Castanea sativa (European Chestnut) Leaf Extracts Rich in Ursene and Oleanene Derivatives Block Staphylococcus aureus Virulence and Pathogenesis without Detectable Resistance

The Mediterranean is home to a rich history of medical traditions that have developed under the influence of diverse cultures over millennia. Today, many such traditions are still alive in the folk medical practices of local people. Investigation of botanical folk medicines used in the treatment of skin and soft tissue infections led us to study Castanea sativa (European Chestnut) for its potential antibacterial activity. Here, we report the quorum sensing inhibitory activity of refined and chemically characterized European Chestnut leaf extracts, rich in oleanene and ursene derivatives (pentacyclic triterpenes), against all Staphylococcus aureus accessory gene regulator (agr) alleles. We present layers of evidence of agr blocking activity (IC50 1.56–25 μg mL-1), as measured in toxin outputs, reporter assays hemolytic activity, cytotoxicity studies, and an in vivo abscess model. We demonstrate the extract’s lack of cytotoxicity to human keratinocytes and murine skin, as well as lack of growth inhibitory activity against S. aureus and a panel of skin commensals. Lastly, we demonstrate that serial passaging of the extract does not result in acquisition of resistance to the quorum quenching composition. In conclusion, through disruption of quorum sensing in the absence of growth inhibition, this study provides insight into the role that non-biocide inhibitors of virulence may play in future antibiotic therapies.


Introduction
Alarming trends in the spread of antibiotic resistance among top pathogens, including Staphylococcus aureus, have placed mankind at the brink of what has been coined as the 'post-S. aureus produces an extensive array of enzymes, hemolysins, and toxins that are essential to its ability to spread through tissues and cause disease [9]. These virulence factors serve a wide scope of purposes in the infection process, including disruption of the epithelial barrier, inhibition of opsonization by antibody and complement, neutrophil cytolysis, interference with neutrophil chemotaxis, and inactivation of antimicrobial peptides [10][11][12][13]. The expression of all of these invasive factors is controlled by cell-density quorum sensing using the autoinducing peptide (AIP) molecule (Fig 1). Like other quorum-sensing signals, AIP accumulates outside the cell until it reaches a critical concentration and then binds to a surface receptor called AgrC, initiating a regulatory cascade. Since AIP controls the expression of accessory factors for S. aureus, this regulatory system has been named the accessory gene regulator (agr), and the majority of the proteins necessary for this quorum-sensing system to function are Fig 1. Schematic of the Staphylococcus aureus accessory gene regulator system. The agr locus has been investigated in detail and is known to contain two divergent transcripts named RNAII and RNAIII [9]. The RNAII transcript is an operon of four genes, agrBDCA, that encode factors required to synthesize AIP and activate the regulatory cascade. Briefly, AgrD is the precursor peptide of AIP, AgrB is a membrane protease involved in generating AIP, AgrC is a histidine kinase that is activated by binding AIP, and AgrA is a response regulator that induces transcription of both RNAII and RNAIII. The RNAIII transcript yields a regulatory RNA molecule that acts as the primary effector of the agr system by up-regulating extracellular virulence factors and down-regulating cell surface proteins [65]. The agr pathway is illustrated here with potential target sites for 224C-F2. encoded in the agr chromosomal locus [9,14]. Applying inhibitors to quench this communication system to attenuate pathogenicity and virulence lies at the core of this approach [15,16].
Agr plays a key role in S. aureus pathogenesis. For example, SSTIs are the most common type of infection caused by S. aureus [17,18]. These range from minor inflammatory conditions to more invasive infection, and most of these cases are associated with the formation of abscesses, the hallmark of a S. aureus infection. Through the use of genetic and agr-inhibiting tools, the agr system's importance to abscess formation has been confirmed [19][20][21][22][23]. The bulk of the phenotype is due to agr-dependent secreted virulence factors as demonstrated with studies on sterile supernatants from wild type and agr mutant strains [20,24,25]. Interference with the agr system through the use of competing AIPs or AIP-sequestering antibodies decreased abscess formation [20,21,23]. These findings provide direct support for the notion that agr-targeted therapies could be an option for the development of skin infection treatments. Looking at other types of infections, agr mutants also display attenuated virulence in mice in the establishment of pneumonia and mortality [26][27][28][29], and in a systemic bloodstream infection model [30].
Given the importance of the agr system in pathogenesis, it has become the target of a number of anti-virulence chemical approaches [31]. With the extracellular exposure of the AgrC receptor, chemists have developed receptor antagonists that successfully inhibit the system in vitro and quench a S. aureus mouse skin infection [32][33][34]. Since there are different groups of the agr system (4 alleles), broad spectrum inhibitors were developed to extend the applicability of the antagonist. To the best of our knowledge, these leads were never pursued in a comprehensive way for therapeutic development, perhaps because they are labile synthetic peptides and possess poor bioavailability or pharmacokinetic properties. Other recent leads have included AgrA inhibitors, savirin [35] and the polyhydroxyanthraquinones [36,37], AgrC antagonists solonamide A and B [38] and the AgrB inhibitor ambuic acid [39]. The present study represents the first in-depth analysis of botanical natural product inhibitors for agr first identified in the Quave et al. 2011 screening paper on quorum quenching Italian medicinal plants [4]. We hypothesize that by using a complex botanical composition to target quorum sensing rather than growth inhibition, the typical pitfalls of classical antibiotics can be avoided by limiting impact on the cutaneous microbiome and avoiding generation of resistance.

Collection and crude extraction of plant materials
Fresh leaves of the European Chestnut (Castanea sativa Mill., Fagaceae) were collected from wild populations in the months of May-July (2012-2014) in the Rionero-Alto Bradano region of the Basilicata Province in southern Italy following standard guidelines for collection of wild specimens [40]. Collections were made on private land with the permission of the landowner. Voucher specimens (CQ-309) were deposited at the Herbarium Lucanum (HLUC) at the Universitá della Basilicata in Potenza, Italy and the Emory University Herbarium (GEO) in Atlanta, GA, USA. The specimens were identified using the standard Italian Flora [41] and identification was confirmed at HLUC. European Chestnut leaves were shade-dried, ground with a blender, and vacuum sealed with silica packets prior to shipment to the US (under USDA permit P587-120409-008) for extraction and analysis. Upon arrival at the lab, leaves were further ground into a fine powder with a Thomas Wiley Mill at a 2 mm mesh size (Thomas Scientific).

Extraction and purification of QSI-containing fractions
Crude methanol extracts (Extract 224) of the ground leaves were created by maceration of the plant materials at room temperature using a ratio of 1g dry leaves:10 mL MeOH for two successive periods of 72 hours, with daily agitation. Filtered extracts were combined, concentrated at reduced pressure and a temperature <40°C with rotary evaporators, and lyophilized before being re-suspended in water and partitioned in succession with hexane, ethyl acetate and butanol (all solvents acquired from Fisher Chemical, Certified ACS). The resulting nonaqueous partitions were dried over anhydrous Na 2 SO 4 , concentrated in vacuo, and lyophilized before testing for activity.
The most active partition (ethyl acetate, extract 224C) was subjected to further fractionation using a CombiFlash Rf+ (Teledyne ISCO) flash chromatography system using a RediSep Rf Gold silica column. Extract 224C was bonded to Celite 545 (Acros Organics) at a 1:4 ratio and dry-loaded using a RediSep dry load cartridge. The mobile phase consisted of (A) hexane, (B) EtOAc, and (C) MeOH. The linear gradient begins with 100% A for 6.3 column volumes (CV), and then increased to 50:50 A:B by25.3 CV, and increased to 100% B at 63.3 CV, which was held until 69.6 CV, and then to 70:30 B:C at 88.6 CV, which was held until 94.9 CV. The chromatography was monitored at 254 and 280 nm, as well as via ELSD. The resulting fractions were combined into 5 fractions. Following further bioassay testing, it was determined that the fraction which eluted from 30-40 CV (224C-F2) was most active. The full extract fractionation scheme is presented in

Characterization by HPLC and LC-FTMS
An analytical HPLC-method was developed for the purposes of characterization of 224 and fractions. The analysis was performed on an Agilent 1260 Infinity system running OpenLab CDS ChemStation (Agilent Technologies, Santa Clara, CA, USA) with an Agilent ZORBAX Eclipse XDB-C18 (250 mm x 4.6 mm, 5 μm) column with compatible guard column at a column temperature of 40°C. Mobile phase reagents were HPLC-grade and purchased from Fisher Scientific, except for the Type 1 water, which was obtained from an EMD Millipore MILLI-Q water system (Billerica, MA). Mobile phase consisted of a linear gradient elution 0.1% formic acid in acetonitrile (A) and 0.1% formic acid in water (B) at a flow rate of 1 mL/ min. Initial conditions were 98:2 (A:B) changing to 70:30 (A:B) at 50 min, to 2:98 (A:B) at 70 min and held until 85min., Samples were prepared in DMSO and 10 μL injections were made. Chromatograms were monitored at 254 nm and 314 nm.
Liquid chromatography-Fourier transform mass spectrometry (LC-FTMS) was performed on 224C-F2 using a Shimadzu SIL-ACHT and Dionex 3600SD HPLC pump with a modification of the previous chromatographic conditions. A 20 μL injection at ambient temperature with 0.1% formic acid in Optima LC/MS acetonitrile (Fisher Scientific) (A) and 0.1% formic acid in water (B) at a flow rate of 1 mL/min. Initial conditions were 98:2 (A:B) changing to 64:36 (A:B) at 12 min, to 52:48 (A:B) at 86 min, 2:98 (A:B) at 102.6 min and held until 117.6 min before returning to initial conditions to equilibrate the column. The data was acquired in MS 1 mode scanning from a m/z of 150-1500 on a Thermo Scientific LTQ-FT Ultra MS in negative ESI mode and processed with Thermo Scientific Xcalibur 2.2 SP1.48 software (San Jose, CA). The capillary temperature was 275.0°C, sheath gas of 60, source voltage and current 5.0 kV and 100.0 μA, and the capillary voltage -49.0 V.
Putative compounds were determined for compounds present in the bioactive active region of 224C-F2's chromatogram (retention time of 21-49 min). The Dictionary of Natural Products (CRC Press) and Scifinder (Chemical Abstracts Service) were searched in May 2015 using similar methodology. The high resolution mass of the compound was determined from the LC-FTMS data and the database searched for all compounds within ± 0.5 Da. The resulting compounds were limited to only those identified in the genus Castanea, for DNP several entries for the misspelling "Castaneae" was also included. The molecular formulas of the remaining compounds were compared to empirical formulas derived from the MS data and those that matched the experimental molecular mass with a delta of less than 100 ppm were evaluated further. Only small molecules were considered for further evaluation. Publications on the remaining small molecules were reviewed and the presence of the compound in the genus was verified.
In addition to examining LC-FTMS data and fragmentation patterns as described above, a number of natural products reported to occur in crude C. sativa leaf extracts [42] were specifically searched for in 224C-F2: chlorogenic acid, ellagic acid hyperoside, isoquercitrin and rutin. Standards of chlorogenic acid and ellagic acid (MP Biomedicals, Solon OH) and hyperoside (Chromadex, Irvine, CA) were run on the analytical HPLC method described above to determine retention times, the others were examined by MS fragmentation patterns and published UV-Vis spectra [43]. Standards were evaluated for purity via HPLC-DAD.
Bacterial strains, plasmids, and culture media S. aureus cultures were grown in Tryptic Soy Broth (TSB) or Tryptic Soy Agar (TSA). Cationadjusted Mueller-Hinton broth (CAMHB) was used for minimum inhibitory concentration (MIC) testing of S. aureus. The bacterial strains and plasmids used in this study are described in Table 1. Escherichia coli cultures were grown in Luria-Bertani (LB) broth or on LB agar plates

Minimum inhibitory concentration (MIC)
Extract 224 and fractions were examined for minimum inhibitory concentrations (MIC) against strains representing the four agr alleles (AH430, AH1677, AH1747, AH1872), biofilm test strain (UAMS-1) and a USA500 strain (NRS385), which was used in δ-toxin quantification experiments. Clinical Laboratory Standards Institute (CLSI) M100-S23 guidelines for microtiter broth dilution testing were followed [44]. Controls include the vehicle, and antibiotics: Kanamycin (Kan) and Amp (MP Biomedicals Inc). All concentrations were tested in triplicate and repeated twice on different days. Briefly, overnight cultures in CAMHB were standardized by OD to 5 x 10 5 CFU/mL, and this was confirmed by plate counts. Two-fold serial dilutions were performed on a 96-well plate (Falcon 35-1172) to achieve a test range of 512-0.25 μg mL -1 for extracts and 64-0.03125 μg mL -1 for Amp and Kan. Plates were incubated at 37°C for 18 hrs. under static conditions. Plates were read at an OD 600nm in a Cytation 3 multimode plate reader (Biotek) at 0 and 18 hrs. post inoculation. The following formula, which takes into account the impact of extract color and vehicle on the OD, was used as previously described [5]: with OD t18 = OD of the test well at 18 hrs., OD t0 = OD of the test well at 0 hrs., OD vc18 = OD of the vehicle control well at 18 hrs, and OD vc0 = OD of the vehicle control well at 0 hrs. MIC 50 and MIC 90 values were assigned based on the concentration at which at least 50 or 90% inhibition of growth was observed as determined by OD, respectively. Growth inhibition of the refined extract, 224C-F2, was also assessed for impact on the normal skin microflora. In all cases, with the exception of P. acnes, the appropriate CLSI method for MIC determination by broth microdilution was employed. Briefly, MICs for Staphylococcus warneri, S. epidermidis, S. haemolyticus and Micrococcus luteus were determined using the above described M100-S23 CLSI method [44] for S. aureus with vehicle and antibiotic controls. Amp and Kan (MP Biomedicals Inc) were used in all staphylococcal tests; Amp, Erm (Sigma Aldrich) and clindamycin, Clin (MP Biomedicals) were used for M. luteus controls. MICs for Streptococcus pyogenes and S. mitis were determined using the M100-S23 CLSI method [44] in CAMHB with 3% lysed horse blood (LHB), incubated at 37°C for 24 hrs under static conditions, with Amp and Erm as antibiotic controls. MICs for Corynebacterium striatum and C. amycolatum followed the M45-A2 CLSI method [45] in CAMHB with 3% LHB, incubated at 35°C for 24 hrs under static conditions, with Amp and Erm as antibiotic controls. MICs for Propionibacterium acnes were based on a previous method [46] using BHI supplemented with 1% dextrose, incubated at 37°C for 72 hrs under static, anaerobic conditions.

Quorum quenching assays with reporter strains
Extracts were tested for quorum quenching activity against all four agr types using previously described [47] agr P3-YFP reporter strains AH1677 (type I), AH430 (type II), AH1747 (type III), and AH1872 (type IV), as well as previously described agr P3-lux (type I) reporter strain AH2759 [36]. Overnight cultures of reporter strains that were grown in TSB supplemented with Cam were inoculated at a dilution of 1:250 into fresh TSB containing Cam. 100 μL aliquots were added to 96-well microtiter plates (Costar 3603) containing 100 μL aliquots of TSB containing Cam and 2-fold serial dilutions (0.1-200 μg mL -1 ) of extracts 224, 224C, and 224C-F2. After mixing, the effective inoculum dilution was 1:500 and the final extract concentrations ranged from 0.05-100 μg mL -1 , with a final DMSO concentration of 1% (v/v) in all wells. Four dilution series were prepared for each reporter/extract combination, and in addition 4 mock vehicle (DMSO) dilution series were included for each reporter strain. Microtiter plates were incubated at 37°C with shaking (1000 rpm) in a Stuart SI505 incubator (Bibby Scientific, Burlington, NJ) with a humidified chamber. Fluorescence (top reading, 493 nm excitation, 535 nm emission, gain 60) and optical density (OD) readings at 600 nm, or luminescence and OD 600 readings in the case of reporter AH2759, were recorded at 30 min increments using a Tecan Systems (San Jose, CA) Infinite M200 plate reader.

Hemolytic activity by red blood cell lysis assay
The quorum quenching activity of extracts was assessed by measuring the hemolytic activity of culture supernatants on rabbit red blood cell lysis. Overnight cultures of an Erm sensitive variant of USA300 strain LAC, AH1263 [48] and an hla::Tn551 (AH1589) mutant of AH1263 [49] were inoculated 1:500 into 5 ml of TSB (in 17x150 mm culture tubes) containing extracts 224, 224C, or 224CF2 at concentrations of 6.25, 12.5, 25, 50 and 100 μg mL -1 . In all tubes containing extract the mock vehicle (DMSO) concentration was held constant at 1% (v/v). Vehicle control tubes containing 1% DMSO were similarly prepared for AH1263, AH1589 well as for an Δagr:: tetM (AH1292) mutant of AH1263 [49]. All tubes were incubated at 37°C with shaking (250 rpm), and growth was monitored by periodically transferring 100 μL of culture to a 96-well microtiter plate and reading OD 600 in a Tecan Systems (San Jose, CA) Infinite M200 plate reader. Following 6 hrs of incubation, 600 μL of each culture was filter sterilized using cellulose acetate SpinX 0.22 μm filters (Corning).
To quantify hemolytic activity, the filter sterilized culture supernatants were serially diluted in 2-fold steps (from 0.04-100%) in TSB, and 50 μL aliquots were dispensed in quadruplicate into 96-well microtiter plates. Rabbit erythrocytes, prepared from defibrinated blood (Hemostat Laboratories, Dixon, CA) by washing 3 times with 1.1x PBS and resuspending in 1.1x PBS at 1% (v/v), were added to the microtiter plates at 50 μL per well (yielding a final erythrocyte concentration of 0.5% (v/v)). The erythrocytes and culture supernatants were mixed thoroughly and incubated statically at room temperature for 2 hrs. Hemolysis was detected by the loss of turbidity as measured at OD 630 using a Tecan Systems (San Jose, CA) Infinite M200 plate reader. Relative hemolytic activities were obtained by using KaleidaGraph 4.1.3 (Synergy Software, Reading, Pa., USA) to perform 4-parameter logistic fits of the turbidity data in order to determine the concentration of supernatant that resulted in 50% red blood cell lysis.

Western blot for alpha-hemolysin
An overnight culture of S. aureus AH3052 Δspa was inoculated into 5 mL of TSB at 1:500 and grown at 37°C with shaking (250 rpm), in the presence of either DMSO or one of the extracts (224, 224C or 224C-F2) at concentrations of 6.25, 12.5, 25, 50 and 100 μg mL -1 . Following 8 hours of incubation, 600 μL of each culture was filter sterilized using a cellulose acetate SpinX 0.22 μm filter (Corning) and the filter sterilized media was stored at -20°C. The filtered media was electrophoresed on 13% SDS-PAGE gels and transferred to nitrocellulose membranes (Bio-Rad). Membranes were blocked overnight at 4°C in TBST (20 mM Tris [pH 7.5], 150 mM NaCl, 0.1% Tween 20) with 5% nonfat dry milk then washed 3 times with TBST. Hla was detected using a polyclonal rabbit anti-Hla antibody (Shlievert Lab, University of Iowa) at a 1:5000 dilution and a goat anti-rabbit HRP secondary antibody (Jackson ImmunoResearch Laboratories) at a 1:20000 dilution. Blots were incubated at RT for 5 min with Supersignal West Pico Chemiluminescent Substrate (Thermo Scientific) then exposed to film for 30 min.

Quantification of δ-toxin by HPLC
Overnight cultures of S. aureus NRS385 were standardized by OD to a starting density of 5 x 10 5 CFU mL -1 in TSB, and this was verified by plate counts. The standardized culture was added to 14 mL test tubes containing the extract or vehicle control, for a final tube to volume ratio of 1:10. All extracts were examined at sub-MIC 50 concentrations to avoid impact of growth inhibition on quorum sensing. Cultures were incubated at a 45°angle at 37°C while shaking (275 rpm) for 15 hrs, and then placed on ice until cultures were centrifuged (13,000 rcf x 5 min) into a pellet using a bench-top refrigerated (4°C) centrifuge. Supernatants were carefully removed and sterile filtered with a 0.22 μm nylon syringe filter (Membrane Solutions, Dallas, TX). Each supernatant was divided into equal aliquots for freezing at -20°C until needed for HPLC quantification of δ-toxin, toxicity testing on HaCaT cells and AIP I quantification.
Frozen supernatant samples were defrosted to room temperature and transferred to HPLC autosampler vials. Resolution of the de-formylated and formylated δ-toxin peaks was achieved on an Agilent 1260 Infinity system with a Resource PHE 1-mL (GE Healthcare, Uppsala, Sweden) analytical column, as previously described [4,50]. Briefly, 500 μL of supernatant was injected onto the column. The toxins were eluted at a flow rate 2 mL min -1 using a gradient of two solvent systems: (A) 0.1% trifluoracetic acid (TFA) in water and (B) 0.1% TFA in acetonitrile (ACN). The mobile phase was 10% B for 3 min., 90% B for 7.5 min., 100% B for 2 min. and 0% B for 2 min. Peak integration was at 214 nm, with de-formylated and formylated δtoxin recorded at a retention time of 6.4 and 6.8 min, respectively. Total peak height and areas were recorded. Peak identities were confirmed by running the same chromatographic method on the previously described LC-FTMS system in negative ESI mode and comparing the de-formylated and formylated δ-toxin ions to published values [51].

Resistance passaging
To determine the ability of S. aureus to generate resistance to the quorum quenching effects of 224C-F2, cultures were exposed to sub-MIC concentrations (16 μg mL -1 ) of extract for 15 hrs, as described above, the OD 600 taken, and cultures centrifuged. The cell-free supernatant was removed and frozen for later HPLC quantification of δ-toxin as described above. The cell pellets were then reconstituted in TSB to an OD equivalent of 5 x 10 5 CFU mL -1 with extract (or vehicle control) added, and incubated while shaking as described above. This process was repeated for a total of 15 passaging days.

Biofilm assay
Extract 224 and fractions were examined for impact on S. aureus biofilm formation using a human plasma protein-coated assay as previously described [6,52] using strains UAMS-1 [53] (a PFGE USA200 osteomyelitis isolate, agr type III) and its isogenic sarA mutant, UAMS-929, which has a biofilm deficient phenotype and serves as a positive control. We also included the natural product-based anti-biofilm composition "220D-F2", which has been shown to inhibit biofilm formation in both Staphylococcus aureus [6] and Streptococcus pneumoniae [54], as a positive drug control. Briefly, following inoculation and addition of appropriate media (containing extract or vehicle alone), 96-well plates (Falcon 35-1172) were incubated for 22 hrs at 37°C. The wells were gently washed with phosphate-buffered saline (PBS), fixed with ethanol, stained with crystal violet, rinsed in tap water, and the stain eluted into ethanol and transferred to a new plate prior to quantification of the eluate at an OD 595 with a Cytation 3 multimode plate reader (Biotek).
The cytotoxic effects of bacterial supernatants and a positive control, Staurosporine (Sigma), were further examined with the Viability/Cytotoxicity Assay Kit (Biotum, Hayward, CA). Cells were plated in 24-well plates with glass coverslips. Cells were plated and grown to 90-95% confluence glass coverslips in 24 well plates (Costar 3526) before the addition of treatments. Cells were either treated with 14% (v/v) spent bacterial supernatants or 7.1μM staurosporine for 3 hrs, and then stained following manufacturer's fluorescence microscopy protocol. After staining, the glass slides were mounted using ProLong Gold and fluorescence was assessed using a DMRXA2 microscope (Leica) with narrow band pass Texas Red and FITC filters. Images were collected with ORCA-ER digital camera (Hamamatsu) and processed using Simple PCI software (Hamamatsu) and ImageJ software (National Institutes of Health Research Services Branch, Bethesda, MD, USA).

Mice and S. aureus skin infection model
C5Bl/6 dams were purchased from Charles Rivers (Wilmington, MA). Mice were allowed to acclimate to the BSL-2 level animal housing facility at the University of Iowa (Iowa City, IA) for at least seven days, prior to their inclusion in this study. All animal work described herein was approved by and conducted in accordance with the recommendations of Animal Care and Use Committee at the University of Iowa (IACUC # 1205097). At D0, [8][9][10][11][12] week old mice were anesthetized with isoflurane, abdominal skin was carefully shaved with an Accu-Edge microtome blade (Sakura-Finnetek, Torrance, CA) and exposed skin was cleansed by wiping with an alcohol prep pad (Covidien, Mansfield, MA). For inoculum preparation, a USA 300 MRSA strain (AH1263) or its deletion mutant (AH1292) were grown in TSB medium overnight at 37°C in a shaking incubator set to 200 rpm. Log-phase bacteria were obtained after a 2 hr subculture of a 1:100 dilution of the overnight culture in TSB. Bacterial cells were pelleted and resuspended in DPBS to a concentration of 1x10 8 CFUs/45 μL. 50 μL inoculum suspensions containing 1x10 8 CFUs and either 224C-F2 (5 μg, or 50 μg diluted in DMSO) or DMSO alone were injected to intradermally into abdominal skin using 0.3 mL/31 gauge insulin syringe (BD, Franklin Lakes, NJ). Infectious dose was confirmed by plating serial dilutions of inoculum on TSA and counting ensuing colonies after overnight culture. Baseline body weights of mice were measured before infection and every day thereafter for a period of 7 days. For determination of lesion size, digital photos of skin lesions were taken daily with a Canon Rebel Powershot (ELPH 330 HS) and analyzed via ImageJ software (National Institutes of Health Research Services Branch, Bethesda, MD, USA). Following infection, mice were monitored daily for signs of overt distress that had been pre-established as humane endpoint criteria e.g., weight loss exceeding 20% of baseline (D0) body weight, hunching, loss of mobility and ruffled fur. As no such signs of distress were observed in the present study, all animals were euthanized via continuous administration of 100% CO 2 at the experimental end point.

Statistical analysis
All tests were performed in triplicate and repeated on at least two different occasions. Pair-wise testing was performed based on the Student's t test in Microsoft Excel.

Isolation of a highly bioactive fraction: 224C-F2
Fractionation of the crude Castanea sativa leaf extract (224) was guided by measures of bioactivity, selecting for fractions that exhibited quorum quenching with little to no growth inhibitory activity, Fig 2. This was measured through use of reporter strains for agr types I-IV. To create fractions for testing, extract 224 was suspended in water and partitioned in succession using hexane, ethyl acetate and butanol. The ethyl acetate partition (224C) was determined to be the most bioactive under these testing parameters and was selected for further fractionation with a flash chromatography system using a gradient of hexane, ethyl acetate and methanol. The most active fraction (224C-F2) was selected for further testing and chemical characterization, described below.
224C-F2 inhibits S. aureus quorum sensing across the diversity of agr alleles A number of in vitro assays were employed to guide fractionation of the natural product composition and to evaluate efficacy in blocking S. aureus quorum sensing mediated virulence. Growth inhibitory impact of the extracts was assessed with traditional static MIC assays ( Table 2); growth inhibition was also tracked in the fluorescent reporter assays for agr activity (Fig 3). A slightly higher level of growth inhibition was observed in the static MIC assays over that observed in the super-aerated reporter assay, but in all reporter strains, the MIC remained >100 μg mL -1 for 224C-F2. Limited biofilm inhibitory activity of the extracts was noted ( Table 2).
Quorum quenching effects for 224C-F2 were observed at IC 50 values of 1.56-25 μg mL -1 , depending upon the strain tested ( Table 3). The most potent quorum quenching activity was observed for agr III (IC 50 of 1.56 μg mL -1 ), and the least for agr IV (IC 50 of 25 μg mL -1 ). Significant inhibition of agr was observed for all agr alleles at sub-inhibitory concentrations for growth, indicating that the quorum-quenching activity is due to specific interference with agr, and not simply the result of a false positive due to growth inhibition.
To verify the observed quorum quenching activity, downstream translational products of the quorum sensing system were assessed. HPLC quantification of δ-toxin (Fig 4A) from the supernatant of a heavy producer of exotoxins (NRS385, a USA 500, agr I, HA-MRSA isolate) revealed significant reduction (p<0.01) in production of δ-toxin in 224C-F2 treated cultures at doses as low as 0.25 μg mL -1 (Fig 4B).
To verify the block in production of additional exotoxins, cultures of strain LAC (AH1263, a USA300, agr I, CA-MRSA isolate) and its isogenic agr (AH1292) and hla (AH1589) mutants were grown in the presence of the extracts and their supernatants were examined in a rabbit red blood cell lysis assay. In this assay, the majority of RBC lysis is attributed to the presence of α-hemolysin in the culture supernatant. The presence of some lytic activity in the Δhla vehicle control suggests that some additional hemolytic activity (~18%) may be due to additional toxins in the supernatant, phenol soluble modulins (PSMs), in particular. Treatment of wild type with 224C-F2 resulted in significant (p<0.001) reduction in hemolytic activity in wild type strain at 6.25 μg mL -1 , and almost total loss of hemolytic activity at the concentration of 100 μg mL -1 . Treatment of the Δhla mutant demonstrated nearly total loss of hemolytic activity at 6.25 μg mL -1 (Fig 4C). Similar to the hemolysis assessment, when USA300 is exposed to increasing doses of all extracts (224, 224C, and 224C-F2), the level of α-hemolysin protein production is markedly attenuated, with the most potent activity exhibited by 224C-F2 (Fig 4D).

224C-F2 blocks S. aureus damage to human keratinocytes
In addition to monitoring the activity of each agr allele and detecting specific downstream products (e.g. α-hemolysin and δ-toxin), we also broadened our scope to capture virulence impact data on any other exotoxins that could be produced through this system. To do this, we exposed HaCaT cells to the sterile-filtered supernatants of treated and control cultures. The difference in cytotoxicity as detected by LDH assay was very clear (p<0.001) for all extracts (224,  224C, and 224C-F2) in comparison to control, and this was evident at doses as low as 0.25 μg mL -1 (Fig 5A). Likewise, images of the HaCaT cells following exposure to the supernatants reaffirmed the lack of exotoxins in the supernatants in 224C-F2 treated cultures (Fig 5B).

224C-F2 does not inhibit the growth of common skin bacteria
We investigated the potential of 224C-F2 to create a state of dysbiosis by inhibiting the growth of specific members of the normal skin microflora. While our studies were restricted to assessing the MICs of Actinobacteria and Firmicutes, we did find that 224C-F2 has little to no growth inhibitory activity against the Actinobacteria (Corynebacterium amycolatum, C. striatum, Micrococcus luteus, and Propionibacterium acnes) and Firmicutes (Staphylococcus epidermidis, S. haemolyticus, S. warneri, Streptococcus mitis, and S. pyogenes) tested (Table 4) at the concentrations required for quorum quenching activity in S. aureus. Of these species, S. warneri was the most sensitive, with an MIC 50 of 32 μg mL -1 ; the MIC 90 was not detectable at the range tested (4-512 μg mL -1 ).

Repeated exposure to 224C-F2 does not lead to resistance
Antibiotic resistance is a major concern in any anti-infective drug discovery initiative. Here, we hypothesized that targeting bacterial virulence with a multi-component botanical therapypotentially containing multiple actives acting on multiple targets-would not be very likely to generate resistance. As reporter strains can lose their effectiveness in tracking activity over multiple passaging days (e.g. due to loss of the plasmid), we chose to design a new method for tracking the quorum quenching efficacy of our lead composition (224C-F2). This was achieved through use of a high toxin output strain (NRS385) that has been shown to consistently produce high levels of δ-toxin in the supernatant. Bacterial growth was monitored by OD 600 and δ-toxin was quantified by HPLC. Data for total peak area measured by HPLC ( Fig 6A) and area adjusted for slight differences in daily OD (Fig 6B) both reflect significant differences between the levels of δ-toxin produced by the treated versus control cultures for 15 days of passaging. Moreover, no trends in the shift of this observation towards resistance were noted.

224C-F2 is nontoxic to HaCaT cells and mouse skin
To investigate the potential for cytotoxic or irritant effects of C. sativa leaf extracts, we treated immortalized human keratinocyte cells with up to 512 μg mL -1 of each extract. In all cases (224, 224C, 224C-F2), cytotoxicity (>30%) was only observed at doses at 8-10 times greater than the dose range necessary for quorum quenching activity, and which also corresponded with the rise in toxicity of vehicle treatment alone (DMSO), with no significant difference in cytotoxicity between the vehicle and extracts (Fig 7A). With regards to the potential for irritant or necrotic effects on murine skin, mice were injected intradermally with either 5 μg or 50 μg and monitored for any visible changes in the skin morphology and weight loss. No changes were noted any day at up to 6 days of post-injection follow-up (Fig 7B).

224C-F2 attenuates MRSA-induced illness in an in vivo skin infection model
The agr quorum sensing system controls staphylococcal virulence factor expression and is required for necrotic skin lesion formation following cutaneous challenge [29,35,37]. Having demonstrated the quorum sensing inhibiting activity of 224C-F2 in vitro (Figs 3-6), we next assessed the efficacy of this composition in a mouse model of S. aureus skin infection. When delivered at the time of infection, 224C-F2 decreased the area of resultant ulcers in a dosedependent manner (Fig 8A and 8B). In addition, 224C-F2 administration significantly attenuated infection-induced morbidity (assessed by weight loss) compared to vehicle treated controls ( Fig 8C). Importantly, mice receiving intradermal injection of 224C-F2 alone did not exhibit any overt signs of dermal irritation or clinical illness e.g., weight loss, malaise, hunching, coat ruffling (Fig 7B and data not shown). Together these data corroborate the in vitro findings and suggest that 224C-F2 impairs MRSA pathogenesis without manifesting local or systemic toxicity.

Chemical characterization of 224C-F2
The percent yield of extract from the dry leaves was 43.98% for extract 224, 2.716% for 224C and 1.155% for 224C-F2 (Fig 2). LC-FTMS analysis of 224C-F2 revealed the presence of at least 94 compounds ( Table 5). The greatest quorum quenching effects of 224C-F2 were observed in the retention time region of 21-49 min (Fig 9), suggesting the presence of several distinct quorum quenching compounds (data not shown). Specifically, there are 22 compounds found in this region, 10 present at >1% relative abundance. These correspond to peak numbers, predicted formulas, and relative abundances of: Putative structures for 7 peaks were determined to be pentacyclic triterpenes (specifically, oleanene and ursene derivatives) based on accurate mass analysis, fragmentation patterns, and comparison with natural product databases (Fig 10), and these collectively represent 16.37% in relative abundance. Of note, while present at relative abundance levels of <1% each, the putative structures of gallotannins (32,33,34) and ellagitannins (39) were also identified in the most active region of 224C-F2 (Fig 11). 224C-F2 was also examined by HPLC-DAD and LC-FTMS for the presence of 5 compounds reported to be found in crude C. sativa leaf extracts [42], and it was determined that 224C-F2 does not contain chlorogenic acid, ellagic acid, hyperoside, isoquercitrin, or rutin.

Discussion
The ethnobotanical approach to drug discovery [55] was used here to identify Castanea sativa leaves as a potential source new anti-infective agents. Through design of a bioactivity-guided fractionation strategy based on limited growth-impact coupled to quorum sensing inhibition, we were successful in creating a highly efficacious botanical composition with universal quenching activity for all agr alleles. To the best of our knowledge, the present work represents the first in-depth investigation of European Chestnut leaf extract for its quorum quenching and anti-virulence effects since its identification as a potential quorum quenching lead [4]. Furthermore, this is the first report of the quorum quenching effects of a botanical composition rich in ursene and oleanene derivatives (Fig 10) against S. aureus. Additional compounds identified in the most active region (at <1% relative abundance each) included putative gallotannins, which share a tri-galloyl structure with varying core sugars (32,33,34), and a putative ellagitannin (39). It is possible that in addition to the pentacyclic triterpenes present in 224C-F2, hydrolysable tannins also contribute to the extract's quorum quenching activity.
European Chestnut leaf extracts have been the focus of a number of studies centered on evaluation of its activity in scavenging reactive oxygen species [42,56] and cytoprotective effects, specifically with regards to protection from UV-damage in skin cells [57]. The examination of European Chestnut leaf extracts with a patch test revealed that with respect to irritant effects, such extracts can be considered as safe for topical applications [58]. The integration of C. sativa leaf extracts into cosmetic compositions has also been patented, and is based on the antibacterial and reactive oxygen species (ROS) scavenging effects of the extract [59]. Our safety studies in both human keratocytes (HaCaT cells) and murine skin (Fig 6) have reconfirmed that this version of European Chestnut leaf extract (224C-F2) can be considered safe for topical applications based on its lack of cytotoxic and irritant effects.
Several layers of evidence in support of the efficacy of C. sativa leaf extracts in blocking S. aureus virulence have been presented. Specifically, we have demonstrated that European Chestnut leaf extracts are effective in blocking production of the translational products of RNAIII, including a number of exotoxins. Overall virulence was quenched as demonstrated by the lack of cytotoxic effects elicited by supernatants of cultures treated with the extract. Importantly, using an in vivo model, we have demonstrated efficacy in attenuating dermonecrosis, even in the absence of adjuvant antibiotics.
This inhibition of virulence and pathogenesis was accomplished without posing growth inhibitory pressures on not only S. aureus, but also a panel of common members of the human cutaneous microbiome. A robust skin microflora is critical to skin barrier health and prevention of disease onset. The majority of the bacterial cutaneous microbiome is represented by Actinobacteria, Firmicutes, Proteobacteria and Bacteroidetes [60]. Much like cases of dysbiosis in gut microflora, broad-spectrum activity against the skin microflora also holds the potential for fostering an environment amenable to the proliferation of pathogenic bacteria [61]. The presence of commensals, like Staphylococcus epidermidis, is essential to state of host innate immunity [62]. Thus, it is noteworthy that 224C-F2 specifically blocks S. aureus virulence without adding selective pressures on major representatives of the cutaneous microbiome.
While it is debatable whether virulence inhibitors will ever serve as stand-alone therapeutics, many agree that their application as adjuvants to existing lines of antibiotics could be a critical tool in this era of rising antibiotic resistance. Specifically, by inhibiting agr, such a therapy effectively blocks the production of an entire suite of diverse staphylococcal toxins, ranging from immune-attacking PSMs, pore-forming hemolysins, and a number of other proteases and lipases that damage the host tissue and weaken the host immune response. This will be of particular relevance to patients faced with toxin-mediated infection, including staphylococcal scalded skin syndrome (esp. in neonates), abscesses, necrotizing fasciitis, sepsis, atopic dermatitis (eczema) and more.
In conclusion, we have demonstrated that a folk-medical treatment for skin inflammation and SSTIs that does not demonstrate "typical" antibacterial activity (bacteriostatic or bactericidal) nevertheless shows great potential for development as a therapeutic due to its ability to specifically target and quench S. aureus virulence. The results of this study are important not only to future antibiotic discovery and development efforts, but are also vital to the validation of this previously poorly understood traditional medicine as an efficacious therapy, and not simply an unsubstantiated relict of folklore. Importantly, this composition was non-toxic to human keratinocytes and no dermatopathology was noted upon administration to murine skin. Moreover, the composition did not inhibit growth of the normal skin microflora, suggesting that its disruptive action on the cutaneous microbiome would be minimal to nil. Future work will focus on evaluation of individual actives within the composition with the aim of determining whether a complex mixture, such as 224C-F2 or a single compound will prove most effective against all agr alleles and which will be least likely to develop resistance when administered under multiple selective pressures, such as for in vivo administration as an antibiotic adjuvant.