Introduction
Understanding the factors that drive patterns of genetic variation among plant populations is central in biology because genetic diversity is the raw material on which evolution acts. Quantifying population differentiation, which is most frequently done using the fixation index FST (Wright,1951; see Holsinger & Weir, 2009; Meirmans & Hedrick, 2011 for a review of FST and related metrics), is important for understanding the first stages of allopatric speciation (Harvey, Singhal, & Rabosky, 2019; Templeton, 1981), as well as the basis of local adaptation (Leimu & Fischer, 2008; Linhart & Grant, 1996), and provides critical information for conservation genetics (Ellstrand, 1992; Ellstrand & Elam, 1993; Kramer & Havens, 2009). Life history traits are expected to influence population genetic structure in seed plants (Duminil et al., 2007; Hamrick & Godt, 1996; Loveless & Hamrick, 1984). However, previous studies have given conflicting results as to the importance of specific traits, such as pollination mode, seed dispersal mode, mating system, and growth form (e.g., Duminil et al., 2007; Hamrick & Godt, 1996), and only one study has compared patterns of FST variation between latitudinal regions (Dick, Hardy, Jones, & Petit, 2008). Furthermore, little is known about the relative importance of these factors. Below, we discuss prior evidence for each of these factors in turn, and then detail our approach to test them all together in a single analysis that also accounts for phylogenetic relatedness.
Pollination mode is predicted to affect population genetic structure, because pollen dispersal is critical to moving alleles between plant populations. Previous reviews have lumped different pollination mutualists together as animal pollination and compared them to wind pollination (Hamrick, Godt, & Sherman-Broyles, 1992; Loveless & Hamrick, 1984), revealing that wind tends to reduce genetic structure. Although the idea has not been tested on a broad scale, it has long been thought that different types of animal pollinators should also lead to differences in population genetic structure due to differences in their movement patterns and pollen carry-over capacity (Castellanos, Wilson, & Thomson, 2003). In fact, direct measures of pollen dispersal reveal that volant vertebrates and large bees transport larger proportions of pollen from individual trees to longer geographic distances than small insects (Dick et al. , 2008). Given these results, we predict that small insects restrict gene flow among plant populations and increase FST, compared to large insects, vertebrates, or wind.
Seed dispersal mode is also expected to influence plant population genetic structure because, like pollination mode, it directly affects the movement of alleles and thus gene flow among populations. Strong evidence suggests that limited dispersal increases fine-scale spatial genetic structure in plants (Gelmi‐Candusso, Heymann, & Heer, 2017) and in other organisms (Aguillon et al., 2017), which in consequence might scale up and lead to greater population genetic structure (Hamrick & Trapnell, 2011). In fact, reviews of the allozyme literature suggest that seed dispersal by wind and ectozoochory results in lower FST than dispersal by gravity and endozoochory due to greater gene flow among populations from long distance dispersal events (Hamrick & Godt, 1996; Hamrick, Murawski, & Nason, 1993). However, Duminil et al. (2007) found that dispersal mode was not a significant predictor of FST. The lack of consistency among studies encourages further work with larger sample sizes to fully understand the role of seed dispersal mechanisms on population genetic structure.
Unlike pollination and seed dispersal modes, the effect of mating system on plant population genetic structure has been well-established in previous broad-scale studies (Duminil et al., 2007; Loveless & Hamrick, 1984), which suggest that it is the most important predictor of FST variation. Mating system affects inbreeding, which lowers within-population variation, inflating between-population FST values (Charlesworth, 2003). Duminil, Hardy, and Petit (2009) found that the outcrossing rate and the inbreeding coefficient, which measures biparental inbreeding and selfing, are both significant predictors of FST in seed plants. Both selfing and inbreeding increase inbreeding depression and induce purging of deleterious alleles, reducing effective population size and increasing genetic drift, which can ultimately lead to fixation of different alleles in different populations (Angeloni, Ouborg, & Leimu, 2011; Wright, Ness, Foxe, & Barrett, 2008). In contrast, outcrossing increases gene flow within populations, potentially intensifying pollen-mediated gene flow among populations, which counteracts genetic drift and thus decreases population genetic structure (Duminil et al., 2009; Ellstrand, 2014).
Growth form is also an important predictor of population genetic structure. Broad-scale analyses (Duminil et al., 2009; Hamrick et al., 1992) have found strong associations between growth form and FST, with woody plants tending to have lower FST than herbaceous plants. The mechanism that causes this association is unclear, however, and may actually be driven by correlations between growth form and other factors. For example, Duminil et al. (2009) found that growth form only affects FSTindirectly, through its influence on outcrossing rate (tm) and inbreeding coefficient (FIS); woody growth form is associated with greater tm and lower FIS. However, Hamrick and Godt (1996) reviewed the allozyme literature for over 300 species and found that when considering outcrossing plants, woody plants show lower levels of FST than herbs, which suggests that growth form directly affects gene flow among populations, decreasing population genetic structure. This could be because in trees greater geographic distance is presumably required for genetic differences to be detected among populations than in herbs, given that trees are larger than herbs. Thus, when considered at similar geographic scales, we predict that herbs have populations with greater genetic differentiation than trees.
Finally, the latitudinal region in which a plant occurs could also affect its population genetic structure due to differences among regions in spatial and climatic landscapes. In general, geographic heterogeneity and seasonal asynchrony over short distances are considerably higher in the tropics than in the temperate zones (Esquerré, Brennan, Catullo, Torres‐Pérez, & Keogh, 2019; Ricklefs, 1977; Stein, Gerstner, & Kreft, 2014), which may act to disrupt mating among conspecific subpopulations, and thus limit gene flow (Martin, Bonier, Moore, & Tewksbury, 2009; Quintero, González-Caro, Zalamea, & Cadena, 2014). Additionally, genetic drift could have a more prominent role in the tropics than in the temperate zones, due to the fact that most species in the tropics occur at low population densities and thus should have lower effective population sizes than in temperate zones (Dick et al., 2008; ter Steege et al., 2013). In fact, although their sample size was limited and phylogenetic autocorrelation was not accounted for, Dick et al. (2008) found that tropical trees have on average higher FSTvalues than temperate trees. Given all of the above effects, we predict that FST is higher in the tropics than in the temperate zones.
Previous studies have not included all of the aforementioned factors together when modeling patterns of population genetic structure in seed plants (Duminil et al., 2007; Hamrick et al., 1992; Hamrick & Godt, 1996; Loveless & Hamrick, 1984; Nybom & Bartish, 2000). Furthermore, the most thorough study of FST in seed plants was over a decade ago (Duminil et al., 2007) and thus could not take advantage of the wealth of population genetic studies published since then. Here we reviewed publications to assemble a 337-species database of seed plants with the goal of evaluating the factors predicted to best explain variation in plant population genetic structure. We focused on studies that used nuclear markers because their genetic structure should reflect both pollen and seed movement (due to biparental inheritance), unlike chloroplast markers, which only reflect seed movement (due to maternal inheritance) (McCauley, 1994). We examined five ecological factors, including pollination mode, seed dispersal mode, mating system, growth form, and latitudinal region, while controlling for phylogenetic autocorrelation. We also accounted for variables pertaining to the sampling scheme that have been shown to affect FST values for plants (Nybom & Bartish, 2000) and other systems (Blasco-Costa & Poulin, 2013; Pascual, Rives, Schunter, & Macpherson, 2017; Riginos, Douglas, Jin, Shanahan, & Treml, 2011); namely, genotyping technique, distance between populations, and sample size. Using multiple regressions, we asked: (Q1) What set of life history traits promote population divergence in seed plants? (Q2) Do patterns of variation in FST differ between latitudinal regions? (Q3) What are the relative importance of these factors in explaining variation in FST?