It is now commonplace for studies of molecular biogeography to employ a diverse suite of genetic markers, including loci both in the mitochondrial genome (mtDNA) and throughout the nuclear genome (nuDNA). This variety of genetic information is, in many cases, now complemented with broad taxon sampling encompassing a large geographic scope. In most studies that employ a diverse array of genetic markers and a robust sampling effort, the patterns observed between different genetic marker types generally align (Avise 1994). This is true for comparisons between species as well as phylogeographic structure that arises within species – the localities that harbour deep splits between mtDNA clades also have corresponding differences in the nuclear genome (Zink & Barrowclough 2008). This observation is one reason the ‘barcoding of life’ project has proved successful: clades identified in mtDNA are generally concordant with other phenotypic and genetic information (e.g. 94% of taxonomic bird species in North America have concordant mtDNA clusters; Kerr et al. 2007). However, concordant patterns between mtDNA and nuclear DNA are not always observed (Funk & Omland 2003; Chan & Levin 2005). In fact, the number of studies that report discordant patterns between mtDNA and nuclear markers, while not large, is increasing, especially within the last decade, as more researchers have been able to use both types of markers in combination.
Discordance between mtDNA and nuDNA can be most broadly defined as a significant difference in the patterns of differentiation between these two marker types. Most commonly, these conflicts can be in the overall amount of differentiation or in how these markers reconstruct relationships among groups. This type of discordance is expected because the mitochondrial genome is haploid and uniparentally inherited in most animals (but see Hoeh et al. 1991), and therefore has a fourfold smaller effective population size (Hudson & Turelli 2003; Zink & Barrowclough 2008). This means that mtDNA will complete the process of lineage sorting, where ancestral polymorphisms are lost over time, faster than nuDNA, as this rate is inversely proportional to the effective population size (Funk & Omland 2003). While the inheritance properties of mtDNA make it more likely than any single nuclear marker to accurately reflect recent divergence (Zink & Barrowclough 2008), studies that rely solely on mtDNA to infer phylogenetic relationships risk generating gene trees that do not represent the true relationships among taxa (Edwards & Bensch 2009). The prevalence of incomplete lineage sorting in contributing to discordant patterns between mtDNA and nuDNA has been discussed extensively (Funk & Omland 2003; Zink & Barrowclough 2008; McKay & Zink 2010), and the primary resolution is that, where feasible, researchers should include multiple independent loci to generate robust phylogenetic relationships (Edwards & Bensch 2009).
Even if numerous nuclear loci are employed, mito-nuclear discordance can also arise if there are differences in how selection acts on the mitochondrial genome as compared to the nuclear genome or if there is biased movement of either marker type driven by demographic asymmetries, such as sex-biased dispersal (Rheindt & Edwards 2011). For instance, despite the long-held assumption that variation in mtDNA is primarily neutral, a number of studies have identified intra- and interspecific variation in the proteins encoded by genes in the mitochondrial genome that authors have attributed to natural selection (Bazin et al. 2006; Meiklejohn et al. 2007; Edwards 2009; Ballard & Rand 2005; Ballard & Melvin 2010; Scott et al. 2011). If selection for mtDNA variants varies geographically, then discordant patterns between mtDNA and nuDNA can arise (Irwin 2012). In addition to differences in the adaptive landscapes for nuDNA and mtDNA, demographic asymmetries can also create discordant patterns and distributions of these different marker types. For instance, female-biased dispersal or disparities in range size or abundance between hybridizing groups can promote the dispersal of mtDNA in the absence of concordant movement of nuDNA (Funk & Omland 2003).
Distinguishing between incomplete lineage sorting and these other types of discordance can be difficult (McKay & Zink 2010). One important distinction, however, is that discordance that arises from incomplete lineage sorting is not expected to leave any predictable biogeographic pattern (Fig. 1; Funk & Omland 2003). Therefore, in cases where there are strong geographic inconsistencies between patterns in mtDNA and nuDNA, incomplete lineage sorting can usually be ruled out. This type of discordance, referred to more generally as biogeographic discordance, can result from clines in mtDNA being displaced from nuclear DNA in both their location and/or their width (Fig. 2). Biogeographic discordance can be extensive, such as the complete replacement of mtDNA of one species by another (i.e. ‘mitochondrial capture’), or more limited, where mtDNA haplotypes show a higher frequency in a given population than would be expected from nuDNA markers.
Two general situations can lead to biogeographic discordance between mtDNA and nuDNA: following isolation and hybridization or in situ (i.e. secondary versus primary contact). Most of the taxa that display patterns of biogeographic mito-nuclear discordance are groups that were isolated for long periods of time and are either currently in secondary contact or have experienced range contact at some point in their past. During this period of isolation, it is assumed that divergent groups accumulated mutations in both their mitochondrial and nuclear genomes, which increased to high frequency via selection, drift or some combination of the two (i.e. ‘genetic draft’; Hudson & Turelli 2003). Upon secondary contact, these groups formed hybrid zones, interbreeding to varying extents, and mtDNA–nuDNA discordance was promoted by divergent patterns of gene flow between the two genomes.
It has also been suggested that mito-nuclear discordance can arise in the absence of geographic isolation, where mitochondrial types show strong frequency differences between localities that potentially arose in the face of gene flow (Irwin 2002; Ribeiro et al. 2011). In these cases, patterns in the nuclear genome, combined with the biogeographic history of the taxa, suggest a narrow mtDNA divide that may not be the product of geographic isolation followed by secondary contact. This pattern is consistent with a scenario where selection favours one mitochondrial variant over another in a given area; in some cases, these differences may be associated with important environmental characteristics (Cheviron & Brumfield 2009; Irwin 2012).
In many cases, discordant biogeographic patterns have been used to infer the potential drivers of discordance. For situations where the mtDNA of one taxon shows complete fixation in another or where a mtDNA cline centre is displaced and/or wider as compared to nuDNA, a number of processes have been inferred: (i) adaptive introgression of mtDNA; (ii) demographic disparities; (iii) sex-biased asymmetries; (iv) hybrid zone movement; (v) Wolbachia infection; and (vi) human introductions. Adaptive processes can create discordance if selection favours mutually beneficial mitochondrial variants and promotes introgression upon secondary contact. Demographic disparities can generate discordance if there are large differences in population or range size between two taxa, especially if there is the potential for very small population sizes to influence mtDNA frequency by sampling effects (i.e. genetic drift), promoting asymmetric introgression (i.e. Currat et al. 2008). A subset of more general demographic differences, systems with female-biased dispersal propensity, behavioural differences in mating likelihood and differential production of offspring can promote mtDNA introgression because of its matrilineal inheritance. Hybrid zone movement can also create discordance when the majority of nuclear markers (in addition to phenotypic traits) shift their geographic location, leaving a wake of mtDNA behind (Rohwer et al. 2001). In insects, Wolbachia infection is a potentially important driver of discordance, where mating incompatibilities can arise between individuals with and without this cytoplasmic endosymbiotic parasite and mtDNA hitchhikes (i.e. infected males mated with uninfected female are incompatible, whereas infected females mated with uninfected males suffer less fitness loss; Jiggins 2003). It has also been recognized that human actions can facilitate secondary contact and generate some of the demographic asymmetries outlined previously by moving individuals (i.e. Perry et al. 2001) or by facilitating interaction via habitat alteration, potentially generating discordance.
Biogeographic discordance is also distinguished if secondary contact and hybridization generate more structuring in mtDNA and/or narrower geographic clines as compared to nuDNA. These are likely produced by either nuclear introgression and/or sex-biased asymmetries. Sex-biased asymmetries in this context can be driven by male-biased dispersal, mating behaviour or sex-biased offspring production. This latter scenario (sex-biased offspring production) has been the focus of theoretical and empirical investigations, as it is a specific prediction of ‘Haldane’s rule’ (reviewed in Coyne & Orr 2004). This theory posits that, following secondary contact and interbreeding between divergent taxa, if one sex suffers a fitness loss, it will more often be the heterogametic sex. It follows that in those systems that where females are the heterogametic sex (i.e. ZW systems), as in Aves and Lepidoptera, mtDNA will be less likely to introgress between divergent groups as compared to other taxonomic groups with XY sex determination (such as mammals), and subsequently, mtDNA will have a narrow cline (Coyne & Orr 2004).
Given the increase in interest and the availability of molecular markers, biogeographic patterns of mito-nuclear discordance are being identified more readily in many systems (Edwards & Bensch 2009). However, in most cases, the processes driving such patterns are still unknown. Here we attempt to address some of these knowledge gaps by reviewing recent progress in our understanding of mito-nuclear discordance in animal taxa. Our goals in this synthesis are to: (i) review known cases of biogeographic mito-nuclear discordance in animal systems, (ii) to summarize the geographic patterns in each instance and (iii) to identify common drivers of discordance in various groups. Our treatment differs from previous articles in both scope and inclusiveness, as the primary criterion for our survey is only that the systems display a strong biogeographic signal of mito-nuclear discordance (see Box 1 for a discussion of previous treatments of mito-nuclear discordance). We focus on biogeographic discordance because these cases are much more likely to be associated with other complementary historical, biological and ecological information that can be used to reveal the underlying processes driving discordance.