Analysis of two-component systems in group B Streptococcus shows that RgfAC and the novel FspSR modulate virulence and bacterial fitness.

UNLABELLED: Group B Streptococcus (GBS), in the transition from commensal organisms to pathogens, will encounter diverse host environments and, thus, require coordinated control of the transcriptional responses to these changes. This work was aimed at better understanding the role of two-component signal transduction systems (TCS) in GBS pathophysiology through a systematic screening procedure. We first performed a complete inventory and sensory mechanism classification of all putative GBS TCS by genomic analysis. Five TCS were further investigated by the generation of knockout strains, and in vitro transcriptome analysis identified genes regulated by these systems, ranging from 0.1% to 3% of the genome. Interestingly, two sugar phosphotransferase systems appeared to be differentially regulated in the TCS-16 knockout strain (TCS loci were numbered in order of their appearance on the chromosome), suggesting an involvement in monitoring carbon source availability. High-throughput analysis of bacterial growth on different carbon sources showed that TCS-16 was necessary for the growth of GBS on fructose-6-phosphate. Additional transcriptional analysis provided further evidence for a stimulus-response circuit where extracellular fructose-6-phosphate leads to autoinduction of TCS-16, with concomitant dramatic upregulation of the adjacent operon, which encodes a phosphotransferase system. The TCS-16-deficient strain exhibited decreased persistence in a model of vaginal colonization. All mutant strains were also characterized in a murine model of systemic infection, and inactivation of TCS-17 (also known as RgfAC) resulted in hypervirulence. Our data suggest a role for the previously unknown TCS-16, here named FspSR, in bacterial fitness and carbon metabolism during host colonization, and the data also provide experimental evidence for TCS-17/RgfAC involvement in virulence. IMPORTANCE: Two-component systems have been evolved by bacteria to detect environmental changes, and they play key roles in pathogenicity. A comprehensive analysis of TCS in GBS has not been performed previously. In this work, we classify 21 TCS and present evidence for the involvement of two specific TCS in GBS virulence and colonization in vivo. Although pinpointing specific TCS stimuli is notoriously difficult, we used a combination of techniques to identify two systems with different effects on GBS pathogenesis. For one of the systems, we propose that fructose-6-phosphate, a metabolite in glycolysis, is sufficient to induce a regulatory response involving a sugar transport system. Our catalogue and classification of TCS may guide further studies into the role of TCS in GBS pathogenicity and biology.

INTRODUCTION

Streptococcus agalactiae, or group B Streptococcus (GBS), is a human pathogen that causes septicemia and meningitis in neonates (1, 2), and it is also a known cause of bovine mastitis (3). In this decade, GBS remains the dominant cause of infant morbidity and mortality in the United States (4), and it has also been recognized as an important cause of infections in immunocompromised patients and the elderly (5). GBS is a commensal in the rectovaginal tract of 20% to 30% of healthy women (6). However, by vertical transmission intrapartum, it may transition to an invasive pathogen, resulting in pneumonia, sepsis, and meningitis (7). The physiopathology of GBS infections implies that this bacterium encounters several very different microenvironments during colonization and the infectious process. The transition of the organism from a commensal niche (e.g., vaginal tract) to invasive niches (e.g., blood, lung, brain, and other organs) is likely to require adaptive changes, and a well-known way for bacteria to monitor and respond to their environment is by the use of two-component signal transduction systems (TCS).

TCS are typically organized in operons that encode a sensing histidine kinase (HK) and a response regulator (RR). The HK harbors an N-terminal input domain that recognizes a specific stimulus. The information is then transduced through intramolecular conformational changes, resulting in the phosphorylation and activation of the C-terminal transmitter domain. This domain, in turn, activates its cognate receiver, encoded by the N-terminal domain of the RR. Once activated, the RR gives rise to an intracellular response through the C-terminal effector (or output) domain. This response typically results in differential gene expression. Sequence homology is generally a poor predictor of sensing mechanisms or specific stimuli, while domain architecture may provide clues that assist in assigning HKs to one of three groups: extracellular sensing, membrane sensing, and cytoplasmic sensing (8).

Overall, the role of TCS in GBS pathogenesis is not well understood. The most studied system is CovRS, an important regulator in Streptococcus spp. In GBS, the CovRS regulon extends to 7% to 27% of the genome (depending on growth conditions) and includes different functional categories, such as cell envelope, cellular processes, metabolism, and virulence factors (9–12). The RRs CiaR (13), Sak189 (14), and LiaR (15) are reported to directly influence in vivo virulence in GBS. Moreover, the TCS DltRS (16) and RgfAC (17, 18) have been shown to affect gene expression in GBS.

In the present study, we aimed to better understand the role of TCS in GBS pathophysiology by adopting a stepwise screening strategy. First, inventory, comparative genomics analysis, and sensing mechanism classification were performed through a bioinformatics approach. Second, by transcriptional analysis and identification of output domains related to virulence, we selected five systems for further study. We generated TCS knockout mutants and analyzed their transcriptome in vitro, as well as their contribution to virulence and colonization in vivo. One TCS was further characterized, and we propose evidence for a specific stimulus-response circuit.

RESULTS

Identification and comparative genomics of TCS.

For the purposes of our study, we focused our main attention on serotype V GBS isolates. Type V is the most common capsular serotype associated with invasive infection in nonpregnant adults and has increased among neonatal invasive disease strains in recent years, accounting for approximately one-third of clinical isolates in the U.S. population (19). To identify possible TCS, genes predicted to encode HKs or RRs in the GBS type V strain 2603 V/R genome were collected from the P2CS database (http://www.p2cs.org) (20) and the loci were compared with those in a second type V genome (strain CJB111). A total of 38 genes (21 systems) were identified (Table 1). Seventeen of these loci contained a cognate pair of HK/RR-encoding genes, and an additional 4 TCS genes were orphans in strain 2603 V/R, while the corresponding numbers of cognate pairs and orphans in strain CJB111 were 19 and 2, respectively. Each TCS locus, whether paired or orphaned, was assigned a number in order of its appearance on the chromosome. The predicted TCS proteins were then used as queries to identify corresponding gene loci in 251 other GBS genomes (draft and complete), representing isolates from different host species (predominantly human and bovine). Overall, the 21 TCS were very well conserved (>98% identity) across the genomes. Our analysis was unidirectional, and we do not exclude the presence of additional TCS. We have likely underestimated the number of fully functional orthologous systems, due to a number of TCS sequences being located in the vicinity of contig breaks. With the very limited variability observed in a large set of genomes, the TCS identified here appear to be part of the GBS core genome. TCS-17 (also known as RgfAC) presented an interesting case of variability. While TCS-17/RgfAC is complete and conserved in 64 strains, 118 strains have an orphan TCS-17/RgfAC where only the RR/RgfA is present, and another 61 strains had polymorphisms that inactivated the RR/RgfA. Moreover, orphaned and complete loci may represent two distinct families. While the protein sequence conservation is typically high (99% identity) when comparing RRs within a family (e.g., orphaned versus orphaned), a similar comparison across families shows a decreased sequence conservation (79% identity). Genomes with an orphan TCS-17 may have undergone recombinatorial events, as they show larger deletions in the locus with short remnants of putative hk or rr genes.

TABLE 1 

Comparative genomics of TCS

TCS locus	Locus tag in 2603 V/R	Type	No. of strains with:			No. of proteins with:		Avg % identity (SD)	JCVI designation[f]	Gene name(s) (reference)[g]
			Ortholog:		Stop[c]	Ambiguity[d]	Variance[e]
			Present[a]	Absent[b]
1	SAG0123	RR	240	4	0	9	6	99.02 (0.46)	SAM0116
	SAG0124	HK	246	3	0	4	17	98.61 (0.58)	SAM0117
2	SAG0182	HK	240	3	6	4	21	99.53 (0.22)	SAM0183
	SAG0183	RR	248	3	1	1	6	99.06 (0.48)	SAM0184
3 (orphan)	SAG0310	HK	198	3	12	40	27	98.17 (3.32)	SAM0322
			246	3	2	2	30	97.87 (2.39)	SAM0323
4	SAG0321	HK	250	3	0	0	9	99.28 (0.29)	SAM0333	liaSR (15)
	SAG0322	RR	248	3	0	2	4	99.30 (0.25)	SAM0334
5	SAG0393	RR	242	10	0	1	4	99.13 (0.39)	SAM0401
	SAG0394	HK	239	3	1	10	11	99.36 (0.28)	SAM0402
6	SAG0616	RR	226	26	0	1	2	99.54 (0.00)	SAM0583	sak188/sak189 (14)
	SAG0617	HK	209	25	4	15	11	97.23 (4.53)	SAM0584
7 (orphan)	SAG0712	RR	240	2	1	10	7	99.11 (0.38)	SAM0733
8	SAG0719	RR	241	4	7	1	7	99.28 (0.19)	SAM0741
	SAG0720	HK	247	4	1	1	16	99.53 (0.13)	SAM0742
9	SAG0976	RR	237	4	9	3	12	98.86 (0.56)	SAM0983
	SAG0977	HK	247	4	0	2	12	99.04 (0.41)	SAM0984
10	SAG0984	HK	250	3	0	0	20	99.47 (0.18)	SAM0991	ciaRH (13)
	SAG0985	RR	250	3	0	0	4	99.34 (0.24)	SAM0992
11	SAG1016	RR	248	3	1	1	14	98.91 (0.45)	SAM1027
	SAG1017	HK	244	3	1	5	25	99.12 (0.39)	SAM1028
12	SAG1327	HK	249	3	0	1	15	99.27 (0.27)	SAM1289
	SAG1328	RR	248	3	1	1	10	98.80 (0.65)	SAM1290
13	SAG1624	HK	249	1	1	2	27	99.35 (0.38)	SAM1583	covRS (10)
	SAG1625	RR	252	1	0	0	6	99.19 (0.32)	SAM1584
14	SAG1791	HK	244	0	3	6	17	99.34 (0.23)	SAM1775	dltRS (16)
	SAG1792	RR	232	1	17	3	15	98.70 (0.54)	SAM1776
15 (orphan)	SAG1922	RR	239	5	6	3	8	99.54 (0.30)	SAM1860
16	SAG1946	RR	248	3	1	1	16	99.11 (0.28)	SAM1885
	SAG1947	HK	245	3	3	2	27	99.12 (0.39)	SAM1886
17 (orphan)			123	128	0	2	8	88.26 (10.33)	SAM1896	rgfAC (17)
	SAG1957	RR	182	3	61	7	12	98.71 (0.72)	SAM1897
18	SAG1960	HK	246	4	2	1	23	99.10 (0.46)	SAM1900
	SAG1961	RR	239	4	0	10	12	98.92 (0.42)	SAM1901
19	SAG2054	RR	249	2	1	1	15	98.69 (0.60)	SAM1962
	SAG2055	HK	237	2	7	7	21	99.14 (0.36)	SAM1963
20	SAG2122	RR	237	14	0	2	14	98.85 (0.42)	SAM2038
	SAG2123	HK	236	13	2	2	19	98.87 (0.61)	SAM2039
21	SAG2127	HK	228	21	0	4	15	99.26 (0.30)	SAM2043
	SAG2128	RR	223	12	6	12	3	99.52 (0.21)	SAM2044

Strains with putatively functional ortholog (i.e., >75% identity for at least 90% of the query sequence).

Strains where no ortholog was identified.

Strains where proteins were truncated and possibly inactivated.

Proteins with nucleotide ambiguities or proximity to contig break.

Distinct variant proteins.

J. Craig Venter Institute (formerly TIGR) gene names of hk and rr genes in the CJB111 strain used in the experimental studies.

Previously reported gene name and citation.

A minimum spanning tree (MST) using the 40 TCS genes and all their respective variants was created to identify potential TCS profile clusters (Fig. 1). Interestingly, bovine isolates appeared to contain TCS profiles that clustered together and were relatively more distant from human isolates. TCS-17 again constituted a particular example in which RR variants appeared to group according to the host species from which they were isolated (see Fig. S1 in the supplemental material). Variant 2 was found in 43% of bovine strains versus 6% of human strains. Conversely, variant 4 was found in 41% of human strains and only 13% of bovine strains.

FIG 1 

Minimum spanning tree (MST), obtained using the goeBurst algorithm and based on the TCS protein similarity in 253 GBS genomes. Isolates are from different hosts, as indicated, and each node represents a TCS profile. The sizes of the nodes are proportional to the numbers of isolates with each profile.

Domain architecture for TCS in the genome of CJB111.

For further experimental analysis, we shifted our focus to the serotype V strain CJB111, which is more virulent than 2603 V/R in mouse models and for which a custom microarray was available in-house. The CJB111 TCS gene identifiers are listed in Table 1. We analyzed the domain architecture of all the HK proteins in the genome of CJB111 and predicted the mechanism of stimulus perception according to criteria reported in the literature (8) (Fig. 2). Nine of the HKs contained an N-terminal domain with two transmembrane helices with a spacer of 50 to 300 amino acids, typical for extracellular (EX) sensing. Eight HKs contained 2 to 6 predicted transmembrane-spanning regions separated by short spacers and were predicted to be membrane sensing (TM). Only one HK was classified as cytoplasmic (C) sensing.

FIG 2 

Domain architecture of histidine kinases. Protein sequences were analyzed using SMART http://smart.embl-heidelberg.de, and PFAM domains were included. The scale bar shows the sequence length in amino acids. Blue vertical bars represent putative transmembrane helices, and pink squares represent low-complexity regions. Various functional domains are indicated by the remaining colored elements and using the SMART nomenclature (e.g., GAF, PAS, HAMP, PAC). The SMART HATPase_c domain in TCS-17/RgfAC is an outlier homolog and was found by using the schnipsel database. Sensing mechanisms were manually predicted as transmembrane (TM), extracellular (EX), and cytoplasmic (C).

We also observed that all the RR proteins had a DNA binding output domain, indicating a role as direct transcriptional regulators. Interestingly, the LytTR output domain was the second most common output domain in GBS (15%), while it accounts for a smaller percentage (4%) in prokaryotes in general. LytTR-type output domains have been noted for the control of virulence factors in several important bacterial pathogens (20).

Transcriptional analysis of TCS.

Transcriptional analysis was performed in early logarithmic (EL) and early stationary (ES) phase using a custom microarray chip designed on the CJB111 genome. We noted that several TCS were significantly upregulated in ES phase, and 3 of these (TCS-2, TCS-16, and TCS-21) were upregulated 4-fold or more. As many TCS respond to stress conditions like those encountered in stationary phase (lack of nutrients, low pH, and accumulation of toxic metabolites), we hypothesized that these systems could be involved in bacterial stress responses and selected them as targets for mutagenesis. To further understand the levels of expression of each selected TCS during bacterial growth in vitro, we measured the gene transcripts of the HKs at four time points (representing early, mid-, and late logarithmic and early stationary phases) using a real-time quantitative (qRT)-PCR analysis (see Fig. S2 in the supplemental material). These data confirmed that TCS-2, TCS-16, and TCS-21 are upregulated only in ES phase.

Transcriptional regulators with the LytTR-type output domains control the production of virulence factors in several important bacterial pathogens (20). We therefore selected as mutagenesis targets the three TCS containing an RR with a LytTR-type DNA binding domain, i.e., TCS-2 (upregulated at ES phase), TCS-11, and TCS-17/RgfAC. Thus, a total of five TCS exhibiting either the presence of a LytTR domain and/or upregulation during ES were selected for further experimentation (Table 2).

TABLE 2 

CJB111 expression of selected TCS

TCS locus	Locus tag	Type	Fold change[a]	P value[b]
2	SAM0183	HK	6.1	7.4 × 10−4
	SAM0184	RR[c]	5.6	6.0 × 10−3
11	SAM1027	RR[c]	0.9	NS
	SAM1028	HK	1.5	NS
16	SAM1885	RR	4.5	3.8 × 10−3
	SAM1886	HK	5.5	1.1 × 10−3
17	SAM1896	HK	1.1	NS
	SAM1897	RR[c]	1.6	NS
21	SAM2043	HK	37.0	6.0 × 10−8
	SAM2044	RR	19.5	6.3 × 10−5

Relative expression of the TCS in ES compared to EL growth phase.

NS, the fold change was considered not significant if the P value was >0.05.

Contains a LytTR output domain.

Expression microarray analysis of TCS mutants.

The 5 selected TCS loci were modified genetically by in-frame deletion of the genes encoding RRs. There were no apparent differences in colony size, hemolysis, or other macroscopic features between the wild-type (WT) strain and the five isogenic mutant strains on blood agar or tryptic soy agar plates (data not shown). Mutant strains grown in complex medium exhibited growth curves identical to that of the wild-type parental strain (see Fig. S3 in the supplemental material).

Global transcriptional analysis was performed by microarray technology. Gene expression of all of the predicted genes (n = 2,232) in the CJB111 genome was performed by comparing the WT strain with each of the five mutant strains in ES phase (Fig. 3A). Experimental design, chip validation, quality control, and data analysis were performed as described in Materials and Methods. Further validation of the microarray data was performed by qRT-PCR, using probes for 9 independent transcriptional units (see Fig. S4 in the supplemental material). Overall, all five mutants exhibited differential transcription of genes in comparison to the gene transcription in the wild type when using a permissive threshold (±2-fold change and a P value of <0.05). The number of regulated genes ranged from 2 (Δrr17 strain; see Materials and Methods) to 66 (Δrr11 strain). Two mutants (the Δrr11 and Δrr21 strains) showed predominantly downregulated genes, suggesting that these RRs act as activators. We observed that some genes appeared to be regulated in two or more independent mutant strains (Fig. 3B). In particular, 20 genes were regulated in both the Δrr21 and the Δrr11 strain (19 genes downregulated and 1 gene upregulated in both strains).

FIG 3 

Microarray analysis of TCS mutant strains. (A) Bars represent the numbers of genes up- or downregulated in early stationary (ES) phase compared to their transcription in the WT strain. The threshold used is ±2-fold change and a P value of ≤0.05. (B) Venn diagram showing overlaps of gene regulation between the mutants. The fields show the number of individual genes significantly regulated in single or multiple mutant strains.

A more stringent threshold was applied to identify the most highly regulated genes (±4-fold change and a P value of <0.05). Three mutants (Δrr2, Δrr11, and Δrr16 strains) showed a total of 18 highly regulated genes, and in two of them, such genes were identified adjacent to the TCS locus (Table 3). In particular, the Δrr2 mutant exhibited downregulation of SAM0185 and SAM0186, encoding two proteins similar to LrgAB in Streptococcus mutans. In this species, LrgAB is reported to be under the control of an adjacent TCS (homologous to TCS-2) and was suggested to have a role in the control of virulence and biofilm formation (21, 22). The Δrr16 mutant showed strong downregulation of an adjacent phosphotransferase system (PTS) and concomitant upregulation of another PTS system in a different locus. All of the highly regulated genes were subjected to confirmation by qRT-PCR, using one probe set per transcriptional unit (Table 3). Moreover, qRT-PCR experiments were repeated on chromosomally complemented strains where the deleted rr gene was replaced with the WT form, which confirmed that the WT phenotype was restored (Table 3).

TABLE 3 

Highly regulated genes in early stationary phase

Mutation	Locus tag	Annotation	Gene	Fold change in microarray	P value[a]	KO strain/KI strain fold changes in qRT-PCR[b]
Δrr2	SAM0185	LrgA family subfamily, putative	lrgA	−11.5	4.0 × 10−5	−6.14/−1.18
	SAM0186	LrgB family protein	lrgB	−8.8	1.2 × 10−4
Δrr11	SAM0011	Phosphoribosylformyl-glycinamidine synthase		−4.0	2.2 × 10−4
	SAM0012	Amidophosphoribosyl-transferase	purF	−4.0	3.0 × 10−3	−2.23/+2.98
	SAM0013	Phosphoribosylformyl-glycinamidine cycloligase	purM	−4.3	2.4 × 10−4
	SAM0014	Phosphoribosylglycinamide formyltransferase	purN	−4.3	1.8 × 10−3
	SAM0064	Ribosomal protein S5	rpsE	−4.4	1.8 × 10−5	−2.01/+1.56
	SAM1026	Carbon starvation protein, putative	cstA	−5.7	1.6 × 10−5	−10.95/+2.11
	SAM1057	Conserved hypothetical protein		−5.1	1.8 × 10−5
	SAM1058	Conserved hypothetical protein		−5.2	5.8 × 10−7
	SAM1059	Carbamoyl-phosphate synthase, large subunit	carB	−7.4	3.0 × 10−7	−4.72/−1.69
	SAM1060	Carbamoyl-phosphate synthase, small subunit	carA	−4.2	4.2 × 10−2
Δrr16	SAM1715	PTS system, IIC component	fruA-2	+4.0	1.4 × 10−4	+3.25/−1.06
	SAM1716	PTS system, IIA component	fruA-2	+4.1	2.6 × 10−4
	SAM1887	PTS system, IID component		−31.6	1.6 × 10−7	−324.0/+1.72
	SAM1888	PTS system, IIC component		−72.0	1.0 × 10−7
	SAM1889	PTS system, IIB component		−55.7	1.0 × 10−6	−6.14/−1.18
	SAM1890	PTS system, IIA component, putative		−52.0	1.8 × 10−7

Inclusion threshold is a P value of ≤0.05 and fold change of ±4 compared to the WT strain).

Fold changes in knockout/complemented strains compared to the WT strain.

Phenotype microarray screening shows that TCS-16 influences carbon source utilization.

Our microarray analysis showed differential regulation of two different PTS (SAM1715 to SAM1716 and SAM1887 to SAM1890) in the Δrr16 mutant strain, suggesting a possible involvement in the monitoring of carbon source availability. The WT and Δrr16 strains were grown on 192 different single carbon sources using the phenotype microarray (PM) technology (see Fig. S5 in the supplemental material). One compound showed a clear difference between the WT and mutant strain. The WT strain exhibited growth comparable to that of the positive control when supplied with fructose-6-phosphate (Fru-6-P), while the Δrr16 strain showed no growth. The experiment was repeated using the complemented (knock-in [KI]) KIrr16 strain, which showed a growth profile indistinguishable from that of the WT strain (Fig. S5).

To confirm this phenotype in an independent system and to understand if it could be extended to other hexose-6-phosphate sugars, we used a chemically defined medium (CDM) (23) where glucose, fructose, mannose, or the corresponding hexose-6-phosphate counterpart of each was used as the primary carbon source (Fig. 4). The WT, Δrr16, and KIrr16 strains grew well on glucose, fructose, and mannose, while only the WT and KIrr16 strains grew on Fru-6-P. The other phospho-sugars did not allow any bacterial growth. We concluded that TCS-16 is necessary for the growth of strain CJB111 on Fru-6-P as the primary carbon source.

FIG 4 

Growth curves of WT (blue), Δrr16 (red), and KIrr16 (green) strains in CDM supplemented with 10 mg/ml of glucose or glucose-6-phosphate, fructose or fructose-6-phosphate, or mannose or mannose-6-phosphate. Growth curves are from triplicate samples, and the background (negative control, no sugar added) was subtracted.

TCS-16 gene regulation in response to extracellular Fru-6-P.

We subsequently investigated the impact of Fru-6-P on the transcription of TCS-16 and its adjacent PTS system by using qRT-PCR. Bacteria grown in CDM were subjected to a substitution of glucose for Fru-6-P upon entry into mid-logarithmic phase (see Materials and Methods). Compared to its transcription in bacteria exposed to glucose, the transcription of the hk16 gene increased 22-fold when bacteria were exposed to Fru-6-P (Fig. 5A). This response was observed in the WT, absent in the Δrr16 strain, and partially restored in the complemented KIrr16 strain. Concomitantly (Fig. 5B and C), a similar pattern was observed for the adjacent PTS system (3,000-fold upregulation) and for the SAM1402 gene (2.5-fold upregulation), a putative hexose-6-P transporter gene that we identified in the genome of CJB111. A down-regulation of the SAM1402 gene (1.5-fold) in the Δrr16 strain was also observed in the microarray.

FIG 5 

Results of qRT-PCR with probes for SAM1885 (HK) (A, D), SAM1888 (PTS, IIC component) (B, E), and SAM1402 (putative hexose-phosphate transporter) (C, F) using RNA extracted from WT, Δrr16, and KIrr16 strains. (A to C) Bacteria grown in CDM were subjected to a pulse of either glucose (Glc) or fructose-6-phosphate (Fru-6-P) as the main carbon source for 30 min prior to RNA extraction. Bars represent gene expression when exposed to Fru-6-P relative to gene expression during growth in glucose. (D to F) Bacteria grown in CDM were subjected to a pulse of PBS or bacterial lysate for 30 min prior to RNA extraction. Bars represent gene expression when cells were exposed to bacterial lysate relative to gene expression in the PBS control. The experiment was performed twice. Bars represent technical triplicates, and error bars show the standard errors of the means.

Fru-6-P is an intracellular metabolic intermediary of the glycolytic pathway. Bacterial lysis constitutes a possible scenario where Fru-6-P would be found in the extracellular space. We wanted to investigate whether TCS16 would respond under such conditions. Bacteria were grown in CDM until mid-logarithmic phase, and the medium was changed to CDM without glucose in the presence or absence of bacterial lysate (see Materials and Methods). The transcriptional levels of TCS-16, its adjacent PTS system, and the putative hexose-6-P transporter were then analyzed by qRT-PCR. The hk16 gene transcription increased 10-fold when bacteria were exposed to bacterial lysate compared to its transcription in the control (Fig. 5D). Concomitantly, an upregulation of the PTS system (200-fold) and the SAM1402 gene (5-fold) was observed (Fig. 5E and F). These responses were absent in the Δrr16 strain and totally restored in the complemented KIrr16 strain.

TCS-16 influences vaginal persistence in mice.

In a recent work, carbon catabolite repression in Streptococcus pyogenes was shown to influence asymptomatic colonization of the murine vaginal mucosa, suggesting that the availability of carbon sources may be subject to monitoring by the bacteria (24). We subsequently investigated whether TCS-16 could play a role in bacterial survival during colonization, utilizing the Δrr16 mutant strain in a mouse model of GBS vaginal colonization (25, 26). CD-1 mice in estrus were inoculated with ~1 × 107 CFU in the vaginal lumen and, on successive days, bacteria were recovered by swabbing and quantified by serial dilution and plating on selective medium. Interestingly, the Δrr16 mutant exhibited decreased persistence in the vaginal tract (Fig. 6). The statistically significant differences were observed at later time points during the experiment (P = 0.004, day 5; P = 0.02, day 7) and suggested a gradual decline in colonization with the mutant strain, while the WT remained relatively stable throughout the experiment. Identical experiments were performed with the remaining mutant strains, and the results were similar to those with the WT strain, underlining that the phenotype seen in the Δrr16 mutant was unique among the strains tested.

FIG 6 

Murine vaginal colonization model. The vaginal lumen of CD-1 mice was inoculated with 107 CFU of WT or Δrr16 bacteria, and bacterial persistence was determined by counting viable cells. Results shown are from two independent experiments. The detection limit is represented by the dashed line. Horizontal bars represent medians, and the Mann-Whitney U test was used for statistical analysis. *, P < 0.05; **, P < 0.005.

TCS-17/RgfAC influences virulence in a murine model of systemic infection.

To further investigate the role of the selected TCS in GBS pathogenesis, we compared the relative levels of virulence of CJB111 and the Δrr mutants using an in vivo mouse model of infection (27, 28). CD-1 mice were infected intravenously with 1.5 × 107 CFU of the WT or mutant strains. Bacteremia was confirmed in all mice by counting viable cells in blood samples collected 24 h postinfection, and there was no significant difference in bacterial loads between the WT and mutant strains (data not shown). Over the course of infection, animals were sacrificed at individual endpoints (when moribund) or upon termination of the experiments, and blood, brain, and lung tissues were collected for bacterial counts. Overall, the only mutant strain with a virulence phenotype was the Δrr17 strain, infection with which exhibited a significantly higher mortality than was observed during infection with the WT strain (P = 0.003, log-rank test) (Fig. 7A). Interestingly, no significant differences were observed in bacterial loads in brain, blood, and lung tissues when comparing the WT and Δrr17 strains (Fig. 7B to D). Blood and brain tissues from mice infected with the Δrr17 strain contained more individual samples with very high bacterial counts, but such samples were exclusively from moribund mice (Fig. 7B and C). None of the other mutant strains showed an appreciable difference from the WT strain (data not shown).

FIG 7 

Murine intravenous challenge model. (A) Kaplan-Meyer survival plot of mice infected with 1.5 × 107 CFU of bacteria. The log-rank test was used for statistical analysis. (B to D) Bacterial counts (CFU) in blood, brain, and lung tissues from individual mice. Horizontal bars represent medians. Mice were euthanized when moribund (red) or at the end of the experiment (black).

DISCUSSION

In the present study, we conducted genome-wide inventory, classification, and comparative genomics of TCS in GBS and concluded that they are highly conserved and part of the core genome. Nonetheless, the profile of TCS allelic variants differed somewhat when comparing human GBS isolates with those from bovine and other hosts. In our classification of sensing mechanisms, only one HK was classified as cytoplasmic sensing, while this is typically a more frequent category (8). A comparison of the number of TCS (corrected for genome size) in various pathogenic and nonpathogenic Lactobacillales spp. was also performed. In the nine species examined, GBS and S. pyogenes showed the highest frequencies of TCS, with medians of 38 and 36 TCS components, respectively (data not shown), compared to a typical range of 15 to 32 for the other Lactobacillales. This suggests that the two pathogens may require particular fine tuning of transcription in response to changing environments.

Transcriptome analysis was performed on five selected systems, and we observed a limited number of genes that were highly regulated for three TCS. TCS-2 regulates the adjacent lrgAB operon, similar to what has been described for S. mutans (22). TCS-11 regulates genes involved in purine metabolism and carbon starvation. TCS-16 highly regulates two PTS operons and is discussed in more detail below.

TCS-17/RgfAC has already been described in GBS strain O90-R, where it negatively regulates the transcription of C5a peptidase (scpB), a known virulence factor (17). Moreover, an independent group demonstrated that in clonal complex 17 (CC17) strains, RgfAC negatively regulates the fbsA gene, encoding fibrinogen binding proteins (18). FbsA may have a role in protecting the bacteria against opsonophagocytosis, promoting adhesion to lung epithelial cells, and increasing survival in human blood (29, 30). We did not observe any upregulation of the above-described genes in the Δrr17 strain, which may be due to differences in the experimental protocol (time points) or in the genetic background (CJB111 belongs to CC1 [31]). Nonetheless, the Δrr17 strain was hypervirulent in our murine model of systemic infection, consistent with a potential upregulation of virulence factors in the absence of this transcriptional regulator. TCS-17 was also the only system where specific allelic variants were distributed differently between human and bovine isolates, suggesting a potential host-specific role during infection.

The second system of particular interest was TCS-16. Transcriptome analysis showed prominent downregulation of an adjacent operon that encodes a putative PTS (Man/Fru/Sor family) and concomitant upregulation of another PTS (Fru family). TCS-16 was classified as an extracellular sensing system, and we hypothesized that it may be involved in monitoring and responding to the availability of nutrients (8). When bacteria were subjected to a large variety of different carbon sources in chemically defined medium, one compound, Fru-6-P, resulted in a complete growth defect, which was fully restored when the Δrr16 mutation was complemented. Further experiments in vitro, using phosphorylated and nonphosphorylated hexose sugars, confirmed these results. We thus conclude that a functional TCS-16 is necessary for growth on Fru-6-P. Investigation of the transcriptional events in the locus showed that, upon exposure to extracellular Fru-6-P, there is induction of TCS-16 and a concomitant drastic upregulation of the adjacent PTS operon. As the Δrr16 strain showed no such response, while the complemented strain exhibited a partial restoration, we conclude that a functional TCS-16 is necessary for autoinduction and regulation of the adjacent PTS operon. We propose that extracellular Fru-6-P is a signal for TCS-16 and that the gene locus be named fspSR, for fructose-six-phosphate sensor histidine kinase and response regulator. Despite the magnitude of the influence on PTS transcription, the growth defect observed in the Δrr16 (ΔfspR) strain is difficult to explain in terms of a direct link between the upregulated PTS and the utilization of Fru-6-P as an energy source. To our knowledge, PTS-dependent import of phosphorylated sugars has not been described. However, other such uptake mechanisms are known, and UhpT (major facilitator superfamily) in Escherichia coli represents an example where the controlling TCS (UhpAB) is necessary for the growth of E. coli upon Glc-6-P and Fru-6-P (32, 33). We identified a homologue of UhpT in the genome of CJB111 (SAM1402), and analysis of the protein suggests it could function as a hexose-6-P transporter (http://www.tcdb.org/). SAM1402 was among the genes significantly downregulated in the Δrr16 (ΔfspR) mutant transcriptome, and subsequent experiments confirmed that TCS-16/FspSR upregulates the SAM1402 gene in response to extracellular Fru-6-P. The biological relevance of our link between Fru-6-P and TCS-16/FspSR is difficult to ascertain, as the availability of Fru-6-P in the extracellular milieu is presumably limited or unknown. Nevertheless, our in vivo screening showed reduced vaginal persistence of the Δrr16 (ΔfspR) strain in mice. The PTS and associated carbon metabolism pathways may have an impact on in vivo fitness and virulence, as previously demonstrated for several different pathogens (34–38). One possibility is that Fru-6-P may be released from dying microorganisms in the complex microbiota of the vagina or in stationary-phase in vitro cultures, and GBS could consequently initiate a scavenging response involving upregulation of sugar transporters. This hypothesis was supported by our data showing that there is an induction of the TCS-16 response upon exposure to bacterial lysate. This response was observed in the WT and complemented strains but not in the Δrr16 (ΔfspR) strain, confirming that a functional TCS-16 is necessary to upregulate sugar transporters in the presence of lysed GBS components. Moreover, in a previous work using a different strain, the PTS operon described above, together with several other PTS, were highly downregulated under high-glucose conditions, supporting a role for this and similar systems under conditions where nutrients are relatively scarce (12). Another speculative possibility is that FspSR may be activated upon entry/invasion of eukaryotic cells, through the presence of Fru-6-P as a central metabolite in glycolysis.

We have listed and classified TCS in GBS for further study and conclude that these TCS are very well-conserved intraspecies but with allelic profiles that vary somewhat according to the host. Our results provide new insights into four previously unknown TCS but also provide the first in vivo data supporting a role for RgfAC in virulence. Finally we identified FspRS, a new TCS that is involved in vaginal persistence and responds to fructose-6-phosphate by the upregulation of genes involved in sugar transport.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

GBS strain CJB111 (Carol Baker Collection, Division of Infectious Diseases, Baylor College of Medicine, Houston, TX) and its isogenic derivatives were grown in Todd-Hewitt broth (THB medium; Difco Laboratories) at 37°C, 5% CO2. Tryptic soy broth (Difco Laboratories) with 15 g/liter agar (TSA) was used as the solid medium. Max Efficiency DH5α competent E. coli cells (Invitrogen) were used for transformation, propagation, and preparation of plasmids. E. coli was grown at 37°C with agitation (180 rpm) in Luria-Bertani (LB; Difco Laboratories) broth or on 15 g/liter agar plates (LBA). Erythromycin (Erm) was used for selection of GBS (1 µg/ml) or E. coli (100 µg/ml) cells containing the pJRS233-derived plasmids used for mutagenesis (see below).

Strains CJB111, CJB111Δrr16, and CJB111KIrr16 were also grown in chemically defined medium (CDM) (23) or in CDM where glucose was replaced with 10 g/liter glucose-6-phosphate, fructose, fructose-6-phosphate, mannose, or mannose-6-phosphate (Sigma). Briefly, bacteria grown on THB plates were suspended in 5 ml of phosphate-buffered saline (PBS) until the optical density at 600 nm (OD600) reached 0.3. Samples were then diluted 1:10 in PBS, and 88 µl of bacteria was added to 1.2 ml of medium with or without different carbon sources. Bacteria were grown in 96-well plates with 200 µl per well for 48 h at 37°C, and the OD600 was monitored automatically every 20 min in an automated reader (Sunrise; Tecan).

Construction of TCS isogenic mutants.

To construct the knockout strains, a pJRS233 shuttle vector (39) containing each TCS locus with an in-frame deletion in the response regulator gene (rr) was constructed using a splicing by overlap extension PCR (SOEing-PCR) strategy (see Table S1 in the supplemental material) (40). Briefly, the up- and downstream regions of an rr gene were produced from CJB111 genomic DNA and then joined. The resulting fragment was cloned into pJRS233 using BamHI and XhoI restriction sites.

Five plasmids, pJRS233Δrr2, pJRS233Δrr11, pJRS233Δrr16, pJRS233Δrr17, and pJRS233Δrr21, were obtained, each containing an insert with 700 to 800 bp upstream and downstream from the in-frame-deleted rr gene. The CJB111 genes thus inactivated were SAM0184 (rr2), SAM1027 (rr11), SAM1885 (rr16), SAM1897 (rr17), and SAM2044 (rr21).

To construct the respective chromosomally complemented knock-in (KI) strains, each wild-type locus was amplified and cloned into pJRS233 using BamHI and XhoI restriction sites (see Table S1 in the supplemental material). These plasmids were designated pJRS233KIrr2, pJRS233KIrr11, pJRS233KIrr16, pJRS233KIrr17, and pJRS233KIrr21.

An insertion/duplication and excision mutagenesis strategy was used to obtain either the in-frame deletions in the response regulator genes or the chromosomal replacements in the knockout mutants. In brief, pJRS233Δrr2, -Δrr11, -Δrr16, -Δrr17, and -Δrr21 plasmids purified from E. coli were used to transform CJB111 by electroporation, as previously described (41), except that we used M9 medium without glycine and Casamino acids. Transformants were selected by growth on TSA containing Erm at 30°C for 48 h. Integration was performed by growth of transformants at 37°C (nonpermissive temperature for the suicide shuttle vector) with Erm selection. Excision of the integrated plasmid was performed by serial passages in THB at 30°C and parallel screening for Erm-sensitive colonies on plates. Mutants were verified by PCR sequencing of the TCS loci. We obtained thus five knockout strains, each having had the response regulator gene of a specific TCS inactivated, which we designated CJB111Δrr2, CJB111Δrr11, CJB111Δrr16, CJB111Δrr17, and CJB111Δrr21, abbreviated herein as the Δrr2 strain, Δrr11 strain, etc.

To obtain the chromosomally complemented strains, plasmids PJRS233KIrr2, -KIrr11, -KIrr16, -KIrr17, and -KIrr21 were purified from E. coli and complementation of the respective mutant strains was performed as described above. The result was replacement of the deleted rr gene with the WT form. The complemented strains were designated CJB111KIrr2, CJB111KIrr11, CJB111KIrr16, CJB111KIrr17, and CJB111KIrr21, abbreviated herein as the KIΔrr2 strain, KIrr11 strain, etc.

Bioinformatics methods.

The identification of histidine kinases (HKs) and response regulators (RRs) of putative TCS in the genome of GBS strain 2603 V/R was carried out using the P2CS (http://www.p2cs.org) database (42). A TCS was defined as an orphan if the predicted HK or RR gene had no adjacent partner in the locus or if the cognate partner contained any mutations (e.g., stop codons or frameshifts) that compromised the predicted gene product.

Gene sequences encoding the TCS were downloaded from the publically available genome of 2603 V/R in GenBank, except for the RR from TCS-3 and HK from TCS-17, which were downloaded from the publically available genome of CJB111. The genes used are listed in Table 1. All gene sequences were aligned against complete or draft genome sequences using FASTA version 3.4t25. A gene was considered present if the identity was >75% on at least 90% of the query sequence. Genes interrupted at the end of contigs, alignments with over 95% identity on less than 40% of the query length, and hits divided into pieces on the same contig were classified as having “ambiguity in alignment.”

For polypeptide analyses, all of the genes were translated into proteins and those having mutations leading to truncated proteins were discarded. A numeric identifier was assigned to each different putatively active polypeptide for each gene. For each strain, a polypeptide profile of TCS was created. The goeBurst algorithm (43) was used to generate a minimum spanning tree (MST) from the profiles using Phyloviz (44). In this tree, each vertex represents a specific profile and each edge connects the profile to its closest ones.

Functional domains in each HK protein were identified by SMART and/or PFAM (http://smart.embl-heidelberg.de). Subsequently, HKs were manually assigned a signaling mechanism according to proposed criteria (8). Extracellular-sensing HKs typically have two TM helices with an intervening extracytoplasmic domain of 50 to 300 amino acids, lacking large cytoplasmic linker regions; transmembrane-sensing HKs are characterized by the presence of 2 to 20 transmembrane regions connected by very short intra- or extracellular linkers; cytoplasmic-sensing HKs include either membrane-anchored proteins that do not fulfill the criteria described above and have input domains (e.g., PAS, GAF) localized on the intracellular side or soluble HKs with no transmembrane regions.

Expression microarray analysis.

Expression microarray analysis was performed using a custom-made Agilent 60-mer oligonucleotide microarray. Oligonucleotide sequences were selected with the Agilent eArray system, using 6 probes per gene. The complete set of annotated genes (2,232) in the CJB111 genome was thus covered (ftp://ftp.jcvi.org/pub/data/Microbial_Genomes/s_agalatiae_cjb111/). RNA samples used for microarray analysis were prepared as follows. Bacteria for RNA extraction (five mutant strains and the isogenic WT strain) were grown in triplicate cultures on three separate occasions. Hence, for each strain, nine independent GBS cultures (nine biological replicates obtained on three separate days) served as the source of RNA. Bacteria were harvested twice, at an OD600 of 0.3 (EL phase) and an OD600 of 1.8 to 1.9 (ES phase, ~15 min into growth arrest). To rapidly arrest transcription, 10 ml of bacteria was cooled on ice and added to 10 ml of frozen THB medium in a 50-ml conical tube. GBS cells were then collected by centrifugation for 15 min at 4,000 rpm, 4°C, and resuspended in 800 µl of TRIzol (Invitrogen). Bacteria were disrupted mechanically by agitation with lysing matrix B in 2-ml tubes (DBA Italia) using a homogenizer (FastPrep-24; MP Biomedicals) for 60 s at 6.5 m/s for two cycles, and kept on ice for 2 min between the cycles. Samples were then centrifuged for 5 min at 8,000 × g, 4°C, and RNA was extracted with a Direct-zol RNA miniprep kit (Zymo Research) according to the manufacturer’s instructions. RNA samples were treated with DNase (Roche) for 2 h at 37°C and further purified using the RNeasy minikit (Qiagen), including a second DNase treatment on the column for 30 min at room temperature, according to the manufacturer’s instructions. Five micrograms of pooled RNA from each triplicate was labeled (Cy5-dCTP/Cy3-dCTP; Euroclone) and purified (Wizard SV gel and PCR clean-up system; Promega). Labeled RNA from the three independent days of RNA extraction was then pooled and used for microarray analysis. Labeling, chip hybridization, washing, scanning, and data extraction were performed according to the procedures described by Agilent for two-color microarray-based gene expression analysis. After data acquisition, slide normalization was performed by a lowess (locally weighted scatterplot smoothing) algorithm, as implemented by the Agilent Feature Extraction software, version 9.5.3. The average log ratio of fluorescence signals [log2(Cy5/Cy3)] of probes corresponding to the same gene was computed within each slide. Student’s t test statistics were performed to test for differentially expressed genes.

qRT-PCR analysis.

cDNA was prepared using the reverse transcription system (Promega) by using 500 ng of RNA per reaction mixture volume. Real-time quantitative PCR (qRT-PCR) was performed on 50 ng of cDNA that was amplified using LightCycler 480 DNA SYBR green I master (Roche). Reactions were monitored using a LightCycler 480 instrument and software (Roche). The transcript amounts under each condition were standardized to the transcription of an internal control gene (gyrA) and compared with standardized expression in the wild-type strain (cycle threshold [ΔΔC] method). Validation of microarray data by qRT-PCR was performed on a selection of 9 genes differentially expressed in mutant strains, using the RNA prepared for microarray analysis (see above) and primers reported in Table S1 in the supplemental material. Moreover, we prepared new RNA samples from WT, knockout, and complemented strains (three biological replicates obtained on three different days) to generate cDNA for qRT-PCR confirmation of highly regulated genes.

For RNA extraction from bacteria grown with different carbon sources, strains were first inoculated into CDM and grown at 37°C until an OD600 of 0.5 was reached. Cultures were harvested by centrifuging at 4,000 × g, 4°C, for 10 min, and each pellet was resuspended in 10 ml of CDM without glucose and split in two. Each 5-ml sample was then supplemented with 10 mg/ml of glucose or fructose-6-phosphate. The bacteria were then incubated for 30 min at 37°C, and RNA was extracted as described for the microarray analysis.

Bacterial lysate was obtained from a 10-ml culture of the WT strain in late logarithmic phase (OD of 1). Briefly, bacteria resuspended in PBS were disrupted mechanically by agitation with lysing matrix B in 2-ml tubes (DBA Italia) using a homogenizer (FastPrep-24) for 60 s at 6.5 m/s for four cycles. The tubes were kept on ice for 2 min between the cycles. Samples were centrifuged for 5 min at 8,000 × g, 4°C, and the supernatant was collected and designated as lysate. For the RNA extraction, strains were grown in CDM until reaching an OD600 of 0.5, harvested, split in two, and resuspended in 5 ml of CDM without glucose. Each sample was supplemented with 800 µl of PBS or with 800 µl of lysate for 30 min at 37°C before RNA extraction.

Phenotype microarray analysis.

Phenotype microarray (PM) technology uses the irreversible chemical reduction of a patented dye as a reporter of active metabolism (45). The CJB111, Δrr16, and ΚIrr16 strains were used with microplates PM1 and PM2A (Biolog), testing 192 different single carbon sources. Bacteria on THB agar plates were scraped off and suspended in 5 ml of inoculation fluid (IF-0; Biolog) to an OD of 0.3. Samples were then diluted 1:10 in IF-0 to obtain a turbidity of 85% (OD of ~0.02 to 0.03) and kept on ice. Growth medium constituents were prepared according to Biolog’s procedures for S. agalactiae and other Streptococcus species. PM microplates were inoculated by adding 100 µl bacterial suspension/well and then incubated for 48 h at 37°C in an OmniLog reader (Biolog). Data points were collected every 15 min. Data were then analyzed using the OmniLog software. The unspecific background (well A1 in each plate, negative control) was subtracted from the other samples.

In vivo and in vitro infection experiments.

Animal studies were approved by the Office of Lab Animal Care at San Diego State University and conducted under accepted veterinary standards. Animal experiments were performed in compliance with the Novartis Animal Health and Welfare guidelines.

We adopted a previously described murine model of hematogenous GBS meningitis (28), with the difference that preliminary dose-ranging experiments established the 50% lethal dose (LD50) dose of CJB111 as 1.5 × 107 CFU per mouse (outbred 6-week-old male CD1 mice; Charles River Laboratories). The mice were followed for 2 weeks. Also, since histopathology was not performed, we considered the experiment a model of systemic infection.

The in vivo mouse model of GBS vaginal colonization has been described previously (26). Female CD1 mice (8 to 16 weeks old) were obtained from Charles River Laboratories and used for colonization assays. In vitro adhesion was performed as previously described (26) using the immortalized human vaginal epithelial cell line VK2/E6E7, obtained from the American Type Culture Collection (ATCC CRL-2616) (46). Five to 10 passages were used for all cell assays. Cells were maintained at 37°C in a 5% CO2 atmosphere in keratinocyte serum-free medium (KSFM; Invitrogen).

Microarray data accession numbers.

The microarray chip layout was submitted to EBI ArrayExpress and is available under accession number A-MEXP-2382. Experimental results are available under accession number E-MTAB-2423.

Distribution of TCS-17 protein variants for SAM1896 (the HK) and SAM1897 (the RR). Specific polypeptide variants are indicated by numbers. Also included in the analysis are strains where the proteins are absent (Abs) or where the genome sequence presents ambiguous areas (Amb; e.g., contig breaks) and where a stop codon (Stop) terminates the polypeptide. Download

Figure S1, EPS file, 0.1 MB

qRT-PCR analysis of selected TCS in four different phases of growth, early (EL), mid- (ML), and late logarithmic (LL) phases and early stationary (ES) phase. Each bar represents three biological replicates, and error bars show the standard errors of the means. Fold change is relative to the expression level of each system in EL phase. Download

Figure S2, EPS file, 0.1 MB

Growth curves of WT and isogenic mutant strains in THB medium. Curves represent the means of triplicate samples. Download

Figure S3, EPS file, 0.1 MB

Correlation between microarray and qRT-PCR. Values obtained for the two methods were plotted on the x and y axes. Each dot represents the fold change of mutant versus WT for a highly regulated gene, indicated with an arrow. The mutant strains used were Δrr2 (red), Δrr11 (green), and Δrr16 (blue) strains. Download

Figure S4, EPS file, 0.1 MB

Phenotype microarray analysis on WT, Δrr16, and KIrr16 strains was performed using PM1 and PM2A plates (http://www.biolog.com/pdf/pm_lit/PM1-PM10.pdf). Kinetic plots are shown in red (WT) or in green (mutant and complemented strains), and superimposition is shown in yellow. Metabolic differences between WT and Δrr16 mutant or WT and KIrr16 mutant are shown in superimposed kinetic plots obtained by growing bacteria on PM1 (A and C) or PM2A (B and D). Download

Figure S5, EPS file, 1.8 MB

Oligonucleotides used in this study.

Table S1, DOC file, 0.1 MB.