Prezygotic barriers effectively limit hybridization in a rapid evolutionary radiation

Introduction

A growing body of research suggests that hybridization is more prevalent across the tree of life than previously thought, challenging traditional views of speciation as a strictly bifurcating process (Mallet et al., 2016; Dagilis et al., 2022). Genome-scale data have illuminated the extent to which ancient hybridization has reshaped the genomes of many extant species (Moran et al., 2021), including our own (Sankararaman et al., 2016), spurring interest in the evolutionary causes and consequences of hybridization.

Hybridization is common in plants (Stebbins, 1969; Whitney et al., 2010) and has shaped genetic variation across numerous plant lineages (Arnold, 1997). However, varying rates of hybridization among different plant clades suggest this process has played a more significant role in some lineages than others (Whitney et al., 2010; Barker et al., 2016). In angiosperms, hybridization frequently precedes evolutionary innovations and the origin of new lineages. Allopolyploidization, a prevalent mode of hybrid speciation in plants, is implicated in the origin of over 10% of species in a survey of 47 vascular plant genera (Barker et al., 2016). By contrast, homoploid hybrid species, which arise without a change in ploidy, are thought to be rarer and require strong extrinsic barriers to occur (Buerkle et al., 2000; Abbott et al., 2010).

Introgression, another common outcome of hybridization, has been the focus of recent research on adaptation and speciation in plants (Le Corre et al., 2020; Todesco et al., 2020; Nelson et al., 2021). Introgression involves the transfer of genetic material through backcrossing, where hybrids mate with pure individuals. The extent and direction of introgression can inform our understanding of reproductive isolation and adaptation. For example, asymmetric introgression, where backcrossing favors one parental species over the other, can reflect asymmetry in reproductive isolation (Arnold et al., 2008). Additionally, the proportion of the genome inherited through interspecific gene flow can shed light on the strength and genetic basis of reproductive barriers (Borge et al., 2005; Currat & Excoffier, 2011), though neutral abiotic factors and demographic and genetic processes may also influence the extent and direction of introgression (Currat et al., 2008; Bertola et al., 2020).

Recent studies increasingly highlight the adaptive role of introgression in transferring beneficial alleles among species (Stankowski & Streisfeld, 2015; Lewis et al., 2019; Todesco et al., 2020; Baiz et al., 2021). Adaptive introgression is believed to underlie the rapid diversification of many adaptive radiations, including Darwin's finches (Grant & Grant, 2019), Heliconius butterflies (Pardo-Diaz et al., 2012), and Rhagoletis fruit flies (Feder et al., 2005). While reproductive barriers are critical to the balance between hybridization and speciation, we are still in the early stages of identifying the specific barriers that influence the frequency and outcome of hybridization over macroevolutionary timescales.

Phylogenomic datasets and coalescent-based phylogenetic approaches have significantly enhanced our ability to resolve complex evolutionary histories, including rapid and recent evolutionary radiations (Meyer et al., 2017; Irisarri et al., 2018; Suvorov et al., 2022), providing valuable resources for subsequent analyses of introgression. Additionally, genome-scale datasets and advances in analytical approaches have improved our ability to document fine-scale patterns of hybrid ancestry across the genomes of diverse species (Sankararaman et al., 2016; Schumer et al., 2018) and characterize patterns of introgression across entire clades (Morales-Briones et al., 2018a; Suvorov et al., 2022). There are numerous approaches that detect introgression within a phylogenetic context, ranging in complexity from ‘test-based’ approaches that fit a simple phylogenetic tree model to a species quartet (e.g. D-statistic) to parameter-rich ‘model-based’ approaches that fit complex historical graph models (e.g. phylogenetic network inference). The D-statistic and related tests (Green et al., 2010; Durand et al., 2011) are performed on three in-group taxa or populations at a time, and an outgroup taxon or population is used to polarize ancestral states. To characterize introgression across a phylogeny, tests may be conducted exhaustively across all possible trios of populations or taxa. By contrast, model-based Bayesian and likelihood methods utilize data from all samples simultaneously to extensively explore phylogenetic network space and jointly estimate the species tree topology and hybridization events. While powerful, network inference is computationally challenging, and current approaches are limited to small datasets. Alternatively, systems of three and four-sample tests are computationally tractable and generally more robust to common model violations (Hibbins & Hahn, 2022). Still, they can become difficult to interpret when large numbers of correlated tests are significant. In this study, we leverage both model-based and test-based methods to detect introgression across the spiral ginger (Costus, Costaceae) radiation of Central and South America (hereafter referred to as American Costus).

Costus represents one of the fastest known plant radiations, with recent estimates suggesting that 78 species have arisen over the last c. 3 million years (Myr) (Vargas et al., 2020; Maas et al., 2023). Recent phylogenetic analyses (Valderrama et al., 2020; Vargas et al., 2020) have provided well-supported hypotheses for this clade despite extensive genealogical discordance. However, the extent to which hybridization contributed to observed levels of gene tree conflict is unknown. American Costus species are widely interfertile in experimental crosses (Kay & Schemske, 2008) and retain consistent ploidy and high-genomic synteny across the deepest phylogenetic nodes (Harenčár et al., 2023). Whereas Costus species are often found in broad geographic sympatry (Vargas et al., 2020), hybrid zones are rare in the field (Chen & Schemske, 2015), forming only occasionally in areas of anthropogenic disturbance (Sytsma & Pippen, 1985; Surget-Groba & Kay, 2013). Hybridization may be curtailed in Costus by predominantly prezygotic reproductive isolating barriers, especially ecogeographic and floral isolation, which have been studied extensively in the genus (Kay & Schemske, 2003; Kay, 2006; Chen & Schemske, 2015). Prior research has shown that potential pollen flow among sympatric species is significantly reduced between species with divergent pollination syndromes (bee vs hummingbird) (Kay & Schemske, 2003) and even between morphologically divergent species that share pollinators (Kay, 2006; Chen, 2013). A recent ancestral state reconstruction of pollination syndrome suggests that there have been numerous independent transitions from euglossine bee to hummingbird pollination characterized by evolutionary convergence in key floral traits that increase visitation by hummingbirds or deter bees (Kay & Grossenbacher, 2022).

We used phylogenetic approaches to detect clade-wide signatures of introgression across 54 American Costus taxa using a phylogenomic dataset of 756 genes. Our analyses ranged from simple four-taxon tests to parameter-rich model-based approaches. We detected evidence for a moderate number of introgression events in American Costus, including one event that mapped to one of the earliest divergences in the radiation. All but one of these events occurred between taxa or lineages that share the same pollination syndrome (e.g. both bee-pollinated or both hummingbird-pollinated), suggesting that pollinator specialization plays an important role in maintaining species boundaries.