Syst. Biol. 50(4):513���524, 2001 Bacterial Species and Speciation FREDERICK M. COHAN Department of Biology, Wesleyan University, Middletown, Connecticut 06459-0170 , USA E-mail: fcohan@wesleyan.edu Abstract.���Bacteria are profoundly different from eukaryotes in their patterns of genetic exchange. Nevertheless, ecological diversity is organized in the same way across all of life: individual organisms fall into more or less discrete clusters on the basis of their phenotypic, ecological, and DNA sequence characteristics. Each sequence cluster in the bacterial world appears to correspond to an ���ecotype,��� de ned as a population of cells in the same ecological niche, which would all be out-competed by any adaptive mutant coming from the population. Ecotypes, so de ned, share many of the dynamic properties attributed to eukaryotic species: genetic diversity within an ecotype is limited by a force of cohesion (in this case, periodic selection) different ecotypes are free to diverge without constraint from one another and ecotypes are ecologically distinct. Also, ecotypes can be discovered and classi ed as DNA sequence clusters, even when we are ignorant of their ecology. Owing to the rarity and promiscuity of bacterial genetic exchange, speciation in the bacterial world is expected to be much less constrained than in the world of animals and plants. [Clonal structure ecotype periodic selection sexual isolation species concept.] Wherever there is life, there are bacte- ria (Margulis and Schwartz, 1998 Madigan et al., 1999). Free-living bacteria are found in every environment that supports eukary- otes (Madigan et al., 1999). Also, no meta- zoan or metaphyte is known to be free of bac- terial commensals or pathogens (Madigan et al., 1999). The most physically extreme habitats capable of supporting life can sup- port extremophilic bacteria (e.g., Brock, 1978 Madigan and Marrs, 1997). Finally, anoxic habitats can support a great diversity of bacteria whose physiology does not require oxygen (Fenchel and Finlay, 1995). There is clearly tremendous ecological diversity within the prokaryotic world. The ecological diversity among prokary- otes is patterned in much the same way as in eukaryotes: Individual organisms fall into discrete clusters on the basis of their pheno- typic, ecological, and DNA sequence char- acteristics (Sneath, 1977 Broom and Sneath, 1981 Bridge and Sneath, 1983 Sneath and Stevens, 1985 Barrett and Sneath, 1994 Em- bley and Stackebrandt, 1997 Goodfellow et al., 1997). For example, Barrett and Sneath (1994) found that 315 strains from the Neisse- riaceae (the family including the pathogens that cause gonorrhea and meningitis) fall into 31 distinct clusters on the basis of metabolic and other phenotypic characteristics. Like- wise, sequence-based surveys of bacterial di- versity yield discrete clusters in sequence space (Palys et al., 1997), as is also seen in eukaryotes (Mallet, 1995). In the world of the highly sexual eukary- otes (including most plants and animals), the clustered pattern of diversity has long provided the primary basis for demarcat- ing species in practice. For many decades, species have been demarcated as phenotypic clusters (Sokal and Crovello, 1970), and more recently, sequence data have allowed species to be demarcated as clusters in genotypic or sequence space (Mallet, 1995). Species demarcation in bacteria has re- lied exclusively on the phenotypic and genotypic clustering of bacterial strains (Goodfellow et al., 1997). Bacterial species have long been understood to be pheno- typic clusters (based largely on metabolic characters Sneath, 1977). Numerical taxon- omy has clearly demonstrated that bacterial strains fall into discrete phenotypic clusters and has provided a means for discovering species and for classifying unknown strains into already characterized species (Broom and Sneath, 1981 Bridge and Sneath, 1983). Genotypic clustering has largely replaced phenotypic clustering as a primary criterion for demarcating bacterial species (Wayne et al., 1987). For decades, clustering based on whole-genome DNA���DNA hybridization between pairs of strains has served to distin- guish species (Johnson, 1973 Wayne et al., 1987). More recently, systematists are us- ing clustering of 16S rRNA (Stackebrandt and Goebel, 1994) and protein-coding (Palys et al., 1997, 2000) gene sequences as criteria for species demarcation. 513

514 S YSTEMATIC BIOLOGY VOL. 50 Although clustering provides a taxo- nomic tool for classifying organisms into species, systematists and evolutionary biol- ogists have widely believed that species are more than just clusters of very closely re- lated and very similar organisms. Species are believed to have some fundamental dy- namic properties as well. For instance, dif- ferent species are thought to exploit differ- ent ecological niches (e.g., Ecological Species Concept Van Valen, 1976). Also, genetic di- vergence within a species is thought to be constrained by one or more forces of cohe- sion (Cohesion Species Concept Meglitsch, 1954 Templeton, 1989), most frequently ge- netic exchange (Biological Species Concept Mayr, 1982). Different species are thought to have separate evolutionary fates, in that they are free to diverge without constraint from one another (Evolutionary Species Concept Simpson, 1961 Wiley, 1978). In the world of plants and animals, one species is thought to split into two when the two groups have diverged suf ciently in their reproductive and ecological characteristics (e.g., Eldredge, 1985): The reproductive divergence prevents genetic exchange from reversing the genetic divergence between incipient species (Mayr, 1982), and the ecological divergence is nec- essary to allow competitive coexistence be- tween the incipient species (Gause, 1934 May, 1973). Here I address whether clusters observed in the bacterial world share the dynamic properties attributed to eukaryote species, and whether the mechanisms driving the origins of new species in bacteria might be shared with the eukaryotes. I will demon- strate that, despite basic differences in the nature of genetic exchange between bacte- ria and eukaryotes, bacterial species share many of the fundamental properties of eu- karyotic species. Moreover, bacterial species can be demarcated in practice by the same sequence-cluster criteria used in eukaryotic systematics. G ENETIC EXCHANGE IN PROKARYOTES The character of genetic exchange in prokaryotes differs profoundly from that in the most highly sexual eukaryotes, in sev- eral ways. First, recombination in bacteria is extremely rare. Several laboratories have taken a retrospective approach to determin- ing the historical rates of recombination in nature. Based on surveys of diversity in allozymes, restriction-recognition sites, and gene sequences, recombination rates have been estimated from the degree of associa- tion between genes or between parts of genes (Hudson and Kaplan, 1985 Hudson, 1987 Hey and Wakeley, 1997). This approach has shown that a given gene segment is usually involved in recombination at about the same rate at which it is mutated, or less (Selander and Musser, 1990 Maynard Smith et al., 1993 Whittam and Ake, 1993 Roberts and Cohan, 1995). Second, recombination in bacteria is much more promiscuous than is the case for plants and animals. Bacteria do not exchange genes as often as animals and plants, but when they do they are not nearly as fussy about their choice of partners. Animal groups typ- ically lose the ability to exchange genes en- tirely by the time their mitochondrial DNA sequences are 3% divergent (Avise, 2000), al- though some animal subspecies as divergent as 16% can exchange genes in nature (Moritz et al., 1992). Bacteria, in contrast, can undergo homologous recombination with organisms as divergent as 25% in DNA sequence (and possibly more) (Duncan et al., 1989 Roberts and Cohan, 1993 Zawadzki et al., 1995 Vulic et al., 1997). There are, nevertheless, some important constraints on bacterial genetic ex- change. Recombination that depends on vec- tors, such as bacteriophage-mediated trans- duction or plasmid-mediated conjugation, is limited by the host ranges of the respective vectors. Also, restriction endonuclease activ- ity can greatly reduce the rate of recombi- nation (Edwards et al., 1999 Milkman et al., 1999), although not in the case of recom- bination by transformation (Trautner et al., 1974 Cohan et al., 1991). Finally, homolo- gous recombination is limited by the resis- tance to integration of divergent DNA se- quences, because mismatch repair tends to reverse integration of a mismatched donor��� recipient heteroduplex (Rayssiguier et al., 1989 Vulic et al., 1997) and because inte- gration requires a 20���30-bp stretch of nearly perfectly matched DNA (Hsieh et al., 1992 Rao et al., 1995 Majewski and Cohan, 1998, 1999a). In both Gram-positive and Gram- negative bacteria, the rate of recombination decays exponentially with donor-recipient sequence divergence (Roberts and Cohan,