Capacity To Utilize Raffinose Dictates Pneumococcal Disease Phenotype

S. pneumoniae is a component of the commensal nasopharyngeal microflora of humans, but from this reservoir, it can progress to localized or invasive disease with a frequency that translates into massive global morbidity and mortality. However, the factors that govern the switch from commensal to pathogen, as well as those that determine disease tropism, are poorly understood. Here we show that capacity to utilize raffinose can determine the nature of the disease caused by a given pneumococcal strain. Moreover, our findings provide an interesting example of convergent evolution, whereby pneumococci belonging to two unrelated serotypes/lineages exhibit SNPs in separate genes affecting raffinose uptake and utilization that correlate with distinct pathogenic profiles in vivo. This further underscores the critical role of differential carbohydrate metabolism in the pathogenesis of localized versus invasive pneumococcal disease.

teen SNPs and indels present within protein coding sequences of 4559 and 947 resulting in a change in the predicted amino acid sequence are listed in Table 1. The genes affected included those predicted to be involved in metabolism and energy production, transcriptional regulation, transporters, and putative virulence factors. Among the latter category, an SNP resulting in a L43P substitution was identified in cpsE, which encodes the glycosyl transferase that initiates assembly of the capsular polysaccharide (CPS) repeat unit. However, we have previously shown that there is no difference in total CPS production between 4559 and 947 (7). The SNP in the putative plasmin and fibronectin-binding protein gene pfbA is also a conservative T318M substitution. The protein encoded by iga is truncated in 4559 compared to 947, but only by four amino acids. On the other hand, the nanB sequence in 947 has a premature stop codon that truncates the protein by 330 amino acids (47% of the 4559 protein), presumably inactivating the gene product. Mutagenesis studies have previously shown that NanB contributes to colonization of both the upper and lower respiratory tract of mice, albeit to a lesser extent than the major neuraminidase NanA (8). Nevertheless, 4559 and 947 colonize the nasopharynx equally well (7). Interestingly, SNPs were identified in two metabolic genes, coding for ATP-dependent 6-phosphofructokinase (pfkA) and a glycogen synthase (glgA), as well as in two helix-turn-helix (HTH)-type transcriptional regulators, scrR and rafR, involved in metabolism of sucrose and raffinose, respectively. Given the importance of carbohydrate metabolism to S. pneumoniae (9), we employed a phenotypic microarray to compare the capacity of 4559 and 947 to metabolize over 100 different carbohydrates (see Materials and Methods). The only difference observed between the ear and blood isolates was a reduced capacity of the former (947) to grow in medium containing raffinose as the sole carbon source (data not shown).
Genetic differences between ear/blood isolate pairs that are common to two unrelated serotypes/ST lineages would be strong candidates for determinants of tissue tropism. Genomic comparisons were therefore also made between two serotype 3 ST180 ear and blood isolates (strains 180/2 and 180/15, respectively), which like the serotype 14 ST15 isolates, have previously been shown to exhibit distinct tissue tropism in mice in accordance with clinical isolation site (6). There were no differences in ARs between the two strains, while SNPs and indels impacting the deduced amino acid sequence for 27 genes were identified (Table 2). Interestingly, there were no affected genes in common with those in Table 1. However, an I227T SNP was detected in the serotype 3 rafK gene, encoding the ATP-binding protein component of the raffinose Raffinose Metabolism Determines Pneumococcal Virulence ® ABC transporter. RafK is known to be essential for activation of other raf operon genes, and the SNP identified in ST180 isolates is located in the conserved regulatory domain motif 1 (10). Thus, potential defects in raffinose uptake/metabolism appear to be a common feature of ear isolates from both serotypes/lineages. Blood isolates utilize raffinose more efficiently than ear isolates. In view of the SNPs in genes associated with raffinose metabolism between ear and blood isolates in two unrelated serotypes/STs and the fact that the serotype 14 ear and blood isolates differed only in their ability to metabolize raffinose on phenotypic microarray analysis, in vitro growth phenotypes were further investigated. Strains 4559 and 947, as well as another pair of serotype 14 ST15 blood and ear isolates (4534 and 51742, respectively), were grown in a chemically defined medium (CDM) with either glucose or raffinose as the sole carbon source (designated CDMϩGlc and CDMϩRaf, respectively) ( Fig. 1). In CDMϩGlc, there were no significant differences in growth rates between blood and ear isolates. However, in CDMϩRaf, the two blood isolates grew at a higher rate and to a higher final culture density (optical density at 600 nm [OD 600 ]) than either of the serotype 14 ST15 ear isolates. Similarly, there was no significant difference in growth rates of the serotype 3 ST180 ear and blood isolates (180/15 and 180/2, respectively) in CDMϩGlc, but the blood isolate grew better than the ear isolate in CDMϩRaf (Fig. 1). Thus, defective growth in raffinose appears to be a common defect in ear isolates relative to serotype/ST-matched blood isolates.
The raffinose uptake/utilization operon in S. pneumoniae comprises genes encoding transcriptional regulators (rafR and rafS), an ␣-galactosidase (aga), the ABC transporter substrate-binding protein and two cognate permeases (rafE, rafF, and rafG), a sucrose phosphorylase (gtfA), and a protein of unknown function (rafX), as well as the ATP binding protein component of the transporter (rafK), which is independently located in the genome (11) (Fig. 2). To determine if the difference in ability to utilize raffinose  . Similar growth studies were also performed for serotype 3 ST180 strains 180/15 (blood isolate) and 180/2 (ear isolate). OD 600 was measured every hour for 12 h. Data are mean OD 600 Ϯ standard deviation (SD) from triplicate assays. Raffinose Metabolism Determines Pneumococcal Virulence ® was then extracted, and levels of aga, rafG, and rafK mRNA, representative of each of the three rafR-regulated transcriptional units, were then measured relative to 16S rRNA by quantitative real-time reverse transcription-PCR (qRT-PCR). In every case, expression levels for all three genes were significantly greater in the blood isolates than in the respective ear isolates (Fig. 3). As further confirmation, blood and ear isolates belonging to serotype 23F ST81 were also tested for growth in CDMϩGlc and CDMϩRaf, as well as for expression of aga, rafG, and rafK (Fig. 4). Again, the blood isolate grew to a higher OD 600 than the ear isolate in CDMϩRaf, but not in CDMϩGlc. Moreover, expression of all three raf genes was significantly higher in the blood isolate than in the ear isolate.
The SNP in 947 rafR is responsible for its raffinose phenotype. In order to test whether the distinct in vitro and in vivo phenotype of 947 relative to 4559 was attributable to the SNP in rafR, allelic-exchange mutagenesis was performed in 4559 and 947, generating a 4559 derivative with its rafR allele replaced by that from 947 (designated 4559 947rafR ) and a 947 derivative expressing the 4559 rafR allele (947 4559rafR ) (see Materials and Methods). Growth assays in CDMϩGlc showed no significant differences in growth rates between 4559, 947, 4559 947rafR , and 947 4559rafR . However, in CDMϩRaf, growth of 4559 947rafR was at least as poor as that of 947, while growth of 947 4559rafR was similar to that of 4559 (Fig. 5A). Expression of aga, rafG, and rafK was then examined in 4559, 947, 4559 947rafR , and 947 4559rafR by qRT-PCR after 30 min of growth in CDMϩRaf. For all three genes, expression levels in 947 4559rafR were indistinguishable from those in 4559, while expression in 4559 947rafR was essentially the same as that in 947 (Fig. 5B). Thus, exchange of rafR alleles between 4559 and 947 significantly impacts both growth phenotype and raf operon gene expression in CDMϩRaf.
Virulence phenotypes of 4559 and 947 and their rafR exchange mutants. In order to determine whether the marked difference in virulence phenotypes of 4559 and FIG 3 Expression of raffinose pathway genes by serotype 14 and 3 blood and ear isolates. The indicated strains were grown in CDMϩGlc to an OD 600 of 0.2, washed and resuspended in CDMϩRaf, and then incubated at 37°C for a further 30 min. RNA was then extracted, and levels of aga, rafG, and rafK mRNA were analyzed by qRT-PCR using 16S rRNA as an internal control (see Materials and Methods). The data presented are the means Ϯ SD from three independent experiments. *, P Ͻ 0.05, **, P Ͻ 0.01, and ****, P Ͻ 0.0001, by unpaired t test.
947 is also directly attributable to the SNP in rafR, 4559, 947, 4559 947rafR , and 947 4559rafR were tested in a murine intranasal challenge model. Groups of Swiss mice were challenged with 10 8 CFU of each strain, and bacterial loads were quantitated in various tissues 24 h postchallenge (Fig. 6). No significant differences in bacterial numbers in the nasopharynx were seen between any groups (Fig. 6), and no bacteria were detected in the blood of any mice (data not presented). However, 4559 was better able than 947 to persist in the lungs of infected mice, with significantly higher geometric mean (GM) bacterial load (P Ͻ 0.0001) and a significantly greater proportion of infected animals (14/16 versus 6/16; P Ͻ 0.01) (Fig. 6). On the other hand, bacterial loads of 947 in the ear were significantly greater than that for mice challenged with 4559 (P Ͻ 0.01), and the proportion of infected mice was also significantly greater (16/16 versus 7/16; P Ͻ 0.001) (Fig. 6). A similar trend was also seen in the brain (Fig. 6), in accordance with our previous report (4).
Exchanging the rafR alleles has a striking impact on virulence phenotype. In the lungs, both the GM CFU and proportion of infected mice for the group challenged with 4559 947rafR were significantly lower than those for the 4559 group (P Ͻ 0.0001 and P Ͻ 0.001, respectively). Indeed, the virulence phenotype of 4559 947rafR was indistinguishable from that of 947. Conversely, the GM bacterial load and proportion of infected mice for the 947 4559rafR group were significantly greater than those for the 947 group (P Ͻ 0.01 and P Ͻ 0.05, respectively); there were no significant differences in these parameters between the 947 4559rafR and 4559 groups. In the ear, both the GM Raffinose Metabolism Determines Pneumococcal Virulence ® CFU and proportion of infected mice for the group challenged with 4559 947rafR were significantly greater than those for the 4559 group (P Ͻ 0.05 and P Ͻ 0.01, respectively). Conversely, both the GM CFU and proportion of infected mice for the group challenged with 947 4559rafR were significantly lower than those for the 947 group (P Ͻ 0.01 in both cases). Moreover, there was no significant difference in either GM bacterial loads or proportions of infected mice between the 4559 947rafR and 947 groups or between the 947 4559rafR and 4559 groups (Fig. 6). A similar pattern is seen in the brain; the GM CFU for the 4559 947rafR group was significantly greater than those for either the 4559 or 947 4559rafR groups (P Ͻ 0.01 in both cases). Moreover, there was no significant difference in either GM bacterial loads or proportions of infected mice between the 4559 947rafR and 947 groups or between the 947 4559rafR and 4559 group (Fig. 6). Collectively, these data show that swapping the rafR allele between 4559 and 947 leads to a switch in their respective virulence profiles, and thus, the D49G SNP in rafR is entirely responsible for the observed difference in tissue tropisms between the serotype 14 ST15 blood and ear isolates.
Mutagenesis of rafK in serotype 3 ST180 blood and ear isolates. Attempts to construct rafK exchange mutants of serotype 3 ST180 blood and ear isolates (180/15 and 180/2, respectively) analogous to the rafR exchange mutants constructed for the serotype 14 strains were not successful. Thus, the impact of the SNP in RafK could not be directly tested. However, we were able to delete the native rafK genes from both type 3 strains (designated 180/15 ΔrafK and 180/2 ΔrafK, respectively). Both mutants were incapable of growth in CDMϩRaf, and expression of aga and rafG was virtually undetectable by qRT-PCR; rafK expression was also undetectable, as expected (Fig. 7). Thus, the ear isolate 180/2 exhibits a phenotype that is intermediate between that of the blood isolate 180/15 and either of the two ΔrafK mutants, consistent with partial functionality of the ear isolate RafK. S. pneumoniae rafK deletion mutants have previously been shown to be outcompeted by the wild type in the murine lung and nasopharynx (10,12). Similarly, in the present study, bacterial loads in the lungs, blood, ear, and brain were also lower for mice challenged with the 180/2 and 180/15 ΔrafK mutants relative to those challenged with the respective wild types at 48 h after intranasal challenge (result not presented). This indicates that even the intermediate level of raffinose pathway gene expression exhibited by ear isolate 180/2 contributes to virulence.

DISCUSSION
Pneumococci are strictly fermentative bacteria, relying solely on carbohydrate metabolism for energy and growth (13). However, carbohydrate availability differs between host niches, and so the ability to respond to and utilize distinct carbohydrates is crucial for pneumococcal fitness in vivo. The S. pneumoniae genome encodes 21 phosphotransferase systems (PTSs) and up to 8 ATP binding cassette (ABC) transporters for the import of carbohydrates (9,14), accounting for roughly 30% of all transport systems. Previous studies have shown that several of these carbohydrate transporters, present in both the core and accessory genome, impact pneumococcal virulence. For example, a sucrose PTS and ABC transporter system of serotype 4 pneumococci have been shown to play roles in murine colonization and pneumonia, respectively (15), Raffinose Metabolism Determines Pneumococcal Virulence ® while transporters for carbohydrates such as glucose, galactose, and mannose were shown to impact invasive pneumococcal disease (16)(17)(18).
The present study further underscores the critical role played by differential carbohydrate metabolism in pneumococcal pathogenesis. It demonstrates that reduced capacity to utilize raffinose does not simply reduce pneumococcal virulence, but rather changes the nature of disease caused. In multiple serotypes/ST lineages, ear isolates had defective growth in CDMϩRaf and reduced expression of raffinose pathway genes relative to their serotype/ST-matched blood isolates. Exchange of rafR alleles between ear and blood isolates of serotype 14 ST15 reversed these in vitro phenotypes. Moreover, rafR exchange caused blood isolates to now cause otitis media and meningitis rather than pneumonia, after intranasal challenge, and conversely cause ear isolates to now target the lungs. This striking switch in in vitro and in vivo behaviors was attributable to a single, nonconservative SNP (D249G) in RafR, identifying this residue as one of critical functional importance. Significantly, the region of RafR from amino acids 226 to 268 comprises a conserved signature sequence for the AraC/XylS family of transcriptional regulators (11).
Interestingly, in spite of exhibiting similarly distinct in vitro and in vivo phenotypes, the serotype 3 ST180 blood and ear isolates did not share the SNP in rafR, but rather had an SNP in rafK, which encodes the ATPase required for raffinose uptake via the ABC transport system encoded by rafEFG. RafK-mediated uptake of raffinose has previously been shown to be essential for induction of the raf operons in S. pneumoniae D39 (10). Attempts to construct rafK exchange mutants in this lineage (analogous to the serotype 14 ST15 rafR exchange mutants) were not successful. However, rafK deletion mutants of both 180/2 and 180/15 were obtained. Whereas the wild-type ear isolate 180/2 exhibited reduced growth in CDMϩRaf and expression of aga, rafG, and rafK relative to the wild-type blood isolate 180/15, both rafK deletion mutants were unable to grow in CDMϩRaf at all, and expression of any of the raf operon transcripts was undetectable. Clearly, the RafK allele carried by 180/2 retains partial function. The I227T SNP that distinguishes the RafK alleles of 180/2 and 180/15 is located in the conserved regulatory domain motif 1. This domain is believed to be involved in the interaction between RafK and the enzyme dihydrolipoamide dehydrogenase (DLDH), which has been shown to modulate raffinose uptake and raf operon expression in S. pneumoniae D39 (10). In the murine model, both rafK deletion mutants exhibited reduced bacterial loads in multiple host niches relative to their respective wild-type strains, consistent with previous reports (10,12).
Our findings provide an interesting example of convergent evolution, whereby pneumococci belonging to two unrelated serotypes/lineages exhibit SNPs in separate genes, each affecting raffinose uptake and utilization, which in turn correlate with distinct pathogenic profiles in both mice and humans (the latter by inference from the clinical isolation site). In S. pneumoniae D39, induction of expression of the raf operon gene aga required the presence of raffinose; reduced but nevertheless significant aga expression also occurred in a rafR knockout mutant (11). Thus, raf operon expression in pneumococci can be impacted either by defects in raffinose import (e.g., due to a defective RafK), such that insufficient exogenous raffinose (if present) is internalized to induce raf expression, or by functional defects in the transcriptional activator RafR, such that baseline levels of expression induced by the presence of raffinose are not further upregulated. The raf operons are part of the core genome of S. pneumoniae, and BLASTX analysis of available genomes shows that there is between 1% and 3% deduced amino acid sequence variation within any of the raf genes. Thus, SNPs are widespread, but it is not known which (if any) of these other SNPs impact the capacity to import or utilize raffinose or the virulence phenotype.
Notwithstanding the results presented above, the precise mechanism whereby differential raffinose uptake/utilization determines the virulence phenotype is uncertain. Raffinose is a plant-derived trisaccharide present in many staple foods, particularly beans and soy (19,20). Although humans are unable to metabolize it, dietary raffinose is known to be absorbed by the intestinal epithelium (21), raising the possibility of at least small amounts being present on mucosal surfaces. As part of the present study, we confirmed that expression of aga, rafG, and rafK was not detectable by qRT-PCR when pneumococci are grown in vitro in media lacking raffinose. However, expression of all three genes was detected in RNA extracts of mouse lung tissue 6 h after intranasal challenge with either of the serotype 14 ST15 blood or ear isolates (results not shown). Since raffinose is the only known inducer of the raf operon in S. pneumoniae, this finding is strongly indicative of the presence of bioavailable raffinose in the murine lung. A potential complicating factor is that RafEFG is reported to be also capable of importing stachyose (14), while RafK has been reported to also energize uptake of sialic acid and maltotetraose via unrelated transporters (12). Thus, the SNPs observed in the present study could have pleiotropic effects. However, no differences in metabolism of these sugars between the serotype 14 ST15 blood and ear isolates were observed using phenotypic microarray analysis, and the serotype 3 ST180 strains were unable to grow in CDM with stachyose or sialic acid as the sole carbon source. Moreover, there was no significant difference in the growth rates of the ST180 blood and ear isolates when grown in CDM with maltotetraose (results not presented). A particularly intriguing finding of the present study was that lower raffinose uptake/utilization by the ear isolates provided an advantage over blood isolates in the ear compartment. Interestingly, exogenous raffinose has recently been shown to promote biofilm formation by Streptococcus mutans by promoting aggregation of extracellular DNA into the biofilm matrix. Biofilm formation was unaffected by deletion of the ␣-galactosidase gene agaL, indicating that the effect was unrelated to metabolism of any internalized raffinose (22). Thus, it is conceivable that the reduced capacity of S. pneumoniae ear isolates to assimilate (and thereby deplete) raffinose from the middle ear mucosa may similarly promote pneumococcal biofilm formation in that niche, leading to otitis media. Further studies are in progress in our laboratory to elucidate the precise molecular mechanism whereby fine-tuning of levels of expression of raffinose uptake and utilization genes can have such a profound impact on pathogenic profiles of clinical isolates of S. pneumoniae.

MATERIALS AND METHODS
Bacterial strains and growth conditions. The S. pneumoniae strains used in this study are listed in Table 3. Cells were routinely grown in casein-based, semisynthetic liquid medium (CϩY) (23) or serum broth (SB) as required. Growth assays were performed using a chemically medium (CDM) comprising RPMI 1640 medium (Sigma), supplemented with amino acids, vitamins, choline, and catalase as described previously (24), with either 0.5% glucose or 0.5% raffinose. Bacteria were plated on Columbia agar supplemented with 5% (vol/vol) horse blood (BA) with or without gentamicin (40 g/ml), kanamycin (500 g/ml), or streptomycin (150 g/ml) (as required) and incubated at 37°C in 5% CO 2 overnight. For gene expression analyses, strains were grown in CDMϩGlc medium to an OD 600 of 0.2, before being incubated in CDMϩRaf for 30 min.
Genome sequencing. S. pneumoniae strains were grown to mid-exponential phase in Todd-Hewitt broth supplemented with 1% yeast extract. Genomic DNA (gDNA) was extracted using the Qiagen genomic DNA buffer set with 100/g Genomic Tips according to the manufacturer's instructions, except mutanolysin (20 U) and sodium deoxycholate (0.1%) were included to aid cell lysis. The gDNA was sequenced at the Ramaciotti Centre for Genomics (University of New South Wales, Sydney, Australia) on an Illumina MiSeq (250-bp paired-end reads), as well as a PacBio RSII instrument using one SMRT cell per strain, a 20-kb insert library, and the P6 polymerase and C4 sequencing chemistry.
Bioinformatic analyses. The Artemis Comparison Tool was used to compare genomes (25). MiSeq reads of 4559 and 947 and 180/15 and 180/2 were mapped to the assembled reference genome of the opposing strain with BOWTIE2 version 2.2.6 (26). Variant calling was then performed using SAMTools version 0.1.18 (27), and variants were mapped to coding sequences of the reference strain using BEDTools version 2.25.0 (28). Single nucleotide polymorphisms (SNPs) and insertions/deletions (indels) were filtered for those with scores of 100 or greater. Artemis was used to visualize SNPs and indels (29). Sanger sequencing was performed to confirm the SNPs in rafR and rafK (Australian Genome Research Facility, Adelaide).
Phenotypic microarrays. Carbon phenotype microarray analysis was performed on the serotype 14 ST 15 strains, using the PM microplates PM1 and PM2A (Biolog, Inc.), which tested for the catabolism of 190 different carbon sources. Each well of the microarrays contained a different carbon source. Briefly, cells were suspended in the provided buffer (as per the manufacturer's instructions) to an A 590 of 0.37. One hundred microliters of this suspension was added to the wells, and the A 590 was measured after 17 h of incubation at 37°C. Catabolism was measured through the reduction of a colorless tetrazolium dye by NADH, produced during catabolic activity. Absorbance values above 0.65 after subtraction of that for the zero carbon source blank were considered positive.
Growth assays. Each tested strain was grown in CDM supplemented with either 0.5% Glc (CDMϩGlc), 0.5% Raf (CDMϩRaf), or no sugar (CDM) and then incubated at 37°C for 12 h in 96-well flat bottom plates (Costar). The OD 600 was measured every 15 min using a SPECTRAmax M2 spectrophotometer (Millennium Science). All experiments were conducted in triplicate and repeated at least two times.
qRT-PCR. Differences in levels of gene expression were assayed by one-step relative quantitative real-time RT-PCR (qRT-PCR) in a Roche LC480 real-time cycler essentially as described previously (30). The specific primers used for the various genes are listed in Table 4 and were used at a final concentration of 200 nM per reaction. As an internal control, primers specific for 16S rRNA were employed. Amplification data were analyzed using the comparative critical threshold cycle (2 ϪΔΔCT ) method (31).
Mutagenesis. The rafR gene swap between serotype 14 ST15 4559 and 947 strains, to produce 4559 947rafR and 947 4559rafR , was achieved via allelic exchange mutagenesis utilizing the Janus cassette, as described previously (32,33). This involved a three-step process in which endogenous rpsL (which confers Serotype 23F ST81 Ear This study streptomycin sensitivity) was first replaced with the streptomycin-resistant rpsL1 allele by direct transformation of the blood and ear isolates. The Janus cassette (comprising a kanamycin resistance marker and a dominant counterselectable rpsL ϩ marker) was then used to replace the native rafR gene by direct transformation with a linear PCR product comprising the Janus cassette flanked by sequences 5= and 3= to rafR (selecting on kanamycin). In the final step, the Janus cassette in Kan r /Strep s transformants is replaced by transformation with the alternative rafR allele and flanking sequences, counterselecting on streptomycin (loss of the Janus cassette reinstates the Strep r phenotype). Gene swap constructs were confirmed by Sanger DNA sequencing (AGRF, Adelaide). The rafK gene was also deleted from serotype 3 ST180/2 and ST180/15 by direct transformation with a linear DNA fragment comprising an erythromycin resistance cassette flanked by sequences 5= and 3= to rafK generated by overlap PCR, essentially as previously described (34). The primers used are listed in Table 4. Mutant constructs were confirmed by PCR. Animal studies. Animal experiments were approved by the University of Adelaide Animal Ethics Committee. Groups of outbred 6-week-old female Swiss (CD-1) mice were anesthetized by intraperitoneal injection of pentobarbital sodium (Nembutal; Rhone-Merieux) and challenged intranasally (i.n.) with 50 l of bacterial suspension containing approximately 1 ϫ 10 8 CFU in SB (7). The challenge dose was confirmed retrospectively by serial dilution and plating on BA. Mice were euthanized by CO 2 asphyxiation at 24 h, and then tissue samples (lungs, nasopharynx, brain, ear, and blood) were harvested and pneumococci enumerated in tissue homogenates as described previously via serial dilution and plating on plates containing BA plus gentamicin (35).