Molecular Microbiology (1999) 32(5), 1090±1102
Multiple hok genes on the chromosome of Escherichia coli Kim Pedersen and Kenn Gerdes* Department of Molecular Biology, Odense University, Campusvej 55, DK-5230 Odense M, Denmark.
hok-like genes may be induced by an as yet unknown signal. Introduction
Summary The hok/sok locus of plasmid R1 mediates plasmid stabilization by the killing of plasmid-free cells. Many bacterial plasmids carry similar loci. For example, the F plasmid carries two hok homologues, ¯m and srnB, that mediate plasmid stabilization by this specialized type of programmed cell death. Here, we show that the chromosome of E. coli K-12 codes for ®ve hok homologous loci, all of which specify Hoklike toxins. Three of the loci appear to be inactivated by the insertion elements IS150 or IS186 located close to but not in the toxin-encoding reading frames (i.e. hokA, hokC and hokE ), one system is probably inactivated by point mutation (hokB ), whereas the ®fth system is inactivated by a major genetic rearrangement (hokD ). In the ECOR collection of wild-type E. coli strains, we identi®ed hokA and hokC loci without IS elements. A molecular and a genetic analysis show that the hokA and hokC loci specify unstable antisense RNAs and stable toxin-encoding mRNAs that are processed at their 38 ends. An alignment of the mRNA sequences reveals all the regulatory elements known to be required for correct folding and refolding of the plasmid-encoded mRNAs. The conserved elements include fbi that ensure a long-range interaction in the full-length mRNAs, and tac and antisense RNA target stem±loops that are required for translation and rapid antisense RNA binding of the processed mRNAs. Consistently, we ®nd that the chromosome-encoded mRNAs are processed at their 38 ends, resulting in the presumed translationally active mRNAs. Despite the presence of all of the regulatory elements, the chromosome-encoded loci do not mediate plasmid stabilization by killing of plasmid-free cells. The chromosome-encoded mRNAs are poorly translated in vitro, thus yielding an explanation for the lacking phenotype. These observations suggest that the chromosomal Received 21 January, 1999; revised 15 March, 1999; accepted 19 March 1999. *For correspondence. E-mail
[email protected]; Tel. (45) 6557 2413; Fax (45) 6593 2781. Q 1999 Blackwell Science Ltd
In recent years, a large number of genes that mediate programmed cell death in bacteria have been identi®ed and analysed (Jensen and Gerdes, 1995; Naito et al., 1995; Yarmolinsky, 1995; Gerdes et al., 1997; Holcik and Iyer, 1997; Gotfredsen and Gerdes, 1998; Grùnlund and Gerdes, 1999). The function of these genes has primarily been ascribed to their ability to mediate plasmid maintenance by killing of plasmid-free cells. However, bacterial chromosomes encode numerous genes that are homologous to the plasmid-encoded killer genes. Two types of loci that mediate plasmid stabilization by post-segregational killing (PSK) have been described. One type, the toxin±antitoxin gene systems, encodes a stable toxin and an unstable protein antitoxin (Jensen and Gerdes, 1995; Holcik and Iyer, 1997). The antitoxins form tight complexes with the toxins and thereby neutralize their toxicity. However, because the antitoxins are degraded by cellular proteases (Lon or Clp), the plasmidfree cells will experience a decay of the antitoxins. This, in turn, leads to activation of the toxins and cell killing. A large number of such loci have been identi®ed on plasmids, but it becomes increasingly evident that they are abundant on bacterial chromosomes as well (Gotfredsen and Gerdes, 1998). Recently, we identi®ed homologues of the E. coli relBE toxin±antitoxin locus in Gram-positive and in Gramnegative eubacteria, and in Archeae (Grùnlund and Gerdes, 1999). More surprisingly perhaps, plasmid-encoded restriction modi®cation cassettes also mediate plasmid stabilization by PSK (Kulakauskas et al., 1995; Naito et al., 1995). The other type of PSK genes are regulated by antisense RNA. Here, the toxins are encoded by stable mRNAs, whose translation is inhibited by unstable antisense RNAs. The paradigm member of this gene family is hok/sok of plasmid R1, whose genetic organization is shown in Fig. 1. The locus encodes a very stable mRNA, which speci®es the toxic Hok (host killing) protein that can kill the cells by damaging the cell membrane (Gerdes et al., 1986a; b). Translation of hok is regulated by Sok-RNA (suppression of killing), an unstable antisense RNA of 63 nucleotides (nts) that is complementary to the hok mRNA leader (Gerdes et al., 1990b; Nielsen et al., 1991; Thisted
E. coli hok homologues 1091 q
lacI and the LacI-regulated PA1/O4/O3 promoter (Lanzer and Bujard, 1988) upstream of a multiple cloning site (mcs). The construction of pKP219 is described in Experimental procedures. Without IPTG, transcripts from the
Fig. 1. Structural organization, RNAs and regulatory elements of the hok/sok system from plasmid R1. Genetic nomenclature: mok, mediation of killing; hok, host killing; sok, suppression of killing; sokT, Sok antisense RNA target; fbi, foldback inhibition element; tac, translational activation element, FL full length; TR, truncated. Numbers refer to the coordinates in the mRNAs.
et al., 1994a). Sok-RNA inhibits translation of the mok reading frame that overlaps with the hok gene (see Fig. 1). As translation of hok is coupled to translation of mok, Sok-RNA inhibits hok translation indirectly (Thisted and Gerdes, 1992). The complex molecular model that explains activation of hok translation in plasmid-free cells is shown in Fig. 2 and is described in detail in the legend. In this article, we describe the analyses of new hokhomologous loci from the E. coli chromosome. We show that E. coli K-12 contains ®ve hok-like loci, four of which encode all of the regulatory elements as described previously for hok/sok of R1. In E. coli K-12, three of these loci are inactivated by IS elements. Screening of the ECOR collection of E. coli wild-type strains reveals hok-homologous loci that are not inactivated by IS-elements. Molecular and genetic analyses point to the conclusion that the hokhomologous genes may be induced by an as yet unknown signal that affects mRNA translation. Results The chromosome of E. coli K-12 encodes ®ve hok-homologous genes The gef and relF genes of E. coli K-12 encode Hok-like proteins that are toxic and that confer similar gross morphological changes upon host cells as Hok of R1 (Gerdes et al., 1986b). Recently, genes gef and relF were denoted hokC and hokD respectively (Gerdes et al., 1997). Using the database search program BLAST (Altschul et al., 1997), we identi®ed three additional hok-homologous genes on the chromosome of E. coli K-12. Thus, E. coli K-12 encodes ®ve hok homologues denoted hokA K12 to hokE K12 . The ®ve chromosomally encoded Hok proteins are aligned with the other known homologues in Fig. 3. Genes hokAK12 , hokBK12 , and hokCK12 encode active Hok-like polypeptides The unit copy number cloning vector pKP219 contains Q 1999 Blackwell Science Ltd, Molecular Microbiology, 32, 1090±1102
Fig. 2. Molecular model that explains activation of hok translation in plasmid-free cells. A. Folding pathway of hok mRNA. During transcription, a metastable hairpin at the mRNA 58 end prevents formation of the tac stem (translational activation). Formation of the tac stem in the nascent transcript would be expected to lead to premature activation of translation or antisense RNA binding. In the full-length transcript, the fbi-element (foldback inhibition) pairs with tac thereby locking the mRNA into an inert con®guration: the SDmok element is base paired to ucb (upstream complementary box) and the Sok-RNA target (sokT ) is shielded by the foldback structure. During steady state, a pool of inert full-length mRNA accumulates in plasmid-containing cells. The full-length mRNA is activated by slow 38 processing, which removes the fbi element. The removal of fbi triggers a refolding of the 58 end of the mRNA with the formation of the tac and antisense target hairpins. B. In plasmid-carrying cells, the truncated mRNA is rapidly bound by Sok-RNA, which prevents translation. Subsequently, the RNAs form a duplex that is cleaved by RNase III. C. In plasmid-free cells, in which the unstable antisense RNA has decayed, translation of the truncated hok mRNA is allowed, thus leading to cell killing. The RNA structures shown were veri®ed by nucleotide covariations in phylogenetically related RNAs (Gultyaev et al., 1997) and by mutational and structural analyses (Thisted et al., 1995; Franch and Gerdes, 1996; Franch et al., 1997). The model was recently described in a review (Gerdes et al., 1997).
1092 K. Pedersen and K. Gerdes Fig. 3. Alignment of the known Hok-homologous toxins. A fully conserved amino acid in the consensus sequence is underlined and highly conserved amino acids are shown in bold. See Fig. 5 for the origin and the sequence of the corresponding genes.
promoter are not detectable by Northern analysis (data not shown). However, with IPTG, strong transcription is induced towards the cloning site. We PCR ampli®ed the hokA K12 , hokB K12 and hokC K12 genes and cloned the resulting DNA fragments into pKP219, which resulted in plasmids pKP612 (138 to 368), pKP611 (147 to 451) and pKP613 (136 to 370) respectively (1 refers to the ®rst nucleotide at the mRNA 58 ends; see Figs 4 and 5). The plasmids were transformed into strain CSH50 and subsequently tested in an induction experiment. Addition of IPTG to cells containing pKP612, pKP611 or pKP613 yielded arrest of cell growth, rapid host cell killing and the typical Hok-induced changes in cell morphology. These results show that genes hokA K12 , hokB K12 and hokC K12 encode polypeptides that, resembling Hok of R1, are very toxic to E. coli host cells.
Similarly, the hokC K12 locus (formerly gef ) located at 0.4 min between nhaA and dnaJ contains an IS186 element 22 bp downstream of the toxin-encoding reading frame (Fig. 4). Subtraction of the DNA sequence of the IS186 element revealed a hokC system with all of the
Insertion elements in the hokA, hokC and hokE loci of E. coli K-12 The hokA K12 gene is located at 80.1 min between glyS and cspA, see Fig. 4 (Blattner et al., 1997). We noticed that an IS150 element had inserted 32 bp upstream of the start codon of hokA K12 . Subtraction of the DNA sequence of the IS element revealed a hok-homologous system with a potential promoter for a hokA mRNA, fbi and tac elements in the mRNA, and a processing stem±loop just downstream of the hokA reading frame (see Fig. 5). A potential antisense RNA gene is also present in the K-12 sequence. However, the base pairs between IS150 and hokA K12 , which would be expected to encode the 10 sequence of the antisense RNA promoter and the start of a mok-homologous reading frame (mokA), is missing. The missing bases are consistent with the proposal that the insertion element introduced a deletion of 39 bp in a complete hok-homologous system. Thus, subtraction of the IS150 element from the E. coli K-12 sequence indicates that the element may have transposed into a system that had all of the regulatory elements as described for the hok/sok system of R1. This interpretation is corroborated by ®ndings described later.
Fig. 4. Structural organization of the hokA , hokB, hokC, hokD, and hokE loci of E. coli K-12. An IS150 element is located 32 bp upstream of hokA . A deletion of 39 bp is marked with a K. In hokC and hokE, an IS186 element is located 22 and 21 bp downstream of the toxin-encoding reading frames respectively. The start codon of the mokE homologue is missing (shown with a dotted line). The hokD system is transcribed in the operon with the relB and relE genes (a D marks the presumed deletion of the upstream part of the hokD K12 locus). Q 1999 Blackwell Science Ltd, Molecular Microbiology, 32, 1090±1102
Fig. 5. Alignment of the mRNAs from the hok family. hok is from plasmid R1; ¯m and srnB are from plasmid F; pnd16 and pnd483 are from plasmid R16 and R483; hokA is from E. coli C, hokB is from K-12, hokC is from ECOR24, hokD and hokE are from K-12 and hokX is from E. coli B; hokH is from Hafnia alvei. The alignment emphasizes the main regulatory elements and visualizes the stem±loop structures present at various times in the mRNA life cycle (i.e. the nascent, the full-length and the truncated transcripts). The main stem±loop structures are shown with arrows and are numbered in the following order: I, metastable; II, extended ucb/SDmok ; III, dcb/SDhok ; IV, processing; V, ucb/SDmok ; VI, top of the tac stem in full-length mRNA; VII, antisense target; VIII, tac stem in truncated mRNA; IX, shortened ucb/SDmok ; X, shortened dcb/SDhok . The conserved sequence elements are underlined and sequences involved in longrange base pairing are shown with asterisks (*). The sequences fbi and anti sokT 8 are paired with tac and sokT 8 in the full-length mRNA (structures XI and XII respectively), whereas antihok is paired with the TIR of hok (reading frame in italic) in the truncated mRNA (structure XIII). The sequences in bold indicate the promoter elements of the antisense RNAs in the complementary DNA strand. The alignment is based on the previously described conserved secondary structural elements and coupled covariations (Gerdes et al., 1997; Gultyaev et al., 1997) and was assisted by the computer programs MFOLD and PILEUP to include the new sequences (Wisconsin GCG package 9.1).
E. coli hok homologues 1093
Q 1999 Blackwell Science Ltd, Molecular Microbiology, 32, 1090±1102
1094 K. Pedersen and K. Gerdes regulatory elements as previously described for the hok system of plasmid R1, including fbi, tac, promoter elements for hokC mRNA and SokC antisense RNA and a mok-homologous reading frame that we denote mokC (formerly orf69 ) (see Fig. 5). This suggests that the hokA and hokC loci of E. coli K-12 were inactivated by IS-elements. The hokE K12 gene is located at 13.1 min between genes nfnB and entD. Interestingly, an IS186 element is located 21 bp downstream of the hokE K12 reading frame (see Fig. 4). Subtraction of the DNA sequence of the IS186 element again revealed a hok-homologous system that contains all of the regulatory elements as previously described for hok of R1, except for the start codon of a mok-homologous reading frame (see Fig. 5). Thus, the chromosome of E. coli K-12 contains three hok-homologous loci that probably have been inactivated by insertion elements. The hokB K12 locus is located between trg and cysB at 32.2 min (Fig. 4). This system contains all the regulatory elements as described for hok, such as fbi, tac, appropriate promoter sequences and a second overlapping reading frame that we denote mokB. The ®fth homologue, hokD K12 (formerly relF ; Gerdes et al., 1986b), is encoded by the third gene of the relB operon at 35.4 min (Fig. 4). Interestingly, the ®rst two genes of the operon encode a toxin±antitoxin system (Gotfredsen and Gerdes, 1998). A potential fbi sequence is located at the appropriate position downstream of the hokD K12 reading frame, whereas the upstream regulatory elements are missing. Thus, the hokD K12 gene may be a relic of a previously intact hok-homologous gene system. Two additional chromosomal hok loci have been identi®ed. In E. coli B, a hok locus that we denote hokX is located adjacent to cysH (Ostrowski et al., 1989). The HokX homologue is aligned with the other Hok-like toxins in Fig. 3. Curiously, an IS186 element is also located 21 bp downstream of the hokX reading frame. The upstream, regulatory part of hokX is also present in K-12, including a putative antisense RNA gene, but the toxin-encoding reading frame is missing. Furthermore, in the enterobacterium Hafnia alvei, we identi®ed a hok-homologous gene system just downstream of the ldc gene (Fecker et al., 1986). The system, denoted hokH, is missing a mok-homologous reading frame (Fig. 5). Identi®cation and cloning of hokA and hokC loci from other E. coli strains Using PCR, we screened strain collections for potentially intact hokA and hokC loci. By `intact', we mean hok-homologous genes without closely linked insertion elements. We found that E. coli C and 38 out of the 72 wild-type strains of the ECOR collection (Ochman and Selander, 1984) encode a hokA system without an IS150 element
(K. Pederson and K. Gerdes, unpublished). All strains contained the hokA gene at the same chromosomal position. Using PCR, we cloned the hokA C system of E. coli C into the R1 cloning vector pOU82, thus resulting in pKP110 ( 294 to 466). The mcs region of the pOU82 vector is only weakly transcribed. However, to avoid effects of fortuitous expression of HokAC , we used the Hok-resistant E. coli K-12 strain NWL37 as the recipient in the cloning procedure. The DNA sequence of the hokA system of E. coli C revealed all the known regulatory elements, as described previously for hok/sok of R1 (see Fig. 5). A schematic drawing of the hokC locus is shown in Fig. 4. Comparisons between the hokA loci of E. coli K-12 and C indicate that the IS150 element present in the K-12 sequence probably caused a deletion of 39 bp, which removed the start of mokA K12 and the 10 sequence of the antisense RNA promoter. The comparison also revealed a number of single base pair substitutions. In conclusion, E. coli C encodes a potentially active hok-homologous gene system (see Discussion ). A second screening of the entire ECOR collection revealed that 28 out of the 72 wild-type E. coli strains encode a hokC system without an IS186 element. We cloned the intact hokC system from ECOR24 into pOU82, resulting in pKP208 ( 640 to 400). As in the case of hokA C , the DNA sequence of hokC ECOR24 revealed an intact hok-like system with all the regulatory elements, as described previously for the hok system of R1 (Fig. 5). Thus, the wild-type ECOR24 strain and a number of other wild-type E. coli strains contain hokC systems without a linked IS186 element (data not shown). Comparison of the hokC sequences of E. coli K-12 and ECOR24 revealed only a few differences at the level of single base pairs. The new hokA C and hokC ECOR24 genes were cloned into the expression vector pKP219, resulting in the plasmids pKP602 and pKP608 respectively. Inductions experiments showed that the hokA C and hokC ECOR24 genes also encode active toxins. For comparison, the hokB locus of E. coli K-12 was also PCR ampli®ed and cloned into pOU82, thus resulting in pKP302 ( 1090 to 451). Smaller DNA fragments encoding the entire hokA C ( 114 to 466), hokB K12 ( 74 to 451) and hokC ECOR24 ( 225 to 400) genes were also generated by PCR and cloned into pOU82, thus resulting in pKP101, pKP301 and pKP201 respectively. Again, the Hok-resistant strain NWL37 was used as recipient in the cloning procedure. For reasons unknown, we were not able to transfer plasmids pKP101 (hokA C ) and pKP201 (hokC ECOR24 ) to Hoksensitive K-12 strains such as CSH50 or MC1000. In the case of the hokB K12 carrying plasmids pKP301, we encountered no such problem, and none of the plasmids carrying longer hokA C- and hokC ECOR24-encoding fragments yielded cloning problems. Similar toxicity was expressed from Q 1999 Blackwell Science Ltd, Molecular Microbiology, 32, 1090±1102
E. coli hok homologues 1095 short hokA- and hokC- encoding fragments derived from other ECOR strains without IS150 or IS186 respectively (data not shown). Northern analyses of RNAs encoded by short and long hokA C and hokC ECOR24 DNA fragments produced similar amounts of mRNA and antisense RNAs (not shown). Thus, we believe that the toxicity expressed by the short versions of the DNA fragments may re¯ect cloning artefacts caused by the removal of the hokA and hokC genes from their natural context. Curiously enough, the toxicity expressed by the short DNA fragments encoding hokA C and hokC ECOR24 was relieved by cloning the fragments into high copy number plasmids (not shown). We believe that this relief is caused by the concomitant increased concentration of the antisense RNAs, which repress translation of the cognate toxin-encoding mRNAs. The hokAC , hokBK12 , and hokCECOR24 loci encode stable mRNAs that are processed at their 3 8 ends Total RNA was prepared from NWL37 carrying plasmids pKP402 (hokA C ), pKP408 (hokB K12 ) or pKP406 (hokC ECOR24 ) before and after the addition of rifampicin to growing cells. The RNA samples were analysed by Northern blotting. Figure 6A shows that the hokA C locus encodes a stable mRNA of < 362 nucleotides (nts). This size of full-length hokA C RNAs is in accordance with the tac and fbi sequences in Fig. 5. The long half-life of the full-length mRNA is consistent with the proposal that the tac and the fbi elements mediate a 58 to 38 long-range interaction, as described for the hok mRNA of R1 (see later). After the addition of rifampicin, a processed hokA C mRNA species of < 330 nt slowly appeared. The difference in sizes of the two mRNAs is consistent with a 38 exonucleolytical removal of the 32 nts in the 38 end of the full-length hokA C mRNA that encodes the fbi element. Thus, by inference, the processing pattern indicates mRNA 38 processing as previously described for the plasmidencoded hok-homologous mRNAs (Gerdes et al., 1990b; Nielsen et al., 1991; Thisted et al., 1994b; Franch and Gerdes, 1996). The 58 end of the hokA C mRNA was determined by primer extension analysis and is consistent with this inference (data not shown). The 58 end is located downstream of the 10 and 35 promoter sequences and the tac element is located in the very 58 end of the mRNA, as in the case of hok mRNA of R1. A putative RNase III cleavage product of the mRNA±antisense RNA duplex is also indicated in Fig. 6A. The hokB K12 and hokC ECOR24 mRNAs exhibited similar processing patterns as those described above for hokA C mRNA (Fig. 6B and C). Both mRNAs were stable and were slowly processed to slightly shortened versions, the sizes of which are consistent with 38 exonucleolytical removal of the fbi encoding sequences. However, the 38 Q 1999 Blackwell Science Ltd, Molecular Microbiology, 32, 1090±1102
Fig. 6. Stability and processing patterns of the hok-like mRNAs shown by Northern analyses. Parts A, B and C show the hokA C , hokB K12 and hokC ECOR24 mRNAs respectively. Cells were grown in LB medium to an OD 450 of
