Comprehensive Transcriptome Profiles of Streptococcus mutans UA159 Map Core Streptococcal Competence Genes

Published transcriptome surveys provide valuable views into the breadth of streptococcal competence-specific genes; however, for most species, including the competence model S. pneumoniae, transcriptome data for competence development have come mostly from early techniques of microarray analysis that have technical limitations, such as the use of probes that are often restricted to annotated open reading frames (ORFs), the use of probes with low spatial resolution, and the use of cDNA preparations that result in artifactual antisense signals. In S. mutans, newer high-density tiling arrays and RNA-sequencing methods have already been used in transcriptome surveys of the competence response, but none of them was designed specifically to distinguish the different regulons that are activated in response to CSP. The tiling array study was restricted to responses resulting from long exposure to CSP and lacked full coverage, including probes for the comS gene (28), whereas in the RNA-sequencing study, short exposure to CSP was investigated in a medium that does not support activation of competence by CSP (27). Therefore, to capture a more complete picture of competence regulation in S. mutans, we chose a strategy of probe design and sample collection and preparation that would allow comprehensive mapping of transcripts in both the early and late phases of the CSP response under conditions in which CSP induces competence.

We employed six strategies to minimize confounding errors known to affect transcriptional profiling. (i) To improve the specificity of hybridization signals, we employed 385,000 overlapping 50-mer oligonucleotide tiling probes optimized for uniqueness, T_m, and probe length, with a resolution of approximately 10 bp, as described previously (29). (ii) To minimize the artifactual “antisense” signals that arise during cDNA preparation, RNA was instead directly chemically labeled (30). (iii) To maximize mRNA signals, the RNA was depleted of rRNA. (iv) To minimize loss of sRNAs and possible short transcripts coding for small peptides such as ComS and ComC, microRNA purification protocols were employed. (v) To minimize signals from irrelevant metabolic changes that might occur during the experiment, we limited cultures to the early log phase. (vi) Finally, we chose to use the mature CSP18 pheromone (31), instead of the precursor CSP-21 peptide used in previous transcriptome surveys (15, 28, 32), in order to minimize any response delay due to peptide processing steps (31, 33, 34).

RNA preparations were made from cultures of the wild type (WT) in the early and late stages of the response to CSP and from cultures of a comS mutant in the late stage. The most appropriate times to distinguish the early and late global responses during competence development were selected on the basis of measurements of growth, DNA incorporation dynamics, and expression of selected early and late genes after CSP supplementation of cultures in tryptone soya broth (TSB) (Fig. 2). TSB was chosen because this medium supports competence development stimulated by synthetic CSP, but endogenous competence development is absent or restricted to low levels and is independent of comC (35). During the first 120 min of incubation, cultures with and without CSP grew at equal rates and remained far from stationary phase. The culture treated with CSP became capable of rapid DNA uptake (Fig. 2A), while competence remained very low in the parallel culture without added CSP throughout 4 h. Transformation was first detected at 100 min after CSP addition, reached maximal levels at 3 h, and declined by 4 h. Using a gene fusion reporter, we found that cipB expression, used as an indicator of early gene expression, was strongly dependent on CSP and increased as an immediate response, replicating previous findings (33, 34). In contrast, expression of sigX (also CSP dependent) began only after a delay of approximately 50 min (Fig. 2B). Thus, (i) the strong dependence of both the early expression of bacteriocin genes and the late development of competence on CSP in TSB and (ii) the time difference between the two responses established optimal conditions for study of the temporally specific effects of CSP on the transcriptome. Because endogenous competence induction was absent within the first 3 h and CSP-induced DNA incorporation was robust by approximately 2 h, at a time when growth had not been inhibited by CSP, we further evaluated the use of CSP exposure times of 10 and 100 min. Reverse transcription (RT)-PCR assays confirmed that the early and late transcriptional responses were activated at 10 and 100 min, respectively (Fig. 2C to E). Expression of the ComE-regulated gene SMU.1914 (cipB, nlmC) had already increased dramatically at 10 min compared with that in the culture without CSP, but there was only a slight increase in the expression of the SigX-dependent genes comGA and comEC. By 100 min, comGA and comEC expression had increased by more than 200-fold, while expression of SMU.1914 continued at an elevated level. Thus, 10 and 100 min were selected as the earliest suitable sampling times for studying the immediate and delayed transcriptional responses to CSP in the absence of any gross effect of CSP on the growth rate.

RNA was extracted from cultures under the six conditions selected (UA159 for 10 or 100 min and the comS mutant for 100 min, with and without CSP) in duplicate experiments and analyzed for strand-specific genome-wide gene expression as described in Materials and Methods. Inspection of the resulting profiles gives several indications that the sampling strategy yielded expression patterns of significantly improved quality. Using directly labeled RNA to exclude the artifactual antisense signals often observed with conventional cDNA preparations offered clear benefits, as it revealed that several of the genes previously classified as upregulated in response to CSP were indeed upregulated, but only in the antisense direction. We also observed mRNAs of short length in our preparations, including that for the CSP-encoding gene comC (approximately 190 bp), indicating successful isolation of short transcripts. Moreover, by carefully matching the CSP-treated and untreated samples in relation to the incubation time and culture density, we could clearly distinguish changes specific to CSP exposure from nonspecific changes occurring during growth. We found that in the control samples without CSP, the expression of 98 genes differed between cultures grown for 10 min and those grown for 100 min (see Table S2 in the supplemental material). None of the genes required for competence, such as those for DNA uptake and recombination, were identified in this group, indicating that the changes were not competence related. To determine whether this information would contribute to a better definition of the overall response to the CSP pheromone, we investigated whether genes previously defined as CSP induced include the group of genes identified here as nonspecific to the CSP response. In fact, 26 of the genes showing changes associated with growth were among genes previously defined as part of the CSP response (see Table S2) (15).

Among the 1,961 protein-encoding genes annotated in the UA159 genome sequence, a subset of 83 were represented by a >2-fold expression increase during a 100-min CSP exposure (Table 1; see Table S3 in the supplemental material). Only five genes exhibited downregulation, but all in the low range between −2.1- and −2.3-fold changes. The CSP induction ratios were, in general, similar to those previously observed in a transcriptome survey examining a competent subfraction of the S. mutans population (28) and higher than in surveys using mixed populations exposed to CSP for 120 min (see Fig. S1 in the supplemental material) (15, 28). Approximately 160 genes that were reported as differentially regulated in response to CSP in one of the mixed-population studies (15) were not confirmed in either this study or in the study using sorted competent cells (see Fig. S1B and Table S4 in the supplemental material). We conclude that such genes are unlikely to be reproducible parts of the late CSP response and may represent experimental error, indirect effects of competence, or metabolic changes unrelated to the response to CSP. This interpretation is strengthened by the observation that the induction levels of most genes were uniformly higher in the present data and in the previous study using sorted cells and by the fact that the former transcriptome did not provide information on direction of transcripts and had a limited set of probes for each of the ORFs. Finally, the induced genes encoding bacteriocin and bacteriocin immunity proteins were generally induced at higher ratios here than in the previous studies using mixed or sorted populations of CSP-stimulated cells. This was particularly valuable in view of our aim to differentiate early and late responses associated with the activation of distinct regulons during the CSP response.

A comparison of the expression levels of individual upregulated genes in samples treated with CSP for 10 and 100 min is shown in Table 1. The induced genes form two classes with different temporal patterns of expression; one class was induced at both times, and a second, larger, class was upregulated at 100 min but was not perceptibly affected at 10 min. For convenience in discussion, we designate the former early genes and the latter late genes. As they are likely to have different modes of regulation, transcriptionally active regions (TARs) in the two classes are described separately below.

All five genes previously reported individually to be upregulated in the early phase of the CSP response (nlmAB, nlmD, immB, and cipB), either by means of reporter constructions or by RT-PCR (18, 28, 36), were among the early upregulated genes (Fig. 3). The genes in the early induction class clustered in four chromosomal regions. One locus contains the genes for bacteriocin NlmAB and the cognate bacteriocin immunity protein SMU.152 (37), with transcription continuing past a predicted terminator and through SMU_153, encoding a hypothetical protein. The three intergenic sequences among these genes were also upregulated, suggesting that the four genes are part of a single transcript. Two TARs, encoding the bacteriocin NlmD (BsmC, SMU.423) and the ImmB immunity protein (SMU.925), appear to be essentially monocistronic, with readthrough past terminator elements into a total of six downstream neighboring genes (Fig. 3). For nlmD, the induction of the downstream genes was already evident at 10 min, whereas for immB, the induction of downstream ORFs SMU.926 to SMU.928 was seen only at 100 min. Although they were late genes, they form a single TAR initiated at the early gene immB, thus appearing to be part of the same regulon as immB and other early genes, as discussed below. The fourth early upregulated region comprises a 13-kb island, including genes for four bacteriocins (BsmB, -L, and -K and CipB), two immunity proteins (ImmA and SMU.1909), five proteins of unknown function, and an unusually large proportion of apparent intergenic regions. Analysis of expression changes in the antisense direction reveals that a marginally upregulated region extends through SMU.1902 to SMU.1897, apparently as a result of transcriptional readthrough.

Comparative analysis of the RNA harvested from CSP-treated cells at 100 min to that isolated from parallel untreated controls reveals that the four TARs of the early CSP response comprising 20 genes continued to be upregulated at this time. The TAR initiated at immB was extended to include three other genes (SMU.926 to SMU.928) at this time. In addition, 60 other genes organized in 22 TARs were induced in the sense direction and another three TARs were induced exclusively in the antisense direction by at least 2-fold (Table 1; Fig. 3). Among these, one TAR was initiated at the 3′ end of the SMU.60 gene, upstream of comR, and extended to the start of comR. Five others included regions of antisense transcription (initiated at comS, cilC, pilC, radC, and comE), and at six loci, TARs (initiated at SMU.325, SMU.351, SMU.504, lytF, comE, and ssbB) extended past terminators in the sense direction, as predicted by DOOR (38), to include downstream ORFs. The three TARs induced only in the antisense direction of the ORFs start at SMU.691, SMU.1853, and rl16. Over half of the TARs comprise genes upregulated by 10- to 132-fold.

Three of the late upregulated genes encode known regulators of competence-specific transcription in S. mutans (ComS, SigX, and ComED). Consistent with a positive feedback loop mediated by XIP (Fig. 1), there was strong induction of the XIP-encoding gene comS at 100 min. The TAR initiated at comS is 8 kb long, comprising six other ORFs, one of them transcribed in the antisense direction (SMU.63). SigX, long known as part of the response to CSP in streptococci (11), is represented by a one-gene TAR upregulated by 16-fold. Finally, the comED genes are among the late genes and are part of a TAR that extends in the 3′ direction to include the transcription of comC in the antisense direction. No induction of comC in the sense direction was observed. Of the remaining 19 late-induced TARs initiated by genes transcribed in the sense direction, 12 include at least one gene known to be involved in competence; 4 encode proteins for DNA transport, 6 encode proteins for DNA recombination, 1 encodes a protein for DNA methylation, and 1 encodes a protein for autolysis (Fig. 3).

ComR and ComS form a positive feedback circuit that enhances the synthesis of ComS and concomitantly upregulates sigX (6). Although only two copies of the complex ComR box recognition motif have been detected in the S. mutans genome, the inference that ComR acts only at those two sites has not been experimentally tested in S. mutans. To test this inference and investigate the place of ComS in the CSP response more thoroughly, transcriptome analysis of a comS deletion mutant was performed in parallel with the analysis of WT UA159 described above (Table 1; see Tables S4 and S5 in the supplemental material). During CSP treatment of the comS mutant for 100 min, none of the late-phase genes identified in the WT were upregulated, but the profile of expression seen at 10 min in the WT was closely recapitulated (Table 1). This pattern supports the inference made from previous studies of several late genes that comS is an integral link between the early response and all elements of the late response.

The single gene that was upregulated in the comS culture but not in the 10-min WT sample was SMU.926. As shown in Fig. 3, SMU.926 most probably represents a readthrough from the early gene SMU.925 (immB) rather than a distinct late-activated promoter. Similarly, the single gene upregulated in the 10-min sample but not upregulated in the ΔcomS mutant culture is SMU.1902c (bsmK). This had the lowest degree of upregulation (2.1-fold) among the early genes and is located at the 3′ end of a long TAR extending from the early gene cipB, which was induced by 16-fold, again indicating a possible readthrough past a cryptic terminator site. In fact, the transcripts at all four early loci extended across predicted terminators and operon borders, indicating readthrough as a result of strong activation or incorrect terminator predictions.

For a comprehensive identification of the members of each of the three principal competence regulons, we combined information on the temporal responses to CSP and on the effect of comS deletion described above with a thorough analysis of the transcriptome map in the vicinity of conserved sequences recognized by ComE, ComR, and SigX (Table 1; Fig. 3 to 5). We then compared this information with the transcriptome data from five previous surveys investigating competence in S. mutans to search for conserved responses (Fig. 6). Two of the transcriptomes compared long exposure times (120 min) of strain UA159 to CSP with nontreated samples (15, 28); the third transcriptome was also for strain UA159, but the comparison was between the WT and a CSP response-defective cipB mutant also exposed to CSP for 120 min (32); the fourth one was for strain UA140 comparing the WT to an hdrR overexpression strain that shows induction of late competence genes in the absence of CSP supplementation (39); and the fifth one used short (up to 30 min) exposure of UA159 to CSP or XIP in a defined medium in which CSP responses are not linked to competence (27).

To delimit the ComE regulon, we first compared the transcriptome profiles of the early response of the WT to CSP and of the comS mutant exposed for 100 min to CSP, as illustrated in Fig. 4A and B. The expression pattern at 10 min is expected to reveal the immediate response of genes transcribed via ComE activation, before regulons dependent on ComR are upregulated, whereas data from the late response of a comS mutant, in which downstream regulons remain silent because of a lack of the linking XIP peptide, would potentially reveal direct targets of ComE with a slower response, as well as other downstream regulons independent of the ComR-ComS link. We found near identity between the early response in the WT and the late response in the comS mutant. All four early-induced TARs initiated at nlmA, nlmD, immB, and cipB are preceded by the direct repeat (DR) identified previously as a putative ComE binding site required for CSP-dependent bacteriocin expression (18) and subsequently identified (40) as a tight binding site for purified ComE protein (Fig. 4 and 5). Other DR sites that bind purified ComE have been suggested to be functional, but they appeared to be nonfunctional under the conditions examined here, including the sites distal to bsmB (SMU.1906), comC (sense direction), and cslAB (Fig. 5A). The results indicate that all of the targets of ComE respond immediately to CSP and establish for the first time that ComE has no additional targets activated at late times independent of the ComR-ComS link.

We further examined the conservation of the CSP response by comparing the set of genes within the induced TARs with genes within or in the vicinity of these regions showing changed expression in previous studies (Fig. 6). Although the one published transcriptome study evaluating the early CSP response (27) was conducted under conditions of culture in CDM that do not support induction of competence by CSP, it did identify 19 of the 23 genes within the ComE regulon, as determined here (Fig. 5). Notably, similar to our results, in that study, comED and comC expression was only slightly or not induced by CSP, a finding also corroborated by a previous transcriptome study using a sigX mutant (15). Despite the methodological limitations of the latter study, including lack of information on the direction of the transcripts and low levels of induction by CSP, all of the 23 early genes induced in the sense direction in this study were identified among the induced genes reported there (Fig. 4C). We conclude that the UA159 genome contains only four TARs that are directly regulated by ComE, defining the ComE regulon. A total of 23 genes are transcribed in the sense direction, but direct experimentation is needed to assess the biological significance of distal genes or of the antisense transcripts in all four TARs, which are the source of the greatest variation among the different transcriptome studies.

Genes transcribed directly via activation of ComR in the CSP response can be identified from transcription differentials between a sigX mutant and a comS mutant in the late phase of the CSP response. Both data sets would reflect upregulation of genes of the ComE regulon, which is an early response, but only the sigX mutant would allow increased late expression of the transcripts of the ComR regulon. Although the information derived from a previous survey of gene expression in a sigX mutant (15) did not address transcript polarity or intergenic sequences, we can nonetheless use the ORF expression information available there to uncover candidate members of the ComR regulon. The scatterplot in Fig. 4C compares the genes that were upregulated in the 100-min comS mutant culture in this study to those previously reported as upregulated in a sigX mutant after a similar period of CSP exposure. The genes listed as upregulated in both transcriptomes represent the ComE regulon, as described above. The remaining seven genes, which were upregulated in the sigX mutant but not in the comS mutant or within a 10-min CSP exposure, are thus candidates for the ComR regulon. Of these, we exclude three (SMU.2037, SMU.2038, and SMU.799c) that, although induced to low levels in the sigX mutant, were not upregulated in the late response of the WT in either of the transcriptomes here (see Table S6 in the supplemental material) or in two other previous transcriptomes (27, 28) and are not preceded by the ComR box inverted repeat (IR). The remaining candidate genes, SMU.63, SMU.64, SMU.65, and SMU.66, are contiguous with comS, which was not itself represented in the microarray used for analysis of the sigX mutant. These four genes were also upregulated in the 100-min WT transcriptome, along with the downstream genes SMU.67 and SMU.68, forming a single TAR with comS. The present data reveal that SMU.63, immediately downstream of comS but in the inverse orientation, was transcribed only from the noncoding strand, indicating strong readthrough past comS and suggesting that the entire upregulated region downstream of comS represents a single readthrough mRNA (Fig. 3). Examination of the transcription map for the WT at 100 min reveals that this TAR starts close to the ComR box IR that has been described as the binding site for ComR (Fig. 5B) (41), further supporting a role for ComR in driving the transcription of this region.

The final member of the ComR regulon is sigX, which was expressed as a late gene, but not in the comS mutant (and was not probed in the sigX mutant). As for comS, the transcription start site at sigX mapped to a ComR box IR motif (Fig. 5B), providing support for a direct regulatory role for ComR in the expression of both comS and sigX. A previous transcriptome study of CSP-induced competence in S. mutans reported a TAR extending downstream of sigX that, given its position and the general variation observed in the expression of the 3′-terminal genes in several of the induced TARs in different transcriptomes, may represent a readthrough (Fig. 5). We conclude that just two TARs make up the ComR regulon, one initiated at comS but often extending downstream to encompass two to five additional genes and the second encompassing sigX, with an occasional downstream readthrough.

Because sigX is expressed late in the CSP response, genes restricted to late expression and not expressed in a sigX or comS mutant are candidates for the SigX regulon. The late class of genes is easily distinguished from the early class, as illustrated in the scatter plots in Fig. 4D and E. Discounting the genes belonging to the ComE and ComR regulons identified above (Fig. 4A to C), 23 TARs comprising a total of 78 ORFs transcribed in either the sense or the antisense direction are candidate members of the SigX regulon (Fig. 3, bottom). To distinguish direct from indirect regulation by SigX, we focused on the apparent presence of the highly conserved noncanonical −10 promoter element recognized by SigX polymerase, the SigX box, near the start of these TARs. Twenty-one of the 23 candidate TARs are preceded by a SigX box, as illustrated in the transcriptome maps in Fig. 5C. Of these, 7 represent SigX boxes not previously described and the remaining 16 include mostly SigX boxes that have been predicted in previous S. mutans studies and that were confirmed here by transcriptome mapping. Five other SigX boxes have been suggested in regions preceding competence-induced genes (SMU.431, SMU.504, SMU.507, SMU.925, and SMU.1904) (39), but these boxes were not confirmed in our study. At least one gene in each of these TARs (those in the sense direction) has been previously reported as upregulated in transcriptome surveys of the S. mutans response to CSP (Fig. 6). In the cases where there was some variation among the different transcriptomes regarding the set of genes that were induced within the SigX-controlled TAR regions, these usually involved genes in the 3′ termini of the TARs. Only two TARs, one for SMU.109 (possibly initiated at SMU.108) and another extending from SMU.166 to SMU.168, were upregulated in multiple transcriptomes but lack an apparent upstream SigX box (Fig. 6). These two are thus the only candidates for indirect regulation by SigX. We conclude that the SigX regulon comprises 23 TARs, with conserved upregulation of genes toward the 5′ ends of the TARs and a certain level of variation among different transcriptome surveys toward the set of genes activated in the 3′-terminal regions of the TARs.

During their evolution from a common ancestor, the streptococci became specialized for survival in diverse hosts and diverse sugar-rich niches, diverging not only by accumulation of mutations but also by gains and losses of many genes and by extensive shuffling within the genome (42). Throughout this evolution, competence for genetic transformation has been a conserved trait dependent on the alternative sigma factor SigX and the SigX regulon (12). The maintenance of the SigX regulon amid pervasive genetic change provides a natural genus-wide survey that can be used to distinguish conserved from variable components of the regulon. To mine the data provided by this natural experiment, we compared the competence-specific transcriptome data sets that are available for six species, representing five of the six major groups of species in this genus. In Fig. 7, we display alignments of competence-specific TARs in S. mutans with homologous regions that are upregulated in response to CSP in S. pneumoniae, S. gordonii, and S. sanguinis or by XIP in S. pyogenes and S. thermophilus (9, 13, 14, 43–46). Since induction levels are apparently low in S. pyogenes (9, 46) and transformation is observed only under particular biofilm conditions (47), the absence of core induced genes in their transcriptomes was not used as evidence for the lack of a core response. However, induction of S. pyogenes genes within core regions strengthened the classification of genes into the SigX core response. Inspection of the alignments reveals a broad pattern of synteny adjacent to the SigX box motifs but also variation in both gene presence and the observed length of associated TARs. Overall, the streptococcal SigX regulons contain a minority of genes invariably subject to SigX regulation, organized into three groups, (i) core genes induced in more than four streptococcal species (dut, radA, ccs50, cbf1, comFA, comFC, yfiA, cilC, comEA, comEC, coiA, pepB, pilC, dprA, radC, ssbB, comGA to comGH, ack, cinA, and recA), (ii) core genes in the DpnII group of strains (dpnA and dpnB), and (iii) core genes that have a domain in common but are not necessarily orthologs (lytF in S. mutans, S. gordonii, and S. sanguinis and cbpD in S. pneumoniae, S. thermophilus, and S. pyogenes), which we define as the core SigX regulon. This information was used to delineate a model of the transcriptional organization of the S. mutans response to CSP in peptide-rich medium (Fig. 8). A larger number of accessory genes found in SigX-dependent transcripts in some species but not in others are exemplified in Table S7 in the supplemental material. A notable feature of the core regulon is that core genes are usually at the 5′ extremity of a TAR, which immediately suggests that transcription of the accessory genes may reflect variation in terminator presence and efficacy. Indeed, in no case that we are aware of has an accessory gene downstream of a core gene been shown to provide a function in transformation. Additional accessory SigX-dependent genes are found in TARs that lack any core gene, but these are few in S. mutans, as only 4 of 22 TARs initiated by genes transcribed in the sense direction are not induced in the other species (3′ end of SMU.60, comED, SMU.1400, and possibly SMU.2076).

S. mutans UA159 and the isogenic mutants used in this study are presented in Table 3. Cultures of S. mutans were grown in TSB (Oxoid) at 37°C in 5% CO₂ and stored at −80°C in TSB supplemented with 15% glycerol.

CSP was used in the form of CSP18 (NH₂-SGSLSTFFRLFNRSFTQA-COOH), synthesized by GenScript (GenScript Corporation, NJ), with a purity of >95%. The lyophilized peptide was reconstituted in distilled water at 175 µg·ml⁻¹ and stored in small aliquots at −20°C.

Transformation experiments were performed in the absence or presence of CSP18 (50 nM). Cultures grown overnight at 37°C in 5% CO₂ were diluted to an optical density at 600 nm (OD₆₀₀) of 0.04. From this point, incubation proceeded at 37°C in ambient air. Upon reaching an OD₆₀₀ of 0.065, the cultures were distributed into Eppendorf tubes (1.2 ml) and CSP was added to a final concentration of 50 nM. At different times, a 100-µl aliquot was used for OD₆₀₀ measurements, and a 100-µl sample was pelleted and frozen for RNA extraction as described below. Another 100-µl portion was diluted 1:2 in fresh TSB containing replicative plasmid pVA838 DNA (final concentration of 1 µg·ml⁻¹). After a 20-min incubation with plasmid DNA, recombinant DNase I (Roche) was added at a final concentration of 10 U·ml⁻¹ and incubation proceeded for 40 min before dilution and plating on THB agar with or without erythromycin at a final concentration of 20 µg·ml⁻¹. The plates were incubated at 37°C in 5% CO₂ for 48 h before the counting of visible colonies.

Bacterial samples for real-time PCRs were collected at 10 and 100 min after the addition of CSP as described above. Total RNA was extracted with the High Pure RNA isolation kit (Roche, Mannheim, Germany) according to the manufacturer’s recommendation, except that the cells were incubated at 37°C for 20 min in 200 µl of lysis buffer containing 10 mM Tris (pH 8), 20 mg of lysozyme ml⁻¹, and 100 U of mutanolysin ml⁻¹. DNase I was used during RNA extraction to remove the remaining DNA. Complementary DNA templates were prepared from RNA with the Transcriptor First Strand cDNA synthesis kit (Roche Diagnostics GmbH, Mannheim, Germany) in accordance with the manufacturer’s protocol. Controls without reverse transcriptase were included. Expression of cipB and comGA was examined by real-time PCR with primer pairs FP156 (TGCTCTAGGTGCTGGGCAAG)-FP157 (GAGCTCCTCCGATTCCTCCA), FP166 (ATTGGCAACAAGAGGGAATG)-FP167 (TCTTGCTGACGCAAAACATC), and FP128 (AGAAACCGCCAGAGCTGTTA)-FP129 (CCACGCAAAGCATTTTGTAA), respectively. To normalize the data, primer pair FP299 (CCATGACCATCAACCAACAT)-FP300 (ATCAGCGCGTATTACAGGTG) was used to amplify a portion of gyrA. Assays were carried out with quantitative PCR master mix for SYBR green I. Data were collected and compared with the software and graphics program MxPro (Stratagene).

RNA samples were from the WT UA159 and the comS deletion mutant, grown in the presence or absence of CSP in 100-ml volumes of TSB as described for the transformation assay. WT cultures were collected by centrifugation at 10 and 100 min after CSP addition, and comS deletion mutant cultures were collected at 100 min. At each time, samples without CSP were included as controls. Two independent biological replicates were obtained for each condition, giving a total of 12 samples. Immediately after centrifugation (9,000 × g, 2.5 min, 4°C), the pellets were frozen in liquid nitrogen. RNA was prepared as previously described, with a few modifications (30). Briefly, the pellets were lysed with mutanolysin-lysozyme, followed by RNA extraction with the mirVana miRNA isolation kit (Ambion). This kit was used to enhance the rate of recovery of short transcripts (down to 10 nucleotides). The samples were then treated with Turbo DNase (Ambion) and analyzed for quality with a Bioanalyzer. Samples with remaining DNA, as determined by PCR with primers for ccpA (FP297, [GTAGGTGTGGTTATCCCTAATATTGC] and FP298 [ATAAATCGGCTGACTGATAGATGTC]), were retreated with Turbo DNase, and repurified until no DNA was detected. The MICROBExpress kit (Ambion) was then used for mRNA enrichment. RNA was then fragmented and Cy3 labeled (Mirus Label IT µArray Cy3 labeling kit; Mirus); this was followed by hybridization to the microarray probes. UA159 genomic reference DNA was purified with the DNeasy Blood and Tissue kit (Qiagen), with mutanolysin-lysozyme treatment for the lysis step, followed by fragmentation and labeling (Mirus Label IT µArray Cy3 labeling kit).

The genomic tiling microarray was constructed with probe sets designed from both forward and reverse complement strands of the entire target genome of S. mutans UA159 as described by Høvik and Chen (29). A total of 385,000 optimized probes covered the entire genome, including ORFs and intergenic regions. The probes were printed on high-density microarrays by Roche NimbleGen. To block nonspecific binding of RNA molecules, RNase-free bovine serum albumin (500 µg·ml⁻¹) was added to the prehybridization solution and salmon sperm DNA (100 to 700 µg·ml⁻¹) was included in both prehybridization and hybridization solutions. Hybridization was conducted at a temperature of 42°C in the presence of 25% formamide. NimbleScan v2.5 software was used for spot feature extraction from the scanned images, followed by normalization and analysis as previously described (30). Briefly, the nonspecific background was estimated from the intensity of the intergenic sequence probes and of the genomic DNA reference and used for corrections due to sequence-specific factors (68). Normalization between arrays was done with the vsn algorithm (29). The log₂ means of the normalized signal intensities from each condition were used for downstream processes.

A Hidden Markov support vector machine (69) was used to identify the boundaries of TARs on the basis of a set of training data derived from both ORFs and intergenic regions. The expression level of annotated genes was determined by averaging the nucleotide intensities of probe signals within the length of the gene (68). Differential expression at the ORF level was measured as the difference between the log₂ mean probe signal intensities of the control and CSP-treated samples from two independent biological experiments, except for the 10-min sample without CSP, in which one of the hybridizations failed. The P values were calculated with the SAM software (68) at default settings by performing 10 permutations for the inclusion of two sets of repeats. Genes that exhibited a >2-fold mean signal intensity difference (with a P value of <0.05) were registered as differentially expressed.

Original and normalized microarray data used in this study were deposited in the NCBI Gene Expression Omnibus database (http://www.ncbi.nlm.nih.gov/geo ) under accession no. GSE70067 . The transcriptome profiles are also available for browsing at the Microbial Transcriptome Database website (http://bioinformatics.forsyth.org/mtd/dataset=RNAseq_smut_comS ).