Assessing the genetic structure of microbial populations.

A strenuous debate continues to rage over the use of DNA typing as forensic evidence (1-5). One of the most contentious issues has been how to calculate the probability of a coincidental match between the DNA of a suspect and DNA taken from the scene of a crime (and thought to belong to the perpetrator). In principle, this probability depends upon the genetic structure of the human population, including the extent to which alleles are associated statistically due to limited recombination among genetically distinct subpopulations. More quietly, the genetic structure of microbial populations has been the subject of growing interest for more than a decade. No microbe faces conviction in a court of law for its offenses, although many have caused harm that would be the envy of any mass murderer. But an understanding of the genetic structure of microbial populations is relevant to tracking down the sources of epidemic outbreaks of pathogens, to managing the spread of resistance to drugs, and to improving the efficacy and safety of genetically engineered microorganisms intended for environmental applications. The basic issue is whether bacteria and certain other microorganisms, including parasitic protozoa, exist as a series of asexual clones or as promiscuous, freely intermixing populations. The rapid spread of antibiotic-resistance genes led to a widespread view that genetic exchange was prevalent among bacteria in nature (6) and that clonal growth was an artifact of laboratory studies. But antibiotic-resistance genes are typically encoded by horizontally transmissible plasmids, and these genes have come under extraordinarily intense selection (7). Hence, antibiotic-resistance genes may not give a complete picture of the population structures of bacteria and other microorganisms (8). In this issue of the Proceedings, Maynard Smith et al. (9) shed some new light on the analysis of population genetic structure in bacteria and parasitic protozoa. At first glance, their results are discomforting, because they show that conclusions regarding population structure sometimes depend on details of how one analyzes the genetic data. Fortunately, however, the differences that emerge depending on the method of analysis are illuminating and may reveal important biological processes. The data sets analyzed by Maynard Smith et al. (9) are based on multilocus enzyme electrophoresis, whereby alleles at different loci are distinguished on the basis of differences in the mobility of their resulting proteins in an electrical field. Though the resolution of the method is now regarded as crude relative to DNA sequencing or fingerprinting, it has the virtue that hundreds of isolates can be screened for differences at tens of loci quickly and inexpensively (10). In 1973, Milkman (11) performed the first systematic survey of electrophoretic variation in a microorganism, showing that Escherichia coli harbored a tremendous wealth of molecular genetic variation. In 1980, Selander and Levin (12) showed that, despite this tremendous allelic variation, there was much less genotypic variation in E. coli than one might have expected, because certain alleles were almost invariably found in association with one another-e.g., the two-locus haploid genotypes AB and ab might both be common, whereas aB and Ab are absent or extremely rare. The scarcity of certain multilocus genotypes relative to expectations under free recombination is an indication of linkage disequilibrium. Later studies strengthened Selander and Levin's inference that linkage disequilibrium in E. coli was due primarily to infrequent recombination, rather than population subdivision or other factors that could create disequilibrium. In particular, geographical variation accounts for only a small fraction of genetic variation in E. coli (13), and the same multilocus electrophoretic genotypes are found to persist stably over several decades (14). Since Selander and Levin's benchmark paper (12), clonal population structures have been reported for many other bacterial species, including numerous pathogens as well as environmentally important bacteria (15, 16). It has also been reported that some parasitic protozoa have clonal population structures, raising the possibility that clonality may be common in this group of organisms as well (17). But even as the clonal structure of bacterial populations was being elevated to the status of a paradigm, there were also some indications that things might not be quite so simple. Dykhuizen and Green (18) showed that phylogenies constructed for different chromosomal genes (using DNA sequences) from the same set of E. coli isolates were significantly different from one another. This result makes sense only if recombination between lineages has been sufficiently common to disrupt the associations between loci, but not so common as to preclude constructing gene phylogenies in the first place. And two studies published last year suggest that recombination among chromosomal genes may be much more frequent in some other bacterial species. Istock et al. (19) found only slight linkage disequilibrium in a population ofBacillus subtilis taken from one site in an Arizona desert; what disequilibrium did exist was much less extreme than in a correspondingly localized population of E. coli (collected from a single host, ref. 20). They concluded that recombination must be much more frequent relative to binary fission in B. subtilis than in E. coli. Souza et al. (21) investigated Rhizobium leguminosarum nodulating wild and cultivated beans at several different geographical scales, ranging from individual host plants to throughout the Western Hemisphere. Although they observed extremely strong linkage disequilibrium in the data set as a whole, most of this disequilibrium could be attributed to geographical subdivision. Souza et al. (21) also found that a local population of R. leguminosarum showed much less extreme disequilibrium than a local population of E. coli, suggesting that recombination may be much more frequent in Rhizobium than in E. coli. Brown et al. (22, 23) proposed a useful index to describe linkage disequilibrium, which is based on the distribution of allelic mismatches between pairs of organisms over several loci. For example, if one organism has haploid genotype A1B1C1D1El and another has genotype A2B1C2D2E1, then the number of mismatches between them is three. The mean number of pairwise mismatches among a set of organisms is a measure of genetic distance. The variance in the number of pairwise mismatches, relative to that expected under the hypothesis of panmixia (i.e., random association of alleles), provides an index that allows clonality to be evaluated statistically. This