INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

Streptococcus pneumoniae is the major cause of community-acquired bacterial pneumonia, and it is also responsible for otitis media and meningitis (2). Capsular polysaccharides were the first virulence factors to be identified. The capsule is thought to protect the bacteria from the host immune system by preventing phagocytosis (17). Purified capsular extracts do not have an inflammatory or toxic effect (31, 32). Among the proteins considered to be virulence factors (17, 45) are pneumolysin (3, 7, 12), autolysin (4, 12, 56), hyaluronidase (5), pneumococcal surface protein A (PspA) (8), PsaA (6), neuraminidase (10), immunoglobulin A1 (IgA1) protease (46, 59), and pyruvate oxidase (55), although for some of them a role in virulence has not been demonstrated.

Recent advances in the field of bacterial pathogenesis have allowed a large-scale identification of new virulence genes in different bacterial species. The methods developed are based on the concept that specific gene products are required for each stage of an infection process and that their expression is often regulated by the different environmental conditions encountered in the host. Mahan et al. (38) developed a system called IVET (in vivo expression technology) aimed at identifying bacterial genes that were preferentially expressed in the host during infection and were poorly transcribed under laboratory conditions. IVET was originally developed for use with Salmonella typhimurium (38) and then applied to Vibrio cholerae (11) and Pseudomonas aeruginosa (58). Hensel et al. (28) expanded the concept of tagging originally developed by Walsh and Cepko (57) to monitor the fate of clonally related neocortical cells during brain development, developing a strategy to identify virulence genes by negative selection. This system, called STM (signature-tagged mutagenesis), exploits a pool of transposons in which each transposon is tagged with a unique DNA sequence so that the resulting insertion mutants are marked with a different DNA sequence. This permits the identification of bacteria recovered from hosts infected with a mixed population of mutants.

Tagged insertion mutants are combined into pools, which are used to infect the animals. At a defined time point, bacteria are recovered from the animals. Tag sequences are amplified from each pool by using a radioactive label before and after the infection. These two labeled tag probes are hybridized to filters containing spotted genomic DNA from all mutants of the corresponding pool. Mutants whose tags are positive for hybridization with the probe from the original pool and negative with the one from the recovered bacteria are considered to be virulence attenuated. This system was originally used to identify genes involved in virulence in S. typhimurium (28) and recently applied to Staphylococcus aureus and V. cholerae (15, 42).

S. pneumoniae has been studied for many years, yet its virulence mechanisms are not fully understood (17). Therefore, we have modified the original STM methodology to discover novel virulence factors in S. pneumoniae. Initially, we attempted to use for this purpose the encapsulated type 3 strain GP119, obtained by transformation from the nonencapsulated avirulent laboratory strain Rx1. Surprisingly, when GP119 was tested for virulence in a mouse septicemia model, it was found to be not virulent (unpublished observations); this result indicates that in factors in addition to the capsule are required for virulence. We then selected the encapsulated type 19F strain G54, a recent clinical isolate. This strain was chosen because of its high transformation rate (47) induced by using the 17-residue competence-stimulating peptide (26) and its virulence in a mouse pneumonia model, as assessed in our laboratory. The following two major changes were introduced in the original STM protocol (28).

(i) A library of tagged mutants was constructed by insertion-duplication mutagenesis using short random genomic DNA fragments inserted in a suicide plasmid vector bearing the molecular tag. The original transposon mutagenesis technique (28) was not applied to S. pneumoniae, since we did not expect transposons to insert randomly as shown for Streptococcus mutans (24) and Lactococcus lactis (49).

(ii) While in the original method (28) the filters corresponding to each pool had spots of genomic DNA from each mutant obtained by transferring bacterial colonies to the filters (colony hybridization), we used filters containing amplified tags from each mutant. This modification was necessary since in our hands colony hybridization generated a high background giving rise to false positives.

Using the modified STM technique, we identified 126 putative virulence genes from S. pneumoniae. Some of them corresponded to previously described pneumococcal virulence factors, while others showed homology to virulence genes found in other bacteria. In addition, we have identified genes not previously known to be involved in virulence.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Media. Escherichia coli strains were grown in Luria broth (LB; Difco). S. pneumoniae strains were grown in Todd-Hewitt broth (Difco), and tryptic soy broth (TSB; GIBCO) or agar plates. When needed, these media were supplemented with ampicillin (100 µg/ml), chloramphenicol (8 µg/ml), or 5% defibrinated sheep blood (Unipath; Oxoid).

Animals. Male BALB/c mice weighing 16 to 18 g (Charles River, Como, Italy) were used.

Competence factor 1. The 17-amino-acid peptide used to stimulate competence was synthesized by Neosystem Laboratoire (Strasbourg, France) according to the published sequence (26).

Construction of the tagged library. A pool of single-stranded 89-bp DNA molecules containing a central stretch of 40 random bp flanked by two constant sequences was generated by oligonucleotide synthesis. The constant regions contained HindIII and KpnI sites (Fig. 1). The sequence of the random 89-mer oligonucleotides used was 5'-CTAGGTACCTACAACCTCAAGCTTN₄₀AAGCTTGGTTAGAATGGGTACCATG-3'. The constant regions allowed the sequence tags to be amplified by PCR using the upper primer (5'-TACCTACAACCTCAAGCT-3') and the lower primer (5'-TACCCATTCTAACCAAGC-3'). The double-stranded library of tags was generated by extending the 18-mer primer P1 (5'-CATGGTACCCATTCTAAC-3') annealed to the random 89-mer template. Ten micrograms of P1 primer was mixed with 3 µg of 89-mer random template in 14 µl of annealing buffer (T7 sequencing kit; Pharmacia) and annealed as indicated by the manufacturer. T7 DNA polymerase (8 U) and 3 µl of 5 mM deoxynucleoside triphosphates (dNTPs) were added and incubated for 30 min at 37°C. The final population of double-stranded DNA molecules was digested with KpnI. The corresponding DNA band was recovered from a 7% polyacrylamide gel, subcloned into the KpnI site of pR326 vector (3 µg) (16) (Fig. 1), and used to transform E. coli DH5 by electroporation with a Bio-Rad apparatus according to the manufacturer's instructions. Ten electrotransformation preparations, each containing 100 µl of cells and approximately 300 ng of vector DNA, were pooled, and a small aliquot was plated to assess transformation efficiency. The remainder was added to 500 ml of LB with ampicillin and grown for 8 h with shaking. Tagged DNA plasmids (tagged library) were then extracted and stored at 20°C.

View larger version (14K): [in this window] [in a new window]   FIG. 1.   (A) Construction of a pool of pR326 plasmid DNA tags. A pool of double-stranded DNA tags was generated by extending with T7 DNA polymerase the 18-mer primer P1 annealed to random 89-mer templates (see Materials and Methods). Each tag is a variable sequence of 40 bp (N40), where N is A, C, G, or T, flanked by two constant arms of 20 bp which contain restriction sites for HindIII and KpnI. These arms allow PCR amplification with the upper primer (U.P.) and lower primer (L.P.) of each tag. Tags were digested with KpnI and ligated in the vector pR326, generating a pool of tagged plasmids. (B) Construction of a library of S. pneumoniae-tagged mutants by homologous recombination. S. pneumoniae chromosomal DNA was mechanically sheared; fragments ranging from 500 to 700 bp were size selected and ligated with the pool of pR326 plasmids DNA tags. The ligation mix was directly transformed in S. pneumoniae G54; homologous recombination events lead to insertion-duplication mutagenesis, with the plasmid carrying the cat gene and the tag inserted into the inactivated gene in the chromosome. Total genomic DNA was extracted from mutant clones, digested with a restriction endonuclease, and circularized by ligation. The genomic flanking region outside the cat gene was amplified with the inner set of primers cat0 and cat770; 4 ng of the product of the first PCR was reamplified with the nested pair of primers cat1 and cat720. The resulting single-PCR product was then sequenced.

Tagged genomic library construction. S. pneumoniae chromosomal DNA was isolated from G54 (47) as reported by Dillard and Yother (19).

Filter preparation. The individual DNA tag from each mutant was amplified by PCR as follows. The mutants were grown in 3 ml of TSB containing 8 µg of chloramphenicol per ml to an OD₆₀₀ of 0.6; 1 ml of this culture was centrifuged, the bacterial pellet was resuspended in 100 µl of H₂O and then heated at 96°C for 5 min; 4 µl of lysate was used as the template. The PCR mixture (50 µl) contained the primers (1 µM each), dATP, dTTP, dGTP, and dCTP (0.2 mM each), 4 µl of DNA solution, Taq DNA polymerase (1 U), and 5 µl of Taq polymerase 10× buffer (Perkin-Elmer Cetus). Twenty cycles of denaturation (60 s at 94°C), annealing (60 s at 50°C), and polymerization (10 s at 72°C) were performed on a Perkin-Elmer Cetus thermocycler. The amplification products were checked on a gel, and the same amount of DNA tags was spotted onto duplicate filters.

Pneumonia and septicemia animal models. The animal model developed by Canvin et al. (12) was adapted for this work. The experiments were carried out in accordance with prevailing legislation (Decreto legislativo italiano no. 116), company policy on the care of use of animals, and related codes of practice.

Probe preparation and hybridization. Five-microgram aliquots of chromosomal DNA extracted from S. pneumoniae in vitro and in vivo pools (19) were used as PCR templates to generate labeled probes. Genomic DNA was amplified by PCR as described above. The amplified DNA was separated on a 3% low-melting-point agarose gel, and DNA bands of approximately 80 bp were recovered. Then 5 µl of melted agarose was used as a template for a second PCR to generate the labeled probes. The PCR mixture (30 µl) contained primers (1 µM each), dATP, dTTP, and dGTP (0.2 mM each), 0.05 mM dCTP, 5 µl of [³²P]dCTP (5,000 mCi/mmol), Taq DNA polymerase (1 U), and 3 µl of Taq polymerase 10× buffer (Perkin-Elmer Cetus). Twenty cycles of denaturation (30 s at 92°C), annealing (60 s at 50°C), and polymerization (10 s at 72°C) were performed. The amplified DNA was digested with HindIII; the 40-bp fragment was recovered from a 6% low-melting-point agarose gel and added to 5 ml of hybridization buffer (5× SSC [1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate], 2% blocking reagent [Boehringer], 0.02% sodium dodecyl sulfate). The filters were hybridized for 14 h at 50°C in a Hybaid hybridization oven range, washed twice in washing buffer (5× SSC, 0.02 sodium dodecyl sulfate) for 30 min at 42, 50, and 55°C, dried, and autoradiographed with Kodak XAR-5 film at 80°C for 24 to 72 h.

Southern blot analysis. Southern blot experiments and ³²P DNA probe labeling were performed according to standard methods (54).

Localization of the integration site in the virulence-attenuated mutants. Inverse PCR was used to amplify the genomic DNA segment flanking the chloramphenicol resistance (cat) gene in the S. pneumoniae mutants. Two micrograms of genomic DNA was digested separately with each of the enzymes KpnI, HincII, ClaI, and HindIII for 8 h at 37°C. Digested genomic DNA was diluted 100-fold and self-ligated. One microliter of ligation mixture was used as a template for PCR using primers cat0 (5'-ACAGCTTCCAGATCCATATCC-3') and cat770 (5'-TATCCCACTTTATCCAATTTT-3') (Fig. 1). The PCR mixture (50 µl) contained primers (1 µM each), dATP, dTTP, dGTP, and dCTP (0.2 mM each), Taq DNA polymerase (1 U), and 5 µl of Taq polymerase 10× buffer (Perkin-Elmer Cetus). Twenty cycles of denaturation (30 s at 92°C), annealing (60 s at 50°C), and polymerization (1.5 min at 72°C) were performed. One microliter of the PCR reaction was used as the template in a nested reaction performed under the same conditions as described above, using primers cat1 (5'-TGACGTTGAGCCTCGGAACCCATCG-3') and cat720 (5'-GTTGAACCATTATATCACATT-3') (Fig. 1). The average size of the amplified DNA product was 500 bp. The PCR products were sequenced with an automated sequencer (ABI Prism 377) as instructed by the manufacturer. Predicted amino acid sequences from all six reading frames of the DNA sequences obtained above were subjected to similarity search of the SWISS-PROT TREMBL database, using the BLAST programs (1).

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Construction of a tagged S. pneumoniae chromosomal DNA library. The pR326 vector (16) is unable to replicate in S. pneumoniae and carries two markers: one for ampicillin resistance and the other for chloramphenicol resistance (cat). The latter is under the control of the pneumococcal promoter amiA and therefore can be expressed in S. pneumoniae. A mixture of double-stranded DNA tags was ligated into plasmid pR326. The ligation mixture was used to transform E. coli DH5 to obtain a pool of pR326 plasmid DNA tags (Fig. 1A). The central variable sequence allows the potential generation of 10¹⁷ different tags. Thirty ampicillin-resistant colonies were picked at random, and the variable 40-bp region of the tag was sequenced by using both upper and lower primers. None of the 30 plasmids tested contained identical 40-bp sequences (data not shown). The tagged genomic library was constructed by cloning fragments of S. pneumoniae chromosomal DNA from the G54 strain in the pool of tagged plasmids (Fig. 1). Therefore each fragment was associated with a unique DNA sequence tag. The G54 strain was transformed with this tagged genomic library. Since pR326 cannot replicate in S. pneumoniae, Cm^r colonies could be generated only by integration of the donor vector at the region of the chromosome which was homologous to the pneumococcal DNA fragment inserted in the vector (Fig. 1B). Using this procedure, we selected 1,250 Cm^r mutants.

View larger version (70K): [in this window] [in a new window]   FIG. 2.   Southern blot analysis of mutants. Genomic DNAs obtained from clones 23, 187, 985, and 1177 were digested with EcoRV (E), or with EcoRV and NcoI (E/N). The EcoRV site is absent in pR326; the NcoI site is unique to the cat gene. Five micrograms of digested DNA was then applied on 0.8% agarose gel, electrophoresed, blotted on nylon filter, and subjected to Southern hybridization analysis using the cat gene as a probe. G54 genomic DNA did not hybridize to the probe (data not shown). The positions of size markers are shown in kilobases on the right.

Screening for virulence genes. The mutant library was analyzed for reduced virulence in a mouse pneumonia model system (12). The 1,250 mutants were grouped in 25 pools of 50 clones; for each pool, duplicate filters containing spots of the individually amplified tags were prepared. The pooled mutants were grown in liquid medium, an aliquot was then removed, and genomic DNA was extracted; this represented the inoculated in vitro pool (Fig. 3). The bacteria were inoculated in mice intranasally, and six animals were used for each pool. After 32 h, the mice were killed and the lungs were recovered; then the bacteria were titrated and cultured in liquid medium. The bacteria recovered (approximately 10⁵ from each mouse) were pooled, and the DNA was extracted; this represented the recovered in vivo pool (Fig. 3). The tags from the in vitro and the in vivo pools were amplified by PCR using ³²P label in the reaction. The invariant arms of 20 bp were released by digestion with HindIII (Fig. 1), and the radiolabeled 40-bp central sequences were used as a probe against the duplicate filters. Figure 4 shows as an example two filters hybridized with the in vivo and in vitro tags; tags that hybridize to the probe from the inoculated in vitro pool but not to the probe from the recovered in vivo pool represent mutants with attenuated virulence. From the screen of our 1,250 S. pneumoniae mutants, we identified 200 clones whose tag did not hybridize to the probe obtained from the recovered pools, showing reduced virulence in vivo.

View larger version (31K): [in this window] [in a new window]   FIG. 3.   Design of virulence gene screening. S. pneumoniae chromosomal DNA was fragmented, size selected, and ligated with the pool of pR326 plasmid DNA tags. Homologous recombination events lead to insertion-duplication mutagenesis. DNA tags were individually amplified from all mutants and applied to duplicate filters in groups of 50; 50 mutants representing the tags on each filter were pooled and grown in a flask (20 ml of broth). Fifteen milliliters was removed for DNA extraction (in vitro pool). Fifty microliters was used to inoculate mice intranasally; after 32 h, the mice were sacrificed and their lungs were removed and homogenized. The recovered bacteria (in vivo pool) were plated by serial dilutions. The bacteria recovered (approximately 50,000 CFU from each mouse) were pooled and grown in a flask, and then the genomic DNA was extracted. The tags of the in vitro and in vivo pools were amplified and labeled with [32P]dCTP. Variable regions were recovered by HindIII digestion and used to probe twin filters; DNA tags that hybridize to the family of probes from the in vitro pool but not from the in vivo pool correspond to mutants negatively selected in vivo.

View larger version (64K): [in this window] [in a new window]   FIG. 4.   Identification of avirulent mutants. Two identical filters containing the same amount of individually amplified DNA tags (from A1 to L5) were hybridized to labeled DNA tags from the in vitro pool and the in vivo pool. The filters shown exemplify the many analyzed. Tags that weakly hybridized to the probes from both the in vitro pool and the in vivo pool (D1, E1, F1, G1, D2, C5, and E3) probably corresponded to slow-growing mutants and were not considered. Tags that hybridized to the probe from the in vitro pool but not to the probe from the in vivo pool (I5, L5, F5, L1, A2, and C4) corresponded to mutants with attenuated virulence and therefore were unable to survive in the host.

Sequence analysis of the virulence-attenuated clones. To identify the genes responsible for virulence attenuation, we first attempted to clone the DNA regions flanking the cat insertion point in E. coli either by recovering plasmid pR326 or by cloning the resistance marker and its flanking region into a low-copy-number vector. In many cases, the plasmids were not stable and were severely rearranged. Instability of plasmids containing streptococcal DNA in E. coli is well documented (14, 18, 40). To circumvent this problem, the regions flanking the cat insertion points were identified by inverse nested PCR and subsequently sequenced as described in Materials and Methods. Sequence analysis of the 200 clones revealed that they originated from insertions in 126 different genes. Some of the mutants resulted from different insertions within the same gene, whereas in other cases the mutants had identical insertion points within the same gene.

Analysis of septicemia of the virulence-attenuated clones. To evaluate rapidly the attenuated virulence of the individual mutants in a different in vivo model, all 126 clones were individually analyzed in a mouse septicemia model. The mutants fell into three classes, as shown in Table 1. Those mutants showing the same virulence as G54 (intraperitoneal 50% lethal dose of 10²) were classified as high, since all of the 15 treated mice died within 2 days. Mutants showing the same virulence attenuation as the nonencapsulated PR212 strain (intraperitoneal 50% lethal dose of >10⁵) were classified as low, since all of the 15 treated mice survived the 4-day duration of the experiment. Mutants falling in between the two classes were defined as medium. Thirty-six mutants were shown to belong to the low-virulence class, 52 were classified as having medium virulence, and 38 were as virulent as the parent strain (G54) and thus classified as having high virulence. The analyzed mutants had different in vitro growth rates; however, we did not find a correlation between the loss of virulence and slow-growing bacteria since several mutants that grew very slowly in vitro showed the same virulence as G54 wild-type strain. We conclude that in vivo and in vitro growth rates do not correlate.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

Characterization of microbial genes that are specifically required for the establishment and maintenance of an infection is an important aspect of understanding the mechanism by which microbial pathogens cause diseases. Large-scale isolation of virulence genes has been performed by either the IVET (38, 39) or STM (15, 28, 42) method.

The approach that we used to identify new virulence genes in S. pneumoniae was validated by the fact that among the mutants unable to survive in mice, insertions were found in previously identified or predicted pneumococcal virulence genes (Table 1). There are contradictory reports about the role of autolysin in virulence; some authors reported that autolysin-negative mutants were less virulent than the parent strain (4), while others claimed that autolysin played no role (56). Our first screen demonstrates the importance of autolysin for establishing pneumonia, while intraperitoneal inoculation of the same mutant shows that it is not necessary for septicemia. The discrepancies between the published data and ours could be due to differences in the genetic methodologies. The role of neuraminidase in streptococcal pathogenesis has been inferred from immunization studies showing that the purified protein partially protected mice against pneumococcal infection (36). A hypothetical role of hyaluronidase in virulence has been suggested since this enzyme hydrolyzes hyaluronic acid, an important component of connective tissue (5). Our results are the first genetic evidence that these genes play an important role during pneumonia infection. No significant reduction in virulence was observed when mutants SPN-605, SPN-673, and SPN-1264 were analyzed in a mouse septicemia model. In this model, the adhesion and colonization steps are not required since bacteria are introduced directly in the bloodstream. It is thus likely that these genes are involved in steps specific for pneumonia infection. We did not find in our screen genes involved in capsule biosynthesis. The mutagenesis that we performed was not saturating since we screened only 1,250 insertional mutants. However, we were surprised not to find at least one mutant mapping in the 20-kb-long capsule cassette.

Neither the pneumolysin gene nor a pneumolysin homologue which can be identified by sequence homology in the genome (this gene might code for a second hemolysin recently identified in S. pneumoniae [13]) was found in our screen. This could be due to the fact that mutants deficient in virulence proteins acting extracellularly such as the hemolysins might not be detected by this methodology since they will be complemented by the other strains in the pool.

We found a CBP whose inactivation (SPN-227) resulted in virulence attenuation in both pneumonia and septicemia models. LytA and PspA are the best-characterized CBPs: LytA is the major cell wall amidase and is required for cell autolysis (50); PspA has been characterized as a protective antigen, although its function is unknown (60, 61). Additional CBPs have been identified recently in S. pneumoniae (25, 52): SpsA described by Hammerschmidt et al. (25), CbpA described by Rosenow et al. (52), and PcpA (GenBank accession no. Z82001) identified by A. Sanchez-Beato, and J. Garcia. From genomic analysis, we know that our CBP is different from the above-mentioned genes that map at other loci on the G54 chromosome.

From our in vivo screen, we identified a mutant (SPN-1810) with an insertion in the IgA1 protease; this enzyme belongs to the class of extracellular endopeptidases which specifically cleave human (52) but not mouse (48) IgA1. It has recently been found that in Neisseria gonorrhoeae, IgA1 protease is important for survival of this pathogen within epithelial cells due to its ability to cleave LAMP1, a major integral membrane protein of lysosomes (35). Since inactivation of IgA1 leads to reduced virulence in a mouse model, we suggest that IgA1 protease in S. pneumoniae may play a role in pathogenesis by cleaving some substrate(s) other than immunoglobulins. Further support for this hypothesis comes from the fact that SPN-1810 has attenuated virulence in both pneumonia and septicemia models; thus, the IgA1 protease could play roles in addition to counteracting the mucosal defenses (a specific step in pneumonia infection). Alternatively, the phenotype observed could be ascribed to a polar effect on the expression of downstream genes. From our genome analysis, it was possible to identify only one downstream open reading frame (ORF) since our contig ends with this ORF. Surprisingly, this ORF codes for a gene showing homology to S. pneumoniae iga (29% identity and 49% similarity). This iga-like ORF belongs to the same operon as iga, and we cannot tell if there are termination signals downstream. This gene, the one identified by STM (mutant SPN-1471 [Table 2], and iga (59) constitute a family of iga-like genes stressing the importance of its function for S. pneumoniae.

The ability to adapt to the host environment is a key component of pathogenicity. Many genes whose products are involved in metabolic pathways and nutrient uptake have indeed been identified following our negative in vivo selection. De novo synthesis of purines has been shown to be a strict requirement for bacterial growth and persistence under the nucleotide-limiting conditions in the host (38); also purine auxotrophy attenuates many pathogenic bacteria (58). Accordingly, insertions in four genes belonging to the purine biosynthetic pathway (purE, purK, purC, and purL) were found. Three insertions in genes involved in glutamine metabolism, glnA (glutamine synthase), glnQ, and glnH (high-affinity glutamine transport system), were identified, suggesting that reduced availability of glutamine synthesis or loss of high-affinity glutamine transport leads to virulence attenuation in S. pneumoniae. Klose and Mekalanos (34) reported that only the simultaneous inactivation of glutamine synthesis (glnA) and high-affinity transport (glnH) reduced virulence in S. typhimurium. Since salmonellae also have a low-affinity glutamine transport system, a mutant defective in both glnA and glnH can survive in vitro when supplemented with a high level of glutamine but not in vivo due to the low level of this metabolite in animal tissues. Glutamine is a key metabolite since it is the major donor of amino groups in the amino acid biosynthetic pathway and it is required in the first two steps of purine biosynthesis. The virulence defects of S. pneumoniae support the hypothesis that the concentration of glutamine in the host is a critical parameter for in vivo survival and show that import and biosynthesis are both required in vivo.

Nutrient and amino acid uptake seems to be a critical factor for in vivo survival of S. pneumoniae. Two mutants unable to grow in mice had insertions in potA and potD, which encode the ATPase component and substrate binding lipoprotein, respectively, of the spermidine/putrescine ABC transport system. Spermidine and putrescine are polyamines whose synthesis uses arginine as a precursor. A mutant in arginine transport in L. monocytogenes (33) and more recently a mutant in arginine biosynthesis in V. cholerae (11) were both shown to be attenuated in virulence. One mutant (SPN-1802) had an insertion in a gene with strong homology to msmK, a component of a multiple-sugar uptake system of S. mutans (53). S. pneumoniae is limited in sugar fermentation ability, being able to use only glucose, raffinose, lactose, threalose, and maltose as sources of energy (30, 44). The MsmK homologue in S. pneumoniae is likely to play a role in the transport of those sugars.

One gene involved in anaerobic metabolism, adh (encoding alcohol dehydrogenase) was identified in this study. S. pneumoniae is a facultative anaerobe, and its energy-yielding metabolism is fermentative. Alcohol dehydrogenase catalyzes the reduction of acetyl coenzyme A first to acetaldehyde and then to ethanol. These reactions are needed to regenerate reducing ability (NADH) from glycolysis under anaerobic conditions. Our data suggest that as infection progresses, streptococci must adapt to low-oxygen tension environments, and a fully active anaerobic metabolism is required for survival in the host.

One insertion (SPN-1055) maps in a gene whose product shows similarity to ClpC, one of the regulatory subunits of Clp proteases. Clp proteases are a family of highly conserved proteins involved in ATP-dependent proteolysis (23). The Clp protease in E. coli is a two-component system where the larger regulatory subunit (ClpA, ClpB, or ClpX) has ATPase activity and the smaller subunit (ClpP) has proteolytic activity. The Clp protease is involved not only in degradation of toxic denatured proteins but also in controlling the half-life of unstable proteins to modulate gene expression. In fact, Clp proteases have been shown to be important in modulation of stress response, stationary phase, and competence in B. subtilis (41). Recent evidence suggests that Clp proteases also may be involved in modulating virulence. A ClpP homologue has been shown to be essential for survival of S. typhimurium in the spleen (28), the regulatory subunit ClpC is required for survival of L. monocytogenes in macrophages (51), and more recently a ClpX homologue has been found to be important for virulence in S. aureus (42). It is not clear what role is played by these Clp proteins in the virulence process or what pathways are affected, but it is tempting to suggest a common mechanism of action. Identification of the gene products that are regulated by the S. pneumoniae ClpC homologue will shed light on the role of this regulatory protein in virulence.

The ability to keep spontaneous mutation levels low seems to be crucial for in vivo survival. A mutant (SPN-1634) showing insertions in the S. pneumoniae mutX gene (the homologue of E. coli mutT) (43) and a mutant (SPN-1411) whose insertion maps in a gene similar to recJ were identified. MutX is an NTPase whose role is to hydrolyze 8-oxo-dGTP to 8-oxo-dGMP, thus preventing its incorporation into DNA (21, 43). RecJ is a single-stranded DNA binding protein involved in DNA recombination/repair (37). It is known that during infection, pathogenic bacteria may encounter a variety of oxygen radicals produced by macrophages and neutrophils. These oxygen species are likely to induce mutations in the DNA or to oxidize the NTP pool. The ability to repair DNA damage has indeed been shown to be essential for full S. typhimurium (27) and S. aureus (42) virulence. The recombination-deficient recA and recBC mutants of S. typhimurium have been shown to be sensitive to the oxidizing compounds generated by the oxidative burst of macrophages or phagocytes (9, 27).

The results presented in this paper clearly show that S. pneumoniae virulence is a multifactorial process and that, in addition to factors involved in specific steps in pneumonia infection, many other gene products are required to confer to the bacterial cells the ability to grow and survive in the host. Most of the mutants isolated that bore genes involved in metabolic functions, transport, and DNA modification showed reduction in virulence in both the pneumonia and septicemia models. This finding is consistent with the hypothesis that these functions are needed for bacterial adaptation to the different host environments and not for specific steps during infection. The adaptation of IVET and STM technologies to several distinct bacterial pathogens is revealing that different pathogens causing different diseases may have virulence determinants in common. Indeed, many of the genes identified in this work as important for S. pneumoniae survival in the host were also identified as virulence determinants in other pathogens (S. typhimurium, S. aureus, V. cholerae, and P. aeruginosa).

Insertions in several genes that show no homology to known genes in databases or that have similarity to hypothetical proteins could potentially encode novel factors that may play important roles in the bacterial infection process. The identification of the entire ORF and search for specific motifs or domains may give clues about the putative function of these proteins.

Despite the possibility that some of our mutants show polar effects, our data provide strong evidence that the ability to sense and respond to complex and overlapping signals encountered in the host is a key point in pathogenicity. This ability results in the elaboration of regulatory, metabolic, and virulence functions that contribute to growth and persistence in the host. We are currently constructing nonpolar mutations to exclude the involvement of neighboring genes in virulence. We believe that the approach described here can serve as a basis for a more comprehensive understanding of S. pneumoniae pathogenicity.