Superintegrons (SIs) are chromosomal genetic elements containing assemblies of genes, each flanked by a recombination sequence (attC site) targeted by the integron integrase. SIs may contain hundreds of attC sites and intrinsic instability is anticipated; yet SIs are remarkably stable. This implies that either selective pressure maintains the genes or mechanisms exist which favour their persistence in the absence of selection. Toxin/antitoxin (TA) systems encode a stable toxin and a specific, unstable antitoxin. Once activated, the continued synthesis of the unstable antitoxin is necessary for cell survival. A bioinformatic search of accessible microbial genomes for SIs and TA systems revealed that large SIs harboured TA gene cassettes while smaller SIs did not. We demonstrated the function of TA loci in different genomic contexts where large-scale deletions can occur; in SIs and in a 165 kb dispensable region of the Escherichia coli genome. When devoid of TA loci, large-scale genome loss was evident in both environments. The inclusion of two TA loci, relBE1 and parDE1, which we identified in the Vibrio vulnificus SI rendered these environments refractory to gene loss. Thus, chromosomal TA loci can stabilize massive SI arrays and limit the extensive gene loss that is a hallmark of reductive evolution.
Bacterial genomes are in a constant state of flux and perpetual gene loss and acquisition have a significant impact on bacterial evolution and genome architecture (Jain et al., 1999; 2002; Ochman and Moran, 2001). Reductive evolution, also referred to as the streamlining of bacterial genomes, can occur when selective pressures are insufficient for gene persistence, i.e. dramatic reductions in genome size can arise not from an increased selection for DNA loss but rather from a decreased selection to maintain gene function. However, natural selection favours the evolution of strategies that increase the rate of adaptation, that is, chance favours the prepared genome (Caporale, 1999). Hence, the balance of deletions to insertions is a tradeoff between minimizing the metabolic burden and maximizing the adaptive potential. Thus, despite the imposed genetic burden, the evolution of strategies to minimize gene loss in the absence of selection may prove beneficial for bacterial survival.
Integrons are simple yet potent genetic systems for maximizing the adaptive potential of bacteria (Rowe-Magnus and Mazel, 2001). The integron platform codes for an integrase (intI) that mediates recombination between a proximal primary recombination site (attI) and a target recombination sequence called an attC site (Stokes and Hall, 1989; Hall and Stokes, 1993). The attC site is usually found associated with a single open reading frame (ORF) in a circularized structure termed a gene cassette (Collis and Hall, 1992a; Recchia and Hall, 1995). Insertion of the gene cassette at the attI site places it downstream of a promoter, Pc that is internal to the intI gene. Pc then drives expression of the encoded proteins (Collis and Hall, 1992b; Collis et al., 1993; Levesque et al., 1994). The substantial impact of integrons on bacterial evolution is underscored by the present dilemma in the treatment of infectious disease, as the development of multiple-antibiotic resistance can often be traced to the stockpiling of resistance loci within integrons. Resistance integrons (RIs) harbouring up to eight resistance cassettes have been isolated (Naas et al., 2001) and are in large part responsible for the evolution of multidrug resistance among diverse human, animal and plant pathogenic isolates (Rowe-Magnus and Mazel, 1999; 2002; Rowe- Magnus et al., 2002a). Integron platforms are commonly found embedded within mobile DNA elements such as transposons and conjugative plasmids that can serve as vehicles for the intra- and interspecies transmission of the resistance genes that have been amassed by integrons. The proficiency of this partnership is confirmed by the marked differences in codon usage among cassettes within the same RI, indicating that the antibiotic resistance determinants are of diverse origin.
The evolutionary divergence among the five classes of RI integrase genes indicated that the activity of this system has extended far beyond the 60 years of the antibiotic era (Rowe-Magnus et al., 2001). This postulate was confirmed by the recent discovery of massive ancestral versions, the superintegrons (SIs), in the genomes of diverse proteobacterial species (Mazel et al., 1998; Rowe-Magnus et al., 1999; 2001; 2003; Vaisvila et al., 1999; Coleman et al., 2004). The SIs of Vibrio species can be greater than 120 kb in length (Heidelberg et al., 2000; Chen et al., 2003). As such they contain hundreds of integrated gene cassettes with attC sites that are all arranged in direct orientation. Furthermore, multiple copies of gene cassettes and insertion sequences have been identified in SIs. All are targets for recombination and if selective pressures are insufficient for gene persistence, intervening DNA can be rapidly eliminated to provide compact chromosomes to minimize the genetic burden (Lawrence et al., 2001). Clearly, ample opportunities exist for large-scale deletions and rearrangements to occur, and an inherent instability of SIs is expected. Yet, paradoxically, SIs are remarkably stable. The genetic organization of RIs promotes coexpression of the inserted gene cassettes from a single promoter. Hence, selection for one resistance determinant often co-selects for the maintenance of the entire array. A similar situation for SIs is difficult to imagine. The stability of SIs implies either the existence of selective pressures that maintain the gene cassettes of SIs or mechanisms that favour their persistence in the absence of selection.
Addiction modules (Yarmolinsky, 1995; Engelberg-Kulka and Glaser, 1999) are selfish genetic elements that ensure their propagation by preventing the proliferation of module-free progeny. Two classes of addiction modules have been defined: toxin/antitoxin (TA) and Restriction–methylation (RMS) systems (for review see Jensen and Gerdes, 1995; Kulakauskas et al., 1995; Kusano et al., 1995; Gerdes et al., 1997; 2005; Couturier et al., 1998; Kobayashi et al., 1999; Rawlings, 1999; Handa et al., 2000; Kobayashi, 2001). TA loci are commonly found either on plasmids or within prophages and they have been found to enhance plasmid and phage maintenance in bacterial populations by preventing the proliferation of plasmid or phage-free progeny. They are comprised of two small ORFs that are organized as an operon. The downstream gene specifies a stable toxin while the upstream gene codes for a specific but unstable antitoxin. Once activated, any disruption in the expression of the TA system leads to the accumulation of free toxin due to rapid loss of the antitoxin's neutralizing activity and results in bacterial stasis. Consequently, expression of the operon renders cells ‘addicted’ to the short-lived antitoxin because its continued de novo synthesis is essential for cell growth. RMS fit all the properties of post-segregational killing systems. Once acquired, they become essential for the survival of the bacteria because the long half-life of the nuclease compared with the methylase will eventually cause cell death if the RMS is lost. Thus, the bacteria become dependent on the invading RMS.
We previously identified a TA system that was structured as a gene cassette in the SI of V. fischeri and demonstrated that this cassette encoded an addiction module of the ccdAB family (Rowe-Magnus et al., 2003). We also identified a doc/phd P1 toxin analogue that was structured as a gene cassette in the SI of V. metschnikovii as well as nine SI cassettes encoding higAB, doc/phd and parDE homologues in the V. cholerae SI (Rowe-Magnus et al., 2003). Additionally, two RMS are known to be SI gene cassettes (Vaisvila et al., 1999; Rowe-Magnus et al., 2001). These observations led us to develop the hypothesis that addiction loci coded by SI gene cassettes acted to stabilize these massive arrays (Rowe-Magnus et al., 2003). Here, we conducted a convergent bioinformatic search of the available microbial genomes for SIs and TA systems. This analysis revealed that large SIs contained gene cassettes encoding TA loci while smaller SIs did not. We showed that two TA loci gene cassettes in the V. vulnificus SI: (i) encoded functional TA systems; (ii) were expressed in vivo; (iii) stabilized SI gene cassette arrays by minimizing large-scale deletions without hindering their microevolution and (iv) when highly expressed, can limit the extensive gene loss that is a hallmark of reductive evolution (Mira et al., 2001; Rocha, 2003; Wernegreen, 2005).
Identifying SIs and TA gene cassettes in microbial genomes
Standard blastp analysis using the integron-integrase proteins IntI1 and IntIAVvu (accession numbers AAA92752 and AAN33109 respectively) as bait was performed against the completed or whole genome shotgun sequences of the available microbial genomes as described in the Experimental procedures. This analysis revealed IntI homologues in the genomes of the bacteria listed in Table S2. We used clustalx to generate a protein alignment and compile a dendrogram of the IntI protein sequences. The three Patches (I, II and III) and two Boxes (I and II) of amino acids that define the integrase family (Nunes-Duby et al., 1998) were identified within the catalytic domains of each of the IntI homologues (see Supplementary Figure S1). Each of the homologues also possessed the conserved RHRY tetrad of amino acids located in Boxes I and II that are critical for the recombination activity of IntI1, suggesting that they all catalyse recombination by the same mechanism. The sole exception among the complete IntI sequences was that of Desulfotalea psychrophila LSv54, which carried a G residue in place of the conserved R residue in Box I. A previous alignment of integron integrases with other tyrosine recombinases showed that they possessed an additional stretch of approximately 35 amino acids that flank the Patch III motif (Messier and Roy, 2001). These additional amino acids were also present in the IntI homologues listed in Table S2 and were recently shown to be essential for site-specific recognition by the integrase (MacDonald et al., 2006). Because the catalytic residues were absolutely conserved among all but one of the integron integrases identified, we presumed that they all use the same mechanics for cassette recombination.
Dendrograms showed that, despite their divergence, the integron integrases clearly formed a related but distinct group from other members of the Int family of site-specific recombinases (Fig. 1), represented here by XerC/D, Cre, P2, P22, e14 and λ. Furthermore, the integrases partitioned into class, genus and species-specific clades. For example, we previously showed that the majority of the integron integrases from the γ-proteobacterial class clustered together and then partitioned into Vibrio, Shewanella and Xanthomonad genus-specific clades (Rowe-Magnus et al., 2001). Species-specific clustering was also evident, as the integrases of the closely related V. cholerae, V. mimicus, V. metschnikovii and Listonella anguillarum species clustered together, while the integrases of the more distantly related V. parahaemolyticus, Listonella pelagia, V. vulnificus and V. fischeri species formed separate clades. A similar distribution pattern emerged for the integrase genes of the Pseudomonas, Shewanella and Xanthomonas genera.
Some other notable phylogenetic patterns were revealed. The integrase of Rhodopirellula baltica, a planctomycete partitioned early in the dendrogram and may represent an evolutionary link between the XerC/D recombinases and integron integrases. The integrase of Geobacter metallireducens, a δ-proteobacterium, grouped with the integrases of the β-proteobacterial class, as did the integrases of the γ-proteobacterial Pseudomonads. These cases likely represent examples of the horizontal acquisition of all or part of a SI in these microorganisms. In support of this notion, the integrase protein that was identified on the V. salmonicida plasmid, pRVS1 (Accession No. AJ277063), was 99% identical to the integrase of Pseudoalteromonas haloplanktisTAC125 and the nucleotide sequences of the integron platforms (i.e. the region from the integrase gene to the GTT/A cross-over point of the attI site) from theses two bacteria were 99% identical. The attC sites of the P. haloplanktis SI shared an overall homology of 74%, and were 77% and 72% identical to the attC sites of the first and fifth cassettes of the integron on pRVS1. These results suggest that the integron platform of pRVS1 was acquired by lateral transfer from the integron platform in the genome of P. haloplanktis. To our knowledge, this is the first example of an identifiable SI source for an integron on an extrachromosomal DNA element.
We sought evidence of cassette recombination by using a software program we developed, xxr (Rowe-Magnus et al., 2003), to identify and extract gene cassettes from the respective SIs. Once the intIA gene and the gene cassette array adjacent to it were delimited, we also identified the signature attC sites of each species and then blasted the nucleotide sequences against the respective genomes to identify displaced or disrupted array segments elsewhere in the genome because genomes are dynamic and genomic rearrangements can disrupt the continuity of a SI. In many cases, neither an attI site nor any attC sites could be identified. In other cases, only a small or disrupted gene cassette array was detected.
blastp analysis was used to scan each gene cassette array for toxins and antitoxins belonging to the seven known TA families (Gerdes et al., 2005; Pandey and Gerdes, 2005). The arrays were also scanned for the presence of un-annotated gene cassettes that potentially coded for operons of two small overlapping reading frames characteristic of TA systems. In some cases, we could identify a closely linked, annotated antitoxin partner. If an annotated partner was not apparent, we looked for un-annotated ORFs in the adjacent DNA. Each homologue was carefully checked for the presence of an associated attC site characteristic of an integron gene cassette. All of the integron TA gene cassettes that we identified showed homology to previously described TA family members. The toxin homologues (Table S2), along with the 13 integron TA gene cassettes of the V. cholerae SI (Rowe-Magnus et al., 2003; Pandey and Gerdes, 2005) and representatives of each of the seven known TA families, were included in the construction of the dendrogram shown in Fig. 2. The number of identified TA homologues relative to the size of the SI is presented in Table S2. A clear and simple relationship emerged from this analysis – gene cassettes encoding potential TA systems were present in larger SIs but were absent from smaller SIs.
It was also evident that the incorporation of TA loci into SIs did not prevent their fragmentation. We identified two integron TA gene cassettes in the Photobacterium profundum SI that coded for a ParE (PBPRB1193) and a RelE (PBPRB1201) homologue respectively (Table S2). However, the integrase gene contained a frameshift that rendered it non-functional and the cassette array was disrupted; 15 kb of the array was adjacent to the integrase while another 15 kb fragment of the array was found 490 kb away. Other gene cassettes were scattered throughout the genome.
We also observed that, unlike V. cholerae (Pandey and Gerdes, 2005), the SI-containing bacteria examined here all harboured at least one chromosomal TA locus that was not structured as an integron gene cassette. For example, V. vulnificus CMCP6 had two chromosomal loci that were homologues of RelE (VV21493) and HipA (VV12047). In P. profundum, a RelE homologue (PBPRA1506) was identified on chromosome I but it was not structured as a gene cassette. A second ParE homologue (PBPRC0063) was also identified on a resident plasmid. Three additional TA loci belonging to the HipA family, TDE1634, TDE2105 and TDE2050, were identified in the chromosome of Treponema denticola but they were not structured as gene cassettes. These TA homologues were included in Fig. 2. Thus, the presence of chromosomal TA systems is likely to be a common feature of the genomes of free-living microbes (Pandey and Gerdes, 2005).
Functional analysis of TA loci in the V. vulnificus SI
Five of the annotated gene cassettes of the V. vulnificus CMCP6 SI showed homology to TA loci or fit the criteria of a TA operon: vv12409/vv12410, vv12432/vv12433, vv12463/vv12464, vv12525/vv12526 and vv12546/vv12547. We designed specific primers for the putative toxin gene of each pair for use in PCR with V. vulnificus 75.4T genomic DNA as matrix. Primers towards vv12410 (vv12410Nde and vv12410Xba) generated a PCR product. We were able to amplify a homologue of vv12525 from genomic DNA as well with the primer pair vv12525Nde and vv12525Xba. A second PCR was then conducted on genomic DNA with primers VVR1 or VVR2 and vv12410Xba and VVR1 or VVR2 and vv12525Xba to amplify each TA pair. These fragments were cloned and sequenced. The insert of pCR2.1::VvuSI-12525/12526 was 737 bp and carried a single gene cassette coding for two small overlapping reading frames. The 1100 bp insert of pCR2.1::VvuSI-12409/12410 contained three ORFs structured as two gene cassettes. Sequence alignments showed that the putative proteins coded by vv12410 and vv12433 were identical to one another and that they were 94% identical at the amino acid level to the vv12410 homologue we amplified. The closest homologues of Vv12410, Vv12433, Vv12547 and Vv12525 were among the 13 TA loci identified as SI gene cassettes in the genome of V. cholerae by Pandey and Gerdes (2005). Vv12410 and Vv12433 were homologues of RelE6, Vv12547 was related to RelE3 and Vv12525 was a homologue of ParE2. Vv12464 was a homologue of HigB of Salmonella typhimurium LT2 (Fig. 2).
As the overexpression of many different proteins can result in phenotypic toxicity, even if the protein does not function as a toxin per se, and the cloning of overlapping genes into the same expression vector could negatively impact on the expression of 3′ ORFs (the putative toxins, here), we opted to clone the putative toxin and antitoxin of each pair into separate, compatible expression vectors (see Supplementary Figure S2) to preclude these possibilities. We amplified vv12410 and vv12525 and cloned each into the isopropyl β-D-thiogalactopyranoside (IPTG)-inducible expression vector pTRC99A2. In both cases, growth of E. coliπ1 cells was evident in the absence of IPTG but ceased in the presence of IPTG. The ability of Vv12409 and Vv12526 to neutralize the toxicity of Vv12410 and Vv12525, respectively, was tested by cloning vv12409 and vv12526, into the pTRC99A-compatible expression vector, pQTπb. This placed expression of the antidotes under the control of the arabinose-inducible Pbad promoter (Guzman et al., 1995). E. coliπ1 cells that harboured both pTRC99A2::vv12410 and pQTπb::vv12409 or pTRC99A2::vv12525 and pQTπb::vv12526 formed colonies in the presence of IPTG when arabinose was also included in the media but no colonies were visible when arabinose was omitted from the media. These results demonstrated that Vv12409 and Vv12526 nullified the toxic effects of Vv12410 and Vv12525 respectively. Accordingly, we designated the vv12409/vv12410, vv12432/vv12433 and vv12546/vv12547 gene pairs relBE1, relBE2 and relBE3 respectively. The vv12525/vv12526 and vv12463/vv12464 gene pairs have been named parDE1 and higBA1 respectively.
The antidote specifically neutralizes the activity of the cognate toxin
We conducted cross-talk assays to determine the neutralizing activity of each antidote towards a non-cognate toxin. To do this, E. coli carrying either pQTπb::relB1 or pQTπb::parD1 were grown in media with arabinose to induce expression of the antidote. Each strain was then transformed with the pTRC99A2 derivatives that carried the non-cognate toxin, that is, pTRC99A2::relE1 was transformed into the E. coli/pQTπb::parD1 strain and pTRC99A2::parE1 was transformed into the E. coli/pQTπb::relB1 strain. Transformants were then cultured in media containing arabinose and IPTG to induce toxin expression. No bacterial growth was evident for either of the two non-cognate antidote/toxin pair combinations that we tested. These results indicated that each antidote specifically neutralized its cognate toxin.
The relBE1, parDE1 and intIAVvu are transcribed in vivo
Integron gene cassettes are unusually compact genetic elements. In general, the start codon of the ORF is situated very close to 5′ flanking attC site, leaving little intergenic space for regulatory sequences. As a consequence, most gene cassettes contain a translation initiation site (RBS) but lack identifiable promoter sequences (Collis and Hall, 1995). The sequence of the relBE and parDE gene cassettes from V. vulnificus 75.4T and CMCP6 revealed that in each case the intergenic region from the end of the 5′ flanking attC site to the start codon of the antitoxin was of sufficient length to accommodate a promoter. Upon closer examination, we were able to discern potential transcription initiation signals (−10 and −35 sequences) within this region (Fig. 3). A common feature of TA operons is their autoregulation due to the binding of the TA complexes to inverted repeats that overlap the −10 or −35 sequences (Roberts et al., 1993). Inverted repeats that overlapped the −10 and −35 sequences could be identified within each promoter region and may constitute binding sites for the respective TA complexes. These observations suggested that the relBE and parDE gene cassettes were expressed in vivo. RT-PCR was used to determine if relBE1 and parDE1 were transcribed in vivo. The primers used targeted the toxin gene of each TA system and the integron-integrase gene, intIAVvu (Table S1). The ribosomal rplT gene was used as a positive control. A product was not obtained when PCR was carried out on DNAse I-treated RNA, while RT-PCR on the same sample yielded a product of the expected size for the relBE1, parDE1, intIAVvu and the rplT genes (Fig. 4). These results indicated that both TA systems were being transcribed in vivo and this is the first demonstration that the integrase gene of a SI, intIAVvu, is expressed in vivo.
TA systems stabilize SIs
As the integrase gene was transcribed in vivo, its activity could lead to destabilization of the SI gene cassette array. pWEBTNC::VvuSI carried a 36 kb fragment of the V. vulnificus 75.4T SI. The annotated sequence of pWEBTNC::VvuSI did not reveal an operon characteristic of a TA system. To test the effect of IntI1 expression on the stability of pWEBTNC::VvuSI, we transformed the plasmid into strain 1811 to create strain 1822. IntI1-mediated recombination was induced with IPTG and contextual changes in pWEBTNC::VvuSI were monitored by Southern analysis using the V. vulnificus attC site as a probe (Fig. 5). While no change in the size of pWEBTNC::VvuSI was observed under non-inducing conditions in strain 1822, losses of up to 30 kb of DNA from pWEBTNC::VvuSI were observed after induction of intI1 expression. This indicated that SI gene cassette arrays were subject to large-scale, integrase-mediated deletion. Two of these SI gene cassettes of pWEBTNC::VvuSI were tagged with the relBE1 and parDE1 TA systems by cloning into unique NruI and StuI sites to create pWEBTNC::VvuSI/relBE1/parDE1. This plasmid was transformed into strain 1811 to create strain 1908 and the above integrase-mediated excision assay was repeated. Cassettes loss from pWEBTNC::VvuSI/relBE1/parDE1 following induction of intI1 expression was suppressed, indicating that the TA systems inhibited the large-scale deletion of gene cassettes by IntI1. To detect any SI cassette relocation event representative of integrase-mediated microevolution, we utilized a three-plasmid system to monitor cassette trafficking: the first plasmid, pTRC99Tcoriπ::intI1, was a pir-dependent, TcR derivative of the pTRC99A expression vector that carried the intI1 gene under the control of the inducible trc promoter; the second was pWEBTNC::VvuSI/relBE1/parDE1; the third was the conjugative plasmid R388, which harbours a naturally occurring class 1 target integron, In3 (Martinez and de la Cruz, 1988). In3 contains the dfrB2 gene cassette, which confers resistance to trimethoprim, and an ORF of unknown function, orfA. Both plasmids were introduced into E. coli strain π1. Following induction of intI1 gene expression, R388 was conjugated to the E. coli strain β2155 and transconjugants were selected on Luria–Bertani (LB) plates that contained diaminopimelic acid (DAP), kanamycin (Km) and chloramphenicol (Cm). Transconjugants were screened by allele-specific fusion primer PCR (Rowe-Magnus et al., 2002b) with primers attI1.1 and DFRFUS1. The attI1.1 primer annealed adjacent to the start ATG of the intI1 gene. DFRFUS1 was a unique fusion primer designed to hybridize specifically to the VVR–dfrB2 cassette junction that would arise from an IntI1-mediated recombination event of any V. vulnificus SI cassette at the attI1 site of plasmid R388. After cloning of the bulk PCR product, we selected and sequenced the inserts of eight random clones to confirm the attI site location of each cassette. We identified five different cassettes, all of which mapped to various positions in pWEBTNC::VvuSI/relBE1/parDE1. None of the five coded for a TA system. Thus, our detection of five independent recombination events of SI gene cassettes demonstrated that the presence of TA systems as gene cassettes within SIs prevented massive gene loss due to integrase-mediated recombination but did not preclude integrase-mediated microevolution from occurring.
TA systems stabilize genomic DNA
A 165 kb region of the E. coli K12 genome has been shown to be dispensable for growth under standard laboratory conditions (Fukiya et al., 2004). We inserted adjacent attC/AbR cassettes at four specific locations within this region by allelic exchange using the λ red recombination system to create strain 1523 (Fig. 6). Allelic exchange was performed such that all of the attC sites would be in direct orientation with respect to one another. In this configuration, recombination catalysed by IntI1 would lead to the deletion of the Cm, rifampicin (Rf) or spectinomycin (Sp) resistance markers that were linked to each of the attC sites. In parallel, the same four sites were targeted for allelic exchange using attC/AbR constructs that were linked to the relBE1 and parDE1 TA operons we identified (strain 1788). We then introduced the expression vector pTRC99Tc or pTRC99Tc::intI1 into strains 1523 and 1788 to create strains 1832, 1921, 1930 and 1931 (Table 1). Expression from the Ptrc promoter was induced with IPTG and bacteria were enumerated after spreading on plates containing the indicated antibiotics. Loss of the respective antibiotic resistance markers was confirmed by PCR with maker-specific primers on 10 random clones from each plate. PCR using the same primers on the parent strains 1523 and 1788 were used as controls (data not shown). Deletions of the Rf marker occurred at high frequency in strains 1832 and 1921 (Table 1). This suggested that intrinsic instability in the phoE region of E. coli K12 led to deletion of the Rf marker independent of IntI1 activity. The inclusion of the relBE1 locus markedly repressed deletion of the Rf marker in strains 1930 and 1931. This demonstrated that the relBE1 locus attenuated deletion of the flanking chromosomal DNA. Deletion of the Sp marker from strains 1832 and 1921 occurred at frequencies of 22% and 56% respectively. These frequencies were reduced to less than 7% in strains 1930 and 1931 when the parDE1 locus was present. These results indicated that the relBE1 and parDE1 TA systems served to minimize the frequency of deletion events that encompassed the respective system and, in doing so, stabilize long stretches of genomic DNA.
Table 1. IntI1-driven deletion frequencies of the Rf, Sp and Cm markers.
No. of resistant clones (% deletion frequency)a Strain
% deletion frequency = [1 − (#RfR, CmR or SpR clones/KmR clones)] × 100.
The yafQ/dinJ operon in the E. coli K12 genome encodes a relBE TA system
In doing the above experiments, we observed that deletion of the Cm marker from strains 1832 and 1921 was obtained at a lower frequency (12% and 15% respectively) than that of the Rf and Sp markers, even though it would have encompassed the smallest deletion possible of the three. As most of the clones remained CmR, this suggested that the 15 kb region between the Cm and Rf markers was refractory to deletion. We analysed this region and found two genes annotated as dinJ, a damage-inducible protein, and yafQ, a hypothetical ORF, that formed an operon of two small overlapping ORFs similar in organization to the parDE1 and relBE1 systems we identified. Furthermore, YafQ was homologous to RelE1 of V. cholerae (Fig. 2) and we support the proposal by Gerdes et al. (2005) to rename the E. coli yafQ/dinJ gene pair relBE-2. To determine if the yafQ/dinJ operon coded for a TA system, we cloned yafQ into pTRC99A and dinJ into pQTπ. We then transformed pTRC99A::yafQ into E. coliπ1 and transformants were grown under inducing and non-inducing conditions. Cells were readily recovered in the absence of IPTG while no cell growth occurred in the presence of IPTG. However, cells that harboured both pTRC99A::yafQ and pQTπ::dinJ grew in the presence of IPTG when arabinose was also included in the media but not if arabinose was omitted from the media. These results suggested that yafQ encoded a toxin and that dinJ encoded the cognate antitoxin. To determine if the yafQ/dinJ TA pair contributed to stabilizing the Cm marker, we cloned dinJ into pBAD24Gm and transformed this plasmid into strain 1523. Transformants were grown in the presence of 1% arabinose to induce the expression of dinJ. We then introduced pTRC99A::intI1 and clones were cultured in the presence of IPTG to induce IntI1-mediated recombination. Bacteria were enumerated after spreading on plates containing the indicated antibiotics. We did not observe a decrease in the number of CmR clones compared with strain 1523 carrying pBAD24Gm and pTRC99A::intI1 (data not shown). These results suggested that although the yafQ/dinJ operon coded for a TA system of the relBE family, this system did not contribute to stabilization of the proximal Cm marker. Presumably, under the experimental conditions tested here, other gene(s) in the vicinity were required for growth and stabilized the region.
Integrons are adaptable by design because each gene cassette is structured as an independently mobile genetic unit that is subject to episodic selection (Rowe-Magnus et al., 2002b). The vast majority of gene cassettes are promoterless (Recchia and Hall, 1995; Rowe-Magnus and Mazel, 1999) and, in the absence of selective pressure, perpetual integrase-mediated gene acquisition and loss (Collis and Hall, 1992a,b; Collis et al., 1993) would be expected to limit the number of cassettes accumulated in an integron. The impressive lengths that SIs can attain are thus surprising. The gene cache of a SI can equal 3% of the bacterial genome content (Heidelberg et al., 2000; Chen et al., 2003) and the sheer number of accumulated cassettes in large SIs implies either the existence of selective pressures that maintain these gene cassettes or mechanisms that promote their persistence in the absence of selection. In recent years, TA loci have been detected on the chromosomes of numerous free-living bacteria (Pandey and Gerdes, 2005). We have shown that two gene cassettes in the V. vulnificus SI encoded functional TA systems of the relBE and parDE families. Both systems were expressed in vivo, and the activity of the RelBE1 and ParDE1 proteins counterbalanced the extent of deletions catalysed by the class 1 integron integrase, IntI1. Furthermore, the presence of TA gene cassettes in SIs did not preclude integrase-mediated microevolution from occurring; en masse gene cassette loss was suppressed but the exchange of individual cassettes continued. The net effect was stabilization of SI gene cassette arrays in the absence of selection and this activity may in part explain why gene cassettes coding for TA systems are common in large SIs but are absent from smaller SIs. We propose the following model for the stabilization of SI gene cassette arrays by integron TA gene cassettes. TA loci characterized as plasmid stabilization agents ensure partitioning of the plasmid to bacterial progeny because loss of the plasmid, and more specifically the loss of antitoxin expression, causes toxin-dependent post-segregational killing of plasmid-free progeny. SI gene cassettes are integrated in and replicated with the bacterial chromosome. Integrase-mediated excision of an SI array segment encompassing a functioning integron TA gene cassette would be analogous to the post-segregational killing effect of addiction modules on extrachromosomal elements. The result would be cell death.
Previous studies of the five chromosomal TA loci of E. coli MG1655 demonstrated that deletion of the antitoxin partner alone was not possible. However, artificial deletion of the entire TA loci was possible (Christensen et al., 2004; Budde et al., 2006). As viable TA deletion mutants could be recovered, this suggested that chromosomal TA loci lacked the capacity to prevent the loss of flanking chromosomal DNA. A possible explanation for this discrepancy with our results may reside in the level of expression of the TA system in question. Cell death would be anticipated if the deleted TA system was highly expressed, whereas little or no effect would be expected if the deleted system was expressed at a low level. The relBE1 and parDE1 genes of the V. vulnificus SI were highly expressed and suppressed deletion of flanking DNA. Although we did not quantify the level of expression of the yafQ/dinJ loci, overexpression of the DinJ antidote did not result in an increased frequency of IntI1-mediated deletion of the flanking DNA. This would be expected if the YafQ toxin were expressed at a low level. We are currently investigating this hypothesis experimentally.
The neutralizing activity of each antitoxin was restricted to it cognate toxin, as cross-talk between the relBE1 and parDE1 systems was not observed. This has important implications with respect to the stability of SI cassette arrays. The lack of cross-talk between the TA systems in the SI should reduce the likelihood of the loss of DNA that is linked to non-homologous TA systems, because the deletion of a non-homologous TA system from the SI would result in the cessation of growth. But what are the stability dynamics when homologous TA systems are involved? The chpB locus is a chromosomal homologue of the parD locus of plasmid R1 (Bravo et al., 1987; 1988). Diaz and coworkers demonstrated that the antidote of the chromosomal chpB locus in E. coli could neutralize the plasmid-borne toxin of the parD system (Santos Sierra et al., 1998). These types of functional interactions between systems may play an important role in bacterial evolution by permitting the acquisition of plasmids bearing only the toxin partner of a TA operon or by reducing the genetic burden on the bacterium by permitting the loss of DNA that is linked to homologous addiction systems. However, complementation could only be attained by way of a promoter-up mutation that increased the expression level of the chpB system by six times or if the toxin component of the chpB locus was deleted. These results indicated that there was a correlation between the level of antidote expression and the degree of complementation, and that the cognate toxin can mask the basal ability of the antidote to neutralize the toxins of homologous systems. Three relBE homologues were identified in the SI of V. vulnificus CMCP6. Two of these, relBE1 and relBE2, were identical over their entire length, including their promoter regions, and likely represent a recent duplication event. Similar duplication events may be the source of the four HipA homologues in the T. denticola SI (Table S2). However, these loci are clearly not pseudogenes as none of ORFs for the toxins contained a premature stop codon and, as cell death could be prevented by mutational inactivation of the toxin component of a TA operon, the in vivo expression and activity of the toxins indicated that these loci were under selective pressure to maintain their functionality. The identity of the relBE1 and relBE2 gene cassettes necessitates that both operons be expressed to the same level. In this scenario, one locus cannot adequately compensate for the consequences of deleting the other if both are highly expressed because the ratio of antitoxin to toxin would be perturbed and result in the cessation of growth.
Clearly, gene acquisition and gene loss are important facets of bacterial evolution. We have shown that the selfish nature of TA systems can contribute positively to the fitness of bacteria by allowing potentially unstable chromosomal DNA regions to persist, and may in part explain their presence in SIs. In support of this notion, all of the TA loci identified in the V. cholerae genome are structured as gene cassettes in its SI (Rowe-Magnus et al., 2003; Pandey and Gerdes, 2005). The novel mechanism of integron gene cassette recombination, deciphered by Mazel and colleagues (Bouvier et al., 2005) and MacDonald et al. (2006), also promotes the maintenance of integron gene cassettes. Integrase-catalysed recombination occurs between the native double-stranded (ds) attI site and the bottom stand of the attC site, which initially is in single-stranded (ss) form. The ss attC site adopts a folded structure that reconstitutes a ds recombination site that is cleaved by the integrase. Recombination leads to insertion of the gene cassette in ss form. The missing strand must be subsequently synthesized, thus linking integrase-mediated recombination to DNA replication. If the mechanism for gene cassette deletion proceeds by the inverse reaction, then a maximum of half of the progeny would lose the gene cassette per round of deletion, while the other half would retain it. The end result would be the maintenance of any particular gene cassette to some degree within the population, regardless of the number of generations. Thus, both mechanisms contribute to the overall stability of SIs.
Media and growth conditions
Antibiotics were used at the following concentrations: ampicillin (Ap), 100 μg ml−1; Km, 25 μg ml−1; Cm, 25 μg ml−1; Sp, 100 μg ml−1; Rf, 50 μg ml−1 and tetracycline (Tc), 15 μg ml−1 in LB. Trimethoprim (Tp) was used at 87 μg ml−1 in Mueller–Hinton media. IPTG, DAP and l-arabinose were supplemented when necessary to final concentrations of 1 mM, 0.3 mM and 1% respectively. All bacterial strains and plasmids are listed in Tables 2 and 3 respectively.
Table 2. Plasmids used and constructed in this study.
The genomes of microbial species that were identified as containing an integron-integrase gene were downloaded and searched with toxin and antitoxin protein query sequences belonging to the seven known TA families. Each homologue was carefully checked for the presence of an associated attC site characteristic of an integron gene cassette. In some cases, we could identify a closely linked, annotated antitoxin partner. If an annotated partner was not apparent, we looked for un-annotated ORFs in the adjacent DNA. The cut-off E-value used in this analysis was 1e-4.
PCR, cloning, sequencing and phylogenetic analysis
Genomic DNA was isolated by using the Invitrogen DNAzol kit. PCR was conducted as follows: 94°C, 30 s; 60°C, 30 s; 72°C, 2 min, for 30 cycles with Pfu polymerase. PCR-amplified genes were cloned by using the TA-TOPO cloning kit (Invitrogen) and sequenced by The Centre for Applied Genomics (TCAG) at The Hospital for Sick Children (Toronto, Canada). Primers were obtained from Operon Biotechnologies (Huntsville, AL). Protein alignments were generated with the clustalx version 1.8 (Thomson et al., 1997) and dendrograms were compiled by using the neighbour-joining method (computed from 1000 independent trials) of clustalx and treeview (Page, 1996). Amino and nucleic acid sequences were retrieved from GenBank.
Construction of base cloning vectors
Plasmid pUC18Asc2Not was created from pUC18 by inverse PCR (IPCR) with primers pUC18 NotAsc-1 and pUC18 NotAsc-2. The multiple cloning site of this plasmid has been replaced by a NotI site that is flanked by two AscI sites. We used primers VVR1Mfe and VVR2Mfe to amplify a SI attC site from V. vulnificus genomic DNA. The PCR product was treated with MfeI and cloned into the EcoRI site of pNOT218 (Matsumoto-Mashimo et al., 2004) to create pNOT218::VVR.
Construction of base expression vectors
The expression vector pTRC99A was digested with NdeI, treated with Klenow polymerase (New England Biolabs) to fill in the site, self-ligated and transformed into DH5α. The NcoI site of the resulting plasmid, pTRC99AΔNde, was replaced with an NdeI site by IPCR with primers pTRC99AΔNco-1 and pTRC99AΔNco-2. The PCR product was cut with EcoRI, self-ligated and transformed into DH5α to create pTRC99A2. pTRC99Tcoriπ was created as follows: to replace the bla gene of pTRC99A with the tet gene of pBR322, IPCR with the primers pTRC99aoriπΔAp-1 and pTRC99aoriπΔAp-2 was used to amplify the pTRC99 backbone. Primers pBR322tetNot1.1 and pBR322tetNot2 were used to amplify the tet gene from pBR322. Both products were cut with NotI and ligated together. Plasmid DNA was isolated from DH5α transformants and used as the matrix in a PCR with the pTRC99ΔcolE1Asc-1 and pTRC99ΔcolE1Asc-2 primers in a second IPCR. The oriVR6Kγ origin of replication was amplified from pSW23 with the primers ori pir Asc-1 and ori pir Asc-2. Both products were cut with AscI, ligated and transformed into E. coli strain π1. The intI1 gene was amplified from pTRC99A::intI1 (Rowe-Magnus et al., 2002b) with the primers intI1BspHI and intI1Xba, cut with the corresponding enzymes and cloned into pTRC99TcoriVR6Kγ that had been digested with NcoI and XbaI.
To construct an expression plasmid that was compatible with pTRC99A, the Pbad promoter, araC gene and transcriptional terminator of pBAD24 were amplified with primers PbadaraCAscI and rrnBT1T2AscI. IPCR was performed on pSW23 with the pSW23AscI-1/pSW23AscI-2 primer pair to delete the multiple cloning site. The PCR products were cut with AscI, ligated together and transformed into strain π1. The copy number of the resulting plasmid, pQTπb, can be maintained at low or high copy number by transforming it into a pir or pir116 background (Metcalf et al., 1994; 1996).
The plasmid pBADGm was constructed as follows. IPCR was performed using primers pTRC99AΔApAsc1 and pTRC99AΔApAsc2 with pBAD24 as the template. The GmR marker was amplified from pTX1G with primers pTX1GAscXho and pTX1GPflMPmeAsc. Both PCR products were digested with AscI and ligated together.
Cloning and expression of the toxin and antitoxin genes
The relBE1 and parDE1 T/A pairs were cloned and expressed as follows: the relE1 and parE1 and yafQ toxins were amplified with the VV12410Nde/VV12410Xba, VV12525Nde/VV12525Xba and yafQERI/yafQXba primer pairs respectively. The PCR products were treated with the NdeI or EcoRI and XbaI and cloned into the same sites of pTRC99A2 to create pTRC99A2::vv12410, pTRC99A2::vv12525 and pTRC99A::yafQ. The relB1, parD1 and dinJ antitoxins were amplified with the VV12409ERI/VV12409Xba, VV12526ERI/VV12526Xba and dinJERI/dinJXba primer pairs, the PCR products were treated with the EcoRI and XbaI and cloned into the corresponding sites of pQTπb to create pQTπb::vv12409, pQTπb::vv12526 and pQTπb::dinJ. The dinJ fragment was also cloned into the corresponding sites of pBADGm.
Promoter-gene fusions using super-primer PCR
To ensure adequate expression of single copy genes that we integrated into the chromosome of E. coli K12, we opted to fuse each with a consensus σ70 promoter. We chose the Ptrc promoter because the LacI repressor and IPTG can be used to regulate its expression. The primers pTRC99AH3 and pTRC99AATG were used to amplify the region from −40 to the start ATG of the Ptrc promoter. The primer pairs pTRC99Acat/pTX1CcatH3, pTRC99AKm/pTX1KKmH3, pTRC99Aarr2/pCTF104arr2H3 and pTRC99AaadA7/pSW25aadA7H3 were used to amplify the cat, aph, aar2 and aadA7 genes from plasmids pTX1C, pTX1K, pCTF104 and pSW25 respectively (Table 2). Twenty nanograms of the pTRC99AH3/pTRC99AATG PCR fragment was mixed with an equal quantity of the cat, aph, aar2 and aadA7 PCR products in separate reactions that contained all the necessary PCR reagents but no primers. Five PCR extension cycles were run as follows: 94°C for 60 s; 72°C for 60 s. The pTRC99AH3 primer was then added along with the pTX1CcatH3, pTX1KKmH3, pCTF104arr2H3 or pSW25aadA7H3 primer to the appropriate tubes and 25 more cycles of 94°C for 30 s; 60°C for 30 s; 72°C for 90 s were completed to fuse the Ptrc promoter to each of the markers. The cat, aph, arr2 and aadA7 promoter fusions were then cloned into the SmaI site of pNOT218::VVR. To ensure expression of the relBE1 and parDE1 TA cassettes, we created similar Ptrc promoter fusions to these loci. The primers pTRC99APst and pTRC99AATG were used to amplify the region from −40 to the start ATG of the Ptrc promoter. Using V. vulnificus genomic DNA as matrix, we amplified the relBE1 and parDE1 TA loci with the primer pairs pTRC99Avv12409/vv12410Xba and pTRC99Avv12526/vv12525Xba. These fragments were fused with the PstI-containing Ptrc promoter fragment as described above. The amplicons were cut with PstI and XbaI and cloned into the corresponding sites of pNOT218::VVR-arr2 and pNOT218::VVR-aadA7 to create pNOT218::VVR-arr2-relBE1 and pNOT218::VVR-aadA7-parDE1 respectively.
Constructing strains 1523, 1788, 1832, 1921, 1930 and 1931
Four regions of the E. coli K12 genome were targeted for recombination. These regions surrounded the yagI, phoE, yafK and yaiD genes. Primers yagHNot and argFNot, crlNot and proBNot, yafJNot and yafLNot, and cds18Not and yajFNot respectively, were used to amplify each target gene along with 400 bp flanking regions on both sides. The PCR products were cut with NotI and cloned into the NotI site of pNOT218. The yagI, phoE, yafK and yaiD genes were deleted from these plasmids by IPCR with the primer pairs ΔyagI1Asc and ΔyagI2Asc, ΔphoE1Asc and ΔphoE2Asc, ΔyafK1Asc and ΔyafK2Asc, and ΔyaiD1Asc and ΔyaiD2Asc respectively. PCR products were cut with AscI and ligated to the AscI-digested PCR products of pNOT218::VVR-cat, pNOT218::VVR-aph, pNOT218::VVR-arr2, pNOT218::VVR-aadA7, pNOT218::VVR-arr2-relBE1 and pNOT218::VVR-aadA7-parDE1 following their amplification with the primers M13FAsc and M13RAsc. The resulting plasmids were designated pNOT218::yafJ-yafL::VVR-cat, pNOT218::cds18-yajF::VVR-aph, pNOT218::crl-proB::VVR-arr2 and pNOT218::yagH-argF::VVR-aadA7, pNOT218::crl-proB::VVR-arr2-relBE1 and pNOT218::yagH-argF::VVR-aadA-parDE1.
Escherichia coli K12 was transformed with pKOBEGA and cells were made electrocompetent following arabinose induction as described (Chaveroche et al., 2000). The insert of pNOT218::yafJ-yafL::VVR-cat was amplified by PCR with the M13FAsc and M13RAsc primers and electroporated into the above cells. Transformants were selected on LB Cm plates at 30°C. This strain was in turn sequentially transformed with the amplicons from pNOT218::cds18-yajF::VVR-aph, pNOT218::crl-proB::VVR-arr2 and pNOT218::yagH-argF::VVR-aadA7 to create strain 1523. The correctly targeted recombinants were verified by PCR after each round of selection using the VVR2cStu primer in combination with the yafJ, cds18, yafA or yagH recomb primers. To create strain 1788, the same procedure was followed with inserts that were amplified from pNOT218::yafJ-yafL::VVR-cat, pNOT218::cds18-yajF::VVR-aph, pNOT218::crl-proB::VVR-arr2-relBE1 and pNOT218::yagH-argF::VVR-aadA-parDE1.
Plasmids pTRC99Tc and pTRC99Tc::intI1 were electroporated into strains 1523 and 1788 to create strains 1832, 1921, 1930 and 1931 (Table 3). These strains were grown in LB media containing Km and Tc to an OD600 of 1.0, at which point IPTG was added to induce integrase-mediated recombination. Incubation of the bacteria was continued overnight with shaking at 37°C. The following day, a 1 ml aliquot of bacteria from each sample was diluted in LB to an OD600 of 0.6. Serial dilutions of 105−107 were prepared in LB from these normalized samples and 100 μl of each was spread on plates containing the indicated antibiotics. Bacteria were enumerated after incubation of the plates at 37°C for 16 h. Loss of the respective antibiotic resistance markers was confirmed by PCR on 10 clones chosen at random from each plate with the marker-specific pTRC99Acat/pTX1CcatH3, pTRC99AKm/pTX1KKmH3, pTRC99Aarr2/pCTF104arr2H3 and pTRC99AaadA7/pSW25aadA7H3 primer pairs. PCR using the same primers on the parent strains 1523 and 1788 were used as controls.
Constructing a V. vulnificus cosmid library
A cosmid library of V. vulnificus genomic DNA was constructed using the pWEBTNC cosmid cloning kit (Epicentre). Several clones were positive when screened by PCR with the primers VVR1Mfe and VVR2Mfe primers (Table S1). One of these, pWEBTNC::VvuSI, was found to contain an insert of 36 kb and was sequenced.
Constructing of pWEBTNC::VvuSI::relBE1/parDE1
The plasmid contained unique NruI and StuI sites at the 16 kb and 25 kb marks. These site were located within the ORFs of two gene cassettes and were used to tag the cassettes with the arr2-relBE1 and aadA7-parDE1 constructs as follows: the primer VVR2cStu was used in combination with vv12410Stu or vv12525Stu to amplify the arr2-relBE1 and aadA7-parDE1 constructs from pNOT218::VVR-arr2-relBE1 and pNOT218::VVR-aadA7-parDE1 respectively. The PCR products were digested with StuI and the arr2-relBE1 and aadA7-parDE1 constructs were cloned into the NruI and StuI sites, respectively, of pWEBTNC::VvuSI to create pWEBTNC::VvuSI::relBE1/parDE1.
Allele-specific fusion primer PCR
Strain 1953 was streaked out on MH dT Tp Km Ap plates and grown overnight. A single colony was inoculated into 5 ml of the equivalent liquid MH media, without Tp, and grown to an OD600 of 0.8. IPTG was added to induce expression of IntI1 and the sample was incubated overnight. The following day, plasmid DNA was isolated from a 1 ml aliquot of cells and used as matrix in a PCR with primers attI1.1 and DFRFUS1. DFRFUS1 is a unique fusion primer designed to hybridize specifically to the VVR–dfrB2 cassette junction that would arise from an IntI1-mediated recombination event of any V. vulnificus SI cassette at the attI1 site of plasmid R388. The PCR was conducted with Pfu polymerase as described earlier. The PCR products were cloned using the TA-TOPO cloning kit (Invitrogen). Eight colonies were chosen at random for sequence analysis to confirm the attI site location of each cassette. None of the primers alone produced a PCR product nor was a product obtained for non-induced cultures. The amplified fragments were sequenced directly at TCAG.
RNA isolation and RT-PCR
RNA was isolated using the QIAGEN RNeasy® Mini Kit. Briefly, an overnight culture of bacteria was diluted 1/100 into LB. The cultures were grown at 37°C with aeration to an OD600 of 0.5. Cells were pelleted by centrifuging 1 ml of culture at 5000 g for 5 min at 4°C. Cells were resuspended and lysed for 15 min in 100 μl of TE buffer containing 400 μg ml−1 of lysozyme (Sigma). After addition of ethanol, samples were applied to the RNeasy® column and treated with DNase using the RNase-Free DNase Set (QIAGEN, Mississauga, ON) for 1 h. The column was washed and RNA was eluted using RNase-free water. PCR was performed to verify that DNA was not present in the sample. Total RNA was quantified using a Biomate 3 spectrophotometer (Thermo Spectronic). RNA was divided into 5 μl aliquots and stored at −70°C.
The QIAGEN OneStep RT-PCR Kit was used to amplify regions of vvuintIA, relE1, parE1 and rplT mRNA, according to the manufacturer's protocol. The primer pairs used were VvuintIART1/VvuintIART2, VV12410RT1/VV12410RT2, VV12525RT1/VV12525RT2 and VvuL20RT1/VvuL20RT2 respectively. RT-PCR was carried out using a MBS Multiblock Thermocycler System (Thermo Hybaid). Reverse transcription reactions conditions were at 50°C for 30 min. cDNA was amplified using an initial hot start at 95°C for 15 min followed by 30 cycles of denaturation at 94°C for 1 min, annealing at 55°C for 1 min and elongation at 72°C for 1 min. PCR products were visualized by agarose gel electrophoresis.
Plasmid DNA was recovered from samples using the Gene Elute Miniprep Kit (Sigma) according to the manufacturer's instructions. Plasmid DNA was digested with NotI and separated on a 1% agarose gel using a Chef-DR III Pulse Field Gel Electrophoresis system (Bio-Rad). The DNA was transferred to a Hybond-N+ membrane (Amersham Biosciences). The 121 bp signature attC sites of the V. vulnificus SI [formerly referred to as Vibriovulnificusrepeats (VVRs)] were amplified with the primers VVR5 and VVR6 (Table S1). The PCR product was labelled using the AlkPhos Direct labelling Kit (Amersham Biosciences) and was used as a probe to detect DNA fragments containing attC sites by Southern analysis.
Deposition of nucleic acid sequences and accession numbers
All nucleic acid sequences have been deposited at the National Centre for Biotechnology Information GenBank Database under the following accession numbers: EF213116 to EF213121.
This work was supported by funding from the Sunnybrook Research Institute and the Canadian Institutes of Health Research to Dean Rowe-Magnus, and the Institut Pasteur, the CNRS, and the EU (STREP CRAB, LSHM-CT-2005-019023, NoE EuroPathoGenomics, LSHB-CT-2005-512061) to Didier Mazel.