Novel Phenol-soluble Modulin Derivatives in Community-associated Methicillin-resistant Staphylococcus aureus Identified through Imaging Mass Spectrometry*

Background: Phenol-soluble modulins (PSMs) are small peptides of Staphylococcus aureus with immunosuppressive and antimicrobial properties. Results: Imaging mass spectrometry (IMS) identified PSM derivatives with properties different from those of the parent forms. Conclusion: S. aureus generates truncated PSMs with altered antimicrobial and immunostimulatory properties and aureolysin may contribute to processing of some PSMs. Significance: Observations using the technology of IMS expand our understanding of S. aureus PSMs. Staphylococcus aureus causes a wide range of human disease ranging from localized skin and soft tissue infections to potentially lethal systemic infections. S. aureus has the biosynthetic ability to generate numerous virulence factors that assist in circumventing the innate immune system during disease pathogenesis. Recent studies have uncovered a set of extracellular peptides produced by community-associated methicillin-resistant S. aureus (CA-MRSA) with homology to the phenol-soluble modulins (PSMs) from Staphylococcus epidermidis. CA-MRSA PSMs contribute to skin infection and recruit and lyse neutrophils, and truncated versions of these peptides possess antimicrobial activity. In this study, novel CA-MRSA PSM derivatives were discovered by the use of microbial imaging mass spectrometry. The novel PSM derivatives are compared with their parent full-length peptides for changes in hemolytic, cytolytic, and neutrophil-stimulating activity. A potential contribution of the major S. aureus secreted protease aureolysin in processing PSMs is demonstrated. Finally, we show that PSM processing occurs in multiple CA-MRSA strains by structural confirmation of additional novel derivatives. This work demonstrates that IMS can serve as a useful tool to go beyond genome predictions and expand our understanding of the important family of small peptide virulence factors.

Systemic infection by a bacterial pathogen is the result of a stepwise process achieved by circumventing the host immune response. The ability to evade innate immunity, in part, is defined by the metabolic output of the bacterium. Specific components of a pathogen's metabolic output, experimentally proven to contribute to infection, are defined as virulence determinants. Among the virulence determinants used by bacterial pathogens to establish themselves on or within a host are molecular entities possessing adherence, invasion, inflammatory, cytotoxic, or molecular mimicry functions (1)(2)(3). One of the most prominent human pathogens, the Gram-positive coccus Staphylococcus aureus, has evolved to synthesize and express a large number of virulence determinants. This abundance of molecular weaponry contributes to the significant morbidity and mortality associated with S. aureus infections.
A challenge to public health, S. aureus is increasingly resistant to first-line antibiotics, as exemplified by methicillin-resistant strains (MRSA). 4 Hospital-associated MRSA is a longstanding concern within hospitals and healthcare centers, primarily due to transmission between healthcare workers, patients, and surgical instrumentation (4). However, over a decade ago, the epidemiology of MRSA infection shifted, as infections also spread to the community (CA-MRSA), frequently arising in healthy individuals with no predisposed risks (5). CA-MRSA strains typically present in localized skin or soft tissue infections but also can spread systemically to cause potentially life-threatening diseases, including sepsis and endocarditis (6). Soon after the emergence of CA-MRSA, the pore-forming cytotoxin Panton-Valentine leukocidin, present in many strains, was suggested to be the critical underlying determinant of infection (7). Controversy has arisen surrounding this proposal since murine models of infection using isogenic Panton-Valentine leukocidin-deficient CA-MRSA strains failed to show conclusive virulence contributions (8), and the role Panton-Valentine leukocidin plays in the infectious cycle of CA-MRSA continues to be examined in differential infection models with differing results (9).
In 2007, Wang et al. (10) characterized a set of ribosomally encoded peptides related to the phenol-soluble modulins (PSMs) found in Staphylococcus epidermidis and demonstrated their contribution to CA-MRSA virulence in the murine model. Candidate S. aureus PSMs are present within the genome sequences of all of the available sequence strains as short open reading frames, which many gene identification computer algorithms do not easily recognize. Using analytical methods, it was shown under in vitro culture conditions that CA-MRSA strains produced higher concentrations of PSMs compared with hospital-associated MRSA, prompting the hypothesis that these peptides contribute to the unusual degree of virulence among CA-MRSA strains. Targeted mutagenesis studies showed a significant reduction in the ability of PSM knock-out strains to cause infection in murine abscess and bacteremia models. More recently, the same investigators determined the existence of truncated forms of PSMs with antimicrobial activity against other Gram-positive bacteria. It was proposed that the truncated PSMs contribute to CA-MRSA niche establishment by providing a competitive advantage over other microflora, thereby allowing skin colonization, which predisposes to subsequent soft tissue infection (11).
Liquid chromatography and mass spectrometry was previously used to identify PSMs among the S. epidermidis proteins differentially regulated by quorum sensing during growth in liquid culture (12). In this paper, the emerging technique of microbial imaging mass spectrometry (IMS) was applied to a representative strain of the epidemic CA-MRSA USA300 clone to gain insight into the metabolic fingerprint displayed by this organism when grown as colonies on agar. Spatial distribution patterns of the observed ions showed differential arrangements between neighboring ion masses, particularly within the 1000 -3000 Da range. Investigations into the identity of the ions by tandem mass spectrometry demonstrated that they belonged to the PSM␣ family. Ions with spatial distributions localized to the CA-MRSA colony were determined to be the previously characterized full-length PSM␣1 to -4 and the pore forming cytolysin ␦-toxin or PSM␥ (13). The ions with distributions extending beyond the colony were identified as novel truncated PSM␣ derivatives, herein annotated as dPSM␣1 and dPSM␣4. In this work, we present a functional analysis of the antimicrobial, cytotoxic, and immunostimulatory properties of these novel truncated PSMs, study the potential contribution of the highly conserved S. aureus exoprotease aureolysin in their processing, and demonstrate how PSM truncation occurs in multiple MRSA strains by discovery and structural confirmation of other derivatives.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Culture Conditions-S. aureus USA300 strains TCH1516 (14), LAC (15), and UAMS1182 (provided by G. Somerville at the University of Nebraska, Lincoln, NE), and USA1000 strain S. aureus ST59 (16) were used in the study. Additionally, the strains S. epidermidis ATCC35984 (17) and group A Streptococcus serotype M3 strain 4041-05 (Centers for Disease Control) were studied. All strains were grown in Todd Hewitt broth or Todd Hewitt agar (THA) except S. epidermidis, which was grown in tryptic soy broth. All strains were propagated at 37°C unless indicated in each specific method.
Preparation of CA-MRSA Samples for Imaging Mass Spectrometry-THA was prepared and 10 ml of autoclaved agar was poured into Petri dishes under sterile conditions. Colony growth was initiated by spotting 3 l of an overnight CA-MRSA TCH1516 culture on the prepared THA plates. The bacterial colony was allowed to grow for 72 h at 30°C, excised, and transferred to a Bruker MSP 96 MALDI anchor plate. Thereafter, the target plate was uniformly covered with Sigma universal matrix (␣-cyano-4-hydroxycinnamic acid and 2,5-dihydroxybenzoic acid) by the use of a 50-m sieve. Once the sample was uniformly covered with matrix, it was placed in a 37°C oven for 3-4 h or until dried.
Imaging Mass Spectrometry-The Bruker MSP 96 anchor plate containing the sample was inserted into a Microflex Bruker Daltonics mass spectrometer outfitted with the Compass version 1.2 software suite (which consists of FlexImaging version 2.0, FlexControl version 3.0, and FlexAnalysis version 3.0). The sample was run in positive mode, with 600-m laser intervals in xy and 60% laser power. A photomicrograph of the colony undergoing IMS was loaded onto the FlexImaging command window. Three teach points were selected in order to align the background image with the sample target plate. After the target plate calibration was complete, the AutoXecute command was used to analyze the samples. The method setting under the FlexControl panel was ImaingRPpepmix, which consisted of the following settings. Laser: Fuzzy  Zoom Range 1.0% Peak Assignment Tolerance-User Defined-500 pap. After acquisition, the data were analyzed using the FlexImaging software. The resulting mass spectrum was filtered manually in 0.5-3.0-absorbance unit increments with individual colors assigned to the specific masses.
Identification of PSM Derivatives-To prepare samples for MALDI-TOF-TOF analysis, 50 ml of Todd Hewitt broth was inoculated separately with a cell stock of CA-MRSA strain TCH1516, LAC, UAMS1182, or ST59 and grown overnight at 37°C with shaking. The saturated medium was then split into two 50-ml conical tubes and extracted with 25 ml of either 1-butanol or ethyl acetate. Extractions were then concentrated to ϳ500 l by the use of a SpeedVac and subjected to gel filtration chromatography using methanol as the mobile phase. Fractions were collected at 15, 20, 25, 40, and 50 ml. Samples were then dried by a SpeedVac and resuspended in a 1:1 mixture of methanol/water in preparation for high pressure liquid chromatography (HPLC). Samples were further separated and purified by the use of an Agilent Technologies 1200 series high pressure liquid chromatograph using a linear gradient from 5% acetonitrile to 95% in 40 min. Fractions resulting from the HPLC run were brought to complete dryness by a SpeedVac and then resuspended in 10 l of the appropriate acetonitrile/ water ratio, depending on the HPLC elution time, to ensure maximum solubility. One l of each collected fraction was then mixed 1:1 with Sigma-Aldrich universal MALDI matrix resuspended in 78% acetonitrile and 0.1% trifluoric acid. One l of the mixture was plated on an Allied Biosystems (ABI) MALDI target plate for analysis by ABI 4800 MALDI-TOF-TOF. Masses that corresponded to potential dPSMs based on intact mass were isolated and fragmented. Ions that were amenable to tandem mass spectrometry were manually annotated by de novo sequencing. Sequences identified to be dPSM␣1 and dPSM␣4 were purchased as synthetic peptides through American Peptide Corporation.
IMS of Bacterial-Bacterial Interactions-CA-MRSA TCH1516, S. epidermidis, and group A Streptococcus (GAS) were grown to A 600 ϭ 0.4 and plated as adjacent spots on either THA or tryptic soy broth agar. The co-cultures were allowed to grow for 48 h at 37°C. After incubation, the agar surrounding the colonies was excised and transferred to a MALDI target plate in preparation for IMS as previously described.
Antimicrobial Assays-Minimum inhibitory concentration (MIC) assays were performed by broth microdilution. Vancomycin and gentamicin (Hospira, Lake Forest, IL) were used as control antibiotics. MIC was determined as the lowest concentration of antibiotic that inhibited bacterial growth by A 600 . In agar inhibition assays, PSM␣1 (20 g), dPSM␣1 (20 g), PSM␣4 (20 g), and dPSM␣4 (20 g) were resuspended in 70% ethanol, and 20 g of compound in a 10-l volume were spotted in triplicate onto prewarmed agar plates and allowed to dry. GAS (2 ϫ 10 5 cfu) was pipetted onto agar at the location of the compounds, and plates were incubated at 37°C for 2 h. Spots were then excised and placed into 2-ml screw cap tubes con-taining 1 mm silica/zirconia beads in 1 ml of PBS. The samples were homogenized in a mini-BeadBeater-8 for 1 min at full speed twice, and the tubes were placed on ice in between. Homogenized samples were serially diluted in sterile PBS and plated on THA plates for enumeration. Inhibition of bacterial growth by purified compounds was calculated as a percentage of initial inoculum.
Assay for GAS Inhibition on Murine Skin-Permission to undertake animal experiments was obtained from the University of California, San Diego Institutional Animal Care and Use Committee. Flanks of 8 -10-week-old female CD1 mice were shaved. The following day, mice were anesthetized with ketamine/xylazine and placed on heating pads to maintain body temperature. Full-length PSMs and dPSMs were resuspended in 70% ethanol alone, and 20 g of compound in 10 l were pipetted onto premarked spots on the mouse flank and allowed to dry. Five spots per mouse were prepared as follows: negative control (70% ethanol), PSM␣1 (20 g), dPSM␣1 (20 g), PSM␣4, and dPSM␣4. GAS was grown to midlogarithmic phase (A 600 ϭ 0.4), washed with sterile PBS, and diluted to 2 ϫ 10 7 cfu/ml. Ten l (ϳ2 ϫ 10 5 cfu) of bacteria were spotted onto agar plates in triplicate and allowed to dry. Once the spots had dried, agar discs were excised using a 6-mm biopsy punch. One agar disc containing 2 ϫ 10 5 cfu of GAS M3 was placed onto each spot (5 discs/mouse) and affixed to the mouse with a Tegaderm transparent wound dressing. After 2 h, both the skin and the agar disc were placed into 2-ml screw cap tubes containing 1-mm silica/zirconia beads in 1 ml of PBS. The samples were homogenized in a mini-BeadBeater-8 for 1 min at full speed twice, placing the tubes on ice in between. Homogenized samples were serially diluted in sterile PBS and plated on THA plates for enumeration of CFU. Inhibition of bacterial growth by purified compounds was calculated as a percentage of initial inoculum. Statistics analyses for the acquired data sets were performed by one-way analysis of variance.
Neutrophil Lysis Assays-Blood was drawn from healthy donors after informed consent, and neutrophils were isolated using the PolyMorphPrep kit as per the manufacturer's instructions (Axis-Shield, Rodeløkka, Norway). The Promega Cyto Tox 96 non-radioactive cytotoxicity assay kit was used to monitor lactate dehydrogenase release from neutrophils upon incubation with PSMs or dPSMs. Briefly, a round-bottom 96-well plate was used to mix the following reaction mixtures: (i) no cells or medium only, (ii) untreated neutrophils, (iii) lysis control, (iv) DMSO, (v) PSM1 (100 and 10 g/ml), (vi) dPSM1 (100 and 10 g/ml), (vii) PSM4 (100 and 10 g/ml), (viii) dPSM4 (100 and 10 g/ml). Each well contained 2.5 ϫ 10 6 neutrophil cells/ml in RPMI plus 2% heat-inactivated FBS. After mixing, PSM peptides as well as all controls were incubated at 37°C and 5% CO 2 . The lactate dehydrogenase release assay was measured at 60 and 120 min by reading A 490 .
Neutrophil Extracellular Trap Assays-Neutrophils at 2 ϫ 10 5 cells/well were pipetted into a 48-well tissue culture plate in RPMI. Twenty-five nM phorbol 12-myristate 13-acetate stimulation was used as a positive control to monitor extracellular trap formation. Neutrophils incubated in DMSO were used as a background control. PSM1 and PSM4, including their corresponding truncated derivatives, were incubated for 180 min at concentrations of 100 and 10 g/ml. Thereafter, 500 milliunits of micrococcal nuclease in a volume of 50 l/well were added for 10 min. The reaction was then halted with 5 mM EDTA. The plate was then centrifuged at 200 ϫ g for 8 min. After centrifugation, 100 l of the supernatant were transferred to 96-flat bottom well plates in preparation for the Quant-iT Picogreen (Invitrogen) assay. The Quant-iT Picogreen assay was performed as described in the provided protocol. Briefly, Picogreen reagent was diluted 1:200 and was mixed 1:1 with samples. The mixture was allowed to incubate for 5 min at room temperature in dark conditions. After incubation, the excitation 480 nm, emission A 520 nm (fluorescein) for each well was recorded.
Fluorescence Microscopy of Neutrophil Extracellular Traps (NETs)-For visualization of extracellular traps, neutrophils were placed on poly-L-lysine-coated glass slides and treated with PSMs or vehicle controls. NETs were visualized using a rabbit antimyeloperoxidase antibody followed by a secondary goat anti-rabbit Alexa 488 antibody; samples were embedded in 4Ј,6-diamidino-2-phenylindole (DAPI) to counterstain DNA in blue. Mounted samples were examined using an inverted confocal laser-scanning two-photon microscope, Olympus Fluoview FV1000, with Fluoview TM spectral scanning technology (Olympus) and a ϫ20/0.75 Olympus objective.
Hemolysis Assays-Defribinated sheep or human whole blood was washed in PBS and diluted to a final concentration of 1:25 (v/v) in PBS. Whole blood (100 l) was then placed into individual wells of a flat-bottom 96-well microtiter plate (Costar, Corning). PSM1, dPSM1, PSM4, and dPSM4 at a concentration of 100 and 10 g/ml were added directly to the wells, and the mixture was incubated for 60 min at 37°C. After incubation, plates were centrifuged at 500 ϫ g for 10 min, and an aliquot of supernatant (100 l) was placed in a separate microtiter plate to measure hemoglobin absorbance at 450 nm on a microplate reader. One l of the erythrocyte and PSM mixtures was aliquoted into separate Eppendorf tubes and examined by a Riveal contrast microscope (Quorum Technologies) to confirm erythrocyte lysis.
Generation of CA-MRSA Aureolysin Mutant-The wild type strain LAC is an isolate of the MRSA USA300 clone. A markerless aur deletion was created in each strain by transduction of the pKOR1⌬aur plasmid (18) by 80␣ phage and utilization of the pKOR1 knock-out strategy, as described previously (19). To construct the aureolysin complementation plasmid, pDB59 (20) was digested with BamHI and EcoRI to remove the P3 promoter driving YFP. Oligonucleotides CLM525 (5Ј-gatccatcgatcgatgctagcattcgatcatg-3Ј) and CLM526 (5Ј-aattcatgatcgaatgctagcatcgatcgatg-3Ј) were annealed and ligated into pDB59 digested by BamHI and EcoRI in order to introduce a NheI site. This intermediate plasmid was designated pCM46. The aur gene was amplified from the USA300 LAC genome by PCR using primers CLM529 (5Ј-gttgttggatccgttatctcacatatttcaagcattg-3Ј) and CLM522 (5Ј-gttgttgctagcggtataccccatatataagctg-3Ј). Digestion of the PCR product by BamHI and NheI and ligation with pCM46 digested by the same enzymes yielded complementation plasmid pAur. The aureolysin mutant was subsequently transformed with pAur by electroporation.
Dried Droplet MALDI-MS of LAC Strains-Wild type LAC, LAC ⌬aur, and LAC ⌬aur ϩ pAur were prepared and analyzed with MALDI-MS as described previously. Repeated experiments were performed to generate a mass list of the ions observed within a 400 -5000 Da range for each strain. Ultimately, each mass list for a particular strain was compiled and used to identify potential dPSMs based on intact mass.
Recombinant Aureolysin Cleavage Assays-Recombinant aureolysin was purchased from Biocentrum and resuspended as recommended in the factory protocol. Thereafter, activity was assessed by incubating 2 g of aureolysin with 4 g of either PSM1 or PSM4 in 20 mM Tris-HCl, pH 7.8, containing 5 mM CaCl 2 . Control samples were included to further verify aureolysin activity on PSMs. Observation of PSM cleavage was performed by dried droplet MALDI-MS.

RESULTS
Imaging Mass Spectrometry of CA-MRSA-The technique of thin layer agar IMS is an effective tool for characterizing microbial metabolic output within a defined mass range (21,22). The strategy lends a molecular view, in particular spatial distributions, of the different small molecules and peptides an organism produces when grown on solid agar media and exposed to abiotic or biotic factors (23). Our goal was to apply IMS to study the metabolic behavior of the notable human pathogen, CA-MRSA. To this end, a representative strain (TCH1516) of the epidemic CA-MRSA USA300 clone was grown on agar medium and then subjected to IMS in order to gain a molecular fingerprint of the bacterial pathogen within a 400 -3500 Da mass range (supplemental Fig. S1). The resulting IMS data revealed several small molecules (Ͻ1000 Da) and a set of molecular entities falling in the range of peptides (1000 -3000 Da). Many of the observed molecular species had spatial distribution patterns, indicating them to be physically bound to, or contained within, the CA-MRSA colony, whereas others had spatial distribution patterns revealing the ions to be released beyond the colony into its environment. The identities of the molecular entities observed in the mass spectral fingerprint were pursued, with a particular interest in the peptides that displayed a spatial distribution consistent with release.
Identification of PSM Derivatives-To identify the released peptides observed by IMS, CA-MRSA was grown in liquid broth medium overnight and extracted with various solventsolvent systems as a preparatory step. Samples were further processed by HPLC and then analyzed by dried droplet MALDI-MS/MS. Ions within the 1000 -3000 Da range were targeted for fragmentation to gain structural information in the form of a peptide sequence tag (24). Data were collected for those ions that were amenable to collision-induced fragmentation (supplemental Figs. S2-S8). BLAST analysis was performed on the acquired peptide sequence tags searched against the CA-MRSA USA300 TCH1516 genome. The initial BLAST analysis identified two of the acquired sequence tags related to the previously characterized PSM␣1 (10). One of the fragmented peptides was identified as full-length PSM␣1, based on the BLAST analysis and the corresponding observed intact mass at an error of 170 ppm. The released ion detected by IMS at mass 1773 Da corresponded to a five-amino acid truncated form of PSM␣1, with an observed mass error relative to the theoretical value of 112 ppm ( Fig. 1A and supplemental Fig. S2).
BLAST analysis of the remaining peptide sequence tags identified PSM␣2 and PSM␥, whereas a number of others remained unidentified and were not annotated in the CA-MRSA USA300 TCH1516 genome ( Fig. 1C and supplemental Table S1). Therefore, a six-frame translation of the genome was performed, creating a database with protein sequences for every potential open reading frame (ORF). Using the database, the unidentified peptide sequence tags were manually searched by peptidogenomics (25), and PSM␣3 and PSM␣4 were identified (Fig. 1B). Like our observation with the truncated form of PSM␣1, another derivative was identified originating from PSM␣4 (supplemental Fig. S3), again with an IMS spatial distribution extending beyond the bacterial colony (Fig. 1A). With this structural evidence in hand, two PSM␣ derivatives (dPSM␣) at masses 1773 and 1855 Da were commercially synthesized and designated dPSM␣1 and dPSM␣4, respectively, for functional characterization studies.
Antimicrobial Activity of PSM Derivatives-Otto and colleagues (11) suggested that PSM derivatives might serve antimicrobial functions relevant for CA-MRSA niche establishment. To this end, we set up S. aureus bacterial-bacterial interactions, first with the common skin microbe S. epidermidis and second with a leading human skin pathogen, GAS. The CA-MRSA colony achieved similar size and morphology when grown alone or in the co-culture experiment. However, both S. epidermidis and GAS showed inhibition of growth at the bac-terial interface presumably caused by the metabolic output of the CA-MRSA colony (Fig. 2, A and B). When the interactions between CA-MRSA and S. epidermidis or GAS were analyzed by IMS, dPSM␣1 and dPSM␣4 were both localized to the inhibition zones, suggesting that the peptides could contribute to the growth-suppressive phenotype.
Suppression of GAS growth on solid agar was further analyzed using synthetic dPSM␣1 and dPSM␣4, which showed inhibitory activity greater than their parent peptides (Fig. 3A). In MIC testing, peptide dPSM␣1 had activity against GAS at 11 M and S. epidermidis at 42 M but no activity at the highest concentration tested (64 M) against a methicillin-susceptible S. aureus strain (MSSA in Fig. 3C). The parent peptide PSM␣1 showed no activity against any of the strains at the upper concentration limit. Peptide dPSM␣4 showed activity against GAS at 10 M (but not S. epidermidis), whereas the parent peptide PSM␣4 showed no activity. We next tested if bioactivity could be maintained in an in vivo setting. Flanks of CD1 mice were shaved and individually treated with PSM␣1, PSM␣4, dPSM␣1, or dPSM␣4 and subsequently inoculated with GAS. Both dPSM␣1 and dPSM␣4 restricted growth of GAS on mammalian skin at 4 h because a significant decrease in cfu count was produced by both derivatives compared with the parent peptides (Fig. 3B).
Mammalian Cell Lysis Caused by PSM Derivatives-An essential first line of defense in the mammalian innate immune response is the ability of phagocytic cells, such as neutrophils, to accumulate at the site of infection and intercept invading for- eign pathogens (26). One strategy bacterial pathogens have evolved to circumvent neutrophil killing is to secrete enzymes and peptides cytotoxic to the phagocyte. To test if dPSM␣1 and dPSM␣4 exhibited cytolytic activity, neutrophils were isolated from human whole blood and incubated with the peptides at different concentrations and time points. At 10 g/ml through a 90-min exposure, dPSM␣1 showed no cytolytic activity, whereas full-length PSM␣1 parent peptide caused 59% neutrophil lysis. In parallel experiments, the truncated peptide dPSM␣4 caused 13% and its parent peptide PSM␣4 induced 5% neutrophil lysis (Fig. 4A). No neutrophil cytolytic activity of any of the full-length or truncated peptides was detected at a lower concentration of 1 g/ml.
To determine if dPSM␣1 and dPSM␣4 or their parent peptides had hemolytic activity against human erythrocytes (27,28), whole blood was washed and incubated with 10 g/ml peptide for 60 min, and the A 450 was measured to monitor the release of hemoglobin. Peptide dPSM␣1 showed no hemolytic activity, whereas the parent peptide PSM␣1 released 58% hemoglobin compared with the Triton X-100 lysis control. In contrast, the truncated dPSM␣4 had a stronger hemolytic phenotype (52% lysis) than its parent peptide (19% lysis) (Fig. 4B). Similar hemolytic activities were observed using 5% difibrinated sheep blood (Fig. 4B) and validated by microscopy to visualize erythrocyte lysis from each sample (Fig. 4C).
Immunostimulation of Neutrophil Extracellular Traps-Now recognized as a key component of their innate immune function, neutrophils release DNA-based NETs to immobilize bacteria at the tissue site of infection, exposing them to embedded antimicrobial peptides and proteases. Because PSMs and their respective derivatives are released extracellularly in high concentrations by CA-MRSA isolates, we explored their influence on NET formation. As a positive control, neutrophils were incubated with phorbol 12-myristate 13-acetate, and a robust formation of NETs was visualized by microscopy using an anti-myeloperoxidase antibody and DAPI nuclear staining. The parent peptides PSM␣1 and The optical image shows that CA-MRSA has the ability to inhibit the growth (black arrow) of the competing GAS colony. IMS showed that dPSM␣ is released and therefore potentially contributes to growth inhibition. Shown is the spatial distribution of dPSM␣4 (dPSM␣1 is omitted). FIGURE 3. Antimicrobial activity exhibited by PSMs and derivatives. A, inhibition of GAS growth on solid agar. GAS was grown on agar plates in the presence of synthetic PSMs. **, p Ͻ 0.01; ***, p Ͻ 0.001; ns, no significant difference. B, in vivo growth inhibition of GAS on mammalian skin in the presence of PSMs. Inhibition of bacterial growth was calculated as a percentage of the initial inoculum. *, p Ͻ 0.05; ***, p Ͻ 0.001; ns, no significant difference. C, MIC of the synthetic peptides dPSM␣1 and dPSM4. Derivative peptides displayed antimicrobial activity in contrast to their respective parent forms. Numerical values displayed in the table indicate the concentrations tested (M).
PSM␣4 at 10 g/ml both stimulated NET formation to a level comparable with phorbol 12-myristate 13-acetate as observed by microscopy. In contrast, both derivative peptides dPSM␣1 and dPSM␣4 did not demonstrate the ability to stimulate NETs (Fig.  5A). The differential effects on NET stimulation were corroborated quantitatively using Picogreen dye to measure extracellular DNA (Fig. 5B).
MALDI-MS of Multiple MRSA Strain Extracts Identifies Additional PSM Derivatives-Next we set out to determine if other MRSA strains biosynthesized an array of dPSMs similar to those we observed with CA-MRSA USA300 strain TCH1516. To this end, CA-MRSA strains ST59 (USA1000), UAMS1182 (USA300), and CA-MRSA LAC (USA300) were grown, and supernatants were extracted as described previously for MALDI-TOF/TOF analysis. In each strain, several ions were observed, corresponding to dPSM␣ candidates based on intact mass (supplemental Table S2). All ion candidates were subjected to tandem mass spectrometry in order gain structural information in the form of peptide sequence tags, and the generated tags were studied by either BLAST analysis or manually using six-frame translation databases as described above. We found that the majority of the ions were not amenable to tandem mass spectrometry either because the ions were of low concentration or because they were not easily fragmented by collision-induced dissociation. Nevertheless, two additional novel dPSMs, originating from PSM␣2 and dPSM␣4, were structurally verified, as was one novel derivative of PSM supplemental Figs. S9 -S12).
Aureolysin Activity Influences Processing of PSMs-Finally, we performed initial investigations of contributing mechanisms for the proteolytic cleavage of native full-length PSMs into their truncated dPSM form. First, the gene cluster containing the PSMs was examined in various CA-MRSA USA300 genomes to assess adjacent or neighboring genes (supplemental Table S1), recognizing that because PSMs are short peptides, they are commonly missed by sequencing algorithms and often not annotated. We found that PSM␣1 to -4 are located near the cobalamin biosynthesis protein and adjacent to small hypothetical proteins (47 and 42 amino acids) of unknown function. Downstream to these PSMs are proteins annotated as NADH dehydrogenase, PAP2 superfamily phosphatase, carboxyesterase, or hypothetical proteins. The gene encoding PSM␥ is distant from the PSM␣1 gene, contained within the agr gene cluster, and embedded within rnaIII (29). Because no gene immediately adjacent to the PSMs showed clear homology to known proteinases, we explored the potential contribution of the well studied zinc-dependent S. aureus metalloprotease aureolysin. Previous studies have shown that aureolysin is promiscuous in its enzymatic activity, with the ability to cleave its own extracellular cell wall proteins, the human-derived antimicrobial peptide LL-37, and several other substrates (30,31).
To investigate if aureolysin contributes to processing PSMs, we compared the mass spectrometry profile of PSM␣1 and PSM␣4 peptides produced by WT CA-MRSA LAC, its isogenic ⌬aur mutant, and a complemented strain in which the ⌬aur mutant is transformed with an expression vector expressing the aur gene (Fig. 6A). The WT CA-MRSA LAC produced detectable ions at 1855 and 1774 Da as expected; however, the isogenic ⌬aur strain showed only the ability to produce the ion at 1774 Da but not the 1855 Da ion. The presence of the 1855-Da ion was restored in the complemented mutant strain (Fig. 6B). Aureolysin's activity on PSMs was further validated when recombinant protein was incubated with PSM␣4 and PSM␣1. In both cases, peptide truncated products were observed, which were absent in the control samples (supplemental Figs. S13-S15). . Neutrophil and erythrocyte lysis caused by PSMs. A, neutrophils were incubated with 10 g/ml synthetic PSMs and their respective derivatives for 180 min. Lysis was assessed by monitoring the release of lactate dehydrogenase by colorimetric spectroscopy. Absorbances were normalized by the use of a lysis reagent. B, human (white bars) and 5% sheep (black bars) blood was incubated with 10 g/ml synthetic PSMs and their respective derivatives for 60 min. Lysis was then assessed by monitoring the release of heme spectroscopically at A 450 . C, human blood samples incubated with PSMs were further assessed with a Riveal contrast microscope to confirm erythrocyte lysis.

DISCUSSION
We have demonstrated that IMS can be used as a tool to discover previously undetermined biological phenomena by providing a molecular view of the vast number of small molecules and peptides a bacterial colony biosynthesizes during growth. CA-MRSA pathogenesis is the integrated outcome of multiple interactions of bacterial and host factors with different components acting alone or in synergy. The cytolytic peptides known as PSMs were shown to contribute to CA-MRSA pathogenesis, and more recently, the same set of peptides, once proteolytically cleaved, demonstrated a gain of antimicrobial function. Using the strategy of IMS, we were able to uncover previously unidentified PSM derivatives produced by prototypical CA-MRSA strains. The peptides were initially targeted based on the spatial distribution patterns observed by IMS, which showed that the dPSMs were released into the solid agar medium during growth, whereas the parent PSMs remained colony-bound. Once the structural identity of dPSM␣1 and dPSM␣4 was confirmed, we verified that dPSMs have selective antimicrobial activity, a characteristic not displayed by the PSM until proteolytic cleavage of the peptide's N terminus. The potential role that dPSM modifications played in influencing host cells was studied by monitoring neutrophil and erythrocyte lysis, followed by stimulation of NETs. PSM␣1 demonstrated lytic activity against human neutrophils and erythrocytes, but these activities diminished in the truncation to the dPSM␣1 form. Conversely, the modest lytic activities of the dPSM␣4 were significantly greater than those of its parent peptide, PSM␣4. Stimulation of NET production by human neutrophils was readily observed in response to either of the parent PSMs, whereas no detectable stimulation was observed with either dPSM. Thus, the processing of PSM peptides not only influences their ability to diffuse further away from the parent cell but influences their antimicrobial, cytolytic, and proinflammatory capacities. In the two studied PSM peptides, processing was associated with increased antimicrobial activity and decreased NET-stimulating capacity, whereas cytolytic activity was affected differentially in the two instances.
The mechanism for processing PSMs into their truncated form has not previously been investigated, and a candidate FIGURE 5. Immunostimulation of NETs. A, for visualization of NETs, neutrophils were placed on poly-L-lysine-coated glass slides and treated with PSMs or vehicle controls. NETs were visualized using a rabbit antimyeloperoxidase antibody followed by a secondary goat anti-rabbit Alexa 488 antibody; samples were embedded in DAPI to counterstain DNA in blue. B, in parallel with microscopy, the Quant-iT Picogreen assay was used to quantify the extracellular DNA upon stimulation by PSMs and the controls by colorimetric spectroscopy. Error bars, S.D. coordinately regulated protease gene does not appear in the gene cluster containing the PSMs. Our studies using isogenic aureolysin-producing and -deficient strains suggest that expression of the broad spectrum metalloprotease was associated with the appearance of dPSM␣4. Additionally, recombinant aureolysin cleaves PSM␣4 into smaller peptides, although we were not successful in capturing the ion corresponding to dPSM4 at the mass 1855 Da. In contrast, dPSM1 was present in both the WT and aureolysin mutant CA-MRSA strains. We conclude that aureolysin may influence processing of a subset of PSMs. Given that the genome of CA-MRSA encodes several other known proteases, alternative activities probably work together to generate a multitude of dPSM derivatives. The hypothesis is consistent with our analysis of three additional CA-MRSA strains in which a large number of potential dPSMs were identified based on mass alone but not structurally verified, given the peptides were either of low concentration or were not amenable to collision-induced dissociation leading to a peptide sequence tag. Ultimately, the work herein demonstrates that IMS can serve as a useful tool to go beyond genome predictions and expands our understanding of the important family of small peptide virulence factors, such as PSMs in S. aureus, S. epidermidis, and other important human pathogens. Recognition of differences in spatial distribution and biological activities of derivative peptides will aid in elucidating their influence on the establishment of infection and their broader impacts on the pathogen-host interaction.