Strains used in this study are described in Supplementary Table S2. Unless indicated, Staphylococcus spp. and Corynebacterium spp. were routinely cultured using Brain Heart Infusion (BHI) broth or solid agar medium and Escherichia coli was grown on Lysogeny Broth (LB) medium. For indicated experiments, a chemically defined medium (CDM) (Brown and Whiteley, 2007) was used with the addition of 25 mM 2-(N-morpholino)ethanesulfonic acid (MES) buffer at pH 6 or pH 7.5. All cultures were grown at 37°C in standard atmosphere; liquid cultures were shaken at 250 RPM. Antibiotics were used at the following concentrations: ampicillin 100 μg/ml, erythromycin 10 μg/ml, tetracycline 3 μg/ml, fosfomycin 20 μg/ml and chloramphenicol 20 μg/ml for E. coli and 10 μg/ml for S. aureus. We selectively enumerated S. aureus on Mannitol Salts Agar (MSA) and Corynebacterium spp. on BHI agar with 20 μg/ml fosfomycin using colony-forming unit (CFU) measurement via track dilution as described previously (Jett et al., 1997).

DNA and plasmid isolation were performed using standard methods (Ausubel et al., 2002). Restriction endonucleases and DNA modification enzymes were purchased from New England Biolabs. Chromosomal DNA from all bacteria was isolated using DNeasy tissue kits (Qiagen); plasmid isolations were performed using QIAprep spin miniprep kits (Qiagen). For both kits, S. aureus cells were lysed by adding 25 μg of lysostaphin (Sigma–Aldrich) to the kit lysis buffer and incubating at 37°C for 30 m prior to following the manufacturer’s protocol. DNA fragments were purified using QIAquick mini-elute PCR purification kits, and PCR was performed using GoTaq Green (Promega). Primers used in this study are indicated in Supplementary Table S3. DNA sequencing was performed by automated sequencing technology through Macrogen Co. (Macrogen USA).

Markerless gene deletions were generated using a previously established protocol with derivatives of the temperature sensitive S. aureus shuttle vector pKFT (Kato and Sugai, 2011). Surface protein A (spa) deletion was accomplished using pLF048, a pKFT derivative (Foulston et al., 2014). pKFT-derivative vectors were electroporated into S. aureus RN4220 and grown at room temperature for 2 h before plating on BHI tetracycline 3 μg/ml agar medium and incubated at 30°C for ~36 h. Replicating plasmids were purified from S. aureus RN4220 prior to electroporation into S. aureus JE2 and treated as previously described (Kato and Sugai, 2011). Markerless tet^S mutants were confirmed by PCR using primers external to flanking regions of the pKFT derivatives (oKL377 and oKL378 for the spa deletion) followed by amplicon sequencing to verify loss of the targeted gene and religation of the flanking regions.

To express spa in trans in the Δspa mutant, we cloned the spa gene with its native promoter region into pEPSA5 (Forsyth et al., 2002). Using primers oKL554 and oKL555, we amplified the P_spa-spa-containing fragment, then digested it and pEPSA5 with SalI and BamHI (NEB). We electroporated pEPSA5 and pEPSA5-spa into S. aureus RN4220 then repurified and transformed each respectively into KPL2389 to generate the Δspa mutant with the empty vector control (KPL2496) and spa-complemented strain (KPL 2497).

Colony cultures (Zheng and Stewart, 2002) were prepared similar to previously described on agarose medium (Ramsey and Whiteley, 2009). Briefly, CDM 1% agarose solid medium was prepared using 25 mM MES buffered CDM at pH 6.0. Sterile 25 mm 0.2 μm polycarbonate membranes (Millipore #GTTP02500) were placed on solid agarose medium and inoculated with 10 μl of liquid CDM containing either 5 × 10⁶ C. striatum or 1 × 10⁴ S. aureus separately in monoculture or combined in coculture. Inoculated spots were allowed to briefly air dry then transferred to 37°C for 18 h. Next, membranes were aseptically transferred to fresh medium and incubated at 37°C for another 4 h. To harvest cells, membranes were either immediately transferred to RNALater (Ambion) for RNAseq experiments (described below); transferred to 1 ml sterile Phosphate Buffered Saline (PBS), vortexed for 30 s and serially diluted for CFU enumeration; or transferred to 1 ml sterile PBS, vortexed for 30 s, pelleted by centrifugation at 10,000 g for 5 m then resuspended in 300 μl of lysis buffer for β-galactosidase assays (described below).

Bacterial cells resuspended in 1 ml of RNALater (Ambion) (see above) were pelleted by centrifugation (10,000 g at 4°C for 10 m). Cell pellets were stored at -20°C. For lysis, cells were resuspended in 100 μL TE buffer with 10 mg/ml lysozyme and 25 μg/ml lysostaphin and incubated 20 m at 37°C. After adding 900 μl of RNA Bee (Amsbio), the mixture was transferred to 2 ml Lysing Matrix B bead tubes (MP Biomedicals) and bead beat in a FastPrep-24 (MP Biomedicals) 6x at the highest setting for 30 s each, with 30 s of rest on ice between each cycle. Next, 0.2 ml of chloroform was added to each tube and vortexed for 30 s then stored on ice for 5 m prior to centrifugation (12,000 g for 15 m at 4°C). The (upper) aqueous phase was removed and mixed with 0.5 ml of isopropanol in a microcentrifuge tube and incubated 10 m at room temperature (RT). After centrifugation at 12,000 g for 5 m at RT, the supernatant was carefully removed. One ml of ice-cold 75% ethanol was added to each tube, vortexed briefly and then centrifuged for 5 m at 7,500 g. This 75% ethanol wash was repeated, then RNA pellets were air dried at RT for 5–10 m. RNA was resuspended in 30 μl of nuclease-free ddH₂O. DNA contamination was removed by digestion with 10 U of RQ1 DNAse (Promega) in a 100 μl reaction for 30 m at 37°C. RNA was then re-purified using RNABee as in the steps above. Final RNA samples were resuspended in 30 μl of nuclease-free H₂O. RNA integrity was verified by visualization on a SYBR-Safe (Invitrogen) stained agarose gel and RNA quantity was determined by spectrophotometric analysis.

RNAseq libraries were prepared using methods adapted from those described previously (Jorth et al., 2014). Ribosomal RNA was removed with the RiboZero Bacterial kit (Epicentre) per the manufacturer’s instructions. RNA was then fragmented with the NEBNext Magnesium RNA Fragmentation Module (NEB) and ethanol Na-acetate precipitated with the addition of 2 μl of linear acrylamide (NEB). Precipitated RNA was resuspended in 14 μl of nuclease-free H₂O. Sequencing libraries of fragmented RNA were assembled using the NEBNext sRNA library prep kit (NEB) per the manufacturer’s instructions. Assembled libraries were run on 6% polyacrylamide TBE gels and bands between 130 and 300 bp were gel purified and extracted as per the NEBNext sRNA library kit instructions. Libraries were assayed for concentration and integrity on a Bioanalyzer 2100 (Agilent). RNAseq libraries were submitted to Tufts University Core Facility Genomics for sequencing using single-end 50 bp reads on an Illumina HiSeq 2500 system.

Separate quantification of gene expression from S. aureus JE2 was determined by quantitative Reverse Transcriptase PCR (qRT-PCR) from the original non-rRNA-depleted RNA samples and additional experimental samples generated under identical conditions. RNA preparation with DNA removal was performed as above and 0.5 ng of total RNA was used for qRT-PCR with a qScript One-Step SYBR Green qRT-PCR kit (Quanta Biosciences) to determine overall transcript abundance in each experimental condition. We used primers oKL311 and oKL312 to measure spa transcript abundance and oKL339 and oKL340 to measure gyrB abundance as an internal control. Controls lacking reverse transcriptase were also performed to verify lack of contaminating DNA. Comparisons of relative abundance of spa expression normalized to gyrB expression per sample were used to calculate fold change of spa expression and compared to RNAseq data under the same conditions (Table 1).

The S. aureus USA300 FPR3757 genome sequence (Diep et al., 2006) (NCBI accession #NC_007793) was used for RNAseq read alignment. FASTQ files were aligned to the S. aureus JE2 (NC_007793) genome using Rockhopper (McClure et al., 2013) with default settings.

RNAseq sequencing data are available at NCBI in the sequence read archive under BioProject accession number PRJNA335565.

AIP-1 production was performed similarly to methods described previously (Thoendel and Horswill, 2009). Briefly, E. coli strain AH594 carrying an arabinose-inducible AIP-1 synthesizing gene locus (P_BAD-agrBD) was grown overnight at 37°C shaking at 200 RPM in 5 ml of LB with 100 μg/ml ampicillin. A 1:50 dilution of overnight culture was added to 50 ml LB without ampicillin in a 250 ml flask and was shaken at 200 RPM incubated at 37°C for 90 min prior to addition of arabinose to a final concentration of 0.02%. The induced culture was further incubated in the same conditions for 4 h. Next the cell culture was transferred to a sterile 50 ml conical centrifuge tube and cell supernatants were removed after centrifugation at 7,000 g for 10 m at 4°C. Cell supernatants were next passed through a 0.2 μm filter (Millipore, #SCGP00525) to remove remaining cell debris and then passed through a 3 kDa-cutoff size exclusion filter (Millipore #UFC900324). Filtrates were prepared for storage by addition of sterile glycerol and dithiothreitol to concentrations of 10% and 1 mM respectively. Aliquots of 1.5 ml each were stored at -80°C and exhibited full activity for >30 days.

Cell-free conditioned medium was generated from each indicated Corynebacterium spp. after inoculation into BHI cultures and growth at 37°C shaking at 200 RPM for 48 h. Conditioned medium was removed after centrifugation of 48-h cultures for 10 m at 7,000 g and then passed through a 0.2 μm filter (Millipore, #SCGP00525) yielding CFCM, which was either used directly for experiments, frozen at -80°C prior to use or further passed through a 3 kDa molecular weight size exclusion filter (Millipore #UFC900324) then frozen at -80°C. For some samples, CFCM was heat treated at 95°C for 15 m. For other samples, CFCM was Proteinase K (Promega #V3021) treated by addition of 100 μg in 1 ml of CFCM prior to incubation at 55°C for 1 h followed by a 95°C heat inactivation for 10 m prior to testing for the effect on AIP-1.

AIP-1-dependent luminescence assays were performed similarly to those described previously (Jensen et al., 2008; Thoendel and Horswill, 2009). Briefly, 100 μl of an overnight culture of the AIP-1-responsive agrP3-lux luminescence reporter strain ROJ143 grown in 3 ml BHI with 10 μg/ml chloramphenicol was inoculated into 5 ml BHI with 10 μg/ml chloramphenicol and incubated at 37°C, shaking at 200 RPM for 1.5 h. Optical density at 600 nm (OD₆₀₀) was measured and cells were then diluted in BHI to an OD₆₀₀ of 0.1 and 100 μl were added to each well of a 96-well plate to be combined with mixes of CFCM for testing. Meanwhile, 150 μL of E. coli AIP-1-containing CFCM was incubated in a 1.5 ml microcentrifuge tube for 2 h at 37°C with an equal volume of one of the following: BHI or Corynebacterium spp. CFCM. Next, 25 μl of each incubated solution was added to four separate wells containing 100 μl of the ROJ143 reporter strain and incubated for 3 h at 37°C prior to reading luminescence on a BioTek Synergy HT illuminometer with a 0.5 s integration time. Luminescence was quantified as the average of four replicate wells each from 3 independent experiments.

We adapted this assay from a previously published protocol (Weidenmaier et al., 2004) and A549 cell cultures were performed identically with the exception that F-12K medium and 10% Fetal Bovine Serum (Life Technologies) was used. A549 cells were grown to 90–95% confluency (5 × 10⁴/well) in 24-well plates and washed 3x in serum-free F-12K medium before addition of S. aureus. We added a 1:10 dilution of E. coli CFCM containing AIP-1 into 1.5 ml cultures containing 1:100 inoculums from overnight cultures of either wild-type or agrA-deficient S. aureus JE2. A separate set of tubes was prepared identically plus the addition of a 1:2 dilution of 3 kDa-filtered C. striatum CFCM. These cultures were incubated at 37°C shaking at 200 RPM for 4 h. Late exponential cells were washed 3x in BHI and then diluted in serum-free 37°C F-12K medium and added to each well in 500 μl volumes at an MOI of 25. S. aureus cells from each growth condition and A549 cells were coincubated for 1h at 37°C and 5% CO₂. Aliquots of planktonic cells were removed after incubation for CFU enumeration and A549 cell cultures were then washed 3x with sterile PBS. Five hundred microliter of PBS with 0.04% Triton-X100 was added to each well and incubated for 5 m at ambient temperature before epithelial cells were disrupted by vigorous micropipette aspiration. S. aureus-containing suspensions were transferred to microcentrifuge tubes and vortexed for 30 s each before being serially diluted for CFU enumeration. CFU values for attached cells were divided by their respective planktonic CFU values to determine % attachment for each condition. Percent attachment was divided by the % attachment for the WT S. aureus condition to determine fold change differences in attachment for each condition. Attachment was measured from technical triplicates for each condition across three biological replicates from separate A549 cell cultures.

Cultures were grown as described in ‘mono- and coculture assays.’ Polycarbonate membranes were transferred to 1.5 ml microcentrifuge tubes containing 1 ml of ice-cold 95% methanol and stored at -20°C. Cells were removed from membranes by gentle agitation and membranes were discarded. We added 5 μl of cell suspension to poly-lysine coated slides (Corning). These were air dried for ~5 m. To block, 100 μl of PBS with 1% BSA was added to each sample and incubated for 1 h at RT. Blocking solution was removed by pipette after 1 h and an equal volume was added then removed immediately after. Fifty microliter of rabbit α-goat FITC-conjugated IgG antibody (Life Technologies) diluted 1:500 in PBS with 1% BSA was added to the sample and incubated in the dark at RT for 1 h. Antibody solution was removed by pipette and the sample was rinsed 3x with 200 μl PBS with 0.1% Triton X100 which was gently added and removed by pipette. The sample was rinsed 1x with 100 μl PBS and excess liquid was removed by pipette without allowing the sample to air dry. Seven microliter of DAPI mounting medium (Southern Biotech) was quickly added and the sample was covered with a 25 × 25 mm #1 coverslip and incubated in the dark for >10 m at RT.

Each sample was imaged at 1000x magnification on a Zeiss Axio Observer using manufacturers filter settings for DAPI and FITC acquisition and identical exposure times for all samples based on autofluorescence of a non-FITC stained control sample. Three randomly selected fields of view were taken of each sample from three biological replicates for each growth condition. Images were analyzed using ImageJ software (NIH). Due to pronounced differences in DAPI-staining intensity between species, S. aureus was easily quantified from coculture samples by thresholding images in the DAPI channel. Cells positively stained for FITC were also identified by thresholding based on a non-FITC stained control. Using identical thresholding settings, total S. aureus (DAPI) versus SpA-expressing S. aureus (FITC) were counted in all fields of view using the ‘Analyze Particles’ function.

This assay was adapted from a previously published protocol (Pang et al., 2010). We added a 1:10 dilution of E. coli CFCM containing AIP-1 into 1.5 ml cultures containing 1:100 inoculums of either wild-type or agrA-deficient S. aureus JE2. A separate set of tubes was prepared identically plus the addition of a 1:2 dilution of 3 kDa-filtered C. striatum CFCM. These cultures were incubated at 37°C shaking at 200 RPM for 4 h. After incubation, CFCM was generated from each S. aureus culture by passage through a 0.2 μm filter (Millipore, #SCGP00525). Rabbit blood (Hemostat Laboratories, USA) was diluted v/v to 3% in sterile PBS then 70 μl was combined with 30 μl of each S. aureus CFCM in eight replicates for each condition in a 96-well plate. Blood was also combined at the same ratio with BHI as a negative control and BHI was used as a blank for OD₆₃₀ measurements. Plates were incubated at 37°C for 40 m then OD₆₃₀ was read to determine rabbit erythrocyte lysis.

Murine abscesses were generated essentially as described previously (Mastropaolo et al., 2005). Briefly, 6–8 week-old, female, Swiss Webster mice were anesthetized with an intraperitoneal injection of Nembutal (50 mg/kg). The hair on the left inner thigh of each mouse was shaved and the skin was disinfected with 70% alcohol. Mice were injected subcutaneously in the inner thigh with ~10⁶ CFU either S. aureus (wt or Δspa) or ~10⁶ C. striatum, or a combination of both. At 4 days post-infection, mice were euthanized and intact abscesses were harvested, weighed and placed into 2 ml of sterile PBS. Tissues were homogenized, serially diluted and plated on BHI agar with 25 μg/ml fosfomycin for C. striatum enumeration or MSA for S. aureus enumeration, to determine bacterial CFU/abscess. Experimental protocols involving mice were examined and approved by the Texas Tech University HSC Institutional Animal Care and Use Committee.

Assays to detect AIP induction were carried out similarly to those described previously, (Kavanaugh et al., 2007) using the agrP3-gfp AIP inducible reporter vector pDB59 (Yarwood et al., 2004). Briefly, S. aureus reporter strains AH1677, AH430, AH1747 and AH1871 (representing agr types I–IV respectively) were grown overnight in 5 ml BHI cultures at 37°C shaking at 200 RPM. From each culture, 3 ml were passed through a 0.2 μm membrane to generate S. aureus CFCM containing AIP produced by post-exponential phase cultures. We then generated 1.5 ml BHI cultures, which were separately inoculated with a 1:100 dilution of each overnight culture and contained a 1:10 dilution of the cognate AIP-containing overnight CFCM to induce agr-dependent expression of the GFP reporter as positive controls. A separate set of tubes was prepared similarly but with the addition of a 1:10 dilution of S. aureus CFCM containing an AIP type known to inhibit the agr type of the respective reporter strain; these served as negative controls for agr induction. E. coli AIP-1-containing CFCM was used to inhibit agr types II and III while S. aureus CFCM containing AIP-2 was used to inhibit agr types I and IV. Finally, another set of tubes was prepared identically to the positive controls except that these contained a 1:2 dilution of C. striatum CFCM passed through a 3 kDa-cutoff size exclusion filter. These cultures were incubated identically to the overnight cultures for either 4 or 6 h and examined for GFP fluorescence. To quantify GFP fluorescence, 25 μl of culture was combined with 75 μl of sterile BHI in a single well of a 96-well plate with four replicates per culture. GFP activity was determined by first measuring OD at 630 nm (OD₆₃₀) then measuring GFP emission through 485 ± 20 nm excitation and 528 ± 20 nm emission filters on a BioTek Synergy HT plate reader. Fluorescence measurements were divided by OD₆₃₀ to normalize for changes in culture density. Strains of agr-I, III, and IV showed maximal induction 6 h after supernatant additions whereas agr-II was induced at 4 h.

This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocol was approved by the Institutional Animal Care and Use Committee of Texas Tech University Health Sciences Center (Protocol Number: 09039).

We hypothesized that S. aureus gene expression would change in response to C. striatum. To test this hypothesis, we assessed gene expression using RNA sequencing (RNAseq) after growing S. aureus JE2 (Fey et al., 2013) (a USA300 LAC derivative) in either mono- or coculture with C. striatum ATCC 6940 on solid (1% agarose) CDM (Brown and Whiteley, 2007) at pH 6 without glucose at 37°C. This solid-phase, low pH medium partially approximates human skin-surface conditions. There were no significant changes in growth yield, as determined by CFU (colony forming unit) measurement, between mono- and coculture conditions for either species. Table 1 contains select S. aureus genes that were differentially expressed in coculture with C. striatum as compared to monoculture. Overall, 469 genes were differentially expressed ≥2-fold, with statistical significance (Supplementary Table S1).

A striking result from the S. aureus RNAseq data was that roughly half of the differentially expressed genes in cocultivation with C. striatum are also members of the S. aureus agr regulon (Cassat et al., 2006; Queck et al., 2008). The gene whose transcript was most decreased, psmβ1, encodes a phenol-soluble modulin toxin, which is positively regulated by the agr regulon, and psmβ1 downregulation in coculture with C. striatum was validated using a lacZ reporter construct based on its own promoter (Supplementary Figure S1). Further, expression of the entire S. aureus agr operon was decreased in coculture with C. striatum (Table 1). Similarly, the gene whose transcript was most increased in coculture (up 260-fold) is indirectly influenced by the S. aureus agr QS system (Schmidt et al., 2003; Novick and Geisinger, 2008). This is the staphylococcal protein A (spa) gene, which encodes surface protein A (SpA). We validated this increase in spa transcription by qRT-PCR (Table 1). SpA is characterized as an immunoprotective protein that inhibits opsonization and phagocytosis (Forsgren and Sjoquist, 1966; Peterson et al., 1977; Spika et al., 1981) and recent research indicates a role for SpA during nasal colonization. Transcription of spa is elevated in both humans and rodents during nasal colonization compared to in vitro culture (Burian et al., 2010a,b; Krismer et al., 2014). In addition, during experimental in vivo human nasal colonization, spa-deficient mutants are cleared more rapidly than wild-type S. aureus in hosts with a robust nasal immune response and SpA protein levels positively correlate with duration of colonization (Cole et al., 2016). Interestingly, in addition to spa, C. striatum also induced several other S. aureus genes whose expression is increased during in vivo nasal colonization of humans (and of cotton rats), including metI, sbnC, clfB, isdA, and oatA (Table 1) (Burian et al., 2010a,b; Krismer et al., 2014). Overall, we observed altered levels of S. aureus transcripts regulated by agr QS during cocultivation with C. striatum with increased expression of genes known to be upregulated during in vivo nasal colonization and decreased expression of genes known to be upregulated during invasive infection. Based on these data, we hypothesized that, in response to the commensal C. striatum, S. aureus shifts to a commensal state with a decrease in activities positively regulated by the agr QS system and an increase in activities negatively regulated by the agr QS, i.e., diminished expression of virulence factors needed for successful invasive infection and increased expression of surface-associated adhesion factors important for host colonization.

As the first step toward testing the hypotheses above, we asked whether the decrease in expression of genes positively regulated by agr QS in coculture with C. striatum requires cell-cell contact or whether exposure to C. striatum CFCM is sufficient. To address this question, we used a luminescent agrP3 promoter reporter assay (Jensen et al., 2008), which produces light only when agr QS is activated, and measured the effect of C. striatum CFCM on agr QS (Figure 1). This also allowed us to confirm that the S. aureus response is mediated via agr QS. In this assay, S. aureus autoinducing peptide type 1 (AIP-1) that has been heterologously produced in E. coli (Thoendel and Horswill, 2009) induces an AIP-1-responsive agrP3-luxCDABE transcriptional fusion in a S. aureus mutant unable to produce AIP-1 (Jensen et al., 2008). If C. striatum CFCM is sufficient to decrease agr QS then addition of CFCM will decrease luminescence. C. striatum CFCM samples were adjusted to neutral pH to avoid inhibition of agr activity by pH extremes (Regassa et al., 1992). CFCM from C. striatum stationary phase cultures (>36 h old) was sufficient to inhibit S. aureus agr QS (Figure 1, leftmost light gray bar), indicating that cell–cell contact is not required for the observed interaction between S. aureus and C. striatum. In contrast, neither CFCM from the parent E. coli strain used to produce AIP-1 nor a S. aureus mutant with a transposon insertion in agrB (AgrB^-), which is unable to produce AIP, resulted in inhibition of agrP3-lux activity to the extent of C. striatum CFCM (Figure 1, white bars). Further fractionation of C. striatum CFCM revealed that the inhibitory activity passed through a 3 kDa-molecular weight filter and also remained functional after heat treatment at 95°C for 15 min and was resistant to Proteinase K digestion (Figure 1, medium gray bars). Maximal agr QS inhibitory activity required preincubation of AIP-1 with C. striatum CFCM for at least 2 h prior to addition to the reporter strain (Supplementary Figure S2). This eliminates the possibility that inhibition is due to production of a competitive inhibitor of the AgrC sensor kinase by C. striatum, because an inhibitor that binds directly to AgrC would not require preincubation with AIP-1 for activity. These data established that S. aureus decreases activation of the agr QS system in response to C. striatum CFCM likely due to the presence of a small molecule that acts on AIP-1 itself. In subsequent experiments, we assayed for changes in S. aureus’ activity (behavior) after exposure to C. striatum by using CFCM, rather than cocultivation.

In addition to C. striatum, S. aureus encounters a number of other commensal species of Corynebacterium in the nasal passages, on skin and in DFIs, e.g., (Citron et al., 2007; Frank et al., 2010; Wos-Oxley et al., 2010; Oh et al., 2012; Kaspar et al., 2016). To determine whether S. aureus responds similarly to other commensal Corynebacterium spp., we selected C. amycolatum, C. accolens and C. pseudodiphtheriticum, all three of which are common members of nasal microbiota (Yan et al., 2013; Kaspar et al., 2016), as is S. aureus. Also, like C. striatum, C. amycolatum is associated with DFIs (Citron et al., 2007). In addition, we tested a strain of C. glutamicum, which was originally isolated from soil (Abe et al., 1967) and is not associated with human colonization. When exposed to CFCM from these Corynebacterium spp., S. aureus again displayed reduced agrP3-lux reporter activity (Figure 1, light gray bars). These results are consistent with a conserved genus-level interaction between Corynebacterium spp. and S. aureus that is predicted to result in a shift in S. aureus behavior toward colonization and away from virulence. We selected C. striatum for all subsequent experiments to characterize the S. aureus phenotypic response to commensal Corynebacterium spp.

Many of the cell surface proteins that are expressed when the S. aureus agr QS is not activated are involved in adhesion to host cells (Burian et al., 2010b; Foster et al., 2014), and are negatively regulated by activation of agr QS. Therefore, we hypothesized that when grown to high cell density in the presence of C. striatum CFCM, S. aureus would exhibit increased adhesion to host epithelial cells, a phenotype that can serve as a proxy for colonization or commensal behavior (Foster et al., 2014). To test this, we measured S. aureus attachment to A549 human airway epithelial cells, similarly to previously described (Weidenmaier et al., 2004). We cultured wild-type S. aureus in the presence of exogenous AIP-1 plus or minus C. striatum CFCM. As a positive control, we performed the identical experiment with an isogenic S. aureus agrA-deficient mutant in the presence of AIP-1. S. aureus cells in late exponential phase from each condition were normalized to identical optical densities and then added to epithelial cell monolayers at a multiplicity of infection (MOI) of 25. Next, we enumerated total planktonic and epithelial-cell-attached S. aureus by CFU measurement. We compared the proportion of planktonic versus epithelial cell-attached CFUs in each condition and expressed the results as fold change of attached cells from each condition compared to the S. aureus WT. After growth with C. striatum CFCM, the wild-type S. aureus (WT; dark gray bars in Figure 2) exhibited increased adhesion to the respiratory epithelial cells; this increase was comparable to that of an isogenic agrA-deficient mutant (AgrA^-, white bar in Figure 2). The increase in S. aureus adhesion to epithelial cells when exposed to C. striatum is consistent with our hypothesis that S. aureus shifts toward a commensal state in the presence of C. striatum.

The S. aureus spa gene transcript was the most highly increased in coculture with C. striatum compared to monoculture. Traditionally, SpA has been studied in the context of invasive infection; however, using qRT-PCR, Burian et al. (2010b) report that transcript levels of spa are increased in vivo during nasal colonization compared to in vitro culture indicating SpA might play a role in commensal interactions with the host. More recently, Cole et al. (2016) found that increased levels of SpA correlate with longer duration of S. aureus colonization during experimental human nasal inoculation and that spa-deficient mutants are less effective than the wild-type at colonizing some humans. Based on these data, which indicate a likely role for SpA in colonization, we proceeded to explore the effect of the observed increase in spa transcription in response C. striatum. Regulation of spa transcript levels is well studied and complex, depending on several elements (Schmidt et al., 2003; Gustafsson et al., 2009). Like many adhesion factors, SpA is a cell-surface protein that is negatively regulated through agr QS, albeit indirectly. Therefore, we first verified that the increase in spa transcription during cocultivation with C. striatum was indeed dependent on agrA (Supplementary Figure S3). We hypothesized that the increase in spa transcript level when grown with C. striatum would translate to an increase in S. aureus production of functional SpA. To test this hypothesis, we investigated SpA abundance and activity in S. aureus using immunoblot and immunofluorescence assays. First, we stained for SpA via immunoblot with an α-SpA antibody using cell lysates from cultures incubated in the presence of AIP-1 alone or with exposure to C. striatum CFCM. SpA abundance qualitatively increased upon treatment with C. striatum CFCM and SpA was absent in a Δspa mutant (Supplementary Figure S4). We then assayed for SpA activity with a standard IgG binding assay. Using solid-phase culture conditions identical to those in our RNAseq experiments, we quantified the proportion of S. aureus cells exhibiting SpA IgG binding activity in mono- versus coculture by measuring SpA capture of FITC-conjugated IgG antibodies. We found that 60% of wt S. aureus cells had detectable IgG capture in coculture compared to only 10% in monoculture (Figure 3A, dark bars). IgG capture was visible as a ring around cells in both mono- and coculture with C. striatum, consistent with cell-surface expression of SpA (Figures 3B,C). No FITC fluorescence was detected in a Δspa mutant in mono- or coculture (Figure 3A). Fluorescence signal was restored when spa was complemented in trans. The greater proportion of cells with detected surface IgG capture in the Δspa mutant containing a multicopy plasmid with spa under its native promoter suggests that the construct results in expression of spa above wt levels (Figure 3A, white bars). Together these data demonstrate that the C. striatum-dependent increase in S. aureus spa transcript levels (Table 1) results in a phenotypic change with an increased number of cells displaying active SpA on their surface. Active SpA can diminish opsonization and phagocytosis via binding of the Fc portion of IgG, inhibiting serum complement pathway activation (Forsgren and Sjoquist, 1966; Peterson et al., 1977; Spika et al., 1981). Therefore, we also confirmed that the increased level of SpA activity in coculture corresponded to decreased S. aureus phagocytosis (Supplementary Figure S5). Together with the evidence for increased epithelial cell adhesion after exposure to C. striatum, these results, along with reports that spa expression and SpA abundance increases during in vivo nasal colonization (Burian et al., 2010b; Cole et al., 2016), support our hypothesis that S. aureus shifts toward a colonization (commensal) state in the presence of Corynebacterium spp.

The other side of agr QS-regulated activities in S. aureus is an increase in the production of secreted virulence factors at high cell density when agr QS is activated (Novick and Geisinger, 2008; Thoendel et al., 2011). Hemolysis has traditionally served as an approximation of S. aureus virulence factor production and is typically diminished in agr mutants (Blevins et al., 2002). Therefore, we assayed for the rabbit erythrocyte hemolytic activity of CFCM from S. aureus grown with or without C. striatum CFCM. Based on our RNAseq data, we predicted that S. aureus exposed to C. striatum would exhibit decreased hemolysin activity due to repression of genes encoding β and δ-hemolysins (Table 1), which are both positively regulated by agr QS. Using a published method (Pang et al., 2010), we quantified hemolysis as the loss of optical density at 630 nm of rabbit erythrocytes when mixed with CFCM harvested from each S. aureus strain induced with AIP-1, in late-exponential phase, in the presence or absence of C. striatum CFCM (Cst). An agrA-deficient mutant served as a non-hemolytic control (Figure 4, light gray bars). In this assay, wild-type S. aureus production of hemolytic activity was strongly diminished by exposure to C. striatum CFCM (Figure 4, dark gray bars). These results further support our hypothesis that S. aureus shifts away from virulence in the presence of C. striatum.

We have shown that S. aureus agr QS-dependent gene expression is altered in response to commensal Corynebacterium spp. in a manner that shifts its phenotypic behavior toward colonization and away from virulence. Therefore, we hypothesized that in vivo coinfection with C. striatum would attenuate S. aureus’ success when compared to monoinfection. To test this, we compared colony-forming units (CFUs) from mono- and coinfection of S. aureus and C. striatum in a previously established murine subcutaneous abscess model (Mastropaolo et al., 2005; Ramsey et al., 2011). We measured abscess weight and bacterial CFUs 4 days after subcutaneous injection of single- versus dual-species infections. S. aureus CFUs decreased 6.2-fold in coinfection with C. striatum (Figure 5A), whereas C. striatum CFUs increased 20.7-fold in coinfection with S. aureus (Figure 5B). Thus, coinfection with S. aureus and C. striatum resulted in a disadvantage for S. aureus while providing an advantage to C. striatum compared to monoinfection, implying that S. aureus responds to the presence of C. striatum with changes in gene/protein expression that decrease its fitness for invasive infection while C. striatum capitalizes on the presence of S. aureus in this infection model.

Staphylococcus aureus strain JE-2, which was used for all of the above experiments, is an AIP type I strain. S. aureus has four characterized agr types, (agr I–IV) that each make a specific AIP molecule (AIP 1–4) along with its cognate AgrC (reviewed in Novick and Geisinger, 2008; Thoendel et al., 2011). Each AIP type acts as a potent quorum signal inhibitor for other agr type strains, with the exception of AIP-1 and AIP-4 whose structures differ by only a single peptide (D-Y substitution) and which do not inhibit each other (Novick and Geisinger, 2008; Thoendel et al., 2011). We hypothesized that multiple agr types would respond similarly to C. striatum and tested this using a published assay (Kavanaugh et al., 2007) with an AIP-inducible agrP3-gfp transcriptional fusion to monitor agr QS-dependent activation in strains representing all four agr types. In liquid culture, the agr QS system is activated in stationary phase when AIP levels are highest; therefore, as a positive control, we added stationary-phase CFCM from each cognate AIP-type strain as an inducer to mid-exponential-phase cells of the same strain and measured GFP fluorescence (Figure 6, dark gray bars). Strains incubated with a combination of the cognate S. aureus AIP CFCM and an inhibitory S. aureus AIP-type CFCM served as negative controls (Figure 6, white bars). Reporter cultures were incubated with a mix of cognate inducing S. aureus AIP CFCM and C. striatum CFCM to test for inhibition. In response to C. striatum CFCM, the agrP3 activity of agr types I, II, and III decreased, but type IV’s did not (Figure 6, light gray bars). These results indicate that in response to C. striatum, and likely Corynebacterium spp. in general, agr QS decreases in multiple S. aureus agr lineages.