Regulation of pneumococcal epigenetic and colony phases by multiple two-component regulatory systems

Streptococcus pneumoniae is well known for phase variation between opaque (O) and transparent (T) colonies within clonal populations. While the O variant is specialized in invasive infection (with a thicker capsule and higher resistance to host clearance), the T counterpart possesses a relatively thinner capsule and thereby higher airway adherence and colonization. Our previous study found that phase variation is caused by reversible switches of the “opaque ON-or-OFF” methylomes or methylation patterns of pneumococcal genome, which is dominantly driven by the PsrA-catalyzed inversions of the DNA methyltransferase hsdS genes. This study revealed that switch frequency between the O and T variants is regulated by five transcriptional response regulators (rr) of the two-component systems (TCSs). The mutants of rr06, rr08, rr09, rr11 and rr14 produced significantly fewer O and more T colonies. Further mutagenesis revealed that RR06, RR08, RR09 and RR11 enrich the O variant by modulating the directions of the PsrA-catalyzed inversion reactions. In contrast, the impact of RR14 (RitR) on phase variation is independent of PsrA. Consistently, SMRT sequencing uncovered significantly diminished “opaque ON” methylome in the mutants of rr06, rr08, rr09 and rr11 but not that of rr14. Lastly, the phosphorylated form of RR11 was shown to activate the transcription of comW and two sugar utilization systems that are necessary for maintenance of the “opaque ON” genotype and phenotype. This work has thus uncovered multiple novel mechanisms that balance pneumococcal epigenetic status and physiology.

Introduction Streptococcus pneumoniae (pneumococcus) is a commensal at human nasopharynx and also an important opportunistic pathogen that causes pneumonia, meningitis, septicemia, and otitis media [1]. This bacterium is well known for its strain-to-strain genetic variations, which are predominantly driven by natural transformation-mediated horizontal gene transfer. These variations are exemplified by acquisition of the exogenous genes conferring antimicrobial resistance, change of capsule types, and permutation of surface-exposed antigens. In the absence of foreign DNA, the pneumococcus is also capable of phase variation or reversible switch between the opaque (O) and transparent (T) colony variants in clonal populations on translucent agar plates [2,3]. While the O colony variant is specialized in invasive infection (with a thicker capsule and higher resistance to host clearance), the transparent counterpart is characterized by a relatively thinner capsule, more cell wall teichoic acids, and thereby higher airway adherence [2][3][4]. In animal models, the O variant is more virulent in systemic infection while the T variants exhibit higher levels of nasopharyngeal colonization with relatively lower virulence [3,4]. Recent studies have revealed that pneumococcal phase variation is epigenetically defined by reversible switches of the methylomes or genome methylation patterns [5,6], which is controlled by invertase PsrA-catalyzed inversions of three homologous methyltransferase hsdS genes in the colony opacity determinant (cod) locus, also referred to as Spn556II locus [5,7,8].
The cod locus, a type-I restriction-modification (RM) system, consists of the hsdR (restriction endonuclease), hsdM (DNA methyltransferase), psrA (invertase), and three homologous hsdS (hsdS A , hsdS B and hsdS C ) genes [5]. PsrA catalyzes extensive inversions between hsdS A and two downstream homologs (hsdS B and hsdS C ) [9,10]. As a typical sequence recognition or S subunit of the type-I RM DNA methyltransferase (MTase) [11], hsdS A encodes two target recognition domains (TRDs), each of which recognizes a half of the type-I RM methylation bipartite sequence [5]. Our recent study has shown that only hsdS A is actively transcribed and produces a functional S subunit for the MTase, whereas hsdS B and hsdS C merely serve as templates for replacement of one or two TRD domains in the HsdS A MTase by the PsrA-catalyzed inversions [5,9]. The inversions generate six different hsdS A alleles, each of which drives methylation of a unique DNA motif in the genome of S. pneumoniae [5,7]. This site-specific recombination mechanism enables a single pneumococcal cell to generate a mixture of progeny cells each carrying one of the six hsdS A alleles (A1-A6) and thereby producing six distinct methylomes [5]. Only the pneumococci carrying hsdS A1 form O colonies and those with the other five hsdS A alleles produce T colonies [5,6]. Therefore, the reversible switch between O and T colony phases is epigenetically driven by reversible "ON-or-OFF" alteration of the hsdS A1 genotype in the cod locus [5], although the biochemical basis of the colony opacity remains to be characterized.
The two-component signal transduction systems (TCSs) are widely found in bacteria to regulate a variety of cellular processes in response to changes in environmental conditions, such as chemotaxis, capsule production, balance of osmolarity, photosynthesis, sporulation and transformation [12]. A typical TCS is comprised of a histidine kinase (HK) as a sensor and a cognate response regulator (RR) as an effector. The histidine kinase auto-phosphorylates itself at the conserved histidine residue when it senses extracellular changes and the cognate response regulator is subsequently activated by accepting the phosphoryl group from histidine to its conserved aspartic acid [13]. The DNA-binding output domain of activated regulator typically interacts with the promoter(s) of the target genes and activates/represses their transcription. S. pneumoniae has 13 complete TCSs and an orphan regulator (RR14, RitR) [14][15][16]. Previous studies have shown that most of the TCSs are necessary for pneumococcal pathogenicity in animal models [15,17], but the precise functions for most of the TCSs are largely unknown.

Five response regulators modulate pneumococcal colony phases
To determine potential impact of pneumococcal TCSs on phase variation in colony opacity, we generated unmarked deletion mutants of 13 response regulator (rr) genes in strain ST606, a streptomycin-resistant rpsL1 derivative of type-19F strain ST556, except for the essential gene rr02 [14,44]. The rationale was that disruption of the rr gene in bacterial TCS is more likely to yield a functional phenotype [45]. The colony phenotype or ratio between O and T colonies in a single clonally derived population was characterized for each mutant on the catalase-TSA plates. Similar to the parental strain ST606, seven of the 13 rr mutants generated populations that were dominated by O colonies (rr01, rr03, rr04, rr07, rr10, rr12 and rr13) ( Table 1). As an example, the rr10 mutant produced 75.2% O and 24.8% T colonies, which is comparable with the parental strain (79.8% O and 20.2% T). In contrast, deleting the six other regulators (rr05, rr06, rr08, rr09, rr11 and rr14) led to significantly reduced proportion of O colonies as compared with the parental strain ( Fig 1A). The % O values of the rr05, rr06, rr08, rr09, rr11 and rr14 mutants are 60.9, 9.0, 14.5, 17.0, 22.2 and 27.8, respectively (Table 1). It should be noted

PLOS PATHOGENS
Regulation of pneumococcal phase variation that the transparent colonies of the rr05 mutant were morphologically different from the parental counterpart. ST606 produced transparent colonies with a large translucent center and a thin halo around it, but the rr05 counterparts showed a relatively smaller transparent center and a thicker ring (Fig 1A).
To verify the impact of these six regulators on pneumococcal colony opacity, we constructed genetic revertants of these mutants by replacing each deletion with the corresponding wild type gene in its native locus of the genome. The revertants of the rr06, rr08, rr09, rr11 and rr14 mutants displayed a comparable fraction of O colonies as the parental strain (Fig 1B;  S1 Fig; S1 Table). However, the rr05 revertant strain produced 100.0% O colonies. Whole genome sequencing of the rr05 mutant and its revertant revealed a single nucleotide mutation from adenine to cytosine at the 539 th position in the coding region of MYY890, which encodes the substrate-binding protein in the phosphate uptake system [31]. This nonsynonymous mutation led to a substitution of the 180 th glutamine (CAG) with proline (CCG) of this protein. It is worth of mentioning that other pneumococcal strains lacking rr05 also tended to carry spontaneous mutations in our collection. These findings strongly suggested that deleting rr05 somehow renders the genome more prone to spontaneous mutations, which makes it difficult to properly interpret the colony morphology data. The TCS05 was thus excluded from further investigation. These results demonstrated that RR06, RR08, RR09, RR11 and RR14 regulate pneumococcal colony opacity.
We next determined the impact of the five regulators on reversibility between O and T colonies by re-streaking individual O and T colonies of each mutant on catalase-TSA plates. For each of the five mutants, the progeny colonies derived from both the O and T colony seeds showed a comparable ratio between the O and T phenotypes. As an example, the populations derived from the O ( Fig 1C) and T (Fig 1D) seeding colonies of the rr06 mutant showed the most dramatic but similar reduction in the fraction of O colonies (1.9% for O seed, 1.6% for T seed). This finding is consistent with the result from the initial screening of the rr06 mutant

PLOS PATHOGENS
Regulation of pneumococcal phase variation

PLOS PATHOGENS
Regulation of pneumococcal phase variation ( Fig 1B). As a negative control, progeny colonies of the rr10 mutant did not show significant alteration in colony phenotype, regardless the original phenotypes of the seeding colonies ( Fig  1C and 1D). This result indicated that the rr06, rr08, rr09, rr11 and rr14 mutants still retained the capability of phase switching.

Colony phases are regulated by the PsrA-dependent and -independent mechanisms
Our previous study has found that the reversible switch between O and T colony phenotypes is controlled by the PsrA-catalyzed inversions of the three invertible regions in the cod locus [5,9] (Fig 2A). Therefore, enzymatic inactive psrA Y247A mutant (with a point mutation in the catalytic residue tyrosine 247) is locked in either O or T colony phase [9]. To determine whether the impact of RR06, RR08, RR09, RR11 and RR14 on phase variation requires PsrA-mediated inversions, we constructed deletion mutant in each of these genes in the O-locked psrA Y247A strain, which was previously shown to produce only the O colonies [5] (Fig 2B; S2 Fig). Additional deletion of rr14 in this inversion-deficient mutant still resulted in significant changes in colony phenotype. To a less extent, the Δrr14-psrA Y247A mutant formed 66.9% O colonies. This result suggested that RR14 (RitR) regulates pneumococcal colony opacity in a PsrA (hsdS inversion)-independent manner. However, deleting rr06, rr08, rr09 or rr11 in the psrA Y247A background did not yield any obvious impact on the phenotype of the strain. Virtually all of the colonies formed by these four double mutants showed the O phenotype ( Fig 2B; S2 Fig). This result indicated that regulatory impact of RR06, RR08, RR09 and RR11 on phase variation depends on the PsrA-mediated hsdS inversions.

The PsrA-dependent regulators modulate pneumococcal methylome
Because only the methylome specified by the HsdS A1 MTase is required for the formation of O colonies [5], the abovementioned changes of the rr mutants in colony phenotypes strongly

PLOS PATHOGENS
Regulation of pneumococcal phase variation suggested that these response regulators regulate pneumococcal methylome through PsrAdriven inversions. We thus compared the methylomes of strain ST606 and 13 isogenic rr mutants by single molecule real-time (SMRT) sequencing. This trial detected N6-methyladenine (6-mA) in virtually all 664 loci of the Spn556I recognition motif (5'-TCTAG m6 A-3', type II RM) in ST606 and its 13 derivatives (S2 Table), and thereby demonstrated that our sequencing setup was adequate for genome-wide detection of all 6-mA methylated sequences. Consistent with frequent hsdS inversions in the cod locus [5,7], SMRT sequencing revealed 6-mA methylation in the motifs recognized by four of the six hsdS A allelic variants (HsdS A1 , HsdS A2 , HsdS A3 and HsdS A4 ) of the cod locus in ST606 ( Fig 3A; Table 2). Virtually all 2,058 loci of the HsdS A1 motif in the genome were methylated (99.4%) in ST606, but there were much lower methylation percentages for the other motifs (HsdS A2 −38.7%, HsdS A3 −69.5% and HsdS A4 −7.0%); no 6-mA methylation was detected for any loci of the HsdS A5 and HsdS A6 motifs in ST606 or its 13 derivatives. This observation is consistent with the previous finding that single clonal population of ST606 is predominantly made up by the cells possessing the hsdS A1 , hsdS A2 or hsdS A3 allele [5]. As exemplified with the rr10 mutant in Fig 3B, nine of the 13 rr mutants (rr01, rr03, rr04, rr05, rr07, rr10, rr12, rr13 and rr14) exhibited a similar methylome as the parental strain (Table 2, S2 and S3 Tables).
In agreement with significant decrease in the fraction of O colonies in the rr06, rr08, rr09 and rr11 mutants, these strains showed dramatic alteration in the methylome specified by the "opaque ON" HsdS A1 MTase. The rr06 mutant showed the most striking overhaul in the methylome ( Fig 3C). Compared with 99.4% methylation rate of the HsdS A1 motif in ST606 (Fig 3A), none of these sequences were detected as a methylated form in this mutant. The complete loss of methylated HsdS A1 motif loci in this strain was accompanied with increase in methylation rates of the HsdS A2 (2,054 copies) and HsdS A3 (1,468 copies) motifs. Identification of 6-mA methylation in the HsdS A2 and HsdS A3 motifs demonstrated that the HsdM subunit of the cod MTase was functional in the rr06 mutant. The complete loss of HsdS A1 motif methylation in the rr06 mutant was consistent with dramatic reduction in the fraction of the O colonies in this strain (Fig 1; Table 1). Similar to their common characteristics in the colony phenotype (Fig 1; Table 1), the mutants of rr08 and rr09 showed a similar pattern in methylome. Methylation of the HsdS A1 motif loci fell to 63.1% and 44.5% in the rr08 ( Fig 3D) and rr09 ( Fig 3E) mutants, respectively. Both mutants also displayed a moderate increase in methylation of the HsdS A3 motif loci. Interestingly, none of the 2,054 HsdS A2 motif copies were methylated in either rr08 or rr09 (Table 2). In sharp contrast, nearly all of the HsdS A2 motif loci were methylated in the rr11 mutant (94.5%) ( Fig 3F). The methylation rates of the other HsdS A motifs were extremely low (5.5% for HsdS A1 and 6.3% for HsdS A3 ) or undetectable (HsdS A4-6 ) in this strain. The methylation level of various HsdS A motifs in the rr11 mutant is consistent with its colony phenotype (Fig 1). In summary, the SMRT sequencing confirmed the observation with the psrA Y247A -rr double mutants that RR06, RR08, RR09 and RR11 regulate pneumococcal methylome through the PsrA-driven inversions.
In contrary to its T colony-dominant phenotype, the methylome of the rr14 mutant was virtually identical to that of the parental strain (Table 2; S2 and S3 Tables). In particular, virtually all 2,058 copies of the HsdS A1 motif were methylated in the rr14 mutant ( Fig 3G). This result is consistent with their lack of association with PsrA-catalyzed inversions (Fig 2), and further validated our conclusion that RR14 impacts pneumococcal colony morphology through a nonepigenetic mechanism. DNA methylation detection also revealed certain interesting features associated with the Spn556III locus, the other functional type-I RM system in ST556 and many other pneumococcal strains [7,46,47]. A recent study reports that the two hsdS genes in this locus undergo excision and reintegration recombinations in other pneumococcal strains [47]. In addition to a previously identified motif (5'-GAT m6 AN 7 TCA-3') for this system [46], Relative methylation rate of each DNA motif recognized by HsdS A1 (5'-CRA m6 AN 8 CTT-3'/3'-GYTTN 8 G m6 AA-5', 2,058 loci), HsdS A2 (5'-CRA m6 AN 9 TTC-3'/3'-GYTTN 9 m6 AAG-5', 2,054 loci), HsdS A3 (5'-CRA m6 AN 8 CTG-3'/3'-GYTTN 8 G m6 AC-5', 1,468 loci) and HsdS A4 (5'-C m6 ACN 7 CTG-3'/3'-GTGN 7 G m6 AC-5', 888 loci) was calculated in each strain by dividing the number of methylated chromosomal loci for each motif with the total loci of the motif in the ST556 genome. Only the values for the parental strain ST606 (A) and isogenic mutants of rr10 (B), rr06 (C), rr08 (D), rr09 (E), rr11 (F) and rr14 (G) are presented. The results for the other rr mutants are described in Table 2. the parental (5'-GG m6 AN 7 TGA-3') and Δrr13 (5'-GG m6 AN 7 TCA-3') strains each showed a new type-I RM MTase recognition sequence (S3 Table). Based on the published specificities of the HsdS variants for this system [47], it is apparent that these sequences were methylated by two new HsdS alleles generated by DNA excisions in the Spn556III locus. We designated these HsdS alleles as the HsdS 1 (published), HsdS 2 (new in ST606) and HsdS 3 (new in Δrr13) (S3 Table). Complete shift of the methylation activity from the HsdS 1 MTase to HsdS 3 MTase in Δrr13 was caused by the formation of a hybrid hsdS 3 allele in the Spn556III locus as reflected in the SMRT sequencing data. This observation suggested that RR13 influences the hsdS recombinations in the Spn556III locus. However, our preliminary mutagenesis trial did not reveal obvious connection between the Spn556III locus and colony phases. The mutants with unmarked deletion of the entire Spn556III locus displayed a similar methylation rate of the cod HsdS A1 motif to the parental strain.

The PsrA-dependent regulators modulate the colony phases in multiple pneumococcal strains
Based on the significant impact of RR06, RR08, RR09 and RR11 on the directions of phase variation in the ST556 background, we tested the role of these four regulators in P384 (serotype 6A) and ST877 (serotype 35B), two strains that exhibited typical hsdS A1 -dependent phase variation in colony opacity [5]. In agreement with our previous study [5], strain P384 produced 76.2% O colonies on catalase-TSA plates. However, deleting rr06, rr08, rr09 or rr11 led to dramatic reduction in the fraction of O colonies within clonal populations (Fig 4A). The mutants

PLOS PATHOGENS
Regulation of pneumococcal phase variation

Phosphorylated form of RR11 drives pneumococci toward the O phase
We further characterized the molecular mechanism behind the action of RR11 in regulation of pneumococcal phase variation. We first tested whether RR11 depends on phosphorylation state of the 53 rd aspartic acid residue (D 53 ) for its role in phase variation by making a D-to-A point mutation at this site. D 53 has been identified as an amino acid residue to receive phosphoryl group for activation of RR11 [43]. Similar to the rr11 deletion mutant (TH9063), the rr11 D53A strain (TH11425) displayed a "hypo-O" phenotype, in which the proportion of O colonies decreased to 14.5% (Fig 5A). The effect of this point mutation on colony phenotype was successfully rescued by in situ replacing the rr11 D53A allele with the wildtype rr11, suggesting that phosphorylation of RR11 is required for its modulation of phase variation. Previous studies have found that the substitution of the conserved aspartic acid residue (D) for phosphorylation in bacterial response regulators with glutamic acid (E) can trigger proteins in an activated or "constitutively phosphorylated" state in the absence of the cognate sensing kinase [48,49]. We thus constructed a ST606 derivative carrying a rr11 D53E allele by replacing the endogenous rr11 in the original chromosomal locus (TH13757). Consistent with the predicted "constitutive phosphorylation" status of the RR11 protein, the rr11 D53E strain showed a "hyper-O" phenotype with 93.4% of O colonies in a clonally derived population ( Fig 5A). Furthermore, reverting the rr11 deletion with the wildtype rr11 allele fully restored the colony phenotype of the mutant to that of the parental strain. The revertant produced 79.7% O colonies. These genetic and phenotypic characterizations indicated that the phosphorylated form of RR11 is required for the generation and/or maintenance of an O colony-dominated population. We next tested the role of HK11, the sensor histidine kinase partner of RR11, in modulation of pneumococcal colony phenotype by unmarked deletion of hk11 in ST606. The Δhk11 strain formed 41.4% O colonies, a significant reduction from that of the parental strain (80.3%), although it still produced more O colonies than the Δrr11 strain (23.2%) (Fig 5B). This result suggested that the HK11 is involved in phase variation. To verify this finding, we constructed a H-to-A mutation at the 184 th histidine residue of the HK11 because this represents the highly conserved site for auto-phosphorylation of sensing kinases in bacterial TCSs [13]. The hk11 H184A mutant showed only a modest reduction in proportion of O colonies (with 67.8% O colonies) although this reduction was fully restored in the hk11 revertant (Fig 5B). The milder phenotype of the hk11 H184A strain than that of rr11 D53A suggested that RR11 is phosphorylated by an alternative phosphoryl group donor(s) besides HK11. Hentrich et al. reported that intracellular acetyl phosphate generated by pyruvate kinase SpxB is capable of phosphorylating the RR05 (CiaR) response regulator [50]. To elaborate on the function of HK11 in phase variation, we constructed a "hyper-active" mutant of HK11 by making a T-to-P mutation in the 190 th threonine residue. Previous study has demonstrated that the same mutation in the corresponding threonine residue at the 230 th position of HK05 (CiaH) renders the protein constitutively active as the sensor kinase [20]. The "constitutively active" hk11 T190P strain displayed a "hyper-O" phenotype (with 92.2% O colonies), which resembles the phenotype of the "constitutively phosphorylated" RR11 (rr11 D53E ) strain. These lines of evidence showed that the kinase activity of the sensing kinase HK11 is necessary for phosphorylation of RR11 in generation and/or maintenance of an O colony-dominated population.
Finally, we tested the phenotypic impact of the entire tcs11 locus by simultaneous deletion of both the rr11 and hk11 genes in ST606. In a dramatic fashion, the tcs11 mutant lost the ability to produce O colonies (0.4%) and produced only T colonies (Fig 5C). This deficiency was fully complemented to the parental level (81.6%) in the revertant. Taken together, the combinations of gene deletion and complementation unequivocally have demonstrated that the

PLOS PATHOGENS
Regulation of pneumococcal phase variation

PLOS PATHOGENS
Regulation of pneumococcal phase variation TCS11 system regulates the direction of phase variation in favor of the O phase by phosphorylation of the RR11 response regulator.

Phosphorylated RR11 stabilizes the "opaque ON" genotype in a PsrAdependent manner
Because only the hsdS A1 allele, one of the six functional hsdS A alleles derived from the PsrAdriven inversions in the cod locus, enables pneumococci to produce O colonies [5,6], the requirement of TCS11 for the O colony-dominant phenotype implied that this two-component system enriches the fraction of the hsdS A1 -carrying variant in the population. We thus determined relative abundance of the hsdS A1 -carrying variant in the populations of ST606 and various TCS11 mutants by measuring the relative mRNA abundance of the hsdS A1 allele in the context of the six hsdS A allelic variants. The quantitative reverse transcriptase PCR (qRT-PCR) analysis showed 41.9% hsdS A1 allele out of the six potential hsdS A alleles in the parental strain ST606 (Fig 6A). However, the hsdS A1 transcript was reduced to 17.1% and 16.5% in the rr11null and rr11 D53A mutants, respectively, but the significantly decreased representation of the hsdS A1 allele in the RR11-deficient populations was restored to the parental level in the rr11 revertant (rr11 rev). Consistent with its "opaque ON" phenotype, the "hyper-phosphorylated" rr11 D53E mutant showed significantly increased hsdS A1 transcript (65.4%). This result indicated that phosphorylated RR11 enriches the hsdS A1 variant out of the six hsdS A allelic variants.
We next determined if the function of RR11 in regulation of pneumococcal colony phases involves PsrA-catalyzed inversions in the cod locus by constructing multiple rr11 mutations in the invertase-negative psrA Y247A strain. In sharp contrast to remarkable impact of RR11 on the colony opacity phenotype in the wild type pneumococci (Fig 5), all genetic manipulations of rr11 in the psrA Y247A strain did not show obvious impact on the opaque-locked phenotype. All of the rr11 mutants (Δrr11, rr11 D53A and rr11 D53E ) in the psrA Y247A background showed the "O" phenotype ( Fig 6B; S3 Fig), indicating that RR11 is no longer able to impact the colony phases in the absence of the hsdS inversions. To verify the importance of PsrA in the regulation of RR11 on pneumococcal colony phases, we determined the fraction of the hsdS A1 -carrying cells in various genetic backgrounds of psrA and rr11 by measuring the relative abundance of the hsdS A1 mRNA molecules in these strains. As shown in Fig 6C, all of the three rr11 mutants (Δrr11, rr11 D53A and rr11 D53E ) generated in the psrA Y247A background had a uniformly increased level of the hsdS A1 mRNAs as compared with the psrA + ST606 strain. Taken together, these results have demonstrated that RR11 collaborates with the PsrA invertase in regulating the hsdS A allelic configurations in the cod locus.

Multiple RR11-regulated genes are required for stabilization of the O phase
To determine how RR11 regulates pneumococcal colony phases, we compared the transcriptomes between ST606 and Δrr11 mutant by RNA sequencing (RNA-seq). This trial identified a total of 23 genes with significant reduction (>2-fold change) in transcription between ST606 and Δrr11 (Table 3). This result thus revealed that RR11 acts as a transcriptional activator under these conditions. Except for comW, MYY691 and MYY1747, all the RR11-activated genes are concentrated in six apparent operons (Table 3; S4A Fig). While comW encodes the DNA-binding regulator ComW, a transcriptional activator of the competence genes [51,52], virtually all of the other down-regulated genes in the rr11 mutant are associated with substrate uptake and metabolism. However, the rr11 mutant showed a similar level of the psrA mRNA as the parental strain (S4 Table). This result indicated that RR11 does not regulate pneumococcal colony phases through modulating the transcription of psrA.

PLOS PATHOGENS
Regulation of pneumococcal phase variation

Regulation of pneumococcal phase variation
We next determined potential contribution of the RR11-regulated genes to the colony phases. The qRT-PCR result confirmed significant down-regulation of 24 genes within the RR11-regulated loci identified by the RNA-seq (S4B Fig). Further deletion analysis revealed significant shift in the colony phenotype in the mutants of three RR11-regulated gene loci ( Fig  7A). The fraction of the O colonies in the comW mutant was dramatically diminished to 9.3% from 80.7% in the parental strain (Fig 7B). To a less extent, unmarked deletion of the genes in the two sugar phosphotransferase systems (PTS) also led to significant reduction in the fraction of the O colonies (Fig 7A). The mutants of the bgaC (MYY134-139) and ugl (MYY403-408) loci showed 42.4% and 31.9% O colonies, respectively (Fig 7B). While the bgaC locus is implicated in pneumococcal uptake and utilization of galactose and N-acetylgalactosamine [53,54], the ugl gene cluster is necessary for uptake and metabolism of hyaluronic acid [55,56]. The mutagenesis did not detect obvious changes in the mutants of the other RR11-regulated loci as exemplified by the ΔMYY1793-1796 (putative ABC transporter, gtfA locus), ΔMYY1924-1925 (putative ABC transporter, rr11 locus) and ΔMYY2067-2068 (arginine shown as mean ± SEM of a representative experiment. Each experiment was replicated at least twice. B. Colony opacity of the psrA Y247A -rr11 double mutants. ST606 (WT) and its derivatives with either the psrA Y247A allele alone or additional rr11 allelic modifications were processed for colony enumeration, photographing and data presentation as in Fig 1B. C. Relative abundance of the hsdS A1 mRNA in the psrA Y247A -rr11 double mutants. Same as in Fig 6A except

PLOS PATHOGENS
Regulation of pneumococcal phase variation

PLOS PATHOGENS
Regulation of pneumococcal phase variation metabolism, arcA locus) strains (Fig 7B). This result revealed that the competence regulator ComW and two sugar utilization systems are associated with pneumococcal phase variation.
To define how the RR11-regulated genes impact the colony phenotype, we tested if comW and the two sugar utilization loci are involved in the dominance of the hsdS A1 -carrying cells in the ST606 population. Relative abundance of the hsdS A1 allele (out of the six potential hsdS A allelic variants in the cod locus) was compared between ST606 and isogenic mutants of the RR11-regulated gene loci by testing the levels of the hsdS A1 mRNA using qRT-PCR. As compared with the invertase-negative psrA Y247A strain (an hsdS A1 -locked strain producing 100% hsdS A1 mRNA), the parental strain ST606 produced 44.5% hsdS A1 mRNA (Fig 7C). However, the level was reduced to 3.5% and 1.9% in the mutants of the comW and ugl loci, respectively, indicating that comW and the hyaluronate utilization locus are essential for stabilization of hsdS A1 allelic configuration in the cod locus. In sharp contrast to its significant change in colony phenotype (Fig 7B), the mutant of the bgaC locus had a comparable level of the hsdS A1 allele as the parental strain (Fig 7C). This result suggested that the galactose utilization system impacts pneumococcal colony phases in an inversion-independent manner. As represented by the ΔMYY1793-1796 (gtfA locus), ΔMYY1924-1925 (rr11 locus) and ΔMYY2067-2068 (arcA locus) mutants, deleting other RR11-regulated genes did not lead to significant change in the allelic dominance of hsdS A1 in ST606 (Fig 7C). Taken together, these data showed that three of the RR11-regulated gene loci are necessary for stabilization of the O phase. While the comW and hyaluronate utilization loci achieve this function by enforcing the hsdS A1 allelic configuration in the cod locus, the galactose utilization system impacts pneumococcal colony phases in an inversion-independent manner.

Discussion
Phase variation in colony opacity has been well characterized as an important strategy for pneumococcal adaption to various host environmental conditions [3,4]. However, it is completely unknown whether the directions of the reversible switches between O and T colonies are re-balanced in response to environmental conditions. Our systematic screening of the 13 two-component response regulators, for the first time, has demonstrated that the balance between the O and T phases of S. pneumoniae is subjective to modulation by five TCSs in the hsdS inversion-dependent (RR06, RR08, RR09 and RR11) and -independent (RR14) manners. Although the precise environmental signal(s) sensed by each of the regulators remains to be identified, this study uncovers a new level of molecular sophistication and complexity in bacterial fine tuning of its biological activities and behaviors by linking the two-component environmental sensing systems to the epigenetic switching machinery (illustrated in Fig 8). Furthermore, the colony phenotypes of the TCS mutants identified in this work are highly valuable for systematic characterization of pneumococcal two-component systems in the future because their biological contributions remain largely unknown due to the lack of phenotypic hints.
This study has shown that, once being phosphorylated by its cognate sensing kinase HK11 and/or other phosphoryl group donor(s), RR11 indirectly enforces the O colony phase of S. pneumoniae by transcriptional activation of its target genes. The importance of RR11 phosphorylation is supported by the similarly diminished O phase of the rr11-null and phosphorylation deficient rr11 D53A strains. Consistently, the RR11 hyper-phosphorylation rr11 D53E strain displayed a hyper-O colony phenotype. Mutations in the key amino acid residues of the HK11 have also corroborated the significance of RR11 phosphorylation in favoring the O phase. Lastly, complete loss of the O colonies in the TCS11-deficient mutant has provided the striking phenotypic evidence for the essential role of this two-component system in modulating the directions of PsrA-mediated inversions and phase variation. More severe phenotype of the

PLOS PATHOGENS
Regulation of pneumococcal phase variation tcs11-null mutant (0.4% O colonies) than the rr11 counterpart (reduced O colonies) suggests that HK11 may cross-phosphorylate other response regulators (e.g. RR06, RR08 and RR09) that are involved in modulating the directions of the inversion reactions in the cod locus. Cross-phosphorylation of bacterial response regulators has been well documented [57].
RR11 activates the transcription of comW and two sugar utilization gene (bgaC and ugl) loci that are necessary for the maintenance of the O colony phase. Among the 23 RR11-regulated genes identified by RNA-seq, comW and two sugar utilization gene (bgaC and ugl) loci are shown to promote the O phase. The comW and hyaluronate utilization (ugl) loci accomplish this function by modulation of the PsrA-catalyzed inversions in the cod locus because deleting comW or the entire ugl locus not only led to significant loss of the O colonies but also diminished the dominance of the bacterial cells carrying the "opaque ON" hsdS A1 allele in the clonal populations. In contrast, loss of the galactose utilization (bgaC) system only shrank the representation of the O colonies without significant change in the status of hsdS A1 in the mutant population. This information points to the conclusion that the bgaC locus impacts pneumococcal colony phases in an inversion-independent manner. While it remains to be determined if all of the genes in the ugl and bgaC sugar utilization loci are involved in modulation of the colony phases, the existing information has provided more mechanistic hint for the action of ComW.
ComW may regulate hsdS inversions through its physical interactions with unknown enhancer sequence(s) and/or PsrA. ComW is known as a competence-specific protein that stabilizes and activates the alternative sigma factor ComX for transcriptional activation of the natural transformation genes [51,52]. A recent study shows that ComW binds DNA independent of DNA sequence, and its activity in activating pneumococcal transformation depends on the DNA binding activity [58], but this protein has not been associated with any non-competence function. In the context, ComW could act like the HU and/or Fis DNA-binding proteins of Salmonella enterica serovar Typhimurium. HU and Fis (factor for inversion stimulation) are

PLOS PATHOGENS
Regulation of pneumococcal phase variation the nucleoid-associated proteins that promote the Hin recombinase-catalyzed inversions, the best-characterized DNA inversion system in prokaryotes [59,60]. The Hin-mediated inversions of a~1 kb sequence flanked by the two hix sites (inverted repeat sequences recognized by the Hin recombinase) lead to alternative expression of the FljB and FliC flagellins and thereby drives phase (antigenic) variation in two different flagellar serotypes [59,61]. While HU loops the invertible sequence [60,62,63], Fis enhances assembly of the supercoiling-dependent invertasome by binding to a 65-bp enhancer sequence (within the invertible sequence) [64]. The Fis/enhancer complex also determines the direction and rotation of the Hin synaptic complex [65]. It is possible that ComW enhances the formation and activity of the PsrA invertasome through its DNA-binding activity. Alternatively, binding of ComW to its target sequence (s) may regulate the hsdS inversions by altering the supercoiling state of the local sequences. It has been shown that negative DNA supercoiling is essential for the Hin-mediated inversions [60,66].
Multiple lines of experimental evidence have also revealed that RR06, RR08 and RR09 promote the O colony phase by modulating the direction of the PsrA-catalyzed hsdS inversions. The first set of the proof came from the impact of the psrA gene on the colony phenotypes of the rr06, rr08 and rr09 mutants. Although these mutants generated in the psrA + genetic background showed significantly reduced proportions of O colonies, the phase-locked O counterpart derived from the PsrA-deficient strain uniformly produced O colonies. Consistent with their reduced fractions of O colonies, the rr06, rr08 and rr09 mutants also exhibited diminished methylation for the DNA motif recognized by the "opaque ON" HsdS A1 MTase. Although the current data cannot explain how the pneumococci utilize multiple response regulators for regulation of hsdS inversion reactions, these factors must act in a highly coordinated manner in the context of environmental/cellular conditions. As exemplified with TCS11, further identification of the genes regulated by RR06, RR08 and RR09 under these conditions will be necessary for elucidation of the molecular mechanisms governing the actions of these TCSs in pneumococcal phase variation.
The existing information strongly suggests that pneumococcal phase variation in colony opacity is a gross reflection of pneumococcal physiological conditions, such as redox and metabolism. The linkage between pneumococcal colony phases and redox conditions is supported by significant impact of the orphan response regulator RR14 (RitR) on the colony phases. RR14 represses the transcription of the pneumococcal iron uptake (Piu) and nutrient uptake loci [39]. A recent study has shown that RR14 serves as a principal regulator in response to oxidative stress via the cysteine-mediated dimerization, which enhances pneumococcal tolerance to hydrogen peroxide and other oxidants [40]. S. pneumoniae is famous for its production of millimolar range of hydrogen peroxide as a metabolic byproduct of pyruvate oxidase SpxB [67,68]. Consistently, SpxB is under-expressed in the opaque colony variant [69]. Potential association of pneumococcal metabolism with colony phases agrees with the known regulation of the nutrient uptake systems by RR14 [39]. Moreover, significant impact of the bgaC and ugl sugar utilization systems on the O phase could be explained by their roles in cellular metabolism. Lastly, broad impact of the five TCSs on the colony phases implies that the appearance of pneumococcal colonies is defined by vastly different mechanisms/pathways.

Bacterial strains and reagents
The bacterial strains used in this work are listed in S5 Table. S. pneumoniae was grown in Todd-Hewitt broth with 0.5% yeast extract (THY) or on tryptic soy agar (TSA) plate as described [70]. Streptomycin (150 μg/ml) and kanamycin (400 μg/ml) were added in the

PLOS PATHOGENS
Regulation of pneumococcal phase variation medium when necessary. All chemical reagents used in this study were purchased from Sigma (Shanghai, China) unless otherwise indicated. DNA processing enzymes were purchased from New England Biolabs (Beijing, China). All primers were synthesized by Synbio Tech (Beijing, China) and are listed in S6 Table. All Sanger sequencing data were obtained from Ruibiotech company (Beijing, China).

Construction of pneumococcal mutants
Pneumococcal mutants were constructed in streptomycin-resistant strains ST606 (derivative of ST556, serotype 19F), TH6671 (derivative of P384, serotype 6A) and TH6675 (derivative of ST877, serotype 35B) essentially as described [5]. DNA templates, primers, restriction enzymes, resulting strains, genotypes and other details associated with mutant construction are listed in S7 Table. Briefly, JC1 (a modified Janus cassette used in our previous studies) replacement of the target sequences was generated as followed: the up-and down-stream arms of each target sequence were PCR amplified, digested with XbaI and XhoI, and ligated to a XbaI/XhoI-digested JC1 before being transformed into target pneumococcal strains. JC1 was amplified by primers Pr9840 and Pr1098 from genomic DNA of strain TH7919 (TH5445ΔbgaA::JC1) [5]. Unmarked deletions were subsequently constructed in the JC1 replacement strains by transformation with either ligation or fusion PCR products of the upand down-stream target sequences.
The genetic revertants were constructed by transformation of JC1 replacement strains with amplicon of the corresponding wildtype target genes from genomic DNA of ST606. Gene deletions in the psrA Y247A background were similarly constructed in strain TH6552 [5]. The point mutations in rr11 and hk11 were generated by introducing the mutations in the primers used for fusion PCR. Specifically, the codon of D53 (GAT) in rr11 was converted to GCG (A) and GAG (E), respectively; the codons of H184 (CAT) and T190 (ACC) in hk11 to GCC (A) and CCT (P), respectively. The bgaC, ugl, gtfA, rr11 and arcA locus deletion mutants were constructed by unmarked deletion of MYY134-139, MYY403-408, MYY1793-1796, MYY1924-1925 and MYY2067-2068 as indicated in S7 Table, respectively. Mutated sequence in each strain was confirmed by PCR amplification and DNA sequencing.

Microscopic quantification of O and T colonies
Pneumococcal colony opacity was microscopically assessed with colonies grown on catalase-TSA plates (6,000 units catalase/9-cm diameter plate) under 37˚C, 5% CO 2 for 17 hrs as described [71]. Each inoculum was adjusted according to its OD 620nm value to yield~200 colonies per plate for quantification of the O and T colonies for each strain. The ratio between the two types of colonies for each strain was obtained with triplicate plates each time, and subsequently repeated at least twice. The ratio of the progeny colonies generated from individual O and T seeding colonies was determined as previously described [5].

SMRT sequencing
Methylation of pneumococcal genome was assessed by SMRT sequencing essentially as described [5]. Genomic DNA was isolated from colonies grown on catalase-TSA plates using the HiPure Bacterial DNA Kit (Magen, Beijing, China) according to the manufacturer's protocol. Briefly, pneumococci were grown on the plates under 37˚C, 5% CO 2 for 17 hrs before the colonies were scraped off with a glass spreader and washed once with pre-chilled Ringer's solution. Bacterial lysis was achieved in 250 μl Buffer STE supplemented with 10 μl of 10% sodium deoxycholate (DOC) solution before the lysates were processed to remove RNA and proteins. The final DNA extracts were eluted with 35 μl of deionized distilled water and

RNA-seq
RNA sequencing (RNA-seq) was carried out as described with minor modifications [19]. Specifically, pneumococci were cultured on catalase-TSA plates for 17 hrs (a time for the routine colony photographing) before colonies on multiple plates were collected in pre-chilled Ringer's solution and pooled as described for genomic DNA purification. Bacteria in suspensions were pelleted by centrifugation at 4˚C, frozen in liquid nitrogen and stored at -80˚C. Total RNA was extracted from the frozen samples with the Purelink™ RNA Mini Kit (Invitrogen, USA) and further purified with Qiagen RNeasy MinElute spin columns according to the manufacturer's protocol (Qiagen, Germany). RNA-seq was performed at the Novogene Bioinformatics Technology (Beijing, China). Trimmed reads were mapped to the genome of S. pneumoniae ST556 (CP003357.2) using Bowtie 2.3.1 and Tophat 2.1.1. Significant difference was defined by an at least 2-fold change and a P value of < 0.001. All of the raw RNA-seq data presented in this work are available in NCBI's Gene Expression Omnibus (GEO) database (accession GSE137447). The result of each sample represents the means of two independent experiments.

qRT-PCR
Pneumococcal mRNAs were quantified by quantitative real-time reverse transcriptase PCR (qRT-PCR) as described [72]. Briefly, total RNA extracts were prepared as described for RNAseq and used to prepare cDNA pools with iScript™ cDNA Synthesis Kit (Bio-Rad, USA). The 367-bp hsdS A1 allele-specific sequence was amplified with primers Pr16174/Pr16175. As an internal reference for PCR, the 5' non-invertible region shared by the six hsdS A alleles was also amplified using primers Pr16178/Pr16179. The relative abundance of hsdS A1 mRNA was obtained by a two-tier approach. The ΔC T value for each strain was firstly calculated by subtracting the average C T value of the non-invertible hsdS A reactions from the C T values of the hsdS A1 -specific reactions. Because the hsdS A1 -specific sequence (367 bp) was substantially longer than the common hsdS A region (267 bp), the ΔC T value of each strain was further normalized to the counterpart of psrA Y247A by subtracting the average ΔC T of psrA Y247A from that of each strain. Our previous study showed that the loss-of-function mutation in psrA made the psrA Y247A strain genetically locked in the hsdS A1 allelic state [5], and should produce only the hsdS A1 mRNA. The relative abundance of hsdS A1 mRNA of each strain is presented as (2 -ΔΔCT ) % given that the relative abundance of hsdS A1 mRNA in psrA Y247A is 100%. The transcriptional levels of RR11-regulated genes were detected by qRT-PCR with the era gene as an internal control and primers listed in S8 Table. The era gene was amplified with primers Pr7932/ Pr7933, which is commonly used as an internal control [72]. The relative gene expression was

PLOS PATHOGENS
Regulation of pneumococcal phase variation calculated according to the comparative 2 -ΔΔCT method [73] and the ΔΔC T was calculated using the following equation: ΔΔC T = (C T gene of interest -C T era) mutant-(C T gene of interest-C T era) ST606. The data from one representative experiment are presented as mean value of triplicate samples ± SEM (standard error of mean) for each strain. Each experiment was repeated at least twice.

Statistical analysis
The colony ratio data was statistically analyzed by two-sided Chi-square test (means); qRT-PCR, relative abundance of hsdS A1 mRNA data by two-tailed unpaired Student's t test.
Supporting information S1

PLOS PATHOGENS
Regulation of pneumococcal phase variation orientations of the genes in the six RR11-regulated gene loci are indicated with arrowheads; each gene identified with its functional or genomic names below; the number of nucleotides between two adjacent genes marked at the intergenic region. The genes deleted for mutagenesis were shown in bold. B. Transcription of the RR11-regulated genes in the Δrr11 mutant. Transcriptions of comW and the genes in the loci of MYY134-139, MYY403-408, MYY1791-1796, MYY1923-1925 and MYY2067-2068 in the ST606 (WT) and Δrr11 strains were detected by qRT-PCR. Relative transcriptional difference of each gene in the rr11 mutant is calculated by normalizing the C T value of each gene to that of the parental strain. (TIF)