A Technology for Developing Synbodies with Antibacterial Activity

The rise in antibiotic resistance has led to an increased research focus on discovery of new antibacterial candidates. While broad-spectrum antibiotics are widely pursued, there is evidence that resistance arises in part from the wide spread use of these antibiotics. Our group has developed a system to produce protein affinity agents, called synbodies, which have high affinity and specificity for their target. In this report, we describe the adaptation of this system to produce new antibacterial candidates towards a target bacterium. The system functions by screening target bacteria against an array of 10,000 random sequence peptides and, using a combination of membrane labeling and intracellular dyes, we identified peptides with target specific binding or killing functions. Binding and lytic peptides were identified in this manner and in vitro tests confirmed the activity of the lead peptides. A peptide with antibacterial activity was linked to a peptide specifically binding Staphylococcus aureus to create a synbody with increased antibacterial activity. Subsequent tests showed that this peptide could block S. aureus induced killing of HEK293 cells in a co-culture experiment. These results demonstrate the feasibility of using the synbody system to discover new antibacterial candidate agents.


Introduction
While there is no perfect understanding of the forces directing evolution of antibiotic resistance a prominent view holds that these issues are in part the consequence of the widespread use of broad spectrum antibiotics [1,2]. It has been proposed that next generation antimicrobial treatments must focus on: developing pathogen-specific antibiotics, greatly improving diagnostics, and expanding the role of immunotherapy [3]. Along these lines, there has been a resurgence of monoclonal antibody (mAbs) based therapeutic development [4,5]. Historically, antibody therapies were the first effective anti-infective agents (e.g. for pneumonia, meningitis, erysipelas). However, their wide usage is restricted by the high cost of development and production of pathogen specific mAbs and the large number of current antimicrobial drugs on the market. Other groups are developing antimicrobial peptides (APs) as a means of avoiding resistance, as there have been few reports of resistance arising to APs [6]. Despite numerous attempts to develop new AP-based therapeutics using either natural [7,8], optimized via amino acid substitutions [9][10][11][12] or dimeric peptides [13,14], only a few products have reached the market [15]. APs have two limitations. One is that there are relatively few for development [16] (131 APs for Gram-negative bacteria and 283 for Gram-positive peptides in Antimicrobial Peptide Database, February 2012). The second is that they generally have high toxicity and broad activity, which is consistent with their evolutionary origin [17].
Our group has previously developed a class of affinity agents called synbodies that are produced by screening the target of interest against a peptide microarray to discover low affinity peptides that are then joined on a scaffold to produce high affinity, highly specific binding agents [18][19][20]. Synbodies can be easily modified to increase affinity [20,21], have an orthogonal functional group that can be used for conjugation to a wide variety of moieties, and should be ideal lead therapeutic candidates. We sought to extend this platform to bacteria in an attempt to create synbodies with specificity towards a target pathogen that can function as new antibacterial candidates. By targeting the bacterial surface, we should reduce the likelihood of the target bacteria developing resistance.
The discovery platform is similar to the platform we have used to develop synbodies to protein targets but with a few important modifications: 1) whole bacteria are screened against the 10,000 random-sequence peptides microarray; 2) pathogen specific peptides with binding or lytic action are identified from a microarray functional screening assay and 3) combining binding and lytic peptides produces synbodies with activity against a particular pathogen ( Figure 1A). By screening whole bacteria, we have the ability to profile any possible pathogen without selecting specific surface components, such as lipoteichoic acids, proteins and peptidoglycans for Gram-positive bacteria or lipopolysaccharides (LPS) for Gram-negative bacteria. The microarray based functional assay distinguishes between peptides with antimicrobial activity and those that bind without affecting growth providing a large source of pathogen specific peptides that can be identified in a rapid manner. We hypothesize that this technology for developing specific antibiotics to any pathogen will provide a straightforward method for designing new compounds. As a test of this system, we developed a synbody against Staphylococcus aureus (SA), the Gram-positive bacteria that is one of the causative agents of hospital acquired bacterial pneumonia [22] to demonstrate a general method to create new antibacterial candidates.

Bacterial Cell Binding to Peptide Microarrays
The basic elements of the synbody process are shown in Figure 1A. In the first step, labeled bacteria are applied to the peptide microarray, which consists of 10,000 peptides of 20aa length. The composition of each peptide is known, but generated by a random number generator using 19 aa (cysteine was omitted). We found that the (3-aminopropyl)triethoxysilane (APTES) microarray surface chemistry used in earlier published reports of screening proteins, antibodies and carbohydrates [18,[23][24][25] showed high levels of non-specific binding to the interstitial regions of the array and low level binding to peptide spots when whole bacterial cells were screened. To address these issues we developed an alternative microarray surface chemistry using a hyperbranched polymer that reduced non-specific binding and increased the peptide density ( Figure S1). We chose a diverse set of bacteria to test the general applicability of this approach and screened Gram-negative bacteria, Escherichia coli O111:B4 (EC) and Pseudomonas aeruginosa (PA); and Gram-positive bacteria, Streptococcus mutans (SM), Staphylococcus aureus (SA) and Bacillus subtilis (BS) against the 10,000 peptide library. In order to identify peptides that bound to the bacteria surface but did not affect the function of the bacterial membrane, we used Cell Tracker Orange (CTO), an internalizing dye that is activated upon entering bacterial cells and remains inside intact cells. Each bacterium was labeled with CTO and screened against the peptide microarray with each experiment performed in duplicate at with four technical replicates per peptide sequence. The raw images represent the same area of five peptide microarrays processed with different bacterial species ( Figure 1B), which bind in a clearly distinguished pattern. A competition experiment was also performed in which the CTO labeled bacterium was screened in the presence of 206 excess of un-labeled bacteria to confirm the bacterial cell binding. Construction of scatter plots [25,26] for both datasets (binding and competition) (Figure 2A) reveal peptides with signals that are at least twofold higher than those in the presence of excess un-labeled bacteria (black filled circles in Figure 2A). Whole bacterial cell binding can clearly be seen in Figure 2B, where EC binding to selected peptides (highlighted in Figure 2A) was visualized by fluorescent microscopy. This assay is a straightforward method to identify peptides that specifically bind a pathogen and works well for both Gram-negative and Grampositive bacteria (Figure 2A,C).

Identification of Antimicrobial Peptides on a Peptide Microarray
While this technique works well to select peptides that specifically bind surface components on the target bacteria, these peptides might not have intrinsic antibacterial activity. As depicted in Figure 1, this concept of the synbody involves combining a high specificity binding peptide with a peptide that has high antibacterial activity, such as a naturally occurring antimicrobial peptide. However, many naturally occurring APs function by disruption of the bacterial membrane and have been evolutionarily optimized to have broad activity and can be quite toxic [17]. We hypothesized that we could use the peptide microarray to identify both binding peptides and lytic peptides. To do so we used an N-hydroxysuccinimide (NHS) ester activated dye, AlexaFluor-555 (AF) to label the primary amines present in the bacterial membrane. Peptides that bind but do not disrupt the bacterial membrane should have signal when detected with both CTO and AF (CTO+AF+) while peptides with lytic activity cause CTO to leak out of cells once the membrane is disrupted (CTO-AF+). The binding/competition assay was performed with SA labeled with both dyes and peptides that showed binding to SA-CTO (black circles) were inspected on the scatter plot with AF data ( Figure 3A). Most of the peptides repeated in the AF data can be considered as binder-candidates (CTO+AF+). However, there are additional peptides (red circles) in this area that may also have lytic activity (CTO-AF+). Examples of functional assay performance with other bacteria EC, PA, BS, SM can be seen in Supplementary Figure  S2. Examples of microarray profiles of CTO and AF labeled bacteria for four different peptides show that the profile of binding peptides and lytic peptides can be clearly distinguished ( Figure 3B). We then analyzed this data to determine the specificity of binding and lytic peptides across the five bacteria tested. Using a linear classifier, we were able to identify specific binding ( Figure 3C) as well as lytic ( Figure 3D) peptide-candidates for SA.

Validation of in vitro Activity of Microarray Selected Peptides
We then tested the correlation between peptides with microarray predicted activity and their ability to inhibit bacterial growth in vitro. Based on their microarray profile, we chose 40 peptides that were identified as binding peptides, lytic peptides, or nonbinding peptides for EC, PA, SA, SM, and BS and tested each peptide in a standard bacterial growth inhibition assay. In this manner, we could assess the performance of the selection process. Peptides were screened at 100 mM ( Figure 4A) and the number of peptides that inhibited bacterial growth greater than 50% is shown in Table 1 along with the average inhibition of each group. As can be seen, more than 50% of peptides that were predicted to have lytic activity from the microarray assay had inhibitory activity in vitro and those peptides showed stronger growth inhibition compared to either peptide binders or non-binding peptides. The correlation between inhibitory activity of immobilized peptides and their behavior in solution was quite good considering that the density of peptide immobilized on a microarray is quite high. It is possible that the high avidity of the peptide on the microarray causes some peptides to have a lytic phenotype when immobilized but not have activity when free in solution. We next determined the minimum inhibitory concentration (MIC) for a subset of peptides that showed growth inhibition (Table S1). From these data, it can be seen that a number of peptides showed broad bacterial inhibition while other peptides had little to no activity. Microarray predicted properties of peptides were additionally confirmed by growing bacterial cultures on the plates after mixing with lytic or binding peptides ( Figure 4B). Peptides that were predicted to be lytic from the microarray screen inhibited bacteria growth while binding peptides did not. Based on these results, the microarray screen does select peptides with antibacterial activity.

Evaluation of a S. aureus Synbody
To develop a S. aureus synbody, we combined a peptide with activity against SA and a SA specific binding peptide ( Figure 5A) similar to the approach described in [27,28]. Peptide RWRRHKHFKRPHRKHKRGSC (peptide RW) was selected for its activity for SA and PA, while peptide DRIFHKMQHK-PYKIKKRGSC (peptide DR) was selected for its high binding to SA on the peptide microarray. This synbody had two times higher antibacterial activity for SA than the original lytic peptide yet did not inhibit growth of E. coli or S. mutans (Table 2). This was not observed for P. aeruginosa where the MIC of the synbody is roughly the same as that of the lytic peptide. While the synbody had activity against B. subtilis, this bacterium was the most susceptible to growth inhibition by individual peptides (Table S1). We also tested the antibacterial activity against three species of bacteria that were not included in the selection process: the closely related commensal bacteria Staphylococcus epidermidis; and two other Gramnegative bacteria, Escherichia coli O157:B7 and Burkholderia thailandensis. It was found ( Figure S3) that the synbody inhibited the growth of S. epidermidis (MIC = 1.75 mM) and E. coli O157:B7 (MIC = 14 mM) while it did not inhibit the growth of B. thailandensis, which is resistant to many antibacterial peptides [29]. Analysis of bacterial growth kinetics ( Figure 5B) illustrates the advantage of the synbody over the original peptides at 25 mM concentration of each. Noticeable growth of S. aureus in the presence of the peptide RW was detected after 4 hours incubation, while in the presence of synbody noticeable growth was detected only after 16 hours. Interestingly, a mixture (not conjugate) of the two original peptides RW and DR also shows improved activity, with bacterial growth suppressed for 9 hours. However, a mixture of peptides is still less effective than the synbody.  The lytic activity of the synbody and the peptide RW was tested by adding each to a S. aureus culture at ,1/3 of log phase at ,1.7610 9 CFU/ml ( Figure 5C). The growth of S. aureus in the presence of the peptide plateaus while bacteria treated with synbody shows a gradual reduction in OD after synbody addition. This suggests a bacteriostatic rather than bactericidal effect of the synbody given that the synbody was added at a concentration that was 7 times the MIC. To further clarify this effect, we performed a kill curve kinetic study of the synbody and each peptide (Table S2). As can be seen, kanamycin has a rapid bactericidal effect (.3 log 10 reduction in CFU/mL) when added at 2 times and 4 times the MIC. However, neither the synbody nor the individual peptides had a bactericidal effect at 4 times the MIC. Thus, we conclude that the synbody has a bacteriostatic rather than a bactericidal effect on S. aureus.

The Synbody Inhibits S. aureus Induced Cell Death
To ensure that the synbody was not highly toxic like many antimicrobial peptides, we tested the synbody for toxicity against mammalian cells using a standard hemolytic assay against murine red blood cells ( Figure 6A) and a growth inhibition assay against human kidney cells (HEK293) ( Figure 6B). It was found that the  synbody did not lyse mouse red blood cells or inhibit the growth of HEK293 cells. These data demonstrate that we can combine a peptide with rather broad growth inhibition activity with a peptide that specifically binds S. aureus to produce a synbody without a widespread toxic effect. After this preliminary toxicity test, we wished to determine if the synbody could protect mammalian cells from bacterial induced cell death. We used a model system where human embryonic kidney cells (HEK293) were co-cultured with S. aureus to simulate S. aureus infection of kidney cells in transplants [30]. Cell viability was determined by cellular ATP content and was normalized to 100% for the cells alone control. We tested a range of conditions at which S. aureus can grow in tissue culture media and found that there was significant decline in HEK293 viability with as little as 1.25610 3 CFU/ml of S. aureus. The final experimental setup consisted of 1.25610 6 CFU/ml of SA (10006 higher than minimal harmful concentration of bacteria) added alone or with synbody or the individual peptides added to at 25 mM. A 180 mM solution of PenStrep was used as a positive control and provided complete protection in this assay. Treatment with the individual peptides offered no protection from S. aureus induced cell death while the synbody treated cells maintained about 70% cell viability ( Figure 6C). The protective effect of the synbody on human cells was confirmed using light microscopy after 24 hours of treatment ( Figure 6D). Large colonies of S. aureus and dead cells are clearly seen in the untreated image while the synbody treated sample has few colonies of S. aureus present. These results demonstrate that the synbody has a protective effect against a simulated S. aureus infection.

Discussion
Here we demonstrate a simple and general method for screening bacteria on peptide microarrays. We have shown that a variety of bacteria can be panned and simultaneously screened for peptides that bind the bacteria or lyse them, using a two dye system. When these peptides are resynthesized and tested in solution, over half manifest the phenotype selected on the array. When a SA specific binding peptide was combined in one synbody with a broadly lytic peptide, the specificity and activity was improved. This improved synbody was relatively non-toxic on mammalian cells in culture. Finally, we show that the synbody is also effective in an in vitro assay for protecting human cells against S. aureus.
The rise of multiple sources of antibiotic resistance creates urgency not only in the development of new antibiotics but on the development of new systems of producing antibiotic candidates. We are focusing on technologies to create species specific antibiotics that would be difficult to develop resistance to. Ideally, this technology could be applied in a systematic method that is both rapid and cost effective. In this work, we have demonstrated the feasibility of such a system to quickly develop an antimicrobial synbody that has a relatively narrow specificity for Staphylococcus aureus.
While others have detected antimicrobial activity on cellulosetethered peptides using bacteria transformed with luciferase gene [31,32] our method offers several advantages over this method: (i) the functional assay allows detection of lytic peptides as well as peptide-binders directly on microarrays; (ii) the small peptide library was characterized prior to spotting, in contrast to peptides synthesized directly on the cellulose support which are not characterized prior to use; (iii) the hyperbranched polymer surface has higher sensitivity than other surfaces; and finally, (iv) bacteria can be profiled in as little as 2 hours. The unique combination of bacterial profiling on a microarray with subsequent selection of binding versus lytic peptides enables a very straightforward process for selection of pathogen specific candidates. The technology for designing a synbody based on a pathogen's microarray profile that combines two or more peptide characteristics is fundamentally new to the antimicrobial field. The flexibility of the platform suggests that it would be possible to use this system to develop antibacterial candidates to almost any pathogen in a rapid manner. It should also be possible to improve the activity and possibly specificity of the synbody for a target bacterium. As has been demonstrated with proteins [18,20] combining two or more peptides can produce synbodies with much higher affinity. A simple mutagenesis protocol to improve peptide features for binding proteins has also been presented [21] which may also be applicable to bacteria. While peptide therapeutics have issues with in vivo stability and pharmacokinetics, there are numerous advances that have been made in the stabilization of peptide therapeutics, that make the synbody approach a viable source of new antibacterial candidates.

Materials and Methods
Animal experiments were conducted following an animal use protocol (1099R) which was reviewed and approved by the Arizona State University Institutional Animal Care and Use Committee. Human tissue culture experiments were approved by Arizona State University Institutional Review Board (Protocol # 0508000152).
Microarray peptides were synthesized by Alta Biosciences Ltd. (Birmingham, UK). Lead peptides were synthesized in-house or by Sigma Aldrich by using Fmoc chemistry and purified to 95% by HPLC. Spectrophotometric measurements were carried out by using a NanoDropH ND-1000, SpectraMax 190 and M5 (Molecular Devices). Microarrays were scanned with ProScanArray HT microarray scanner (Perkin Elmer).

Bacteria
Escherichia coli O111:B4 (ATCC) was grown at 37uC in Difco nutrient broth medium for non-fastidious organisms (Becton, Dickinson and Company 231000). Bacillus subtilis 1A423, Pseudomonas aeruginosa PAO-1 and Staphylococcus aureus UAB637 (kindly provided by Center for Infectious Diseases and Vaccinology (CIDV), the Biodesign Institute at Arizona State University (ASU)) were grown at 37uC in Luria-Bertani broth medium (LB, Fisher) under aerobic conditions. Streptococcus mutans UAB147 Serotype C (kindly provided by CIDV, ASU) and Streptococcus pneumoniae were grown at 37uC in Todd-Hewitt (TH, Fisher) broth medium with 5% horse blood under anaerobic conditions. Escherichia coli O157:H7 (ATCC) and Burkholderia thailandensis (ATCC) were cultured in Mueller-Hinton broth medium (MH, Fisher) while Staphylococcus epidermidis was cultured in TH broth medium.

Peptide Microarray Preparation
Prior to spotting of the peptide library, polymer slides were prepared by: 1) Cleaning glass slides with Piranha solution (70:30 v/v mixture of concentrated H 2 SO 4 and 30% H 2 O 2 . WARNING: Piranha solution should be handled with caution and can detonate if mixed with significant quantities of oxidizable organic materials) for at least 1 hour at low rotation, followed by rinsing with H 2 O and drying. 2) Slides were silanized in 1% solution of 3glycidoxypropyl-trimethoxysilane in anhydrous toluene for 30 minutes at 40uC, and washed with toluene (3 times). 3) Slides were cured for 40 minutes at 120uC. 4) Slides were coated with a solution of 6 mg/mL polyethylenimine in 10% ethanol for 1 hour at room temperature with agitation. 5) Slides were activated with sulfo-SMCC (Pierce Biotechnology, Rockford, IL, USA; Cat# 22622) to create a maleimide-activated surface. 6) The peptide library was then printed using a contact spotter on the activated slides. The maleimide-activated surface reacts with the sulfhydryl group on the peptide's terminal cysteine to orient the peptides on the surface.

Microarray Assays
Before probing, the slides were treated with 90% trifluoroacetic acid (TFA) to remove non-immobilized peptides, blocking groups and probable organic impurities followed by dimethylfuran (DMF), ethanol and deionized water washing. Then, the slides were placed in a humidified chamber and blocked for 1 hour at room temperature with buffer [3% bovine serum albumin (BSA), 0.014% mercaptohexanol and 0.05% Tween-20 in 16Tris buffered saline (TBS)].
Turbidity of overnight cultures was measured at OD 600 . The CFU/mL value was calculated according to the McFarland Equivalence Turbidity standard (Remel, R20421). Cell cultures were diluted to 8610 8 and washed 2 times with 16 phosphate buffered saline (PBS) buffer with 0.05% FBS (Fetal Bovine Serum, Invitrogen 10091-130). CTO staining solution was prepared by adding 500 mL of pre-warmed appropriate media with 10 mM CTO to one tube of washed cells and incubated in foil wrapped tube for 1 hour at 37uC at 250 rpm. Alexa Fluor 555 NHS ester (Invitrogen A32755) labeling solution was prepared by adding the 500 mL of 16TBS/FBS with the content of one pre-packed dye vial dissolved in 10 mL DMSO to washed cells. The sample was incubated in a foil wrapped tube for 1 hour at room temperature with agitation. After staining/labeling, cells were washed with 16TBS/FBS. The amount of dyes and incubation times vary for different pathogens and needed to be found experimentally.
After blocking the slides were washed with 16TBS-T (1630 inversion in a Coplin jar) and water (3630 inversions in a Coplin jar). The slides were then dried by centrifugation at 1500 rpm for 2 minutes. Agilent hybridization chambers were used to ensure the interaction of the solution (10 8 labeled cells in 16TBS with 0.03% SodimM azide, 3% BSA and 0.05% Tween 20 in total 450 mL) with the microarrays. To subtract false positive non-specific signals driven by dye binding, we conducted competitions with 206 excess of un-labeled cells. Each microarray assay was performed in triplicate. The slides were incubated for 1 hour at 37uC in the rotator (Agilent Technologies). Then slides were washed with 16TBS-T (3630 inversions in a Coplin jar) and water (3630 inversions in a Coplin jar); the solution was changed each time.
Finally, the slides were dried by centrifugation at 1500 rpms for 2 minutes and scanned.

Microarray scanning and data analysis
Microarrays were scanned by using a Perkin-Elmer ProSca-nArray HT Microarray Scanner with the 488, 543 and 633 nm excitation lasers at 100% power and 70% photomultiplier tube gain. Detection was done at 570 nm for Cell Tracker Orange and AF555, at 508 nm for SYTO 9 and at 670 nm for DRAQ5. All scanned images were analyzed by using GenePix Pro 6.0 software (Axon Instruments, Union City, CA, USA). Upon careful visual inspection, bad spots were eliminated by flagging them absent. Median spot intensities were used in further analyses. Imageprocessed data were imported from GenePix for the following statistical analysis of microarray data to GeneSpring 7.2 (Agilent, Inc., Palo Alto, CA, USA). For correct analysis, each slide was normalized either to 50th percentile or by subtracting the local background from median intensity at each spot. Measurements of less than 0.01 were set to 0.01. The microarray profiles collected for each bacterial strain were compared by using scatter plots [25,26]. Binding assay data were plotted versus competition assay data for each dye separately. Peptides that demonstrated at least 2 times higher intensity in binding assay were considered for subsequent analysis. Peptides with signals at both CTO and AF555 dyes (CTO+AF+) were considered as binders. Lytic peptides were selected as those with no signal in the CTO channel and signal in the AF555 channel (CTO-AF+). In order to distinguish specific peptide-binders and APs candidates the profiles of different strains were compared by Venn diagrams.

Fluorescent Microscopy
For the microscopic detection of bacterial binding to the peptide microarray we printed custom slides with 10-20 peptides of interest. All procedures of microarray preparation and processing were the same as described above. After the last wash, 50 mL of 16PBS was applied to the slides and spread under a cover slip. Binding was evaluated using fluorescent microscopy (Olympus BX61), at 660 magnification with immersion oil with Cy3 excitation laser. Digital images were collected using factorysupplied software DP Controller 2.2.1.227 Olympus Corporation.

Synbody Synthesis
Bivalent synbody was synthesized via a modified divergent solid phase peptide synthesis using Fmoc-Lys(ivDde)-OH as the scaffold using the protocols outlined in [19]. Synthesis was performed by removal of the Fmoc-protecting group followed by synthesis of peptide 1 on a-amino group of Lysine through stepwise addition of Fmoc amino acids. Upon completion of peptide 1 synthesis, the Nterminal Fmoc group was substituted with Boc group prior to deprotection of the Ne-(ivDde) protecting group. The stepwise assembly of peptide 2 was then accomplished at Ne-lysine position using stepwise addition of Fmoc-protected amino acids on the peptide synthesizer. The final protected synbody was treated with cleavage cocktail for 2 hours at room temperature and precipitated in cold diethyl ether. The solid was separated from diethyl ether by centrifugation and the top phase decanted off and pellet resuspended with another addition of dry diethyl ether. The cooling and centrifugation processes were done in triplicate, as the construct was dried and dissolved in water for HPLC purification. Finally, the synbody was purified by HPLC and quality was analyzed by MALDI mass spectrometry.

Minimum Inhibitory Concentration (MIC) and Kill Curve Kinetic Analysis
MIC determinations were conducted utilizing the broth microdilution assay according to the Clinical and Laboratory Standard [33] in Mueller-Hinton (MH; Fisher) broth medium at 3562uC for EC, BS, PA, SA; in MH II (cation adjusted) with 5% horse blood for SM. CFU number controls and survivors control in bactericidal kinetics were conducted in MH agar for EC, BS, PA, SA; in MH II agar with 5% sheep blood for SM. For time kill curve studies, aliquots from each sample-treated well were removed at the indicated time point, diluted with 16 PBS, and serial dilutions of each sample were plated on agar plates. The plates were incubated overnight and bacterial colonies were counted. The results reported are the average of two experiments.

Hemolytic assay
The protocol was adopted from Shin et al [34]. Female BALB/ C mice were obtained from Charles River and housed in barrier isolation caging with food and water provided ad libitum. Mouse blood samples were collected via submandibular venipuncture using a 5.0 mm lancet (MEDIpoint, Inc., Mineola, NY) into heparinized tubes. All animal experiments were conducted following an animal use protocol (1099R) that was reviewed and approved by the Arizona State University Institutional Animal Care and Use Committee. Briefly, fresh mice erythrocytes were rinsed three times with PBS, centrifuged for 15 minutes at 900 g and resuspended in PBS. Samples (100 mL) of the suspension (4% in PBS, v/v) were plated in 96-well microtiter plates, after which 100 mL of the appropriate concentration peptide dissolved in PBS was added. Plates were incubated for 1 hour at 37uC and then centrifuged at 1000 g for 5 minutes. Aliqots (100 mL) of the supernatant were transferred to 96-well plates, where hemoglobin release was monitored using a microplate reader (Molecular Devices) by measuring the absorbance at 414 nm. Percent hemolysis was calculated by the following formula: % hemolysis = [(A414in the peptide solution -A414in PBS)/(A414in 0.1% Triton-X 100 -A414in PBS)] 6100. Zero and 100% hemolysis were determined in PBS and 0.1% Triton-X 100, respectively.

Cytotoxicity assay
HEK293 (Human Embryonic Kidney cells were purchased from ATCC) (10 5 /mL cells) were seeded in individual wells of a microtitre plate and incubated for 24 hours at 37uC with 5% carbon dioxide in EMEM supplemented with 10% fetal bovine serum and Penicillin/Streptomycin. Cells were then challenged with synbody or peptides from 0.05 to 100 mM for either 24 or 48 hours. Cell viability was measured spectrophotometrically (450 nm) following the addition a peroxidase-conjugated anti-BrdU antibody, subsequent TMB degradation by peroxidase and stopping by acidic solution, as per the manufacturer's recommendations (Cell Proliferation ELISA, BrdU, Chemicon (Millipore)). Human tissue culture experiments were approved by Arizona State University Institutional Review Board.

Co-culture of human embryonic kidney cells (HEK293) and S. aureus
Done accordingly to protocol [35] with modifications. Bacteria were grown on Mueller-Hinton (MH; Fisher) broth medium at 37uC for overnight. The bacteria were then harvested by centrifugation and the pellet suspended in Eagle Minimum Essential Medium (EMEM) supplemented with 10% fetal bovine serum without antibiotics. The bacterial density was adjusted to 10 8 CFU/mL (stock). Serial dilutions of this concentration were prepared for further experiments.
HEK293 cells were cultured in complete EMEM (Minimum Essential Medium Eagle Media) supplemented with 10% fetal bovine serum and Penicillin/Streptomycin for 2 days in 75 cm 2 cell culture flasks (Greiner) at 37uC in humidified atmosphere containing 5% CO 2 . For experiments, the cells were harvested through trypsin-EDTA treatment, seeded into 96 well tissue culture plates ''Microtest'' (Falcon) at concentration 4610 5 cells/ mL. After 24 hours, the culture medium was replaced by either fresh EMEM without antibiotics (negative control) or S. aureus dilutions. To make the test conditions for the Synbody more restrictive, 10006 excess of minimal harmful concentration of S. aureus (1.25610 6 CFU/mL) was chosen for the set up of the coculture system. Synbody and peptides RW and DR (25 mM) were added at 1 minute after starting the co-culture. After 24 hours of co-culture, the cell viability was determined visually and on the basis of a luminescence ATP detection assay ''ATPlite'' (Perkin Elmer). For the measurement of cellular ATP the cells were lysed (50 mL of mammalian cells lysis solution to 100 mL of cell suspension) for 5 minutes in an orbital shaker at 700 rpm. After lysis, a 50 mL of substrate solution containing Luciferase/Luciferin was added to react with released ATP. After 5 minutes incubation in an orbital shaker at 700 rpm, the emitted light was measured with a luminometer (Clarity, BioTek Instruments, Inc.). The ATP standard sample provided in ATPlite kit was diluted and measured to build a standard curve. Luminescence was converted to the cellular ATP content (nM) using the standard curve. The cellular viability under test conditions was expressed as percent of an untreated control (HEK293 cells). Annotated dark dots are peptide-binders detected previously with CTO for each strain specifically. Peptides are classified ''Binders'' if repeated with AF (CTO+AF+) out of twofold compared to negative control. Other peptides in this area (red dots) have profile ''CTO-AF+'' and classified ''Lytic''. Annotated peptides (black filled circles) within 2-fold change were ignored as CTO false positive signals. Note that some overlap in properties binder/lytic is possible when signal ratio AF/CTO is exceeding 1.5 for the peptide classified as binders and getting less than 2 for lytic peptides. (TIF) Figure S3 Bacterial growth inhibition assay for synbody (red), peptides DR (green) and RW (black) for A) S. epidermidis B) E. coli O157:B7 C) B. thailandensis. (TIF)