Introduction

The adenine nucleotide translocase (Ant), also called ADP/ATP carrier (Aac), mediates the exchange of ADP and ATP across the inner mitochondrial membrane, thus playing an essential role in energy metabolism in eukaryotic cells [1]–[3]. Under respiring conditions, ATP produced within the mitochondria is exported to the cytosol through Ant to support cellular activities. In exchange, ADP is imported to provide a substrate for the conversion of ADP to ATP by ATP synthase. Ant belongs to the mitochondrial carrier family that supports a variety of transport activities across the mitochondrial inner membrane [3], [4]. A typical Ant molecule comprises 300–320 amino acid residues that form six transmembrane helices.

The Ant family proteins are encoded by the nuclear genome. As free living bacteria lack Ant-like molecules, Ant proteins are thought to have evolved from a broad specificity transport family of eukaryotic origin [5]. Most eukaryotes, including unicellular eukaryotes, have multiple Ants. For example, the budding yeast Saccharomyces cerevisiae has three genes (AAC1, AAC2, and AAC3) that encode Ant proteins [6]. Among them, Aac2p is a major isoform that is abundantly expressed during respiration and repressed during fermentation [7]. Aac1p and Aac3p are expressed almost exclusively in aerobic and anaerobic conditions, respectively [6], [8]. Moreover, Aac2p and Aac3p have an ability to import ATP that is sufficient for survival of yeast after loss of the mitochondrial genome whereas Aac1p does not [9]. Thus, different Ants are likely utilized for ATP export and import, in order to efficiently cope with varying external nutrient and oxygen conditions. Interestingly, the S. cerevisiae AAC1 gene appears to have accumulated substitutions at a higher rate that AAC2/3 [10], consistent with the idea that the AAC1 gene underwent a functional change after duplication. Phylogenetic analysis has the ability to provide insight into the functions of the Ant genes in vertebrates.

In multicellular organisms, differential expression of Ant genes depends on a variety of factorsincluding tissue type, developmental stage, and cellular proliferation state. Most vertebrates possess three distinct Ant paralogs that exhibit a relatively high degree of sequence identity. Of these, Ant1 (Slc25a4) is expressed primarily in the heart and skeletal muscle, and is presumed to be suitable for rapid ATP metabolism in heart and skeletal muscles [11]. Genetic inactivation of Ant1 (Ant1−/−) resulted in viable mice; however, these animals developed mitochondrial myopathy and severe exercise intolerance along with hypertrophic cardiomyopathy as young adults [12]. Ant2 (Slc25a5) and Ant3 (Slc25a6) are expressed ubiquitously in somatic tissues; however, Ant2 expression is higher in rapidly growing cells and is inducible in mammals whereas Ant3 appears to be constitutively expressed in all tissues [13], [14]. It should be noted that rodents lack the Ant3 ortholog; instead, the mouse Ant2 ortholog likely carries out the functions of human Ant2 and Ant3 [15], [16].

Utilizing various approaches, we and others identified a fourth member of the Ant gene family, Ant4 (also called Slc25a31, Aac4, and SFEC) in both humans and mice [17]–[19]. The human ANT4 gene was predicted to encode a 315 amino acid protein with a relatively high degree of amino acid sequence identity to the previously identified ANT proteins (73%, 71% and 72% overall amino acid identity to ANT1, ANT2 and ANT3, respectively). Like all other ANT genes, ANT4 encodes a protein that has three tandem repeats of an approximately 100 amino acids containing two transmembrane regions, which is a characteristic shared by all members of the solute carrier family [20]. Further, the Ant4 protein contains a RRRMMM sequence at the end of the fifth α-helix transmembrane domain, a sequences that is conserved in all of the Ant proteins (but not in other Slc25s) that is essential for ADP/ATP transport activity [21]. Indeed, Dolce et al. [18] reported that Ant4 specifically exchanges ADP and ATP, but not other solutes, by an electrogenic antiport mechanism.

We have determined that Ant4 is expressed exclusively in testicular germ cells in adult mice, with its expression being particularly high during meiosis [22]. It should be noted here that Ant4 appears to be expressed in embryonic ovaries as well [23](Lim et al., unpublished observation). Further, Ant4 is essential for male germ cell development in mice, as Ant4 null male mice are infertile and exhibit meiotic arrest around the leptotene stage [22], [24]. The Ant2 gene is always located on the X chromosome of mammals; thus, Ant2 gene expression is repressed during male germ cell meiosis due to meiotic sex chromosome inactivation (MSCI) [22]. On the other hand, the Ant4 gene is always located on an autosome. A plausible hypothesis regarding the primary function of Ant4 is that it acts to compensate for the loss of Ant2 expression during male meiosis in mammals [22]. The meiotic arrest phenotype in Ant4 null male mice is indeed consistent with this compensation theory [24], and this theory would be consistent with the presence of Ant4 only in mammals. Since Ant4 is present in both eutherian and metatherian mammals, this hypothesis would place the duplication sometime between 150 million years ago (MYA), when the eutherian-metatherian divergence occurred [25], [26], and 300 MYA, when synapsids (Synapsida, the clade that includes mammals) diverged from the birds and true reptiles (clade Reptilia) [27].

This hypothesis is indirectly challenged by phylogenetic analyses that suggested a more ancient origin of Ant4 [22]. Here we present direct evidence for a more ancient origin of Ant4. Briefly, the anole lizard has an ortholog of Ant4 that exhibits testicular expression similar to that seen in mammals; however the anole Ant2 gene was not located on a sex chromosome. Moreover, additional phylogenetic analyses confirm that the Ant gene family in animals has a complex history of gene duplications and losses and further indicate that the Ant4 subgroup likely has an ancient origin. Based on these findings, we describe our new hypothesis, which is not necessarily mutually exclusive with the compensation theory, that there is a specific advantage for mammals and lizards to express Ant4 during spermatogenesis and/or in their sperm.

Results

Anole lizard has the Ant4 ortholog

The green anole lizard, Anolis carolinensis, is the first non-avian reptile genome to be sequenced. When we examined conservation of Ant gene orthologs in the anole, we found likely orthologs of all four Ant1, 2, 3 and 4 genes present in humans. Anole Ant1, 2, 3 and 4 genes encode proteins with 90%, 93%, 90% and 79% identity to human orthologs, respectively. Moreover, their intron-exon structure was conserved between the anole and mammals, with four exons observed in Ant1, 2 and 3 and six exons in Ant4 (Fig. 1A). Indeed, the deduced amino acid sequence of anole Ant4 had unique features only seen in the mammalian Ant4, including N-terminal and C-terminal extensions and a RRRMMMQSGE amino acid sequence at the end of the helix H5 region (Fig. 1B). To the best of our knowledge, this was the first time a clear ortholog of Ant4 has been identified outside of mammals.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Anole lizard has orthologs of the Ant1, 2, 3, and 4 genes.

(A) The schematic diagram from the Ensemble database shows the configuration of genes encoding Ant proteins in human and anole lizard. Exons and introns are shown as black boxes and lines, respectively. Numbers represent the amount of amino acid sequence identity that the products of each Ant gene in the human and anole lizard exhibit. (B) Alignment of the inferred amino acid sequences of the products of the Ant1, 2, 3, and 4 genes in mammals (human, cow, mouse) and reptile (anole lizard). Ant amino acid sequences of the selected species were aligned using ClustalW2. Amino acids sequence surrounding a signature motif in the Ant4 protein (RRRMMM) and the N- and C-terminal parts of the proteins are shown. Amino acid residues are numbered from the initiation codon (M) to the termination codon in each Ant protein.

https://doi.org/10.1371/journal.pone.0023122.g001

Consistent with the intron-exon structure and the signature sequences, phylogenetic analyses (Fig. 2A) verified the relationship between anole and mammalian Ant4 and further suggested that the Ant4 lineage is more closely related to the specific invertebrate Ant genes than it is to Ant1, 2, or 3. These analyses suggest the occurrence of a large number of gene duplications (and gene losses, given the inclusion of many organisms with relatively complete genome sequences that were included in our analysis) and further indicate that vertebrate Ant genes are found in a “core group” that also includes some invertebrate sequences. This core group was united by a long branch with a high degree of bootstrap support. Within this group, however, bootstrap support was relatively limited. Since the inclusion of very divergent outgroup sequences has the potential to distort the ingroup topology [29], [30] we excluded the outgroup sequences (along with two very divergent ingroup sequences) and repeated the phylogenetic analysis (Fig. 2B). These analyses confirmed the relationship between a set of tunicate Ant4-like genes and vertebrate Ant4 orthologs. Although the relevant tunicate genes are intronless and lack the signature sequences of vertebrate Ant4 orthologs, the phylogenetic analyses support the hypothesis that the divergence between the Ant4 group and the Ant1-3 group reflects an ancient duplication, substantially predating the origin of vertebrates.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. The maximum-likelihood (ML) estimate of Ant phylogeny supports an ancient origin of Ant4.

(A) Large scale phylogeny of animal Ant homologs obtained by ML analysis of amino acid sequences using the LG+Γ+F model of evolution. Support based upon 500 bootstrap replicates is shown as a percentage adjacent to the relevant branches when it exceeds 50%. This phylogeny included a fungal (Saccharomyces cerevisiae) outgroup, and the root of the tree was assumed to lie between fungi and the animal-choanoflagellate clade (the latter group is represented by Monosiga brevicollis). This phylogeny suggests a number of ancient gene duplications, one of which led to a “core group” that includes the vertebrate Ant homologs. (B) Phylogeney of the “core group” obtained using the same model. Support based upon 500 bootstrap replicates is shown as a percentage adjacent to the relevant branches when it exceeds 50%; support for some clades within well-supported groups (e.g. branches within teleost fish) was omitted in the interest of simplicity. Analyses using this more limited taxon sample were conducted to test the possibility that the position of Ant4 was influenced by the inclusion of divergent sequences. The position of Ant4 was robust to the taxon set analyzed; in fact, the bootstrap support for an ancient origin (the Ant4-tunicate clade) increased for the smaller taxon sample.

https://doi.org/10.1371/journal.pone.0023122.g002

Anole Ant4 is expressed in testis

We next examined the gene expression pattern of Ant1, 2, 3 and 4 in anole lizard. Heart, liver and testis were harvested from male anole lizard, and total RNA was extracted. The amount of Ant mRNA in these organs was then examined using quantitative RT-PCR (Fig. 3). Ant1, which is selectively expressed in heart and skeletal muscle in mammals, was also expressed in anole heart but was not detectable in anole liver and testis. Ant2 and Ant3, which are expressed ubiquitously in mammals, were also expressed in all of the anole tissues examined. Ant4, which is expressed exclusively in testis in adult mammals, was also expressed in anole testis but not in any somatic organs examined. These data indicated that all Ant orthologs exhibit similar patterns of gene expression in the anole and mammals. In particular, testicular expression of Ant4 suggests this gene likely plays a role in anole male germ cells, similar to the role of Ant4 in mice.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Anole Ant4 is expressed in testis.

Ant1, 2, 3 and 4 gene expression in heart, liver, and testis of anole lizard was examined by qRT-PCR. Ant mRNA amounts were normalized based upon the expression of the β-actin mRNA. Error bars indicate standard deviations of triplicate samples.

https://doi.org/10.1371/journal.pone.0023122.g003

Chicken has a degenerate DNA fragment that corresponds to a likely Ant4 pseudogene

Based on phylogeny and intron-exon structure, the anole Ant4 (Slc25a31) gene appears to be an ortholog of mammalian Ant4 genes. Moreover, the gene expression profile suggests that its function in the anole is similar to its function in mammals. This implies that the common ancestor of mammalian and reptilian species had the Ant4 gene. It's logical to ask what happened to the Ant4 gene in birds. To answer this question, we examined chromosomal regions harboring Ant4 in other vertebrates. As shown in Fig. 4A, the human ANT4 gene is located on chromosome 4q28.1, between the INTU gene and the HSPA4L, PLK4, and MFSD8 genes. Chromosomal synteny is conserved between human and anole genomes and these genes flank the 5′- and 3′-end of the Ant4 gene in the anole as well. These data further confirm that the anole Ant4 gene is an authentic ortholog of the mammalian Ant4 genes.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. Degenerate DNA fragment of putative Ant4 gene in the syntenic region of chicken genome.

(A) The syntenic region that includes Ant4 in the human, anole lizard and chicken genomes. The schematic diagram shows Ant4 (red boxes) with chromosomal location flanked by neighboring genes (blue boxes). Translated amino acid sequence of degenerate DNA regions of the putative Ant4 gene loci in the chicken genome was aligned with amino acid sequence of Ant4 of human and anole lizard. The asterisk in the chicken sequence is a stop codon; additional inactivating mutations include a 2-bp frameshift between nucleotide that would encode the D and K at the end of the chicken sequence and a stop codon immediately after the K. (B) Vertebrate phylogeny showing approximate divergence times (in millions of years before present) and the expected changes in ω (see text for definition) for the early and late inactivation models. (C) Phylogeny with branch lengths reflecting numbers of synonymous and non-synonymous substitutions. The light gray lines are included to make it easier to identify the taxon associated with each terminal, they have no biological significance.

https://doi.org/10.1371/journal.pone.0023122.g004

Examining the chicken genome demonstrated the syntenic region including Intu, Hspa4l, Plk4, Mfsd8 genes were present in this order on chromosome 4, although an Ant4 ortholog was absent from this region (Fig. 4A). However, it was possible to identify DNA fragments similar to the Ant4 gene between Intu and Hspa4l genes in the chicken genome and other avian genomes. For example, the sequence of chromosome 4 in the chicken genome (nucleotides 35430626 through 35431987 in build WUGSC 2.1/galGal3) includes two genomic regions which could be translated to 23 and 40 amino acid fragments with a high degree of similarity to human and anole Ant4 proteins. However, there is one nonsense codon and a two bp insertion that creates a frameshift in these regions. Similar degenerate Ant4-like DNA sequences were also evident in the turkey and zebra finch genomes, although the zebra finch sequence only includes a region similar to the second of these exons. An intron-like sequence was present at a site identical to the first intron of the human and anole Ant4 genes in the avian pseudo-coding regions, and these regions retain the GT sequence after the splice donor site and the AG before the splice acceptor (only the AG is retained in the zebra finch since the first exon-like region appears to have been deleted in this taxon). Moreover, the second exon is followed by a GT sequence in the zebra finch (this sequence is AT in the chicken and turkey, suggesting a mutation occurred in these lineages after their divergence from the finch lineage). These data confirm that the common ancestor of reptilian and mammalian species had the Ant4 gene, as suggested by phylogenetic analyses (Fig. 2), and indicate that the Ant4 gene has been degraded during avian evolution. Since we were not able to identify obvious synteny of the region in the zebrafish (Danio rerio) and frog (Xenopus tropicalis) genome, we were not able to elucidate whether degenerate Ant4 pseudogenes are present outside the amniotes.

The evidence suggests the Ant4 gene was inactivated in the common ancestor of the zebra finch (a member of the Passeriformes) and the chicken and turkey (both members of Galliformes). The avian supergroup (Neognathae) that includes both of these orders is very diverse, and includes virtually all orders of birds [31]. Given this constraint, it remains possible that the Ant4 was inactivated at any point before the divergence of these lineages (Fig. 4B). To place the timing of this inactivation in a framework that can be tested using phylogenetic methods, we simplified this to two models, an early inactivation model and a late inactivation model (indicated by the arrows in Fig. 4B). Fitting a set of models of codon evolution to the nucleotide sequence data for Ant4, including the two avian Ant4 pseudogenes, allows us to test these simplified models. The fundamental prediction is that the ω parameter, which describes the relative rates of nonsynonymous and synonymous substitutions, will increase to a value of one upon gene inactivation. As a heuristic, we present a comparison of trees with branch lengths proportional to the amounts of nonsynonymous and synonymous change (Fig. 4C). As expected for pseudogenes, the synonymous branch lengths were similar for birds and other taxa whereas the nonsynonymous branch lengths were much longer in birds (Fig. 4C). The length of the branch leading to the birds in the nonsynonymous tree was also long, a finding more consistent with the early inactivation model.

We tested five different models of codon evolution using this Ant4 data. All of these models involved shifts in the ω parameter, and the model set included one model with two different ω parameters that was more consistent with the early inactivation model and one that was more consistent with late inactivation. This allowed us to put the branch length heuristic in more rigorous framework. The best-fitting model based upon the AIC was the two-ω early inactivation model. However, the alternative model was not excluded from the 95% confidence set of models (Table S1). Nonetheless, the early inactivation model is better supported by the estimates of the ω parameter (Table S1) from other alternative models. Taken as a whole, the majority of the data suggest an early inactivation.

Anole Ant2 is not on a heterogametic sex chromosome

We have previously hypothesized that mammalian Ant4 genes may be conserved to compensate for the loss of X-linked Ant2 (Slc25a5) gene expression during male meiosis. It is important to note that anole lizards are considered male heterogametic reptiles, having XY chromosome differentiation [32], [33]. Thus, we hypothesized the anole Ant2 gene might also be located on the X-chromosome. Since chromosomal localization data are not available for a number of green anole genes, here we determined the relative copy numbers of genes between male and female animals by quantitative genomic DNA PCR (Fig. 5). First, in order to verify the methods, we compared gene dosages of Ant2 in male and female mice. When compared to the control β-actin gene, the relative copy number of the X-linked Ant2 gene was approximately double in female mice when compared to male mice. In contrast, dosage of the autosomal Ant4 gene was similar between male and female mice. These results indicate this is a valid method for determining relative copy number of genes. When we applied the method to anole lizard, relative copy number ratios were approximately one for all Ant genes, including Ant2. In addition, multiple primer sets were used to confirm the results for the Ant2 and Ant4 genes. Similarly, the relative copy number ratio between male and female were examined for Slc25a43, another gene encoding a solute carrier protein that is located adjacent to the Ant2 gene in the syntenic region of both the mouse and anole genomes. As seen with the Ant2 gene, the relative copy number ratio of the Slc25a43 gene was approximately two in female and male mice but approximately one in female and male anole (Fig. 5B). Taken together, these data indicate that the anole Ant2 gene is not located on a heterologous region of the X chromosome, but either on an autosome or on the pseudoautosomal regions of the anole X and Y chromosomes.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 5. Gene dosage analysis in female and male animals.

Relative copy number ratio of Ant1, 2, 3 and 4 genes and Slc25a43 gene in male and female animals were examined by quantitative PCR analysis of genomic DNA of anole lizard and mouse. PCR amplification was normalized to the control β-actin gene. Error bars indicate standard deviations of three independent experiments. (A) Relative gene dosages of the Ant 1, 2, 3 and 4 between female and male in anole lizard and mouse. The result was confirmed with two different sets of primers for anole Ant2 and Ant4 genes. (B) Relative gene dosage of the Slc25a43 between female and male in anole lizard and mouse. The Slc25a43 gene localizes adjacent to the Ant2 gene in both anole and mouse genome (left panel).

https://doi.org/10.1371/journal.pone.0023122.g005

We previously demonstrated the promoter region of the murine Ant2 gene is partially methylated at CpG dinucleotides in female somatic tissues, which is to the result of somatic X-chromosome inactivation of one allele in females. Here we examined CpG methylation in the promoter region of the Ant2 gene in male and female tissue samples from anole lizards and mice. As we previously described, the murine Ant2 promoter was unmethylated in male tail tissue, whereas female tail tissue showed hypermethylation of the Ant2 promoter in approximately half of the samples examined (Fig. 6). In contrast, the anole Ant2 promoter was largely unmethylated in both male and female liver samples. These data are consistent with the idea that Ant2 is located on an autosome or undifferentiated regions of sex chromosomes in anole lizard.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 6. CpG methylation of the anole and mouse Ant2 genes.

The CpG methylation of the Ant2 gene promoter regions in female and male anole lizard and mouse was examined by bisulfite sequencing analysis. The schematic diagrams indicate CpG islands in the Ant2 gene promoter regions examined in anole lizard and mouse. Each row of the circles represents an individual clone of a PCR amplicon in bisulfate sequencing analysis.

https://doi.org/10.1371/journal.pone.0023122.g006

Discussion

Previously, we hypothesized Ant4 emerged at least 150 million years ago [22] based upon gene conservation in both eutherian and metatherian mammals. Here we demonstrated that anole lizard has an authentic Ant4 ortholog based upon a number of criteria, including phylogeny, intron-exon structure, synteny, and pattern of gene expression. Further, degenerate DNA fragments of a putative Ant4 pseudogene were identified in the syntenic region of avian genomes. These indicate the presence of the Ant4 gene in the common amniote ancestor. This indicates that origin of the Ant4 gene clearly predates the divergence of mammals and reptiles, more than 300 MYA.

Phylogenetic analyses suggest that the gene duplication that led to the Ant4 lineage and the Ant1, 2, and 3 lineage may actually be much more ancient than the origin of amniotes. A number of invertebrate lineages were nested within the diversity of vertebrate Ant genes (Fig. 2A), a topology similar to the estimate of phylogeny obtained by Brower et al. [22]. However, there are a number of reasons that one can obtain an incorrect estimate of phylogeny [27], [28], and the branch length heterogeneity evident in the large-scale animal Ant tree is a major cause for concern (especially for the Brower [22] tree that was based upon NJ of protein distances that were corrected by the Kimura [34] formula). To address concerns regarding the phylogeny we used ML with the best fitting model (LG+Γ+F) to estimate the tree for this study, addressing the issue of model adequacy. We also examined two sets of sequences, one of which excluded the divergent outgroups and other relatively long branches (Fig. 2B) because divergent sequences can act as “rogues” that can distort the topology [29], [30], [35]. Rogue taxa are characterized with an unstable phylogenetic placement when relatively subtle details of analyses are changed and they can have a negative effect upon the estimation of both topology and bootstrap support values. Despite our concern regarding the inclusion of divergent, potentially rouge sequences, the estimate of phylogeny actually appeared quite robust to the specific sets of taxa that were included. However, the exclusion of divergent sequences did appear to improve the estimate of phylogeny in relatively subtle ways. Specifically, the topology for both the Ant1 and Ant3 subgroups were more congruent with the species tree (e.g., analyses of the large taxon sample placed the root within tetrapods whereas analyses of the smaller taxon sample placed it between fish and tetrapods, with the latter being more likely to be correct). Significantly, we found a higher degree of bootstrap support for the clade including vertebrate Ant4 and tunicate Ant4-like sequences when the more limited taxon sample was analyzed, consistent with our hypothesis that the Ant4 group is the result of an ancient duplication.

It is clear that all estimates of Ant phylogeny demand a number of duplications and losses to reconcile them with the animal tree of life. Some of the observed incongruence between expectation based upon the species tree and the estimates of Ant phylogeny may reflect the lack of power associated with relatively short protein sequences [36]. Indeed, fairly long sequences can be necessary to have a high probability of obtaining an accurate estimate of phylogeny [37], [38]. In addition to issues associated with the power of phylogenetic analyses to distinguish among alternative trees, the observed branch length heterogeneity (Fig. 2) suggests that the patterns of evolution for Ant genes have undergone substantial changes during animal evolution. More specifically, the branch length for the Ant4 group is somewhat longer than those for the Ant1, 2, 3 group, suggesting that the latter group has retained more of the ancestral functions of animal Ant proteins. Indeed the distance between the anole and human Ant4 (0.3608 amino acid substitutions per site, based upon patristic distances calculated from the tree in Fig. 2B) indicates that the rate of amino acid evolution was approximately 2.4-fold faster for the product of Ant4 than it was for Ant1, 2, and 3 gene products (the mean anole-human distance for the Ant1-3 group was 0.1484 amino acid substitutions per site). These rate estimates based upon amino acid substitutions are consistent with estimates of ω (Table S2), emphasizing the fundamentally different patterns of sequence evolution for Ant4 and Ant1-3 groups.

The observed changes in the patterns of evolution for Ant genes may have impacted the estimate of Ant phylogeny, although the heterogeneity does not appear so extreme as to radically alter our conclusions regarding the evolution of the vertebrate Ant gene family. The basis for the higher rate of nonsynonymous sequence evolution evident in the Ant4 group is unclear, although it is known that genes with sex-biased expression often exhibit higher rates of evolution [39]. The basis for the observed rate differences for sex-biased genes are complex [40], but it is important to note that our observation that Ant4 exhibits a higher rate of evolution is consistent with a role for this gene in male reproduction in various amniotes. Alternatively, the higher rate of Ant4 evolution may reflect changes in selection unrelated to sex-biased expression, either the relaxation of purifying selection, the action of positive selection having acted upon Ant4, or both phenomena. Indeed, when the branch length heterogeneity observed in our Ant trees (Fig. 2) is combined with the evidence for the action of natural selection upon other mitochondrial proteins [41], [42] it seems reasonable to speculate that various Ant proteins have been subject to episodes of positive selection during animal evolution. Regardless of the specifics, both hypotheses regarding Ant4 (relaxed purifying selection or positive selection) are consistent with the idea that Ant4 has undergone a shift in function.

Pseudogenes exhibit relaxed selection on nonsynoymous sites after the relaxation of purifying selection, and ω is expected to approach one after a period of time [43]. Although the fact that the ML estimates of ω for the birds were less than one (ω = 0.5146 for exons 1 and 2 and ω = 0.3921 for exon 2 alone; also see Table S1) might seem more consistent a relatively recent gene inactivation, it is important to note that the confidence interval on estimates ω after the relaxation of purifying selection tends to be quite wide [43]. Instead, the best-fitting models (Table S1) suggest the relaxation of purifying selection upon Ant4 in birds occurred substantially earlier than the Passeriformes-Galliformes divergence at least 87 MYA [37]. It will be interesting to determine whether a functional Ant4 gene is present in any other extant archosaurs (birds and crocodilians) and whether there are any phenotypic correlates of the Ant4 gene loss.

The present study also urged us to reconsider our original hypothesis that was based upon the conservation of Ant4 only in mammals, since it is clearly present both in mammals and anole lizards. Based on the unique chromosomal localization pattern of the Ant genes as described in the introduction, we originally hypothesized that the conservation of Ant4 is driven by the sex chromosomal localization of the Ant2 gene and the subsequent inactivation of Ant2 during male meiosis in mammals. When we identified an authentic Ant4 ortholog in anole lizard, we first expected the anole Ant2 gene to be located on its sex chromosome as well, which would strengthen the “compensation” hypothesis. Among reptiles that have various sex determination systems, anoles are considered to have male heterogametic sex determination like mammals [32], [33]. Although the scenario was rather unlikely as sex chromosomal differentiation presumably developed independently in mammals and reptiles after the bifurcation of the two species, it is possible that the Ant2 gene can be localized to sex chromosomes in both mammals and anoles by chance. Interestingly, an Ant gene homolog was isolated from heteromorphic sex chromosomes in the Japanese frog Rana rugosa [44], an unusual species where different populations have three distinct types of sex chromosomes (homomorphic XY chromosomes, heteromorphic XY chromosomes, and heteromorphic ZW chromosomes). Although the gene was originally annotated as Ant3 in the report [44], it actually shows greater sequence identity to Ant2 rather than Ant3 (91% and 88% to human Ant2 and Ant3, respectively). Placement of the Rana rugosa sequences using phylogenetic criteria was more equivocal, as they formed a clade with putative Xenopus Ant3 ortholog (data not shown). However, the putative Xenopus Ant3 sequence was placed either within fish (Fig. 2) or sister to fish (when the Rana sequence was added); this topology suggests either a local rearrangement in the tree (likely due to the relatively short length of the Ant sequences) or additional gene duplications and losses. The absence of bootstrap support for separating Ant2 and Ant3 (Fig. 2) is consistent with the first hypothesis, which postulates that the relevant branches in our estimate of phylogeny are rearranged relative to the actual evolutionary history. Regardless, it is clear that the sex chromosomal localization of a member of the Ant2/3 group is not unique to mammals. There could be a yet unidentified common mechanism or reason why members of this subgroup of Ant genes have a higher probability of moving to a sex chromosome after their origin. However, it should be noted that Ant2 gene localizes to autosomes in birds (chicken and zebra finch); thus the sex linkage of Ant2 is not a universal phenomenon in vertebrates.

Chromosome maps are not completed to date for Anolis carolinensis, making it difficult to determine whether or not the anole Ant2 gene is localized at a sex chromosome. Instead here we investigated relative gene dosage of the gene between females and males. In mice, dosage of the Ant2 gene as well as its neighbor Slc25a43 gene were approximately 2∶1 between females and males, consistent to the fact that these genes are localized at heterologous regions of the X chromosome. In contrast, in anoles, both Ant2 and Slc25a43 genes doses were 1∶1 between females and males. These data indicate that the Ant2-Slc25a43 region is not likely on a heterologous region of the sex chromosomes. One of two X chromosomes in the somatic cells of female mammals is known to be inactivated accompanying CpG methylation. Indeed, we previously showed that the mouse Ant2 gene was partially methylated in females but not in males [45]. Here, we demonstrated that anole Ant2 gene was unmethylated both in females and males, which is consistent with the idea that the Ant2 gene is not on the anole X chromosome. It should be noted, however, that there is no information available regarding the mechanism (or even the existence) of dosage compensation for sex-linked genes in anoles. Thus, it is unclear whether X chromosomes undergo methylation and/or inactivation in anoles in a similar mechanism as mammals, though it is known that reptilian genomes typically have a CpG methylation ratio comparable to that of other amniotes [46]. Although a definitive conclusion regarding the chromosomal location of Ant2 in Anolis carolinensis will require an improved anole genome assembly, our present data suggest that the Ant2 gene is not likely localized to a heterogametic sex chromosome in anoles.

Our results imply the conservation of Ant4 may not be simply driven by the sex chromosomal localization of Ant2 gene and its subsequent inactivation during male meiosis. Importantly, our data suggest that testicular expression of Ant4 existed prior to and is independent of the sex chromosome linkage of the Ant2 gene in mammals. This prompted us to hypothesize that Ant4 has been functioning originally in testicular germ cells and sperm, in the amniote ancestor and potentially even earlier in evolutionary phylogeny. The degeneration of Ant4 in birds indicates that this Ant paralog is unnecessary for the development of testicular germ cells in this lineage. To this end, one alternative hypothesis could be that Ant4 plays a critical additional role during spermatogenesis and/or in sperm function. This role would be shared by specific groups of organisms such as mammals and anoles (and probably by at least some other squamates) and absent in other groups of organisms including birds. Likewise, the absence of Ant4 orthologs in the frog genome and all teleost fish genomes suggest it is absent from these lineages as well, and the existence of multiple fish genome sequences strongly suggests this absence is real rather than a problem with the completeness (or assembly and annotation) of the available draft genome sequences. Our phylogeny implies that duplication leading to the Ant4 lineage predates the origin of vertebrates, indicating that the absence of Ant4 in this lineage likely reflects gene loss. The loss of Ant4 in birds and fish could reflect the relaxation of selection of this paralog or the existence of a specific advantage linked to the loss of this gene, since some evolutionary innovations can be linked to gene loss [47]. Regardless, it seems clear that conservation of Ant4 in anoles and mammals cannot be explained by the simplest version of the compensation theory, which postulates selection to retain the Ant4 gene after duplication reflects a need to compensate for the loss of Ant2 expression during male meiosis in mammals. Instead it suggests that vertebrate Ant4 orthologs may have additional specific functions associated with male germ cells prior to the need for compensation.

Then, what can be an additional ‘specific’ function of Ant4 that is essential for male germ cells only in anole and mammals? Interestingly, the CatSper genes are also conserved in mammals and the anole lizard, but not in birds, frogs, and bony fishes, similar to Ant4 [48]. CatSper genes encode sperm-specific voltage-gated Ca²⁺ channels that are crucial for sperm hyperactivation and male fertility in mammals [49]. Hyperactivated motility in mammals occurs as the sperm encounter progressively more alkaline environments as they ascend the reproductive tract [50]. This sperm hyperactivation characterized by a large bend angle of head and tail is proposed to be required for sperm to penetrate the cumulus and thick wall of the zona pellucida in mammals [51]. It is proposed that the unique conservation pattern in species having CatSper genes reflect the necessity of sperm hyperactivation for fertility [48]. It has long been appreciated that phylogenetic profiles (patterns of gene presence and absence) can provide information regarding gene function [52]. Although this hypothesis remains speculative, the analogous patterns of gene conservation for Ant4 and Catsper in vertebrates combined with the evidence that Ant4 is expressed in sperm imply Ant4 might be critical for sperm hyperactivation. Since the hyperactivated motility likely requires increased ATP [53], it is reasonable to imagine that the ATP producing machinery, including the product of the Ant4 gene, could be involved in hyperactivation. Although glycolytic ATP is largely implicated in mammalian sperm motility, it is still controversial whether mitochondrial ATP plays a role [54], [55]. As Ant4 null mice exhibit meiotic arrest in male germ cells, we were unable to determine whether Ant4 plays a specific role in mouse sperm after meiotic arrest. It would be interesting to examine whether expression of an Ant2 transgene in Ant4 null mice can fully compensate for the loss of Ant4 or whether it is able to restore male germ cell meiosis but not in sperm hyperactivation. It should be noted, however, that multiple CatSper genes were lost in specific lineages, such as birds and fish, and that CatSper homologs are present in a number of invertebrates [48]. The tunicate Ant4-like genes appear likely to be orthologous to Ant4 based upon phylogenetic criteria, but there is no other evidence (such as intron-exon structure) uniting them with vertebrate Ant4 genes and it is unclear whether their function shows specific similarities to Ant4. Further evolutionary genomic analyses will be required to fully elucidate the similarities and differences among the Ant/Aac proteins expressed by distant species.

The idea that Ant4 has an additional specific function in male gametes is supported by the fact that there are some differences in protein structure between Ant4 and the somatic Ant proteins [17], [18], [22]. There are common features seen only in Ant4, including N-terminal and C-terminal extensions and acidic amino acids between the end of the helix H5 region and the matrix M5 region. In addition, Dolce et al. [18] demonstrated that Ant4 has a lower affinity for adenine nucleotides than the other Ant proteins [56]. The structural and kinetic difference between somatic Ant proteins and Ant4 could imply that selection has optimized the Ant4 protein to function better in the helically arranged sperm mitochondria [57]. Alternatively, they could reflect the existence of a highly specific function associated with the Ant4 protein combined with relaxed selection of other biochemical properties. Interestingly, the ANT4 protein was isolated from the fibrous sheath of the human sperm flagellar principal piece using mass spectrometry proteomics and was shown to co-localize with glycolytic enzymes [19]. This study implied an additional function of Ant4 independent of mitochondrial oxidation. Although we were not able to detect Ant4 protein expression beyond the mid-piece of sperm in both mice and humans (unpublished data), in contrast to the report by Kim et al [19], the possibility that Ant4 protein has an extended localization and/or function is certainly an interesting hypothesis that needs to be clarified in the future.

Materials and Methods

Supporting Information

Acknowledgments

We thank Dr. Katherine Hankowski for critical reading of the manuscript, Drs. Jim Resnick, Steve Sugrue and Paul Oh (U Florida) for helpful suggestions.