The par stability determinant of the Enterococcus faecalis plasmid pAD1 is the first antisense RNA-regulated post-segregational killing system (PSK) identified in a Gram-positive organism. Par encodes two small, convergently transcribed RNAs, designated RNA I and RNA II, which are the toxin and antidote of the par PSK system respectively. RNA I encodes an open reading frame of 33 codons designated fst. The results presented here demonstrate that the peptide encoded by fst is the par toxin. The fst sequence was shown to be sufficient for cell killing, and removal of the final codon inactivated the toxin. In vitro translation reactions of purified RNA I transcript produced a product of the expected size for the fst-encoded peptide. This product was not produced when purified RNA II transcript was added to the translation reaction. Toeprint analysis demonstrated that purified RNA II was able to inhibit ribosome binding to RNA I. These data suggest that fst expression is regulated by RNA II via an antisense RNA mechanism. In vitro translation studies and toeprint analyses also indicated that fst expression is internally regulated by a stem–loop structure at the 5′ end of RNA I. Removal of this structure resulted in better ribosome binding to RNA I and a 300-fold increase in production of the fst-encoded peptide. Finally, RNA II was shown to be less stable than RNA I in vivo, providing a basis for the selective expression of fst in plasmid-free cells.
Bacterial plasmids are inherited independently of the host cell chromosome and, therefore, must encode mechanisms to prevent plasmid loss during cell division. Post-segregational killing (PSK) systems were identified as plasmid stability mechanisms that ensure the stable inheritance of plasmids by selectively killing any plasmid-free segregants that arise during cell division (Jensen and Gerdes, 1995; Gerdes et al., 1997). These systems produce two components: a stable toxin and an unstable antidote that inhibits either the expression or the activity of the toxin. Because the antidote of PSK systems is unstable, plasmid-free cells are essentially programmed for cell death; no longer having a means of replenishing their supply of antidote, the toxin eventually kills the cell. Homologues of PSK systems have also been identified on the Escherichia coli chromosome, where they may function in stress response to amino acid starvation (Gerdes et al., 1986; Aizenman et al., 1996; Pedersen and Gerdes, 1999; Gerdes, 2000).
Two different types of PSK systems have been identified in bacteria: proteic systems and antisense RNA-regulated systems. In proteic systems, both toxin and antidote are proteins, and the antidotes neutralize the toxins by forming tight complexes with them. The differential decay rate between toxins and antidotes for these systems is dependent upon cellular proteases (Lon or Clp). Details of several proteic PSK systems have been reviewed recently (Jensen and Gerdes, 1995; Gerdes, 2000)
In antisense RNA-regulated systems, the role of the antidote is played by a small, unstable transcript that binds to the toxin-encoding RNA and inhibits its translation (Gerdes et al., 1997). The best-characterized system of this type is the hok/sok locus of plasmid R1 (Gerdes et al., 1986). This locus encodes a stable mRNA encoding the Hok (host killing) protein. Translation of hok is regulated by the sok (suppressor of killing) antisense RNA (Gerdes et al., 1990; Nielsen et al., 1991). Hok expression is also regulated by secondary structures within the hok transcript. Hok mRNA contains a fold-back inhibitory (fbi) sequence at its 3′ end, which forms a complex secondary structure preventing the translation of Hok and also interfering with the binding of the Sok–antisense RNA (Thisted et al., 1995; Franch and Gerdes, 1996). As the hok–sok duplex is rapidly cleaved by RNase III (Gerdes et al., 1992), this is necessary to ensure that a pool of hok message is available to be activated in plasmid-free cells. The hok message is slowly degraded from the 3′ end by host RNases, eventually resulting in removal of the fbi. Once the fbi region has been processed, hok can either be bound by the sok antisense RNA or, in plasmid-free cells, can be translated to produce toxin. A metastable hairpin has also been shown to form at the 5′ end of the nascent hok transcript. This structure sequesters the mok translation initiation region until transcription of hok is completed and the fbi region can suppress translation (Nagel et al., 1999).
Par is a 400 bp determinant required for the stable inheritance of the Enterococcus faecalis plasmid pAD1 (Weaver et al., 1993). The par determinant has been shown to stabilize plasmids in the host population by a PSK mechanism (Weaver, 1995; Weaver et al., 1996; 1998) and is the first plasmid stability system of this kind identified in a Gram-positive organism, with the exception of restriction-modification systems (Naito et al., 1995). The par locus encodes two small RNA transcripts, designated RNA I and RNA II, which are essential for its function (Weaver and Tritle, 1994). RNA I and RNA II are convergently transcribed from promoters located at each end of par, and terminate at a common bidirectional terminator (Fig. 1). The RNAs are transcribed in opposite directions across a pair of direct repeat sequences located at the 5′ end of each gene. This 5′ region and the shared terminator region provide regions of complementarity at which the two RNAs could interact. Computer-predicted RNA secondary structure analyses indicate that these regions are present on exposed loops in both RNAs making them accessible for interaction (Fig. 2). Additionally, RNA I codes for the peptide Fst (faecalis plasmid-stabilizing toxin). The translation initiation signals for fst are adjacent to the 5′ complementary repeats, suggesting that its translation would be inhibited by interaction with RNA II. Thus, the organization of the par locus suggests that RNA II regulates the translation of the RNA I-encoded fst via an antisense RNA mechanism.
Several lines of evidence support the hypothesis that par is a PSK system with RNA I encoding the toxin and RNA II the antidote (Weaver, 1995; Weaver et al., 1996; 1998). Until now, we have been uncertain whether the Fst peptide encoded by RNA I is translated and is responsible for par-induced killing. In this report, we provide evidence showing that this small peptide is indeed the par toxin. In addition, we show that RNA II is capable of preventing the translation of fst by inhibiting ribosome binding. Finally, we demonstrate that RNA II is less stable than RNA I, as would be expected for the antidote of a PSK system.
The par toxin is encoded by fst
The RNA I gene has been shown to be the toxic component of the par system but, as discussed above, whether the RNA I-encoded peptide (Fst) or some other attribute of RNA I is responsible for this toxicity has not been determined. To resolve this question, we sought to determine whether fst was sufficient for cell killing. As shown in Table 1, pDAK704, a pAM401 construct containing the complete RNA I gene, could not be introduced in E. faecalis UV202 cells unless RNA II was already present on another plasmid (pDAK611). To determine whether the Fst peptide was sufficient for cell killing, the 3′ non-coding sequence of RNA I was removed, so that only the promoter, Shine–Dalgarno (SD) sequence and fst open reading frame remained. This shortened RNA I gene was cloned into pAM401 creating pDAK614. Transformation experiments with this clone gave results similar to those for the wild-type RNA I clone (Table 1), indicating that it was still lethal to E. faecalis and that RNA II was still protective. In contrast, a construct in which the final codon of fst was replaced by a stop codon (pDAK615) could be introduced into E. faecalis cells in the absence of RNA II. The transformation efficiency of pDAK615 was consistently, but not significantly, lower than that of the pAM401 vector control in the absence of RNA II (P > 0.05), but transformant colonies were much smaller than those of pAM401, suggesting that the truncated fst product might retain some toxicity, but is clearly much less lethal than the full-length peptide. Interestingly, transformation efficiencies into the RNA II-containing strain (pDAK611) with both pDAK704 and pDAK614 were significantly lower than with pAM401 or pDAK615 (P < 0.05).
Table 1. Transformation of an E. faecalis UV202 strain carrying pDL278 or pDAK611.
Transformation efficiency (cfus µg−1 ± SD)
pDAK611 is a pDL278 clone containing the RNA II gene. Electroporatable E. faecalis cells were transformed with 250 ng of plasmid DNA. pDAK704 contains wild-type RNA I; pDAK614 contains just fst; and pDAK615 contains a one-amino-acid truncation of fst.
. The difference in transformation efficiencies between these plasmids and pAM401 was determined to be statistically significant (P < 0.05).
As discussed above, the organization of par and the sequences of RNA I and RNA II suggest that RNA II functions as an antisense RNA to suppress fst translation. To determine if this was true, an in vitro translation system was devised using purified RNA I as template (see Experimental procedures). An in vitro translation reaction of wild-type RNA I produced a faint protein band of approximately 3.5 kDa, close to the predicted molecular weight of 3.73 kDa for Fst (Fig. 3, lane 1). This was the only specific translation product for this reaction and was not present when heterologous RNAs were translated (data not shown). The addition of an equimolar concentration of purified RNA II to the translation reaction significantly reduced the amount of translation from RNA I (data not shown), and the addition of a fivefold molar excess of purified RNA II resulted in undetectable amounts of translation (Fig. 3, lane 2). An RNA II mutant known to be incapable of RNA I interaction (Greenfield and Weaver, 2000) was not able to interfere with fst translation (Fig. 3, lane 3), indicating that translational inhibition of fst results from specific interaction of RNA I with RNA II.
Computer secondary structure analysis of RNA I indicated that the fst SD sequence is secluded within a stem–loop at the 5′ end of the transcript (Fig. 2). This structure may function to repress translation of the toxic peptide during transcription and/or in the completed transcript to allow a pool of inactive message to accumulate, as has been shown for hok (Thisted et al., 1995; Franch et al., 1996). To determine whether this structure inhibited translation, a DNA template was constructed that lacked the upstream portion of the 5′ stem–loop, so that it could not be formed in the RNA I transcript. Translation of purified RNA transcribed from this template resulted in a nearly 300-fold increase in peptide synthesis over equimolar amounts of wild-type RNA I (compare Fig. 3, lanes 1 and 4). RNA II was capable of inhibiting translation of the 5′-deleted RNA I derivative at the same concentration used to inhibit wild-type RNA I, in spite of the higher translation levels. The RNA II mutant was unable to inhibit translation (Fig. 3, lane 6).
RNA II inhibits fst translation by preventing ribosome binding
The positioning of complementary regions around the putative translation initiation region of fst suggested that RNA II inhibits translation by interfering with ribosome binding. To test this hypothesis, ribosomal toeprint assays (Hartz et al., 1988) were performed on RNA I in the presence and absence of RNA II. This technique also allowed a comparison of the accessibility of the SDfst in the full-length RNA I with that of the 5′-deleted RNA I. As the 5′ stem–loop sequesters the SD sequence, the 5′-deleted RNA I derivative should bind ribosomes more efficiently than full-length RNA I. Purified wild-type and 5′-deleted RNA I transcripts were incubated with or without 30S ribosomal subunits. An oligonucleotide complementary to the RNA sequence 50 nucleotides downstream of the fst start codon was used to prime the reverse transcriptase reaction back towards the ribosome binding site. A reverse transcriptase stop (toeprint) at +15 nucleotides relative to the start codon is indicative of ribosome binding. As shown in Fig. 4, a band corresponding to a toeprint is present at 15–16 nucleotides downstream of the predicted start codon of fst (Fig. 4, lane 2). The toeprint is significantly increased when the 5′-deleted derivative of RNA I is used (Fig. 4, lanes 6–10), consistent with the increased translation observed in vitro. When purified RNA II was added to the reaction mixture in a 1:2 ratio of RNA II to RNA I, toeprint intensity decreased ≈ fourfold for both the full-length and the 5′-deleted RNA I (Fig. 4, lanes 3 and 8). The ribosomal toeprint was nearly undetectable when RNA II was added to the toeprint assay at or above a 1:1 ratio. (Fig. 4, lanes 4–5 and 9–10). These results demonstrate that (i) the GUG start codon of fst is a legitimate translation start site; (ii) the RNA I 5′ stem–loop represses ribosome binding; and (iii) RNA II inhibits ribosome binding to RNA I.
Differential stabilities of the par RNAs
If RNA II functions as an antidote via an antisense RNA mechanism, then translation of fst should be regulated by the differential stabilities of RNA I and RNA II. By analogy with the hok/sok system, RNA II should be less stable than RNA I; its selective degradation in plasmid-free segregants would lead to the expression of fst and death of the host cell. To determine whether RNA II is less stable than RNA I, RNA was purified from E. faecalis OGIX cells containing pDAK2300K (a pAD1 miniplasmid containing par;Weaver and Clewell, 1991) at varying time intervals after exposure to rifampicin to inhibit transcription initiation. The results shown in Fig. 5 demonstrate that RNA II levels declined continually after the addition of rifampicin, whereas RNA I levels remained essentially constant. Densitometric analysis of multiple experiments determined the half-life of RNA II to be approximately 10 min, whereas the half-life of RNA I was found to be > 1 h. Thus, the RNA II antidote is indeed less stable than its target, RNA I.
The par determinant has been shown to stabilize the E. faecalis plasmid pAD1 by a PSK mechanism. Par encodes two genes for this function, RNA I and RNA II. RNA I encodes a small open reading frame (fst) for a 33-amino-acid peptide. In this communication, evidence is presented establishing the fst peptide as the toxin component of the killing system. The fst sequence is sufficient for cell killing, and deletion of the last amino acid inactivates the toxin. When purified RNA I transcript was added to an in vitro translation reaction, a product of the appropriate size for the fst-encoded peptide was produced. Ribosomal toeprint analysis also demonstrated that the predicted SD and start codon sequences do indeed bind the ribosomal 30S subunit. These results demonstrated that the peptide encoded on RNA I, and not some other aspect of RNA I structure, is responsible for post-segregational killing. Currently, the target of the par toxin is not known. Homology searches with the peptide sequence reveal no significant similarity to any of the known PSK system toxins or any other proteins.
The results presented here also demonstrate that fst expression is regulated post-transcriptionally by RNA II, as the PSK hypothesis for par function predicts. Previously, RNA II had been shown to protect against induced expression of RNA I (Weaver et al., 1996) and, here, we demonstrate that fst is not translated when purified RNA II transcript is added to in vitro translation reactions of purified RNA I. Toeprinting analysis of RNA I indicated that RNA II blocks translation by interfering with ribosome binding to the translation initiation region (TIR) of fst. These data, together with the organization and sequence of RNA I and RNA II within par, indicate that RNA II regulates fst expression by an antisense RNA mechanism. The details of antisense RNA interaction are described in the accompanying paper (Greenfield and Weaver, 2000).
Fst translation also appears to be regulated by secondary structure at the 5′ end of RNA I. Computer modelling of RNA I secondary structure predicts that the SDfst is sequestered within a 5′ stem–loop. In vitro translation studies as well as toeprint analyses indicated that this stem–loop restricted fst translation. A 5′-deleted RNA I derivative, which was incapable of forming this structure, was translated 300 times more efficiently than wild-type RNA I. This mutant was also better able to bind ribosome than wild-type RNA I. This type of regulation may be similar to that observed in the hok transcript, in which structures are formed to repress both translation and antisense RNA interaction (see above; Franch and Gerdes, 1996; Franch et al., 1997; Nagel et al., 1999). Interestingly, the par RNA I 5′ stem–loop is adjacent to the 5′ complementary sequences known to be involved in RNA II interaction (Greenfield and Weaver, 2000). Therefore, this structure is predicted to interfere with both fst translation and rapid RNA II binding.
In order for fst to function effectively as a PSK toxin, its translation must be regulated by the differential stabilities of the toxin and antidote. This means that, in antisense RNA-regulated systems, the toxin-encoding mRNA must be more stable than the antisense RNA. In this report, we also demonstrated that RNA II was indeed less stable than RNA I. This was shown by adding rifampicin (rif) to cells harbouring a pAD1 miniplasmid encoding par. After the addition of rif to stop transcription, RNA I levels within the cell remained essentially constant, whereas RNA II levels declined steadily. Thus, all the results presented here support the conclusion that the par toxin is the Fst peptide and that its expression is regulated at the translational level by RNA II.
Bacterial strains, media and culture conditions
The recombination-deficient E. faecalis strain UV202 (Yagi and Clewell, 1980) was used for the transformation of RNA I constructs, and the strain OGIX (Ike et al., 1983) was used for RNA stability analysis. E. coli strains JM101 (Sambrook et al., 1989) and DH5-alpha (Gibco BRL) were used during plasmid constructions and DNA manipulations. For routine cultivation, E. coli cells were grown in Luria broth (Sambrook et al., 1989), and E. faecalis cells were grown in Todd–Hewitt broth (Difco Laboratories). Antibiotics (Sigma Chemical) were used at the following concentrations: ampicillin, 100 µg ml−1; chloramphenicol, 10–25 µg ml−1; tetracycline, 10 µg ml−1; erythromycin, 10 µg ml−1; spectinomycin, 100 µg ml−1; rifampicin, 300 µg ml−1.
DNA purification, manipulation and transformation
Small-scale plasmid purifications from E. faecalis were performed using the modified alkaline lysis procedure described previously (Weaver and Clewell, 1988) and, from E. coli, using the method of Holmes and Quigly (1981). Large-scale plasmid purifications were from E. coli using the Qiagen midi-prep protocol according to the manufacturer's instructions.
Plasmid constructs were introduced into E. faecalis by electroporation using a cell porator electroporation system I and voltage booster (Gibco BRL) according to the manufacturer's instructions. E. faecalis cells were prepared as described by Cruz-Rodz and Gilmore (1990). Transformation of E. coli was performed using subcloning efficiency DH5-alpha-competent cells (Gibco BRL) according to manufacturer's instructions.
Plasmids used and constructed in this study are listed in Table 2. The fst constructs, pDAK614 and pDAK615, were created by polymerase chain reaction (PCR). PCR amplification was accomplished with Taq DNA polymerase (Promega) in a Precision Scientific genetic thermal cycler using a MgCl2 concentration of 2.0 mM. Thirty cycles with a melting temperature of 91°C for 30 s, an annealing temperature of 50°C for 30 s and an extension temperature of 72°C for 60 s were used for amplification. The 5′ primer (BamHI-1310: GCGGATCCGGCTGTCTAGCAAGCAG) is complementary to sequences upstream of the RNA I promoter (Weaver, 1995). The 3′ primers used were: RNA I-33, TTACTTTCGGCTATCGTCTT; and RNA I-32, TTATCGGCTATCGTCTTCCT. PCR products were cloned into the pGEM-t easy vector system (Promega) according to the manufacturer's instructions and sequenced (LoneStar Labs) to ensure that the proper product was created. The products were then subcloned into the E. coli–E. faecalis shuttle vector pAM401 (Wirth et al., 1987).
DNA templates for in vitro transcription were generated by PCR using a 5′ T7 primer engineered to replace the native promoters with a T7 promoter. PCR amplification was performed as described above. The amplified RNA I and RNA II products were individually cloned into the pGEM-T vector system (Promega) and sequenced to ensure the presence of the proper product. Both the 5′ and the 3′ primers also contained restriction enzyme sites to digest the template before the transcription reaction. RNA I 5′ T7 primer: CGGGATCCTGTAATACGACTCACTATAAGGGAATGCGGCAGCTCG; RNA I 3′ primer: CGTCTAGATGAAAAGCAATCCCTACGGCGA; RNA II 5′ primer: CGGGATCCTGTAATACGACCTCAACTATAGGGAAAAAGGTGCGAAACG; RNA II 3′ primer: CGAAGCTTTGAAAAAAAAGCAATCCTATTCGCCG; restriction site are italicized and the T7 promoter underlined. To produce the 5′-deleted RNA I mutant, the following 5′ primer (CGGGATCCTGTAATACGACTTCACTATAAGGGGATTGGAGGTGTGT) was used, in which the promoter and SD sequence remained identical, but the intervening sequence was removed. The RNA II mutant used in translation inhibition experiments was created using the following 5′ primer: CGGGATCCTGTAATACGACCTCAACTATAGGGAAAAGTGCGAAAGCACTAATTATCGTACAGATAAC (altered nucleotides are in bold).
RNA preparation and Northern blotting
RNA was purified from cultures of E. faecalis using RNAzol B (tel-test) as described previously (Weaver and Tritle, 1994). RNA was electrophoresed through a 5% polyacrylamide−6 M urea gel (FMC Bioproducts) in 1 × TBE running buffer (89 mM Tris, pH 8.3, 2 mM EDTA, 89 mM boric acid) at 325 V. After electrophoresis, RNA was transferred to Nytran Plus nylon membrane (Schleicher and Schuell) as described by Weaver et al. (1996). Membranes were prehybridized for 2 h at 42°C in Rapid-hyb hybridization buffer (Amersham Life Science) and probed for 2 h at 42°C. The RNA I- and RNA II-specific probes used were as described previously (Weaver et al., 1996). Probes were end-labelled using T4 polynucleotide kinase (Promega) and [γ-32P]-ATP (ICN) according to the manufacturer's instructions. Blots were washed at room temperature in a 1 × SSC−0.1% SDS solution and exposed to Kodak Biomax X-ray film (Fisher Scientific) at −70°C.
Preparation of par RNAs in vitro
DNA templates were digested with the appropriate restriction enzymes and ethanol precipitated before transcription. Transcripts were synthesized using T7 RNA polymerase (New England BioLabs) according to the manufacturer's directions. Transcription reactions were run on 6% polyacrylamide−6 M urea gels for 45 min at 350 V. Transcripts were gel purified by elution with 0.5 ml of elution buffer (0.1 M NaOac, pH 5.7, 10 mM EDTA, 0.5% SDS) at room temperature for 4 h (Thisted et al., 1995). Gel debris was removed by centrifugation. The supernatant was extracted with phenol–chloroform, ethanol precipitated and resuspended in DEPC (diethyl pyrocarbonate; Sigma Chemical)-treated water.
In vitro translations
Translations were performed in vitro using the E. coli T7 30S extract system for linear DNA (Promega). RNA I (15 pmol) or 5′ deleted RNA I transcript was translated according to the manufacturer's protocol. For translation inhibition experiments, a fivefold molar excess of RNA II was added and allowed to bind for 30 min at 37°C in TMK buffer (20 mM Tris acetate, pH 7.5, 10 mM magnesium acetate, 200 mM potassium glutamate; Mikkelsen and Gerdes, 1997) before the addition of the S30 protein extract. Translation reactions were precipitated with four volumes of acetone at 0°C, dried and redissolved in 2 × SDS sample buffer (per 10 ml: 2.0 ml of glycerol, 2.0 ml of 10% SDS, 0.25 mg of bromophenol blue, 2.5 ml of Tris HCl−0.4% SDS and 0.5 ml of beta-mercaptoethanol). Samples were denatured at 100°C for 5 min before loading on SDS gels. Discontinuous tricine–SDS gels were used as described by Schagger and von Jagow (1987). The stacking gel contained 4% acrylamide, and the separation gel contained 16% acrylamide. Electrophoresis was performed at 30 V in the stacking gel and was increased to 95 V after the sample had reached the separation gel. Total run time was approximately 20 h. The gel was fixed in 50% methanol, 20% acetic acid and 10% glycerol for 1 h, dried for 4 h and autoradiographed at −70°C. The intensity of the bands was determined using a Packard Instant Imager and a phosphorimaging software package.
Ribosomal toeprinting analysis
The 30S ribosomal subunits were prepared as described by Franch et al. (1997). The primer used for toeprint analysis was RNAITP: CCAACACACGAGAATCCAATTTCC. 5′ end-labelled primer (0.4 pmol) was annealed to 0.4 pmol of RNA I transcript in a buffer containing 60 mM NH4Cl, 10 mM Tris-acetate, pH 7.5, 8.5 mM beta-mercaptoethanol, 10 mM MgCl2, 10 units of RNAquard RNase inhibitor (Pharmacia) and 100 µM dNTPs. Two microlitres of 0.2 µM 30S ribosomes was added and incubated for 10 min at 37°C, followed by the addition of 1 µl of 25 µM uncharged tRNAfmet (Sigma) for 5 min. Two units of AMV-RT were added, and the reaction was allowed to proceed for 20 min at 37°C. Reactions were stopped by ethanol precipitation and subsequently resuspended in 10 µl of FD buffer (92% formamide, 17 mM EDTA, 0.025% xylene cyanol and 0.025% bromophenol blue). For reactions containing antisense RNA, the indicated amount of RNA II transcript was added to the reaction buffer before the addition of the 30S ribosomal subunits and incubated for 15 min at 37°C. The RNA sequence reaction was made by adding 1 µl of 40 µM ddATP/ddCTP or 1 µl of 60 µM ddGTP/ddTTP to the extension mixtures. Products were run on 8% polyacrylamide−6 M urea sequencing gels, which were subsequently dried and exposed to X-ray film at −70°C.
Differential stabilities of par RNAs
Stationary phase cultures of OG1X (pDAK2300K) were diluted 1:10 in THB and grown for 1 h before the addition of rifampicin (rif). Samples were taken immediately before the addition of rif and at various time points thereafter. Total RNA was prepared, electrophoresed and blotted as described above. For this experiment, the RNA II-specific probe was labelled with a higher specific activity in order to detect small amounts of RNA II: > 6000 Ci mmol−1 for the RNA II-specific probe and > 3000 Ci mmol−1 for the RNA I-specific probe. Densitometry was performed using a Molecular Dynamics PDSI-PC laser densitometer and the imagequant 4.2 software package.