Analysis of RecA-independent recombination events between short direct repeats related to a genomic island and to a plasmid in Escherichia coli K12

Bacterial strains, plasmids, and growth conditions

E. coli K12 BZB1011 is a spontaneous nalidixic acid resistant derivative of W3110, a strain whose genome has been sequenced (GenBank: AP_009048). Most strains used are BZB1011 derivatives, carrying deletions proceeding from selected mutants of the Keio Collection (Baba et al., 2006). In these latter, gene deletions with replacement by a kanamycin resistance (kan) cassette were introduced into BZB1011 by P1 transduction, selecting the kanamycin resistant clones (Table S1). When required, double deletion mutants were constructed in two steps: the kanamycin cassette of the first mutated gene was eliminated as indicated previously (Datsenko & Wanner, 2000) and the cured mutant was then transduced with a P1 grown on the second deletion mutant, selecting the kanamycin resistant clones. Multiple deletion mutants were constructed with further rounds of the same procedure. E. coli K12 JC8679 is a recB recC sbcA mutant (Coli Genetic Stock Center strain 6490). Two mutant derivatives of BW25113, the parent E. coli K12 strain for the Keio Collection, were used: BW25113 recA56 and BW25113 topB recA56, this latter bearing the ΔtopB761::kan mutation from the Keio Collection. The recA56 allele was introduced by transducing BW25113 and BW25113 topB (JW1752) with P1 grown on a srl:: Tn10 recA56 strain from our laboratory collection, selecting the tetracycline resistant clones and then assaying the recA defficiency by testing their UV sensitivity.

Plasmid pEX2000 is a pBR322 derivative that carries the H47 GI included in a 16,823 bp-chromosomal DNA segment from strain H47 (GenBank: AJ_009631). Plasmid pΔint1 carries the same DNA segment but deleted for most of the H47 GI, keeping the attachment sites and the external chromosomal adjacent sequences (Azpiroz, Bascuas & Laviña, 2011). pUY-FRT was constructed by cloning a 1,182 bp HindIII fragment from pKD3 (GenBank: AY_048742) into the pUC13 vector. The insert contains the chloramphenicol resistance gene (cat) flanked by the FRT sites from S. cereviseae. Since pKD3 requires the pir gene product to be provided in trans for its replication (Datsenko & Wanner, 2000), pUY-FRT was constructed to avoid this requirement.

Bacteria were grown in LB medium at 37 °C. Antibiotics were added at the following final concentrations: kanamycin, 30 µg/ml; ampicillin, 50 µg/ml, chloramphenicol, 60 µg/ml, and tetracycline, 20 µg/ml.

Plasmid purification, PCR assays and DNA sequencing

Plasmid DNA was extracted with “QIAprep Spin Miniprep Kit” (Qiagen, Hilden, Germany). DNA bands were extracted from gels using “QIAEXII Gel extraction kit” (Qiagen, Hilden, Germany). For PCR reactions, primers and conditions employed are explained in Table 1. For sequencing purposes, PCR products were extracted from gels as explained. Both strands of the OUT and IN amplicons from the pUY-FRT model were sequenced using the primers employed for their amplification. DNA sequencing was performed at the “Molecular Biology Unit” of the Pasteur Institute of Montevideo.

Excision of the H47 genomic island

Recombination underlying the H47 GI mobility was studied through the detection of the island excision from a multi-copy recombinant plasmid, pEX2000, in the context of E. coli K12 cells. Plasmid pEX2000 carries a DNA segment from the E. coli H47 chromosome containing the H47 GI flanked by its direct repeats and by adjacent chromosomal sequences on both sides. The direct repeats, attL (148 bp) and attR (144 bp), are imperfect and share four main segments of homology of 13, 11, 20 and 17 bp (Fig. 1B). Recombination between them leads to the excision of the H47 GI, which is assessed by PCR using primers “out” in order to detect the occurrence of the empty attC site (amplicon OUT_H47) (Fig. 1A). The assays were performed in the context of the E. coli K12 strain BZB1011 and in a set of BZB1011 derivative mutants affected in recombination functions (Table S1). The purpose was to identify the mutant/s where excision did not occur, thus revealing the gene/s involved in the recombination process. Plasmid pEX2000 was introduced into each of these strains and then was extracted to be used as template in PCR reactions. The sequence of the OUT amplicons corresponding to the mutant strains was compared with that from the wild-type BZB1011 strain, i.e., with a mix of four overlapping sequences with the predominance of that corresponding to recombination in site III (Fig. S1).

Considering that genomic islands usually employ site-specific recombinases to mediate their mobility, the search first concentrated in mutant strains deficient for such enzymes. Although the H47 GI attachment sites did not reveal significant homology with any known target for a site-specific recombinase, we did not discard this possibility since it has been reported that some of these enzymes are able to work although with low efficiency—on secondary sites that differ in their sequence from the primary one (Menard & Grossman, 2013). Therefore, the excision of the H47 GI was analyzed in mutants deficient for each of the 16 site-specific recombinases encoded by the E. coli K12 chromosome, most of them related to prophages (intF, intD, ybcK, intE, pin, intR, pinR, pinQ, intQ, intS, intA, xerD, xerC, intB, fimB and fimE). Mainly based on homologies among these enzymes, double and multiple mutants were constructed and included in the assays (Table S1). In all cases the OUT_H47 amplicon appeared and always revealed the same pattern of overlapping sequences as that proceeding from the wild-type BZB1011 strain (Fig. S1).

Given the previous results, the involvement of site-specific recombination in the H47 GI mobility was discarded. An important fact was that recombination always took place in regions of total homology between the attachment sites, indicating that this condition would be necessary. Therefore, we examined the possible involvement of several genes that participate in homologous recombination.

First, the RecA-independent RecET homologous recombination pathway, encoded by the defective Rac prophage present in E. coli K12, was analyzed (Kolodner, Hall & Luisi-DeLuca, 1994). It was considered that the recET genes might have a basal level of expression, as has been proposed for E. coli K12 crytic prophage genes (Wang et al., 2010). The excision of the H47 GI was assayed in BZB1011 mutant derivatives recE (deficient for exonuclease VIII) and recT (deficient for the recombinase function). None of these contexts affected the recombination pattern. The same result appeared using the recB recC sbcA strain JC8679, a genetic background where the RecET pathway is known to be induced (Kolodner, Hall & Luisi-DeLuca, 1994).

It was also considered that recombination under study, although RecA-independent, could share some functions with the homologous recombination pathways. Following this idea, the excision of the H47 GI was surveyed in a set of mutants defective for proteins that participate in homologous recombination (recB, recC, recD, recF, recO, recR, recN, recJ, recQ, recG, sbcB, sbcC, sbcD, uvrD, ruvA, ruvB, ruvC, radA, recX, exoX, rarA, seqA and helD). The double mutant recT recA, deficient for all known homologous recombination pathways, was constructed and also included in the assays. The result was that in neither mutant the excision of the island was detectably affected: the OUT_H47 product always appeared as a weak band in gels and its sequence exhibited the same pattern as that of the amplicon from the wild-type strain (Fig. S1). Therefore, none of the functions analyzed related to homologous recombination appeared to be involved in the H47 GI mobility.

In vivo search for the H47 GI excision

Although in vitro PCR experiments consistently supported the existence of recA-independent recombination events between the attL and attR sites flanking the H47 GI, this phenomenon still awaited an in vivo confirmation. This would imply the detection and characterization of at least a clone carrying the replicative recombination product, i.e., the deleted molecule. Obviously, the in vivo detection of the other putative product, the excised circular form, would be much less likely because of the non-replicative condition of this molecule. It should be noted that the model of the H47 genomic island was rather complex since it encodes the production of an antibiotic activity, which would exert a deleterious effect upon cells that had lost the island in a recombination event. For this reason, we investigated this phenomenon not only in strains carrying the entire island (pEX2000) but also in strains with a plasmid derivative in which most of the intervening sequence between the att sites had been deleted (pΔint1).

After being propagated in E. coli K12 cells (BZB1011 recT recA), plasmids were extracted, and important amounts of DNA (1–2 µg) were run in gels in an attempt to perceive new bands that could correspond to the recombination products. We did not succeed to see anything other than the plasmids’ original forms.

It has been described that E. coli mutants affected in the topB gene, encoding topoisomerase III, have an increased frequency of spontaneous deletions occurring at short direct repeats (Whoriskey, Schofield & Miller, 1991; Schofield et al., 1992; Uematsu, Eda & Yamamoto, 1997). Therefore, we repeated the assays in the context of strains carrying a topB deficient allele. BZB1011 topB was constructed, but its recA derivative resulted nonviable. Considering that the topB context would be affected in DNA supercoiling, the nalidixic acid resistant strain BZB1011 might not be suitable for this study since it most probably contains a mutant DNA gyrase. For this reason, the gyr wild type BW25113 context was employed. BW25113 recA and BW25113topB recA were constructed and transformed with pEX2000 and pΔint1. In all these strains, although recombination between the att H47 sites was PCR-confirmed, no extra-bands appeared when plasmids DNA was analyzed in gels.

H47 GI-related attC sequence survey in data banks

Since no product of the H47 GI excision could be detected in vivo, a clue in this sense was searched for in data banks. This type of survey had been done before, mainly focusing on the presence of the H47 GI in other strains besides E. coli H47 (Azpiroz, Bascuas & Laviña, 2011). Now, we looked for the four types of attC sequences (I–IV), corresponding to recombination events in each of the four segments of identity between the attL and attR sites. The search revealed that these sequences are widespread among pathogenic E. coli strains and also appear in two strains of Salmonella enterica, always being present in the chromosome. Although these sequences exhibited a certain degree of variability, several attC sites could be recognized and distinguished from attL and attR. They were further confirmed as attC by analyzing their adjacent sequences (about 500 nucleotides on each side), which should correspond to those surrounding the H47 GI. Most of them were attC_III, resulting from recombination at the most extensive site of homology between the direct repeats. Among these, six identical matches were found (E. coli strains Sanji, SEC470, RS76, PCN033, Santai O157:H16, and S. enterica Heidelberg str. SL476). There was a match with attC_II (E. coli ETEC H10407) and another single match with attC_IV (E. coli FHI23). No sequence of the attC_I type appeared. Therefore, these findings strongly supported the idea that the H47 GI is indeed a mobile element due to recombination between its attachment sites.

RecA-independent recombination in the model of plasmid pUY-FRT

In view that our efforts to detect in vivo the H47 GI excision were unsuccessful, the studies were extended to a simpler genetic model. It was a multi-copy plasmid, pUY-FRT, carrying the cat gene (for chloramphenicol resistance) flanked by a perfect 46 bp-direct repeat. The repeated sequences proceeded from S. cereviseae and contained the FRT target for the site-specific recombinase Flp. Recombination between the repeats would lead to the deletion of the intervening cat-containing DNA segment. The assays were carried out in the absence of the Flp enzyme in E. coli K12 cells. Specifically, plasmid pUY-FRT was propagated and then extracted from BZB1011 recT recA, BW25113 recA and BW25113 topB recA.

The experiments began by transforming with pUY-FRT the three E. coli K12 mutant backgrounds and chloramphenicol resistant transformants were selected so as to ensure that the assays started with clones carrying the original non-deleted form of the plasmid. After isolation in the same medium, clones were grown in liquid LB Ap, which was supposed to be a permissive condition for the propagation of the deleted plasmid.

First, plasmid DNA was used as template in PCR reactions devoted to detect the excision of the DNA segment between the repeats, i.e., the cat gene. As with the H47 GI, two types of reactions were carried out to detect the two possible products: an OUT_FRT amplicon of 381 bp in the deleted plasmid, and an IN_FRT amplicon of 335 bp in the excised circular form. In both types of reactions, amplicons of the expected size were produced and appeared in the three contexts assayed. The DNA sequence of these PCR products exhibited the expected recombination pattern, i.e., a single repeat flanked by recombined sequences.

For in vivo analysis, ca. 500 ng of plasmid DNA were loaded in gels. This time, a new small band was barely seen in plasmid preparations proceeding from the topB⁺ backgrounds, being more evident in the BW25113 recA context, and this same band appeared clearly stronger when the plasmid came from the topB deficient cells. We presumed that it could correspond to the deleted plasmid molecule, which would be called pUY-FRTΔ (Fig. 2A). Then, this smaller band, proceeding from BW25113 recA and BW25113 topB recA, was extracted from gels and used to transform BW25113 recA with selection for ampicillin-resistance. In each case, a few clones grew and, when assayed in the presence of chloramphenicol, all proved to be sensitive. Plasmid DNA was extracted from a clone of each type of transformant and was then digested with EcoRI and HindIII. Both plasmids were found to be identical in their size and restriction profile, which corresponded to those of a deleted plasmid derived from pUY-FRT by recombination between its direct repeats (Fig. 2B). DNA sequencing confirmed this structure: the sequence was read through the recombined repeat and beyond it more than 300 bp on each side.

Finally, it should be mentioned that no trace of the excised DNA could be seen.