Natural Transformation Facilitates Transfer of Transposons, Integrons and Gene Cassettes between Bacterial Species

We have investigated to what extent natural transformation acting on free DNA substrates can facilitate transfer of mobile elements including transposons, integrons and/or gene cassettes between bacterial species. Naturally transformable cells of Acinetobacter baylyi were exposed to DNA from integron-carrying strains of the genera Acinetobacter, Citrobacter, Enterobacter, Escherichia, Pseudomonas, and Salmonella to determine the nature and frequency of transfer. Exposure to the various DNA sources resulted in acquisition of antibiotic resistance traits as well as entire integrons and transposons, over a 24 h exposure period. DNA incorporation was not solely dependent on integrase functions or the genetic relatedness between species. DNA sequence analyses revealed that several mechanisms facilitated stable integration in the recipient genome depending on the nature of the donor DNA; homologous or heterologous recombination and various types of transposition (Tn21-like and IS26-like). Both donor strains and transformed isolates were extensively characterized by antimicrobial susceptibility testing, integron- and cassette-specific PCRs, DNA sequencing, pulsed field gel electrophoreses (PFGE), Southern blot hybridizations, and by re-transformation assays. Two transformant strains were also genome-sequenced. Our data demonstrate that natural transformation facilitates interspecies transfer of genetic elements, suggesting that the transient presence of DNA in the cytoplasm may be sufficient for genomic integration to occur. Our study provides a plausible explanation for why sequence-conserved transposons, IS elements and integrons can be found disseminated among bacterial species. Moreover, natural transformation of integron harboring populations of competent bacteria revealed that interspecies exchange of gene cassettes can be highly efficient, and independent on genetic relatedness between donor and recipient. In conclusion, natural transformation provides a much broader capacity for horizontal acquisitions of genetic elements and hence, resistance traits from divergent species than previously assumed.


Introduction
The acquisition and dissemination of antibiotic resistance in Gram-negative bacteria is frequently facilitated by integrons [1][2][3]. Integrons contain genetic determinants for site-specific recombination and promoters driving the expression of gene cassettes [4]; the integrase (IntI) encoded by the integron facilitates site-specific acquisitions and excisions of gene cassettes within the integron [5][6][7][8]. The gene cassettes often encode antibiotic resistance, however cassettes conferring other metabolic functions to bacteria have also been described [9][10][11]. Integrons and integrases have been found to be present in approximately 9 to 17% of the sequenced bacterial genomes [1,12]. Class 1 integrons are the most widely disseminated type in commensals and pathogens of human and animal origins [13][14][15][16][17][18], and have also been found in soil and in aquatic ecosystems [19][20][21][22][23]. This class of integrons is characterized by two conserved regions, the 59 conserved segment (59-CS), which includes the integrase gene (intI1), the adjacent recombination site (attI1) and the promoter (Pc), and the 39 conserved segment , which contains the qacED1 gene (encoding an incomplete version of a quaternary ammonium compound resistance), the sulI (encoding resistance to sulfonamides) and the orf5 [2,24].
Highly similar class 1 integrons have been found in both Gramnegative and -positive bacteria [25,26] and integrons with the same composition and organization have been found in unrelated bacterial species and strains in geographically distinct areas [27][28][29][30]. The exceptional broad potential of integrons to disseminate among pathogenic bacterial strains and species is remarkable because they only harbor functions for genomic integration and excision and do not encode functions that enable horizontal transfer between bacterial cells. Class 1 integrons are often present in plasmids, transposons and insertion sequences [12,31,32], and their dissemination is considered to depend on horizontal mobility of the genetic element or the genetic region they reside drawing on both transduction and conjugational processes [33][34][35]. However, horizontal movement of class 1 integrons genetically linked to nonconjugative elements [36] or incomplete mobile genetic elements and transposons [20,32,37] remains to be explained. Only few studies have attempted to experimentally examine how (nonconjugative) transposons, integrons or gene cassettes move horizontally. These include reports on the recruitment of gene cassettes [38], the acquisition and integration of synthetic gene cassettes by natural transformation [39] and transposition of the integron In33 [40,41]. Horizontal transfer of non-conjugative transposons seems to rely on linkage to conjugative elements or on transduction by bacteriophages [42,43].
Here we have investigated the potential for transposons, integrons and gene cassettes, supplied as fragmented DNA substrates, to move horizontally between bacterial species via natural transformation. Our model system relied on exposing the transposon-and integron-free and naturally-transformable bacterium Acinetobacter baylyi [44][45][46] to purified DNA or cell lysates obtained from the integron-carrying Gram-negative bacteria: A. baumannii, Citrobacter freundii, Enterobacter cloacae, Escherichia coli, Escherichia fergusonii, Pseudomonas aeruginosa, Salmonella enterica serovar Rissen and serovar Typhimurium. The exposure of naturally competent A. baylyi cells to DNA from these sources led to the acquisition of novel resistance traits as well as entire integrons and transposons. Transposition-based integration occurred between unrelated hosts, whereas both transposition and homologous recombination facilitated acquisitions from related host species.
Both the donor strains and the transformant isolates were characterized by antibiotic resistance profiling, targeted PCR, DNA sequencing by extensive primer walking, genome sequencing of two transformants, pulse-field gel-electrophoresis (PFGE) and Southern blot hybridization. A pairwise growth competition assay was undertaken to determine the impact of the acquired integrons on relative fitness. The integron-carrying transformants of A. baylyi were also used in subsequent transformation assays to confirm the nature of the initial interspecies gene transfer and determine further intraspecies transfer frequencies.
Some of the transformants of A. baylyi that were confirmed to have taken up the integron from the heterologous donor sources were also used as a source of donor DNA in subsequent transformation experiments (i.e. isolates SD1, SD2, SD3, SD4, SD5 and SD6; see Table 1). One of the integron-carrying transformants of A. baylyi (isolate SD2) was also used as a recipient bacterium in subsequent transformation assays (see Table 1).
Other derivatives of the A. baylyi BD413 strain employed as recipients (Table 1) were SD9 (a DrecBCD DsbcCD double mutant constructed as described by Harms and Wackernagel [49]) and the SD2 derivatives KOI (intI1::cat) and RAM (recA::cat) constructed as follows: an internal segment (724 bp) from the intI1 gene of A. baumannii 064 was PCR-amplified using the primers intI1-f (59-AGCTTACGAA CCGAACAGGC-39) and INCINTF (59-TGATGCCTGC TTGTTCTACG-39) and Phusion DNA polymerase (Finnzymes, Finland), according to the manufacturer's instructions, and inserted into the SmaI site of pACYC177, resulting in pACYC177-int36. Next, a 1077 bp segment covering the cat (chloramphenicol resistance) gene from pACYC184 was amplified with primers cat-f (59-CTCCGCTAGC GCTGATG-TCC-39) and cat-r (59-GTAGCACCAG GCGTTTAAGG-39) using Phusion polymerase and inserted into the singular PvuII site located in the intI1 segment of pACYC177-int36, resulting in pACYC177-int-cat. This plasmid was HincII-linearized and used to naturally transform A. baylyi SD2, giving strain KOI (verified by PCR). In parallel, an internal 938 bp segment of the recA gene of ADP1 was PCR-amplified with primers recA-f (59-AGCAAGG-CAT TACAAGCTGC-39) and recA-r (59-AATTCTGTAG AAATCTGAGG-39) and Phusion and inserted into the HincII site of pUC19, giving pUC19-recA. The 1077 bp cat segment was cloned into the singular HincII site of pUC19-recA (located in the center of the recA fragment), resulting in pUC19-recA-cat which was XmnI-linearized and used to inactivate recA of A. baylyi SD2 by natural transformation to yield strain RAM (verified by PCR). The non-transformable strain KOC4 was constructed by transformation of BD413 by DNA from a strain carrying a DcomFECB::dhfr allele (trimethoprim-resistant) [50] (A. Utnes, unpublished data).
A. baylyi was cultivated in Luria-Bertani (LB) medium with rifampicin (R) 50 mg/ml; wild-type donor bacteria as well as transformants were grown, and selected for in LB supplemented

Author Summary
Genetic elements, such as transposons and integrons, frequently carry antimicrobial resistance determinants and can be found widely disseminated among pathogenic bacteria. Their distribution pattern suggests dissemination through horizontal gene transfer. The role of natural transformation in horizontal transfer of genetic elements other than those that are self-replicative (plasmids) has remained largely unexplored. We have tested if natural transformation can facilitate transfer of transposons and class 1 integrons between bacterial species. We here provide experimental evidence showing that natural transformation can be a general mechanism for dissemination of genetic elements that by themselves do not encode interspecies transfer functions (e.g. transposons, insertion sequences). We demonstrate that antibiotic resistance determinants present in such genetic elements can spread by natural transformation between species of clinical interest. We show by quantitative data that interspecies exchange of resistance gene cassettes is highly efficient among integron-containing strains and species. Our study also provides a plausible explanation for how sequence-conserved integrons can become distributed among bacterial species. with antibiotics according to their phenotype: ampicillin (AM; 25 mg/ml), cefotaxime (CTX; 10 mg/ml), chloramphenicol (Cl; 5 or 10 mg/ml), gentamicin (CN; 10 mg/ml), kanamycin (K; 10 mg/ ml), R (25 mg/ml) spectinomycin (SC; 10 mg/ml) and trimethoprim (W; 250 mg/ml). A. baylyi cells and transformants were grown at 30uC, and the different donor bacteria at 37uC.

DNA extraction
Genomic DNA used in the transformation assays (10 mg) was isolated from bacterial cultures using anion exchange columns (QIAGEN, Germany) columns according to the manufacturers protocol and resuspended in EB buffer, pH 8.5 (QIAGEN, Germany). Plasmid DNA was isolated using a Plasmid Mini Kit (QIAGEN, Germany). The DNA concentration was measured with a UV/VIS spectrophotometer (6405 Spectrophotometer, Jenway, England) or a Nanodrop ND-1000 (Nanodrop Technologies, USA).
For the preparation of supernatants (lysate) of the heat-treated bacterial cell suspensions, 5 ml overnight cultures of the bacteria were centrifuged at 20,0006g for 5 min and resuspended in water (50 ml), followed by heat treatment at 80uC for 15 min. The raw lysate was centrifuged, and the supernatant containing DNA was collected [51]. The lack of viable cells was confirmed by streaking aliquots on LB plates.

Natural transformation assays
The recipient cells were prepared, and experiments were performed on nitrocellulose filters placed on agar-surfaces, as previously described [52,53]. Each transformation assay was repeated between two and ten times (each assay was done in triplicate). Transformation assays were done with: i) A. baylyi BD413 as recipient and DNA extracted from various wild-type integron-carrying species; ii) with A. baylyi BD413 as recipient and DNA extracted from integron-carrying A. baylyi transformants; iii) with integron-carrying A. baylyi transformant SD2 as recipient and DNA extracted from various wild-type integron-carrying species; iv) with an integrase deletion carrying A. baylyi transformant (KOI) as recipient and DNA extracted from various wild-type integroncarrying species; v) with an integron-carrying A. baylyi recA deletion (RAM) recipient and DNA extracted from various wild-type integron-carrying species; vi) with the double mutant strain A. baylyi SD9 (DrecBCD DsbcCD) as recipient; vii) and with competence mutant strain A. baylyi KOC4 (DcomFECB) as a recipient. Selection of the transformants was done with different antibiotics and concentrations (see above in ''bacterial strains and culture'' and Table 1), according to the known or established resistance levels of the donors as determined by resistance typing and MIC determination.
A positive control was included to verify recipient cell competence and reproducible experimental conditions (transformation of A. baylyi BD413 by DNA from A. baylyi KTG which contains a chromosomally located nptII [kanamycin resistance] gene [47,53]). As negative control recipient cells were streaked on LBR plus selective antibiotic without addition of the donor DNA. The transformation frequencies were calculated for each transformation assay and are given as the number of transformants divided by the number of viable recipient cells.
The following E-tests were used: ampicillin, cefotaxime, ceftazidime, gentamicin, kanamycin, spectinomycin, sulphamethoxazole, and tobramycin.  [24] or for the conserved regions of the class 1 integrase gene, IntI1 [55]. The DNA amplification program consisted of an initial denaturation step (94uC, 5 min) followed by 35 cycles of denaturation (94uC, 1 min), annealing (55uC, 1 min) and extension (72uC, 5 min), and a single final extension of 16 min at 72uC for detection of integrons [28] or was performed for 30 cycles of denaturation at 94uC for 30 s, annealing at 65uC for 30 s and extension at 72uC for 45 s, followed by a final extension time of 10 min at 72uC [55], for amplification of the integrase gene.

PCR-based detection of integrons, gene cassettes, insertion sequences and other resistance determinants
Rearrangements of gene cassettes during integron transfer and transformant cultivation were tested by PCR combining one primer for the 59-CS region and one primer that binds in the distal gene cassette in the integron of the donor bacterium.
Transformants with reduced susceptibility to some antibiotics but without a positive detection of the entire integron by PCR, were screened for the acquisition of different resistance determinants. PCRs were performed that targeted the gene cassettes present in the variable region of the integrons and the genes present in the 39-CS of the integron of the corresponding donor bacteria, with specific primers for each cassette. The presence of the insertion sequence ISAba1 [56] was screened for by PCR in the recipient A. baylyi, in A. baumannii donors and in transformants with reduced susceptibility to ampicillin. For all PCR analyses, DNA extracted from the recipient and donor bacteria were used as negative and positive controls, respectively.
Transformants with reduced susceptibility to ampicillin, but that did not yield positive PCR products for the presence of integrons, were also tested for b-lactamase activity using a qualitative chromogenic method, with nitrocefin disks (AB Biodisk, Sweden) according to the manufacturer's instructions. The randomly selected transformants showed reduced susceptibility to ampicillin; donors and recipient bacteria were used as positive and negative controls for b-lactamase production.
The genomes of transformants (St)3 and (AbII)3 (Table 1) were sequenced on the Roche/454 GS FLX Titanium platform, using one full plate and multiplexing of three 8 kb paired-end libraries. Between 137,080 and 283,658 sequence reads were generated per genome, resulting in single-scaffold assemblies with a length of 3,614,029 bp; and 3,667,429 bp and average sequencing depths of 13-fold, and 28-fold for (AbII)3, and (St)3, respectively. Sequence trimming, assembly, gene finding and annotation were performed with the automated CloVR-Microbe pipeline [58,59], which is part of the Cloud Virtual Resource (CloVR) appliance [60] developed in the CloVR project (http://clovr.org). Briefly, raw sequence data were filtered and trimmed for quality and adaptor removal, and assembled with Celera Assembler [61]. Gene predictions and functional annotations were carried out using the tools and decision process described in the IGS Standard Operating Procedure for Automated Prokaryotic Annotation [62]. The annotated assemblies resulted in between two and 20 scaffolds, i.e. one or more contigs bridged by paired-end reads. In each case, only one scaffold was larger than 10,000 bp. None of the smaller contigs showed significant sequence similarity to plasmid and/or phage sequences and were considered assembly artifacts. Fasta and annotated Genbank files of all three assemblies are available from the authors.

Pulse-field gel-electrophoresis (PFGE) and Southern blot hybridization
Genomic DNA was prepared in agarose blocks and digested for 3 h at 37uC with the endonuclease I-CeuI (New England Biolabs, Beverly, MA) that specifically recognize the rRNA operons [63]. The I-CeuI fragments were separated in a 1% agarose by PFGE using a CHEF-DR III apparatus (Bio-Rad, Hercules, Calif.) at 15uC, 6 V/cm with pulse time ramped from 20 s to 120 s over 11 h, followed by ramping from 60 s to 100 s for 11 h (adapted from Liu et al., [63]). The separated DNA was transferred by vacuum blotting (Vacugene XL, Pharmacia Biotech) to a positively charged nylon membrane (Roche, Germany) as described by Sambrook et al. [64]. The 16S rRNA [65] and intI1 [55] probes were amplified by PCR and labeled by using a PCR digoxigenin (DIG) probe synthesis kit (Roche Diagnostics, Basel, Switzerland). A DIG luminescent detection kit (Roche) was used according to the manufacturer's instructions. Co-hybridization for the 16S RNA and for the class 1 integrase intI1 was performed at 68uC.

Reverse-transcriptase PCR (RT-PCR)
Total RNA was isolated using NucleoSpinH Triprep (Macherey-Nagel, Germany) according to the manufacturer's instructions. Residual genomic DNA was removed by treatment with rDNase (Macherey-Nagel, Germany), followed by RNA ethanol precipitation. Reverse transcription was performed with 1 mg of RNA using the MonsterScript 1st-Strand cDNA synthesis kit with random 9-mer primers (Epicentre, USA), and the resulting cDNA was used in PCR reactions. The primers HS464 and HS463a [55] amplified a 473-bp fragment of intI1, and primers 16SF and 16SR [66] an approx. 1500-bp fragment of 16S rRNA. Isolated RNA was included in the PCR analyses to verify absence of DNA, and PCR targeting the 16S rRNA genes was used to confirm successful cDNA synthesis.

Biological fitness measurements
The relative fitness (W) of integron-carrying transformants was estimated by pairwise competition experiments between transformants and the untransformed recipient strain in S2-minimal medium for 24 h, as previously described [53]. Individual transformants containing one out of five distinct integrons with the following resistance genes were evaluated: bla IMP-5 (

Location and composition of integrons in the donor bacteria
The recipient strain A. baylyi BD413 does not carry identifiable integrons [44] as also confirmed by our own results from PCR and genome and direct DNA sequencing of strain BD413. The availability of multidrug resistant and integron-containing isolates in our own strain collection determined the initial selection of strains used as a source of donor DNA in our investigation. In order to be able to determine the stability of the integrons from the donor genomes during transformation of A. baylyi, the composition and genomic context of all integrons in the donor genomes was assessed by integron-specific PCR, and DNA sequencing by primer walking. PCR with primer pairs specific for the amplifi-cation of integron gene cassettes yielded a single product, which in all cases, upon sequencing, contained at least one known antibiotic resistance gene. The direct sequencing of each PCR product indicated the presence of a single type of integron in each donor genome. In A. baumannii 064, the integron harbored a central region of 763 bp, including the aadB gene (CDS 597 bp), which is known to encode an aminoglycoside adenyltransferase, responsible for gentamicin, kanamycin and tobramycin resistance. The integron of S. enterica serovar Rissen 486 contained a central region of 1913 bp, with two gene cassettes: dfrA12 (CDS 498 bp), and aadA2, (CDS 792 bp). The dfrA12 gene encodes a dihydrofolate reductase that confers resistance to trimethoprim, and the aadA2 gene encodes an aminoglycoside adenyltransferase, responsible for streptomycin and spectinomycin resistance.
The genomic location of the integrons in the donor genomes was determined by PFGE. Co-hybridization of the intI1 and 16S rRNA probes was interpreted as indicating chromosomal location, whereas hybridization only with the intI1 probe was interpreted as indicating a plasmid location of the integron. A plasmid location of the integron was shown in the donors S. enterica serovar Typhimurium 490, E. cloacae C2R371 and C. freundii C16R385, while integrons in the A. baumannii 064, S. enterica serovar Rissen 486, E. coli C10R379, A. baumannii 65FFC [28] and P. aeruginosa SM [67] were located on the chromosome. The A. baumannii 064 and S. enterica serovar Rissen 486 strains both contain a plasmid, as observed by agarose gel electrophoresis (data not shown). E. fergusonii AS041A2 also harbored a plasmid and repeated PFGE results were not conclusive for the determination of the location of the integron in this strain.
Sequencing of the flanking regions of the integrons revealed that their insertion sites varied in the donors, and that the integrons were often linked to transposable elements. In S. enterica serovar Typhimurium 490, the integron was inserted in an intact Tn21like transposon [36]. The integron of A. baumannii 064 was inserted in a Tn1721-like transposon [68,69], flanked by the insertion sequence IS26 on both sides, forming an IS26-composite transposon. The A. baumannii 65FFC integron was embedded in a defective Tn402-like transposon, flanked by a 439 bp Miniature Inverted-repeat Transposable Element (MITE) on both sides [70].
In S. enterica serovar Rissen 486, the tnpA, tnpR and tnpM genes were detected next to the 59-CS region of the integron. The chrA gene, which codes for a putative truncated chromate ion transporter, was found flanking the 39-CS region. The closest homologue to the tnp genes was found in the transposon Tn1721, suggesting the integron in serovar Rissen 486 may be localized inside a transposon. However, for technical reasons, we were unable to obtain additional DNA sequence by primer walking.
In P. aeruginosa SM, tnpR and tnpM genes were identified adjacent to the 59-CS flanking region of the integron. The genes flanking the 39-CS region could not be determined. In this case, the tnp genes were 100% identical to the genes found in the Tn5051-like transposon; also in this case indicating that the integron is located in a transposon.
The flanking regions of the integrons of C. freundii C16R385, E. cloacae C2R371, E. coli C10R379 and E. fergusonii AS041A2 were not determined.

Interspecies transfer of integrons/gene cassettes into wild-type A. baylyi
Phenotypic characterization of transformants. Natural transformation of A. baylyi BD413 by integron-containing DNA from the related species A. baumannii occurred at frequencies up to 1.6610 27 transformants per recipient over a 24 h period. Natural transformation was also seen after exposure to integron-containing DNA of strains from the unrelated bacterial species C. freundii, E. coli, E. cloacae, K. pneumoniae, P. aeruginosa, S. enterica serovar Rissen, and S. enterica serovar Typhimurium, at frequencies ranging from 4.4610 29 to 2.1610 27 ( Table 2). The only two integroncontaining DNA sources tested and not capable of transforming wildtype A. baylyi BD413 above the detection limit of ,1 transformant per 10 9 bacteria were two Escherichia isolates (E. coli C10R379 and E. fergusonii AS041A2). In addition to purified DNA, the supernatant of a heat-killed bacterial cell suspension also gave rise to bacterial transformants ( Table 2) at comparable frequencies. Thus, DNA purity seems to be of limited importance. As expected, natural transformation of the competence deficient mutant A. baylyi KOC4 as a recipient did not yield transformants.
Antimicrobial susceptibility profiles were determined for selected transformants (Table S1). The susceptibility profiles of some of the transformants were altered in ways that could not be explained by the resistance phenotypes encoded by the transferred complete integron. The hypothetical transfer of single gene cassettes rather than the entire integron could have explained the reduced susceptibility of transformants if such cassettes would have been inserted next to a host region providing promoter functions. However, none of the resistance genes, present as part of the integrons of the donor bacteria, could be identified by genecassette specific PCR analyses in the tested atypical transformants.
Transfer of the insertion sequence ISAba1 could hypothetically explain reduced ampicillin susceptibility in some transformants, due to induction of expression of the A. baylyi ampC gene [56]. However, this insertion sequence was not detected by PCR in the transformants tested. The transfer of other b-lactamase determinants unrelated to the integron could also explain reduced ampicillin susceptibility in some transformants. Observed transfer of b-lactamase was detected in two transformants resulting from exposure to DNA of A. baumannii 064 DNA [(AbII)4 and (AbII)L1]; however, the molecular basis for the resistance was not determined. See footnote of table S1 for an explanation of the abbreviations of transformants.
Genotypic characterization of transformants. Among all the phenotypes with increased resistance observed after exposure of the BD413 strain to DNA extracted from various integroncarrying species, four transformants were shown to have acquired the complete integron from S. enterica serovar Typhimurium 490 [named SD1, (St)1, (St)2 and (St)3], and another four transformants the entire integron from A. baumannii 064 [named SD2, (AbII)1, (AbII)2 and (AbII)3], as confirmed by PCR ( Figure S1A) and DNA sequencing. The DNA sequence of the integron and integron flanking regions of the same transformants were determined. In all cases, acquisition of complete integrons plus additional flanking DNA was shown. Flanking acquired DNA fragments were up to 23,000 bp long and included in several cases transposable elements or other genes with important phenotypes, such as a chloramphenicol resistance gene (catA2). Figure 1 summarizes the genetic composition of the DNA acquired by the individual A. baylyi transformants.
In the transformants (St)1 and SD1, obtained after exposure to DNA of S. enterica serovar Typhimurium 490, the incorporated DNA region was found to consist a copy of transposon Tn21-like, which also contained the integron. In both cases, target site duplications of 4 and 5 bp, respectively, were identified providing evidence for that the (St)1 and SD1 transformants resulted from active transposition of Tn21-like (figure S2A). The transformant SD1 also showed a partial duplication of the merR gene of the transposon. In transformant (St)1, the integron-containing transposon had inserted into the A. baylyi gene ACIAD1773, which encodes a putative transport protein permease ( Figure 1A). In The transformation frequency obtained with DNA from positive control source (A. baylyi KTG with the ntpII gene) was 4.7610 24 . Negative control experiments, consisting of the same experimental conditions, with the exception of that no DNA was added, were included in each transformation assay. c) The transformation frequency is, when calculated as the frequency per exposed unit of DNA: 1.  transformant SD1, the integron-containing transposon had inserted into gene ACIAD3230, encoding a putative proton/ sodium-glutamate symport protein ( Figure 1B).
Whole-genome sequence analysis of A. baylyi transformant (St)3 revealed that the integron had transferred as part of a Tn21-like transposon. The incorporated DNA was found in the transformant genome in a region containing a 54 kb stretch of DNA not present in the published A. baylyi ADP1 strain. This DNA stretch has not been described before, and appears to be a prophage region, inserted in the ACIADtRNASer_34 gene. The transposon present in transformant (St)3 also showed rearrangement compared with its donor, S. enterica serovar Typhimurium 490. In (St)3, the tnpR, tnpA, ybjA, catA1 and IS1 genes were located at the 39-CS region, flanking the mer operon, whereas in the previously described Tn21 sequences, they are located in the 59-CS region ( Figure 1E). The rearrangement was confirmed by sequencing as well as by PCR.
DNA sequencing by primer walking also showed that transformant (St)2 acquired the integron embedded in the Tn21-like transposon, without rearrangements in this case, and the insertion had also occurred in the 54 kb region indicated above.
The A. baylyi transformant SD2, obtained after exposure to A. baumannii 064 DNA, also acquired considerable stretches of genomic donor DNA flanking the complete integron ( Figure 1C). Extensive primer walking identified a crossover in the lrp gene of both the donor and recipient genomes, 5,428 bp from of the 39-CS region of the integron. The second crossover junction was not identified but, based on primer walking results, had to be located at least at a distance of 10,500 bp from the 59-CS region of the integron; the acquired DNA in this region included genes similar to the Tn1721 transposon. The high nucleotide similarities between the lrp gene of the donor and the recipient and the characteristics of the crossover junctions suggest that recAdependent homologous (also called homeologous, heterologous, or heterogamic) recombination had facilitated the acquisition of the integron-containing DNA, a process recently investigated in detail for A. baumannii donor and A. baylyi recipient cells [53].
Whole genome sequencing of the A. baylyi transformant (AbII)3 revealed that it had acquired approximately 18,625 bps of DNA sequence from the A. baumannii 064 genome, which contained a IS26-composite transposon ( Figure 1D); the horizontally acquired fragment was submitted to GenBank under the accession number JX041889. The insertion contained the integron embedded in a transposon-like region including a tet (tetracycline resistance) gene and a tnpA transposase gene (74% similar to the tnpA gene of Tn1721), and further flanked on both sides by the IS26 elements. A target site duplication of 8 bp on both sides of the IS26 insertion suggests transfer of the entire 18,625 bp segment occurred by transposition facilitated by IS26 sequences, into the ACIAD0480 gene of A. baylyi (encoding a putative membrane protein). Analysis of additional transformants, exposed to DNA of A. baumannii 064 by primer walking allied with PCR confirmed that the integrons acquired by transformants (AbII)1 and (AbII)2 were also located within a Tn1721-like transposon, and flanked by a IS26 element. Although the exact locations in the A. baylyi genome were not determined for the latter, the obtained DNA sequence suggests a similar composition and mode of transfer as for transformant Ab(II)3.
The possible transfer of whole plasmids was investigated in the transformants collected after exposure to DNA from the plasmidharboring S. enterica serovar Typhimurium 490 or A. baumannii 064. Although plasmids were clearly identified in both of the donor species by agarose gel-electrophoresis (data not shown) or PFGE, the same plasmids could not be detected in the transformants ( Figure S3). The transformants SD1, (St)3, SD2 and (AbII)1 were shown by PFGE and probe hybridization to carry the acquired integron in the chromosome. It is noted that co-hybridization with the two probes targeting intI1 or the 16S rRNA genes did not take place for DNA extracted from transformants (St)1 and (St)2. Chromosomal location is nevertheless assumed because the larger fragment, where the integron of these two transformants is located, is not expected to hybridize with the 16S rRNA probe even though it is chromosomal. This is due to the opposite orientation of the rRNA operons at nucleotides 1660700-1666022 and 2941950-2947363 (access. number CR543861), leading to a 1.28 MB I-CeuI fragment without the 16S rRNA gene. A. baylyi BD413 has seven rRNA operons, producing seven fragments after digestion with I-CeuI [71], which can be seen for strain BD413 and transformants (St)1 and (St)2 ( Figure S3A), lanes 2, 4 and 6, respectively. Variable locations, as well as different numbers, of the hybridizing band in the PFGE results of transformants (AbII)2 and (AbII)3 ( Figure S3) led to several repetitions of the experiments; however both transformants repeatedly displayed a variable chromosomal location of the integrase sequence (data not shown), also confirmed by full genome sequence of one of these transformants, (AbII)3. This variability may hypothetically be due to gene amplification, which is reported to occur in A. baylyi [72,73], followed by intra-genomic rearrangements. Rearrangements are often associated with repetitive sequences [74], which are known to represent 1.6% of the A. baylyi genome [44].
Effects of the RecBCD exonuclease. As several of the observed interspecies integron transfers appeared to have taken place through transposition, a transient existence of a linear double-stranded (ds) DNA intermediate in the cytoplasm was assumed. Such intermediates are thought to be substrates for the dsDNA-attacking exonuclease RecBCD [75,76], and recent observations suggest that RecBCD also removes dsDNA-intermediates occurring during transformation (K. Harms et al., unpublished data). We deleted the recBCD operon (and the RecBCD suppressor genes sbcCD to obtain a wildtype-like viability; [49]) in the strain BD413, yielding strain SD9 which was employed as recipient in transformation experiments. Three of the SD9 transformants showed acquisition of integron from the S. enterica serovar Typhimurium 490. The low transformation frequencies obtained with this strain do not suggest that lack of RecBCD (and/or SbcCD) protects transformable cells from transposon activity initiated by linear double-stranded DNA present in the cytoplasm.

Intraspecies transfer of integrons/gene cassettes in A. baylyi
To investigate whether acquisition of DNA in a first transformation would impact transformation efficiencies in subsequent experiments, transformation of A. baylyi BD413 by DNA extracted from A. baylyi transformants that previously acquired an integron (SD1, SD2, SD3, SD4, SD5 and SD6) was performed. Intraspecies transformation was 10-to 1,000-fold more efficient than interspecies transfer of the same integrons (Table 2).
Crossover junctions in the A. baylyi genome sequence were detected in 6 out of 6 tested transformants, suggesting that the observed intraspecies gene transfer occurred via homologous Extensive DNA sequencing by primer walking of 6 transformants suggested that the composition of the donor DNA segments acquired in the initial interspecies transformation assay were maintained after subsequent intraspecies transformation ( Figure 1A and C).
All of the tested transformants showed expected antimicrobial susceptibility profiles (Table S1) and had acquired the complete integron, which was confirmed by PCR ( Figure S1C). The conserved composition of the acquired integron and the flanking regions suggest homologous recombination facilitated DNA exchange between the two A. baylyi genomes involved.

Interspecies transfer of integrons/gene cassettes into integron-containing A. baylyi
To determine the impact of integrons resident in the recipient bacteria on the overall transformation efficiencies, we exposed integron-containing bacteria to chromosomal DNA with integrons of different compositions. Interspecies transformation of integroncontaining A. baylyi recipient cells was 10-to 100-fold more efficient than interspecies transformation of wildtype A. baylyi cells ( Table 2). The presence of integron sequences in the recipient cells (A. baylyi SD2) led to efficient replacement of gene cassettes rather than to the accumulation of additional integrons or gene cassettes in all tested transformants (76 out of 76) (figure S2B). Replacement was shown by changes in antimicrobial susceptibility profiles (Table S1) and the obtained fragment sizes by integron-specific PCR assays ( Figure S1B). Recipient strains with integron-encoded aadB (SD2, KOI, RAM) lost the corresponding aminoglycoside resistance profiles upon transformation by genomic DNA containing different integrons (Table S1). Gene loss and substitution of the entire integron in the transformant strain was confirmed by PCR ( Figure S1B). Rearrangements of gene cassettes within the integrons were not observed among the analyzed transformants. The expression of the acquired integrase gene in A. baylyi transformant SD2 was not detected in RT-PCR analyses ( Figure S4), which might explain the absence of observable recombination of gene cassettes among the limited numbers of transformants examined.
Transformation frequencies of the recA mutant RAM were between 10 4 -to 10 5 -fold lower when compared to the recAproficient recipient strain SD2 (Table 2), suggesting homologous recombination as mechanism responsible for efficient substitutive recombination.
Absence of effects of the integron-encoded integrase. To investigate whether the integrase was required for integronacquisition or affected transformation efficiencies, experiments were repeated using the intI1::cat-inactivated integron-carrying A. baylyi strain KOI as a recipient. The resulting transformation frequencies were similar to those obtained with the recipient (SD2) harboring a functional integrase gene (Table 2). Of the 73 analyzed transformants, 70 acquired the entire gene cassette composition from the integron of the donor bacteria ( Figure S1D). Further nucleotide analysis of the intI1 region of these transformants revealed that the inactivated intI1 region had been replaced with a functional copy from the donor bacteria in 4 out of 10 transformants after exposure to DNA from A. baumannii 65FFC, in 17 out of 21 from P. aeruginosa SM, in 10 out of 21 from S. enterica Rissen 486 and in 14 out of 21 from S. enterica Typhimurium 490. The remaining integrons maintained the inactive integrase of the recipient. The lack of an effect of the intI1-encoded integrase in the recombination process was also seen when the donor DNA contained an inactive integrase (Table 2); thus excluding a role of the integrase inter-and intraspecies acquisition of integrons or gene cassettes in A. baylyi recipients already containing an integron.

Discussion
Horizontal gene transfer (HGT) is a key driver of bacterial adaptation and evolution [77][78][79]. A number of studies demonstrate that a significant amount of bacterial genomes are affected by HGT events [80][81][82][83]. In this study we examined to what extent DNA substrates and natural transformation can lead to interspecies transfer of distinct genetic elements encoding site-specific recombination mechanisms.
Annealing of complementary DNA strands in the cytoplasm is considered necessary for recircularization and hence stable uptake of plasmids in competent bacterial species [84] (K. Harms, unpublished data). Drawing on this general observation, we hypothesized that annealing of other single-stranded DNA taken up into the cytoplasm of competent bacterial cells would allow transient expression of genes present on such linear fragments. For instance, expression of recombinase genes present in mobile genetic elements in these double-strand DNA fragments could lead to integration of the elements into the host chromosome. To what extent natural transformation of bacteria can facilitate intergenomic mobility of genetic elements being part of species-foreign, linear DNA fragments can be tested.
Indeed, we provide experimental data demonstrating that natural transformation can facilitate interspecies transfer of integrons and transposons that is not limited by the genetic relatedness of the donor with the host. Exposure of A. baylyi to DNA of an integron-containing Salmonella strain led to horizontal transfer of its integron due to transposition of a Tn21-like transposon that contained the integron. The Tn21-like transposon was found as single inserts at different chromosomal locations in 3 different examined transformant genomes. Exposure of A. baylyi to DNA of an integron containing A. baumannii strain also led to transformants with single chromosomal integrations of the integron. The integrations occurred by either DNA sequence similarity-based homologous recombination, or by genetic linkage to and movement of IS26 elements. Resistance profiling, PCRbased target amplification and extensive DNA sequencing confirmed the acquisition and chromosomal insertion of complete integrons and its flanking DNA by A. baylyi transformant cells. The general acquisition process involved is illustrated in Figure S2A.

Interspecies transfer into integron free A. baylyi cells
Purified DNA substrates of A. baumannii 064 and S. enterica serovar Typhimurium 490 transformed A. baylyi cells at frequencies of 10 28 and 10 27 over a 24 h time period. In all examined cases, the composition and order of gene cassettes in the donor bacteria was maintained in the transformant cells. Also DNA lysates could transform A. baylyi cells suggesting variability in DNA purity is of minor importance for HGT, as observed previously [51].
DNA sequencing of the flanking regions of the integrons in 6 initial transformants obtained after exposure to species-divergent donor DNA verified their incorporation into the A. baylyi BD413 chromosome. The genomic locations of the integrons were also maintained in subsequent transformants (n = 6) obtained after exposure to DNA isolated from the initial transformants. Sequencing of the flanking regions of the transferred integrons also revealed that not only the integron was acquired from the donor bacteria, but also additional DNA flanking the integrons. The length of the acquired, continual DNA fragments could be up to 23,000 bps long. The sequencing of the acquired DNA in 6 transformants obtained after exposure to DNA of various species, as well as Southern hybridization of a total of 8 transformants, revealed that integration had occurred in different regions of the A. baylyi genome.
Most interestingly, stable integration of the integron containing DNA in the A. baylyi chromosome had occurred by several mechanisms, depending of the flanking regions of the integrons. Transposition-based insertions were observed for integrons acquired from unrelated species. Specifically, DNA sequencing revealed that exposure to DNA substrates of S. enterica serovar Typhimurium 490 resulted in chromosomal integration of the acquired DNA due to the activity of transposase genes flanking the integrons; such as transposon Tn21-like. Integrons are often embedded in transposon Tn21 [36], and many different gene cassette arrays have been reported for integrons present in Tn21like transposons [85]. The horizontal dissemination of this transposon is a major contributor to the dissemination of integrons. Here we have experimentally demonstrated that natural transformation can facilitate interspecies movement of this transposon. For a related species (A. baumannii DNA donor), homologous recombination-based insertions were observed, as expected [53]. However, in this latter exposure scenario, also sitespecific recombination due to IS26-associated transposition was observed.
Similarly to Tn21-like, IS26 elements are often associated with antibiotic resistance genes and presumed active in Acinetobacter [86]. There are several reports of IS26-associated integrons [87,88], and transposition of IS26-composite transposons has been suggested [89]. The presence in the recipient of nonfunctional IS26 could also allow the capture of additional resistance genes by homologous recombination-mediated DNA integration, if a homologous IS26 region is present in the donor DNA fragments. Such recombination scenario could explain, for example, the dissemination of the chromosomally-located integron-IS26 in A. baumannii in South Korea [88,90]. A previous bioinformatics-based study of Partridge et al. [41] also indicated transposition as a possible mechanism of integron movement. However, no experimental tests were performed.
The biological properties of natural transformation as a mechanism that can facilitate DNA translocation over bacterial membranes and site-specific recombination (e.g. active transposition events of transposons) in bacterial genomes could be further utilized. For instance, both model and applied systems can be developed using small DNA fragments with site-specific recombination functions as donor DNAs. Moreover, possible interactions between recombinases encoded by the incoming DNA and host enzymes should be investigated. For example, the A. baylyi ADP1 strain harbors two 2 prophages, a number of transposases, and 6 copies of IS1236, with two of them flanking Tn5613 [44].
The exposure of A. baylyi cells to the various DNA substrates resulted in surprisingly high transformation frequencies, as determined by phenotypic screening on antibiotic-containing growth media. However, subsequent genotypic screening (PCR) showed that only some of the transformants with changed antibiotic resistance profiles had acquired entire integrons. The observation of variable changes in susceptibility patterns, after exposure to various sources of bacterial DNA, suggest the broad potential of and the presence of a yet not described range of transferable resistance determinants in bacteria, of unknown mechanistic and genetic nature. Our study suggests that exposure of competent bacteria to heterologous sources of DNA may produce complex changes in resistance profiles, not necessarily predictable from the known resistance genes present in a given donor isolate. In two examined cases, changes in resistance profile were due to the acquisition of b-lactamases as determined by the nitrocefin test. The broader effects of the composition of the A. baylyi genome after the heterologous recombination with DNA from the related species A. baumannii will be determined in a separate study. A recent study by Mell et al. [91] reported transfer of numerous chromosomal polymorphisms between Haemophilus influenza genomes after natural transformation.

Inter-and intraspecies transfer into integron-containing A. baylyi cells
Natural transformation of the integron-carrying A. baylyi SD2 recipient with DNA isolated from different species of donor bacteria gave comparable transformation frequencies to those obtained with DNA of the same species. Thus, interspecies acquisitions of integrons that take place at low initial frequencies can be followed by high-frequency interspecies dissemination. Interestingly, all tested transformants acquired the gene cassette composition of the integron of the donor bacterium, rather than acquiring additional new gene cassettes in the existing integron. This result suggests that homologous recombination replaces gene cassettes through gene replacement caused by crossover junctions forming at the conserved end segments of the integron ( Figure  S2B). A role of the intI1 integrase was not observed in our studies.
Transformation experiments performed with recipient cells containing an integron with the integrase gene inactivated (strain KOI), and also with recipient cells containing an integron, but with the recA gene inactivated (strain RAM) that resulted four orders of magnitude reduction in transformation frequencies, confirm the role of homologous recombination [53,92] in the replacement of integrons and gene cassettes in these recipient bacteria. Partridge et al [41] also suggested replacement of a gene cassette array in Pseudomonas aeruginosa to have occurred by homologous recombination. Here we provide experimental evidence that homologous recombination can efficiently replace gene cassette arrays. Mobile genetic elements thus provide sufficient DNA similarity for homologous recombination to occur between otherwise unrelated bacterial species. The presence of DNA similarity can therefore likely contribute to the formation and spread of chromosomally-located complex mosaic regions, often formed by several resistance genes and mobile elements [93].
Recent publications demonstrate that many integrase genes in integrons carry LexA binding sites in the vicinity of their promoter regions and can be controlled by the host LexA protein (transcriptional repressor of the SOS response [94,95]). In organisms harboring lexA alleles, SOS induction can lead to increased integrase activity resulting in increased frequencies of cassette rearrangements [94,95]. SOS induction of the expression of the integrase by both conjugation [95] and natural transformation [96] has been recently shown in V. cholerae. In contrast, and unlike many Enterobacteriaceae, the SOS response in A. baylyi is not regulated by a LexA orthologue [97]. The integrase gene in our recipient strain is therefore derepressed or regulated by other unknown factors, as expression was not found in the RT-PCR.
The absence of detected rearrangements of gene cassettes in our study might be explained by the absence (or low) expression of the integrase in the recipient A. baylyi SD2, and by the fact that the homologous recombination events will occur much more frequently than integrase-catalyzed cassette recombination events. Therefore, rare rearrangements of gene cassettes may have occurred in our model system but would not be detectable in the limited number of transformants that were genetically characterized. The recent work published by Baharoglu et al. [95] on conjugative induction of SOS and its ability to trigger cassette recombination showed that cassette rearrangements occurred at very low frequencies (e.g. 10 28 ) in V. cholerae). The relative low frequency of integrase-facilitated cassettes recruitment and rearrangements, compared with the frequency of homologous recombination, can also help explain why some gene cassette arrays are common and shared between different bacterial species [98].

General considerations
Intervention policies aimed at combating antibiotic resistance often rely on the assumption that resistance genes are costly and will be lost from pathogenic bacterial populations when antibiotic pressure is removed [99]. Our relative fitness measurements suggest that the acquired integrons and the resulting changes in resistance profiles did not cause major fitness reductions of transformants. It is noted that the fitness measurements represent the combined effect of all transferred DNA segments, as well as effects on the transformant genome due to insertion effects.
Natural competence is described in at least 60 species of bacteria, including several species present in clinical environments [100]. However, the proportion of species able to acquire DNA by natural transformation remains to be clarified, as many species or strains may have the capacity without such being detectable under laboratory conditions that are limited in sensitivity, time and by environmental variables [100]. The occurrence of naturally transformable species within the pool of uncultivable bacteria also remains to be fully explored. Interestingly, several species or strains, including E. coli, A. baumannii and Pseudomonas stutzeri, regarded as non-competent have later been shown to be naturally transformable under certain circumstances [101][102][103][104]. Natural transformation could, for instance, be responsible for the dissemination of chromosomal integrons in P. aeruginosa [105]. Recently, the acquisition of (synthetic) gene cassettes by natural transformation in a Pseudomonas species [39] was also shown; indicating that homology-independent DNA acquisitions might be more general phenomenon. It is also worth noting that chromosomal super-integrons are mainly found in Vibrio species, which are naturally transformable [9,100].
Our results indicate that natural transformation of DNA fragments is not necessarily limited by requirements of high DNA sequence similarity for stable integration to occur; and that genetically unrelated bacteria can exchange genetic material (such as integron-containing transposons) through natural transformation provided the transferred DNA fragments encodes functions for site-specific recombination. As observed in our studies, the genetic signature of events of natural transformation can be limited to the site-specific insertions of transposons. However, the observation of transposons or insertion sequences in a given bacterial genome is usually not causally attributed to events of natural transformation. Thus, retrospective DNA sequence analysis would not have considered natural transformation as the causal mechanism until now.
This study suggests a significant potential for the interspecies spread of mobile genetic elements such as transposons, and insertion sequences through natural transformation. Hence, presenting a new pathway for horizontal dissemination of antimicrobial resistance determinants present in integrons embedded in transposons.