Domestication and the mitochondrial genome: comparing patterns and rates of molecular evolution in domesticated mammals and birds and their wild relatives.

Studies of domesticated animals have led to the suggestion that domestication could have significant effects on patterns of molecular evolution. In particular, analyses of mitochondrial genome sequences from domestic dogs and yaks have yielded higher ratios of non-synonymous to synonymous substitutions in the domesticated lineages than in their wild relatives. These results are important because they imply that changes to selection or population size operating over a short timescale can cause significant changes to the patterns of mitochondrial molecular evolution. In this study, our aim is to test whether the impact on mitochondrial genome evolution is a general feature of domestication or whether it is specific to particular examples. We test whether domesticated mammals and birds have consistently different patterns of molecular evolution than their wild relatives for 16 phylogenetically independent comparisons of mitochondrial genome sequences. We find no consistent difference in branch lengths or dN/dS between domesticated and wild lineages. We also find no evidence that our failure to detect a consistent pattern is due to the short timescales involved or low genetic distance between domesticated lineages and their wild relatives. However, removing comparisons where the wild relative may also have undergone a bottleneck does reveal a pattern consistent with reduced effective population size in domesticated lineages. Our results suggest that, although some domesticated lineages may have undergone changes to selective regime or effective population size that could have affected mitochondrial evolution, it is not possible to generalize these patterns over all domesticated mammals and birds.

Introduction

Does domestication influence rates and patterns of molecular evolution? Analysis of single-nucleotide polymorphisms from the dog nuclear genome suggests a higher ratio of non-synonymous to synonymous alleles relative to the wolf, which has been interpreted as the signature of relaxed selection and reduction in effective population size associated with domestication (Cruz et al. 2008). Similarly, studies have found that rice (Lu et al. 2006) and a laboratory strain of yeast (Gu et al. 2005) have higher ratios of nonsynonymous to synonymous changes (dN/dS) than their wild relatives. Comparison of dog, yak, pig, and silkworm mitochondrial genomes with their respective wild relatives have also shown that the domesticated lineages have higher dN/dS than their wild relatives (Björnerfeldt et al. 2006; Wang et al. 2011; Hughes 2013).

These studies raise the possibility that domestication has significant effects on molecular evolution. If true, this would demonstrate that rates and patterns of molecular evolution are labile on relatively short timescales. It is widely assumed that all domesticated lineages were established less than 15,000 years ago, so any detectable effects of domestication on molecular evolution must be due to recent changes having a significant and measurable impact on molecular evolution. Domesticated lineages might therefore provide an interesting case study for the influence of population changes or alteration of selective regime on patterns and rates of molecular evolution. On a practical level, observation of widespread impacts of domestication on molecular evolution would suggest that caution must be exercised when estimating the date of origin of domesticated lineages from molecular data or when including sequences from domesticated lineages in dating analyses.

Broadly speaking, there are three ways that domestication could affect patterns of molecular evolution: artificial selection, relaxed selective constraints, and reduced effective population size in domesticated lineages. Direct or indirect selection for traits during domestication may increase the rate of nonsynonymous substitutions at specific loci associated with selected traits (e.g., coat color in pigs) (Fang et al. 2009). Similar effects may be detected in loci that are linked to sites under artificial selection, as selective sweeps can drive fixation of neutral or nearly neutral linked alleles (Innan and Kim 2004; Kim and Nielsen 2004; Rubin et al. 2010). Artificial selection could also have genome-wide impacts on the rates and patterns of molecular evolution if selection for novelty promotes the evolution of mechanisms that increase the production of variation. For example, Burt and Bell (1987) found that domesticated mammals have higher chiasmata frequencies than other mammals with similar ages of maturity, which they suggested reflects “adaptation to an environment characterized by intense selection in small populations for novel combinations of traits.” Otto and Barton (2001) also found several examples across different kingdoms that suggest a link between artificial selection regimes and increased recombination. Strong directional selection pressure and/or reduced effective population size could potentially increase the mutation rate (Sniegowski et al. 1997; Lynch 2010, 2011), though any increase in production of novel traits comes at the cost of a higher rate of deleterious mutations (King and Kashi 2007). While mitochondrial genomes of mammals and birds rarely if ever recombine, if domestication does indirectly select for generation of variation through recombination or mutation (Burt and Bell 1987; Denamur and Matic 2006; Dobney and Larson 2006; Bromham 2009), it could potentially influence rates of molecular evolution.

Relaxed selection could influence molecular evolution in domesticated lineages by permitting a greater proportion of nonsynonymous mutations to persist. Some of the traits that experience relaxed selection during domestication may be related to changes in environmental conditions and lifestyle (Clutton-Brock 1999; Björnerfeldt et al. 2006; Driscoll et al. 2009; Rubin et al. 2010). For example, the higher proportion of nonsynonymous changes in the mitochondrial genomes of dogs (Björnerfeldt et al. 2006) and domestic yaks (Wang et al. 2011) has been attributed to relaxed selection on metabolic efficiency in domesticated lineages, due to humans changing their habitat, selecting for tameness, and providing protection from predators.

Domesticated populations may often experience reductions in effective population sizes due to inbreeding and genetic bottlenecks (Vilà et al. 2005; Xia et al. 2009). Reduced effective population size increases the chance of fixing slightly deleterious mutations through drift, which should be reflected in increased dN/dS (Kimura and Ohta 1971; Ohta 1992). This effect is thought to account for patterns such as the correlation between body size and dN/dS in mammals (Nikolaev et al. 2007; Popadin et al. 2007; Nabholz et al. 2013). Domesticated lineages may undergo extreme bottlenecks on foundation. However, the domestication process has likely occurred over long periods of time and may have included few or many bottlenecks interspersed with introgression and population expansion (Allaby et al. 2008; Meyer and Purugganan 2013). This process could allow a lineage to recover from dramatic bottlenecks (Vilà et al. 2005). For example, although Taurine cattle may have originally descended from less than one hundred female founders (Bollongino et al. 2012), the high level of current genetic diversity has led to estimates of an ancestral wild population of 90,000 (MacEachern et al. 2009). Ongoing selective breeding and narrowing of the breeding pool may have also reduced effective population size in some domesticated lineages (Medugorac et al. 2009). For example, dogs are likely to have experienced a prehistoric bottleneck from wolves (Vilà et al. 1997), but it is likely that some dog populations have experienced more severe bottlenecks in recent history from breeding pressure (Wayne and Ostrander 2007).

Changes in population structure or conditions during domestication may be expected to have significant impacts on molecular evolution. However, the generality of the relationship between domestication and patterns of molecular evolution has not been established. Is it confined to a few well-studied examples, or is it a more general feature of all domesticated lineages? Not all studies support higher nonsynonymous rates in domestic lineages. For example, Rokas (2009) found a lower dN/dS in the proteome of a domesticated fungus compared with its wild relative. Here, we aim to ask whether increased dN/dS is a general feature of the mitochondrial genomes of domesticated lineages by comparing sequences from the maximum available number of phylogenetically independent comparisons of domesticated mammals and birds and their wild relatives.

We focus on the mitochondrial genome for several reasons. The animal mitochondrial genome has a higher rate of molecular evolution than the nuclear genome (Rand 1994; Ballard and Whitlock 2004), so is more likely to reflect any recent changes in rates and patterns of molecular evolution than the nuclear genome. The mitochondrial genome also has a smaller effective population size than the nuclear genome because it is haploid, rarely if ever recombines, and is maternally inherited (Harrison 1989; Moore 1995; Rokas et al. 2003; Ballard and Whitlock 2004), so it is expected to have a higher rate of fixation of nearly neutral substitutions (Ohta 1992), which are thought to dominate mitochondrial genome evolution (Rand and Kann 1996; Bazin et al. 2006). As our aim is to include as many independent domestic lineages as possible, there is a much wider availability of mitochondrial genomes than whole nuclear genome sequences.

To test whether domesticated animals have significantly different patterns of molecular evolution in mitochondrial genomes, we compared complete or nearly complete mitochondrial genome sequences between 16 phylogenetically independent comparisons of domesticated mammals and birds and their close wild relatives. We took two complementary approaches to analyze the data. We used a sister pairs approach to compare branch length, synonymous and nonsynonymous differences, and their ratios in wild and domesticated lineages. We also analyzed all taxa together in a single phylogenetic (“whole tree”) analysis. We found no evidence of a consistent difference between rates and patterns of molecular evolution in the mitochondrial genomes of domesticated mammals and birds and their wild relatives.

Materials and Methods

Selection of Comparisons

We defined domesticated lineages as genetically distinct populations of organisms that have been purposely bred to suit the needs of the domesticator (Blumler et al. 1991; Diamond 2002). We identified the wild relatives of each domesticate from the literature and collected information on the age and history of each domestication event (see supplementary material, Supplementary Material online). We verified using published sources that the chosen wild relative and domestic populations could be identified as well-supported, independent lineages from genetic data and that the domesticated and wild taxa were considered distinct based on morphology, behavior, or geography.

To maintain phylogenetic independence among comparisons of domesticates and their wild relatives, we did not include multiple domesticated lineages that share the same wild relatives. For example, the llama and alpaca are suspected of sharing a wild relative (Kadwell et al. 2001; Cui et al. 2007), so we could only use one of these domesticates in our study. However, we were able to obtain whole mitochondrial genomes associated with two independently domesticated lineages for the dog (Björnerfeldt et al. 2006) and the pig (Wu et al. 2007). For both the dog and pig, the two domesticate–wild comparisons were analyzed as quartets, where one domesticate–wild relative pair acted as the outgroup for the other comparison.

DNA Sequences

We found 16 comparisons of domesticates and their wild relatives with complete or nearly complete mitochondrial genome sequences available on GenBank (www.ncbi.nlm.nih.gov/genbank/, last accessed May 2013). For each comparison, we collected a complete or nearly complete mitochondrial genome sequence for the domesticate, its wild relative, and a closely related outgroup (see supplementary table S3, Supplementary Material online, for accession numbers and alignment lengths). We preferentially collected sequences for the most closely related wild relative and outgroup for each domesticated lineage for which we could obtain a complete or nearly complete mitochondrial genome sequence. We preferentially selected sequences from published articles that explicitly stated whether the sequences came from wild or domesticated individuals. Sequences were not always available for the closest known wild relative, so in some cases we had to choose a more distant wild taxon. We conducted analyses with and without these more distant comparisons (for details see supplementary material, Supplementary Material online). Similarly, in some cases, there is evidence in the literature for population bottlenecks in the wild relatives, and this parallel change may make it harder to detect any effect of reduction in effective population size in the domesticated lineages. We repeated the sister pair and whole tree analyses excluding these comparisons to account for these potentially problematic comparisons.

We used a single mitochondrial genome to represent each taxon. This is because we wished to maximize the number of independent comparisons included in order to gauge general patterns of mitochondrial evolution in domesticated lineages. Multiple sequences are available for relatively few appropriate comparisons, and in many cases the lineages are not clearly monophyletic, which complicates the comparison of rates of substitution or levels of polymorphism (Hughes 2013). Use of a single sequence also avoids the problem of node density effect (Hugall and Lee 2007), especially because the level of polymorphism or number of substitutions may be overestimated in domesticated lineages if a greater number of sequences from domesticated lineages are included than sequences from the wild relatives. By using only a single sequence per lineage, we are unable to distinguish between substitutions (present in all members of a population) and polymorphisms (present in some but not all members of a population).

Sister Pairs Analysis

We aligned the mitochondrial sequences (including protein-coding genes, rRNA, tRNA, and control region sequences) for each domesticate–wild relative comparison and outgroup. We also constructed alignments of only protein-coding genes for estimating nonsynonymous (dN) and synonymous (dS) substitution rates. All alignments were performed by eye in Geneious (Drummond et al. 2011). We deleted any sites or codons that contained gaps in either the domesticate or wild relative sequence so that each base was comparable between sister species and thus informative for a sister pairs analysis.

For the whole genome alignments for each comparison, we estimated branch lengths in BASEML (Yang 2007) using the TN93 substitution model and unconstrained rates (clock = 0 in PAML). We estimated dS, dN, and dN/dS for the protein-coding sequences in CODEML in PAML (version 4.4b, Yang 2007), using the F3x4 codon frequency model (clock = 0). We tested for significant differences in branch length for each comparison using a likelihood ratio test (LRT).

We combined all 16 independent comparisons into a single analysis in order to ask whether the domesticated lineages have consistently different patterns of molecular evolution than their wild relatives. Each independent comparison contributed one data point to a nonparametric analysis of the differences in branch length, dN, dS, and dN/dS between domesticates and their wild relatives. We used both a sign test and the Wilcoxon signed-ranks test (Wilcoxon 1946).

As older divergences have had more time to accumulate substitutions, it may be that the power to detect a significant difference increases over time. If this were the case, we expect that if we compare age of domestication (years before present) or divergence of each sister pair (sum of domesticate and wild relative branch lengths) with the difference between domesticate and wild relative in dN/dS, dN, dS, and total substitution rate, we would find that the older or more divergent comparisons are more likely to show a positive association between domestication and molecular evolution. To test this prediction, we used Spearman’s rank correlation to test for an association between mean age of domestication (measured in years before present, table 1) and differences in branch lengths, dS, dN, and dN/dS between domesticates and wild relatives. We also used Spearman’s rank correlation to test for an association between the genetic distance between domesticated and wild lineages (measured as the sum of both the domestic and wild branches in each comparison) and differences in branch length, dS, dN, and dN/dS.

Table 1

Comparison of Mitochondrial Genomes from 16 Domesticated Mammals and Birds and Their Wild Relatives

Domesticate (Common Name)	Domesticate (Scientific Name)	Wild Relative (Scientific Name)	Mean Domestication Age	dN/dS			dS			dN			Branch Length
				D	W	Sign	D	W	Sign	D	W	Sign	D	W	Sign
Dog	Canis lupus familiarisa	Canis lupusa	13,500	0.1673	0.1006	+	0.0083	0.0082	+	0.0014	0.0008	+	0.0045	0.0047	−
Dog	Canis lupus familiarisa	Canis lupusa	13,500	0	0	−	0.0038	0.0019	+	0	0	−	0.0016	0.0010	+
Cat	Felis catus	Lynx rufus	9,750	0.0330	0.0278	+	0.2811	0.3106	−	0.0093	0.0086	+	0.0814	0.0905	−
Horse	Equus caballus	Equus przewalskii	6,000	0.0441	0.3065	−	0.0057	0.0030	+	0.0002	0.0009	−	0.0021	0.0013	+
Donkey	Equus asinus	Equus asinus somalicus	5,000	0.0465	0.0648	−	0.0325	0.0234	+	0.0015	0.0015	−	0.0098	0.0065	+
Pig	Sus scrofa domesticusa	Sus scrofaa	9,000	0.1728	0.1932	−	0.0136	0.0096	+	0.0024	0.0019	+	0.0056	0.0037	+
Pig	Sus scrofa domesticusa	Sus scrofaa	9,000	0.0575	0.0497	+	0.0129	0.0074	+	0.0007	0.0004	+	0.0033	0.0018	+
Goat	Capra hircus	Capra falconeri	11,000	0.0662	0.0601	+	0.0474	0.0476	−	0.0031	0.0029	+	0.0141	0.0176	−
Sheep	Ovis aries	Ovis ammon	8,500	0.0562	0.0391	+	0.0369	0.0596	−	0.0021	0.0023	−	0.0135	0.0155	−
Cow	Bos taurus	Bos gaurus	9,000	0.0360	0.0400	−	0.1566	0.1486	+	0.0056	0.0059	−	0.0405	0.0426	−
Yak	Bos grunniens	Bos grunniens	4,750	0.0870	0.0634	+	0.0090	0.0123	−	0.0008	0.0008	+	0.0034	0.0038	−
Water buffalo	Bubalus bubalis	Bubalus depressicornis	5,000	0.0476	0.1009	−	0.0447	0.0429	+	0.0021	0.0043	−	0.0099	0.0162	−
Llama	Lama glama	Lama guanicoe	6,000	0.0431	0.0171	+	0.0378	0.0313	+	0.0016	0.0005	+	0.0113	0.0080	+
Bactrian camel	Camelus bactrianus	Camelus ferus	5,000	0.0591	0.0730	−	0.0305	0.0329	−	0.0018	0.0024	−	0.0089	0.0102	−
Chicken	Gallus gallus	Gallus gallus gallus	8,250	0.1083	0.1098	−	0.0033b	0.0033b	+	0.0004b	0.0004b	−	0.0011	0.0010	+
Goose	Anser anser	Anser albifrons	3,500	0.0511	0.0302	+	0.0310	0.0443	−	0.0016	0.0013	+	0.0071	0.0122	−
		No. positive comparisons				8			10			8			7
		Total no. comparisons				16			16			16			16
		Sign test P				1			0.45			1			0.80
		Wilcoxon signed-rank test P				0.75			0.78			1			0.32

Note.—The age of domestication was calculated from the mean of published estimated ranges of timing of domestication events in years before present (see supplemental material, Supplemental Material online, for sources and details on the domesticate–wild comparisons chosen). Estimates of synonymous (dS) and nonsynonymous (dN) substitutions rates, dN/dS, and total substitution rate (substitutions per site) were estimated in PAML v 4.4b (Yang 2007). Comparisons with significantly different branch lengths between domesticated and wild lineages (see Materials and Methods) are presented in bold. Columns marked with D represent estimates for domesticates, and columns labeled W represent estimates for wild relatives. The sign columns represent the sign of the difference between domesticate and wild relative values. Positive symbols represent values that are larger in domesticates compared with their wild relatives, and negative symbols represent smaller values in domesticates.

aDomesticate–wild relative comparisons represented by two independent lineages (the dog and the pig).

bWhile these values are nearly equal, there is a small difference reflected in the direction of the sign of the dS (chicken 0.003297, red junglefowl 0.003293) and dN (chicken 0.000357, red junglefowl 0.000362). When the chicken comparison is removed from the data set, the sign and Wilcoxon signed-ranks tests are still not significant.

Whole Tree Analysis

In addition to the sister pairs approach, we performed a whole tree analysis where we combined the domesticated and wild taxa together into a single phylogeny. Because not all sequences could be confidently aligned between birds and mammals, we created three different alignments: 1) all sequences for all bird taxa; 2) all sequences for all mammal taxa; and 3) protein-coding sequences for all birds and mammals. The D-loop region was excluded from the whole tree analysis because it could not be confidently aligned across all taxa and was not available for several of the domesticate–wild relative comparisons.

For each of these three alignments, we estimated a phylogeny using the following procedure. First, we established data partitions for each alignments using a greedy search in PartitionFinder v1.0.1 (Lanfear et al. 2012), with linked branch lengths, constraining the models of evolution to those available in RAxML, and using AICc for model selection (a measure of AIC corrected for small sample sizes, Hurvich and Tsai 1989). In PartitionFinder, we defined initial data blocks that separated protein-coding genes by gene and codon position. For alignments 1 (all bird genes) and 2 (all mammal genes), we treated the 12S and 16S rRNA genes as separate data blocks and combined all tRNA sequences into one data block. Then, using the best partitions identified with PartionFinder, we analyzed the three alignments in RAxML version 7.0.4 (Stamatakis et al. 2008) to estimate a maximum likelihood phylogeny for each alignment, with 1,000 bootstrap replicates generated using the rapid bootstrapping algorithm. For the phylogenies based on alignments 1 and 2, we estimated branch lengths in BASEML (Yang 2007) using the REV model, unpartitioned data, and no molecular clock (clock = 0). For the phylogeny based on the protein-coding genes for birds and mammals, we used CODEML (Yang 2007) to estimate dN/dS in domesticated and wild lineages using the F3x4 codon frequency model, unpartitioned data, and no molecular clock (clock = 0).

For all phylogenies, we then tested for a significant difference in branch length between domesticated lineages and nondomesticated lineages using a LRT, comparing a one-rate model, where all taxa have the same rate, and a two-rate model, where one rate was estimated for all domesticates and a second rate for all wild relatives. A significant result from the LRT would allow us to reject the hypothesis of uniform rates over the phylogeny.

All alignment and data files used in this analysis are available on the Dryad Digital Repository http://doi.org/10.5061/dryad.d85ng) and can also be obtained from the corresponding author.

Results

We analyzed differences in branch length, synonymous (dS) and nonsynonymous (dN) differences, and dN/dS for 16 sister pairs between domesticated birds and mammals and their wild relatives using a sign test and Wilcoxon signed-rank test (table 1). We found no evidence for a consistent difference between domesticated and wild lineages in branch length (sign test P = 0.80, Wilcoxon signed-ranks P = 0.32), synonymous rates (dS: sign test P = 0.45, Wilcoxon signed-ranks P = 0.78), nonsynonymous rates (dN: sign test P = 1.00, Wilcoxon signed-ranks P = 1.00), nor dN/dS (sign test P = 1.00, Wilcoxon signed-ranks P = 0.75).

Six out of 16 comparisons showed a significant difference in branch length between the domesticated and wild lineages (presented in bold in table 1). In three of these comparisons (llama and both pig lineages), the domesticated lineages had a significantly longer branch length. In the remaining three comparisons (sheep, cow, and goose), the wild relative had a significantly longer branch length.

A Spearman’s rank correlation test revealed no evidence of a correlation between the age of the domestication event and direction of the difference between domesticated and wild lineages in branch length (ρ = 0.01, P = 0.97), dS (ρ = 0.08, P = 0.76), dN (ρ = 0.40, P = 0.13), nor in dN/dS (ρ = 0.14, P = 0.62). We also found no evidence of a correlation between domestication age and genetic distance between sister pairs (ρ = 0.01, P = 0.96), suggesting that, in the mitochondrial genome, the older comparisons included in this study do not always have the greatest genetic distance.

We found no significant relationship between genetic distance (sum of wild and domesticate branch lengths) and difference in dS (ρ = −0.26, P = 0.34), dN (ρ = 0.07, P = 0.80), or dN/dS (ρ = 0.07, P = 0.79), but we did find a significant negative relationship between genetic distance and difference in branch length (ρ = −0.65, P = 0.01). This suggests that in the most divergent comparisons, the wild relative is more likely to have the longer branch length. The relationship is robust to the removal of either the cat or the cow comparisons, which are the most divergent comparisons (supplementary fig. S1, Supplementary Material online); however, removing both of these comparisons makes the relationship nonsignificant. This result suggests that the net amount of molecular change between the sequences could influence the chance of detecting a difference in rate between the domesticated and wild relatives, but that this effect is unlikely to be responsible for our failure to detect more genetic change in domesticated lineages, as the relationship is in the opposite direction (greater genetic distance is associated with longer branches in the wild relative).

For the whole tree analysis, we found no significant difference between the one- and two-rate models for any of the three alignments we tested: 1) no significant difference in dN/dS for the alignment of protein-coding genes for all birds and mammals (P = 0.42); 2) no significant difference in branch length for whole genome alignment for all birds (P = 0.98); 3) no significant difference in branch length for whole genome alignment for all mammals (P = 0.95).

We repeated the sister pair and whole tree analyses, removing comparisons for which we were unable to use the closest wild relatives (cat, goat, cow, water buffalo, and goose; either because of sequence availability or because the closest relative is extinct) or where we found evidence in the literature that the wild relatives have experienced genetic bottlenecks (camel, pigs, horse, donkey, and water buffalo, see supplementary material and supplementary tables S1 and S2, Supplementary Material online). Only a small part of the divergence in the distant comparisons (cat, goat, cow, water buffalo, and goose) may actually correspond to molecular changes influenced by domestication, which could make these comparisons less informative. Furthermore, if reduced effective population size influences molecular rates in domesticates, we may have had difficulty detecting that signal when comparing a domesticate with a wild relative that has also experienced reduced effective population size. We repeated the analyses removing comparisons with suspected bottlenecks in the wild relatives: camel, pigs, horse, donkey, and the water buffalo. We repeated this analysis with and without the water buffalo as the wild relative, the lowland anoa, has only recently experienced a genetic bottleneck (see supplementary material, Supplementary Material online). In addition to experiencing a recent genetic bottleneck, the lowland anoa is an island endemic, which could be associated with a reduced effective population size and, thus, increased molecular rates (Woolfit and Bromham 2005).

When repeating the sister pairs analysis without comparisons where the wild relative has experienced bottlenecks (camel, pigs, horse, donkey, and the water buffalo), we found that domesticates have a significantly higher dN/dS than their wild relatives (Wilcoxon signed-ranks P = 0.02). Therefore, it is possible that in the pairs with bottlenecks in the wild relatives, reduced effective population size has had parallel effects in both domesticated lineages and their wild relatives, reducing the chance of detecting differences between them. All other alternative sister pair and whole tree analyses were not significant (supplementary tables S1 and S2, Supplementary Material online).

Discussion

We find no evidence for a general and consistent difference in the tempo and mode of mitochondrial molecular evolution of domesticated birds and mammals when compared with their wild relatives. Given that higher dN/dS has been reported for a number of domestic lineages, why do we fail to find evidence for a general increase in dN/dS across all the domestic lineages included in this study?

It is possible that lack of statistical power has prevented us from identifying significant differences in some comparisons. Our power is unavoidably limited by the nature of the question. We are unable to include more comparisons because there are relatively few fully domesticated animal lineages, and we had to leave some lineages out of this study due to lack of sequence data from appropriate wild relatives (e.g., turkey: see supplementary material, Supplementary Material online). It may be informative to apply this comparative approach to domesticated plants, which are more diverse. Furthermore, all domestication events are young on an evolutionary scale, so there has been only a short period of time for differences in tempo and mode of molecular evolution to make a detectable impression on patterns of sequence differences.

If the relatively small number of sequence differences between recently diverged genomes was obscuring a result, then we would expect the six sister pair comparisons with a significant difference in branch length to be more likely to show longer branch lengths, or higher dN/dS, in the domesticated lineage. But only half of the comparisons with a significant difference in branch length show more genetic change in the domesticated lineage, a pattern indistinguishable from chance. We also find that in the more divergent comparisons (those with a greater net genetic distance between the domestic and wild lineages), it is the wild relative that is more likely to have a longer branch length. Older domesticated lineages, which have had more time to accumulate evidence of distinct patterns of molecular evolution, do not show a greater tendency to have higher rates of change than their wild relatives. So we do not think that lack of power to detect differences in rate of change explains the lack of a consistent pattern in our comparisons. However, it may be possible that bottlenecks in wild relative populations may impact our power to detect a difference in dN/dS between domesticated lineages and their wild relatives.

One way to increase power to detect changes in the tempo and mode of molecular evolution in domesticated lineages is to take a population-level approach, with multiple individual samples for each domesticate and wild lineage. Recent population-level studies have found increased dN/dS or ratio of nonsynonymous to synonymous diversity (πN/πS) in a number of domesticated lineages compared with their wild relatives (Wang et al. 2011; Hughes 2013). However, these studies have included an uneven number of domesticated and wild samples (254 from the dog vs. 19 from the wolf; 59 from the domestic pig vs. 27 from wild boar; 41 from the domestic chicken vs. 17 from the red junglefowl in Hughes 2013, and 51 domestic yaks vs. 21 wild yaks in Wang et al. 2011). Many short, recently diverged branches can increase estimates of dN/dS (Rocha et al. 2006), so higher dN/dS is more likely to be reported if an analysis includes more branches in a domesticate population than a wild one.

To avoid the measurement bias due to the node density effect, we only sampled one individual per domesticated and wild lineage. Choosing only one sequence per lineage also helps us to avoid the problem of lack of monophyly in analyses of population-level data. Backcrossing and interbreeding with wild relatives can shape the molecular evolution of domesticated and wild lineages (Vilà et al. 2005), and these processes may have varied substantially between lineages. For example, Hughes (2013) reported that the phylogenies of domesticated and wild lineages of chickens, dogs, and pigs are not monophyletic but intermixed, which could be a signature of ancestral polymorphisms or interbreeding in these populations. We have attempted to minimize this effect on our results by choosing wild lineages that may not be the closest relative but have less chance of being influenced by recent introgression (see Materials and Methods and supplementary material, Supplementary Material online). However, by choosing only one sequence per lineage, we are unable to distinguish substitutions from polymorphisms. Our approach could mask higher rates of change in the domesticated lineage if wild lineages consistently retained comparatively more ancestral polymorphisms.

If the majority of substitutions in the mitochondrial genome are neutral or slightly deleterious, rather than under positive selection, then we would expect dN/dS estimates in the mitochondrial genome to be higher within species than between species (Hasegawa et al. 1998; Rand and Kann 1998; Weinreich and Rand 2000; Ho et al. 2005). Therefore, population-level estimates of mitochondrial dN/dS that do not account for the effect of ancestral polymorphism are expected to be higher than those estimated at the lineage level. As such, we would expect our dN/dS estimates to be lower than those from population-level studies. Concordant with these population-level studies, we found a higher dN/dS in one dog, one pig, and the yak comparison. Although we cannot compare our dN/dS estimates with the πN/πS reported in Hughes (2013), our dN/dS estimate for the domesticated and wild yaks are, as expected, lower than those reported by Wang et al. (2011) (our dN/dS for wild yaks: 0.06, their dN/dS for wild yaks: 0.07, our dN/dS for domesticated yaks: 0.09, their dN/dS for domesticated yaks 0.23).

It could be argued that the housekeeping genes of the mitochondria are unlikely to experience a dramatic change in selective regime, which could explain why we found no consistent pattern associated with domestication in the mitochondrial genome. Actually, many studies of domestication report changes in traits associated with metabolism (Xia et al. 2009; Gibbons et al. 2012). For example, selective sweeps in chickens raised for meat production are connected to genes associated with growth, appetite, and metabolic regulation (Rubin et al. 2010). It is therefore possible that artificially selected traits could be associated with growth and metabolism, which could potentially increase dN in mitochondrial loci (MacEachern et al. 2009; Rubin et al. 2010; Akey et al. 2010; Amaral et al. 2011; Kijas et al. 2012). However, our study is designed to detect changes in genome-wide rates of change, rather than focusing on the effect of selection on particular genes.

Our results do not preclude an impact of domestication on patterns of mitochondrial evolution, but they do suggest that there is no consistent, detectable difference between all domesticated lineages and their wild relatives. It may be that domestication can influence mitochondrial molecular evolution, but that it does not do so consistently and uniformly across all domesticated lineages in comparison to their wild relatives. Each domestication history has involved different levels of human intervention, and the observed genetic and morphological changes in domesticated lineages are variable (Zeder 2006). For example, it has been suggested that domestic sheep and cats may have undergone less severe genetic bottlenecks than other domesticated animals (Driscoll et al. 2007; Kijas et al. 2009), but as both of these lineages have higher dN/dS estimates (table 1), this does not seem to provide an explanation for the lack of a general pattern of higher dN/dS across domesticated lineages. Similarly, some domesticated lineages, like the horse, cat, and camel, may have experienced less artificial selection than others (Clutton-Brock 1999; Driscoll et al. 2009), yet the horse and cat have higher dN/dS than their wild relatives, and the camel has lower dN/dS.

In addition to considering the heterogeneity of processes affecting the domesticated lineages, population processes in the wild relatives may also impact on our ability to detect changes in the tempo and mode of molecular evolution in domesticated lineages. If similar changes have occurred in both the domesticated lineages and their wild relatives, then we may be unable to detect a significant difference between them. In particular, some wild relatives may have experienced significant genetic bottlenecks. For example, the wild relative of the water buffalo, the lowland anoa, is an island endemic, which could be associated with a reduced effective population size and, thus, increased dN/dS (Woolfit and Bromham 2005). Other examples of wild relatives that may have undergone population size reduction are the wild Bactrian camels (Hare 1997; Silbermayr et al. 2010), wild boar (Scandura et al. 2008), Przewalski’s horses (Clutton-Brock 1999; Vilà 2001), and the Somali wild ass (Moehlman 2002). When we analyzed a reduced set of comparisons, removing comparisons where we found evidence that the wild relative had undergone a population bottleneck, we found that domesticated lineages had a higher dN/dS than their wild relatives. Although the sample size for this test is small (N = 10), this result is consistent with the hypothesis that domestication reduces a lineage’s effective population size and thus may increase the accumulation of slightly deleterious, nonsynonymous changes in the mitochondrial genome.

In this analysis of 16 domesticated mammals and birds, we find no evidence of a general, consistent pattern in the rates or patterns of molecular evolution in the mitochondria. However, we do find that in a subset of comparisons, there is evidence of higher dN/dS in domesticated lineages, which may be a signature of changes in effective population size. We conclude that differences in dN/dS between particular domesticated lineages and their wild relatives in the mitochondrial genome (Björnerfeldt et al. 2006; Wang et al. 2011) are best explained by specific factors in the biology or domestication history of particular lineages and not a generally predictable result of domestication.

Supplementary Material

Supplementary material, figure S1, and tables S1 and S2 are available at Genome Biology and Evolution online (http://www.gbe.oxfordjournals.org/).

Acknowledgment

This work was funded by the Australian Research Council.