New Insights into the Anti-pathogenic Potential of Lactococcus garvieae against Staphylococcus aureus Based on RNA Sequencing Profiling

The bio-preservation potential of Lactococcus garvieae lies in its capacity to inhibit the growth of staphylococci, especially Staphylococcus aureus, in dairy products and in vitro. In vitro, inhibition is modulated by the level of aeration, owing to hydrogen peroxide (H2O2) production by L. garvieae under aeration. The S. aureus response to this inhibition has already been studied. However, the molecular mechanisms of L. garvieae underlying the antagonism against S. aureus have never been explored. This study provides evidence of the presence of another extracellular inhibition effector in vitro. This effector was neither a protein, nor a lipid, nor a polysaccharide, nor related to an L-threonine deficiency. To better understand the H2O2-related inhibition mechanism at the transcriptome level and to identify other mechanisms potentially involved, we used RNA sequencing to determine the transcriptome response of L. garvieae to different aeration levels and to the presence or absence of S. aureus. The L. garvieae transcriptome differed radically between different aeration levels mainly in biological processes related to fundamental functions and nutritional adaptation. The transcriptomic response of L. garvieae to aeration level differed according to the presence or absence of S. aureus. The higher concentration of H2O2 with high aeration was not associated with a higher expression of L. garvieae H2O2-synthesis genes (pox, sodA, and spxA1) but rather with a repression of L. garvieae H2O2-degradation genes (trxB1, ahpC, ahpF, and gpx). We showed that L. garvieae displayed an original, previously undiscovered, H2O2 production regulation mechanism among bacteria. In addition to the key factor H2O2, the involvement of another extracellular effector in the antagonism against S. aureus was shown. Future studies should explore the relation between H2O2-metabolism, H2O2-producing LAB and the pathogen they inhibit. The nature of the other extracellular effector should also be determined.


INTRODUCTION
Interest in Lactic Acid Bacteria (LAB) as bio-preservation agents against foodborne pathogens has been growing for the last 15 years (Ito et al., 2003;Batdorj et al., 2007;Charlier et al., 2009;Adesokan et al., 2010). Considering their beneficial properties, the dairy industry could find it useful to employ Lactococcus garvieae strains as starter or adjunct cultures, provided the strains are safe (Fernández et al., 2010). Aquatic strains of L. garvieae are frequently referenced as fish pathogens (Eldar and Ghittino, 1999;Vendrell et al., 2006) but other strains are regarded as opportunistic pathogens for humans (Aguado-Urda et al., 2011;Ortiz et al., 2014). Despite its ubiquity in foods such as milk (Devriese et al., 1999;Villani et al., 2001;Callon et al., 2007) and dairy products of various origins (Flórez and Mayo, 2006;Fortina et al., 2007;Alomar et al., 2008b;El-Baradei et al., 2008;Jokovic et al., 2008;Alegría et al., 2009;Sip et al., 2009;Monfredini et al., 2012;Pangallo et al., 2014;Morandi et al., 2015), to our knowledge, L. garvieae has never been involved in a foodborne disease outbreak. L. garvieae is of interest for bio-preservation owing to its ability to inhibit the growth of staphylococci, particularly Staphylococcus aureus, as has been observed in milk, in cheese and in vitro (Alomar et al., 2008a,b;Delbes-Paus et al., 2010;Delpech et al., 2015). The transcriptome response of S. aureus to this inhibition has already been explored (Delpech et al., 2015). It is associated with a repression of the stress response (especially H 2 O 2 response) and of cell division genes and with modulation of the expression of virulence genes (particularly agrA, hld, and enterotoxin-encoding genes) genes. However, the molecular mechanisms of L. garvieae underlying the antagonism against S. aureus have never been explored.
Since L. garvieae produces low amounts of acetic and lactic acid in milk (Alomar et al., 2008a,b;Nouaille et al., 2009;Sip et al., 2009;Delbes-Paus et al., 2010;Delpech et al., 2015;Morandi et al., 2015), its antagonism against S. aureus is not associated with acidification. It is also unlikely to be associated with nutritional competition (Alomar et al., 2008a). With high aeration, S. aureus inhibition is mainly associated with hydrogen peroxide (H 2 O 2 ) production by L. garvieae, as already observed in vitro (Delbes-Paus et al., 2010;Delpech et al., 2015). With low aeration, the inhibition is weaker and H 2 O 2 is not detected. The addition of catalase (an H 2 O 2 -degrading enzyme) partly suppressed the inhibition (Delbes-Paus et al., 2010). Delbes-Paus et al. (2010) therefore suggested that at least one other molecule from L. garvieae must be involved in the residual inhibition of S. aureus in the absence of H 2 O 2 . The role and nature of this molecule have not yet been determined.
In Gram-positive bacteria, the aeration level and the presence of H 2 O 2 generally strongly affect the expression of H 2 O 2 metabolism genes. The expression of major H 2 O 2 degradation genes (ahpC, ahpF, gshR, gpo, trxA, and trxB) and their related proteins is strongly induced in L. lactis (Pedersen et al., 2008) under aeration and in Bacillus subtilis in the presence of H 2 O 2 (Mostertz et al., 2004). As regards the major H 2 O 2 synthesis genes, sodA is generally regulated in the same way as H 2 O 2degradation genes while the pyruvate oxidase (pox) gene is not affected by these parameters. Biological functions of L. lactis related to aerobiosis, i.e., O 2 response, menaquinone metabolism and stress response, are variably affected by the presence of S. aureus (Nouaille et al., 2009). However, little is known about the H 2 O 2 -metabolism of L. garvieae as an anti-pathogenic process.
In this study, we aimed to characterize the antagonism of L. garvieae against S. aureus in vitro in greater depth. Firstly, we investigated the presence of potential antagonist molecules in H 2 O 2 -free L. garvieae supernatants and their impact on S. aureus growth. Secondly, the transcriptome of L. garvieae with different aeration levels and in different biotic environments (absence or presence of S. aureus) was determined through RNA sequencing, an accurate and efficient method for revealing bacterial transcriptome profiles (Pinto et al., 2011). The resulting data led us to a better understanding of L. garvieae hydrogen peroxide metabolism and to some initial hypotheses on the nature of a new inhibition effector.

Strains and Culture Conditions
Lactococcus garvieae N201 and S. aureus SA15, isolated from raw milk, were obtained from the INRA UR545 collection (Alomar et al., 2008a). Both strains were aerobically grown in Brain-Heart Infusion broth ("BHI", Biokar Diagnostic, Pantin, France) for 20 h, at 30 • C for L. garvieae and at 37 • C for S. aureus. They were then inoculated separately or in co-culture at 10 6 cells.mL −1 for S. aureus and 10 7 cells.mL −1 for L. garvieae into BHI buffered at pH = 7 with phosphate buffer KH 2 PO 4 , 3H 2 O/K 2 HPO 4 at 0.1 mol.L −1 (KH 2 PO 4 , 3H 2 O, Riedel-de-Haen, Honeywell GmbH, Seelze, Germany; K 2 HPO 4 , Merck KGaA, Darmstadt, Germany) previously equilibrated at 30 • C. Pure cultures and co-cultures of both strains were performed with either a high or a low aeration depending on the experiment. Low aeration cultures were set in static, fully filled and sealed, 50-mL Nunc EZ Flip conical centrifuge tubes (Sigma-Aldrich, St. Louis, MO, USA). High aeration was obtained by a mechanical shaking at 150 rpm on 50-mL cultures in 250-mL Erlenmeyers. All cultures were incubated at 30 • C for 24 h in an Infors HT Minitron (Infors AG, Bottmingen, Switzerland). The cultivable cell counts were determined after plating for each sampling time as described by Delpech et al. (2015).

Impact of H 2 O 2 -Free Co-culture Supernatants on S. aureus Growth
The potential presence of other antistaphylococcal molecules in L. garvieae N201 and S. aureus SA15 co-culture supernatant and their impact on S. aureus planktonic growth were investigated. After adding catalase at 400 U.mL −1 , a pure culture of S. aureus SA15, a co-culture of S. aureus SA15 and L. garvieae N201 and plain BHI were incubated with low aeration as described above. After 6 h of incubation, 40 mL of each tube were centrifuged at 9,600 × g for 10 min at 4 • C and supernatants stored at 4 • C until utilization. Four milliliters of the remaining S. aureus culture were centrifuged at 7,500 × g for 10 min. The cell pellet was resuspended at a concentration of ∼10 6 cells.mL −1 in 20 mL of fresh 2X-concentrated BHI buffered at pH = 7. The supernatants were filtered through a cellulose acetate membrane (pore size 0.45 µm; GVS S.p.A., Zola Predosa, Italy). They were either treated with proteinase K (AMRESCO LLC, Solon, OH, USA, ref: 0706-100MG) and pronase E (Merck KGaA, ref: 537088), i.e., with two proteases, or treated with lipase (Sigma-Aldrich, ref: L3126-100G), or treated with α-amylase (Sigma-Aldrich, ref: A3176-500KU), or not treated at all. Each enzyme was added at 0.2 mg.mL −1 from stock solutions at 10 mg.mL −1 prepared in 200 mM of phosphate buffer at pH = 7. Treatment with proteases consisted of a first incubation of the supernatants with proteinase K at 50 • C for 2 h 15 and a second incubation with pronase E at 37 • C for 2 h 15. Treatments with lipase and α-amylase consisted of an incubation of the supernatants with the enzyme at 37 • C for 5 h. After each incubation step, the enzyme was inactivated by heating (95 • C during 10 min) and all supernatants were filtered again through a cellulose acetate membrane (pore size 0.45 µm; GVS S.p.A.). Prepared supernatants and S. aureus culture in 2X-BHI were then distributed at 1:1 volume ratio (total volume = 1.5 mL) in a CytoOne 24-well cell plate covered with a lid (ref: CC7672-7524; STARLAB, Hambourg, Germany). The cell plate was incubated for 25 h in a SAFAS Xenius XC spectrophotometer (SAFAS Monaco, Monaco) and thermostated at 30 • C with a Julabo CryoStat (JULABO Gmbh, Seelbach, Germany). S. aureus growth in each well was determined by measuring the OD 600 every 15 min for 1 h and then every 30 min for 24 h. Before each OD 600 measurement, the cell plate was shaken at 5 Hz for 20 s with an orbital diameter of 6 mm. The whole experimental design was repeated three times. To identify sample means that were significantly different from each other, statistical analyses were performed on values at 9, 12, 15, 18, 21, and 24 h, using Statistica software (StatSoft) with a single-factor analysis of variance (ANOVA) followed by a Newman-Keuls post hoc test.

Sample Preparation for RNA Analysis
Pure cultures and co-cultures of L. garvieae N201 and S. aureus SA15 were grown with high or low aeration, as specified above. After 3, 6, 9, and 24 h of incubation, 40 mL of each culture were centrifuged at 9,600 × g for 10 min at 4 • C. Hydrogen peroxide concentrations and pH values were determined on the supernatants by enzymatic reaction and spectrophotometry as described by Delbes-Paus et al. (2010). The cell pellets were immediately frozen in an ethanol bath and stored at −80 • C. Extraction of total RNA from the frozen cell pellets was performed as described by Delpech et al. (2015). For each sample of total RNA obtained with one culture condition, a first part of the aliquot was used for RNA sequencing after rRNA depletion and a second part of the aliquot was used for RT-qPCR analyses. The whole experimental design was repeated three times.

Determination of L. garvieae Transcriptome Changes by RNA Sequencing
Ribosomal RNAs were removed from the total RNA (2 × 5 µg of RNA by sample) using a RiboZero Magnetic Kit for Gram Positive Bacteria (Illumina Inc., San Diego, CA, USA) according to the manufacturer's instructions. The quality and concentration of RNA in each sample were assessed using a RNA 6000 pico kit (Agilent Technologies, Santa Clara, CA, USA). The following steps were performed by the MGX Platform (Montpellier GenomiX, CNRS, Montpellier, France) using Illumina kits and devices (Illumina Inc.): construction of the mRNA library using a TruSeq Stranded mRNA Sample Preparation Kit; cluster generation with the cBot system using a Cluster Generation Kit; hybridization of the sequencing primer on the flow-cell; 50-bp single-read sequencing using a HiSeq 2000 device with SBS technology; informatic pretreatments, i.e., image analysis with the HiSeq Control Software and Real-Time Analysis component, base-calling with the RTA software and demultiplexing with CASAVA (Illumina). The RNAseq data are available from NCBI GEO datasets under the accession number GSE74030.
The quality scores across all bases of all reads and the N (non-attributed bases) content across all bases were determined for each condition with the FastQC software from the Babraham Institute 1 . Both analyses showed good quality values (data not shown). The reads were next aligned simultaneously on reference genomes, i.e., L. garvieae N201 (unpublished, GOLD project ID = Gp0034836 and NCBI BioProject ID = 184287) and S. aureus MW2 (RefSeq number = NC_003923.1), using the BWA package (Li and Durbin, 2009) with a seed of 32 bases and a maximum of two mismatches tolerated on the seed. Using the Samtools suite  we excluded from further analyses those reads with low alignment quality scores (MAPQ index < 20). Reads which mapped on multiple sites (between 0.3 and 3.2% depending on the sample) and reads which did not map on any site considering the stringency applied (between 8.7 and 15.5% depending on the sample) were excluded from further analysis (see Supplementary Table 1). Reads overlapping genes were counted with the HTSeq Count software in Union mode (Anders et al., 2015). Differentially expressed genes were identified using the Bioconductor R 2 packages EdgeR, DESeq and DESeq2 (Anders and Huber, 2010;Robinson et al., 2010;Love et al., 2014). Genes with less than 15 reads (from the three biological replicates of two compared samples) were filtered and thus removed from the analysis. Data were normalized using the Relative Log Expression (RLE) normalization factor for EdgeR and the DESeq normalization factor for DESeq and DESeq2. Gene expression changes with adjusted p-value of less than 0.05 (by the FDR method from Benjamini-Hochberg) were declared differentially expressed. Differentially expressed genes highlighted by at least one of the three packages were considered for further investigation.
Genes were sorted into two lists: one of genes differentially expressed in both pure culture and co-culture and one of genes differentially expressed only in pure culture or only in co-culture. These lists were separately subjected to a Blast2Go analysis (Conesa et al., 2005). Blast2Go analysis consisted first of a BlastX 3 on the nr database with 20 hits and a maximum E-value of 1.0E −15 . The resulting data were enriched with an InterPro scan analysis (Zdobnov and Apweiler, 2001). Blast hits of each sequence were then mapped with Gene Ontology terms (Gene Ontology Consortium, 2008) annotated with a maximum E-value of 1.0E −6 , a cut-off of 55 and a GO weight of 5. The number of genes involved in each Biological Process GO category was calculated from combined graphs with no filter and a score alpha of 0.6. Since more than 450 different biological processes were identified for each category of genes, we excluded biological processes involving less than nine genes. When several biological processes involved the same genes and were associated with comparable functions, we considered only the one most relevant to our scientific hypotheses.
For a deeper analysis of S. aureus direct effect on L. garvieae gene expression regardless of aeration level, data mining was undertaken using the regression functions in Microsoft excel. Expression values were systematically plotted for each pair of samples (co-culture versus pure culture), a standard residual value was determined from the regression analysis for each gene across multiple pair wise comparisons and the mean value was determined. This mean standard residual value was used to rank the genes and identify the strongest gene expression differences. A standard residual cut-off of 5 was used as the threshold for significance to account for false discovery using a Bonferroni correction and approximating the standard residual to an equivalent p-value for the number of comparisons.

Determination of Gene Expression by RT-qPCR
Total RNA was retro-transcribed using a High Capacity cDNA Reverse Transcription kit (Invitrogen, Life Technologies, Carlsbad, CA, USA) following the supplier's instructions. Ct of genes of interest (see Supplementary Table 2) were determined by RT-qPCR as described by Delpech et al. (2015). All primers were designed using PrimerExpress R software (Applied Biosystems R , Life Technologies). By comparison with the Ct of the tufB reference gene stable in our conditions (data not shown), expression of a gene of interest ("goi") was calculated using the formula introduced by Pfaffl (2001).
The influence of two experimental factors was studied: the presence of S. aureus (with high or low aeration, gene expression in co-culture divided by gene expression in pure culture) and aeration level (in presence or absence of S. aureus, gene expression in shaken condition divided by gene expression in static condition). Statistical analyses were performed using Statistica software (StatSoft, Inc., StatSoft France, Maisons-Alfort, France) by single-factor ANOVA followed by a Newman-Keuls post hoc test.
Since the inhibition of S. aureus may not be related solely to hydrogen peroxide, we looked for another possible antistaphylococcal molecule produced by L. garvieae in the culture supernatants. We monitored S. aureus growth (OD 600 ) over 24 h in either non-inoculated BHI or in supernatants prepared from H 2 O 2 -free cultures (presence of catalase) with low aeration (either pure culture of S. aureus or co-culture of L. garvieae and S. aureus). In order to avoid any nutritional competition, BHI at a final concentration of 1X was added to each culture.
Growth of S. aureus in supernatant from S. aureus pure culture was comparable to that in plain BHI (data not shown). During the stationary phase, the OD 600 measured in S. aureus cultures in supernatant from co-culture was lower than that in supernatant from S. aureus pure culture (see Data Sheet 1). This inhibition was still observed when these supernatants were treated with proteases, lipase or α-amylase, the main macromolecule-degrading enzymes. pH values remained between 6.9 and 7.1 in all cultures (data not shown).

Determination of Aeration and S. aureus
Effects on L. garvieae Transcriptome by RNA Sequencing To identify the best conditions for studying the transcriptome of L. garvieae N201 with respect to its capacity to inhibit S. aureus SA15, we followed the growth of the two bacteria in pure cultures and co-cultures in BHI for 24 h under high or low aeration conditions ( Table 1).
The growth of L. garvieae was not affected by S. aureus. The growth of S. aureus was inhibited by L. garvieae at both aeration levels. With high aeration, maximal inhibition was observed from 9 to 24 h (difference with pure culture of 4 and 4.5 log CFU/ml, respectively, at 9 and 24 h). At 9 h, H 2 O 2 concentration reached a peak concomitant with the lowest S. aureus concentration. With low aeration, inhibition was weaker than with high aeration and was observed later, at 24 h. Concomitantly, H 2 O 2 was not detected in these cultures. pH values remained between 6.9 and 7.1 in all cultures (data not shown).
Considering these data, we analyzed the L. garvieae transcriptome in pure culture and in co-culture with S. aureus after 9 h of incubation with high and the low aerations. RNA sequencing generated a number of reads per sample, ranging from 12,365,133 to 15,670,131 depending on the sample after the initial quality filter (see Supplementary  Table 1). Between 83.1 and 91.0% of these reads mapped correctly onto the reference genomes. Considering the direct effect of S. aureus on L. garvieae gene expression, analysis of the mapped reads using the Bioconductor R packages failed to identify expression differences (data not shown). Additional regression analyses highlighted 39 genes differentially Frontiers in Microbiology | www.frontiersin.org

Effect of Aeration Level on L. garvieae Biological Processes
After RNA sequencing data treatments, 22 biological processes (named according to Gene Ontology termes) related to L. garvieae genes differentially expressed depending on aeration level in both pure culture and co-culture were identified using Blast2Go (Figure 1). Changes in gene expression are shown in Supplementary Table 4. Most of the biological processes affected were related to fundamental growth functions (Figures 1A,B) and nutrition ( Figure 1C). With high aeration, two genes (ilvA, LCGN_1922) related to "threonine metabolism" were repressed while two genes (LCGN_1919, LCGN_1920) were over-expressed.
Fourteen genes related to "transport" processes were differently expressed depending on aeration level both in pure culture and co-culture, including genes involved in the transport of metals (lead, cadmium, zinc, copper and/or mercury) and vitamins (riboflavin and folate).
Genes and GO biological processes related to O 2 and H 2 O 2 metabolism were also affected. "Oxidation-reduction process" and "response to stress" biological processes involved more repressed genes (17 and 8, respectively) than overexpressed genes (6 and 1, respectively). Three genes related to H 2 O 2 metabolism (ahpF, pox, and spxA1) and one gene related to O 2 consumption (lox) were repressed with the high aeration level. Several stress response genes (hrcA, groES, groEL, dnaK, dnaJ, grpE, clpB, and five genes belonging to the universal stress protein family) were repressed. Genes related to peroxide resistance (ohrA and ohrR) were strongly overexpressed under high aeration conditions. Electron Transport Chain (ETC) genes (cydB, menH, and ubiE) were overexpressed.

Response to Aeration
We identified 27 Gene Ontology biological processes related to genes differentially expressed depending on aeration level either in pure culture or in co-culture (Figure 2).
Changes in gene expressions are shown in Supplementary  Table 4. Most of the biological processes affected were related to fundamental growth functions (Figures 2A,B) or nutrition ( Figure 2C). Under high aeration conditions, four lysine metabolism genes (LCGN_0575, LCGN_0576, LCGN_0577, LCGN_0578) and two threonine metabolism genes (LCGN_0576 and LCGN_0577) were over-expressed exclusively in pure culture while four galactose metabolism genes were overexpressed exclusively in co-culture (LCGN_1809, LCGN_1810, LCGN_1811, and LCGN_1812). Eight genes involved in pyruvate and carbohydrate metabolism and the citrate cycle (glmS, glmM, galE, PTS-Man-EIIC and EIID pdhA, pdhB, DLAT) were repressed by S. aureus. With low aeration, the enolase encoding gene was repressed by S. aureus.
Four biological processes related to transport functions were affected ( Figure 2D): "transport, " "transmembrane transport, " "organic substance transport" and "ion transport." In pure culture, genes involved in the transport of cobalt/zinc/cadmium (LCGN_1867) and an unspecified monosaccharide (LCGN_0332) were over-expressed with high aeration while a gene involved in copper transport (LCGN_1427) was repressed. Two genes of the fur regulon (ferrous iron transport) were over-expressed with high aeration, one of them in pure culture (feoA) and one in co-culture (feoB).
With low aeration, the H 2 O 2 synthesis gene sodA was up-regulated by S. aureus. The H 2 O 2 degradation gene ahpC was repressed by high aeration in co-culture but not in pure culture. In pure culture, two genes related to O 2 consumption, noxE and LCGN_0208, were repressed and six genes related to ETC, cydA, cydC, menB, menC, menD, and LCGN_0364, were over-expressed. FIGURE 1 | Lactococcus garvieae biological processes involving genes differentially expressed depending on aeration level in both pure culture and co-culture. Bars represent the number of genes significantly overexpressed (left) or repressed (right) with high aeration compared to low aeration, according to at least one of the three R packages used (EdgeR, DESeq, and DESeq2). After the Blast2Go analysis and data filtering, biological processes were manually sorted into several categories: (A) processes related to fundamental growth function, i.e., global cellular, metabolic and biosynthetic processes, (post-)transcriptional and (post-)translational functions, (B) processes potentially associated with fundamental growth functions but also potentially associated with other metabolisms, (C) nutrition-related processes and (D) transport processes.

Impact of Aeration on the Expression of L. garvieae Genes Potentially Involved in the Antagonism against S. aureus
To complete the RNA-seq data obtained only at 9 h, we determined the expression of ten genes from the cultures previously used for RNA-seq analyses, at 6, 9, and 24 h by RT-qPCR. These genes, potentially involved in the antagonism mechanisms, were involved in O 2 consumption (noxE and lox), H 2 O 2 synthesis (pox and sodA), H 2 O 2 degradation (ahpC, ahpF, gpx, and trxB1) and resistance to other peroxides (ohrA and ohrR).
The presence of S. aureus induced only slight changes in their expression (data not shown). With high aeration, ahpC and pox were over-expressed 2.4-fold and 2.0-fold in the presence of S. aureus at 6 h. With low aeration, S. aureus induced a 3.1-fold over-expression of trxB1 at 9 h and a 3.3-fold repression of gpx at 24 h.
Conversely, RT-qPCR results showed major differences in H 2 O 2 -metabolism gene expression between the two aeration levels ( Table 2). The expression of ohrA was strongly induced under high aeration conditions in both pure culture and co-culture at 9 h. While RNA sequencing identified trxA2 as up-regulated with high aeration in both pure culture and co-culture at 9 h (Supplementary Table 4), RT-qPCR identified other H 2 O 2 -degradation genes (ahpC, ahpF, gpx, and trxB1) as repressed at 6, 9, or 24 h. The lactate mono-oxygenase gene lox seemed slightly repressed with high aeration but this modification was significant only at 9 h in co-culture. In pure culture with high aeration, the expression of noxE reached a peak at 6 h when it was 11.0 times higher than with low aeration. With low aeration, noxE expression gradually increased over time. No significant difference in the expression of H 2 O 2 -synthesis genes (pox and sodA) according to aeration level was observed.

DISCUSSION
This study aimed to improve understanding of the mechanisms underlying the antagonism of L. garvieae against S. aureus, especially as regards H 2 O 2 -related pathways.
The high aeration level and the resulting high H 2 O 2 concentration were associated with stronger inhibition of S. aureus, confirming previous observations under the same conditions (Alomar et al., 2008a;Delbes-Paus et al., 2010;Delpech et al., 2015). It was already known that this inhibition is associated with a modulation of the expression of S. aureus virulence genes and a repression of the H 2 O 2 response, stress response and cell division genes of S. aureus by L. garvieae and with high aeration (Delpech et al., 2015). However, the transcriptome adaptation of L. garvieae to prevailing aeration conditions during this interaction had not been explored until FIGURE 2 | Lactococcus garvieae biological processes involving genes differentially expressed depending on aeration level exclusively in pure culture or exclusively in co-culture. Stacked bars represent the number of genes significantly overexpressed (left) or repressed (right) with high aeration compared to low aeration in pure culture only (black bars) or in co-culture only (white bars), according to at least one of the three R packages used (EdgeR, DESeq, and DESeq2). After the Blast2Go analysis and data filtering, biological processes were manually sorted into several categories: (A) processes related to fundamental growth function, i.e., global cellular, metabolic and biosynthetic processes, (post-)transcriptional and (post-)translational functions, (B) processes potentially associated with fundamental growth functions but also potentially associated with other metabolisms, (C) nutrition-related processes and (D) transport processes. this study. It is known that difference in aeration is associated with drastic modifications of the L. lactis transcriptome (Pedersen et al., 2008;Dijkstra et al., 2014) and consequent metabolic adaptations (Jensen et al., 2001;Nordkvist et al., 2003;Pedersen et al., 2008;Dijkstra et al., 2014;Larsen et al., 2016a,b). In accordance with these findings, we found that different aeration levels were associated with significant differences in the L. garvieae transcriptome in biological processes related to fundamental growth functions, nutritional and metabolic adaptations and transport functions. Moreover, L. garvieae genes involved in fundamental growth functions were slightly repressed by S. aureus under both aeration levels suggesting an impact of S. aureus on L. garvieae metabolism. While L. garvieae is a catalase-negative bacterium, the main S. aureus enzyme involved in H 2 O 2 dismutation is the O 2 -forming catalase KatA (Cosgrove et al., 2007). The S. aureus catalase may affect the transcriptomic response of L. garvieae to different aeration conditions by modulating O 2 and H 2 O 2 concentrations. It is known that the katA gene of S. aureus SA15 is repressed by L. garvieae in high aeration conditions but not in low aeration conditions (Delpech  et al., 2015). This may explain why the presence of S. aureus (via its catalase production) modulated the L. garvieae transcriptome response under different aeration conditions. Indeed, with low aeration, the L. garvieae H 2 O 2 synthesis gene sodA expression was slightly higher in presence of S. aureus. With high aeration, the main enzymes involved in O 2 consumption of L. lactis (NADH-oxidase NoxE and ETC enzymes (Tachon et al., 2010), were repressed only in L. garvieae pure culture. The expression of the L. garvieae noxE reached a peak at 6 h (beginning of exponential growth) and then decreased until 24 h. This suggests that L. garvieae NoxE is the main enzyme responsible for O 2 consumption at the beginning of exponential growth with high aeration, as observed for L. lactis NoxE (Lopez de Felipe and Hugenholtz, 2001;Tachon et al., 2010). The induction of most of the L. garvieae ETC genes (cydA, cydB, cydC, menB, menC, menD, menH, and LCGN_0364) under high aeration conditions at 9 h (beginning of stationary phase) suggested that ETC may be involved in O 2 consumption by L. garvieae during the stationary phase, as already shown with L. lactis (Tachon et al., 2009(Tachon et al., , 2010. Previous studies have showed that hydrogen peroxide production by L. garvieae depends on aeration level and plays a key role in S. aureus inhibition. As regards S. aureus, Delpech et al. (2015) suggested that the stronger inhibition with a high aeration level may be caused by the higher concentration of H 2 O 2 associated with the L. garvieae-induced repression of S. aureus genes involved in the H 2 O 2 response (katA and sodA at 6 and 9 h) and cell division (mraZ, mraW and potentially the dcw cluster). As regards L. garvieae, our RNAsequencing and RT-qPCR analyses showed an overexpression of the H 2 O 2 -degradation genes ahpC, ahpF, gpx, and trxB1 under low aeration conditions compared to high aeration, while the expression of H 2 O 2 synthesis genes pox and sodA remained stable. The expression of the main H 2 O 2 degradation genes of Gram-positive bacteria is generally induced more under aeration and in the presence of H 2 O 2 (Mostertz et al., 2004;Pedersen et al., 2008). Also, H 2 O 2 metabolism may be very different in L. garvieae compared to other Gram-positive bacteria. Although the difference in H 2 O 2 concentration between the two aeration levels was probably primarily conditioned by the availability of O 2 , our transcriptome results suggest that it was also associated with a control of H 2 O 2 degradation by L. garvieae rather than with a control of H 2 O 2 synthesis. Since the AhpCF peroxyredoxin system was repressed, the OhrAR system was probably essential for the resistance of L. garvieae to ROS (other than H 2 O 2 ) under the high aeration conditions. The widespread organic hydroperoxide detoxifying system ohrAR, known to be over-expressed under high O 2 conditions (Mongkolsuk et al., 1998;Fuangthong et al., 2001;Chuchue et al., 2006;Oh et al., 2007;Atichartpongkul et al., 2010;da Silva Neto et al., 2012;Clair et al., 2013), was consistently induced in L. garvieae at the high aeration level.
The fact that S. aureus population levels were lower in the stationary phase in H 2 O 2 -free supernatant from a coculture of S. aureus and L. garvieae revealed the presence of a new molecule involved in this inhibition. This effector is extracellular, is produced by L. garvieae during its exponential growth phase and can reduce the population level of S. aureus during its stationary growth phase. In view of the results of our enzymatic treatments on supernatants, the inhibitory gap during the stationary phase was probably caused neither by hydrogen peroxide, nor by a protein, nor by a lipid, nor by a polysaccharide. The only putative bacteriocin identified in the L. garvieae N201 genome was homologous to garvieaecin Q (GarQ, data not shown), a class IId bacteriocin (Tosukhowong et al., 2012). Class IId bacteriocins are generally sensitive to protease treatments as stringent as the one we used in this study (Kuo et al., 2013;Song et al., 2014), suggesting that garvieacin Q is unlikely to be the effector we are seeking. This effector may instead be related to several genes identified by RNA seq as being regulated by aeration level, such as genes involved in metal homeostasis (e.g., siderophores (Hannauer et al., 2015), chemical and ionic equilibrium (Doyle et al., 1975;Chudobova et al., 2015), transport of vitamin-related compounds (Schlievert et al., 2013) and export of unknown proteins. For example, the differential expression of ferrous ion transport encoding genes has already been observed in L. lactis under oxidative stress in milk (Larsen et al., 2016b). It may also be related to signaling molecules (stress, quorum sensing). It is known that L. garvieae can modify the expression of several S. aureus genes involved in environment-sensing systems like the agr system, CodY or two-component systems SaeRS and SrrAB (Delpech et al., 2015).
RNA-seq revealed variations in the expression of the codY gene (involved in nutritional adaptation (Guédon et al., 2001;Ercan et al., 2015), and of several nutritional-related metabolisms (lysine, threonine, mannitol, aspartate, ribose, fructose, and galactose). This suggests that L. garvieae adapts its nutritional behavior to the prevailing aeration level and the presence or absence of S. aureus. In a rich medium like BHI, there should be little nutritional competition. It is known that L. garvieae can consume all the L-threonine in micro-filtered milk in less than 3 h (Alomar et al., 2008b) and that S. aureus growth could be inhibited by L-threonine depletion (Pohl et al., 2009). However, we showed that the antagonism of L. garvieae against S. aureus was not associated with nutritional competition for L-threonine in micro-filtered milk (see Supplementary Table 5).

CONCLUSION
RNA sequencing analyses revealed a L. garvieae transcriptome adaptation to aeration level. This adaptation differed depending on the presence or absence of S. aureus. Our findings show that the control of autogenic H 2 O 2 levels by L. garvieae was probably carried out by H 2 O 2 degradation genes rather than H 2 O 2 synthesis genes. Our study also leads us to suggest that an unidentified effector was involved in the inhibition of S. aureus in the stationary phase. The potential inhibitory role of metals, siderophores and signal molecules (e.g., stress signal, quorum sensing) generated by L. garvieae should be investigated. In order to promote the use of H 2 O 2 -producing bacteria as bio-preservation agents, future studies should explore the relation between H 2 O 2 -metabolism, H 2 O 2 -producing LAB and the pathogen they inhibit.

AUTHOR CONTRIBUTIONS
PD carried out all the experiments, excluding preparation of cDNA libraries and RNA sequencing, and drafted the manuscript helped by CD and SB. ER, ED, and GB analyzed RNA sequencing data. SN performed the RNA sequencing (preparation of libraries and the sequencing itself). GG, M-CM, CD and SB conceived the study. All authors read and approved the final manuscript.