Exploiting Cooperative Pathogen Behavior for Enhanced Antibiotic Potency: A Trojan Horse Approach

Antimicrobial resistance poses an escalating global threat, rendering traditional drug development approaches increasingly ineffective. Thus, novel alternatives to antibiotic-based therapies are needed. Exploiting pathogen cooperation as a strategy for combating resistant infections has been proposed but lacks experimental validation. Empirical findings demonstrate the successful invasion of cooperating populations by non-cooperating cheats, effectively reducing virulence in vitro and in vivo. The idea of harnessing cooperative behaviors for therapeutic benefit involves exploitation of the invasive capabilities of cheats to drive medically beneficial traits into infecting populations of cells. In this study, we employed Pseudomonas aeruginosa quorum sensing cheats to drive antibiotic sensitivity into both in vitro and in vivo resistant populations. We demonstrated the successful invasion of cheats, followed by increased antibiotic effectiveness against cheat-invaded populations, thereby establishing an experimental proof of principle for the potential application of the Trojan strategy in fighting resistant infections.


INTRODUCTION
The fight against microbial pathogens is almost completely reliant on the development of new antibiotics.However, the long-term effectiveness of this strategy is threatened due to the evolution of resistance by pathogens 1,2 .Hence, there is a growing interest in developing "evolution-proof" strategies to fight against infections [3][4][5][6] .The idea of utilizing social cheats as therapeutic agents has been proposed based on theoretical and empirical studies on cooperative virulence and cheat invasion [4][5][6] .However, experimental evidence supporting the effectiveness of cheat therapies is limited.
Bacterial virulence is a complex phenomenon that frequently relies on cooperative interactions facilitated by the release of public goods [7][8][9] .This cooperative behavior can be targeted directly, similar to conventional drug treatments 3,6 , or it can be exploited by utilizing cheats [4][5][6] .Cheats are cells that do not contribute to the production of public goods but still benefit from cooperation.Consequently, cheats can invade populations by capitalizing on their fitness advantage [9][10][11][12] .Furthermore, cheat invasion is expected to lead to a reduction of virulence due to reduced cooperation within the population 4 .

Empirical evidence of cheat invasion in Pseudomonas aeruginosa has been demonstrated
through in vitro experiments [11][12][13][14] .Moreover, studies by Harrison et al. 15 and Rumbaugh et al. 16 demonstrated the invasion potential of P. aeruginosa cheats and subsequent reduction in virulence using animal infection models.Additional evidence of cheat invasion has been shown using animal models of Salmonella enterica serovar Typhimurium and Staphylococcus aureus infections 17,18 .Expanding on these findings, cheats in pathogen populations can be further manipulated to become Trojan cheats: strains which act as vehicles for medically beneficial alleles such as antibiotic sensitivity or genes which quench quorum QS signals (Fig. 1).The concept of employing cheats as Trojan horses has gained traction over the past 15 years 4,5,16 as a way of "socially engineering" populations, similar to genetic engineering at the individual genome level.

Antibiotic-resistant populations multiply and cause severe infections (left panel). A Trojan horse approach can be used to introduce medically beneficial alleles (e.g. antibiotic sensitivity) into antibiotic-resistant pathogenic populations to reduce total bacterial load and virulence (right panel).
Why would cells fail to evolve resistance to cheat intervention strategies in the same way that they evolve resistance to antimicrobial treatments?The answer to this question is wellunderstood from evolutionary theory.The simplest way to avoid exploitation is to refrain from engaging in cooperative actions 5,19 .However, by avoiding cheating, bacterial cells become less virulent, and the overall goal is still achieved.Additionally, once invaded by cheats, cooperators are unable to invade a population of cheats because rare cooperators face a severe fitness disadvantage.These individuals bear all the costs to provide a valuable public good, while the benefits are enjoyed by neighbors who bear none of the costs 13 .Therefore, a cooperative (and consequently more virulent) cell must overcome significant evolutionary and mechanistic constraints to resist cheating.
Here we present a proof of principle for application of cheats as part of a Trojan horse strategy to drive a selected beneficial allele to populations of cooperators.We used cheats of P. aeruginosa, defective in QS signaling, to invade and introduce antibiotic sensitivity in populations of resistant cooperators to improve antibiotic effectiveness.First, we demonstrated the invasive ability of a Trojan strain in vitro using head-to-head competition assays in both liquid and viscous media (increased viscosity and spatial structuring are expected in real-life infections; and has been shown to affect cooperator cheat dynamics 20,21 ).
We then exposed the Trojan invaded populations to antibiotic selection and showed improved antibiotic efficacy.Finally, we provided evidence of a successful application of the Trojan invasion strategy to improve antibiotic efficiency using a murine chronic wound infection model.

Bacterial strains and media
For all experiments, wild type P. aeruginosa PA14 strain was used as the parental strain.The Trojan strain was an isogenic QS signal blind mutant of PA14 generated by insertion mutagenesis in the lasR locus lasR::Gm (PA14ΔlasR) 16 .A multi-drug resistance plasmid, pAMBL2 22 was used to obtain antibiotic-resistant variants.To generate resistant PA14 and PA14ΔlasR strains an electroporation protocol was used 23 .Minimum inhibitory concentrations (MICs) to kill the resulting transformants were 2048 µg/mL, 16384 µg/mL, 1365 µg/mL and 192 µg/mL for streptomycin, carbenicillin, ceftazidime and meropenem, respectively.MICs of sensitive stains were 8 µg/mL, 53 µg/mL, 1 µg/mL and 0.5 µg/mL for streptomycin, carbenicillin, ceftazidime and meropenem, respectively.
To grow cultures, cells were inoculated from frozen stocks into fresh King's broth media [KB, 20 g proteose peptone No3, 10 mL glycerol, 1.5 g K 2 HPO 4 .3H 2 O and 1.5 g MgSO 4 .7H 2 O, per liter of dH 2 O].Then cultures were incubated overnight at 37°C with 200 rpm shaking.Next, the overnight cultures were centrifugated, supernatant was discarded, and the pellet was washed twice in M9 minimal salts media [6.8 g Na 2 HPO 4 , 3 g KH 2 PO 4 , 0.5 g NaCl, and 10 g NH 4 Cl, per liter of dH 2 O] to remove residual carbon sources from the KB media.Cell density was assessed by measuring absorbance at 600 nm, and each set of cultures were standardized to the same density before the competition assays.

In vitro competition assays in defined QS media
To determine the relative fitness of cheats, a defined QS medium (QSM) was used.PA14 and Trojan strains were inoculated in a 10:1 ratio in 50 ml falcon tubes containing 2 ml QSM [1% BSA w/v and 0.1% CAA w/v in M9 solution] as described previously 11,24 .Then, culture tubes were incubated at 37°C for 24 or 48 h with shaking at 50 rpm.To control for social effects, both strains were also grown in monocultures.Relative frequencies at the start and at the end of experiments were determined by plating on LB agar [25 g LB Miller agar per liter of dH 2 O].After 24 h incubation at 37°C, plates were removed and kept at room temperature for an additional 24 h.This allows distinguishing QS cheats from cooperators based on differences in colony morphologies.While cooperators form raised colonies with fuzzy edges, lasR mutant colonies show characteristic colony flattening, sharper edges and surface iridescent sheen 25 .To change the viscosity and spatial structuring of the media, competition assays were performed in QSM supplemented with 0.5 % agar.

Antibiotic survival assays
To determine antibiotic survival of populations before or after competitions in QSM, 10 µl samples were spotted in triplicates onto LB agar plates supplemented with varying concentrations of antibiotics.The inoculated plates were incubated at 30°C for 24 h and then at room temperature for an additional 24 h, which made colony counting more reliable by preventing overgrowth.To obtain survival rate data, triplicate counts were averaged, and percentage of survival was calculated for each data point by the formula CFU(antibiotic)/CFU(no-antibiotic)*100.

Murine chronic wound model assays
The murine chronic wound model utilized in this study has been previously described 26,27 .For the experiments herein, female Swiss Webster mice were anesthetized using intraperitoneal injection of ketamine and xylazine.The backs of the mice were shaved, and a full-thickness skin excision measuring 1.5 cm in diameter was administered.To cover the wounds, semipermeable polyurethane dressings were applied, followed by the inoculation of bacteria on top of the wounds, beneath the dressings.The initial inoculation of bacteria was performed at a concentration of 10 5 CFU/mL, with strains inoculated according to experimental group.
For co-inoculated mixed infections, a Trojan to PA14 R ratio of 1:10 was used for initial inoculation.For delayed-mixed infections, PA14 R infection was allowed to establish for 72 hours.At this point, the delayed-mixed infection groups received an additional 10 6 CFU per 100 µL of Trojan strain, which corresponded to approximately 10% of the wound burden of PA14 R at the 72-hour mark post-infection.The remaining groups were sham inoculated with PBS.On day 7 post-infection, the mice were euthanized.The wound beds were harvested and subjected to ex vivo treatment with either a vehicle control (PBS) or an antibiotic solution containing 1.75 mg/mL of streptomycin and 8 mg/mL of carbenicillin.The difference in dosage compared to in vitro experiments is due to the recalcitrant nature of biofilms in wounds, which reduces the efficacy of antibiotics even on genotypically susceptible infections 28 .Following antibiotic treatment, the wound beds were washed thoroughly to prevent antibiotic carryover, resuspended in PBS, and subsequently homogenized.Serial dilutions were prepared for CFU quantification, which was calculated per gram of tissue.To determine the population compositions, each homogenized tissue sample was plated on Pseudomonas Isolation Agar (PIA) or PIA supplemented with 200 µg/ml ceftazidime, which selects for the presence of the resistance plasmid.All animal experiments were conducted in accordance with protocol 07044 approved by the Animal Care and Use Committee at Texas Tech University Health Sciences Center.

Statistics
Relative fitness (w) of the cheats, which is the change in frequency relative to the cooperators during competition assays, was obtained by w =x 2 (1-x 1 )/x 1 (1-x 2 ) where x 1 is the initial and x 2 is the final proportion of cheats in each sample 13 .A value of w > 1 indicates that the cheat has a higher, and a value of w < 1 indicates that the cheats have a lower fitness than cooperators.
Each competition assay was replicated six or eight times.All analyses were performed in R statistical software (http://www.R-project.org).

Growth and cheating in QSM
We confirmed that lasR encoded extracellular protease activity is required for optimal growth in QSM 11,24 by comparing growth of antibiotic resistant PA14 (PA14 R ) and sensitive PA14ΔlasR mutants (Trojan cheats) in QSM without antibiotic selection.After 24h incubation, monocultures of Trojan cheat were significantly lower in density than PA14 R monocultures (Fig. 2a).We obtained similar results when we cultivated antibiotic sensitive PA14 S and resistant PA14ΔlasR R strain variants in QSM (Fig. S1).
In head-to-head competition assays, when Trojan cheats were co-inoculated with PA14 R , they successfully invaded the populations, increasing in relative frequency, even when carrying the costly resistance plasmid (Fig. 2b).We did not observe any significant impact of the resistance plasmid on monoculture growth (Fig S1).These results display that lasR encoded extracellular protease activity is required for maximal growth of PA14 in QSM.More importantly, these results show the capacity of Trojan cheats to exploit cooperation and invade mixed cultures.In addition, we showed that the presence of the antibiotic resistance plasmid is costly and can influence the level of cheat invasion.

Antibiotic susceptibility of Trojan cheat invaded populations
To test the efficacy of antibiotics against Trojan invaded and uninvaded populations, we compared population survival at the beginning and at the end of competitions after exposing to varying doses of carbenicillin.We hypothesized that following the expansion of a sensitive population, Trojan invaded PA14 R populations would show increased susceptibility to antibiotics.As shown in Figure 3, our results confirmed this hypothesis; mixed cultures of PA14 R and Trojan were more susceptible than pure PA14 R monocultures.Notably, while PA14 R monocultures were able to resist carbenicillin doses up to 200 µg/ml, sensitive Trojan monocultures could not grow above doses higher than their MIC (Fig. 3a).Consequently, this led to a significant decline in survival rates for susceptible and mixed populations when exposed to carbenicillin at MIC (Fig. 3b).Moreover, post-invasion mixed cultures were even more susceptible compared to their initial state (Fig. 3b).A similar reduction in survival rates was observed when we replicated the experiments using different antibiotics to which the pAMBL2 plasmid confers resistance (Fig. S2).These findings emphasize the potential to induce antibiotic sensitivity in target populations as a result of successful Trojan cheat invasions.As a result, Trojan cheat invasion has the potential to enhance the efficacy of antibiotics.

Increased viscosity hinders Trojan cheat effectiveness.
In natural infections, where populations are likely to be more structured than in our liquid media conditions, accessibility to public goods may be limited for invading Trojan cells 20 .To assess the effectiveness of Trojan invasion in spatially structured populations, we conducted competitions in QSM supplemented with agar.In comparison to unstructured media, Trojan cheats exhibited a significantly lower relative fitness and invaded the populations to a lesser degree in viscous media (Fig. 4a).This reduced invasion observed in viscous media also translated into a decreased efficacy of antibiotic treatment.Mixed populations in viscous media were less susceptible to streptomycin treatment than mixed populations in unstructured media (Fig. 4b).Trojan strain can invade infections in vivo.
After confirming Trojan invasion in unstructured and viscous in vitro environments, we proceeded to examine the population dynamics of the Trojan strain in combination with PA14 R in an in vivo chronic wound infection assay.Following seven days of infection, we observed no significant difference in the total bacterial load in the wounds between single and mixed infections (Figure 5a).However, in both co-inoculated populations and delayed-mixed populations, the Trojan strain displayed a higher relative fitness than PA14 R (Figure 5b) and successfully invaded up to 75% of the populations by the end of the assay.These results demonstrate the invasive capability of QS mutants in mixed infections.Importantly, our findings also highlight the invasive ability of QS mutants even when introduced to an already established PA14 R infection.Mann-Whitney U test W = 0, p = 0.002671; W = 0, p = 0.002778 for co-inoculated and delayed-mixed populations, respectively.

Antibiotic treatment reduces bacterial load in Trojan-invaded populations ex vivo
Having demonstrated that the Trojan strain could invade PA14 R infections, we proceeded to investigate the antibiotic susceptibility of these invaded populations.Similar to in vitro observations, we expected that the increased frequency of the invading Trojan subpopulations would lead to a reduction in the total bacterial load in mixed infections following antibiotic treatment.
In single infections of the Trojan strain, we observed higher susceptibility to antibiotics, resulting in a three-log reduction in bacterial load (Figure S3a).In contrast, PA14 R infections displayed resistance to antibiotic treatment (Figure S3b).However, unlike our in vitro assays, complete killing of the sensitive populations proved challenging due to the resilient nature of wound extracts.When we treated the wound beds collected from both initially co-inoculated and delayed-mixed infections, we observed a 1-2 log decrease in total bacterial load compared to uninvaded single PA14 R infections (Fig. 6).The reduction in total bacterial load can be attributed to the elimination of the susceptible Trojan subpopulation, as we noted a comparable decrease in Trojan load in mixed populations (Figure S4).
These results demonstrate that Trojan invasion in vivo can induce antibiotic sensitivity in an otherwise resistant infection, highlighting the potential for manipulating antibiotic resistance of a target population through Trojan invasion.

DISCUSSION
We provide experimental evidence supporting the Trojan horse approach as a potential medical intervention strategy.Following the successful invasion of Trojan strains when introduced into both in vitro and in vivo populations of PA14 R (Figures 2, 4 and 5), a larger bacterial population was eliminated with antibiotics in comparison to clonal populations and single infections of PA14 R (Figures 3 and 6).While we observed successful invasion by cheats under the tested conditions, it is crucial to take into account various factors known to influence cooperator-cheat dynamics 4,13,14,10,11,20,29 when considering the viability of this strategy.These factors include population viscosity, the relative frequency of cheats, population density, the nature of selection, and the mechanism of cheating.
Our findings from the in vivo experiments (Figure 5) not only confirmed earlier observations of lasR cheat invasions of PA14 16 , but also offer new insights into potential applications of Trojan cheats to combat infections.We demonstrated the feasibility of introducing antibiotic sensitivity to a target population of chronic PA14 infection, leading to increased antibiotic susceptibility in that population (Figure 6).Moreover, we showed for the first time, to our knowledge, the possibility of introducing Trojan cheats into an established infection.In a therapeutic intervention, a Trojan cheat would be introduced in this fashion to an ongoing infection, rather than co-occurring from the onset, thus emphasizing the application potential of this therapeutic strategy.
While we successfully demonstrated the potential of employing Trojan cheats, it's important that the total bacterial load after antibiotic treatment remained substantial.Even after reducing a significant portion of the genotypically susceptible Trojan population, the remaining bacterial load exceeded the clinically defined infection threshold for wounds, commonly set at 10 5 CFU/g of tissue 29,30 (Figure 6, S3, and S4).Therefore, certain limitations must be addressed before advancing to clinical applications.These necessary improvements include improving invasion levels, and improved targeting of the resident population.
One approach to enhance cheat invasion is through engineering more effective invaders.
While single-trait mutants serve as a promising starting point, theoretically, it is possible to create even better invaders that exploit multiple cooperative traits simultaneously 31 .Using such enhanced cheats would have the added benefit of further reduced virulence, as many cooperative traits are associated with virulence 32 .To improve the Trojan strain that we used in this work, one could consider using double QS mutants, such as lasR/rhlrR, instead of the single lasR mutant.Furthermore, mutants capable of exploiting both QS and other social traits, like siderophore production 12,15 , could be explored.Depending on the specific mechanisms of cooperation and cheating, increasing the extent of cheating could also be considered as a viable option 33 .Additionally, genome reduction strategies could be applied to enhance overall strain fitness and invasive capabilities 34 .Since biofilms have been extensively documented to increase antibiotic tolerance 35,36 , one can also adopt several approaches to disperse biofilms to improve cheat invasion and antibiotic effectiveness.These approaches include mechanical removal or using molecular dispersal agents like matrixdegrading enzymes, anti-biofilm peptides, and dispersal molecules to disrupt biofilms 37 .
One major concern in almost all biocontrol methods is the potential restoration of virulence once the target population is invaded by external agents.In the case of antibiotic sensitivity introduction, the primary worry is the Trojan population acquiring resistance through horizontal gene transfer.Preventing this scenario involves implementing fail-safe mechanisms while constructing Trojan cheats.These mechanisms could include synthetic inducible systems for toxin-antitoxin or phage lytic protein expression to trigger programmed self-destruction upon invasion and antibiotic treatment 38,39 .
In conclusion, this study presents initial experimental evidence demonstrating the potential application of cheats to socially engineer and deliver medically advantageous alleles, such as antibiotic sensitivity, to target populations.The aforementioned technical improvements and the development of robust and secure delivery strategies will facilitate the clinical implementation of Trojan strains as a novel alternative strategy to fight infections.

Figure 1 .
Figure 1.Trojan horse cheat intervention strategy to fight resistant infections.

Figure 4 .
Figure 4. Spatially structured media hinders cheat invasion and antibiotic effectiveness.a) Relative fitness of Trojan cheat in spatially structured and unstructured conditions.Competition assays were performed in the presence and absence of agar in QSM without antibiotic selection for 48 h.n = 8 populations per condition.Two-sample t-test, t = 4.21099, d.f.= 14, p = 0.000872 b) Antibiotic survival of Trojan cheat invaded populations exposed to 12.5 µg/ml streptomycin.The table shows the results of post hoc pairwise comparison of survival rates at 12.5 µg/ml streptomycin.MIC Trojan = 8-12 ug/ml.All mixed populations start with an initial ratio of 10:1 cooperator to cheat.

Figure 5 .
Figure 5. Trojan cheats can invade resistant infections in vivo.a) Total bacterial load of the wound extracts at 7 days as CFU/g of wound tissue.n=6 per group.Kruskal-Wallis chisquared = 5.8933, df = 3, p = 0.1169 b) Relative fitness of Trojan strain (w) was determined after collection of wounds on day 7 from CFU/g gram tissue measurements.Values of w > 1 indicates that the cheat has a higher fitness than the cooperator.All mixed populations start with an initial ratio of 10:1 PA14 R to Trojan cheat at the time of mixing.n = 6 populations.

Figure 6 .
Figure 6.Trojan-invaded mixed infections are more susceptible to antibiotic treatment.Infected wound beds were harvested and treated ex vivo with antibiotics (1.75 mg/mL streptomycin + 8 mg/mL carbenicillin) or with PBS.Total bacterial load was expressed as CFU/g wound tissue.a) Total bacterial load of antibiotic-treated wound extracts obtained from infections co-inoculated at time 0. Mann Whitney-U test W = 2, p = 0.01307 b) Total bacterial load of antibiotic-treated wound extracts obtained from delayed-mixed infections.n= 6 populations.Mann Whitney-U test W = 3, p = 0.02024.

Figure 3. Antibiotic susceptibility of Trojan cheat invaded populations. Cultures
grown in QSM without antibiotic were spotted on antibiotic agar media and survival percentages were determined from CFUs.a) Antibiotic survival curves: Solid black: PA14 R monocultures at 24 h; solid gray: Trojan cheats at 24 h; dashed red: Mixed populations at the beginning of competition; dashed green: Mixed populations after 24 h competition in QSM.All mixed shows the results of post hoc pairwise comparison of survival at 50 µg/ml carbenicillin.MIC Trojan = 50 µg/ml.