Copyright 0 1995 by the Genetics Society of America Separating Population Structure from Population History: A Cladistic Analysis of the Geographical Distribution of Mitochondrial DNA Haplotypes in the Tiger Salamander, Ambystoma tigrinurn Alan R. Templeton, Eric Routman and Christopher A. Phillips * Department of Biology, Washington University, St. Louis, Missouri 63130-4899 Manuscript received October 17, 1994 Accepted for publication February 16, 1995 ABSTRACT Nonrandom associations of alleles or haplotypes with geographical location can arise from restricted gene flow, historical events (fragmentation, range expansion, colonization), or any mixture of these factors. In this paper, we show how a nested cladistic analysis of geographical distances can be used to test the null hypothesis of no geographical association of haplotypes, test the hypothesis that significant associations are due to restricted gene flow, and identify patterns of significant association that are due to historical events. In this last case, criteria are given to discriminate among contiguous range expansion, long-distance colonization, and population fragmentation. The ability to make these discriminations depends critically upon an adequate geographical sampling design. These points are illustrated with a worked example: mitochondrial DNA haplotypes in the salamander Ambystoma tign‘num. For this example, prior information exists about restricted gene flow and likely historical events, and the nested cladistic analyses were completely concordant with this prior information. This concordance establishes the plausibility of this nested cladistic approach, but much future work will be necessary to demonstrate robustness and to explore the power and accuracy of this procedure. 0 NE of the principal aims of population genetics is the measurement of the amount and patterns of genetic variation found within and among subpopu- lations of interbreeding organisms to study gene flow, genetic drift, system of mating, mutation, and natural selection. Traditionally, inferences about microevolu- tionary forces have been based upon the number of alleles (or haplotypes) , their frequencies, and their geographical distribution. For example, WRIGHT (1969) developed hierarchical Fstatistics as a tool to study gene flow, genetic drift and system of mating from data on allele and genotype frequencies and geography. After estimating these F statistics, microevolutionary parame- ters could be estimated by relating the Fstatistics to an underlying model. For example, in the “island model” of gene flow in which the population is subdivided into many “islands” of inbreeding effective size N with a rate of exchange of m per generation at random over all islands, Wright’s &, should be equal to 1 / [ 4N( m + p) + 13 where p is the mutation rate. Hence, if &, and p are estimated, it is possible to estimate Nm, the effective number of migrating individuals per genera- tion. One serious limitation of this approach is that the Corresponding authm: Alan R. Templeton, Department of Biology, Washington University, St. Louis, MO 63130-4899. E-mail: [email protected] sity, San Francisco, CA 94132. sity, 607 E. Peabody Drive, Champaign, IL 61820. ‘Present address: Department of Biology, San Francisco State Univer- ‘Present address: Illinois Natural History Survey, Center for Biodiver- relation of 4, to underlying microevolutionary parame- ters changes with different models of population struc- ture. For example, if the populations are in a onedimen- sional habitat (such as in or along a river) and all dispersal is limited to exchanges between adjacent popu- lations at a rate of m/2 per generation, then F,, = 1/ [ 4N( 2mp) ’’* + 13 ( KIMURA and WEISS 1964). Many other models exist, each with its own relationship be- tween &, and underlying inbreeding effective size, muta- tion, and gene flow parameters. Consequently, one ma- jor limitation of the use of F,, (and related statistics, e.8, LYNCH and CREASE 1990; HUDSON et. al. 1992a) is that the data used to estimate the F statistics often do not indicate which model of gene flow is appropriate for the populations being studied. This is further complicated by the fact that the various models of gene flow are not necessarily alternatives; one part of a species range may be restricted to onedimensional, stepping stone gene flow, whereas another part may fit the two-dimensional continuous, isolation by distance model. Even worse, the geographical genetic variation measured by &, and re- lated statistics may have nothing to do with current pat- terns of gene flow at all. For example, LAFSON (1984) pointed out that when a population expands into or colonizes a new geographic area, genetic homogeneity could be created within the recently colonized area that does not reflect current patterns of gene flow. Similarly, two populations may have been fragmented in the past and currently have no gene flow whatsoever, yet their shared ancestry can create F,, values less than one that Genetics 140 767-782 (June, 1995)768 A. R. Templeton, E. Routman and C. A. Phillips would erroneously imply gene flow. Hence, the gee graphical pattern of genetic variation is influenced by population structure, by population history, and by com- binations of these structural and historical factors. This paper is concerned with separating population structure from population history as sources of geographical asso- ciations with genetic variants. The effects of population history and structure have been studied by spatial autocorrelation ( SOKAL et al. 1989a,b; SLATKIN and ARTER 1991; EPPERSON 1993), principal component analyses ( e.g., AMMERMAN and CA- VALLI-SFORZA 1984), and multidimensional scaling ( LESSA 1990). These approaches can be thought of as the analogue of “phenetic” approaches in systematics, as they are based upon some sort of genetic distance or identity measure. In contrast, we will outline a cladistic approach in this paper that uses a character-state based haplotype evolutionary tree. This paper will be limited to genetic surveys using DNA sequence or restriction site data with samples derived from a finite number of distinct sampling locations. With such genetic surveys, it is
View Full Document
Unlocking...