Molecular Ecology (2014) doi: 10.1111/mec.12837 Understanding the spectacular failure of DNA barcoding in willows (Salix): Does this result from a trans-specific selective sweep? DIANA M. PERCY,*† GEORGE W. ARGUS,‡ QUENTIN C. CRONK,*† ARON J. FAZEKAS,§ PRASAD R. KESANAKURTI,§ KEVIN S. BURGESS,¶ BRIAN C. HUSBAND,§ STEVEN G. NEWMASTER,§ SPENCER C.H. BARRETT** and SEAN W. GRAHAM*† *Department of Botany, University of British Columbia, Vancouver, BC, Canada V6T 1Z4, †Biodiversity Research Centre, University of British Columbia, Vancouver, BC, Canada V6T 1Z4, ‡Canadian Museum of Nature, PO Box 3443 Stn “D”, Ottawa, ON, Canada K1P 6P4, §Department of Integrative Biology, University of Guelph, Guelph, ON, Canada N1G 2W1, ¶Department of Biology, Columbus State University, Columbus, GA 31907-5645, USA, **Department of Ecology & Evolutionary Biology, University of Toronto, 25 Willcocks Street, Toronto, ON, Canada M5S 3B2 Abstract Willows (Salix: Salicaceae) form a major ecological component of Holarctic floras and consequently are an obvious target for a DNA-based identification system. We sur- veyed two to seven plastid genome regions (~3.8 kb; ~3% of the genome) from 71 Salix species across all five subgenera, to assess their performance as DNA barcode markers. Although Salix has a relatively high level of interspecific hybridization, this may not sufficiently explain the near complete failure of barcoding that we observed: only one species had a unique barcode. We recovered 39 unique haplotypes, from more than 500 specimens, that could be partitioned into six major haplotype groups. A unique variant of group I (haplotype 1*) was shared by 53 species in three of five Salix subgenera. This unusual pattern of haplotype sharing across infrageneric taxa is suggestive of either a massive nonrandom coalescence failure (incomplete lineage sorting), or of repeated plastid capture events, possibly including a historical selective sweep of hap- lotype 1* across taxonomic sections. The former is unlikely as molecular dating indi- cates that haplotype 1* originated recently and is nested in the oldest major haplotype group in the genus. Further, we detected significant non-neutrality in the frequency spectrum of mutations in group I, but not outside group I, and demonstrated a striking absence of geographical (isolation by distance) effects in the haplotype distributions of this group. The most likely explanation for the patterns we observed involves recent repeated plastid capture events, aided by widespread hybridization and long-range seed dispersal, but primarily propelled by one or more trans-species selective sweeps. Keywords: chloroplast capture, DNA barcoding, Malpighiales, molecular dating, phylogeogra- phy, selective sweep Received 19 December 2012; revision received 29 May 2014; accepted 4 June 2014 This is often inferred by observations of strong incon- Introduction gruence between plastid and nuclear gene trees, or dis- A partial failure of plastid data to track species bound- agreements between gene trees and classically defined aries is fairly common in phylogenetic plant studies species (Xu et al. 2012; Yu et al. 2013). The processes (Percy et al. 2008; Starr et al. 2009; Hassel et al. 2013). underlying this incongruence likely contribute to the somewhat lower success of DNA barcoding markers as Correspondence: Diana M. Percy, Department of Life Sciences, an identification tool in plants vs. animals (Fazekas Natural History Museum, Cromwell Road, London SW7 5BD, et al. 2009), possibly reflecting more slowly evolving UK, Fax: +44 (0)2079425229; E-mail: [email protected] markers, or more problematic levels of hybridization, © 2014 John Wiley & Sons Ltd 2 D. M. PERCY ET AL. introgression or incomplete lineage sorting (Hollings- confusion from incomplete lineage sorting and facilitate worth et al. 2011). Here, we report on patterns of varia- species fingerprinting. However, comparative studies tion in DNA barcoding markers in the genus Salix. The have shown that while many plant groups may be ame- spectacular failure of DNA barcoding that we observe nable to reasonably precise species identification using in willows may require an explanation involving trans- DNA barcoding, others are less so (reviewed in Hol- specific selection. This in turn may have important con- lingsworth et al. 2011). Indeed, it is now acknowledged sequences for the interpretation of incongruence in that plant species are generally harder to barcode than other taxonomically complex plant groups. most animal species. This may reflect various phenom- Willows (Salix; Salicaceae), a genus of shrub and tree ena, including large effective population sizes, periods species, are a significant component of Holarctic ecosys- of rapid speciation, larger disparity in dispersal rates of tems (Ager & Phillips 2008; Argus 2010; Myers-Smith paternal and maternal haploid genomes and higher lev- et al. 2011). They are important indicators of riparian els of hybrid speciation or introgression (Fazekas et al. habitats, and many ecological studies include estimates 2009; Hollingsworth et al. 2011). of the diversity and abundance of willows because of One or more of these phenomena may have an their contribution to ecosystem function, community impact on the success of DNA barcoding in willows. dynamics and assemblage (Myers-Smith et al. 2011). Our overall goal in this study was to document the They are also extensively planted for habitat restoration, extent to which plastid DNA barcodes (CBOL Plant erosion control, and windbreaks, and the physiological Working Group 2009) can be implemented in Salix. The adaptations and ecological resilience of willows make more specific aims of this study were to: (i) establish them valuable species for use in conservation and envi- the extent of species identification that is possible ronmental projects (Kuzovkina & Quigley 2005; among western North American Salix using DNA bar- Kuzovkina & Volk 2009). code markers; (ii) to document haplotype patterns by The genus contains ~450 species worldwide, includ- sampling widely and deeply (within and among spe- ing localized and widespread species with extensive cir- cies); (iii) to establish a probable diversification time- cumpolar distributions. However, Salix species are scale among plastid haplotypes (e.g. in the context of notoriously difficult to identify based on morphology. historic environmental and climatic shifts in North Many species are vegetatively similar and exhibit sub- America); and (iv) to explore and evaluate hypotheses stantial heteroblasty (the production of markedly differ- for the unusual patterns of molecular diversity that we ent juvenile and adult leaves; Rechinger 1992; Zotz et al. observed. 2011). Furthermore, Salix populations are dioecious, and taxonomic keys often require examination of both sta- Methods minate and pistillate individuals. In many species, flow- ering (catkin production) occurs before leaf production, Specimen sampling presenting an additional challenge when using keys that consider both vegetative and reproductive traits. Salix species are subdivided into five subgenera (four Collectively, these factors can limit the characters avail- according to Chen et al. 2010) and are mostly found in able to identify individuals sampled at a single period arctic, boreal and temperate regions (Argus 2010). Salix of development. Finally, hybridization may blur bound- is a large, diverse and widespread genus, and our sam- aries between individual species, and hybrid offspring pling covers a wide range of the geographic (from can exhibit highly variable morphologies (Mosseler North America, Asia, and Europe) and phenotypic 1990; Hardig et al. 2000). diversity and includes a fair representation of the broad The development of a molecular identification system taxonomic diversity of the genus, with representatives for willows, such as DNA barcoding (CBOL Plant from 27 of the ~40 sections (Table 1). Our species sam- Working Group 2009), would be a useful tool for pling focused on western North America, the most spe- applied and basic research in the genus. The organellar cies-rich region of the North American willow flora. We genomes (plastid or mitochondrial) are currently the sampled extensively within British Columbia, the prov- source of choice for plant and animal DNA barcodes. ince with the greatest willow diversity in Canada, with These genomes are haploid and typically uniparentally supplemental sampling in adjacent regions (e.g. Yukon, transmitted and have smaller effective population sizes Alberta) and more widely across northern temperate than those in the diploid nuclear genome (and the dioe- regions, including species from central and eastern Can- cious sexual system of Salix would be expected to ada, the United States, Europe, Mexico and Japan. We reduce effective population size further, Maynard Smith sampled a total of 546 individuals, representing 71 spe- 1978). Loci in organellar genomes therefore undergo cies, 10 identified hybrids and an additional 10 more rapid coalescence, which in turn should reduce individuals that could not be identified to species due © 2014 John Wiley & Sons Ltd HAS A SWEEP FOILED DNA BARCODING IN WILLOWS? 3 Table 1 Salix subgenera and sections (according to Flora of Our sampling included several voucher specimens col- North America, Argus 2010) sampled for this study are shown lected between 1946 and 2000; individual herbarium (species sampled and number of individuals sampled per samples of Salix with well preserved green leaves were taxon relative