Collective Resistance in Microbial Communities by Intracellular Antibiotic Deactivation

1Antibiotics are indispensable for fighting bacterial infections. Yet the rapid emergence of resistance during the last decades renders current drugs increasingly ineffective and poses a serious threat to human health. Drug action and bacterial resistance mechanisms are well understood in population assays of isogenic cultures in vitro. However, ecological factors and cell physiological parameters in natural environments influence the impact of antibiotics. Streptococcus pneumoniae (pneumococcus) is an important human pathogen that resides in complex and dynamic host environments. The bacterium primarily populates the nasopharynx of healthy individuals, together with numerous commensal microbiota, and often alongside disease-associated species, including Staphylococcus aureus, Moraxella catarrhalis, and Haemophilus influenzae.

2While an individual pneumococcal cell competes for limited resources with all other bacteria present in the niche, it may also benefit from a community setting. In a collective effort, bacteria become recalcitrant to antibiotics when forming biofilms that represent a physical constraint for drug accessibility. Additional population-based survival strategies involve the phenotypic diversification of an isogenic population, either to preadapt for environmental changes (bet-hedging) or to enable division of labor. Because the impact of most antibiotics is growth rate dependent, a bifurcation into growing and nongrowing cells increases the drug tolerance for the latter fraction, commonly referred to as persisters. Cell-to-cell communication represents another way to react to antibiotic inhibition by allowing bacteria to coordinate a common response; S. pneumoniae, for example, activates the developmental process of competence whereupon it may acquire resistance. A quorum-sensing mechanism that compromises antibiotic effectiveness was also found in evolved Escherichia coli cultures, in which cells of increased resistance induce drug efflux pumps in susceptible cells via the signaling molecule indole.

3As an alternative to reduced drug susceptibility, bacteria can also clear lethal doses of antibiotics from their environment. High cell densities and thus the presence of many drug target sites may be sufficient to lower the concentration of active compound by titration of free drug molecules. Furthermore, antibiotic degradation via β-lactamase enables growth not only of resistant cells but also of susceptible cells in their vicinity, even across species, as demonstrated for amoxicillin-resistant H. influenzae and susceptible S. pneumoniae. This mechanism is of direct relevance to clinical medicine and is alternatively referred to as passive or indirect resistance (from the perspective of susceptible cells) or collective resistance (from the perspective of mixed populations).

4Here, we describe another mechanism by which bacteria survive antibiotic therapy without obtaining genetic resistance, with the example of the bacteriostatic antibiotic chloramphenicol (Cm) and the opportunistic human pathogen S. pneumoniae. We show that Cm-resistant pneumococci expressing the resistance factor Cm acetyltransferase (CAT) can provide passive resistance for Cm-susceptible pneumococci by intracellular antibiotic deactivation. CAT covalently attaches an acetyl group from acetyl coenzyme A (acetyl-CoA) to Cm and thus prevents the drug from binding to bacterial ribosomes. Intracellular CAT in resistant bacteria can potently detoxify an entire environment in growth culture, semisolid surfaces of microscopy slides, or in a mouse infection model, supporting the survival and growth of genetically susceptible bacteria in the presence of initially effective Cm concentrations. Our results expand recent findings on the basis of E. coli growth cultures and indicate a potential clinical relevance of passive Cm resistance.

5Resistances to all currently prescribed antibiotics have been identified in clinical isolate strains of S. pneumoniae. Genes that transfer antibiotic resistance can be classified according to their mode of action. One class keeps the cytoplasmic drug level low by preventing drug entry or by exporting drug molecules. Another class alters the targeted enzymes by modifying their drug binding sites or by replacing the entire functional unit. A third class alters the drug molecules themselves. Only members of the latter group are potential candidates for establishing passive resistance. In the pneumococcus, resistance genes that deactivate antibiotics include aminoglycoside phosphor- or acetyltransferases and cat. To date, β-lactam antibiotic-degrading enzymes have not been reported in S. pneumoniae genomes or plasmids.

6Standard therapy of pneumococcal infections does not include aminoglycosides because of the relatively high intrinsic resistance of S. pneumoniae to members of this antibiotic family. In contrast, Cm, a member of the World Health Organization Model List of Essential Medicines, is regularly prescribed throughout low-income countries for infections with S. pneumoniae and other Gram-positive pathogens due to its broad spectrum, oral availability, and excellent tissue distribution, including the central nervous system. Recently, the antibiotic was also discussed as candidate for a comeback in developed nations due to spreading resistances against first-line agents.

7To test whether passive resistance emerges from antibiotic-deactivating resistance markers with S. pneumoniae, we used the drug-susceptible clinical isolate D39. We constructed an antibiotic-susceptible reporter strain expressing firefly luciferase (luc) and antibiotic-resistant strains expressing single-copy genomic integrated kanamycin 3′-phosphotransferase (aphA1), gentamicin 3′-acetyltransferase (aacC1), and chloramphenicol acetyltransferase (cat). Resistant and susceptible cells were grown at a one-to-one ratio, and optical density (both strains) and bioluminescence (emitted by susceptible cells only) were measured (Fig 1). Expression of cat, but not aphA1 or aacCI, conferred passive resistance to susceptible cells (as observed by increased luminescence in mixed populations compared with assays of susceptible cells only; S1 Fig), mirroring prior investigations of antibiotic deactivation by resistant isolates of S. pneumoniae. Aminoglycosides permeate the bacterial cell only at low frequency; high permeability, however, was recently shown to represent an important precondition for the establishment of passive resistance, explaining why the phenomenon could not be observed with aphA1 and aacCI expression.

8To characterize the observed Cm collective resistance in more detail, we used the Cm-susceptible strain D-PEP2K1 (from here on CmS), which constitutively expresses luc and the kanamycin resistance marker aphA1, and the Cm-resistant strain D-PEP1-pJS5 (from here on CmR), which expresses cat from plasmid pJS5 (see Methods). Luminescence allowed for the real-time estimation of growth (or inhibition) of the CmS population, and kanamycin resistance allowed for the monitoring of their viable cell count by plating assays in the presence of kanamycin. Cm represses the growth of susceptible pneumococci at a minimal inhibitory concentration (MIC) of 2.2 μg ml−1, and during Cm exposure, luminescence from luc expression of susceptible pneumococci was previously shown to decrease at a rate that depends on the applied Cm concentration. However, when CmS was co-inoculated with CAT-expressing CmR, luminescence (indicative for growth or inhibition of the CmS cell fraction) recovered, both for a Cm concentration slightly above the MIC (3 μg ml−1; Fig 2A) and even for a Cm concentration of more than two times the MIC (5 μg ml−1; S2 Fig). Luminescence recovery in mixed population assays (CmR + CmS) exceeded the values measured with CmS monoculture by up to 10-fold (Fig 2A and S2 Fig), and plating assays (with kanamycin) revealed that the difference in viable cell count was 1,000-fold greater after 8 h of cocultivation (Fig 2B and S2 Fig). Although Cm is commonly regarded as bacteriostatic, bactericidal activity has also been demonstrated against S. pneumoniae, explaining the observed decrease in viability of CmS monoculture (Fig 2B and S2 Fig).

9To confirm that CmR cells actually deactivate Cm in the growth medium, we analyzed culture supernatant (S) by high-performance liquid chromatography (HPLC). As shown in Fig 2C, within 4 h of growth, CmR cells entirely converted an initial Cm concentration of 5 μg ml−1, as evidenced by the disappearance of the corresponding Cm peak at wavelength 278 nm. New peaks (at later elution times) appeared and gradually increased in HPLC profiles of S collected after 1, 2, and 4 h of cultivation; these peaks were previously shown to correspond to acetylated Cm derivates (1- and 3-acetylchloramphenicol).

10Next, we focused on whether the initial amount of CAT-expressing CmR cells was important for the survival and growth of CmS cells during drug treatment. To test this, we inoculated microtiter plate wells with a fixed number of CmS cells (inoculation at optical density [OD] 0.001, corresponding to ~1.5 × 106 colony-forming units per ml [CFUs ml−1]) while varying the number of CmR cells (Fig 2D). High inoculation densities of CmR cells (OD 0.01) resulted in a fast recovery of luminescence activity of CmS cells; however, the peak of luminescence was lower compared to intermediate CmR inoculation densities. This difference can be explained by cells reaching the carrying capacity of the growth medium before the pool of Cm is completely deactivated; luciferase expression activity was previously shown to slow down when cultures reach high cell densities (above ~OD 0.05). Relatively low CmR inoculation densities (OD 0.0001) also limited luminescence recovery of CmS cells during cocultivation. This finding likely reflects fewer CmR cells requiring more time to deactivate Cm, resulting in increased time spans of CmS drug exposure. Prolonged drug exposure of susceptible pneumococci was previously shown to result in increasing lag periods after drug removal, indicating a more severe perturbation of cell homeostasis. The time span before outgrowth of CmS cells consequently consists of both the period required for drug clearance (by CmR cells) and the period required to reestablish intracellular conditions allowing for cell division.

11To test whether Cm processing by CAT is an intracellular process, or if it takes place after secretion or cell lysis, we examined the potential of the S and the cytosolic content of CmR cells to deactivate Cm (assay scheme in Fig 2F). Precultured CmR cells were diluted to OD 0.02 and translation activity was blocked by adding 1 μg ml−1 tetracycline ([Tc]; S. pneumoniae D39 MIC: 0.26 μg ml−1) for 1 h at 37°C to prevent ongoing protein synthesis and thus CAT expression. Next, the Tc-treated culture was split into three fractions: cell pellet (P) and S, separated via centrifugation, and cell culture lysate (L), obtained by sonication. The P was resuspended in C+Y medium containing 3 μg ml−1 Cm (and 1 μg ml−1 Tc), and 3 μg ml−1 Cm was added to the S and the L, followed by incubation at 37°C. After 2 h, the remaining cells and cell debris were removed by centrifugation and filtration, and the treated medium was used to test cell growth of a Tc-resistant variant of the CmS strain. Neither the S nor the L could support growth of CmS, whereas medium preincubated with the P did (Fig 2E). Together, these experiments indicate that CAT is only active inside living cells, in which acetyl-CoA is present.

12Because the abovementioned experiments were performed in bulk assays, we wondered whether CAT-expressing bacteria would also efficiently deactivate Cm, and thus support the growth of susceptible cells, in a more complex environment, such as on semi-solid surfaces. To do so, we spotted CmR cells together with Cm-susceptible D-PEP33 cells expressing green fluorescent protein (GFP) on a matrix of 10% polyacrylamide C+Y medium containing 3 μg ml−1 Cm. Indeed Cm-susceptible D-PEP33 cells were able to grow and divide under these conditions (S3 Fig).

13S. pneumoniae cohabitates the human nasopharynx with other bacteria, such as S. aureus. Therefore, we investigated whether CAT-expressing S. aureus could also support growth of Cm-susceptible S. pneumoniae in environments containing Cm. As shown in Fig 3 and S1 Movie, all S. aureus cells grew and divided from the starting point of the experiment, whereas S. pneumoniae CmS cells did not grow initially. However, after 8 h, a fraction of CmS cells grew out to form microcolonies. Note that CmS cells spotted in the absence of S. aureus did not grow under these conditions (S2 Movie).

14The observation that CmS cells grow only when Cm-deactivating cells are present in their close vicinity (Fig 3) suggests that the establishment of collective resistance requires CmS and CmR bacteria to be present in the same niche. However, such coexistence is subject to ecological constraints (e.g., the competitive exclusion principle), particularly if susceptible and resistant strains compete for the same limiting resource. We therefore developed an ecological model to assess the scope for coexistence between CAT-producing bacteria and an antibiotic-susceptible strain (S1 Text). Consistent with this objective, we employed a minimalist modeling strategy and disentangled the qualitative effects of different factors (antibiotic stress, relative cost of Cm degradation and density regulation by ecological resource competition) from the interaction between CmS and CmR bacteria rather than aiming for a precise quantitative reconstruction of the experimental conditions. In fact, in contrast to natural environments (such as the human nasopharynx) that provide ample opportunities for coexistence because of spatial structure and concentration gradients of multiple resources, the model considers a worst-case scenario for coexistence: the two populations are assumed to grow in a well-mixed, homogeneous chemostat environment and are limited by the same resource. Nonetheless, we found that coexistence between CmR and CmS bacteria was feasible (Fig 4A and 4B), albeit under a restricted range of conditions (Fig 4C and S4 Fig). A mathematical analysis of the model (S1 Text) indicates that resistant and susceptible bacteria can establish a stable coexistence when CAT expression has a modest fitness cost. Without such a cost, the CmR strain is predicted to outcompete the CmS strain in the presence of antibiotics. Conversely, if the cost of expressing resistance is too high, the CmS strain will be the superior competitor. Interestingly, the model furthermore predicts parameter ranges that result in the extinction of mixed populations during drug treatment, while CmR populations on their own could survive (S4 and S5 Figs). A second condition for coexistence demands that the CmR population has a significant impact on the extracellular Cm concentration in its ecological niche. This requires that the population density reached at steady state must be high, so that coexistence can be stabilized by frequency-dependent selection, generated by a negative feedback loop between the relative abundance of drug-deactivating cells and the level of antibiotic stress in the environment.

15We note that competitive exclusion acts at a local scale in structured environments, where the presence of spatial gradients in Cm and resources may help to create refuges in which either strain can escape competition from the other. In addition, we expect that coexistence between resistant and susceptible bacteria would be promoted in vivo by previously evolved ecological niche partitioning between co-occurring species.

16A general prediction from our mathematical model (S1 Text) is that coexistence of CmS and CmR in the presence of the antibiotic is precluded when the production of CAT carries no fitness cost; we expect this prediction to apply likewise in more complex environments with spatial and/or temporal heterogeneity in Cm concentrations. However, in vitro, in short-term experiments, we did not observe any obvious fitness cost for CAT expression (such as reduced growth rates or a reduced maximum cell density; Fig 2). Nevertheless, in vivo, a fitness cost might come into existence because resistant cells that grow rapidly might be preferentially targeted by the host innate immune system, as previously shown for commensal and pathogenic bacteria, including E. coli and S. aureus. We tested the activity of the human antimicrobial peptide LL-37 in dependency of Cm treatment and found, indeed, increased killing efficiency against CAT-expressing S. pneumoniae (S6 Fig). Furthermore, although collective resistance could be successfully demonstrated in vitro, the phenomenon might not occur in more complex environments in vivo, such as in an animal infection model, because of a different flux balance between local Cm deactivation and restoration of effective drug concentrations via diffusion from surrounding tissues. To examine whether a coexistence between CmS and CmR is possible under therapy in vivo, we performed intratracheal infection of 8-wk-old female CD-1 mice with CmS alone and the combination of CmS + CmR.

17In the absence of Cm treatment, we observed no significant difference in the amount of viable bacteria recovered from the lungs 24 h after infection with CmS alone versus CmS + CmR at a one-to-one ratio (Fig 5A). When mice were given three doses of 75 mg kg−1 Cm once every 5 h, mice infected with CmS alone demonstrated a significant drop of one log-fold versus the untreated control. This is in contrast to mice coinfected with CmS + CmR, in which Cm treatment did not significantly reduce the number of viable bacteria recovered from the lung versus the untreated control (Fig 5A). In the one-to-one mixed infection, approximately equal numbers of CmS and CmR cells were recovered in the absence of Cm treatment: 46% CmS and 54% CmR (Fig 5B). Surprisingly, with Cm treatment, 6 out of 14 animals had a dramatic increase in the percentage of CmS cells versus CmR cells. No pneumococcal colonies recovered could grow in both Cm- and kanamycin-containing media, excluding the possibility that horizontal gene transfer of the cat gene occurred during coinfection. Together, these results show that passive Cm resistance and the coexistence of resistant and susceptible cells also occur in vivo, associated with a fitness cost to the CmR niche members benefiting the CmS subpopulation.

18This work elucidates that CAT, which is commonly found as a resistance marker in the human microbiome, can effectively protect Cm-susceptible pneumococci from the activity of the drug within local environments occupied by CAT-expressing cells. Because of its potency, long shelf life, and low cost, Cm remains a mainstay of broad-spectrum antibiotic therapy in several countries in Africa, the Indian subcontinent, and China. The rise of multidrug resistance among human pathogens has also provoked interest in reevaluating Cm for certain serious infections in developed countries. This work points out some caveats in using Cm to target human pathogens on mucosal surfaces because CAT-expressing commensals might provide passive resistance.

19CAT can only deactivate Cm inside living cells (Fig 2E and 2F), presumably because it needs acetyl-CoA to acetylate and deactivate the target drug. We show that Cm deactivation and collective resistance via CAT is not limited to S. pneumoniae, because CAT-expressing S. aureus can also support the local growth of pneumococci in the presence of initially effective Cm concentrations (Fig 3). Collective resistance by CAT does not only occur in vitro but also in vivo in a mouse pneumonia model (Fig 5). Strikingly, when Cm-treated mice were coinfected with CAT-expressing and Cm-susceptible pneumococci, the susceptible bacteria outcompeted the resistant ones (Fig 5). We previously showed that the susceptibility of bacteria towards antimicrobial peptides, produced by the host innate immune system, is markedly diminished in the presence of bacteriostatic antibiotics; Cm-inhibited E. coli, for example, are less efficiently cleared by the human peptide LL-37. We could demonstrate that this mechanism also takes place in S. pneumoniae (S6 Fig). When Cm was added to LL-37 treatment of pneumococci, the number of CmS cells recovered was one log-fold higher compared with CmR cells (S6 Fig). As shown before, this effect occurs because bacteriostatic antibiotics, such as Cm, inhibit the growth of susceptible bacteria and thereby reduce the susceptibility to host antimicrobial peptides that target bacterial division; Cm-resistant bacteria, in contrast, maintain fast growth in the presence of Cm and are therefore more rapidly killed by host antimicrobial peptides in vivo. This phenomenon may therefore represent a contributing factor underlying our findings of the mouse pneumonia model. In this framework, rapidly growing CmR cells would suffer immune clearance, while the initially nongrowing CmS are less efficiently targeted by host defense factors. Once the Cm concentration has dropped sufficiently, CmS cells can outgrow and outcompete the diminished CmR population.

20Our work with CAT and pneumococcus extends the known phenomenon of passive resistance via β-lactamase expression and expands on recent findings of collective resistance of bacterial communities. Intracellular antibiotic deactivation requires a high drug permeability, and it is worth noting that this—in general desired—drug characteristic can also represent a risk factor for the effectiveness of an antibiotic therapy. Passive resistance could also appear with other antibiotic-degrading resistance factors in other bacteria and may even emerge for synthetic antibiotic compounds. In light of numerous reports of prevalence of drug resistance in pathogens, successful antibiotic therapy might become increasingly complicated with the occurrence of collective resistance. The phenomena could furthermore give rise to multidrug resistance of bacterial communities, in which individual resistances are expressed in different bacterial community residents. Our mathematical model, however, predicts that collective resistance is only sustainable when resistance expression comes at a (modest) fitness cost (S4 Fig), and competitive exclusion is avoided by strong ecological feedback or alternative mechanisms (such as spatiotemporal structure or previously evolved niche partitioning). Nevertheless, even if coexistence is limited, the prolonged survival of susceptible cells within resistant communities may already represent an issue by increasing the opportunity for horizontal gene transfer during antibiotic selection pressure. Passive resistance might consequently represent an important factor towards the development of genetic multidrug resistance.

21S. pneumoniae CmS, a Cm-susceptible D39 derivate strain that constitutively expresses luc and a kanamycin resistance marker was used throughout. The Tc-resistant variant of this strain contained the Tc resistance gene tetM integrated at the bgaA locus, obtained via transformation with pPP1. luc has a reported half-life of 3 min in S. pneumoniae, and luminescence therefore gives real-time information on the level of gene expression activity. S. pneumoniae CmR, expressing CAT from plasmid pJS5, was used as standard for a Cm-resistant strain. Initial experiments were carried out with the Cm-resistant strain D-PEP1C3 that expresses CAT from a strong synthetic promoter. S. pneumoniae D-PEP33 expressing GFP was used as a Cm-susceptible strain in time-lapse microscopy experiments. S. aureus experiments were performed with strain LAC pCM29 that constitutively expresses CAT and GFP.

22S. pneumoniae and S. aureus cells were grown in C+Y medium (pH 6.8), supplemented with 0.5 μg ml−1 D-luciferin for luminescence measurements, at 37°C. Pre-cultures for all experiments were obtained by a standardized protocol, in which previously exponentially growing cells from −80°C stocks were diluted to OD (600 nm, path length 10 mm) 0.005 and grown until OD 0.1 in a volume of 2 ml medium inside tubes that allow for direct in-tube OD measurements. To determine the number of colony-forming units (CFUs), S. pneumoniae cells were plated inside Columbia agar supplemented with 3% (v v−1) sheep blood and incubated overnight at 37°C.

23Costar 96-well plates (white, clear bottom) with a total assay volume of 300 μl per well were inoculated to the designated starting OD value. Microtiter plate reader experiments were performed using a TECAN infinite pro 200 plate reader (Tecan Group) by measuring every 10 min with the following protocol: 5 s shaking, OD (595 nm, path length 10 mm) measurement with 25 flashes, luminescence measurement with an integration time of 1 s.

24In mixed population assays (shown in Fig 2A), all cultures were inoculated with CmS cells to an initial cell density of OD 0.001. CmR cells were inoculated to the same density, and control cultures without CmR cells contained equal amounts of Cm-sensitive D39 wild-type cells to correct for unspecific effects such as drug-titration via cellular Cm binding.

25S were obtained by CmR cultivation (inoculation at OD 0.001) in the presence of 5 μg ml−1 Cm in microtiter plates (as described above). Four wells were sampled and pooled per time point (combined volume of 1.2 ml), centrifuged to remove cells, and filtered through a 0.2 μm filter. HPLC analysis was carried out using an Agilent 1260 Infinity system (Agilent Technologies) with ultraviolet (UV) detection at 278 nm (maximum absorbance of Cm). An Aeris Peptide XB-C18 column (Phenomenex) with 3.6 μm particle and a size of 250 × 4.60 mm was used. Reversed-phase chromatography was carried out at a constant flow rate of 1 ml min−1, with the mobile phase consisting of solution A: 10 mM sodium acetate buffer (pH 6.0) containing 5% acetonitrile (v v−1) and solution B: acetonitrile 0.1% TFA, according to the following protocol: 100 μl sample loading, 3 min 10% B, 20 min gradient 10% to 50% B, 1 min gradient 50% to 95% B, 3 min 95% B, 1 min gradient 95% to 10% B, 6 min 10% B.

26A Nikon Ti-E microscope equipped with a CoolsnapHQ2 camera and an Intensilight light source was used. Time-lapse microscopy was carried out by spotting pre-cultured cells on 10% polyacrylamide slides inside a Gene Frame (Thermo Fisher Scientific) that was sealed with the cover glass to guarantee stable conditions during microscopy. The polyacrylamide slide was prepared with growth medium containing 3 μg ml−1 Cm. Images of fluorescing cells were taken with the following protocol and filter settings: 0.3 s exposure for phase contrast, 0.5 s exposure for fluorescence at 440–490 nm excitation via a dichroic mirror of 495 nm, and an emission filter of 500–550 nm. Temperature during microscopy was controlled by an Okolab climate incubator, and images were taken every 10 min during 20 h at 37°C.

27The murine pneumonia model was performed with slight modifications as previously described. Based on pilot experiments, we estimated that the number of animals required to observe a statistical difference between the groups would exceed the technical limit of animals that could be inoculated and treated per day. Therefore, the experiment was split into 2 d with the original pool of animals randomized to each group at the start of the multi-day experiment. Prior to statistical analysis, the data were combined. Note that all intratracheal infections were performed in a blinded fashion with respect to Cm or vehicle treatment. Eight-wk-old female CD1 mice (Charles River Laboratories) with an average body weight of 28 g were used. Fresh cultures of CmS and CmR were started in 10 ml of Todd-Hewitt broth containing 2% yeast extract (THY) and 10 ml of THY supplemented with 5 μg ml−1 Cm, respectively. Cultures were grown at 37°C in a 5% CO2 incubator until OD (600 nm) 0.6. Bacteria were washed twice with PBS via centrifugation at 3,220 × g at room temperature and concentrated in PBS to yield 3.5 × 107 CFU in the inoculation volume of 40 μl. For mixed infections, an equal volume of concentrated CmS and CmR pneumococci were combined. Mice were anesthetized with 100 mg kg−1 ketamine and 10 mg kg−1 xylazine. Once sedated, the vocal chords were visualized using an operating otoscope (Welch Allyn), and 40 μl of bacteria was instilled into the trachea during inspiration using a plastic gel loading pipette tip. Mice were placed on a warmed pad for recovery. After 1 h, one intraperitoneal injection of Cm 75 mg kg−1 or vehicle controls was given, followed by two additional doses spaced 5 h apart. Mice were sacrificed with CO2 24 h after infection.

28To enumerate total surviving bacteria in the lungs, both lung lobes were removed and placed in a 2 ml sterile micro tube (Sarstedt) containing 1 ml of PBS and 1 mm silica beads (Biospec). Lungs were homogenized by shaking twice at 6,000 rpm for 1 min using a MagNA Lyser (Roche), with the specimens placed on ice as soon as they were harvested. Aliquots from each tube were serially diluted for CFU enumeration on THY plates. To determine whether or not a colony was CmS or CmR, individual colonies from the THY plates were picked and transferred into 100 μl of THY media in 96-well plates. The 96-well plates were incubated overnight at 37°C in a 5% CO2 incubator. After overnight incubation, wells were mixed, and 5 μl of media from each well was transferred into 100 μl of THY containing 15 μg ml−1 Cm or 100 μg ml−1 kanamycin. The 96-well plates were once again incubated overnight at 37°C in a 5% CO2 incubator, and wells were finally assessed for the presence or absence of a bacterial P. Cm (≥98% purity; Sigma) for animal injection was prepared as follows: 40 mg ml−1 of Cm was dissolved in 800 μl of 70% ethanol in PBS to make a 50 mg ml−1 stock solution. This stock solution was diluted in PBS to 3.75 mg ml−1 for intraperitoneal injection into mice at 75 mg kg−1.

29This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The corresponding protocol entitled “Mouse Models of Bacterial Infection and Innate Immunity” (#S00227M) was approved by the Institutional Animal Care and Use Committee of the University of California, San Diego (Animal Welfare Assurance Number: A3033-01). All efforts were made to minimize suffering of animals employed in this study.

30The model describes the dynamic of a coculture of CAT-expressing CmR and CmS bacterial cells growing in the presence of Cm in a chemostat environment. The two strains, with population densities xr and xs, respectively, compete for a growth-limiting resource, z. Cm is assumed to inhibit growth; we separately keep track of the intracellular concentrations of Cm (ys in susceptible cells and yr in resistant cells) and its concentration in the extracellular medium ym. The equations for the growth of the two bacterial populations and the growth-limiting resource are given by where r is the maximum growth rate of CmS cells, η is the relative growth efficiency of CmR cells, c is the resource consumption rate, and kz and hy, respectively, are the half-saturation and inhibitory constants of the growth function. Time, resource concentration, and cell densities have been scaled relative to the flow rate of the chemostat, the resource concentration in the inflow medium, and the number of cells that fit in the chemostat volume, respectively, in order to reduce the number of free parameters (see S1 Text for details).

31The concentrations of Cm in the different compartments, which have been scaled relative to the Cm concentration in the inflow medium, change according to the equations:

32The processes described by the terms on the right-hand side include inflow of Cm into the medium, passive transport of Cm between compartments at rate p, outflow from the chemostat, degradation of Cm by CAT in CmR cells (according to Michaelis–Menten kinetics with maximum rate d and half-saturation constant ky), and concentration changes due to fluctuations in the volume of the compartments. Eqs (1) and (2) were solved numerically using Mathematica (Wolfram) or simulation software written in C++ (used for the numerical bifurcation analysis, based on a Runge–Kutta integration algorithm with adaptive step-size control).