Streptococcus pneumoniae TIGR4 Phase-Locked Opacity Variants Differ in Virulence Phenotypes

A growing number of bacterial species undergo epigenetic phase variation due to variable expression or specificity of DNA-modifying enzymes. For pneumococci, this phase variation has long been appreciated as being revealed by changes in colony opacity, which are reflected in changes in expression or accessibility of factors on the bacterial surface. Recent work showed that recombination-generated variation in alleles of the HsdS DNA methylase specificity subunit mediated pneumococcal phase variation. We generated phase-locked populations of S. pneumoniae TIGR4 expressing a single nonvariant hsdS allele and observed significant differences in gene expression and virulence. These results highlight the importance of focused pathogenesis studies within specific phase types. Moreover, the generation of single-allele hsdS constructs will greatly facilitate such studies.

A major limitation in the field of pneumococcal phase variation research is the lack of genetically "locked" strains that are incapable of switching phenotype. Typical studies use bacterial populations enriched for each phenotype via selection of a single colony and serial passage until the majority of colonies are uniformly one phenotype (8). While useful, such strains are not genetically "locked" into a particular phase and can freely switch back and forth. Recent work has shown that pneumococcal phase variation of colony opacity occurred via site-specific recombination of the three methylase sequence specificity genes (hsdS, hsdS=, and hsdS==) in the SpnD39III and Spn556II type I restriction-modification (R-M) systems (19,20). Specifically, recombination of the three genes produced six predicted hsdS alleles that can generate six bacterial subpopulations with distinct colony phenotypes, virulence, DNA methylation patterns, and changes in gene expression (19,20).
Like many other bacteria, pneumococci encode multiple R-M systems to provide a key defense mechanism against invasion of foreign DNA (e.g., bacteriophage DNA) and protection of host DNA (via methylation) from endogenous restriction enzyme cleavage. The pneumococcal SpnD39III and Spn556II type I R-M systems include genes hsdS, hsdM, and hsdR and a site-specific recombinase gene. The hsdS, hsdM, and hsdR genes are cotranscribed into protein subunits HsdS, HsdM, and HsdR, which assemble into a heteromeric enzyme that can methylate and cleave double-stranded DNA. The HsdS subunit is the determinant of DNA methylase sequence specificity and has two different DNA target recognition domains separated by conserved regions that recognize "split" sequences (e.g., 5=-CRAAnnnnnnnnCTG-3=) (19). The HsdM subunit is responsible for DNA methylation at the appropriate adenine within the DNA recognition sequence (see the underlined adenine in the sequence example). The HsdR subunit is responsible for DNA cleavage and translocation through the bound heteromeric complex (21). This hsd type I R-M system is used by many bacterial pathogens, including Escherichia coli, Mycoplasma pneumoniae, Staphylococcus aureus, S. pneumoniae, and others (19,(22)(23)(24). In this study, we sought to assess the role that the six predicted hsdS alleles might play in colony phenotype, gene expression, and pathogenesis using a mouse model for nasal colonization. The S. pneumoniae TIGR4 genetic background strain was chosen since it is a commonly used model strain for which genomic sequence data are available (25).

RESULTS
Pneumococcal hsdS genetic loci. Examination of the six published hsdS allele sequences (A, B, C, D, E, and F, corresponding to GenBank accession numbers KJ955483, KJ955484, KJ955485, KJ955486, KJ398403, and KJ398404, respectively) and comparative analyses of their translated protein sequences led us to define the coding sequence for each HsdS target recognition domain (TRD). The pneumococcal strain D39 HsdS protein was divided into four sections: inverted repeat 1R (IR1R; codons 1 to 92), TRD 1.1 (codons 93 to 232), IR2R (codons 234 to 336), and TRD 2.1 (codons 337 to 522). With this information in hand, we annotated the homologous hsdS genetic locus in pneumococcal strain TIGR4 (GenBank accession number AE005672.3) (see Table S1 in the supplemental material). Like S. pneumoniae D39, S. pneumoniae TIGR4 harbored all the TRDs necessary to produce the six predicted hsdS alleles (Fig. 1), and their DNA sequences were 100% identical. The only coding mutation identified was a single amino acid substitution (Ile11Val) in the IR1R region of the hsdS gene. The orientations and locations for TRD 2.1, TRD 2.2, TRD 2.3, and the site-specific recombinase creX differed in S. pneumoniae TIGR4. Additionally, the last 47 bp of TRD 2.3 overlapped the last 47 bp of the inverted TRD 2.2 sequence, making mutational studies particularly challenging. Examination of other published pneumococcal genomes revealed that the six TRDs identified in pneumococcal strain D39 (GenBank accession number CP000410.1) were represented in other strains (data not shown). Their gene orientations and locations were either highly similar to those seen with strain D39 (e.g., strain AP200; GenBank accession number CP002121.1) or quite different (e.g., strain 70585; GenBank accession number CP000918.1), highlighting the genetic variability of this region.
Creation of S. pneumoniae TIGR4 hsdS variants. Strain MBO15 was created by replacing all genes located between S. pneumoniae TIGR4 SP_0504 and hsdM with a Janus cassette (26,27). Overlap extension PCR was used to create the six predicted hsdS alleles, which were then cloned in a location adjacent to spectinomycin resistance gene aad9 and to flanking genes glnA, SP_0503, and SP_0504 (Fig. 2) Tables S1 and S2. Impact of individual hsdS alleles on colony phenotype. Pneumococcal colony phenotypes are determined by visualizing colonies on a translucent agar under oblique lighting (8). Generally, opaque colonies appear as domes of solid color, while transparent colonies appear as colonies that are clear or that have a dense center and translucent halo (Fig. 3A). The colony phenotypes for recombinant hsdS strains A to F were determined using strain S. pneumoniae TIGR4 and its unencapsulated derivative TIGR-JS as an opaque control and a transparent control, respectively. Briefly, bacterial stocks were grown on nonselective Trypticase soy agar supplemented with catalase for 18 to 20 h. At least 50 colonies were visualized under a dissection microscope, and their phenotypes were recorded. Strains S. pneumoniae TIGR4, A, and B had larger, 100% opaque colonies, while strains C, D, E, F, and TIGR-JS had smaller, 100% transparent colonies (Table 1). No mixed populations or colony variants were observed for any of the strains.
In order to quantitatively assess and compare colony phenotypes, at least 100 individual colonies of each strain were visualized with phase-contrast microscopy and photographed and their diameters measured (see Materials and Methods). Within each strain, the colonies were nearly identical in size ( Fig. 3B) and had nearly identical diameters (Fig. 3C). Interestingly, the average colony diameters were very similar between the opaque strains S. pneumoniae TIGR4, A, and B and also between the transparent strains C, D, E, and F (Fig. 3C). The opaque and transparent colony sizes differed by~25%. In order to test whether colony phenotypes were stable on passage, the colonies imaged in Fig. 3 were harvested and passed at least five consecutive days in vitro. All strains maintained their colony phenotype (data not shown). Together, these findings show that the recombinant hsdS variants were genetically "locked" in the opaque (strains A and B) or transparent (strains C, D, E and F) phase.
The phase phenotype did not affect adherence to human epithelial cells. In a recent study, phase-locked strains in S. pneumoniae ST556 (representing a serotype 19F background) were shown to differ in their levels of adherence to immortalized epithelial cells; in particular, the opaque phase had reduced adherence to human lung (A549) and nasopharyngeal (Detroit 562) epithelial cell lines (20). This is consistent with many previous studies using enriched populations (9,14,(28)(29)(30). To determine if our phaselocked strains had similar differences in adherence, bacterial adherence studies were performed essentially as described previously (30,31) using immortalized human lung (A549) and human bronchoepithelial (HPE14 and CFBE-4o) cell lines. Surprisingly, we did not observe significant differences in adherence (data not shown).
Opaque-phenotype variants had diminished biofilm formation. Opaque colony variants have decreased adherence to host cells, animal nasal surfaces, and plastic presumably due to overexpression of capsular polysaccharide, which can mask subcapsular ad- hesins (9,30,32). To determine whether our phase-locked mutant strains differed in capsule expression, we performed enzyme-linked immunosorbent assays. We found that the hsdS variants produced less capsule than S. pneumoniae TIGR4 and that the transparent variants produced less capsule than the opaque strains ( Fig. 4). To determine whether the strains differed in their ability to form biofilms in vitro, we performed static biofilm assays essentially as previously described (30). Total biomass and bacterial viability within biofilms were assessed by crystal violet staining and by conventional colony counting, respectively, at the 4 h and 24 h time points. Although no difference in the levels of biomass was detected at either time point (data not shown), we observed diminished bacterial counts within biofilms for the opaque strains (S. pneumoniae TIGR4, A, and B) after 24 h (Fig. 5). On the basis of these results, we conclude that opaque variants have a diminished capacity to survive within mature biofilms. Colonization, persistence, and virulence of phase-locked mutants. Nasal colonization studies were conducted with S. pneumoniae TIGR4; opaque variants A and B; and transparent variants C, D, E, and F. Briefly, BALB/c mice were infected intranasally with~10 6 CFU, and at 3, 5, 7, and 14 days postinfection, groups of mice were euthanized and the nasopharynx and right lung lobe were excised, homogenized, and plated on blood agar containing gentamicin (4 g/ml) for bacterial counts. All mice infected with the parental S. pneumoniae TIGR4 strain showed nasal carriage that did not result in bacteremia, which is consistent with our prior infection studies. Remarkably, only mice infected with transparent strains C and D appeared visibly sick and had considerable weight loss (data not shown). The mortality rates differed for each of the phase-locked strains (Fig. 6A). All strains maintained nasal colonization for 14 days (Fig. 6B). The opaque strains had a small colonization defect on day 3 and a reduced bacterial load on day 14. Interestingly, transparent strains C and D were more virulent and caused the majority (61.5%) of the lung infections (Fig. 6C), particularly at the early time points. The levels of the lung infections decreased steadily, with no lung infections detected on day 14.  Although strain F had stable colonization, infection with this strain did not progress to lung infection ( Fig. 6B and C). On the basis of these results, we conclude that S. pneumoniae phase variation mediated by hsdS alleles C and D had a significant impact on pneumococcal colonization, persistence, and virulence.
Phase-locked variants maintained phenotype for 2 weeks in vivo. Previous in vitro studies found that serial passage of the phase-locked hsdS mutants in vitro did not significantly affect colony size or phenotype (data not shown). Therefore, on days 7 and 14, nasal and lung tissue homogenates from the animal experiments were plated on Trypticase soy agar containing gentamicin for determinations of colony phenotype. At least 50 individual colonies were imaged ( Fig. 7A and C), and their diameters were measured. The average colony diameters were nearly identical for the S. pneumoniae TIGR4, A, and B strains and for transparent strains C, D, E, and F ( Fig. 7B and D). The opaque and transparent colony sizes differed by~39%. These findings showed that all strains tested were phase-locked and could maintain their colony phenotype after passage in an animal for 14 days.
HsdS-dependent changes in gene expression. Bacterial DNA methylation can result in epigenetic changes that can affect gene expression. Previous studies of S. pneumoniae D39 and S. pneumoniae ST556 and their phase-locked variants determined that each hsdS allele was associated with distinct DNA methylation patterns and with altered virulence gene expression (19,20). For example, the S. pneumoniae D39 variant that harbored hsdS allele B had reduced expression of the luxS gene and the capsular polysaccharide synthesis operon (19). In this study, RNA sequencing was used to determine whether our panel of hsdS variants had unique transcriptomes. Data sets were deposited in the NCBI GEO database (accession number GSE103364).
Changes in gene expression compared to that seen with S. pneumoniae TIGR4 hsdS deletion mutant MBO15 were detected for all strains. The data were searched for a set of genes that had similar altered expression between the transparent (C, D, E, and F) and opaque (A and B) strains. Although no such genes were detected for the transparent strains, opaque strains A and B had reduced expression of 10 genes, with about half being associated with glutamine biosynthesis and ABC transporters (  (33). Strain C had increased expression of three genes (SP_1383, SP_2175, and SP_2176) known to be involved in D-alanylation of cell wall teichoic acids. Finally, strain D had reduced expression of six genes (SP_0957, SP_2002 and SP_2003, and SP_1796 to SP_1798) that encoded ABC transporters. Strains B and E had differential gene expression whereas strain F had no altered gene expression compared to strain MBO15. Thus, we concluded that the six hsdS alleles altered gene expression of potential virulence factors in the different phase-locked strains. Four genes were selected for validation with real-time PCR (Table S4). While most of the transcriptome sequencing and reverse transcription-PCR data were similar, incongruent data were observed for strains TIGR4 (genes SP_0148 and SP_2176), E (SP_0148 and SP_2176), and F (SP_1797).

DISCUSSION
Research in the field of pneumococcal phase variation has been hampered by the lack of availability of phase-locked strains. Recent genetic findings revealed that site-specific recombination of the hsdS DNA methylase targeting subunit in a type I restriction-modification locus could result in six distinct phase-locked subpopulations with different colony opacity phenotypes and gene expression characteristics (19,20). In this study, we used a genetic approach to produce six phase-locked subpopulations of S. pneumoniae TIGR4 and to examine their colony sizes, ability to form biofilms, attachment to host cells, and persistence in nasal colonization. Analysis of the phenotypes could provide insights into the mechanism(s) (e.g., contribution of phase-specific virulence factors) that facilitates the adaptation of pneumococci to different host environments.
This study demonstrated that recombinations of the hsdS gene in the S. pneumoniae TIGR4 type I restriction-modification system resulted in six distinct bacterial population derivatives that were 100% phase-locked in the opaque (strains A and B) or transparent (strains C, D, E, and F) phenotype. Importantly, our hsdS variants maintained their phenotype over multiple consecutive passages in vitro and for 2 weeks in vivo. Unlike those reported for S. pneumoniae D39 hsdS derivatives (19), our colony phenotype determination did not detect mixed bacterial populations for any S. pneumoniae TIGR4 hsdS derivative. These findings were supported by another study which produced the same six recombinant hsdS derivatives in several different strain backgrounds, including S. pneumoniae TIGR4, and determined that they were 100% phase-locked (20). One major difference between our studies was that the opaque phenotype was linked only to expression of hsdS allele E in reference 20 but was linked to hsdS alleles A and B in this study. Another major difference was that their S. pneumoniae TIGR4 strain and its six hsdS derivatives all produced 100% opaque colonies (20). Although this supported the finding that our S. pneumoniae TIGR4 was 100% opaque, it is unclear why our results differed in colony phenotype determinations. One explanation for these differences could simply be the genetic and phenotypic diversity of the strains (34,35). Epigenetic changes in gene expression could alter certain surface-expressed moieties that may be missing in the genomes of different strains. Despite these differences, we can conclude that each hsdS-locked genotype limited intrastrain variation to a single phenotype in this study.
Transparent variants have been shown to have increased biofilm formation and adherence to epithelial cells (9,14,(28)(29)(30)32). While we saw no differences in adherence to immortalized epithelial cells or in the early stages of biofilm formation, the opaque phase-locked variants (A and B) showed a significant decrease in bacterial counts within biofilms at later stages. It is possible that the opaque strains had altered expression of surface factors that altered their adherence to plastic surfaces. Although the phase phenotype affected biofilm formation and viability, it played a less significant role in adherence to host cells and stable colonization.
Bacterial virulence has been linked to opaque colony phenotypes (8). On the basis of previous studies, we expected that the opaque strains would have a colonization defect but increased invasiveness. Unexpectedly, we found that transparent phaselocked strains C and D were hypervirulent and caused the majority of lung infections. This finding does not seem to fit with the results seen in the field, where transparent variants are considered generally nonvirulent. One explanation may be that it is difficult to compare a phase-locked strain population to heterogeneous populations. Another explanation may be that the phase-locked strains differed in gene expression.
It has been demonstrated that hsdS derivatives in S. pneumoniae D39 (19) and six other strains (20) had distinct genome methylation patterns with the exact same methylation sequences for each allele. For example, both studies found that hsdS genes encoding TRD 1.1 and TRD 2.2 methylated the same adenine in the motif 5= CRAANN NNNNNNCTT 3=. DNA methylation is known to result in epigenetic changes that can alter gene expression. We performed RNA-seq studies to help to identify potential virulence factors that may differ in each phase. We found that all strains in our study had differential gene expression compared to S. pneumoniae TIGR4 hsdS locus deletion strain MBO15. ABC transporter gene SP_0148 was the most highly downregulated gene in opaque strains A and B. Interestingly, deletion of SP_0148 in S. pneumoniae TIGR4 was shown to attenuate virulence in pulmonary infections (36). This may help explain why opaque variants A and B were less virulent than the transparent strains. Of the 18 genes upregulated only in variant C, 3 (SP_1383, SP_2175, and SP_2176) were involved in D-alanylation of cell wall teichoic acids. This type of cell wall modification has been shown to increase Gram-positive bacterial resistance to antimicrobial peptides (via an increase in the net surface positive charge) (37). This finding may help to partially explain why variant C was so virulent in vivo. Notably, variant D showed reduced expression of six ABC transporter genes (SP_0957, SP_2002 and 2003, and SP_1796 to SP_1798). The S. pneumoniae TIGR4 genome encodes 73 ABC transporters (38), and mutation of genes SP_1796 to SP_1798 in TIGR4 was shown to not affect virulence in vivo (36). Therefore, reduced expression of that ABC transporter in this study may have had only a minimal effect during the course of the infection.
One potential limitation in this study was that we created single-allele hsdS derivatives of only one strain, S. pneumoniae TIGR4. However, with our recombinant hsdS constructs in hand, it will be straightforward to produce variants in other pneumococcal backgrounds. Another limitation is that we did not assess DNA methylation patterns for each variant. On the basis of our findings, we hypothesize that, as has been shown in previous studies (19,20), the DNA methylation patterns likely differ in the recombinant hsdS variants.
Since we have phase-locked strains that differ in colony size, biofilm formation, and persistence in vivo, it would be interesting to investigate the contribution of specific virulence factors to each phenotype. By deleting and complementing various factors within the phase(s) in which they are predominantly expressed, we have the potential to more directly and specifically address contributions to virulence or virulence-related phenotypes. Further, since genes glnA and hsdM are conserved among different pneumococcal strains, the same genetic constructs used in this study can be transformed into multiple genetic backgrounds. This would be useful to study the contribution of certain virulence factors in a serotype-specific manner.
Epigenetic phase variation associated with variable expression or the specificity of DNA methylase is emerging as a common theme for generation of phenotypic diversity in bacterial populations. Our construction of single-allele hsdS mutant substrains offers the possibility of refining molecular pathogenesis work by specifically studying virulence factors within one, or a few, phase types. This study demonstrated that the phase-locked S. pneumoniae TIGR4 hsdS strains offer a useful model for the study of the contribution of colony phenotype to multiple aspects of host-pathogen interactions.

MATERIALS AND METHODS
Pneumococcal culture conditions. A list of all pneumococcal strains used in this study is provided in Table 1 in the supplemental material. Bacteria were cultured in a humid atmosphere at 37°C in 5% CO 2 . Pneumococci were grown on tryptic soy agar (Difco, BD Diagnostics) supplemented with 5% sheep's blood (Hemostat) and 4 g/ml gentamycin (Sigma) or in Todd-Hewitt broth (Difco, BD Diagnostics) supplemented with 0.5% yeast extract (THY medium; Difco, BD Diagnostics) at 37°C. Freezer stocks were made in 18 to 20% glycerol.
Construction of six recombinant hsdS alleles. The recombinant hsdS alleles were produced via overlap extension PCR using S. pneumoniae TIGR4 genomic DNA, primers F1-F10 and R1-R8 (see Table S2 and Fig. S1 in the supplemental material), and TaKaRa Ex Taq polymerase (Clontech). The primer pairs, DNA targets, and expected PCR amplicon sizes are shown in Table S3. For example, to create hsdS allele A (TRD 1.1 and 2.1), PCR amplicon "A1" (gene target TRD 2.1) was produced with primer pair F4/R3 (~0.5 kb), and "A2" (IR2R through hsdM) was produced with primer pair F8/R8 (~2.3 kb). The "A1" and "A2" PCR amplicons were mixed together (50 to 100 ng each) and amplified with primer pair F4/R7 (~2.6 kb) to produce the final hsdS allele, allele A. Recombinant hsdS alleles B through F were similarly produced with their specific primer pairs. All PCR amplicons and final constructs were gel purified to reduce the possibility of parental chromosomal contamination during the overlap extension PCR procedure.
for colony imaging. Plates were incubated for 18 to 20 h at 37°C with 5% CO 2 . All animal experiments were approved by the University of Alabama at Birmingham Institutional Animal Care and Use Committee (Animal Project number IACUC-20589).
Gene expression analyses. For each pneumococcal strain, three independent cultures were grown in THY medium to mid-log phase (OD 600 of 0.8), centrifuged (12,000 rpm for 5 min), and resuspended in 1 ml PBS containing lysozyme (2 mg/ml). Total RNA was extracted using a NucleoSpin RNA II kit (Macherey-Nagel, Germany) per the manufacturer's recommendations. Frozen (Ϫ80°C) RNA samples were sent to the UAB Heflin Center for Genomic Sciences for RNA-seq analyses using an Illumina HiSeq 2500 platform (Illumina). The rRNA was removed from the total bacterial RNA using ribosome reduction for Gram-positive bacteria (ArrayStar, Rockville, MD). The purified RNA was then processed using a SureSelect strand-specific RNA-Seq library prep kit from Agilent Technologies (San Diego, CA) following the manufacturer's protocol. The resulting libraries were quantitated with quantitative PCR (Kapa Biosystems, Woburn, MA), and the size distributions of the insertions were checked using an Agilent 2100 BioAnalyzer and a high-sensitivity DNA chip. The library concentration was normalized to 2 nM, and sequencing on a HiSeq 2500 platform with paired-end 50-bp sequences was performed under standard conditions. Sequencing generated 61-bp single-end RNA-Seq reads with an average depth of 16.8 ϩ 2 M reads. The sequencing depth translated to~475ϫ coverage (based on the length of the S. pneumoniae TIGR4 genome [2.1 MB]). The quality of FASTQ files were checked using tool FASTQC. Reads were mapped to the reference S. pneumoniae TIGR4 genome (accession number NC_003028.3) using Bowtie2 (96% reads mapped), and transcript abundance was calculated using the summarizeOverlaps function in R package Genomi-cAlignments. Finally, differential expression analyses were carried out using R package DeSeq2. All data were compared to data from strain MBO15 (an S. pneumoniae TIGR4 hsdS deletion variant). The RNA-seq data were deposited in the NCBI GEO database (accession number GSE103364).
Real-time PCR. Target genes SP_0806, SP0148, SP_1797, and SP2176 were selected for validation. Custom oligonucleotides and probes conjugated to 6-carboxyfluorescein were designed using the IDT PrimerQuest Tool (Table S3). Real-time PCR was performed using a TaqMan RNA-to-Ct 1-Step kit (Applied Biosystems catalog no. 4392938) in triplicate 20 l reaction mixtures containing 2ϫ master mix, a 900 nM concentration of each primer, 100 nM fluorescently labeled probe, and RNA at a final concentration of 10 ng/l. Amplification and detection were performed using a QuantStudio 3 real-time PCR system (Thermo Scientific). PCR conditions were 48°C for 2 min and 95°C for 10 min, followed by 40 cycles of 95°C for 15 s and 60°C for 1 min. ROX1 was used as the passive reference dye. The transcription level of each gene was normalized to the gyrB reference gene (41), and the results were analyzed using the comparative threshold cycle (C T ) method. The experiments were repeated twice.