Proteic toxin–antitoxin (TA) loci were first identified in bacterial plasmids, and they were regarded as involved in stable plasmid maintenance by a so-called ‘addiction’ mechanism. Later, chromosomally encoded TA loci were identified and their function ascribed to survival mechanisms when bacteria were subjected to stress. In the search for chromosomally encoded TA loci in Gram-positive bacteria, we identified various in the pathogen Streptococcus pneumoniae. Two of these cassettes, sharing homology with the Escherichia coli relBE locus were cloned and tested for their activity. The relBE2Spn locus resulted to be a bona fide TA locus. The toxin exhibited high toxicity towards E. coli and S. pneumoniae, although in the latter, the chromosomal copy of the antitoxin relB2Spn gene had to be inactivated to detect full toxicity. Cell growth arrest caused by expression of the relE2Spn toxin gene could be reverted by expression of the cognate antitoxin, relB2Spn, although prolonged exposition to the toxin led to cell death. The pneumococcal relBE2Spn locus is the first instance of a chromosomally encoded TA system from Gram-positive bacteria characterized in its own host. We have developed a bioluminescence resonance energy transfer (BRET) assay to detect the interactions between the RelB2Spn antitoxin and the RelE2Spn toxin in vivo. This technique has shown to be amenable to a high-throughput screening (HTS), opening new avenues in the search of molecules with potential antibacterial activity able to inhibit TA interactions.
Chromosomally encoded toxin–antitoxin (TA) loci constituted by a protein pair exist in many microorganisms, including Gram-negative (G–) and Gram-positive (G+) bacteria and Archaea (Gerdes, 2000). Leaving aside the systems that are encoded by bacterial plasmids, which are considered as mechanisms to ensure plasmid maintenance (Jensen and Gerdes, 1995), recent biocomputing analysis showed that, depending upon the life-stile of the bacteria, the number of chromosomal TA loci in a given species may vary from none to more than 40 (Pandey and Gerdes, 2005). Among this profusion of chromosomal TA loci, those from Escherichia coli K-12 have been the most widely studied, especially the relBE locus. In E. coli, two other homologues of relBE have been described, namely relBE-2 (or yefM–yoeB), and relBE-3 (or dinJ–yafQ; Gerdes, 2000; Grady and Hayes, 2003; Cherny and Gazit, 2004; Pandey and Gerdes, 2005). In addition there are two other TA loci in E. coli K-12 called mazEF (chpA) and chpBK (chromosomal homologue of plasmid-encoded genes; Masuda et al., 1993). Cytotoxins MazF and ChpB/K (MazF-2) both belong to a toxin family that is distinct from the RelE family (Gerdes, 2000) but very common in bacteria (Pandey and Gerdes, 2005). By contrast, relBE loci are very common both among Bacteria and Archaea (Gotfredsen and Gerdes, 1998; Pandey and Gerdes, 2005).
The relBE gene pair of E. coli has the regulatory components and genetic organization of a typical TA operon (Jensen and Gerdes, 1995). The antitoxin gene (relB) is located upstream the toxin gene (relE) and both are translated from one mRNA transcript. The antitoxin (RelB) counteracts RelE toxic activity by direct protein–protein interaction (Gotfredsen and Gerdes, 1998; Galvani et al., 2001). Furthermore, RelB also represses relBE transcription, RelE acting as a corepressor (Gotfredsen and Gerdes, 1998). While RelE is stable, RelB is degraded by Lon proteases and its presence in the cell therefore requires its continuous synthesis (Christensen et al., 2001). Under conditions of balanced growth, the expression of the operon is minimal, although RelB is more abundant than RelE mainly because transcription and initiation of translation of the antitoxin are stronger than that of the toxin (Gotfredsen and Gerdes, 1998). Under stress conditions, like amino acid starvation, de novo synthesis of RelB diminishes and the degradation of the protein is also enhanced because of the activation of cellular proteases. As a consequence, the concentration of the antitoxin will be reduced and free RelE molecules will appear in the cytoplasm. Thus, the toxicity of RelE is neutralized by RelB as long as protein synthesis proceeds normally, whereas RelE is activated when protein synthesis is impaired (Christensen et al., 2001). The mechanism of action of RelE has remained unknown until recently. Induced over-expression of relE indicated that the toxin inhibited cell growth and reduced the fraction of cells that were able to colony formation on agar plates (Christensen et al., 2001). Furthermore, when transcription of relE was blocked and relB was over-expressed, the bacteria rapidly regained their ability to make proteins and form colonies (Pedersen et al., 2002). Inhibition of protein synthesis by RelE was shown to be due to cleavage of mRNA codons in the ribosomal A site in a codon-specific manner, with preference for the stop codon UAG. Furthermore, recovery of protein synthesis in vivo after inhibition by RelE was mediated by tmRNA, suggesting that they can be part of a mechanism that minimizes the delay before bacteria can resume rapid growth after the end of starvation period (Pedersen et al., 2003). Consequently, it was assumed that, over a time window, inhibition of protein synthesis by RelE resulted in cell stasis rather than in cell death. Thus, the physiological role of the relBE operon would be to induce a condition of growth arrest that could modulate the global rates of macromolecular synthesis during nutritional stress (Pedersen et al., 2002). However, other chromosomal TA systems, like MazEF, which also triggers cell growth arrest by cleaving mRNA have been considered as involved in programmed cell death (Engelberg-Kulka and Glaser, 1999; Amitai et al., 2004; Engelberg-Kulka et al., 2004).
In general, all the TA systems can be considered as targets for antibiotics (Engelberg-Kulka et al., 2004), which make them a valuable tool when searching for new ways to deal with bacterial infections due to the need of novel approaches to develop new drugs (Carey, 2004). One of the worldwide important pathogens is the G+ bacterium Streptococcus pneumoniae (pneumococcus), because it is responsible for over one million human deaths per year, being an important threat to children and immunologically compromised patients (Bartlett et al., 1998). The appearance of penicillin-resistant strains has contributed to an increased importance of this pathogen (Baquero, 1996; Pallares et al., 2000; Walsh and Amyes, 2004). Several clinical approaches, including vaccination, have been proposed, although the effectiveness of these approaches is arguable (Shann, 1990; Alonso de Velasco et al., 1995; Whitney, 2000). Seeking for ways to tackle pneumococcal infections, we searched for and found various putative TA cassettes in the chromosome of this bacterium. Two of them, sharing homology with the relBE systems of E. coli, were detected in the chromosome of non-capsulated, non-pathogenic strain R6 (Hoskins et al., 2001), but also in the capsulated, pathogenic Tigr4 strain (Tettelin et al., 2001). We termed these two putative loci relBE1Spn and relBE2Spn, and set out to their characterization.
Cloning two DNA regions encompassing the toxin genes of both operons allowed us to demonstrate that only RelE2Spn exhibited toxicity towards E. coli, whereas RelE1Spn was harmless, making the former a bona fide candidate to further analyses. Here we present a detailed genetic and functional characterization of the relBE2Spn operon, and demonstrate that both genes relB2Spn and relE2Spn are organized into a single operon. Over-expression of relE2Spn both in S. pneumoniae and E. coli led to cell growth arrest, although in the G+ host it was needed to inactivate the gene encoding the antitoxin to achieve full toxicity, showing that the presence of even a single copy of the relB2Spn antitoxin gene was enough to partially counteract the arrest in cell growth provoked by the cognate toxin. In the case of E. coli, and as a difference with the E. coli RelE toxin, over-expression of the pneumococcal toxin during long periods of time led to cell death, as judged by the loss of cell viability and the inability of cell growth recovery in E. coli cultures even after the induction of the antitoxin synthesis. In addition, bioluminescence resonance energy transfer (BRET) assays were used to assess direct protein–protein interactions between the toxin and the antitoxin in vivo and, indeed, BRET signals could be detected. We believe that this is the first instance of a chromosomally encoded TA system from a G+ bacteria ever studied.
Location of the relBE2Spn locus
Two TA loci, termed relBE1Spn and relBE2Spn, homologous to the relBE locus of E. coli K-12 were identified in the chromosome of S. pneumoniae (Gerdes, 2000). DNA fragments encoding the putative toxin genes, relE1Spn and relE2Spn, were generated by polymerase chain reaction (PCR) and cloned into the E. coli expression vector pNM220 (Gotfredsen and Gerdes, 1998). Regulated transcription of both genes showed that only relE2Spn was toxic to E. coli (see below) whereas transcription of relE1Spn was innocuous (K. Gerdes, unpublished). Homology molecular modelling of both putative toxins on the three-dimensional structure of RelE from Pyrococcus horikoshii (Takagi et al., 2005) was performed. Whereas RelE2Spn could be successfully modelled, this was not the case for RelE1Spn, because it exhibited marked differences in the distribution of the charges as well as in the residues involved in the toxin activity (not shown). Although these differences could explain the lack of toxicity of RelE1, this does not exclude that relBE1Spn encodes an intact relBE locus because we did not actually show its expression in E. coli.
We located the relBE2Spn locus in the chromosome of the two available genomes of S. pneumoniae, namely strains R6 and Tigr4 (Hoskins et al., 2001; Tettelin et al., 2001). In the former strain, the relB2Spn gene is termed spr1104 (coordinates 1105420–1105118), whereas relE2Spn is named spr1103 (coordinates 1105125–1104862), both located at 1105.4 kb and 1105.1 kb respectively. In Tigr4 (capsulated, pathogenic strain), the antitoxin gene is called SP1244 (coordinates 1156966–115768) and the gene encoding the toxin is SP1223 (coordinates 1156710–1156964), both located at 1157.3 and 1157.0 kb respectively. The nucleotide sequence of the relBE2Spn loci of both strains is identical (see http://genolist.pasteur.fr/StreptoPneumoList/index.html).
Genes relB2Spn and relE2Spn are organized in an operon
The structure of the pneumococcal relBE2Spn locus is such that both genes seem to constitute a single operon, as it is the case for other relBE loci (Bech et al., 1985). The antitoxin is encoded by the upstream gene of the operon and the last codons of the relB2Spn reading frame overlap with the first AUG codon of the toxin (Fig. 1A). This linkage, found in other TA loci, may indicate translational coupling between the relB and relE genes, in agreement with the finding that plasmid- and chromosomally encoded relE genes are expressed at lower levels than those of relB (Gotfredsen and Gerdes, 1998; Gronlund and Gerdes, 1999; Gerdes, 2000). Experimental confirmation of the genetic organization of the pneumococcal relBE2Spn cassette in an operon was confirmed using a reverse transcription polymerase chain reaction (RT-PCR) analysis (Fig. 1B). The RT technology was used to synthesize cDNAs that were complementary to the relBE mRNA from three independent preparations of pneumococcal RNAs. The cDNAs were PCR-amplified by the use of two DNA primers (relB2N and relFc) that would anneal to either the 5′-end of relB2Spn or the 3′-region of relE2Spn respectively (Fig. 1A). The resulting DNA products (lanes 1–3) were, in all three cases, a single 480 bp DNA-fragment, identical to that obtained using these primers and chromosomal DNA as template (lane 4), showing the transcriptional linkage between relB2Spn and relE2Spn into an operon. In the control reactions, the same RNA samples were added but the RT was not performed, and no PCR DNA fragment was detected (lanes 5–7).
To determine the transcription initiation point of the relBE2Spn operon, primer extension analyses were performed. Total RNA was prepared from either the wild type (wt), or an isogenic strain that harbours a deletion of the relB2Spn gene (ΔrelB2Spn; see below). The results showed the presence of a major band of 102 nucleotide (nt) in both strains, indicative of a single transcription initiation site (Fig. 2). Inspection of the sequence around the +1 site (Fig. 2), showed the existence of almost consensus −35 (5′-TTGACt-3′) and −10 (5′-TAcAAT-3′) boxes separated by a 17 nt long spacer, which matches with other pneumococcal promoters previously identified (Sabelnikov et al., 1995). As usual in bacterial promoters, the distance from the −10 and the +1 position in promoter Prel is 11 nt whereas the +1 position in the non-template strand is a purine (Lewis and Adhya, 2004). Downstream the +1 site, and before the ATG first codon of the relB2Spn gene, there is a region that matches well (with the exception of the two C residues just after the A at +1 position) with the optimal initiation of translation region derived from 460 sequences of S. pneumoniae, as deduced from the data taken from Bio-ToolKit300 (http://www.changbioscience.com/virtualab.html). As TA proteins generate a complex that autoregulate their own synthesis (Ruiz-Echevarría et al., 1991), we searched for inverted repeats (IR) that could indicate putative binding sites for transcriptional regulators, including the RelBE protein complex that would bind to these sequences. Two IRs were found in the relBE2Spn sequence: IR-1, an almost perfect 13-bp IR located upstream the −35 region and IR-2, a 6 bp perfect palindrome placed downstream the −35 and the −10 regions (Fig. 2). Binding of the RelBE2Spn proteins to any of these two IR could result in repression of transcription of the operon, although at present we cannot postulate which of the two, if any, of the IRs could be the binding site of these pneumococcal TA proteins. These types of regulatory sequences and their role in the TA autoregulation have been identified in another relBE system of E. coli (Gronlund and Gerdes, 1999).
Further information on the activity of the Prel promoter was obtained by a transcriptional fusion, using the gene encoding the green fluorescent protein (GFP) as a reporter. A DNA fragment containing the Prel promoter was synthesized by PCR, cloned in the E. coli vector pJDC9 (Chen and Morrison, 1988), and fused to the gfp gene (Nieto et al., 2000). The resulting plasmid was termed pJPrelGFP. To compare the activity of the Prel promoter with that of a well characterized pneumococcal promoter, PM (Nieto et al., 1997), another plasmid-borne gfp-transcriptional fusion was constructed. This second plasmid, termed pJPMGFP (Nieto et al., 2001), harbours the gfp gene under the control of the pneumococcal PM promoter. Measurement of the fluorescence in E. coli harbouring either of the plasmids showed that promoter Prel was more than 15 times stronger than PM (Table 1). In both cases, the promoters were under unregulated conditions, because Prel was uncoupled of autoregulation by the RelBE2Spn complex (Gotfredsen and Gerdes, 1998), and PM is repressed by the product of gene malR (encoding the MalR transcriptional repressor) which is located in the chromosome of S. pneumoniae (Nieto et al., 1997). The Prelgfp cassette was also cloned into the pneumococcal plasmid pLS1 (Lacks et al., 1986) to construct plasmid pLS1PrelGFP. In this host, the fluorescence due to the activity of the Prel promoter was measured in both pneumococcal hosts (wt and relB2 mutant) and compared with the activity of the PM promoter under inducible condition (maltose-grown cultures). In both strains it was evident that PM promoter was three to four times stronger than promoter Prel (Table 1). The different fluorescence of pneumococcal strains (wt and ΔrelB2, see below) harbouring plasmid pLS1PrelGFP (175 vs. 300 units, Table 1) suggested to us that partial autoregulation due to the presence of the chromosomal copy of relB2Spn existed. Preliminary results showed an increase of fluorescence in the wt strain, but not in the antitoxin-deficient strain (both harbouring pLS1PrelGFP) under stress conditions (not shown). In addition, when the protein synthesis was inhibited in the wt strain, an increase in the level of relBE2Spn mRNA was also observed (not shown). Fluorescence of control cultures (grown under non-inducible conditions) gave values below a threshold of 40 units. Taken these results together, it was evident that in both bacterial species, E. coli and S. pneumoniae, the DNA region encompassing the mapped Prel promoter was able to support expression of the gfp reporter gene.
Table 1. Measurement of the strength of promoter Prel by transcriptional fusions.
Fluorescence of both E. coli and S. pneumoniae cells harbouring plasmids in which gene gfp was placed under the control of promoters Prel or PM.
E. coli TOP10
E. coli TOP10
E. coli TOP10
S. pneumoniae wt
S. pneumoniae wt
S. pneumoniae wt
Construction of tools to characterize the relBE2Spn locus
Preliminary experiments indicated that expression of relE2Spn gene in an S. pneumoniae wt background resulted only in a relatively modest delay in cell growth (not shown), in agreement with reports indicating that the presence of a single copy of the E. coli relB antitoxin gene could account for a partial titration of multiple copies of the toxin gene expressed from a multicopy plasmid (Pedersen et al., 2002; Christensen and Gerdes, 2004). Thus, a pneumococcal relB2Spn null strain unable to synthesize the antitoxin was constructed by means of a gene replacement strategy developed for pneumococci (López et al., 1982). A truncated relB2Spn gene of 200 bp, lacking three codons from the 5′-end and 11 codons from the 3′-end, was cloned to construct plasmid pTrB2 (Fig. 3A). This plasmid is a derivative of pUC18 that harbours the truncated relB2Spn fused to a chloramphenicol-resistance (cat) gene (with its own transcriptional terminator), permitting selection for resistance to chloramphenicol (CmR). In addition, pTrB2 is unable to replicate in S. pneumoniae. DNA from pTrB2 was used to transform competent pneumococcal wt cells, and transformants were rescued by selection to CmR PCR-analysis of the chromosomal region showed that homologous recombination between the incoming DNA and the chromosome of the recipient was achieved (Fig. 3A). The strain thus constructed contained two truncated relB2Spn genes and an intact relE2Spn gene that was placed away from its natural promoter (Fig. 3B). The construction was confirmed by determination of the nucleotide sequence of two chromosomal DNA fragments of 1.4 kb and 0.6 kb obtained by PCR using as a primers relB2p and M13-reverse oligonucleotides (olg.rv) or M13-forward (olg.fw) and relFc oligonucleotides respectively. The former fragment was sequenced with C-ter, relB2p and M13-reverse oligonucleotides and the latter with M13-forward, N-ter, relFc oligonucleotides (Fig. 3B).
The next stage was to design a vector to over-express the pneumococcal toxin. To this end, a new plasmid, termed pSE23, was constructed, in which the relE2Spn toxin gene was placed under the control of the strong maltose-inducible PM promoter (Nieto et al., 2001). This promoter directs transcription of the pneumococcal malMP operon (involved in the metabolism of maltosaccharides). Expression of malMP is negatively regulated by the product of gene malR, also belonging to the pneumococcal mal regulon (Puyet and Espinosa, 1993). MalR repressor is inactivated by maltose (Nieto et al., 1997; 2001), a feature that has been exploited by us to construct regulated pneumococcal expression vectors (Nieto et al., 2000), although the present new vector yielded higher levels of expression (not shown). Thus, induction of the pneumococcal toxin was achieved by growing the cultures in the presence of maltose as carbon source.
RelE2Spn inhibits cell growth and colony formation in S. pneumoniae
Competent cells of S. pneumoniae (wt and ΔrelB2Spn strains) were transformed with DNA from either pSE23 or the control plasmid pLS1PMGFP. Clones harbouring the plasmids were selected and characterized. The four resulting strains were grown in medium containing sucrose (repressed conditions) to an initial OD650 = 0.02 and the cells were collected and washed twice with sugar-free medium. Cells were suspended at the same density in medium containing either maltose (to induce synthesis of RelE2Spn) or sucrose, and growth was followed by determination of the OD650 of the cultures. The results showed that overproduction of RelE2Spn under inducible conditions inhibited cell growth (Fig. 4A), which was concomitant with a severe reduction in the number of colony-forming units (cfu; Fig. 4B). Cessation of growth and reduction in the cfu ability of the cultures is a feature of over-production of the toxin in the TA systems (Sat et al., 2001; Pedersen et al., 2002; Grady and Hayes, 2003; Zielenkiewicz and Ceglowski, 2005). In the case of the pneumococcal RelE2Spn toxin (Fig. 4B), the final number of colony-forming units per millilitre in the mutant strain dropped from 109 (uninduced conditions) to 4 × 104 (induced conditions), that is a 100 000-fold reduction in the viability of the cultures. In the wt strain, however, this reduction was more modest (about a 1000-fold final reduction). Such a difference in toxicity can be explained because the wt strain has single chromosomal relB2Spn gene, whose product would counteract the cytotoxic effect of the toxin (see also Table 1). When induction of the system was performed at lower maltose concentrations, intermediate levels of toxicity were observed (not shown).
The ability of the pneumococcal strains (wt and ΔrelB2Spn) harbouring the toxin-producing plasmid pSE23 to recover from growth arrest subsequent to a toxic shock was next tested. In these experiments, the cells were grown in maltose-containing medium (conditions of induction) for 8 h, so that cell growth was totally arrested. Then, cells were washed and transferred to sucrose-containing medium (non-induced conditions), and growth was followed as above. The results showed that both strains resumed growth after a lag period (Fig. 4C). Recovery of the wt cells took place after 60 min, while it took 250 min for the mutated strain to resume growth, demonstrating that toxin-induced stasis in the cells that lacked the antitoxin was much more severe (Fig. 4B). After 300 min of growth, wt cells started to lyse because of the potent autolysin encoded by S. pneumoniae (López and García, 2004).
From the above results, we can draw the following conclusions: (i) the relBE2Spn operon is a bona fide chromosomal TA system; (ii) RelE2Spn has a dosage-dependent potent toxic effect; (iii) deletion of the chromosomal relB2Spn gene increases the toxic effect, indicating that even a single copy of the antitoxin can be enough to, at least partially, neutralize the toxicity of RelE2Spn; (iv) overproduction of the toxin in S. pneumoniae inhibits cell growth, and (v) cells lacking the antitoxin needed more time to recover from the RelE2Spn inhibitory effect than the wt.
Broad-range spectrum of relBE2Spn: toxicity towards E. coli
Escherichia coli was employed as an additional host to: (i) test the heterologous toxicity of the pneumococcal toxin; (ii) determine whether the cognate antitoxin could counteract the possible toxic effect, and (iii) develop a system amenable to detection of the interactions between the toxin and its antitoxin counterpart. This was a logical approach, because of the availability of many expression vectors developed for this bacterium. To achieve these objectives, translational fusions with reporter genes were made. The reporters used encoded the enhanced yellow fluorescent protein (EYFP) and the Renilla luciferase (Rluc) and were fused with the pneumococcal toxin or antitoxin respectively. The plasmid constructions were designed in such a way that E. coli cells harbouring them would synthesize N-terminally fused proteins, namely EYFP–RelE2Spn toxin and Rluc–RelB2Spn antitoxin. The eyfp::relE2Spn fusion was cloned in the pFUS2 plasmid (Lemonnier et al., 2003) and placed under the control of the araBAD promoter (PBAD), which is inducible by arabinose and repressed by glucose. The resulting plasmid was termed pYE304. The Rluc::relB2Spn translational fusion was cloned into plasmid pNM220 (Gotfredsen and Gerdes, 1998) and placed under the control of the IPTG-inducible Plac promoter, and the resulting plasmid was termed pLB404 (Fig. 5A). Plasmids with the inducible promoters and only the reporter genes were used as a controls, namely pY104 (PBAD-eyfp) and pL204 (Plac-Rluc). The E. coli cells transformed with pYE304 DNA were selected in plates with 0.4% glucose (repressed conditions of PBAD) to minimize toxin expression, because transformants selected in glucose-free TY plates, although generated tiny colonies, they failed to further growth.
Next, the toxicity of the EYFP–RelE2Spn fusion was assayed. No significant differences in the growth rates of cultures grown under PBAD-induced (+arabinose) or repressed (+glucose) conditions were observed when the cells harboured the control pY104, nor when the cells harbouring the toxin-expressing pYE304 were grown in medium with glucose. However, a total arrest in the cell growth was observed in E. coli cells harbouring pYE304 after arabinose addition, demonstrating that the EYFP–RelE2Spn fused protein retains its toxic activity (Fig. 5B). When the cell viability of these cultures was determined (Fig. 5C), it was apparent that the growth arrest due to the expression of the EYFP–RelE2Spn fusion protein was concomitant with a severe decrease in the number of viable cells (almost five orders of magnitude), demonstrating the potent toxic effect of the pneumococcal RelE2Spn toxin on the G– host.
To determine whether the toxicity of the pneumococcal toxin towards E. coli could be reverted by the RelB2Spn antitoxin, cells containing the above mentioned plasmids were streaked on 0.2%-arabinose plates with or without IPTG. Cells harbouring the control plasmids did not show differences of growth. However, when the cultures harbouring only plasmid pYE304 or the pair of plasmids pYE304 + pL204, or pYE304 + pLB404 were tested, only those cells that carried the TA pair were able to grow in the arabinose-IPTG-induced conditions (see supplementary data in Fig. S1). To follow the growth rate and viable counts of the cultures, cells harbouring both pYE304 and pLB404 were grown in glucose-containing medium to repress toxin expression. When the culture reached an OD600 of 0.02, it was divided into two. One of them continued to grow only with glucose as a carbon source (control culture), whereas the other culture received arabinose to the final concentration of 0.4%. The results showed that cells were unable to resume growth once synthesis of the toxin was triggered, being more pronounced when the inducer was added to 0.4% (Fig. 6A) than when arabinose was used at 0.2% (not shown). No effect was observed in the control culture. At each time, the number of cfu were determined by plating on glucose medium in the presence or absence of IPTG (to induce the antitoxin), and plates were incubated more than 24 h at 37°C. The results showed that, contrary to the drastic reduction in viable counts when the cells harboured only the pneumococcal toxin gene (Fig. 5B), the number of cfu in cells harbouring the toxin and the antitoxin genes was moderately reduced (Fig. 6B), even without antitoxin synthesis induction (-IPTG). Not changes were observed when arabinose was used at 0.2% (not shown). These results were unexpected because they are somewhat different than those reported when the E. coli RelB and RelE were coexpressed from two different plasmids, in which overproduction of RelE during 4 h reduced 1000-fold the number of cfu in plates lacking IPTG (no antitoxin being synthesized), as compared with plates supplemented with IPTG (Pedersen et al., 2002), whereas in similar conditions, only a 10-fold reduction in the number of cfu was observed for the pneumococcal toxin (Fig. 6B). A similar result was obtained when an E. coliΔrelBE strain was used (not shown).
The above differences could be explained by a number of reasons: (i) there was a basal relB2Spn expression from the Plac promoter under non-inducible conditions due to leakage of the system, so that even under antitoxin non-inducible conditions there were enough pneumococcal antitoxin molecules to titrate the toxin; (ii) the RelB2Spn antitoxin might be more stable than its E. coli homologue perhaps because is not well recognized by the E. coli proteases either as such or as a fusion protein, and (iii) the incubation time was longer in our experimental conditions, so that we observed in plates without the antitoxin inductor (IPTG) the appearance of tiny colonies which, after further incubation, could grow, thus increasing the number of cfu. In plates with IPTG, no increase in the cfu counts was found even after prolonged periods of incubation. This observation agrees with previous results reported for E. coli relBE system in which the same antitoxin expression vector (pNM220) was used (Keren et al., 2004).
To increase the amount of free pneumococcal toxin (not titrated by the cognate antitoxin) E. coli cultures harbouring plasmids pLB404 and pYE304 were grown for five or six generations in glucose-free medium, so that promoter ParaBAD was under non-repressed conditions. Then, arabinose 0.4% and glucose 0.4% were added (toxin induction) and incubation continued during 3, 7 and 10 h. Viable cells were counted before arabinose addition and after the different times of toxin induction. The results showed a gradual reduction in cfu formation even after induction of the antitoxin with IPTG (Fig. 7A). The reduction in viable cells was already detectable at time 0 in plates without IPTG, indicating some leakage in toxin synthesis, even though arabinose was not added. Concomitant with the cfu reduction, the ability of the cells to be resuscitated by the antitoxin was also gradually lost (Fig. 7B). This behaviour contrasted with the response of E. coli cells to their cognate RelBE TA system, in which the cell viability was retained and the bacteria were still alive after antitoxin production even after a long period of toxin overproduction (Pedersen et al., 2002).
Direct TA interaction: BRET assays
The above results demonstrated that fusion toxin and antitoxin proteins retained the ability to interact, so that they were amenable to the analysis of the BRET interactions. BRET occurs from energy transfer between luminescent donor and fluorescent acceptor proteins. This phenomenon is strictly dependent on the molecular proximity between energy donors and acceptors, making it a technique of choice to study the interaction between two proteins. Protein partners (pneumococcal toxin and antitoxin) were tagged with EYFP or Rluc respectively. When the donor and acceptor are in close proximity, the light generated from the substrate transformation by luciferase excites the EYFP, which will then emit fluorescence at its characteristic wavelength, the non-radioactive energy transfer between the excited Rluc and the EYFP thus permitting the study of both spatial- and orientation-relationships between the two partners. Rluc and EYFP do not naturally interact with one another so the BRET signal is a consequence of the spatial proximity between the two proteins under study. The implementation of the BRET assay was carried out with a positive control plasmid harbouring Rluc and EYFP fused in frame (Xu et al., 1999; 2003); a high BRET signal using coelenterazine as luminogenic substrate was rapidly detected and found to be constant for up to 60 min after the addition of the substrate (Fig. 8A). Plasmids harbouring the relB2Spn and relE2Spn gene fusions to Rluc (pLB404) and to EYFP (pYE304), respectively, were cotransformed in E. coli and used to assess the TA interaction (see Fig. 5A). Cells were grown in the presence or absence of inducers and the BRET ratio determined. The results obtained showed that the plasmids render an inducible and high BRET signal (Fig. 8B). To confirm the specificity of this signal, cotransformed E. coli cells with pYE304 and pLB404 or pL204 were grown under inducing conditions for both plasmids (arabinose for pYE304 and IPTG for plasmids pLB404 and pL204). A negligible BRET ratio was detected in cells harbouring only the fused toxin compared with that elicited by cells transformed with the toxin and antitoxin fusions. These results strongly suggest that the direct interplay between RelB2Spn and RelE2Spn is the responsible for the BRET signal (Fig. 8C). To support this observation, we performed experiments in which E. coli cells cotransformed with plasmids pYE304 and pLB404 were subjected to stress conditions, and BRET signal was measured in a time-course experiment up to 1100 min. The results showed a consistent reduction from a 100% value (non-stressed cells) to about 30% (stress conditions) that could me measured at 120 min after stress induction and that remained all the period tested (Supplementary material, Fig. S2). These results support that interaction between the pneumococcal toxin and its antitoxin counterpart does exist and can be measured in vivo. Furthermore, they also add a new dimension of the BRET assays, which can be used to measure not only protein–protein interactions but dissociation of the interactions as well.
The S. pneumoniae relBE2Spn locus is the first reported chromosomally encoded TA system, present in G+ bacteria, that has been characterized as such in its own host. The system has a general behaviour similar to that reported for the E. coli relBE (Gerdes et al., 2005), although we have found differences in the cell-death mediated by the pneumotoxin and resuscitation by the antitoxin (Figs 6 and 7). Under our present experimental conditions, we could reveal that the pneumococcal TA system can act as a killing system at least in the E. coli heterologous host. Based on sequence homologies, numerous putative proteic TA loci have been postulated to exist in G+ bacteria (Mitenhuber, 1999; Gerdes, 2000; Grady and Hayes, 2003; Cherny and Gazit, 2004; Pandey and Gerdes, 2005). However, only three plasmid-encoded TA systems have been characterized so far, namely the ωɛζ module of Streptococcus pyogenes plasmid pSM19035 (Ceglowski et al., 1993; Meinhart et al., 2001; 2003; Zielenkiewicz and Ceglowski, 2005),and the TA cassettes of plasmids pRUM from Enterococcus faecium (Axe–Txe) (Grady and Hayes, 2003) and p256 of Lactobacillus plantarum (Sorvig et al., 2005). Overproduction of the toxins of these modules inhibited E. coli cell growth, but they exerted a limited rate of cell death in E. coli. In addition, these cassettes are plasmid-encoded and promote segregational stability, so that they could be considered as a post-segregational killing system. The chromosomal TA loci may represent cell growth modulators during stress conditions, so that toxin expression in their natural host, contrary to those encoded by plasmids, does not necessarily leads to cell death (Hayes, 2003; Alonso et al., 2006). The relBE2Spn locus was identified in the pneumococcal chromosome because of its homology with the relBE system of the E. coli chromosome (Gerdes, 2000). The relBE2Spn gene pair has the regulatory components and genetic organization of a typical TA system, because: (i) they are organized as an operon; (ii) the antitoxin gene (relB2Spn) is placed upstream of the toxin gene (relE2Spn), and (iii) both are translated from one mRNA transcript (Fig. 1). With such structure, the RelB2Spn antitoxin would bind to RelE2Spn and they should form, like in the E. coli relBE, a non-toxic protein complex (Galvani et al., 2001). As reported for other TA systems (Gronlund and Gerdes, 1999), the RelBE2Spn complex would regulate its own transcription by binding to an operator region (presumably the IR sequences; Fig. 2). The relBE2Spn resulted to be a functional TA system in the cognate host because overproduction of the RelE2Spn toxin resulted in growth arrest and a severe decay in colony formation (Fig. 4), which was more marked in the ΔrelB2Spn mutant (100 000-fold) than in the wt (1000-fold). This differential reduction in cfu agrees with the different capacity of cell growth recovery between the wt and the mutant. Thus, after shifting pneumococcal cultures from inducible to repression conditions for toxin synthesis, the wt strain required about 150 min to double its optical density, whereas the mutant needed more than 360 min (Fig. 4C).
The host range of relBE2Spn as a TA in a heterologous host was evaluated in E. coli by construction of plasmids with translational fusions between EYFP–RelE2Spn and Rluc–RelB2Spn (Fig. 5A). In the case of the toxin, fusion of a 239-residue protein (EYFP) to the 87-residue of the RelE2Spn did not modify the cytotoxic activity of the latter protein. Therefore, inhibition of cell growth and decay in viable counts were observed in E. coli cultures expressing the EYFP-toxin fusion protein (Fig. 5), demonstrating a broad host activity of the RelE2Spn toxin. This is understandable because RelE2Spn inhibits translation in E. coli by mRNA cleavage (Christensen and Gerdes, 2003). Similarly, the fused Rluc–RelB2Spn protein also retained its antitoxin activity, as shown by its ability to partially counteract the RelE2Spn toxic effect (Fig. 6B and Fig. S1). When higher amounts and longer periods of toxin synthesis were tested, a gradual decrease in the number of cfu was found, so that after 10 h of continuous toxin synthesis, a 100-fold decrease in viable cells was observed, even in plates supplemented with IPTG (Fig. 7). However, within its genetic context, we believe that the single chromosomal copy of the relE2Spn gene will provide a much lower level of toxin than in the plasmid multicopy situation, in which the amount of free toxin should lock the ribosomes impairing its rescue by tmRNA. In these conditions translation would be very low, favouring the senescence processes and finally cell death (Bogosian and Bourneuf, 2001; Desnues et al., 2003; Nystrom, 2003). In this sense, it has been shown that E. coli cells were more sensible to heat shock, oxygen radicals and osmotic stress during the RelE-induced stasis period (Pedersen et al., 2002). As in the case of other TA loci, RelE2Spn should be involved in promoting cell growth arrest under stress conditions (perhaps a host immune response, giving the nasopharynx niche of the bacteria).
A group of techniques, commonly referred to as Resonance Energy Transfer, provide an exciting new approach to study protein–protein interactions. These technologies can be carried out both in vitro (with purified proteins) and in vivo (with intact cells). BRET is currently being used as the method of choice to unambiguously assess interactions between therapeutically relevant proteins. For instance, recent works have shown that BRET is a reliable, quantitative and highly sensitive method to monitor molecular interactions between different proteins of pharmacological interest (Eidne et al., 2002; Issad et al., 2002; Mercier et al., 2002; Babcock et al., 2003; Germain-Desprez et al., 2003). Here we have used BRET to assess protein–protein interactions between the RelE2Spn toxin and the RelB2Spn antitoxin in E. coli. The in vivo assays were carried out with whole cells transformed with the BRET reporter plasmids. The high BRET signal that was detected with the inducible plasmids expressing the TA pair indicates that the S. pneumoniae TA proteins were stably expressed in E. coli (Fig. 8). This result, together with the observation of growth inhibition and a decrease in the cfu counts in E. coli cultures by the pneumotoxin, highlights three interesting aspects: (i) the pleiotropic effect of TA systems among different prokaryotic species; (ii) validates the use of BRET to study heterologous systems, and (iii) suggests that the mechanisms underlying the role of the TA systems can be more complex than envisaged. The stability of the BRET ratio with time seems to be the consequence of the constant diffusion of the coelenterazine substrate through the cell membrane. BRET signal arising from cells expressing Rluc and RelE2Spn fused to EYFP was lower than that generated by cells expressing RelB2Spn and RelE2Spn fused to Rluc and EYFP respectively. This difference in the BRET ratio points out the specificity of the assay and highlights the TA-driven nature of the results. The specificity, reproducibility and robustness of the BRET assay developed in this study seem to make it a useful tool to identify mutants affected in TA interaction (as shown by the feasibility to measure decreases in the BRET signals when cells were subjected to stress; Fig. S2) and to define domains in the proteins involved in the complex formation. Furthermore, it would be a method to screen molecules capable of disrupting TA interactions in a high-throughput fashion and thus to discover new potential antimicrobials.
Bacterial strains, growth conditions and plasmid constructions
Strains and plasmids used in this study are listed in Table 2. E. coli cultures were grown in TY medium (Maniatis et al., 1982) with selection for erythromycin resistance (ErmR, 500 µg ml−1), ampicillin resistance (AmpR, 150 µg ml−1), or kanamycin resistance (KmR, 50 µg ml−1). S. pneumoniae cells were grown in AGCH medium (Lacks, 1968), supplemented with 1% sucrose or maltose and 0.25% yeast extract (YE), with selection for resistance to tetracycline (TetR, 1 µg ml−1) and to chloramphenicol (CmR, 2 µg ml−1). All cultures were grown at 37°C.
Plasmids used in this work were constructed as follows
pJPrel. A DNA fragment spanning promoter Prel was amplified by PCR from the chromosome of S. pneumoniae wt strain using primers: rel2P5 (5′-CGGAATTCCGATCAGGTTCTTA CGCTTGGCG-3′) and prelBam (5′-CGGGATCCTTTTTAA TGGTAACACCATTG′-3). The 350 bp PCR fragment was digested with EcoRI and BamHI before ligation into the equivalent sites of plasmid pJDC9.
pJPrelGFP. An XbaI–SphI DNA fragment from the plasmid pCL1GFP containing the gfp gene was ligated into the equivalent sites of pJPrel.
pLS1PrelGFP. A 1075 bp fragment containing the PrelGFP fusion was obtained by digestion of pJPrelGFP with ClaI, made blunt ended with Klenow enzyme, and then digested with EcoRI. The fragment was cloned in plasmid pLS1 previously digested with HindIII, made blunt ended with Klenow enzyme and further digested with EcoRI.
pTrB2. A PCR-DNA fragment obtained from the chromosome of S. pneumoniae encoding a deleted relB2Spn gene, lacking the codons encoding the 3 first amino acids and the codons encoding the last 11 residues, was generated using primers: relB2N (5-TGCTCCCGGGCTATTACATTAAAAGTTT CTGAAGCTG-3′) and relB2C (5′-CGCGAATTCCTTCCCAA GTAATGGGTTCAACTCC-3′). The 210-bp DNA fragment was digested with SmaI (blunt end) and EcoRI and ligated into pCR2.1cat, digested with EcoRV (blunt end) and EcoRI.
pLS1RPXMSpnrelE2. The relE2Spn gene with its own SD was amplified from the wt pneumococcal strain by PCR using primers: relE2N (5′-CGCGGATCCGATGCATGATTTAGGCT TGAAGGATGAATA-3′) and relE2C (5′-CGCGAATTCGAAT GAAAATTTACTTGAAAAAAGTAATTC-3′). The 300-bp PCR product and an EcoRI–SphI DNA fragment from plasmid pCL1 (harbouring promoters PX-PM) were digested with EcoRI made blunt-ended with Klenow enzyme and then digested with BamHI. Both DNA fragments were cloned into plasmid pLS1 digested with EcoRI and HindIII and with the ends made blunt with Klenow enzyme.
pSE23. A HindII–AvaI DNA fragment from pLS1RPXMSpnrelE2, containing only the PM promoter and relE2Spn was ligated into the equivalent sites of pLS1.
pY104. A DNA fragment containing the eyfp gene was amplified by PCR from plasmid pT7RLUC-EYFP (a gift of Dr Carl Johnson, Vanderbilt University, Nashville, TN, USA) using primers: EYFPN (5′-CCGGAATTCTAAAGGGAGGAAA AACATATGGTGAGCAAGGGC-3′), and EYFPC (5′-GCG GGTACCACCCTTGTACAGCTCGTCCATGCCGAGAGT-3′). The 717-bp DNA fragment contained the eyfp reading frame combined with an optimized SD, but lacked its termination codon. It was digested with EcoRI and KpnI before ligation into the equivalent sites of pFUS2.
pYE304. A DNA fragment containing the relE2Spn gene was amplified by PCR from the chromosome of S. pneumoniae, using primers: rel FN (5′-GCGGGTACCAT GAATAATTTGTATAAATTAGTTCCAACA-3′) and relFC (5′-CGCGGATCCGAATGAAAATTTACTTGAAAAAAGTAATTC-3′). The 300-bp PCR product was digested with KpnI and BamHI and ligated into pBEYFP digested with KpnI and BglII (compatible ends with BamHI). Upon addition of arabinose, an EYFP–SpnReLE2 fusion protein is synthesized.
pL204. A DNA fragment containing the Rluc gene was amplified by PCR from plasmid pT7RLUC-EYFP (a gift of Dr Carl Johnson, Vanderbilt University, Nashville, TN, USA) using primers: LUCN (5′-CGCGGATCCTAAAGGGAGGA AAAACCATGACTTCGAAAGTTTATG-3′) and LUCc (5′-CCG GAATTCGCTGCAGGTACCCCGTTGTTCATTTTTGAGAAC-3′). The 933-bp PCR product included rluc fused to a optimized SD but lacked its termination codon. This DNA fragment was digested with EcoRI and BamHI before ligation into the equivalent sites of pNM220.
pLB404. A DNA fragment containing the relB2Spn gene was amplified by PCR from the chromosome of S. pneumoniae, using primers: relLcN (5′-CGGGGTACCAT GACTACTATTACATTAAAAGTTTCTG-3′) and relLcC (5′-CGCGAATTCTAAAACGTCTTGTTGGAACTAATTTATAC-3′). The 300-bp PCR fragment was digested with KpnI and BamHI before ligation into the equivalent sites of pNMRLUC. The resulting pLB404 produces a Rluc–RelB2Spn fusion protein upon addition of IPTG.
Mapping of initiation of transcription start points
Total RNA was isolated from S. pneumoniae and primer extension assays were performed as described (Puyet and Espinosa, 1993) using a [32P]-labelled relBE2Spn-specific primer relRNA (5′-GAAACTCCTTCAAACTTAGCC-3′), its 3′-end being located 56 nt from the A included in the relB2Spn ATG initiation codon.
Reverse transcription PCR
Total RNA from S. pneumoniae (400 ng) and primer relFc were mixed and denatured by incubating 5 min at 70°C, and then placed on ice. This mixture was added to a 20 µl RT reaction containing 4 µl of cDNA synthesis buffer, 5mMDTT, 40 units RNase OUT, 1mMdNTP mix, 10 µM gene-specific primer, 15 U Thermo-Script RT (Invitrogen) and incubated for 60 min at 50°C. Reverse transcription was terminated by incubation during 5 min at 85°C. RNA template was removed by addition of 2 units of RNase H and incubated at 37°C, 20 min. An aliquot (10%) of the cDNA synthesis reaction (2 µl) was used for each PCR reactions. PCR reactions were carried out using Taq DNA Polymerase (Invitrogen) and, as primers, relFc and relB2N. Annealing temperature was 50°C with an extension time of 2 min at 72°C and 30 cycles of amplification.
Measurement of GFP activity
Escherichia coli cells harbouring plasmids pJPrelGFP or pJPMGFP or S. pneumoniae strains carrying plasmids pLS1PMGFP or pLS1PrelGFP were grown to middle exponential phase (OD600 = 0.4 or OD650 = 0.4 for E. coli or S. pneumoniae respectively). Cells were sedimented by centrifugation and suspended in PBS Buffer (10 mM Na2HPO4, 140 mM NaCl, 3 mM KCl, pH = 7.2). Fluorescence was determined on a LS-50B spectrophotometer Perkin Elmer Life Sciences by excitation at 488 nm and detection of emission at 515 nm. As background controls, cells of E. coli or S. pneumoniae harbouring plasmids pJDC9 or pLS1, respectively, were used. The fluorescence values obtained were subtracted from those obtained from cells harbouring plasmids with the gfp gene. All experiments were performed at least three times.
Capacity of growth cell recovery in S. pneumoniae
Streptococcus pneumoniae wt cultures were exponentially grown to OD650 = 0.4 in sucrose AGCH medium supplemented with 0.25% YE and Tet (1 µg ml−1), washed and diluted to an OD650 = 0.02 in the same medium but supplemented with 1% maltose as carbon source. The cultures were grown at 37°C for about 8 h. After this time the OD ratio sucrose/maltose was 0.9/0.24 or 1.050/0.4 for the mutant and the wt cells respectively. Both cultures were washed twice with AGCH and suspended in a volume of the same medium containing 1% sucrose to a final OD650 = 0.2. This was considered to be t = 0 and the optical densities of both cultures were followed for the indicated periods of time. Cultures of the ΔrelB2Spn mutant strain were treated similarly, but Cm (2 µg ml−1) was added to the growth media.
Bioluminescence resonance energy transfer assays
The BRET assays were performed in vivo with whole cells with modifications of the method described by (Xu et al., 1999). E. coli cells cotransformed with pYE304 and pLB404 or pNMRluc (as a control) were grown aerobically in Luria–Bertani with the appropriate antibiotic at 37°C overnight. Cells were diluted in the same medium until an OD600 of 0.5, and induced with IPTG and d-arabinose for 3 h. Then, cells were centrifuged, washed twice with BRET buffer (PBS pH 7.4, CaCl2 0.1 g l−1, MgCl2·6H2O 0.1 g l−1, glucose 1 g l−1) and adjusted to an OD600 of 1.0. The assay was carried out in black and white polystyrene 96-well plates (Wallac, Perkin-Elmer Life Sciences) at room temperature. Each well contained 25 µl of cells and 25 µl of BRET buffer. The reaction was initiated by the addition of 50 µl of 1× coelenterazine substrate (Promega, Madison, WI, USA) and the fluorescence emission measurements were recorded 10 min after (time required for the substrate to enter the cells), using the Victor3V 1420 multilabel reader (Wallac, Perkin Elmer) instrument. Results are means of at least two experiments performed in triplicate. For the time course experiments, fluorescence was recorded for up to 60 min after the addition of the substrate; for the rest of the experiments, BRET was measured 10 min after the addition of the substrate. For this BRET pair, the emission filters were 510 nm and 535 nm. The intensity of both wavelengths was automatically recorded and the 535/510 (BRET measurement) ratio calculated. The calculated BRET ratio indicates the occurrence of protein–protein interaction in vivo.
We thank M.T. Alda for technical help, M. García de Lacoba for molecular modelling, M. Lemonnier for generous gift of plasmid pFUS2 and C. Johnson for kindly providing us with plasmids pT7/EYFP and pT7RLUC-EYFP. Research financed by the European Union (Grant QLK3-CT-2001-00277), the Spanish Ministry of Education and Science (Grant BFU2004-00687/BMC) and the Comunidad Autónoma de Madrid (Grant GR/SAL/0659/2004).