Differences between genes and species in the rate of evolution from ancestral to extant species
We sequenced 84 amplicons from 15 autosomal genes in 14 representatives of the Bovinae subfamily resulting in ~52,800 base pairs of sequence in each species, which we have submitted to Genbank [3
]. The sequence of the common ancestor was inferred and used to count the synonymous and nonsynonymous substitutions between the ancestral sequence and each extant species. An analysis of variance tested the significance of differences in the rate of evolution on the 14 branches of the evolutionary tree (analysis GLM1 in Methods). Table shows a summary of the p-values from the analysis of variance. Pairwise comparisons between the ancient and contemporary samples in GLM 1 yielded significant (p < 0.05) effects for amplicon (a) for all of the variables tested. However, difficulties arise with this analysis due to the lack of independence for substitutions between lineages. Because some mutations have occurred in a common co-ancestor, these mutations are being considered more than once causing the F-test to be anti-conservative. This problem does not affect the F-test for gene because it compares gene and amplicon within gene mean squares (see Methods). That is, we test if the difference between genes is greater than the variation between amplicons within genes. Significant gene effects were detected for dN and dN/(dN+dS). Thus, different genes have accumulated different numbers of amino acid changing mutations, which may be due to different levels of selective constraint. Table presents the estimated effects for GLM 1 examining the variation in genes for dN and dN/(dN+dS). Amongst the most rapidly evolving genes are 5HT1F and EGF, while the slowest evolving genes are IGFBP2 and IGFBP5.
Significance tests for the effect of gene (g), amplicon within gene (a) and species (s).
Estimated gene effects in bold (and s.e.) are deviations from mean values from GLM 1 minus amplicon y = u + g + s + e effect for dN and dN/(dN+dS), where g and s = gene and species effects, respectively and e = residual error.
Surprisingly, a significant difference between species for the assumed neutral mutation rate in dS and dI was detected. The differences between species for dS were found to be restricted to the Water buffalo, which show a significantly higher rate of silent site evolution, while the breeds of B. taurus were amongst the lowest when compared to the ancestral sequence. The significant effect detected in species for the number of intronic substitutions per site (dI) was the result of a higher rate of substitution in the Eland lineage (results not shown). This is likely an anomaly in the ancestral sequence, as the Eland outgroup's divergence time from the ancestral sequence is not similar to other members of the Bovini tribe.
The rate of radical (Dr) to conservative mutations (Dc) was not significantly different between genes or species, because most amino acid changing mutations were conservative in nature. Therefore, Dr/Dc is not a useful measure of molecular evolution for this study.
In figure we present a partial phylogenetic reconstruction from over 21,000 noncoding sites and 1,500 substitutions at these sites. As can be seen in the phylogenetic tree one river buffalo (BubB) appears to be more closely related to a swamp buffalo (BubC) than to other river buffalo, which may be due to recent introgression with a swamp buffalo. Despite this anomaly, the majority of species sampled appear to be related as we expected. More details regarding the relationships between these animals can be found in MacEachern et al [3
]. We also compare the neutral phylogenetic tree with the variation found at amino acids with the number of amino acid changing mutations per site (dN). The highest number of amino acid changes per site identified was in Holstein (Figure , 0.0068). Where a direct contrast between domestic and undomesticated animals was possible, like domestic cattle and water buffalo, the breeds under more intense artificial selection (Holstein and Hereford or Murrah, Asian water and swamp buffalo) have accumulated more amino acid changes than less intensely selected breeds (Tuli and the nondescript breed of Indian water buffalo or species like the African buffalo). However, as described in GLM 1 (Table ) these differences between species are not quite significant. Surprisingly, the dN rate in Gaur is higher than Mithan, suggesting a faster mutation rate in this species.
Figure 1 Comparison of a neutral phylogeny and amino acid variation, neighbour joining analysis for members of the Bovini tribe (Anc: Ancient, Ban: Banteng, Bis: Bison, BubB & BubC: Asian buffalo, water and swamp type, Ela: Eland, Gau: Gaur, Her: Hereford, (more ...)
The model GLM 2 (see Methods) tested the interaction between gene and species (g.s) on rates of evolution. No significant interaction was detected between gene and species for any of the dependent variables, indicating that the differences detected for genes in GLM 1 were not particular to any one species. Therefore, high substitution or evolutionary rates for a given gene were common across the entire phylogeny and the differences in evolutionary rate between genes were most likely a result of different levels of selective constraint amongst genes rather than species.
Comparison of substitutions at synonymous and noncoding sites and CpG dinucleotides
Synonymous and noncoding sites were examined for differences in the number of substitutions per site (dS and dI, respectively). We identified a significant difference between the number of substitutions per site for dS and dI, with a significantly higher rate of substitution detected at synonymous sites. Overall a ~30% higher rate of substitution at synonymous sites was detected, which was determined to be significant (p < 0.001). An examination of whether CpG dinucleotides were over represented in substitutions at synonymous or noncoding sites uncovered a significantly higher proportion of silent substitutions at synonymous sites that involve CpG dinucleotides than at noncoding sites (Table , p < 0.001). The difference between synonymous and noncoding substitutions per site decreases when substitutions involving CpG dinucleotides are removed from the analysis with only a 10% higher rate of substitution at synonymous sites, however, this difference is still significant (Table , p < 0.001).
Table 4 Summary of the total number of sites, substitutions and substitutions per site at synonymous and noncoding regions, substitutions involving CpGs, relative proportion of substitutions involving CpG dinucleotides and the number of substitutions per site (more ...)
Genetic distance between species and breeds
Pairwise comparisons for dI and dS between all 14 Bovini representatives are given in table . The lowest values are for comparisons within a species (i.e. within Indian water buffaloes or within Bos taurus) and the highest are between Eland and other species. This finding is also represented graphically in figure , with a significant linear relationship identified between dS and dI for pairwise comparisons summed across all genes. The simplest explanation is that both dI and dS are largely neutral substitutions that accumulate over greater evolutionary time.
Pairwise comparisons between Bovinae representatives for the number of silent substitutions per site summarised for all genes, with intronic substitutions (dI) below the diagonal and synonymous substitutions (dS) above the diagonal in bold
the rate of substitution at synonymous sites (dS) plotted against the rate of substitution at noncoding sites (dI) for all paiwise comparisons within the Bovini tribe.
In table the pairwise comparisons for dN/dS and dN are presented, lower and upper diagonal, respectively. The patterns of divergence for dN are similar to those found for dI and dS with the smallest values found between breeds of Domestic cattle and buffalo, while the largest values were consistently found between species. Pairwise comparisons for dN/dS show a different pattern, with some of the largest values detected between breeds of B. taurus and other closely related species from the Bovina subtribe. Interestingly, pairwise comparisons between buffalo breeds or species that have been separated over similar time frames often show much smaller evolutionary ratios (BubB vs BubC and Ind vs Mur)
Pairwise comparisons between all Bovinae representatives for the ratio of nonsynonymous to synonymous substitutions per site (dN/dS) summarised for all genes below the diagonal and dN above the diagonal in bold
A significantly negative relationship was found between the evolutionary ratio (dN/dS) and the number of substitutions in noncoding DNA (dI) (Figure ). The relationship appears to be quadratic, which suggests that there is a relative abundance of dN:dS over short time scales and that negative selection is more effective at removing deleterious mutations over increasing evolutionary time. Thus, dN/dS is variable, but includes some high values when dI is small. Conversely, at high dI, dN/dS is approximately 0.1.
Relationship between the evolutionary ratio (dN/dS) plotted against the number of substitutions for noncoding DNA per site (dI) for all pairwise comparisons within Bovini.
Divergence, polymorphism and domestication
If selective constraint is reduced over small time scales consistently higher rates of molecular evolution should be seen between very closely related species. Likewise, if this phenomenon is common for polymorphisms the highest rates of evolution should be seen within species. Alternatively, if domestication has increased the evolutionary rate in B. taurus by relaxing selective constraint, higher rates of divergence between species should be seen in domesticated species when compared to their undomesticated counterparts and a higher rate of polymorphism within B. taurus should be found when compared to less intensely domesticated species such as the Water buffalo.
Pairwise comparisons presented in table between closely related species from the Bovina subtribe have identified no large evolutionary ratios (dN/dS > 0.1) for two undomesticated animals that recently shared a common ancestor (Gaur vs Bison, Gaur vs Banteng, Bison vs Banteng). Therefore, the negative relationship detected in figure may be the result of comparisons animals that have undergone a relatively intense domestication event. Pairwise comparisons, between closely related species where at least one species has a history of domestication (Table , bottom left) (Mithan, Yak and Domestic cattle) were generally higher than pairwise comparisons between undomesticated species. The values for dN/dS appear to increase again when two domestic animals are compared (Table , bottom right). In fact all dN/dS comparisons with values > 0.15 involve one or two domestic species (Table , bottom left and right).
Pairwise comparisons between domestic and nondomestic members of the Bovina subtribe for dN/dS calculated for all mutations and sites sampled.
Because the Bovini tribe diverged in a star like phylogeny ~2MYA [3
], comparisons between undomesticated animals (Gaur vs Bison) should be comparable to domesticated species (Yak vs Mithan). However the high rates detected between breeds of B. taurus
could be the result of changes in the time of divergence.
To examine if selective constraint in B. taurus had recently been relaxed, we compared within species variation for dN/dS between Domestic cattle and buffalo. High evolutionary rates were detected (Table , lower diagonal) for Hereford vs Holstein and Hereford vs Tuli, with the highest ratio detected between Hereford and Tuli. In comparison, the evolutionary ratios for buffalo were generally much smaller, with Ind vs Mur showing low evolutionary ratios and BubB vs BubC showing the smallest evolutionary ratio. However, some large ratios were found in buffalo, highlighting the low accuracy of dN/dS ratios between close relatives. If intense domestication and artificial selection in B. taurus is affecting dN/dS rates, then dN/dS between breeds of Domestic cattle should be significantly larger than those detected between breeds of buffalo. Therefore, we have tested for significance between the ratio of nonsynonymous and synonymous substitutions within both species. None of the comparisons were significant, however, a significant difference was nearly detected for dN/dS between Hereford vs Tuli and BubB vs BubC (Table , lower diagonal, χ2 test, p = 0.068).
Holstein polymorphism data was collected from 8 animals in an effort to detect if differences existed between breeds of B. taurus that may be the result of recent artificial selection. Table summarises the number of polymorphisms and the evolutionary rate detected in Holsteins. Only 3 nonsynonymous and 5 synonymous mutations were found in the genes 5HT1F, 5HT6 and EGF. The dN/dS ratio within Holstein is similar to that between breeds of B. taurus. Both are higher than dN/dS in species divergence, so a McDonald-Kreitman test gives no evidence of positive selection driving species divergence. This might be because dN/dS within B. taurus is higher as a result of ineffective negative or purifying selection.
Summary of total Holstein polymorphisms for all genes
A number of neutral or nearly neutral substitutions were identified from the dataset that appear to be the result of homologous mutations in a common ancestor that have subsequently been randomly assorted in modern species [3
]. In total these ancestral polymorphisms corresponded to 7 nonsynonymous and 26 synonymous mutations and a dN/dS rate of 0.065.
The McDonald-Kreitman test was modified to compare dN/dS between species to dN/dS in ancient polymorphisms on the assumption that the latter are neutral. Table shows the results of chi-squared tests in which Ka:Ks in a series of species differences and species polymorphisms is compared to Ka:Ks in ancient substitutions. In general, pairwise comparisons between Mithan and other members of the Bovidae have higher dN/dS than the rate in aberrant polymorphism, with significant differences detected between Mithan and Holstein cattle. This finding could be mistaken as a sign of rapid evolution in domestic cattle. However, examining the other pairwise comparisons it appears that Mithan is generating the difference with near significant results for the majority of members in the Bovidae (Table ) and consistently large evolutionary ratios (dN/dS) in table (lower diagonal).
Modified McDondald-Kreitman test examining variation between ancient polymorphism and pairwise comparisons between species and breeds for the ratio of nonsynonymous (Ka) and synonymous (Ks) polymorphisms
We also applied the test to identify differences in the rate of polymorphism for B. taurus and B. bubalis. Comparisons of the polymorphisms within Domestic cattle and Water buffalo with the rate of ancient polymorphisms should help identify whether any differences in evolutionary rate can be attributed to a relaxation of selective constraint. A significant difference was detected for the number of polymorphisms between Hereford and Tuli when compared with the number of ancient polymorphisms (Table ). The same comparison was made for polymorphisms found between both types of Water buffalo (Ind vs Mur and BubB vs BubC); however, no significant differences were detected (results not shown). Therefore, it appears that during the domestication of B. taurus there has been an accumulation of nonsynonymous mutations above those found at aberrant sites, which may result from an accumulation of deleterious alleles associated with domestication and a relaxation of selective constraint.