Autoinducer 2 Signaling via the Phosphotransferase FruA Drives Galactose Utilization by Streptococcus pneumoniae, Resulting in Hypervirulence

The primary habitat of S. pneumoniae is the human upper respiratory tract, where the principal sugar available for use as a carbon source is galactose (17, 18). However, previous studies of the growth properties of S. pneumoniae luxS mutants have employed media that contain glucose, and no growth defects were reported (16, 19, 20). We therefore commenced by comparing the growth phenotypes of serotype 2 S. pneumoniae strain D39 with that of its luxS deletion mutant (D39ΔluxS) and a luxS-overexpressing strain (D39luxS⁺) in C+Y medium (21) containing either glucose (C+Y-Glc) (Fig. 1A) or galactose (C+Y-Gal) (Fig. 1B). In C+Y-Glc, the growth kinetics of D39 and D39ΔluxS were virtually identical, while there was a slight growth retardation of the D39luxS⁺ strain (Fig. 1A). In contrast, in C+Y-Gal, there was a marked increase in the generation time and a reduction in the final cell density of both D39ΔluxS and D39luxS⁺ compared to those of the D39 parent strain (Fig. 1B). This suggested that either too little or too much luxS expression may be detrimental in this medium. To distinguish between true QS effects mediated by AI-2 from the indirect consequences of luxS disruption on the activated methyl cycle, we then examined the capacity of various concentrations of exogenous purified DPD (AI-2) to complement the growth defect of D39ΔluxS in C+Y-Gal. Supplementation of C+Y-Gal with 10 μM AI-2 partially restored growth, but supplementation with lower (4 μM) or higher (200 μM) AI-2 concentrations had no impact whatsoever (Fig. 1C). Furthermore, the growth defect of D39luxS⁺ in C+Y-Gal could not be complemented by any of these concentrations of AI-2 (result not presented). Taken together, these findings indicate that extracellular AI-2 is either sensed or imported into the cell and that this process appears to be under carbon catabolite repression.

We then determined whether the known virulence defect of a luxS mutant relative to its wild-type parent D39 (15) could be complemented by administration of exogenous synthetic AI-2. Mice were challenged i.n. with S. pneumoniae D39, D39ΔluxS, or D39luxS⁺ with or without AI-2 administered i.n. at time zero and at 6, 12, and 18 h postinfection. Mice challenged with D39ΔluxS had significantly lighter bacterial loads in the nasopharynx, lungs, and blood at 24 h postchallenge than those challenged with D39 (Fig. 2). However, i.n. administration of AI-2 significantly increased the bacterial loads in all host niches, such that the virulence of D39ΔluxS with AI-2 was indistinguishable from that of D39 without AI-2, while D39 with AI-2 became hypervirulent. In contrast, mice challenged with D39luxS⁺ had significantly lighter bacterial loads in the nasopharynx than D39-challenged mice (Fig. 2A) and no bacteria could be detected in the lungs or blood of these mice. Moreover, unlike the findings with D39- and D39ΔluxS-infected mice, i.n. administration of AI-2 did not increase the numbers of D39luxS⁺ bacteria in any of the host niches (Fig. 2A to C).

Histological analysis of the infected lungs at 12 h postinfection reflected the above findings (Fig. 3A). Lungs from mice challenged with D39ΔluxS exhibited less severe inflammation than those from mice challenged with D39, as evidenced by hematoxylin-and-eosin (H&E)-stained sections, resulting in a lower mean pathological score (Fig. 3B). Two out of three mice challenged with D39ΔluxS showed no or only mild inflammation signs in the lung tissue (only a minor degree of swelling of alveolar/bronchiole walls and minor lung tissue leukocyte infiltration), similar to those of uninfected mice (Fig. 3A, D39ΔluxS panel). Importantly, AI-2 treatment of both D39- and D39ΔluxS-infected mice increased the degree of inflammation, such that the lungs of mice challenged with D39ΔluxS and treated with AI-2 were similar to those of mice challenged with D39, while the degree of inflammation in the lungs of mice challenged with D39 and treated with synthetic AI-2 was higher than that of mice challenged with D39. AI-2 treatment increased the mean pathological score of lungs challenged with D39ΔluxS to 5.0, compared with 4.3 for those challenged with D39, while AI-2 treatment of D39-challenged mice increased the mean pathological score to 7.0 (Fig. 3B).

As further confirmation that exogenous AI-2 increases virulence, mice (n = 5 per group) challenged with D39 and treated with AI-2 as described above had a median survival time of 53 h, compared to 73 h for those given phosphate-buffered saline (PBS) (P < 0.05; Mann-Whitney U test). Furthermore, at 48 h, the geometric mean bacterial loads in the blood were 5.83 × 10⁸ and 7.53 × 10⁶ CFU/ml, respectively (P < 0.01).

Thus, we conclude that the virulence defect of D39ΔluxS is attributable to loss of AI-2 signaling rather than any metabolic side effect of perturbation of the activated methyl cycle and that exogenous AI-2 directly increases virulence. The nonvirulent phenotype observed in D39luxS⁺ and the lack of in vivo efficacy of exogenous AI-2 in this strain are consistent with the in vitro growth defect observed in C+Y-Gal, as well as the inability of exogenous AI-2 to restore growth (Fig. 1C).

Bacterial loads in a given tissue type are determined not only by the growth rate but also by the rate of clearance by innate immune cells. We therefore examined the capacity of THP-1-derived human macrophages to internalize the various strains with or without exogenous AI-2. The numbers of intracellular bacteria in macrophages infected with D39ΔluxS and D39luxS⁺ were significantly greater than (roughly double) that of wild-type D39-infected cells (Fig. 4A). Preincubation of the bacteria with synthetic AI-2 significantly decreased the ability of the macrophages to internalize D39 or D39ΔluxS, but there was no effect on the uptake of D39luxS⁺ (Fig. 4A). Thus, exogenous AI-2 significantly increases the resistance of pneumococci to internalization by macrophages.

In view of these findings, we hypothesized that the stimulatory effects of exogenous AI-2 on virulence of S. pneumoniae D39, and the attenuating effect of luxS mutation, could be due to opposite impacts on capsular polysaccharide (CPS) production, as the capsule is known to be strongly antiphagocytic (22). Accordingly, we measured the total amount of type 2 CPS in cells grown in C+Y-Glc or C+Y-Gal. Production of type 2 CPS was significantly lower in the D39ΔluxS and D39luxS⁺ strains than in wild-type D39 in C+Y-Gal (Fig. 4A), which is consistent with the attenuated virulence of these mutants. Moreover, exogenous AI-2 restored CPS production by D39ΔluxS to a level commensurate with that of the wild type and induced a significant increase in CPS production by D39 (Fig. 4B), but again, AI-2 had no effect on D39luxS⁺. Thus, CPS production in vitro paralleled the hypervirulence of D39 in the presence of exogenous AI-2 in vivo. However, when cells were grown in C+Y-Glc, there were no significant differences in either bacterial growth or CPS production among D39ΔluxS, D39luxS⁺, and wild-type D39, and exogenous AI-2 had no effect whatsoever (Fig. 4C). Moreover, the total CPS production by D39 in C+Y-Glc was less than half of that seen in cells grown in C+Y-Gal (result not shown), consistent with previous observations (23).

We then performed a transcriptomic analysis to examine the global effect of exogenous AI-2 on gene expression patterns in D39 and D39ΔluxS (Table 1). For these studies, cells were grown in C+Y without added sugars and RNA was extracted 30 min after the addition of 10 μM AI-2. Addition of AI-2 caused significant changes in the expression of 22 genes in D39 but only 9 genes in D39ΔluxS (adjusted P < 0.05). Importantly, no differences in the expression of any of the genes in the CPS biosynthesis (cps) locus were observed in either strain. These data indicate that the impact of AI-2 on total CPS production, and hence virulence, is independent of cps transcription. However, AI-2 did induce significant differences in the expression of four genes/operons in both strains, notably, an operon (spd_0771-0773) comprising lacR1, fruB, and fruA (encoding the lactose phosphotransferase system repressor, phosphofructokinase, and fructose PTS enzyme IIABC, respectively), piuA (iron-binding lipoprotein), adhE (an iron-containing aldehyde-alcohol dehydrogenase), and brnQ (branched-chain amino acid transport system II carrier) (Table 1). The differential expression of these genes was confirmed by real-time quantitative reverse transcription (qRT)-PCR analysis (Fig. 5). Interestingly, in both D39 and D39ΔluxS, fruA and adhE were repressed in the presence of AI-2 whereas brnQ and piuA were induced. However, the transcription of all four genes was lower in D39ΔluxS than in D39.

We therefore constructed D39 derivatives with deletion mutations in the four AI-2-regulated genes identified above to determine whether any play a direct role in AI-2 sensing by S. pneumoniae. The fruA gene was a particularly promising candidate, as it encodes a fructose-PTS uptake system component that might be capable of recognizing a ketopentose like AI-2. Moreover, PTS uptake systems typically phosphorylate their cargo, resulting in intracellular retention and accumulation, a critical feature of cell density-dependent signaling molecules such as AI-2 (12). In the first instance, we used CPS production in C+Y-Gal as a readout of AI-2-mediated signaling. Similar to the earlier observations for D39 and D39ΔluxS, exogenous AI-2 increased the total CPS in D39ΔpiuA and D39ΔadhE. Strikingly, AI-2 had no impact whatsoever on CPS production by D39ΔfruA (Fig. 6). Moreover, when tested in the murine i.n. challenge model (n = 5 per group), D39ΔfruA was at least as virulent as D39, but importantly, treatment with exogenous AI-2 had no effect on the loads of D39ΔfruA bacteria in the nasopharynx, blood, and lungs at 24 h postchallenge. Respective geometric mean bacterial loads in the absence or presence of AI-2 were 7.94 × 10³ versus 1.10 × 10⁴ CFU in the nasopharynx, 4.07 × 10⁵ versus 1.55 × 10⁵ CFU/ml in the blood, and 7.11 × 10⁵ versus 1.20 × 10⁵ CFU in the lungs. However, the impact of the brnQ deletion on CPS production could not be assessed in the D39ΔbrnQ strain because of a major in vitro growth defect.

Collectively, the above findings indicate that AI-2 signaling in S. pneumoniae is dependent on FruA and suggest that the impacts on growth, CPS production, and virulence may be mediated via effects on Gal uptake and/or catabolism. In S. pneumoniae, there are two pathways for Gal uptake and utilization: a PTS (SPD_0559-0561) whereby Gal-6-phosphate (Gal6P) is fed into the glycolytic pathway via the tagatose-6-phosphate pathway and a Gal-specific ABC transporter (SPD_0088-0090) that delivers free Gal into the Leloir pathway. The latter pathway is critically important, as it generates Glc-1-phosphate (G1P), from which various UDP-activated sugar precursors required for CPS synthesis are derived (18, 23). We therefore examined the transcription of spd_0088 and Leloir pathway genes galR (transcriptional regulator), galK (galactokinase), and galT (galactose-1-phosphate uridyltransferase), as well as lacD (tagatose 1,6-diphosphate aldolase, the last enzyme of the tagatose-6-phosphate pathway), in D39 and D39ΔluxS grown in a fully synthetic chemically defined medium (CDM) (24) with Gal as the sole carbon source (CDM-Gal) with or without 10 μM AI-2. Transcription of cps2A, the first gene of the locus encoding type 2 CPS biosynthesis, was also assessed. No differences in cps2A transcription between D39 and D39ΔluxS were seen, with or without AI-2 in CDM-Gal, consistent with the previous RNA-seq data (result not shown). However, the expression of galR, galK, and galT was significantly lower in D39ΔluxS than in D39. Moreover, addition of AI-2 significantly increased the expression of all three genes in both strains. Indeed, exogenous AI-2 increased the expression of the Leloir pathway genes in D39ΔluxS such that it was essentially identical to that in D39 without added AI-2, thereby fully complementing the luxS mutation (Fig. 7B to D). Expression of the Gal ABC transporter spd_0088 gene showed a trend identical to that of the Leloir pathway genes, but the differences did not reach statistical significance (Fig. 7A). However, the tagatose pathway gene lacD did not respond to exogenous AI-2 in either strain (Fig. 7E). Interestingly, unlike the findings in medium without added sugars (Fig. 5), in CDM-Gal, expression of fruA by D39 and D39ΔluxS in response to AI-2 mirrored that of the Leloir pathway genes (Fig. 7F). Importantly, in D39ΔfruA, AI-2 had no effect whatsoever on the expression of Leloir pathway genes galR, galK, or galT or of spd_0088 (Fig. 7A to D). Thus, FruA is essential for the transcriptional response of galactose uptake and utilization genes to exogenous AI-2 that results in upregulation of CPS production and increased virulence. It is important to note that FruA itself plays no known catalytic role in Gal uptake or catabolism or CPS synthesis. Thus, our findings that the baseline Leloir pathway gene expression, total CPS production, and virulence of D39ΔfruA are similar to or even higher than those of D39 do not conflict with the hypothesis that FruA transduces the AI-2 signal.

In order to account for the lower CPS production and avirulence of the luxS-overexpressing strain observed in Fig. 2 and 4, we then examined spd_0088, galT, and fruA expression in D39luxS⁺ versus D39 (Fig. 8). In D39, all three genes were significantly upregulated in the presence of exogenous AI-2. However, all three genes were significantly downregulated in D39luxS⁺ relative to D39 and expression did not increase in the presence of AI-2. These findings further illustrate the tight nexus between galactose uptake/catabolism and CPS production. These data also indicate that there is an optimum AI-2 concentration range above which responses to the QS molecule are switched off. This was further examined by measuring spd_0088 and galT expression in D39 and D39ΔluxS exposed to higher AI-2 concentrations. In D39, addition of 60 μM AI-2 upregulated spd_0088 and galT, but these genes were significantly downregulated at 80 and 120 μM AI-2, respectively (Fig. 9). In D39ΔluxS, exogenous AI-2 remained stimulatory at concentrations of up to 120 μM, but significant downregulation occurred at 200 μM. This higher inhibitory threshold is presumably attributable to the absence of baseline endogenous AI-2 synthesis in the luxS mutant.

The S. pneumoniae strains used in this study are listed in Table 2. Strains were routinely cultured on Columbia blood agar base supplemented with 5% horse blood at 37°C in 95% air–5% CO₂. Blood agar plates were supplemented with 0.2 μg/ml erythromycin or 200 μg/ml spectinomycin for the selection of mutant strains. Growth experiments with pneumococci in liquid culture were performed with a casein-based semisynthetic medium (C+Y) (21). For animal challenge experiments, pneumococci were grown in nutrient broth supplemented with 10% (vol/vol) horse serum. Gene expression experiments were performed with a CDM (38) supplemented with catalase (10 U/ml; Sigma) at 37°C under static conditions. The AI-2 precursor molecule, DPD, was purchased from Omm Scientific Inc., TX.

Genes were deleted from S. pneumoniae D39 and replaced with an erythromycin or spectinomycin resistance cassette by transformation with a linear DNA fragment constructed by overlap extension PCR (39) with primers listed in Table 3.

Analysis of pneumococcal growth was performed in 96-well microtiter plates (200-μl culture volume) at 37°C. Culture optical density at 600 nm (OD₆₀₀) was measured at 30-min intervals over a 24-h time period with a Spectromax spectrophotometer (Molecular Devices).

Our animal experiments were approved by the University of Adelaide Animal Ethics Committee. Female outbred 4- to 6-week-old CD-1 (Swiss) mice were anesthetized and inoculated i.n. with 10⁷ CFU of wild-type D39, D39ΔluxS, or D39luxS⁺ in 50 µl of medium as described previously (40). Two groups of five mice were used for each strain tested, and at time zero and 6, 12, and 18 h postinfection, one group received 1 μg of AI-2 in 10 µl of PBS i.n. (without anesthesia), while the other group received 10 μl of PBS as a control. Nasal wash, nasopharyngeal tissue, lung, and blood samples were collected and processed as previously described at 24 h postinfection. Samples were homogenized where appropriate, serially diluted, and plated onto blood agar for enumeration of viable pneumococci as previously described (40, 41). Statistical analyses of log-transformed data were performed with a two-tailed Student t test; a P value of <0.05 was defined as statistically significant.

Twelve hours after an i.n. challenge or no challenge (control), lungs were removed from euthanized mice, fixed in 4% formaldehyde overnight at 4°C, embedded in paraffin, sectioned, stained with H&E, and examined by light microscopy (Axiophot; Zeiss, Germany). Digital images were captured and processed with a CCIZ soft imaging system and analySIS LS Research software. Lung pathological scores (maximum = 8) for each mouse were determined by a blinded observer according the following scheme: congested capillaries, scored 1 to 2; thickened alveolar wall, scored 1 to 2; secretions and hemorrhage in alveolar and bronchiolar spaces, scored 1 to 2; parenchymal hemorrhage and neutrophil infiltration, scored 1 to 2.

All tissue culture media and reagents were obtained from Gibco. THP-1 cells (ATCC TIB-202) were grown in 95% air–5% CO₂ at 37°C in complete RPMI medium (RPMI with phenol red, supplemented with 10% fetal calf serum, 10 mM HEPES, 50 IU/ml penicillin, and 50 µg/ml streptomycin). Ten-centimeter dishes were seeded with 3.5 × 10⁶ THP-1 cells and differentiated by adding phorbol 12-myristate 13-acetate to a final concentration of 10 ng/ml and incubating them for 3 days at 37°C in 95% air–5% CO₂. Differentiated cells attach to the plastic surface. Following the 3 days of incubation, cells were washed twice with complete RPMI and then rested at 37°C in 95% air–5% CO₂ in complete RPMI for a further 3 days. At 90% confluence, differentiated THP-1 cells were incubated with bacteria (100 µl) at a 10:1 ratio of bacteria to THP1 cells. After 45 min, plates were washed and cells were reincubated with 10 μg/ml penicillin and 200 μg/ml gentamicin (Sigma, Germany) for 30 min to kill extracellular pneumococci. Intracellular bacteria were enumerated after lysis with 0.01% Triton X-100 by plating serial dilutions on blood agar.

Pneumococcal CPS samples were prepared as previously described (42). In brief, deoxycholate-lysed pneumococci were treated by adding 100 U of mutanolysin (Sigma, Australia), 50 U of DNase I (Roche, Australia), and 50 μg of RNase A, incubating them overnight at 37°C, and treating them with 100 μg/ml proteinase K at 56°C for 4 h. Uronic acid was then quantitated colorimetrically as described previously (42).

S. pneumoniae strains were grown in C+Y without added sugars to an OD₆₀₀ of 0.15 and then treated with or without 10 μM AI-2 for 30 min. RNA extraction was performed with a Qiagen RNase minikit according the manufacturer’s protocol. Each sample was derived from three independent cultures (biological replicates) and pooled prior to transcriptome sequencing (RNA-seq) analysis. Total RNA was submitted to the Australian Genome Research Facility (Melbourne, Australia) for sequencing. Briefly, the Epicentre Bacterial Ribozero kit (Illumina) was used to reduce the rRNA content, and then the Ultra Directional RNA kit (New England Biolabs) was used to generate bar-coded libraries. Prepared libraries were then sequenced with the Illumina HiSeq 2500. Trimmed RNA-seq fastq files were then mapped to the D39 reference genome with BOWTIE2 version 2.2.3 (43). Counts for each gene were obtained with the aid of BEDTools (44), and differential gene expression was examined with DESeq (45).

For qRT-PCR analyses, cells were grown in CDM-Gal to an OD₆₀₀ of 0.2 and treated with or without 10 μM AI-2 for 30 min. RNA was extracted from triplicate cultures as described above, but individual rather than pooled extracts were analyzed. Gene expression was assayed by one-step relative real-time qRT-PCR in a Roche LC480 real-time cycler essentially as described previously (40). The specific primers used for the various genes are listed in Table 3, and they were used at a final concentration of 200 nM. As an internal control, primers specific for 16S rRNA were employed. Amplification data were analyzed by the comparative critical threshold (2^−ΔΔCT) method (46) and are presented as total expression relative to that of the 16S rRNA gene.