Taxon-specific or universal? Using target capture to study the evolutionary history of rapid radiations

Abstract

Target capture has emerged as an important tool for phylogenetics and population genetics in nonmodel taxa. Whereas developing taxon-specific capture probes requires sustained efforts, available universal kits may have a lower power to reconstruct relationships at shallow phylogenetic scales and within rapidly radiating clades. We present here a newly developed target capture set for Bromeliaceae, a large and ecologically diverse plant family with highly variable diversification rates. The set targets 1776 coding regions, including genes putatively involved in key innovations, with the aim to empower testing of a wide range of evolutionary hypotheses. We compare the relative power of this taxon-specific set, Bromeliad1776, to the universal Angiosperms353 kit. The taxon-specific set results in higher enrichment success across the entire family; however, the overall performance of both kits to reconstruct phylogenetic trees is relatively comparable, highlighting the vast potential of universal kits for resolving evolutionary relationships. For more detailed phylogenetic or population genetic analyses, for example the exploration of gene tree concordance, nucleotide diversity or population structure, the taxon-specific capture set presents clear benefits. We discuss the potential lessons that this comparative study provides for future phylogenetic and population genetic investigations, in particular for the study of evolutionary radiations.

Keywords: Tillandsia; Bromeliaceae; phylogenomics; plant radiation; population structure; target capture.

FIGURE 1

Effects of (a) putative gene copy number, (b) gene GC content, (c) average exon size, and (d) per cent of identity on bait efficiency in Bromeliad1776 bait set, measured as the number of high‐quality reads uniquely mapping to bait target region across samples. Continuous variable was binned and y‐values higher than 1000 excluded for visualization in b–d

FIGURE 2

Coalescent‐based species trees generated ASTRAL‐III for samples enriched with Bromeliad1776 (left) and Angiosperms353 (right, flipped for mirroring), on 269 and 1600 genes for each set, respectively. Node values represent local posterior probabilities (pp) for the main topology and are equal to 1 unless noted otherwise. Pie charts at the nodes show levels of gene tree discordance: the percentages of concordant gene trees (blue), the top alternative bipartition (green), other conflicting topologies (red) and uninformative gene trees (grey). At bottom, length and average bootstrap support for gene trees from either data set, according to the design of the bait set used for enrichment: Angiosperms353 (right) and Bromeliad1776 (left). Each gene was considered a single genomic window

FIGURE 3

Principal component analysis (PCA) plot for samples of Tillandsia subgenus Tillandsia enriched with two bait sets: (a) Angiosperms353 (1025 variants); (b) Bromeliad1776 (32,941 variants). Colours indicate different species according to legend

FIGURE 4

Population structure of 5 Tillandsia subgenus Tillandsia species from 14 sampling locations inferred with the ADMIXTURE software. Samples were enriched with either of two bait sets: Angiosperms353 (9804 variants after LD‐pruning) and Bromeliad1776 (42,613 variants after LD‐pruning), showing values of K = 2 to K = 9. Colours represent genetically differentiated groups, while each accession is represented by a vertical bar

FIGURE 5

Distribution of Tajima's D and synonymous (π _S) nucleotide diversity within each species for the Bromeliad1776 kit