Of the 13 two-component signal transduction systems (TCS) identified in Streptococcus pneumoniae, two, ComDE and CiaRH, are known to affect competence for natural genetic transformation. ComD and ComE act together with the comC-encoded competence-stimulating peptide (CSP) and with ComAB, the CSP-dedicated exporter, to co-ordinate activation of genes required for differentiation to competence. Several lines of evidence suggest that the CiaRH TCS and competence regulation are interconnected, including the observation that inactivation of the CiaR response regulator derepresses competence. However, the nature of the interconnection remains poorly understood. Interpretation of previous transcriptome analyses of ciaR mutants was complicated by competence derepression in the mutants. To circumvent this problem, we have used microarray analysis to investigate the transition from non-competence to competence in a comC-null wild-type strain and its ciaR derivative after the addition of CSP. This study increased the number of known CSP-induced genes from ≈ 47 to 105 and revealed ≈ 42 genes with reduced expression in competent cells. Induction of the CiaR regulon, as well as the entire HrcA and part of the CtsR stress response regulons, was observed in wild-type competent cells. Enhanced induction of stress response genes was detected in ciaR competent cells. In line with these observations, CSP was demonstrated to trigger growth arrest and stationary phase autolysis in ciaR cells. Taken together, these data strongly suggest that differentiation to competence imposes a temporary stress on cells, and that the CiaRH TCS is required for the cells to exit normally from the competent state.
Competence, a physiological state in which exogenous DNA can be internalized, is a prerequisite for genetic transformation in Streptococcus pneumoniae. Genetic transformation is likely to play a significant role in the lifestyle of this bacterium by favouring genetic plasticity. A survey of several parameters that govern recombination rate in S. pneumoniae suggested that many of them have been optimized in order to favour genetic exchanges (Claverys et al., 2000). Intra- and interspecies genetic exchanges are believed to help this human pathogen to adapt to its host and bypass host defences. However, this does not imply that genetic transformation is the only purpose of competence development. Therefore, unravelling the signals that trigger competence may help to answer the question of the role(s) of competence.
The core part of the S. pneumoniae competence regulatory cascade is now quite well understood (for a review, see Claverys and Håvarstein, 2002). A quorum-sensing mechanism consisting of an unmodified peptide pheromone, the comC-encoded competence-stimulating peptide (CSP) (Håvarstein et al., 1995), its secretion apparatus, ComAB (Hui et al., 1995) and a two-component regulatory system (TCS) that responds to external CSP, ComDE (Pestova et al., 1996) is used to regulate competence for natural genetic transformation. ComD, a membrane-bound histidine kinase, is the CSP receptor that activates its cognate response regulator, ComE. ComE binds specifically to a target site consisting of two 9 bp imperfect direct repeats separated by a stretch of 12 nucleotides (Ween et al., 1999).
Comparison of the kinetics of CSP-induced transcription of comCDE and of the first identified competence-regulated operon, cinA–recA–dinF (Martin et al., 1995), revealed that comCDE was first induced strongly, but briefly (about 5 min) (Alloing et al., 1998). Then, while comCDE transcription was rapidly shut off, induction of the cinA–recA–dinF operon was observed to occur. Transcriptome studies using microarrays confirmed this pattern of transcription (Peterson et al., 2000; Rimini et al., 2000) and showed that the whole competence regulon is made of two classes of com genes, early (i.e. comCDE-like) and late (i.e. cinA–recA–dinF-like). The former class probably relies on ComE∼P for expression, whereas the latter depends on the comX-encoded alternative sigma factor σX (Lee and Morrison, 1999; Luo and Morrison, 2003). Late com genes are expressed through a common regulatory element designated the cin-box or com-box (Campbell et al, 1998; Pestova and Morrison, 1998), a conserved −10 sequence (TACGAATA) representing a typical binding site for the alternative σX factor.
To understand better the early control of competence development, we set up a genetic screen for mutants that upregulate comCDE or cup (for competence up) mutants (Martin et al., 2000). Mutations were generated in a strain harbouring a comC::lacZ transcriptional fusion by in vitro mariner transposon mutagenesis of the chromosome, and cup mutants were selected on the basis of increased β-galactosidase activity on medium on which competence is normally repressed. The isolation of four independent mariner insertions in the ciaR gene provided strong evidence that CiaR affected the regulation of comCDE expression (Martin et al., 2000). CiaR is the response regulator of a TCS identified in a screen for β-lactam resistance determinants in laboratory mutants (Guenzi et al., 1994). CiaRH was the first identified among 13 TCS of S. pneumoniae (Lange et al., 1999) and was linked to competence by the finding that a ciaHT230P change likely to mimic an activated histidine kinase led to complete inhibition of competence (Guenzi et al., 1994). Although inactivation of ciaR or ciaH by plasmid insertion was reported to have no effect on competence (Guenzi et al., 1994), a careful analysis of ciaR mutants, isolated as spontaneous revertants of ciaHT230P with a restored ability to transform, revealed that these mutants were derepressed for competence (Giammarinaro et al., 1999). Two recent transcriptome analyses confirmed the induction of the competence regulon in ciaR mutant cells (Sebert et al., 2002; Mascher et al., 2003). However, because of the Cup phenotype conferred by ciaR mutations, these studies did not allow examination of the consequences of the absence of CiaR on the transition from non-competence to competence or a comparison of non-competent wild-type and ciaR cells.
Our study was undertaken with the aim of specifically investigating these aspects in the hope that it would reveal events occurring at the onset of competence when CiaR is missing. The study took advantage of microarrays designed to represent each of the annotated open reading frames (ORFs) present in S. pneumoniae strains TIGR4 (Tettelin et al., 2001) and R6 (Hoskins et al., 2001) developed through a tripartite collaboration (St Georges’ Hospital, London, Norwich and Glasgow, UK; McCluskey et al., 2004). Abolition of spontaneous competence in the wild type and its ciaR derivative was achieved through deletion of the entire CSP-encoding comC gene. Genome-wide transcription profiling was used to analyse the CSP-induced transition from non-competence to competence in wild-type and ciaR cells, and to compare the transcriptome of wild-type and ciaR non-competent cells. The possible involvement in the early control of competence development of HtrA, a stress response extracellular protease under CiaR control, and of a CiaR-binding region previously identified ≈ 1 kb upstream of comCDE (Mascher et al., 2003) were also examined. Finally, the consequences of competence development in the absence of CiaR on exponential phase growth and on stationary phase autolysis were documented.
Results and discussion
Streptococcus pneumoniae genome microarrays and experimental design
The purpose of our experiment was to evaluate the impact of the absence of the CiaR response regulator on the transition from non-competence to competence, as well as to identify genes with altered expression in non-competent ciaR cells. To eliminate the derepression of competence that characterizes ciaR mutants, a ciaR mutation was introduced into a comC0 strain. Mutation of comC (see Experimental procedures) is not expected to affect other cellular functions, unlike comA mutations, for example, which could affect the export of other double-glycine leader peptides (Claverys and Håvarstein, 2002) as well as CSP.
Cultures of the wild-type strain R1502 (comC0, ssbB::luc) and its isogenic ciaR derivative R1577 (Table 1) were grown in parallel to an OD492 of ≈ 0.045 and sampled immediately before and 10 min after the addition of CSP (500 ng ml−1) for purification of RNA (see Experimental procedures). This dose of CSP was sufficient to induce full competence in both cultures (Fig. 1). It was noticeable that, although similar competence levels were achieved with both strains, growth of ciaR cells was severely affected after CSP addition for at least 40 min (Fig. 1). This observation is discussed later. To analyse the kinetics of CSP response with the wild-type (comC0) strain under the same experimental conditions, samples were taken 5, 10, 20 and 30 min after CSP addition.
Table 1. . Bacterial strains, plasmids and primers used in this study.
Strain, plasmid or primer
Relevant genotype and description
Source or reference
. Small letters in the oligonucleotide sequence indicate 5′ nucleotide extension.
A and C refer, respectively, to the anti- and co-transcribed orientation of the minitransposon antibiotic resistance cassette with respect to the target gene.
TCAAACATATGGAGGCAAATATG; upstream from htrA; 2036386–2036408; AE007317
GCTTTCTTTACATAGATGTCAAT; downstream of htrA; 2037609–2037587; AE007317
GAAAAATAATCCACAATGTTACC; downstream of parB; 2038535–2038513; AE007317
Microarrays were designed to represent each of the annotated ORFs present in S. pneumoniae strains TIGR4 (Tettelin et al., 2001) and R6 (Hoskins et al., 2001) to create a composite, inclusive array for these two strains (McCluskey et al., 2004). The microarray thus comprises 2236 TIGR4 and 117 R6 gene-specific targets (see Experimental procedures). Gene expression profiles of competent and non-competent cells (or of non-competent wild-type and ciaR cells) were compared in two-colour microarray experiments using fluorescent probes, labelled with Cy3 and Cy5, consisting of the first-strand cDNA made from randomly primed total cellular RNA (Eisen and Brown, 1999). The relative representation of a transcript between the two RNA pools was assayed by measuring the fluorescence intensities of the two dyes at a given target on the array. Each comparison with a given RNA pair was repeated twice with the dyes swapped. Determination of fluorescence intensity, image treatment and further analysis were as described in Experimental procedures. Figure S1 in the Supplementary material shows scatter plots for comparisons between non-competent ciaR and wild-type cells (Fig.S1, top), between competent and non-competent ciaR (Fig.S1, bottom) and between competent and non-competent wild-type (Fig.S1, centre) cells.
Genome-wide expression profiling identifies new genes upregulated in wild-type competent cells
Microarray comparisons of the transcription profiles of competent and non-competent wild-type and ciaR cells revealed a number of induced genes much larger than expected (105 genes; 5.1% of the R6 genome). Genes were retained if they demonstrated a statistically significant (P ≤ 0.05) change in expression of at least twofold. With this criterion, the study increased the number of known competence-induced genes from ≈ 47 to 105.
A comparison between our list of induced genes and that of Peterson et al. (2004) (124 induced genes) revealed that 91 loci were detected in both studies. For four of the loci identified by Peterson et al. (2004), change in expression was slightly below the twofold cut-off in our study (radC, def, spr1827–1828). The remaining differences (29 loci) can be tentatively attributed to the distinct microarrays, but also to the strains used. Indeed, although both set of strains are reported to be R6 derivatives, they have different histories (Tiraby et al., 1975) and are clearly not strictly related. For example, although D39 from which R6 was derived is a type 2 strain, strain Cp contains the type 3 capsule genes as its parent was obtained by transforming R36A (the original D39 derivative) with the chromosomal DNA of the type 3 Avery's strain A66 (Pearce et al., 2002); strain Cp is also deficient for mismatch repair (Prudhomme et al., 1989). Thus, the lack of detection of the induction of malMP in the study by Peterson et al. (2003) is probably due to the presence in strain Cp of the malM511 allele, a point mutation that abolishes amylomaltase activity (Lacks, 1968). Either this mutation destabilizes the malMP mRNA or, because of the absence of amylomaltase, cytoplasmic accumulation of some substrate of the maltose permease results in repression by MalR, thus preventing competence-dependent (and CiaR-dependent; see below) induction of malMP.
Fourteen loci upregulated in our study were not detected by Peterson et al. (2004). Among these, adhE, which encodes an alcohol dehydrogenase, was CSP induced in our strains, whereas it was depressed in strain Cp (Supplementary material, TableS1). Differences in medium composition could possibly account for this striking observation.
. When more than two ORFs, only the first and the last are indicated, with total number of ORFs given between parenthesis. ORF numbering is that of R6 (Hoskins et al. , 2001). To reduce length, ‘ spr000 ’ was deleted throughout (e.g. spr0001 is now 1 , spr0025 is 25 , etc…).
. Values are indicated for each gene for clusters containing up to three genes. When there are more than three genes, values between parentheses correspond to the average for all genes.
. Average of values obtained for this ORF with two different probes (Table S1).
. Value considered as non-significant in our study.
For each gene kinetics, the maximum value is undered.
A complete list of CSP-responsive genes with relative values of the ratios for the wild-type and ciaR strains, and incorporating references to previous identifications and a comparison with data from Peterson et al. (2004) is available in the Supplementary material (TableS1). An in-depth description and analysis of the entire set of genes upregulated in S. pneumoniae competent cells is outside the scope of this study. Readers are therefore referred to the paper by Peterson et al. (2004) for a comprehensive discussion of these genes.
Depressed genes in wild-type competent cells
Microarray comparisons of the transcription profiles of competent and non-competent wild-type and ciaR cells also revealed a number of genes with transiently reduced expression (depressed genes). The criterion for selection of depressed loci was similar to that for upregulated ORFs (see above). A previous genome-wide transcriptome analysis revealed a dozen genes with reduced expression in competent cells, mainly genes encoding ribosomal proteins of both 30S and 50S subunits (Rimini et al., 2000). However, our study revealed a number of depressed genes much larger than expected (42 loci; 2.1% of the genome; TableS1). The depressed clusters comprise genes encoding products involved in lipid metabolism (spr0375 to spr387), in transport of K+ (trkH) and manganese (psaBCA), and in the F1F0 ATPase (atpAGDC) (TableS1). The reduced expression of trkH, which was previously attributed to the absence of CiaR (Sebert et al., 2002), was observed in both wild-type and ciaR competent cells.
Intriguingly, the number of differences between our list and that of Peterson et al. (2004) (64 depressed genes) is much larger than for the upregulated loci as only six loci were common to both (manLMN, glnRA, thrS; TableS1). For example, whereas the rps and rpl genes in the interval spr0187–spr0211 were classified as depressed by Peterson et al. (2003), their expression was reduced by less than twofold in our study. Conversely, clusters of depressed genes were detected only in our study (e.g. spr0375 to spr387, psaBCA and atpAGDC). The reasons for the differences are unclear. They could be attributed to the distinct microarrays, but it is more likely that they resulted from variations in medium composition and, as already discussed above, in the strains used. Thus, a readily explainable difference concerned aliA, which was found to be depressed only in our study. This gene encodes an oligopeptide-binding lipoprotein (Alloing et al., 1994). The deletion of the 5′ moiety of aliA in the Cp strain (Dillard et al., 1995; Alloing et al., 1996) readily accounts for its non-detection by Peterson et al. (2004). In any case, the poor overlap between the two lists of depressed genes suggests that the phenomenon is more dependent on the experimental conditions than on competence development per se.
Induction of stress response genes in wild-type competent cells
Microarray analysis of the wild-type strain during competence induction revealed strong induction of genes that classically make up the bacterial heat shock regulons, groESL, dnaK–grpE–dnaJ, clpL and htrA (Table 2). The GroEL/GroES chaperone system and the DnaK chaperone with its co-chaperones DnaJ and GrpE form a functional network, which also assists the folding of newly synthesized proteins in eubacteria (Hartl and Hayer-Hartl, 2002). ClpL is an ATPase belonging to the family of Clp/HSP100 proteins which, together with the proteolytic subunit ClpP, constitutes the multimeric Clp protease (Schirmer et al., 1996). Clp-mediated proteolysis not only removes abnormal proteins that accumulate during stress conditions, but can also control the half-life of regulatory proteins (Gottesman and Maurizi, 1992). HtrA is a predicted surface-expressed serine protease belonging to the trypsin family (Sebert et al., 2002). Members of the HtrA family have been implicated in the degradation of denatured surface proteins and are generally important for stress resistance (Pallen and Wren, 1997).
Among the set of stress response genes induced at competence, only two had been identified previously as upregulated at competence, dnaK and clpL. Induction of the former (to 5.6-fold) was reported to be delayed until 15 min after the addition of CSP in an experiment in which the late com genes were induced at 10 min (Rimini et al., 2000). In another study, its expression was considered to be unaffected (Peterson et al., 2000). The clpL gene was (under the name clpB) recently classified as a competence-related cia-regulated gene by Mascher et al. (2003), with no indication of the expression range.
The induced stress response genes are known to be regulated by two different proteins, HrcA (Naberhaus, 1999) and CtsR (Derréet al., 1999; for a review, see Schumann et al., 2002). HrcA binding sites have been identified immediately upstream of the hrcA–grpE–dnaK–dnaJ and groESL operons. The groESL operon is also controlled by CtsR which, in addition, controls five other genes, clpL, clpE, clpP and ctsR–clpC (Chastanet et al., 2001; Kim et al., 2001). Among members of the CtsR regulon, only groESL and clpL were induced 10 min after the addition of CSP, whereas clpE possibly displayed late induction (Tables 2 and 3). Other known stress response genes, such as clpX (which encodes another ATPase subunit of the Clp protease) and ftsH (which encodes a single-chain AAA protease), the regulation of which is not known in the model Gram-positive organism Bacillus subtilis, were not induced at 10 min. However, the latter showed slight induction at 20 min (Table 3). The observation that stress response genes differed in induction not only between regulons but also within the same regulon argues strongly in favour of the hypothesis that different stress signals are produced during differentiation to competence.
Interestingly, induction of the stress response genes, hrcA–grpE–dnaK–dnaJ, groESL and possibly clpL (Table 2), was significantly enhanced in ciaR cells, with the exception of the htrA–parB operon (Table 2), which is discussed in the next section. This observation suggested that ciaR competent cells suffered a more intense stress than the wild type, an interpretation consistent with the observation that CSP induces growth arrest in ciaR cultures (Fig. 1). We therefore propose that the CiaRH TCS is required for the cells to cope with competence-induced stress.
Induction of the CiaR regulon in wild-type competent cells
Together with ciaRH, spr0931, axe1, malMP and dltABCD, htrA–parB constitute a group of 12 genes located at six clusters sharing the unique property of being upregulated in competent wild type while remaining essentially unchanged in ciaR competent cells (Table 4). The observation that these loci required CiaR for induction and that they were maximally induced 10 min after CSP addition (Table 3) suggested that they belong to the CiaR minimal regulon. This is consistent with a previous finding that restriction fragments overlapping the control region of four of them (axe1, malMP, dltABCD and htrA–parB) were bound by the CiaR protein (Mascher et al., 2003).
Table 4. . Changes in gene expression resulting from the absence of CiaR at competence.
For each gene, the maximum value is shown in bold.
Phosphotransferase system, mannose-specific EII
? upstr. cbr8
Conserved hypothetical protein
spr1768-1769 ( cyl locus)
CylM protein, cytolytic toxin system
Amylomaltase; maltose transport
d -alanylation of lipoteichoic acid
Serine protease; chromosome segregation protein
lic (5 genes)
Involved inteichoic acid metabolism
TCS03 histidine kinase and response regulator
Interestingly, the ciaRH operon itself belongs to this group. Its specific induction at competence was not reported in previous transcriptome analyses (Sebert et al., 2002; Mascher et al., 2003). Its identification in this study validates our use of an approach based on the analysis of the transition from non-competence to competence. Induction of ciaRH in competent cells had been reported only by Rimini et al. (2000), who identified it as one of three spots still showing increased expression 30 min after CSP addition. In our study, ciaRH showed maximum induction at 10 min (Table 3). Despite the failure to identify a CiaR-binding region (CBR) overlapping the control region of ciaRH (Mascher et al., 2003), there is little doubt that CiaR directly controls the expression of its operon. spr0931 is, with spr0782, another candidate member of the CiaR regulon for which no CBR could be identified (Mascher et al., 2003). No induction of spr0782 or of the lic operon (Table 4) was detected in either our study or the study by Peterson et al. (2004). The failure to identify the lic cluster is not surprising as its expression was known to occur at the end of the exponential phase in the wild-type strain R6 (Mascher et al., 2003), and our samples were collected from cultures at 0.06 OD492 (see Experimental procedures). Together, these data suggest that CiaR is necessary but not sufficient for induction of the lic operon. A similar interpretation could account for the apparent lack of induction of spr0782 (Table 4).
Kinetics of induction of ciaRH and of candidate members of the CiaR regulon were similar (Table 3), suggesting that phosphorylation of CiaR rather than an increase in CiaR concentration is responsible for concomitant induction of the entire regulon.
Changes in expression in competent cells resulting from the absence of CiaR
Two other clusters with previously identified CBR exhibited altered expression in ciaR cells, manLMN and spr1768–spr1769 at the cyl locus (Table 4). Both clusters were reported to be upregulated in ciaR cells (Mascher et al., 2003). Interestingly, our data revealed that the term ‘upregulation’ covered widely different situations, while remaining consistent with the conclusion that CiaR is acting as a repressor at these loci (Mascher et al., 2003). Thus, only the cyl locus was really induced in ciaR but not in wild-type competent cells (Table 4), whereas the manLMN locus was repressed at competence in the wild type but remained unchanged in ciaR cells (Table 4).
Another TCS, hk03–rr03, showed slight induction (less than twofold) in wild type but was significantly upregulated in ciaR competent cells (Table 4). It is possible that some of the effects attributed to the lack of CiaR are indirectly mediated through activation of TCS03. Response regulator 03 could be responsible for altered gene expression in the absence of CiaR by interacting directly with some of its own targets.
Apart from modification in expression of CiaR-dependent loci and the increased induction of hk03–rr03 and of stress response genes, no difference in competence-regulated genes was apparent between the ciaR and wild-type cells (TableS1 and Table 2), consistent with another report (Mascher et al., 2003). This clearly indicated that the CiaRH TCS does not directly interfere with the central part of the competence regulatory circuit (i.e. ComE- or ComX-dependent induction).
Changes in expression resulting from the absence of CiaR in non-competent cells
Only ciaH, spr0931 and htrA appeared as depressed (0.5, 0.5 and 0.7-fold reduction respectively) when wild-type and ciaR non-competent cells were compared (data not shown). The other loci previously concluded to be depressed in ciaR cells (spr0931, licB, malP; Sebert et al., 2002; Mascher et al., 2003) were not seen to be so in our study. We propose that the severe downregulation reported previously for htrA (27-fold) and parB (spo0J; 11.5-fold) (Sebert et al., 2002) resulted from the comparison of competent ciaR and wild-type cells. Indeed, the extent of downregulation for htrA is much higher when calculated for competent (Table 2) than for non-competent cells (data not shown).
No locus was upregulated more than twofold in ciaR non-competent cells, except ciaR (2.3-fold increase). This increase in expression is readily explained by the presence of cassette-specific transcripts extending into the ciaR region.
CiaR and the early control of competence induction
The Cup phenotype of ciaR mutants indicated that the CiaRH TCS is involved in the early control of competence induction (see Introduction). HtrA was suggested as a potential participant in this control system (Sebert et al., 2002). Our finding that htrA was depressed in non-competent ciaR cells (see above) lends some weight to this proposal. It could do so either directly by degrading the secreted CSP at the surface of the cell or through some other functional connection (Sebert et al., 2002). To check whether inactivation of htrA leads to competence derepression, the gene was inactivated by in vitro mariner mutagenesis (see Experimental procedures). Spontaneous competence development was then analysed in parallel cultures of an htrA mutant, a ciaR mutant and the wild-type parent, grown under acid conditions (Fig. 2). Spontaneous competence induction is strongly dependent on the initial pH of the culture medium. For instance, initial pH values between 6.8 and 8.0 affected the timing of the occurrence and the level of competence (Chen and Morrison, 1987). Figure 2 shows that, as expected, the ciaR strain develops competence early in the exponential phase of growth at two suboptimal pH values, whereas competence development of the wild type is diminished, and that of the htrA mutant is even more so. We conclude that inactivation of htrA does not upregulate competence (Fig. 2). It is therefore unlikely that the downregulation of htrA in ciaR cells is responsible for the Cup phenotype of ciaR mutants.
Because CiaR bound a restriction fragment partly overlapping orfL and hrtA, it was suggested recently that CiaR binding in this region less than 1 kb upstream of comCDE (cbr18; Fig. 2) could interfere with ComE binding to the comCDE promoter (Mascher et al., 2003). A deletion of the putative CiaR binding site located in cbr18 was obtained (see Experimental procedures), and its effect on spontaneous competence development was examined. The deletion did not derepress competence, but rather reduced it (Fig. 2), strongly suggesting that the binding of CiaR to the orfL–htrA intergenic region is not required for repression of competence.
These data suggest that neither the downregulation of htrA in ciaR cells nor the binding of CiaR to cbr18 can account for the Cup phenotype of ciaR mutants. It remains possible that the CiaR protein directly regulates comAB rather than comCDE, as it has been shown that upregulation of comAB is sufficient to confer a Cup phenotype (Martin et al., 2000). However, transcriptome analysis did not reveal any change in comAB expression in ciaR non-competent cells (data not shown). Indirect effects of the CiaRH TCS on competence should also be considered. Competence induction might be affected because activation of CiaRH results in alteration of surface properties that interferes with CSP production or recognition (Claverys and Håvarstein, 2002; Mascher et al., 2003). Alternatively, the steady-state concentration of some key competence regulatory protein could be affected through long-term secondary effects of the inactivation of ciaR.
Competence induction, CiaR and the control of autolysis in S. pneumoniae
Transcriptome analysis indicated that CSP was triggering the induction of several stress response genes, a possible consequence of the severe physiological changes associated with differentiation to competence (Table 2). Recent experiments to document the spontaneous release of chromosomal DNA into the medium have suggested that one of these changes is autolysis (Steinmoen et al., 2002; M. Moscoso and J. P. Claverys, in preparation). Thus, the amount of DNA released from cells grown in competence-permissive C medium was found to be much higher than that from cells in competence non-permissive acidic C medium (Fig. 3). Noticeably, DNA release did not stop when competence declined. Instead, the amount of released DNA in the competent culture reached a maximum in stationary phase (Fig. 3), suggesting that competence could trigger delayed autolysis. As ciaR cells released significantly higher amounts of DNA than the wild type (our unpublished observations), we hypothesized that a ciaR strain could be particularly prone to autolysis. Growth curves of a wild-type strain and its ciaR derivative in competence-permissive medium indeed suggested a direct connection between competence and stationary phase autolysis (Fig. 4).
To demonstrate that competence was responsible for stationary phase autolysis, a comC0 wild-type strain and its ciaR derivative were grown in C medium, and half of each culture received CSP after 100 min. In the absence of CSP, the two strains exhibited very similar growth curves. Autolysis was initiated in both cases only ≈ 5 h after the cells entered the stationary phase and was completed some 60 min later (Fig. 5A). No noticeable change was induced by CSP in the wild-type culture (Fig. 5B). In contrast, the addition of CSP had dramatic effects on the ciaR strain (Fig. 5C and D). As shown in Fig. 5D, about 18 min after CSP addition, growth was greatly reduced for ≈ 60 min (from OD492 of 0.21 to 0.25), then resumed at a reasonable rate for 80 min before autolysis was initiated at OD492 of 0.49. Note that a maximum OD492 of 0.72 was obtained in the non-competent culture, as well as in both wild-type cultures. Autolysis was completed 140 min later. These data demonstrate that CSP induces post-competence stationary phase autolysis in ciaR cultures and suggest that the CiaRH TCS is required to cope with the physiological changes induced during differentiation to competence and for normal exit from competence.
It is noteworthy that growth arrest at competence was observed for ciaR cells in the experiment in Fig. 5C and D (and in Fig. 1), but not that in Fig. 4. We tentatively attribute the variation in the intensity of the effects of competence to the fact that in the latter experiment cells were comC+. As ciaR cells are derepressed for competence, they had to adapt in the preculture to ‘steady-state’ competence compatible with growth, whereas comC0 cells were completely ‘naïve’ with respect to competence induction before CSP addition.
Intriguingly, the growth rate of a ciaR mutant but also of an rr04 mutant was reported previously by Lange et al. (1999) to decline significantly in mid-exponential phase. The latter exhibited a 3-h-long growth arrest (Lange et al., 1999). It is tempting to speculate that this unusual growth delay in mid-exponential phase was also consecutive to competence induction. The HK04–RR04 TCS could be also required for normal exit from competence.
Further discussion – regulation of stress responses in S. pneumoniae
Induction of the heat shock regulon determined by microarray analysis was reported previously for a ΔclpP mutant (Robertson et al., 2002). Upregulation of the competence regulon was also detected in this study, in complete agreement with a previous report that inactivation of clpP conferred a Cup phenotype (Chastanet et al., 2001). However, a number of differences can be seen. Thus, manN was induced; among the mal genes, malR, which encodes the maltose repressor, was also induced but not malMP; and htrA was downregulated (Robertson et al., 2002). These differences suggested that the CiaRH TCS was not activated in competent ΔclpP cells and, indeed, induction of ciaRH was not detected (Robertson et al., 2002). These differences argue strongly in favour of the occurrence of clpP-specific defects that could be related to the alteration of the competence-induced stress response in the absence of ClpP, a key component of the stress response network.
Significance of stress response induction in competent cells
As several chromosomal mutations (i.e. alterations of cell wall, lipid membrane composition or purine pools and perturbation of stress response; Claverys and Håvarstein, 2002) leading to upregulation of the comCDE competence control operon could be viewed as mimicking stress signals, it was suggested that competence could be the S. pneumoniae substitute for the SOS system and that the CSP could act as an alarmone (Martin et al., 2000; Claverys and Håvarstein, 2002). The observation that stress regulons are induced after CSP addition (Table 2) would fit this interpretation. The competence-induced growth arrest detected in ciaR cultures (and sometimes also in wild-type cultures, see Fig. 5B; our unpublished observations) would be analogous to the situation in Escherichia coli in which, after SOS induction, septation is transiently blocked to prevent untimely cell division and resumes when the SOS-inducing signal disappears (Huisman et al., 1984).
Alternatively, or in addition, growth arrest could be the consequence of physiological changes accompanying the differentiation to competence. Thus, cell surface changes have been reported to occur at competence (Tomasz and Zanati, 1971). Such alterations could generate stress signals and subsequent cell division arrest until the physiological alterations disappear.
CiaR, competence and autolysis
Several lines of evidence suggest that competent cells are more prone to autolysis that non-competent cells: (i) competence was shown to activate protoplast formation (Seto and Tomasz, 1975) and, conversely, non-competent mutants were reported to form protoplasts less readily (Lacks and Neuberger, 1975); (ii) there is a tendency of cells collected soon after competence to lyse (Morrison and Baker, 1979; our unpublished observations); and (iii) chromosomal DNA release in competent cultures of S. pneumoniae is, at least in part, dependent on the presence of the major autolytic enzyme, LytA (Steinmoen et al., 2002; M. Moscoso and J. P. Claverys, manuscript in preparation). Nevertheless, contrary to a previous report (Trombe et al., 1992), competent cells of the wild-type strain were not found to undergo rapid autolysis in stationary phase (Fig. 5B). In addition, investigation of the behaviour of our wild-type strain and of strain Cp in CAT medium (the strain and medium originally used by Trombe et al., 1992) did not reveal competence-associated fast stationary phase autolysis (data not shown). It is possible that autolysis was triggered in Trombe and co-workers’ experiment because of the use as competence activator of culture supernatants that could contain autolysin in addition to CSP.
In our study, fast autolysis of competent cells was only observed for ciaR competent cells (Fig. 5C and D). The existence of a connection between the CiaRH TCS and autolysis was already suggested by the observation that the R6ciaHC306 mutant (which mimics activation of the CiaRH TCS) was resistant to sodium deoxycholate-induced autolysis and to treatments that induce protoplast formation (Giammarinaro et al., 1999). Also, a ciaR strain was reported to display an increased rate of autolysis in the stationary phase (Lange et al., 1999). This was attributed to constitutive expression of the autolytic enzyme LytA in the ciaR mutant, independent of competence (Lange et al., 1999). In contrast, our data clearly establish a direct connection between competence and autolysis, as CSP induces lysis of ciaR cells in stationary phase (Fig. 5C and D).
It is unlikely that competence-induced transcription of lytA and lytC is responsible for CSP-induced stationary phase autolysis. First, the previously reported 3.5-fold induction of lytC (under the name cbp6; Rimini et al., 2000) was not confirmed in our transcriptome analysis or in that of Peterson et al. (2004). Secondly, a strain engineered in order to prevent CSP-induced expression of lytA showed no difference in its ability to release DNA in competent cultures (M. Moscoso and J. P. Claverys, in preparation). Therefore, autolysis is probably caused by alteration(s) in the cell wall and/or the cytoplasmic membrane accompanying the differentiation to competence (Tomasz and Zanati, 1971). Such alterations could be partly responsible for the induction of stress response genes detected in transcriptome analysis (Table 2). Attenuation of cell envelope modifications during post-competence growth would be essential to prevent stationary phase autolysis. We propose that the CiaRH TCS plays a critical role at this stage and is therefore required for the cells to exit normally from competence.
Our observations readily account for the fact that, although mutation of ciaRH has been obtained in the laboratory during a search for mutants with increased resistance to β-lactam antibiotics, ciaRH mutants were never observed among natural clinical isolates resistant to β-lactams (Hakenbeck et al., 1999). ciaRH mutants would probably be unable to survive under the stress-generating and competence-inducing conditions in the host.
Although wild-type cells are more resistant to post-competence autolysis than ciaR cells, they can nevertheless experience limited autolysis, as suggested by DNA release experiments (Fig. 2; Steinmoen et al., 2002; M. Moscoso and J. P. Claverys, in preparation). This competence-triggered tendency to lyse could contribute to the virulence of S. pneumoniae through the release of virulence factors such as peptidoglycan, teichoic acid and lipoteichoic acid (Nau and Eiffert, 2002), pneumolysin or chromosomal DNA itself (Chatellier and Kotb, 2000). Thus, competence might serve not only to allow genetic plasticity through transformation but also to increase virulence whenever cells encounter environmental conditions in the host that do not permit normal exit from competence and so provoke autolysis.
Bacterial strains, growth and transformation conditions
Streptococcus pneumoniae strains and plasmids used in this study are described in Table 1 . For induction of competence, cells grown in Todd–Hewitt + Y, in CAT or in C + Y (at the indicated pH) medium ( Bergéet al., 2002 ) to OD 550 of 0.4, then centrifuged, resuspended in fresh C medium containing 15% glycerol and kept frozen at −70°C, were gently thawed and inoculated with a 100-fold dilution in C + Y medium. For experiments with comC 0 strains, synthetic CSP1 (500 ng ml −1 ) was added at the indicated time. Monitoring of competence was achieved using a transcriptional fusion between a gene specifically induced at competence ( comC or ssbB ; Table 1 ) and the Photinus pyralis luc gene, which encodes the firefly luciferase, as described previously ( Bergéet al., 2002 ). A direct correlation between luciferase activity and transformation level was observed under widely varying levels of competence ( Bergéet al., 2002 ). For transformation, after the addition of DNA, cells were incubated at 30°C. Transformants were selected by plating on CAT agar supplemented with 4% horse blood, followed by challenge with a 10 ml overlay containing the appropriate antibiotic concentration, after phenotypic expression for 120 min at 37°C. Antibiotic concentrations used for selection of transformants were: chloramphenicol (Cm), 4.5 µg ml −1 ; erythromycin (Ery), 0.05–0.2 µg ml −1 ; kanamycin (Kan), 250 µg ml −1 ; spectinomycin (Spc), 100 µg ml −1 ; streptomycin (Sm), 200 µg ml −1 ; tetracycline (Tc), 1 µg ml −1 .
Construction of comC0, htrA::spc and ΔorfL–htrA strains
A complete deletion of comC0 was generated by fusing the ATG of comC with the comD ORF. This was achieved by taking advantage of a previously constructed strain, R1036 (Table 1), in which the comC gene was substituted with Janus (Sung et al., 2001). Janus is a kan–rpsL+ cassette that confers dominant SmS in a SmR background. Replacement of Janus by an arbitrary segment of DNA during a second transformation restores SmR (and KanS) (Sung et al., 2001). Strain R1036 was therefore transformed with PCRcomC0, a polymerase chain reaction (PCR) fragment overlapping the comC::Janus segment in which the ATG of comC was fused with the comD ORF. This fragment was generated in two steps: first using BM47–BM113 (comC upstream region) and BM54–BM114 (comC downstream region) primer pairs (Table 1) and R800 chromosomal DNA as a template and, secondly, BM54 and BM47 primers and a mixture of the two products from the first reaction as template. SmR transformants were selected, and their comC region was sequenced to obtain strain R1500. A SmS derivative, R1501, was then constructed by transforming strain R1029 (Table 1) with R1500 DNA and screening for a KanS transformant by replica plating.
Insertions of the spc (SpcR) gene cassette in htrA were generated by in vitro mariner mutagenesis as described previously (Martin et al., 2000). Briefly, the plasmid used as a source for the 1145 bp spc (SpcR) mariner mini-transposon was pR412 (Table 1). Plasmid DNA (≈ 1 µg) was incubated with an htrA PCR fragment (≈ 1 µg) generated with primers BM40 and SphtrA6 (Table 1), in the presence of purified Himar1 transposase (Lampe et al., 1996), in a total volume of 40 µl leading to random insertion of the mini-transposon within the fragment. Gaps in transposition products were repaired as described previously (Akerley et al., 1998), and the resulting in vitro-generated transposon insertion library was used to transform S. pneumoniae. Location and orientation of the mariner cassette was determined through PCRs using primers MP127 or MP128 (Table 1) in combination with either of the two primers used to generate the htrA PCR fragment. Cassette–chromosome junctions were sequenced using the Thermo Sequenase cycle sequencing kit (USB) and primer MP128. Strain R1452, which carries the htrA::spc5Cmariner insertion (position 615 with respect to the ATG of htrA), was constructed by transformation of strain R825 with a PCR fragment generated with the SphtrA1 and SphtrA3 primer pair (Table 1) to eliminate any mutation that would be located in htrA flanking regions.
A deletion extending from within orfL (position 2035810 on the R6 genome; Hoskins et al., 2001) to within htrA (position 2037205) was obtained through selection of a spontaneous insertion-deletion mutation (Prudhomme et al., 2002), after transformation of the wild-type strain R800 with pXF520 plasmid DNA. This non-replicative plasmid was designed to fuse lacZ to comC while maintaining a fully functional copy of the comCDE region downstream of the fusion, and confers CmR (Pestova et al., 1996). After transformation with pXF520, illegitimate recombination events sometimes occur that lead to simultaneous insertion of vector sequences and deletion of chromosomal sequences flanking the pneumococcal insert in pXF520 on either side (Alloing et al., 1998; Prudhomme et al., 2002). CmR transformants unable to express lacZ (white colonies on transformation permissive medium; Martin et al., 2000) were selected and analysed by PCR as described previously (Prudhomme et al., 2002). A candidate clone was retained, and the limits of the deletion were sequenced. The orfL–htrA (2035810–2037205) deletion and concomitant insertion of a 1.68-kb-long segment from pXF520 containing the cat gene was introduced into strain R825 to generate strain R1510 (Table 1).
Construction of the S. pneumoniae DNA microarray
DNA fragments of individual ORFs present in S. pneumoniae strain TIGR4 (also known as JNR7/87) (Tettelin et al., 2001) and the R6 strain-specific genes (Hoskins et al., 2001) were amplified using ORF-specific primers (MWG Biotech). The primers (22-mers) were designed using the approach of Hinds et al. (2002) with a predicted melting temperature of 60°C (Tm 60). In total, 2236 TIGR4 and 117 R6 gene-specific probes were amplified by PCR for spotting on the microarray.All PCRs were carried out using HotStart Taq (Qiagen) and a Primus-HT PCR machine (MWGAG Biotech) using the following parameters: 95°C for 15 min, followed by 30 cycles of 95°C for 30 s, 50°C for 30 s, 72°C for 120 s, and a final incubation at 72°C for 600 s. Genomic TIGR4 and R6 DNA (2.5 ng) was used as template, and ≈ 100 pmol of each primer in a reaction volume of 50 µl. For each primer pair, the successful PCR amplification of a single DNA fragment of the correct length was checked by electrophoresis in agarose gels. In the initial high-throughput PCR, a small percentage of reactions (≈ 5%) were either unsuccessful or gave poor product yield. However, these missing probes were obtained subsequently by optimizing the PCR conditions or by designing new primers. Ten probes encoding ORFs from Campylobacter jejuni with no significant DNA sequence homology to S. pneumoniae were also included on the array as negative control features. All PCR products were purified using SigmaSpin 2TM Post-Reaction Clean-Up plates (Sigma), dried and resuspended in a half volume of 3× SSC containing 0.01% sarkosyl before spotting on the microarray. PCR products were spotted on GAPS II-coated slides (Corning) using an in-house Stanford-designed arrayer (see http://cmgm.stanford.edu/pbrown/mguide/index.html for associated software and protocols). Each glass slide contains two arrays each with two probes (or features) for each gene.
Preparation of probes, hybridization and data analysis
Total RNA was extracted from cells that were collected by centrifugation and frozen in liquid nitrogen. Frozen pellets were mechanically disrupted in a microDismembrator (Braun), then frozen cell powder was resuspended in RLT buffer from the Qiagen RNeasy mini kit, and further RNA purification was performed using this kit. Fluorescent probes were prepared by reverse transcription of 5–10 µg of total RNA and purified on the columns, using the CyScribe kit for direct labelling (Amersham).
For hybridization, the protocol developed at the Bacterial Microarray Group from St George's Hospital Medical School (Stewart et al., 2002) was used. For each RNA sample, two RT reactions were performed, using different dyes. Each comparison with a given RNA pair was repeated twice with the dyes swapped. The intensity of fluorescence of the microarrays was determined with a laser scanner (Axon), and the genepix software was used for image treatment. Background correction and further analysis of raw data were made using bioplot and bioclast scripts developed at the Plateforme Transcriptome in Toulouse (http://bio71.gba.insa-tlse.fr). Signal normalization with the all spots’ mean was used. A t-test P-value of mutant versus wild-type (or competent versus non-competent) log10 (relative fluorescence) of < 0.05 was considered to be statistically significant. The normalized data for the complete gene sets together with the calculated ratios are also available (contact J. P. Claverys).
The authors wish to thank Dave Lane for critical reading of the manuscript, and Chantal Granadel for expert technical assistance, Jason Hinds and the Bacterial Microarray Group at St George's Hospital Medical School (London) for providing detailed protocols for the preparation of probes and hybridizations, and Sergeï Sokol from the Plateforme Transcriptome (Toulouse) for help with microarray data analysis. Miriam Moscoso was the recipient of a Marie-Curie Individual Fellowship (QLK2-CT-1999-51509). This research was financed in part by the European Union (grant QLRK 2000-00543) and by the Programme de Génomique Microbienne du Génopôle de Toulouse.
Fig. S1. Scatter diagrams for data from non-competent ciaR versus non-competent wild type (top), competent wild type versus non-competent wild type (centre) and competent ciaR versus non-competent ciaR (bottom). The lines of best fit are shown, together with lines that denote the twofold limits.