Pneumococcal galactose catabolism is controlled by multiple regulators acting on pyruvate formate lyase

Catabolism of galactose by Streptococcus pneumoniae alters the microbe’s metabolism from homolactic to mixed acid fermentation, and this shift is linked to the microbe’s virulence. However, the genetic basis of this switch is unknown. Pyruvate formate lyase (PFL) is a crucial enzyme for mixed acid fermentation. Functional PFL requires the activities of two enzymes: pyruvate formate lyase activating enzyme (coded by pflA) and pyruvate formate lyase (coded by pflB). To understand the genetic basis of mixed acid fermentation, transcriptional regulation of pflA and pflB was studied. By microarray analysis of ΔpflB, differential regulation of several transcriptional regulators were identified, and CcpA, and GlnR’s role in active PFL synthesis was studied in detail as these regulators directly interact with the putative promoters of both pflA and pflB, their mutation attenuated pneumococcal growth, and their expression was induced on host-derived sugars, indicating that these regulators have a role in sugar metabolism, and multiple regulators are involved in active PFL synthesis. We also found that the influence of each regulator on pflA and pflB expression was distinct in terms of activation and repression, and environmental condition. These results show that active PFL synthesis is finely tuned, and feed-back inhibition and activation are involved.


Results
Transcriptome analysis of ΔpflB. To identify transcriptional regulators of pflB, the transcriptional profile of Δ pflB was obtained during growth on galactose relative to the wild type D39 strain. This analysis revealed differential expression of 113 genes in the mutant relative to the wild type ( Table 1). The microarray results were validated for selected genes by quantitative real time PCR, and a similar trend of expression was obtained ( Table 2). Out of these 113 genes, 62 were down-regulated and 51 were up-regulated in Δ pflB. The notable gene classes with differential expression were consistent with PFL's role in galactose metabolism, pyruvate dissimilation, and catabolic and anabolic reactions, and included those coding for galactose hydrolysis and sugar uptake (n = 27), genes implicated in cell shape (n = 6), and fatty acid biosynthesis and acetate dissimilation (n = 13). Interestingly, the expression of loci containing Leloir pathway genes (n = 7) decreased in the mutant, whereas tagatose pathway genes were not affected, implying that the two galactose catabolic pathways are governed by distinct regulatory processes.
The expression of seven genes annotated as transcriptional regulators was also either significantly up-or down-regulated in the mutant relative to the wild type (Table 1). These were, as annotated in the D39 genome (http://www.ncbi.nlm.nih.gov), MarR (SPD_0379), GlnR (SPD_0447), GntR (SPD_1524), SPD_1594, PlcR (SPD_1745), CcpA (SPD_1797), and NmlR (SPD_1637). These regulators are known to be involved in diverse metabolic functions in bacteria, from the control of sugar metabolism (CcpA), fatty acid synthesis (MarR), control of glutamine metabolism (GlnR), and virulence gene expression (PlcR), to sensing environmental and nutritional cues (GntR) 12 . Differential expression of the seven transcriptional regulators, including CcpA, led to the hypothesis that some of these regulators mediate the expression of pflA or pflB. We could not test our hypothesis for SPD_1594 because, despite repeated attempts, this gene could not be mutated, implying that it is likely to be essential 13 . CcpA, GlnR, and GntR interact with the putative promoters of pflA and pflB. The direct interaction of transcriptional regulators with the putative promoters (P) of pflA (PpflA) and pflB (PpflB), was determined by electrophoretic mobility shift assay (EMSA). For this, recombinant CcpA, PlcR, NmlR, GntR, GlnR, and MarR were purified ( Supplementary Fig. S1), and their interactions with PpflA and PpflB were studied. The results showed that, CcpA and GlnR bound to both PpflA and PpflB. Interaction of transcriptional regulators with their target DNA was concentration-dependent and specific, as the regulators could not bind to a gyrB-specific probe ( Fig. 1a-d). It should be also noted that CcpA binding to PpflA and PpflB occurred without the addition of HPr[Ser-P], which is reported to be essential for CcpA-DNA interaction in other lactic acid bacteria 14,15 . GntR could also form complexes with PpflB ( Supplementary Fig. S2) but its inducibility with different sugars, and its involvement in regulation of pflA and pflB could not be established with subsequent reporter assays. In addition, the other transcriptional regulators did not interact either with PpflA or PpflB (data not shown).
PFL activity leads to the generation of up to 35 mM formate when the pneumococcus is propagated on 55 mM galactose 2 . Hence, we tested whether sodium formate would alter the binding affinity of CcpA, and GlnR for PpflB. For this, the recombinant proteins were preincubated with 10 mM sodium formate before addition of the DNA probe. The results showed that formate enhanced the binding of CcpA to PpflB such that with sodium formate Kd was 0.35 ± 0.09 μ M, and without it was 1.27 ± 0.1 μ M (p < 0.01) ( Fig. 2 and Supplementary Fig. S3). The effect of formate was specific because the addition of 10 mM Tris-HCl did not have any effect on binding (data not shown). On the other hand, formate reduced GlnR (Kd: 0.79 ± 0.1 μ M) affinity for the pflB probe relative to without formate (Kd: 0.31 ± 0.09 μ M) (p < 0.01) (Fig. 2).
CcpA and GlnR bind to promoters containing cre sites. After establishing the interaction of CcpA and GlnR with PpflA and PpflB, we set out to determine the binding site of these regulators. Initially, we mapped the putative binding sites on PpflA and PpflB ( Supplementary Fig. S4), and identified multiple sites resembling a cre consensus sequence, the known binding site for CcpA from Bacillus subtilis (TGWAARCGYTWNCW, where N is any base, W is A or T, R is A or G, Y is C or T) 16,17 . At each putative promoter site, cre1 had the highest level of homology to the consensus sequence, 12 out of 14 nucleotides being similar. Therefore, EMSA was performed to determine whether cre1 had any role in binding of CcpA, and GlnR. For this, a promoter probe that excluded the cre1 sequence was produced and designated as PpflA(cre1 − ) and PpflB(cre1 − ), which was similar in size, approximately 100 bp, to the PpflA and PpflB.
EMSA analysis showed that CcpA and GlnR could not bind to PpflB(cre1 − ), indicating that the 14 nucleotide cre sequence located in PpflB contains the binding site for these regulators (Fig. 3a). Conversely, the results show that CcpA and GlnR bind to both PpflA and PpflA(cre1 − ) (Fig. 3b), However, they had higher binding affinities for PpflA than for PpflA(cre1 − ), which has an additional putative cre-like sequence. Host-derived sugars induce the expression of transcriptional regulators. The inducibility of the regulators by different sugars was determined using transcriptional reporter assays in anaerobically grown cultures. For this, the putative promoter region of each transcriptional regulator was fused to a promoterless lacZ in the wild type D39 background, and β -galactosidase activity in resulting reporter strains, PccpA::lacZ-wt, and PglnR::lacZ-wt, was determined in the presence of glucose, galactose, mannose, or N-acetyl glucosamine ( Table 3). All sugars induced the transcriptional regulator-driven β -galactosidase activity relative to expression in the absence of sugar (p < 0.01), but the highest inducer was found to be galactose. In addition, the responsiveness of the regulators to 10 mM sodium formate, which does not affect pneumococcal growth, was also determined. For this, the reporter strains were grown in the presence of glucose because formate production was found to be below the detection limit on glucose 2,18 , hence endogenous formate would not interfere with the assay. The results showed that the addition of sodium formate, but not Tris-HCl, induced β -galactosidase activity in PccpA::lacZ-wt by 2.4-fold, but decreased the activity in PglnR::lacZ-wt by 4.3-fold compared to the activity on glucose alone (Table 3) (p < 0.01). These data indicate that ccpA and glnR expression are responsive to formate, and that the absence of formate must be partly responsible for the transcriptional profile of Δ pflB. CcpA, and GlnR have different roles in expression of pflA and pflB. To test the regulatory roles of CcpA, and GlnR on pflA and pflB, PpflA::lacZ and PpflB::lacZ fusions were introduced into both wild type D39, and the mutants Δ ccpA, and Δ glnR. The resulting strains PpflA::lacZ-wt, PpflA::lacZ-Δ ccpA, PpflA::lacZ-Δ glnR, PpflB::lacZ-wt, PpflB::lacZ-Δ ccpA, and PpflB::lacZ-Δ glnR were tested during anaerobic growth in CDM supplemented with 55 mM of galactose. The results show that β -galactosidase activity in PpflA::lacZ-Δ ccpA and PpflA::lacZ-Δ glnR increased by 2.8-and 2.2-fold, respectively, compared to the activity in PpflA::lacZ-wt ( Fig. 4a) (p < 0.0001), suggesting that CcpA and GlnR repress the transcription of pflA. On the other hand, β -galactosidase activity in PpflB::lacZ-Δ glnR increased by 7.2-fold, whereas in PpflB::lacZ-Δ ccpA it decreased by 28.7-fold compared to the wild type (p < 0.0001) (Fig. 4b), indicating that GlnR repress, and that CcpA activates pflB on galactose. On the other hand, on glucose both CcpA and GlnR were found to repress both pflA and pflB,     To rule out the possibility that the expression data was influenced by lacZ fusion, pflA, and pflB expression was also determined by real time quantitative reverse transcriptase PCR (qRT-PCR) in Δ ccpA and Δ glnR grown on galactose relative to wild type. The results obtained with qRT-PCR were consistent with those of the transcriptional reporter assays, ruling out any interference by lacZ (Supplementary Table S1). Next, we investigated whether CcpA and GlnR could play a regulatory role in each other's expression. This hypothesis was tested by use of EMSA and LacZ reporter assays. EMSA showed that both CcpA and GlnR could bind to PccpA, indicating that ccpA regulates its own transcription, and that GlnR is involved in ccpA expression (Fig. 5). This binding was specific as CcpA and GlnR could not bind to the putative promoter region of gyrB (lane 2 in Fig. 1a-d). The interaction between CcpA and GlnR was further investigated by LacZ reporter assays. It was shown that β -galactosidase activity in PccpA::lacZ-Δ ccpA and PccpA::lacZ-Δ glnR decreased both on glucose, 2.5-and 10.3-fold, and on galactose, 30.7-and 1.4-fold, respectively, compared to the activity of PccpA::lacZ-wt (Table 4) (p < 0.001). This result shows that both on glucose and galactose CcpA increases its own expression, and GlnR activates ccpA, particularly on glucose.
CcpA's role in regulation of glnR was also determined. CcpA could not bind to PglnR, which was consistent with the lack of a discernible cre sequence in the promoter region. The reporter assay results show that β -galactosidase activity in PglnR::lacZ-Δ ccpA decreased by 1.7-and 3.8-fold on glucose and galactose, respectively, compared to PglnR-lacZ-wt. This result suggests that CcpA increases glnR expression in the presence of both glucose and galactose, and GlnR does not effect its own expression, since β -galactosidase activity in PglnR::lacZ-Δ glnR on glucose (89.7 MU ± 4.4, n = 3) and galactose (572.3 MU ± 4.2, n = 3) was similar to that of PglnR::lacZ-wt (with glucose 93.4 MU ± 1.8, and with galactose: 565.0 MU ± 7.1, n = 3 for both) (Table 4). Moreover, the results showed that GlnR activates ccpA expression both on glucose and galactose (Table 4). Growth and end product analysis. To further investigate CcpA, and GlnR's role in pneumococcal sugar metabolism, the growth profiles and fermentative end-products of Δ ccpA, and Δ glnR were determined. Moreover, to evaluate the added impact of CcpA and GlnR mutation on pneumococcal metabolism, a Δ ccpAΔ glnR mutant was also constructed and tested. On glucose, Δ ccpA, Δ glnR, and Δ ccpAΔ glnR had significantly reduced growth rates (0.36 h −1 ± 0.01, 0.47 h −1 ± 0.01, and 0.31 h −1 ± 0.01 respectively, n = 3) compared to wild  type D39 (0.67 h −1 ± 0.01) (p < 0.0001 for all strains). Despite a low rate of growth, Δ ccpA growth yield (max change in OD500), was similar to that of wild type, while Δ glnR had a lower yield than the wild type. In addition, Δ ccpAΔ glnR had a lower growth rate and yield than each single mutant (p < 0.01) (Fig. 6a). The reason for growth attenuation of strains with a glnR mutation, Δ glnR and Δ ccpAΔ glnR, is not known but it may be linked to GlnR's involvement in glucose transport and metabolism. On galactose, the growth rate and yield of all strains decreased approximately to half or less than that on glucose (Fig. 6b). Δ ccpA, Δ glnR, and Δ ccpAΔ glnR had significantly reduced growth rates (0.24 h −1 ± 0.01, 0.26 h −1 ± 0.01, and 0.07 h −1 ± 0.01, respectively, n = 3) compared to the wild type strain (0.35 h −1 ± 0.01, n = 3) (p < 0.01). Δ glnR had an extended lag phase and started to grow after 11 h, which may indicate its involvement in sensing of galactose. Moreover, there was significant difference in the rate of growth between Δ ccpAΔ glnR and each single mutant (p < 0.001), and the growth yield (max change in OD500) of the double mutant (0.21 ± 0.02, n = 3) was significantly lower than that of the single mutants (0.79 ± 0.07 for Δ ccpA, 0.74 ± 0.06 for Δ glnR, n = 3) (p < 0.01).
In the presence of glucose the wild type, and Δ glnR strains displayed typical homolactic behaviour with lactate as the main fermentation product, and minor amounts of formate and acetate (Fig. 6c). Loss of CcpA in Δ ccpA and Δ ccpAΔ glnR caused a shift from homolactic to mixed-acid fermentation. In Δ ccpA, a 4.4-fold increase in the yield of formate and 23.5-fold in acetate was detected relative to the wild type. On galactose, the wild type, and Δ glnR displayed a mixed-acid fermentation pattern with lactate, formate and acetate as the end products. However, Δ ccpA and Δ ccpAΔ glnR had a homolactic product pattern with lactate production in Δ ccpA (29.1 mM ± 0.9, n = 3) being significantly higher than in the wild type (p < 0.0001), whereas formate and acetate production in these strains was significantly decreased compared to the wild type (p < 0.0001) (Fig. 6d). Therefore, the results showed that both transcriptional regulators are important for growth on both glucose and galactose, and CcpA has a major impact on the composition of the mixed acid profile. show the standard error of the mean for three individual measurements each with three replicates. Significant differences were seen comparing the growth rates, and the fermentative profile of mutant strains to the wild type D39 using ANOVA followed by Dunnett's multiple comparison test. **p < 0.01 and ****p < 0.0001. Virulence studies. In vitro assays indicated CcpA and GlnR are involved in pflA and pflB regulation, and they have a regulatory influence on each other's expression. Therefore, we determined the contribution of CcpA and GlnR to nasopharyngeal colonization, and virulence in mouse models of pneumococcal infection. The results show that the median survival time of mice infected intranasally with Δ ccpA, Δ glnR, and Δ ccpAΔ glnR (168 h ± 17.3, 168 h ± 13.0 and 168 h ± 0.0 respectively, n = 10) was significantly longer than the wild type infected group (35 h ± 13.2, n = 10) (p < 0.001 for all strains) (Fig. 7a). The introduction of intact copies of ccpA and glnR into Δ ccpA and Δ glnR, respectively, reconstituted the virulence of these strains with the median survival times of mice infected with Δ ccpAComp (37 h ± 2.3, n = 10) and Δ glnRComp (45 h ± 19.5, n = 10), being not significantly different from the wild type infected cohort (p > 0.05), eliminating the possibility of a polar effect of the mutations (Fig. 7a).

Discussion
This study provides new insights into how the important human pathogen S. pneumoniae regulates its galactose metabolism, which is known to have a major effect on pneumococcal survival in vivo 6 . Pyruvate formate lyase is a key enzyme for pyruvate dissimilation in S. pneumoniae. In addition to its role in pneumococcal energetics, the products of the PFL-catalysed reaction, formate, and acetyl CoA, play a key role in anabolic reactions. For example, while formate is essential for serine biosynthesis by hydroxymethylation of glycine, acetyl-CoA is important for fatty acid biosynthesis 2,19 . Until this study, very little was known about the transcriptional regulation of genes involved in active PFL synthesis.
Our results show that multiple regulators modulate the expression of pflA and pflB. While CcpA and GlnR are involved in regulation of both pflA and pflB, GntR could only bind to putative promoter region of pflB. The other regulators identified to be differentially expressed in Δ pflB could bind neither to PpflA nor to PpflB, presumably their differential expression in Δ pflB was due to indirect effect of pflB mutation. CcpA was found to repress pflA and pflB on glucose, and activate pflB and repress pflA expression on galactose. This opposing regulatory influence of CcpA on pflA and pflB could indicate that constitutive pflA expression sufficiently activates PFL synthesis. This is supported with the fact that the association rate between PFL-AE and PFL is low for biological interactions because of large conformational changes when these two enzymes interact. Hence a high level pflA expression would be wasteful because of a low association rate 20 . Moreover, intracellular calculations in E. coli revealed that PFL-AE is almost entirely in the PFL-AE/AdoMet PFL pyruvate complex, and its activation requires reduction by flavodoxin to initiate catalysis. Therefore, despite its constitutive expression, the existing PFL-AE can be promptly activated in this complex when needed. It was also noticed that induction level of PpflA is lower than PpflB on galactose (Fig. 4). The expression data are very likely reflecting the amount of PFL-AE required to activate PFL, because PFL is completely activated by 0.01 equivalents of PFL-AE in 100 minutes 21 . The opposing effects of transcriptional regulators on pflA and pflB must be an important pneumococcal adaptation mechanism in tissues with different sugar profiles and oxygen concentrations. For example, in an oxygenated environment active PFL is not critical as it is sensitive to oxygen. On the other hand, in tissues with high galactose content but limited oxygen concentration, the pneumococcus requires mechanisms to induce pflB expression.
In other Gram positive bacteria CcpA binding requires HPr-[Ser-P], while the pneumococcal CcpA bound to a cre sequence in the absence of HPr-[Ser-P] 22 . However, we found that 10 mM sodium formate enhanced CcpA affinity for PpflB, consistent with its induction by sodium formate in the transcriptional reporter assays. Although the inducibility of ccpA expression by different carbon compounds had been shown previously 11,23 , this is the first demonstration of the responsiveness of this master regulator to formate. It is not clear how formate enhances CcpA binding and stimulates the expression, but given that formate is a potent reducing agent, with a redox potential of − 420 mV 24 , it is very likely due to its effect on the conformation or multimerization of protein structure. Addition of formate increased ccpA-, and decreased glnR transcription, indicating the responsiveness of S. pneumoniae to formate. Formate's role as a transcriptional signal also shows that the absence of formate is partly responsible for the transcriptional profile of Δ pflB. In addition, we previously demonstrated that fatty acid composition of S. pneumoniae changes in Δ pflB, which cannot produce formate, also indicating the importance Each point is the mean of data from ten mice. (c) Pneumococcal strains defective in ccpA and glnR were less able to colonize nasopharynx. Mice were infected approximately with 1 × 10 5 CFU pneumococci. At day 0 and day 7, five mice were culled, and bacterial CFU/mg were determined by serial dilutions of nasopharyngeal homogenates. Each column represents the mean of data from five mice. Error bars show the standard error of the mean. *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001. of formate on transcription and protein synthesis 2 . Formate was reported to have a major influence on bacterial gene expression, and acts as a diffusible signal to induce Salmonella enterica serovar Typhimurium invasion 25 . In Staphylococcus aureus, it was suggested that formate production is necessary for formyl-tetrahydrofolate synthesis, which is required for protein and purine synthesis 18 . Further work is required to understand the mechanism by which formate impacts on the pneumococcus. We did use galactose similarly to formate to determine its effect on the affinity of transcriptional regulators to PpflA and PpflB. However, no difference could be detected.
The cre1 within PpflB has been identified to be the binding site for GlnR. GlnR bound two cre sites in PpflA but here it had a higher binding affinity for cre1 than cre2. GlnR binding affinity for pflB was shown to decrease in the presence of 10 mM sodium formate, consistent with a decrease in PglnR driven β -galactosidase activity with sodium formate. This is the first demonstration of interaction of proteins, other than CcpA and HPr, with cre. Given that cre sites are plentiful in Gram-positive bacteria, it is plausible that some of these targets are regulated by multiple regulators, either independently or in concert with CcpA.
Interestingly, although CcpA plays a role in glnR regulation, as determined by the LacZ reporter assay, PglnR does not have an identifiable cre-like sequence. Therefore, it is possible that CcpA-mediated glnR regulation requires the involvement of another, unknown, protein(s). It is known that several genes with no cre or un-functional cre have been reported to be indirectly regulated by CcpA [26][27][28] . Our results showed that both CcpA and GlnR have a significant role in pneumococcal virulence and colonization. The results obtained in this study on CcpA's impact on pneumococcal colonization and virulence are in line with previous reports 11,23 . However, our in vivo results for GlnR differ from the study of Hendriksen et al., who could not attribute a role for GlnR in pneumococcal virulence and colonization 29 . One reason for this discrepancy can be linked to the difference in dose-preparation, for example, while in this study we did not passage the bacterial inoculum through peritoneum before infection, Hendriksen's study used passaged bacteria. In vivo passaging before infection can introduce mutations to a bacterial population due to strong within-host selective pressure 30 .
The GntR family of transcriptional regulators is a group of transcriptional repressors distributed among diverse bacterial groups, and mediate bacterial response to nutritional and/or environmental signals 31 . In addition to CcpA and GlnR, GntR could also bind to PpflB. However, GntR could not bind to PccpA and PpflA, even though both have a cre-like sequence. Unlike cre1 in PpflB (TGAAATCGGTTACT, conserved residues are underlined), which is categorised as a high-affinity cre box because it contains all four strongly conserved residues, G2, C7, G8, and C13, the cre sequences in PpflA, and PccpA contain only 2 or 3, respectively, conserved residues 16,17 . Therefore, it is likely that GntR only recognizes highly conserved cre boxes. PgntR-driven β -galactosidase activity was not induced by any of the tested sugar nor Δ gntR is attenuated in growth on galactose (data not shown). This suggests that in addition to sugar dependent regulation, sugar independent regulation of pflA and pflB may take place. On galactose, formate production by PFL activity decreases GlnR affinity for PpflB and GlnR increases ccpA transcription. The increased formate production increases CcpA affinity for PpflB, and CcpA either alone or after interacting with GlnR binds to PpflB and increases pflB transcription. However, CcpA also binds to PpflA and represses its expression to fine-tune the level of active PFL. Increased formate production then creates a positive feed back loop, and CcpA self-regulates its own expression. Glu: glucose, Gal: galactose, SF: sodium formate. The green arrow indicates increase in expression, whereas the red arrow is for decreased expression. *Posttranslational activation of PFL by PFL-AE has been previously reported 9,47 .
Based on the available data we propose the model for pflA and pflB regulation shown in Fig. 8. Growth on galactose increases PFL activity, which leads to increased production of formate, and also induces glnR expression. When the formate level reaches a certain threshold, GlnR repression over pflA and pflB expression is removed, and GlnR then binds to cre, up-regulating ccpA transcription. In addition, the increased formate production also increases ccpA expression as well as CcpA affinity for PpflB, leading to an increase in pflB transcription. However, CcpA and GlnR also bind to PpflA, and repress pflA expression to fine-tune the level of pflA transcript. On the other hand, on glucose, GlnR increases ccpA expression, and CcpA activates its own expression. This, in turn, results in repression of both pflA and pflB, and pyruvate is dissimilated mainly to lactate by lactate dehydrogenase activity 1,2 .
Galactose is found in high concentration on host glycoconjugates in the nasopharynx, where the pneumococcus resides, and the microbe is able to cleave and utilize galactose from these sources 8 . However, galactose is not efficiently catabolized by the pneumococcus, possibly due to the possession of low-affinity galactose importers 6 , and/or to a slow association rate between nascent PFL and PFL-AE for the formation of active PFL 32 . In addition, at the transcriptional level the expression of pflA and pflB is down-regulated by the regulators identified in this study, with the exception of CcpA, in the presence of galactose. Taken together, these data indicate that the pneumococcus has evolved to catabolize galactose at a lower rate than other sugars, such as glucose. It can be speculated that low galactose catabolic rate allows the microbe to maintain a stable relationship with its host. This can be linked to galactose catabolism and its role in pneumococcal virulence. For example, the pneumococcus produces more capsule, an essential virulence determinant, on galactose than it does on glucose 10 .
This study demonstrates that PFL synthesis is fine-tuned at the transcriptional level by multiple regulators and various environmental conditions. PFL activity is known to be crucial for galactose catabolism, which is linked to pneumococcal colonization and virulence. Detailed knowledge of regulation of pneumococcal sugar metabolism may allow identification of targets for new antiinfectives against this important pathogen.

Materials and Methods
DNA manipulations. Standard methods were used for chromosomal DNA isolation, restriction enzyme digestion, cloning, transformation, agarose gel electrophoresis, and sodium dodecyl sulphate polyacrylamide gel electrophoresis 33 . Plasmids were extracted using QIAprepSpin Miniprep Kit (Qiagen, UK), and PCR products were purified using the Wizard ® SV Gel and PCR Clean-Up System (Promega, UK).
Bacterial strains, plasmids and growth conditions. The strains and plasmids used and constructed in this study are listed in Supplementary Table S2. Pneumococcal strains were grown microaerobically at 37 °C as described previously 2 . For anaerobic growth, an anaerobic cabinet was used. Pneumococci were also grown in chemically defined medium with selected sugars (CDM) 1 . Growth was monitored by measuring the optical density at 500 nm (OD500). Growth rates (μ ) were calculated through linear regressions of the plots of ln(OD500) versus time during the exponential growth phase. When required, the pneumococcal growth medium was supplemented with spectinomycin (100 μ g/ml), tetracycline (15 μ g/ml), and kanamycin (250 μ g/ml).
Escherichia coli strains, used for cloning and protein expression, were grown in Luria Bertani (LB) broth with shaking at 37 °C or on LB agar plates. For E. coli cultures, ampicillin and kanamycin at 100-and 50 μ g/ml, respectively, were used when required.
Microarray analysis. S. pneumoniae D39 and its isogenic mutant strain Δ pflB, deficient in pyruvate formate lyase (PFL) activity, were grown anaerobically in CDM supplemented with 55 mM galactose as the main carbon source. The experiments were repeated with four biological replicates. The MicroPrep software package was used to obtain the microarray data from the slides. CyberT implementation of a variant of t-test (http://bioinformatics. biol.rug.nl/cybert/index.shtml) was performed and false discovery rates (FDRs) were calculated 27 . For differentially expressed genes, p < 0.001 and FDR < 0.05 were taken as standard. For the identification of differentially expressed genes a Bayesian p-value of < 0.001 and a fold change cut-off twofold were applied. All other procedures for the DNA microarray experiments and data analysis were performed as described before 34 . Gene splicing by overlap extension (gene SOEing). Allelic replacement mutagenesis was achieved by transformation with an in vitro mutagenized SOEing construct. To do this, spectinomycin (Spec R ) or kanamycin (Kan R ) resistance gene cassettes 35,36 were amplified using the primes Spec-F and Spec-R or Kan-F and Kan-R primers (Supplementary Table S3). The primers LF-SOEX-F/LF-SOEX-R and RF-SOEX-F/RF-SOEX-R (where X indicates the gene code) were used to amplify the left and right flanking regions of each target gene, generating PCR products of approximately 600 bp in length. The amplified antibiotic resistance cassette and the PCR products flanking the target gene were then fused using LF-SOEX-F and RF-SOEX-R primers. The fused product was gel-purified after electrophoresis (Qiagen) and transformed into S. pneumoniae 37 . The transformants were selected on blood agar base supplemented with 5% (v/v) defibrinated horse blood containing the appropriate antibiotic and the mutations were confirmed by PCR and sequencing. The double mutant was constructed by transformation of amplicons representing the mutated region of glnR into the ccpA mutant.
Genetic complementation of mutants. To eliminate the possibility of polar effects, the mutant strains were genetically complemented using plasmid pCEP 36 . The wild type copy of mutated genes with their upstream sequence were amplified with X-Comp-F and X-Comp-R primers (where X indicates the gene code) (Supplementary Table S3). The amplicons were digested with NcoI-BamHI or SphI-BamHI and were ligated into similarly digested pCEP. A sample of ligation mixture was transferred into One Shot ® TOP10 competent E. coli cells (Invitrogen, UK). The transformants were selected on kanamycin-containing LB agar plates, and successful cloning was confirmed by PCR using Comp-Seq-F and Comp-Seq-R primers. The recombinant plasmid was purified, and a portion was transformed into the appropriate mutant strains. The complemented strains were designated as Δ ccpAComp, and Δ glnRComp.
Quantification of extracellular metabolites. Samples (2 ml) of anaerobic cultures growing in CDM containing 1% (w/v) glucose or galactose were collected at late exponential phase during growth, centrifuged (6700 × g, 10 min, 4 °C), filtered (Millex-GN 0.22 μ m filters) and the supernatant solutions were stored at − 20 °C until analysis with commercial kits for lactate, formate, and acetate detection (Megazyme, Ireland). The amount of organic acids produced by different strains was normalized to 1 × 10 8 cells. The spent culture supernates were obtained at late exponential phase and this corresponded to the following OD500s for the different strains grown on glucose and galactose, respectively: for D39 1.9 and 1.1; for Δ ccpA 1.8 and 0.7; for Δ glnR 1.2 and 0.6; for Δ ccpAΔ glnR 0.7 and 0.2.
Purification of recombinant proteins. The genes of interest were amplified and cloned into the hexa histidine-tagged, ampicillin resistant plasmid pLEICS-01 (PROTEX, University of Leicester) using the primers (Indicated with C-F/R in Supplementary Table S3)  Electrophoretic mobility shift assay (EMSA). In silico analysis, using the bacterial promoter prediction tool (BPROM) 39 and the Motif-based sequence analysis tools (MEME), was used to predict the presence of regulatory elements in the putative promoter regions of target genes 40 . Based on in silico analysis, primers (indicated by EMSA-F/R in Supplementary Table S3) were designed to amplify 95-100 bp DNA fragment representing putative promoter sites upstream of each gene.
EMSA was performed according to the protocol described previously 41 using Molecular Probes fluorescence-based EMSA kit (Invitrogen). Briefly, 5X FY binding buffer (20 mM Tris-HCl pH 7.5, 30 mM KCl, 1 mM DTT, 1 mM EDTA pH 8.0 and 10% v/v glycerol) was prepared to incubate the promoter probes and proteins. The binding reaction was set up by mixing a constant amount of target promoter probes (~30 ng), and increasing amounts (0.1-0.5 μ M) of purified and dialysed His-tagged proteins. The binding reaction was incubated at room temperature for 20 min in a total volume of 20 μ l, and then analysed on an 8% w/v non-denaturing polyacrylamide gel. After electrophoresis, gels were stained with SYBR ® Green EMSA gel stain (Invitrogen) and visualized using a Typhoon Trio + scanner (GE Healthcare Life Sciences) with a 526 nm short-pass wavelength filter.

Construction of lacZ-fusions.
Chromosomal transcriptional lacZ-fusions to the target promoters were constructed via double crossover in the bgaA gene by using the integration plasmid pPP1 42 . As the extracellular BgaA enzyme is responsible for removal of galactosides from complex carbohydrates, inactivation of the bgaA gene will not change the growth of S. pneumoniae in media and will not compromise regulatory studies 42 . The putative promoter regions were amplified using the primers modified to incorporate SphI and BamHI sites (indicated with Fusion-F/R in Supplementary Table S3). The amplicons were digested with SphI and BamHI and were ligated into similarly digested pPP1. All plasmid constructs were confirmed by sequencing. β-galactosidase assays. β -galactosidase activity was measured as described before 43 using cells grown anaerobically in CDM at 37 °C supplemented with 55 mM of the selected sugars, or 10 mM sodium formate, and harvested in the mid-exponential phase of growth.
Total RNA purification and quantitative RT-PCR. The extraction of RNA was done by the Trizol method using mid-exponential phase cultures, as described previously 44 . Before use the RNA was treated with DNase using a TURBO DNA-free ™ Kit (Ambion, UK), and subsequently purified with an RNeasy Mini Kit (Qiagen).
First strand cDNA was synthesized using approximately 1 μ g of DNase-treated total RNA, immediately after isolation, random hexamers and 200 U of SuperScript III reverse transcriptase (Invitrogen) at 42 °C for 55 min as described previously 2 . Three independent RNA preparations were used for qRT-PCR analysis. The transcription level of specific genes was normalized to gyrB transcription, which was amplified in parallel using the primers in Supplementary Table S4 (indicated with RT-F/R tags). cDNA samples obtained from individual biological samples were analysed in triplicates. The results were analysed by the comparative threshold cycle C T method 45 . Differences in expression of twofold or greater relative to control were considered as significant.
In vivo virulence studies. To evaluate the virulence of pneumococcal strains, nine to ten-week-old female MF1 outbred mice (Charles River, UK) were lightly anesthetized with 3% v/v isoflurane and oxygen mixture, and an inoculum of 50 μ l containing approximately 2 × 10 6 CFU in PBS was administered dropwise into the nostrils. Disease signs in infected animals were monitored regularly 2,46 . When the mice become lethargic, they were culled by cervical dislocation, "survival time" was defined as the time to reach the lethargic state. Mice that were alive 7 days after infection were deemed to have survived the infection. To monitor the development of bacteraemia, approximately 20 μ l of venous blood was obtained from each mouse at predetermined time points after infection, and viable counts were determined, as described above. Colonization experiments were done principally as described previously, using 5 × 10 5 CFU of S. pneumoniae in 10 μ l PBS, given intranasally 8 . Survival data were analysed by the Mann-Whitney test. Colonization and bacteraemia data were analysed by analysis of variance followed by Tukey's multiple comparisons test. Statistical significance was considered to be a p-value of < 0.05.

Ethics statement.
Mouse experiments at the University of Leicester were performed under appropriate project (permit no. 60/4327) and personal (permit no. 80/10279) licenses according to the United Kingdom Home Office guidelines under the Animals Scientific Procedures Act 1986, and the University of Leicester ethics committee approval. The protocol was approved by both the U.K. Home Office and the University of Leicester ethics committee. Where indicated, the procedures were carried out under anesthetic with isoflurone. Animals were housed in individually ventilated cages in a controlled environment, and were frequently monitored after infection to minimize suffering. Every effort was made to minimize suffering and in bacterial infection experiments mice were humanely culled if they became lethargic.