Introduction
The divergence of functional traits is usually influenced by multiple factors, and disentangling the relative contributions of these factors is important for understanding the evolutionary drivers of traits. Fruit type is a fundamental trait in plant biology, as it is a key component for seed dispersal, thereby linking reproduction of individual plants to regeneration of ecological communities (Howe & Smallwood 1982; Jordano et al. 2007; Seale & Nakayama 2020). Fruits can be dichotomously classified as fleshy or dry. The composition of fleshy versus dry fruits varies greatly across communities and climate regions, with an increasing proportion of fleshy-fruited species occurring with both decreasing latitude and altitude at a global scale (Willsonet al. 1989; Almeida-Neto et al. 2008; Chen et al.2017; Albert et al. 2018; Zhao et al. 2018). Although many factors have been proposed to explain this pattern, the relative importance of each factor has not yet been resolved. Here, we analyze a very large dataset to disentangle the relative contributions to the variation in fruit type of three factors that are most often included as possible explanations for fleshy versus dry fruits: plant growth form, environmental/climate factors (as summarized by the climate region in which species occur), and phylogenetic conservatism (evolutionary history).
Spatial variation in climate mediates the latitudinal and altitudinal patterns in functional traits of plants (Reich & Oleksyn 2004; Moleset al. 2007; Diaz et al. 2016). In general, precipitation is positively associated with the proportion of fleshy fruits (Correaet al. 2015; Chen et al. 2017; Zhao et al. 2018), as water availability is a crucial factor for the development of fleshy fruits (Coombe 1976). High temperature is also associated with an increasing proportion of species producing fleshy fruits (Willsonet al. 1989; Zhao et al. 2018), possibly because succulent fleshy fruits cannot survive freezing conditions (Burke et al.1976). By analyzing the effect of 15 climatic variables on the latitudinal pattern in fruit types of 4008 Australian species, Chenet al. (2017) showed that wet, warm, and stable climates predicted a higher proportion of fleshy-fruited species. Furthermore, frugivory may be another possible explanation, as tropics usually hold more species and greater abundance of fruit-eating animals than temperate regions (Kissling et al. 2009; Onstein et al.2017).
Plant growth form is often a predictor of other functional traits, for example, seed mass, stem specific density, and leaf area (Moles et al. 2005; Diaz et al. 2016). Fruit type is also strongly related to plant growth form, showing that trees and shrubs are more likely to produce fleshy fruits, while herbs are more likely to produce dry fruits (Chen et al. 2004; Bolmgren & Eriksson 2005; Cortes-Floreset al. 2013). de Queiroz (2002) proposed that the cost of producing fleshy fruits may reduce seed output, which would have proportionally more negative effects on small and short-lived plants. Furthermore, in the context of apparency theory (Feeny 1976), large and long-lived woody species that produce fleshy fruits may better advertise to frugivores than herbaceous fleshy species. Using a global dataset, Moles et al. (2009) showed a remarkable relationship between latitude and plant height, indicating that more trees occurred in the tropics than in the temperate regions. In this way, the latitudinal gradient in fruit type could be a reflection of the latitudinal gradient in growth form.
Phylogenetic signal of specific functional traits may be a consequence of either adaptation among phylogenetically related species that share similar environments or phylogenetic conservatism (Blomberg et al. 2003). Although the shift between fleshy and dry fruit occurred independently in many plant lineages as a response to environmental filters (Bolmgren & Eriksson 2005), a significant phylogenetic signal can still be detected (Herrera 2002; Chen et al. 2017), indicating that phylogenetic conservatism is a potential factor explaining fruit type. We therefore may expect the different composition of fruit types among floras to reflect their evolutionary histories. Furthermore, plant growth form may be also phylogenetically controlled, as are many other plant functional traits such as seed mass (Moleset al. 2005), leaf traits (Pearse & Hipp 2009) and root traits (Valverde-Barrantes et al. 2017). Therefore, we further expected that the phylogenetic conservatism might play an important role in determining the divergence of fruit type.
In summary, plant growth form, climatic factors, and phylogenetic conservatism may account, at least in part, for the pattern of fleshy versus dry fruit types (Bolmgren & Eriksson 2005; Cortes-Floreset al. 2013; Chen et al. 2017). However, disentangling the relative contributions of each factor is a challenge. Two common methods have been performed to partition variance of a target functional trait among possible predictor variables. Phylogenetic generalized least squares models (PGLS) with the phylogenetic tree treated as the error covariance structure can be used to compare the relative contributions among predictor variables after accounting for phylogenetic non-independence of species; however, PGLS models do not directly give a way to partition sources of variation explaining a trait. Nested ANOVAs and linear mixed models with several taxonomical levels (e.g., order, family, and genus) treated as nested random effects can be used to disentangle the relative contributions among multiple predictors, including phylogeny (Chen et al. 2004; Valverde-Barranteset al. 2017). However, these models only consider the nested hierarchy of a few taxonomical levels, which oversimplifies the phylogeny of related taxa. Furthermore, interpreting the partition of contributions of multiple factors to the total explained variance (R2) in the linear mixed model may be confounded when there is correlation between predictor variables (e.g., latitude and growth form). Therefore, to assess the roles of plant growth form, climate region, and phylogenetic conservatism on whether species produce fleshy or dry fruit, we used a phylogenetic partial R2that measures the reduction of explained variance when each factor is removed from a full model containing all other predictors (Ives & Li 2018; Ives 2019). Thus, the partial R2 values give the amount of variation explained by a predictor by asking how much of the explained variation is lost when the factor is removed.
In this study, we give a quantitative analysis of the relative contributions of plant growth form, climate region, and phylogenetic conservatism to the variation of fruit type with a data set including 9,370 species from Southwest China which represent a tropical, subtropical, and temperate flora. We predicted that (i) woody species produce more fleshy fruits than herbs; (ii) more fleshy-fruit species occur in tropical than in subtropical and temperate floras climate regions; and (iii) fruit type is phylogenetically conserved. Because we found that phylogenetic conservativism in fruit type was extremely strong, we further investigated whether this was due to ancient or recent phylogenetic relationships towards the base or tips of the phylogeny, respectively. Finally, we investigated the phylogenetic conservatism of climate region and plant growth form: if fruit type were to show greater phylogenetic conservatism than either of these predictor variables, then it would not be surprising that they are poor predictors, because the greater phylogenetic conservatism of fruit type would limit its evolutionary response to the predictors.