Transcriptional and Phenotypic Characterization of Novel Spx-Regulated Genes in Streptococcus mutans

In oral biofilms, two of the major environmental challenges encountered by the dental pathogen Streptococcus mutans are acid and oxidative stresses. Previously, we showed that the S. mutans transcriptional regulators SpxA1 and SpxA2 (formerly SpxA and SpxB, respectively) are involved in stress survival by activating the expression of classic oxidative stress genes such as dpr, nox, sodA and tpx. We reasoned that some of the uncharacterized genes under SpxA1/A2 control are potentially involved in oxidative stress management. Therefore, the goal of this study was to use Spx-regulated genes as a tool to identify novel oxidative stress genes in S. mutans. Quantitative real-time PCR was used to evaluate the responses of ten Spx-regulated genes during H2O2 stress in the parent and Δspx strains. Transcription activation of the H2O2-induced genes (8 out of 10) was strongly dependent on SpxA1 and, to a lesser extent, SpxA2. In vitro transcription assays revealed that one or both Spx proteins directly regulate three of these genes. The gene encoding the FeoB ferrous permease was slightly repressed by H2O2 but constitutively induced in strains lacking SpxA1. Nine genes were selected for downstream mutational analysis but inactivation of smu127, encoding a subunit of the acetoin dehydrogenase was apparently lethal. In vitro and in vivo characterization of the viable mutants indicated that, in addition to the transcriptional activation of reducing and antioxidant pathways, Spx performs an important role in iron homeostasis by regulating the intracellular availability of free iron. In particular, inactivation of the genes encoding the Fe-S biogenesis SUF system and the previously characterized iron-binding protein Dpr resulted in impaired growth under different oxidative stress conditions, increased sensitivity to iron and lower infectivity in rats. These results serve as an entryway into the characterization of novel genes and pathways that allow S. mutans to cope with oxidative stress.


Abstract
In oral biofilms, two of the major environmental challenges encountered by the dental pathogen Streptococcus mutans are acid and oxidative stresses. Previously, we showed that the S. mutans transcriptional regulators SpxA1 and SpxA2 (formerly SpxA and SpxB, respectively) are involved in stress survival by activating the expression of classic oxidative stress genes such as dpr, nox, sodA and tpx. We reasoned that some of the uncharacterized genes under SpxA1/A2 control are potentially involved in oxidative stress management. Therefore, the goal of this study was to use Spx-regulated genes as a tool to identify novel oxidative stress genes in S. mutans. Quantitative real-time PCR was used to evaluate the responses of ten Spx-regulated genes during H 2 O 2 stress in the parent and Δspx strains. Transcription activation of the H 2 O 2 -induced genes (8 out of 10) was strongly dependent on SpxA1 and, to a lesser extent, SpxA2. In vitro transcription assays revealed that one or both Spx proteins directly regulate three of these genes. The gene encoding the FeoB ferrous permease was slightly repressed by H 2 O 2 but constitutively induced in strains lacking SpxA1. Nine genes were selected for downstream mutational analysis but inactivation of smu127, encoding a subunit of the acetoin dehydrogenase was apparently lethal. In vitro and in vivo characterization of the viable mutants indicated that, in addition to the transcriptional activation of reducing and antioxidant pathways, Spx performs an important role in iron homeostasis by regulating the intracellular availability of free iron. In particular, inactivation of the genes encoding the Fe-S biogenesis SUF system and the previously characterized iron-binding protein Dpr resulted in impaired growth under different oxidative stress conditions, increased sensitivity to iron and lower infectivity in rats. These results serve as an entryway into the characterization of novel genes and pathways that allow S. mutans to cope with oxidative stress.

Introduction
Dental caries remains one of the most prevalent infectious diseases affecting billions of people worldwide [1]. Caries results from an ecological imbalance of the oral flora caused by interactions of specific bacteria (e.g. Streptococcus mutans) with salivary proteins and dietary carbohydrates (e.g. sucrose). The cariogenic potential of microorganisms is largely based on three factors: (i) the ability to form biofilms on the tooth surface, (ii) the ability to produce weak acids mainly lactic acid, and (iii) the ability to rapidly adapt to environmental stresses such as large and fluctuations in pH, oxygen tension and nutrient availability [2]. In addition to fulfilling these central requirements, S. mutans appears as one of the most dominant species during the early and intermediate stages of caries development, strengthening its association with dental caries initiation and progression [3,4].
Well-documented work from a number of laboratories has established that S. mutans is well-equipped to adapt to low pH values by activation of a robust physiological response to acidification that includes, among other responses, upregulation of the membrane-associated F-ATPase, induction of pathways that contribute to cytoplasm buffering and changes in membrane fatty acid composition [2]. In addition to acid tolerance, mounting evidence indicates that the ability of S. mutans to cope with the reactive oxygen species (ROS) generated by their own metabolism as well as by other oral species can also impact its pathogenic potential. In fact, it has been shown that, in dental plaque, there is an inverse correlation between the total numbers of S. mutans and of members of the mitis group (e.g. S. sanguinis and S. gordonii), known to produce large quantities of H 2 O 2 from the metabolic reduction of oxygen. Specifically, members of the mitis group are often associated with oral health; a series of elegant in vitro studies showed that H 2 O 2 produced by S. gordonii or S. sanguinis serves as a "chemical weapon" antagonizing the growth of S. mutans [5]. Likewise, S. oligofermentans, a recently identified oral streptococci isolated from plaque of caries-free patients, was shown to antagonize the growth of S. mutans through a lactate oxidase that generates H 2 O 2 from lactic acid [6]. Finally, H 2 O 2 present in certain oral hygiene and tooth bleaching products may represent another source of peroxide stress for oral bacteria [7].
As a facultative anaerobe that lacks catalase and a full electron transport chain, oxygen metabolism in S. mutans is thought to occur through two flavin-based enzymes, Nox and AhpF [8]. However, biochemical and physiological characterization of AhpF (an H 2 O 2 -forming NADH oxidase) and Nox (an H 2 O-forming NADH oxidase) indicates that the latter is the stronger of the two implicated in oxygen metabolism [9,10]. Loss of Nox leads to a decreased ability to metabolize oxygen and elevated expression of genes involved in ROS detoxification [9]. In conjunction with the AhpC peroxidase, AhpF was shown to be part of an alkyl hydroperoxide reductase system working as a flavoprotein dehydrogenase, which supports peroxide reduction by AhpC [8]. Other ROS scavenging and protective systems are also present in S. mutans including superoxide dismutase (SodA), thiol peroxidase (Tpx), the thioredoxin reductase system (TrxA/TrxB), glutathione synthase (GshAB), glutathione oxidoreductase (Gor), and the iron-binding peroxide resistance protein Dpr [11][12][13][14].
Spx is a global regulator ubiquitously found in low GC Gram-positive bacteria that is involved in stress survival, principally by serving as a transcriptional activator of genes involved in thiol homeostasis and detoxification [15]. Spx lacks a DNA-binding domain and its function depends on direct interactions with the RNA polymerase α-subunit, which can result in either positive or negative regulation [16,17]. There is now clear evidence that two or more Spx paralogs, with overlapping but also unique regulatory functions, contribute to adaptive stress responses of streptococcal species and, more recently, Bacillus anthracis [18][19][20][21][22][23].
Previously, we identified and characterized two Spx proteins (SpxA1 and SpxA2, previously named SpxA and SpxB, respectively) in S. mutans [18,19]. Inactivation of spxA1 significantly impaired growth and survival under acid and oxidative stress conditions. Whereas stress tolerances were generally not impaired in the ΔspxA2 strain, the stress sensitivities of the double ΔspxA1ΔspxA2 strain was more pronounced than in the ΔspxA1 strain [19]. Microarray profiling of the ΔspxA1, ΔspxA2 and ΔspxA1ΔspxA2 strains further indicated that SpxA1 plays a primary role in activating oxidative stress genes whereas SpxA2 appears to have a secondary role in the regulation of these same genes [19]. Notably, nearly every known gene of S. mutans with a proven role in oxygen metabolism or oxidative stress was found to be under Spx control. For example, transcription of ahpC, ahpF, dpr, gor, nox, sodA, tpx and trxB was downregulated in the ΔspxA1 strain and, in most cases, further downregulated in the double mutant ΔspxA1ΔspxA2 [19]. Despite the large number of genes identified in our microarray analysis, a relatively small number of genes followed this regulatory pattern, i.e. downregulated in both ΔspxA1 and ΔspxA1ΔspxA2 strains. We reasoned that some of the uncharacterized genes under SpxA1 control, and to a lesser extent SpxA2, might be involved in oxidative stress management. We performed transcriptional and phenotypic characterization of selected Spxregulated genes previously identified in our microarray analysis to further investigate oxidative stress management under Spx regulation in S. mutans. Our results strongly suggest that, in addition to reducing and antioxidant pathways, Spx also controls metal ion homeostasis, thus serving as a starting point for the characterization of novel genes and pathways that allows for S. mutans to cope with oxidative stresses.

Bacterial strains and stress growth conditions
The bacterial strains used in this study are listed in Table 1. S. mutans UA159 and its derivatives were routinely grown in brain heart infusion (BHI) or trypticase soy agar (TSA) at 37°C in a 5% CO 2 atmosphere or under anaerobic conditions (BBL Gaspack system, BD). Where appropriate, kanamycin (1 mg ml -1 ), erythromycin (10 μg ml -1 ) or spectinomycin (1 mg ml -1 ) was added to the growth medium. For mRNA quantifications, cultures were grown to OD 600 = 0.4, at which point control samples were harvested by centrifugation, while experimental samples were exposed to 0.5 mM H 2 O 2 for 5 and 15 min before harvest. Harvested pellets were stored at -80°C until use. To generate growth curves, strains were grown overnight under anaerobic conditions and diluted 20-fold in pH 7.0-buffered BHI containing H 2 O 2 (0.2 mM), methyl viologen (MV, 5 mM), diamide (0.5 mM). Cultures were incubated at 37°C in a 5% CO 2 atmosphere and the OD 600 recorded at selected intervals. To test the toxicity of iron to S. mutans strains, bacterial cultures grown to OD 600~0 .3 were diluted 1:100 in fresh BHI supplemented with 5 mM FeSO 4 . After 18h of incubation at 37°in 5% CO 2 , a 100 μl aliquot of each culture was serially diluted and plated onto BHI agar for CFU counting. To test the sensitivity of S. mutans UA159 and its derivatives to streptonigrin in disc diffusion assays, a uniform layer of exponentially-grown cells was spread onto BHI agar plates using a sterile swab, and filter papers (Whatman paper, 1 mm in diameter) saturated with 20 μl of a solution of 0.125 μg μl -1 of streptonigrin were placed onto the agar. All plates were incubated at 37°C in a 5% CO 2 for 48 h after which the diameter of the inhibition halo was measured.

Construction of mutant strains
Standard DNA manipulation techniques were used as previously described (Sambrook et al., 1989). The S. mutans strains lacking the various genes selected for this study (smu143c smu144c, smu248, smu540, smu570, smu929, smu1296 and smu1497) were constructed using a PCR ligation mutagenesis approach (Lau et al., 2002). Briefly, PCR fragments flanking each target gene were obtained and ligated to non-polar kanamycin (NP-Km R ), polar kanamycin (O-Km R ), or polar erythromycin (Em R ) cassettes, and the ligation mix was used to transform S. mutans UA159. Double mutants were obtained by amplifying the mutated spxA1 region in the previously constructed ΔspxA1 strain (spectinomycin-resistant, Spc R ) and using this PCR product to transform the different single mutants. Mutant strains were isolated on BHI plates supplemented with the appropriate antibiotic. The deletions were confirmed as correct by PCR sequencing of the insertion site and flanking region. Primers used to generate mutants are listed in S1 Table. Real-time quantitative RT-PCR Total mRNA was obtained using our standard protocols [24]. Briefly, RNA was isolated from homogenized S. mutans cells by repeated hot acid-phenol:chloroform extractions, and the nucleic acid precipitated with 1 volume of ice-cold isopropanol and one-tenth volume of 3 M sodium acetate (pH 5) at 4°C overnight. RNA pellets were resuspended in 80 μl nuclease-free H 2 O, and treated with DNase I (Ambion) at 37°C for 30 minutes. The RNA was purified again using the RNeasy mini kit (Qiagen), including a second on-column DNase treatment that was performed as recommended by the supplier. RNA concentrations were determined in triplicate using a Nanovue Spectrophotometer (GE Healthcare) and run on an agarose gel to verify RNA integrity. For qRT-PCR, cDNA templates were created from 0.4 μg of RNA using the Super-Script first-strand synthesis system (Invitrogen). Reactions were carried out on a StepOnePlus real-time PCR system (Life Technologies) according to protocols described elsewhere [24]. Gene-specific primers (S2 Table) were designed using Beacon Designer 2.0 (Premier Biosoft International) to amplify region of each gene 85 to 200 bp in length.

Protein Purification
The S. mutans SpxA1 and SpxA2 were expressed in E. coli as recombinant His-tagged fusion proteins using the pET-16B expression vector (EMD Millipore) as described elsewhere (Kajfasz et al, submitted). Purification of the recombinant proteins was performed by column chromatography using Ni-NTA resin (Qiagen) following the manufacturer's instructions. A crude extract of S. mutans RNAP was obtained based upon the methods described by Seepersaud et al for the extraction of RNAP from Streptococcus agalactiae, which utilizes the heparin binding properties to enrich for RNAP [25]. Briefly, S. mutans cells were grown in BHI broth to OD 600 0.5. Cells were harvested, resuspended in protoplast preparation buffer (0.3 M potassium phosphate buffer, pH 7.0; 40% sucrose; 0.5 U μl -1 mutanolysin) and incubated at 37°C for 90 min. Protoplasts were harvested, and the pellets resuspended in lysis buffer (50 mM Tris HCl, 10 mM MgCl 2 , 0.1 M DTT, 0.1 mM EDTA, 10% glycerol, 1 mM PMSF, pH 8.0), followed by 3 rounds of 15 seconds of sonication, intersperced by 1 minute intervals on ice. Following centrifugation, supernatants were applied to Affi-Gel heparin resin (Bio-Rad) and eluted with a gradient of a NaCl gradient of 0.1 to 1 M. The elutions were run on 10% SDS-PAGE and the fractions containing a visible β subunit (134 kDa) were pooled, and the salt concentration adjusted to 0.1 M NaCl by buffer exchange. The protein was then applied to Macro-Prep High-Q ion exchange resin (Bio-Rad) and eluted with a gradient of 0.1-0.8 M NaCl. The desired fractions were again pooled and dialyzed (10 mM Tris HCl, 10 mM MgCl 2 , 100 mM NaCl, 0.1 mM EDTA, 50% glycerol, pH 8.0). Protein concentration was determined using the bicinchoninic acid assay (Pierce).

In vitro transcription assay
A linear DNA template for each gene of interest was generated by PCR using primers designed to amplify from the promoter region to 70 to 200 base pairs of the coding region (S3 Table). Following amplification, the DNA fragments were purified with a QIAQuick PCR Purification Kit (Qiagen). The in-vitro transcription (IVT) reactions were performed as described elsewhere [26]. Briefly, 10 nM of each individual purified promoter template, 25 nM S. mutans RNAP, and 25 nM B. subtilis σ A (a gift from Peter Zuber, Oregon Health Science Center) were incubated with or without purified 75 nM ofSpxA1 or SpxA2 in reaction buffer (10 mM Tris-HCl, 50 mM NaCl, 5 mM MgCl 2 ; pH 8.0) containing bovine serum albumin (BSA, 50 μg ml -1 ) to a final volume of 20 μl for 10 min at 37°C. A nucleotide mixture (200 mM ATP, GTP, and CTP, 10 mM UTP, and 5 μCi [α-32 P] UTP) was added and the incubation proceeded for an additional 3 to 5 minutes. Stop solution (1 M ammonium acetate, 0.1 mg ml -1 yeast RNA, 0.03 M EDTA) was added and the mixture was precipitated with ethanol at 4°C for 3 hours. The nucleotide pellet was resuspended in formamide dye (0.3% xylene cyanol, 0.3% bromophenol blue, 12 mM EDTA dissolved in formamide). The samples were heated at 90°C for 2 min, then placed on ice before applying to an 8% polyacrylamide urea gel. The transcripts were visualized with Bio-Rad Quantity One Fx software following overnight exposure to an Imaging Screen-K (Bio-Rad).

Competition assay on solid media
The ability of Streptococcus gordonii to inhibit growth of S. mutans via H 2 O 2 production was assessed as described previously by Kreth et. al. [5]. Briefly, 8 μl of an overnight culture of S. gordonii DL-1 was spotted on the center of a BHI agar plate incubated at 37°C for 16 h. The following day, 8 μl of S. mutans cultures were spotted near the S. gordonii spot and incubated for an additional 16 h before visualizing the ability of the different S. mutans strains to grow in proximity of S. gordonii. To ascertain that any growth inhibition was due to the production of H 2 O 2 by S. gordonii, a control condition was included in which 8 μl of catalase (0.75 μg μl -1 ) was immediately spotted on top of the S. gordonii culture.

Rat colonization
The ability of the mutant strains to colonize the teeth of rats was evaluated using an established model of dental caries [27]. Pathogen-free Wistar rat pups 19 days of age were purchased from CEMIB/ University of Campinas (credited by ICLAS-International Council for Laboratory Animal Science). Rats were first screened to ensure a lack of indigenous mutans streptococci as previously described [27], and infected separately for three consecutive days by means of cotton swab with actively growing S. mutans strains. Animals were fed a cariogenic diet (Diet-2000 containing 56% sucrose) and 5% sucrose water (wt/vol) ad libitum to enhance the infection by S. mutans. The experiment proceeded for 15 days, at the end of which the animals were killed by CO 2 asphyxiation, and the lower jaws removed for microbiological assessment. The number of mutans streptococci recovered from the animals was expressed as CFU ml -1 of jaw sonicate. This study was reviewed and approved by the University of Campinas Committee on Animal Resources (CEUA Protocol #2639-1).

Biofilm assay
Biofilm formation of the constructed S. mutans mutants was assessed by growing cells in wells of polystyrene microtiter plates using a semi-defined biofilm medium (BM) [28]. Strains grown in BHI medium to OD 600~0 .5 were diluted 1:100 in BM containing 1% sucrose and added to the wells of the microtiter plate. Plates were incubated at 37°C in a 5% CO 2 atmosphere for 24 h. After incubation, plates were washed twice with water to remove planktonic and loosely bound bacteria, and adherent cells were stained with 0.1% crystal violet for 15 min. The bound dye was extracted with 33% acetic acid solution, and biofilm formation was then quantified by measuring the optical density at 575 nm. To exclude the possibility that differences in bound cells obesrved were due to variations in growth rates and growth yields, the biofilm data was normalized by the final OD 600 growth.

Statistical analysis
Student's t tests were performed to assess the distributions of the real-time PCR quantifications. One-way analysis of variance (ANOVA) was performed to verify the significance of the growth curves, growth in FeSO 4 , streptonigrin disc inhibition and biofilm quantifications. The animal study data were subjected to ANOVA in the Tukey-Kramer Honest Standard Deviation (HSD) test for all pairs. In all cases, P-values 0.05 were considered significant.

Transcription of several Spx-regulated genes is induced by H 2 O 2 stress
Nine genes selected for this study were previously identified in a microarray study as being positively regulated by the S. mutans SpxA1 protein [19]. In addition to the genes positively regulated by SpxA1, one gene under SpxA1 negative control, smu570 predicted to encode the second gene (feoB) of the FeoABC iron transporter system [29], was also analyzed. The selected genes with their respective assigned functions and expression ratios from our previous microarray analysis are shown in Table 2. Recently, we showed that treatment with 5 mM H 2 O 2 strongly induced the expression of known oxidative stress genes (e.g. ahpC, gor, nox, sodA, tpx and trxB), and that this regulation was strongly dependent on SpxA1 (Kajfasz et al., submitted). Here, we used quantitative real-time PCR (qPCR) to evaluate the transcriptional responses of the ten selected SpxA1-regulated genes (smu127, smu143c, smu144c, smu248, smu540, smu570, mu929c, smu1296, smu1297 and smu1645 to H 2 O 2 stress (Fig 1). A heat map depicting the transcript fold change in relation to the parent UA159 under control growth conditions (T0) is also shown (S1 Fig). In the parent strain, transcription of smu127, smu143c, smu144c, smu248 (sufD), smu540 (dpr), smu929, smu1296 and smu1297 was significantly induced by H 2 O 2 at both 5 and 15 min time points. Transcription of smu1645 was also moderately induced by H 2 O 2 but the differences were not statistically significant. Nevertheless, transcriptional levels of smu1645 were significantly lower in all Δspx stains confirming that this gene is also under Spx control. Among the H 2 O 2 -induced genes, smu1296 and smu1297 respectively, showed the most robust responses ranging from 10 to 100-fold induction after stress. Transcription of all H 2 O 2 -induced genes was strongly dependent on SpxA1 and, with the exception of smu127, SpxA2. In agreement with our previous microarrays, expression of smu570 (feoB) was induced in the absence of SpxA1 but slightly repressed in ΔspxA2. Also within the expected, feoB transcription was repressed by H 2 O 2 in the two SpxA1-positive strains (UA159 and ΔspxA2). Altogether, these results reveal that transcription of the genes investigated is controlled by both SpxA1 and SpxA2, induced by H 2 O 2 in an SpxA1-dependent manner and, therefore, likely to participate in oxidative stress responses.
IVT assays confirm direct regulation of smu144c, dpr and smu929c by Spx To confirm whether the regulatory effects exerted by the two Spx proteins were direct or indirect, we performed IVT reactions in the presence or absence of purified SpxA1 or SpxA2 proteins. Our results demonstrate that the addition of purified SpxA1 to the reactions enhanced transcription of smu144c/143c, dpr and smu929c (Fig 2). Addition of SpxA2 also enhanced dpr transcription. However, SpxA2, at equal or greater concentrations than SpxA1, had no apparent effects on smu144c/143c or smu929c transcript levels. In the case of the sufA operon, a transcript was seen in the absence of Spx protein, but the addition of SpxA1 or SpxA2 did not enhance transcription. Despite multiple attempts, we were unable to detect transcripts upstream smu127, feoA, smu1296 and smu1645 (data not shown). Given that transcripts for these genes have been detected by microarrays and qPCR, it is possible that these transcripts are not abundant and/or short-lived for detection via IVT assay. Another possible explanation is that co-factors or other regulatory proteins, not included in the in vitro system, may be required for efficient transcription of these genes. Growth characteristics of mutant strains Upon confirmation that the selected genes were transcriptionally regulated by Spx and, in most cases, induced by H 2 O 2 stress, we sought to determine the effect that deletion of each individual gene would have upon S. mutans ability growth under oxidative stress. The smu1297 gene, encoding an enzyme with 3'-phosphoadenosine-5'-phosphate (pAp) activity, was previously identified in a transposon mutagenesis library to be involved in the superoxide stress response [30]. Because Smu1297 already has an assigned role in oxidative stress, the smu1297 gene was not further investigated. Deletion mutations of eight of the genes described above were readily obtained but a viable smu127 mutant strain was not. The smu127 gene is predicted to encode the E1 component of the subunit α of the acetoin dehydrogenase (Adh) complex. The genetic organization of the eight mutated genes is depicted in (S2 Fig). The smu143c and smu144c genes, encoding, respectively, polypeptide deformylase and a putative transcriptional regulator, are separated by only 25 bp and co-transcribed (S3 Fig). Likewise, smu1296 encoding a glutathione S-transferase (GST) homologue was shown to be co-transcribed with smu1297 [30]. To avoid downstream polar effects, smu144c and smu1296 were inactivated with a non-polar cassette that contains a promoterless aphA3 Km R gene without transcription termination sequences to allow transcription readthrough into downstream sequences [31]. Real-time PCR quantifications confirmed expression of the downstream smu143c and smu1297 genes in the Δsmu144 and Δsmu1296 strains, respectively (S4 Fig). The Em R marker was used to replace the smu143c and smu929c genes; smu929c is the last gene of a bicistronic operon (smu930c-smu929c). The smu569-571 and smu247-251 genes are arranged in operons with other genes involved in iron transport (FeoABC) and iron-sulfur (Fe-S) cluster assembly (SUF), respectively. To disrupt the FeoABC and Suf systems, the feoA/feoB genes (smu569-570) was replaced by the nonpolar Km R cassette (Δfeo) whereas the first two genes in the suf operon (sufC/sufD) were individually inactivated using a polar Km R cassette such that transcription of the downstream genes was interrupted (Δsuf). The smu1645 gene, encoding a putative tellurite resistance protein, is transcribed in the opposite orientation of its surrounding genes and was also inactivated using the polar Km R marker. Finally, we used the Em R marker to inactivate the smu540 gene, encoding the iron-binding peroxidase resistance Dpr protein. The importance of Dpr to S. mutans oxidative stress tolerance has been relatively well documented [11,32,33], and the Δdpr strain was used as a benchmark in downstream studies. The growth characteristics of the mutant strains under different conditions are shown in Fig 3. Under standard non-stressful conditions, e.g. brain heart infusion (BHI) broth at 37°C in a 5% CO 2 atmosphere, all strains, with the exception of the Δsuf strain with a doubling time of 2.56 ± 0.23 h, grew as well as the parent UA159 strain (doubling time 1.32 ± 0.07 h). Next, we evaluated the ability of the mutant strains to grow in a 5% CO 2 atmosphere in the presence of H 2 O 2 , methyl viologen (MV, a quaternary ammonium compound that generates superoxide radicals) and diamide (a specific oxidant of thiols). In the presence of 0.2 mM H 2 O 2 , the Δsuf strain showed an extended lag phase and slow growth rates whereas the Δdpr strain was completely unable to grow under this condition. In the presence of diamide, growth rates and growth yields of the Δsuf were severely affected (doubling time 3.54 ± 0.41 h compared to 1.48 ± 0.08 for the parent strain). The Δsuf strain also displayed long lag and slow growth rates in MV. Surprisingly, the Δdpr strain that failed to grow in H 2 O 2 was able to grow in the presence of diamide or MV, albeit with extended lag phases. Growth kinetics of the remaining strains, Δsmu143c, Δsmu144c, Δfeo, Δsmu929, Δsmu1296 and Δsmu1645, in the presence of the different oxidative stress agents was nearly identical to that of the parent strain.
We also tested the ability of our panel of mutants to grow in the presence of the peroxigenic commensal Streptococcus gordonii. In agreement with its inability to grow in the presence of H 2 O 2 (Fig 3), growth the Δdpr strain was inhibited by S. gordonii (Fig 4). Despite an extended lag phase in media containing H 2 O 2 , growth of the Δsuf strain was not inhibited by S. gordonii. Also in agreement with growth curve results, the remaining mutant strains were not inhibited by S. gordonii, at least under the conditions tested (Fig 4 or data not shown).

Iron sensitivity of the mutant strains
Functional annotation of genes selected for this study suggest that some of the genes may be involved in metal ion homeostasis, which includes genes encoding a polypeptide deformylase metalloprotein (smu143c), an Fe-S cluster assembly system (smu247-smu251), iron-binding protein (dpr), iron transport system (feoABC) and tellurite resistance protein (smu1645). Metal ions, in particular iron, are essential for life functions, but cytoplasmic levels must be tightly controlled as iron can serve as a catalyst for the production of damaging hydroxyl radicals in the presence of H 2 O 2 via Fenton reaction. Next, we tested the ability of our panel of mutant strains to cope with iron by assessing their ability to grow in the presence of 5 mM FeSO 4 or streptonigrin, an iron-activated antibiotic [34] (Fig 5). Strains Δsmu143, Δsuf and Δdpr displayed a growth impairment in the presence of FeSO 4 , whereas Δsmu144, Δsuf, Δdpr, Δsmu1296 and Δsmu1645 showed increased sensitivity to streptonigrin based on disc diffusion assays. In agreement with its assigned role in iron transport, the Δfeo strain showed increased tolerance to streptonigrin but not to FeSO 4 . Because smu1645 encodes a putative tellurite resistance protein, we also compared the minimum inhibitory concentration (MIC) of tellurite between the parent UA159 and Δsmu1645 strains. In accordance with its predicted function, the Δsmu1645 strain had a lower MIC for tellurite than UA159 (3.9 μM versus 15.6 μM for UA159, data not shown).  The Δsuf and Δdpr strains showed an impaired ability to colonize the teeth of rats Next, we tested the ability of the Δsmu143, Δsuf, Δdpr, Δfeo, Δsmu1296 and Δsmu1645 strains to colonize the teeth of Wistar rats (Fig 6). Following a 15-day infection period, colonies of S. mutans were recovered from the jaws of infected animals by plating the jaw sonicates on MS agar. In agreement with the growth defect and general stress sensitivity, the infectivity of the Δdpr and Δsuf strains was significantly reduced (P < 0.05). The differences observed between the remaining mutant strains were not statistically significant. Because the infectivity of S. mutans in the oral cavity has been shown to be directly tied to its ability to form biofilms on tooth surfaces through a sucrose-dependent mechanism [35], we also evaluated the capacity of each strain to form biofilms on microtiter plates in the presence of sucrose. The Δsmu143, Δdpr and Δsuf and ΔfeoB strains accumulated more biofilm (P < 0.05) when grown in sucrose when compared to UA159 (Fig 7). Therefore, the lower infectivity of the Δdpr and Δsuf strains in the teeth of rats does not appear to correlate with a deficiency in biofilm formation.
A phenotype enhancement screen confirmed the physiologic relevance of the Fe-S cluster assembly system but failed to assign new roles to the other Spx-regulated genes Due to functional redundancy and genetic buffering, it is not unusual that single gene deletions do not have any detectable phenotype. This is readily observed in systematic gene deletion libraries [36,37] whereby gene functions are rarely assigned based on single deletions. A useful way to overcome this limitation is the construction of paired mutations that may eliminate or reduce genomic buffering effects [38,39]. Recently, the Zuber lab utilized such an approach to uncover the role of Spx-regulated genes of unknown function in B. subtilis [40]. Specifically, double mutant strains were created by introducing mutations on selected Spx-regulated genes into the spx mutant background followed by a screen for enhanced oxidative stress sensitivity when compared to the spx single mutant. Here, we utilized a similar approach by inactivating our genes of interest, one at a time, in the ΔspxA1 strain. We have previously shown that the ΔspxA1 strain grew poorly under oxidative stress conditions [19], thus we evaluated the ability of ΔspxA1 single and double mutants to grow under these conditions. While we were able to isolate a ΔspxA1Δsuf double deletion strain, this strain grew very poorly under standard laboratory conditions and was not amenable to further phenotypic characterization. In most cases, the double mutants did not display slower growth rates or hypersensitivity against the oxidative stress agents tested (H 2 O 2 , MV or diamide) when compared to the ΔspxA1 single mutant (Fig 8). The only exceptions were the ΔspxA1Δsmu143c and ΔspxA1Δsmu929c strains that, unexpectedly, grew better than ΔspxA1 in the presence of diamide and the ΔspxA1Δsmu144c strain that showed slower growth than ΔspxA1 in the presence of MV. We tested whether the increased diamide resistance of the ΔspxA1Δsmu143c and ΔspxA1Δsmu929c strains was a result of elevated spxA2 expression but in, both case, the transcriptional levels of spxA2 were identical to the parent UA159 strain (S5 Fig). At this time, the reasons for mutations on either the smu143 or the smu929 genes serving to alleviate the diamide sensitivity of the ΔspxA1 strain are unknown. One possible explanation is that loss of Smu143 or Smu929 triggered some yet to be determined oxidative stress mechanism that compensates for the loss of SpxA1.

Discussion
While the significance of acid production, acid tolerance and biofilm formation to the cariogenic potential of S. mutans has been well established and have been studied in great detail [2], the role of bacterial oxygen metabolism and oxidative stress survival in caries have thus far received limited attention. Over the years, the primary notion that the dental plaque environment was virtually anaerobic has been replaced by evidence that, as a whole, the microbial plaque community has a high capacity to reduce oxygen thereby generating a variety of toxic byproducts. The relevance of oxidative stress survival to the pathophysiology of S. mutans has been further supported by in vitro, in vivo and clinical studies, showing an inverse relationship between the presence of H 2 O 2 -generating streptococci from the mitis group and S. mutans [5,[41][42][43].
Spx proteins are conserved global transcriptional regulators of Gram-positive bacteria that function as positive regulators of oxidative stress genes [15] that, more recently, have been linked to the virulence potential of E. faecalis, S. mutans and other streptococcal species [19,22,23,44]. Previously, using microarrays, we showed that most genes with a proven role in the oxidative stress response of S. mutans were under SpxA1 positive regulation, and, to a lesser extent, SpxA2 [19]. Specifically, expression of genes involved in oxygen metabolism (nox), thiol stress (trxA and trxB) and ROS detoxification (ahpC, ahpF, sodA and tpx) among others was significantly lower in strains lacking spxA1, spxA2 or both. Here, we sought to uncover the function of previously uncharacterized genes that, in the microarrays, displayed the same regulatory profile of the known oxidative stress genes, namely smu127, smu143, smu144, smu247, smu929c, smu1296, smu1297, and smu1645. In addition, we included one previously characterized oxidative stress gene, smu540 (dpr), to serve as a phenotypic benchmark as well as to gain additional knowledge about the function of Dpr in S. mutans. An exception to the trend of positive regulation by Spx was the inclusion of the smu569/570/571 transcriptional unit encoding the FeoABC iron transport system, which was by our microarrays to be negatively regulated by SpxA1. Given the well-defined role of iron in hydroxyl radical formation as a catalyst of the Fenton reaction, it is therefore expected that negative regulation of iron uptake by Spx is an integral part of the oxidative stress response.
The transcriptional data obtained from our microarrays provided insight into the role of Spx proteins as regulators under basal conditions, when the cells were not subjected to oxidative stress. Here, quantitative real time PCR analysis of cells exposed to H 2 O 2 further supported that the genes studied herein are regulated by Spx and encode bona fide oxidative stress proteins. Exposure to H 2 O 2 resulted in increased expression of each of these genes (with the exception of, as expected, feoB), demonstrating a role for each of them in defending the cell against oxidative stresses. However, in the absence of SpxA1, exposure to oxidative stress did not trigger increased expression of these genes, demonstrating their regulatory dependence upon SpxA1. With the exception of smu143c (polypeptide deformylase), smu144c, (transcriptional regulator) and smu929c, which encodes a hypothetical membrane protein conserved among Gram-positive cocci, all other genes are predicted to encode proteins involved in oxidative stress responses in other bacteria. Among the latter group of genes, the aforementioned dpr and feoB are predicted to participate in oxidative stress defenses by binding to iron (Dpr) or by limiting iron uptake (FeoB). The FeoABC system can be found in both Gram-positive and Gram-negative bacteria and has been implicated in virulence of several Gram-negative species as well as Streptococcus suis [45,46]. With the exception of the increased resistance to streptonigrin, an antibiotic whose toxicity depends on the availability of intracellular free iron, the Δfeo strain phenocopied the parent strain including a nearly identical ability to colonize the teeth of rats fed a cariogenic diet. This may be explained, at least in part, by iron being non-specifically transported by other metal-binding ABC transporters. The iron-binding Dpr has been extensively studied in bacteria and shown to mitigate H 2 O 2 lethality by preventing the Fenton reaction [11,33]. In agreement with previous reports [11,14,47], the Δdpr strain was hypersensitive to H 2 O 2 and, based on indirect evidence, including growth in the presence of iron and tolerance to streptonigrin, cannot maintain iron homeostasis. Interestingly, growth of the Δdpr strain was not affected by diamide, and only modestly inhibited by MV. However, the different behavior of the Δdpr mutant in the presence of H 2 O 2 , diamide or MV is on par with a proteomic study in S. aureus that revealed a small overlap in the kinds of proteins produced when cells are exposed to these oxidants [48]. Diamide specifically oxidizes thiol groups, causing an increase in formation of intra-or extramolecular disulfide bonds leading to the accumulation of misfolded and aggregated proteins as a result of thiol oxidation. In this case, hydroxyl radical formation due to Fenton reaction should play an indirect role in cell viability. MV, also known as paraquat, generates O 2 radicals in the cell that can directly mobilize Fe-S clusters or can be converted to H 2 O 2 . It is conceivable that the Δdpr strain can partially overcome the detrimental effects of MV by protecting Fe-S cluster enzymes, or by activating multidrug efflux systems. In fact, an ABC transporter, dubbed VltA/VltB for viologen transporter, was shown to mediate resistance to MV as well as other quaternary ammonium compounds in S. mutans [49]. Most importantly, the Δdpr mutant showed reduced ability to colonize the teeth of rats thereby providing, for the first time, direct evidence that Dpr is involved in the virulence of S. mutans. The most dramatic phenotypes were observed in the Δsuf strain, in which the entire suf operon was inactivated with a polar kanamycin cassette. The Δsuf strain displayed impaired growth under all conditions tested, increased sensitivity to iron and lower infectivity in rats. The SUF machinery is one of three Fe-S cluster assembly systems identified in bacteria but can also be found in plant chloroplasts [50]. Proteins containing Fe-S clusters as a prosthetic group are widely distributed in nature and perform essential biological processes including electron transfer, substrate binding/activation, transcriptional and translational regulation and iron storage [51]. ROS cause destabilization of Fe-S clusters affecting important cellular processes and stimulating the Fenton reaction due to an increase in intracellular free iron. Two Fe-S assembly systems are present in E. coli: the housekeeping ISC and the stress-inducible SUF, whereas only the SUF machinery is present in S. mutans and other members of the Firmicutes phylum [52]. Considering that the SUF system is the only pathway for assembly and repair of Fe-S clusters in S. mutans, it was not entirely unexpected that fitness and viability of the Δsuf strain was significantly impaired. Interestingly, while transcriptionally induced by H 2 O 2 stress and dependent on the SpxA1 and SpxA2 regulators for optimal expression (see Fig 1), our IVT assays indicate that neither SpxA1 nor SpxA2 directly induce transcription of the suf operon. The most logical interpretation of these results is that Spx exerts indirect control over suf gene expression perhaps by controlling the transcription of another regulator. Of note, the suf operon of E. coli is induced by oxidative stress and is under the control of three transcriptional regulators, IscR, Fur and OxyR [53,54].
Although strains lacking either smu144c or smu143c were able to grow as well as the parental strain under oxidative stress conditions, it is important to note that the smu144c/143c genes, which are likely co-transcribed, were: (i) induced by H 2 O 2 treatment, (ii) directly regulated by SpxA1, and apparently (iii) unable to maintain iron homeostasis. The smu144c genes encode a putative transcriptional regulator of the Crp/Fnr family conserved in many streptococcal species. Members of the Crp/Fnr family utilize Fe-S clusters to sense and respond to oxygen. The smu43c codes for a polypeptide deformylase (PDF) metalloenzyme, which removes the formyl group from the N-terminal methionine of newly synthesized proteins. The crystal structure of Smu143c has been solved (pdb3L87) and shown to coordinate binding to ferrous ion. Given that the PDF process is essential in bacteria but not in eukaryotes, PDF inhibitors are considered promising targets for the development of new antimicrobials [55].
In conclusion, we have shown that the Spx proteins of S. mutans control the expression of several additional genes involved in oxidative stress management. It appears that in addition to the transcriptional activation of oxygen metabolism, ROS scavengers and thiol repair enzymes, Spx performs an important role in iron homeostasis by regulating the intracellular availability of free iron. As part of a continued effort to assign a functional role for the genes characterized in this study, an expanded phenotype enhancement screen based on the double spx strain (ΔspxA1ΔspxA2) is currently under way.
Supporting Information S1