Coordinated regulation of transcription by CcpA and the Staphylococcus aureus two-component system HptRS

The success of Staphylococcus aureus as a pathogen is due in part to its ability to adapt to changing environmental conditions using signal transduction pathways, such as metabolite- responsive regulators and two-component systems. S. aureus has a two-component system encoded by the gene pair sav0224 (hptS) and sav0223 (hptR) that regulate the hexose phosphate transport (uhpT) system in response to extracellular glucose-6-phosphate. Glycolytic intermediates such as glucose-6-phosphate are important carbon sources that also modulate the activity of the global metabolite-responsive transcriptional regulator CcpA. Because uhpT has a putative CcpA binding site in its promoter and it is regulated by HptR, it was hypothesized the regulons of CcpA and HptR might intersect. To determine if the regulatory domains of CcpA and HptRS overlap, ccpA was deleted in strains SA564 and SA564-ΔhptRS and growth, metabolic, proteomic, and transcriptional differences were assessed. As expected, CcpA represses hptS and hptR in a glucose dependent manner; however, upon CcpA derepression, the HptRS system functions as a transcriptional activator of metabolic genes within the CcpA regulon. Importantly, inactivation of ccpA and hptRS altered sensitivity to fosfomycin and ampicillin in the absence of exogenous glucose-6-phosphate, indicating that both CcpA and HptRS modulate antibiotic susceptibility.


Introduction
Staphylococcus aureus is a common commensal bacterium and a versatile pathogen that can infect nearly any anatomic site. As a commensal, S. aureus persistently colonizes the nares of 12-30% of the population, while 16-70% of people are intermittent carriers [1][2][3]. The prevalence of S. aureus and its increasing antibiotic resistance contributed to a 50% increase in diagnoses from 9.17 to 13.79 per 1000 hospitalizations from 1999 to 2005 [4]. This prevalence and a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 the success of S. aureus as a pathogen are strong indicators of its ability to adapt and survive in diverse conditions. S. aureus survival in a host requires nutrients, avoidance of the immune system, and proliferation within a changing environment. To survive environmental challenges, S. aureus must be able to respond to changing environmental cues [5][6][7]. Rapid responses to environmental challenges are achieved by alterations in enzymatic activity and through transcriptional and translational initiation changes [6]. Feedback and feedforward enzymatic changes are induced by changes in substrate, cofactor, and/or inhibitor concentrations, while transcriptional and translational changes are mediated through metabolite, metal ion, and/or cofactor-responsive regulators (e.g., CcpA, Fur, Rex), small RNAs (e.g., RsaE), alternative sigma factors (e.g., σ B ), and two-or three-component regulators (e.g., KdpD-KdpE) [8][9][10][11][12][13][14][15][16][17].
The metabolite-responsive carbon catabolite protein A (CcpA) belongs to the LacI protein family and regulates transcription in response to changes in metabolite concentrations such as glucose, fructose, and glycerol [8,18]. These carbon sources can be converted into the activated glycolytic intermediates glucose-6-phosphate and fructose-1,6-bisphosphate through glycolysis or gluconeogenesis. Glucose-6-phosphate and fructose-1,6-bisphosphate promote phosphorylation and activation of the CcpA co-regulator, the histidine containing protein (HPr) [12,13,19]. CcpA activity is also regulated through phosphorylation by serine/threonine protein kinase 1 (Stk1) in response to changes in intracellular organic phosphate [20][21][22]. Other transcriptional regulators directly bind to metal ions such as Fur (Fe 2+ ) or dinucleotides such as Rex (NAD + /NADH) [14,15,18]. Together these regulators alter transcription of genes involved in diverse cellular processes such as metabolism, virulence factor synthesis, and antibiotic resistance. There are also non-protein regulators that respond to environmental stimuli such as small regulatory RNAs (e.g., RsaE), which bind to RNA and alter translation initiation or stability of target mRNAs [16,17]. Interestingly, transcription of several members of the Rsa family of sRNAs (rsaA, rsaD, rsaF, rsaE) are regulated by the alternative sigma factor B (σ B ), which confers promoter specificity for RNA polymerase to target genes, in response to physical and chemical stresses [16,17,23,24]. Environmental stresses also activate other regulators such as two-and three-component systems.
Two-and three-component signal transduction systems, such as the S. aureus KdpD-KdpE two-component system, serve as a stimulus response coupling mechanism that allows bacteria to alter transcription of genes necessary for survival in response to changing environmental conditions [10,11]. The two-component systems are generally comprised of a sensor histidine kinase and a response regulator [11]. Signal sensing is through the input domain in the sensor histidine kinase that activates phosphorylation of a histidine in the autokinase domain. The phosphoryl group is subsequently transferred to an aspartate in the receiver domain of the response regulator, increasing output domain activity, and altering the transcription of response-regulated genes [11]. In the example of the KdpD-KdpE two-component system, changes in environmental potassium can alter the concentration of intracellular potassium and phosphate, affecting the ionic strength of the cytoplasm, and/or the ATP concentration [25]. In response, the KdpD-KdpE two-component system alters transcription of potassium ion transport genes (i.e., kdpA, kdpB, and kdpC) [26,27]. Activation of KdpD results in autophosphorylation of histidine 673, which then transfers the phosphoryl group to aspartate 52 on the response regulator KdpE [27]. In addition to activating transcription of the kdpABC operon, KdpE alters transcription of S. aureus virulence genes, such as protein A, α-toxin, and aureolysin [27][28][29]. Interestingly, kdpDE transcription is under the control of another twocomponent system (AgrC-AgrA) that responds to cell density and metabolic signals [29]. In total, S. aureus are hypothesized to have 16 two-component systems, many of which are named and at least partially described [9,30]. Of the partially defined two-component systems in S. aureus, one is encoded within the gene pair sav0224 (hptS) and sav0223 (hptR). The HptRS two-component system is in a predicted three-gene operon downstream of sav0225 (hptA), a putative metal binding protein. Recently, it was suggested that HptS and HptR regulate transcription of the hexose phosphate transport gene (uhpT) in response to changes in the availability of exogenous glucose-6-phosphate [31,32]. The hexose phosphate transport system functions as an inorganic phosphate-coupled hexose phosphate sugar transporter for sugars such as glucose-6-phosphate, fructose-6-phosphate, and mannose-6-phosphate [33,34]. This transport system is medically relevant because it is one of only two known transporters (i.e., UhpT and GlpT) that mediate the uptake of the antibiotic fosfomycin [35,36]. In an S. aureus USA300 lineage strain, inactivation of either hptS or hptR increased fosfomycin resistance, reduced transcription of uhpT, and reduced growth when cultivated in a chemically defined medium containing glucose-6-phosphate as the primary carbon source [31]. Interestingly, in S. aureus strain NCTC8325, it was demonstrated that HptA binds to glucose-6-phosphate and may function to activate the HptRS two component system [32]. Although cultivation in a chemically defined medium containing glucose-6-phosphate indicated that HptRS regulated the UhpT transporter, it is likely that cultivation on a single carbohydrate source caused widespread changes in metabolism. Commonly, carbohydrate-dependent metabolic adaptation is regulated in part by CcpA [8,13], thus linking separate regulatory systems that balance S. aureus responses to the availability of rapidly catabolizable carbon sources.
Not only is glucose-6-phosphate the proposed effector molecule of HptA, it is also an effector of CcpA [13]. This is noteworthy because the uhpT gene promoter contains both a HptR (-67 to -96 bp), and a putative CcpA binding site (cre site; -61 to -46 bp), suggesting uhpT gene transcription is coordinately regulated by HptR and CcpA [18,32]. Support for CcpA-mediated transcriptional regulation of uhpT was observed when inactivation of CcpA abolished the glucose-dependent increase in uhpT mRNA [8]. Interestingly, transcription of the hptR, hptS, and hptA operon is induced in S. aureus strain COL under anaerobic cultivation, which is consistent with reduced CcpA activity during anaerobic growth due to a decrease in phosphorylation of the HPr co-regulator [8,37,38]. These observations indicate the CcpA and HptR regulons overlap. In total, it is likely that CcpA and HptR form an undefined, complex regulatory network that responds to environmental changes to regulate uhpT and other genes. To determine if HptRS is within or interacts with the CcpA regulon, ΔccpA, ΔhptRS, and ΔhptRS/ ccpA mutants were constructed in S. aureus strain SA564 and were assessed for growth, metabolic, proteomic, and transcriptional differences.

Bacterial strains and cultivation conditions
Bacterial strains, bacteriophage, and plasmids used in this study are listed in Table 1. Escherichia coli strain DH5α was cultivated in lysogeny broth (LB) or on LB containing 1.5% agar. S. aureus strains were grown in tryptic soy broth (TSB) containing 0.25% dextrose (BD Biosciences), TSB without dextrose (TSB-dex; BD Biosciences), or on TSB containing 1.5% agar (TSA). Bacterial pre-cultures were inoculated 1:100 from overnight cultures into TSB or TSBdex, incubated at 37˚C, aerated at 225 rpm using a flask-to-medium ratio of 10:1, and grown for 2 h. These exponential growth phase pre-cultures were centrifuged for 5 min at 5,000 rpm (4,272.5 X g) at 22˚C and suspended in 1-2 mL of medium. Primary cultures were inoculated into 100 mL pre-warmed TSB or TSB-dex, at an absorbance at 600 nm (A 600 ) of 0.01, and incubated at 37˚C, aerated at 225 rpm, with a flask-to-medium ratio of 10:1 (aerobic) or 10:8 (microaerobic/anaerobic). The A 600 and pH were recorded hourly. The data were analyzed using a 2-way ANalysis Of VAriance (ANOVA) using SigmaPlot version 11.1, from Systat Software, Inc. (San Jose California USA).
For antibiotic disk diffusion assays, bacterial strains were inoculated from a single colony into 2 mL TSB in a 14 mL culture tube and incubated overnight at 37˚C and aerated at 225 rpm. A 2 h pre-culture was prepared as described. After pre-incubation, the A 600 were recorded and bacteria were diluted to 0.5 McFarland units (i.e., A 600 = 0.08; 1-5 x 10 8 CFU/ mL). This suspension was diluted 1:10 and 100 μL was spread onto TSA plates and antibiotic disks (Becton, Dickinson and Company) were placed on each plate. Plates were incubated at 37˚C overnight. Zone of inhibition was recorded by measurement of the space between the edge of the disk and the first colony. The data were analyzed using a 2-way ANOVA.
For antimicrobial susceptibility tests, the broth microdilution method of the Clinical and Laboratory Standards Institute (CLSI) was utilized, with the exception that no glucose-6-phosphate was added to the medium [43]. Briefly, bacterial strains were inoculated from a single colony on TSA into cation adjusted Mueller Hinton broth (CAMHB) with 50 mg/L Ca 2+ and 12.5 mg/L Mg 2+ , and incubated at 37˚C under aeration at 225 rpm [43]. 2 h pre-cultures were prepared as described. The pre-cultures were diluted to produce a final inoculum of 1-5 × 10 5 CFU/mL in 96-well flat bottom cell culture plates. The plates were incubated at 37˚C for 20 h and the absorbance at 595 nm (A 595 ) was recorded in a plate reader. The minimum inhibitory concentration (MIC), as defined by CLSI, is the lowest concentration of antimicrobial agent that completely inhibits growth of the organism in microdilution wells as detected by the unaided eye [43]. To eliminate human error, an A 595 � 0.05 was considered to reflect no bacterial growth; hence, the MIC was defined as the lowest concentration of antimicrobial agent that inhibits growth at A 595 � 0.05. The data were analyzed using a 2-way ANOVA.

Construction of hptRS and ccpA mutants in S. aureus
Primers ( Table 2) used for PCR amplification of S. aureus strain SA564 genomic DNA were designed using the SA564 genome sequence (NCBI Reference Sequence: NZ_CP010890.1) as a template. Inactivation of hptRS in strain SA564 was accomplished by deleting a 2,380 bp region, which included all of sav0223 (hptR) and the majority of sav0224 (hptS), using the gene splicing by overlap extension technique [44]. Primers MUT0223CF2 and MUT0223Y were used to amplify a 1,416 bp region upstream of sav0223 (fragment 1) and primers MUT0224AR and MUT0224X were used to amplify a 1,849 bp region that included 147 bp at the beginning of sav0224 and the region downstream of sav0225 (fragment 2). Primers MUTO224AF and MUT0223CR, were used to sew fragments 1 and 2 together to create the spliced gene fragment with KpnI and BamH1 restriction sites for digestion and ligation into plasmid pBT-2. The size AGTCAGGTCGACGAAGGAGGGATTCGTCATGTTG SAV0224-ermB-r AGTCAGGTCGACGCGACTCATAGAATTATTTCCTCCC PCR and DNA Sequencing confirmation Primers of the spliced gene fragment was confirmed by restriction digestion. The ermB cassette from plasmid pEC4 was amplified using primers SAV0224ermB-F and SAV0224ermB-R and inserted between the cloned sav0225 and sav0222 DNA by digestion with BsiW1 to create plasmid pBT2-Δsav0223/0224::ermB. This plasmid was electroporated into E. coli strain DH5α, harvested, and transformed into S. aureus strain RN4220. S. aureus strain RN4220Δsav0224/ 0223::ermB was constructed using the temperature shift protocol of Foster et al. [45]. The Δsav0223/0224::ermB mutation from strain RN4220-Δsav0224/0223::ermB was transduced into S. aureus strain SA564 using S. aureus phage φ11 to make strain SA564-Δsav0224/0223:: ermB. The resulting transductants were selected for erythromycin resistance and confirmed by PCR. To reduce the likelihood of spurious mutations from selection and transduction, the Δsav0224/0223::ermB allele was backcrossed into wild-type strain SA564 using transducing phage 80α to create SA564-ΔhptRS::ermB (ΔhptRS). Deletion of genes sav0224/0223 were confirmed by DNA sequencing (Eurofins MWG Operon) using primers SAermB375R, SA0222-f-seq500, SAV0222-fseq1000, MUT0223CF, SAMUT0223R400, SAMUT0224ermB222F, and SAMUT0224AF. The ccpA::tetM allele was transduced from S. aureus strain Newman ccpA:: tetM using phage 80α into strain SA564-ΔhptRS::ermB to make strain SA564-ΔhptRS::ermB/ ccpA::tetM (ΔhptRS/ccpA). Successful transduction was confirmed by PCR of both ccpA and hptRS mutant regions using primers acuC-R, SAV1652-F and SAV0225-R, SAV0222-F.

Determination of acetate, glucose, and lactate levels in medium supernatant
Bacterial cultures (1 mL) were harvested hourly during cultivation and centrifuged for 5 min at 20,800 x g at 4˚C. The cell-free media were transferred to microfuge tubes and stored at -20˚C until needed. Glucose, acetate, and ammonia concentrations in the culture media were determined (n = 3) with kits purchased from R-biopharm.

Reverse transcription PCR
For quantification of mRNA by reverse transcription PCR (RT-qPCR), RNA isolation and RT-qPCR were performed as described [46]. Briefly, cDNA (20 ng/reaction) was used for realtime amplification using primers listed in Table 2. 16S rRNA was used as an internal reference and transcript levels relative to 16S RNA were determined by the comparative threshold (ΔΔC T ) method (Bio-Rad) using two technical replicates per plate. All experiments were performed in biological triplicate for each gene. Experimental setup and data analysis were carried out using a Bio-Rad CFX96 real-time PCR detection system and CFX Manager software version 3.1 following the minimum information for the publication of quantitative real-time PCR (MIQE) guidelines [47]. The data for each biological replicate (n = 3) was compiled using the gene study tool within the BioRad CFX manager software package. In order to assess significance, a p-value cutoff of 0.05 (p � 0.05), and fold change greater than 1.5 (FC � 1.5), were utilized to determine if there was a statistically significant difference between the transcript abundance of the gene when strains were compared (Tables A-D in S5 File).

Protein collection and processing
Bacteria (10 A 600 units) were harvested after 2 and 6 h cultivation by centrifugation and suspended in 1 mL of lysis buffer containing 50 mM tris-HCL, 8 M urea, and 1.5 mM phenylmethylsulfonyl fluoride. The bacterial suspensions were lysed for 40 s at a setting of 6 m/s in a FastPrep instrument (MP biomedical) and the lysate was clarified by centrifugation for 5 min at 20,800 x g at 4˚C. Cell-free lysates were stored at -80˚C until use. Protein concentrations were determined using a Modified Lowry Protein Assay Kit (Fisher Scientific) and diluted to a final concentration of 235 μg/mL in ultrapure water. Protein samples were mixed with a 3X volume of cold acetone and incubated at -20˚C for 30 min and centrifuged for 2 min at 15,000 x g to pellet the precipitated proteins. Protein pellets were washed twice with cold acetone, dried 15 min in a speed vacuum, and stored at -80˚C. Protein samples were solubilized in 20 μL denaturing buffer (25 mM ammonium bicarbonate, pH 8.0; 10 mM tris (2-carboxyethyl) phosphine hydrochloride (TCEP); 5% sodium deoxycholate) and incubated for 10 min at 60˚C. Thiol alkylation was achieved by adding 5 μl alkylation buffer (100 mM iodoacetamide in water) and incubating the samples for 1 h at room temperature. Following alkylation, samples were diluted with 275 μL dilution buffer (25 mM ammonium bicarbonate, pH 8), 2 μL trypsin solution (1 mg/mL) was added, and incubated overnight at 37˚C. To stop the reaction, 10 μL of 10% trifluoroacetic acid (TFA) was added, incubated for 30 min, and centrifuged at 15,000 x g. Supernatants were transferred to microfuge tubes and used for LC-MS E analysis.

LC-MS E analysis
All analyses were carried out using a Waters nanoAcquity UPLC System coupled to a Waters Synapt G2 mass spectrometer. The mobile phases were composed of Solvent A (0.1% formic acid in H 2 O) and Solvent B (0.1% formic acid in acetonitrile). Injection volume was 2 μL. The source temperature was set at 70˚C. Cone gas flow was maintained 6 L/min. Acquisition range was 50-2,000 m/z. MS E data was collected with alternating low (4 eV) and elevated (ramp from 17 to 42 eV) energy over a 100-1500 m/z range. Spectra and peptide identification statistical analysis were carried out using Progenesis software (Waters Corporation, MA) and was searched against all S. aureus Mu50 predicted proteins (NCBI database).

Power analysis
Experimental variability was assessed using a pilot proteome study of strains SA564 and SA564-ΔhptRS cultivated in TSB. A power analysis was conducted to determine the number of biological replicates needed to achieve the probability of detecting a true difference between strains during the exponential growth phase (2 h) and post-exponential growth phase (6 h) when one exists. It was determined that 5 biological replicates of each strain, growth phase (hour), and cultivation condition (media) combination would achieve a minimum of 74% power assuming dispersion (variability) of 0.05 and 99% power assuming dispersion of 0.01 (

Statistical analysis of the total cytosolic proteome
The relative normalized abundance of each peptide was extracted as a .txt file from Progenesis for statistical analysis. Statistical analysis was performed using R 3.3 software, R Core Team (2017) [48]. The supporting information file 1 contains examples of the R code corresponding to the proteomic analysis (S1 File). Final data sets were exported into Microsoft Excel files for final organization and interpretation. Peptide data were visualized in a histogram and the relative normalized peptide abundance had a unimodal, right skewed, distribution with responses greater than or equal to 0. For this reason, data were modeled using a generalized linear model that assumed the relative normalized peptide abundance followed a gamma distribution with strain (s), growth phase (h), and media (m) as fixed effects and the relative normalized abundance as the response variable [49]. Peptides with a 0 abundance value, due to an abundance not included in the gamma distribution, were assigned a value of 0.0001 to facilitate modeling and statistical analysis. Peptide abundance and variability were combined and protein abundance was modeled and analyzed for interactions between variables of strain (s), media (m), and/or growth phase (h) (Tables B-J in S2 File). A significant interaction would indicate that the effect of one variable depends on the magnitude of change of another variable. For any significant interaction (p � 0.2) the simple effects, the effect of a single variable given the level of another variable on protein abundance, were assessed (p � 0.05; Figures Table G in S3 File). In order to quantify any differences between SA564, SA564-ΔhptRS, SA564-ΔccpA and SA564-ΔhptRS/ccpA the relative foldchange between strains was computed and exported for proteins with a significant interaction or main effect that included strain. All proteins that were identified as significantly (p � 0.05) different between the strains were contained within the 3-way interaction of variables strain, media, and growth phase (s x m x h), two-way interaction of strain and growth phase averaged over media (s x h | m), and the main effect of strain (Figures F, H, L in S6 File; Tables A-C in S4 File). Only proteins, with at least two peptides identified within the LC-MS E analysis, a significant interaction or a main effect (p � 0.2), and with a significant difference (p � 0.05) between the strains were taken into account for subsequent interpretation and transcriptional analysis (Table 3; Tables D, E in S4 File).

Deletion of hptRS does not alter growth or sensitivity to stressors
To assess the growth and viability of strain SA564-ΔhptRS relative to the isogenic wild-type strain SA564, the growth and pH profiles were determined (Fig 1). The growth of strains SA564 and SA564-ΔhptRS were equivalent when cultivated in TSB and TSB-dex (Fig 1A and  1B). Similarly, strains SA564 and SA564-ΔhptRS had equivalent growth when cultivated anaerobically or in iron-limited TSB ( Figure A in S6 File). Strains SA564-ΔhptRS and SA564 responded comparably when cultivated in TSB with oxidative stress-inducing compounds (i.e., streptonigrin, hydrogen peroxide, diamide, and paraquat) or membrane de-couplers (i.e., carbonyl cyanide 3-chlorophenylhydrazon E and nigericin) (Figures B, D in S6 File). In addition, the antibiotic resistance profiles were equivalent when compared using disk diffusion

Deletion of ccpA alters the accumulation and/or depletion of ammonia, acetate, and glucose in the culture media
When cultivated in TSB, inactivation of ccpA slightly reduced the growth rates (μ) of strains SA564-ΔccpA (μ = 1.432 h -1 ) and SA564-ΔhptRS/ccpA (μ = 1.474 h -1 ) relative to strain SA564 (μ = 1.673 h -1 ; Fig 1A). Concomitant with the reduced growth rates in the ccpA inactivated strains, ammonia accumulation was slightly increased and the rate of glucose consumption was slightly decreased during the exponential phase (Fig 2A and 2E), indicating that the reduced growth rate is accompanied by reduced carbon flow through glycolysis and increased amino acid catabolism (Fig 3). Interestingly, while acetate accumulation in strains SA564, SA564-ΔhptRS and SA564-ΔccpA cultivated in TSB were similar, strain SA564-ΔhptRS/ccpA had a significantly altered acetate accumulation and depletion profile (Fig 2C). Similarly, ammonia accumulation was significantly different when strains SA564-ΔccpA and SA564-ΔhptRS/ccpA were compared (Fig 2A and 2B). The absence of a growth phenotype in strain SA564-ΔhptRS, but the presence of a growth phenotype in strain SA564-ΔhptRS/ccpA that is different from strain SA564-ΔccpA, indicates there is an antagonistic relationship between HptRS and CcpA. As expected, these differences decreased when the strains were cultivated in TSB-dex (Figs 1B, 1D, 2B and 2D). While these observations are consistent with previous reports on CcpA, they establish a previously unobserved linkage between HptRS and CcpA co- CcpA and HptRS co-regulation of transcription regulation [6,8,40,50,51]. To determine the extent of the regulatory linkage between CcpA and HptRS, total cytosolic proteomes were analyzed by mass spectroscopy during the exponential (2 h) and post-exponential (6 h) growth phases in the presence or absence of glucose.

Proteomics overview (experimental design and statistical analysis)
Cytosolic protein fractions were harvested for five biological replicates for each strain, cultivated in TSB and TSB-dex, during the exponential (2 h) and post-exponential (6 h) growth phases for a total of 80 samples, and analyzed by LC-MS E . Using LC-MS E , a total of 33,026 spectral peaks were collected and analyzed. The raw data were deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD008708 [52]. The spectra were compared against S. aureus strain Mu50 predicted proteins (NCBI database) using Progenesis QI, and yielded 11,971 identifiable spectral peaks, with sequences that were assigned to 501 proteins. Comparative quantitative analysis was completed for each protein incorporating the relative normalized abundance for each peptide, yielding a relative abundance for each protein. Proteins identified by two or more peptides (n = 438 proteins) were analyzed for interactions between experimental variables [i.e., strain (s), media (m), and growth phase (h); Tables B-F in S2 File]. Several groups of proteins emerged within this data set based on the interaction between the variables: (i) Proteins with an abundance that was altered between bacterial strains, cultivation media, and growth phases were defined as having a 3-way interaction (s x m x h) ( . All proteins that were selected for further analysis were input into the Protein ANnotation THrough Evolutionary Relationship (PANTHER) classification system and sorted by molecular function (Fig 4) [53]. Of the 105 proteins input into PANTHER, the majority were predicted to have catalytic activity (Fig 4) [8,51,54].
To determine if proteins with significant interactions or main effects contained statistically significant differences (p � 0.05) between the strains, growth phases, growth medium, or a combination of these variables, multiple comparisons were calculated and p-values generated for the simple effects (combination of variables) or main effect of each variable (Tables A-G in S3 File). For a given protein, if a significant 3-way interaction was observed, then the effects of each variable were considered (p � 0.05; 3-way multiple comparison); if no significant 3-way interaction was observed but a significant 2-way interaction was observed, then the effects of each variable within the 2-way analysis was considered (p � 0.05) averaging over the third variable (2-way multiple comparison); and if no significant 3-way or 2-way interactions were observed, then the effect of strain was considered (p � 0.05) averaged over media and growth phase (main effect comparison). A p-value cutoff of 0.05 was used to determine if there was a statistically significant difference between the effects of strain (s), media (m) and growth phase (h). With this cutoff, it was determined that all proteins identified as being significantly altered were contained within the 3-way interaction (s x m x h), 2-way interaction (s x h | m), and the main effect of strain (s), hence an emphasis was placed on the simple effects or main effect within these categories for comparisons between strains (Table J in S2 File; Tables A, C, G in S3 File). To quantify specific differences in protein abundance between strains SA564, SA564-ΔhptRS, SA564-ΔccpA and SA564-ΔhptRS/ccpA fold changes (i.e., SA564/mutant or mutant/ mutant) were computed for each protein with a significant interaction or main effect (Table 3; Tables A-E in S4 File). In total, 13 proteins were identified as having significant 3-way interactions, 53 proteins had significant 2-way interactions between strain and growth phase, and 124 proteins were identified for further analysis due to the main effect of strain, providing a high probability of identifying specific proteomic differences within and between strains SA564, SA564-ΔhptRS, SA564-ΔccpA and SA564-ΔhptRS/ccpA (Table J in S2 File).

Media and temporal changes in strain SA564
S. aureus adaptations to glucose-rich and -limited environments have been studied [8,51,55]. Cultivation of S. aureus in glucose-rich media increases the accumulation of proteins involved in glycolysis and protein synthesis. In contrast, cultivation in glucose-limited media increased the abundance of proteins involved in the tricarboxylic acid cycle, amino acid catabolism, and gluconeogenesis. Although less well-studied, temporal protein changes during cultivation in medium containing glucose have also been examined [55]. Temporal changes occur when glucose is depleted and incompletely oxidized carboxylic acids become the primary carbon sources (Fig 3). In response to changing carbon sources, the proteome of S. aureus must adapt to supply energy and metabolic precursors for growth, which causes a large temporal rearrangement in protein accumulation. Many of these temporal changes are mediated in part by CcpA [8,12,51]; hence, several of the significant proteins found in the 3-way and 2-way interactions are translated from genes that are known or predicted to be regulated by CcpA (Tables A-D in S3 File; Table D  2) is significantly increased during the post-exponential growth phase in strain SA564 cultivated in TSB and TSB-dex relative to the exponential phase (p � 0.05). The media and temporal differences in strain SA564, and previous observations on cultivation media and temporal differences in other S. aureus strains [8,55], establish an important baseline for understanding the proteomic changes associated with CcpA and HptRS transcriptional regulation.

Protein abundances are similar between strains SA564 and SA564-ΔhptRS
Consistent with the growth and pH profiles, deletion of hptRS did not significantly (p � 0.05) alter cytosolic protein abundance in response to changing growth phases or media composition in strain SA564-ΔhptRS relative to strain SA564 for proteins identified as having significant 3-way (s x m x h, p � 0.2) or 2-way (s x h | m, p � 0.2) interactions, or main effect of strain (s, p � 0.2) (Tables A-C in S4 File). These results indicated that HptRS may be inactive under these conditions or the genes regulated by HptRS are also under the control of another glucose-dependent transcriptional regulator, such as CcpA. To determine if CcpA masks the effect of hptRS genetic inactivation, the proteomes of strains SA564-ΔccpA and SA564-ΔhptRS/ccpA were analyzed.

Many differentially regulated proteins in strain SA564-ΔhptRS/ccpA are common to the CcpA regulon
For proteins having significant 3-way (s x m x h, p � 0.2) or 2-way (s x h | m, p � 0.2) interactions, or a main effect of strain (s, p � 0.2), 98 proteins were identified as significantly different between SA564-ΔhptRS/ccpA and SA564 (p � 0.05) ( Table E in S4 File). Of these proteins, 46.9% have been experimentally determined or predicted to be regulated by CcpA [8,18]. The majority (n = 48) of significantly different proteins found in SA564-ΔccpA (n = 56 proteins) were also found in SA564-ΔhptRS/ccpA relative to SA564 (Tables D, E in S4 File). Interestingly, while CcpA-regulated proteins were common to both SA564-ΔhptRS/ccpA and SA564-ΔccpA, several proteins (n = 12) had different fold changes in protein abundance relative to SA564. As examples, pyruvate formate lyase (EC: 2.3.1.54) and alcohol dehydrogenase (EC: 1.1.1.1) were present in greater abundance in strain SA564-ΔccpA relative to SA564, yet lower in abundance in strain SA564-ΔhptRS/ccpA relative to SA564. These data suggest an antagonistic regulatory relationship between HptRS and CcpA and that certain enzymatic/protein differences caused by inactivation of ΔhptRS can only be visualized by removing CcpA-regulated proteins from the proteome of strain SA564-ΔhptRS/ccpA. To do this, significant differences in protein abundance in strain SA564-ΔhptRS/ccpA relative to SA564-ΔccpA were analyzed.
Significant differences were also observed within pyruvate, amino acid metabolic, and TCA cycle enzymes during the post-exponential phase (p � 0.05; Table 3). As examples, pyruvate formate lyase B, and acetate kinase had significantly increased protein abundances in strain SA564-ΔccpA relative to SA564-ΔhptRS/ccpA. Amino acid metabolic enzymes such as alanine dehydrogenase, tetrahydrodipicolinate acetyltransferase (EC: 2.3.1.89), and ornithine transcarbamoylase had significantly increased protein abundances in strain SA564-ΔccpA relative to strain SA564-ΔhptRS/ccpA. These differences in pyruvate metabolic and amino acid catabolic protein abundance are consistent with the physiological differences in acetate and ammonia accumulation between strains SA564-ΔccpA and SA564-ΔhptRS/ccpA (Figs 1 and 2; Table 3). Additionally, the abundance of the TCA cycle enzyme isocitrate dehydrogenase was significantly decreased in strain SA564-ΔccpA relative to strain SA564-ΔhptRS/ccpA (p � 0.05; Table 3). Taken together, these data suggest that HptR functions to antagonize transcription of a subset of CcpA-regulated genes, which give rise to alterations in protein accumulation. To assess this possibility, the mRNA abundance of select genes was determined to gauge transcriptional regulation of genes encoding proteins that were significantly different in the proteome.

CcpA and HptRS regulate genes encoding for proteins identified as being significantly different in proteomic analysis
Proteomic analysis demonstrated that genetic inactivation of the genes coding for the transcriptional regulators ccpA alone or in combination with hptRS resulted in differential protein abundances. The most likely explanation for these results is that transcription of the genes coding for the differentially-regulated proteins was altered. To assess this possibility, mRNA transcript levels were determined for select genes using RT-qPCR. Genes for pyruvate formate lyase B (pflB; sav0226) and alanine dehydrogenase (dha1; sav1439) were selected for analysis due to their significant 3-way interaction (s x m x h; p � 0.2), and having a protein abundance fold-change greater than 1.5 (FC � 1.5) in strain SA564-ΔccpA relative to SA564-ΔhptRS/ccpA (Table 3). Alcohol dehydrogenase (adh; sav0605) was assessed due to its significant 2-way interaction (s x h | m; p � 0.2) and having a protein abundance fold-change greater than 1.5 (FC � 1.5) in strain SA564-ΔccpA relative to SA564-ΔhptRS/ccpA (Table 3). Lastly, the 1-pyrroline-5-carboxylate dehydrogenase gene (rocA; sav2554) was assessed due to its significant 3-way interaction (s x m x h; p � 0.2), and also having a protein abundance fold-change above 1.5 (FC � 1.5) in strains SA564-ΔccpA and SA564-ΔhptRS/ccpA relative to SA564 (Tables D, E in S4 File). In addition, rocA was chosen as a control due to the lack of a difference in protein abundance between strains SA564-ΔccpA and SA564-ΔhptRS/ccpA, which indicated this effect was caused by CcpA alone. An additional consideration for choosing these genes was that pflB, dha1, adh, and rocA have been previously identified or predicted to be transcriptionally regulated by CcpA [18,51,56].
The mRNA abundances of pflB, dha1, and adh in SA564-ΔccpA were significantly decreased (FC � 1.5; p � 0.05) during the exponential growth phase (Fig 5A and 5B) and significantly increased (FC � 1.5; p � 0.05) during the post-exponential growth phase (Fig 5C  and 5D) relative to SA564 in either TSB or TSB-dex (Tables A, B in S5 File). CcpA is a transcriptional repressor of pflB, adh, and dha1, therefore ccpA inactivation should increase mRNA abundance [8]. Interestingly, an increase mRNA abundance was only observed during the post-exponential growth phase for strain SA564-ΔccpA in TSB and TSB-dex (Fig 5C and  5D). As expected, post-exponential growth phase transcription of pflB, adh, and dha1 was significantly increased (p � 0.05) in strain SA564-ΔccpA cultivated in TSB relative to that cultivated in TSB-dex (Table D in S5 File). Together, these data suggest there are one or more transcriptional activators whose activity is repressed by glucose, or more appropriately glucose depletion, and that activates post-exponential growth phase transcription of pflB, adh, and dha1 when CcpA repression is relieved.
Consistent with there being an activator of pflB, dha1, and adh transcription, a significant reduction (FC � 1.5; p � 0.05) in mRNA abundance of pflB, dha1, and adh in strain SA564-ΔhptRS/ccpA relative to strain SA564-ΔccpA, irrespective of media or growth phase, (Fig 5; Table C in S5 File) was observed, indicating that HptRS activated transcription of these genes upon CcpA de-repression or inactivation. During the post-exponential growth phase, when cultivated in TSB, strain SA564-ΔhptRS/ccpA produced transcripts for pflB, dha1, and adh, indicating that other regulators are also involved in glucose-dependent transcription of these genes (Fig 5C). In contrast to pflB, dha1, and adh mRNA levels, significant exponential growth phase increases (p � 0.05) in mRNA abundance of rocA were observed in ccpA inactivated CcpA and HptRS co-regulate transcription of genes during the exponential (2 h) and post-exponential (6 h) growth phases. The relative mRNA abundance during the exponential (A, B) and post-exponential (C, D) growth phases for genes hptR, hptS, hptA, dha1, adh, ccpA, pflB, rocA, and uhpT was determined by RT-qPCR. Data are the mean and standard error of the mean for 3 biological replicates of SA564 (black bars), SA564-ΔhptRS (red bars), SA564-ΔccpA (green bars), and SA564-ΔhptRS/ccpA (blue bars) cultivated in TSB (A, C) and TSB-dex (B, D). The data were normalized using 16S rRNA and are plotted relative to SA564. A statistically significant difference (p � 0.05) between SA564-ΔccpA and SA564-ΔhptRS/ccpA is represented with an ( � ).
https://doi.org/10.1371/journal.pone.0207161.g005 strains relative to SA564, irrespective of hptRS inactivation (Fig 5A and 5B; Tables A, B in S5 File). During the post-exponential growth phase, there were only small differences in rocA mRNA levels between strains SA564-ΔccpA and SA564-ΔhptRS/ccpA (Fig 5C and 5D), confirming that rocA is primarily regulated by CcpA. Taken together, these data demonstrate that transcription of pflB, dha1, and adh is co-regulated by CcpA and HptRS, while rocA is largely regulated by CcpA. These data also demonstrate the CcpA and HptR regulons intersect and that CcpA functions as a transcriptional repressor, while HptR functions as a transcriptional activator.

Predicted membrane-associated proteins are also regulated by CcpA and HptRS
Histidine kinase HptS, the protein of unknown function HptA, and UhpT are predicted to be membrane-associated and similar to homologous proteins found in E. coli, hence they are unlikely to be present in the cytosolic fraction used for proteomic analysis [32,57]. To determine if CcpA regulates transcription of hptS, hptA, and uhpT, mRNA abundance was assessed in the exponential and post-exponential growth phases from bacteria cultivated in TSB or TSB-dex (Fig 5). Post-exponential growth phase transcription of hptS and hptA was significantly (p � 0.05) increased in strain SA564-ΔccpA relative to strain SA564 cultivated in TSB ( Fig 5C; Table A in S5 File). As expected, the magnitude of this effect was lessened when the bacteria were cultivated in TSB-dex ( Fig 5D, Table B in S5 File). These results are consistent with a hypothesis put forth previously, whereby CcpA can mediate repression even during cultivation in the absence of glucose [8]. Transcription of uhpT was significantly increased (p � 0.05) in strain SA564-ΔccpA relative to strain SA564 during the exponential and postexponential growth phases (Fig 5; Tables A, B in S5 File). Importantly, this transcriptional activation was abolished in strains SA564-ΔhptRS and SA564-ΔhptRS/ccpA, demonstrating the HptRS two-component system activates uhpT transcription. In addition, the post-exponential growth phase mRNA level of uhpT was increased (FC = 2.92) in strain SA564 cultivated in TSB-dex relative to TSB, suggesting that HptRS transcriptional activation can occur without exogenous glucose (Table D in S5 File). These data reveal a complex regulation of uhpT transcription involving CcpA repression and HptRS activation, partially in response to changes in the availability of glucose or glucose-derived metabolites. The significance of these observations relate to the importance of UhpT in susceptibility to the antibiotic fosfomycin [36]. To investigate if genetic inactivation of ccpA or hptRS altered antibiotic susceptibility through their coordinated regulation of transcription, the antibiotic susceptibility of each strain was analyzed using broth microdilution and disk diffusion assays.
As reported, CcpA mutants also have increased susceptibility to beta-lactam antibiotics [40,61]. Consistent with this observation, inactivation of ccpA increased susceptibility to penicillin and ampicillin, but also to kanamycin, neomycin, and chloramphenicol relative to strains SA564, SA564-ΔhptRS, and SA564-ΔhptRS/ccpA (p � 0.05; Figs 6B and 7). Inactivation of hptRS in the ccpA mutant background restored the resistance profile to a level similar to the wild-type strain, suggesting that HptRS antagonizes CcpA-mediated transcriptional regulation of genes that are involved in antibiotic susceptibility (Figs 6 and 7).

Conclusions
Gene regulation by multiple transcriptional regulators is common and overlapping regulation impedes the identification of genes controlled by individual regulators. On this latter point, deletion of genes coding for the sensor histidine kinase (hptS) and the response regulator (hptR) produced no observable phenotypic effects on growth, metabolism, or protein abundance. In contrast, inactivation of both ccpA and hptRS caused significant changes in pyruvate metabolism, amino acid metabolism, and TCA cycle protein abundance, which affected growth and metabolism relative to the single mutants. By combining rigorous statistical analyses with exhaustive proteomic data, it was possible to determine that HptRS and CcpA function antagonistically to co-regulate transcription of a subset of CcpA-regulated genes (Fig 8). This approach led to the identification of a small group of proteins altered by inactivation of hptRS. These data provided focus to determine which protein alterations were the result of transcriptional changes due to hptRS inactivation. Lastly, these data provide important In the absence of glucose, CcpA de-repression increases mRNA abundance of sav0225 (hptA), sav0224 (hptS), and sav0223 (hptR). Transcription is further increased by the presence of metabolites potentially derived from gluconeogensis. Upon activation by glucose-6-phosphate, or another unknown stimulus, the HptRS system functions to activate transcription of genes normally repressed by CcpA, including uhpT. Increased transcription of uhpT is correlated with increased transport of the antibiotic fosfomycin; hence, coregulation of transcription by CcpA and HptRS directly alter fosfomycin susceptibility, and susceptibility to other antimicrobials.
https://doi.org/10.1371/journal.pone.0207161.g008 information about the complex regulation of antibiotic susceptibility and how "standard" susceptibility assay conditions can influence the outcomes of these assays.