Introduction

Gene flow across distantly related bacterial groups (horizontal gene transfer) is a major feature of bacterial evolution (Maynard Smith et al, 1991; Campbell, 2000; Ochman et al, 2000; Gogarten et al, 2002). This evolution need not be slow. The intense selection pressure imposed on microbial communities by worldwide antibiotic use reveals that new multiresistance plasmids can arise from diverse origins and spread in less than five decades (Hartl and Dykhuizen, 1984; Davies, 1994). In this case, the antibiotic resistance genes that spread so rapidly were mostly carried by mobile genetic elements.

Genetic exchange of DNA that is not self-transmissible can occur within a similar timeframe (Maynard Smith et al, 1991; Bowler et al, 1994), but such genetic exchanges occur by homologous recombination and so are usually restricted to very closely related bacteria. Thus the primary genetic barrier against genetic exchange of DNA that is not self-transmissible is dissimilarity in the bacterial DNA sequences concerned (DuBose et al, 1988; Rayssiguier et al, 1989; Maynard Smith et al, 1991; Vulic et al, 1997; Feil et al, 2001). The rate of homologous recombination is high enough to obliterate the phylogenetic signal among some close bacterial relatives (Feil et al, 2001), and for the nucleotide sites of some species, homologous recombination is a far more likely prospect than a point substitution (Guttman and Dykhuizen, 1994; Feil et al, 1999, 2000, 2001). Recent experiments show that close DNA sequence similarity can also facilitate the stable acquisition of linked nonhomologous foreign DNA sequences (Nielsen et al, 2000a, 2000b; de Vries and Wackernagel, 2002; Kay et al, 2002; Prudhomme et al, 2002).

The synthetic vectors that are heavily used in industry and research are mosaics of mobile DNA elements from natural bacterial isolates (Pouwels et al, 1988). Care was taken to render these vectors nontransmissible by removal of the genes necessary for their mobility (Berg et al, 1975; NIH guidelines for research involving recombinant DNA molecules: Appendix I, April 2002). Owing to their natural origins, artificial vectors still bear a residual similarity to DNA of naturally occurring bacteria. If this residual similarity were to enable recombination between bacterial DNA and only the parts of the artificial vectors that are homologous, there would most likely be no discernable effect on the bacterium involved. However, a plethora of different DNA sequences is inserted into artificial vectors and this raises the question whether homology between vector and bacterial DNA can facilitate the uptake of these foreign DNA inserts. In this review, we address this question by considering the extent of vector similarity to natural bacterial sequences, the requirements of homology-mediated foreign DNA uptake, the persistence of plasmid DNA and its bacterial uptake in the environment, and the possible effects of natural selection.

The natural homologies of artificial vectors

How similar are the sequences of artificial vector DNA to those of naturally occurring bacterial DNA? We use the relatively small cloning vector pUC18 (a pBR322 derivative) to address this question, because it is one of the simplest synthetic vectors, is commonly used, and is similar in sequence and organisation to many other artificial vectors.

The 2686 bp plasmid vector pUC18 is a mosaic of three different naturally occurring DNA sequences: an origin of replication and its flanking DNA from the Escherichia coli ColE1 plasmid, an antibiotic resistance gene and its flanking DNA from the transposon Tn3 of a naturally occurring resistance plasmid, and part of a β-d-galactosidase gene (LacZ) and flanking DNA from E. coli chromosomal DNA that may have originated from transposon Tn951 (Cornelis et al, 1978). The remaining 55 bp originate from the mostly synthetic multiple cloning site (MCS, or polylinker region originally developed in the bacteriophage M13) into which recombinant DNA is inserted (Figure 1) (Pouwels et al, 1988). The ‘M13’ universal primers that are generally used for PCR of recombinant DNA bind to sites that are actually homologous only to the β-d-galactosidase gene.

Figure 1
figure 1

The natural origins of plasmid vector pUC18. These were summarised using the details given in the GenBank entries of pUC18 (Accession: L08752), cloning vector pBR322 (J01749), cloning vector M13mp18 (X02513), the E. coli lactose operon (J01636), transposon Tn3 (V00613), the ColE1 plasmid (NC_001371) and references therein, and by aligning these sequences with pUC18, using the NCBI ‘BLAST 2 sequences’ facility at <www.ncbi.nlm.nih.gov/BLAST/>.

Full size image

The largest component of pUC18 is the Tn3-like region containing the 860 bp TEM-1 β-lactamase gene (blaTEM-1), which confers resistance to the antibiotic ampicillin, and other penicillins (β-lactam antibiotics). The Tn3-like region of pUC18 (1209 bp) is almost identical (97.5–99.8% similarity) to all reported natural TEM-type DNA sequences (Table 1). These include the TEM-type β-lactamases that are resistant to the new generation of synthetic β-lactams and those that are resistant to inhibitors manufactured to suppress β-lactamase activity. These highly related TEM-type sequences have spread, on Tn3-like transposons, to an extremely broad range of bacterial taxa. The TEM-type sequences deposited in GenBank alone (Table 1) represent species of commensal and pathogenic bacteria of the Enterobacteriaceae, other Proteobacteria such as Pseudomonas aeruginosa, Stenotrophomonas maltophilia, Neisseria meningitidis, Neisseria gonorrhoeae (Table 1) and the phylogenetically distant gum-dwelling Capnocytophaga ochracea of the Flavobacteria (Rosenau et al, 2000). TEM-type β-lactamases are carried by plasmid-borne transposons, to which pUC18 has at least 99% similarity, including Tn1, Tn2, Tn3, Tn801, Tn841 and TnSF1 (Table 1).

Table 1 Examples of homology between naturally occurring bacterial DNA and pUC18
Full size table

The broad distribution of these sequences is less surprising considering that β-lactams are the most commonly prescribed antibiotics worldwide, and that the single most common form of genetic resistance to these antibiotics is TEM-type β-lactamase production (Therrien and Levesque, 2000). Early work on antibiotic resistance has shown that TEM-type β-lactamases occur in a broad range of bacterial species, in many different countries, and on many different conjugative resistance plasmids, whose incompatibility groups include FI, FII, Iα, N, A-C complex, H, L, M, P, S, T, W, X, and Y (Matthew and Hedges, 1976). The Tn3-like region of pUC18, and most ampicillin-resistant vectors, is derived from one of the first antibiotic resistance factors to be isolated in 1963 (Datta and Kontomichalou, 1965; Sutcliffe, 1978), and it is not surprising that one of the first natural isolates of an antibiotic resistance gene represents one of the most ubiquitous types.

The pUC18 ColE1-like replication region has at least 83% sequence identity to the replication origins of many naturally occurring, small, transmissible plasmids of the Enterobacteriaceae (Table 1). In some cases similarities flank the pUC18 cloning site, such that they also flank any DNA that is inserted into the vector. This is also the case for the β-d-galactosidase (LacZ) region (Table 1 and Figure 1).

The cloning vector pUC18 is not unusual among synthetic vectors. Most of the artificial vector sequences in the GenBank vector database share pUC18's ColE1-like origin of replication (671 of 911 sequences), and its Tn3-like region conferring ampicillin resistance (648 of 911). Other genes commonly used as selective markers for vectors are also derived from mobile, naturally occurring bacterial DNA sequences. For example, neomycin resistance (nptII) is conferred by a region of transposon Tn5, chloramphenicol resistance is from transposon Tn9, and kanamycin resistance (nptI) is from transposon Tn903 (Pouwels et al, 1988).

A hitchhiker's guide to homology

How much homologous DNA sequence is required for the bacterial acceptance of how much foreign DNA? In a recent experiment, de Vries and Wackernagel (2002) grew Gram-negative naturally competent bacteria (Acinetobacter sp. BD413) in the presence of linear naked DNA. As little as 183 bp of homology between the linear naked DNA and a plasmid in the bacteria was enough to facilitate detectable levels of foreign DNA uptake. Further experiments, which used 1 kb of homology, showed that this homology enabled the frequent bacterial acquisition of a working bleomycin resistance gene that itself had no homology to the plasmid. This 1 kb of homology frequently anchored the stable integration of as much as 2.6 kb of foreign DNA and about 1 kb on average (de Vries and Wackernagel, 2002).

Prudhomme et al (2002) present very similar findings for Streptococcus pneumoniae, a naturally competent Gram positive bacterium, where less than 1 kb of one-sided homology is sufficient for the chromosomal uptake of over 2 kb of foreign DNA. In the absence of DNA homology no stable integration of the naked DNA with bacterial DNA could be detected, but when homology to bacterial DNA flanks both sides of the foreign DNA the rate of uptake is higher than in the case of one-sided homology (de Vries and Wackernagel, 2002; Prudhomme et al, 2002) and close to the rate at which naked DNA is able to enter the bacterium (Nielsen et al, 2000a, 2000b; de Vries et al, 2001). Prudhomme et al also show that a single homologous region (942 bp) facilitates the integration of whole plasmids into bacterial chromosomes at a 100-fold greater rate than foreign DNA uptake from the same plasmid when linearised.

Several studies estimate the degree of similarity required for homologous recombination in different types of bacteria. The frequency of recombination of DNA molecules in bacteria, such as the enterobacteria (Vulic et al, 1997, 1999), Bacillus (Majewski and Cohan, 1998), and Streptococcus (Majewski et al, 2000), increases exponentially as the nucleotide similarity of the participating DNA molecules increases. Most wild-type bacteria abort effective recombination with DNA that is over 1–2% diverged from their own; however, the constraints on sequence similarity for homologous recombination can be relaxed to ∼20% divergence in certain ‘mutator’ individuals (Rayssiguier et al, 1989; Vulic et al, 1997, 1999; Majewski et al, 2000), and others predict recombination between molecules that are up to 30% diverged (Townsend et al, 2003). Mutator individuals make up 1–2% of natural E. coli and Salmonella populations (LeClerc et al, 1996; Matic et al, 1997), though one study of patients with cystic fibrosis, found that, on average, 20% of the pathogenic isolates of P. aeruginosa examined showed mutator phenotypes (Oliver et al, 2000). Homologous recombination in mutators could explain the finding that S. pneumoniae strains became resistant to antibiotics in the last 50 years through changes to their penicillin-binding protein genes by at least two independent recombination events, involving DNA sequences that were up to 20% divergent (Maynard Smith et al, 1991).

Most of the homologies described in Table 1 are between pUC18 and Tn3-like transposons. Most of these are less than 1% diverged from naturally occurring bacterial DNA sequences and span more than 1 kb. If artificial vectors and their inserts were available to bacteria, such sequence similarity would seem sufficient to facilitate the bacterial uptake of insert DNA. The sequences with which pUC18 has Tn3-like homology are on broad host-range conjugative plasmids and are associated with broad host-range mobile elements (Table 1). If new DNA sequences could be associated with such mobile elements, this may increase the chance of it spreading to other bacteria. The homology of pUC18 to plasmids with ColE1-like origins of replication is also sufficiently long for the uptake of inserted DNA sequences, but natural and artificial sequences are up to 17% diverged so homology-facilitated uptake of foreign DNA would seem likely only in mutators. Homologies of ColE1-like origins of replication are generally associated with small nonconjugative plasmids (Table 1) so even if recombination did occur with these, it probably would not significantly increase the chances of spread to other bacterial species.

Bacterial competence and naked DNA

Are artificial vectors and their inserts available to free-living bacteria? If DNA is associated with the appropriate mobile elements in living bacteria, genetic exchange can proceed by conjugation or transduction, and thus reach a broad range of bacteria at a potentially high rate. Artificial cloning vectors are not mobile, however, and the laboratory microorganisms that propagate them are killed prior to their environmental release. Recombinant DNA will not be propagated in the environment unless free-living bacteria pick it up by natural transformation. Given that most of the mobile elements with which artificial vectors have homology are the very ones that carried antibiotic resistance genes across bacterial phyla, the product of an initial natural transformation event could then spread in a relatively unimpeded way.

Some bacteria, for example S. pneumoniae, Acinetobacter sp. Strain BD13, Pseudomonas stutzeri, N. gonorrhoeae (reviewed in Lorenz and Wackernagel, 1994), are naturally competent, that is they are able to accept naked DNA from the environment into their cells. The various experiments that show how homology facilitates natural transformation, also illustrate how readily these bacterial species can absorb naked DNA (Nielsen et al, 2000a, 2000b; de Vries et al, 2001; de Vries and Wackernagel, 2002). It is likely that some naturally competent bacteria carry elements homologous to artificial vectors because of the broad host-range of the elements to which pUC18 is homologous (Table 1; Matthew and Hedges, 1976).

There is tight homology between pUC18 and mobile elements in Neisseria species (Table 1), which are naturally competent. However, Neisseria (and Haemophilus) only take up DNA that is connected to specific uptake sequences (Solomon and Grossman, 1996). In Neisseria the specific uptake sequence required is 10 nucleotides long and there is no exact match to this in the pUC18 DNA sequence. In other species, such as Acinetobacter and P. stutzeri, DNA uptake is not sequence specific.

Several genes resembling those that encode the cellular machinery required for natural transformation exist in bacterial species that are not known for their natural competence, for example in E. coli, Salmonella typhi, Klebsiella pneumoniae, and P. aeruginosa (Finkel and Kolter, 2001). Some species are also known to be naturally competent in particular situations (Lorenz and Wackernagel, 1994; Demaneche et al, 2001b) and so could potentially absorb artificial DNA if it is available in their environment. For example, E. coli may become competent in mineral water, when hungry, when in human food, and when in soil that is struck by lightning (Bauer et al, 1999; Finkel and Kolter, 2001; Woegerbauer et al, 2002; Demaneche et al, 2001a), and Pseudomonas fluorescens is competent in soil but not in vitro (Demaneche et al, 2001b).

Could free-living bacteria be exposed to recombinant DNA in the environment? DNA can persist in the environment for thousands of years (Austin et al, 1997), though probably not in a form where the genes it may encode are readily utilizable. Experiments so far have demonstrated the transformation and unharmed activity of genes after a few days in soil (Nielsen et al, 1997; Sikorski et al, 1998).

There are no recommendations or guidelines on the disposal of recombinant DNA, only on the disposal of living genetically modified organisms (e.g. NIH guidelines for research involving recombinant DNA molecules (2002): Sections I-B and III-F-1, and Appendices G and K). Recombinant DNA is most likely poured down the sink in large and diverse quantities and genes can stay intact even when heated or chemically treated along with the killed microbial biomass from which they came (Andersen et al, 2001).

Natural selection

Most of the DNA that is inserted into artificial vectors would probably not benefit free-living bacteria. Without a phenotypic benefit to recipient bacteria, it is unlikely that non-homologous DNA picked up from artificial vectors would spread or persist (Campbell, 2000; Berg and Kurland, 2002), and it is unclear why such acquisitions would be of any concern. Given the large-scale genome sequencing of deadly microorganisms or the scope of research into virulence, however, there are almost certainly DNA inserts released in artificial vectors that would have a chance of persistence and would be of concern if acquired by some bacteria.

Even genetic variants that do not confer a beneficial trait sometimes spread among bacteria if linked to an advantageous locus (Guttman and Dykhuizen, 1994). The plasmids and transposons to which artificial vectors have homologies generally confer a whole suite of potentially beneficial traits, for example, multiple drug resistance or pathogenic characteristics (Table 1). Close linkage to such traits increases the chances of natural recombinants persisting in the environment even if newly acquired foreign DNA does not confer an immediate selective advantage on recipient bacteria.

Conclusions

Many factors influence the risk that the natural homology of artificial vectors could facilitate the bacterial uptake of recombinant DNA inserts. As several of these factors are notoriously difficult to quantify in an open environment, the degree to which this risk is realistic remains unknown. The processes involved occur at low frequencies, but could act on a large number of different molecules, bacteria and potential environmental situations. Given the mobility of the elements to which artificial vectors have homology, the acquisition of an undesirable trait that is beneficial to bacteria need only happen once for potentially far-reaching consequences. Most DNA that is inserted into artificial vectors would most likely pose no risk to human health or the environment even if acquired by free-living bacteria.