A novel streptococcal cell-cell communication peptide promotes pneumococcal virulence and biofilm formation

Abstract

Streptococcus pneumoniae (pneumococcus) is a major human pathogen. It is a common colonizer of the human respiratory track, where it utilizes cell-cell communication systems to coordinate population-level behaviors. We reasoned that secreted peptides that are highly expressed during infection are pivotal for virulence. Thus, we used in silico pattern searches to define a pneumococcal secretome, and analyzed the transcriptome of the clinically important PMEN1 lineage to identify which peptide-encoding genes are highly expressed in vivo. In this study, we characterized virulence peptide 1 (vp1), a highly expressed Gly-Gly peptide-encoding gene in chinchilla middle ear effusions. The vp1 gene is widely distributed across pneumococcus as well as encoded in related species. Studies in the chinchilla model of middle ear infection demonstrated that VP1 is a virulence determinant. The vp1 gene is positively regulated by a transcription factor from the Rgg family and its cognate SHP (short hydrophobic peptide). In vitro data indicated that VP1 promotes increased thickness and biomass for biofilms grown on chinchilla middle ear epithelial cells. Further, the wild-type biofilm is restored with the exogenous addition of synthetic VP1. We conclude that VP1 is a novel streptococcal regulatory peptide that controls biofilm development and pneumococcal pathogenesis.

Abbreviated summary

Streptococcus pneumoniae asymptomatically colonizes the human nasopharynx and disseminate to other tissues causing mild to severe disease. It utilizes cell-cell communication systems to coordinate population-level responses. However, how this bacteria cause disease is poorly understood. We describe the small peptide VP1 and its role in biofilm development and pathogenesis in a clinically relevant pneumococcal strain. Our findings will shed light on the biological processes that lead to microbial virulence.

Introduction

A wide range of bacterial pathogens cause disease by population-level behaviors coordinated by communication across cells (Bassler, 2002; Bassler and Losick, 2006; Federle and Bassler, 2003). Streptococcus pneumoniae (pneumococcus) is a major human pathogen that engages in social behaviors. Secreted pneumococcal peptides have been shown to coordinate competence and bactericidal activity, and have been implicated in the extensive genomic plasticity and adaptability of this species (Claverys et al., 2006; Claverys and Havarstein, 2007; Dawid et al., 2007; Havarstein et al., 1995a). However, additional signaling molecules that control pneumococcal behaviors in the context of the human host are poorly understood.

Bacterial biofilms are recalcitrant to antibiotic treatment, and thus pose a major threat to human health (Stewart and Costerton, 2001). Pneumococcus forms biofilms on mucosal surfaces such as the nasopharynx during asymptomatic colonization, the middle ear during otitis media, and the sinus during rhinosinusitis (Hall-Stoodley et al., 2006; Hoa et al., 2009; Post et al., 2007; Sanderson et al., 2006). Pneumococcal biofilms play a central role in drug resistance, as cells in this mode of growth display decreased sensitivity to antibiotics (Marks et al., 2012). Further, the biofilm is an environment propitious for transformation. The available DNA in the matrix serves as a substrate for transformation, leading to strain evolution and the spread of vaccine-escape and drug-resistant genotypes (Chao et al., 2014; Croucher et al., 2012; Trappetti et al., 2011; Vidal et al., 2011). Pneumococcal biofilms play a key role in the virulence potential of an infection. Transcriptional changes induced during a biofilm mode of growth have been associated with pneumococcal colonization (Sanchez et al., 2011). Furthermore, bacteria dispersed from a biofilm display increased propensity for tissue dissemination and pathogenesis, when compared to cells from the biofilm layer or planktonic cultures (Chao et al., 2015; Marks et al., 2013). In summary, pneumococcal biofilms are of high clinical importance for drug-resistance and pathogenesis.

Bacteria orchestrate community behaviors to cause disease (Popat et al., 2008). In Gram-positive bacteria, these behaviors mostly rely on signaling cascades regulated by extracellular peptides (Li and Tian, 2012). Many of the Gram-positive ribosome-synthesized peptides that initiate cell-cell communication signaling cascades, can be categorized based on sequence features (Cook and Federle, 2014). The pneumococcal CSP and BIP belong to the double glycine (Gly-Gly) peptide family, a group distinguished by a cleavable N-terminal domain with a double glycine motif at the cleavage site (Havarstein et al., 1995a). The pro-peptide is processed and exported by an ATP-binding cassette (ABC) transporter, and cleavage occurs immediately after the double glycine (Havarstein et al., 1995b). The secreted forms are active and bind the input domain of a surface-exposed histidine kinases from a two component system (TCS) (de Saizieu et al., 2000; Pestova et al., 1996). Another distinct peptide group is represented by the pneumococcal PhrA, which belongs to the family of PlcR (phospholipase C regulator) cognate peptides, a group secreted by the Sec pathway and internalized by transporters from the conserved oligopeptidase permease family (Bouillaut et al., 2008). Finally, pneumococci also encode peptides belonging to the family of Rgg (regulator gene glucosyltransferase) cognate peptides, specifically the short hydrophobic sequences (SHPs) (Fleuchot et al., 2011). The SHPs fall into three categories, identified based on the nature of the charged residue and/or organization relative to the Rgg (Fleuchot et al., 2011). Pneumococcal genomes encode SHPs from groups I and III (Fleuchot et al., 2011). These SHPs are predicted to be processed and secreted by ABC transporters and internalized by members of the conserved oligopeptidase permease family (Chang and Federle, 2016). The molecular components of trafficking provide clues regarding the likelihood that signaling peptides are functional across strain and species barriers (Besset et al., 2013). The conserved features of these peptide families set the stage for in silico pattern searches to identify novel signaling molecules.

Secreted peptides can exert their influence via multiple different types of signal transduction systems. The best-characterized signaling pathway is the TCS. In this architecture, the secreted peptide binds a surface exposed histidine kinase triggering autophosphorylation and subsequent phosphotransfer to a cognate response regulator (Cozzone, 1988). There are thirteen TCSs in the pneumococcal core genome (Lange et al., 1999; Throup et al., 2000). However, cognate peptides have been identified for only two of these systems: CSP for ComDE and BIP for BlpDE (for the remainder, the molecule(s) that directly activate the kinases are unknown) (de Saizieu et al., 2000; Pestova et al., 1996). More recent studies have revealed a distinct signal transduction architecture, where an active form of a secreted peptide is internalized into cells and directly binds a transcription factor from the RRNPP (Rap, Rgg, NprR, PlcR, PrgX) superfamily (Rocha-Estrada et al., 2010). There are multiple RRNPP systems encoded in the pneumococcal pangenome (Bortoni et al., 2009; Hoover et al., 2015; Kadam et al., 2017), and cognate peptides have been identified for two of the regulators (PhrA for TprA and PhrA2 for TprA2) (Hoover et al., 2015; Kadam et al., 2017). Thus, pneumococcal strains make use of distinct peptide-induced signal transduction pathways during infection.

Pneumococcus can asymptomatically colonize the nasopharynx. However, in a percentage of cases pneumococcal strains disseminate to other tissues and cause moderate to severe disease (Kadioglu et al., 2008). Bacterial dissemination is often initiated as a response to extracellular stresses caused by changes in microbiome composition, immune defenses, and/or nutritional changes (Chang et al., 2015; Dalton and Scott, 2004). Free sugars are scarce in the nasopharynx and upper respiratory tract, such that pneumococci survive by breaking down host mucins into galactose and other free complex sugars. In contrast, levels of glucose and/or free sugars are substantially higher in the lower respiratory tract and the blood leading to tissue-specific differences in pathways utilized for the uptake and processing of nutrients (Paixão et al., 2015; Terra et al., 2010). In pneumococcus, CodY, GntR, and CcpA are master regulators involved in amino acid metabolism, carbohydrate metabolism, and iron uptake (Carvalho et al., 2011; Hendriksen et al., 2008a, 2008b). Further, RRNPP regulators have been associated with metabolic control in multiple species (Chang et al., 2011; Cook et al., 2013; Cook and Federle, 2014; Hoover et al., 2015; Lasarre et al., 2012). Transcriptomic studies demonstrate that the PlcR-type regulator TprA (spd1745) is regulated by CcpA (Carvalho et al., 2011), and that a Rgg-type regulator (spr0144) is regulated by CodY and by effectors of GntR (Hendriksen et al., 2008a, 2008b; Hoover et al., 2015). Thus, pneumococcal RRNPP regulators are likely to play an important role in nutritional response, and consequently in population level behaviors that promote either dissemination or asymptomatic colonization.

Regulators in the TCS and RRNPP signaling pathways can monitor the extracellular environment and thus are well positioned in the cellular circuitry to control microbial-host interactions, intra- and inter-strain interactions, biofilm development, and virulence. We hypothesized that bacterial communication peptides involved in virulence are secreted and highly expressed in vivo. We employed comparative genomics to identify putative secreted peptides in the pneumococcal pangenome, and prioritized candidates based on expression levels in vivo. In this study, we characterize the virulence peptide 1 (vp1). The gene encoding VP1 is highly expressed during infection and regulated by CodY and Rgg. VP1 enhances biofilm development and activates the pneumococcal virulence potential.

Results

Identification and prioritization of Gly-Gly peptides in the pneumococcal pangenome

We reasoned that secreted peptides highly expressed during infection and/or highly induced relative to planktonic growth are critical in pneumococcal virulence. Thus, we undertook a two-pronged approach: identify secreted peptides in the pneumococcal genome and measure their expression in a model of infection.

To identify secreted peptides, we focused our attention on the pneumococcal peptides CSP and BIP. Both are exported by a dedicated ABC transporter with a proteolytic domain required for peptide cleavage (Havarstein et al., 1995b). The ABC transporters recognize N-terminal sequences with conserved features, notably the presence of the previously characterized Gly-Gly leader peptide (LSXXELXXIXGG) (Havarstein et al., 1995a). To search for additional secreted pneumococcal peptides, we used Multiple Expectation Maximization for Motif Elicitation (MEME) to generate a position dependent probability matrix that defines the Gly-Gly motif and captures its length and positional variability at each peptide position. Our input consisted of 167 homologues of CSP and BIP, obtained from a NCBI blastP searched of pneumococcal CSP and BIP against the non-redundant protein database. Next, we employed the Motif Alignment and Search Tool (MAST) to search for the MEME motif in a database of coding sequences from 60 streptococcal genomes. This database is composed of thirty-nine S. pneumoniae (encompassing nineteen serotypes and thirty-six MLST types), three S. pseudopneumoniae, nine S. mitis, and seven S. oralis strains, selected to represent much of the genomic diversity of pneumococcus and viridans, plus two S. infantis strains as representatives of more distantly related genomes (Table S1). The output of this MAST search included 192 protein sequences. This expanded set of sequences was used as input to generate a MEME motif, and subsequently search the same database with MAST (Table S2). Reiterative use of MEME and MAST returned a list of 716 proteins, of which 696 encode the motif at the N-terminus and 98% are peptides with less than 130 residues. To illustrate the domain conserved across these sequences we generated a MEME logo using the 696 sequences (Fig. 1A). This set of sequences from the MAST output corresponds to a predicted secretome.

Figure 1. Identification of a predicted secretome of Gly-Gly family peptides, and analysis of peptide expression during host infection.

A. Sequence logo derived from 696 predicted secreted peptides. Amino acids are represented by one letter abbreviation and color coded as follows: blue, basic; red, acidic; green, polar; yellow, neutral. Height of amino acids is proportional to the fraction of the observed frequency relative to the expected frequency in the streptococcal database. B. NanoString RNA count averages for the set of predicted secreted peptides in PMEN1 strain PN4595-T23 in mid-log planktonic cultures (left column) and effusions extracted from the chinchilla middle ear 48h post-inoculation (right column). As positive controls, we utilized the quorum sensing peptide encoding gene phrA, and the predicted lanthiopeptides lanA and lcpA genes, which are highly expressed during in vivo expression. RNA counts were normalized to the geometric mean of the housekeeping genes metG and gyrB, and sorted by expression level. The VP1-encoding gene is highlighted in red. Upstream refers to the relative position of the nanoString probe. ID’s are represented in the model PMEN1 genome (strain ATCC 700669, GenBank FM211187), and corresponding IDs in PN4595-T23 are listed in Table 1 and Table S3.

The predicted secretome was organized into 25 clusters using sequence similarity and genomic localization (Table 1). These clusters display variable distributions across strains from the same species and across species. As expected, this set of clusters contains the input molecules CSP and BIP. In addition, this set also includes previously identified peptides, such as seven peptides localized into multiple blp loci (Lux et al., 2007; Reichmann and Hakenbeck, 2000), the bacteriocidal molecules CibA and CibB (Guiral et al., 2005), and the LanA lantibiotic (Hoover et al., 2015). We also identified thirteen peptides that have not yet been functionally characterized. Twelve of these are present in some subset of pneumococcal strains as determined using blastP against the NCBI database of Streptococcus pneumoniae, and one is present in S. oralis and S. mitis but appears to be absent in pneumococcal strains. The pneumococcal peptides range from rare in the species to distributed in the vast majority of strains. We hypothesize that these peptides correspond to the Gly-Gly components of the pneumococcal secretome.

Table 1.

List of Putative Secreted Peptides. IDs represent strain PN4595-T23, followed by the orthologue in strain ATCC 700669. If gene is absent in these PMEN1 strains, the IDs are provided for representative strains TIGR4, BS9BS68, and Hungary19A. For peptides observed outside the selected set of pneumococcal genomes; IDs correspond to S. mitis B6 or S. oralis subspecies tigurinus TAZ3a. Bold: Input for original MEME. Blue: previously characterized genes.

Blast ID	Annotation
CGSSp4595_0134 (SPN23F01520)*	VP1
CGSSp4595_0386 (SPN23F04030)*	hypothetical
CGSSp9BS68_07272	hypothetical
SPH_0224	hypothetical
SP_0924	hypothetical
SPH_0218	hypothetical
SP_0115	hypothetical
CGSSp4595_0119 (SPN23F01370)*	CibB
CGSSp4595_0120 (SPN23F01380)*	CibA
CGSSp4595_1950 (SPN23F19710)*	LanA
CGSSp4595_2265 (SPN23F22700)*	CSP2 (comC)
CGSSp4595_0465 (SPN23F04791)	BIP (blpC)
CGSSp4595_0469 (SPN23F04830)*	BlpI
CGSSp4595_0470 (SPN23F04840)*	BlpJ
CGSSp4595_0471 (SPN23F04850)	BlpK
CGSSp4595_0472 (SPN23F04860)*	putative bacteriocin
CGSSp4595_0038 (SPN23F00570)*	bacteriocin
SP_0539	BlpM
SP_0540	BlpN
WP_001814148 (SPH_0227)	hypothetical
smi_0057.1	hypothetical
smi_0059.1	hypothetical
smi_0061	hypothetical
HMPREF1112_0784	hypothetical
H354_08695	hypothetical (not present in pneumococcus)

^*analyzed by nanoString.

To prioritize the pneumococcal peptides for characterization, we measured their transcription in the chinchilla model of human middle ear infection. To this end, we selected an isolate from the clinically important PMEN1 lineage (PN4595-T23). The PMEN1 lineage is widespread and multidrug-resistant, and includes vaccine-escape isolates (Croucher et al., 2011; Nesin et al., 1998; Wyres et al., 2013). To measure gene expression, we employed nCounter nanoString technology, which provides an automated, highly sensitive enumeration of pathogen mRNA transcripts in infected tissues (Geiss et al., 2008). Our probe set included eleven peptides identified from the MEME/MAST and encoded in PMEN1 (Table 1, Table S3).

The PMEN1 strain PN4595-T23 was inoculated transbullarly into the chinchilla. RNA was extracted from effusions from chinchilla ears 48h post-transbullar inoculation. A previously uncharacterized peptide downstream of an Rgg/SHP system fits our criteria for the identification of novel virulence factors in that it is the most highly expressed Gly-Gly peptide-encoding gene in effusions, and appears highly induced relative to planktonic growth; we termed it virulence peptide 1 (VP1) based on analysis presented below. The level of vp1 is 1.3 times that of epsA (exopolysaccharide A) and 1.6 times that of cbpA (choline binding protein A), which are highly expressed during infections (Ogunniyi et al., 2002; Orihuela et al., 2004) (Table S3). We also measured peptide expression in planktonic cells grown to mid-log in rich media, to identify genes differentially expressed in vivo relative to in vitro. The average number of RNA molecules for vp1 in vivo was 36.15 ± 1.69-fold higher than in planktonic culture (Fig. 1B). We conclude that the levels of vp1 during middle ear infection are comparable to levels of transcripts highly expressed in vivo, and thus VP1 may play a role in infection.

VP1 is a novel virulence determinant

To evaluate the virulence properties of VP1, we employed the chinchilla model of human middle ear infection. Cohorts of 10 chinchillas were injected transbularlly with either the parental wild-type (WT) strain or the isogenic vp1 deletion mutant (Δvp1). Chinchillas inoculated with Δvp1 displayed a significant decrease in mortality relative to animals inoculated with the WT strain (P-value = 0.023) (Fig. 2A). Further, relative to the WT, the Δvp1 displayed a significant reduction in dissemination to the brain (P-value = 0.035), and although not significant, reduced dissemination to the lungs (P-value = 0.152) (Fig. 2B). These findings suggest that VP1 plays a role in pneumococcal in vivo survival and pathogenesis.

Figure 2. VP1 is a virulence factor.

Pneumococcal WT or Δvp1 strains were injected bilaterally into the tympanic bullae of chinchillas and mortality and dissemination were scored. A. Kaplan-Meier survival curves of chinchillas. B. Percent of chinchillas with bacterial dissemination to the brain or the lungs, as determined by CFU counts in these tissues at time of death. Statistical significance was calculated using paired two-tailed Student’s t-test (* P-value < 0.05, & P-value = 0.152).

VP1 plays a role in pneumococcal biofilm development

In middle ear infections pneumococci adopt a biofilm mode of growth, thus we investigated whether vp1 influences biofilm development (Hall-Stoodley et al., 2006). To this end we utilized chinchilla middle ear epithelial (CMEE) cells (Raffel et al., 2013). We established that PMEN1 pneumococcal cells form a robust biofilm on CMEE substrate sixteen hours post-inoculation (Fig. S1). We compared biofilm biomass, thickness and roughness between WT and Δvp1 strains using confocal imaging (Fig. 3A). The Δvp1 displays approximately 2-fold reduction in biofilm biomass and thickness when compared to the WT (Fig. 3A, B). Furthermore, we demonstrate that the WT phenotype is rescued in a vp1 complemented strain (Δvp1:vp1) (Fig. 3A, B). The difference in biofilm growth is not the result of variation in growth rate, as WT, Δvp1, and Δvp1:vp1 strains display the same rate of growth in the epithelial growth media used for the pneumococcal-CMEE biofilm assays (Fig. 3C). Thus, we conclude that vp1 contributes to the development of biofilm thickness on epithelial cells.

Figure 3. VP1 enhances biofilm development on CMEE cells.

A. Representative images assembled from confocal micrographs of biofilms of WT, Δvp1, or Δvp1:vp1 grown on CMEE cells imaged 16h post-seeding. Top panels: confocal image of the lower end of the biofilm, just above the mammalian cells. Middle panels: 3D projection pseudo-colored to highlight biofilm thickness, according to scale on the left. Lower panels: side view of a biofilm reconstruction. Axis represent distance in mm. B. Biofilm biomass and average thickness were computed from confocal images. Data represent the mean ± S.E.M. of at least three independent experiments. Statistical significance was calculated using 2-way ANOVA analysis with Tukey correction. * P-value = 0.0039, ** P-value ≦ 0.0007, ns = not significant. C. Growth curve of WT, Δvp1, or Δvp1:vp1 in epithelial cell media demonstrating that differences in biofilm development are not due to differences in growth rates (n = 3). Scale bar is conserved across all figures.

Synthetic VP1 localizes to the pneumococcal surface

VP1 is part of the Gly-Gly family of peptides. Known Gly-Gly peptides are processed and secreted. Subsequently, their active form signals through kinases. Thus, we tested whether synthetic VP1 (labeled with FITC at the N-terminus) added extracellularly can bind to PMEN1 cells. As positive controls, we utilized two FITC-labeled peptides known to interact with PMEN1 cells. The first was the CSP pherotype CSP2, a Gly-Gly peptide that binds to the surface exposed histidine kinase ComD and initiates competence (Havarstein et al., 1995a). The second was PhrA2, a cognate peptide to TprA2 (RRNPP superfamily) that is proposed to be internalized into the cytosol (Kadam et al., 2017). As a negative control, we used FITC-CSP1, as this peptide does not bind to the PMEN1 ComD variant (Pozzi et al., 1996). The distribution of VP1 is distinct from the cytosolic localization of PhrA2 (Fig. 4A). Instead, we find that VP1 distribution resembles that of CSP2, both are observed on ~95% of the cells and dispersed on the cell surface (Fig. 4A, B). Further, the VP1 signal appears internal to capsular staining and external to DNA staining (Fig. S2). These localization experiments are consistent with surface localization of VP1.

Figure 4. VP1 peptide binds to the surface of pneumococcal cells.

A. Confocal micrographs of early biofilms. Left Panel: WT cells stained with FITC-labeled VP1, PhrA2, CSP2, CSP1, or no peptide. Peptides were added to 16h biofilms for 1h prior to fixation and imaging. Middle Panel: Indirect immunofluorescence using antibodies targeting the pneumococcal capsule. Fluorescence intensity across the width of the bacterial cell (white line) was measured on representative cells and results are plotted on the right panels. B. Quantitation of the percentage of cells positive for FITC-VP1 and FITC-CSP2. Statistical significance was calculated using unpaired two-tailed Student’s t-test. * P-value < 0.0001, ns = not significant.

The Δvp1 strain displays reduced biofilm biomass and thickness relative to the WT (Fig. 3). If VP1 exerts its function by binding to the pneumococcal surface to trigger a signaling cascade, the Δvp1 defect should be restored to WT by addition of exogenous VP1 to biofilm cultures. To test this prediction, we grew Δvp1 biofilms in the presence of synthetic VP1. The synthetic peptide rescued the biofilm defect in a dose-dependent manner (Fig. 5A, B). Together, these findings suggest that VP1 is a secreted peptide, which plays a role in biofilm development and exerts its effect by binding to the surface of producer and neighboring cells.

Figure 5. Synthetic VP1 rescues the Δvp1A biofilm defect.

A. Representative 3D projection of biofilms on CMEE cells, pseudo-colored to highlight biofilm thickness according to the scale on the left. (i) WT, (ii) Δvp1, (iii) Δvp1 + 2.5 nM sVP1, (iv) Δvp1 + 10 nM sVP1, (v) Δvp1 + 50 nM sVP1. B. Biomass and average thickness computed from micrographs. sVP1 stands for synthetic VP1 peptide. Data represent the mean ± S.E.M. of at least four independent experiments. * P-value = 0.0055, * P-value < 0.0001, ns = not significant; statistical significance was calculated using 2-way ANOVA analysis with Tukey correction.

The VP1 operon is regulated by a Rgg signal transduction pathway

To gather insight on the regulation of vp1, we analyzed the organization of the vp1 genomic neighborhood in strain PN4595-T23 (Fig. 6A). Immediately downstream of vp1 is a predicted protease belonging to the family of CAAX proteases and bacteriocin-processing enzymes (CPBP) and a predicted efflux pump, denoted vpoB (vp1 operon gene B) and vpoC (vp1 operon gene C), respectively. Another 453 bp downstream is a predicted azlC permease for transport of branched amino acids, denoted vpoD. Upstream of vp1 is a gene encoding a predicted rgg regulator. A gene encoding a predicted short hydrophobic peptide (shp) is transcribed from the complementary strand, where its sequence overlaps with that of rgg. To define the operon structure of this region, we performed PCR analyses on cDNA. The analysis suggests that rgg is one transcriptional unit, and that vp1vpoBCD is a second transcriptional unit (Fig. S3). No product was detected for primers spanning azlC and downstream genes, suggesting that azlC demarks the 3′end of the vp1 transcript. In silico analyses of the predicted promoter region for rgg and for the vp1 operon revealed putative CodY binding-boxes (“AACTATCAGAATATT” and “GATTTTCTAAAATAA”, respectively) (Fig. 6A). This is consistent with previous work in which a deletion of codY led to a 2.5 fold increase in the levels of rgg, vp1, vpoB and vpoC (Hendriksen et al., 2008a). Finally, a region resembling a Rgg-box (TTGAAGGAGTGTAA) is located upstream of vp1, suggesting that Rgg may regulate the vp1 operon. This genomic organization is consistent with a role for CodY in the negative regulation of the vp1 operon, and a regulatory link between rgg and vp1.

Figure 6. Genomic organization and vp1 regulation by a Rgg/SHP.

A. Genomic organization of vp1 genomic locus. The locus was adapted from a RAST annotation of the genomic region. Genes in the same color are in the same transcriptional unit, and IDs correspond to gene names in strain ATCC 700669. Thin rectangles correspond to predicted binding sites for CodY and Rgg. Predicted Rgg binding site is “TTGAAGGAGTGTAA”. Putative functions are indicated below each gene. B. qRT-PCR analysis of transcript levels in the Δrgg strain relative to WT strain of vp1 and neighboring genes relative to gapdh. C. Analysis of transcript levels in WT and WT + 0.68 mM SHP synthetic peptide. * P-value ≤ 0.0001 and ** P-value = 0.0002 with n = 3. Normalized to levels of gapdh. Statistical significance was calculated using 2-way ANOVA analysis with Bonferroni correction. The figure displays the IDs in strain ATCC 700669, the corresponding IDs in strain PN4595-T23 are: CGSSp4595_0133 (SPN23F01510), CGSSp4595_0134 (SPN23F01520), CGSSp4595_0135 (SPN23F01530), CGSSp4595_0136(SPN23F01540), CGSSp4595_0137 (SPN23F01550), CGSSp4595_0138 (SPN23F01560), and CGSSp4595_0139 (SPN23F01570).

The relative position of rgg and predicted shp suggest their gene products function as a cognate pair. In other streptococcal species, SHPs bind to their cognate Rgg activating transcription (Fleuchot et al., 2011). In PMEN1, the predicted shp encodes a twenty-five-amino acid peptide (MKKQILTLLKIVAEIIIILPFLTNR) and is encoded in the opposite strand to and partially overlaps rgg. Moreover, the relative location of vp1 and rgg combined with the presence of a putative rgg-box upstream of vp1 are consistent with regulation of vp1 by Rgg. We hypothesized that Rgg activates vp1 expression, and that activation is SHP-dependent. To test this hypothesis we first searched for an in vitro condition with high levels of vp1 expression; we tested rich media (Columbia broth) and chemically defined media supplemented with glucose (CDM-Glu) (Carvalho et al., 2013), as well various stages of growth (Fig. S4). The highest level of vp1 expression was observed in CDM-Glu at early-log phase. Thus, using this condition, we compared the expression levels of the genes in the vp1 transcriptional unit between WT and Δrgg strains. All four genes were five-fold lower in the deletion mutant relative to the WT, suggesting that Rgg is a positive regulator of the vp1 operon (Fig. 6B). Given that SHPs induce Rgg activity (Chang et al., 2011), we tested the effect of SHP on gene expression. Addition of exogenous SHP to the WT strains leads to over two-fold induction of all the genes in the vp1 operon (Fig. 6C). Together, these data suggest that a cognate Rgg/SHP pair positively regulates vp1 expression.

There are multiple VP1 alleles distributed in multiple streptococcal species

The intra and inter species influence of vp1 depends on the distribution and variation of vp1 alleles and its receptor. Epidemiologic studies reveal that infection with multiple pneumococcal strains occur in an estimated 20–40% of infections (Everett et al., 2012; Rodrigues et al., 2013; Saha et al., 2015). Furthermore, multiple streptococcal species, such as S. mitis and S. oralis, commonly colonize the human nasopharynx and there are many examples of DNA exchange between pneumococcus and these species (Chi et al., 2007; Croucher et al., 2012; Dowson et al., 1993; Eutsey et al., 2015; Johnston et al., 2013; Valentino et al., 2014). We have analyzed the distribution of vp1 across pneumococci and related species: vp1 is widely distributed in pneumococcal strains (Fig. 7). Using thirty-five pneumococcal genomes from nineteen serotypes and thirty-six MLSTs, we determined that VP1 is encoded in thirty-three strains (Table S1). The exceptions are strain CGSSp14, where the region encoding VP1 appears to be replaced with a lantibiotic biosynthetic locus, and strain CDC10870 where it appears to be replaced with a domain from a different pneumococcal peptide (KGI31098.1). The thirty-three sequences can be grouped into three allelic groups (Fig. S5). In all instances the alleles are syntenic, with rgg upstream and vpoBC downstream. Some strains encode two adjacent copies of the protease (vpoB). This distribution suggests that the function of VP1 is widely conserved. Further, signaling may be limited to subsets of strains based on specific interactions between each variant and its cognate receptor.

Figure 7. Distribution of vp1 alleles in Streptococcus pneumoniae and related streptococci.

Specie and strain conservation of vp1. Maximum likelihood phylogenetic tree of streptococcal genomes was generated from the core genome with bootstrap values displayed on the branches. Green circles, red squares, and blue stars represent distinct allelic types; black triangles represent multiple allelic types identified outside of pneumococcal strains. Species are colored coded as follows: gray, S. infantis; beige, S. oralis; green, S. mitis; pink, S. pseudopneumoniae; light blue, distinct phyletic group of S. pneumoniae; blue, S. pneumoniae.

The distribution of vp1 in related streptococci is highly variable (Fig. 7, Fig. S6). Two out of three S. pseudopneumoniae genomes encodes a vp1 allele, and one differs from those captured in the pneumococcus genomes. Approximately half of the S. mitis genomes encode vp1 alleles; in this species, there are multiple alleles including some observed in pneumococcus. This gene appears less frequently in the S. oralis clade, where only one S. oralis (S. tigurinus subspecies) encodes a vp1 allele. The absence of a clear correlation between allelic type and phylogeny of the species is consistent with gene transfer of the vp1 region. The wide distribution of this gene, suggests vp1 is a key component of pneumococcal virulence and dissemination.

Discussion

Our in silico screen suggests that pneumococci utilize multiple cell-cell communication peptides during infection. Pneumococcal CSP and BIP control competence and bacteriocidal activities via cell-cell interactions and their role in multi-strain communities is well documented (Claverys et al., 2006; Claverys and Havarstein, 2007; Dawid et al., 2007; Havarstein et al., 1995a). In this study, we present peptides predicted to be part of the pneumococcal secretome; all encode a Gly-Gly secretion motif consistent with export and cell-cell interactions (Table 1). We characterize VP1 as a signaling molecule, highly expressed in vivo, and with a role in biofilm formation and pathogenesis.

Multiple lines of evidence support the hypothesis that vp1 and the genes in the vp1 operon sense amino acid levels in a cellular community. First, vp1 and downstream genes are negatively associated with levels of glnA and glnP, which are critical players in glutamine/glutamate metabolism (Hendriksen et al., 2008b). Second, genes in the vp1 operon are negatively regulated by the nutritional regulator CodY, and CodY-mediated inhibition is enhanced in the presence of branched-amino acids (Hendriksen et al., 2008a). Third, vpoD belongs to a protein family predicted to function in the transport of branched-chain amino acids (AzlC, score: 1e⁻³¹). It will be fascinating when future work reveals the extracellular signal(s) that control the high vp1 levels we have observed in vivo.

Rgg (SPN23F01510), the positive regulator of vp1vpoBCD, has multiple molecules controlling its transcription and activity. Analogous to the vp1 operon, rgg is also negatively regulated by CodY and glutamine/glutamate metabolism (Hendriksen et al., 2008a, 2008b). In this manner, it appears that Rgg amplifies the regulation of vp1 operon. Presumably, once inhibition by these nutritional regulators is released, transcription of rgg is activated and serves to further enhance transcription of the vp1 operon. While it is likely that rgg receives transcriptional input from other sources, these remain to be identified. Regarding the activity of Rgg, we observe that it is modulated by SHP. The predicted promoter region for shp is on the opposite strand and overlapping with the vp1 promoter, and appears to encode a putative CodY-box. In this manner, SHP levels may also be under the control of this nutritional regulator and/or amino acid levels. Rgg/SHP quorum-sensing systems have been extensively characterized in Group A streptococci (Aggarwal et al., 2015, 2014; Chang et al., 2011), and a Rgg has been characterized in pneumococcus and implicated in response to oxidative stress (Bortoni et al., 2009). Based on the GAS orthologues, we predict this SHP is processed and secreted from producer cells, accumulates in the extracellular environment in a density-dependent manner, and is imported by the AmiABCDE transporter (Fleuchot et al., 2011). Once in the cytosol, SHP presumably binds Rgg, stimulates this positive regulator and induces transcription of the Rgg regulon (Fig. 8).

Figure 8. Model for VP1 regulation and for cell-cell communication in Streptococcus pneumoniae.

Regulation: Rgg is a positive regulator of vpo1vpoBCD. Rgg activity is induced upon binding to its cognate SHP. We predict that SHP is a quorum-sensing molecule, activated during export and functional upon internalization. Red: rgg, green: shp. Signaling: VP1 is processed and secreted into the extracellular milieu, from where it binds to the surface of producing and neighboring cells. VP1 influences biofilm and virulence. We propose VP1 binds to a surface molecule triggering a signaling cascade. Blue: genes in the vp1 operon, and blue circle: VP1 peptide.

The organization of the vp1 operon is conserved across strains and species, further supporting the functional association between these co-regulated genes. Encoded immediately downstream are a transmembrane protease (SPN23F01530) and a transporter (SPN23F01540) (Fig. 6A). The protease is part of the Abi protease family (Pfam: Abi, PF02517, score: 7.3e⁻¹⁷). The presence of a protease is common to many Gly-Gly peptides (Dawid et al., 2007); while their function is unknown, these proteins are presumed to play a role in self-immunity against bacteriocins. This functional annotation is consistent with a role for this protease in VP1 regulation via control of degradation. The vpoC encodes a transporter from the major facilitator superfamily 1 (PF07690; score: 3.6e⁻¹¹). The functional link between this transporter and VP1 is still unclear. We predict that it is not part of the VP1 import or export machinery. The Gly-Gly motif in VP1 is consistent with export via an ABC-transporter. Similarly, the peptide family combined with the surface localization of FITC-VP1 suggests VP1 is not internalized but instead binds to the cell surface. The MSF1 transporter (Pao et al., 1998) may be related to controlling molecules upstream or downstream of VP1. The role of the molecules in the VP1 locus and the identity of the VP1 receptor are currently under investigation.

Comparative genomics of pneumococcal strains revealed three distinct groups of VP1 alleles, as well as additional alleles in related streptococci (Fig. 6B). The non-clustered distribution of alleles across the species tree is consistent with extensive vp1 gene transfer across both strains and species. The presence of variants raises the question: how do these alleles interact? For signaling and bacteriocidal molecules, specificity has been documented between the signal and the receptor and between the bacteriocins and immunity system, respectively (Capra et al., 2010; Dawid et al., 2007; Maricic et al., 2016). Further, specificity has also been observed for bi-functional molecules that play a role in both cooperation and competition (Anderson et al., 2014). Thus, it is likely that the allelic variations in vp1 are important in multi-strain infections by increasing cooperative behaviors across strains that share the same allele, and/or increasing competitive behaviors across strains with diverse alleles.

In this study, we have characterized VP1 as a signaling molecule. Our findings invite exciting questions regarding VP1 regulation and function. What are the in vivo conditions that activate vp1 expression? What are the downstream components required for VP1 signaling? What effector molecules are downstream of this signaling? Is VP1 a good target for anti-infective therapies? Future work on VP1 will provide answers to these questions, and in doing so should advance our understanding of pneumococcal biology in the context of human infections.

Experimental Procedures

In silico screen for predicted secreted peptides

To generate a set of predicted secreted peptides we utilized MEME to generate a position-dependent probability matrix from a set of secreted peptides and MAST to search for the MEME motif in a database of streptococcal genomes (Bailey et al., 2009). The MAST database was comprised of a set of 134,662 sequences in sixty genomes: thirty-nine S. pneumoniae, three S. pseudopneumoniae, nine S. mitis, seven S. oralis strains, and two S. infantis strains. The thirty-nine pneumococcal genomes capture nineteen serotypes and thirty-six MLST types, and include isolates from North America, South America, Africa, Europe, and Asia (Table S1). This set was assembled in multiple steps. First, we started from a large-scale pneumococcal pangenome study and selected at least one genome from each branch of a maximum likelihood tree build from the core genome (Donati et al., 2010). To this set of twenty-six genomes, we added two genomes (WL677 and WL400) isolated from PCV-7 immunized children to capture samples of vaccine escape strains emerging in the post-vaccine era (Frazão et al., 2013). Further, recent studies have revealed that a subset of non-encapsulated isolates form a distinct phyletic group basal to other pneumococcal strains (Croucher et al., 2013; Valentino et al., 2014). Thus we added isolates from this group, ending up with thirty-two pneumococcal genomes (Antic et al, unpublished) (Keller et al., 2013) (light blue, Table S1). Finally, for the MAST, we added seven genomes from isolates from eye infections and non-encapsulated strains that are distributed throughout the phylogenetic tree (Antic et al, under review). Thus, the final set of pneumococcal strains for our MAST search corresponds to thirty-nine genomes. Previous phylogenetic work on pneumococcus and viridans reveals that pneumococcus, S. pseudopneumoniae, and S. mitis form a large lineage, which is well-separated from S. oralis and S. infantis (Kilian et al., 2014). Using this phylogenetic work as a guide we selected twenty-one strains distributed with these four non-pneumococcal species.

Our first MEME input consisted of 167 homologues of CSP and BIP. This set was obtained using NCBI blastP (e-value cutoff of 1e⁻⁴) to search the non-redundant protein database using query sequences of pneumococcal CSP and pneumococcal BIP. The MEME motif was normalized to the amino acid frequency of the coding sequences in the database of sixty genomes. Next, we employed MAST to search for the MEME motif in the database of streptococcal genomes. Importantly, this method is limited to annotated coding sequences, such that peptides that have not been predicted by annotation programs will be overlooked in this analysis. The output of this MAST search consisted of 192 coding sequences. This expanded set of sequences was used as input to generate a second MEME motif, and subsequently search the same database with MAST (Table S2). The reiterative use of MEME and MAST converged on a list of 716 proteins, of which 696 encode the motif at the N-terminus and 98% are coding sequences less than 130 residues in length. The strong enrichment of the motif at the N-terminus of short coding sequences is consistent with the hypothesis that the identified motif function in the secretion of small peptides.

The next step was to organize the 696 coding sequences into clusters based on sequence similarity and conservation of the genomic locus. To this end, we generated a maximum parsimony tree from this predicted secretome, using iQ-TREE and the cpREV amino acid exchange model (Trifinopoulos et al., 2016). The branches were manually curated using RAST to visualize the genomic neighborhoods and to ensure that alleles in the same groups shared the same genomic location (Overbeek et al., 2014). One cluster was curated out of the set based on its annotation as a v-type ATP synthase subunit G. This analysis revealed 25 clusters (Table 1).

Bacterial strains and culture conditions

Wild-type S. pneumoniae strain PN4595-T23 (GenBank ABXO01), graciously provided by Drs. Alexander Tomasz and Herminia deLancastre, was used as a PMEN1 representative (Hiller et al., 2011). This isolate was recovered in 1996 from the nasopharynx of a child in Portugal, and corresponds to a low passage number. We avoided strain ATCC 700669, as we have evidence that our aliquot of this strain has incurred mutations during serial passage (unpublished). Strain PN4595-T23 was modified to generate a vp1 deletion mutant (Δvp1), the vp1 complemented strain (Δvp1:vp1), and a rgg deletion mutant (Δrgg) (Table S4). For growth on solid media, strains were streaked onto Trypticase Soy Agar II plates containing 5% sheep blood (BD BBL, New Jersey, USA). For growth in liquid culture, colonies from a frozen stock were grown overnight on TSA plates and inoculated into Columbia broth (Remel Microbiology Products, Thermo Fisher Scientific, USA). The media was supplemented with antibiotics as needed. Cultures were incubated at 37°C and 5% CO2 without shaking. Experiments in chemically defined media (CDM) were performed utilizing previously published recipe (Carvalho et al., 2011), and glucose was used at a final concentration of 55mM. Growth in CDM was initiated by growing a pre-culture in Columbia broth for 2–3 hours and back dilution to OD₆₀₀ 0.1 to initiate a culture.

Growth curves and CFU counts

Bacteria were inoculated from overnight plates into 30 ml of liquid cultures, and samples were taken at constant intervals to evaluate bacterial growth. Liquid cultures were grown statically and monitored by optical density at 600 nm (OD₆₀₀) using Nanodrop 2000c spectrophotometer (Thermo Scientific). Bacterial counts were assessed by streaking dilution on TSA or TSA-antibiotic plates.

Strain construction

The Δvp1 and Δrgg strains were generated by site-directed homologous recombination where the target region was replaced with the spectinomycin-resistance gene (aadR) (Al-Bayati et al., 2017; Eutsey et al., 2015). The spectinomycin-resistance gene was inserted on the complementary strand, to decrease the likelihood of polar effects. The Δvp1:vp1 complemented strain was generated by inserting the vp1 gene fused to the kanamycin resistance cassette into the spectinomycin cassette of the Δvp1 strain. Briefly, approximately 2kb of flanking region upstream and downstream of the deletion target were amplified from the parental strain by PCR using Q5 2× Master Mix (New England Biolabs, USA) generating flanking regions, and the spectinomycin resistant gene was amplified from the plasmid pR412 (kindly provided by Dr. Donald Morrison). Assembly of the transforming cassette was achieved by either sticky-end ligation of restriction enzyme-cut PCR products or by Gibson Assembly using NEBuilder HiFi DNA Assembly Cloning Kit. The resulting constructs were transformed into PN4595-T23 and confirmed using PCR and DNA sequencing. Primers used to generate the constructs are listed in Table S5.

Bacterial transformations

For all bacterial transformations, approximately 1μg of transforming DNA was added to the growing culture of the target strain at OD₆₀₀ of 0.05, supplemented with 125 μg ml⁻¹ of CSP2 (sequence: EMRISRIILDFLFLRKK; purchased from GenScript, NJ, USA), and incubated at 37°C. After 4 hours, the treated cultures were plated on Columbia agar containing the appropriate concentration of antibiotic for selection, spectinomycin, 100 μg ml⁻¹, kanamycin 150 μg ml⁻¹). Resistant colonies were cultured in Columbia broth, the region of interest was amplified by PCR and the amplimer was submitted for Sanger sequencing (Genewiz, Inc., USA) to verify the sequence of the mutants. The strains generated in this study are listed in Table S4.

Treatment with synthetic peptides

Bacterial biofilms were treated with synthetic peptide labeled at the N-terminus with FITC and custom ordered from GenScript, (NJ, USA) at 90% purity. The peptides are as follows: VP1 (FGTPWSITNFWKKNFNDRPDFDSDRRRY), CSP1 (EMRLSKFFRDFILQRKK), CSP2 (EMRISRIILDFLFLRKK), and PhrA2 (VDLGLAD). Each peptide was independently added during biofilm seeding, and biofilms were incubated at 37°C, 5% CO₂ for 16 hours. For experiments where different concentration of peptides were compared in parallel, the original culture was seeded onto separate CMEE cultures, and each one was treated with the relevant peptide concentration, in addition to a no-peptide control. Using a single parent culture ensured minimal variation across comparing treatments. Planktonic cultures were treated with synthetic SHP (MKKQILTLLKIVAEIIIILPFLTNR) custom ordered from GenScript, (NJ, USA) at 90% purity for 1 hour.

Studies in the chinchilla OM model

Young adult chinchillas (Chinchilla laniger) weighing 400–600 g were obtained from R and R Chinchilla Inc., Ohio. All chinchilla experiments were conducted with the approval of the Allegheny Singer Research Institute (ASRI) Institutional Animal Care and Use Committee (IACUC) at the Allegheny Hospital Animal Facility. Cohorts of ten chinchillas were used for each strain, and maintained in BSL2 facilities. Animals were anesthetized on day 0 by subcutaneous injection of 0.1 ml of a sedative solution (ketamine hydrochloride 100 mg ml⁻¹, xylazine hydrochloride 30 mg ml⁻¹ and acepromazine 5 mg ml⁻¹). For inoculation, 100 μl of pneumococcal suspension containing 100 CFUs was injected bilaterally into the tympanic bullae. Animals were maintained in observation for up to 10-days post-inoculation. During these ten days animals with severe acute infection perished and animals with severe signs of pain and illness were euthanized by administering an intra-cardiac injection of 1 ml potassium chloride after regular sedation. Animals showing prolonged signs of discomfort were administered with pain relief (Rimadyl, 0.1 ml of 50 mg ml⁻¹). We evaluated mortality, time to death, and spread of bacteria to the brain and the lungs. Tissue dissemination was tested by plating homogenized tissue on TSA plates with 5% sheep blood to establish pneumococcal presence. Handling, euthanasia, necropsy, and sample collection was performed in accordance with the IACUC protocols.

Quantification of pneumococcal gene expression during infection

For RNA extraction from in vivo experiments, chinchillas were euthanized 48h post-inoculation of PN4595-T23. Effusions were siphoned out from the middle ear though a small opening generated in the bulla, and flash frozen in liquid nitrogen to preserve bacterial RNA. Lysis was performed by resuspending the cells in an enzyme cocktail (2 mg ml⁻¹ proteinase K, 10 mg ml⁻¹ lysozyme and 20 μg ml⁻¹ mutanolysin), and bead beating with glass beads, acid-washed 425–600 μm (Sigma) and 0.5mm Disruption Beads made by Zirconia/Silica in FastPrep®-24 Instrument (MP Biomedicals, USA). These lysates were frozen for nanoString analyses. RNA extraction from in vitro mid-log planktonic cultures were performed as previously described (Kadam et al., 2015). The RNA concentration was measured by NanoDrop 2000c spectrophotometer (Thermo Fisher Scientific, USA) and its integrity was confirmed on gel electrophoresis, final concentrations were approximately 200 ng/μl for in vivo samples and 60 ng/μl for in vitro samples. RNA was hybridized onto the nCounter chip according to the manufacturer recommendations. RNA counts were normalized to the geometric mean of the housekeeping genes gyrB and metG (Carvalho et al., 2011; Kim et al., 2013) using manufacturer’s software, nSolver. The in vitro and in vivo levels were compared using Student’s t-test in the GraphPad Prism 6 tool and the heat map was generated using nSolver.

RNA purification, RT-PCR and qRT-PCR

Bacterial samples were collected at different ODs, and RNALater (Thermofisher) was used to ensure RNA preservation and quality. Next, bacterial pellets were lysed with 1× enzyme mix (lysozyme, proteinase K, and mutanolysin) in TE buffer (10 mM Tris·Cl, 1 mM EDTA, pH 8.0) for at least 10 min. Finally, total RNA was isolated using the RNeasy kit following manufacturer instructions. Contaminant DNA was removed by incubating total RNA samples with DNase (2U/μl) 37°C for at least 15 min. Any remaining DNA contamination was checked by PCR of gapdh (no visible band should be observable in RNA only samples). RT reaction was performed using 1 μg of total RNA using SuperscriptVILO kit for 1 h. Products were amplified using either Q5 polymerase or OneTag polymerase (New England Biolabs). For quantitative RT–PCR analysis, 5 ng of total cDNA was subjected to real-time PCR using PowerUp SYBR Green Master Mix in the ABI 7300 Real Time PCR system (Applied Biosystems) according to the manufacturer’s instructions. All qRT-PCR amplification was normalized to pneumococcal gapdh and expressed as fold change with respect to WT strain and untreated bacterial cells. PCR and qPCR primers are listed in Table S5. Primers were obtained from IDT.

Mammalian cells

3T3 cells were generously obtained from Dr. Jonathan W. Jarvik (Carnegie Mellon University). Cells were propagated in DMEM media (Corning), supplemented with 5% of FBS and Glutamine 20 mM (Corning). 3T3 cells were maintained at 37º C and supplemented of 5% carbon dioxide at pH 7. 3T3-conditioned media was filtered using 0.45 μm pore size filter and store at 4ºC. Cells were propagated as previously described (Raffel et al., 2013). Briefly, cells were grown on Seeding Media (DMEM/High Glucose, Ham’s F12, HEPES 25 mM, Glutamine 2 mM, Hydrocortisone 0.4 μg ml⁻¹, Isoproterenol 2 μg ml⁻¹). Sixteen hours later, the media was replaced with 3T3-conditioned media containing EGF 10 ng ml⁻¹ (Sigma) and supplemented with penicillin/streptomycin. For biofilm assays, fresh CMEE media was utilized. Bacterial cells were cultured in Nunc Lab-Tek 4-chamber slides or glass-bottomed dishes (MatTek Corporation) coated with an organic polymeric matrix (100 μg ml⁻¹ BSA (Sigma), Collagen type 1 rat tail 20 μg ml⁻¹ (Millipore), 10 μg ml⁻¹ fibronectin rat tail (Biomedical Technologies), in phosphate-buffered saline (PBS) to facilitate CMEE cell adhesion. All assays involving co-cultivation with bacteria were carried in absence on antibiotics.

Biofilm formation assays

Starter cultures of pneumococcal strains were inoculated from frozen glycerol stocks into Columbia broth, serially diluted in the same media to an OD₆₀₀ of 0.010–0.015 to start final cultures. When cultures reached an OD₆₀₀ of 0.025, each bacterial strain was seeded in tissue culture plates containing attached, confluent CMEE cells. To promote biofilm growth, the plates were incubated aerobically in humidified conditions at 37º C and 5% carbon dioxide for 16 h. Then, the supernatants were carefully aspirated, and biofilms were washed with PBS to remove non-adherent and or weakly adherent bacteria. Subsequently, samples were fixed with 4% PFA (Electron Microscopy Sciences). The fixing solution was removed; samples were washed twice with PBS and prepared for confocal microscopy.

Confocal microscopy

Biofilms were prepared as mentioned above. Briefly, biofilms were fixed and stained with SYTO59 dye for 30 min. Biofilms were washed and kept in PBS buffer before imaging. Confocal microscopy was performed on the stage of Carl Zeiss LSM-510 META DuoScan or MP ConfoCor 3 confocal microscopes, using 561 nm laser line for SYTO59 dye. Stacks were captured every 0.5 or 1 μm and reconstructed in Carl Zeiss Black Edition and Fiji (ImageJ. 2.0.0rc48). Laser intensity and gain were kept the same for all images. The fluorescence FITC-VP1 peptide was incubated with 16h-preformed biofilms of the WT strain PN4595-T23. Then biofilms were fixed and stained with rabbit anti-pneumococcal capsule antibody (AbD Serotec) and detected with goat anti-rabbit IgG (Alexa Fluor 568). Confocal microscopy was performed on the stage of Carl Zeiss LSM-880 META FCS, using 488 nm laser line for the FITC fluorescence tag and 561 nm laser line for the fluorescence Alexa tag. RGB scanning profiling was employed to validate the peptides localization relative to the capsule marker using Fiji (ImageJ. 2.0.0rc48). Laser intensity and gain were kept the same for all stacks and images.

Statistical analysis

Statistical significance was assessed using Student’s two-tailed t-test for independent samples, Logrank test for animal survival experiments (Mantel-Cox method), One-tail Fisher’s Exact Test for bacterial systemic dissemination, and Two-way ANOVA with Tukey multiple comparison tests. p values less than 0.05 were considered statistically significant.

VP1 strain distribution and promoter analysis

To determine the distribution of vp1 across pneumococcal strains we used blastP to search for the VP1 sequence in a database of streptococcal genomes (Table S1). This database includes thirty-two genomes that include nineteen serotypes and thirty-six MLST types (as detailed in “In silico screen for predicted secreted peptides”). Further, since our work was performed in PMEN1 we expanded the set with three additional PMEN1 genomes (Hiller et al., 2011). Sequences were aligned with Jalview (Waterhouse et al., 2009) revealing three putative pherotypes in pneumococcus as well as four additional variants outside of pneumococcus. To confirm the organization of VP1 sequences, all the coding sequences were clustered using a maximum parsimony tree in iQ-TREE; it confirmed the organization into seven types (Fig. S6). The VP1 alleles are presented in the context of a maximum parsimony species tree, generated from the core region of streptococcal genomes (Kadam et al., 2017). Each allelic type is colored-coded, and the visualization was generated using the Interactive Tree of Life (iTOL) (Letunic and Bork, 2016).

We searched in silico for Rgg and CodY binding regions. The putative Rgg-binding box upstream of the vp1 operon was identified using the consensus for the Rgg-box (as presented in RegPrecise (Novichkov et al., 2013)). Specifically, we identified “TTGAAGGAGTGTAA” at position −17 of the vp1 gene. Next, given that the orthologs of rgg and vp1 in D39 (spd0144 and sp0145, respectively) are differentially expressed in a codY deletion mutant relative to the WT strain (Hendriksen et al., 2008a), we searched for a CodY-box upstream of these genes. In rgg, a CodY binding box is identified upstream of Rgg using RegPrecise on the TIGR4 genome (the representative pneumococcal genome in this database); it corresponds to the sequence “AACTATCAGAATATT” (score of 4.7) and an identical region is present at position −69 of rgg in PN4595-T23. In vp1, the sequence “GATTTTCTAAAATAA” at position −113 closely resembles the L. lactis CodY consensus as well as predicted CodY binding boxes in pneumococcus (den Hengst et al., 2005).

Supplementary Material

Supplementary Materials

Figure S1: Effect of CMEE cells presence on biofilm capacity of strain PN4595-T23. Comparison of 16h biofilms grown in CMEE media with or without pre-cultured basal layer of epithelial cells. A. Representative 3D images assembled from confocal images from biofilms seeded on glass or CMEE cells. The 3D projections are pseudo-colored to highlight biofilm thickness, as described in the scale. B. Biofilm biomass and average thickness were computed from confocal images. Data represent the mean ± S.E.M. of at least three independent experiments. Statistical significance was calculated using unpaired two-tailed Student’s t-test. * P-value < 0.0001.

Figure S2: VP1 peptide binds to the surface of pneumococcal cells. Fluorescence intensity across the width of the bacterial cell (white line) was measured on representative cells and compared to capsular staining (red anti-capsule) and DNA straining (blue DAPI). Red green blue (RGB) profiles (right panel) show that VP1 appears to bind internally to the capsule and externally to DNA, consistent with membrane binding.

Figure S3: PCR performed on cDNA and genomic DNA to demonstrate transcriptional units. Prior to cDNA synthesis, all RNA samples were DNase-treated and subjected to a PCR check using primers for gapdh gene to ensure total elimination of DNA. Only when no amplification was observed in the gapdh check PCR was the cDNA synthesized. Lanes 1–5 and 11–14 are PCRs on cDNA template, and lanes 6–10 and 15–18 on gDNA template. Primes were labeled according the locus they bind to. A. Primers are as follows: lanes 1 and 6, rggF fwd and rggF rev; lanes 2 and 7, rgg-a fwd and vp1-a rev; lanes 3 and 8, vp1-b fwd and vpoB-a rev; lanes 4 and 9, vp1-b fwd and vpoC-a rev; lanes 5 and 10, vpoC-a fwd and vpoD-a rev. B. Primers as follow: lanes 11 and 15, spn23f01560 fwd and spn23f01570 rev; lanes 12 and 16, vpoC fwd and vpoD rev; lanes 13 and 17, vpoCb fwd and spn23f01570 rev; lanes 14 and 18, vpoD fwd and spn23f01570 rev. C. Genomic locus schematic indicate the primer binding sites corresponding to the bands on the gel. Solid lines indicate expected amplification products. Dashed lines indicate no observed amplification product. Numbers indicates lane numbers from A and B.

Figure S4: Determination of experimental condition for high levels of vp1 expression. Bacterial were pre-cultured in rich media (Columbia broth) and switched to CDM-Glu. RNA samples were isolated at 1h, 2h, or 3h after the switch and relative expression was measured by qRT-PCR. Expression was normalized to levels of gapdh. Data represent the mean ± S.E.M. of at least two independent experiments. Statistical significance was calculated using unpaired two-tailed Student’s t-test. * P-value < 0.05 and ** P-value < 0.01.

Figure S5: Alignments of VP1 sequences from Streptococcus pneumoniae. Residues are colored according to Zappo’s color scheme. Black box indicates predicted mature peptide. Red arrow indicates predicted cleavage site. Products of vp1 alleles 1, 2 and 3 are shown.

Figure S6: Alignment of seven VP1 sequences from Streptococcus sp. All share the N-terminal motif, predicted to function in processing and export. Top three variants represent coding sequences present in pneumococcal genomes: D39, Taiwan19A, and ATCC 700669 respectively. The next four variants are from related species: S. pseudopneumoniae IS7493, S. mitis SK564 and SK569, and S. oralis subspecies tigurinus 2426). The overall conservation of the positively charged C-termini suggests this region is critical for VP1 signaling.