Regulation of Glutamine and Glutamate Metabolism by GlnR and GlnA in Streptococcus pneumoniae*

Several genes involved in nitrogen metabolism are known to contribute to the virulence of pathogenic bacteria. Here, we studied the function of the nitrogen regulatory protein GlnR in the Gram-positive human pathogen Streptococcus pneumoniae. We demonstrate that GlnR mediates transcriptional repression of genes involved in glutamine synthesis and uptake (glnA and glnPQ), glutamate synthesis (gdhA), and the gene encoding the pentose phosphate pathway enzyme Zwf, which forms an operon with glnPQ. Moreover, the expression of gdhA is also repressed by the pleiotropic regulator CodY. The GlnR-dependent regulation occurs through a conserved operator sequence and is responsive to the concentration of glutamate, glutamine, and ammonium in the growth medium. By means of in vitro binding studies and transcriptional analyses, we show that the regulatory function of GlnR is dependent on GlnA. Mutants of glnA and glnP displayed significantly reduced adhesion to Detroit 562 human pharyngeal epithelial cells, suggesting a role for these genes in the colonization of the host by S. pneumoniae. Thus, our results provide a thorough insight into the regulation of glutamine and glutamate metabolism of S. pneumoniae mediated by both GlnR and GlnA.

Regulation of nitrogen metabolism in bacteria is closely connected with the intracellular levels of glutamine and glutamate, the main nitrogen donors in the cell. Glutamine is formed from glutamate and ammonium by glutamine synthetase (GlnA), which is a major way for the cell to assimilate ammonium. Glutamate can be formed either by glutamate dehydrogenase from 2-oxoglutarate and ammonium or by glutamate synthase, which converts glutamine and 2-oxoglutarate into two molecules of glutamate.
Several studies indicate that nitrogen metabolism, especially glutamine metabolism, is important for the virulence of various bacterial pathogens (1)(2)(3). Signature-tagged mutagenesis screens suggest that genes involved in glutamine metabolism, glnQ and glnA, are likely to play a role in the virulence of S. pneumoniae as well (4 -6). However, so far, glutamine metabolism and the way in which it is regulated have not been studied in this human pathogen.
In the well characterized Gram-positive bacterium Bacillus subtilis, regulation of nitrogen metabolism is carried out mainly by CodY, GlnR, and TnrA (7). The latter two are members of the MerR family of regulators, and both recognize the same operator sequence: 5Ј-TGTNAN 7 TNACA-3Ј. TnrA functions during growth on a poor nitrogen source (e.g. solely glutamate) when it activates or represses expression of various genes involved in nitrogen metabolism (8 -12). GlnR represses its own operon glnRA (13), the ureABC operon (encoding urease) (14,15), and tnrA (7) in the presence of a good nitrogen source, like glutamine.
Genetic experiments have shown that genes regulated by GlnR and TnrA are constitutively expressed in a mutant of glnA (8, 16 -18). An explanation for this observation came with the discovery that in vitro DNA binding by TnrA is blocked by feedback-inhibited GlnA (19). Although it has been suggested that GlnA also controls the DNA binding activity of GlnR, this has never been shown. In fact, B. subtilis GlnR has a high affinity for DNA on its own (13).
B. subtilis CodY functions as a repressor of genes involved in nitrogen metabolism (20) but also of carbon and energy metabolism (21), motility (22), and competence development (23). In the lactic acid bacterium Lactococcus lactis, CodY represses genes of the proteolytic system and several amino acid transport and metabolism genes, among others gltA and gltD, which are involved in glutamate biosynthesis (24 -26).
Analysis of the S. pneumoniae R6 (27) and TIGR4 (28) genomes revealed that they contain genes encoding orthologs of GlnR and CodY but not of TnrA. Furthermore, Streptococcus pneumoniae contains a putative ortholog of glnA, several predicted glutamine uptake systems, and a predicted biosynthetic glutamate dehydrogenase (27,28). In contrast to B. subtilis and L. lactis, a gene encoding glutamate synthase is not present. This suggests that S. pneumoniae has various ways to secure sufficient cellular glutamine levels, either by uptake from the environment or by de novo synthesis.
In this study, we report on the important role of GlnR and GlnA in the regulation of glutamine and glutamate metabolism in S. pneumoniae and present indications for a role of GlnR targets in pneumococcal virulence.

EXPERIMENTAL PROCEDURES
Strains, Media, and Growth Conditions-Strains used in this study are listed in Table 1 and were stored in 10% glycerol at Ϫ80°C. S. pneumoniae was grown essentially as described (29): on plates in a flame-pot, giving an elevated CO 2 concentration, or in liquid medium as standing cultures. L. lactis and Escherichia coli were grown as described previously (29). Kanamycin and tetracycline were used in concentrations of 500 and 2.5 g/ml for S. pneumoniae, respectively. Ampicillin was used in a concentration of 100 g/ml for E. coli. Chemically defined medium with a final pH of 6.4 was composed as described (29), except that sodium citrate was used in the buffer instead of ammonium citrate and that glutamine was omitted from the amino acid mixture. Glutamine, glutamate, and ammonium were added as specified under "Results." Induction of gene expression with nisin was performed as described, using a stock solution of nisaplin, containing 20 mg/ml nisin (29).
DNA Isolation and Manipulation-Primers used in this study are listed in Table 2. Primers were based on the genome sequence of strain S. pneumoniae R6 (27). Unless otherwise indicated, chromosomal DNA of S. pneumoniae D39 was used as a template for PCR amplification. All DNA manipulations were done as described (29).
Construction of glnR, glnA, glnRA, zwf, and glnP Mutants of S. pneumoniae-The glnR-stop mutant (TK102) was constructed using plasmid pORI280 as follows. Primer glnR-stop 1 with two point mutations, leading to two premature stop codons at codon positions 20 and 21 in the glnR reading frame, was used in combination with primer glnR_R6 -1 to PCR-amplify a fragment comprising the upstream part and the beginning of glnR. A second PCR product, comprising the rest of the glnR gene and part of the downstream sequence was produced with primers glnR-stop 2 and glnR-3. These PCR products were complementary by 20 bp covering the position of the stop codons and were used as a template in a PCR with primers glnR_R6-1 and glnR-3. The resulting product was cloned as an XbaI, BglII fragment in pORI280, giving plasmid pTK20. pTK20 was used to introduce the mutations into the chromosome of S. pneumoniae D39 as described (29), giving strain TK102. The mutations led to the disappearance of a HincII site, on the basis of which the proper mutant could be identified. The mutations were further verified by DNA sequencing.
The glnA deletion strain (TK103) was generated by allelic replacement mutagenesis, removing 1300 bp of the glnA open reading frame (ORF). 3 A PCR fragment, generated with primers Spec_pORI38-Fp and Spec_pORI38-Rp on the spectinomycin resistance gene from pORI38, was cloned into the HindIII site of pORI28 in the same orientation as the erythromycin gene on this vector, yielding pORI28spec1. Next, the 3Ј-flanking region of glnA, amplified with primer pair glnA_R6 -3/glnA_R6 -4 (886 bp), was cloned into the NcoI/BglII sites of pORI28spec1, giving pTK18. pTK18 was cut with NdeI/AatII, and a PCR fragment generated with primers glnA_R6 -1/glnA_R6 -2 (808 bp), which was digested with the same enzymes, was ligated to it. This ligation mixture was used to generate a PCR product with primers glnA_R6 -1 and glnA_R6 -4, which was transformed to S. pneumoniae D39. Spectinomycin-resistant clones were examined for the presence of the glnA deletion by PCR and Southern blotting. The zwf deletion mutant (TK107), removing 1416 bp of the zwf ORF, was constructed in a similar way as the glnA mutant, using primers G6PDH-4/G6PDH-5 (660 bp) and G6PDH-6/G6PDH-7 (610 bp).
To construct the glnRA mutant (TK104), the upstream part of glnR, amplified with primer pair glnR_R6 -1/glnR_R6 -2 (883 bp), was cloned into the XbaI/BamHI sites of pORI28spec1, giving pTK22. pTK22 was used together with pTK18, which contains the glnA_R6 -3/glnA_R6 -4 PCR product cloned into the NcoI/BglII sites of pORI28spec1, as a template in a PCR with primers glnR_R6 -1 and glnA_R6 -4. In this way, a PCR product was obtained containing the spectinomycin resistance gene flanked by the upstream and downstream sequence of glnRA. The resulting PCR product was transformed to D39. The deletion was confirmed by PCR and Southern blotting.
L. lactis 108 was used as the cloning host for plasmid pTK19. All other constructs were made in E. coli EC1000.
To construct the glnP deletion mutant (TK106), removing 2080 bp of the glnP ORF, a PCR fragment, generated with primer pair Ery-rev/Ery-for on the erythromycin resistance gene from pORI28, was fused to the flanking regions of glnP, which were PCR-amplified with primer pairs glnPKO-1/gln-PKO-2 (628 bp) and glnPKO-3/glnPKO-4 (610 bp) by means of overlap extension PCR (30). The resulting PCR product was transformed to S. pneumoniae D39, and clones were checked for the presence of the mutation by PCR. In the same way, a deletion mutant of gdhA, removing 1311 bp of the gdhA ORF, was constructed in D39 using primer pairs gdhAKO-1/ gdhAKO-2 (479 bp) and gdhAKO-3/gdhAKO-4 (498 bp).
Construction of capsuleless derivatives of D39 and its glnA, glnR, glnRA, and glnP mutants was done as described (31), using primers PE21 and FI4. Mutants were checked by PCR and appearance. In addition, they adhere several orders of magnitude better than the encapsulated mutants.
The lacZ fusions to the gdhA promoter were constructed in pPP2 with primer pair PgdhA-2/PgdhA-4, giving a PCR prod- uct comprising the full-length promoter (PgdhA-1) and primer pair PgdhA-3/PgdhA-4, resulting in a PCR product without the predicted GlnR operator (PgdhA-2), using E. coli EC1000 as the cloning host. The constructs were introduced in S. pneumoniae strains D39 and TK102.
DNAse I footprinting was done essentially as described (34). 150,000 cpm of [␥-32 P]ATP-labeled PCR products of the glnR and glnP promoters, made with primer pairs R6_PglnR_FP/ R6_glnR-7 (244 bp) and R6_glnP-1/R6_glnP-GFP1 (235 bp), respectively, were used as probes in 40 l of binding buffer (EMSA) containing 5 mM glutamine and purified H 6 -GlnR and H 6 -GlnA as specified under "Results." Enzyme Assays-Cell-free extracts, used for the determination of glutamine synthetase (GlnA), biosynthetic glutamate dehydrogenase (GdhA), and glucose-6-phosphate dehydrogenase (Zwf) activity, were made from 1 or 2 ml of cells harvested in exponential phase of growth, which were resuspended in 250 l of 20 mM Tris (pH 7.5) and disrupted by shaking for 1 min with 400 mg of glass beads (75-150 m) in a Biospec Mini-Bead-Beater-8. After the removal of cell debris by centrifugation, cellfree extracts were used in a concentration of one-tenth to onetwentieth of the total volume of the assay mixture. The A 600 at which cells were harvested was used to calculate the enzyme activity per A 600 unit. S.D. values were calculated from at least three independent replicate experiments. GlnA activity (transferase reaction) was determined as described (35). Biosynthetic glutamine synthetase activity of purified H 6 -GlnA (ATP ϩ L-glutamate 3 ADP ϩ P i ϩ L-glutamine) was determined as described (36). Biosynthetic GdhA activity (2-oxoglutarate ϩ NADPH ϩ NH 4 ϩ 3 glutamate ϩ NADP ϩ ) was determined at 30°C in a reaction mixture containing 50 mM Tris-HCl (pH 8.5), 35 mM 2-oxoglutarate, 80 mM NH 4 Cl, and 0.3 mM 2Ј-NADPH by monitoring the decrease in absorption at 340 nm (A 340 ) caused by oxidation of NADPH. Catabolic GdhA activity was measured at 30°C in a reaction mixture containing 50 mM Tris-HCl (pH 7.5), 70 mM L-glutamate, and 0.5 mM NADP ϩ , by monitoring the increase in A 340 . Activity of Zwf (glucose 6-phosphate ϩ NADP ϩ 3 gluconolactone-6-phosphate ϩ NADPH) was measured in buffer containing 1 mM NADP ϩ , 2 mM glucose 6-phosphate, 10 mM MgCl 2 , 1 mM dithiothreitol, and 20 mM Tris-HCl (pH 8.0) by monitoring the increase in A 340 at 30°C. The activity of ␤-galactosidase was determined as described (37), except that cells were permeabilized with a final concentration of 0.06 mg/ml cetyltrimethyl ammonium bromide. During growth for ␤-galactosidase assays, no antibiotic selection was imposed.
Transcriptome Analysis Using S. pneumoniae DNA Microarrays-DNA microarray experiments were performed essentially as described (38). RNA was isolated from 50 ml of cells grown to midexponential phase of growth (A 600 ϭ 0.3) in M17 containing 0.25% glucose (GM17) containing 0.5 mg/ml glutamine (GM17Gln). Cells were harvested by centrifugation for 1 min at 10,000 rpm at room temperature. Cell pellets were immediately frozen in liquid nitrogen and stored at Ϫ80°C. Pellets were resuspended in 500 l of 10 mM Tris-HCl, 1 mM EDTA (T 10 E 1 ), pH 8.0, after which 50 l of 10% SDS, 500 l of phenol/chloroform, 500 mg of glass beads (75-150 m), and 175 l of Macaloid suspension (Bentone) were added.
Synthesis of cDNA and indirect Cy-3/Cy-5-dCTPs labeling of 15-20 g of total RNA was performed with the CyScribe Post labeling kit (Amersham Biosciences) according to the supplier's instructions. Hybridization (16 h at 45°C) of labeled cDNA was performed in Ambion Slidehyb #1 hybridization buffer (Ambion Europe) on superamine glass slides (Array-It; SMMBC), containing technical replicates of amplicons representing 2,087 ORFs of S. pneumoniae TIGR4 and 184 ORFs unique for S. pneumoniae R6. DNA microarrays were produced essentially as described (38,39). Amplicon sequences are available on the World Wide Web at molgen.biol.rug.nl/ publication/glnRAspn_data. Slides were scanned using a GeneTac LS IV confocal laser scanner (Genomics Solutions).
DNA Microarray Data Analysis-Slide images were analyzed using ArrayPro 4.5 (Media Cybernetics Inc., Silver Spring, MD). Processing and normalization (LOWESS spotpin-based) of slides was done with the in-house developed MicroPrep software as described (38,40). DNA microarray data were obtained from three independent biological replicates hybridized to three glass slides, of which one was a dye swap. Expression ratios of mutant strain over the wild-type strain were calculated from the measurements of at least five spots. Differential expression tests were performed on expression ratios with a local copy of the Cyber-T implementation of a variant of the t test. False discovery rates were calculated as described (38). A gene was considered differentially expressed when p was Ͻ0.001 and false discovery rate was Ͻ0.05 and when at least five measurements were available. The DNA microarray data are available on the World Wide Web at molgen.biol.rug.nl/publication/ glnRAspn_data. In addition, they have been deposited in the Gene Expression Omnibus GEO (available on the World Wide Web at www.ncbi. nlm.nih.gov/projects/geo/query/acc. cgi?accϭGSE5088).
Reverse Transcription-PCR-RNA isolation and cDNA synthesis were performed as described above, except that aminoallyl-dUTP was replaced by dTTP during cDNA synthesis. To confirm the absence of DNA contamination, reactions were also carried out without reverse transcriptase. 100 ng of cDNA was used for each PCR, and after 20 amplification cycles (30 s, 95°C; 30 s, 52°C;, 60 s, 72°C) with primers G6PDH4 and G6PDH5, reactions were analyzed on 1% agarose gels.
Adhesion Assays-Adhesion of pneumococci to epithelial cells was studied essentially as described previously (41). Briefly, the human pharyngeal cell line Detroit 562 (ATCC CCL-138) was cultured in RPMI 1640 without phenol red (Invitrogen) containing 1 mM sodium pyruvate and 10% fetal calf serum. Aliquots of bacteria, grown to midexponential phase in GM17 and stored until use at Ϫ80°C, were thawed rapidly, har-  vested by centrifugation, and resuspended to 1 ϫ 10 7 colonyforming units/ml in RPMI 1640 medium without phenol red containing 1% fetal calf serum. Monolayers of Detroit 562 in 24-well tissue culture plates were washed twice with 1 ml of phosphate-buffered saline, after which 1 ml of bacterial suspension was allowed to adhere for 2 h at 37°C in a 5% CO 2 atmosphere. Subsequently, nonadherent bacteria were removed by three washes with 1 ml of phosphate-buffered saline, and the epithelial cells were detached by treatment with 200 l of 25% trypsin, 1 mM EDTA in phosphate-buffered saline. Detroit 562 cells were lysed by the addition of 800 l of ice-cold 0.025% Triton X-100 in phosphate-buffered saline, and appropriate dilutions were plated on blood agar plates to count the number of adherent bacteria. This colony-forming unit count was first corrected mathematically to account for small differences in count in the initial inoculum, after which data were normalized so that the adhesion of the wild-type strain TK136 was expressed as 100%. Wild type and mutant pneumococci grew comparably in RMPI medium without Detroit 562 cells. All experiments were performed in triplicate and repeated at least three times. Significant differences between wild type and mutants were calculated by the Mann-Whitney t test ( p Ͻ 0.05).

Prediction of Putative GlnR Operators in S. pneumoniae-B. subtilis
GlnR is known to repress genes that contain two copies of the inverted repeat 5Ј-TGTNAN 7 TNACA-3Ј in their promoters (13,15). This repeat is also present in the promoter regions of the Lactobacillus rhamnosus (42) and Bacillus cereus (43) glnRA operons. Since the GlnR binding box seems so well conserved between species, we screened the entire genome of S. peumoniae R6 for the presence of putative GlnR operators  using Genome2D (44). Predicted operators with the highest similarity to the B. subtilis consensus sequence were present in the promoter regions of glnR (spr0443); glnP (spr1120), encoding a glutamine ABC transporter substrate-binding protein; gdhA (spr1181) encoding a NADP(H)-specific glutamate dehydrogenase; and arcA (spr1955, spr1956), encoding arginine deiminase (Fig. 1). Re-searching the R6 genome with a weight matrix built from these putative operators did not reveal additional putative GlnR operators.
The Regulon of GlnR and GlnA in S. pneumoniae-To investigate the role of GlnR in S. pneumoniae in more detail, we constructed a glnR mutant in strain D39. To preserve the glnRA operon structure, two consecutive stop codons were introduced in the beginning of the glnR ORF, specifying amino acids in the middle of the predicted helix-turn-helix DNA-binding motif.
DNA microarray analyses were performed of S. pneumoniae D39 wild-type and its isogenic glnR mutant grown in the nitro-gen-rich medium GM17, supplemented with 0.5 mg/ml glutamine (GM17Gln). This amino acid is assumed to be a co-repressor of GlnR in B. subtilis (18), and we expected it to also induce repression of GlnR targets in S. pneumoniae. The operons/genes that were most highly up-regulated in the glnR mutant were glnRA, glnPQ, and gdhA (Table 3), all of which have a GlnR operator in their promoter regions (Fig. 1). The arcA gene, which also contains a putative GlnR operator in its promoter, was only weakly up-regulated. Remarkably, also zwf, encoding the key enzyme glucose-6-phosphate dehydrogenase of the pentose phosphate pathway, was up-regulated. This gene lies downstream of and in the same orientation as glnPQ.
To investigate the influence of GlnA on the expression of the identified GlnR targets, a comparison of the transcriptomes of wild-type D39 with its isogenic glnA mutant, grown in GM17Gln, was performed. This showed that glnR, glnP, glnQ, and gdhA were, like in the glnR mutant, also up-regulated in the glnA mutant (Table 3), indicating that GlnA is necessary for the functioning of GlnR.
Besides the genes mentioned above, SP2063, encoding a predicted LysM domain-containing protein, not directly involved in glutamine metabolism, was 2-fold down-regulated in both the glnR and the glnA mutant. Interestingly, two degenerate GlnR boxes (5Ј-TGTGACAGAGACCTAACA-3Ј and 5Ј-TGTTATTAG-CGTCAACA-3Ј) are present in the promoter region of this gene.
In both the glnR and glnA mutant, genes predicted to encode proteins involved in pyrimidine metabolism (SP1275, SP1276, SP1277, and SP0954) were moderately up-regulated, which seems logical, since glutamine is a precursor of pyrimidine. However, since no GlnR operator could be identified upstream of any of these genes, the up-regulation is likely to be an indirect effect caused by altered intracellular glutamine/glutamate levels. Furthermore, a number of other genes of various functions were moderately up-regulated in either the glnR or the glnA mutant.
Chromosomal transcriptional lacZ fusions were used to confirm that in both the glnR and the glnA mutant, expression of glnA, glnP, zwf, and gdhA was derepressed (Figs. 2A, 3,  A and B, and 4A). In addition, enzymatic activity assays showed that the observed effects on transcription corresponded with altered activities of GlnA, GdhA, and Zwf in the glnR and glnA mutants (Figs. 2B, 3C, and 4B). No strong effect of the glnR or the glnA mutation on the expression of a chromosomal ParcA-lacZ transcriptional fusion was observed in a range of different media (data not shown).
Nitrogen Source-dependent Regulation of glnRA and glnPQ by GlnR-To address the role of GlnA substrates in the observed GlnR-dependent regulation in S. pneumoniae, expression of the chromosomal glnA-lacZ and glnP-lacZ fusions was measured in a chemically defined medium (CDM), to which glutamine, glutamate, and ammonium were added in varying amounts ( Fig. 2A). In CDM with only glutamate, glnA expression was similar in the wild type and the glnR mutant. In contrast, glutamine led to repression of glnA expression in the wild type already at a relatively low concentration (0.25 mM). A higher glutamine concentration did not lead to stronger repression in the wild type. However, when besides glutamine ammonium was included, glnA expression could be further repressed. The combination of glutamate and ammonium also gave rise to repression of glnA expression in the wild-type. None of the above combinations caused repression of glnA expression in the glnR mutant. GlnA enzymatic activity is regulated in the same way (Fig. 2B).
Regulation of glnP in response to glutamate, glutamine, and ammonium is very similar to that of glnA (Fig. 3A). Derepression of glnP expression is also seen in the glnA mutant, albeit to a somewhat lower extent than in the glnR mutant. This could indicate that, in the absence of GlnA, GlnR is still able to exert a weak repressive effect on the expression of glnP.
Regulation of zwf by GlnR and GlnA Occurs via the glnP Promoter-The DNA microarray results showed that zwf, which lies downstream of glnPQ, is also up-regulated in the glnR and glnA mutants. Reverse transcription-PCR demonstrated that zwf lies on the same transcript as glnPQ and thus forms an operon with these genes (data not shown). To examine whether zwf is only transcribed from PglnP or also from a possible promoter in the glnQ-zwf intergenic region, the latter was cloned upstream of lacZ in pORI13 and introduced in the RepA ϩ strain D39repA (29) and its glnR mutant. Promoter activity was present in this fragment (ϳ5 Miller units), which was not dependent on GlnR (data not shown). Regulation of zwf by GlnR and GlnA was similar to but weaker than regulation of glnPQ (Fig. 3, B and C), which can be explained by the presence of the second promoter upstream of zwf. Thus, expression of zwf initiates from two promoters, a GlnRdependent promoter upstream of glnP and a second promoter in the glnQ-zwf intergenic region.
Regulation of gdhA by GlnR and CodY-Despite the fully conserved GlnR operator in the gdhA promoter, regulation of gdhA by GlnR and GlnA in GM17Gln and CDM was weaker than regulation of glnPQ and glnRA (Fig. 4, A and  B). However, expression of an ectopic lacZ fusion to the fulllength gdhA promoter (PgdhA-1) and to a truncated version without the GlnR box (PgdhA-2; see also Fig. 1) showed that deletion of the predicted GlnR operator abolished the GlnR-dependent repression of PgdhA (Fig. 4C), demonstrating that the predicted GlnR operator in the gdhA promoter is functional.
Interestingly, in the S. pneumoniae R6 genome, putative CodY operator sequences are present in the promoter regions of, among others, gdhA and zwf (24). To examine whether CodY regulates these genes in S. pneumoniae, the activity of the corresponding enzymes was measured in a codY deletion mutant. No effect of the codY deletion was seen on the activity of Zwf in GM17Gln (data not shown), but activity of GdhA was strongly increased in the codY mutant (Fig. 4D). In a glnRcodY double mutant, GdhA activity was even higher than in the codY mutant, indicating that GlnR and CodY independently repress gdhA in S. pneumoniae (Fig. 4D).
glnPQ Encodes the Main Glutamine/Glutamate Transport Operon in S. pneumoniae-Of the genes encoding predicted glutamine transporters in the R6 genome (27), glnPQ were the only ones found to be regulated by GlnR. To investigate the role of glnPQ in glutamine metabolism, a deletion of glnP, encoding the permease component of the GlnPQ ABC transporter, was constructed in D39. Whereas S. pneumoniae D39 is able to grow in CDM containing glutamine (Fig. 5A) or glutamate (29), but not in their absence, the glnP mutant was not able to grow in CDM with either glutamine (Fig. 5A) or glutamate (data not shown). This phenotype could be complemented by in trans expression of glnPQ from a nisin-inducible promoter (Fig. 5A). Moreover, the addition of the dipeptide Gly-Gln to the CDM also rescued growth of the glnP mutant (Fig. 5A), whereas this was not the case with the dipeptide Phe-Gly (data not shown). These data indicate that glnPQ encode the only actively expressed glutamine and glutamate uptake system in S. pneumoniae under these conditions.
GlnA activity was increased in the glnP mutant in GM17 (Fig.  5B), although to a lower extent as in the glnR mutant (Fig. 2B). To investigate whether regulation of glnA is affected in the glnP mutant, the effect of casitone as the nitrogen source in the medium was tested. Casitone, an enzymatic (pancreatic) digest of casein, consists of casein-derived peptides and contains no free glutamine and only a very low level of free glutamate (available on the World Wide Web at www. bd.com/ds/technicalCenter/misc/ bionutrientmanual.pdf). Growth of the glnP mutant in CDM containing 2% casitone as the only nitrogen source was the same as that of the wild-type strain (data not shown), indicating that the uptake of peptides can bypass the inability to take up glutamine and glutamate. The addition of casitone, like glutamine, to GM17 resulted in an ϳ2-fold reduced GlnA activity in the wildtype strain, but not in the glnP mutant (Fig. 5B). Thus, besides GlnA, also GlnPQ appear to be necessary for efficient repression by GlnR.
Binding of GlnR to PglnP and PglnR Is GlnA-dependent-The transcriptional data presented above show that the activity of GlnR is dependent on GlnA. Therefore, we investigated whether GlnA is required for the binding of GlnR to the glnR and glnP promoters in vitro. For this we used a His-tagged variant of each protein (H 6 -GlnA and H 6 -GlnR). Nisin-induced expression of H 6 -GlnA in strain TK100 restored growth in CDM with glutamate and no glutamine (data not shown). Nisin-induced expression of H 6 -GlnR in the glnR mutant (TK105) led to 5-fold lower GlnA activity in CDM with 5 mM glutamine and 10 mM ammonium (data not shown). In CDM with 10 mM glutamate, the level of repression was still 4-fold, although the effect was weaker at low nisin concentration. With wild-type GlnR, the repressive effect was also 5-fold in CDM with 5 mM glutamine and 10 mM ammonium, but only 2-fold in CDM with 10 mM glutamate. These data indicate that H 6 -GlnA and H 6 -GlnR are functional, although the latter seems to respond in a less sensitive way to its assumed co-repressor glutamine.  glnR and glnP promoters. A and B, EMSA of binding of GlnR to the glnR (PglnR-1) and glnP (PglnP-1) promoter regions and to truncated glnR (PglnR-2) and glnP (PglnP-2) promoters, lacking their respective GlnR operators. H 6 -GlnR and H 6 -GlnA were added as indicated above the panels in concentrations of 400 nM (monomer*) and 1.5 M (monomer*), respectively, and glutamine (Gln), glutamate (Glu), ammonium (Am), and ATP were all added at 5 mM. The higher band seen for the free PglnR probe is probably single-stranded DNA (34). C, DNAse I footprint of PglnR and PglnP in the absence (Ϫ) and presence of 400 nM H 6 -GlnR monomer* and 1.5 M H 6 -GlnA monomer*. Glutamine was present in all reactions in a concentration of 5 mM. Protected regions are indicated by black bars. The arrows indicate hypersensitive sites. Numbers on the left indicate bp positions relative to the translational starts of glnP and glnR. AG, Maxam-Gilbert sequence ladder. *, GlnR is probably active as a dimer (13,15,48), and GlnA is probably active as a dodecamer (49). H 6 -GlnR alone at the concentration shown (400 nM) did not bind to PglnR and PglnP (Fig. 6, A and B), also not in the presence of glutamine. However, in the presence of H 6 -GlnA, binding of H 6 -GlnR to PglnR and PglnP was observed (Fig. 6, A and  B). At a 5-10 times higher H 6 -GlnR concentration (2-4 M), a shifted band at the same position could be observed in the absence of H 6 -GlnA (data not shown), which corresponds with the transcriptional data, showing some weak repression by GlnR in the glnA mutant (Table 3, Fig. 3A). No binding of H 6 -GlnR was seen with the controls, the promoters without their GlnR operators (Fig. 6, A and B).
DNase I footprinting showed that, in the presence of H 6 -GlnA, H 6 -GlnR specifically reduces DNAse I sensitivity of the predicted GlnR operator in PglnP (Figs. 1 and 6C). Remarkably, the protected region in PglnR only partially overlapped with the predicted GlnR operator (Figs. 1 and 6C), suggesting that GlnR binds in a different manner to this promoter than to PglnP.
In contrast to what would be expected from the expression data, GlnA-dependent binding of GlnR to PglnP and PglnR was only weakly stimulated by the addition of glutamine (Fig. 6, A  and B). This could be explained by the observation mentioned above, that H 6 -GlnR seems to be less sensitive to glutamine than the native protein. The addition of the other GlnA substrates glutamate, ammonium, ATP, and AMP alone or in combination did not alter the observed GlnA dependence of the GlnR-DNA interaction at PglnP (Fig. 6A). Thus, although GlnA is required for the binding of GlnR to the GlnR operators in the glnP and glnR promoters, this effect was not modulated by GlnA substrates.
GlnA and GlnP Contribute to Adhesion to Pharyngeal Epithelial Cells-The crucial first step of pneumococcal virulence is the colonization of the nasopharynx. Therefore, we tested the ability of glnR, glnA, glnRA, and glnP mutants to adhere to the human pharyngeal epithelial cell line Detroit 562. These mutants were created in the capsuleless background strain D39 ⌬cps, since unencapsulated strains tend to show higher levels of adhesion (45) (data not shown). The glnP, glnA, and glnRA mutants displayed a significantly decreased adhesion to the pharyngeal epithelial cells compared with the capsuleless wildtype strain (Fig. 7). Thus, both GlnP and GlnA could play a role in colonization of the nasopharynx. However, since most pneumococcal isolates are encapsulated, the actual contribution of these two proteins to virulence remains uncertain.

DISCUSSION
In this study, we characterized the regulation of glutamine and glutamate metabolism mediated by GlnR and GlnA in the human pathogen S. pneumoniae. Previously, GlnR-dependent regulation of nitrogen metabolism has been thoroughly studied in B. subtilis. The only target that B. subtilis GlnR shares with GlnR from S. pneumoniae is glnRA (13). In addition, B. subtilis GlnR is a repressor of the ureABC operon (7,14) and tnrA (7), which are absent in S. pneumoniae (27). Another difference is that whereas GlnR is a repressor of glnPQ and gdhA in S. pneumoniae, in B. subtilis the catabolic glutamate dehydrogenase gene rocG is regulated by CcpA, RocR, and AhrC (46), and the glnQH glutamine transport operon is activated by TnrA (11).
We found that also expression of zwf, encoding a putative glucose-6-phosphate dehydrogenase, is regulated by GlnR. This enzyme catalyzes the first reaction in the pentose phosphate pathway, which provides the cell with NADPH and ribose 5-phosphate, a building block of nucleic acids. Since glutamine is a precursor for the synthesis of nucleotides as well, it might be advantageous for S. pneumoniae to coordinate zwf expression with glutamine metabolism.
Our data suggest that the regulation by S. pneumoniae GlnR depends on a conserved inverted repeat. The B. subtilis GlnR targets contain two copies of the same inverted repeat in their promoter regions (13,15). S. pneumoniae GlnR resembles B. subtilis TnrA in this respect, since TnrA activates or represses promoters containing only one copy of this repeat (11).
The distance between the GlnR operator and the Ϫ35 box in PglnR is 7 bp, and for PglnP it is 16 bp. GlnR boxes are also present at a distance of 5-7 bp from the Ϫ35 in the glnR promoters of the S. pneumoniae relatives Streptococcus pyogenes, Streptococcus agalactiae, and Streptococcus mutans. Moreover, GlnR operators are located in the glnP promoters of S. pyogenes and S. agalactiae, which, like in S. pneumoniae, have a 16-bp spacing with the Ϫ35 sequence. Since the spacing in the glnP promoters is 9 -10 bp longer than in the respective glnR promoters, regulation by GlnR via these operators might be helix side-dependent in these organisms.
The GlnR operator in PgdhA confers a less pronounced GlnRdependent effect than the GlnR operators in PglnP and PglnR, although the inverted repeat is perfectly conserved. The same accounts for the GlnR operator in ParcA. In PglnR and PglnP, there is a stretch of A nucleotides immediately upstream of the repeat and a stretch of T nucleotides between the two half-sites. These stretches might explain the more efficient transcriptional repression of PglnP and PglnR than of ParcA and PgdhA, since AT-rich stretches on these positions of B. subtilis PnrgAB enhance TnrA-dependent transcriptional activation (47).
Both GlnR and CodY function as a repressor of gdhA in S. pneumoniae, of which CodY seems to be the more important regulator. Furthermore, both regulators control gdhA transcription independently of each other, which is in agreement with the location of their operators, that for CodY lying upstream of the Ϫ35 and the GlnR operator downstream of the Ϫ10 in PgdhA. In B. subtilis, CodY controls the cellular nutritional and energy status (20,21). Although GdhA is obviously connected to glutamine metabolism, the observation that gdhA expression is, next to GlnR, also regulated by CodY in S. pneumoniae, might indicate that GdhA is an important control point of the cellular nutritional status in this bacterium.
We show that GlnR DNA binding is dependent on GlnA, in contrast to the situation in B. subtilis, where GlnR alone binds with high affinity to its target promoters in the absence of any effectors (13). Since a high concentration of GlnR alone led to a shifted band at the same position as in the presence of GlnA, it is unlikely that GlnR and GlnA bind as a complex to the DNA. It might be that GlnA induces a conformational change or multimerization of GlnR, which increases its DNA binding affinity. Next to GlnA, also GlnP seems important for activity of S. pneumoniae GlnR. Although GlnR and GlnA alone were sufficient for in vitro binding to the glnP and glnR promoters, it could be that in vivo both GlnPQ and GlnA are needed for optimal activity of GlnR.
Our results and previous STM screens (4 -6) implicate a role for both GlnP and GlnA in pneumococcal adhesion to human pharyngeal cells, which is a prerequisite to invade the host. Previously, GlnQ was shown to be required for adhesion of S. pyogenes to fibronectin and epithelial cells of the respiratory tract (2). However, it remains to be investigated whether the effect of GlnPQ on adhesion by S. pneumoniae and S. pyogenes is caused by a general effect of distorted glutamine metabolism (e.g. on the cell surface composition) or if GlnPQ are directly involved. We are currently analyzing glnR, glnA, glnP, and gdhA mutant strains in several in vivo mouse models to gain more insight into the role of glutamate and glutamine metabolism during infection by S. pneumoniae.