Genetica 116: 359–370, 2002.© 2002 Kluwer Academic Publishers. Printed in the Netherlands.

359

Sexual isolation and speciation in bacteria

Frederick M. CohanDepartment of Biology, Wesleyan University, Middletown, CT 06459-0170, USA (Phone: +1860-685-3482; Fax:+1860-685-3279; E-mail: fcohan@wesleyan.edu)

Received 18 March 2002 Accepted 3 June 2002

Key words: horizontal transfer, species concept, recombination

Abstract

Like organisms from all other walks of life, bacteria are capable of sexual recombination. However, unlike mostplants and animals, bacteria recombine only rarely, and when they do they are extremely promiscuous in theirchoice of sexual partners. There may be no absolute constraints on the evolutionary distances that can be traversedthrough recombination in the bacterial world, but interspecies recombination is reduced by a variety of factors,including ecological isolation, behavioral isolation, obstacles to DNA entry, restriction endonuclease activity,resistance to integration of divergent DNA sequences, reversal of recombination by mismatch repair, and functionalincompatibility of recombined segments. Typically, individual bacterial species are genetically variable for most ofthese factors. Therefore, natural selection can modulate levels of sexual isolation, to increase the transfer of genesuseful to the recipient while minimizing the transfer of harmful genes. Interspecies recombination is optimizedwhen recombination involves short segments that are just long enough to transfer an adaptation, without co-transferring potentially harmful DNA flanking the adaptation. Natural selection has apparently acted to reducesexual isolation between bacterial species. Evolution of sexual isolation is not a milestone toward speciation in bac-teria, since bacterial recombination is too rare to oppose adaptive divergence between incipient species. Ironically,recombination between incipient bacterial species may actually foster the speciation process, by prohibiting oneincipient species from out-competing the other to extinction. Interspecific recombination may also foster speciationby introducing novel gene loci from divergent species, allowing invasion of new niches.

Introduction

The earth is a highly sexual planet. Species in all oflife’s major divisions undergo sexual recombination,at least occasionally (Cohan, 2002). However, sex isnot the same across all of life. While sex in mostanimals and plants is obligately tied to reproduction,and occurs only between very close relatives (i.e., usu-ally members of the same species), this pattern is nottypical of microbial eukaryotes, bacteria, and viruses.For example, bacteria exchange genes only rarely, andwhen they do, they are much less fussy about theirchoice of sexual partners than are plants and animals(Cohan, 2001, in press). As I will demonstrate, therare but promiscuous pattern of recombination seenin bacteria has profound consequences for bacterialadaptation and speciation.

Mechanisms of recombination

Bacteria can recombine by three parasexual mecha-nisms: transduction, conjugation, and transformation(Madigan, Martinko & Parker, 1999). These mecha-nisms introduce DNA unidirectionally, from a donorcell to a recipient cell, and the transferred DNA is gen-erally a small fraction of the genome. None of thesemechanisms is tied to reproduction, so recombinationis much less frequent in bacteria than in most plantsand animals (Cohan, 2001).

In most cases, DNA is integrated into the re-cipient genome through homologous recombination,in which the donor DNA replaces the homologoussegment in the recipient. Occasionally, bacterial re-combination can introduce donor DNA that has nohomologous counterpart in the recipient (heterologous

360

recombination). This can occur through the machineryof homologous recombination, provided that a heter-ologous donor segment is flanked by DNA thatis homologous with the recipient (Harris-Warrick &Lederberg, 1978; Hamilton et al., 1989; Majewski& Cohan, 1999). Alternatively, donor DNA with nohomology anywhere to the recipient may integrateinto the recipient chromosome if it contains an inser-tion sequence, which can integrate anywhere in therecipient chromosome (Van Spanning, Reijnders &Stouthamer, 1995; Ochman, Lawrence & Groisman,2000; Shoemaker et al., 2001).

Transduction is genetic exchange mediated by bac-teriophages (bacterial viruses) (Madigan, Martinko &Parker, 1999; Miller, 2001). Transduction begins dur-ing a viral infection, when a genomic segment fromthe infected cell (the donor) becomes mistakenly pack-aged into the virus. When the virus later infects an-other cell (the recipient), the DNA from the donor cellis injected into the recipient’s cytoplasm. The donorDNA can then be incorporated into the recipient’sgenome by homologous or heterologous recombina-tion. Transduction can transfer a length of donor DNAabout as large as the phage genome, up to severalhundred kilobases.

In conjugation, genetic exchange is mediatedby plasmids, which are self-replicating circles ofDNA that infect bacteria as endosymbionts (Madigan,Martinko & Parker, 1999). Recombination betweenbacteria becomes possible when DNA from a donor’sgenome becomes integrated into an infecting plasmid.The plasmid may then promote its own transfer toanother cell, introducing donor DNA into a potentialrecipient. Plasmid conjugation can transfer larger seg-ments of the genome than is possible in transduction.In both transduction and conjugation, the recipient cellplays a passive role in the uptake of donor DNA. Aswe shall see, this limits a bacterial population’s op-tions for optimizing the level of sexual isolation intransduction and conjugation.

In contrast, the recipient plays an active role in theuptake of DNA in transformation. In transformation,free DNA from the environment is taken up throughpathways coded by the recipient’s genome, and is thenintegrated into the recipient’s chromosome (Dubnau,1999). Some species, such as Neisseria gonorrheae,are always competent to act as recipients of trans-formation (Sparling, 1966; Dubnau, 1999); in gram-positive bacteria, such as Bacillus and Streptococcus,competence must be induced by environmental con-ditions, usually through starvation (Dubnau, 1999).

Transformation introduces very small segments of freeDNA from the environment, up to 1% of the genome,into the recipient (Dubnau, 1999).

While only a minority of species appear to be com-petent to act as recipients of transformation (Lorenz &Wackernagel, 1994), all bacterial species are likely tobe sexual. This is because all bacteria studied are sub-ject to viral infection (Madigan, Martinko & Parker,1999), and mistakes in DNA packaging that lead totransduction are probably inevitable.

Bacterial recombination is rare and promiscuous

One could, in principle, estimate recombination ratesfor bacteria in nature from experiments in the labora-tory. However, the rate of recombination is sensitiveto many environmental parameters, and so it wouldbe difficult to determine a typical rate of recombin-ation in nature from experiments in a limited set ofenvironmental conditions (Cohan, 2002). Instead, mi-crobiologists have taken a ‘retrospective’ approach, inwhich average recombination rates from the past areestimated from the patterns of DNA sequence and al-lozyme variation in contemporary natural populations.The rationale is that low recombination rates can be in-ferred when alleles at different loci show high linkagedisequilibrium, or when different DNA segments yieldcongruent phylogenies (e.g., Hey & Wakeley, 1997).

The retrospective approaches have shown that re-combination in most bacterial species occurs at aboutthe rate of mutation or somewhat higher (Selander &Musser, 1990; Maynard Smith et al., 1993; Roberts& Cohan, 1995; Feil et al., 1999, 2000b). Less com-monly, as in the cases of Helicobacter pylori and N.gonorrhoeae, recombination occurs at least an order ofmagnitude more frequently than mutation (O’Rourke& Spratt, 1994; Suerbaum et al., 1998), but surveymethods do not allow us to determine by how much.We should note that studies of recombination rates innature have been limited largely to human pathogens(with the exception of free-living Bacillus species), sogeneralizations may be limited.

While bacteria do not recombine frequently, theyare promiscuous in their choice of partners in geneticexchange. Bacteria can undergo homologous recom-bination with organisms that differ by at least 25%in DNA sequence (Duncan et al., 1989; Spratt et al.,1992). Moreover, genomic analyses show that a sub-stantial fraction of gene loci in a typical bacteriumhave been acquired from extremely divergent taxa

361

(i.e., from different bacterial divisions, which are com-parable to the most divergent groups of eukaryotes)through heterologous recombination (Rivera et al.,1998; Jain, Rivera & Lake, 1999; Ochman, Lawrence& Groisman, 2000). There appear to be no abso-lute constraints on the evolutionary distances that canbe traversed through recombination in the bacterialworld. Nevertheless, recombination between divergentbacteria is reduced severely by several factors.

Mechanisms of sexual isolation

Ecological isolation. As in the case for animals andplants, ecological differences can reduce the rate of re-combination between bacterial populations, especiallywhen the populations inhabit different microhabitats.The need for physical proximity is most stringent inthe case of conjugation, since this process requirescell–cell contact (Ippen-Ihler, 1989; Majewski, 2001).Physical proximity of donor and recipient also fostersrecombination by transformation, since free DNA re-quired for transformation has limited longevity inthe environment (Leff, McArthur & Shimkets, 1992;Majewski, 2001). Transduction is probably the leastconstrained by physical distance between recipientsand donors, since bacteriophages are often stable inthe environment (Stotzky, 1989).

Ecological differences between populations canseparate populations in time as well as space. Forexample, one ecotype of the human pathogen Strep-tococcus pyogenes is adapted to infecting our throats,primarily in winter, while another ecotype is adaptedto infecting our skin, usually in summer (Bessen et al.,2000).

Behavioral isolation. Sexual isolation can resultfrom differences in behavior analogous to courtship inanimal reproduction. In gram-positive bacteria such asBacillus (Dubnau, 1999) and Streptococcus (Morrison& Lee, 2000), competence for transformation is in-duced by a ‘quorum sensing’ mechanism. Under cer-tain growth conditions, cells produce a pheromone,and competence for transformation is induced onlywhen this pheromone reaches a critical level. There isremarkable specificity in pheromones produced withinBacillus subtilis and Streptococcus pneumoniae, suchthat the pheromones produced by a strain’s close rela-tives frequently do not elicit competence (Pozzi et al.,1996; Tortosa et al., 2001). Therefore, a minoritystrain growing amongst members of another species

(or even other genotypes of the same species, in somecases) is less likely to become competent than the ma-jority strain. Nevertheless, the quorum requirementwill not hinder a majority strain from taking up DNAfrom other strains or species growing in a minority.

Obstacles to DNA entry. In animals and plants, post-mating, pre-zygotic mechanisms prevent a foreignmale gamete from fertilizing a female gamete. Like-wise in bacteria, DNA may be prevented from enteringa potential recipient. In gram-negative transformationsystems, a donor must have a short, taxon-specific up-take sequence to be admitted efficiently into a recipientcell, although the requirement is not absolute (Danner,Smith & Narang, 1982; Goodman & Scocca, 1988).Gram-negative species that are competent to act as re-cipients in transformation are also adapted to act asdonors – the taxon-specific uptake sequences appearabout every 4 kb throughout the genome, much morefrequently than would be expected by chance (Dan-ner, Smith & Narang, 1982). Uptake sequences are notrequired for gram-positive transformation.

In the case of vector-mediated recombination (i.e.,transduction and conjugation), entry of donor DNA islimited by the host range of the vector. Most bacterio-phages can infect strains from only a single speciesor closely related species (Mazodier & Davies, 1991);so transduction is limited to recombination betweenclose relatives. Likewise, most plasmids are thought tohave a narrow host range (Mazodier & Davies, 1991).However, some plasmids have an extremely broad hostrange and are capable of facilitating recombinationacross divisions of bacteria. Most likely, recombina-tion between bacterial divisions has been mediated bybroad-range plasmids (Mazodier & Davies, 1991).

Restriction endonuclease activity. Many bacterialspecies have what amounts to a simple ‘immune’ sys-tem: they produce restriction endonucleases, whichcleave foreign DNA at specific target sequences. Forexample, some strains of Staphylococcus aureus con-tain the restriction enzyme Sau3AI, which cleavesforeign DNA at every 5′-GATC-3′ sequence (Roberts& Macelis, 1992). The cell’s own DNA is protectedby genome-wide modification, often methylation, ofthe target sequence. Thus, if incoming DNA is notproperly modified, the native restriction enzymes arelikely to cleave it.

Restriction endonucleases provide effective pro-tection against bacteriophages and plasmids: cleavageof virus or plasmid DNA kills the infectious particle

362

(Bickle & Kruger, 1993; for a different view, seeRocha, Danchin & Viari, 2001). However, restrictionis only modestly effective as a means of sexual isola-tion, especially in transformation. In the case of B.subtilis, only a minority of strains has restriction activ-ity, and restriction reduces the transformation rate byonly a factor of 2–6 (Cohan, Roberts & King, 1991).Restriction has limited effectiveness in gram-positivetransformation in part because DNA enters the recip-ient in single-stranded form, and restriction enzymesattack only double-stranded DNA. The apparent re-duction in transformation is likely due to cleavageof the recipient’s own chromosome after the donorDNA is integrated (because only one strand of thedonor–recipient heteroduplex is methylated), resultingin the killing of many transformants (Cohan, Roberts& King, 1991; Majewski, 2001).

The potential effect of restriction on interspecificrecombination is actively reduced in Streptococcustransformation (Lacks et al., 2000). A DNA methylasein Streptococcus efficiently methylates single-strandedDNA, and this protein is hyper-expressed upon induc-tion of competence. The cell is thus protected fromkilling itself following integration of foreign DNA(Majewski, 2001).

While restriction protects a cell from plasmid in-fection, it has little effect on the rate of recombinationmediated by infecting plasmids (Vulic et al., 1997).Perhaps cleavage of a plasmid by restriction enzymes,while killing the plasmid, creates a large number ofsmall, recombinogenic fragments (Majewski, 2001).

Integration of donor DNA into the recipient chromo-some. After donor DNA enters the recipient cell, thenext step in recombination is integration of the donorDNA into the recipient chromosome. Here a singledonor strand replaces the homologous strand in therecipient, and the complementary recipient and donorstrands form a heteroduplex (Majewski, 2001).

If integration is to occur, there must be a near iden-tity between the recipient and donor at one or bothends of the donor segment, depending on the taxon.In the case of Bacillus transformation, the terminal18–24 base pairs at both ends of the donor strandmust be nearly identical, such that the heteroduplexDNA at each end is highly stable thermodynamic-ally (Majewski & Cohan, 1999; Majewski, 2001). InEscherichia coli conjugation, near identity of about 27base pairs is required at only one end of the donorsegment. Because the donor and recipient must benearly identical over a short sequence, the interspecific

recombination rate decays exponentially with increas-ing sequence divergence, as observed in Bacillus,Streptococcus, and Escherichia (Vulic et al., 1997;Majewski & Cohan, 1998; Majewski et al., 2000). Inthe case of Bacillus transformation, for example, each5% increment in sequence divergence yields about a10-fold decrease in recombination (Roberts & Cohan,1993; Zawadzki, Roberts & Cohan, 1995).

Reversal of recombination by mismatch repair. Mis-match repair enzymes are adaptations for reduc-ing mutations that occur during DNA replication(Schofield et al., 2001). After replication, mismatchrepair enzymes resolve the physical irregularity inDNA caused by a mismatch between the parentalstrand and the daughter strand. A stretch of severalhundred base pairs flanking the mismatch bulge is re-moved from the newly synthesized daughter strand(targeted because it is not yet methylated by the cell’sDNA modification system), and the gap is replacedusing the parental strand as a template for synthesis.

In some species, such as E. coli and Salmonellaenterica, mismatch repair acts as a powerful force ofsexual isolation between species. Each mismatch de-tected between the heteroduplex strands is perceivedas a replication error, and stimulates removal of sev-eral hundred base pairs from one strand. The strandremoved is usually the donor strand, since it is notyet methylated by the recipient’s modification sys-tem. Knocking out mismatch repair in these speciesincreases inter-species recombination by a factor of1000, and over-expression of mismatch repair re-duces inter-specific recombination by a factor of 100(Rayssiguier, Thaler & Radman, 1989; Matic, Taddei& Radman, 2000). Mismatch repair is responsible fornearly all resistance to recombination between entericspecies.

In gram-positive transformation, mismatch repairhas only a modest effect on sexual isolation. Only16% of sexual isolation between Bacillus species and34% of sexual isolation between Streptococcus speciesis ascribed to mismatch repair (Majewski & Cohan,1998; Majewski et al., 2000).

Functional incompatibility. As in eukaryotes, for-eign alleles are generally expected to have deleteriousfitness effects in bacteria (Majewski, 2001). If the locitransferred are involved in niche-specific adaptation,then recombination may debilitate the specific adapta-tion of the recipient. Even when foreign alleles arenot involved in niche-specific adaptation, they may be

363

incompatible with the genetic background of the re-cipient. The effect of interspecific recombination onfitness is little studied in bacteria, but Vulic (pers.comm.) has recently shown that interspecific hybridsbetween E. coli and Salmonella are frequently less fitthan either species in laboratory culture.

Evolution of sexual isolation

Can sexual isolation in bacteria evolve? We begin witha survey of intraspecific genetic variation in sexualisolation, and then consider how natural selectionmight act on this variation.

Genetic variation in sexual isolation. Strains withina species vary considerably in their levels of sexualisolation from other taxa, owing to variation in fourfactors of sexual isolation. First, strains within aspecies typically differ in the restriction–modificationsystems they bear (Roberts & Macelis, 1992). Forexample, a population survey of B. subtilis and B.mojavensis showed that only a minority of strainshave restriction endonucleases, and strains differ inthe restriction systems they contain (Cohan, Roberts& King, 1991). Different strains of Streptococcus suishave different alleles of the SsuDAT1I gene, whichcode for restriction enzymes with different recog-nition sequences (Sekizaki et al., 2001). A whole-genome comparison of two H. pylori strains showedthat each strain contains four heterologous restriction–modification systems, which are unique to each strain,and are likely to have been acquired by horizontaltransfer from other species (Lin et al., 2001). Thus,variation in restriction–modification systems may bebrought about through modification of existing genes(as in the case of S. suis) and through horizontaltransfer of heterologous genes from other species.

Second, mutations can alter how sensitive het-eroduplex formation is to sequence divergence. Forexample, Zawadzki, Roberts and Cohan (1995) foundthat mutations in recH in B. subtilis yield signif-icantly increased sexual isolation in transformation,most likely by increasing the difficulty of heteroduplexformation between divergent DNA strands. Neverthe-less, no naturally occurring variation in sensitivityto sequence divergence was found within either B.subtilis or B. mojavensis (Zawadzki & Cohan, 1995).

Third, natural isolates of E. coli vary greatly intheir capacity for mismatch repair. About 1% of natur-ally occurring E. coli strains are deficient in mismatch

repair (LeClerc et al., 1996; Matic et al., 1997), andmultiply resistant pathogens show high frequencies ofmismatch deficiency (Oliver et al., 2000). Mismatchrepair has been shown to evolve quickly in liquidculture (Sniegowski, Gerrish & Lenski, 1997; Cooperet al., 2001). Also, when E. coli infects a new mousehost, mutants deficient in mismatch repair increasein frequency, and then after successful colonizationthe mutants decrease in frequency (Giraud et al.,2001).

Horizontal transfer appears to play an importantrole in generating variation in mismatch repair en-zymes. Mismatch repair genes in wild isolates of E.coli have an extraordinary level of mosaicism, contain-ing sequences derived from widely divergent species.This has been interpreted as evidence that mismatchrepair levels have been modulated repeatedly by in-corporation of foreign DNA (Denamur et al., 2000;Brown et al., 2001).

Finally, strains within a species vary in their in-trinsic competence for transformation. Intrinsic com-petence may be defined as the rate at which a straintransforms itself (i.e., when a suitably marked mutantof the recipient strain is used as the donor). For ex-ample, intrinsic competence within a natural popula-tion of B. subtilis varied over four orders of magnitude(Cohan, Roberts & King, 1991). Regardless of thelevel of competence, the rate at which a B. subtilisstrain was transformed by B. mojavensis was alwaysabout six times lower than for intraspecific trans-formation (Cohan, Roberts & King, 1991; Zawadzki,Roberts & Cohan, 1995). Thus, modifying theintrinsic rate of competence alters the capacityfor interspecific recombination. There is consider-able variation in intrinsic competence in other speciesknown to be transformable, including Acinetobactercalcoaceticus (Bergan & Vaksvik, 1983), Pseudo-monas stutzeri (Lorenz & Sikorski, 2000; Sikorski,Teschner & Wackernagel, 2002), and H. pylori (Haas,Meyer & van Putten, 1993).

In summary, sexual isolation is determined by sev-eral factors that can vary within natural populations,owing to mutation of existing genes and acquisitionof foreign genes. While abundant variation in sexualisolation exists, sexual isolation is not free to evolveindependently because of strong pleiotropic effects.Restriction–modification variants likely have a greatereffect on resistance to phages than on sexual isolation.Mismatch repair mutants not only foster interspecificrecombination, but also increase rates of mutation(Vulic et al., 1997). Finally, competence-deficient

364

mutants reduce the rate of intra- as well as interspecificrecombination.

Benefits and costs of sexual isolation. Interspeciesrecombination has a profound impact on bacterialevolution. Generally, transfers of alleles from otherpopulations cause many more nucleotide substitu-tions than mutation. The ratio of recombination- tomutation-caused nucleotide substitutions has been es-timated at 80 for Neisseria meningitidis and 50 forS. pneumoniae (Feil, Enright & Spratt, 2000a) andE. coli (Guttman & Dykhuizen, 1994). Here I out-line the positive and negative impacts of interspecificrecombination on adaptation.

Some adaptations have been acquired from otherbacterial species through homologous recombination,in which a native allele was replaced by a for-eign allele. For example, penicillin resistance hasbeen acquired in this way in S. pneumoniae andN. meningitidis (Maynard Smith, Dowson & Spratt,1991).

Bacteria have obtained many adaptations throughheterologous recombination, in which a novel locus oroperon from another species was acquired (Paul, 1999;Ochman, Lawrence & Groisman, 2000). For example,many of Salmonella’s loci conferring virulence wereobtained from distantly related donors (Groisman &Ochman, 1997). A typical bacterial genome has ac-quired 5–10% of its genes from very divergent species(Ochman, Lawrence & Groisman, 2000), althoughsome obligate intracellular parasites, with little op-portunity for interspecific recombination, show noevidence of having acquired loci from other species(Moran & Wernegreen, 2000). Lawrence (2001) hashypothesized that most niche invasion in bacteria hasoccurred through acquisition of existing adaptationsfrom other species. However, examples of mutation-based niche invasion are known (Sokurenko, Hasty &Dykhuizen, 1999).

While recombination between species has clearlybeen of great benefit in some cases, there is also afitness cost to promiscuous uptake of foreign DNA. In-troduction of foreign DNA may be harmful if it codesfor adaptations that are of value only in the ecologicalor genetic context of the donor organism. The fitnesscosts of randomly chosen interspecific recombinationevents have been little studied, but as previously noted,Vulic (pers. comm.) has found that interspecies trans-conjugants involving transfer between E. coli andSalmonella frequently have reduced fitness.

Can natural selection optimize sexual isolation?How might natural selection affect recombinationbetween species so as to maximize the potentially be-neficial effects while minimizing the fitness costs? Weshould first note that even if the intensity of naturalselection on recombination is extremely weak, theenormous population sizes of bacteria might never-theless allow the characteristics of recombination tobe optimized. For example, the large population sizesof bacteria have allowed natural selection to opti-mize codon usage (where the selection intensity on asingle codon is miniscule); this is a feat beyond thereach of the largest animal populations (Sharp & Li,1986).

First, let us consider the case of sexual isolationin transduction and conjugation, in which recombina-tion is effected by viruses and plasmids, respectively.Majewski (2001) has argued that natural selection act-ing within bacterial populations is more likely to befocused on defenses against infection by the virusesand plasmids, rather than on preventing recombinationcaused by them. This is because infection by virusesand plasmids is generally costly to the cell, althoughthere are exceptions (Desiere et al., 2001). For in-stance, while restriction endonucleases could be underselection for the sexual isolation they provide, theirprincipal fitness effect is likely to be through defenseagainst parasites.

Now let us consider transformation, which bycontrast is a physiological process coded entirely bythe bacterial genome. In Bacillus, transformation isbrought about by over 20 proteins, many of whichhave no known function outside of transformation(Dubnau, 1991). Natural selection might therefore op-timize sexual isolation in transformation without theconstraints associated with transduction and conjuga-tion. How, then, might we expect natural selection tooptimize transformation between species?

Selection for optimal length of transforming segments.Let us suppose that interspecies transfer of most genesis at least slightly deleterious, but that an occasionaltransfer is adaptive. In this case, we should expectselection to favor transformation of short segmentsthat are just long enough to transfer an adaptive seg-ment (Majewski, 2001; Zawadzki & Cohan, 1995).This would allow occasional transfer of a segmentof adaptive value for the recipient, without the co-transformation of flanking DNA segments that arenarrowly adapted to the ecological niche and geneticbackground of the donor.

365

What is the tiniest recombination event that couldeffectively transfer an adaptation? The transfer of asingle nucleotide could transfer an adaptation, but thisis not an important role for recombination; a singlenucleotide substitution could be wrought by muta-tion. Recombination becomes significant in adaptationwhen it brings into a recipient a series of changes thatare adaptive only when they are all acquired together.Thus, an adaptive transfer could involve a small num-ber of changes within a single gene. For example, twopathogenic species of Neisseria have acquired anti-biotic resistance from other species through homol-ogous recombination of segments about 100 basepairs long; the recombined segments were each likelyto contain several adaptive amino acid substitutions(Spratt et al., 1992).

In other cases, much longer recombined segmentsare needed. Because adaptation frequently involvesincorporation of a whole operon (Groisman &Ochman, 1997; Lawrence, 1999; Ochman, Lawrence& Groisman, 2000), recombination by transformationshould be able to transfer segments ranging up toseveral thousand base pairs (Lawrence, 1999). Theoptimal size of the transforming fragment wouldstrike a balance between the benefit of transferringwhole operons, which are useful only if transferredwhole, and the cost of bringing in an excessivenumber of maladaptive genes flanking an adaptivegene.

The balance may be reached when there is a highvariability in the size of segments transferred in trans-formation. In the case of Bacillus, the size of therecombined fragment ranges from several hundredbase pairs to large segments around 10 kb (Dubnau,1991; Zawadzki & Cohan, 1995). Over this range,some adaptations coded on a single gene can be trans-ferred on small segments, without co-transformationof neighboring genes, while whole operons can betransferred on larger segments.

The optimal rate of interspecific transformation.How is natural selection expected to optimize the rateof transformation with divergent donors? The balancebetween the beneficial and deleterious effects of inter-specific recombination will depend on the ecologicalopportunity for using genes of another species, and onthe likelihood that a donor segment would be deleter-ious. In the absence of data on the fitness effects ofinterspecific recombination, we can look to the prop-erties of sexual isolation in transformation to infer thehistory of natural selection.

Several lines of evidence suggest that natural se-lection has reduced sexual isolation in transformation.First, the requirement of near sequence identity atthe ends of a donor–recipient complex appears to re-flect the minimum thermodynamic requirement forassuring accurate alignment of homologous segments(Majewski & Cohan, 1999; Majewski, 2001). Second,in the entire region between the donor ends, sequencedivergence and even non-homology contribute absolu-tely no resistance to heteroduplex formation, at leastin Bacillus (Majewski & Cohan, 1999). Third, mis-match repair has only a minimal effect on transfor-mation between species (Majewski & Cohan, 1999;Majewski et al., 2000; Majewski, 2001). Finally,induction of competence causes hyper-expression ofDNA methylase in Streptococcus, thereby minimizingthe effect of restriction on recombination with otherspecies (Lacks et al., 2000).

Nevertheless, transforming systems do show apreference for bringing the DNA of close relatives intothe cell, owing to the specificity of quorum sensing ingram-positive transformation and the specificity of up-take sequences in gram-negative transformation. How-ever, because all sexual isolation factors subsequent toDNA uptake have been minimized, Majewski (2001)has argued that perhaps the preference for uptake ofclosely related DNA should be interpreted as a positiveforce for acquiring adaptations from close relatives,rather than negative selection against acquiring DNAfrom divergent sources.

The low levels of sexual isolation in transformationsuggest that the rare uptake of beneficial genes fromother species outweighs the more frequent uptake ofdeleterious genes. A bacterium’s strategy for maxim-izing the integration of foreign genes may be likenedto a home run slugger’s strategy of wildly swingingat every plausible pitch. This strategy does not suc-ceed in baseball because the benefit of an occasionalrun or two is outweighed by the cost of nearly alwaysstriking out. However, the bacterium’s strategy is suc-cessful because, as we shall see, the rare home runin recombination yields a novel adaptation, allowinga bacterium to invade a novel ecological niche, and tocreate a new lineage that is irreversibly separate fromthe parental population. In the bacterium’s case, therare home run is rewarded by an immediate WorldSeries championship, dwarfing the harmful effect ofstriking out at nearly every opportunity.

My final aim is to review the roles of recombina-tion and sexual isolation in the origins of speciesdiversity in the bacterial world. We must first consider

366

how a bacterial lineage splits into two lineages thatare ecologically distinct and irreversibly separate, eachwith its own evolutionary tendencies and historicalfate (that is, how one species splits into two, sensuWiley, 1978).

Origins of permanent divergence in bacteria

Consider a definition of species that can accommodatethe peculiar sexual habits of bacteria, in which re-combination is extremely rare but promiscuous. I havepreviously defined an ecotype as a group of bacteriasharing the same ecological niche, such that an ad-aptive mutant (or an adaptive recombinant) from thatniche will outcompete all other members of the eco-type (Cohan, 1994, 2001). Because recombination isextremely rare, each bout of natural selection purgesthe genetic diversity within an ecotype to nearly zero(a purging of diversity called ‘periodic selection’;Atwood, Schneider & Ryan, 1951). All evolution-ary divergence within an ecotype is thus transient,awaiting its demise with the next round of periodicselection.

Permanent divergence can occur only when amutant (or recombinant) invades a new ecologicalniche, such that the mutant is no longer vulnerableto extinction from periodic selection events stemmingfrom the old ecotype. At this point, the genotype in-vading the new niche constitutes a separate ecotype.Diversity within each ecotype is purged by periodicselection from its own adaptive mutants, but peri-odic selection from one ecotype cannot extinguish theother.

I have previously demonstrated that recombinationcannot prevent adaptive divergence between incipientecotypes (Cohan, 1994). This is because recombina-tion in bacteria is rare, even among closest relativeswhere there is no sexual isolation. Even if recombin-ation between nascent ecotypes were to occur at thesame rate as recombination within ecotypes, this rateof recombination would not be sufficient to preventeach ecotype from maintaining and evolving its niche-specific adaptations. If cb is the between-ecotyperecombination rate, and s is the selection intensityagainst a recombinant genotype (i.e., containing oneniche-specific allele from another population), thenthe frequency of maladaptive foreign alleles is cb/s.Because cb ≈ 10−8, selection is expected to keepmaladaptive foreign alleles to negligible levels.

While recombination cannot prevent adaptive di-vergence between ecotypes, recombination can suc-cessfully transfer adaptations that are beneficial inboth the donor and recipient ecotypes. For example,antibiotic resistance has been passed between spe-cies with very different niches, by both homologous(Maynard Smith, Dowson & Spratt, 1991) and heter-ologous (Davies, 1996) recombination. Virulencefactors have been transferred between pathogens withvery different etiologies, such as Yersinia pestis (caus-ing plague) and enteroaggregative strains of E. coli(causing infant diarrhea) (Schubert et al., 1998). Herea transferred iron uptake operon is used by each spe-cies to promote its own form of virulence. In suchcases, the occasional transfer of an adaptation allowsthe recipient to improve its adaptation to its own niche;sharing an adaptation does not reverse the ecologicaldivergence between ecotypes.

Bacterial ecotypes, as defined here, have thequintessential characteristics of species (Cohan, 2001,2002), as understood by all the modern concepts ofspecies (Istock et al., 1996; de Queiroz, 1998). First,each ecotype is subject to a force of genetic cohe-sion, which limits divergence within the ecotype. Inthe case of the highly sexual animals and plants, ge-netic exchange and genetic drift are the predominantforces of cohesion (Avise, 2000; Mayr, 1963). In thecase of bacteria, periodic selection constrains the ge-netic diversity within an ecotype. Second, differentbacterial ecotypes have passed a threshold of diver-gence such that they are no longer bound togetherby any force of evolution: neither periodic selectionnor recombination between ecotypes can prevent fur-ther divergence. Finally, ecotypes are by definitionecologically distinct. In summary, bacterial ecotypesrepresent lineages that are irreversibly separate, eachwith its own evolutionary tendencies and historicalfate (Cohan, 2002, in press; Wiley, 1978).

Consequences of inter-ecotypic recombinationon ecotype formation

In the highly sexual animals and plants, populationscannot develop ecological differences without somedegree of sexual isolation. That is, populations can-not diverge unless the rate of recombination betweenpopulations is greatly reduced compared to that withinpopulations. However, as I have discussed, recom-bination between bacterial populations cannot preventadaptive divergence, even if it occurs at the same

367

rate as recombination within populations. Therefore,evolution of sexual isolation is not a prerequisite forspeciation in the bacterial world (Cohan, 1994). Twobacterial lineages become irreversibly separate by di-verging ecologically, so that they escape each other’speriodic selection events.

Ironically, while recombination between incipientspecies of animals and plants opposes their diver-gence, recombination between newly divergent bac-terial ecotypes may actually have the opposite effect;recombination between ecotypes may foster their di-vergence (Cohan, 2001). Suppose that the ecologicaldivergence between a nascent and a parental ecotyperequires several mutational steps before each eco-type is completely free of periodic selection fromthe other ecotype. In the early stages of such diver-gence, most periodic selection events may be limitedto each ecotype, but occasionally an extraordinarilycompetitive adaptive mutation from the parental eco-type might outcompete all strains from the nascentecotype (as well as strains from its own ecotype). Ge-netic exchange between the ecotypes could preventthis extinction. Through genetic exchange, the nas-cent ecotype could acquire the adaptive mutation fromthe parental ecotype while maintaining its ecologicaldistinctness, and the new ecotype would thereby loseits disadvantage. Genetic exchange can thus foster co-existence of very similar ecotypes until they becomesufficiently divergent that they are not susceptible toeach other’s periodic selection events.

Interspecies recombination can also bring aboutspeciation by introducing adaptations that allow in-vasion of new ecological niches. Whereas plant andanimal species are genetically closed to all other spe-cies (or at least closed to all but the most closelyrelated species with which hybridization is possible), abacterial species is open to gene transfers from a greatdiversity of taxa. So, while plant and animal speciesmust evolve all their adaptations on their own (withthe exception of introgressive hybridization with veryclose relatives), bacteria can take up existing adapta-tions from nearly anywhere in the bacterial world(Ochman, Lawrence & Groisman, 2000). This readyacquisition of new adaptations should make invasionof new niches easier for bacteria than for eukaryotes.

Fortunately, we have experimental systems thatcan assay the importance of inter-species recombina-tion on the origins of bacterial ecotypes. In thesemodel systems, a clone and its descendants are cul-tured in liquid in the laboratory, and are allowedto evolve on their own. In these experimental mi-

crocosms, the starting clone inevitably undergoes anadaptive radiation into multiple ecotypes (Rainey &Travisano, 1998; Treves, Manning & Adams, 1998;Imhof & Schlotterer, 2001).

In microcosms studied up until now, this evolu-tionary diversification has evolved through mutationalone, as the possibilities for recombination wereeliminated in the design of the system. It wouldbe interesting to assay the importance of interspe-cies recombination in fostering speciation, by addingopportunities for recombination to the experimentalsystem. For example, one could allow recombinationamong the descendants of the founding clone, andthen test whether nascent ecotypes persist longer whenthey can exchange genes. Also, by supplying bacterialDNA extracted from a diverse biological community,one could test whether interspecies recombinationincreases the rate of origin of new ecotypes.

If recombination between nascent ecotypes werefound to foster their coexistence, this would havelittle consequence for natural selection on sexual isola-tion. This is because sexual isolation between nascentecotypes is already minimal: sequence divergencebetween nascent ecotypes is near zero, yielding nosexual isolation due to mismatch repair or to difficultyin heteroduplex formation; also, the ecotypes wouldlikely share restriction–modification systems. On theother hand, future experiments might show that in-vasion of new niches is caused largely by uptake ofdivergent DNA from a diverse bacterial community, aspredicted by Lawrence (2001). In this case, we shouldexpect natural selection to place a premium on reducedsexual isolation.

Acknowledgement

I am grateful to Michael Dehn for critically readingthe manuscript and for many helpful suggestions. Thiswork was supported by National Science Foundationgrant DEB-9815576.

References

Atwood, K.C., L.K. Schneider & F.J. Ryan, 1951. Periodic selectionin Escherichia coli. Proc. Natl. Acad. Sci. USA 37: 146–155.

Avise, J.C., 2000. Phylogeography: The History and Formation ofSpecies. Harvard University Press, Cambridge, MA.

Bergan, T. & A.K. Vaksvik, 1983. Taxonomic implications of quant-itative transformation in Acinetobacter calcoaceticus. Zentralbl.Bakteriol. Mikrobiol. Hyg. [A] 254: 214–228.

368

Bessen, D.E., J.R. Carapetis, B. Beall, R. Katz, M. Hibble, B.J.Currie, T. Collingridge, M.W. Izzo, D.A. Scaramuzzino & K.S.Sriprakash, 2000. Contrasting molecular epidemiology of groupA streptococci causing tropical and nontropical infections of theskin and throat. J. Infect. Dis. 182: 1109–1116.

Bickle, T.A. & D.H. Kruger, 1993. Biology of DNA restriction.Microbiol. Rev. 57: 434–450.

Brown, E.W., J.E. LeClerc, B. Li, W.L. Payne & T.A. Cebula, 2001.Phylogenetic evidence for horizontal transfer of mutS allelesamong naturally occurring Escherichia coli strains. J. Bacteriol.183: 1631–1644.

Cohan, F.M., 1994. The effects of rare but promiscuous geneticexchange on evolutionary divergence in prokaryotes. Am. Nat.143: 965–986.

Cohan, F.M., 2001. Bacterial species and speciation. Syst. Biol. 50:513–524.

Cohan, F.M., 2002. Clonal structure: an overview, pp. 158–161in Encyclopedia of Evolution, edited by M. Pagel. OxfordUniversity Press, New York.

Cohan, F.M., in press. What are bacterial species? Ann. Rev.Microbiol. 56.

Cohan, F.M., M.S. Roberts & E.C. King, 1991. The potential forgenetic exchange by transformation within a natural populationof Bacillus subtilis. Evolution 45: 1393–1421.

Cooper, V.S., D. Schneider, M. Blot & R.E. Lenski, 2001. Mech-anisms causing rapid and parallel losses of ribose catabolism inevolving populations of Escherichia coli B. J. Bacteriol. 183:2834–2841.

Danner, D.B., H.O. Smith & S.A. Narang, 1982. Construction ofDNA recognition sites active in Haemophilus transformation.Proc. Natl. Acad. Sci. USA 79: 2393–2397.

Davies, J., 1996. Origins and evolution of antibiotic resistance.Microbiologia 12: 9–16.

de Queiroz, K., 1998. The general lineage concept of species, spe-cies criteria, and the process of speciation, in Endless Forms:Species and Speciation, edited by D.J. Howard & S.H. Berlocher.Oxford University Press, Oxford.

Denamur, E., G. Lecointre, P. Darlu, O. Tenaillon, C. Acquaviva,C. Sayada, I. Sunjevaric, R. Rothstein, J. Elion, F. Taddei, M.Radman & I. Matic, 2000. Evolutionary implications of the fre-quent horizontal transfer of mismatch repair genes. Cell 103:711–721.

Desiere, F., W.M. McShan, D. van Sinderen, J.J. Ferretti & H.Brussow, 2001. Comparative genomics reveals close geneticrelationships between phages from dairy bacteria and patho-genic Streptococci: evolutionary implications for prophage–hostinteractions. Virology 288: 325–341.

Dubnau, D., 1991. Genetic competence in Bacillus subtilis. Micro-biol. Rev. 55: 395–424.

Dubnau, D., 1999. DNA uptake in bacteria. Annu. Rev. Microbiol.53: 217–244.

Duncan, K.E., C.A. Istock, J.B. Graham & N. Ferguson, 1989.Genetic exchange between Bacillus subtilis and Bacillus licheni-formis: variable hybrid stability and the nature of species.Evolution 43: 1585–1609.

Feil, E.J., M.C. Maiden, M. Achtman & B.G. Spratt, 1999. Therelative contributions of recombination and mutation to the di-vergence of clones of Neisseria meningitidis. Mol. Biol. Evol.16: 1496–1502.

Feil, E.J., M.C. Enright & B.G. Spratt, 2000a. Estimatingthe relative contributions of mutation and recombination toclonal diversification: a comparison between Neisseria men-ingitidis and Streptococcus pneumoniae. Res. Microbiol. 151:465–469.

Feil, E.J., J.M. Smith, M.C. Enright & B.G. Spratt, 2000b. Estima-ting recombinational parameters in Streptococcus pneumoniaefrom multilocus sequence typing data. Genetics 154: 1439–1450.

Giraud, A., I. Matic, O. Tenaillon, A. Clara, M. Radman, M. Fons& F. Taddei, 2001. Costs and benefits of high mutation rates:adaptive evolution of bacteria in the mouse gut. Science 291:2606–2608.

Goodman, S.D. & J.J. Scocca, 1988. Identification and arrange-ment of the DNA sequence recognized in specific transforma-tion of Neisseria gonorrhoeae. Proc. Natl. Acad. Sci. USA 85:6982–6986.

Groisman, E.A. & H. Ochman, 1997. How Salmonella became apathogen. Trends Microbiol. 5: 343–349.

Guttman, D.S. & D.E. Dykhuizen, 1994. Clonal divergence in Es-cherichia coli as a result of recombination, not mutation. Science266: 1380–1383.

Haas, R., T.F. Meyer & J.P. van Putten, 1993. Aflagellated mutantsof Helicobacter pylori generated by genetic transformation ofnaturally competent strains using transposon shuttle mutagen-esis. Mol. Microbiol. 8: 753–760.

Hamilton, C.M., M. Aldea, B.K. Washburn, P. Babitzke &S.R. Kushner, 1989. New method for generating deletionsand gene replacements in Escherichia coli. J. Bacteriol. 171:4617–4622.

Harris-Warrick, R.M. & J. Lederberg, 1978. Interspecies transform-ation in Bacillus: sequence heterology as the major barrier. J.Bacteriol. 133: 1237–1245.

Hey, J. & J. Wakeley, 1997. A coalescent estimator of the populationrecombination rate. Genetics 145: 833–846.

Imhof, M. & C. Schlotterer, 2001. Fitness effects of advantageousmutations in evolving Escherichia coli populations. Proc. Natl.Acad. Sci. USA 98: 1113–1117.

Ippen-Ihler, K., 1989. Bacterial conjugation, in Gene Transfer in theEnvironment, edited by S.B. Levy & R.V. Levy. McGraw-Hill,New York.

Istock, C.A., J.A. Bell, N. Ferguson & N.L. Istock, 1996. Bacterialspecies and evolution: theoretical and practical perspectives.J. Ind. Microbiol. 17: 137–150.

Jain, R., M.C. Rivera & J.A. Lake, 1999. Horizontal gene transferamong genomes: the complexity hypothesis. Proc. Natl. Acad.Sci. USA 96: 3801–3806.

Lacks, S.A., S. Ayalew, A.G. de la Campa & B. Greenberg, 2000.Regulation of competence for genetic transformation in Strep-tococcus pneumoniae: expression of dpnA, a late competencegene encoding a DNA methyltransferase of the DpnII restrictionsystem. Mol. Microbiol. 35: 1089–1098.

Lawrence, J., 1999. Selfish operons: the evolutionary impact of geneclustering in prokaryotes and eukaryotes. Curr. Opin. Genet. Dev.9: 642–648.

Lawrence, J., 2001. Catalyzing bacterial speciation: correlatinglateral transfer with genetic headroom. Syst. Biol. 50: 479–496.

LeClerc, J.E., B. Li, W.L. Payne & T.A. Cebula, 1996. Highmutation frequencies among Escherichia coli and Salmonellapathogens. Science 274: 1208–1211.

Leff, L., J. McArthur & L. Shimkets, 1992. Information spiraling:movement of bacteria and their genes in streams. Microb. Ecol.24: 11–24.

Lin, L.F., J. Posfai, R.J. Roberts & H. Kong, 2001. Comparative ge-nomics of the restriction–modification systems in Helicobacterpylori. Proc. Natl. Acad. Sci. USA 98: 2740–2745.

Lorenz, M.G. & J. Sikorski, 2000. The potential for intraspe-cific horizontal gene exchange by natural genetic transformation:sexual isolation among genomovars of Pseudomonas stutzeri.Microbiology 146(Pt 12): 3081–3090.

369

Lorenz, M.G. & W. Wackernagel, 1994. Bacterial gene transferby natural genetic transformation in the environment. Microbiol.Rev. 58: 563–602.

Madigan, M., J. Martinko & J. Parker, 1999. Brock’s Biology ofMicroorganisms. Prentice-Hall, Upper Saddle River, NJ.

Majewski, J., 2001. Sexual isolation in bacteria. FEMS Microbiol.Lett. 199: 161–169.

Majewski, J. & F.M. Cohan, 1998. The effect of mismatch repair andheteroduplex formation on sexual isolation in Bacillus. Genetics148: 13–18.

Majewski, J. & F.M. Cohan, 1999. DNA sequence similarity re-quirements for interspecific recombination in Bacillus. Genetics153: 1525–1533.

Majewski, J., P. Zawadzki, P. Pickerill, F.M. Cohan & C.G.Dowson, 2000. Barriers to genetic exchange between bacterialspecies: Streptococcus pneumoniae transformation. J. Bacteriol.182: 1016–1023.

Matic, I., F. Taddei & M. Radman, 2000. No genetic barri-ers between Salmonella enterica serovar typhimurium and Es-cherichia coli in SOS-induced mismatch repair-deficient cells.J. Bacteriol. 182: 5922–5924.

Matic, I., M. Radman, F. Taddei, B. Picard, C. Doit, E. Bingen,E. Denamur & J. Elion, 1997. Highly variable mutation ratesin commensal and pathogenic Escherichia coli. Science 277:1833–1834.

Maynard Smith, J., C.G. Dowson & B.G. Spratt, 1991. Localizedsex in bacteria. Nature 349: 29–31.

Maynard Smith, J., N. Smith, M. O’Rourke & B.G. Spratt, 1993.How clonal are bacteria? Proc. Natl. Acad. Sci. USA 90:4384–4388.

Mayr, E., 1963. Animal Species and Evolution. Belknap Press ofHarvard University Press, Cambridge.

Mazodier, P. & J. Davies, 1991. Gene transfer between distantlyrelated bacteria. Annu. Rev. Genet. 25: 147–171.

Miller, R.V., 2001. Environmental bacteriophage–host interac-tions: factors contribution to natural transduction. Antonie VanLeeuwenhoek 79: 141–147.

Moran, N.A. & J.J. Wernegreen, 2000. Lifestyle evolution in sym-biotic bacteria: insights from genomics. Trends Ecol. Evol. 15:321–326.

Morrison, D.A. & M.S. Lee, 2000. Regulation of competencefor genetic transformation in Streptococcus pneumoniae: a linkbetween quorum sensing and DNA processing genes. Res. Mi-crobiol. 151: 445–451.

Ochman, H., J.G. Lawrence & E.A. Groisman, 2000. Lateral genetransfer and the nature of bacterial innovation. Nature 405:299–304.

O’Rourke, M. & B.G. Spratt, 1994. Further evidence for the non-clonal population structure of Neisseria gonorrhoeae: extensivegenetic diversity within isolates of the same electrophoretic type.Microbiology 140(Pt 6): 1285–1290.

Oliver, A., R. Canton, P. Campo, F. Baquero & J. Blazquez, 2000.High frequency of hypermutable Pseudomonas aeruginosa incystic fibrosis lung infection. Science 288: 1251–1254.

Pozzi, G., L. Masala, F. Iannelli, R. Manganelli, L.S.Havarstein, L. Piccoli, D. Simon & D.A. Morrison,1996. Competence for genetic transformation in encap-sulated strains of Streptococcus pneumoniae: two allelicvariants of the peptide pheromone. J. Bacteriol. 178:6087–6090.

Rainey, P.B. & M. Travisano, 1998. Adaptive radiation in a hetero-geneous environment. Nature 394: 69–72.

Rayssiguier, C., D.S. Thaler & M. Radman, 1989. The barrier torecombination between Escherichia coli and Salmonella typh-

imurium is disrupted in mismatch-repair mutants. Nature 342:396–401.

Rivera, M.C., R. Jain, J.E. Moore & J.A. Lake, 1998. Genomicevidence for two functionally distinct gene classes. Proc. Natl.Acad. Sci. USA 95: 6239–6244.

Roberts, M.S. & F.M. Cohan, 1993. The effect of DNA se-quence divergence on sexual isolation in Bacillus. Genetics 134:401–408.

Roberts, M.S. & F.M. Cohan, 1995. Recombination and migra-tion rates in natural populations of Bacillus subtilis and Bacillusmojavensis. Evolution 49: 1081–1094.

Roberts, R.J. & D. Macelis, 1992. Restriction enzymes and theirisoschizomers. Nucl. Acids Res. 20: 2167–2180.

Rocha, E.P., A. Danchin & A. Viari, 2001. Evolutionary roleof restriction/modification systems as revealed by comparativegenome analysis. Genome Res. 11: 946–958.

Schofield, M.J., S. Nayak, T.H. Scott, C. Du & P. Hsieh, 2001. Inter-action of Escherichia coli MutS and MutL at a DNA mismatch.J. Biol. Chem. 276: 28291–28299.

Schubert, S., A. Rakin, H. Karch, E. Carniel & J. Heesemann, 1998.Prevalence of the high-pathogenicity island of Yersinia speciesamong Escherichia coli strains that are pathogenic to humans.Infect. Immun. 66: 480–485.

Sekizaki, T., M. Osaki, D. Takamatsu & Y. Shimoji, 2001. Distri-bution of the SsuDAT1I restriction–modification system amongdifferent serotypes of Streptococcus suis. J. Bacteriol. 183:5436–5440.

Selander, R. & J. Musser, 1990. Population genetics of bacterialpathogenesis, pp. 11–36 in Molecular Basis of Bacterial Patho-genesis, edited by B. Iglewski & V. Clark. Academic Press, SanDiego.

Sharp, P.M. & W.H. Li, 1986. An evolutionary perspective on syn-onymous codon usage in unicellular organisms. J. Mol. Evol. 24:28–38.

Shoemaker, N.B., H. Vlamakis, K. Hayes & A.A. Salyers,2001. Evidence for extensive resistance gene transfer amongBacteroides spp. and among Bacteroides and other gen-era in the human colon. Appl. Environ. Microbiol. 67:561–568.

Sikorski, J., N. Teschner & W. Wackernagel, 2002. Highlydifferent levels of natural transformation are associated withgenomic subgroups within a local population of Pseudo-monas stutzeri from soil. Appl. Environ. Microbiol. 68:865–873.

Sniegowski, P.D., P.J. Gerrish & R.E. Lenski, 1997. Evolution ofhigh mutation rates in experimental populations of E. coli. Nature387: 703–705.

Sokurenko, E.V., D.L. Hasty & D.E. Dykhuizen, 1999. Pathoad-aptive mutations: gene loss and variation in bacterial pathogens.Trends Microbiol. 7: 191–195.

Sparling, P.F., 1966. Genetic transformation of Neisseriagonorrhoeae to streptomycin resistance. J. Bacteriol. 92:1364–1371.

Spratt, B.G., L.D. Bowler, Q.Y. Zhang, J. Zhou & J.M. Smith,1992. Role of interspecies transfer of chromosomal genes in theevolution of penicillin resistance in pathogenic and commensalNeisseria species. J. Mol. Evol. 34: 115–125.

Stotzky, G., 1989. Gene transfer among bacteria in soil,pp. 165–222 in Gene Transfer in the Environment, edited byS.B.L.a.R.V. Miller. McGraw-Hill, New York.

Suerbaum, S., J. Maynard Smith, K. Bapumia, G. Morelli, N. Smith,E. Kunstmann, I. Dyrek & M. Achtman, 1998. Free recombina-tion within Helicobacter pylori. Proc. Natl. Acad. Sci. USA 95:12619–12624.

370

Tortosa, P., L. Logsdon, B. Kraigher, Y. Itoh, I. Mandic-Mulec &D. Dubnau, 2001. Specificity and genetic polymorphism of theBacillus competence quorum-sensing system. J. Bacteriol. 183:451–460.

Treves, D.S., S. Manning & J. Adams, 1998. Repeated evolution ofan acetate-crossfeeding polymorphism in long-term populationsof Escherichia coli. Mol. Biol. Evol. 15: 789–797.

Van Spanning, R.J., W.N. Reijnders & A.H. Stouthamer, 1995. In-tegration of heterologous DNA into the genome of Paracoccusdenitrificans is mediated by a family of IS1248-related ele-ments and a second type of integrative recombination event.J. Bacteriol. 177: 4772–4778.

Vulic, M., F. Dionisio, F. Taddei & M. Radman, 1997. Molecularkeys to speciation: DNA polymorphism and the control of ge-netic exchange in enterobacteria. Proc. Natl. Acad. Sci. USA 94:9763–9767.

Wiley, E., 1978. The evolutionary species concept reconsidered.Syst. Zool. 27: 17–26.

Zawadzki, P. & F.M. Cohan, 1995. The size and continuity of DNAsegments integrated in Bacillus transformation. Genetics 141:1231–1243.

Zawadzki, P., M.S. Roberts & F.M. Cohan, 1995. The log-linearrelationship between sexual isolation and sequence divergencein Bacillus transformation is robust. Genetics 140: 917–932.