A Quorum-Sensing System That Regulates Streptococcus pneumoniae Biofilm Formation and Surface Polysaccharide Production

Quorum sensing regulates bacterial social behaviors by production, secretion, and sensing of pheromones. In this study, we characterized a new quorum-sensing system of the Rgg/SHP class in S. pneumoniae D39. The system was found to directly induce the expression of a single gene cluster comprising the gene for the SHP pheromone and genes with putative functions in capsule synthesis. Capsule size, as measured by dextran exclusion, was increased by SHP exposure in R36A, an unencapsulated derivative of D39. In the encapsulated parent strain, overexpression of the gene cluster increased capsule size, supporting the role of Rgg/SHP in the synthesis of surface polysaccharides. Further, we found that biofilm formation on epithelial cells was reduced by overexpression of the system and increased in a mutant with an rgg deletion. Placing surface polysaccharide expression under quorum-sensing regulation may enable S. pneumoniae to tune interactions with the host and other bacteria in accordance with environmental and cell density conditions.

A detailed promoter analysis and a map of the Rgg0939/SHP system in D39 are presented in Fig. S1 in the supplemental material. The predicted binding site for Rgg in the shp promoter of D39 is similar to those previously described for shp3 and shp2 in S. pyogenes and shp1555 in S. agalactiae (16,17). Furthermore, we assessed the presence of this system in other strains of S. pneumoniae by using Rgg0939 as a template and found that 12 (36.4%) of the 33 fully sequenced strains of S. pneumoniae available in the NCBI database carry an Rgg0939-homologous protein with Ͼ86% identity (Table 1). Remarkably, Rgg2 and Rgg3 in S. pyogenes (15) and RovS in S. agalactiae (16) are conserved in nearly all of the strains reported. In draft genomes of S. pneumoniae at the NCBI (7,415 databases), there are 2,343 hits for homologs of Rgg0939 with at least 70% identity and Ͼ70% coverage. Among these, there are 2,308 and 1,765 hits with Ͼ85 and 90% identity, respectively. This indicates that the system is present in a significant number of pneumococcal strains. Further, relevance is added as interspecies communication has been shown to occur in Rgg/SHP systems (21,22). SHP3 in S. pyogenes is, in fact, identical to SHP0939 produced by S. pneumoniae and only one amino acid different from the SHP produced by S. agalactiae. This pheromone has been shown to allow cross-communication with S. agalactiae and even with S. mutans, which has an SHP with a more divergent sequence (21).
The SHP pheromone acts as an autoinducer via Rgg. Pheromone systems most frequently form a positive feedback loop with the regulator, stimulating a fast response and high expression of the pheromone. To investigate whether the putative SHP pheromone sequence upstream of Rgg forms an active autoinducing system, we measured shp mRNA expression in cultures treated with the predicted mature pheromone (DIIIIVGG) and in untreated samples. The full ORF in S. pneumoniae encodes 23 amino acid residues (MKKISKFLPILFLVMDIIIIVGG), and on the basis of studies of S. pyogenes (23), the highest inducing activity is seen in peptides comprising the C-terminal 8 amino acids. Gene expression was nearly 20-fold higher in cultures treated with the specific S. pneumoniae pheromone (Fig. 1), indicating that this is indeed a functional autoinducing system. Rgg/SHP complexes most often activate transcription by binding to the promoter sites that they regulate; however, a previous study showed that the Rgg3 protein of S. pyogenes can also function as a transcriptional repressor (15). In this case, once the pheromone is imported into the cell, it binds to Rgg and removes the protein from the repressor position. Given the high similarity of Rgg0939 to Rgg3 of S. pyogenes, we investigated whether Rgg0939 would also act as a transcriptional repressor of shp. Interestingly, we found that deletion of Rgg0939 does not upregulate shp expression (Fig. 1). In addition, treatment of the Δrgg0939 mutant strain (SP068) with the synthetic pheromone showed no upregulation of shp compared to the control culture, and upon complementation of the strain with rgg0939 (SP101), upregulation was restored. This indicates that Rgg0939 acts as an activator, rather than as a repressor.
A single transcript initiated at shp and extending to 11 downstream genes is upregulated by the Rgg/SHP signaling system. To identify which genes this pheromone system regulates in S. pneumoniae, we assessed the transcriptome response in cultures treated with synthetic SHPs (sSHPs) and untreated controls. For selection of optimal conditions for the sSHP response, we tested the expression of shp during growth in two different media, tryptic soy broth (TSB) and CϩY (24), by real-time PCR. We found that induction of shp by sSHP compared to untreated samples was slightly higher in CϩY than in TSB, particularly at late exponential phase (Fig. S2). The expression of both shp and rgg increased at the stationary growth phase, which seemed to account for the reduced differences in shp expression among treated and untreated samples at this phase. Thus, we decided to collect the samples at late exponential phase by using cultures grown in CϩY. The global mRNA response is shown in Fig. 2.
A group of 12 genes, comprising 12.7 kb extending from position 949,424 to position 962,125 in the genome of D39, was highly upregulated in the treated culture. The database of prokaryotic operons (DOOR) (25) predicts that the region between SPD_0940 and SPD_0948 forms a single operon, with no candidate terminators, and the MicrobesOnline database predicts an operon from SPD_0940 to SPD_0949 (26). While our data cannot exclude the possibility that they actually form a separate operon, it was clear that upon sSHP stimulation, the transcript extended all the way from shp to SPD_0950 ( Table 2). The map of mRNA expression levels in this region shows a continuous transcript corresponding to the forward strand, whereas in the control culture without sSHP, this transcript was virtually absent (Fig. 3). The 11 genes downstream of shp comprise a variety of functions and annotations, such as MnaB (SPD_0940), UDP-4-galactose-epimerase (SPD_0941), a putative xylose isomerase (SPD_0942), an AMP-binding enzyme (SPD_0945), membrane proteins, and trans- ferases. Homologous genes such as those for MnaB, UDP-4-galactose-epimerase, AMPbinding enzymes, membrane proteins, and transferases were also identified downstream of shp in other strains of S. pneumoniae (Table 2). In addition, other genes with distinct functions such as lantibiotic and bacitracin transport, lanthionine biosynthesis protein, and lactococcin processing and transport were identified (Table 1).
To further investigate the function of the Rgg/SHP regulon, we created an overexpression mutant (SP044) by deleting the regulator rgg0939 and inserting a kanamycin resistance cassette without a terminator into the opposite strand (Fig. S3). We confirmed the expected expression profile of the Δrgg (SP068) and Over-EXP (SP044) mutants by reverse transcription (RT)-PCR with specific primers for shp, SPD_0940, SPD_0947, and SPD_0950 (Fig. 4). No growth differences between the mutant and wild-type strains were observed in three different media, TSB, CϩY, and CDMϩϩ (27) (Fig. S4).
Production of surface polysaccharides is increased in the overexpression mutant. The upregulation of genes putatively involved in polysaccharide biosynthesis upon the addition of sSHP led to the hypothesis of a possible role for the Rgg/SHP pheromone system in surface polysaccharide production by S. pneumoniae D39. Among the 11 genes in the Rgg0939 regulon, 2 (SPD_0940 and SPD_0941) seem to have a putative role in polysaccharide production on the basis of a homology search.  SPD_0940 was ϳ155-fold upregulated by sSHP and presents 27.2% identity to the gene encoding the enzyme MnaB, an essential component of the MnaA/MnaB biosynthetic pathway in pneumococcal capsule production of serotypes 12F and 12B. Furthermore, it has 26% identity to a gene in the capsule (cps) operon of Streptococcus suis (cps3N). SPD_0941 presented an upregulation of ϳ15-fold, and a homology search revealed 30% identity to GalE of Escherichia coli, a UDP-galactose-4-epimerase that interconverts UDP-glucose (UDP-Glc) and UDP-galactose (UDP-Gal) and is present in the cps locus of Staphylococcus aureus serotypes 5 and 8 (28). The serotype 2 cps locus encodes five type-specific transferases, one initiating transferase that adds the first sugar to a lipid receptor (cps2E) and four glycosyltransferases (cps2T, cps2F, cps2G, and cps2I) that further link sugars to create repeat units. The two putative transferases present in the Rgg0939 regulon (SPD_0944 and SPD_0949) seemingly do not present homology with the glycosyltransferases present in the cps locus; however, they are similar to enzymes present in cps loci of E. coli (wbbj, 36% identity) and S. agalactiae (neuD, 31% identity).
To examine whether the Rgg/SHP system is involved in surface polysaccharide production, we used the fluorescein isothiocyanate (FITC)-dextran assay, where the zone of exclusion of FITC-dextran indicates polysaccharide thickness (29)(30)(31)(32)(33)(34). A D39 derivative strain without the capsule (R36A) was included as a negative control. R36A originates from experiments reported by Avery et al. (35) and was obtained by 36 serial passages of D39 growing in the presence of anti-type 2 rabbit serum. The sequence of the capsule locus in this strain shows a deletion of ϳ7.5 kb, including the first nine genes of the capsule operon (cpsABCDETFGH) (36). The area occupied by R36A, as determined by light microscopy, was similar to the calculated area in the FITC-dextran assay using fluorescence microscopy. This confirmed that under the conditions used, no surface polysaccharide was produced by the unencapsulated strain (Fig. 5). The exclusion area was larger for D39, as expected for an encapsulated strain, and was not different from that of the Δrgg0939 mutant. In contrast, a significantly larger mean area of FITC exclusion per cell was observed with the overexpression strain (Fig. 5). The results indicate that the overexpression strain produces a thicker surface-associated polysac- charide, providing support for the hypothesis that the Rgg/SHP-regulated genes are involved in surface polysaccharide production. Inactivation of the Rgg/SHP system enhances biofilm formation, whereas overexpression of the Rgg regulon has the opposite effect. In S. pneumoniae, the capsule is thought to block surface proteins important for adhesion, thus impairing biofilm formation (37). To investigate whether the Rgg regulon, which is involved in the synthesis of surface polysaccharides, as demonstrated above, is involved in biofilm formation, we tested the ability of D39 and the deletion and overexpression mutants to form biofilms by using a eukaryotic lung cell line (A549) as the substrate. The overexpression strain formed less biofilm on the biotic surface than D39 did (Fig. 6), while the Δrgg mutant formed significantly more. Such findings corroborated our hypothesis that overexpression of the regulon would impair biofilm formation because of the increased production of surface polysaccharides. Also, the difference in biofilm formation between the rgg deletion mutant and D39 indicates that the system is active under biofilm growth conditions.
Overexpression of the Rgg/SHP regulon impairs fitness in a mouse model of lung infection. We further asked if changes in surface polysaccharide expression could also affect S. pneumoniae fitness in the host. We observed that the overexpression strain had a reduced ability to cause early lung disease (Fig. 7). To assess if the impaired ability of the overexpression strain to form biofilms and cause disease could be affected by its fitness compared to that of the parent strain (38), we inoculated mice with D39 and the overexpression strain. The overexpression strain was recovered in lower numbers from nasal washes than D39 (Fig. 7C). Overall, competitive assays indicate that overexpression of the Rgg/SHP regulon reduces S. pneumoniae fitness in a murine model of lung infection. In contrast to the overexpression mutant, the deletion strain was not significantly affected in virulence.
SHP stimulates exopolysaccharide production in unencapsulated strain R36A. We next asked if treatment with sSHP could increase the production of surface polysaccharides and whether such an effect would depend on the presence of a functional capsule locus. We treated D39, an unencapsulated D39 derivative strain (R36A), and the Δrgg and Over-EXP mutant strains with a specific and a nonspecific sSHP and measured both cell size and FITC-dextran exclusion in all samples. No . Every 12 h, the medium was changed to avoid bacterial lysis. At 36 h, the supernatant was removed, the biofilm was resuspended in rich medium, the wells were scraped, and the material obtained was immediately refrigerated for plating. The y axis shows the number of biofilm CFU recovered from wells as fold change compared to wild-type D39, and each value is the mean and the standard error of the mean of at least three independent biological experiments. ****, P Ͻ 0.0001. The second and third sets show bacteria recovered from BALF and mashed lungs, respectively. The y axis shows the ratio of the bacterial counts recovered to the inoculum. A ratio of Ͼ1 indicates that the overexpression strain was predominant, while a ratio of 0 to 1 indicates that D39 was recovered in higher numbers. *, P Ͻ 0.05. differences in surface polysaccharide expression were observed in any of the strains when a negative-control nonspecific sSHP peptide from Streptococcus salivarius was added to the cultures. Differences in polysaccharide thickness were also not detected after the addition of a specific sSHP to D39 or the derivative Δrgg and Over-EXP mutants (Fig. S5). In untreated R36A, the cell size matched the FITC-dextran exclusion area, as expected; however, upon treatment with the specific sSHP, the FITC-dextran exclusion area increased significantly, reaching the levels of encapsulated D39 (Fig. 8). Such findings corroborate the above-described data showing that the new locus participates in the regulation of surface polysaccharides in S. pneumoniae and that this increase is observed even in the absence of a functional serotype 2 capsule locus.

DISCUSSION
The pneumococcus has been a model organism for the major discoveries regarding natural transformation, culminating with the discovery of DNA as hereditary material (35,39). The discoveries gradually progressed to the description of a competence factor in culture supernatants as the first report of quorum sensing in Gram-positive bacteria (40,41). More recently, Rgg proteins, once seen as exclusive stand-alone regulators, were identified together with SHPs as components of newly described quorum-sensing signaling systems found in the majority of streptococci (12,13). Studies of the Rgg/SHP systems in S. pyogenes and S. agalactiae have revealed that while the mechanisms involved in SHP production and processing are similar, the regulated genes differ (15,16), indicating that the system has evolved to control different social behaviors. Here we present for the first time the characterization of an Rgg/SHP system in S. pneumoniae. By using directional transcriptome sequencing (RNA-Seq), which provides information on the direction of transcripts, this study provides the most detailed map of a transcriptome capturing Rgg/SHP responses, and it shows that one single transcript initiated at shp is regulated by this system. In this transcript, different genes carrying annotations related to polysaccharide synthesis were identified. By stimulating the expression of the rgg regulon, an increase in surface polysaccharide expression was produced, concomitant with a decrease in biofilm formation (Fig. 9).
An emerging theme in social behaviors controlled by quorum sensing is the regulation of surface polysaccharide expression. The first Rgg regulator identified in streptococci, although not associated with any known quorum-sensing peptide signal, was indeed found to control the expression of a glucosyltransferase in Streptococcus gordonii (42). In Pseudomonas aeruginosa, acyl homoserine lactone signals activate, for instance, a cascade of events that, although indirect, lead to increased expression of matrix polysaccharides (43). Direct regulation of polysaccharide synthesis genes has also been described, such as in the Gram-negative bacterium Rhodobacter capsulatus, in which capsule production is controlled by intracellular regulators that bind homoserine lactone quorum-sensing signals (44). The connection between quorum sensing and extracellular polysaccharide production has also been observed for regulators in the bacteria Pantoea stewartii and Vibrio vulnificus that act as repressors of capsule synthesis (33,45). Interestingly, in the Rgg system of S. pyogenes, genes with homology to polysaccharide synthesis genes seem to be activated by SHP (15), but their functions remain mostly uncharacterized. In this study, overexpression of the Rgg target regulon in D39 and stimulation with sSHP in the derivative unencapsulated strain R36A led to increased exopolysaccharide expression. Interestingly, in the D39 wild type, no stimulation by sSHP was observed, which contrasted with the results obtained with the overexpression mutant. This is most probably due to the levels of stimulation of the putative polysaccharide expression genes of the Rgg regulon, which were considerably lower in D39 than in the overexpression mutant.
S. pneumoniae exhibits highly variable capsules, with Ͼ90 pneumococcal serotypes currently recognized that vary in polysaccharide composition and structure from linear repeats of two sugars (serotype 3) (46) to highly complex repeats of multiple sugars with side branches (serotype 12F) (47). The strain used in this study, S. pneumoniae D39 serotype 2, contains six sugars per repeat unit (three rhamnose residues, one glucuronic acid, and two glucose units), and the polymerizing linkage happens between the first and fourth sugars (47,48). The biosynthetic pathway encoded specifically by the cps loci of serotype 2 takes part via UDP-glucose-6-dehydrogenase, a key compound in the biosynthesis of pneumococcal capsules that converts the precursor sugar substrate UDP-glucose (UDP-Glc) into UDP-glucuronic acid (UDP-GlcA) (36,49,50). In contrast, serotypes 12F and 12A employ another biosynthetic pathway for capsule production by using the enzymes MnaA and MnaB. Our data show that the Rgg0939 regulon encloses a homolog of MnaB, which is responsible for the conversion of N-acetylmannosamine (product of a previous reaction catalyzed by MnaA) into N-acetylmannosaminuronic acid (UDP-ManNAcA), an uncommon capsule component present only in serotypes 12F and 12A (48).
It is intriguing, however, that (i) UDP-ManNAcA has not been identified in the capsule structure of D39, (ii) D39 only produces MnaB, and (iii) the gene responsible for MnaB production is located outside the capsule loci. On the other hand, it has been FIG 9 Schematic representation of the Rgg/SHP system in S. pneumoniae D39. The SHP precursor is produced, processed, and exported by PptAB and Eep (16,77). Once it reaches the quorum-sensing threshold, the mature peptide is imported by Opp (12,15,16), and it binds to Rgg for activation of the shp promoter. This, in turn, activates the pheromone positive-feedback loop and an operon formed by 12 genes initiated at the shp gene. High activation of this regulon upregulates polysaccharide synthesis and downregulates biofilm formation.
previously noted that essential enzymes involved in capsule production are not seldom encoded in a locus other than the one involved in capsule regulation. One example is GalU, one of the proteins required for adequate production of UDP-Glc, the main serotype 2 capsule precursor (51,52). Furthermore, GalE, a UDP-4-glucose-epimerase, has been shown to be involved in lipopolysaccharide synthesis in Gram-negative bacteria. UDP-Glc is the initial substrate of the serotype 2 capsule; thus, it is likely that GalE converts UDP-Gal into UDP-Glc, increasing its availability for subsequent polysaccharide production. Although this conversion has not been shown experimentally in S. pneumoniae (53), cells can be metabolizing exogenous galactose as a primary energy substrate for capsule synthesis. In addition, galactose has been recently suggested to be a possible component of the exopolysaccharide matrix formed by S. pneumoniae unencapsulated strain R6, which is a derivative of R36A (54). This can partially explain the increased capsule thickness of the overexpression strain and SHP-stimulated unencapsulated strain R36A. Further studies are under way to investigate the nature of the surface polysaccharide regulated by the identified S. pneumoniae SHP/Rgg system. Extracellular polysaccharides, represented by capsule structures, as well as by loosely associated exopolysaccharides and other forms of polysaccharide structures, have long been known to affect biofilm formation. In S. pneumoniae, capsule expression has been negatively associated with biofilm formation (55), whereas other forms of polysaccharides seem to promote biofilm formation (54). The inhibition of biofilm formation by the capsule is attributed to the capsule effect on blocking the exposure of S. pneumoniae surface adhesins that promote attachment to epithelial cells (37). Our results showed that the biofilm phenotype of the rgg deletion mutant deficient in quorum sensing resembles that of capsule deletion mutants of S. pneumoniae and the close relative Streptococcus mitis, which form thicker biofilms on epithelial cells in the absence of the capsule (55). Comparing the results obtained with those of the dextran exclusion assay, which uses planktonic cells and showed no differences between D39 and the rgg deletion mutant, with the biofilm results showing remarkable differences in biofilm, the results suggest that planktonic conditions do not favor expression of the Rgg regulon. This is further reinforced by the fact that placing the Rgg regulon under the control of a constitutive promoter supporting overexpression of the regulon resulted in reduced biofilm formation. Of note, in this study, biofilm formation was assessed by comparing the mean total numbers of CFU recovered from experiments.
Previous reports have shown a positive correlation of capsule expression and virulence in S. pneumoniae (29,56). In animal models, unencapsulated strains, in general, show a reduced capacity to colonize and cause infections. Most important, though, seems to be the ability of S. pneumoniae to regulate capsule expression. In line with this, Hammerschmidt et al. (37) showed that pneumococci attached to or invading lung epithelial cells downregulate their capsule, minimizing the physical interference for successful adhesion. Also, a recent study showed that both over-and underexpression of the capsule locus reduced S. pneumoniae fitness in infection models (57). While no difference was observed between D39 and the rgg deletion mutant, which may reflect functional redundancy because of the expression of the original capsule, the overexpression strain was recovered in lower numbers from the nasal cavity and bronchoalveolar lavage fluid (BALF). Such an outcome might be explained by a high metabolic cost (number of carbons and high-energy bonds) involved in the production of capsular polysaccharide (29) and the inability to fine-tune capsule expression (57). Remarkably, surface polysaccharides serve several functions for bacteria, including but not limited to escape from phagocytosis by macrophages and neutrophils and protection or increased sensitivity to human defensins (58)(59)(60). Thus, the identified Rgg/SHP quorum-sensing-regulated behavior is likely to have implications beyond those characterized here.
Quorum-sensing signaling of Rgg/SHP systems is often found to be inhibited or decreased in peptide-rich media (16,61). In this study, stimulation of the quorumsensing system was similar in rich, semidefined and defined media. It is important to note, however, that the saturation threshold point in each particular medium was not assessed and that further optimization of experimental conditions may demonstrate different levels of stimulation.
The mature form of the S. pneumoniae D39 SHP (DIIIIVGG) is identical to SHP3 of the Rgg2/3 system in S. pyogenes and to SHP1044 in S. thermophilus. Together with the finding that even slightly divergent sequences can promote interspecies crosscommunication (21,22), it is possible that the Rgg/SHP system has an important role in microbial interactions taking place in multispecies communities. Finally, the finding that S. pneumoniae D39 uses a quorum-sensing system to directly regulate a single locus with a role in surface polysaccharide expression suggests an important function for the system in adaptation and survival. The importance of controlling surface polysaccharide production as a collaborative group behavior is supported by genome analysis showing homologs of this system in other S. pneumoniae strains that have the Rgg/SHP system and by the link between quorum sensing and capsule production in other bacterial species distantly related to streptococci.

MATERIALS AND METHODS
Bacterial strains and media. The S. pneumoniae strains used in this study are listed in Table 3. All strains were stored at Ϫ80°C in Todd-Hewitt broth (THB; Becton, Dickinson and Company, Le Pont de Claix, France) supplemented with 0.5% yeast extract and 30% glycerol. Precultures were prepared as described previously (62). For growth, RNA sequencing, and biofilm experiments, the media used were TSB (Bacto soybean-casein digest medium), THB, CϩY (24), and a chemically defined medium supplemented with 1% TSB and 0.5% choline (CDMϩϩ) (27). For plates, blood agar base number 2 (Oxoid, Hampshire, England) supplemented with 5% defibrinated sheep blood (TCS Biosciences Ltd., Buckingham, United Kingdom) was used. The selective antibiotic used was kanamycin at 500 g·ml Ϫ1 .
Pheromones. The predicted specific S. pneumoniae mature pheromone SHP (NH 2 -DIIIIVGG) was obtained from the GenScript Corporation (NJ) as a custom synthetic peptide and stored as a sterile 10 mM solution in 3% ammonia water at Ϫ20°C. Stock solutions of lyophilized CSP-1 were prepared by resuspending it in distilled water to a concentration of 10 mM and storing it at Ϫ20°C. Working solutions of 1 and 0.1 mM were aliquoted and stored at Ϫ20°C. As a negative-control peptide, SHP from S. salivarius (NH 2 -PYFTGCL) was synthesized (GenScript Corporation, NJ) and stored as a 10 mM solution at Ϫ20°C.
Bacterial growth. For bacterial growth assays, precultures of the strains were centrifuged at 8,000 ϫ g at 4°C for 8 min; suspended 1:10 in TSB, CϩY, or CDMϩϩ; and transferred in 200-l volumes to wells of a 96-well plate. Optical density at 600 nm (OD 600 ) was measured every 15 min at 37°C.
Real-time PCR. Precultures were centrifuged at 8,000 ϫ g at 4°C for 8 min and resuspended 1:10 in TSB, CϩY, or CDMϩϩ, and cultures were incubated at 37°C in 5% CO 2 for 20 min. Cultures were divided in two, with one half being treated with sSHP and the other half being treated with the same volume of 1ϫ phosphate-buffered saline (PBS). Cultures were incubated for different times at 37°C and 5% CO 2 and harvested by centrifugation at 8,000 ϫ g at 4°C for 8 min. RNA was extracted with the High Pure RNA Isolation kit (Roche). The First Strand cDNA Synthesis kit from Thermo Scientific (Fermentas) was used in accordance with the manufacturer's instructions. The primers used are listed in Table S1.
RNA-Seq sample preparation and analysis. S. pneumoniae precultures were centrifuged at 8,000 ϫ g at 4°C for 8 min and resuspended 1:10 in CϩY; 10-ml culture volumes were incubated at 37°C at 5% CO 2 for 20 min; and the cultures were divided in two, with one half being treated with sSHP (1 M, N-DIIIIVGG) and the other half being treated with the same volume of 1ϫ PBS. Cultures were incubated for 2 h at 37°C in 5% CO 2 and harvested at 8,000 ϫ g at 4°C for 8 min. RNA was extracted with the mirVana microRNA isolation kit (Life Technologies, Inc.). After a first extraction, lysed cells were treated with TURBO DNase (Life Technologies, Inc.) and a second RNA extraction was performed. RNA quality was tested on a bioanalyzer with the RNA 6000 Nano kit. The 16S and 23S rRNAs were then removed with the MICROBExpress Bacterial mRNA Enrichment kit (Thermo Fisher Scientific). RNA was treated with the NEXTFlex Directional RNA-Seq kit (dUTP based) for preparation of the DNA library for paired-end sequencing with Illumina HiSeq by the Norwegian Sequencing Centre (http://www.sequencing.uio.no). A sequence in FASTQ format was derived from each sample. Analyses were performed in accordance with previously reported protocols (63,64).
Construction of mutants with antibiotic resistance markers. Knockout deletion mutants were constructed by the PCR mutagenesis method described previously (65). All of the strains constructed and the primers used are listed in Table S1.

High-efficiency transformation and markerless mutant construction.
Encapsulated pneumococcal strains present very low in vitro transformation efficiency levels (66)(67)(68)(69). While low efficacy can still be used for mutant construction carrying markers, for direct genome editing, higher transformation efficiencies are valuable. Thus, to create markerless mutants of D39, we adapted the growth conditions, synthetic peptide, medium, and screening protocol from a recently developed strategy that we optimized for S. mitis, a close relative of S. pneumoniae (62), which resulted in increased efficiency rates of 4 to 21%. This improvement in transformation efficiency allowed for direct markerless deletion of rgg0939 from S. pneumoniae D39. Table S2 contains details of how we assembled the previously described CϩY YB medium (70). An amplicon with 3-kb flanking regions was constructed by overlapping PCR and transformed into D39 in accordance with the protocol for high transformation efficiency shown in Table S3. Initial screening of 40 colonies (Fig. S6A) showed two positive colonies; one was a mixture of both mutant and parent alleles, and the other was a pure mutant. The pure colony was then selected, grown for another 6 h at 37°C, and replated. On the next day, three derivate colonies of this isolate were tested with two sets of primers, positive (FP952/FP1126) and negative (FP952/FP1579) selection, for confirmation of the mutation (Fig. S6B). After confirmation, cultures were stocked at Ϫ80°C supplemented with 30% glycerol. Strain SP068 was complemented with rgg0939 (SP101) by using a markerless amplicon carrying the gene (FP1577/FP1578 amplified from D39). Selection and confirmation of the mutation were performed by screening PCR with two sets of primers, FP952/FP1126 and FP952/FP1579, for confirmation of the mutation.
Measurement of pneumococcal capsule thickness. Degree of encapsulation was determined by measuring the zone of exclusion of FITC-dextran (2,000 kDa; Sigma) by the method of Gates et al. (31), and wet mounts were prepared in accordance with reference 29. The preparations were visualized with a Nikon Eclipse CI-L/S microscope (Nikon) with a 100ϫ objective and photographed with a Zyla 5.5 complementary metal oxide semiconductor camera (Andor). The images were analyzed with NIS-Elements BR 3.0 software (Nikon). For each strain, the mean area of FITC exclusion of 200 cells was determined; long chains and clumps of cells were not included.
Biofilm assay. We used a protocol for biofilm assays previously described (71). Briefly, a lung cell line (A549; Sigma-Aldrich) was grown to 100% confluence on 24-well plates. At the point of full confluence, the cells were fixed with 4% paraformaldehyde for 30 min at room temperature and washed three times with 1ϫ PBS. Bacterial cultures were seeded into a 500-l volume of CDMϩϩ to an OD 600 of 0.1 and incubated for 36 h at 34°C in 5% CO 2 . The medium was changed every 12 h. At the end of 36 h, the supernatant was removed, the wells were scraped, and the material obtained was resuspended in THY and plated. CFU were counted after 24 h of incubation at 37°C in 5% CO 2 .
Animal models. Animal experiments were performed with female CD1 mice (Charles River, Inc.) 6 to 8 weeks old. For mixed pairwise infections, the same numbers of bacteria from stocks of different strains were mixed and diluted to the appropriate concentration. All samples were plated onto blood agar plates containing gentamicin (5 g·ml Ϫ1 ) for differentiation from other species and CFU calculation.
For the early lung disease model, 5 ϫ 10 6 CFU of a strain in 50 l of 1ϫ PBS were inoculated intranasally under isoflurane general anesthesia. At 24 h postinoculation, animals were killed by intraperitoneal pentobarbital injection and the nasal cavity and lungs were each washed with 1 ml of sterile 1ϫ PBS. Samples from the nasal wash and BALF were stored in ice for immediate dilution and plating. To recover the BALF, a small cut in the trachea was made with a scalpel and 1 ml of sterile fresh 1ϫ PBS was inoculated and recovered for plating (72)(73)(74).
For the early lung disease mixed-infection model (competitive assay), 2.5 ϫ 10 6 CFU each of strain D39 and the Over-EXP mutant in 50 l were inoculated intranasally under isoflurane general anesthesia. The same protocol for killing and sample collection was used. The competitive index (CI) was calculated by dividing the outcome ratio by the inoculum ratio (72). A CI of Ͻ1 indicates that the mutant is attenuated compared to D39, and a CI of Ͼ1 indicates that D39 is attenuated compared to the mutant (38).
Ethics statement. Our experiments were approved by the UCL Biological Services Ethical Committee and the UK Home Office (project license PPL70/6510). Experiments were performed in accordance with United Kingdom national guidelines for animal use and care under UK Home Office license.
Statistical analysis. The Human Oral Microbiome Database (75) and GenBank (76) were used to identify and map genes of each strain. Operon prediction was performed with DOOR (25) and the MicrobesOnline database (26). RT-PCR and FITC-dextran exclusion data for two groups were analyzed with Student's t test, while for three or more groups, we used one-way analysis of variance (ANOVA), followed by Tukey's post hoc test. For the in vitro and animal models assays, we used Student's t test and one-way ANOVA, followed by the Tukey or Kruskal-Wallis test, respectively. For the mixed-infection model, CIs were compared with the initial ratio of 1.0 (the predicted CI if there is no difference between the two strains tested) by using Student's t test. Significance was set at a P value of Ͻ 0.05.
Data availability. The RNA-Seq raw read data obtained in this study have been deposited with links to BioProject accession no. PRJNA377718 in the NCBI BioProject database.

ACKNOWLEDGMENTS
This study was partially funded by Norwegian Research Council grant 241011, by the Norwegian Surveillance System for Antibiotic Resistance in Microbes (NORM), and by the Faculty of Dentistry at the University of Oslo. In addition, this study was partially undertaken at UCLH/UCL, which received a portion of its funding from the Department of Health NIHR Biomedical Research Centre funding scheme. J.N.P. was supported by a grant from the Medical Research Council (MR/K00168X/1).