4.1 Mito-nuclear discordance and sex-biased dispersal
At the inter-lineage scale, we observed more genetic structure at mitochondrial than at nuclear loci. This dissimilarity has been observed for numerous Procellariiformes species (e.g. Gangloff et al. 2013, Silva et al. 2015, Welch et al. 2011) as well as other organisms (see Toews & Brelsford 2012 for a review). One likely explanation is a difference in the pace of molecular evolution between mitochondrial and nuclear markers, with the latter having a slower substitution rate than the former, and the former being more polymorphic at the intra-specific scale (see Brown 1985), at least in birds (Helm-Bychowski 1984; Mindel et al. 1996). However, we did observe high levels of intra-lineage diversity for some nuclear loci such as βfib (see also Gangloff et al. 2013, Silva et al. 2015), and therefore the difference of structuration cannot be solely attributed to a difference of marker variability. We therefore suspect that incomplete lineage sorting and retention of ancestral polymorphisms at nuclear loci also contribute. Indeed, effective population size of mitochondrial DNA is four times smaller than that of nuclear DNA due to uniparental inheritance. Lineage sorting will therefore be faster in mtDNA than in nuDNA, being inversely proportional to the effective population size (Funk & Omland 2003). Incomplete lineage sorting is actually thought to be the main cause of mito-nuclear discordance when associated to a pattern of loss of geographic differentiation on nuclear markers (McKay & Zink 2010; Toews & Brelsford 2012). We also found patterns suggestive of introgression in nuclear markers. Hybridisation with introgression has been documented in shearwaters (Genovart et al. 2007, Gómez-Díaz et al. 2009), other Procellariiformes (Brown et al. 2010,) and other seabirds (Gay et al. 2009; Morris-Pocock et al. 2011; Pons et al. 2014). The likelihood of Indian Ocean petrels visiting breeding Atlantic Petrels may be supported by recent tracking of Pterodroma arminjoniana breeding on Round Island (Mauritius), which showed that some individuals foraged around South Trinidad Is, off Brazil, and even in the northern Atlantic (Booth Jones et al. 2017), although flight capacities of Pterodroma are far higher than Puffinus . Introgression can also blur phylogeographic signals by mixing alleles from distinct populations, and is considered as the second main cause of mito-nuclear discordance (McKay & Zink 2010). Incomplete lineage sorting and introgression are however difficult to distinguish, and additional unlinked markers would be required to disentangle these phenomena. Finally, as the mitochondrial markers represent only the female evolutionary history, sex-biased dispersal favoring females may alternatively explain why the population structure inferred based on nuclear markers conflicts with female-inherited mtDNA markers (see Petit & Excoffier (2009). These authors suggested that the markers associated with the most dispersing sex should better delimitate species, as they will show stronger intra-specific gene flow from colonizing lineages, reducing the effects of genetic drift and lowering the probability of fixating introgressed alleles. Dispersal was indeed stronger in females in some populations, particularly in the larger and the putatively ancestral lineage,nicolae . Sex-biased dispersal was however more uncertain forlherminieri and boydi , while for baroli dispersal was inferred to be male-biased. The sample size for baroli was theoretically large enough to robustly detect a bias inF ST, F Is, and Relatedness (Goudet et al. 2002). Sex-biased dispersal may therefore have further contributed to the observed mito-nuclear discordance, at least in some lineages.