We previously characterized the PI-2a locus encoding a heterotrimeric pilus in the ST23 strain NEM316 [14, 15]. Like its allelic counterpart, the PI-2b pilus carries 3 genes coding for the structural subunits (spb1, ap1 and ap2), 2 genes (srtC1 and srt2) expressing Srt enzymes, and two additional genes (orf and lep) encoding a small conserved gene of unknown function and a putative signal peptidase, respectively (Fig 1A).

Surface distribution of the major PI-2b pilin Spb1 in strains A909 and BM110 was analysed by immunofluorescence. Spb1 level on the bacterial surface was higher in A909 than in BM110. As shown for other cell-wall anchored proteins, Spb1 preferentially accumulates at the cell poles in A909, whereas in BM110 only 2 to 4 distinct spots per bacterium were observed (Fig 1B). As expected, Spb1 protein was undetectable on the surface of A909Δspb1 and BM110Δspb1 (Fig 1B). Spb1 expression at the single cell level was further characterized using flow-cytometry (Fig 1C). As shown in Fig 1C, Spb1 is expressed homogenously in both strains, and Spb1 level is about 3 to 5 times higher in A909 than in BM110.

Finally, expression of PI-2b was analysed at RNA level by quantitative reverse transcriptase polymerase chain reaction (qRT-PCR) using primers Spb1/Spb2 specific of spb1 gene (S2 Table). As negative controls, we used the mutants BM110Δspb1 and A909Δspb1 and the housekeeping gyrA gene for normalization. Transcription of spb1 was about 6-fold higher in A909 as compared to BM110 (Fig 1D).

Collectively, these results demonstrated that expression of Spb1 was higher in A909 than in BM110. This 4- to 6-fold difference in the levels of transcription and surface protein expression suggests a difference in the regulation of PI-2b expression in these two strains.

Alignment of the PI-2b locus sequence and its immediate genomic environment in BM110 and A909 revealed a strong conservation (Fig 2A). Except for ap1 encoding the minor pilin of unknown function, the other genes of the PI-2b locus are highly conserved encoding proteins displaying from 98.5 to 100% identity. According to the published A909 genome sequence, AP1 exhibits a premature STOP codon at amino acid 280, leading probably to a non-functional protein. However, re-sequencing of the gene ap1 (sak1441) in strain A909 revealed a missing base at position 839 which suppresses the premature stop codon (our unpublished data). As opposed to the PI-1 and PI-2a pili characterized in GBS, no regulatory gene was found in the vicinity of PI-2b locus (S1 Fig). Immediately upstream from the PI-2b operon lies another large operon of 16 genes (sak1438-1460 in A909) proposed to encode the group B carbohydrate also known as antigen B (AgB) [16].

RT-PCR analyses revealed the presence of a read-through transcript encompassing the last gene (sak1445) of AgB operon and the first gene (orf) of PI-2b locus in strain A909 but not in strain BM110 (Fig 3). This correlates with the absence of a predicted rho-independent terminator sequence in strain A909.

Transcriptional Start Sites (TSS) were identified at the genome level in GBS strains A909 and BM110 using a differential RNA-seq approach as previously described for strain NEM316 [17]. A common TSS was identified 37 bp upstream of the start codon of the first gene (orf) of the PI-2b locus in both strains (Fig 4A). In addition, increased transcription was detected in the intergenic region lying between the AgB and PI-2b operons in strain A909, but not in strain BM110 (Fig 4A), which further suggests read-through transcription in strain A909. In contrast a sharp decrease in coverage is observed in strain BM110 upstream from the 43-bp sequence and is in agreement with the proposed role of this hairpin as transcriptional terminator of AgB operon.

Specific mapping of PI-2b TSS in A909 and BM110 was also performed by primer extension analysis. A single signal located 37 bp upstream from the translation initiation codon of orf was detected in both strains (Fig 4B). The sequence lying upstream of this TSS displayed a canonical extended -10 (TGATATAAT) and semi-canonical -35 (ATGAGT) boxes separated by 16 bp (Fig 4C).

To map precisely the Ppi2b promoter region, fragments encompassing different domains of A909 PI-2b intergenic region (Fig 5A) were cloned upstream from the reporter gene encoding gfp using the vector pTCV-GFP, a low-copy number plasmid and assayed for GFP expression in the non-PI-2b GBS strain NEM316 (ST23). The various constructs are drawn schematically on Fig 5A. Plasmid pTCV4 was used as negative control. Quantification of GFP fluorescence was performed using flow-cytometry in exponentially growing bacteria in TH broth supplemented with erythromycin for plasmid maintenance.

A detectable GFP expression was seen with a 125-bp fragment containing the promoter Ppi2B and about 60 bp upstream from the TSS (pTCV1). Higher level of GFP expression was observed with a larger 406-bp fragment including the whole intergenic region of Ppi2b (pTCV3) (Fig 5B). No detectable GFP expression was seen with a fragment encompassing the region upstream without Ppi2b (pTCV2) which indicated that this segment did not contain another uncharacterized promoter.

To determine if transcription driven from PI-2b promoter region was dependent on GBS-specific regulatory factors, the various transcriptional fusions (pTCV1 to 4) were introduced into Lactococcus lactis strain NZ9000, a non-pathogenic Gram-positive species belonging to the Streptococcaceae family. As shown in Fig 4B, pTCV1 and pTCV2 gave a signal similar to the negative control. A very weak GFP fluorescence activity was measured with the largest promoter fragment (pTCV3). This result suggests that GBS-specific regulatory factors lacking in L. lactis are required for maximal transcription of PI-2b locus.

CovR was previously shown to be the master regulator of virulence genes in GBS [18, 19]. CovR binds to target DNA sequences when phosphorylated at amino acid D53 by cognate histidine kinase CovS [20]. To assess the contribution of CovR in PI-2b expression, we introduced the pTCV3 gfp plasmid in NEM316ΔcovR and CovRD53A mutants (Fig 5C). A 2- to 3-fold increase in GFP expression driven from the Ppi2b promoter was consistently observed in NEM316ΔcovR as compared to WT NEM316 (Fig 5C). However, this effect was not apparent in CovRD53A mutant (Fig 5C). To test the role of covR directly in strain BM110, qRT-PCR analyses comparing mRNA levels of PI-2b first gene (orf) and AgB last gene (san1522) were carried out in BM110 WT, BM110ΔcovR and BM110CovRD53A. As shown in Fig 5C, a 2- to 3- fold increase of PI-2borf was seen in the ΔcovR, but not in CovRD53A mutant, as compared to the WT BM110. In contrast, the levels of san1522, encoding the last gene of antigen B, appeared unchanged in BM110 WT and covR mutants.

Since CovR was shown to directly repress PI-1 pilus transcription in GBS strain 2603V/R [21], we examined PI-1 mRNA levels in BM110 WT, BM110ΔcovR and BM110 D53A by qRT-PCR. In agreement with Jiang results, we showed that transcription of the gene san0698 encoding the PI-1 major pilin in BM110 increased 2.5-fold in the ΔcovR strain and 5-fold in the CovRD53A mutant (S3 Fig). Collectively, these results demonstrate a direct role of CovR in PI-1 regulation in GBS strain BM110 and suggest an indirect role of CovR in PI-2b expression.

The bacterial strains and plasmids used in this study are listed in S1 Table. Primers are listed in S2 Table. GBS ST17 clinical strains isolated from invasive infections (i.e. from blood culture, cerebro-spinal fluid or other sterile sites) were obtained from the French National Reference Center for Streptococci in Paris (http://www.cnr-strep.fr) [5]. GBS NEM316, capsular serotype III ST-23, A909, capsular serotype Ia ST-9 and GBS BM110 capsular serotype III ST17 are well-characterized isolates from human with invasive infections. Seven non-ST17 GBS isolates (HN016, WC135, YM001, GX064, LMG14608, LMG15083, BSU178) whose genome are available were previously described [10, 26]. Escherichia coli DH5α (Gibco-BRL) was used for cloning experiments.

S. agalactiae strains were cultured in Todd Hewitt (TH) broth or agar (Difco Laboratories, Detroit, MI) at 37°C in standing filled flasks and E. coli in Luria-Bertani (LB) medium. Lactococcus lactis NZ9000 was grown in M17 medium supplemented with 1% glucose at 30°C or 37°C. Antibiotics were used at the following concentrations: for E. coli, erythromycin, 150 μg ml⁻¹, kanamycin, 50 μg ml⁻¹; for S. agalactiae, erythromycin, 10 μg; for L. lactis, erythromycin, 5 μg ml⁻¹.

Standard recombinant techniques were used for nucleic acid cloning and restriction analysis. Plasmid DNA from E. coli was prepared by rapid alkaline lysis using the Qiaprep Spin Miniprep kit (Qiagen). Genomic DNA from S. agalactiae was prepared using the DNeasy Blood and Tissue kit (Qiagen). PCR was carried out with Phusion Taq polymerase as described by the manufacturer (Finnzymes). Amplification products were purified with QIAquick PCR purification kit (Qiagen) and verified by sequencing.

Bacterial strains pre-cultured overnight in TH broth at 37°C in standing filled cultures were diluted to OD₆₀₀ = 0.05 in 25 ml TH and grown at 37°C until early exponential growth phase (OD₆₀₀ = 0.3). After centrifugation of 20 mL (4°C, 5,200 g), total RNA extraction was performed using the FastRNA PRO BLUE Kit (MP Biomedicals). RNA quality was assessed by agarose gel electrophoresis, residual DNA from the RNA samples was removed with the TURBO DNase kit (2U/μl, Ambion / Thermo Fischer Scientific) and extracts were stored at -80°C. RNA concentrations were measured using Nanodrop 2000c (Thermo Fischer Price). Reverse transcription (RT) was performed with the iScript cDNA synthesis kit (Bio-Rad). Specific primer pairs were designed to obtain a predicted amplicon size of 170 to 220 bp (see S2 Table) and quantitative PCR (qPCR) was carried out with EvaGreen Universal qPCR Supermix (Bio-Rad) in a CFX96 Touch Real-Time PCR Detection System (Bio-Rad). Relative gene expression levels were calculated with the ΔΔCq method [27] where expression values were normalized with the expression of the housekeeping gene gyrA and to a control sample by the CFX Manager Software v3.0 (Bio-Rad). Each assay was performed in triplicate with three independent cultures.

Total RNA was used as template for primer extension reaction using a radiolabeled specific primer complementary to a sequence located downstream from the putative PI-2b promoter 2bEA3 (S2 Table). For detail, synthetic oligodeoxynucleotides were 5'-end-labelled with [γ-³²P]-ATP (110 TBq mmol^-1) using T4 polynucleotidekinase. 15 μg of RNA and 2 pmol of labelled oligo-nucleotide were annealed in a total volume of 18 μl of reverse transcriptase buffer (50 mM Tris-HCl; 8 mM MgCl₂; 30 mM KCl; 1mM DTT; pH 8.5). The mixture was incubated for 3 min at 65°C and then 1 μl (25 U) of avian myelo-blastosis virus (AMV, Boehringer) reverse transcriptase and 1μl of all dNTPs mix (20 mM each) were added. After 30 min at 42°C, reactions were stopped by the addition of 5 μl of a solution containing 97.5% deionized formamide, 10 mM EDTA, 0.3% xylene cyanol, and 3.3% bromo-phenol blue.

The corresponding Sanger DNA sequencing reactions were carried out by using the same primer and a PCR-amplified fragment containing the PI-2b upstream region (primer pair 2bEA1 and 2bEA2, S2 Table) with the Sequenase PCR product sequencing kit (USB).

Genome-wide mapping of transcription start sites in GBS strains A909 and BM110 was performed by using a differential RNA-sequencing protocol based on selective Tobacco Acid Pyrophosphatase (TAP) treatment and 5' adapter ligation as previously described [17, 28]. Briefly, for each strain, mixes of RNA prepared at mid-exponential and stationary growth phases in a rich culture medium (TH) and at the beginning of the stationary phase in a poor culture medium (RPMI supplemented with glucose 1% and pH-buffered with 50 mM Hepes) were enriched in mRNA with the MICROBExpress Kit (Ambion). dRNA-seq libraries were prepared after or without prior TAP treatment by ligation with a 5' adapter (Illumina TruSeq Small RNA kit) and by reverse transcription using a random primer. After sequencing the reads were aligned to the reference genomes (A909: NC_007432.1; BM110: unpublished complete genome sequence obtained from PacBio sequencing) and analyzed for statistical assignment of TSSs based on the differences between the number of reads originating at each sequence position under TAP+ and TAP- conditions. In addition, strand-specific RNA-seq libraries were prepared from the two strains grown at mid-exponential growth phase by using the Illumina primer ligation method. dRNA-seq and RNA-seq libraries were sequenced on the Illumina GAIIX or HiSeq 2000. Alignments of the reads to the reference genomes were visualized with the IGV genome browser [29].

In frame deletion mutants of spb1 in BM110 and A909 were constructed as previously described [13]. The deletion was confirmed by PCR and sequencing on the genomic DNA of the mutants.

DNA fragment intragenic to spb1 (from amino acids 26–460) was amplified by PCR using genomic DNA of GBS BM110 as a template and primers listed in S2 Table. The amplification product was cloned into the pGEM-T easy vector (TOPO Zero Blunt for Ap1). After verification by sequencing, the fragment was digested with the appropriate enzymes and cloned into pET28a. The resulting plasmid was introduced into E. coli BL21λDE3/pDIA17 for protein expression. Recombinant proteins were purified on Ni-NTA columns. Protein purity was checked on sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE), and accurate protein concentration was determined using the Bradford protein assay. Rabbits polyclonal antibodies were produced and purchased from Covalab.

GBS strains A909 and BM110 were grown overnight in Todd-Hewitt broth. Bacteria were washed twice in phosphate buffered saline (PBS), and incubated for 1h at 4°C with rabbit primary antibody anti-Spb1 diluted in PBS-BSA 1.5% (1/1000). After two washes with PBS, samples were incubated for 30 min with secondary DyLight₄₈₈-conjugated goat anti-rabbit immunoglobulin diluted in PBS-BSA 1.5% (1/1000 dilution; Thermo Scientific Pierce) and Hoechst 33342 (1/1000). Microscopic observations were done with a Nikon Eclipse Ni-U and images acquired with a Nikon Digital Camera DS-U3. Images taken at different color channels were treated and merged in Image J.

To analyze PI-2b expression, exponentially growing bacteria (OD₆₀₀ = 0.3–0.4), were collected and washed twice in PBS. Bacterial pellet was incubated with rabbit anti-Spb1 primary antibody (1/1000) diluted in PBS-BSA 1% for 1 h at 4°C. After two washes with PBS, samples were incubated with the secondary antibody DyLight₄₈₈-conjugated goat anti-rabbit immunoglobulin (Thermo Scientific Pierce) diluted 1/1000 in PBS-BSA 1% for 30 min at 4°C. S. agalactiae and L. lactis strains carrying GFP reporter constructs were grown overnight at 30°C in 2 ml TH broth or M17-glucose, respectively, diluted 20x the following day, and grown at 37°C to exponential phase. One ml samples were harvested, and Hoechst 33342 was directly added in the medium (1/1000), the tubes were left open and incubated 10 minutes under gentle agitation. After washing and resuspension in PBS, samples were acquired on a MACSQuant YGV Analyzer apparatus (Miltenyi Biotec) and data were analyzed using FlowJo X software.

Bacteria were grown in TH medium at 37°C and harvested for protein analysis during late exponential phase of culture. Bacteria were washed in PBS and then resuspended in mutanolysin digestion mix (100Uml^-1 Mutanolysin (Sigma) resuspended in 50 mM Tris-HCl, pH 7.3, 20% sucrose, supplemented with 1x Proteinase Inhibitor Complex (Roche). The digestion was performed for 2 h at 37°C under gentle agitation. After centrifuging at 13,000 g for 10 min at 4°C, supernatants corresponding to the cell wall fractions were analyzed on SDS-PAGE or kept frozen at −20°C.

For analysis of major- and ancillary- pilin expression, cell wall proteins were boiled in Laemmli sample buffer, separated by SDS-PAGE on 4–12% Tris-Acetate Midi Criterion XT Precast gels (Bio-Rad), and transferred to PVDF membrane using the Trans-Blot Turbo transfer pack (Bio-Rad). Immuno-detection was performed as follow: membrane was blocked in TBS–skimmed milk 5% and incubated for 1 h with rabbit primary Spb1 and Ap1 antibodies and then with the secondary Dylight₈₀₀-coupled goat anti-rabbit antibody (Thermo Scientific Pierce). Between the two antibodies and before detection, membranes were extensively washed with TBS + 0.1% Tween 20. Bound pilins were detected using the LI-COR Odyssey Infrared Imaging System (LI-COR Biosciences).

Data were analyzed using GraphPad Prism 6.0 (GraphPad Software, San Diego, California). The significance of differences between the values was determined by Student’s t-test. Significance levels were set at *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.005; ****P ≤ 0.001.