MBE Advance Access originally published online on May 16, 2006
Molecular Biology and Evolution 2006 23(8):1516-1524; doi:10.1093/molbev/msl013
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Research Article |
Spatial Covariation of Mutation and Nonsynonymous Substitution Rates in Vertebrate Mitochondrial Genomes
Oklahoma Biological Survey and Department of Zoology, University of Oklahoma
E-mail: rbroughton{at}ou.edu.
 |
Abstract |
|---|
 TOP Abstract IntroductionMaterials and MethodsResultsDiscussionAcknowledgementsReferences |
|---|
Mitochondrial genomes encode fundamental subunits of the basic energy producing machinery of eukaryotic cells that are under strong functional constraint. Paradoxically, these genes evolve rapidly in general, and there is substantial variation in evolutionary rates among genes within genomes. In order to investigate spatial variation in selection intensity, we conducted tests of neutrality using ratios of synonymous to nonsynonymous substitutions (dN/dS =
) on numerous protein gene segments from fishes and mammals. Values of were very low for nearly all genomic regions. However, values of both and dN varied in a clinal pattern with increasing distance from the light-strand origin of replication. Spatial heterogeneity of nonsynonymous substitution rates exhibits a significantly positive correlation with variation in mutation rates that are related to the mode of mitochondrial DNA replication. The finding that nonsynonymous substitution rates are proportional to mutation rates is expected if a majority of substitutions are selectively neutral or slightly deleterious. Spatial patterns of among-gene variation in nonsynonymous rates were highly similar between fishes and mammals, suggesting that forces governing mitochondrial gene evolution have remained relatively constant over 450 Myr of vertebrate evolution. Conservation of substitution patterns despite major shifts in thermal habit and metabolic demands among taxa implicates a conserved replication mechanism controlling relative mutation rates as a major determinant of mitochondrial protein evolution.
Key Words: natural selection • neutral theory • nonsynonymous substitution • mitochondria
|
Introduction |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionAcknowledgementsReferences |
|---|
Variation in animal mitochondrial genomes has been widely utilized both as a marker in evolutionary studies (Avise 2004
; Ballard and Whitlock 2004) and for understanding metabolic diseases (Wallace 1999), apoptosis (Green and Reed 1998), and aging (Balaban et al. 2005). Mitochondrial proteins are central to the cellular oxidative phosphorylation pathway and are functionally conserved across metazoan phyla (Gray et al. 1999). Functional and structural conservation is more striking among vertebrates where gene content and gene order are nearly invariant. Paradoxically, mitochondrial genes evolve rapidly, up to 10-fold faster than nuclear genes (Brown et al. 1979; Moriyama and Powell 1997), and evolutionary rates among genes vary substantially among genes across the genome (Lynch and Jarrell 1993; Kumar 1996). Rapid evolutionary rates relative to nuclear genes may be due in part to increased exposure to mutagenic oxygen radical species (by-products of electron transport) and reduced effective population size due to haploid maternal inheritance (Birky 2001). However, although there have been numerous studies of the relative phylogenetic utility of different mitochondrial protein genes (e.g., Russo et al. 1996; Zardoya and Meyer 1996; Miya and Nishida 2000), the factors governing variation in evolutionary rates among genes remain less clear.
In general, evolutionary rates are governed by stabilizing (negative) selection, mutation, and directional (positive) selection. Negative selection due to functional constraint is ubiquitous, but its intensity may vary as proteins differ in the number of substitutions that can be tolerated while still maintaining function. Although differential negative selection may allow variation in evolutionary rates, its primary action is to prohibit evolutionary change. Factors that actually cause evolutionary change include positive selection on changes that enhance protein function and fixation of mutations that provide little functional advantage or disadvantage by genetic drift. The relative importance of mutation and positive selection remains unresolved (Kimura 1983; Gillespie 1991; Li 1997), but these forces are predicted to produce different relationships between mutation rates and rates of gene evolution. If neutral mutation is an important driver of evolutionary change, synonymous as well as nonsynonymous substitution rates should be proportional to mutation rates at different loci, although nonsynonymous rates should be generally lower than synonymous rates due to negative selection. Alternatively, adaptive evolution is expected to be gene (or domain) specific, and rates of change driven by positive selection should be essentially independent of the mutation rate as long as mutation is not limiting (Gillespie 1991; Li 1997).
Studies of large numbers of genes have yielded unambiguous cases of positive selection, but these tend to be in the minority relative to genes under some level of negative selection (Endo et al. 1996; Hughes 1999; Clark et al. 2003). For mitochondrial genes, evidence of excess nonsynonymous polymorphism within species relative to divergence between species suggests a dominant role for negative selection (Ballard and Kreitman 1994; Nachman 1998; Rand and Kann 1998; Ballard 2000; Rand 2001). The ratio of the rate of nonsynonymous substitutions (dN) to synonymous substitutions (dS) is a widely used index of selection (denoted ) in interspecific comparisons, where < 1 indicates negative selection, = 1 indicates neutrality, and > 1 indicates positive selection (Yang and Bielawski 2000; Nielsen 2001). For most genes, is much less than 1, indicating strong negative selection, and combined analysis of 12 hominoid mitochondrial protein genes revealed the average to be between 0.04 and 0.05 (Yang et al. 2000), which is near the low end for loci examined to date. Negative selection notwithstanding, the continued rapid evolution of mitochondrial genomes suggests that mutation and/or positive selection are also important, and substantial rate heterogeneity among genes indicates that these factors act differentially across the genome.
Mutation rates are known to vary across the mitochondrial genome and are related to the mode of replication. In the standard model of mitochondrial DNA (mtDNA) replication (fig. 1), one DNA strand (denoted the heavy strand) is displaced and single stranded for a length of time that is determined by the distance from the light-strand origin of replication (OL) (Clayton 1982; Bogenhagen and Clayton 2003). Mutations occur preferentially on the heavy strand in both fishes and mammals (Tanaka and Ozawa 1994; Bielawski and Gold 2002; Raina et al. 2005), suggesting increased exposure to mutations while single stranded. Moreover, rates of synonymous substitutions at both 2-fold and 4-fold degenerate sites are positively correlated with the distance from OL (Reyes et al. 1998; Bielawski and Gold 2002; Faith and Pollock 2003). This indicates that mutation rates are proportional to the amount of time spent in the single-stranded state and vary as a cline with respect to genomic position.
|
The contiguous arrangement of a large block of intronless genes replicated as a single unit, that differentially accrue mutations as a function of genomic position, provides a unique and unprecedented opportunity to investigate the relationship among mutation rates, substitution patterns, and the role of natural selection. Here we characterize synonymous and nonsynonymous substitution rates among protein-coding segments of mitochondrial genomes from fishes and mammals. These parameters are related to genomic position and to relative mutation rates inferred independently via the mode of mtDNA replication. We hypothesize that if positive selection is important in the evolution of mtDNA, then the magnitude of dN should vary independent of the relative mutation rate and genomic position; whereas if neutral mutation is a major factor, dN and mutation rate should be positively correlated among genomic segments. Our results provide new insights into factors driving mtDNA evolution and explain rate heterogeneity among protein-coding genes across the genome as a consequence of the mechanism of replication.
|
Materials and Methods |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionAcknowledgementsReferences |
|---|
Taxa included in this study are shown in figure 2. Several mitochondrial genome sequences were determined in our laboratory. Briefly, genomes were amplified in 2 overlapping segments of approximately 9 kb with the proofreading Herculase polymerase (Stratagene, San Diego, California). These segments were gel purified and sheared to an average size of 1.5 kb via sonication. Sheared fragments were end repaired to create blunt ends and shot-gun cloned in the pGEM-3Z vector (Promega, Madison, Wisconsin). Approximately, 200 random clones were sequenced for each species yielding an average coverage depth of 5x. Contigs were assembled with Sequencher ver. 4.1.4 (Gene Codes, Ann Arbor, Michigan). Any remaining gaps were filled via direct polymerase chain reaction sequencing with primer pairs designed for specific gaps. Sequences were annotated with the DOGMA web application (Wyman et al. 2004
). Additional genome sequences were obtained from the Entrez Genome organelle archive at GenBank (fig. 2). Nucleotide sequences were aligned with ClustalX and modified by eye. Novel sequences have been deposited in GenBank under accession numbers AY341348, AY302574, AY366087, and DQ536421–DQ536432.
|
Concatenated data matrices of 12 heavy-strand protein genes, arranged in genomic order (from ND1 to Cytb), were used for phylogenetic analysis. ND6 was excluded because it is the only protein gene encoded by the light strand and is known to differ substantially in nucleotide composition. All stop codons and regions of reading frame overlap were excluded. An approximately 30-bp region of ambiguous alignment at the 3' end of ND5 in fishes was also excluded. Phylogenetic analysis employed heuristic maximum likelihood searches as implemented in PAUP* ver. 4b10 (Swofford 2002
). The evolutionary model general time reversible (GTR) + 
+ I was indicated as the best by Modeltest ver. 3.06 (Posada and Crandall 1998), and model parameter values were determined on trees obtained via maximum parsimony and minimum evolution using Log-Det distances. For both mammals and fishes, maximum likelihood recovered the same topology regardless of the tree used for model parameterization. We note that the recovered maximum likelihood tree for mammals (fig. 2) differs somewhat from relationships derived from 18 nuclear genes (Murphy et al. 2001). To assess the effect of these topological differences, we estimated dN and dS on both the mitochondrial tree and the nuclear tree pruned to include the mammal taxa analyzed here.
The number of synonymous substitutions per synonymous site (dS) and nonsynonymous substitutions per nonsynonymous site (dN) were estimated for several data partitions. Each data set was partitioned in 2 ways. In one, partitions included either whole genes or portions of genes such that each partition was roughly between 450 and 650 bp. For example, for fishes, the COI gene was divided into 3 partitions of 520 bp, the COII gene was left intact as a single partition of 690 bp, and the ND3 and ND4L genes were combined into a single partition of 642 bp. The portion of ATP8 that does not overlap with ATP6 was only 164 bp and was excluded. Because mitochondrial genes differ substantially in size, this partition scheme incorporated gene boundaries while maximizing similarity in the number of codons sampled from each segment. Analyses based on these partitions are likely to reflect gene- or domain-specific substitution patterns. The other set of partitions included 10 equally sized segments of roughly 1080 bp each without regard for gene boundaries. As the original matrix included all genes in genomic order from ND1 to Cytb, the designation of partitions 1–10 reflects their relative genomic position. These partitions should yield substitution patterns that are more reflective of genomic regions, rather than specific genes or domains, and their larger size should increase the statistical power to detect positive selection (see Anisimova et al. 2001).
Substitution rates were estimated independently for all partitions under topology-dependent codon-based models (Goldman and Yang 1994) with the codeml program in PAML ver. 3.14 (Yang 1997). Estimates of dN and dS for each partition employed the trees recovered in maximum likelihood analyses for the entire data sets (described above). Seven models, including M0, M1a, M2a, M3, M5, M7, and M8 (Yang et al. 2000; Wong et al. 2004), were examined on unpartitioned data. Each model employed a 60 x 60 codon frequency matrix (based on the vertebrate mitochondrial genetic code), with transition/transversion ratios (
) and dN/dS ratios () allowed to vary based on the data. Likelihood ratio tests (LRTs) showed that 2 models (M7 and M8), employing a beta distribution for among site variation in , explained the data significantly better than the other models (data not shown). Substitution rates for each partition were estimated as the sum of branch lengths for dS or dN across the entire tree. Parameters of the beta distribution (p and q) were estimated from the data. The beta distribution was partitioned into 10 discrete frequency classes of equal size with restricted to be between 0 and 1 (M7) or a similar beta distribution but with an additional (11th) class where is allowed to be >1 (M8). To assess the possibility of finding local optima, at least 2 runs of codeml were conducted on each partition of the fish data set. If the likelihood of M8 is significantly better than M7 in LRTs and of the additional class is >1, positive selection is indicated (Nielsen and Yang 1998).
Mutation rates across mitochondrial genomes are determined largely by the amount of time-specific regions spend in the single-stranded state. The time spent single stranded is related to the physical distance to OL and is defined in terms of the point at which the heavy strand is displaced as the replication fork proceeds in one direction and point at which it is made double stranded as synthesis proceeds back in the opposite direction (see fig. 1). We quantify the time spent single stranded with the value Dss. Dss for genomic regions located between OH and OL were calculated as twice the distance (in kb) from the midpoint of a partition to the midpoint of OL. The ND1 and ND2 genes are positioned on the far side of OL, and once they become single stranded, the lagging strand replication complex must proceed around nearly the entire genome before they are made double stranded. Dss for these regions was calculated as twice the distance from the midpoint of the segment to OL, subtracted from the whole-genome size. As a measure of the time spent single stranded, Dss therefore serves as an indirect measure of the relative rate of mutation among genomic segments.
|
Results |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionAcknowledgementsReferences |
|---|
Maximum likelihood estimates of phylogeny derived from the sequences of 12 concatenated mitochondrial protein genes are shown for the included fish and mammal taxa in figure 2. To test our hypothesis that evolutionary forces are influenced by genomic region, we divided the sequence data using 2 partition schemes. In one, the 12-gene data set was divided into 10 equally sized partitions (hereafter 10-part), and in the other, the data were divided into 19 partitions based on gene boundaries (hereafter 19-part) (tables 1 and 2). We examined the intensity and type of selection acting on each partition using maximum likelihood models to estimate dN and dS and their ratio
. Tables 1 and 2 list the results of these neutrality tests. For each partition, estimates of compared under 2 evolutionary models, one (M7) restricting to be between 0 and 1, with the other (M8) allowing an additional class of sites with > 1. Mean estimates of were all below 0.12. The few cases where the less restrictive model (M8) included additional site classes are indicated. In only 2 cases were the data significantly more likely under the M8 model, and there was only one case with > 1 for the extra frequency class. This is a single codon in partition 5 of the mammal 10-part data which had an = 2.122. This position is an asparagine residue at position 48 of human ATP8. The ATP8 protein is comprised of only 68 amino acids, and the entire gene tends to be highly variable among species. Despite this exception, the general result is that, on average, dN is approximately 10-fold lower than dS across mitochondrial genomes. These results suggest that the vast majority of codon sites in mtDNA are under negative selection.
|
|
To examine substitution rate heterogeneity across the genome, we compared dN and dS estimated for each partition with its genomic position in terms of time spent single stranded, Dss (figs. 3 and 4). For consistency of comparisons, all points represent substitution rates estimated under the M7 codon model. For both fishes and mammals and for both the 10-part and 19-part genomic partitions, the relationship between dN and Dss was significantly positive, and the model explains between 34% and 53% of the variance around the least squares regression line. The fit of the points to the regression line is lower in analyses of the 19-part data sets. This may be due, in part, to greater variance in likelihood estimates of dN based on fewer codons per partition. Smaller codon samples might be more sensitive to variation in selection intensity on individual gene regions or protein domains, whereas larger codon samples (10-parts) should be less biased approximations of average regional substitution rates.
|
|
There is an obvious outlier point (studentized residuals >2) in each of the dN plots. The far right-hand data point for the 10-part data (fig. 2A and B) represents the Cytb gene (including all but the first 57 bases at the 5' end), whereas the second point from the right for the 19-part data (fig. 2C and D) is the 5' half of Cytb. The correlation of dN with Dss increases substantially when these outliers were excluded (fish 10-part: R2 = 0.891, P < 0.001; fish 19-part: R2 = 0.494, P < 0.001; mammal 10-part: R2 = 0.837, P < 0.001; mammal 19-part: R2 = 0.517, P < 0.001). The fit of the data to the model (whether or not outliers were excluded) suggests that dN is strongly dependent on genomic position in all regions except the 5' portion of Cytb which may be influenced by additional factors. Two different trees, based on nuclear or mitochondrial sequences, were used for analysis of the mammal 10-part data (fig. 3B) (see Materials and Methods), but there was little difference in estimates of dN, suggesting the codon models are robust to modest differences in tree topology. These results suggest that a significant fraction of among-gene variation in nonsynonymous substitution rate is driven by concordant variation in mutation rates that has been previously established to vary clinally across the genome.
Results for the dS data (fig. 4) show the expected increase of dS with Dss for fishes but not for mammals where there was a slight negative trend. In all cases, there was greater variance among dS estimates than for dN. The lower fit to the model for dS was surprising given that dS is expected to more closely approximate the mutation rate. However, at the levels of evolutionary divergence exhibited by these taxa, it is possible that saturation of synonymous changes may obscure the true rates of synonymous substitution. To explore this possibility, we used the M7 model to estimate dN and dS on the 10-part partitions for a subset of more closely related taxa, a monophyletic group of 9 cypriniform fishes. Results of this analysis were very similar to the analyses for all fish taxa, with both dN and dS positively correlated with Dss (dN: R2 = 0.397, P = 0.051; dS: R2 = 0.379, P = 0.058). The mammal dS data were unexpected as there is less divergence across the mammal tree than among the fish taxa sampled, so saturation should be less pervasive. Because positive selection was negligible, we also analyzed the mammal 10-part data using a model with fewer free parameters. The M1a model allows only 2 classes of sites, negative selection (0) and neutral (1) in propotions 
0 and 1 = 1–0, respectively. In this case, results for dN were nearly identical to those with the M7 model, whereas dS exhibited an essentially flat relationship with Dss (dN: R2 = 0.451, P = 0.033; dS: R2 = 0.380, P = 0.058).
If mitochondrial genomes were evolving according to a molecular clock, we might expect a tight coupling of mutation and evolutionary rates. To assess whether these mitochondrial genomes are evolving in a clock-like manner, we conducted LRTs comparing likelihood values both with and without a clock constraint on the unpartitioned data matrices. The unconstrained trees had higher likelihoods (fish 2
lnL = 1885, df = 34, P < 0.001; mammal 2lnL = 257, df = 19, P < 0.001) and thus, the rate constancy was rejected. Thus, evolutionary rates may vary substantially on different branches of the trees, however, when total rates of change (sum of branch lengths) for nonsynonymous substitutions are compared between fishes and mammals, remarkably similar patterns of rate heterogeneity are revealed among the various partitions (e.g., compare patterns of fig. 2A and B). This suggests that forces influencing among-gene evolutionary rates are conserved among vertebrates.
|
Discussion |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionAcknowledgementsReferences |
|---|
Given their mode of replication, lack of recombination, and compact size, mitochondrial genomes provide a unique window on the evolution of protein genes. Our tests of neutrality using dN/dS ratios included a breadth of taxon sampling and a variety of genomic partitions not previously examined. Values of
were consistently low, suggesting strong negative selection as has been inferred in previous studies (Ballard and Kreitman 1994; Nachman 1998; Rand and Kann 1998; Ballard 2000; Yang et al. 2000; Rand 2001). However, values of did increase with Dss (see tables), similar to the increase of dN. A positive correlation between the fixation probability of nonsynonymous substitutions (estimated by ) and the mutation rate would be surprising. Yet at least some saturation of synonymous substitutions may make the ratio of dN/dS unduly influenced by the numerator, and we suspect that the observed increase of with time spent single stranded may be an artifact of this ratio. We suggest that the poor fit of mammal dS data to the replication model reflects greater divergence among taxa than those used in previous studies (Reyes et al. 1998; Bielawski and Gold 2002; Faith and Pollock 2003) and stochastic effects due to high rates of synonymous substitution and sparse taxon sampling. We demonstrate a strong positive relationship between dN and genomic region in both fish and mammal mtDNA. This relationship is particularly striking when the 5' region of Cytb is excluded. The explanation for reduced dN in part of Cytb is not clear, but dS from the same region was not unusually low, suggesting that the low dN is not due to a reduced mutation rate. The explanation may simply be that there is uniquely strong functional constraint on this portion of the Cytb protein such that there are fewer nonsynonymous sites free to vary.
Our results provide empirical evidence that rates of nonsynonymous substitutions are closely tied to genomic position and hence to independently estimated relative mutation rates. If differential selection, either positive or negative, played a major role in nonsynonymous rates among genes, then evolutionary rates should be driven by forces specific to protein function and should vary by gene or protein domain, independent of genomic location. As selection will act mainly on a protein's ability to interact with other proteins and function efficiently in oxidative phosphorylation, there is no a priori reason to expect that selection intensities should vary in a clinal manner across the genome. It is possible that selection could be dependent on mutation rates if mutations are limiting. If the rate of mutational input is too low, opportunities for positive selection will be limited by, and hence proportional to, mutation rates. However, in the present case, dS for even the most slowly evolving gene, COI, is fairly high (averaging about 50 substitutions per synonymous site among all taxa), suggesting that mutation is not likely to be limiting. Our results do not suggest that positive selection is unimportant but are consistent with a model in which adaptive evolution occurs in rare, episodic bursts on particular lineages. The positive correlation of nonsynonymous substitutions with genomic position does suggest that differential accumulation of nonsynonymous substitutions among loci, most of which are likely to be slightly deleterious, depends largely on variable mutation rates across the genome that are dependent on the time spent single stranded during replication.
There is a broad range of substitution models that may be applied to likelihood estimation of dN and dS. It remains possible that models other than the GTR + gamma and M7 codon models may provide more accurate estimates of dN and dS for our data. For example, Bielawski and Gold (2002) and Krishnan et al. (2004) found that symmetric but nonreversible models with fewer free parameters than the GTR model explained the data quite well. Nonetheless, we have applied several codon models on the unpartitioned data and on subsets of fish or mammal data, and the qualitative results remained unchanged. It therefore appears that the basic result of a positive correlation between dN and genomic position is robust to the model employed.
We note that there is currently some controversy over the model of mitochondrial replication. It has recently been suggested that the mechanism of mtDNA replication is strand coupled in which the heavy strand would not be single stranded for any length of time (Holt et al. 2000; Yang et al. 2002). If this model is the primary mechanism of mtDNA replication, the hypothesis that the observed gradient of substitution rates is based on time spent single stranded would be rejected. However, the standard strand-displacement replication model is supported by the majority of experimental evidence, and it is likely to be the primary mode of replication even if the strand-coupled mechanism if found to occasionally occur (Bogenhagen and Clayton 2003). Thus, the strand-displacement model remains the best hypothesis to explain our results.
The patterns of dN variation among gene regions are remarkably similar between the groups of fishes and mammals we examined. This similarity is observed despite the differences in evolutionary divergence among taxa in the 2 trees. Major differences in metabolic rates between poikilothermic fishes and homeothermic mammals are well known, and substantial variation in environmental conditions and metabolic demands exist within these groups. Forces driving differential rates of protein evolution in different parts of the genome thus exhibit extensive conservation over the approximately 450 Myr since ray-finned fishes and mammals diverged from their last common ancestor. We cannot rule out the possibility that establishment of gene order in the ancestor of vertebrates favored placement of genes less tolerant of nonsynonymous change in regions of lower mutation rate and those with relaxed functional constraint in regions of higher mutation rate (Cytb excepted). However, natural selection acting on individual loci under divergent environmental and physiological conditions is unlikely to result in such similar patterns of nonsynonymous rate heterogeneity. Mitochondrial genomes represent a natural experiment where mutational mechanisms are reasonably well-characterized, and the results indicate a strong dependence of long-term nonsynonymous substitution rates on relative mutation rates across the genome. Thus, the similarity of spatial heterogeneity in nonsynonymous substitution rates between fishes and mammals is likely the result of a conserved replication mechanism driving variation in region-specific mutation rates. This suggests that general genomic mechanisms may be as important as selection acting on individual genes in the evolution of mitochondrial proteins.
|
Acknowledgements |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionAcknowledgementsReferences |
|---|
We thank S. Richter for sequence assembly and annotation and J. Bielawski, G. Naylor, G. Wellborn, and L. Weider for helpful discussions or comments on an earlier version of the manuscript. This work was supported by National Science Foundation grant DEB-0108201 to R.E.B.
|
Footnotes |
|---|
Ziheng Yang, Associate Editor
|
References |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionAcknowledgementsReferences |
|---|
Anisimova M, Bielawski JP, Yang Z. 2001. The accuracy and power of likelihood ratio tests to detect positive selection at amino acid sites. Mol Biol Evol 18:1585–92.
Avise JC. 2004. Molecular markers, natural history, and evolution. Sunderland, MA: Sinauer.
Balaban RS, Nemoto S, Finkel T. 2005. Mitochondria, oxidants, and aging. Cell 120:483–95.[CrossRef][ISI][Medline]
Ballard WJO. 2000. Comparative genomics of mitochondrial DNA in Drosophila simulans. J Mol Evol 51:64–75.[ISI][Medline]
Ballard WJO, Kreitman M. 1994. Unraveling selection in the mitochondrial genome of Drosophila. Genetics 138:757–72.[Abstract]
Ballard WJO, Whitlock MC. 2004. The incomplete natural history of mitochondria. Mol Ecol 13:729–44.[CrossRef][Medline]
Bielawski JP, Gold JR. 2002. Mutation patterns of mitochondrial H- and L-strand DNA in closely related cyprinid fishes. Genetics 161:1589–97.
Birky CW. 2001. The inheritance of genes in mitochondria and chloroplasts: laws, mechanisms, and models. Annu Rev Genet 35:125–48.[CrossRef][ISI][Medline]
Bogenhagen DF, Clayton DA. 2003. The mitochondrial DNA replication bubble has not burst. Trends Biochem Sci 28:357–60.[CrossRef][ISI][Medline]
Brown WM, George M, Wilson AC. 1979. Rapid evolution of animal mitochondrial DNA. Proc Natl Acad Sci USA 76:1967–71.
Clark AG, Glanowski S, Nielsen R et al. (17 co-authors). 2003. Inferring nonneutral evolution from human-chimp-mouse orthologous gene trios. Science 302:1960–3.
Clayton DA. 1982. Replication of animal mitochondrial DNA. Cell 28:693–705.[CrossRef][ISI][Medline]
Endo T, Ikeo K, Gojobori T. 1996. Large-scale search for genes on which positive selection may operate. Mol Biol Evol 13:685–90.[Abstract]
Faith JJ, Pollock DD. 2003. Likelihood analysis of asymmetrical mutation bias gradients in vertebrate mitochondrial genomes. Genetics 165:735–45.
Gillespie JH. 1991. The causes of molecular evolution. New York: Oxford University Press.
Goldman N, Yang Z. 1994. A codon-based model of nucleotide substitution for protein-coding DNA sequences. Mol Biol Evol 11:725–36.[Abstract]
Gray MW, Burger G, Lang BF. 1999. Mitochondrial evolution. Science 283:1476–81.
Green DR, Reed JC. 1998. Mitochondria and apoptosis. Science 281:1309–12.
Holt IJ, Lorimer HE, Jacobs HT. 2000. Coupled leading- and lagging-strand synthesis of mammalian mitochondrial DNA. Cell 100:515–24.[CrossRef][ISI][Medline]
Hughes AL. 1999. Adaptive evolution of genes and genomes. New York: Oxford University Press.
Kimura M. 1983. The neutral theory of molecular evolution. Cambridge, UK: Cambridge University Press.
Krishnan NM, Raina SZ, Pollock DD. 2004. Analysis of among-site variation in substitution patterns. Biol Proced Online 6:180–8.[CrossRef][Medline]
Kumar S. 1996. Patterns of nucleotide substitution in mitochondrial protein coding genes of vertebrates. Genetics 143:537–48.[Abstract]
Li W-H. 1997. Molecular evolution. Sunderland, MA: Sinauer.
Lynch M, Jarrell PE. 1993. A method for calibrating molecular clocks and its application to animal mitochondrial DNA. Genetics 135:1197–208.[Abstract]
Miya M, Nishida M. 2000. Use of mitogenomic information in teleostean molecular phylogenetics: a tree-based exploration under the maximum-parsimony optimality criterion. Mol Phylogenet Evol 17:437–55.[CrossRef][ISI][Medline]
Moriyama EN, Powell JR. 1997. Synonymous substitution rates in Drosophila: mitochondrial versus nuclear genes. J Mol Evol 45:378–91.[CrossRef][ISI][Medline]
Murphy WJ, Elzirik E, Johnson WE, Zhang YP, Ryder OA, O'Brien SJ. 2001. Molecular phylogenetics and the origins of placental mammals. Nature 409:614–8.[CrossRef][Medline]
Nachman MW. 1998. Deleterious mutations in animal mitochondrial DNA. Genetica 102–3:61–9.
Nielsen R. 2001. Statistical tests of selective neutrality in the age of genomics. Heredity 86:641–7.[CrossRef][ISI][Medline]
Nielsen R, Yang Z. 1998. Likelihood models for detecting positively selected amino acid sites and applications to the HIV-1 envelope gene. Genetics 148:929–36.
Posada D, Crandall KA. 1998. Modeltest: testing the model of DNA substitution. Bioinformatics 14:817–8.
Raina SZ, Faith JJ, Disotell TR, Sligmann H, Stewart C-B, Pollock DD. 2005. Evolution of base-substitution gradients in primate mitochondrial genomes. Genome Res 15:665–73.
Rand DM. 2001. The units of selection on mitochondrial DNA. Annu Rev Ecol Syst 32:415–48.[CrossRef][ISI]
Rand DM, Kann LM. 1998. Mutation and selection at silent and replacement sites in the evolution of animal mitochondrial DNA. Genetica 102–3:393–407.
Reyes A, Gissi C, Pesole G, Saccone C. 1998. Asymmetrical directional mutation pressure in the mitochondrial genome of mammals. Mol Biol Evol 15:957–66.[Abstract]
Russo CA, Takezaki N, Nei M. 1996. Efficiencies of different genes and different tree-building methods in recovering a known vertebrate phylogeny. Mol Biol Evol 13:525–36.[Abstract]
Swofford DL. 2002. Phylogenetic analysis using parsimony (*and other methods). Version 4b10. Sunderland, MA: Sinauer.
Tanaka M, Ozawa T. 1994. Strand asymmetry in human mitochondrial DNA mutations. Genomics 22:327–35.[CrossRef][ISI][Medline]
Wallace DC. 1999. Mitochondrial diseases in man and mouse. Science 283:1482–8.
Wong WSW, Yang Z, Goldman N, Nielsen R. 2004. Accuracy and power of statistical methods for detecting adaptive evolution in protein coding sequences and for identifying positively selected sites. Genetics 168:1041–51.
Wyman SK, Jansen RK, Boore JL. 2004. Automatic annotation of organellar genomes with DOGMA. Bioinformatics 20:3252–5.
Yang MY, Bowmaker M, Reyes A, Vergani L, Angeli P, Gringeri E, Jacobs HT, Holt IJ. 2002. Biased incorporation of ribonucleotides on the mitochondrial L-strand accounts for apparent strand-asymmetric DNA replication. Cell 111:495–505.[CrossRef][ISI][Medline]
Yang Z. 1997. PAML: a program package for phylogenetic analysis by maximum likelihood. Comput Appl Biosci 13:555–6.
Yang Z, Bielawski JP. 2000. Statistical methods for detecting molecular adaptation. Trends Ecol Evol 15:496–503.[CrossRef][Medline]
Yang Z, Nielsen R, Goldman N, Krabbe Pedersen A-M. 2000. Codon-substitution models for heterogeneous selection pressure at amino acid sites. Genetics 155:431–49.
Zardoya R, Meyer A. 1996. Phylogenetic performance of mitochondrial protein-coding genes in resolving relationships among vertebrates. Mol Biol Evol 13:933–42.[Abstract]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  M. M. Fonseca, D. Posada, and D. J. Harris Inverted Replication of Vertebrate Mitochondria Mol. Biol. Evol., May 1, 2008; 25(5): 805 - 808. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
N. Howell, J. L. Elson, C. Howell, and D. M. Turnbull Relative Rates of Evolution in the Coding and Control Regions of African mtDNAs Mol. Biol. Evol., October 1, 2007; 24(10): 2213 - 2221. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||