Discussion
Across a broad taxonomic and geographic range of plant species in SW China, plant growth form and climate region had little power to explain whether a plant species had fleshy or dry fruit. In contrast, fruit type exhibited strong phylogenetic conservatism. This conservatism was distributed over the entire phylogeny; in other words, broad phylogenetic relationships established at the base of the phylogeny, as well as recent phylogenetic relationships towards the tips of the phylogeny, all contributed to the strong phylogenetic conservatism of fruit type.
Fleshy fruit pulp has been hypothesized to evolve as a defense against seed predation (Mack 2000), and fleshy fruit are secondarily associated with endozoochory to promote seed dispersal (Howe & Smallwood 1982; Onstein et al. 2017). Therefore, weak phylogenetic conservatism of fruit type might be expected because of the strong selection from seed predators and frugivore seed dispersers. In contrast to this expectation, a strong phylogenetic signal of fruit type was detected in this study which is consistent with some previous studies that focused at different scales of both geography and taxa (Herrera 2002; Chenet al. 2017). Fleming and Kress (2011) reported that the first appearance of the core angiosperm families producing fleshy fruits were earlier than the first appearance of the families of their major fruit eating animals. Therefore, the phylogeny signal residing towards the base of the phylogeny might be independent of coevolution with frugivores, and instead the later-appearing frugivores evolved to take advantage of the preexisting fruit types. Other fruit/seed traits, such as seed size, fruit size and fruit color appear to evolve predominately under the selection pressure of fruit-eating animals; these traits also show phylogenetic signal, which implies that selection may generate phylogenetic signal if related species are under the same selective pressures (Chen et al. 2004; Moles et al. 2005). In our study, the majority of families (67.8%) exclusively produce fleshy or dry fruits, with shifts between dry and fleshy fruit occurring in the remaining families.
We had expected both plant growth form and climate region to explain much of the variation in fruit type. Our analyses suggest that the reason they do not provide much explanation differs between the two factors. Like fruit type, growth form itself shows strong phylogenetic conservatism that is distributed over the phylogeny (Fig. 3). When phylogeny is not included, growth form explains 22.8% of the variation in fruit type, while this drops to 1.5% when phylogeny is included (Table 2). As another useful comparison, the partial R2lik for growth form is 1.7% when the full phylogeny is used, but this increases to 25.1% when the 0.67-threshold phylogeny is used which removes the phylogeny information in the top 1/3 of the phylogeny. These two patterns suggest that fruit type and growth form are evolving largely independently, and their apparent association in the raw data – with 55.6% of the wood species and only 9.5% of the herbaceous species having fleshy fruit – is due largely to the phylogenetic conservatism of both traits.
In contrast to growth form, the climate region in which species occur does not show strong conservatism. What conservatism tropical and temperate climate regions have occurs towards the base of the phylogeny, implying that there are major clades being associated with tropical and temperate regions, but at finer scales towards the tips of the phylogeny, the role of phylogeny diminishes (Fig. 3). Thus, the explanation for climate region having little effect on fruit type is that climate region is relatively phylogenetically labile, changing up the phylogeny more rapidly than fruit type, especially towards the tips.
In contrast to our study, previous studies have shown effects of climate region (or environmental factors in general) on fruit type (Willsonet al. 1989; Bolmgren & Eriksson 2005; Chen et al. 2017; Zhao et al. 2018). There are several possible explanations for our differing results. We confined our study to plants occurring in SW China, a region known for its biogeographic and topographic/environmental diversity (Zhu 2012). Because tropical, subtropical, and temperate climate regions are in close proximity, we suspect that dispersal of species among regions is rapid on the phylogenetic time scales represented by our study plants. This suggests that dispersal and local speciation may be responsible for the lack of phylogenetic conservatism of those traits that determine the climate region in which species live. Thus, the traits that determine success in a climate region are evolutionarily labile, and the trait of fruit type plays at most a minor role in determining success in climate regions.
There may also be methodological explanations for why our results differ from many previous studies, that have investigated phylogenetic patterns for traits using different taxonomic levels, for example in nested ANOVAs (Chen et al. 2004; Valverde-Barrantes et al. 2017). Our approach allows a finer-scale assessment of the role of phylogeny. For example, using the 0.67-threshold phylogeny, in which phylogenetic information above roughly the family level is removed, growth form was a moderately strong predictor of fruit type, explaining 25.1% of the variance. However, using the full phylogeny, this dropped to 1.7%. Therefore, if nested ANOVAs did not account for the fine relationships towards the tips of the phylogeny, then they would miss the full effects of phylogenetic conservatism.
Exploring the drivers of the geographic variation in plant functional traits, especially along latitude or/and altitude gradients, has long attracted attention in eco-biogeography (Reich & Oleksyn 2004; Moleset al. 2007; Moles et al. 2009; Diaz et al. 2016; Valverde-Barrantes et al. 2017). Several models have been introduced to account for phylogenetic non-independence among species, although they have not quantified the contribution of phylogeny itself to trait variation (Swenson & Enquist 2007; Chen et al. 2017). However, quantifying the effect of phylogeny in comparison to other factors is useful (Valverde-Barrantes et al. 2017; this study). Partial R2s can disentangle the relative contributions of both predictor variables (climate region and growth form in this study) and covariances (phylogeny and other possible random effects). Combined with the findings from recent empirical studies on other important plant functional traits (Moles et al. 2005; Swenson & Enquist 2007; Valverde-Barrantes et al. 2017), our results suggest that methods using the full phylogeny rather than earlier methods such as nested ANOVAs have more power to explore the phylogenetic patterns of functional traits.