Estimates of gene flow in Eichhornia paniculata (Pontederiaceae): effects of range substructure

Gene flow in the annual aquatic plant, Eichhornia paniculata, was inferred from estimates of genetic differentiation at 24 isozyme loci among 44 populations from north-east Brazil. Population differentiation, estimated as the correlation among genes of different individuals (phi), was 0.45 (range among loci, 0.10-0.69). Based on Wright's island model, this heterogeneity would result from gene flow equivalent to 0.31 immigrants (Nm) entering each population per generation. The distribution of E. paniculata in north-eastern Brazil is geographically and genetically subdivided, and therefore, the assumption that migrants are a random draw from all populations is likely to be violated for this range-wide estimate of gene flow. We investigated the importance of range subdivision on indirect estimates of gene flow through computer simulation and through a hierarchical analysis of FST and Nm in populations of E. paniculata from northeastern Brazil. Simulations indicated that estimates of Nm in the presence of range substructure consistently underestimated the actual values of gene flow. The degree to which Nm was underestimated increased in proportion to the magnitude of genetic differentiation among range subdivisions. In E. paniculata, northern and southern regions of the geographical range and local clusters within regions were genetically differentiated (phi = 0.10, phi= 0.14, respectively) and there was a strong negative relationship between Nm estimated for pairs of populations and the geographical distance between them. Average estimates of population differentiation decreased and gene flow increased with successive reductions in spatial scale, from the complete range sampled (Nm = 0.31), to regions (Nm = 0.44), to local areas within regions (Nm = 0.64), and to neighbouring population pairs within local areas (Nm = 0.58). Similarity in estimates for neighbouring pairs and local areas suggests that, below the spatial scale of local area, gene flow estimates are not influenced by range substructure and can be considered to occur at random. Our results suggest that range substructure can have a substantial influence on gene flow estimates, and that ecologically relevant rates are likely to be higher than those indicated by range-wide analyses in organisms with geographical subdivision.