Genome sequence analyses of Escherichia coli K-12 revealed four copies of long repetitive elements. These sequences are designated as long direct repeat (LDR) sequences. Three of the repeats (LDR-A, -B, -C), each approximately 500 bp in length, are located as tandem repeats at 27.4 min on the genetic map. Another copy (LDR-D), 450 bp in length and nearly identical to LDR-A, -B and -C, is located at 79.7 min, a position that is directly opposite the position of LDR-A, -B and -C. In this study, we demonstrate that LDR-D encodes a 35-amino-acid peptide, LdrD, the overexpression of which causes rapid cell killing and nucleoid condensation of the host cell. Northern blot and primer extension analysis showed constitutive transcription of a stable mRNA (≈ 370 nucleotides) encoding LdrD and an unstable cis-encoded antisense RNA (≈ 60 nucleotides), which functions as a trans-acting regulator of ldrD translation. We propose that LDR encodes a toxin–antitoxin module. LDR-homologous sequences are not present on any known plasmids but are conserved in Salmonella and other enterobacterial species.
Recently, a computer analysis revealed additional repetitive sequences, PAIR, TRIP and QUAD (Rudd, 1999). Transcription from PAIR and QUAD was observed with DNA microarray and Northern blot analysis (Wassarman et al., 2001). The transcripts from QUAD have the potential to form stable stem–loop structures (Rudd, 1999). As for the previously identified repeat, the physiological functions of the PAIR, TRIP and QUAD sequences are still unknown. Nevertheless, functional analyses of all the repeat sequences are required to understand the organization and plasticity of the E. coli genome.
Four long direct repeat (LDR) sequences were also detected upon completion of the E. coli sequence. Three tandem copies of a 535 bp repeat are present at 27.4 min on the chromosome map (Oshima et al., 1996; Yoshida et al., 2000), known to be the largest intergenic region (1729 bp), and another single fourth copy is located at 79.7 min (Blattner et al., 1997). It is interesting that the LDR sequences are symmetrically positioned on the chromosome, but the reasons for this are unknown. In our recent studies, we obtained another four statistically significant LDR sequences with more than 187 bp matched to LDR-A near the LDR loci (Kawano et al., 2002). LDR sequences are conserved in Salmonella species, suggesting that they play an important physiological role in the cell.
To investigate whether LDRs may have a physiological function in the cell, we analysed the effects of deleting and overexpressing the LDR sequences and characterized the LDR-D sequence. Our results show that LDR-D codes for two small genes, denoted ldrD and rdlD. The ldrD gene product is highly toxic to the host cell upon overexpression, and the rdlD product is a trans-acting antisense RNA that inhibits the translation of the ldrD mRNA. On the basis of structural and functional similarities, we propose that LDRs are members of a toxin–antitoxin gene family. The conservation of the LDR family in enteric bacteria suggests that these repetitive sequences are important to the genomes that carry them.
Characteristics of LDRs in E. coli K-12
Three tandem copies of a 535 bp LDR sequence (LDR-A, -B and -C) are located on the E. coli genome between kdsA and chaA, at the 27.4 min region of the E. coli genetic map. Another 450 bp copy named LDR-D is located at a symmetrical position on the chromosome between yhjU and yhjV, at about 79.7 min (Fig. 1, GenoBase 4.0: http://ecoli.aist-nara.ac.jp/). All the copies have the same direction on the chromosome. LDR-A and -B show 97% identity to each other, and LDR-C is 84% identical to the first two copies. Although the LDR-C sequence was originally identified as a 535 bp repeat, identical to LDR-A and -B, we have redefined it as a 471 bp repeat sequence according to the alignment of these three copies. LDR-D shows 63% identity to the first three copies and is 75% identical to them when these four repeats are aligned in length of LDR-D (Fig. 2). The G+C content of LDR-D is 56%, and the ratio of purines to pyrimidines is 55%.
Deletions of LDR sequences have no effect on host cell growth
To investigate the physiological role of LDR, a deletion mutant series of LDR, ΔLDR-ABC, ΔLDR-D and ΔLDR-ABCD, was constructed (see Experimental procedures). The resultant ΔLDR-ABC mutation was transferred by P1 phage-mediated transduction into a wild-type strain. Transduction frequency of the ΔLDR-D mutation was not influenced by the host strain with or without LDR-ABC sequences (data not shown). These deletion strains showed no clear differences in host cell growth, morphology, nucleoid structures and mutation frequency (data not shown). From these analyses, it can be concluded that LDRs do not have any essential function under laboratory growth conditions.
The ldrD gene is encoded within LDR-D
For further physiological analysis of the LDRs, we analysed the effect on the cell of an increase in LDR copy number. For this purpose, we cloned polymerase chain reaction (PCR)-amplified LDR-D fragments into a SmaI site of plasmid pUC18 (Yanisch-Perron et al., 1985) in both orientations.
In each LDR sequence, GENEMARK software at http://opal.biology.gatech.edu/GeneMark/ (Borodovsky and McIninch, 1993) predicts a small open reading frame (ORF) encoding 35 amino acids (ORF35). p18-D+, a plasmid containing a cloned LDR-D fragment whose putative ORF has the same direction of transcription as the lac promoter of the plasmid, caused a significant growth defect in host cells when stimulated with 1 mM IPTG to induce the lac promoter. In contrast, p18-D–, which contains the fragment in the opposite direction, showed no growth inhibition (Fig. 3A). In order to investigate whether the growth inhibition of the LDR-D sequence is caused by ORF35 in LDR-D, we constructed and analysed deletion mutants of the region upstream of this coding region. As seen in Fig. 3A, plasmid p18-D+X, with a deleted upstream region, still showed the growth inhibition. This short DNA fragment of LDR-D had two potential ORFs in the same frame, ORF35 and ORF28. We therefore isolated suppressor mutants lacking the growth defect, and sequence analysis revealed an 11 bp deletion from the ninth amino acid (alanine) to the 12th (histidine) in ORF35 (Fig. 3A, p18-D+D). To confirm further whether this coding region was responsible for the growth inhibition of LDR-D, an amber mutation was introduced at the second codon of ORF35 by site-directed mutagenesis. As shown in Fig. 3A, plasmid p18-D+A, which carries the amber mutation, lost the growth inhibition activity; expression of plasmid p18-D+A in the supF strain, which suppresses the amber mutation, restored the growth inhibition activity (Fig. 3A). From these results, we concluded that the growth inhibition of the LDR-D sequence is caused by ORF35. This gene was designated as ldrD.
To study the expression of ldrD, the region upstream of ldrD was cloned in front of the lacZ gene of plasmid pMC1403 (Casadaban et al., 1980) with or without the N-terminal coding region of ldrD in the same frame (Fig. 3B). The plasmid pMC-Plac-ATG fused with the promoter region and initiation codon of ldrD showed 10 times higher β-galactosidase activity than the plasmid pMC-Plac-SD lacking the initiation codon of ldrD (Fig. 3B). When fused with the N-terminal region of ldrD carrying the amber mutation (pMC-Plac-ldrD-A), β-galactosidase activity was observed only when the host cell carried the suppressor mutation of supF (Fig. 3B). These results strongly suggest that ldrD encodes a small peptide of 35 amino acids. Thus, the growth defect of the LDR-D sequence may be caused by ldrD.
We did not detect β-galactosidase activity from the plasmid pMC-PldrD-ATG containing the ldrD–lacZ gene fusion under the control of the ldrD promoter (Fig. 3B), even though the ldrD mRNA is transcribed constitutively (see below). Possibly, this transcript is not expressed in the cell because of the high toxicity of LdrD. However, we detected significant β-galactosidase activity (Fig. 3B) when the –10 promoter sequence of rdlD (see next section) in the plasmid was changed to a KpnI restriction enzyme recognition sequence (GGTACC), which reduced the transcription of rdlD (≈ 200-fold) but should not affect a secondary structure of ldrD mRNA, as predicted by MFOLD software at http://bioinfo.math.rpi.edu/~zukerm/ (Mathews et al., 1999; unpublished results).
Detection of transcripts from LDR-D
Northern blot analysis was performed to analyse the transcription from LDR-D on the chromosome. Single-stranded LDR-D RNA probes were prepared to distinguish the transcription direction (see Experimental procedures). As seen in Fig. 4A, lanes 1 and 2, a transcript of ≈ 370 nucleotides (nts) (ldrD mRNA) was identified in the wild-type and the LDR-ABC deletion strain on the same strand as ldrD, and a very weak signal was seen in lane 3, which contained the deletion mutant lacking LDR-D. It seems likely that this weak signal represents transcripts from the LDR-ABC locus detected by means of cross-hybridization. No signal was seen in the LDR-ABCD deletion strain. On the other hand, a transcript of ≈ 60 nts was identified on the strand opposite to the 370 nt transcript in the wild-type and the LDR-ABC deletion strain (Fig. 4B, lanes 1 and 2). This small RNA transcript was designated RdlD RNA (regulator detected in LDR-D). Faint bands of ≈ 290 nt were also seen, but their origins are currently not known. Using single-stranded LDR-ABC probes, RNA products of the same length were detected (data not shown). Curiously, Rdl RNA was detected only when using an identical complementary probe, which suggests that Rdl RNA may hybridize with a counterpart ldr mRNA transcribed in the same region in vivo.
To identify the location of the rdlD gene, Northern blot analysis was performed. The LDR-D region was divided into three regions (I: –42 to +80; II: +139 to +248; III: +303 to +452 in Fig. 5C), and probes were prepared from each region (see Experimental procedures). We could not identify any signal using probes I and III, but a distinct RdlD signal was identified using probe II (data not shown). To investigate whether the predicted –10 and –35 promoter sequences of rdlD actually had promoter activity, a fragment (+198 to +435 in Fig. 5C) was cloned into pMS434 (Hirano et al., 1987), the plasmid for promoter activity (see Experimental procedures). This fragment was found to function as a promoter (data not shown).
Mapping the 5′ end of ldrD mRNA and RdlD RNA by primer extension analysis
Two distinct transcripts were identified from the LDR-D sequence. To map the 5′ end of these two transcripts, primer extension experiments were performed. The ldrD mRNA was started at the A base at position +74 (Fig. 5A). The promoter corresponding to this start has a sequence that conforms closely to the canonical −10 sequence for σ70 (TAGAAT compared with TATAAT). However, no obvious σ70–35 sequence is present. The RdlD RNA started at the G at position +210 (Fig. 5B), which was the position 49 nt upstream of ldrD (+259) in the opposite direction (Fig. 5C). Both sets of σ70-dependent promoters were found upstream of rdlD, and the promoter se-quences were completely conserved among LDR-A, -B, -C and -D sequences (Fig. 2). No ORF is present in rdlD; thus, RdlD RNA may not be translated into a protein product.
Stability of the ldrD and rdlD transcripts
As two independent transcripts were identified on different strands of the same region, one of these transcripts might function as an antisense RNA to regulate the expression of a gene encoded in the other transcript (Gerdes et al., 1997). In the hok/sok system of plasmid R1, the antisense RNA is less stable than the hok mRNA (Gerdes et al., 1988). Northern blot analysis was performed in the presence of rifampicin to prevent the initiation of RNA synthesis in order to test whether one transcript is less stable and might therefore represent an antisense RNA. In general, the half-life of mRNA in E. coli has been reported to be 2–3 min (Belasco, 1993), and that of RdlD RNA was determined to be ≈ 2 min. On the other hand, ldrD mRNA was very stable, with a half-life of ≈ 30 min (Fig. 6). The RNA molecules from LDR-ABC had the same stabilities (data not shown).
The ldrD gene expression is regulated at a post-transcriptional level by the trans-acting rdlD gene product
As shown in Fig. 3B, the ldrD gene on plasmid pMC-ldrD-ATG was significantly expressed when a mutation, to reduce RdlD RNA expression, in the –10 sequence of the rdlD promoter was introduced. To test whether this mutation affects transcription from the ldrD promoter, we checked the expression using a lacZ transcriptional fusion plasmid. No change was observed between wild-type and mutated sequences (data not shown); thus, this mutation possibly regulates ldrD expression at the post-transcriptional level. If the rdlD gene product functions as an inhibitor of ldrD expression, oversupply of RdlD RNA will reduce LdrD synthesis. We constructed a plasmid, pRdlD, carrying part of the LDR-D DNA fragment (+134 to +249) and expressing RdlD RNA by its own promoter, and pRdlD was transformed into lysogen MKN005, which contains a single copy of the ldrD–lacZ translational fusion at the λatt locus with the antisense promoter mutation. The rdlD gene fragment in trans significantly reduced the β-galactosidase activity expressed from MKN005 (Fig. 7A). Therefore, the presence of pRdlD in the lysogen influences the production of the ldrD gene product. The same experiment was carried out with the transcriptional ldrD–lacZ fusion lysogen MKN006. No repression of β-galactosidase activity was observed (Fig. 7B). Therefore, we conclude that RdlD RNA seems to regulate the expression of ldrD at a post-transcriptional level. The mechanism of this antisense regulation is currently unknown, as the antisense RNA does not overlap with the translational initiation region of ldrD.
Overproduction of LdrD causes growth inhibition, loss of cell viability and nucleoid condensation
It is very likely that the growth inhibition seen with the increased number of LDR-D transcripts was caused by overproduction of the ldrD product. To learn more about the physiological function of the ldrD product, we cloned the ldrD DNA fragment into the expression plasmid pTrc99A (Amann et al., 1988) to construct pTr-ldrD and observed its effect on host cell growth. Two minutes after the induction of LdrD synthesis with 1 mM IPTG in mid-log phase, cell growth was arrested, and the numbers of viable cells decreased (Fig. 8A). One hour after induction, the viable cell counts decreased to 0.01% of the total number of viable cells before IPTG addition. When IPTG was removed from the culture by centrifugation, cell growth continued to be impaired. The results indicate that overproduced LdrD contributes to irreversible growth arrest.
To investigate the mechanism by which LdrD synthesis leads to growth inhibition, we performed a microscopic analysis of nucleoid structure using DAPI staining. The results shown in Fig. 8B demonstrate that the addition of 1 mM IPTG to the cells resulted in the rapid condensation of the nucleoid. This phenomenon started within 2 min of the induction, and one or two condensed nucleoids were observed after 30 min (Fig. 8B). After 5 h, further morphological changes, such as linear or circular nucleoids, were detected. The same physiological effect was produced using the complete LDR-D sequence (p18-D+) with 1 mM IPTG, but not by overproduction of ORF28 (data not shown).
Global transcriptional analysis by DNA microarray technique
To examine the physiological effects of LdrD overexpression, we carried out global transcriptional analysis. Total RNA was isolated from exponentially growing cultures (OD600 = 0.3) of strain KP7600 carrying pTr-ldrD before and 2 min after induction with 1 mM IPTG and analysed using DNA microarrays spotted with 4097 genes of E. coli (see Experimental procedures; Nakahigashi et al., 2002). In the duplicate experiments, we labelled the RNA from the control culture with Cy3 dyes and the RNA from induced culture with Cy5 dyes. Of the 3443 genes that reliably showed expression, nine genes were elevated more than threefold, and nine genes were reduced more than threefold after induction with IPTG in both experiments (Table 1). The effects of adding 1 mM IPTG on the whole-genome expression profile of wild-type cells were reported previously (Richmond et al., 1999; Wei et al., 2001). In these experiments, only a few genes, lacZ, lacY, lacA and melA, were induced, and no significant repression was observed. We also observed the upregulation of lacZ, lacY, lacA and additionally ldrD (ORF_ID = JW3507.1 and #_ID = 603#8.1 at GenoBase 4.0 web site), which were probably increased by the release of the LacI-mediated transcriptional repression. These increases provide a control that our microarray analysis was valid.
Table 1. Genes with induced and reduced expression after IPTG treatment.
In addition to lacZ, lacA and ldrD, we observed greater than threefold upregulation of genes involved in the purine metabolism pathway, purE, purK, purT, purF, purM, purL, purH and purD (Table 1). These genes are required for de novo synthesis of IMP (Zalkin and Nygaard, 1996) and are under the control of PurR and its co-repressors, hypoxanthine and guanine (Houlberg and Jensen, 1983; Meng and Nygaard, 1990; Rolfes and Zalkin, 1990). We did not observe a reduction in purR expression (Cy5:Cy3 ratios are 1.4 and 1.5), suggesting that LdrD over-expression might be changing PurR activity by affecting purine metabolism and the levels of hypoxanthine and guanine.
We also detected greater than threefold repression of several genes (Table 1). Most of these encode proteins of unknown function, although several are predicted to be localized to the inner membrane by PSORT software (http://psort.ims.u-tokyo.ac.jp/). The repressed genes in-clude the mglBAC operon, encoding genes involved in β-methylgalactoside transport system (Lin, 1996), dsdX, which is located in the D-serine tolerance locus, encoding a putative membrane protein that may function in D-serine transport (Nørregaard-Madsen et al., 1995), and yjiY, which encodes a homologue of the carbon starvation protein A (cstA), which may be involved in peptide utilization (Schultz and Matin, 1991). Interestingly, both the dsdX and cstA (Cy5:Cy3 ratios are 0.43 and 0.44) promoters are positively regulated by the cAMP–CRP complex (Schultz and Matin, 1991; Nørregaard-Madsen et al., 1995), and a consensus CRP binding site (TGTGAN6TCACA) (Kolb et al., 1993) is also found in the mglB (Hogg et al., 1991) and yjiY promoter regions. The expression of crp was not changed by LdrD induction (Cy5:Cy3 ratios are 0.86 and 0.98). Thus, another direct or indirect effect of elevated LdrD could be a reduction in cAMP levels. Although the proposed affects of LdrD on the hypoxanthine, guanine and cAMP levels need to be tested directly, the results of global transcriptional profiling suggest that overexpression of LdrD leads to physiological alterations in the cell.
Retrieval of LDR family sequences in enteric bacteria by database search and Southern blot analysis
To investigate the distribution of the LDR family in bacteria closely related to E. coli, we examined the conser-vation of LDRs between species. We searched for sequences homologous to LDR-D in several databases and found such sequences in E. coli O157:H7, Salmonella typhi CT18, Salmonella typhimurium LT2, Salmonella paratyphi A, Salmonella enteritidis and Salmonella dublin. These LDR-homologous sequences maintain the promoter, Shine–Dalgarno sequence, initiation site of transcription and termination sequences at appropriate positions (unpublished data). Multiple alignments of LdrD-homologous sequences revealed that the amino acid sequences are highly conserved, showing 40% identity and 71% similarity (Fig. 9). The common structural characteristics of the Ldr peptides suggest that the genes evolved from a common ancestor. Moreover, no deletion or insertion mutations were found in all the predicted ORF region, which strongly suggests that the ORF was under selection pressure during its evolution as a protein-coding gene.
We also carried out Southern blot analysis for 11 enteric species using the LDR-D sequence as a probe. The LDR-ABC- and LDR-D-containing fragments of E. coli W3110 appeared as 6.7 kb and 4.8 kb bands respectively (Fig. 10A). We observed fragments hybridizing with the LDR-D probe in the chromosomes of Salmonella typhimurium, Kluyvera ascorbata, Leclercia adecarboxylata and Citrobacter freundii. Multiple fragments were observed in S. typhimurium and Kluyvera ascorbata, suggesting the existence of multiple copies of LDR homologues. On the other hand, only one clear band was found in C. freundii and L. adecarboxylata. Recently, we identified an ldrD homologue in C. freundii and confirmed that this homologous gene has the same activity as ldrD (unpublished results). In the context of the phylogenetic tree constructed using the gyrB genes (Watanabe et al., 2001), LDR-homologous sequences might have arisen after the speciation of Hafnia alvei and E. coli (Fig. 10B). From this hypothesis, we expected Pantoea agglomerans and Cedecea davisae to have similar sequences; how-ever, no obvious hybridized fragments were detected. This inconsistency will be the subject of further investigations, including the reliability of the phylogenetic tree constructed with the gyrB genes.
We began this work to investigate whether LDRs may be functional units in the cell. To this end, we characterized the LDR sequences (and especially the LDR-D se-quence), constructed LDR deletion mutants and carried out overexpression analysis of LDR fragments. Deletion of these repeats did not result in an obvious phenotype and, thus, they are not essential for the cell. However, induction of transcription at the LDR-D locus from a multi-copy plasmid led to growth inhibition and a concomitant change in nucleoid morphology that was dependent on the direction in which the fragment was cloned into the plasmid. Molecular and genetic analyses suggested the existence of a gene, ldrD, within LDR-D that encodes a cell-killing gene product, and we have shown that the gene product might cause this physiological effect. Northern blot analysis showed that two RNA species are transcribed from the same region with the opposite orientation. The longer transcript is a stable ldrD mRNA, and the other transcript is an unstable small RNA, functioning as antisense RNA. Southern blot analysis also revealed that the LDR homologues were found in enteric bacteria closely related to E. coli.
Identification of ldrD in the LDR-D sequence
A gene finding program, GENEMARK, predicted the existence of a small ORF with 35 amino acids in LDR-D. Genetic analysis using the gene fusion of ldrD and lacZ under the control of the lac promoter strongly suggested that ldrD has the potential to code for a peptide. Several candidates for an initiation codon in the same reading frame as ldrD were tested by lacZ fusion, amber mutation and deletion analysis. A spontaneous mutation appeared in ldrD, and bacteria containing the mutation showed tolerance to the growth inhibition seen after transcriptional induction with IPTG. We performed further genetic experi-ments designed with the amber mutation at the second codon of ldrD. Growth inhibition by ldrD was lost after this mutation and was restored by the supF suppressor mutation (Fig. 3A). Based on these observations, we believe that ldrD is the gene encoding the 35-amino-acid cell-killing peptide. The Shine–Dalgarno sequence was also identified at the appropriate position of ldrD. However, we cannot rule out the possibility that other gene(s) exist in LDR-D. Similarly, the gene-finding program predicted a gene with 35 amino acids on LDR-A, -B and -C, and each predicted gene was highly homologous to ldrD (Fig. 9). We confirmed that each of the ldr genes (ldrA, ldrB, ldrC) had the same activity as ldrD presented in this work (data not shown).
Mechanisms of cell killing
The overproduction of the ldrD gene product leads to nucleoid condensation. Similar nucleoid condensation has been observed in cells treated with chloramphenicol to prevent protein synthesis (Kellenberger et al., 1958). Overproduction of the nucleoid protein H-NS also leads to nucleoid condensation and a strong inhibition of transcription and translation, and results in a drastic loss of cell viability (Spurio et al., 1992). We performed SDS–PAGE analysis to test whether the overproduction of the ldrD gene product caused the inhibition of protein synthesis. Two hours after treatment with chloramphenicol, degradation of total protein was observed; in contrast, overproduction of the ldrD gene product had no effect on the protein band patterns within 2 h (data not shown). These data suggest that overexpression of the ldrD gene product may not lead to the inhibition of protein synthesis in the cell. Chloramphenicol also causes the nucleoids to appear as hollow spheres with the denser material in the outer region, whereas expression of the ldrD gene product leads to particularly dense and homogeneous nucleoids with the appearance of filled spheres, similar to the effect of overproduction of H-NS (Spurio et al., 1992). Therefore, it seems likely that the mechanism of nucleotide condensation is different between chloramphenicol treatment and overproduction of ldrD gene product.
Condensation of nucleoid structure is a quick reaction that is observed within 2 min of induction of ldrD gene product synthesis. This speed, together with the failure to detect the ldrD gene product using SDS–PAGE, makes it seem unlikely that this condensation is caused by the accumulation of ldrD gene product. The ldrD gene product might interact with an unknown target that is important for maintaining normal nucleoid structure and cell growth. The physiological function of the LdrD peptide involved in the phenotype is at present unknown; however, micro-array analysis suggests that overexpression of LdrD leads to physiological alteration in purine metabolism (Table 1). Thus, the identification of the specific molecular target(s) of the small peptide will be the important next subject of investigation.
Analogy between the LDRs and the hok/sok system of plasmid R1
We observed a small antisense RNA complementary to the ldrD mRNA leader region (Fig. 4B). Similar genetic organization was found in the hok/sok system of plasmid R1, which functions in the stable inheritance of plasmids by selectively killing any plasmid-free cells. This phenomenon is referred to as post-segregational killing (PSK) (Gerdes et al., 1986). Five loci homologous to the hok/sok system have been found on the E. coli K-12 chromosome ( Pedersen and Gerdes, 1999). However, all the hok/sok homologues on the chromosome have lost PSK activity, probably through inactivation by insertion sequence (IS) elements, point mutation and a genetic rearrangement; the chromosome-encoded mRNA is insufficiently translated in vitro, thus explaining the absence of the PSK phenotype (Pedersen and Gerdes, 1999).
There are several similarities between the LDRs and hok/sok on the chromosome of E. coli: (i) the encoding of a small killing peptide (LdrD: 35 amino acids; HokA: 51 amino acids); (ii) the transcription of a highly stable mRNA and an unstable antisense RNA; (iii) the formation of a stable mRNA secondary structure (ldrD mRNA: 374 nts in length, ΔG = –177.24 kcal mol−1; hok mRNA: 433 nt in length, ΔG = –197.02 kcal mol−1; predicted using MFOLD software); and (iv) existence on the chromosome as multiple copies. In spite of their similarities, we failed to observe any homology between the LDRs and the hok/sok system within the amino acid or DNA sequences. Therefore, we tested the ability of the LDRs to mediate plasmid stabilization using a conditionally unstable plasmid containing a temperature-sensitive replicon, but no PSK activity was observed (data not shown). Also, no LDR-homologous sequences were found in plasmids by database analysis. These results suggest that the LDRs are genetic elements that are not used for stabilizing plasmid inheritance and are probably not involved in PSK in chromosomeless cells. The physiological function of the LDRs is currently unknown, but we think that they have the genetic organization of a toxin–antitoxin (TA) module, in which the regulators are unstable trans-acting antisense RNAs that inhibit the translation of toxin-encoding stable mRNAs. The instability of the antisense RNA induces translation of the toxin-encoding mRNAs. Gerdes (2000) noted that chromosomal TA loci are beneficial to cell survival by being part of the global cellular response to environmental stress, rather than being cell-killing modules, and also proposed that the TA loci are induced after amino acid and/or carbon source starvation, and co-ordinately regulate DNA replication and protein synthesis. Thus, our results suggesting that the LDRs may possess a function other than plasmid stabilization by PSK are in agreement with the above hypothesis.
Bacterial strains, media and culture conditions
The E. coli strain XL1-Blue (Bullock et al., 1987) was used during plasmid constructions and DNA manipulations. The W3110 derivative strain KP7600 (lacIqZΔM15) was used as a host strain of the growth inhibition assay and the β-galactosidase assay. The single copy number of the ldrD–lacZ fusion in lysogen, MKN005 and MKN006, was veri-fied by PCR (Powell et al., 1994). E. coli cells were grown in Luria–Bertani (LB) broth (Sambrook et al., 1989). Antibiotics (Sigma Chemical) were used at the following concentrations: ampicillin, 50 μg ml−1; rifampicin, 300 μg ml−1. IPTG was added to a final concentration of 1 mM.
DNA purification and manipulation
Construction of plasmids, transformations and purification of plasmids were performed according to the methods of Sambrook et al. (1989). The nucleotide sequencing reactions were performed using the ABI Prism BigDye terminator cycle sequencing ready reaction kit and analysed on the ABI Prism 310 genetic analyser according to the manufacturer's instructions.
The DNA fragment of LDR-D from E. coli W3110 was PCR amplified using primer LDR-D-F (5′-CTA CCA GTC TGG CGG CCG CCC GGA TAT G-3′) and primer LDR-D-R (5′-CGC CTG AAC GGC GGC CGC TGG GTG CCT TAC-3′). The NotI site added in the primers is in italics. The PCR product was inserted into the SmaI site of the plasmid pUC18 to yield p18-D+ and p18-D–. The orientation of the inserted DNA was checked by DNA sequencing. The PCR product was digested with XbaI, and the resulting 5′ overhangs were end-filled using T4 polymerase and nucleotides. The 287 bp fragment was purified from agarose gel using MagExtracter –PCR and Gel Clean up– (Toyobo). The DNA fragment was inserted into the SmaI site of pUC18 to yield p18-D+X. The orientation of the inserted DNA was verified by DNA sequencing. To construct p18-D+A, site-directed mutagenesis of ldrD was carried out using the p18-D+ plasmid as template DNA, PfuTurbo DNA polymerase (Stratagene) and primer SDM-F (5′-CAC AGG CGG CTA TAT GTA GTT CGC AGA TGG-3′) and primer SDM-R (5′-CCA GCT CTG CGA ACT ACA TAT AGC CGC CTG TG-3′) according to the procedure described in the Quick-Change site-directed mutagenesis kit (Stratagene). The sequence of the resulting mutated ldrD gene was verified by DNA sequencing. To construct pTr-ldrD, the plasmid pTrc99A (Amersham Pharmacia Biotech) was digested with NcoI, and the resulting 5′ overhangs were end-filled using T4 polymerase and nucleotides. A fragment carrying the ldrD gene was amplified by PCR with the sense primer ldrD-F (5′-ACG TTC GCA GAG CTG GGC ATG G-3′) and the antisense primer ldrD-R (5′-TTA CTT CCG CTT GTT CAG CCA GTT C-3′) and inserted into the NcoI blunted site. To construct pRdlD, the rdlD gene fragment of E. coli W3110 was amplified by PCR using primer rdlD-F (5′-GGT GGC AAAGCTTGC TGG AGA GAG AAA ACC CCC GC-3′) and rdlD-R (5′-CAT ATA GAAGCTTGT GTT GTA ATG ACA ACG TTT CGC-3′). The HindIII restriction enzyme recognition site added in the primers is underlined. The fragment was then digested with HindIII and cloned into the corresponding site of pACYC184. The Plac-SD DNA fragment from p18-D+ was PCR amplified using primer Plac-F (5′-GAC AGA ATT CCC GAC TGG AAA GCG GGC-3′) and primer SD-R (5′-CAT ATA GGGATCCGT GTT GTA ATG ACA ACG TTT CGC-3′). The EcoRI and BamHI restriction enzyme recognition sites added in the primers are in italics and underlined respectively. The BamHI-digested PCR product was inserted into the SmaI–BamHI site of the lacZ reporter plasmid pMC1403 to yield the plasmid pMC-Plac-SD. The Plac-ATG DNA fragment from p18-D+ was PCR amplified using primer Plac-F and primer ATG-R (5′-CTC TGC GGGATCCAT ATA GCC GCC TGT GTT G-3′). The BamHI restriction enzyme recognition site added in the primer is underlined. The BamHI-digested PCR product was inserted into the SmaI–BamHI site of pMC1403 to yield the plasmid pMC-Plac-ATG. The Plac-ldrD-A DNA fragment from p18-D+A was PCR amplified using primer Plac-F and primer ldrD-A-R (5′-CAT GAC AGGATCCTT CCG CTT GTT CAG CCA G-3′). The BamHI restriction enzyme recognition site added in the primer is underlined. The BamHI-digested PCR product was inserted into the SmaI–BamHI site of pMC1403 to yield the plasmid pMC-Plac-ldrD-A. The PldrD-ATG DNA fragment from p18-D+ was PCR amplified using primer PldrD-F (5′-GGA GGC GGAATTCTA CAA AAT TGC G-3′) and primer ATG-R. The EcoRI restriction enzyme recognition site added in the primer is underlined. The EcoRI- and BamHI-digested PCR product was inserted into the corresponding sites of pMC1403 to yield the plasmid pMC-PldrD-ATG. To construct pMS-PrdlD, the rdlD promoter fragment of E. coli W3110 was amplified by PCR using primer PrdlD-F (5′-GAA CGA AAA AGG CCCTCGAGT TTC CCC CCT GC-3′) and PrdlD-R (5′-GGG GCT AAG CTT GAC TCT AGA CCA C-3′). The XhoI and HindIII restriction enzyme recognition sites added in the primers are underlined. The fragment was then digested with XhoI and HindIII and cloned into the corresponding sites of pMS434. The correct ligation was verified by DNA sequencing.
Construction of LDR null mutants
For LDR-A, -B and -C deletions, the plasmid pChaA-KdsA was constructed by cloning a 1.5 kb SacI–NotI terminal chaA DNA fragment and a 1.6 kb PstI–NotI terminal kdsA DNA fragment amplified by PCR from E. coli W3110 as a template between the SacI and PstI sites of pUC18. A chloramphenicol cassette amplified by PCR from pHSG396 was inserted into the NotI site of pChaA-KdsA to yield pChaA-Cm-KdsA. For the LDR-D deletion, the plasmid pYhjU-YhjV was constructed by cloning a 1.5 kb BamHI–NotI terminal yhjU DNA fragment and a 1.2 kb SacI–NotI terminal yhjV DNA fragment amplified by PCR from E. coli W3110 as a template between the SacI and BamHI sites of pUC18. A kanamycin cassette amplified by PCR from pHSG396 was inserted into the NotI site of pYhjU-YhjV to yield pYhjU-Km-YhjV. The nucleotide sequence of the entire cloned DNA region was verified by DNA sequencing. Homologous DNA fragments were obtained from these plasmids, which were digested with SacI and PstI (pChaA-Cm-KdsA), and SacI and BamHI (pYhjU-Km-YhjV). The linear DNA fragment was introduced into the recD strain FS1576 by electroporation and integrated into the chromosome by homologous recombination (Russell et al., 1989), with subsequent selection for kanamycin or chloramphenicol resistance. P1 transduction was performed to transfer the LDR deletion mutation from the recD strain to a wild-type strain W3110. Transducing P1 lysate was prepared by growing P1 vir on the above mutant strains in R-top agar on R plates (per litre: 10 g of Bacto tryptone, 1 g of Bacto yeast extract, 12 g of Difco agar, 8 g of NaCl; after autoclaving, 2 ml of 1 M CaCl2 and 5 ml of 20% glucose were added). The cells to be transduced were grown in LB medium and resuspended with MC buffer (0.1 M MgSO4, 5 mM CaCl2). The cells were infected with the diluted P1 lysate, with subsequent selection for kanamycin and/or chloramphenicol resistance. All LDR deletion mutants were shown by Southern blot analysis, PCR and DNA sequencing to have acquired the desired mutations.
The β-galactosidase activity was determined according to Miller (1992). A sample of 900 μl of Z buffer (60 mM Na2HPO4, 40 mM NaH2PO4, 10 mM KCl, 1 mM MgSO4, 50 mM β-mercaptoethanol, pH 7.0), 100 μl of cultures, 20 μl of chloroform and 10 μl of 0.1% (w/v) SDS were mixed at room temperature. Triplicate aliquots were taken of each culture at an OD600 of ≈ 0.5–0.7.
Preparation of total RNA
The culture was grown at 37°C in LB medium to an OD600≈ 0.5. The culture sample (20 ml) was transferred to a 50 ml tube, and ≈ 10 ml ice blocks were added to the tube. The cells were pelleted by centrifugation at 5000 g for 3 min at 4°C and resuspended in 500 μl of solution A (0.5% SDS, 20 mM sodium acetate, 10 mM EDTA, pH 5.5) and mixed thoroughly. Then, 500 μl of phenol (equilibrated in 20 mM sodium acetate, 10 mM EDTA, pH 5.5) was added to the suspension. After vortexing, the mixture was incubated for 5 min at 60°C and transferred to the 1.5 ml tube. The phases were separated by centrifugation, and the aqueous phase was transferred to the 1.5 ml tube. The RNA was precipitated with 1 ml of 99% ethanol, and the RNA pellet was washed with 1 ml of 70% ethanol. The RNA pellet was resuspended in 400 μl of solution A, and ethanol precipitation was repeated twice more. The final RNA pellet was dissolved in 100 μl of TE (10 mM Tris-HCl, 1 mM EDTA, pH 8.0). The RNA concentration was determined by measuring absorbance at 260 nm, and its purity was checked by means of the A260/A280 ratio. The RNA sample was stored at –80°C. We usually obtained ≈ 600 μg of total RNA from the 20 ml culture.
Northern blot analysis
Ten micrograms of total RNA was loaded into each lane of a 1% (w/v) agarose–formaldehyde gel. After electrophoresis, RNA was blotted onto a positively charged nylon membrane (Roche Diagnostics) and hybridized with a digoxigenin (DIG)-labelled RNA probe at 68°C overnight. The DIG-labelled RNA probes were generated by in vitro transcription with T7 RNA polymerase and DIG RNA labelling mix (Roche Diagnostics) using T7 promoter sequence-tagged PCR products as templates. The LDR-D+ probe is complementary to ldrD mRNA (504 nt from –56 to +448). The LDR-D- probe is complementary to RdlD RNA (537 nt from –46 to +491). The hybridization and detection protocols used were those provided by Roche Diagnostics.
RNA half-life determination
A 200 ml culture of strain W3110 was grown at 37°C in LB medium to OD600≈ 0.5. Rifampicin stock solution was added to a final concentration of 300 μg ml−1. Total RNA was extracted from 20 ml aliquots of culture taken at 0, 2, 5, 15, 30, 60 and 100 min after the addition of rifampicin. A Northern blot analysis was performed using the DIG-labelled LDR-D+ and LDR-D– RNA probes. Quantification of the signal was performed using a Densitograph Lumino-CCD apparatus and ATTO densitograph software provided by ATTO.
A 32P-labelled oligo DNA primer (0.5 pmol with respect to the ldrD mRNA transcription start site, 0.25 pmol with respect to the RdlD RNA transcription start site) from within the LDR-D sequence (+132 to +161 for ldrD mRNA; 5′-ACG TGC GGG GGT TTT CTC TCT CCA GCA ACC-3′), within the trpB gene (downstream from the HindIII site on the plasmid pMS434 for the RdlD RNA; 5′-TGT CTT TAT CGC CGC GAC CGG AAA GGT TGC-3′) was mixed with total RNA (30 μg for ldrD mRNA, 6 μg for RdlD RNA) in 10 μl of 2× annealing buffer [20 mM Tris-HCl, (pH 8.3), 2 mM EDTA, 500 mM KCl], then heated to 85°C for 1 min and allowed to anneal for 60 min at 65°C and 90 min at room temperature. The primer was extended using 200 units of Superscript II (Invitrogen) at 50°C for 60 min after the addition of 60 μl of extension buffer [47 mM Tris-HCl, (pH 8.3), 17 mM KCl, 5 mM MgCl2, 13 mM dithiothreitol (DTT), 0.33 mM each dNTP]. Reactions were stopped by the addition of 2 μl of 0.5 M EDTA and 2 μl of RNase A (10 μg ml−1), and they were incubated at 37°C for 30 min and 70°C for 10 min. The samples were precipitated with ethanol and resuspended in 10 μl of loading buffer [80% formamide, 5 mM EDTA, (pH 8.0), 0.05% xylene cyanol, 0.05% bromophenol blue]. The extension products were analysed by electrophoresis in an 8% acrylamide sequencing gel containing 7 M urea, followed by autoradiography. The DNA sequence ladder of the LDR-D, synthesized using the same labelled primers and a Takara Taq cycle sequencing kit (Takara Shuzo), was analysed alongside the extension primers.
Fluorescence microscopy of cells and nucleoids
Escherichia coli cells from a culture were collected by centrifugation and washed once with 1 ml of 10 mM Tris-HCl (pH 7.5). The washed cells were suspended in 0.1 ml of 10 mM Tris-HCl, and 1 ml of 77% ethanol was added for fixation. The fixed cells were collected by centrifugation, washed with the same buffer, spread on a glass slide pretreated with poly L-lysine and dried at room temperature. DAPI (4′,6-diamidino-2-phenylindole) solution (1 μg ml−1, 150 mM NaCl in 50% glycerol) was dropped on the sample. A clean glass coverslip was put on the drop. Images were taken by combined fluorescence and phase-contrast microscopy (Zeiss Axioskop 2 fluorescence microscope, equipped with Plan-Apochromat 100×/1.4 oil immersion lens and SPOT cooled CCD camera; Diagnostic Instrument). The images were processed using ADOBEPHOTOSHOP (version 6.0) software.
RNA isolation for DNA microarray
Preparation of total RNA for DNA microarray was performed using basically the same hot phenol method as used for Northern blot analysis. The differences are that the hot phenol extraction process was repeated once, and the final RNA pellet was dissolved in a DNase solution (100 mM sodium acetate, 50 mM MgSO4) containing 5 units of RNase-free DNase I (Takara Shuzo) and incubated at room temperature for 1 h. Then, phenol–chloroform extraction and RNA precipitation with ethanol were performed. Purified total RNA was subjected to 1% agarose gel electrophoresis to check for degradation and whether the 23S and 16S ribo-somal RNA were recovered without the contamination of genomic DNA.
Preparation of DNA microarrays
We used custom-made microarrays of DNA molecules on glass slides that had been prepared by Takara Shuzo. The array contains the 4097 independent genes of the E. coli genome that have been cloned previously from E. coli K-12 W3110, the so-called archive clones (Mori et al., 2000). Each gene on the slide was completely amplified by PCR using vector-specific primers targeting both sites of the integrated gene fragment (primer 1: 5′-ATC ACC ATC ACC ATA CGG ATC CGG CCC TGA-3′; primer 2: 5′-TTC TTC TCC TTT ACT GCG GCC GCA TAG GCC-3′). The PCR-amplified fragments contained the DNA region spanning from the second to the last codon of all genes in E. coli. The DNA concentration was more than 0.1 mg ml−1. Furthermore, all PCR fragments were confirmed by DNA sequencing and were spotted at high density in duplicate on a single slide, avoiding the need for slide-to-slide correction. In addition to the genes mentioned above, there were 24 spots of human transferrin receptor gene as a negative control on the slide; E. coli genomic DNA and fluorescent position marker were spotted as negative and positive control and positional marker to judge the spotting error. The microarray of the E. coli genome is now available from Takara Shuzo.
Fluorescent-labelled cDNA preparation, array hybridization and the capture of data
Fluorescent-labelled cDNA probes were prepared by random priming methods. Reverse transcriptase reactions were performed with reverse transcriptase XL (Life Science) and 4 nmol of either Cy3-dUTP or Cy5-dUTP (Amersham Biosciences) using 30 μg of total RNA. Labelled cDNA probes were purified by Centri-sep (Princeton Separations), phenol–chloroform extraction and ethanol precipitation. After drying, the cDNA probe was dissolved in 9 μl of water. Both Cy3- and Cy5-labelled cDNA probes were then added to a final volume of 23 μl of hybridization buffer (4× SSC, 0.2% SDS, 5× Denhardt's solution, 100 ng ml−1 salmon sperm DNA) and denatured by heating at 98°C for 2 min. The denatured cDNA probe was applied to the microarray prehybridized with 100 ng ml−1 salmon sperm DNA under a coverslip. Hybridization was carried out at 65°C for 16 h. Slides were washed at 60°C with 2× SSC for 5 min, then at 60°C with 0.2× SSC containing 0.1% SDS and, finally, at room temperature with 0.2× SSC. The slides were scanned for fluorescent intensity using a GMS 418 array scanner (Genetic Microsystems) and recorded to 16 bit image files. The signal density of each spot in the microarray was quantified using IMAGENE software (BioDiscovery). At least two independent experiments were performed on each growth condition. The duplicate microarray samples were analysed in an experiment.
Data analysis of microarrays
To distinguish reliable data from the background, we corrected each spot for the local background, which was determined as the mean pixel intensity in the circle surrounding each spot and a mean value of the intensity of the 24 negative control spots, and determined a standard deviation (SD). The corrected signal intensity of each spot was then classified into three groups. Group 1 consisted of data where both the Cy3 and the Cy5 signal intensities derived from the control and experiment cDNA, respectively, were greater than the mean +1 SD of the negative controls. Group 2 consisted of spots where either, but not both, the Cy3 or the Cy5 signal intensities were greater than the mean +1 SD of the negative controls. Group 3 consisted of spots where both the Cy3 and the Cy5 signal intensities were lower than the mean +1 SD of the negative controls. The distribution of ratios of all spots including group 1 was nearly log-normal. We then normalized the intensity of all spots in group 1. This was done by calculating the following ratio for the spots in group 1: mRNA level from KP7600 carrying pTr-ldrD (2 min) labelled by Cy5/mRNA level from KP7600 carrying pTr-ldrD (0 min) labelled by Cy3. The mean of the entire ratio in group 1 was defined as 1.0, and the individual spots were divided by the mean to yield a comparative value. The ratio of the group 2 spots could not be determined because of the lack of either a Cy3 or a Cy5 fluorescent signal. Spots in this group with high intensity (over 1000; it was a sufficiently high intensity value to be detected precisely by the GMS 418 array scan-ner) of Cy3 or Cy5 were considered to represent altered expression levels in the experiment sample relative to the control sample. Spots in group 3 were considered to be undetectable spots. The expression ratios of all the genes are available at http://ecoli.aist-nara.ac.jp/xp_analysis/ldrD_2/.
Southern blot analysis
Genomic DNA from Hafnia alvei, Providencia alcanifaciens, Moellerella wisconsensis, Leminorella grimontii, Pantoea agglomerans, Leclercia adecarboxylata, Kluyvera ascorbata, Cedecea davisae, Citrobacter freundii, Salmonella typhimurium LT2 and Escherichia coli W3110 were use. Genomic DNA was digested with EcoRV, and 2–4 μg of DNA was loaded into each lane of a 0.8% (w/v) agarose gel. After electrophoresis, DNA was blotted onto a positively charged nylon membrane (Roche Diagnostics) and hybridized with the DIG-labelled RNA probe at 37°C overnight. The hybridization and detection protocols used were those provided by Roche Diagnostics.
DNA alignment and computer analysis
DNA alignment between E. coli sequences and those of Salmonella were carried out using the GenBank web site (http://www.ncbi.nlm.nih.gov:80/), and preliminary sequence data were obtained from The Institute for Genome Research web site (http://www.tigr.org). The data for the S. typhimurium LT2 and S. paratyphi A sequences were obtained from the Washington University Consortium sequencing project web site (http://genome.wustl.edu/gsc/Projects/bacteria. shtml). The S. typhi CT18 sequences were obtained from the Sanger Centre sequencing project web site (http://www.sanger.ac.uk/Projects/S_typhi/). The S. enteritidis and S. dublin sequences were obtained from the University of Illinois sequencing project web site (http://www.salmonella.org/). The LDR-D sequence was used in a BLAST search as the query, and the alignment was expanded manually from the BLAST alignment using the contig for the unfinished genome fragment.
We thank M. Kitagawa for technical advice, T. Miki for providing strain KP7600, and H. Inokuchi for the gift of several phages for constructing supF strains. We also thank Y. Masuda for making a web site to present microarray data. We appreciate the helpful comments of S. Hiraga, C. Wada, T. Horiuchi, S. Moriya and K. Matsubara. We are grateful to G. Storz for critical reading of the manuscript. We are indebted to intriguing hok/sok work by K. Gerdes and colleagues, which gave us a lot of ideas for performing this work. This work was supported by a Grant-in-Aid for Scientific Research on Priority Areas from the Ministry of Education, Culture, Sports, Science and Technology of Japan and a grant from CREST, JST (Japan Science and Technology).