompT response to cecB. (B) Example of replicate in which ΔphoPQΔrcsB mutant failed to grow in the presence of cecB.

Figure S8 Inoculum effect in the ∆3 mutant

MIC assays were conducted at the indicated dilution of an overnight culture of E. cloacae ∆3 mutant.