Two-generation analysis of pollen flow across a landscape V: a stepwise approach for extracting factors contributing to pollen structure.

Patterns of pollen dispersal are central to both the ecology and evolution of plant populations. However, the mechanisms controlling either the dispersal process itself or our estimation of that process may be influenced by site-specific factors such as local forest structure and nonuniform adult genetic structure. Here, we present an extension of the AMOVA model applied to the recently developed TWOGENER analysis of pollen pool structure. This model, dubbed the Stepwise AMOVA (StAMOVA), focuses on determining to what extent ecological, demographic, and/or environmental factors influence the observed genetic variation in spatially separated pollen pools. The analysis is verified for efficacy, using an extensive battery of simulations, illustrating: (1) how nonuniform adult genetic structure influences the differentiation of spatially separated pollen pools, and (2) how effectively the Stepwise analysis performs in carrying out the appropriate corrections. Finally, the model is applied to a Quercus alba data set, from which we have prior evidence that the adult genetic structure is nonuniformly distributed across the sampling landscape. From this data set, we show how the Stepwise model can be applied to remove the effects of spatial adult genetic structure on pollen pool differentiation and contrast these results with those derived from the original TWOGENER analysis.

Two-generation analysis of pollen flow across a landscape. I. Male gamete heterogeneity among females.

Gene flow is a key factor in the spatial genetic structure in spatially distributed species. Evolutionary biologists interested in microevolutionary processess and conservation biologists interested in the impact of landscape change require a method that measures the real time process of gene movement. We present a novel two-generation (parent-offspring) approach to the study of genetic structure (TwoGener) that allows us to quantify heterogeneity among the male gamete pools sampled by maternal trees scattered across the landscape and to estimate mean pollination distance and effective neighborhood size. First, we describe the model's elements: genetic distance matrices to estimate intergametic distances, molecular analysis of variance to determine whether pollen profiles differ among mothers, and optimal sampling considerations. Second, we evaluate the model's effectiveness by simulating spatially distributed populations. Spatial heterogeneity in male gametes can be estimated by phiFT, a male gametic analogue of Wright's F(ST) and an inverse function of mean pollination distance. We illustrate TwoGener in cases where the male gamete can be categorically or ambiguously determined. This approach does not require the high level of genetic resolution needed by parentage analysis, but the ambiguous case is vulnerable to bias in the absence of adequate genetic resolution. Finally, we apply TwoGener to an empirical study of Quercus alba in Missouri Ozark forests. We find that phiFT = 0.06, translating into about eight effective pollen donors per female and an effective pollination neighborhood as a circle of radius about 17 m. Effective pollen movement in Q. alba is more restricted than previously realized, even though pollen is capable of moving large distances. This case study illustrates that, with a modest investment in field survey and laboratory analysis, the TwoGener approach permits inferences about landscape-level gene movements.

CALEDONIAN SCOTS PINE: ORIGINS AND GENETIC STRUCTURE.

Monoterpene and isozyme loci, used as markers to study the genetic structure of Scots pine (Pinus sylvestris L.) native to Scotland, showed that the endemic populations are not genetically impoverished, in spite of severe contraction in range and numbers as a result of both natural and anthropogenic causes. On the contrary, variability in the relict populations is almost the highest: of any plant species studied, with average heterozygosities of 0.33 for monoterpenes (rive loci) and 0.30 for isozymes (16 loci). The overwhelming proportion of this variability (> 95 %) was within populations, even though significant differences in gene frequencies of many individual loci existed among populations. Multiple-locus comparison of gene frequencies among populations, resolved by canonical variate analysis, showed no coherent geographic pattern of differences from population to population or region to region, with one major exception: certain populations in northwestern Scotland (Wester Ross) were distinct from all others and each other. The pattern of variability of the biochemical markers was consistent with that of metrical and physiological traits reported in the literature. These traits, in turn, show relatively little genetic affinity between contemporary Scottish and continental European populations. The genetic evidence, together with the anomalous distribution of pine pollen in the British Isles during the Holocene, suggests that the Caledonian race of Scots pine originated endemically from more than one refugium after the last glaciation.