Mutation hot spots in mammalian mitochondrial DNA
Nicolas Galtier1, David Enard, Yoan Radondy, Eric Bazin, and Khalid Belkhir
Author Affiliations
Centre National de la Recherche Scientifique, Unité Mixte de Recherche (CNRS UMR) 5171-“Génome, Populations, Interactions, Adaptation,” Université Montpellier 2, 34095 Montpellier, France
Abstract
Animal mitochondrial DNA is characterized by a remarkably high level of within-species homoplasy, that is, phylogenetic incongruence between sites of the molecule. Several investigators have invoked recombination to explain it, challenging the dogma of maternal, clonal mitochondrial inheritance in animals. Alternatively, a high level of homoplasy could be explained by the existence of mutation hot spots. By using an exhaustive mammalian data set, we test the hot spot hypothesis by comparing patterns of site-specific polymorphism and divergence in several groups of closely related species, including hominids. We detect significant co-occurrence of synonymous polymorphisms among closely related species in various mammalian groups, and a correlation between the site-specific levels of variability within humans (on one hand) and between Hominoidea species (on the other hand), indicating that mutation hot spots actually exist in mammalian mitochondrial coding regions. The whole data, however, cannot be explained by a simple mutation hot spots model. Rather, we show that the site-specific mutation rate quickly varies in time, so that the same sites are not hypermutable in distinct lineages. This study provides a plausible mutation model that potentially accounts for the peculiar distribution of mitochondrial sequence variation in mammals without the need for invoking recombination. It also gives hints about the proximal causes of mitochondrial site-specific hypermutability in humans.
Mitochondrial DNA sequence variation in animals is notoriously characterized by a high amount of homoplasy, i.e., phylogenetic/genealogic conflict between sites. This is true between species, decreasing the efficiency of mitochondrial markers for phylogenetic analyses (see Springer et al. 2001; Delsuc et al. 2003), and also within species, as reported in many population genetic and phylogeographic surveys (see Vigilant et al. 1991; Vandewoestijne et al. 2004). Given the prevalence of mitochondrial data in molecular biodiversity, it is essential to understand the reasons for such a high amount of discrepancy between sites and its consequences on the interpretation of data sets. Eyre-Walker et al. (1999) took this point of view when analyzing third-codon-position variations between 29 nearly complete human mitochondrial genomes. They argued that the high amount of homoplasy they found was due to recombination between mitochondrial lineages—a strong claim challenging the “dogma” of maternal, clonal mitochondrial inheritance in animals. This report initiated a controversy.
There are two major mechanisms potentially explaining the occurrence of homoplasy within species. The first one is recombination. When partial genetic exchanges occur between distantly related individuals, the various segments of the recombined molecules are phylogenetically incongruent, because they actually have distinct genealogical histories. Alternatively, homoplasy can be generated by convergence due to multiple mutations. If two distantly related individuals independently receive the same mutation at site i, then site i will wrongly support their grouping, in conflict with other sites in the data set. The high amount of homoplasy in mitochondrial DNA could therefore be due to the phylogenetic noise introduced by mutation hot spots. In principle, an obvious difference between the two models is the expected distribution of the number of distinct states taken by polymorphic sites: Mutation hot spots, not recombination, should generate three- or four-state polymorphisms. Mitochondrial DNA, however, undergoes more transitions (C↔T and A↔G changes) than transversions, so that a majority of two-state polymorphisms is expected under both the recombination and the hot spots hypothesis.
A number of studies have attempted to demonstrate the occurrence of recombination in animal mitochondria. Recombination first appeared supported in humans by linkage disequilibrium (Awadalla et al. 1999) and geographic (Hagelberg et al. 1999) analyses, but these studies were criticized (Kivisild and Villems 2000; Hagelberg 2003) and/or could not be reproduced when the data set increased in size (Ingman et al. 2000; Innan and Nordborg 2002). Given the cytological evidence for maternal mitochondrial inheritance in animals (see Birky 1995), several investigators called for prudence before invoking recombination and suggested that the mutational hypothesis should be favored until it is formally rejected (Hey 2000). Recent reports, however, provided indirect evidence for mitochondrial recombination in several animal species (Piganeau et al. 2004;Gantenbein et al. 2005; Tsaousis et al. 2005), as well as direct proof of paternal leakage (Schwartz and Vissing 2002) and subsequent recombination (Kraytsberg et al. 2004) in one human.
Curiously, despite the importance of the debate, the alternative mutation hot spots hypothesis has not been examined in depth. In their seminal article, Eyre-Walker et al. (1999) considered various mutational models potentially explaining homoplasy in humans, i.e., unidirectional and bidirectional mutation hot spots. They found that not one of them was supported by the data. Stoneking (2000) and Pesole and Saccone (2001) showed that very recent somatic and germline mutations in the human mitochondrial control region tend to occur at evolutionarily hypervariable sites, suggesting that these sites are true hot spots. Aside from these reports, the debate has focused on recombination, mutation hot spots being invoked when the recombination hypothesis was not firmly supported. In particular, no response has been provided to Eyre-Walker et al.'s (1999) rejection of various mutation hot spots models in human coding regions.
In this article, we take the point of view of trying to detect mutation hot spots from mitochondrial DNA sequence variation, and asking whether they can explain the high level of homoplasy observed within species. Mutation hot spots, if any, should result in the co-occurrence of polymorphisms between closely related species, assuming that a hot spot in species 1 is still hot in species 2. They should also imply a correlation across sites of within-species and between-species variability—a hot spot should contribute both to polymorphism and divergence and be variable both within and between species. By using mammals as a model taxon, we take a phylogenetic approach to check these predictions of the mutation hot spots hypothesis and to try to elucidate the causes of the high level of homoplasy in mitochondrial DNA.
+++++
+++++
