5-Fluorouracil blocks quorum-sensing of biofilm-embedded methicillin-resistant Staphylococcus aureus in mice

Abstract Antibiotic-resistant pathogens often escape antimicrobial treatment by forming protective biofilms in response to quorum-sensing communication via diffusible autoinducers. Biofilm formation by the nosocomial pathogen methicillin-resistant Staphylococcus aureus (MRSA) is triggered by the quorum-sensor autoinducer-2 (AI-2), whose biosynthesis is mediated by methylthioadenosine/S-adenosylhomocysteine nucleosidase (MTAN) and S-ribosylhomocysteine lyase (LuxS). Here, we present a high-throughput screening platform for small-molecular inhibitors of either enzyme. This platform employs a cell-based assay to report non-toxic, bioavailable and cell-penetrating inhibitors of AI-2 production, utilizing engineered human cells programmed to constitutively secrete AI-2 by tapping into the endogenous methylation cycle via ectopic expression of codon-optimized MTAN and LuxS. Screening of a library of over 5000 commercial compounds yielded 66 hits, including the FDA-licensed cytostatic anti-cancer drug 5-fluorouracil (5-FU). Secondary screening and validation studies showed that 5-FU is a potent quorum-quencher, inhibiting AI-2 production and release by MRSA, Staphylococcus epidermidis, Escherichia coli and Vibrio harveyi. 5-FU efficiently reduced adherence and blocked biofilm formation of MRSA in vitro at an order-of-magnitude-lower concentration than that clinically relevant for anti-cancer therapy. Furthermore, 5-FU reestablished antibiotic susceptibility and enabled daptomycin-mediated prevention and clearance of MRSA infection in a mouse model of human implant-associated infection.


INTRODUCTION
Molecular communication among bacteria by means of small diffusible signaling molecules, known as quorum sensing, serves to coordinate inter-and intra-population behavior (1). The most common quorum-sensing communication signal is autoinducer-2 (AI-2) (2,3), which controls virulence and biofilm formation in various human pathogens, such as methicillin-resistant Staphylococcus aureus (MRSA), thereby contributing to their antibiotic resistance (4). In the industrialized world, the rise of antibiotic-resistant nosocomial infections has reached crisis proportions; in particular, implant-associated MRSA infections (5) account for almost 50% of infections that occur following prosthetic surgery, and are associated with dramatic morbidity and exploding healthcare expenditures (6,7). This situation calls for immediate (8) and concerted action to develop alternative treatment options that can replace or complement antibiotics (9). Molecular interference with the pathogens' quorum sensing, i.e., quorum quenching (10), by using degradative enzymes (11) or small-molecular inhibitors to block the synthesis and release of quorum-sensing signaling compounds, is one possible approach to switch off biofilm formation (12,13) and reduce or eliminate antibiotic resistance (14). In this context, AI-2 is synthesized in two sequential steps: 5 -methylthioadenosine nucleosidase (MTAN)-catalyzed hydrolysis of cellular S-adenosylmethionine (SAM) to afford S-ribosylhomocysteine (SRH), followed by Sribosylhomocysteine lyase (LuxS)-mediated conversion of SRH to AI-2 (15). Thus, inhibition of MTAN and/or LuxS is expected to quench the quorum-sensing capability of targeted pathogens and consequently to attenuate their virulence and antibiotic resistance.
Screening for enzyme-inhibiting compounds is standard practice in drug discovery and development, and typically involves microscale activity assays (16), combinatorial shuffling in nanocompartments (17) and virtual screening methods (18). However, most of these in-vitro strategies provide little or no information on the function, toxicity, bioactivity and bioavailability of the hit compounds and most hits will fail during later stages of drug development (19). On the other hand, human cell-based assays may provide an all-inone opportunity to detect bioavailable, cell-permeable, noncytotoxic and target-specific functional drug candidates (20)(21)(22). Prime examples would be the anti-tuberculosis drugs that are currently being validated by a Bioversys-GlaxoSmithKline consortium in phase-I clinical trials (23). 5-Fluorouracil (5-FU) was licensed in 1962 for the treatment of various common and aggressive cancers, including colon cancer, breast cancer and pancreatic cancer, and is on the World Health Organization's List of Essential Medicines, which defines the most effective and safe medicines needed in a public health system (https://www. who.int/medicines/publications/essentialmedicines/en/). 5-FU principally acts as a thymidylate synthase inhibitor, blocking the synthesis of thymidine monophosphate (dTMP), which is a nucleoside required for DNA replication (24). Administration of 5-FU triggers apoptosis of rapidly dividing cancer cells by depriving them of dTMP.
Here, we describe the development and application of a high-throughput screening platform for inhibitors of MTAN and/or LuxS, built on a cell-based assay employing engineered human cells programmed to constitutively secrete AI-2. Notably, we found that one of the hit compounds, 5-FU, blocks quorum-sensing by MRSA and prevents biofilm formation at an order-of-magnitude-lower concentration than that clinically relevant for anti-cancer therapy. Furthermore, 5-FU enables daptomycin-mediated prevention and clearance of MRSA infection in a mouse model of human implant-associated infections. We believe this finding has the potential to be rapidly translated into clinical use. LuxS (MF688636) and MTAN (MF688635) transgenes derived from E. coli were codon-optimized for stable expression in mammalian cells. For the generation of double stable HEK-293-derived cell line HEK-293 AI2 transgenic for simultaneous constitutive P hEFI␣ -driven MTAN and P hCMVdriven LuxS-eYFP expression, 250,000 cells were first transfected with 2000 ng of pFS168 (P hEFI␣ -MTAN-pA) and selected in culture medium containing 20 g/ml blasticidin (cat. no. ant-bl-1; InvivoGen, San Diego, CA, USA) for two weeks. Following expansion of single clones by limiting dilution for another two weeks and validation of functional MTAN expression, the best-performing clone (HEK-293 sFS25c21 ) was co-transfected with 1660 ng of fluorescent S-ribosylhomocysteinase (LuxS-eYFP; P hCMV -luxS-eYFP-pA, pFS169) and 340 ng of the zeocin resistance encoding plasmid pZeoSV2(+). After 17 days of selection in medium containing 200 g/ml (w/v) zeocin (cat. no. ant-zn-1; Invivogen, San Diego, CA, USA) and 20 g/ml blasticidin, resistant monoclonal cells (HEK-293 sFS26cx ) were obtained by limiting dilution cloning and screened for AI-2 activity in their supernatants. The best performing clone 12 was named HEK-293 AI2 . The cell line was regularly tested for the absence of mycoplasma.

AI-2 quantification for screening
Compound-treated cells were cultivated for 24 hours and the cell culture supernatants (10 l) were transferred to dry black polystyrene 384-well plates (Corning, New York, USA, cat. no. 3571). The AI-2 activity was quantified by adding 40 l per 384-well of V. harveyi MM32 AI-2 reporter strain, diluted 1:500 in AB-Medium from a stationary overnight culture grown in Luria Marine (LM)medium containing 20 g of NaCl, 10 g of Bacto Tryptone (Difco Laboratories), and 5 g of yeast extract (BBL) (25). Plates were sealed with BREATHseal™ (cat. no. 7.676 050; Greiner Bio-One, Frickenhausen, Germany) and incubated for 4 h at 30 • C and 200 rpm on a Multitron Pro shaker (Infors, Bottmingen, Switzerland). Bioluminescence was measured on an Envision 2104 Multilabel plate reader (Perkin Elmer, Waltham, USA) with 1 s integration time. For each plate, the luminescence data were normalized to the active controls (0% activity: Novobiocin; 100% activity: DMSO) measured in 8-fold replicates on each plate and expressed as % activity of DMSO control cells (100% activity). The initial hit inclusion criteria were a toxicity score of less than 50% and an AI-2 activity reduction below the average of all compound activity three times the standard deviation (-49.39% of activity).

Z' determination
The statistical effect size was calculated according to Z = 1 − ( 3(σ p +σ n ) (μ p −μ n ) ), based on the equation described by Zhang et al. (1999) with the means () and standard deviations () of the luminescence resulting from positive control (p; novobiocin, 10 M) and negative control (n; DMSO) supernatants.

Resazurin-based viability assay
To monitor viable cells with active metabolism, we employed a fast resazurin assay. In short, 1 l of PrestoBlue reagent (cat. no. A13261; Thermo Fisher, Basel, Switzerland) was added directly to each well of a 384-well plate by a Multidrop Combi Reagent Dispenser (cat. no. 5840300; Thermo Fisher, Basel, Switzerland). The plate was incubated for 1 h (37 • C, 5% CO 2 ), and then the fluorescence was measured with an Envision 2104 Multilabel plate reader (Perkin Elmer, Waltham, USA) at excitation and emission wavelengths of 560/9 nm and 590/20 nm, respectively. To calculate the relative cell viability, the fluorescence of DMSO-treated cells was set to 100%.

Criteria for further evaluation of hits
Compounds for further evaluation were selected according to the following combined inclusion criteria: (i) lack of re-ported antibiotic activity according to PubChem, (ii) IC 50 < 4 M, (iii) lack of short-term cytotoxicity (viability decrease < 25%).
The test strain E. coli RP437 was grown overnight in LB. Its optical density was measured on a Novaspec II photometer (Pharmacia, Freiburg, Germany) and the culture was diluted to a final OD 600 of 0.1 the next morning. The bacteria were challenged with test compounds for 3 h, and then AI-2 activity in the culture supernatant was determined. Specifically, 90 l of the AI-2 reporter strain V. harveyi (MM32), grown overnight in LM medium and then prediluted 1:500 in AB-medium containing 30 g/ml kanamycin, was added to 10 l of E. coli supernatant (10% v/v) in black-bottomed Fluotrac 200 96-well plates (Greiner Bio-One, Frickenhausen, Germany), which were shaken at 200 rpm, at 30 • C. The AI-2 induced bioluminescence of MM32 was measured after 5 h. Chemically synthesized DPD served (Omm Scientific, Dallas, Texas, USA) as a positive control.

Quantification of biofilm prevention
To quantify the formation of bacterial biofilms, bacteria were grown for 24 h in the presence of potential inhibitors in 96-well polystyrene plates. The plates were then washed three times with distilled water. Remaining cells were stained with 0.1% Crystal Violet solution (5% methanol, 5% isopropanol) and further washed to remove excess dye. Crystal Violet was redissolved in 20% acetic acid solution and the absorbance of the solution was measured at 600 nm.
To quantify the amount of adherent bacteria, 2 -3 × 10 5 CFU/ml MRSA were grown for 24 h in the presence of different 5-FU concentrations. After incubation under static conditions at 37 • C, the plate was washed twice with PBS. Adherent bacteria were removed with 100 l PBS and appropriate dilutions were plated on Mueller-Hinton agar (Becton Dickinson AG, Allschwil, Switzerland) plates overnight at 37 • C. Three independent experiments were performed, each performed in triplicate, and the mean val- ues were calculated. Dimethyl sulfoxide (DMSO) was used as a control.

Evaluation of 5-FU for prophylaxis in a murine tissue cage infection model
To evaluate the prophylactic efficacy of 5-FU against MRSA infection, we used our murine model of foreignbody infection, which closely mimics human implantassociated infections. This model of foreign-body infection (26,27) was established with the approval of the Kantonale Veterinaeramt Basel-Stadt, Switzerland (permit no. 1710). Experiments were conducted according to the regulations of Swiss veterinary law and performed in the animal house of the Department of Biomedicine, University Hospital Basel, Switzerland. Healthy wild-type female C57BL/6 mice at 13 weeks of age (JanvierLabs, France) kept under specific pathogen-free conditions (biosafety level 2) were anesthetized, followed by the subcutaneous implantation of sterile cylindrical Teflon tissue cages (8.5 × 1 × 30 mm; volume: 1.9 ml) with 130 regularly spaced holes (Angst + Pfister AG, Zurich, Switzerland). They were housed in a 12 hr light/dark cycle (light from 7 am to 7 pm) in a temperaturecontrolled room (24 • C) with free access to regular mice chow and water. Mice were randomly assigned to experimental groups, which were not involved in previous procedures. The mice were randomized into groups, which were treated as follows: saline (untreated growth control; n = 7), saline with 5.2844 mg/kg DPD (n = 4), 50 mg/kg daptomycin (DAP) (Novartis, Basel, Switzerland) (n = 11), 10 mg/kg 5-FU with and without 50 mg/kg DAP (each n = 8), 40 mg/kg 5-FU with or without 50 mg/kg DAP (n = 16 and n = 13, respectively), 40 mg/kg 5-FU with 1 M DPD (n = 6), or 40 mg/kg 5-FU with 50 mg/kg DAP and 5.2844 mg/kg DPD (n = 8). DAP was given intraperitoneally immediately before implantation, and 5-FU and DPD was given immediately after implantation directly into the lumen of the cage. Afterwards, the cages were infected with 965 CFU of MRSA 43300. At 2 days post-infection, tissue cage fluid was collected, and the planktonic bacterial load was evaluated by plating on blood agar plates. In additional, mice were sacrificed and the tissue cages were explanted under aseptic conditions. The explanted tissue cages were washed twice with phosphate-buffered saline followed by 30 s vortexing, sonication for 3 min at 130 W and another 30 s vortexing to release adherent bacteria from the biofilm. Quantification of adherent bacteria was performed by plating appropriated dilutions on blood agar plates. To determine the prevention rate, the presence of any re-growth of MRSA was examined by further incubation of the cage in tryptic soy broth (TSB) for 48 h at 37 • C. MRSA re-growth represents therapy failure, and the prevention rate was defined as the percentage of cages without growth in each treatment group.

Statistical analysis
All in vitro data were analyzed with an unpaired parametric t test and expressed as mean and standard deviation (SD). All in vivo data were analyzed with the nonparametric Mann-Whitney U test because they did not show a normal distribution in the Shapiro-Wilk normality tests. Data are expressed as median and interquartile range (IQR). For all assays, a P value <0.05 was considered statistically significant. Statistical analysis was performed using Prism 9 (GraphPad Software, USA). Figure 1 shows a schematic illustration of the platform for screening bioavailable, non-cytotoxic and targetspecific small-molecular drugs quenching the production of the biofilm-and virulence-promoting bacterial quorumsensing molecule AI-2 (28,29). For the primary screen, we generated the double-transgenic human cell line HEK AI-2 constitutively co-expressing the codon-optimized bacterial AI-2-synthesizing enzymes 5 -methylthioadenosine nucleosidase (MTAN) and S-ribosylhomocysteine lyase (LuxS). In HEK AI-2 , MTAN and LuxS tap into the activated methylation cycle of human cells to convert Sadenosylmethionine via S-ribosylhomocysteine into AI-2 (30). We previously showed in a proof-of-concept study (31) that the modular biosynthetic AI-2 production platform can function in HEK cells, and AI-2 production is significantly decreased by the addition of known MTAN or LuxS inhibitors such as immunicillin-A (32) and Sribosylhomocysteine analogues (33). The decrease of AI-2 production induced by those inhibitors was not due to a decrease of cell viability (31).

Design of the cell-based AI-2-specific quorum-quencher screening platform
We confirmed that AI-2 production and secretion by engineered HEK AI-2 cells could be precisely and reliably quantified in the culture supernatant by addition of the AI-2sensitive reporter bacterium V. harveyi (MM32), which allows rapidly profiling of the quorum-sensing molecule by means of bioluminescence-based assay (34). The functional combination of HEK AI-2 -mediated AI-2 production with V. harveyi (MM32)-based AI-2 quantification provides a potent mammalian cell-based assay platform for the detection of bioavailable, non-cytotoxic and target-specific smallmolecular drugs quenching the production of AI-2 ( Figure  2A). To set a benchmark for larger-scale screening, the cellbased AI-2 production assay was validated by using the V. harveyi-killing antibiotic novobiocin (35); this completely shut down the bioluminescence of V. harveyi (MM32) (Figure 2B). In parallel, the viability and metabolic integrity of HEK AI-2 were profiled by means of resazurin assay to detect test-compound-associated cytotoxicity (see Supplementary Figure S1).

Discovery of quorum-quenching compounds in a chemical library
Capitalizing on the robust assays for both bioactivity (Z' value of 0.71) and viability (Z' value of 0.73), we upscaled our system for compatibility with an automated high-throughput industrial screening platform and tested a library of over five thousand commercially available compounds with validated mechanisms of action. Probing this chemical library in duplicate yielded 66 hit compounds (hit rate: 1.2%) that were bioavailable and showed quorumquenching activities, while lacking substantial metabolic impact or cytotoxicity towards human target cells ( Figure   2C). To mitigate off-target effects, only compounds decreasing AI-2 production by over 50% while maintaining human cell viability and metabolic integrity >50% were considered for a secondary screen of the dose-response relationship in human cells.

Validation and potency quantification of quorum-quenching hit compounds
The AI-2-specific quorum-quenching activity of the hit compounds was confirmed by dose-dependence analysis using the HEK-293 AI2 -V. harveyi (MM32) AI-2 detection platform to determine the half-maximal inhibitory concentration (IC 50 ) of the individual compounds. Among the quorum-quenching drug candidates from the primary screen, 30 compounds inhibited AI-2 production with IC 50 values lower than 5 M (Table 1), and among them, 11 compounds dose-dependently inhibited AI-2 production even in the submicromolar range (Supplementary Figure S2). Hit compounds that were listed in public databases as having antibiotic activity were excluded from consideration as they would decrease the bioluminescence simply by killing the reporter strain, V. harveyi (MM32) ( Table 1). Fortunately, nine of the most potent AI-2 production-inhibiting compounds did not show any antibiotic activity. These included formycin A (36) and MT-DADMe-ImmA (37), which have previously been shown to exhibit quorum-quenching activity by inhibiting MTAN ( Table 2). Rediscovery of established quorum-quenching compounds is a potent validation of this cell-based quorum-quenching drug-discovery platform.

Target specificity of validated hit compounds
Although the cell-based assay for the discovery of AI-2 biosynthesis inhibitors was designed to reveal compounds targeting the synthetic LuxS/MTAN methylation cycle bypass without affecting the metabolism of human cells, potential off-target effects impacting related metabolic pathways or bacterial growth still require careful examination. Therefore, we first examined the candidates' interference with the growth of V. harveyi and E. coli. Indeed, many of the tested quorum-quenching drug candidates dosedependently decreased the growth of V. harveyi ( Figure 3A-H), and clofarabine showed the most potent effect (TC 50 = 4.4 × 10 −7 ) ( Figure 3A). On the other hand, 5-FU, a licensed anti-cancer therapeutic that has not been reported to show quorum-quenching activity, reduced the bioluminescence without decreasing the growth of V. harveyi ( Figure  3C), and also interfered with E. coli quorum sensing (Supplementary Figure S3a-d). 5-FU was therefore chosen for follow-up in vivo studies.

5-FU is a quorum quencher effective against MRSA in vitro and in vivo
Staphylococci, including Staphylococcus aureus (SA) and the coagulase-negative Staphylococcus epidermidis (SE), have evolved quorum-sensing systems that enable cell-tocell communication. SE is a normally harmless commensal bacterium found on the skin, but under certain conditions, especially when a medical device is involved, it can become invasive and colonize the device. To validate the staphylococcal AI-2-specific quorum-quenching capacity of 5-FU, we first investigated its in vitro activity against SE 1457 wild type. 5-FU reduced AI-2 activity dose-dependently (Supplementary Figure S4a). To examine whether synthetic AI-2 supplementation would restore the phenotype impaired by 5-FU, the AI-2 precursor DPD was added directly to SE in combination with increasing 5-FU concentrations. In the presence of DPD, the AI-2 activity returned almost to baseline (Supplementary Figure S4b).
Even though SE exhibits considerably higher AI-2 activity than MRSA, the clinically relevant human pathogen MRSA (7,38) expresses more virulence factors that facilitate its spread and survival. To confirm the AI-2-specific quorum-quenching activity of 5-FU, we next examined its in vitro and in vivo activity against MRSA ATCC 43300. 5-FU dose-dependently reduced MRSA's AI-2 production and quorum-sensing capacity ( Figure 4A) and significantly decreased bacterial growth at the concentration of 0.1 M (Figure 4B), which is an order of magnitude lower than the 5-FU concentration that is clinically relevant for anti-cancer therapy (39). At this concentration, 5-FU not only reduced the number of adherent MRSA but also significantly decreased biofilm formation ( Figure 4C, D). To rule out possible artefacts caused by the vehicle (DMSO) used for 5-FU administration, the influence of DMSO was monitored in parallel. We observed that DMSO started to impact growth and biofilm formation at the highest concentration applied (corresponding to 10 M 5-FU), though the effect did not reach statistical significance (data therefore not shown).
To examine whether the decrease of AI-2 activity caused by 5-FU is reversible, we supplemented the 5-FUchallenged MRSA with the AI-2 precursor DPD. The added DPD counteracted the effect of 5-FU, while showing little influence on the baseline AI-2 activity of MRSA 43300 ( Figure 5). These findings support the idea that 5-FU slows down AI-2 biosynthesis rather than blocking AI-2 sensing.
To assess 5-FU's quorum-quenching anti-infective potential in vivo, we employed the foreign-body mouse infection model, which simulates human implant-associated infections (40). Mice were prophylactically treated with 5-FU (10 mg/kg or 40 mg/kg) in combination with or without daptomycin (DAP). All drug concentrations were within the human therapeutic dosage range (41). In order to assess if the effect of 5-FU is mediated by inhibition of AI-2 production, we simultaneously administered synthetic AI-2 (DPD). While treatment with 5-FU or DAP alone failed to clear planktonic MRSA infections ( Figure 6A), simultaneous prophylactic treatment of animals with 5-FU and DAP prevented MRSA infection and cleared the pathogen ( Figure 6A). Likewise, prophylactic administration of 5-FU and DAP prevented biofilm formation and completely eradicated MRSA, though neither 5-FU nor DAP alone was sufficient to contain the infection ( Figure 6B, C). Interestingly, the presence of DPD reversed the 5-FU-dependent inhibition of planktonic ( Figure 6A) and adherent bacteria ( Figure 6B) when co-administered with DAP. Similarly, DPD decreased the prevention rate to the level observed for DAP monotherapy ( Figure 6C).

DISCUSSION
Overcoming antibiotics resistance is one of the most critical, complex and pressing healthcare challenges of the 21st century. The rapid development of resistance, as well as the poor profitability of these life-saving drugs, which often lose efficacy even before coming off patent, is making more-of-the-same development of novel antibiotics or new antibiotic derivatives scientifically questionable and economically non-viable (38,42,43). Therefore, new antiinfective strategies, as well as novel blueprints for smallmolecular drug discovery, are urgently needed to cope with the globally increasing prevalence of multidrug-resistant pathogenic bacteria (17,44). Several proof-of-concept studies have appeared, suggesting the use of recombinant phages to target bacteria (45), small-molecular drugs to switch off antibiotic resistance genes in Mycobacterium tuberculosis (22,23), antibiotic adjuvants to increase efficacy (46), or immuno-mimetic designer cells to detect and kill multidrug-  (40). In particular, studies on quorumquenching drugs that interfere with the pathogens' interspecies and intra-population molecular communication to coordinate persistence (47), virulence (18) and biofilm formation (10,48) are gathering momentum, because nonkilling drugs that interfere with quorum-sensing should impose low selection pressure, and may eliminate, reduce or delay the emergence of resistance. Non-limiting examples of quorum-quenchers include brominated furanones (49), autoinducer-degrading enzymes (50) and the established antibiotic azithromycin (51). In addition, coatings functionalized with 5-FU have shown promising results in phase-1 clinical trials for the treatment or prevention of implant-associated infections (52); however, the molecular target(s) of 5-FU's anti-infective activity has remained elusive until now. Despite the great prospects for new quorumquenching drugs, every new class of drugs may have side effects or show off-target activities. Therefore, expanding the therapeutic space of licensed drugs with a proven track record of tolerability by finding alternative targets and activities may be a rapid and efficient strategy for bridging the gap until novel approaches can be brought into clinical use. Drug discovery has not seen much conceptual progress in recent decades. Structure-function predictions, drugtarget fitting and high-throughput screening have become increasingly sophisticated, but hit-to-lead development has remained challenging due to limited bioavailability, poor pharmacokinetics and/or cytotoxicity of many drug candidates (53). Thus, although cell-based assays are more expensive to set-up and operate, they may offer advantages over classic in-vitro drug screening, in that they can validate the function, cytotoxicity, bioactivity and bioavailability of drug candidates in an all-in-one test format. Nonlimiting proof-of-principle examples of cell-based screening assays include the discovery of novel anti-infective activities (20,21), peptides ameliorating chronic pain (54), and drug   candidates that switch-off antibiotic resistance genes (23). Indeed, a drug switching off the antibiotic resistance of Mycobacterium tuberculosis, a century-old plague, is currently under phase I clinical trial by a Bioversys-GlaxoSmithKline consortium (23). These small-molecular drugs are based on early hits in cell-based assays revealing compounds that switch off the pathogen's intrinsic resistance to the last-line antibiotic ethionamide (22). In contrast to classical antibiotics which eliminate target pathogens, thereby creating a strong selective pressure that drives the emergence of resistance, there is hope that antibiotics co-administered with compounds that switch off antibiotic-resistance may reduce the selective pressure and thus delay the onset of resistance. But, as this approach still focuses on actively killing the pathogen, resistance is still expected to arise eventually, as in classical antibiotic therapies. Consequently, it is conceivable that attenuation, rather than killing, of the pathogen may become a valid strategy to avoid the selection of resistant populations in future anti-infective therapies. Following Darwinian principles, attenuation of pathogenic traits to increase fitness is typically observed in the co-evolution of host-pathogen interactions. Therefore, non-killing interventions leading to attenuation of host-pathogen interactions may open up new opportunities for anti-infective therapies.
Studies of host-pathogen interactions at the molecular level are frequently done in co-culture systems of bacterial pathogens and host cells (31,55). Such co-cultures not only provide new insight into host-pathogen crosstalk, but also at the same time provide a framework for cellular assays for the discovery of drugs interfering with the host-pathogen interaction. However, the application of synthetic biology principles to engineer mammalian cells with functionalized bacterial circuits and user-defined drug targets has largely eliminated the need for host-pathogen co-cultivation, and instead has enabled the development of simple cell-based assays, thereby increasing the repertoire of drug-screening tools (12,56). For example, pioneering transfer of pathogenderived repressor-operator gene switches into mammalian cells enabled hit discovery of novel streptogramin antibiotics (21) and anti-microbacterial drugs (22,57). Using the same principle for engineering drug-target-specific assay cell lines (22,(58)(59)(60), but replacing the reporter gene with a therapeutic effector gene, has resulted in novel cell-based anti-infective therapies (40,55). For example, implant-associated infections by S. aureus could be prevented as well as cleared by using engineered human-cell implants that detected the presence of the pathogen and coordinated a rapid, reversible and dose-dependent peptide expression response to eliminate biofilm as well as planktonic multidrug-resistant MRSA (40). Overall, designer cell lines incorporating critical pathogen-derived circuits can serve as (i) cell-culture based systems to reveal molecular information on host-pathogen interactions, (ii) cell-based therapies to prevent and cure bacterial infections and (iii) cell-based assays to screen novel anti-infective drugs. Running such cell-based assays in an academicindustrial context using libraries of compounds with established activities is a successful strategy for repurposing licensed compounds with a therapeutic track record, enabling rapid clinical application, and so is particularly relevant for last-line anti-infective therapies. Indeed, this approach identified the licensed chemotherapeutic pyrimidine analogue 5-fluorouracil, one of the WHO essential medicine for chemotherapy (61), as a potent antimicrobial drug. But, although empirical data show that 5-FU has antibiotic-boosting activity, the molecular mechanism has remained a mystery for decades (62,63,(64)(65)(66), and consequently, a rational basis for its therapeutic application as an anti-infective has been lacking. Now, we have shown that 5-FU quenches the quorum-sensing activity in an in vivo model of implant-associated MRSA infections to such an extent that the infections could be completely cleared by the last-line antibiotic daptomycin, which achieves <20% clearance rate when used alone (26). Notably, the addition of the AI-2 precursor DPD reinstated this low rate of clearance and prevention by neutralizing the 5-FU-dependent quorum quenching.
The cell line used for initial screening incorporated E. coli-derived AI-2 biosynthetic enzymes that share sequence identities of 35% for LuxS and 53% for MTAN with the S. aureus homologues. However, AI-2 interference of identified hit compounds was subsequently confirmed in diverse bacterial species, so we believe this is not problematic. Furthermore, the anti-infective effect of 5-FU originally discovered in engineered human cells was also observed in SE, which exhibits pronounced AI-2-mediated quorum sensing and stronger biofilm formation than MRSA (26). Quorum sensing is inherently linked to biofilm formation, so it is not unexpected that both traits respond to 5-FU in a similar manner.
Although it is still too early to speculate about any common molecular basis for simultaneous interference with bacterial and tumor growth, the observation that a simple pyrimidine can block both DNA synthesis and quorumsensor biosynthesis may represent a valuable clue for the development of future anti-infective strategies. Indeed, we propose that the anti-neoplastic drug 5-FU could as of now be considered for anti-infective therapy of implantassociated MRSA infections.

DATA AVAILABILITY
All data and materials are available upon request. Sequences of key expression vectors have been deposited in GenBank: pFS83, MF688636; pFS84, MF688635.

SUPPLEMENTARY DATA
Supplementary Data are available at NAR Online.