Abstract
The Japanese medaka fish Oryzias latipes has an XX/XY sex-determination system. The Y-linked sex-determination gene DMY is a duplicate of the autosomal gene DMRT1, which encodes a DM-domain-containing transcriptional factor. DMY appears to have originated recently within Oryzias, allowing a detailed evolutionary study of the initial steps that led to the new gene and new sex-determination system. Here I analyze the publicly available DMRT1 and DMY gene sequences of Oryzias species and report the following findings. First, the synonymous substitution rate in DMY is 1.73 times that in DMRT1, consistent with the male-driven evolution hypothesis. Second, the ratio of the rate of nonsynonymous nucleotide substitution (dN) to that of synonymous substitution (dS) is significantly higher in DMY than in DMRT1. Third, in DMRT1, the dN/dS ratio for the DM domain is lower than that for non-DM regions, as expected from the functional importance of the DM domain. But in DMY, the opposite is observed and the DM domain is likely under positive Darwinian selection. Fourth, only one characteristic amino acid distinguishes all DMY sequences from all DMRT1 sequences, suggesting that a single amino acid change may be largely responsible for the establishment of DMY as the male sex-determination gene in medaka fish.
MOST animals have two sexes. However, whether an undifferentiated embryonic gonad eventually develops into a testis or an ovary is determined by genetic, environmental, or both factors, depending on the species concerned. For placental mammals, in which females have two X chromosomes and males have one X and one Y chromosome, male sex is determined by a Y-linked gene SRY (Bertaet al. 1990; Sinclairet al. 1990). It has been proposed that SRY originated by divergence from its X-linked homologous gene SOX3, probably due to the lack of recombination between the two genes (Stevanovicet al. 1993; Foster and Graves 1994). It is difficult to delineate the initial steps that led to the emergence of SRY, because SRY likely originated before the separation of placental and marsupial mammals (Fosteret al. 1992; Foster and Graves 1994) about 170 million years ago (MYA; Kumar and Hedges 1998). The identification of the male sex-determination gene DMY in the Japanese medaka fish Oryzias latipes (Matsudaet al. 2002; Nandaet al. 2002) provides an excellent opportunity for such a study of the origin of sex-determination genes. Similar to the situation in mammals, O. latipes has XX/XY sex chromosomes. But, in contrast to the small size of mammalian Y, medaka Y is identical in morphology to X. Because the Y chromosome usually undergoes evolutionary degeneration when it no longer recombines with X (Nei 1970; Charlesworth and Charlesworth 2000), the observation in medaka suggests a recent origin of its Y as well as its sex-determination system. In fact, the Y-specific region in medaka originated from transposition of an autosomal region and it spans ∼280 kb, containing only one functional gene, DMY (Nandaet al. 2002). A naturally occurring insertion in DMY that leads to the truncation of the gene results in XY females (Matsudaet al. 2002). A second mutation that reduces the level of DMY expression also results in a high frequency of XY females (Matsudaet al. 2002). These observations demonstrate the necessity for DMY in male sex determination and strongly suggest it to be the medaka counterpart of the mammalian SRY gene (Matsudaet al. 2002; Nandaet al. 2002). A recent survey of several different strains, however, identified some male medaka fish that do not possess DMY, suggesting the possibility of other contributing factors in male sex determination (Nandaet al. 2003). DNA sequence analysis shows that DMY is a paralog of an autosomal gene DMRT1, which encodes a conserved zinc-finger transcription factor that is found in mammals, birds, reptiles, fruitflies, and nematodes (Zarkower 2001; Matsudaet al. 2002; Nandaet al. 2002). Phylogenetic and evolutionary analyses further indicated that the duplication event was recent, as the DMY gene is found only in two closely related species (O. latipes and O. curvinotus), but is absent from other species of the genus Oryzias so far examined (Kondoet al. 2003; Matsudaet al. 2003). In this work, I analyzed the publicly available DNA sequences of the DMY and DMRT1 genes from Oryzias in an attempt to address the following questions. First, when did the duplication occur? Second, is the mutation rate in DMY higher than that in DMRT1? This question is raised because DMY is located on Y, which is always in males, while DMRT1 is in males half the time and in females half the time. The hypothesis of male-driven evolution asserts that the mutation rate is higher in males than in females due to more rounds of cell divisions in the male germ line (Liet al. 2002). Third, does DMY exhibit a high rate of protein evolution? I raise this question because the mammalian male sex-determination gene SRY shows a rapid pace of sequence evolution among species of primates and rodents (Tucker and Lundrigan 1993; Whitfieldet al. 1993). Fourth, is there any difference in amino acid substitution pattern between DMY and DMRT1, as such a difference may provide information on the critical regions of the protein that confers the new function of DMY? Finally, how many amino acid substitutions may be responsible for DMY’s new role in sex determination?
MATERIALS AND METHODS
The DMY and DMRT1 gene sequences from O. latipes HNI strain, O. latipes Carbio strain, O. curvinotus, and the DMRT1 sequence of O. celebensis were obtained from GenBank. The GenBank accession numbers are listed in the legend to Figure 1. The protein sequences were aligned using CLUSTAL X (Thompsonet al. 1997) and the DNA sequences were then aligned following the protein alignment. It was found that there is a one-nucleotide deletion near the end of the coding region in the O. latipes HNI DMY gene, which results in a frameshift and earlier termination of the open reading frame (see Figure 1). The nucleotide sequence alignment was thus manually adjusted by incorporating this deletion, and the sequence region that is downstream of this deletion is not used in the evolutionary rate analysis because of the difficulty in assigning synonymous and nonsynonymous differences. Gene trees were reconstructed using the neighbor-joining method (Saitou and Nei 1987) implemented in MEGA2 (Kumaret al. 2001), as well as the likelihood method (Felsenstein 1981) implemented in PAUP* (Swofford 1998). Several different distance measures or substitution models (Jukes-Cantor, Kimura’s two parameter, Tajima-Nei, and Tamura-Nei; Nei and Kumar 2000) were used. The bootstrap test (Felsenstein 1985; 2000 replications for neighbor joining and 200 replications for likelihood) was used to examine the reliability of the reconstructed trees. Four-cluster analysis was performed by the PHYLTEST program (Rzhetskyet al. 1995). Ancestral gene sequences were reconstructed for interior nodes by the distance-based Bayesian method (Zhang and Nei 1997). Numbers of synonymous (s) and nonsynonymous (n) substitutions on tree branches were counted as in Zhang et al. (1997). A maximum likelihood method (Yang 1998) was used to analyze changes in w = dN/dS in the evolution of DMY and DMRT1 genes. Here dN and dS refer to the numbers of nonsynonymous and synonymous substitutions per site, respectively. The likelihood method (Yanget al. 2000) was also used for testing positive selection on individual sites.
RESULTS
Phylogenetic relationships of DMRT1 and DMY genes of Oryzias species: The DMY gene was found only in O. latipes (HNI and Carbio strains) and O. curvinotus, but not in other species of the Oryzias genus so far examined (Kondoet al. 2003; Matsudaet al. 2003). This would suggest that the gene duplication that gave rise to DMY occurred in the common ancestor of O. latipes and O. curvinotus after this ancestor diverged from other Oryzias species (Matsudaet al. 2003). To verify this hypothesis, I first aligned the DMY and DMRT1 sequences (Figure 1) and then reconstructed a gene tree by the neighbor-joining method (Figure 2a). The tree shows the clustering of DMY sequences from the two strains of O. latipes. However, the DMY gene of O. curvinotus does not cluster with those of O. latipes. Rather, it clusters with O. curvinotus DMRT1. Nevertheless, this grouping has only 50% bootstrap support, suggesting a possibility that it may not reflect the true relationships among the sequences. A likelihood analysis resulted in the same tree with slightly different bootstrap numbers (Figure 2a). From these analyses, it appears that the DMY genes from the two strains of O. latipes form a cluster (with 100% bootstrap support) and the DMRT1 genes from O. latipes form another cluster (with >97% bootstrap support). It is the relations of these two clusters and the DMY and DMRT1 genes of O. curvinotus that are in question. I then used a distance-based four-cluster analysis, which is particularly suitable for evaluating the relationships among four monophyletic groups (Rzhetskyet al. 1995). This analysis shows that none of the three possible groupings of the above four clusters is significantly better than the other groupings (P > 0.2; Figure 2 legend). That is, there is not sufficient phylogenetic information in the data that can resolve the relationships among the four clusters. It is well known that sex-determination mechanisms are highly variable among animals (Zarkower 2001). It is thus very unlikely that an identical sex-determination system evolved twice in two closely related species by independent duplications of the same gene. Therefore, it is parsimonious to assume that DMY originated only once in Oryzias, as depicted in Figure 2b. Further analyses are based mainly on this assumption. However, to examine whether the results obtained are robust, I also repeated all the analyses without using the two sequences of O. curvinotus, as the tree becomes well resolved when these sequences are excluded (Figure 2a).
A higher rate of synonymous substitution in DMY than in DMRT1: To study the rate of nucleotide substitution in the evolution of DMY and DMRT1, I inferred the ancestral gene sequences on all interior nodes of the tree shown in Figure 3, which has the same topology as the tree of Figure 2b. Because the sequences are relatively closely related, this inference is expected to be reliable. In fact, the computed average posterior probability is >99% for all the ancestral sequences. I then counted the numbers of synonymous and nonsynonymous substitutions on each tree branch (Figure 3a). There are a total of 31.6 synonymous substitutions in DMY, but only 18.3 in DMRT1 for all the branches since the duplication event. The difference between the two numbers is statistically significant (P = 0.03, one-tailed z-test), indicating a higher rate of synonymous substitution in DMY than in DMRT1. Because synonymous changes are more or less neutral and are generally immune to selection (except in the case of weak selection on codon usage in certain organisms), the above result suggests a higher mutation rate in DMY than in DMRT1. Since DMY is on the Y chromosome, which is always in males, and DMRT1 is on an autosome, which is in males half the time and in females half the time, my observation suggests a higher rate of mutation in males than in females. Note that this inference is not affected by the fact that some males do not have DMY (Nandaet al. 2003) as long as DMY is always in males, which has never been violated. The ratio of the mutation rate in males to that in females may be estimated by αm = 1/(2A/Y - 1) (Miyataet al. 1987), where A/Y is the mutation rate ratio between autosome and Y chromosome and is 18.3/31.6 = 0.579 for this data set. Thus, αm is 6.32. This result supports the male-driven evolution hypothesis, which asserts that the mutation rate is higher in males than in females (Miyataet al. 1987; Liet al. 2002). When the two sequences of O. curvinotus are excluded from the analysis (Figure 4), A/Y = 8.3/15.1 = 0.550 and αm is 10.1.
—Protein sequence alignment of DMY and DMRT1 genes of medaka fish (Oryzias species). The DM domain is boxed. Amino acid substitutions in DMY that occurred in the common ancestor of O. latipes and O. curvinotus (see tree b of Figure 2) are indicated by an asterisk. Sequences downstream of the vertical line are not used in the evolutionary rate analyses because of a frameshifting deletion in O. latipes (HNY) DMY. Carbio and HNI are two strains of O. latipes. The sequences are obtained from GenBank, with the accession numbers as follows: O. latipes (Carbio) DMY, AY129240; O. latipes (HNI) DMY, AY129241; O. curvinotus DMY, AB091695; O. latipes (Carbio) DMRT1, AF319994; O. latipes (HNI) DMRT1, AY157712; O. curvinotus DMRT1, AB091696; and O. celebensis DMRT1, AY239587.
—Phylogenetic relationships of seven DMY and DMRT1 gene sequences of medaka fish. (a) The same tree generated by the neighbor-joining and likelihood methods. The neighbor-joining tree was reconstructed with Kimura’s (1980) two-parameter distance, whereas the likelihood tree was reconstructed under the general-reversible model with variable rates of substitutions across sites (plus a class of invariant sites). Bootstrap percentages from neighbor-joining and likelihood analyses are shown on interior branches. (b and c) Alternative trees when all possible relations among groups A, B, C, and D are considered. In a four-cluster analysis (see text), tree a is found to be better than tree c, which is in turn better than tree b; but none of the differences among the three trees are statistically significant (P > 0.2).
—Nucleotide substitutions in the evolution of DMY and DMRT1 of medaka fish for (a) the entire protein, (b) the DM domain, and (c) the non-DM regions. The tree with a single origin of DMY, as depicted in Figure 2b, is used. The numbers of synonymous (s) and nonsynonymous (n) substitutions for each tree branch are shown on branches. The sums of n for all branches involving DMY (after gene duplication), DMRT1 (after duplication), and DMRT1 (before duplication) are indicated by respective Σn values. Σs values are similarly obtained. N and S are potential numbers of nonsynonymous and synonymous sites of the sequences, respectively. For the entire protein, N = 523.0 and S = 212.0. For the DM domain, N = 143.7 and S = 54.3. For non-DM regions, N = 379.3 and S = 157.7.
It is important to know when DMY originated. The divergence times for the species concerned are unknown, making it difficult to use DMRT1 or DMY data for time estimation directly. It has been estimated that the nucleotide mutation rate in mammals is on average 2.2 × 10-9/site/year (Kumar and Subramanian 2002). If the mutation rate in medaka DMRT1 is similar to the above rate, it can be estimated that the duplication occurred (7.15/212.0)/(2.2 × 10-9) = 15.3 MYA. Here 7.15 is the average number of synonymous substitutions in DMRT1 from duplication to present, and 212.0 is the number of synonymous sites in the gene, estimated by the modified Nei-Gojobori method (Zhanget al. 1998). Ellegren and Fridolfsson (2003) estimated the synonymous substitution rates for four autosomal genes from salmonid fish Oncorhynchus kisutch and O. tshawytscha, which were estimated to be separated ∼5 MYA (Ooharaet al. 1997). Weighted by the sequence length, these rates averaged 3.6 × 10-9/site/year. If the synonymous substitution rate in medaka DMRT1 is similar to this rate, I estimated that DMY originated (7.15/212.0)/3.6 × 10-9 = 9.4 MYA. When I exclude the two sequences from O. curvinotus and repeat the above estimation, the date of duplication becomes 5.7–9.2 MYA. Although these estimates may have large errors, it appears that the age of DMY is on the order of 10 MY.
A higher rate of nonsynonymous substitution in DMY1 than in DMRT1: The evolutionary analysis also reveals a higher rate of nonsynonymous substitution in DMY than in DMRT1 (Figure 3a), as the total number of nonsynonymous substitutions since gene duplication is significantly greater in DMY (66.4) than in DMRT1 (21.7; P < 10-5, two-tailed z-test). This phenomenon is due in part to the elevation of mutation rate in DMY. It may also arise from an alteration in natural selection on DMY, compared to that in DMRT1. The n/s ratio is greater in DMY (66.4/31.6 = 2.10) than in DMRT1 (21.7/18.3 = 1.19) since gene duplication (Figure 3a). However, the difference is statistically insignificant (P = 0.12, Fisher’s test, Zhanget al. 1997). I noted that the n/s ratio for DMRT1 after gene duplication (1.19) is similar to that before the duplication (31/31 = 1.00), and thus computed an average ratio [(21.7 + 31)/(18.3 + 31) = 1.07] for DMRT1. When this number is compared to the n/s ratio for DMY, I found that the former is significantly smaller than the latter (P = 0.02, Fisher’s test). These results suggest that the acceleration in nonsynonymous substitutions of DMY is due to the elevation of mutation rate and a change in natural selection on the gene. To further confirm this result, I performed a likelihood analysis (Yang 1998). The log likelihood value under the model of equal w (=dN/dS) among all branches of the tree in Figure 3a is -1929.68, significantly lower than that (-1926.38) under the model of one w for DMY and a different w for DMRT1 (P = 0.01, likelihood ratio test). This two-w model, however, is not significantly worse in fitting the data than a three-w model (log likelihood =-1926.27) that assigns one w for DMY, one for DMRT1 after gene duplication, and one for DMRT1 before duplication (P = 0.64, likelihood ratio test). These results are consistent with the above nonlikelihood-based analyses. Similar results are obtained when the two sequences from O. curvinotus are excluded, as shown in Figure 4a.
—Nucleotide substitutions in the evolution of DMY and DMRT1 of medaka fish for (a) the entire protein, (b) the DM domain, and (c) the non-DM regions. The DMY and DMRT1 genes from O. curvinotus are excluded from the analysis because of their uncertain phylogenetic positions (Figure 2). The numbers of synonymous (s) and nonsynonymous (n) substitutions for each tree branch are shown on branches. The sums of n for all branches involving DMY (after gene duplication), DMRT1 (after duplication), and DMRT1 (before duplication) are indicated by respective Σn values. Σs values are similarly obtained. N and S are potential numbers of nonsynonymous and synonymous sites of the sequences, respectively. For the entire protein, N = 523.0 and S = 212.0. For the DM domain, N = 143.7 and S = 54.3. For non-DM regions, N = 379.3 and S = 157.7.
Altered rates of nonsynonymous substitution in the DM domain: The DM domain (Figure 1) is the DNA-binding domain in DMRT1 and DMY. DNA-binding domains are usually the most conserved parts of transcriptional factors. For example, the DNA-binding homeodomain is conserved among 13 cognate Hox genes of vertebrates, some of which diverged over 600 MYA (Zhang and Nei 1996). DMRT1 appears to follow this rule as well. There are 0.049 nonsynonymous substitutions per nonsynonymous site in the DM domain on all the branches linking DMRT1, in comparison to 0.120 in the non-DM regions (Figure 3, b and c). The former is significantly smaller than the latter (P = 0.008, Fisher’s test). Surprisingly, for DMY, the opposite is observed. There are 0.180 nonsynonymous substitutions per nonsynonymous site in the DM domain of DMY, in comparison to 0.107 in the non-DM regions (Figure 3, b and c), the difference between them being statistically significant (P = 0.02, Fisher’s test). A direct comparison between DMY and DMRT1 can be made by a 2 × 2 table with the numbers of nonsynonymous substitutions in the DM domain of DMY (25.8), DM domain of DMRT1 (7), non-DM regions of DMY (40.6), and non-DM regions of DMRT1 (45.7), respectively. This comparison reveals a significant excess of nonsynonymous substitutions in the DM domain of DMY (P = 0.001, Fisher’s test). However, the dN/dS ratio for the DM domain of DMY is 0.95, not significantly different from 1 (P > 0.5, Fisher’s exact test). Thus, it is unclear whether the accelerated nonsynonymous substitution in the DM domain of DMY is due to a relaxation of functional constraints or positive Darwinian selection. To distinguish the two hypotheses, I used a likelihood method, which may be more powerful in detecting positive selection at individual sites (Yanget al. 2000). In this analysis, I compared the likelihoods for the data set of three DMY sequences under models 1 and 2. Here model 1 assumes that the dN/dS ratio for a given site is either 0 or 1, while model 2 adds an extra class of sites to model 1. The result showed that model 2 fits the data significantly better than model 1 (χ2 = 14.3, 2 d.f., P = 0.0008), with an additional class of dN/dS = 9.8. Three codons were identified to be under positive selection with posterior probabilities >95%. To examine whether this result is robust against possible violations of assumptions made in models 1 and 2, I conducted a second test by comparing models 7 and 8. Model 7 assumes that the dN/dS ratio follows a β-distribution between 0 and 1, while model 8 adds an extra class of sites to model 7. Model 8 was found to fit the data significantly better than model 7 (χ2 = 18.54, 2 d.f., P = 0.0001), with an additional class of dN/dS = 7.3. Here, the above three codons and two additional codons were found to be under positive selection with >95% posterior probability. Thus, the positive selection hypothesis is supported by likelihood. Although both tests favor the positive selection hypothesis, the results should be interpreted with caution as false detections of positive selection by likelihood have been reported recently (Suzuki and Nei 2001, 2002).
When the two sequences from O. curvinotus are excluded from the analysis (Figure 4), qualitatively similar results are obtained. In particular, the dN/dS ratio (3.25) for the DM domain of DMY is significantly >1 (P = 0.03, Fisher’s exact test). This finding, based on a more conservative test (Zhanget al. 1997), provides additional and strong evidence for the action of positive selection on the DM domain.
Characteristic amino acids of DMY: There were no synonymous substitutions in DMY after its origin from duplication but before the split of O. latipes and O. curvinotus (Figure 3), suggesting that the duplication occurred shortly before the species separation (Matsudaet al. 2003). Interestingly, there were three amino acid substitutions during this short period of time, all located in the DM domain (Figures 1 and 3). Because O. latipes and O. curvinotus have the same sex-determination mechanism, it may be inferred that one or more of the three substitutions are important to the establishment of DMY as a sex-determination gene in medaka. Among the three sites, only one (position 26 in O. latipes Carbio DMY) has remained invariant among orthologous DMY sequences. Thus, this site may be most critical to the function of DMY. This site is Thr in the three DMY sequences and Ser in the four DMRT1 sequences shown in Figure 1. I examined this site in all fish DMRT1 sequences available in GenBank and found that all of them have Ser. These sequences include green spotted pufferfish Tetraodon nigroviridis (AY152820), pufferfish Takifugu rubripes (AJ295039), platyfish Xiphophorus maculates (AF529187), tilapia Oreochromis niloticus (AF203489), trout Oncorhynchus mykiss (AF209095), halibut Hippoglossus hippoglossus (CAD44607), sturgeon Acipenser transmontanus (AY057061), cod Gadus morhua (CAD44608), zebrafish Danio rerio (AF080622), and wrasse Halichoeres tenuispinis (AAO18650). Furthermore, I examined this site in DMRT1 sequences of all nonfish vertebrates that are available in GenBank, including human Homo sapiens (NM_021951), mouse Mus musculus (NM_015826), rat Rattus rattus (AAK57706), pig Sus scrofa (AF216651), and chicken Gallus gallus (AAF19666), and found that all have Ser. Thus, the Ser at this position has been conserved among all vertebrate DMRT1 sequences so far examined, suggesting its functional importance. The substitution from Ser to Thr in DMY may therefore have significant functional consequences.
DISCUSSION
In this article, I analyzed the rate and pattern of nucleotide substitutions in DMY, a newly emergent sex-determination gene in medaka fish, and made several observations. First, there is evidence for an elevation of the mutation rate in the Y-linked DMY, compared to its autosomal mother gene DMRT1. This result is in support of the male-driven evolution hypothesis, with the estimated male/female mutation rate ratio αm being about ∼6–10. The male-driven evolution hypothesis has gained substantial support from evidence in mammals (Miyataet al. 1987; Shimminet al. 1993; Changet al. 1994; Nachman and Crowell 2000; Liet al. 2002; Makova and Li 2002; but see McVean and Hurst 1997; Bohossianet al. 2000) and some evidence in birds (Ellegren and Fridolfsson 1997; Kahn and Quinn 1999; Carmichaelet al. 2000). It has also gained support from a comparison of the sequence data of the Y-linked GH-2Y gene and autosomal GH-2 gene of salmonid fish (Ellegren and Fridolfsson 2003). Because the estimated αm values have large errors from either the salmon or the medaka data set, I combined the two data sets and estimated that αm = 7.7, with the 95% confidence interval of (1.36, ∞). This confidence interval was determined using computer simulation as follows. The observed total number of synonymous substitutions on the two Y-linked genes (DMY and GH-2Y) is 52.6 and the corresponding number in the homologous autosomal genes (DMRT1 and GH-2) is 29.7. Because nucleotide substitutions may be considered a Poisson process, the number of synonymous substitutions on the two Y-linked genes is a Poisson variable with a mean of 52.6 and the corresponding number for the autosomal genes is an independent Poisson variable with a mean of 29.7. I generated 10,000 pairs of Poisson random variables and then estimated the 95% confidence interval for αm. As discussed by Ellegren and Fridolfsson (2003), it is currently unknown in fish how many cell divisions per generation occur in the male and female germ lines, but males produce much more sperm than females produce eggs and it is probable that there are more rounds of cell divisions in the male germ line. However, it should be stressed that this conclusion on male-driven evolution in fishes is tentative, as it is derived from only two genes. There appears to be substantial variation in mutation rate across genomic regions and more data are necessary to confirm this finding.
Second, a higher dN/dS value (0.85–1.17, depending on whether O. curvinotus is used) is found in DMY, in comparison to that (0.42–0.48) in DMRT1. This elevation is particularly prominent in the DM domain (0.95–3.25 vs. 0.24–0.19). It is interesting to note that SRY, the mammalian sex-determination gene on the Y chromosome, is also known to have a high dN/dS ratio in primates and rodents (Tucker and Lundrigan 1993; Whitfieldet al. 1993; Pamilo and O’Neill 1997; Wanget al. 2002). However, the rapid evolution is limited to regions that are outside the DNA-binding HMG domain. It is still debatable whether the high dN/dS ratio in SRY is due to positive selection or relaxation of functional constraints, and SRY genes of different species may be under different forms of selection (Wanget al. 2002). Recent studies favor the view that Y-linked genes generally have elevated dN/dS in comparison to their homologous genes on X or autosomes (Sandstedt 2003; Tuckeret al. 2003). This may be due to several reasons, such as changes in gene function or expression and alleviated intensity of natural selection caused by lack of recombination or reduced effective population size for Y-linked genes. In this case of DMY, however, it is the DNA-binding DM domain that shows high dN/dS values. Both likelihood- and nonlikelihood-based methods suggest the action of positive selection. The selective agent, however, is not immediately clear. It is interesting to relate this finding to the observation that the DNA-binding homeodomains of several homeobox genes of mammals and Drosophila evolve rapidly by positive selection (Sutton and Wilkinson 1997; Tinget al. 1998; Wang and Zhang 2004) and all these homeobox genes are located in the X chromosome and expressed in the testis (Wang and Zhang 2004). The rapid evolution of these otherwise conserved DNA-binding (DM and homeobox) domains is intriguing. Their involvement in sexual differentiation and reproduction suggests the possibility that the Darwinian selection that acts on them is related to differential reproductive success and possibly speciation (Tinget al. 1998).
Third, I estimated that the duplication that generated the DMY gene took place ∼10 MYA. The split of O. latipes and O. curvinotus followed shortly (assuming the tree of Figure 2b). There is only one amino acid position that is conserved among DMY sequences but differs between DMY and DMRT1. It is possible that this single amino acid substitution (Ser26Thr) played a major role in the establishment of DMY as the primary sex-determination gene in medaka fish. In this context, it is interesting to note the high plasticity of sex-determination mechanisms in animals (Zarkower 2001). For example, single mutations in the tra-1 gene of the nematode Caenorhabditis elegans can change the sex-determination mechanism entirely (Hodgkin 1983) and strains of C. elegans with different sex-determination mechanisms have been constructed in the laboratory (Hodgkin 2002). Several placental mammals are known to lack the SRY gene (Justet al. 1995) or have multiple SRY genes (Lundrigan and Tucker 1997). In the future, it will be interesting to examine the functional effect of Ser26Thr in the DMY gene of medaka fish and test whether this single amino acid substitution was the key to the origin of a new sex-determination gene.
Among vertebrates, SRY and DMY are the only sex-determination genes so far identified. A number of features are surprisingly similar between them, including the existence of close paralogs in the genome, higher dN/dS ratios than those of their paralogs, possible actions of positive selection, and occasional disappearances of the genes in males (Justet al. 1995; Nandaet al. 2003). These features probably reflect the low functional and developmental constraints on the master controllers of sex-determination pathways and the easiness for natural selection to modify them. Studying additional species may reveal whether these observations are general to all master controllers of sex-determination pathways.
On a final note, the late geneticist Susumo Ohno pioneered evolutionary studies of sex chromosomes and gene duplication (Ohno 1967, 1970). He would be pleased to see the simple, yet extraordinary path to the emergence of the medaka sex chromosome and sex-determination mechanism, which vividly demonstrates the theory of evolution by gene duplication that he championed (reviewed in Zhang 2003).
Acknowledgments
I thank David Webb for valuable comments on the manuscript. This work was supported by a startup fund from the University of Michigan and a research grant from the National Institutes of Health (GM-67030).
Footnotes
-
Communicating editor: S. Yokoyama
- Received October 11, 2003.
- Accepted January 7, 2004.
- Copyright © 2004 by the Genetics Society of America
