High variability and non-neutral evolution of the mammalian avpr1a gene
© Fink et al. 2007
Received: 16 April 2007
Accepted: 27 September 2007
Published: 27 September 2007
Skip to main content
© Fink et al. 2007
Received: 16 April 2007
Accepted: 27 September 2007
Published: 27 September 2007
The arginine-vasopressin 1a receptor has been identified as a key determinant for social behaviour in Microtus voles, humans and other mammals. Nevertheless, the genetic bases of complex phenotypic traits like differences in social and mating behaviour among species and individuals remain largely unknown. Contrary to previous studies focusing on differences in the promotor region of the gene, we investigate here the level of functional variation in the coding region (exon 1) of this locus.
We detected high sequence diversity between higher mammalian taxa as well as between species of the genus Microtus. This includes length variation and radical amino acid changes, as well as the presence of distinct protein variants within individuals. Additionally, negative selection prevails on most parts of the first exon of the arginine-vasopressin receptor 1a (avpr1a) gene but it contains regions with higher rates of change that harbour positively selected sites. Synonymous and non-synonymous substitution rates in the avpr1a gene are not exceptional compared to other genes, but they exceed those found in related hormone receptors with similar functions.
These results stress the importance of considering variation in the coding sequence of avpr1a in regards to associations with life history traits (e.g. social behaviour, mating system, habitat requirements) of voles, other mammals and humans in particular.
The genetic bases of complex phenotypic traits like differences in social and mating behaviour among species and individuals remain largely unknown . Most such traits are probably under polygenic control and the contribution of each gene to the phenotype is often very difficult to assess . Even for genes with large effects, it is highly challenging to identify the causes of particular phenotypic differences because genetic variation is rarely restricted to dichotomous polymorphism in a gene [e.g. [3–6]]. Genetic variation at a locus is not only shaped by locus- or site-specific selective processes but also by the evolutionary history of the particular species or population.
One of the best examples of a single gene with large effects involved in very specific phenotypic and behavioural differences is the arginine-vasopressin receptor 1a (avpr1a). This gene has been proposed to play a key role in controlling variation in mammalian social behaviour [7–11], and it has been particularly well-studied for its role in the formation of mating systems in rodents from the genus Microtus [12–15]. Phenotypic differences between species in arginine-vasopressin 1a receptor (V1aR) distribution in the brain and contrasting social behaviour were largely attributed to the presence of a repetitive expansion in the regulatory region upstream of the gene [12–15]. The transfer of the entire avpr1a gene region including the repetitive expansion or the coding region from a monogamous vole to non-monogamous voles and other rodents resulted in modified V1aR distributions and changes in social behaviour . Additionally, monogamous voles showed increased affiliative behaviour (measured as time spent in contact with other voles, see in ) after injection of the arginine-vasopressin (AVP) hormone in the brain, while non-monogamous voles displayed unchanged social behaviour . However, AVP has two main roles: it controls higher cognitive functions such as memory and learning in the brain, and it acts peripherally by facilitating water absorption in the kidney and by contracting smooth muscle cells from blood vessels . The impact of hormones on behavioural variation may therefore also depend on environmental conditions [see ]. A recent study showed further that neither social nor genetic monogamy are strictly associated with the presence of the repetitive expansion in the regulatory region of avpr1a in voles and other mammals .
In contrast to polymorphism in the regulatory region of avpr1a, variation in the coding part is assumed to be low and functionally negligible [14, 19, 20]. Rodent avpr1a patterns have been proposed as mammalian model systems for the study of the role of hormone receptors in the formation of complex social interactions, including human social disorders . Studies of human avpr1a have mainly focused on variation in the non-coding upstream region of the gene, and associations have been reported with autism [22–24], eating behaviour , self perception  and even creative dance performance . Single nucleotide polymorphisms in the human avpr1a gene have been detected [22–24, 27], but it is unknown if they affect the encoded protein. Previous studies of the Microtine avpr1a have not explicitly studied levels and patterns of variation in the coding region at the inter- or intra-specific level. Its potential influence on social behaviour and interactions in voles and other mammals remains therefore totally unknown.
We use here an evolutionary approach to investigate variation in parts of the coding region of the mammalian avpr1a gene. We analyse patterns of nucleotide and amino acid (AA) polymorphism in the Microtus genus represented by 24 species from three continents (Europe, North America, Asia), and compare it to the avpr1a diversity found in various higher mammalian taxa. Furthermore, we examine rate variation among the functionally important regions – the ligand binding site or the G-protein binding domain  – and other parts of the V1aR, and we test for the role of selection in shaping variability in the avpr1a gene.
Origin of rodent samples and avpr1a sequences
High Tatra Mountains
Gene bank sequence
Gene bank sequence
23 AA substitutions involved radical (at least one change between physico-chemical classes considering polarity, charge and volume; see in ) and 10 conservative (all three categories reveal the same physico-chemical characteristics for the interchangeable AAs) changes. Ten of the radical changes were found in the ligand binding and the G-protein binding domains (see Figure 2).
Results of avpr1a selection tests performed with the software PAML
positively selected sites
M0, one ratio
p 0 = 0.9600, ω0 = 0.0602
p 1 = 0.0400, ω1 = 1
p 0 = 0.9600, ω0 = 0.0602
p 1 = 0.0199, ω1 = 1
p 2 = 0.0200, ω2 = 1
p 0 = 0.1311, ω0 = 0.0406
p 1 = 0.7474, ω1 = 0.0407
p 2 = 0.1215, ω2 = 0.4609
p = 0.29056, q = 2.75313
M8, beta and ω
p 0 = 1.0000, p = 0.29056, q = 2.75313
p 1 = 0.0000, ω = 1.0000
M0, one ratio
p 0 = 0.8433, ω0 = 0.0452
p 1 = 0.1568, ω1 = 1
p 0 = 0.8433, ω0 = 0.0452
p 1 = 0.1356, ω1 = 1
p 2 = 0.0212, ω2 = 1
p 0 = 0.5084, ω0 = 0.0000
p 1 = 0.3169, ω1 = 0.1063
p 2 = 0.1747, ω2 = 0.4661
p = 0.23059, q = 1.71469
M8, beta and ω
p 0 = 1.0000, p = 0.23059, q = 1.71466
p 1 = 0.0000, ω = 2.02300
Branch specific models:
MA, foreground branch = M. montanus
MA, foreground branch = A. terrestris
MA, foreground branch = C. glareolus
MA, foreground branch = A. sylvaticus
MA, foreground branch = M. musculus
MA, foreground branch = R. norvegicus
MA, foreground branch = O. aries
MA, foreground branch = B. taurus
MA, foreground branch = C. familiaris
MA, foreground branch = M. mulatta
MA, foreground branch = P. troglodytes
MA, foreground branch = H. sapiens
MA, foreground branch = M. domestica
The high diversity found at the nucleotide level resulted in high AA diversity after translation with all species showing unique AA sequence types. Most changes occurred in the two functionally important regions of the V1aR: the ligand binding domain and the G-protein binding domain (Figure 1). The latter region included many AA deletions and insertions, resulting in length variation among mammals. Except for a 3 AA long deletion in several rodents (M. montanus, A. terrestris, C. glareolus), the other insertions and deletions occurred in single species only.
Despite the evidence for positive selection on the ligand binding domain, further tests rather suggested generally negative selection on avpr1a. PAML detected significant rate variation among the lineages (M0 vs M3: 2Δ l = 178.537, df = 4 p < 0.05), where 88% of all sites were under strong purifying selection, while 12% showed relaxed purifying selection acting on these sites (Table 2). PAML revealed no evidence for positive selection overall (M1 vs M2: 2Δ l = 0, df = 2 p > 0.05; M7 vs M8:2Δ l = 0.0004, df = 2 p > 0.05; Table 2). HyPhy detected five negatively selected sites in functionally important regions and 24 in-between (codon positions 33, 41, 47, 48, 52, 69, 71, 77, 87, 107, 119, 120, 125, 136, 138, 146, 152, 159, 178, 184, 198, 216, 223, 227, 230, 250, 251, 254, 279).
Considering the phylogenetic background of the species provided further evidence for non-neutral evolution of the avpr1a gene. For the mammalian branches, evolutionary models allowing for selection (MA) were not significantly better than models not incorporating selection (M1; see likelihood values Table 2). Codons with dN/dS ratios exceeding 1 were detected mainly in the G-protein binding domain (231–274), with only two species showing positively selected sites outside (O. aries, C. familiaris; positions 191, 228; see Figure 4).
Our analyses of the avpr1a gene, shown to have high behavioural impact in the genus Microtus as well as in other mammals [12, 30], revealed high nucleotide and protein diversity. Variation within Microtus involved many radical physico-chemical amino acid substitutions and deletions, which were located at functionally important regions of the V1aR. The pattern indicates positive selection on few codons in the ligand binding domain and possibly in the G-protein binding domain, but purifying selection on the majority of the gene.
Genetic variability in the coding region of the avpr1a gene appears much higher and evolutionarily much more important than previously suggested [30, 31]. DNA sequences of just two M. ochrogaster and M. montanus individuals were taken as evidence that the Microtine avpr1a gene was highly conserved [14, 20, 31]. However, our analyses reveal not only high levels of genetic variation in the coding region between mammalian species, but also within the genus Microtus. We detected up to 23 polymorphic positions in the first exon of the gene within a single Microtus individual compared to other closely related species, whereas studies of human avpr1a revealed a few synonymous and non-synonymous SNP in humans [22, 27]. Population data from at least one Microtus species will be necessary to allow more detailed comparisons with human variation. However, the apparent difference between voles and humans may be explained in part by the longer evolutionary history of Microtus voles of at least several hundred thousand years with about three generations per year  and generally elevated mutation rates in rodents compared to primates [32, 33]
Many AA replacements in the Microtine avpr1a gene involved radical physico-chemical changes, and several vole species showed deletions and insertions of AAs in this hormone receptor gene. Such length-variation in coding DNA within or among closely related species is remarkable, because it is usually restricted to non-coding DNA, where it may influence in particular cases the expression of genes but is generally functionally and selectively neutral [34–37]. Additionally, we detected considerable length variation in the G-protein binding region between different mammalian species. The diversity found among mammals might influence signal transduction, since already single AA changes can lead to differences in receptor activation in this region .
Amino acid positions identified as crucial for either ligand binding or G-protein activation in humans, mouse and rat are mainly conserved among Microtus species as well as among other mammals. A highly conserved triplet (Asp148-Arg149-Tyr150) with a role in signal transduction in many G-protein coupled receptors was conserved across all individuals analysed . Additionally, a glycosylation site (Asn27) with a crucial role in protein folding or stabilization  remained conserved among all mammals and voles. Glu185, supposed to be involved in agonist and peptide as well as non-peptide antagonist binding to the V1aR , was highly conserved among mammals except for one Microtus individual (M. richardsoni), which showed a mutation to His185. This alteration together with an additional mutation at a glycosylation site (Asn198->Thr198) could lead to dysfunctions . It is unclear if this would apply to voles since analyses of the specific roles of these sites in Microtus are lacking.
The observed high level of diversity and the detection of indels are unlikely to be due to gene duplications of avpr1a or the occurrence of pseudo-genes. avpr1a is a single copy gene in humans  and in rat , and a duplication has been found only in M. ochrogaster . We cannot exclude that some of the detected variation stems from the presence of very recently duplicated sequences in some species. However, contrary to the truncated and clearly divergent version of avpr1a in M. ochrogaster, we found no indication for non-functionality in any of the sequences, such as reading frame shifts due to insertions or deletions of single nucleotides or the presence of premature stop codons . This suggests that at least the majority of avpr1a gene variants is functional.
The potential functional relevance of variation in Microtine avpr1a is further emphasized by the detection of alleles coding for different protein types within the same individuals. It is worth noting that this involved again the ligand and G-protein binding domains. However, it is currently unclear if heterozygous individuals express different protein variants in the same tissue or if there is tissue specific expression [see ]. Gene expression studies are needed to investigate this further, since some receptor functions can as well be substituted by other receptors for hormones which are involved in very similar pathways [30, 43]. This distinction would be of pharmaceutical importance for human health [see [44, 45]]. Given the high variability in protein types, Microtus voles could serve as ideal models to study the processing and the expression of avpr1a gene products and their functional consequences in homo- and heterozygous individuals since they can be bred in the laboratory .
We further hypothesize that variation in the coding sequence of avpr1a might be related to life history traits (e.g. mating system, habitat), given the peripheral role of the V1aR in water retention in the kidney , and the social relevance when expressed in the brain [7–10]. Although kidney inefficiency was suggested as a major reason for the restriction of some Microtus species to moist habitat , we could not detect any connection between receptor type and a given habitat. Species occupying dry habitats (e.g. M. multiplex, M. arvalis, M. lusitanicus, M. pinetorum, M. nivalis, [47–50]) or wet habitats (e.g. M. agrestis, M. rossiaemeridionalis, M. tatricus and M. oeconomus [51–53]) showed no specific V1aR types or AA change corresponding to habitat requirements, nor were they phylogenetically closely related (Figure 3). Additionally, there was no general association of AA variation in Microtus avpr1a with social or genetic mating system. The socially monogamous (defined by observational studies, see in ) species M. ochrogaster, M. multiplex and M. pinetorum [51, 52, 55] shared neither a protein type nor a particular AA change. Similarly, neither showed the socially non-monogamous species such as M. montanus, M. californicus and M. richardsoni [51, 53, 56] nor the genetically non- monogamous species (M. arvalis, M. agrestis and M. ochrogaster, [18, 57–59]) identical AA alterations or protein types among each other that could potentially be associated with behavioural patterns. It is obvious that the high level of variation in the genus Microtus makes it very difficult to detect any direct associations of protein types or AA changes with basic life history traits. Since inter- and intra-specific variation in mating behaviour and habitat usage exist (e.g. for M. ochrogaster see ), studies on protein variation within species are needed to investigate locally adapted receptor types and their correlation to life history traits.
Statistical tests for selection indicated mostly purifying selection on the transmembrane partition of V1aR but also a minor role for positive selection in shaping avpr1a diversity in mammals. The sliding window analysis detected positive selection mainly in the ligand binding domain and an increased number of synonymous and non-synonymus changes in the G-protein binding domain. Branch-specific tests across mammals detected positively selected sites mainly in the G-protein binding domain. It is unclear why selection tests failed to detect a deviation from neutrality overall, but the background of purifying selection might be too high in comparison to the positive selection acting on particular domains to allow the signal to be picked up. Additionally, positive selection evidenced by ω > 1 may be difficult to detect as selection could also be acting on synonymous sites [60–62].
The impact of positive selection on avpr1a diversity is less evident within the Microtus genus than in evolutionarily less related mammalian taxa. Overall selection tests remained mostly inconclusive, which may be caused by a lack of power due to the high rate of speciation and the short divergence times between Microtus species . Interestingly, the number of AA variants within Microtus was still significantly higher in the ligand and G-protein binding domains than in the transmembrane regions. This pattern might reflect relaxed selective constraints at the N-terminus and at the G-protein binding domain, and stronger evolutionary constraints on the transmembrane region due to structural limitations because of the embedding in the lipid layer. Alternatively, we suggest that this high diversity in functionally important domains of the gene is compatible with balancing selection maintaining high allele diversity in selected regions [63, 64]. However, we shall need detailed studies at the population level to asses the potential impact of this type of selection on avpr1a diversity in Microtus further .
The speed of evolutionary change in avpr1a is difficult to assess because of the lack of data on nuclear mammalian genes with similar taxonomic and geographic scope. The only directly comparable data set from a nuclear gene covering several Microtus species comes from the p53 tumor-suppressor gene . Variation in this gene is much lower than in avpr1a with only a few silent mutations in coding regions . Additionally, nucleotide diversity in the fast-evolving nuclear genes IRBP and RAG1 within the mouse genus is lower than in Microtine avpr1a if the longer divergence time in Mus is taken into account (5 to 6 mya for Mus, see in  vs. 1.2 to 2 mya for Microtus, see in ). Variation in avpr1a appears here even higher after translation of the nucleotide sequences into AAs, because variation in Mus is reduced to 9% variable positions in IRBP and 4% in RAG1  whereas 11% variable positions in the AA sequences remain in the vole avpr1a gene.
Our comparison of all currently available homologous nuclear genes for the mouse-rat-vole trio showed for the avpr1a gene a relatively high synonymous substitution rate but a comparatively low non-synonymous substitution rate (Figure 6). It is unclear to which extent this comparison is somewhat biased by a generally stronger interest and more published sequences of genes with high mutation rates (e. g. MHC, BRCA [68, 69]). It is nevertheless noteworthy that this comparison revealed higher nucleotide and protein diversity in avpr1a than in other related hormone receptors with similar functions (e.g. oxytocin, see in [16, 20, 70]; serotonin, see in [71, 72]; corticothropin, see in [73, 74]).
Our analyses show that genetic diversity in the avpr1a gene is much higher than previously claimed, and that part of this variation might be functionally relevant. We provide evidence for extensive variation in avpr1a at all taxonomic levels of mammals, with many changes in functionally important regions. We suggest that positive selection acting on these operative domains helps to maintain variation despite the presence of overall purifying selection. The role of balancing selection, particularly within the genus Microtus, should nevertheless deserve further investigation at the intra-specific level. The effects of genetic variation in avpr1a on phenotypic traits like mating systems, social behaviour or habitat requirements in Microtus and other mammals are far from being characterized. As this study shows, it seems particularly important to characterize abundant genotypic and phenotypic variation thoroughly before establishing general causal links between genotypes and phenotypes.
The V1aR is encoded by two exons: exon1 (~970 bp) and exon2 (~290 bp). We sequenced part (792 bp) of the first exon of the avpr1a gene since this fragment covers the two functionally important regions (ligand and G-protein binding domains) of the receptor. Sequences were analysed for 24 Microtus species which cover the entire Palearctic range of the genus (Europe, North America, Asia; Table 1). Tissue samples were obtained by live trapping with Longworth small mammal traps (Penlon Ltd), or from ecologists studying the species. Genomic DNA was extracted using a standard phenol-chloroform protocol  or Magnetic beads (MagneSil™ BLUE, Promega). We used two sequences from GenBank from M. ochrogaster and M. montanus (Accession numbers AF069304 and AF070010) to confirm locus identification.
Moreover, we sequenced three rodent taxa (Arvicola terrestris, Apodemus sylvaticus and Clethrionomys glareolus, see Table 1) and retrieved additional mammalian avpr1a sequence information from GenBank  and Ensembl  to compare sequence diversity and substitution rates for the avpr1a gene in mammals. Accession numbers in GenBank are: BC024149 for Mus musculus, NM_053019 for Rattus norvegicus, L41502 for Ovis aries, U19906 for Homo sapiens; Accession numbers in Ensembl:ENSCAFG00000000339 for Canis familiaris ENSBTAG00000007175 for Bos taurus, ENSMMUG00000000549 for Macaca mulatta, ENSPTRG00000005167 for Pan troglodytes, and ENSMODG00000014334 for Monodelphis domestica.
We amplified avpr1a sequences in a reaction volume of 25 μl in a GeneAmp® PCR System 9700 (Applied Biosystems) using Quiagen Taq polymerase. We used two primer pairs for amplification and sequencing reactions: V1aR-5'exon-ProtF 5'-GAGCTTAGGACAGGCTTTCTCG-3' and V1aR-5'exon-ProtR 5'-CGATCACGAAGGTCATCTTCAC-3', Mus-Mic-exon1f 5'-CCGACAGCATGAGTTTCC-3' together with Mus-Mic-exon1r 5'-CCACATCTGGACGATGAAGA-3'. The PCR amplification profile included an initial denaturation step at 92°C for 2 min, followed by 40 cycles of denaturation at 95°C for 1 min, annealing at 55°C for 1 min and extension at 72°C for 90 sec. A final extension step of 72°C for 10 min was performed. Amplified fragments were controlled for size on a 1.5% agarose gel by comparing them with a 100 base pair (bp) ladder (Invitrogen). After cleaning with GenElute™ PCR clean-up kit (Sigma) and dissolving products in 50 μl bi-distilled water, the sequencing reaction was carried out in a 10 μl reaction volume. Terminator Ready Reaction Mix 'Big Dye' Version 3.1 from Applied Biosystems was used. Both strands were sequenced using the following PCR conditions: An initial step of denaturation at 96°C for 10 sec, followed by 30 cycles of denaturation at 96°C for 10 sec, annealing at 55°C for 10 sec, and extension at 72°C for 4 min 30 sec. The products were cleaned using a DyeEx 96 spin kit (Quiagen), and were separated and detected on an ABI Prism 3100 Genetic Analyser from Applied Biosystems.
PCR products of individuals showing heterozygous sites in direct sequencing were cloned using the Qiagen PCR Cloning Kit. Purified PCR products were quantified in a Spectrophotometer (Gene Quant pro RNA/DNA Calculator, Biochrom) and approximately 65 ng of the product were ligated into pDrive Cloning Vector (Qiagen) in 10 μl reactions. Reactions were incubated for 45 min at 4°C before heat shock transformation into QIAGEN EZ Competent Cells. An additional incubation step of 45 min at 37°C with shaking was done before plating to allow recombinant growth. Cells were plated onto Kanamycin-IPTG-X-Gal agar and cultured for 17 h at 37°C. Ten positive clones per individual were randomly selected and further grown in LB broth for 17 h at 37°C with shaking. Plasmid miniprep columns (QIAprep® Spin Miniprep Kit, Qiagen) were used to purify each clone before sequencing with both M13 universal 5'-GTAAAACGACGGCCAGT-3'and M13 reverse 5'-CAGGAAACAGCTATGAC-3' primers. Sequencing conditions were as follows: An initial step of denaturation at 90°C for 50 sec, followed by 25 cycles of denaturation at 90°C for 10 sec, annealing at 50°C for 10 sec, and extension at 60°C for 4 min. After a final cleaning step with a DyeEx 96 spin kit (Quiagen), the sequences were run on an ABI Prism 3100 Genetic Analyser from Applied Biosystems.
Sequences were aligned using the Clustal W algorithm  implemented in the program BioEdit 5.0.9 , and were revised manually. Shared sequence types were detected using the program Arlequin 3.1 . Phylogenetic relationships among sequenced chromosomes were reconstructed by obtaining neighbour-joining (NJ)  and maximum likelihood (ML) trees rooted with Monodelphis domestica for the mammalian taxa and rooted with Arvicola terrestris for the Microtus genus with 10,000 bootstrap replicates in Mega 3  and Paup 4.0 b . For the ML analysis, Modeltest 3.06  implemented in Paup 4.0 b  was used to estimate the most suitable model of DNA substitution, by performing hierarchical likelihood ratio tests to compare 52 different models and by applying the Akaike Information Criterion . For the Microtus genus, the best substitution model was the transversion model with gamma distribution (TVM+G) with the following parameters: Substitution rate matrix: A↔C 2.7903; A↔G and C↔T 9.3807; A↔T 1.1820; C↔G 0.9720; G↔T 1.0000; and gamma distribution shape parameter 0.1986. The base frequencies were estimated as: A: 0.1801, C: 0.2951, G: 0.2963, T: 0.2285.
For the mammalian phylogeny, the best substitution model was the general time reversible model with invariable sites and gamma distribution (GTR+I+G) [86, 87]. The following parameters for the model were estimated: Substitution rate matrix: A↔C 1.7329; A↔G 5.3823; A↔T 0.6124; C↔G 1.4182; C↔T 3.8055; G↔T 1.0000; proportion of invariable sites 0.4474 and gamma distribution shape parameter 3.0860. The base frequencies were estimated as: A: 0.1566, C: 0.3344, G: 0.3228, T: 0.1862.
The nucleotide sequences were translated into AA sequences in Mega 3 using the universal code. The positions of the AA changes were determined using the structural model of the arginine-vasopressin 1a receptor of Mus musculus . To determine whether changes are equally distributed across the model, we applied Chi-Square tests for the different structural regions (ligand binding domain, transmembrane regions and G-protein binding domain). AA changes were classified as radical or conservative by comparing physicochemical properties of AAs such as charge, polarity and volume following Zhang .
To test for a link between V1aR types and phylogenetic relationships between Microtus, we checked for branch specific AA changes of avpr1a on a mitochondrial cytochrome b gene phylogeny [see in ] with sequences obtained from GenBank (accession numbers: AF119280, AF159400, AF163890 –AF163891, AF163893, AF163896, AF163900 –AF163901, AF163903–AF163906, AF187230, AY167210, AY220028, AY220770, AY513788, AY513798, AY513816, AY513819, AY513829, AY513837, AY513840, AY513845). To contrast the synonymous and non-synonymous diversity found in the avpr1a gene to other nuclear genes, we performed an exhaustive GenBank search for all annotated gene sequences available for Microtus species (up to december 20th, 2006). The resulting 31 sequences were aligned with homologous genes from Mus musculus and Rattus norvegicus and synonymous and non-synonymous substitution rates for each gene were computed with Mega 3.
We tested for regions under positive selection along the mammalian avpr1a by estimating the ratio ω of non-synonymous changes (dN) over synonymous changes (dS) per site. We used a sliding window approach with a window size of 30 and a step size of 10 with the program DnaSP 4.10 to compare mammalian species against the marsupial Monodelphis domestica.
To further test for the impact of selection on particular sites in avpr1a, we used a maximum likelihood approach with the single likelihood ancestor counting (SLAC) method implemented in HyPhy which makes no assumption about rate variation between lineages [88–90]. Further statistical tests for selection involved the computation of lineage-specific ratios of ω using codon-based maximum likelihood methods implemented in the program "codeml" from the PAML package . As a basis for these analyses, we used a phylogenetic tree tested for consistent topology between ML and NJ as well as with data from 3rd codon positions only [see ].
We used likelihood ratio tests in PAML to compare different neutral (MO, M1, M7) and selection (M2, M8) models of DNA sequence evolution of avpr1a. In all these tests, two times the log-likelihood difference (2Δ l) between models is compared to a χ2 distribution with the number of degrees of freedom (dF) equal to the difference in the number of parameters between the models . We tested for rate heterogeneity among lineages by comparing the one ratio model M0 against the discrete model M3 where different rates are allowed . This test is mainly used to check for rate variation of ω, but it can also be used to detect positive selection . Additionally, the neutral model M1 with two ratio classes of ω (< 1 and 1) was compared to the selection model M2 which allows for an additional class where ω > 1 . A similar comparison was carried out between a neutral model assuming a beta distribution of ω (M7), and a model with similar characteristics but allowing for positively selected sites (M8) . We performed branch specific tests to examine whether avpr1a evolves differently in the higher mammalian taxa by comparing the neutral model M1 with model MA which allows for positively selected sites on a pre-selected branch [91, 94].
We thank I. Dupanloup for helpful discussions and S. Tellenbach for technical assistance. We are grateful to the following people and institutions for providing access to samples: Museum of Vertebrate Zoology of the University of California, A. Bannikova, S. Braaker, R. Burri, Bündner Naturmuseum, F. Catzeflis, C. Conroy, B. Cushing, T. Derting, M. Jaarola, T. Maddalena, N. Martinkova, J.-P. Müller, M. Pfunder, R. Pita, J. Suchomel, J. Robovsky, J. Runge, L. Vinciguerra, P. Vogel. The Swiss National Science Foundation partly financed this study (project no. 112072).
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.