Molecular Ecology (2006) 15, 2007–2012 doi: 10.1111/j.1365-294X.2006.02896.x © 2006 Blackwell Publishing Ltd Blackwell Publishing Ltd COMMENT Shared alleles in sympatric oaks: recurrent gene flow is a more parsimonious explanation than ancestral polymorphism C. LEXER,* A. KREMER † and R. J. PETIT † *Jodrell Laboratory, Royal Botanic Gardens, Kew, Richmond, Surrey TW9 3DS, UK, †INRA, UMR Biodiversité Gènes & Ecosystèmes, 69 Route d’Arcachon, 336112 — Cestas, France Keywords : ancestral polymorphism, hybridization, interspecific gene flow, oaks, Quercus , selection Received 22 September 2005; revision accepted 22 November 2005 A recent paper in this journal (Muir & Schlötterer 2005) addresses the likelihood of shared ancestral polymorphism vs. recurrent gene flow in two closely related sympatric European species, Quercus robur (pedunculate oak) and Quercus petraea (sessile oak) and concludes that the low genetic differentiation among these species results from shared ancestry rather than from high rates of gene flow. Although we agree that the two oaks might share alleles due to common ancestry, we consider that the classical hypothesis of weak interspecific barriers is more parsi- monious to explain the limited interspecific divergence observed at most loci. We further argue that these oaks represent one of the best examples of selection operating on a subset of loci to effectively maintain species integrity despite high promiscuity. Our arguments are supported by an abundant literature published in both specialized and more general journals, a re-analysis and re-interpretation of Muir & Schlötterer’s own data, and recent data on a genomic scan for species differentiation in oaks (Scotti- Saintagne et al . 2004). In contrast to what is observed at organelle markers (see below), differentiation between Q. robur and Q. petraea exists at some nuclear markers (Bodénès et al . 1997a; Gömöry 2000; Gömöry et al . 2001; Mariette et al . 2002; Scotti-Saintagne et al . 2004; Muir & Schlötterer 2005) whereas morphological data demonstrate that the two species are broadly consistent across their range (Kremer et al . 2002). Muir & Schlötterer identify three hypotheses that may explain the maintenance of interspecific nuclear and phenotypic divergence despite the presence of shared alleles: (1) low levels of interspecific gene flow, (2) selection operating on a subset of loci and effectively maintaining species integrity, despite important gene flow, and (3) shared ancestral polymorphism. However, they fail to address hypothesis (2). As we will argue below, consideration of the relative roles of gene flow and selection is essential. Moreover, their study was based on a limited number of loci. More convincing evidence to address this topic became available roughly at the same time the Muir & Schlötterer (2005) study was published (Scotti-Saintagne et al . 2004). Here, we contrast the results of these two studies and review the literature to confront the different hypotheses raised by Muir & Schlötterer. We argue that, whilst selection at a limited number of loci maintains species differentiation, a major proportion of the nuclear genome is permeable to interspecific gene flow. Our argu- ments are threefold: (i) interspecific differentiation is unevenly distributed within the nuclear genome, (ii) there is evidence from both studies that increased differentia- tion at a limited number of loci is due to selection, and (iii) the opportunity for interspecific gene flow has now been widely shown by controlled hybridization experiments and parentage analyses in natural populations. Genomic distribution of interspecific differentiation Muir & Schlötterer (2005) present pairwise F ST values for 20 individual microsatellite loci between Q. petraea and Q. robur. Most of their estimates differ significantly from zero although individual values are low (Fig. 1b; mean: 0.052 ± 0.057 SD, median: 0.034, range: 0.008 – 0.232). Interestingly, F ST varies over three orders of magnitude, and when the three most divergent loci are omitted (loci 7, 96 and 112), the average F ST decreases to 0.031 (median: 0.023). Hence, data interpreted as ‘genome-wide differentiation’ by Muir & Schlötterer is more likely to reflect large variation in F ST among different loci in the genome (Fig. 1). These findings are corroborated by the much larger ‘genome scan’ experiment by Scotti-Saintagne et al . (2004). Correspondence: Christian Lexer, Fax: +44 (0)20 83325310; E-mail: c.lexer@kew.org