In Vitro and In Vivo Models of Staphylococcus aureus Endophthalmitis Implicate Specific Nutrients in Ocular Infection

Purpose To define global transcriptional responses of Staphylococcus aureus and its codY mutant (CodY is a transcription regulator of virulence and metabolic genes in response to branched-chain amino acids) when growing in bovine aqueous (AH) and vitreous humor (VH) in vitro, and to investigate the impact of codY deletion on S. aureus virulence in a novel murine anterior chamber (AC) infection model. Methods For the in vitro model, differential transcriptomic gene expression of S. aureus and its codY mutant grown in chemically defined medium (CDM), AH, and VH was analyzed. Furthermore, the strains were inoculated into the AC of mice. Changes in bacterial growth, electroretinography and inflammation scores were monitored. Results Bovine AH and VH provide sufficient nutrition for S. aureus growth in vitro. Transcriptome analysis identified 72 unique open reading frames differentially regulated ≥10-fold between CDM, AH, and VH. In the AC model, we found comparable growth of the codY mutant and wild type strains in vivo. Average inflammation scores and retinal function were significantly worse for codY mutant-infected eyes at 24 h post-infection. Conclusion Our in vitro bovine AH and VH models identified likely nutrient sources for S. aureus in the ocular milieu. The in vivo model suggests that control of branched-chain amino acid availability has therapeutic potential in limiting S. aureus endophthalmitis severity.


Introduction
Staphylococcus aureus is a commensal bacterium on the skin and mucosa, but is also a leading cause of infections in humans. When opportunistic pathogens infect sterile sites, they adapt, proliferate in the host, and exhibit virulence. The host becomes the sole source for nutrients. For many gram-positive bacteria, CodY provides an important regulatory link between nutrient availability and virulence factor production [1]. CodY controls expression of virulence and metabolic genes in response to the availability of branched-chain amino acids (BCAA) and GTP through Agr, a global regulator of the staphylococcal virulon [2]. In the presence of GTP and/or BCAA, CodY shows a higher affinity for its DNA targets, while in the absence of nutrients, there is a decrease in the GTP and BCAA levels causing decreased affinity of CodY to the DNA and thus induction of its regulon. In S. aureus, CodY regulates its regulon either indirectly via the quorum sensing regulator Agr or independent of the Agr system and through its direct binding to its DNA targets [2]. In general, CodY-regulated genes trigger adaptation to starvation [2][3][4][5][6][7][8][9] as well as play a role in virulence of pathogenic bacteria [3,4,10].
The human eye as well as eyes of animals like rabbits and mice possess sterile anterior and posterior compartments, which contain aqueous and vitreous humor, respectively. Infectious endophthalmitis [inflammation within the eye) is a complication of penetrating trauma to the eye and intraocular procedures such as cataract surgery that can lead to blindness [11][12][13][14]. The visual prognosis following infection depends greatly on the virulence of the causative organism, visual acuity at presentation, and the efficacy of antimicrobial treatment [15]. S. aureus is the second most common cause of acute postoperative infection following intraocular surgeries and is also associated with significant visual loss [15]. Given the presence of BCAA in human aqueous and vitreous humors [16,17] and the findings that CodY controls directly and indirectly S. aureus virulence genes such as hla and agr [1,2] which contribute to virulence in animal models of endophthalmitis [18,19], CodY may play a role in regulating S. aureus virulence during endophthalmitis.
S. aureus growth and virulence in animal models of endophthalmitis has been assessed, most frequently by intravitreal (posterior chamber) injection [18,19]. S. aureus strains grow in vivo to different extents, depending on the strain used and the inoculation site (anterior versus posterior compartments) [20,21]. For example, following injection into the anterior chamber, Balzli, et al. found that among 9 S. aureus isolates injected into the anterior chamber of rabbit eyes, only one strain, UMCR1, grew [20]. Wu, et al. [21] and Kowalski, et al. [22] found that S. aureus grew in the anterior chamber of rabbits, and used that model to test antibiotic efficacies. Several other studies have found that S. aureus can grow to high densities in the vitreous, and studied the contribution of toxins, the global virulence regulators Agr and Sar [18,23], and other cell wall components to pathogenesis. S. aureus clearly survives in the human eye, given that S. aureus can be recovered from the aqueous and vitreous humors of patients who develop endophthalmitis [24]. It is unknown what nutrient sources S. aureus utilizes during infection of the human eye. In this study, we use aqueous and vitreous fluids extracted from commerciallyobtained bovine eyes as ex vivo endophthalmitis models for S. aureus, and define global transcriptional responses of S. aureus to growth in these media. Our goal was to identify genes that are consistently and highly differentially regulated by S. aureus during growth in pooled bovine AH and VH samples. We additionally interrogate the impact of codY deletion on S. aureus gene expression during growth in these media, as well as its impact on S. aureus virulence in a novel murine anterior chamber (AC) infection model.

Materials and Methods
Strains and growth media S. aureus strains used are listed in Table 1. S. aureus was routinely cultured in brain heart infusion (BHI) or on BHI agar. All cultures were incubated at 37uC. For microarray experiments, S. aureus were grown in chemically defined Socransky's medium [25] supplemented with 20 mM glucose (referred to here as CDM), or bovine aqueous or vitreous humor (AH and VH, respectively). CDM contains 76 mM leucine, 85 mM valine, and 76 mM isoleucine. Bacterial growth was assessed by monitoring optical density at 600 nm (OD 600 ) using a Biotech Synergy 2 microplate reader or by serial dilution and plating on BHI agar to obtain colony forming units per milliliter (CFU/mL).

Bovine AH and VH collection
AH and VH were extracted from commercially available bovine eyes (Sierra for Medical Science, Whittier, CA) and pooled as described previously [26]. Typical volumes of AH and VH recovered per bovine eye were 0.5-1 mL and 3-4 mL, respectively. AH was filter-sterilized with a 0.45 mm HT Tuffryn membrane sterile acrodisc syringe filter (Pall Life Sciences, Batavia, IL). VH was filter-sterilized with a 0.45 mm PES sterile filter (Whatman, Clifton, NJ). Sterile AH and VH were stored at 280uC until use. For microarray experiments, aspirates were pooled to a total of 40 mL achieve adequate volume.

Microarray analysis
S. aureus strains were struck from freezer stock onto BHI agar and incubated overnight. Colonies were used to inoculate CDM, AH or VH broth cultures, which were incubated overnight and then used to inoculate new CDM, AH and VH broth cultures to an initial OD 600 of 0.02-0.03. Bacteria were harvested for microarray analysis and semi-quantitative RT-PCR analyses during exponential growth, at an OD 600 of 0.4-0.5 for CDM and 0.15-0.2 for AH and VH. Cells were stabilized with RNAProtect (Qiagen) and RNA was extracted using the RNA Bee reagent (TelTest, Inc.) per the manufacturer's instructions. Absence of DNA contamination was verified by PCR using primers targeting the 16S rRNA gene (For, 59-AAC TCT GTT ATT AGG GAA GAA C-39; Rev, 59-CCA CCT TCC TCC GGT TTG TCA CC-39). cDNA synthesis, fragmentation, biotin labeling and hybridization to Affymetrix S. aureus GeneChips were performed as previously described [27]. Hybridization and scanning of GeneChips were performed at the University of Iowa DNA Core. Microarray experiments were performed in duplicate. Affymetrix GeneChip data was analyzed with GeneChip Operating Software (GCOS version 1.4). Probe sets with statistically significant change calls (increased or decreased; p#0.05) between control and test conditions were considered for further analysis, and fold change cut-offs were applied as described in the text. Microarray data have been deposited in ArrayExpress under accession numbers E-MTAB-2928.
For the Affymetrix S. aureus GeneChip, probe set IDs (for example, sa_c10261s8939_a_at) are used instead of gene names or ORF designations. To convert probe set IDs to genomic loci, we downloaded target DNA sequences corresponding to differentially expressed probe sets from the NetAffx Analysis Center (www. affymetrix.com/analysis/index.affx). Target sequences were compared to available S. aureus sequences in GenBank using NCBI BLAST (http://blast.ncbi.nlm.nih.gov). Transcription unit and metabolic pathway predictions were obtained from the BioCyc Staphylococcus aureus COL database (http://biocyc.org/ organism-summary?object=SAUR93062). Where appropriate, predicted protein products of differentially expressed genes were analyzed for putative functions using NCBI protein BLAST and Pfam 26.0 (http://pfam.janelia.org). Subcellular localization of proteins was predicted using PSORTb version 3.0 (http://www. psort.org/).

Semiquantitative RT-PCR
Semiquantitative reverse transcription (RT)-PCR was performed using Superscript II reverse transcriptase as outlined by the manufacturer (Invitrogen). S. aureus RNA was used to make cDNA with priming by random hexamers. cDNA was purified

Murine AC infection model
Female C57BL/6J mice were obtained from the Charles River Laboratory (Boston, MA). All animals were treated according to the guidelines of the Association for Research in Vision and Ophthalmology Resolution on the Use of Animals in Research. The protocol was approved by Massachusetts Eye and Ear Infirmary's Institutional Animal Care and Use Committee (IACUC). Mice were anesthetized by intraperitoneal injection of ketamine (62.5 mg/kg) and xylazine (12.5 mg/kg). Animals were euthanized at the appropriate time points by CO 2 asphyxiation.
S. aureus colonies obtained after growth on BHI agar were cultured overnight in BHI broth and subcultured 1:500 with fresh BHI broth and grown to an OD 600 nm of 0.4-0.8, pelleted by centrifugation at 10,000 rpm, and washed twice with PBS. The ACs of the right eyes of 6-8-week-old female mice were inoculated with 1 mL of S. aureus culture using 35 gauge needle on a nanofil syringe (World Precision Instruments, Inc.), just anterior to the limbus without touching the iris. For S. aureus MS7, 2 mM isopropyl b-D-1-thiogalactopyranoside (IPTG) was added to subcultures and included in PBS washes. The left eyes were left untreated and served as internal controls for electroretinography (ERG). Experiments were performed at least in duplicate with a minimum of 3 animals per experimental group. Animals were given water with 12 mM IPTG for one week prior to infection with S. aureus MS7. Quantification of in vivo bacterial growth, slitlamp examination and ERG were performed as described previously [28]. Briefly, intraocular inflammation was graded using the following criteria: 0, normal; 1, small amount of fibrin on the pupil; 2, iris partially covered with fibrin and/or hypopyon; 3, iris covered with fibrin and/or hypopyon; 4, pupil not visible. The retinal function in the infected eye was measured using ERG and was defined as the ratio of the b-wave (measured from the trough of the a-wave to the peak of the b-wave) amplitude of the experimentally treated eye to that of the contralateral untreated eye, times 100.

Histological analysis
Enucleated eyes were fixed in buffered formalin solution and histological analysis was performed by Excalibur Pathology Inc. (Oklahoma City, OK). Pathology slides were examined for signs and extent of inflammation in the different compartments of the eye, and disruption in retinal architecture.

Statistical analysis
Normality tests were performed on all data sets. The data were analyzed with an unpaired t-test if the distribution was Gaussian, or with the nonparametric Mann-Whitney test if the data were not normally distributed. P#0.05 was set as the basis for rejection of

S. aureus SA564 in vitro growth in bovine AH and VH
S. aureus SA564 is a clinical isolate that was previously used to evaluate a role for codY in S. aureus virulence regulation [29]. As a first step in understanding S. aureus physiology and metabolism during endophthalmitis, we evaluated S. aureus SA564 growth in pooled AH and VH harvested from commercially obtained bovine eyes. S. aureus SA564 was also grown in Sokransky's medium, a buffered, defined medium supplying amino acids, vitamins, nucleobases, metals, and other growth factors [25], supplemented with 20 mM glucose as a carbon source. For the purposes of this manuscript, Sokransky's medium with 20 mM glucose is referred to as CDM. Representative growth curves for S. aureus SA564 as assessed by OD 600 nm are shown in Fig. 1. These data demonstrate that CDM, pooled bovine AH, and pooled bovine VH support in vitro growth of S. aureus SA564. We reproducibly observed a clumping phenotype early in S. aureus in vitro AH growth, resulting in reduced cell density measurements obtained by OD 600 nm , as shown in Fig. 1. Also of note, one pooled AH sample did not support robust S. aureus in vitro growth (data not shown), possibly as a result of antimicrobials such as antibiotics or inflammatory factors present in one or more individual AH samples. We were unable to obtain additional information from the vendor about the health and history of cows used in this study.

Transcriptome analysis of AH-and VH-grown S. aureus SA564
We used Affymetrix GeneChips to examine gene expression of S. aureus SA564 during growth in bovine AH and VH, using CDM-grown SA564 as a control. Cells were harvested for microarray analysis during exponential growth; representative time points are indicated by arrows in Fig. 1. Microarrays were Vertical lines indicate genes computationally predicted to be in the same transcriptional unit. Gene names in brackets were assigned in this study using data shown in Table S1. b Fold change in expression of genes during S. aureus SA564 growth in AH as compared to growth in glucose-supplemented CDM; a positive number indicates an upregulation of the gene during growth in AH. Standard deviation is shown in parentheses. Fold changes $3 and ,10 are shown and italicized. c Fold change in expression of genes during S. aureus SA564 growth in VH as compared to growth in glucose-supplemented CDM; a positive number indicates an upregulation of the gene during growth in VH. Standard deviation is shown in parentheses. Fold changes $3 and ,10 are shown and italicized. d Fold change in expression of genes during S. aureus SA564 growth in VH as compared to growth in AH; a positive number indicates an up-regulation of the gene during growth in VH. Standard deviation is shown in parentheses. Fold changes $3 and ,10 are shown and italicized. e At least two differentially expressed probe sets were assigned to these ORFs. Data for all probe sets are shown in Table S1. performed in duplicate for each growth condition. The S. aureus GeneChip was designed using genome sequence from the S. aureus strains N315, Mu50, NCTC 8325, and COL, and queries over 3300 ORFs and intergenic regions [30]. Genome sequence data are not available for S. aureus SA564. For transcriptome experiments with wild-type SA564 and its isogenic codY mutant CDM7 (discussed further below), we obtained statistically significant hybridization over the background for an average of 64.8% of probe sets (range, 57-71.8% over 12 chips), corresponding to ,3763 probe sets queried. We performed three comparative analyses of wild-type S. aureus SA564 transcriptomes: AH-grown cells compared to CDM-grown cells (to model growth of S. aureus in the anterior chamber), VH-grown cells compared to CDM-grown cells (to model growth of S. aureus in the posterior chamber), and VHgrown cells compared to AH-grown cells (to model transcriptional changes potentially occurring after translocation of S. aureus from the anterior to posterior chambers). A fold change cut-off of 10 was used to consider the most highly differentially regulated genes in each condition. For the S. aureus GeneChip, probe set IDs (for example, sa_c10261s8939_a_at) are used instead of gene names or ORF assignments. To convert differentially expressed probe set IDs to meaningful S. aureus genomic loci, we compared target DNA sequences corresponding to differentially expressed probe sets to S. aureus sequences in GenBank (see Materials and Methods).
A total of 78 unique probe sets corresponding to 72 ORFs, regulatory RNAs and intergenic regions were differentially regulated at least 10-fold across the three comparisons (Table 2  and Table S1). Table S1 is an expanded version of Table 2 showing probe set IDs, BLAST hit distribution among S. aureus COL, Mu50, N315, and NCTC 8325 genomes, and fold change data for every gene shown in Table 2, irrespective of meeting the fold change cut-off of 10. Fold changes $3 and ,10 are also shown in Table 2 and are italicized. Two probe sets identified as being differentially regulated in the VH versus CDM analysis, sa_i7808d_x_at and sa_i9119u10r_x_at, query similar sequence at non-syntenic regions of multiple S. aureus genomes and could not be assigned to a single genomic locus (Table S1). Eleven of the differentially expressed probe sets identified for the two comparisons using CDM-grown cells as controls had high standard deviations (Table S1). Further investigation of the CDM control arrays revealed that those 11 probe sets were themselves differentially expressed between the two CDM controls (Table  S1). Data for those 11 probe sets are shown only in Table S1. No other potential conflicts were detected in the control CDM arrays.
The differentially expressed genes identified by our microarray analysis can be divided into three categories: (1) genes putatively involved in transport or metabolism of nutrients; (2) putative or confirmed virulence and/or biofilm factors; and (3) probable prophage or genomic island genes of unknown significance. Perhaps not surprisingly, most differentially expressed genes in the analysis were assigned to the first category.
Several putative transcriptional units are highly up-regulated during growth in AH as compared to CDM, and are also downregulated during VH growth compared to AH, suggesting that the corresponding carbon substrates may be specific to or more abundant in AH. These include SACOL0308-0310, putatively involved in pseudouridine transport and catabolism; SA-COL0311-0312, encoding an operon required for sialic acid catabolism in S. aureus [30]; and SACOL0400-0403, putatively involved in ascorbate uptake. In addition to these genes, SACOL0176-0179, SACOL0192-0195, and SACOL0200 are upregulated during growth in AH, and each appear to involved in the uptake of and transcriptional response to sugars or phosphosugars (Table S1).
We additionally identified genes putatively involved in lysine biosynthesis from aspartate (SACOL1360, bioA, bioD) and gluconeogenesis (gpm, SACOL2527) as being up-regulated during growth in both AH and VH as compared to CDM, suggesting that their regulation is specific to growth in ocular fluids. Another gene putatively involved in gluconeogenesis, gapA2, was also differentially regulated, but its up-regulation was specific to AH growth.
We also observed differential regulation of a set of putative prophage and/or genomic island genes (Table 2). Interestingly, all of these genes were down-regulated in both AH and VH relative to growth in CDM, suggesting that increased expression of these genes is CDM-specific. The significance of this is unknown. Additionally, because genome sequence is not available for SA564 and thus the extent of its accessory genome is unknown, we cannot exclude the possibility of AH-and/or VH-specific differential regulation of prophage, plasmid and island genes that are not represented on the S. aureus Affymetrix chip.
Our microarray analysis was verified using semiquantitative RT-PCR to verify expression of a few genes discussed including tst, cidA and nanA (Figure 2). A more intense signal was observed for tst and nanA from AH samples when compared to CDM, and cidA signal was more intense in VH when compared to CDM. The results are consistent with the differential regulation of those genes in our microarray analysis.
The S. aureus SA564 codY mutant in CDM, AH, and VH We speculated that codY might have a role in ocular infections given the presence of leucine, isoleucine and valine in human AH and VH [16,17,31,32] and given the possibility that those substrates might become depleted during in vivo S. aureus growth. We first explored the effect of codY deletion on SA564 transcriptional responses to CDM, bovine AH and bovine VH, using the previously described SA564 codY-mutant strain, CDM7 [29]. Using GeneChip analysis, we identified 130 probe sets as being differentially expressed $5-fold, corresponding to 125    ORFs, regulatory RNAs and intergenic regions (Table 3 and  Table S2). Genes differentially regulated by the S. aureus SA564 codY mutant as compared to the wild-type strain during exponential growth in CDM are similar to those previously identified for S. aureus Newman (1) and S. aureus UAMS-1 (2) codY mutants during exponential growth in a chemically defined medium and tryptic soy broth, respectively (Table S2). As for Newman and UAMS-1, codY inactivation in SA564 results in an up-regulation of amino acid metabolic and virulence genes including BCAA metabolism (ilvDBC, leuABCD, ilvA), hemolysins (hlb, hld), and phenol-soluble modulins (SACOL1186, SACOL1187). As expected based on previous studies on CodY regulation in SA564 [29], we observed up-regulation of icaA, RNAIII and hld. We also identified expression of an enterotoxin, capsular polysaccharide biosynthesis proteins, metalloproteinase aureolysin and others as being affected by codY deletion. All other genes are shown in Table 3. Interestingly, of 117 ORFs, regulatory RNAs and intergenic regions identified as being differentially regulated in CDM when comparing S. aureus codY-mutant to SA564, all but one (SA1817, the enterotoxin gene) are core to the four S. aureus strains used to generate the GeneChip (Table S2) [33]. However, it is possible that as yet unknown S. aureus SA564-specific genes are under CodY control.
Of the 117 ORFs, regulatory RNAs and intergenic regions identified as being differentially regulated in the SA564 codY mutant relative to the wild-type strain during growth in CDM, 55 of those were also identified as being differentially regulated during growth in VH (Table 3 and Table S2). Of the 117, only 23 were identified as being differentially regulated during growth in AH. We were curious as to why this expression pattern was observed in AH, and whether it could be explained by a relief of CodY repression occurring during growth in AH at the cell density chosen for our microarray experiments. To explore this further, we returned to the microarray analysis of the SA564 wild-type strain grown in AH, as compared to CDM. Expression data for all differentially expressed genes for the SA564 codY mutant during growth in CDM were extracted from each of the four SA564 AH versus CDM analyses (Table S2). Analysis of these data revealed that genes previously identified as being under CodY control [1,2,5,29] were de-repressed in one AH sample (AH2), but not the other (AH1), relative to CDM (Table S2), most likely as a result of BCAA becoming depleted from the pooled AH2 sample. Speculatively, these data suggest that, in vivo in the anterior chamber, where BCAA are present [16,31], CodY repression may limit virulence of S. aureus during early stages of infection when cell densities are likely to be low. Future studies that track BCAA concentrations and expression of CodY-regulated genes in ex vivo AH samples over the course of S. aureus growth could be used to explore this further.
We additionally observed AH-specific up-regulation of lrgAB and VH-specific up-regulation of cidA. lrgAB and cidA are involved in coordination of cell death and autolysis, in addition to their role in biofilm development through the release of genomic DNA that becomes a structural component of the biofilm matrix. Note that lrgAB expression varied between the two AH gene expression experiments, with low albeit significant up-regulation observed during growth in one AH sample (1.4-2.6 fold upregulated compared to CDM controls), and comparatively higher up-regulation during growth in another AH sample (17.1-27.8fold up-regulated compared to CDM controls).
codY deletion enhances S. aureus virulence in a murine AC infection model To assess a potential role for CodY in endophthalmitis, we began by examining the potential for intraocular growth of S. aureus after injection into the murine AC. Approximately 5610 3 CFU of S. aureus SA564 or codY-mutant were injected into the murine AC, and bacterial growth was assessed after 24 h by extraction and homogenization of the entire eye (Table 4). In all cases, viable S. aureus were recovered (SA564 range, 4610 2 -2.3-10 6 CFU; codY mutant range, 1610 2 -1.3610 6 CFU). Average in vivo growth yields of SA564 and CDM7 were similar (4.6610 5 CFU for SA564; 2610 5 CFU for CDM7). Thus, after introduction into the anterior chamber, the murine eye was a permissive environment for the survival of each strain. Because the entire eye was homogenized for these experiments, it is unknown whether growth occurred in the anterior chamber, the posterior chamber, or both. We next assessed the impact of codY on inflammation (Fig. 3A) and retinal responsiveness (Fig. 3B) at 24 h after infection. The distribution of inflammation scores were significantly different for eyes infected with the two strains (p,0.001; one-tailed Wilcoxon rank-sum test), with a higher average inflammation score for the codY mutant (3.6 versus 2.1) (Fig. 3A). Average retinal responsiveness was lower for eyes inoculated with the codY mutant (45% of control versus 80% of control), and the distribution of percent retinal responsiveness values was significantly different between eyes infected with the codY mutant and those infected with SA564 (p = 0.001, one-tailed Wilcoxon rank-sum test) (Fig. 3B). Representative histology images are shown in Fig. 4. As seen in the figure, the eye infected with the codY mutant shows more inflammation in the cornea, anterior chamber, and vitreous, as well as disruption of retinal architecture. Collectively, these data suggest that CodY regulation of its target genes limits S. aureus disease in the murine anterior chamber infection model.
Similar experiments were performed using previously described MS7, codY complemented strain [29]. In this vector, codY expression is under the control of a leaky IPTG-inducible promoter (P SPAC ) [29]. codY expression from pTL6936 appears to be leaky, as partial complementation was observed for an S. aureus UAMS-1 codY mutant in the absence of IPTG [34]. We observed increased retinal responsiveness and decreased inflammation as assessed by slitlamp for murine eyes inoculated with 10 3 -10 4 CFU MS7, compared to SA564 and CDM7 (Fig. 4A-B), suggesting that in vivo complementation of the codY occurred. However, decreased in vivo growth yields were observed for MS7 compared to SA564 and CDM7 (Table 4; range, ,100-2.5610 5 CFU; average, 1.9610 4 CFU). Thus it is unknown whether the increased retinal responsiveness and decreased inflammation observed for MS7 were due to complementation of the codY lesion or to an in vivo growth defect of this strain. MS7 does not have an in vitro growth defect relative to CDM7 as assessed by growth in CDM; the average doubling time of MS7 is 49.060.5 min, compared to 50.760.3 min for CDM7 (Fig. 1).

Discussion
CodY controls expression of virulence and metabolic genes in response to branched-chain amino acids (BCAA) and GTP. This makes it an important regulatory link between nutrient availability and virulence factor production [1]. However, little is known about its contribution to virulence in vitro and in vivo in the ocular milieu.
AH is a complex mixture of electrolytes, organic solutes, growth factors, cytokines, and proteins including BCAA that provide the metabolic requirements to the avascular tissues of the anterior segment. It is produced from the non-pigmented ciliary body epithelium through simple diffusion as well as active transport of ions and solutes and exits the anterior chamber mainly via the trabecular meshwork. The volume of human AH in the anterior chamber generally turns over once every 100 minutes replenishing nutrients that have been taken up by avascular ocular tissues and carries away metabolic wastes [16,17,31].
In this study, we demonstrated that bovine AH and VH provide adequate nutrition for growth of S. aureus SA564, and result in differential gene expression when compared to each other, and to a defined medium. While our in vitro bovine AH and VH models lack the nutritional replenishment and immune response that would be characteristic of in vivo growth environments, the models are useful in that they allow for the identification of nutrients that S. aureus specifically detects and responds to in the ocular milieu, in particular, sialic acid, ascorbate, and pseudouridine.
Pseudouridine is a nucleoside present in RNAs of humans and other animals [32]. It has been detected in tRNAs of the bovine lens [34]. SACOL0308, SACOL0309, and SACOL0310 share protein sequence homology and conserved protein domains with the YeiC, YeiN, and YeiM proteins, respectively, of Escherichia coli (Table S1) [35]. YeiC is a pseudouridine kinase, and YeiN is a pseudouridine-59-phosphate glycosidase. Together, YeiC and YeiN comprise a pathway for the catabolism of pseudouridine to uracil and ribose-5-phosphate [35]. YeiM is a predicted nucleoside transporter and may be involved in uptake of pseudouridine from the environment. It is possible that S. aureus SA564 catabolizes pseudouridine for energy (via ribose-5-phosphate), and/or for uracil scavenging in AH.
N-acetylneuraminate, another highly upregulated gene in AH, is the primary sialic acid moiety present in mammalian tissues, and sialic acid modification of human cell surfaces is used as a ''self versus non-self'' signal to the immune system, allowing for discrimination of cell types, among other functions of sialic acids [36]. In the eye, sialic acid is distributed in all structures, including cornea, sclera, AH, trabecular meshwork, lens, VH and retina, and its concentration seems to increase with aging [37]. SACOL0312 and SACOL0311 encode a putative sodium:solute symporter protein (NanT) and N-acetylneuraminate lyase (NanA), respectively. NanA converts N-acetylneuraminate to N-acetylmannosamine and pyruvate [38]. Recently, it was shown that nanA and nanT are co-transcribed in S. aureus strain AH1263 [30]. Further, nanA and nanT are required for S. aureus growth with sialic acid [30]. Our microarray results suggest that S. aureus SA564 transports and catabolizes sialic acid during growth in bovine AH.
Ascorbate (vitamin C) is abundant in the eye and present at much higher concentrations than in the plasma [39]. It is actively transported by the iris-ciliary body into the AH and serves as an antioxidant to protect the eye against light-induced free radical damage [40]. SACOL0400-SACOL0403 encode a putative ascorbate uptake transport system (ulaABC) and a predicted transcriptional antiterminator (Table 2 and Table S1). In E. coli, the PTS-like UlaABC system (alternatively named SgaTBA) transports ascorbate with concomitant phosphorylation, trapping ascorbate-6-phosphate in the cell [41]. S. aureus may transport ascorbate to prevent Fenton reaction damage occurring as a result of extracellular iron reduction by ascorbate under aerobic conditions [42].
Several virulence factors were specifically up-regulated during growth in AH, including an enterotoxin (sec3), the toxic shock syndrome toxin (tst), and a phenol soluble modulin ( Table 2). The toxic shock syndrome toxin and the SEC enterotoxin are superantigens that are important in infections such as infective endocarditis and pneumonia [43][44][45]. Immunization against those exotoxins was found to protect against those serious illnesses [44,46]. Phenol soluble modulins have also been found in animal models to have an essential role in bacteremia and skin infections [47]. The AH-specific up-regulation of these factors may facilitate translocation of S. aureus or S. aureus-produced factors into the posterior chamber and/or retinal damage during S. aureus endophthalmitis.
As for the analysis of the codY mutant, our microarray results demonstrate that codY deletion impacts expression of metabolic and virulence genes in S. aureus SA564. However, genes affected by codY were not consistent across the two pooled AH samples used here, suggesting that BCAA became depleted from one of the samples. These data indicate that, in vivo in the AC, where BCAA are present and replenished by AH turnover continuously, CodY repression may limit virulence of S. aureus during early stages of infection when cell densities are likely to be low. Consistent with this proposal, deletion of codY enhanced virulence of S. aureus in a murine AC infection model, as assessed by retinal function measurements, degree of inflammation in the eye, and histological assessments of ocular tissue damage. The microarray results suggest a role for enterotoxin (sec3), the toxic shock syndrome toxin (tst), and a phenol soluble modulin in endophthalmitis progression. Collectively, these data suggest that CodY repression of its target genes limits S. aureus disease in the murine AC infection model.
In conclusion, we used novel in vitro and in vivo infection models to characterize the behavior of S. aureus during endophthalmitis, one a nutritional model utilizing bovine ocular fluids as media for S. aureus growth ex vivo, and one an in vivo infection model evaluating endophthalmitis progression after S. aureus injection into the murine AC. We identified metabolic pathways that may be important for S. aureus endophthalmitis, specifically sialic acid, ascorbate, and pseudouridine metabolism. We are now directly assessing the roles of these pathways in the pathogenesis of S. aureus endophthalmitis. We additionally identified several virulence factors whose expression was activated by growth in ocular fluids, suggesting that transcriptional regulation of these genes is influenced by specific nutrients present in the eye. Our in vivo endophthalmitis model, a murine AC infection model, revealed a link between the BCAA-responsive transcriptional regulator CodY and experimental endophthalmitis progression. More specifically, relief of CodY repression of its target genes (by deletion of codY) enhanced S. aureus pathogenesis in the murine eye. Interestingly, this result suggests that it may be possible to use BCAA to mitigate S. aureus endophthalmitis progression by supplementing BCAA in eye drops postoperatively or by adding BCAA to the infusion solution that goes through the eye during intraocular surgery.

Supporting Information
Table S1 An expanded version of Table 2 with probe set IDs, BLAST hit distribution among S. aureus COL, Mu50, N315, and NCTC 8325 genomes, and fold change data for every gene shown in Table 2, irrespective of meeting the fold change cut-off of 10. (PDF)