OUP user menu

Sexual isolation in bacteria

Jacek Majewski
DOI: http://dx.doi.org/10.1111/j.1574-6968.2001.tb10668.x 161-169 First published online: 1 May 2001


Bacteria exchange genes rarely but are promiscuous in the choice of their genetic partners. Inter-specific recombination has the advantage of increasing genetic diversity and promoting dissemination of novel adaptations, but suffers from the negative effect of importing potentially harmful alleles from incompatible genomes. Bacterial species experience a degree of ‘sexual isolation’ from genetically divergent organisms – recombination occurs more frequently within a species than between species. In this review, I outline the sources and mechanisms of sexual isolation within the context of selective pressures acting on different types of recombination events.

  • Recombination
  • Sexual isolation
  • Bacterium
  • Selection pressure

1 Introduction

Yes, bacteria do have sex – sex being defined here as any exchange of genetic material. Contrary to obligate sexual organisms, however, where sex is inseparably coupled with reproduction, bacteria are facultatively sexual. Bacteria can have sex when they ‘feel’ like it; the only price they pay is dictated by natural selection pressures and the lessons they have learned from those pressures during their evolutionary history. Not only can bacteria have sex, they are extremely promiscuous. They exchange genes rarely but the rules as to who an appropriate partner might be are not set in stone. There are documented cases of exchange of genetic material between bacteria, archea, plants and yeast [1].

Inter-specific recombination in bacteria is an important part of adaptive evolution [2,3]. Genetic exchange can modify existing alleles or introduce new genes into a population. The adaptive advantage of such events is opposed by the cost of bringing in maladaptive foreign alleles. It is expected that the frequency of recombination between species should depend largely on the cost versus benefit of acquisition of foreign genes.

It has been shown that the frequency of genetic exchange decreases rapidly as a function of relatedness of the DNA donor and recipient (e.g. [4,5,6,7,8]). We term this decrease of inter-specific, as compared to intra-specific recombination, sexual isolation. In this review, I examine the mechanisms of sexual isolation in the context of known recombinational systems, and discuss the possible reasons for the existence of barriers to genetic exchange between species.

2 Barriers to interspecies recombination

Exchange of genetic material in bacteria occurs through three mechanisms: natural transformation, conjugation and transduction. Transformation is a process by which some naturally competent species take up free DNA available in the environment. Conjugation is mediated by conjugative plasmids or transposons; after formation of a pilus resulting in physical contact between the donor and recipient cells, the donor chromosomal DNA may be co-mobilized by the conjugative element and transferred into the recipient. Transduction, mediated by bacteriophage, is usually a result of mispackaging of bacterial DNA into the phage capsid; upon infection of a new cell by the phage, the donor DNA is released into the new host. All of the above processes induce the action of the host recombination machinery [9,10,11], often resulting in integration of the donor DNA into the recipient chromosome.

A successful genetic exchange requires the following conditions to be satisfied: (1) availability of the donor cell or free donor DNA to the recipient; (2) uptake of donor DNA by the recipient cell; (3) escape of donor DNA from the recipient's restriction enzymes; (4) formation of a donor–recipient heteroduplex DNA molecule; (5) escape of the heteroduplex from the host mismatch repair system; (6) functionality of the donor gene product in the new genetic background. Sexual isolation between divergent strains results from a decreased probability of satisfying any one, or more, of the above conditions. The following sections discuss sexual isolation occurring at the steps described above, in the context of the three means of DNA acquisition.

2.1 Physical proximity

Each microbial cell capable of genetic recombination is most likely to be in close proximity of other cells, or free DNA, of its own strain or species. Even in highly dynamic cultures, where the bacterial colony structure is not maintained, the recombining DNA is most likely to originate from organisms sharing the same environment; that is, species that are ecologically similar. In addition, some recombining species are induced to recombination only in the presence of their own kind.

The naturally transformable Gram-positive bacteria Bacillus subtilis and Streptococcus pneumoniae are induced to competence by the secretion of extracellular pheromones [10,12]. The signaling molecules and their receptors are highly species- and even strain-specific. Hence, competence is highest in the presence of high concentrations of pheromone-producing cells of compatible pherotypes. As a result of this quorum-sensing mechanism, the environment of a competent cell will be biased towards containing homo-specific DNA.

In contrast, other naturally competent species do not exhibit the quorum-sensing behavior. The Gram-negative Neisseria gonorrhoeae is constitutively competent, and is able to take up DNA independently of its growth phase and external signals. Another Gram-negative bacterium, Haemophilus influenzae becomes competent under defined physiological conditions, but is not sensitive to external signaling [13]. In those species, the available DNA is likely to be constrained only by the ecological diversity of the environment.

Conjugation is highly limited by physical proximity, since cell-to-cell contact is required for DNA transfer. In natural habitats, one would expect a predominance of homo-specific matings. Species with the broadest habitat range are most likely to encounter a diversity of cells for inter-specific conjugation.

Finally, transduction is less dependent on physical proximity than conjugation, since virions are able to persist in the environment and protect the enclosed DNA from degradation. Once again, however, the majority of cells available for infection are likely to belong to the species present in the immediate proximity.

2.2 DNA entry

Some recombination systems are selective at the point of DNA uptake by the cell. If preferential uptake does occur, it is biased towards the uptake of DNA from the same, or closely related species.

The DNA receptors of H. influenzae and N. gonorrhoeae recognize specific oligonucleotides that were found to be overrepresented in the genomes of the respective organisms. The 11-bp uptake sequence in Haemophilus is found in about 600 (expected number by chance is <1) copies, occurring roughly every 4 kb [14]. The 10-bp uptake sequence of Neisseria is similarly overrepresented. Note that the 4-kb interval is smaller than the size of fragments typically taken up on transformation, allowing for efficient uptake of every segment of the genome. The uptake signal sequences are species-specific; the Neisseria signal is not recognized by Haemophilus and vice versa. However, it has been shown that the recognition system is ‘leaky’ and that uptake of non-specific DNA will occur in both species, albeit at a much lower frequency [15]. Some specificity for DNA uptake has also been demonstrated in Azotobacter vinelandii, Pseudomonas stutzeri, and Campylobacter coli [5,13].

In contrast, many other naturally transformable species do not exhibit uptake specificity. In the Gram-positive B. subtilis and S. pneumoniae [9], the Gram-negative Acinetobacter calcoaceticus [16], and the cyanobacteria, Synechococcus and Anacystisis nidulans, take up DNA non-preferentially from all sources [13]. In those systems, during transformation DNA divergence is not a barrier to uptake by the cell.

In conjugative systems, where cell–cell contact and fusion of cell membranes is necessary, presumably limits must exist as to how similar the two cells must be for DNA transfer to take place. Remarkably however, certain conjugative plasmids and transposons are known to possess very broad host-ranges. For example, the RK2 plasmid can replicate and mobilize genetic elements in both Gram-positive and Gram-negative species. Trieu-Cuot et al. [17] demonstrated the transfer of RK2-like plasmid from Escherichia coli to several Gram-positive species. The Ti plasmid of Agrobacterium tumefaciens is able to mobilize DNA and promote transfer from bacteria to plants [18]. The conjugative transposon Tn916 is found in both Gram-positive and Gram-negative species and has been demonstrated to transfer bi-directionally between several species across the subdivision [19]. Thus, even vast differences in cell wall and membrane composition are not insurmountable barriers to conjugation between species. Moreover, certain conjugative vectors survive and become replicated in their new host (although that condition is not necessary for a stable DNA exchange between cells). Scarce data are available as to how the frequency of conjugation varies within and between species. Investigation of Hfr matings between E. coli and Salmonella typhimurium found no decrease in the frequency of inter-specific matings as compared to homo-specific matings [20]. However it is expected that successful conjugation between highly divergent species occurs at significantly reduced frequencies.

Transducing phage enter the bacterial cell by attaching to the host surface receptors and are usually highly host-specific. Nevertheless, mutants defective for host-specificity exist even within normally ‘specialist’ phage populations, such as the E. coli phage P1. Besides transmission within the family Enterobacteriaceae, a transfer of P1 between E. coli and Myxobacterium xantus, a member of the delta-subdivision of Proteobacteria, has been documented [21]. In addition, broad-host-range phage have been detected in natural environments, and shown to be capable of generalized transduction between genera. A wild-type bacteriophage Sn-T, isolated from the sheathed bacterium Sphaerotilus natans, is able to infect members of Enterobacteriaceae, phototrophic bacteria and the Pseudomonadaceae. It can also mediate transduction in Pseudomonas and presumably other members of the above families [22]. Even so, of the three DNA exchange methods, the barriers to foreign DNA entry are strongest for transduction. Most of the well-studied phage are highly host-specific. Although broad-host-range-transducing phage exist in both Gram-positive [23] and Gram-negative species, their range is much more limited than that of conjugative elements. No phage-mediated transfer of genetic information between Gram-positive and Gram-negative bacteria has been documented.

2.3 Restriction enzymes

Numerous bacterial species harbor restriction–modification (R–M) systems. Restriction enzymes cut double-stranded DNA at specific sites, while modification enzymes either modify the host DNA to protect it from restriction or modify foreign DNA to target it for restriction. Short fragments of DNA resulting from restriction are easier targets for intracellular nucleases [24]. It is appealing to argue that R–M systems evolved specifically to protect the host from the damaging action of foreign DNA (although other theories, such as selfish evolution of those systems, exist [25]). The R–M is a general barrier to genetic exchange – it does not discriminate on the basis of how divergent the donor is from the recipient, only whether or not it carries a compatible DNA modification system. However, the R–M systems are surprisingly ineffective as a barrier to inter-specific recombination.

In most naturally transformable species, such as S. pneumoniae, B. subtilis, and A. calcoaceticus, DNA is transported across the cell as a single strand [9,13] rendering it inaccessible to restriction nucleases, which act only on double-stranded DNA. In H. influenzae, where donor DNA is taken up in double-stranded form, the DNA enters a protected state by being encased in a membrane vesicle [13,26], and hence restriction does not have a significant effect on homo-specific or hetero-specific transformation in Haemophilus [27]. In B. subtilis, it has been experimentally determined that strains harboring R–M systems experience only slightly reduced transformation rates as compared to non-restricting strains [28]. This reduction was of the order of a factor of 6, and is independent of the divergence of the transforming DNA from the recipient. In S. pneumoniae, not only is the transforming DNA taken up as a single strand, but it has been shown that in strains harboring the DpnII R–M system, the dpnA gene (encoding a single-stranded DNA methylase) is induced on competence, presumably in order to further protect the transforming DNA on integration [29]. Even though DNA integrated by homologous recombination is protected from restriction by the complementary, methylated host strand, highly mismatched or non-homologous DNA remains unpaired and may be subject to restriction when the other strand is replicated. Overexpression of the methylase and subsequent timely methylation of the donor strand could prevent such events [10].

In conjugative DNA transfer, the donor DNA enters the recipient cell as a single strand and is initially protected from restriction. However, once the complementary strand is replicated, the resulting double-stranded form, while unmodified by the host enzymes, is susceptible to restriction. Early investigations of R–M systems in E. coli suggested that restriction poses a major barrier to interspecies conjugation [30]. However, in a recent series of experiments investigating inter-specific conjugation between E. coli and related species, no detectable effect of the type-I R–M system on the frequency of marker recombination was found [20,31]. Matic et al. [11] propose that transient saturation of restriction due to large amount of chromosomal DNA mobilized by the F plasmid (up to hundreds of kilobases), could be responsible for the low efficiency of the R–M system. Alternatively, restricted DNA may retain recombinational potential, and produce a mosaic of integrated donor segments, rather than one continuous tract.

Again, of the three modes of DNA transfer, transduction is the one most severely affected by the host R–M system [32]. Transducing phage are double-stranded DNA phage and, after injecting their DNA into the host, are immediately subject to attack by restriction endonucleases. Edwards et al. [24] demonstrated that the frequency of infection by phages P22 and λ increases dramatically when the host restriction system is inactivated by heating. Different restriction systems exhibit varying efficiencies in preventing infection. The SEN system of Salmonella enterica is particularly potent, reducing infection of P22 by a factor of 3×103, and transduction by at least 102. However, many phages carry their own defenses against restriction including: inactivation of host endonuclease or methylase (phage T3, see [33]), or self-methylation [34]. Jensen et al. [22] found that a large fraction of broad-host-range phage isolated from the environment were resistant to type I and type II restriction endonucleases, most likely due to modification of viral DNA. Finally, it is not clear as to what the fate of the DNA transduced by the phage is. While restriction destroys the infective potential of the virus, restricted chromosomal DNA may remain recombinogenic. Zahrt and Maloy [6] argue that short DNA fragments are quickly degraded by exonucleases, such as the RecBCD enzyme but, as in the case of conjugation, retain some ability to recombine with the host chromosome.

2.4 DNA sequence divergence

Stable inheritance of foreign DNA usually requires recombination of the donor genetic material into the host chromosome. (Note that there may be some exceptions, as plasmids will be stably maintained in the presence of selective agents, and in time may co-evolve into a stable and selectively advantageous symbiosis with the host [35]). Integration may occur via homologous or illegitimate, non-homologous recombination. Non-homologous recombination is thought to be extremely infrequent [36], but may be responsible for introduction of entirely novel elements into the recipient genome [1]. Being mediated by phage or transposon encoded integrases, and possibly the repair of double strand breaks, non-homologous recombination is not dependent on DNA sequence divergence. Homologous recombination is the most common mode of integration. The fragments integrated during homologous recombination may contain insertions, deletions, or even novel genes present within an otherwise homologous DNA segment [37]. Hence homologous recombination may mediate introduction of novel genetic elements, even though the most widespread effect is probably the modification of already existing genes. In all known systems, the frequency of homologous recombination decreases considerably with DNA sequence divergence between the donor and the recipient [4,5,6,8,31,38].

In B. subtilis transformation, sexual isolation, measured as the ratio of homogamic (using the recipient's own DNA) to heterogamic (using divergent donor DNA) transformation frequency, increases exponentially with DNA sequence divergence [8]. This exponential relationship seems to be ubiquitous in recombinational systems throughout the bacterial world. A similar relationship was observed in the natural transformation of S. pneumoniae with DNA from related Streptococcus strains and species [38]. Recent experiments with the naturally transformable Gram-negative species P. stutzeri [5], show a rapid increase of sexual isolation with sequence divergence (2–3 exponential units, for 10% DNA sequence divergence). The sexual isolation in P. stutzeri may plateau at high DNA sequence divergence, but experiments with more divergent strains will be necessary to verify that hypothesis. In conjugation between E. coli, S. typhimurium and related enteric species, Vulic et al. [31] found the same, approximately exponential relationship. In the yeast Saccharomyces cerevisiae, Datta et al. [39] demonstrated that the frequency of mitotic crossovers decreased exponentially as a function of the number of mismatches in the recombining sequences. Finally, in P22 phage-mediated transduction between two very closely related species, S. typhimurium and Salmonella typhi, variation in DNA as low as 1–2% is able to reduce the frequency of DNA integration by a factor of 106[6].

The resistance to integration caused by mismatched DNA is most likely due to difficulty in two subsequent stages of recombination: the formation of a recombinant joint, and escape from the host mismatch repair system. The molecular steps of genetic recombination have been studied extensively in E. coli and S. cerevisiae and to a lesser extent in other systems. In E. coli, initial processing of the donor DNA is necessary to produce free single-stranded ends. The RecBCD enzyme digests and unwinds double-stranded DNA until a characteristic χ sequence is encountered. As a result, a free single-stranded 3′ DNA end is produced. This free end is invasive and is able to displace one of the recipient strands and initiate the recombination process (see [40], for discussion). In Bacillus transformation, no such pre-processing is necessary since the DNA enters the cell as a single strand and is immediately recombinogenic. In fact, no homolog of the RecD protein, which is necessary for exonuclease activity, is found in Bacillus. All subsequent recombination functions are dependent on the function of the RecA protein. The recombinant joint is subject to extension, branch migration mediated by the RuvA and RuvB proteins, and finally editing. In E. coli, the editing process involves the methylation-directed mismatch repair system, controlled by the MutS and MutL proteins. In S. pneumoniae, the editing function is carried out by the homologous, nick-directed HexAB mismatch repair system. If mismatches are detected, the mismatch repair system may reject the entire donor strand and abort the recombination process.

2.4.1 Recombinant joint formation

It has been demonstrated that a short segment of near-identity between the donor and recipient is necessary to initiate recombination [41]. Once a stable joint is formed, branch migration is believed to proceed relatively easily through highly mismatched regions. The near-identity requirement is probably due to RecA-mediated homology search criteria, and stability of the resulting recombinant joint. Majewski and Cohan [42] found a preference for recombination within relatively high GC content stretches, suggesting that a thermodynamic binding threshold may need to be satisfied. In E. coli the length of the initiating sequence is about 27 bp [41]. In Bacillus two mismatch free regions of 20 bp or more are required at both end of the donor strand [42]; most likely one end is necessary for initiation, while the other end for successful resolution of the recombinant molecule. In Streptococcus, the total length of identity has been estimated as at least 27 bp [38]; it is not known whether one or two flanking regions of identity are required. It can be shown [31,43] that the probability of encountering such a mismatch-free region at the invasive end of a donor DNA strand decreases exponentially as a function of DNA sequence divergence. Thus, the requirement for a stable recombinant joint may be responsible for the widely observed exponential relationship between sexual isolation and sequence divergence.

In B. subtilis, Majewski and Cohan [42] demonstrated that by transforming donor strains with PCR-amplified constructs consisting of a divergent donor segment flanked by regions of identity with the recipient, sexual isolation is almost totally eliminated. A similar result has been obtained with classical genetic methods by Harris-Warrick and Lederberg [44]. Hence, in Bacillus, the major barrier to interspecies DNA exchange by transformation is the resistance to initiate recombination between divergent sequences. A comparable effect was observed in S. pneumoniae transformation, where 34% of sexual isolation is caused by the mismatch repair system (see below), while the remainder is most likely the result of difficulty in forming the recombinant joint [38]. In other experimental organisms, such as E. coli and the yeast S. cerevisiae, at least some of the resistance to recombination cannot be explained by the action of mismatch repair system and DNA editing, and is likely to be caused by the instability of mismatched DNA [39,40].

2.4.2 Mismatch repair system

The mismatch repair system has been proven effective in correcting single mismatches resulting from replication errors and recombination in both Gram-positive and Gram-negative bacteria [7,45]. It is also effective in preventing recombination between mismatched sequences in S. cerevisiae [46].

However, in the studied transformable species, mismatch repair has been shown to be a poor barrier to inter-specific recombination. In Bacillus, inter-specific recombination rates are not highly dependent on the presence or absence of the mutS and mutL genes, the essential components of mismatch repair [43]. At most 16% of sexual isolation may be attributed to mismatch repair. In Streptococcus, the HexAB mismatch repair system is believed to be easily saturated by multiple mismatches [10,37,38,47]. It appears to be particularly ineffective when transforming DNA has intermediate levels of divergence (about 5%) from the recipient, but at all divergence levels is much less effective than would be expected from its ability to correct single mismatches in homo-specific transformation. Majewski et al. [38] estimated that in hetero-specific transformation of S. pneumoniae, not more than 34% of observed sexual isolation could be attributed to mismatch repair.

Inter-specific conjugation between E. coli and S. typhimurium (about 16% overall DNA sequence divergence) is much more dependent on the action of mismatch repair [7]. In fact, almost the entire barrier to inter-specific recombination is due to mismatch repair, since mating frequency is unaffected by divergence and restriction is not a significant factor. Removal of the recipient mismatch repair increases the frequency of successful recombination by a factor of 103–104, while overexpression of mismatch repair proteins may further reduce recombination by 102, as compared to wild-type strains [20,31]. In comparison, other factors, such as the stability of the recombinant heteroduplex molecule, account for only a 10-fold reduction in hetero-specific, as compared to homo-specific, recombination [40].

Transduction may be severely affected by DNA sequence divergence. A 1–2% sequence divergence between S. typhi and S. typhimurium reduces P22-mediated recombination frequency by a factor of 106[6]. It is unlikely that one mismatch in 100 bp has such a significant affect on donor strand invasion and stability of the recombinant joint. The strong recombination barrier is caused by the combined action of the mismatch repair system and the RecBCD enzyme [6]. Inactivation of mismatch repair results in up to 103-fold increase in transduction rates, while mutations in the recD gene increase transduction by 101–102. Double mutants for both the mutS and recD genes experience a 106-fold increase in inter-specific recombination. It is believed that mismatches slow down the branch migration process, and the mismatch repair system has a role in aborting recombination. The liberated donor DNA molecule is then subject to degradation by the RecBCD endonuclease, effectively preventing initiation of recombination at the same locus.

In contrast, a recent study of PBS1-mediated transduction between B. subtilis and Bacillus mojavensis (4.6% DNA sequence divergence at the rifampicin resistance marker locus), Feldgarden and Cohan (unpublished results) found only a 3-fold reduction in interspecies as compared to intraspecies transduction. This may be the result of the Gram-positive mismatch repair system (as illustrated by Streptococcus and Bacillus) being generally less effective as a source of sexual isolation than its Gram-negative counterpart (as seen in Enterobacteriaceae).

2.5 Functional compatibility of donor DNA with the recipient genome

Even after successful integration, the donor gene must be expressed and be compatible with the host genetic makeup in order to produce viable recombinant cells and survive selection pressures. In general, genes that have co-evolved with other components of their respective host genomes will be poorly adapted to interact with divergent genomes. Inter-specific hybrids (e.g. those resulting from conjugation between E. coli and S. typhimurium, M. Vulic, personal communication) have invariably reduced fitness under laboratory conditions. Similar reduction in fitness is likely to occur in nature. The reduction in fitness is most likely to depend on the phylogenetic and ecological distance between recombining species. Recombination allows each competent cell access to the entire spectrum of mutations present within the DNA pool. Many of those mutations may be selectively advantageous. However, as the distance between species increases, the possibility of a deleterious or maladapted DNA segment co-recombining with the beneficial one also increases. Reasonably strong selection pressures must exist in order for any single recombinant to be fixed in a population. Those pressures must be much stronger for an interspecies recombinant to become fixed. Predictably, fragments integrated in inter-specific recombination, and particularly transformation, are generally small [2,48]. The negative selection pressures are likely to be lowest for transformation (recombining fragment size is smaller than 10 kb), intermediate for transduction (up to 100-kb donor fragments), and highest for conjugation (fragments of several hundred kilobases may be transferred). Nevertheless, when positive selection pressures are strong, as can be illustrated by the spread of antibiotic resistance genes (e.g. the tetracycline resistance gene carried by the Tn916 transposon, [49]) DNA can be successfully and stably transferred even between the Gram-positives and Gram-negatives.

3 Why does sexual isolation exist in bacteria?

The frequency of gene exchange in bacteria is low. Even in the naturally competent B. subtilis, the rate of recombination has been estimated at 10−7 events per gene segment (∼1 kb) per genome per generation [2]. However, the influence of recombination on bacterial evolution is considerable. In natural bacterial populations, a single nucleotide change is about 50 times more likely to occur through recombination than mutation [50,51]. Thus, it should be expected that the mechanisms of DNA exchange and sexual isolation have evolved in order to balance the benefit of bringing in new alleles and genes, and the negative cost of foreign alleles disturbing the integrity of local genome adaptations. We may ask whether the observed sexual isolation is in fact an active mechanism for preventing excessive interspecies recombination or simply an inevitable consequence of differences between cells and DNA substrates participating in genetic exchange.

Transduction and conjugation are primarily functions of extrachromosomal agents: phage, plasmid and transposons. They experience negative selection pressures other than the disadvantage of acquiring maladaptive foreign alleles. Transduction is the most costly process. Phage kill bacteria. The phage–bacteria relationship shows signs of a classic arms race, evolution of weapons and defenses. It is most probable that R–M systems have evolved as a defense to phage infection (or plasmid infection). One restriction cut suffices to kill a phage, while it is less effective against co-transduced DNA. The phage, in turn, have developed numerous mechanisms to escape restriction: loss of restriction sites [33], self-methylation and modification, activation of host modification, or inhibition of host restriction enzymes [34]. The situation is somewhat similar in case of plasmids and conjugative elements. Even though such elements are not immediately destructive to the host, they do carry a cost associated with the replication of extra DNA and the expression of extra genes.

‘Selfish’ DNA elements may evolve independently of bacterial chromosomes, and often as antagonists of their host. Host DNA exchange resulting from their activities is likely to be incidental [3]. In contrast, natural transformation is fully encoded by the host chromosomal genes, and may have evolved specifically as a machinery for acquisition of new genes [3,9] (for contrasting views see [52,53]).

The cost of acquiring new genes by transformation is low, relative to transduction and conjugation. It is solely due to the potentially maladaptive value of foreign alleles, versus both the destructive or costly actions of phage/plasmid and the disadvantage of acquiring foreign alleles in transduction/conjugation. In addition, DNA fragments integrated upon transformation are small and have a low probability of co-transforming costly alleles [48]. During the uptake process, exogenous donor DNA is broken down into fragments smaller than 10 kb [9]. It is possible that such an uptake mechanism evolved in response to the advantage of integrating small fragments of foreign DNA and minimizing the negative cost of co-integration of several genes.

Natural transformation employs other mechanisms for actively reducing host barriers to recombination: protection from restriction of transforming DNA in transformasomes in Haemophilus and induction of the dpnA-encoded methylase in Streptococcus. In Streptococcus and Bacillus, mismatch repair has been shown to be ineffective as a source of sexual isolation. While this has not been demonstrated in other transformable species, judging by similar observed frequencies of hetero-specific transformation (Table 1), mismatch repair is not likely to be a major source of sexual isolation in Haemophilus and Pseudomonas. In Bacillus, the barriers to inter-specific transformation act almost exclusively at the DNA integration stage, and may be entirely overcome by providing the foreign strand with regions of identity necessary for initiation of recombination [42,44]. No further manipulation of host enzymes is necessary suggesting that the barriers to inter-specific recombination may already be at a minimum level dictated by the thermodynamics of homologous DNA pairing [42].

View this table:
Table 1

Comparison of sexual isolation across species and modes of DNA transfer

Recombination systemSpecies used recipient/donorSexual isolationaDivergence at 16SrRNAb (%)Divergence at marker gene (%)Reference
Transformation in HaemophilusH. influenzae/Haemophilus aphrophilus1035.7<14.8d[4]
Transformation in PseudomonasP. stutzeri/Pseudomonas mendocina1032.89.5c[5]
Transformation in StreptococcusS. pneumoniae/Streptococcus sanguis1032.115.1c[37]
Transformation in BacillusB. subtilis/Bacillus licheniformis1032.015.4c[8]
Transformation in BacillusB. subtilis/B. mojavensis30.34.6c[8]
Transduction in BacillusB. subtilis/B. mojavensis30.34.6cFeldgardenf
Transduction in SalmonellaS. typhimurium/S. typhi1060.5∼1–2e[6]
Conjugation in EscherichiaE. coli/S. typhimurium1052.0∼16e[30]
  • aSexual isolation is defined as the ratio of the frequency of homogamic recombination (i.e. using the recipient's own DNA) and heterogamic recombination (using the divergent donor DNA).

  • bDetermined from 16SrRNA sequences available from Entrez, http://www.ncbi.nlm.nih.gov.

  • cEstimated from complete or partial sequences of the marker locus in the original citations.

  • dEstimated from Entrez sequences of a 1750-bp interval spanning rpsL (the streptomycin resistance locus) and efg of H. influenzae and Haemophilus ducreyi. The H. aphrophilus sequence is not available, but judging by relative divergence at 16SrRNA it is likely to be slightly less divergent from H. influenzae than is H. ducreyi.

  • eWhole genome estimates based on sequences of several available genes, from original citations. Whole genome divergence may be more appropriate than marker-specific divergence for transduction and conjugation, since integrated fragments are large.

  • fFeldgarden and Cohan, unpublished results.

Transformable species exhibit some preference for acquiring DNA from closely related organisms. The quorum-sensing mechanisms of Streptococcus and Bacillus, and the specific uptake of DNA in Haemophilus and Neisseria suggest that transformation is most favorable within a closely related group, and may be more costly outside of it. Studies of recombinant genes in Neisseria meningitidis and S. pneumoniae [10,54] show that frequent exchange of small genetic fragments between closely related strains results in variability in capsular types, whereas exchange with related species results in formation of mosaic penicillin resistant pbp genes, demonstrating immediate advantages of such exchanges between close relatives.

Nevertheless, even such preference for closely related DNA may be viewed as a positive force for acquiring helpful genes, rather than a negative selection against incorporating detrimental ones. All subsequent barriers to recombination during transformation are fairly inefficient: restriction enzymes are ineffective and not selective against foreign DNA; mismatch repair is much less active than in conjugation; recombinant joint formation may be limited only by thermodynamics; incorporation of small DNA fragments minimizes functional incompatibility. In view of the above, it is reasonable to propose that for naturally competent species, sexual isolation is an incidental result of the processes of recombination. Sexual isolation between species does exist, but it is unlikely to be the result of natural selection against acquisition of foreign DNA.

The situation appears different in non-transformable species, such as E. coli. The barriers to inter-specific exchange are much stronger than in naturally competent species (see Table 1 for comparison). In fact, it has been suggested, that the mismatch repair system may have evolved specifically as a method of sexual isolation [11,20]. It is possible that the greater costs of acquiring foreign genes through conjugation and transduction, such as incompatibility of the large DNA fragments and infection by novel selfish DNA elements, have resulted in actively increased sexual isolation in non-transformable species. It is worth noticing that in B. subtilis both transformation and transduction experience the same, low level of resistance to inter-specific recombination (Table 1, Feldgarden and Cohan, unpublished results). A possible explanation is that, as a naturally transformable species, Bacillus has adapted to efficiently integrate foreign DNA. The low level of resistance to transduction may be the result of recombination mechanisms evolved for the purpose of transformation.

While it is tempting to define general rules for the nature of sexual isolation in the bacterial world, a detailed look at distinct examples and mechanisms of isolation shows that the barriers to genetic exchange range from weak and most likely incidental (B. subtilis) to strong and deliberate (E. coli, S. typhimurium). Clearly, our full understanding of sexual isolation in the bacteria world must include the context of ecology, means of DNA acquisition, and the extent to which each species depends on genetic exchange in its evolution.


I would like to thank Frederick Cohan for critical reading of the manuscript and Mike Feldgarden for sharing their latest, unpublished results. I would also like to thank Marin Vulic for help with understanding some of the aspects of his research. This review was made possible by the Human Genome Institute Research Grant HG00008.


  1. [1].
  2. [2].
  3. [3].
  4. [4].
  5. [5].
  6. [6].
  7. [7].
  8. [8].
  9. [9].
  10. [10].
  11. [11].
  12. [12].
  13. [13].
  14. [14].
  15. [15].
  16. [16].
  17. [17].
  18. [18].
  19. [19].
  20. [20].
  21. [21].
  22. [22].
  23. [23].
  24. [24].
  25. [25].
  26. [26].
  27. [27].
  28. [28].
  29. [29].
  30. [30].
  31. [31].
  32. [32].
  33. [33].
  34. [34].
  35. [35].
  36. [36].
  37. [37].
  38. [38].
  39. [39].
  40. [40].
  41. [41].
  42. [42].
  43. [43].
  44. [44].
  45. [45].
  46. [46].
  47. [47].
  48. [48].
  49. [49].
  50. [50].
  51. [51].
  52. [52].
  53. [53].
  54. [54].
View Abstract