The Escherichia coli relBE genes belong to a new toxin–antitoxin gene family


Kenn Gerdes E-mail; Tel. (65) 57 24 13; Fax (65) 93 27 81.


Toxin–antitoxin systems are defined as a group of plasmid- and chromosome-encoded loci that specify a cell toxin and a protein antitoxin. Plasmid-encoded toxin–antitoxin systems stabilize their replicons by killing plasmid-free cells. Here, we show that the relBE genes of Escherichia coli K-12 have all the basic features previously connected with toxin–antitoxin systems: (i) relE encodes a cytotoxin lethal or inhibitory to host cells; (ii) relB encodes an antitoxin that prevents the lethal action of the relE-encoded toxin; (iii) the relBE genes stabilize a mini-R1 test plasmid; and (iv) the RelB antitoxin autoregulates the relBEF operon at the level of transcription. Using database searching, we found relBE homologues on the chromosomes of E. coli K-12, Haemophilus influenzae and Vibrio choleraeA fifth relBE homologue was identified on the enterotoxin encoding E. coli plasmid P307. Indirect evidence suggests that the toxicity of RelE may be related to the inhibition of protein synthesis. Based on these observations, we propose a model that explains the delayed relaxed phenotype associated with mutations in relB.


Bacterial toxin–antitoxin systems constitute a diverse group of two-component systems in which one component is a toxin and the other an antitoxin. The plasmid-encoded systems were called ‘proteic plasmid stabilization systems’ and have been studied in some detail (reviewed by Jensen and Gerdes, 1995). Both the plasmid- and the chromosome-encoded systems mediate plasmid stabilization by the killing of plasmid-free cells. Cell killing, and thus plasmid stabilization, is a consequence of the differential stability of the toxins and the antitoxins: the toxins are metabolically stable, whereas the antitoxins are degraded by cellular proteases (Lon or Clp) (Tsuchimoto et al., 1992; Van Melderen et al., 1994; Lehnherr and Yarmolinsky, 1995). As long as both proteins are produced, the antitoxins neutralize their cognate toxins by forming tight complexes with them. However, when production of the proteins halts, as in plasmid-free segregant cells, for example, the instability of the antitoxins leads to the activation of the toxins. This phenomenon was previously coined post-segregational killing (Gerdes et al., 1986a), and the phenotype is basically similar to that connected with the antisense RNA-regulated post-segregational killing systems (i.e. the hok-like killer genes, as reviewed by Gerdes et al., 1990; 1997).

The genetic organization of the proteic plasmid stabilization systems is paradigmatic: the antitoxin and toxin genes are encoded by adjacent genes in operons and, in all cases but one, the antitoxin-encoding genes are located upstream of the toxin genes (Jensen and Gerdes, 1995; Tian et al., 1996). In all cases, the antitoxins exert negative transcriptional autoregulation of the operons and, in many, the toxins were found to act as co-repressors of transcription (Jensen and Gerdes, 1995).

The plasmid-encoded proteic systems (ccd of F, pem/parD of R100/R1, parDE of RK2, phd/doc of P1) were identified on account of their ability to prevent the formation of plasmid-free progeny (Ogura and Hiraga, 1983; Bravo et al., 1987; Tsuchimoto et al., 1988; Lehnherr et al., 1993; Roberts et al., 1994). The cellular targets of the toxins are known in two cases only. Thus, CcdB of F is a DNA gyrase poison that traps the gyrase in an inactive complex (Bernard and Couturier, 1992; Bernard et al., 1993). The kid (killing determinant) protein of parD of R1 interacts with DnaB and thereby prevents initiation of replication at oriC (Ruiz-Echevarría et al., 1995).

On the basis of sequence similarity, a number of chromosomally encoded toxin–antitoxin systems have been identified. Thus, the E. coli K-12 chromosome encodes two systems, denoted chpA and chpB, that are homologous to pem of plasmid R100 (Masuda et al., 1993). When present on plasmids, the chpA and chpB genes exert plasmid stabilization caused by post-segregational killing. The chpA locus (also called mazEF ) is located just downstream of the relA gene (Metzger et al., 1988). Physiological experiments have suggested that, during amino acid starvation, the mazEF promoter is inhibited by a (p)ppGpp-dependent mechanism (Aizenman et al., 1996). Therefore, it has been proposed that amino acid starvation might lead to the induction of the mazEF system and cell killing. The chpA/mazEF system may thus specify a function that is beneficial to the host cell or to the cell culture during starvation and perhaps other conditions of stress.

The E. coli relB operon encodes three genes, relB, relE and relF (Bech et al., 1985; see Fig. 1). We have shown previously that relF encodes a hok homologue and that production of the RelF protein leads to rapid cessation of cell growth, arrest of respiration and collapse of the cell membrane potential (Gerdes et al., 1986b). Therefore, the relF gene was renamed hokD (Gerdes et al., 1997). The physiological significance of the RelF protein is not known, but translation of relF is probably inhibited post-transcriptionally, as significant amounts of the relBEF mRNA is present in growing cells (Bech et al., 1985).

Figure 1.

. Genetic structure of the E. coli relBEF operon at 34.4 min of the E. coli map. ttp denotes the transcriptional termination point downstream of the relF (hokD ) gene. Arrow pointing right indicates the relB promoter mapped by Bech et al. (1985). Relevant restriction sites are shown. The EcoRI site shown to the left is from a synthetic multiple cloning region.

The relB gene was defined by mutations that conferred a so-called ‘delayed relaxed’ phenotype upon host cells (Lavallé, 1965; Lavalléet al., 1976; Diderichsen et al., 1977; Bech et al., 1985). The delayed relaxed mutants resume synthesis of stable RNA (tRNA and rRNA) approximately 10 min after the initiation of amino acid starvation. This is in contrast to relaxed mutants (defective in relA), in which stable RNA synthesis continues after amino acid starvation without any lag. Furthermore, the relB mutants recovered very slowly after starvation (i.e. virtually no growth took place for about 3 h after release from amino acid starvation). This inhibition of cell growth was attributed to the accumulation of a factor in relB mutants that inhibited translation, most probably a protein (Lavalléet al., 1976). From indirect experiments, Bech et al. (1985) suggested that relB did not encode the translational inhibitor itself, but rather a negative regulator of the inhibitor.

Here, we show that the second gene in the relB operon, relE, encodes a cytotoxin whose overproduction is lethal to host cells. The relB gene was found to encode an antitoxin that prevents the lethal action of RelE. When present on a plasmid, the relBEF operon was able to stabilize the inheritance of a mini-R1 test plasmid. Furthermore, the relB operon was found to be autoregulated by RelB, and RelE acted as a co-repressor of transcription. By database searching, we found four additional relBE homologous gene systems on chromosomes and plasmids from Gram-negative bacteria. These results show that the relBE genes constitute a new family of gene systems that belongs to the proteic plasmid stabilization systems described previously by Jensen and Gerdes (1995).


The relE gene encodes a cytotoxin

The low-copy-number cloning vector pNDM220 contains lacI q and the LacI-regulated pA1/O4/O3 promoter (Lanzer and Bujard, 1988) upstream of a multiple cloning site. Without IPTG, transcripts from the promoter are not detectable by Northern analysis (data not shown). However, with IPTG, strong transcription is induced towards the cloning site. The relE gene of E. coli K-12 (Bech et al., 1985) was polymerase chain reaction (PCR) amplified and inserted into the multiple cloning site of pNDM220, resulting in pMG223 (for the construction of plasmids, see Experimental procedures). Plasmid pMG223 was established in MC1000, which contains a chromosomal copy of the relBEF operon. However, it was not possible to transform pMG223 into strain JS115, which carries a deletion of the relBEF operon (ΔrelB ). Therefore, the induction experiments described below were performed in MC1000.

Strain MC1000/pMG223 was grown in LB at 37°C. After 2 h of induction with IPTG, the viable counts decreased approximately 600-fold (Fig. 2B). The decline started almost immediately and continued exponentially. However, growth resumed after approximately 2 h. On plates containing IPTG, viable counts decreased even further (data not shown). The optical density (OD450) increased during the first 20 min after the addition of IPTG, and then the culture became stationary (Fig. 2A). The addition of IPTG to growing cells containing the vector plasmid had no effect (not shown). These results indicate that the relE gene encodes a cell toxin or an inhibitor of cell growth.

Figure 2.

. Demonstration of toxin and antitoxin activity of RelE and RelB respectively. Optical density at 450 nm and viable counts as a function of time for strains MC1000/pMG223 (relE +) (A and B), MC1000/pMG223/pMG2201 (relB  control plasmid) (C and D) and MC1000/pMG223/pMG2202 (relB + plasmid) (E and F) are shown. At time zero, transcription of relE on plasmid pMG223 was induced by the addition of IPTG (1 mM). Filled symbols indicate that IPTG was added. As seen in (E) and (F), the presence of relB on a second plasmid counteracted relE-mediated cell killing.

The relB gene encodes an antitoxin

Plasmid pMG2202 is a pBR322 derivative that contains the relB gene expressed from its own promoter. Strains MC1000/pMG223 (pA1/O4/O3::relE +)/pMG2202 (relB +) and MC1000/pMG223/pMG2201 (relB ) were subjected to a physiological growth experiment as described above. As seen from 2Fig. 2E and F, the presence of the high-copy-number relB-carrying plasmid prevented relE-dependent cell killing. The antitoxin effect was dependent on an intact relB reading frame, as a plasmid carrying the promoter region and the first part of the relB reading frame did not counteract the relE-mediated cell killing (Fig. 2C and D).

The relB gene encodes a protein of approximately 10 kDa

To detect protein products from the relBE reading frames, coupled in vitro transcription/translation experiments were performed on plasmids containing either relB (pMG221), relE (pMG223) or both (pMG224). As seen in Fig. 3, a product of approximately 10 kDa was produced by pMG221. This is in accordance with the predicted size of 9.1 kDa for RelB and with the previously published size of RelB (Bech et al., 1985). Despite several attempts, we were not able to detect a relE-specified protein (expected size of 11.2 kDa), even when pUC19-derived templates were used in the transcription/translation experiments. A possible explanation for this could be that RelE co-migrates with another peptide. Alternatively, the protein is not produced in sufficient amounts by the in vitro reactions to be detected by our gel system. That relE is indeed translated was shown by relE–lacZ translational fusions (see below).

Figure 3.

. Visualization of the RelB protein by coupled in vitro transcription–translation. The plasmids used were: pNDM220 (control), pMG221 (relB ), pMG223 (relE ) and pMG224 (relBE ). The proteins were labelled with [35S]-methionine. The arrow indicates the position of RelB in the tricine-SDS–polyacrylamide gel.

Unexpectedly, the intensity of the protein bands produced by the relE-encoding plasmid was always significantly lower than those produced by the other plasmids (pMG223, lane 3 in Fig. 3). Furthermore, significant amounts of low-molecular-weight polypeptides were also detected in the reaction containing the relE-encoding plasmid. As this observation was reproducible, it may indicate that RelE interferes with the in vitro transcription/translation reaction. Alternatively, RelE could mediate non-specific degradation of the polypeptides produced in the reaction. When a relBE-carrying plasmid was used in the in vitro translation reaction (pMG224), no decrement in band intensity was observed (Fig. 3, lane 4). This observation is consistent with the in vivo antitoxin effect of RelB described above.

The relBEF operon stabilizes a mini-R1 replicon

Because of the structural similarity between the relBE genes of the relBEF operon and the proteic plasmid stabilization systems, we investigated whether the operon could stabilize a test plasmid. The relBEF operon was cloned into the unstable mini-R1 plasmid pOU82, and the loss frequency of the plasmid was measured in a strain in which the relBEF and lac operons had been deleted (JS115Δ). As seen in Fig. 4, a fourfold stabilization was observed for the relBEF-carrying plasmid (pMG8204). Thus, the relBEF operon was indeed able to stabilize a plasmid replicon. When a frameshift mutation was introduced into relE (pMG8204mE), the ability to stabilize the mini-R1 replicon was abolished, whereas a frameshift in relF (pMG8204mF) had no effect. These results indicate that the stabilization phenotype is dependent on relE but not on relF.

Figure 4.

. Genetic stabilization of pOU82 by the relBEF operon of E. coli K-12. The plasmids used were pMG8204 (carrying relBEF ), pMG8204mE (pMG8204 with a frameshift mutation in relE ), pMG8204mF (pMG8204 with a frameshift mutation in relF ) and pOU82, the unstable mini-R1 control plasmid. A. Percentage of plasmid-carrying cells as a function of number of generations grown without selection for the plasmid. B. The loss frequencies and folds of stabilization compared with pOU82. Numbers in (B) are averages of four independent experiments and showed less than 5% variation.

The relBEF operon did not confer plasmid stabilization in wild-type (wt) E. coli strains (which have a copy of relBEF on their chromosome). This is consistent with the post-segregational killing model, as the continued synthesis of RelB antitoxin in plasmid-free cells should prevent the activation of RelE (see Discussion).

The relB gene encodes an autoregulator of transcription

Different segments of the relBEF operon were inserted into pOU254, a low-copy-number lacZ transcriptional fusion vector. The fusion plasmids were designated pMG4001 (relB′–lacZ ), pMG4002 (relBE′–lacZ ), and pKG4003 (relBElacZ ), see 5Fig. 5A. The basal activity of the promoter when both RelB and RelE were absent (pMG4001) was 8400 Miller units in the ΔrelB Δlac strain (JS115Δ), indicating that prelB is relatively strong. The promoter activity was reduced approximately 130-fold in the presence of relB in cis (pMG4002) and 70-fold in the presence of relB in trans on a pBR322 plasmid (pMG4001 + pMG2202; see Fig. 5A). This indicates that RelB is a repressor of the relB promoter.

Figure 5.

. Analysis of the regulation of the relB promoter by lacZ transcriptional and translational fusions. lacZ activities of the E. coli strain JS115Δ (ΔrelBΔlac) containing the plasmids shown. A. Regulation in cis by relB and relE. B. Regulation in trans. An asterisk (*) denotes a frameshift mutation in relE. A prime indicates that the gene is truncated. prelB and pA1/O4/O3 denote the relB promoter and the pA1/O4/O3 promoter respectively. The fold of repression is calculated relative to the activities of pMG4001 and pKG4001 for transcriptional fusions and translational fusions respectively. The β-galactosidase activities are averages of at least four independent measurements.

Using translational instead of transcriptional lacZ fusions, comparable folds of repression were obtained, thus indicating that the regulation of the relBE genes occurs at the level of transcription (i.e. pKG4001 versus pKG4002 in Fig. 5A).

The relE gene encodes a co-repressor of transcription

A transcriptional fusion containing the intact relBE genes upstream of lacZ (pKG4003 in Fig. 5A) yielded a very low level of expression (approximately 1 unit), corresponding to an 8400-fold reduction in transcription rate (compared with the unrepressed promoter). This result indicates that relE might encode a co-repressor of transcription. To obtain more solid evidence for this inference, a frameshift mutation was introduced into the relE gene of pKG4003, resulting in pKG4003mE. As seen in 5Fig. 5A, the mutation resulted in a 34-fold increase in expression, thus supporting the view that RelE indeed acts a co-repressor.

To conduct trans donation experiments, we constructed plasmids that overexpressed RelB (pMG2221) or RelB plus RelE (pMG2224) from the strong pA1/O4/O3 promoter (Lanzer and Bujard, 1988). By using this foreign promoter, we avoided having the autoregulation loop of the relBEF operon interfere with the expression levels of the RelB and RelE proteins. The β-galactosidase activity expressed by the transcriptional fusion plasmid pMG4001 was measured in the presence of pMG2221 and pMG2224. As seen in 5Fig. 5B, trans donation of RelB alone led to a 20-fold repression of the promoter, whereas RelB plus RelE reduced expression even further (34-fold). Taken together, these results indicate that the product of relE is a co-repressor of the relB promoter.

A new toxin–antitoxin gene family in Gram-negative bacteria

Database searching using the BLAST program (Altschul et al., 1990) revealed four complete gene systems that are homologous to relBE of E. coli K-12. The four systems were located on E. coli K-12 (dinJ and yafQ ; Blattner et al., 1997), H. influenzae (relB and relE ; Fleischmann et al., 1995), V. cholerae (ORF9 and ORF11; Franzon et al., 1993) and the enterotoxin-encoding E. coli plasmid P307 (relB and relE ; Saul et al., 1989). In all cases, the antitoxin homologues were located upstream of the toxin-encoding genes. A fifth relB homologue without a corresponding toxin partner was identified on the chromosome of E. coli K-12 (yafN ; Blattner et al., 1997). These new putative toxins and antitoxins are aligned in 6Fig. 6A and B, and their properties are listed in Table 1. The degree of identity ranges from 20% to 48% among the RelB homologues and from 14% to 55% among the RelE homologues. All the systems are located on the chromosomes of related Gram-negative bacteria, except for the one on the P307 plasmid. The dinJyafQ system and the relB homologue (yafN ) are present in the same region of the E. coli chromosome. Four of the amino acid positions in the RelE alignment are fully conserved (Leucine-44 and -55, Glycine-59 and Arginine-61; Fig. 6B). The alignment of the RelB homologues shows that these proteins are considerably more divergent than the RelE homologues (Fig. 6A).

Figure 6.

. Alignments of the RelB and RelE homologues found by BLAST searches. Amino acids identical in at least 50% of the sequences are shaded black. If more than 50% of the amino acids at a given position are similar, a light grey background is used. A. Alignments of the six RelB homologues. The homologues have the following Entrez accession numbers (from top to bottom): 132283, 984582, 984588, 516610, 1075064 and 481093. B. Alignments of the five RelE homologues that have the following Entrez accession numbers: 132284, 984581, 516611, 1175293 and 396846.

Table 1. . relB and relE homologues. a. The molecular weights of the proteins were calculated from the predicted amino acid sequences.Thumbnail image of

The sizes and pIs of the proteins are remarkably similar. Thus, the molecular weights of the antitoxins (RelB homologues) range from 9.1 kDa to 11.2 kDa, and they are all very acidic (pIs from 4.8 to 5.5). Likewise, the toxins (RelE homologues) are also of similar sizes (10.8–11.9 kDa), but very basic (pIs from 9.5–9.9). The only exception is RelE of H. influenzae, which has a pI of 6.9. The highly different pI values of the toxins and antitoxins may indicate that these proteins, as is the case with several proteins described in proteic plasmid stabilization, are able to interact directly with their cognate protein partner (Tam and Kline, 1989; Ruiz-Echevarría et al., 1995; Johnson et al., 1996).


The observations described here show that the relBEF operon of E. coli K-12 specifies a toxin–antitoxin system: (i) relE encodes a cytotoxin that is lethal or inhibitory to host cells; (ii) relB encodes an antitoxin that prevents the lethal action of the toxin; (iii) the relBEF operon stabilizes a mini-R1 test plasmid in a strain deleted of the relBEF operon but not in a wild-type strain carrying relBEF on the chromosome; (iv) the RelB antitoxin autoregulates the relBEF operon at the level of transcription; and (v) the RelE toxin acts as a co-repressor of transcription.

The plasmid stabilization phenotype was dependent on active RelE toxin (Fig. 4). Thus, our observations suggest that plasmid stabilization is a consequence of post-segregational killing (i.e. killing of plasmid-free cells). The phenotype can be explained if the RelB antitoxin is more unstable than the RelE toxin: cells that lose the relBEF-carrying plasmid at cell division experience decay of the antitoxin, which thus leads to activation of the toxin and killing of the cells. In several cases, this simple model has been shown to be valid for the proteic plasmid stabilization systems (Jensen and Gerdes, 1995). Furthermore, the model is in accordance with the observation that relBEF does not stabilize the mini-R1 plasmid in wild-type strains (relB +) that continue to produce RelB antitoxin even in plasmid-free cells. We are now attempting to identify factors that might lead to activation of the relE-encoded toxin.

The finding of several chromosomally encoded homologues of relBE was surprising. Thus, E. coli K-12 contains a complete relBE homologue (dinJyafQ ) and a third relB homologue (yafN ) that has no closely linked toxin partner. The E. coli plasmid P307 contains an relBE homologue that mediates stabilization of P307 mini-replicons (H. Grønlund, unpublished observations). Furthermore, the chromosomes of H. influenzae and V. cholera contain complete relBE homologues. The E. coli dinJ gene (damage inducible) was identified in a computer search for LexA binding sites (Harlow et al., 1994). A sequence with strong similarity to known LexA binding sites was found in the region between the −10 element of a potential dinJ promoter and the Shine–Dalgarno sequence of dinJ. The presence of a LexA binding site leaves the possibility that dinJ–yafQ expression could be induced during the SOS response.

The E. coli relB operon was discovered as a result of the observation that certain point mutations in relB resulted in the so-called delayed relaxed response. Wild-type cells exhibit stringent control of stable RNA synthesis after amino acid starvation, i.e. the rate of synthesis of stable RNA is rapidly and severely reduced (reviewed in Cashel et al., 1996). The stringent response is dependent on starvation-induced synthesis of (p)ppGpp, and relA mutants (defective in (p)ppGpp accumulation) exhibit the so-called relaxed phenotype, i.e. the synthesis of stable RNA continues at an almost normal rate after amino acid starvation (Stent and Brenner, 1961). A relaxed phenotype can also be seen if an otherwise stringent strain is subjected to certain inhibitors of protein synthesis (Kurland and Maalø, 1962). The relB mutants isolated previously by different laboratories (Lavallé, 1965; Mosteller and Kwan, 1976; Diderichsen et al., 1977) exhibited the so-called delayed relaxed response, in which synthesis of stable RNA resumes approximately 10 min after the initiation of amino acid starvation (reviewed in Cashel et al., 1996). The relB mutants exhibited another characteristic phenotype: after termination of amino acid starvation (by refeeding with amino acids), the mutants resumed growth very slowly, suggesting that a growth-inhibiting substance may have accumulated during the period of starvation.

Based on our observations, we propose a plausible and testable model that explains the molecular basis of the delayed relaxed response. The relB mutants were selected by a scheme that favoured cells that would recover slowly after a period of induction of the of the stringent response (i.e. by the addition of 5-fluorouracil and subsequent ampicillin enrichment during the recovery phase) (Diderichsen et al., 1977). The promoter mapping studies by Bech et al. (1985) showed that the relB101 mutation had an increased level of relBEF mRNA during steady state, indicating that the mutation reduced the repressor activity of RelB. Cells carrying the relB101 mutation therefore probably contain increased amounts of RelE. During amino acid starvation, the rate of protein synthesis in general is severely reduced. Therefore, the intrinsic (and presumably increased) instability of the RelB101 mutant protein could lead to the activation of RelE. If RelE is a translational inhibitor, then the activation of RelE would lead to a decreased protein synthesis, which, in turn, would lead to recharging of tRNA and shutdown of (p)ppGpp synthesis (which is dependent on vacant A-sites on the ribosomes). This, then, would lead to the resumption of stable RNA synthesis as observed in the delayed relaxed response. Consistently, the slow resumption of protein synthesis after refeeding a relB mutant strain that has been starved of amino acids could be explained by the persistence of increased levels of activated RelE inhibitor. Thus, although we do not exclude other explanations, the delayed relaxed response may now be best explained by the activation of a translational inhibitor during amino acid starvation. We are now addressing this hypothesis experimentally.

The function of the chromosomally encoded homologues of the proteic plasmid stabilization systems is not yet known. The mazEF system was proposed to mediate programmed cell death during amino acid starvation (Aizenman et al., 1996). It has been speculated that such stress-induced cell killing might be beneficial to the cell population at large (Nyström, 1998). In the same line, induction of the relBE genes during the stringent response may have some, as yet uncovered, beneficial function. We are now pursuing this question further.

Experimental procedures

Bacterial strains

The E. coli K-12 strain MC1000 (Casadaban and Cohen, 1980), which contains a chromosomal copy of the relBE genes, was used as the standard cloning strain and when a chromosomal copy of the relB operon was required. The E. coli K-12 strain JS115 (leu, thy, thi, supE, ΔrelB ), which contains a deletion covering the entire relB operon, was kindly provided by J.-P. Bouche. JS115Δ is a P1 transduction derivative of JS115 containing a deletion of the lac operon. Thus, JS115Δ is ΔrelBEF and Δlac. The latter strain was used for the regulatory studies of relBE.

Plasmids used (Table 2)

Table 2. . Plasmids used and constructed. Ap, ampicillin resistance; Tc, tetracycline resistance. *A frameshift mutation (Yanisch-Perron et al. (1985); Bolivar (1978); Thisted and Gerdes (1992)).Thumbnail image of

Plasmid pOU254 is a low-copy-number mini-R1 transcriptional fusion vector carrying the lacZ gene of pTL25 (Linn and Ralling, 1985). Plasmid pOU253 is a corresponding translational fusion vector isogenic with pOU254. Both fusion vectors are segregationally stable because of the parA system of plasmid R1 (Dam and Gerdes, 1994).

Plasmid pNDM220 is a low-copy-number mini-R1 expression vector carrying a multiple cloning site placed between the LacI-regulated pA1/O4/O3 promoter (Lanzer and Bujard, 1988) and two transcriptional terminators.

Plasmid pBD2430 is a pUC8 derivative carrying the complete relBEF operon and gene IV located downstream of relF (O. H. Karlström, unpublished). The E. coli DNA present in pBD2430 is shown in Fig. 1.

Plasmids constructed (Table 2)

Plasmid pKG4001.

pBD2430 was digested with EcoRI and XhoI, and the fragment carrying the relB promoter (Fig. 1) was inserted into pOU253 producing an in frame translational fusion between relB and lacZ. Thus, pKG4001 carries a relB ::lacZ translational fusion.

Plasmid pKG4002.

pBD2430 was digested with EcoRI and Bst 1107I (Fig. 1), and the resulting fragment was inserted into pOU253 producing an in frame translational fusion between relE and lacZ. Thus, pKG4002 carries relB and a relE ::lacZ translational fusion.

Plasmid pKG4003.

pBD2430 was digested with EcoRI and HincII, and the fragment carrying the relBE operon, truncated between relE and relF (Fig. 1), was inserted into EcoRI–Bbr PI-digested pOU254. The resulting plasmid carries a transcriptional fusion of the relBE operon to lacZ.

Plasmid pMG221.

relB was amplified by PCR on pDB2430 (Fig. 1) with the primers relB1 (5′-CCCCCGGATCCGTAATTACAAGAGGTGTAAGAC) and relB2 (5′-CCCCCCTCGAGGTCGACTCAGAGTTCATCCAGCGTCACACGTACTGG). The resulting fragment was cloned into pNDM220 using the BamHI and Sal I restriction sites, producing a RelB expression plasmid.

Plasmid pMG223.

relE was amplified by PCR on pBD2430 (Fig. 1) with primers relE1B (5′-CCCCGGATCCATAAGGAGTTTTATAAATGGCGTATTTTCTGGATTTTGACG) containing the parA Shine–Dalgarno sequence (Gerdes and Molin, 1986) and relE2 (5′-CCCCCCTCGAGGTCGACTCAGAGAATGCGTTTGACCGC-3′). The resulting relE-carrying fragment was inserted into pNDM220 using the BamHI and Sal I restriction sites. Plasmid pMG223 expresses RelE upon the addition of IPTG.


relB and relE were amplified by PCR on plasmid pBD2430 (Fig. 1) with primers relB1 (5′-CCCCCGGATCCGTAATTACAAGAGGTGTAAGAC-3′) and relE2 (5′-CCCCCCTCGAGGTCGACTCAGAGAATGCGTTTGACCGC-3′) and inserted into pNDM220 using the BamHI and Sal I restriction sites. Plasmid pMG224 produces RelB and RelE upon induction with IPTG.

Plasmid pMG1904.

pBD2430 was digested with MluI and filled with T4 DNA polymerase. The plasmid was then digested with EcoRI, and the resulting fragment carrying relB, relE and relF (Fig. 1) was inserted into EcoRI–SmaI-digested pUC19, producing pMG1904. This plasmid carries the entire relBEF operon on an EcoRI–BamHI restriction fragment.

Plasmid pMG2201.

This plasmid contains the EcoRI–Eco47III fragment from pBD2430 (Fig. 1) inserted between the EcoRI and ScaI sites of pBR322. Plasmid pMG2201 carries the relB promoter.

Plasmid pMG2202.

pBD2430 was digested with EcoRI and Bst 1107I, and the relB-carrying fragment (Fig. 1) was inserted into pBR322 EcoRI–ScaI. The resulting plasmid carries the relB promoter and relB.

Plasmid pMG2203.

pDB2430 was digested with EcoRI and Pst I, and the fragment containing relBE (Fig. 1) was inserted into pBR322 digested with EcoRI and Pst I. The resulting plasmid carries the relB promoter, relB and relE and is compatible with the mini-R1 derivatives.

Plasmid pMG2221.

pMG221 was digested with Aat II and Pst I, and the resulting fragment was inserted into Aat II–Pst I-digested pBR322. Thus, pMG2221-carrying cells expresses RelB upon the addition of IPTG.


pMG224 was digested with Aat II and Pst I, and the resulting fragment was inserted into Aat II–Pst I-digested pBR322. Thus, pMG2224-carrying cells expresses RelB and RelE upon the addition of IPTG.

Plasmid pMG8204.

pMG1904 was digested with EcoRI and BamHI, and this fragment carrying relB, relE and relF was inserted into EcoRI–BamHI-digested pOU82, which is a segregationally unstable mini-R1 plasmid.

Plasmid pMG8204mE.

pMG1904 was digested with BanII, the ends were polished with T4 DNA polymerase and the plasmid was religated. The resulting plasmid carries the relBEF operon with a 4 bp deletion in relE. The resulting plasmid, pMG1904mE, was then digested with EcoRI and BamHI, and the relBEF fragment was inserted into EcoRI–BamHI-digested pOU82.

Plasmid pMG8204mF.

pMG1904 was digested with EcoNI, the ends were filled with T4 DNA polymerase and the plasmid was religated. The resulting plasmid carries the relBEF operon with a 1 bp insertion in relF. The resulting plasmid, pMG1904mF, was then digested with EcoRI and BamHI, and the relBEF fragment was inserted into EcoRI–BamHI-digested pOU82.

Plasmid pMG4001.

pBD2430 was digested with EcoRI and XhoI, and the resulting fragment carrying the relB promoter and part of relB (Fig. 1) was inserted into pOU254. The resulting plasmid carries a transcriptional fusion between the relB promoter and lacZ.

Plasmid pMG4002.

pBD2430 was digested with EcoRI and Bst 1107I, and this fragment carrying the relB promoter, relB and part of relE (Fig. 1) was inserted into EcoRI–Bbr PI-digested pOU254. The resulting plasmid carries a relB + transcriptional fusion.

Growth media and antibiotics

The growth medium was LB medium (Bertani, 1951) or A + B minimal medium (Clark and Maalø, 1967) supplemented with 0.2% glucose and 1% casamino acids. For growth on solid media, LA plates were used. LA is LB containing 15 g l−1 agar. All media were supplemented with 50 μg ml−1 thymine for growth of the strain JS115ΔrelB. Antibiotics were added at the following concentrations: ampicillin, 30 μg ml−1; and tetracycline, 10 μg ml−1. When indicator plates were used, Xgal (5-bromo-4-chloro-3-indolyl-β-D-galactoside) was added to a final concentration of 40 μg ml−1.

Conditions of cell growth

Cells were diluted in LB + antibiotics from an overnight culture to an OD450 of 0.005. The cultures were grown at 37°C until an OD450 of 0.4, and then diluted to an OD450 of 0.01 in 37°C LB containing 1 mM IPTG and antibiotics. Samples for OD450 measurements and viable counts were taken at the time points indicated. Viable counts were made by plating dilutions of the cultures onto LA plates containing the proper antibiotics.

Coupled in vitro transcription and translation

The reactions were performed using the E. coli S30 extract system for circular DNA as described by the supplier (Promega). DNA (4 μg) was used in all reactions. The reactions were run on a 16% Tricine–SDS–PAGE gel essentially as described by Schäger and von Jagow (1987).

Loss frequencies

Loss frequencies (LF) were determined as described by Gerdes et al. (1985). The cells were grown in non-selective medium, and the fraction of plasmid-bearing cells was determined by plating on non-selective indicator plates containing Xgal.

β-Galactosidase assays

β-Galactosidase assays were performed essentially as described by Miller (1972).

Homology search

BLAST searches were performed at the GENESTREAM BLAST network server CRBM, Montpellier, France. Standard conditions were used, except that the Blossum 80 matrix was used.

Note added in proof

We discovered recently that re/BE homologues are abundant in Gram-positive bacteria and in Archae (H. Grønlund and K. Gerdes, unpublished).


We thank Hyon Choy for the construction of JS115Δlac, and Rasmus Bugge Jensen for suggestions for improvement of the manuscript. Strain JS115 (ΔrelBEF ) was donated by J.-P. Bouche. This work was supported by the Center for Interaction, Structure, Function and Engineering of Macromolecules of the Danish Biotechnology Program.