Two higBA loci in the Vibrio cholerae superintegron encode mRNA cleaving enzymes and can stabilize plasmids


E-mail; Tel. (+45) 65 50 24 13; Fax (+45) 65 50 24 67.


Vibrio cholerae codes for 13 toxin–antitoxin (TA) loci all located within the superintegron on chromosome II. We show here that the two higBA TA loci of V. cholerae encode functional toxins, HigB-1 and HigB-2, whose ectopic expression inhibits cell growth of Escherichia coli, and functional antitoxins, HigA-1 and HigA-2, which counteract the toxicity of the cognate toxins. Three hours of ectopic expression of the HigB toxins resulted in bacteriostasis without any detectable loss of cell viability. The HigB toxins inhibited translation by cleavage of mRNA. Efficient mRNA cleavage occurred preferentially within the translated part of a model mRNA and only when the mRNA was translatable. Promoter analysis in V. cholerae and E. coli showed that the two higBA loci are both transcribed into bi-cistronic mRNAs and that the higBA-2 mRNA is leaderless. Transcription of the two higBA loci was strongly induced by amino acid (aa) starvation in V. cholerae and E. coli, indicating that the regulatory mechanisms of transcriptional induction are conserved across the two species. Both higBA loci stabilized a test-plasmid very efficiently in E. coli, raising the possibility that the loci contribute to maintain genetic stability of the V. cholerae superintegron. Based on these results we discuss the possible biological functions of the TA loci of V. cholerae.


Vibrio cholerae is an aquatic bacterium that is pathogenic because of its capability to proliferate in and secrete cholerae toxin into the human small intestine. Therefore, V. cholera cells encounter and must cope with dramatic environmental changes. During transitions from rapid growth conditions (in e.g. the human intestine) to slow growth or dormancy in oligotrophic habitats, bacteria rapidly downregulate their metabolic rates. To rapidly adapt to such dramatic changes in growth conditions, bacteria have evolved sophisticated checkpoint mechanisms that adjust the rates of protein and RNA synthesis. The best understood of these mechanisms is the ‘stringent response’ that is elicited by amino acid (aa) starvation that triggers a rapid increase in the cellular concentrations of penta- (pppGpp) and tetra-guanosinephosphates (ppGpp), here collectively called P4G. When Escherichia coli and related organisms, such as V. cholerae, encounter aa starvation, P4G is synthesized by P4G synthetase I (RelA), which is activated by uncharged tRNA present at the ribosomal A-site (Cashel et al., 1996; Haralalka et al., 2003; Silva and Benitez, 2006). In contrast, carbon-starvation inhibits P4G hydrolase (SpoT) that is responsible for P4G degradation and thereby also causes accumulation of P4G (Murray and Bremer, 1996). The increased P4G level directs RNA polymerase away from the synthesis of stable RNA (rRNA and tRNA) towards the synthesis of biosynthetic operon mRNAs (Paul et al., 2005) thus to allow the cell to synthesize aa de novo. The mechanism by which P4G specifically inhibits stable RNA promoters is now understood at the molecular level (Paul et al., 2004a). The bacterial stringent response was recently reviewed (Paul et al., 2004b; Magnusson et al., 2005).

The molecular changes elicited by an increased P4G level primarily alter the global cellular transcription pattern, whereas a mechanism that directly inhibits translation during nutrient starvation is not known. However, toxin–antitoxin (TA) loci, such as relBE and mazEF of E. coli, reduce the global level of translation during aa starvation (Christensen et al., 2001; 2003). TA loci may thus be regarded as stress-response-elements that fine-tune the global level of translation. It should be emphasized that relBE and mazEF do not block but rather, reduce translation during nutritional stress (Christensen et al., 2001; 2003).

Toxin–antitoxin loci invariably consist of two genes in an operon encoding a ‘toxin’ and an ‘antitoxin’ respectively (Gerdes et al., 2005). The downstream gene codes for the toxin and the upstream codes for the antitoxin, except in the cases of higBA loci, which exhibit a reversed gene order (Tian et al., 1996a;Pandey and Gerdes, 2005). Ectopic expression of the toxins, such as RelE or MazF, severely reduces cell growth and prevents colony formation. However, cell viability can be rescued by later overproduction of the cognate antitoxins (Pedersen et al., 2002). Thus, the toxins are bacteriostatic rather than bacteriocidal. The antitoxins combine with and neutralize the toxins by direct protein–protein contact. For example, MazF toxin and MazE antitoxin form an F2-E2-F2 heterohexameric complex that is non-toxic (Kamada et al., 2003). The antitoxins are DNA binding proteins that bind to their promoter-regions and autoregulate transcription of their own TA operon. In most cases, the toxins act as co-repressors of transcription, that is, the TA complexes usually repress TA operon transcription much more efficiently than the antitoxins alone (Gotfredsen and Gerdes, 1998; Magnuson and Yarmolinsky, 1998; Marianovsky et al., 2001; Zhang et al., 2003a). Thus, during steady-state cell growth, TA operon transcription is usually highly repressed because of efficient binding of the TA complexes to their own promoter regions.

Shifts to oligotrophic growth-conditions confer Lon or Clp protease-mediated degradation of the antitoxins, leading to both an increased transcription-rate of the TA operons and activation of the toxins (Christensen et al., 2001; 2003; 2004). In turn, activation of the toxins reduces the global cellular level of translation. The mechanism of translational inhibition by toxins is of considerable interest. We showed that RelE of E. coli encodes an mRNA cleaving enzyme that recognizes mRNA codons positioned at the ribosomal A-site and cleaves the mRNA between the second and third bases of the codon (Pedersen et al., 2003). Interestingly, purified RelE cleaves mRNA in vitro only when the mRNA is positioned at the ribosomal A-site. In contrast, the RelE homologue YoeB of E. coli and the unrelated MazF toxin both cleave naked mRNA in vitro, independently of the presence of ribosomes (Zhang et al., 2003b; Kamada and Hanaoka, 2005). The molecular basis for this difference is not yet understood. Both RelE and MazF are activated and cleave mRNA during aa starvation but do not reduce cell viability during starvation (Christensen and Gerdes, 2003; Christensen et al., 2003).

The V. cholerae genome consists of two circular chromosomes of 3.0 and 1.1 MB respectively (Heidelberg et al., 2000). The smallest of these, chromosome II, harbours a large integron element of 125.3 kb encoding 216 open reading frames, called the V. cholerae superintegron (SI). The functions of the majority of the genes in the SI are unknown and the genes either have no counterparts in the databases or are homologous to unassigned open reading frames from other organisms. However, a few genes have been characterized and of these, genes involved in adaptive functions such as pathogenecity, antibiotic resistance, immunity proteins and restriction endonucleases seem to be over-represented (Rowe-Magnus and Mazel, 2001). The original annotation of the V. cholerae genome revealed three TA loci, two higBA (host inhibition of growth) and one phd/doc, in the SI (Heidelberg et al., 2000). higBA and phd/doc loci were originally identified on plasmids because of their ability to increase plasmid maintenance (Lehnherr et al., 1993; Tian et al., 1996a). Recently, we found that V. cholerae has in total 13 TA loci (seven relBE, three parDE, two higBA and one phd/doc), all located in the SI (Pandey and Gerdes, 2005). As shown in Fig. 1A, all of the 13 TA loci each has closely linked attC sites, suggesting that the TA loci can be acquired as bona fide integron cassettes.

Figure 1.

Localization and promoters of the two higBA loci in the V. cholerae super-integron.
A. The positions of the 13 TA loci in the V. cholera super-integron are shown relative to the closest to attC site on the right; the two higBA loci are in blue. Arrows indicate direction of transcription of the TA operons. The boundaries of the integron were determined here by the intI4 gene and the rightmost attC site (attC-R). attC-L is the leftmost attC site of a total of 179 attC sites of the super-integron.
B. The promoter region of higBA-1 locus (also called VCA0391/VCA0392) showing the transcriptional start site as mapped in Fig. 7 and the presumed translational start site of higB-1 at an ATG codon. We do not exclude that higB-1 starts at the GTG four codons upstream of the ATG codon.
C. The promoter region of higBA-2 locus (also called VCA0468/VCA0469) showing the transcriptional start site as mapped in Fig. 7. Broken arrows pointing rightwards indicate transcription start points. The −10 and −35 sequences are underlined. Start codons are also indicated. The start codon of higB-2 coincides with the very 5′ end of the mRNA, indicating that the mRNA is leaderless. Opposing arrows in red and blue indicate inverted repeats that may function as operator elements for HigA and/or HigBA complex binding. The start codons of the higB genes suggested here were supported by blast analyses.

By comprehensive database mining we have in total identified 145 higBA loci in ≈ 200 bacterial genomes (D.P. Pandey and K. Gerdes, 2005 and unpubl. data). Initially, hig was discovered as a TA locus that stabilizes plasmid Rts1 (Tian et al., 1996a). As with many other TA loci, HigA antitoxin autoregulates transcription of the higBA operon via binding to the promoter upstream of higB and HigB augments transcriptional repression mediated by HigA (Tian et al., 1996b; 2001). The biological function and molecular target of the chromosome-encoded HigB toxins are not known.

Here, we present a molecular and genetic analysis of the two higBA loci in the V. cholerae SI. We show that both loci code for fully functional proteins, that is, the HigB toxins inhibit the growth of E. coli, and the HigA antitoxins counteract the inhibitory effect of their cognate toxins. Both higBA loci stabilize plasmids very efficiently in E. coli. The V. cholerae HigB toxins, as well as the original HigB toxin of plasmid Rts1, inhibit translation by cleavage of mRNA. Both higBA loci are transcriptionally activated by aa starvation. These results raise the possibility that the V. cholerae TA loci function as stress-response-elements that help the cells survive during shift from rapid cell growth (in the human intestine) to the very slow growth or dormant state characteristic of bacterial cells in oligotrophic aquatic environments. However, our results are also consistent with the hypothesis that TA loci function to stabilize the genes in the V. cholerae SI as suggested previously (Rowe-Magnus et al., 2003).


higBA-1 and higBA-2 are functional TA loci that encode bacteriostatic toxins

The genes higB-1 and higB-2 of the V. cholera SI were inserted downstream of the arabinose-inducible pBAD promoter of pBAD33 (Guzman et al., 1995), resulting in plasmids pMCD3 (pBAD::higB-1) and pMCD6 (pBAD::SD-opt::higB-2). Similarly, pMCD7 contains the higB gene from plasmid Rts1. Viable counts of strains MG1655/pMCD3, MG1655/pMCD6 and MG1655/pMCD7 were followed after the addition of arabinose (Fig. 2A). As seen, viable counts decreased dramatically in cells expressing either of the higB genes, with a ≈ 104-fold reduction 1 h after addition of arabinose. Thus cells that experience HigB production were not able to form colonies, even without further induction of higB transcription on the LA plates (the pBAD promoter was repressed by glucose in the solid medium). Expression of RelE from E. coli yielded a very similar result (Fig. 2A), as described previously (Gotfredsen and Gerdes, 1998). Thus, the higB genes of V. cholerae encode functional proteins.

Figure 2.

higBA-1 and higBA-2 are toxin–antitoxin loci that encode translational inhibitors.
A. Escherichia coli MG1655 strains carrying one of the following plasmids: pBAD33 (vector) (diamonds), pMG3323 (pBAD::relE) (squares), pMCD3 (pBAD::higB-1) (triangles), pMCD6 (pBAD::higB-2) (crosses), or pMCD7 (pBAD::higB) (circles) were grown exponentially in LB medium containing Cml (50 μg ml−1) at 37°C. At time zero, arabinose (0.2%) was added to induce transcription from the pBAD promoter. Cells were diluted and plated on solid medium containing Cml (50 μg ml−1) and glucose (0.2%) at the time points indicated.
B and C. Strains MG1655/pMCD3 (pBAD::higB-1)/pMCD1 (pA1/O4/O3::higA-1) (B) and MG1655/pMCD6 (pBAD::SD-opt::higB-2)/pMCD2 (pA1/O4/O3::higA-2) (C) were grown exponentially in LB medium at 37°C. Transcription of the higB genes was induced with arabinose (0.2%) at time zero. The samples were spread on LA plates containing 30 μg ml−1 Amp, 50 μg ml−1 Cml and 0.2% glucose with IPTG (2 mM; diamonds) or without IPTG (squares).
D. Effect on translation after induction of transcription of higB-1, higB-2 and relE from MG1655/pMCD3 (pBAD::higB-1)/pMCD1 (pA1/O4/O3::higA-1) (triangles), MG1655/pMCD6 (pBAD::SD-opt::higB-2) (crosses), or MG1655/pMG3323 (pBAD::relE) (squares). The strains were grown exponentially in M9 minimal medium and transcription of the toxins was induced by addition of arabinose (0.2%) at time zero. The rates-of-translation were determined at the time points indicated and the pre-induction rates were set to 100%. All curves were produced three times with similar results. The effect of relE transcription on the rate-of-translation was described previously (Christensen et al., 2001).

The putative antitoxin genes higA-1 and higA-2 were inserted into pNDM220 (a low-copy-number R1 vector) downstream of the strong, LacI-regulated promoter (pA1/O3/O4). The strains MG1655/pMCD3 (pBAD::higB-1)/pMCD1 (pA1/O3/O4::higA-1) and MG1655/pMCD6 (pBAD::SD-opt::higB-2)/pMCD2 (pA1/O3/O4::higA-2) were grown exponentially in LB medium. At time zero, arabinose was added to induce higB transcription. Cells were then plated on solid medium with or without IPTG, which induces higA transcription, and glucose (to repress further higB transcription). As seen, viable counts decreased rapidly on plates without IPTG (after induction of the higB genes). In contrast, cells from the same culture could be completely rescued if higA transcription was induced on the plates, even after 3 h of higB induction (Fig. 2B and C, +IPTG). Therefore, both higA genes encode efficient antitoxins, capable of counteracting the bacteriostatic effect mediated by their cognate HigB toxins. There was no decrease in growth rate or viable counts if transcription of the higA and higB was induced simultaneously (data not shown).

higBA loci do not cross-talk

To investigate whether the HigA antitoxins could neutralize non-cognate HigB toxins, MG1655 was transformed with either pMCD3 (pBAD::higB-1) and pMCD2 (pA1/O3/O4::higA-2) or pMCD6 (pBAD::SD-opt::higB-2) and pMCD1 (pA1/O3/O4::higA-1). The cells were grown exponentially and at time zero, higB transcription was induced by addition of arabinose. To assay the effect of the HigA antitoxins, the cells were plated on LA plates with or without IPTG, and the number of colony forming cells were counted (Fig. 3). In both cases, the decline in colony forming units caused by induction of higB transcription was not counteracted by induction of the non-cognate higA gene. Therefore, the two higBA loci do not cross-talk. The lack of cross-talk is consistent with the low degree of similarity between the two loci (HigB-1 and HigB-2 are 26% similar while HigA-1 and HigA-2 are 17% similar).

Figure 3.

higBA loci do not cross-talk. Strains MG1655/pMCD3 (pBAD::higB-1)/pMCD2 (pA1/O4/O3::higA-2) (A) and MG1655/pMCD6 (pBAD::SD-opt::higB-2)/pMCD1 (pA1/O4/O3::higA-1) (B) were grown exponentially in LB medium at 37°C. At time zero, arabinose was added to induce transcription of the higB genes. Culture samples were diluted and spread on LA plates containing 30 μg ml−1 Amp, 50 μg ml−1 Cml and 0.2% glucose with IPTG (2 mM; diamonds) or without IPTG (squares).

higBA1 and higBA-2 of V. cholerae stabilize plasmids in E. coli

The two higBA loci including their putative promoter regions were cloned into the unstable (par-deficient) R1 test-plasmid pRBJ200 that has a loss-rate of approximately 0.02 per cell per generation. The effects of the higBA loci were accurately measured in long-term plasmid-stability assays (80 generations) in the E. coli K-12 strain MC1000. As seen in Fig. 4, both higBA loci stabilized the R1 test-plasmid very efficiently. The higBA-1-carrying plasmid (pMCD10) had a loss-rate of lower than 3 × 10−5 per cell per generation, corresponding to a ≈ 600-fold increase in genetic stability. The higBA-2-carrying plasmid (pMCD12) had a loss-rate of 7 × 10−4 per cell per generation, corresponding to a ≈ 25-fold increase in stability. The plasmid stabilization phenotype has been described previously for both plasmid and chromosomal encoded TA loci, and can be explained by the post-segregational-killing (PSK) model (Jensen and Gerdes, 1995). By inference, these results show that HigB toxins are stable and that the HigA antitoxins are unstable (see Discussion).

Figure 4.

higBA-1 and higBA-2 efficiently stabilize plasmids in E. coli. Genetic stabilization of pRBJ200 by the higBA-1 or higBA-2 operons of V. cholerae. The plasmids used were pMCD10 (carrying higBA-1) (squares) and pMCD12 (carrying higBA-2) (crosses). The logarithm of the percentage of plasmid-carrying cells was plotted as a function of the number of generations grown without selection for the plasmid. LF (loss frequencies)-values were calculated as described (Gerdes et al., 1985).

Both HigB toxins inhibit global translation in E. coli

The molecular targets of the two HigB toxins are unknown. We measured the rates of protein, DNA and RNA syntheses after overproduction of the HigB toxins. It was not possible to grow strain MG1655/pMCD3 (pBAD::higB-1) in minimal medium, probably because of a low level of leaky expression of higB-1 from the pBAD promoter. Therefore, the experiments were carried out using strains MG1655/pMCD3 (pBAD::higB-1)/pMCD1 (pA1/O3/O4::higA-1) and MG1655/pMCD6 (pBAD::SD-opt::higB-2), which were both able to grow in synthetic medium. Figure 2D shows the relative rates-of-translation after induction of transcription of the higB genes. As seen, ectopic transcription of both higB genes rapidly reduced the rate-of-translation to a basal level (< 0.1% for higBI and < 10% for higB-2). The effect of higB-1 and higB-2 transcription on the global level of translation was similar to that of relE transcription (Fig. 2D). The rate-of-replication was also affected by induction of transcription of higB-1 or higB-2, and resembled that of other known inhibitors of translation [i.e. RelE and Chloramphenicol (Cml)], and can be explained by run-out synthesis of chromosome replication in that translation is required to initiate new rounds of replication (data not shown). The rate-of-transcription was not seriously affected by ectopic transcription of higB-1 or higB-2 (data not shown).

Both HigB toxins cleave mRNA

The stability of a model mRNA was monitored during overexpression of the HigB toxins from V. cholerae and HigB from plasmid Rts1. The lpp mRNA of E. coli was chosen as a test substrate, as it has been used in previous analyses of toxins and is a well-characterized, small (322 nt) and relatively stable message that expresses the abundant but non-essential housekeeping lipoprotein (Nakamura et al., 1980; Nilsson et al., 1984). Figure 5 shows Northern analyses of lpp mRNA before and after transcriptional induction of higB of plasmid Rts1 (panel A, right), higB-1 and higB-2 of V. cholerae (panel B). Indeed, the lpp mRNA was rapidly degraded shortly after induction of transcription of the higB genes, with an effect similar to that mediated by transcription of relE (panel A, middle). A non-translated lpp mRNA (start codon changed from AUG to AAG), however, was not cleaved by relE (panel A, middle), in keeping with previous results (Christensen and Gerdes, 2003). Similarly, none of the three HigB toxins cleaved the non-translated version of lpp mRNA (panel B). We conclude that HigB toxins preferentially cleave mRNAs that are translated.

Figure 5.

The higB toxins cleave a translated model lpp mRNA. Cells of MG1655Δlpp/MG3323 (pBAD::relE), MG1655Δlpp/MCD7 (pBAD::higB), MG1655Δlpp/pMCD3 (pBAD::higB-1) and MG1655Δlpp/pMCD6(pBAD::SD-opt::higB-2) containing either of the plasmids pSC710 (wt lpp) or pSC711 (ATG start codon of lpp changed to AAG) were grown exponentially in LB medium at 37°C. To induce transcription of the toxin genes, arabinose (0.2%) was added at time zero. As a control, the strain MG1655Δlpp/pSC710 (wt lpp) was treated with Cml (50 μg ml−1) at time zero, to inhibit translation. Total RNA samples were fractionated by PAGE, and lpp mRNA was visualized by Northern analysis. Numbers are time points of cell sampling relative to inhibition of translation by either Cml or arabinose.

HigB cleavage patterns are similar but not identical to that of RelE

The mRNA cleavage patterns of the HigB toxins were investigated by primer extension analysis of the 5′ end of lpp mRNA before and after induction of higB transcription and compared with that of RelE (Fig. 6). RNA from wild-type cells treated with chloramphenicol was also included as a control for toxin-specific changes in the RNA band-pattern. As seen, the three HigB toxins showed similar, yet distinct specific cleavage patterns. As with RelE, all three HigB toxins cleaved lpp mRNA at several sites within the coding region, whereas no cleavages were observed in the non-translated leader. Most of the HigB-mediated cleavages were located between the second and third base of a codon (HigB of Rts1: AA↓A, AA↓A; HigB-1: AA↓A, AC↓U, AA↓A, GC↓G, GC↓A; HigB-2: AA↓A, AC↓U, AA↓A, CU↓G, CU↓G, GC↓G, CU↓G, CU↓G, CU↓G). Two cleavage sites mediated by HigB-1 were detected between two codons (GUA↓AUC, AGC↓AAC). Only one cleavage site was observed between the first and second base of a codon (HigB-1: U↓GC).

Figure 6.

mRNA cleavage patterns of HigB toxins. Primer extension analysis of lpp mRNA after induction of transcription of relE, higB (Rts1), higB-1 and higB-2 from the following strains: MG1655/pMG3323 (pBAD::relE), MG1655/pMCD7 (pBAD::higB), pMCD3 (pBAD::higB-1) and MG1655/pMCD6 (pBAD::SD-opt::higB-2). Strains were grown exponentially in LB medium at 37°C and total RNA prepared from cell samples taken at the time points indicated after the addition of 0.2% arabinose. Primer extension was performed with primer lpp 21. Cleavages by the different toxins are indicated with the following symbols: RelE (squares), HigB (circles), HigB-1 (triangles) and HigB-2 (diamonds).

Both higBA loci are transcribed from single promoters upstream of the higB genes

To map the higBA promoters, primer extension analyses were performed on RNA prepared from the clinical V. cholerae strain N16961 and from E. coli strain MC1000 containing plasmids pMCD11 (attC higBA-1::lacZ) or pMCD13 (attC higBA-2::lacZ) (see Fig. 7C and D for relevant genetic content of the plasmids). Because the higBA locus of Rts1 has been reported to have promoters immediately upstream of both higB and higA (Tian et al., 1996b), we performed primer extension analysis that would detect promoters immediately upstream of both the higB and the higA genes. Specific bands consistent with transcriptional start sites were detected upstream of both higB genes (phigB-1 in panel A and phigB-2 in panel B of Fig. 7). In both cases, inspection of the DNA sequences upstream of the +1 site revealed the presence of putative −35 and −10 boxes (Fig. 1B and C). Interestingly, the transcriptional start site of higBA-2 coincides with the A base of the higB AUG start codon (Fig. 1C), indicating that the higBA-2 mRNA is leaderless. This conclusion is consistent with the lack of a functional Shine–Dalgarno (SD) sequence upstream of higB-2. It should be noted that there are no other possible in frame start codons in the higB-2 reading frame.

Figure 7.

Promoter analyses.
A and B. Primer extension analyses of the higBA-1 and higBA-2 promoters. Cells of V. cholerae N16961 and E. coli K-12 MC1000 containing one of the plasmids pMCD11 (attC higBA-1::lacZ) or pMCD13 (attC higBA-2::lacZ) were grown exponentially in LB medium at 37°C. At time zero, Cml (50 μg ml−1) or SHT (1 mg ml−1) was added. Primer extension reactions with 32P-labelled primers were performed on total RNA purified at the time points indicated. Primers higBI-PE1 and higBII-PE2 were used for mapping of the higBA-1 and the higBA-2 promoters respectively.
C and D. Different DNA fragments carrying the higBA genes were inserted upstream of lacZ in transcriptional fusions. Putative transcriptional terminators downstream of the higA genes were not included in the fusion, thus to assay promoter activity within the DNA fragments inserted. Plasmids pMCD11 and pMCD13 contain the upstream attC sites, the higB promoters and higBA; pMCD10 and pMCD12 contain the higB promoters and higBA; pMCD14 and pMCD15 contain higBA. attC sites of upstream integron elements (relative to the phigB promoters) are shown as yellow boxes.

Because of the unusual compact genetic organization of integron elements, we also mapped the higBA promoters using lacZ fusions (Fig. 7C and D). This mapping shows that both loci are transcribed from promoters between the upstream attC sites and the translational start codons and is thus consistent with the promoter-mapping by primer-extension. The very low LacZ activity expressed by cells carrying pMCD14 an pMCD15 indicated that there was no significant transcription initiation in the region upstream of (or within) the higA genes. We conclude that the higBA loci are transcribed from single promoters located immediately upstream of the higB genes.

Amino acid starvation induces transcription of both higBA loci

In E. coli, transcription of TA loci, such as relBE and mazEF, is induced when the cells experience nutritional stress (i.e. aa starvation) (Christensen et al., 2001; 2003). We asked whether the higBA were regulated in a similar way. Cells of V. cholerae N16961 and E. coli MC1000 carrying pMCD11 (attC higBA-1::lacZ) or pMCD13 (attC higBA-2::lacZ) were grown exponentially and at time zero exposed to serine hydroxamate (SHT). SHT induces aa starvation by competitive inhibition of seryl tRNA synthetase (Tosa and Pizer, 1971). The promoter activity was monitored by semi-quantitative primer extension reactions (Fig. 7A and B). As seen, the amounts of both transcripts increased significantly after addition of SHT. Thus, transcription of the higBA operons is induced by aa starvation. Inhibition of translation by chloramphenicol also induced higBA transcription (Fig. 7A and B), consistent with the predicted instability of the HigA antitoxins. Therefore, the transcription patterns of the hig loci of V. cholerae are very similar to those of relBE and mazEF of E. coli (Christensen et al., 2001; 2003), both in V. cholerae and E. coli.


Prokaryotic TA loci have a very complex phylogeny and, based on sequence similarities of the toxins they encode, were divided into seven distinct families (Pandey and Gerdes, 2005). Three of the toxin families, RelE, HigB and ParE exhibit weak but significant sequence similarity, indicating a common, ancestral origin (Anantharaman and Aravind, 2003; Pandey and Gerdes, 2005). This was unexpected, given the fact that RelE and ParE have distinct cellular targets (mRNA and DNA gyrase respectively). Recently, we identified 145 higBA loci in ≈ 200 bacterial genomes (Pandey and Gerdes, 2005) (D.P. Pandey and K. Gerdes, unpubl. data). Therefore, it was of considerable interest to identify the cellular target and biological function of the higBA loci. We show here that the two higBA loci in the V. cholerae SI encode functional TA loci that can stabilize plasmids in E. coli very efficiently (Fig. 4). Plasmid stabilization by TA loci has been shown to occur by so-called post-segregational cell killing or inhibition of cell growth (Jensen and Gerdes, 1995; Jensen et al., 1995). This phenotype is a consequence of the differential stabilities of toxins and antitoxins: antitoxins are invariably more unstable than their cognate toxins (Gerdes et al., 2005). Thus, when cells lose a plasmid carrying a TA locus, the new-born plasmid-free cells experience decay of the antitoxin. As a result, the toxin is activated and proliferation of the plasmid-free cells will be inhibited. Therefore, the observed plasmid stabilization by the V. cholerae higBA loci indirectly shows that the HigA antitoxins are metabolically unstable (likely to be degraded by cellular proteases, such as Lon or Clp) and that the HigB toxins are significantly more stable. Moreover, the observation shows that the HigB toxins are activated in plasmid-free segregant cells.

Ectopic expression of the V. cholerae and Rts1 HigB toxins efficiently inhibited the global rate-of-translation in E. coli (Fig. 2 and data not shown). Concomitantly, an abundant housekeeping mRNA (lpp) was rapidly degraded (Fig. 5) by cleavage in its translated region (Fig. 6). Cleavage of the lpp mRNA was abolished (or at least highly reduced) by a single point mutation in the start codon of the mRNA, showing that all three HigB toxins preferentially cleave translated mRNA in vivo. RelE cleavage of lpp mRNA also depended on translation [Fig. 5 and the study by Christensen and Gerdes (2003)]. Thus, although the HigB and RelE toxins exhibited different mRNA cleavage patterns (Fig. 6) they are likely to function by similar biochemical mechanisms. HigB-1 and HigB-2 elicited different cleavage patterns and the cleavage pattern of HigB-2 was more close to that of RelE, in accordance with the fact that HigB-2 is more similar to RelE than HigB-1 (Pandey and Gerdes, 2005). As in the case of RelE, HigB cleavage of mRNA depended on translation (Fig. 5). By contrast, mRNA cleavage by MazF and related toxins apparently was independent of translation (Zhang et al., 2003b; 2004; Munoz-Gomez et al., 2004; 2005) although a non-translated mRNA was cleaved inefficiently by MazF in vivo (Christensen et al., 2003).

The V. cholerae SI is tightly packed with genes (one gene per 600 bp and at least 179 attC sites) (Rowe-Magnus et al., 2003), leaving relatively little intergenic space for regulatory sequences. The start nucleotides of the higBA promoters were identified by primer extension reactions (Fig. 7C and D) and corresponded well with suitable −10 and −35 boxes in the DNA (Fig. 1B and C). Inverted repeats that overlap with the −10 and/or −35 boxes are putative binding sites for HigA and/or the HigBA complexes that may autoregulate transcription of the higBA operons. The non-translated leader of the higBA-1 mRNA consists of 31 bases (Fig. 1B) and has a relatively weak SD sequence (AAG) located five bases upstream of the start codon, suggesting that the higB-1 gene is translated relatively inefficiently. By contrast, the 5′ end of the higBA-2 mRNA coincides with the A base of the predicted start codon of the higB-2 gene (Figs 1C and 7D), indicating that the mRNA is leaderless. Leaderless mRNAs are relatively rare in Gram-negative bacteria (Moll et al., 2002), raising the possibility that the leaderless higBA-1 mRNA might have a specialized function. Alternatively, the leaderless mRNA has evolved as a result of the compact genetic organization of the SI.

The presence of 13 TA loci in the V. cholerae SI raises the obvious question of the cellular function of all these genes. A number of hypothesis that try to explain the biological functions of plasmid- and chromosome-encoded TA loci have been proposed (Gerdes et al., 1986; Hiraga et al., 1986; Kobayashi, 1998; Engelberg-Kulka and Glaser, 1999; Lewis, 2000). Early on, the theory of programmed cell death (PCD) explained how plasmid-encoded TA loci increase plasmid maintenance (Jensen and Gerdes, 1995), and these observations may have misled some researchers to believe that chromosome-encoded TA loci also elicit PCD (Aizenman et al., 1996). Recently, however, we obtained solid evidence that ectopic expression of RelE and MazF is not lethal but rather, induces a bacteriostatic condition that can be reversed by later overproduction of the corresponding antitoxins (Pedersen et al., 2002). Similarly, the HigB toxins induce a bacteriostatic condition that can be reversed by later induction of antitoxin transcription (Fig. 2B and C). Moreover, it is well known that aa starvation does not kill E. coli cells (Cashel et al., 1996), a fact that we also find ample evidence for (data not shown). Thus, we have not found convincing evidence to support the PCD theory.

A second hypothesis, primarily forwarded by Ichizo Kobayashi and coworkers and Didier Mazel and coworkers (Kobayashi, 1998; Rowe-Magnus et al., 2003), suggests that not only plasmid-encoded but also chromosome-encoded TA loci may function to increase the genetic stability of closely linked DNA. Chromosome-encoded TA loci may also stabilize neighbouring DNA because the loss of any given TA locus (by e.g. illegitimate recombination) results in a cessation of antitoxin production, toxin activation and retardation of host cell growth. In this way, cells that lose any given TA locus (and its closely linked loci) will have a selective disadvantage and TA loci and their closely linked regions will therefore be maintained in a growing bacterial population. This hypothesis was conceived by the observation that plasmid-encoded TA loci in many cases increase the maintenance of their replicons. TA loci stabilize plasmids to varying degrees (Boe et al., 1987; Jensen et al., 1995). Thus, ccd of F stabilizes its replicon two- to fivefold (Boe et al., 1987), phd/doc stabilizes P1 eightfold (Lehnherr et al., 1993), parD of R1 stabilizes R1 twofold (Bravo et al., 1987). The relBE locus of plasmid P307 stabilizes an R1 test-replicon c. fivefold (Grønlund and Gerdes, 1999), similar to the stability conferred by the E. coli chromosomal relBE locus (Gotfredsen and Gerdes, 1998). These observations raise the possibility that TA loci, whether they are present on chromosomes or plasmids, may increase the genetic stability (i.e. maintenance) of neighbouring genes. The SI of V. cholerae has ≈ 180 directly repeated, almost identical attC sequences and may thus be recombinogenic. The efficient plasmid stabilization conferred by both V. cholerae higBA loci thus raises the possibility that the TA loci in the SI increase the maintenance of the genes to which they are closely linked, a hypothesis that may now be tested.

A third hypothesis, that is not mutually exclusive with the gene stabilization hypothesis, states that TA loci function as stress-response-elements that help the cells cope with nutrient starvation (Gerdes, 2000). To our view, this hypothesis is supported by experimental data and consistent with phylogenetic data. First, almost all free-living bacteria have numerous TA loci in their chromosomes while obligatory intracellular organisms have very few or none (Pandey and Gerdes, 2005). For example, Mycobacterium tuberculosis (Mtb) has 60 TA loci, while Mycobacterium leprae has none. M. leprae is an obligatory intracellular organism that descended from Mtb by reductive evolution (Cole et al., 2001). Thus, during the evolution into an obligatory intracellular organism, M. leprae lost all TA loci, consistent with the proposal that TA loci are beneficial to Mtb, which has both an extracellular and an intracellular lifestyle. Secondly, nutrient deprivation, such as aa or carbon starvation, induces strong transcription of TA loci (Christensen et al., 2001; 2003; Christensen and Gerdes, 2004) and also induce RelE and MazF mRNA cleavage activity (Christensen and Gerdes, 2003;Christensen et al., 2003). Recently it was shown that transcription of several TA loci in Sulfolobus sofataricus, which has at least 26 TA loci, was induced by heat shock (Tachdjian and Kelly, 2006). As shown here, transcription of the higBA loci of V. cholerae was induced by aa starvation, both in V. cholerae and in E. coli (Fig. 7C and D). Induction of transcription of higBA loci during starvation raises the possibility that HigB toxin activity is also increased in this condition. This is because TA loci are autoregulated by the antitoxins and the TA complex. Increased higBA transcription reflects a lower level of antitoxin and thus predicts a higher free fraction of corresponding toxin. In turn, the HigB toxins may reduce the global level of translation during starvation. These results are thus consistent with the proposal that higBA function as stress-response-elements that are induced by nutrient starvation.

How do the cells benefit from reduced translation? Obviously, cellular metabolism must be tightly regulated during dramatic changes in environmental conditions. Thus, almost all bacteria elicit the P4G-dependent stringent response during aa starvation (Cashel et al., 1996). Cell lacking RelA do not synthesize P4G as a response to aa starvation and exhibit growth defects and increased translational error rates during growth transitions (O'Farrell, 1978). P4G binds to RNA polymerase that confers dramatic changes the global transcription pattern and the effect of P4G on translation is thus indirect. However, TA loci, such as relBE, yefM yoeB, higBA and mazEF are all induced by aa starvation and encode mRNA cleaving enzymes that can reduce the global rate-of-translation. In turn, the reduced translation-rate reduces the drain on charged tRNA and thereby reduces the translational error-rate (Wagner et al., 1982; Sørensen, 2001; Elf et al., 2003). In this view, TA loci function primarily as translational quality control elements.

Yet a fourth hypothesis, primarily forwarded by Kim Lewis and coworkers and Tom Hill and coworkers, states that TA loci may function to induce a dormant cellular condition termed persistence (Keren et al., 2004; Korch and Hill, 2006). Persisters are non-growing, dormant cells and their state may be related to the so-called viable-but-non-cultural condition (VBNC) that is common to environmental bacteria, including V. cholerae (for a review, see Rice et al., 2000). The molecular mechanism(s) that lead to the VBNC state has not yet been identified and the VBNC state may also be viewed as a non-specific condition of cells losing viability (Nyström, 2003a,b). Interestingly, however, TA locus mRNAs were over-represented in cell cultures enriched for persister cells and ectopic expression of RelE dramatically increased the number of cells that exhibited multidrug resistance (persisters) (Keren et al., 2004). Thus, the potential of RelE and similar toxins to inhibit translation raises the possibility that the toxins in some cases serve to induce the VBNC state. If true, the VNBC state induced by TA locus activation may be viewed as an extreme case of growth control in the above-described stress-response-hypothesis.

Experimental procedures

Growth conditions and media

Cells were grown in either Luria–Bertani (LB) broth or M9 minimal medium supplemented with aa in defined concentrations and 0.2% glucose or 0.5% glycerol at 37°C. When appropriate, the medium was supplemented with ampicillin (30 or 100 μg ml−1) or Cml (50 μg ml−1).

Bacterial strains and plasmids are listed in Table 1 and DNA primers in Table 2

Table 1.  Bacterial strains and plasmids used and constructed.
Strains/plasmidsGenotype/plasmid propertiesaReferences
  • a. 

    mcs, multiple cloning site; T7, phage T7 RNA polymerase promoter.

V. cholerae N16961Wild-type Heidelberg et al. (2000)
MG1655Wild-type E. coli K-12 Xiao et al. (1991)
MG1655ΔlppLacks lpp Christensen and Gerdes (2003)
MC1000Δ(ara-leulac rpsL 150 Casadaban and Cohen (1980)
 pNDM220R1 bla lacIq pA1/O4/O3 Gotfredsen and Gerdes (1998)
 pBAD33p15 cat araC pBAD Guzman et al. (1995)
 pRBJ200R1 bla mcs-lacZYA Jensen et al. (1995)
 pMCD1pNDM220 pA1/O4/O3::higA-1This work
 pMCD2pNDM220 pA1/O4/O3::higA-2This work
 pMCD3pBAD33 pBAD::higB-1This work
 pMCD5pBAD33 pBAD::higB-2This work
 pMCD6pBAD33 pBAD::SDopt::higB-2This work
 pMCD7pBAD33 pBAD::higB (from Rts1)This work
 pMCD10pRBJ200 higBA-1This work
 pMCD11pRBJ200 attC higBA-1This work
 pMCD12pRBJ200 higBA-2This work
 pMCD13pRBJ200 attC higBA-2This work
 pMCD14pRBJ200 higBA-1 (promoter less fragment)This work
 pMCD15pRBJ200 higBA-2 (promoter less fragment)This work
 pMG3323pBAD33 pBAD::relE Christensen and Gerdes (2003)
 pSC710R1 bla par lpp Christensen and Gerdes (2003)
 pSC711R1 bla par lpp ATG1AAG Christensen and Gerdes (2003)
 pSC333pGEM3, bla, T7::lpp Christensen and Gerdes (2003)
Table 2.  DNA primers.

Plasmids constructed.  pMCD1. higA-1 of V. cholera strain N16961 (Heidelberg et al., 2000) was amplified from chromosomal DNA with primers higAI-1 and higAI-2. The polymerase chain reaction (PCR) product was digested with BamHI and EcoRI and inserted into pNDM220. Cells carrying this plasmid express HigA-1 upon addition of IPTG.

pMCD2. higA-2 of V. cholera N16961 was amplified from chromosomal DNA with primers HigAII-1 and higAII-2. The PCR product was digested with BamHI and EcoRI and inserted into pNDM220.

pMCD3. higB-1 of V. cholera N16961 was amplified from chromosomal DNA with primers higBI-1 and higBI-2. The PCR product was digested with SalI and HindIII and inserted into pBAD33.

pMCD5. higB-2 of V. cholera N16961 was amplified from chromosomal DNA with primers higBII-1 and higBII-2. The PCR product was digested with SalI and HindIII and inserted into pBAD33.

pMCD6. higB-2 of V. cholera N16961 was amplified from chromosomal DNA with primers higBII-opt and higBII-2. The PCR product, which encodes an optimized SD sequence upstream the start codon of higB-2 was digested with SalI and HindIII and inserted into pBAD33. The resulting plasmid contains the higB-2 gene with an efficient SD sequence (from parM of plasmid R1) downstream of the PBAD promoter.

pMCD7. Gene higB of plasmid Rts1 was amplified from purified plasmid with primers higB-Rts1-up1 and higB-Rts1-down. The PCR product was digested with SalI and HindIII and inserted into pBAD33.

pMCD10. The higBA-1 locus of V. cholera N16961 was amplified from chromosomal DNA with primers higBAI-up and higBAI-down. The PCR product was digested with BamHI and XhoI and inserted into pRBJ200.

pMCD11. The higBA-1 locus, including the upstream attC site, of V. cholera N16961 was amplified from chromosomal DNA with primers higBAI-attC-up and higBAI-down. The PCR product was digested with BamHI and XhoI and inserted into pRBJ200.

pMCD12. The higBA-2 locus of V. cholera N16961 was amplified from chromosomal DNA with primers higBAII-up and higBAII-down. The PCR product was digested with BamHI and XhoI and inserted into pRBJ200.

pMCD13. The higBA-2 locus, including the upstream attC site, of V. cholera N16961 was amplified from chromosomal DNA with primers higBAII-attC-up and higBAII-down. The PCR product was digested with BamHI and XhoI and inserted into pRBJ200.

pMCD14. The higBA-1 locus, without the upstream promoter region of V. cholera N16961 was amplified from chromosomal DNA with primers higBAI-up2 and higBAII-down. The PCR product was digested with BamHI and XhoI and inserted into pRBJ200.

pMCD15. The higBA-2 locus, without the upstream promoter region of V. cholera N16961 was amplified from chromosomal DNA with primers higBA2-up2 and higBA2-down. The PCR product was digested with BamHI and XhoI and inserted into pRBJ200.

Loss frequencies

MC1000 containing the test plasmids (each of which contained the entire β-galactosidase gene) was grown overnight in LB broth with ampicillin selection at 37°C. Overnight cultures were diluted 1/100 000 into fresh LB medium without antibiotics and grown at 30°C with shaking. Every 12 h each culture was re-inoculated at 1/100 000 into fresh LB broth to maintain exponential growth. Aliquots were removed at 12 h, serially diluted and plated on nutrient plates containing 50 mg l−1 of 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (Xgal). Blue plasmid containing colonies and white plasmid-less colonies were counted, and the percentage of plasmid-containing cells was plotted as a function of the number of generations without selection.

Rates of protein, RNA and DNA syntheses

Cells were grown at 37°C in M9 minimal medium + 0.5% glycerol and aa in defined concentrations to an optical density (OD450) of 0.5. The cultures were diluted 10 times and arabinose was added at an OD of 0.3. Samples of 0.5 ml were added to 5 μCi of [35S]methionine (protein synthesis), 2 μCi [methyl-3H]thymidine (DNA synthesis), or 0.1 μCi [2-14C]uracil (RNA synthesis). After 1 min of incorporation, samples were chased for 10 min with 0.5 mg of cold methionine, 0.5 mg of cold thymidine, or 0.5 mg of cold uracil respectively. The samples were harvested and resuspended in 200 μl of cold 20% trichloracetic acid (TCA) and were centrifuged at 20 000 g for 30 min at 4°C. The samples were washed twice with 200 μl of cold 96% ethanol. Precipitates were transferred to vials and the amount of incorporated radioactivity was counted in a liquid scintillation counter.

Northern and primer extension analyses

Cells were grown in LB at 37°C. At an OD450 of 0.5 the cultures were diluted 10 times and grown to an OD of 0.5 and transcription of the toxins was induced by addition of 0.2% arabinose. To inhibit translation, Cml (50 μg ml−1) was added and to induce aa starvation SHT was added (1 mg ml−1). For Northern analysis, total RNA was fractionated by PAGE (6% low-bis acrylamide), blotted to a Zeta-probe nylon membrane and hybridized with a single-stranded 32P-labelled riboprobe, complementary to lpp. The radioactive probe was generated using linearized plasmid DNA of pSC333. The primer lpp 21 was used for cleavage analysis of lpp RNA. For promoter mapping analyses, primer extension was performed with the primers higBI PE1, higAI PE1, higAI PE2, higBII PE1, higBII PE2, higAII PE1 and higAII PE2.


We thank Matt Waldor for sharing unpublished results and the members of the Gerdes group for stimulating discussions. This work was supported by The Danish National Research Foundation via Centre for MRNP Biogenesis and Metabolism.