A genomic analysis of two-component signal transduction in Streptococcus pneumoniae

Authors


Martin K. R. Burnham, E-mail Martin_K_Burnham@sbphrd.com; Tel. (+1) 610 917 6349; Fax (+1) 610 917 7901.

Abstract

A genomics-based approach was used to identify the entire gene complement of putative two-component signal transduction systems (TCSTSs) in Streptococcus pneumoniae. A total of 14 open reading frames (ORFs) were identified as putative response regulators, 13 of which were adjacent to genes encoding probable histidine kinases. Both the histidine kinase and response regulator proteins were categorized into subfamilies on the basis of phylogeny. Through a systematic programme of mutagenesis, the importance of each novel TCSTS was determined with respect to viability and pathogenicity. One TCSTS was identified that was essential for the growth of S. pneumoniaeThis locus was highly homologous to the yycFG gene pair encoding the essential response regulator/histidine kinase proteins identified in Bacillus subtilis and Staphylococcus aureus. Separate deletions of eight other loci led in each case to a dramatic attenuation of growth in a mouse respiratory tract infection model, suggesting that these signal transduction systems are important for the in vivo adaptation and pathogenesis of S. pneumoniae. The identification of conserved TCSTSs important for both pathogenicity and viability in a Gram-positive pathogen highlights the potential of two-component signal transduction as a multicomponent target for antibacterial drug discovery.

Introduction

Streptococcus pneumoniae is associated with a range of community-acquired diseases, which include otitis media, pneumonia and meningitis (Paton et al., 1993; Paton, 1996). Despite the growing significance of such infections, little is known about the factors that govern the physiological responses of this organism, or the genetic repertoire that the pneumococcus employs to create disease. In spite of the advent of antibiotics, mortality from pneumococcal disease has remained unacceptably high (Gillespie, 1989). Moreover, the recent increase in pneumococci resistant to penicillins and other therapeutic agents has highlighted the need for novel antimicrobials directed at new targets. To this end, researchers are beginning to unravel the molecular basis of streptococcal pathogenicity; the virulence of this organism is now known to be dependent upon a series of key determinants, including capsular polysaccharide, pneumolysin, extracellular enzymes and autolysis (Paton et al., 1993; Polissi et al., 1998). Several of these attributes are thought to be regulated in a cell density-dependent manner, although the genetic factors that govern the pathogenicity of S. pneumoniae remain largely unknown (Tuomanen and Masure, 1997).

Histidine–aspartate phosphorelay systems, also known as two-component signal transduction systems (TCSTSs), are one mechanism that bacteria use to sense and respond to their environment. Arguably the most widespread means of signal transduction in bacteria, these systems have been shown to regulate a wide variety of cellular responses, including osmoregulation, photosynthesis, chemotaxis, sporulation, antibiotic production and pathogenicity, in a number of different bacteria (Appleby et al., 1996). Perhaps more importantly, evidence is now accumulating that links phosphorelay signalling with the cell cycle and bacterial viability (Quon et al., 1996; Fabret and Hoch, 1998). TCSTSs are typically composed of two signalling proteins: a sensor kinase and a cognate response regulator. Upon receipt of a specific stimulus, the kinase domain of the sensor kinase protein is activated, resulting in the autophosphorylation of a conserved histidine residue. Once phosphorylated, the histidine kinase donates the high-energy phosphate group to a cognate response regulator. Phosphorylation of a conserved aspartate residue that lies within the receiver domain of the regulator is thought to lead to structural changes in the protein, allowing the response regulator to mediate changes in gene expression or protein function (Egger et al., 1997).

Histidine kinases are generally composed of a highly variable transmembrane-spanning amino-terminal sensor kinase and a more conserved carboxyl-terminus composed of the transmitter and kinase domains. Within the carboxyl-terminal domain are four, or occasionally five, recognizable amino acid motifs or boxes, which are named after the single-letter amino acid code for the most conserved residues, the H, N, D, F and G boxes. The H box centres on a strictly conserved histidine residue, the site of autophosphorylation (Stock et al., 1989). The amino-terminal receiver domain of response regulators typically contains a stretch of four hydrophobic residues followed by the D1 and D2 boxes, each of which contains a conserved aspartate residue, and an additional K box, which contains an invariant lysine residue. The conserved aspartate, which lies within the D2 box, is considered to be the site of phosphorylation (Volz, 1993).

Four probable histidine–aspartate phosphorelay systems have been identified previously in S. pneumoniae. The ciaRH gene pair has been implicated in cefotaxime resistance of S. pneumoniae and appears to play a role in the regulation of competence (Guenzi et al., 1994; Zahner et al., 1996). A second two-component system, comDE, regulates competence in response to changing concentrations of a signal peptide, encoded by the comC gene (Pestova et al., 1996; Cheng et al., 1997). Tuomanen and Masure (1997) reported previously that this locus is not required for pneumococcal pathogenicity, autolysis or adherence. The pnpSR TCSTS gene pair has been linked to the regulation of competence in S. pneumoniae (Novak et al., 1999a), and a fourth TCSTS encoded by the vncSR gene pair has recently been associated with the development of vancomycin tolerance (Novak et al., 1999b). Although two-component signal transduction has been highlighted as a potential antimicrobial target (Barrett and Hoch, 1998; Barrett et al., 1998), as yet, His-Asp phosphorelays have not been linked to the regulation of viability or pathogenicity of S. pneumoniae. As a first step towards understanding the role of phosphorelay cascades in the regulation of pathogenicity and viability in the pneumococcus, we have undertaken the identification, genetic disruption and preliminary analysis of each and every novel TCSTS of S. pneumoniae.

Results

Identification and analysis of His–Asp gene pairs

A total of 13 TCSTS gene pairs and a single orphan response regulator were identified through bioinformatic analysis of the S. pneumoniae genome sequence databases (The TCSTS sequences identified in this study are accessible through http://www.blackwell-science.com/mmi) (Fig. 1). Among the four S. pneumoniae strains examined, the estimated levels of sequence coverage of these genomes ranged from 90% to > 99%. Thus, it is likely that the TCSTS sequences presented here represent the full complement of two-component loci harboured by S. pneumoniae. Partial or truncated histidine kinase (hk) and response regulator (rr) genes were originally identified in two contigs (539hk/rr and 489rr) and, in each case, inverse polymerase chain reaction (PCR) was used to complete the sequence of the TCSTS loci and identify neighbouring genes.

Figure 1.

. Phylogenetic trees using neighbour joining of the histidine kinases and response regulators from E. coli (ECO), B. subtilis (BSU), Streptococcus gordonii (SGO), Lactobacillus sakei (LSA) and S. pneumoniae (SPN). In each case, either the designated gene name or ORF identification number is given. The TCSTSs derived from S. pneumoniae are highlighted in green, blue (attenuated) and red (essential). The hk and rr proteins have been assigned to the Agr, Lyt, Nar or Pho subfamilies on the basis of homology; numbers represent the percentage occurrence of nodes in 1000 random bootstrap replications of maximum parsimony and neighbour-joining tree methodologies. Only values ≥ 50% are shown. Scale bar represents 0.1 expected amino acid substitutions per site. Details of phylogenetic methods and results are given in the text.

Comparisons of S. pneumoniae, Bacillus subtilis and Escherichia coli TCSTSs suggest that S. pneumoniae harbours genes belonging to four previously described subfamilies. Two hk/rr pairs belong to the Nar subfamily, seven hk/rr pairs and a single response regulator fall into the Pho subfamily, two hk/rr pairs are in the Lyt subfamily and the two remaining hk/rr pairs form part of a new subclass not previously reported, which we designate the Agr subfamily. Separate hk and rr analyses for each of the four subclasses resulted in eight phylogenetic data sets, alignments ranged from 120 to 235 amino acids in length, depending upon the proteins (available upon request from K. K.). Maximum parsimony (MP) and neighbour-joining (NJ) analyses generally resulted in congruent tree topologies. MP analysis showed relatively few minimal length (ml) trees for subfamilies of hk [Nar (ml = 1099; one tree), Pho (ml = 2456, three trees). Lyt (ml = 579, two trees), Agr (ml = 498, one tree)] and rr [Nar (ml = 1248; one tree), Pho (ml = 2613, three trees). Lyt (ml = 321, one tree), Agr (ml = 496, one tree)]. Both NJ and MP methods produced similar levels of bootstrap for the various nodes (Fig. 1).

Of 14 genes predicted to encode response regulators, all but one were found adjacent to genes encoding histidine kinases. For each gene pair, the open reading frames (ORFs) encoding both the hk and rr overlapped, and no putative promoters or transcriptional terminators could be identified within either gene, indicating that each hk/rr is likely to be transcribed as a single transcriptional unit. Moreover, both the hk and rr of each gene pair clustered within the same subfamily. Thus, we believe that both components of each S. pneumoniae TCSTS gene pair act as partners within a single His–Asp phosphorelay unit. Interestingly, S. pneumoniae lacks TCSTS pairs belonging to the chemotaxis (Che), citrate uptake (Cit) and hybrid kinase families. Analysis of both gene locale and gene phylogeny indicated that the 479hk/rr, 492hk/rr, 474hk/rr and 481hk/rr gene pairs appear to be orthologous to the yvfTU, yycFG, yvqCE, phoPR His–Asp gene pairs of B. subtilis respectively. Of these systems, only two have been studied previously in B. subtilis : phoPR is responsible for phosphate regulation (Hulett, 1996); and yycFG appears to be essential for in vitro growth (Fabret and Hoch, 1998). Four gene pairs (494hk/rr, 481hk/rr, 491hk/rr and 498hk/rr) correspond to TCSTS pairs (ciaRH, pnpRS, vncRS, comDE respectively) that have been identified previously in S. pneumoniae (Guenzi et al., 1994; Pestova et al., 1996; Cheng et al., 1997; Novak et al., 1999a,b).

As some TCSTS gene pairs are adjacent to the genes that they control, ORFs that neighbour each TCSTS locus were identified. Further sequence analysis revealed five gene pairs (486hk/rr, 474hk/rr, 479hk/rr, 492hk/rr and 539hk/rr) and the single response regulator (489rr), which appear to be transcriptionally coupled to unrelated ORFs. The histidine kinase gene 492hk seems to be linked to a homologue of the B. subtilis yycJ ORF (55% sequence identity); two other loci (486hk/rr, 539hk/rr) appear to be translationally coupled to ORFs that are homologous to proteins of unknown function. The response regulator 489rr appears to be translationally coupled to a gene encoding a putative phosphogluconate dehydrogenase enzyme; both the 474hk/rr and the 479 hk/rr gene pairs are coupled to genes encoding components of ABC transporters. A further three TCSTS gene pairs are flanked by, but do not appear to be transcriptionally linked to, genes encoding ABC transporters (486hk/rr, 491hk/rr, 481hk/rr).

Phylogenetic analysis of the TCSTSs also reveals that the 486hk/rr pair shares considerable homology with, but is not identical to, components involved in the regulation of cell–cell signalling in other Gram-positive bacteria (Fig. 1). Adjacent to the histidine kinase/response regulator locus, but transcribed convergently, are homologues of the pneumococcal proteins ComA (65% identity) and ComB (66% identity), thought to be responsible for the export of the competence-stimulating peptide (CSP) in S. pneumoniae (Pestova et al., 1996; Cheng et al., 1997). In addition, an ORF encoding a peptide homologous to the lactacin F lantibiotic LafA borders the ComAB gene pair (Muriana and Klaenhammer, 1991).

Mutagenesis of two-component signal transduction loci

In an effort to examine the importance of each locus identified, we attempted to inactivate each of the TCSTSs in S. pneumoniae by allelic replacement and then to determine the phenotype of each mutant with respect to viability and pathogenicity. For each gene deletion described here, the entire target gene(s) was replaced with the constitutively expressed ermAM gene cassette, from which the transcriptional termination signals had been removed. In all cases, the inserted cassette was transcribed in the same orientation as the target gene, assuring transcription of regions downstream of the deletion and thereby minimizing the potential for polarity effects. This mutagenesis strategy resulted in a stable insertion that eliminated the possibility of residual activity resulting from incomplete gene inactivation and minimized disruption of any adjacent genes. Successful transformation of S. pneumoniae R6 and recombination of the target gene typically yielded 102–103 erythromycin-resistant transformants μg−1 DNA. In contrast, no transformants were obtained when S. pneumoniae was transformed with an erythromycin resistance cassette designed to delete the essential gene, uppS (Apfel et al., 1999).

In an effort to minimize ‘cross-talk’ between the non-cognate TCSTSs, we attempted to delete either the entire phosphorelay gene pair or the gene encoding the response regulator. A total of 26 genes or gene pairs was successfully disrupted using this strategy, and each of the deletion mutations was stably maintained in the absence of selection. The growth rate, colony and cell morphology of each S. pneumoniae 100993 TCSTS mutant was examined but, remarkably, no major differences between the wild type and mutants could be ascertained; only the ciaRH deletion resulted in a small but significant decrease in growth rate (Table 1; for further details, see http://www.blackwell-science.com/mmi). Successive attempts at deleting one response regulator (492rr) failed to generate erythromycin-resistant mutants. Deletion of the 492 histidine kinase or the downstream yycJ homologue did not appear to be lethal, nor did the mutants exhibit any obvious growth defects in vitro (Table 1). To study the essentiality of the putative essential response regulator further, the upstream region of the 492rr was replaced by the constitutive promoter, PermAB, and the fusion construct was cloned into a streptococcal plasmid vector, pDL277. The final construct, designated pDL492, was transformed into S. pneumoniae R6. After plasmid complementation, renewed attempts to delete the 492rr gene were successful, confirming that this gene is essential for growth and viability of S. pneumoniae.

Table 1. . Summary of S. pneumoniae TCSTS mutations (alternative gene names are given in parentheses).Thumbnail image of
  • a

    Where possible, the corresponding B. subtilis homologues and the degree of attenuation in the RTI model are indicated. The in vitro growth rate of selected mutants is expressed as doubling time (in min). ND, not done; Y, gene deletion successfully constructed. *Mutant forwarded for in vitro and in vivo growth assessment; E, essential for viability; WT, not attenuated.

  • Analysis of S. pneumoniae TCSTS mutants in a respiratory tract model of infection

    To ascertain the role of His–Asp phosphorelay signalling in the pathogenicity of S. pneumoniae, the various S. pneumoniae 0100993 TCSTS deletion mutants were used to infect groups of mice intranasally using a standard murine respiratory tract infection (RTI) model. Bacterial load was enumerated from lung tissue samples and used to examine directly the capacity of the bacteria to grow and survive within the host, therefore providing an accurate assessment of growth attenuation in vivo. During initial experiments, we noted that the inactivation of histidine kinase genes alone resulted in substantially less attenuation than that obtained with strains harbouring the corresponding response regulator deletion. Thus, where possible, we have included attenuation data that relate to the deletion of either a response regulator or a complete TCSTS pair. Inactivation of eight individual loci (478hk/rr, 480hk/rr, 481hk/rr, 486hk/rr, 484hk/rr, 539hk/rr, 489hk/rr and ciaRH ) resulted in a dramatic reduction in in vivo cell number compared with the wild type (Fig. 2). Of these mutants, the 481rr deletion exhibited the greatest degree of attenuation, with a mean bacterial count of 1.74 log10 cfu lung−1, indicating that this mutant is almost completely attenuated. The wild type, in contrast, gave a mean bacterial count of 7.39 log10 cfu lung−1. Both the 480hk/rr and ciaR mutants were somewhat less attenuated than the 481rr mutant, with mean bacterial counts of 2.2 log10 cfu lung−1 and 1.97 log10 cfu lung−1 respectively.

    Figure 2.

    . The virulence of S. pneumoniae TCSTS mutants in a mouse RTI model of infection. Groups of five mice were challenged intranasally with ≈105 cfu of wild-type pneumococcus and defined TCSTS mutants. Forty-eight hours after inoculation, mice were sacrificed and lung tissue removed. The mean viable bacterial counts (cfu lung−1) recovered from homogenized lung tissue are indicated. The degree of attenuation was assessed as the reduction in mean bacterial counts (cfu lung−1) recovered from animals inoculated with the S. pneumoniae 0100993 TCSTS deletion mutants when compared with lung tissue sampled from animals infected with the parent strain.

    Surprisingly, the 492hk deletion mutant did not exhibit any appreciable attenuation when tested in the RTI model. In contrast, inactivation of the apparently linked downstream pneumococcal yycJ gene homologue resulted in a 3 log attenuation of bacterial growth in vivo. The four remaining mutants displayed little or no appreciable attenuation when tested in our model (< 102).

    The RTI model used in this study allowed us to examine the effects of specific mutations at early presymptomatic and later symptomatic stages of infection. To this end, mice were infected separately with equal doses of five S. pneumoniae 0100993 attenuated mutants (478hk/rr, 480hk/rr, 489rr, 486rr and 486hk). Lung tissue was removed at predetermined intervals, and bacteria were enumerated (Fig. 3). The viable counts of the 480hk/rr mutant dropped throughout the time course, resulting in a 2 log drop within the first 12 h of infection and a 5 log attenuation at 48 h. Thus, the 480hk/rr mutant appears to be unable to establish or maintain a successful infection. In contrast, the bacterial load of the 486hk, 486rr, 489rr and 478hk/rr mutants remained constant or increased during the first 24 h of infection and only began to drop in the symptomatic phase of infection, indicating that these loci may play a more important role in the later stages of infection.

    Figure 3.

    . Evaluation of the growth attenuation of selected S. pneumoniae TCSTS mutants throughout the duration of a mouse RTI model of infection. S. pneumoniae TCSTS mutants previously shown to be attenuated in the mouse RTI model of infection were administered to pools of five mice intranasally. At predetermined time intervals, mice were sacrificed, and the lungs were aseptically removed and homogenized in 1 ml of PBS. The numbers of viable bacteria extracted at each time point are shown. For ease of comparison, the corresponding S. pneumoniae control infection is included.

    Discussion

    Thirteen TCSTS gene pairs and one unpaired response regulator have been identified in each of the four independent isolates of S. pneumoniae examined as part of this study. We believe that this represents the full complement of TCSTSs in this important pathogen. Although the number of TCSTSs found in S. pneumoniae is less than in either B. subtilis (34/35 hk/rr genes) or E. coli (29/34 hk/rr genes), their presence suggests that His–Asp phosphorelay signal transduction is an important mechanism for regulating gene expression even in highly adapted pathogens such as S. pneumoniae. Ten of the TCSTSs can be considered to be novel; a further four loci correspond to the pnpRS, comDE and ciaRH and vncSR gene pairs identified previously in S. pneumoniae (Guenzi et al., 1994; Pestova et al., 1996; Novak et al., 1999a,b).

    Mutagenesis studies and complementation experiments demonstrated that one response regulator (492rr) is essential for cell growth in S. pneumoniae. Surprisingly, inactivation of the cognate histidine kinase had little or no effect on growth in vivo or in vitro. Both the histidine kinase and the response regulator of this system are homologous to the YycFG TCSTS pair identified recently in Staphylococcus aureus and B. subtilis (Fabret and Hoch, 1998; Martin et al., 1999). In each case, both the histidine kinase, YycG, and the response regulator, YycF, were found to be indispensable for growth. The fact that the S. pneumoniae 492hk mutant displayed no significant reduction in growth rate in vitro and was not attenuated in vivo suggests that the essential activity of the 492 response regulator protein may be phosphorylation independent, despite the high homology shared by YycF homologues from S. pneumoniae, S. aureus and B. subtilis. Perhaps the 492rr is more likely to retain activity in the absence of its cognate histidine kinase by phosphorylation either from a non-cognate kinase or from some other phosphate donor. This phenomenon has been well documented in the past (Wanner, 1992; McCleary et al., 1993).

    Genome database analysis has revealed that the Gram-positive bacteria Streptococcus pyogenes, Enterococcus faecalis and Lactococcus lactis, in addition to B. subtilis, S. aureus and S. pneumoniae, harbour YycFG homologues, suggesting that this TCSTS fulfils an essential role in a wide variety of Gram-positive bacteria. The response regulator (492rr) belongs to the OmpR subfamily of response regulators, which contain a winged helix–turn–helix motif suggested as being important for DNA binding and interaction with the α-subunit of RNA polymerase (Martinez-Hackert and Stock, 1997). By analogy with other response regulators that belong to this group, it seems likely that the 492rr acts to regulate gene expression at the level of transcription.

    Sensor kinases belonging to the YycG group possess a PAS/PAC domain within the cytoplasmic N-terminal region. Originally identified as components of circadian clock regulators in eukaryotes, these domains have recently been identified in a variety of sensor proteins from both prokaryotes and eukaryotes (Zhulin et al., 1997; Zhulin and Taylor, 1998). Prokaryotic proteins containing PAS/PAC domains have been demonstrated to respond to oxygen, light and redox, and it has been proposed that the PAS/PAC region mediates this activity by binding cofactors such as FAD or NAD (Ponting and Aravind, 1997). Martin et al. (1999) recently proposed that the staphylococcal yycFG locus is responsible for maintaining membrane integrity or permeability and, thus, in Gram-positive bacteria, YycFG homologues may serve to co-ordinate membrane growth with the cellular redox potential.

    We have demonstrated that seven of the S. pneumoniae TCSTS gene pairs (484hk/rr, 481hk/rr, 486hk/rr, 478hk/rr, 539hk/rr, 480hk/rr, ciaR/H ) and a single response regulator (489rr) appear to be important for growth and pathogenicity in the mouse respiratory tract infection model. Our studies have indicated that deletions of these loci appear to have little detrimental effect on in vitro bacterial growth in nutrient-replete conditions. Phylogenetic analysis of the 481hk/rr reveals that this regulatory pair shares considerable homology with the PhoP/R proteins of B. subtilis. PhoP/R regulate the expression of a series of genes important for continued growth and survival during phosphate limitation (Hulett, 1996). Examination of the 481hk reveals that this gene is flanked by a homologue of the B. subtilis pstS gene, which encodes part of a high-affinity phosphate uptake system (Liu et al., 1998). Recently, Novak et al. (1999a) reported the identification and preliminary analysis of a streptococcal TCSTS gene pair that apparently corresponds to the 481hk/rr gene pair, which they termed pnpSR. The authors were unable to demonstrate a link between pnpSR and phosphate regulation. However, the gene organization of the 481hk/rr locus and the high homology with phoPR suggest strongly that the 481hk/rr gene pair is responsible for phosphate regulation in S. pneumoniae. The dramatic attenuation of the S. pneumoniae 481rr deletion mutant in the respiratory tract infection may indicate that phosphate plays a key role in both the onset and the continuation of pneumococcal infection.

    In S. pneumoniae, key pathogenic determinants, including adherence and autolysis, appear to be regulated in relation to growth phase. Tuomanen (1997) postulated that both attributes are triggered by an unidentified cell–cell signalling cascade. Our data indicate that these different stages of infection may be regulated (at least in part) by different TCSTS loci. Time course infection studies identified three TCSTS loci (486hk/rr, 478hk/rr and 489rr) that appear to be important for in vivo growth during the later phase of infection. One of these loci, the 486hk/rr pair, shares considerable homology with components involved in the regulation of cell–cell signalling in other Gram-positive bacteria (Dunny and Leonard, 1997). In such systems, a modified signal peptide is exported by a dedicated ATP-binding cassette (ABC transporter) and functions as the stimulus for a His–Asp phosphorelay cascade. Adjacent to the 486hk/rr locus, but transcribed convergently, lie genes encoding homologues of an ABC transporter and an ORF predicted to encode a signalling peptide. The structural organization of the 486hk/rr locus coupled with the in vivo studies indicate that this pair may regulate aspects of pathogenicity in relation to cell density.

    Neither the attenuation data nor sequence analysis of the six remaining TCSTS loci necessary for disease progression (478hk/rr, 539hk/rr, 480hk/rr, 484hk/rr, ciaRH and 489rr) revealed any obvious features that provide clues relating to their various roles during infection. The ciaRH gene pair has been implicated previously in the regulation of competence and cefotaxime resistance in S. pneumoniae. Our studies indicate that ciaRH is also important for S. pneumoniae RTI pathogenesis, as deletion of the ciaR gene led to a 3 log attenuation of bacterial growth in vivo. The role of ciaRH in competence development, antibiotic susceptibility and virulence is not yet understood; however, our data appear to link the regulation of pathogenicity with the developmental life cycle of the pneumococcus and indicate that the ciaRH pair may act as important pleiotropic regulators in S. pneumoniae.

    We were surprised to identify so many TCSTS loci, which apparently play a role in the onset and maintenance of the respiratory tract infection and yet seem to be unimportant for growth in vitro. The number of TCSTS mutants attenuated in the RTI model is an indication of both the complexity of the respiratory environment and the adaptive processes that are important for bacterial survival. This contrasts with the nutrient-replete conditions used in our in vitro testing procedures, and other models may offer a less stringent environment for the pneumococcus. We are therefore continuing to evaluate the growth of these mutants in alternative models of infection.

    Using a genomic-based approach, we have described the identification and systematic disruption of the entire TCSTS gene complement of S. pneumoniae. We have defined those S. pneumoniae TCSTSs important for both pathogenicity and viability in a respiratory tract infection. These studies have provided further support for the contention that two-component signal transduction systems are appropriate targets for the development of antibacterial drugs.

    Experimental procedures

    Identification and sequence analysis of two-component signal transduction genes

    Escherichia coli was selected as the source of query sequences, as the biochemical and genetic characteristics of TCST pathways in this species are well known (reviewed by Stock et al., 1989). Two separate searches of the complete genome sequence of E. coli were performed with the program psi-blast (applying an EXPECT threshold of 0.001) using as query sequences either the transmitter domain of the histidine kinase PhoR or the receiver domain of PhoB, the associated response regulator. For the histidine kinase search, convergence was reached after five iterations revealing a total of 31 ORFs with significant sequence similarity to PhoR (P(N) < 1.0e-05), whereas the response regulator search converged after three iterations revealing 37 ORFs with significant similarity to PhoB.

    Using additional rounds of psi-blast, the recovered E. coli histidine kinase and response regulator proteins were then used to search a total of four S. pneumoniae genomic databases for TCSTS gene homologues. The databases used include the publicly available TIGR sequence library derived from S. pneumoniae type 4 (The Institute for Genomic Research website at http://www.tigr.org) and three proprietary databases with sequence information drawn from both pathogenic (23F and 0100993) and non-pathogenic (R6) strains of S. pneumoniae. A subject protein was considered to be a putative histidine kinase only if the H, N, D and G boxes, and ideally the F box, were all present. A particular ORF was considered to be homologous to an E. coli response regulator if all three conserved regions, D1, D2 and K boxes, were found in the proper orientation.

    Similar alignment and phylogenetic methodologies were applied to histidine kinase and response regulator data sets. The homologous carboxyl termini of the histidine kinase and the homologous amino termini of the response regulator proteins were initially aligned using the program clustalw v1.7 (Thompson et al., 1994) with the blosum62 (Henikoff and Henikoff, 1992) similarity matrix and gap opening and extension penalties of 10.0 and 0.05 respectively.

    Specific histidine kinase and response regulator protein clusters were identified and named based upon hk/rr pairs of known function in B. subtilis and E. coli. A full-length sequence from each subgroup was then used as a query sequence in a psi-blast search against the non-redundant database. A profile produced from the converged psi-blast run was used as input into a second psi-blast run against the database of the respective subgroup of E. coli, B. subtilis and S. pneumoniae histidine kinase or response regulator proteins. This produced a highly accurate multiple sequence alignment for each subgroup. The multiple sequence alignments were refined further manually using the program seqlab of the GCG v9.0 software package (Genetics Computer Group, Madison, WI, USA).

    Phylogenetic trees were constructed by neighbour-joining (NJ) and maximum parsimony (MP) methods for each set of alignments. NJ trees were based on pairwise distances between amino acid sequences, using the programs neighbor and protdist of the phylip 3.57c package [program distributed by the author: http://evolution.genetics.washington.edu/ phylip.html. J. Felsenstein (1993), Department of genetics, University of Washington, Seattle, WA, USA]. The dayhoff program option was invoked in the latter program, which estimates the expected amino acid replacements per position (EAARP) using a replacement model based on the dayhoff 120 matrix. The programs seqboot and consense were used to estimate the confidence limits of branching points from 1000 bootstrap replications. MP analysis was carried out using the software package paup [D. L. Swofford (1999). paup*, Phylogenetic Analysis using Parsimony (*and other methods), Version 4. Sinauer Associates, Sunderland, MA, USA]. Given the large size of the data set, it was not possible to search exhaustively for the total number of minimal length trees. Instead, the numbers and lengths of minimal trees were estimated from 100 replicate random heuristic searches, whereas confidence limits of branch points were estimated by 1000 bootstrap replications.

    Bacterial strains and plasmids

    S. pneumoniae 0100993 (NCIMB no. 40794; serotype 3) is a plasmid-free, encapsulated clinical isolate kindly provided by Mr D. Felmington (University College Hospital, London, UK). S. pneumoniae R6 (Lacks and Hotchkiss, 1960) is a non-encapsulated wild-type strain. S. pneumoniae strains were maintained in Todd–Hewitt broth plus 0.5% yeast extract (THB-YE), blood agar plates (Difco) or AGCH medium (Lacks, 1966). Cultures were supplemented with erythromycin (1 μg ml−1) or spectinomycin (120 μg ml−1). E. coli DH5α was grown in Luria broth (LB; Difco) or Luria agar (LA; Difco) with ampicillin (100 μg ml−1) or spectinomycin (100 μg ml−1) as required. pDL277 is a StreptococcusE. coli shuttle cloning vector (Dunny et al., 1991). Plasmid DNA isolation, restriction analysis, ligation, transformation and Southern blot analysis were performed as described by Ausubel et al. (1997).

    Inverse PCR

    Inverse PCR reactions were performed according to Guidolin et al. (1994). Chromosomal DNA (5 μg) was digested to completion with appropriate restriction enzymes. The cleaved DNA was then self-ligated at a concentration of 25 μg ml−1. Amplification reactions (30 cycles, annealing temperature 50°C), performed with 250 ng of ligated DNA and 25 pmol of each divergent primer, were used to generate PCR fragments. The PCR products were sequenced using an Applied Biosystems sequencing system to extend the sequence of the TCSTS gene fragments; new PCR primers were then designed for further rounds of inverse PCR.

    Construction of S. pneumoniae TCSTS mutants

    DNA fragments of 500–1000 bp, which flank the target gene(s), were fused to a 920 bp erythromycin resistance cassette (ermAM ) (Martin et al., 1987; Lunsford and London, 1996) using cross-over PCR (Link et al., 1997). These were then used to transform S. pneumoniae R6 according to published protocols (Håvarstein et al., 1995). Transformants were selected on AGCH agar containing erythromycin (1 μg ml−1) and analysed using PCR to confirm correct integration of the engineered construct. Chromosomal DNA was extracted from the S. pneumoniae R6 mutants as has been described previously (Paton et al., 1986) and used to transform the virulent strain S. pneumoniae 0100993. Correct allelic replacement in both the S. pneumoniae R6 and 0100993 strains was then verified using Southern blot analysis. Using this procedure, the entire target gene was deleted with minimal disruption of upstream or downstream sequences. To assess the stability of each TCSTS deletion before in vivo and in vitro assessment, S. pneumoniae 0100993 mutants were grown in both the presence and the absence of erythromycin to an OD650 of 0.3. Aliquots of 100 μl were removed and used to inoculate 4 ml of fresh AGCH. This procedure was repeated three times before strains were plated on blood agar plates. Colony counts in the presence and absence of erythromycin were used to evaluate the stability of each mutant before being forwarded for future testing. The S. pneumoniae 0100993 TCSTS mutants were used in all subsequent in vivo and in vitro test procedures (refer to http://www.blackwell-science.com/mmi for an overview of the mutagenesis procedures used and details of the oligonucleotide primers used in this study).

    To determine the growth rate of the S. pneumoniae 0100993 mutants, 100 μl aliquots of exponential phase cultures were diluted in 4 ml of AGCH media (1:40 dilution) and incubated at 37°C until the OD650 reached 0.29–0.31. Cultures were rediluted in 4 ml of fresh AGCH and incubated at 37°C. Bacterial growth was monitored using A650 readings at 15 min intervals, until the A650 reached ≈0.3. Cultures were rediluted for a third time and A650 readings measured until the cultures reached stationary phase (refer to http://www.blackwell-science.com/mmi for more experimental data).

    Plasmid complementation

    The promoter region of the ermAM gene was amplified by PCR to generate both an NdeI restriction site at the translation initiation codon of ermAM and a unique SmaI site upstream of the ermAM promoter. The promoter fragment was digested with NdeI/SmaI, then ligated to pDL277 to create pDLErm. The complete coding sequence of the response regulator 492rr was amplified using PCR. The primers were used to introduce an NdeI restriction site, which overlapped with the start codon of the 492rr gene, and an Sst I restriction site, which followed the translation stop codon. PCR fragments digested with Sst I and NdeI were ligated to pDLErm, similarly digested and subsequently transformed into E. coli DH5α. The resulting construct was designated pDL492. DNA sequence analysis was used to confirm that the native promoters of the TSCTS gene had been replaced with the ermAM promoter and ribosome-binding site (RBS). pDL492 was then transformed into S. pneumoniae 0100993 according to the transformation described earlier. Fresh attempts to delete the 492rr gene were then made using existing mutagenic constructs.

    Animal models

    All procedures were performed in accordance with protocols approved by the SmithKline Beecham Institutional Animal Care and Use Committee, and met or exceeded the standards of the American Association for the Accreditation of Laboratory Animal Care (AAALAC), the United States Department of Health and Human Services and all local and federal animal welfare laws.

    Bacteria for in vivo studies were prepared from frozen stocks by inoculation of tryptic soy agar (TSA) plates containing 5% sheep blood and overnight growth at 37°C in 5% CO2. Bacteria were recovered from the plates, resuspended in phosphate-buffered saline (PBS) and adjusted to A600≈0.90 for respiratory tract infections.

    Animals (male CBA/J mice weighing 14–16 g) were anaesthetized with isoflurane (3%), and 50 μl of the prepared bacterial inoculum (≈105 cfu ml−1) was administered via intranasal instillation using a pipettor. Mice were sacrificed at 48 h after inoculation by CO2 overdose. Lungs were removed aseptically, homogenized in 1 ml of PBS, and viable bacteria were enumerated by serial dilution and plate counts. All animals were provided with food and water ad libitum throughout the experiment.

    Footnotes

    1. Magainin Pharmaceuticals Inc., 5110 Campus Drive, Plymouth Meeting, PA 19462, USA.

    2. Protein Design Labs Inc., 34801 Campus Drive, Fremont, CA 94555, USA.

    Acknowledgements

    The authors would like to thank D. McDevitt, M. Zalacain, P. F. Chan, R. D. Lunsford and L. Palmer for helpful discussions and advice throughout this study. D. Jaworski and M. Wang conducted some initial cloning experiments. The SmithKline Beecham sequencing and oligonucleotide biosynthesis unit provided oligonucleotide primers and conducted sequencing reactions. We are grateful to Incyte Pharmaceuticals, Inc. for providing the genome sequence of S. pneumoniae strains 23F and R6 to SmithKline Beecham. Sequence data from S. pneumoniae type 4 was provided by The Institute of Genomic Research (http://www.tigr.org). The work was funded by DARPA grant no. N65236-97-1-5810. The content of this publication does not necessarily reflect the position or the policy of the US Government, and no official endorsement should be inferred.

    Ancillary