1. Introduction
Despite their independent evolution, sex chromosomes in different organismal groups have similar properties (Bull, Reference Bull1983): recombination is restricted between X and Y chromosomes, and the male-specific non-recombining Y chromosome (or female-specific W chromosome in the case of female heterogamety) exhibits genetic degeneration via the loss of functional genes and the accumulation of repetitive DNA (reviewed in Charlesworth & Charlesworth, Reference Charlesworth and Charlesworth2000). In most mammals and Drosophila species Y chromosomes are small and heterochromatic, and contain only a few genes (reviewed in Lahn et al., Reference Lahn, Pearson and Jegalian2001). Even plant Y chromosomes, which are typically young (only few million years old), show signs of genetic degeneration (Guttman & Charlesworth, Reference Guttman and Charlesworth1998; Liu et al., Reference Liu, Moore, Ma, Ackerman, Ragiba, Yu, Pearl, Kim, Charlton, Stiles, Zee, Paterson and Ming2004; Filatov, Reference Filatov2005b).
Y chromosome degeneration is thought to occur due to ‘sheltering’ of the Y-linked genes with X-linked homologues and complete linkage of genes on the non-recombining Y chromosome. Lack of recombination exacerbates such population genetic processes as hitch-hiking (fixation of deleterious mutations due to linkage to favourable mutations spreading in the population; Rice, Reference Rice1987) and background selection (selection against deleterious mutations at linked genes, resulting in reduced effective population size and stochastic accumulation of mildly deleterious mutations; Charlesworth et al., Reference Charlesworth, Charlesworth and Morgan1995), leading to reduced effective population size of the Y-linked genes and accumulation of deleterious mutations by Muller's ratchet (Charlesworth & Charlesworth, Reference Charlesworth and Charlesworth1997). The loss of genetic diversity in the non-recombining regions is consistent with the results of the studies of DNA polymorphism in various organisms (reviewed in Charlesworth & Charlesworth, Reference Charlesworth and Charlesworth2000), including the white campion (Filatov et al., Reference Filatov, Moneger, Negrutiu and Charlesworth2000, Reference Filatov, Laporte, Vitte and Charlesworth2001; Ironside & Filatov, Reference Ironside and Filatov2005), which have found low levels of DNA diversity at Y-linked loci compared with X-linked and autosomal genes in most organisms studied.
The X chromosome (or Z in case of female heterogamety) continues to recombine and does not degenerate, which can compensate to some extent for the loss of functional genes from the Y (or W) chromosome. Dosage compensation has independently evolved in mammals and Drosophila to correct for the differences in the amount of product of sex-linked genes in males and females (reviewed by Straub & Becker, Reference Straub and Becker2007). Even the recently evolved neo-sex chromosomes of Drosophila miranda show signs of evolving dosage compensation (Martin et al., Reference Martin, Franke, Bashaw and Baker1996; Bone & Kuroda, Reference Bone and Kuroda1996). Once the Y-linked copy of a gene is lost, selection may favour an elevated expression level of the homologous X-linked genes in males, leading to partial dosage compensation. Thus, one may expect that advantageous mutations frequently spread in X-linked genes (a process termed ‘selective sweep’ by Maynard Smith & Haigh, Reference Maynard Smith and Haigh1974) once they are under additional selective pressure to compensate for degenerate Y-linked homologues. However, no cases of recent selective sweeps, or other evidence of positive selection in X-linked genes associated with the degeneration of Y-linked homologues, have been reported previously.
The plant genus Silene has evolved sex chromosomes in a small cluster of dioecious Silene species (section Elisanthe: S. latifolia, S. dioica, S. diclinis, S. heuffelii and S. marizii), while the rest of the genus is non-dioecious (exept S. otites, which apparently evolved dioecy independently from Elisanthe) and lacks sex chromosomes. Silent divergence between dioecious S. latifolia and hermaphroditic S. conica is approximately 15%, suggesting that S. latifolia sex chromosomes are probably not older than 15 million years (Filatov & Charlesworth, Reference Filatov and Charlesworth2002) and most likely evolved from a pair of autosomes (Filatov, Reference Filatov2005a). As YY plants (having no X) are usually inviable (Ye et al., Reference Ye, Installe, Ciupercescu, Veuskens, Wu, Salesses, Jacobs and Negrutiu1990), the S. latifolia Y chromosome has probably degenerated to some extent. However, most genes isolated from S. latifolia X chromosome have intact Y-linked homologues, suggesting that genetic degeneration has not proceeded too far and most Y-linked genes are still intact. The first gene isolated from the S. latifolia X chromosome, MROS3X, was reported to have a dysfunctional Y-linked copy (Guttman & Charlesworth, Reference Guttman and Charlesworth1998). However, more recently it was demonstrated that MROS3X/Y genes are members of a multicopy gene family with several copies on the autosomes and sex chromosomes (Kejnovsky et al., Reference Kejnovsky, Vrana, Matsunaga, Soucek, Siroky, Dolezel and Vyskot2001). Thus, the Y-linked degenerate copy of MROS3 is not necessarily the remnant of the Y-linked homologue of MROS3X, but it may be a paralogue translocated from the autosomes. High MROS3X/Y divergence (>30%) also supports this view. All the X-linked genes isolated more recently do have intact Y-linked homologues (Delichère et al., Reference Delichère, Veuskens, Hernould, Baarbacar, Mouras, Negrutiu and Moneger1999; Atanassov et al., Reference Atanassov, Delichère, Filatov, Charlesworth, Negrutiu and Monéger2001; Moore et al., Reference Moore, Kozyreva, Lebel-Hardenack, Siroky, Hobza, Vyskot and Grant2003; Filatov, Reference Filatov2005b; Bergero et al., Reference Bergero, Forrest, Kamau and Charlesworth2007). Furthermore, all these Y-linked genes except SlssY show no signs of genetic degeneration.
The pair of sex-linked genes that do show signs of degeneration in the Y-linked copy, the SlssX and SlssY genes, encode proteins with a strong similarity to spermidine synthases and silent divergence between these genes is 8% (Filatov, Reference Filatov2005b). Although the Y-linked copy of this gene is transcribed, it contains several mutations that change conserved amino acid positions known to be important for spermidine synthase activity and are expected to affect functionality of the protein (Filatov, Reference Filatov2005b). The X-linked copy of this gene is intact and the loss of function of the Y-linked copy might have imposed an additional burden on the X-linked copy to produce enough protein in males to ‘shelter’ the loss of function of the Y-linked copy.
The previous work focused on genetic mapping of the X-linked genes (Filatov, Reference Filatov2005a) revealed unusually low diversity in the SlssX gene, suggesting a recent selective sweep in this gene. This study aims to test the hypothesis of the sweep in the SlssX gene comparing patterns of DNA polymorphism in the SlssX gene and other X-linked genes: SlX1, SlX4 and DD44X. The first two genes are located relatively far away from the SlssX (7·4 and 17·9 cM, respectively), while DD44X is very closely (<1 cM) linked to SlssX (Filatov, Reference Filatov2005a). As previous analyses of polymorphism in SlX1, SlX4 and DD44X genes revealed no indication that these genes may have been affected by a recent selective sweep (Filatov et al., Reference Filatov, Moneger, Negrutiu and Charlesworth2000; Laporte et al., Reference Laporte, Filatov, Kamau and Charlesworth2005; Ironside & Filatov, Reference Ironside and Filatov2005), they can be used as useful reference loci to be compared with SlssX.
2. Materials and methods
To compare amounts and patterns of DNA polymorphism in S. latifolia X-linked genes, I used a set of S. latifolia individuals collected from across Eurasia (Table 1). Sampling was designed to cover the entire range of the species. A small quantity of leaf material was taken from each plant and frozen. DNA was extracted using the magnetic beads-based protocol of the ChargeSwitch Plant DNA kit (Invitrogen). Fragments of DD44X and SlssX genes were amplified from the genomic DNA of male plants (sequences for all PCR and sequencing primers are provided in Table 2). The DD44X was amplified using the primers DD44X+1 and DD44XY2.1R. The SlssX gene was amplified with the primers Slss+6X and Slss-18X. Other primers listed in Table 2 were used for sequencing of internal regions of the PCR products amplified from DD44X and SlssX genes. Amplifications were performed under the following PCR conditions: initial denaturation at 94°C for 2 min followed by 38 cycles of 93°C for 20 s, 53°C for 30 s, and 68°C for 4 min. Expand high-fidelity DNA polymerase (Roche) was used for both DD44X and SlssX genes. Because males possess only a single X chromosome, DD44X and SlssX PCR products amplified from male genomic DNA were hemizygous and were sequenced directly using a BigDye v3.1 sequencing kit and a 3700 automated sequencer (Applied Biosystems). All polymorphic sites were double-checked manually. Chromatograms were checked and corrected, contigs assembled and alignments constructed using ProSeq3 software (Filatov, Reference Filatov2002). The sequences have been submitted to GenBank under accession numbers EU265839–EU265926.
Table 1. The list of Silene samples used in this study
Plus and minus signs indicate whether the SlssX and DD44X genes were sequenced for a particular sample.
Table 2. PCR and sequencing primers
The program ProSeq3 was used to calculate two measures of nucleotide diversity (π (Nei, Reference Nei1987) and Watterson's (Reference Watterson1975) θw) for DD44X and SlssX. The sequence diversity of SlssX was compared with that of DD44X and two other X-linked genes using the HKA test (Hudson et al., Reference Hudson, Kreitman and Aguade1987), implemented in DnaSP (Rozas et al., Reference Rozas, Sanchez-DelBarrio, Messeguer and Rozas2003). A single DD44X and SlssX sequence from the hermaphroditic species S. vulgaris was used as an outgroup. Bias of the site-frequency distribution in the SlssX and DD44X sequences was tested using Tajima's test (Tajima, Reference Tajima1989) as implemented in ProSeq3. The excess of high-frequency derived polymorphisms was estimated using Fay & Wu's (Reference Fay and Wu2000)H statistic implemented in DnaSP. The deviation of the number of distinct haplotypes in the sample from neutral expectation was tested using Fu's (Reference Fu1997) F s statistic implemented in DnaSP. The significance of all the neutrality tests was calculated using coalescent simulations without recombination, which is a conservative approach. The coalescent simulations were conducted using DnaSP and ms (Hudson, Reference Hudson2002) programs. The population-scaled recombination rate, ρ, was estimated using the composite likelihood method of McVean et al. (Reference McVean, Awadalla and Fearnhead2002). The minimum number of recombination events, R m (Hudson & Kaplan, Reference Hudson and Kaplan1985) was conducted using DnaSP.
Estimation of the time since the selective sweep was estimated using the approach proposed by Przeworski (Reference Przeworski2003). A program msHH implementing this approach was kindly provided by M. Przeworski. This program was used to generate samples from the posterior distribution of the time since the selective sweep, as well as the selective coefficient for the mutation that caused the sweep. The location of the target of positive selection, as well as the selective coefficient of the adaptive mutation, was estimated using the composite likelihood method of Kim & Stephan (Reference Kim and Stephan2002). The program clsw (Kim, Reference Kim2005) was used to find the maximum composite likelihood estimates of the population-scaled selective coefficient and the position of the beneficial mutation as well as the likelihood ratio of the sweep versus neutrality models.
3. Results
(i) DNA polymorphism
Sequencing of a 3·5 kb long fragment of the SlssX gene from 42 individuals revealed only 39 nucleotide polymorphisms (Table 3), which corresponds to a per-nucleotide population-scaled mutation rate of θw=0·0029 (all sites) when measured using Watterson's (Reference Watterson1975) estimator. This level of polymorphism is surprisingly low compared with the previously studied X-linked genes SlX1, SlX4 and DD44X (θw=0·02, 0·024 and 0·026, respectively), and is similar to the level of polymorphism observed in S. latifolia Y-linked genes, which is significantly reduced compared with polymorphism in X-linked genes (Filatov et al., Reference Filatov, Moneger, Negrutiu and Charlesworth2000, Reference Filatov, Laporte, Vitte and Charlesworth2001; Laporte et al., Reference Laporte, Filatov, Kamau and Charlesworth2005; Ironside & Filatov, Reference Ironside and Filatov2005).
Table 3. DNA polymorphism in four S. latifolia X-linked genes
a Number of individuals analysed.
b Number of alignment positions analysed.
c Observed number of polymorphic sites.
d Average heterozygosity per nucleotide (Nei, Reference Nei1987).
e Fu's (Reference Fu1997) Fs statistic.
f Minimal number of recombination events (Hudson & Kaplan, Reference Hudson and Kaplan1985).
g Population-scaled recombination rate (Hudson, Reference Hudson2001).
h Jukes–Cantor (Jukes & Cantor, Reference Jukes, Cantor and Munro1969) divergence from outgroup S. vulgaris.
Estimates of DNA polymorphism in a gene may be affected by sampling, local mutation rate, demography (of the populations), and non-neutral evolution of the gene or in the flanking regions nearby. The sampling used in the current study to measure DNA polymorphism in the SlssX gene differed from the previous studies (e.g. see Table 1 in this study and in Ironside & Filatov, Reference Ironside and Filatov2005). Thus comparisons of SlssX data with other genes have to be treated with some caution. To exclude the effects of sampling differences on estimates of DNA polymorphism, I sequenced a 3·5 kb long region from the DD44X gene using the same samples as for the SlssX gene. Regions containing polymorphic insertions/deletions (indels) were excluded from the analyses, which reduced the number of alignment positions that can be analysed for DNA polymorphism to 1104 (Table 3). The observed per-nucleotide average heterozygosity in the new DD44X dataset, π=0·028 was similar to that observed previously (π=0·023; Ironside & Filatov, Reference Ironside and Filatov2005). Thus, the sampling difference between this and the previous studies is unlikely to be the cause of the unusually low DNA diversity in the SlssX gene.
Any difference in mutation rates between the SlssX and DD44X genes may also be a cause of variation in the level of polymorphism among the genes. If the SlssX gene is a mutational cold-spot, then one would expect to observe less polymorphism in that gene. Differences in mutation rate between different genomic regions can be corrected for by comparing homologous regions with an outgroup (in this case the non-dioecious species S. vulgaris). Silent mutations should be predominantly neutral, so the rate of divergence between species will be proportional to the mutation rate. When the diversity differences between SlssX and DD44X were corrected using the divergence from homologues in S. vulgaris, the difference remained significant (as did comparisons for three other X-linked genes; Table 4). The comparison of DNA diversity in the SlssX and DD44X genes with two other previously studied S.latifolia X-linked genes demonstrates that DD44X contains similar amounts of polymorphism as the SlX1 and SlX4 genes, while there is a significant lack of diversity in the SlssX gene, compared with the three other X-linked genes (Table 4).
Table 4. Pairwise HKA test results for four X-linked genes from S. latifolia. S. vulgaris was used as an outgroup for all pairwise HKA tests
a Data from Filatov et al. (Reference Filatov, Laporte, Vitte and Charlesworth2001).
b Data from Laporte et al. (Reference Laporte, Filatov, Kamau and Charlesworth2005).
c Data from this study.
(ii) Recombination rates
Although the physical distance between the SlssX and DD44X genes is not known, it has been demonstrated that these genes are tightly linked on the X chromosome – no recombinants between these genes have been observed previously (Filatov, Reference Filatov2005a). However, from the linkage disequilibrium pattern, the region containing SlssX and DD44X genes is not a recombinational cold-spot. Both genes contain pairs of polymorphisms with all four combinations of alleles, which could have ocurred either due to multiple mutations at the same site, or due to recombination. The minimal number of recombination events, estimated using the four-gamete test of Hudson & Kaplan (Reference Hudson and Kaplan1985) is 8 and 23 in the SlssX and DD44X genes, respectively. In accordance with the view that the DD44X–SlssX region is not a recombinational cold-spot, there is no significant linkage disequilibrium between the two genes despite their tight genetic linkage.
Although it is clear that recombination does occur in the DD44X–SlssX region, the population-scaled recombination rate per nucleotide, ρ (=3N er for X-linked genes, where r is the recombination rate and N e is the effective population size), estimated using the composite likelihood approach (Hudson, Reference Hudson2001; McVean et al., Reference McVean, Awadalla and Fearnhead2002), is more than an order of magnitude lower for SlssX compared with DD44X (Table 3). The population recombination rate, however, depends on the effective population size (which in turn is scaled to polymorphism). Hence, the lower ρ in SlssX may be due to a lower effective population size in the region and not a lower recombination rate. Normalization of ρ by the population-scaled mutation rate, θ (=3N eμ, where μ is the mutation rate), provides a way to correct for differences in the amounts of polymorphism in different genes. Using this correction, the difference between the SlssX and DD44X genes remains (0·092 and 0·746, respectively). Thus, although the four-gamete test (Hudson & Kaplan, Reference Hudson and Kaplan1985) and the composite-likelihood estimator of ρ detect recombination in the SlssX gene, recombination rate at this locus is probably lower than in DD44X. Nevertheless, if genetic diversity in the SlssX gene were affected by a recent selective sweep (see below), then linkage disequilibrium may not reflect the true recombination rate in this gene, as selective sweeps can generate fairly strong linkage disequilibrium (Stephan et al., 2006; McVean, Reference McVean2007), resulting in an underestimation of the local recombination rate.
(iii) Frequency-spectrum-based neutrality tests
The significantly reduced DNA diversity in the SlssX gene, compared with the other three X-linked genes, suggests that diversity may have been reduced by a selective sweep in or near the SlssX gene. A selective sweep is expected to bias the frequency spectrum of mutations in regions adjacent to the selected mutation (Braverman et al., Reference Braverman, Hudson, Kaplan, Langley and Stephan1995), as the spread of the adaptive allele should wipe out genetic diversity and all new mutations that arise after the sweep will be at low frequency. Indeed, there is a significant excess of mutations present only once in the sample (singletons) in SlssX, but not in DD44X. Tajima's D (Tajima, Reference Tajima1989), a neutrality test devised to detect this kind of bias in the frequency spectrum, is significant for SlssX, but not for DD44X (Table 3). Although the bias in the frequency spectrum can be caused by demographic factors, such as population expansion, demography cannot specifically affect SlssX alone.
The spread of an advantageous allele in a population is expected to drag some closely linked polymorphisms to high frequency, which is a characteristic footprint of positive selection (Fay & Wu, Reference Fay and Wu2000; Przeworski, Reference Przeworski2002). Using the S. vulgaris Slss gene sequence as an outgroup to establish the ancestral state of the polymorphic sites, nine of 42 polymorphisms in the S. latifolia SlssX dataset are high-frequency derived polymorphisms. Fay & Wu's (Reference Fay and Wu2000)H statistic demonstrates that there is a significant excess of such polymorphisms in the SlssX (H=−10·91, P<0·01), but not in DD44X (H=1·33, NS).
(iv) Timing of the selective sweep in SlssX gene
The significantly reduced DNA diversity in the SlssX gene compared with other X-linked genes in S. latifolia and the excess of high- and low-frequency polymorphisms in the frequency spectrum of the SlssX gene strongly suggest a recent selective sweep in or near the SlssX gene. The recovery of genetic diversity after a selective sweep takes about 4N e generations. The bias in frequency spectrum by Tajimas' D statistic is almost undetectable after approximately 2N e generations, while the excess of high-frequency polymorphisms detectable by Fay & Wu's H statistic disappears after 0·5–1N e generations (Przeworski, Reference Przeworski2002). As the signature of the sweep is still detectable by both D and H statistics the sweep might have occurred less than 1N e generations ago.
To estimate the time of the sweep more precisely, I used the approach proposed by Przeworski (Reference Przeworski2003). This method summarizes polymorphism data using three summary statistics: the number of polymorphic sites, the value of Tajimas' D and the number of observed haplotypes. Given the observed values of these statistics, the method produces a sample from the posterior distribution of the time since the selective sweep, as well as the selective coefficient for the mutation that caused the sweep (s). The sample of size 2000 from the posterior distribution of the number of generations since the selective sweep (T gen) forms a relatively narrow peak centred at 30 000 generations (Fig. 1A). On the other hand, the sample from the posterior distribution of the selective coefficient forms a fairly wide peak between 0 and 0·05, with the maximum at 0·005 (Fig. 1B), demonstrating that the selective sweep might have been caused by a beneficial mutation with advantage of less than 5%; however, there is not enough power to obtain a more precise estimate of the strength of selection using this method. These estimates are robust to the choice of the prior values for population size (data not shown), and to the choice of the position of the adaptive mutation in the sequence (K; see Fig. 1).
Fig. 1. Estimation of time since the selective sweep and the strength of selection using the method of Przeworski (Reference Przeworski2003). Shown are 2000 samples from the posterior distributions of the time since the selective sweep (A) and the selective coefficient for the mutation that caused the sweep (B). The four series correspond to different positions of the selected mutation (K) in the SlssX at the beginning of the gene, 100 bp, 1000 bp and 5000 bp upstream of the gene.
To compare the posterior distributions of T gen for different S. latifolia X-linked genes, I applied the same analysis to the DD44X, SlX1 and SlX4 datasets. The distributions of T gen for these genes are fairly wide and are centred at around 650 000 generations, while the distribution for the SlssX gene is in sharp contrast with the three other X-linked genes (Fig. 2). The wide peaks of the DD44X, SlX1 and SlX4 genes may reflect an ancient demographic event, such as population expansion (e.g. Taylor & Keller, Reference Taylor and Keller2007), as they give very similar estimates of T gen for the three genes, while the narrow peak in the SlssX gene corresponds to a gene-specific event, providing further support for the recent selective sweep in SlssX.
Fig. 2. Samples from posterior distributions of the time since the selective sweep or population expansion for SlssX, DD44X, SlX1 and SlX4 S. latifolia datasets.
(v) Selective coefficient and location of the target of the sweep
A selective sweep is expected to influence a neutral linked locus if recombination between the selected and neutral loci is less than half the selection coefficient, s/2 (Maynard Smith & Haigh, Reference Maynard Smith and Haigh1974; Stephan et al., 1992). As the DD44X and SlssX genes are very closely linked (<1 cM; Filatov, Reference Filatov2005a), and the sweep in or near the SlssX gene apparently does not affect the DD44X gene, the selective coefficient of the positively selected mutation might have been relatively low (probably less than 1–2%). To estimate the strength of selection in or near the SlssX more precisely and to predict the location of the target of the sweep, I used the composite likelihood method of Kim & Stephan (Reference Kim and Stephan2002). This method compares the observed patterns of nucleotide variation under neutrality and the selective sweep models and allows for a composite likelihood ratio test (LR) of the selective sweep scenario. It also estimates the most likely position of the target of positive selection as well as the population-scaled selective coefficient (1·5N es for an X-linked gene and 2N es for an autosomal gene).
The results of this analysis for the SlssX gene are presented in Table 5. As was noted above, the estimate of recombination rate in this gene could have been biased by the selective sweep, thus the analysis was conducted with a range of different values of the per nucleotide population-scaled recombination rates (ρ=3N er). Regardless of the recombination rate used or whether data with folded or unfolded frequency spectra (i.e. not distinguishing or distinguishing ancestral and derived alleles, respectively) were employed, the composite likelihood ratio tests reject a neutral model in favour of the selective sweep model. The estimate of the strength of selection that might have caused the sweep depends on the amount of recombination assumed in this analysis. With the ‘observed’ value of ρ=0·0006 (estimated for the SlssX gene using the composite likelihood method of McVean et al., Reference McVean, Awadalla and Fearnhead2002), the population-scaled selective coefficient (1·5N es) is about 5. Assuming the effective population size of S. latifolia is ~106 (e.g. Taylor & Keller, Reference Taylor and Keller2007), the selective coefficient of the mutation that has caused the selective sweep in the SlssX gene is of the order of 3×10−6, which is much lower than the estimate of the selective coefficient obtained by the method of Przeworski (Reference Przeworski2003), which yielded a posterior distribution with the maximum at s=0·005. Higher recombination rates assumed for SlssX give higher estimates of 1·5N es (Table 5). Given the difficulty of obtaining a reliable estimate of recombination rate in the SlssX gene and the width of the peak for s (Fig. 1B) obtained by the method of Przeworski (Reference Przeworski2003), it is not possible to conclude that the two methods produce estimates that are significantly different from each other. Both methods agree that the selective advantage of the mutation that might have caused the selective sweep in the SlssX gene is lower than 1% (and probably much lower).
Table 5. Composite likelihood ratio analysis of positive selection in SlssX gene using the method of Kim & Stephan (Reference Kim and Stephan2002)
a Population-scaled recombination rate (Hudson, Reference Hudson2001).
b Likelihood ratio test comparing nested models with and without positive selection.
c Probability of rejecting neutrality in the likelihood ratio test.
d Estimated population-scaled selective coefficient.
e Inferred position of the target of positive selection.
The inferred position of the target of positive selection varies only slightly with the different frequencies of recombination assumed (Table 5). In most cases the target of selection is located between nucleotide position 2400 and 2850 of the dataset, with one case at position 3466 when a very low recombination rate was assumed. The composite likelihood profile for different candidate positions of the target of positive selection is shown in Fig. 3. This profile was calculated for unfolded frequency spectrum data assuming the ‘observed’ ρ=0·0006. Profiles for other recombination rates and for the folded frequency spectra are similar (data not shown). With this profile the highest composite likelihood values are obtained for a selection target around nucleotide position 2700 of the dataset, which corresponds to exon 7 of the SlssX gene (exon numbering according to Filatov, Reference Filatov2005b). However, inspection of polymorphisms in exon 7 revealed only a silent singleton at position 2767 of the dataset and no fixed differences between S. latifolia and S. vulgaris. Thus, although there is only one significant (P<0·05) major peak of the likelihood ratio identified by the Kim & Stephan (Reference Kim and Stephan2002) approach, it does not correspond to any apparent amino acid replacement that could have been a target of positive selection. Although the second highest peak at around nucleotide position 3500 is not significant, it may suggest that the target of selection lies downstream of the studied region of the SlssX gene. Estimates of the selective coefficient (1·5N es) that might have caused the selective sweep in or near the SlssX gene depend on the recombination rate assumed for the SlssX gene and vary between 794 and 5 (Table 5).
Fig. 3. Composite likelihood ratio profile for S. latifolia SlssX gene calculated by the method of Kim & Stephan (Reference Kim and Stephan2002). Unbroken horizontal line shows the 95% percentile for the distribution of composite likelihood ratio under the neutral model.
4. Discussion
In this paper I argue that Silene latifolia X-linked gene SlssX has undergone a selective sweep in the recent past. Although the causes of this sweep remain unknown, the loss of function of the Y-linked homologue, SlssY (Filatov, Reference Filatov2005b), suggests that the X-linked copy may be under selection to compensate for a dysfunctional copy on the Y chromosome. As the Y-linked copy is still transcribed and does not contain any stop codons and frame shifts, dysfunctionalization of the SlssY protein might have occurred very recently. This might have imposed an additional burden on the X-linked SlssX copy to produce enough protein in males to ‘shelter’ the loss of function of the Y-linked copy. Although this is by no means direct evidence for selection for dosage compensation, the finding of a selective sweep in an X-linked gene with a degenerate Y-linked copy may be suggestive of selection to compensate for the loss of function of the Y-linked gene.
The spread of an adaptive allele in a population is expected to result in a loss of genetic variation at linked sites around the selected locus, as the fixation of the advantageous allele will result in the fixation of neutral linked variants (Maynard Smith & Haigh, Reference Maynard Smith and Haigh1974). In the presence of recombination, not all sites on the chromosome are completely linked. These sites may recombine in or out of the background of the spreading advantageous allele. As a result, sites that are further away (in terms of recombination distance) from the selected site will be less affected by the selective sweep. On average, this action results in the formation of a ‘V-shaped valley’ of polymorphism centred on the selected locus (e.g. Kim & Stephan, Reference Kim and Stephan2000). The width of such a valley depends on how quickly the sweep occurs (i.e. on the magnitude of the selective coefficient), as well as on the recombination rate in the region. The new mutations ocurring in the region of reduced polymorphism after the selective sweep are initially present in the population at low frequency, hence the biased frequency spectrum is another characteristic ‘signature’ of the selective sweep. In this paper I reported reduced DNA diversity and biased frequency spectrum in the S. latifolia X-linked gene SlssX, which contrasts with other X-linked genes, including the closely linked DD44X. The observed pattern of polymorphism is strongly suggestive of a relatively recent selective sweep in or near the SlssX gene. Other explanations, such as population expansion that can also cause bias in the frequency spectrum, seem unlikely given that the low polymorphism and excess of low-frequency polymorphisms are specific for the SlssX gene. Moreover, the significant excess of derived high-frequency polymorphisms in SlssX detected by Fay & Wu's (Reference Fay and Wu2000)H statistic is fairly specific to selective sweeps and not expected under a population expansion scenario. Sometimes population structure may create a significant Fay & Wu's H (Przeworski, Reference Przeworski2002); however, multiple genes are expected to be affected, while H is significant only for SlssX.
As the most likely region of the target of selection identified by the method of Kim & Stephan (Reference Kim and Stephan2002) does not contain any amino acid replacements that could be obvious targets of positive selection, the selective sweep could have been driven by a mutation beyond the studied region. It could be a mutation in an unknown neighbour gene that is closer to SlssX than DD44X. However, given fairly low gene density in plant genomes, a more likely possibility is a mutaion in SlssX or in regulatory sequences in introns or untranslated regions that may affect the level of SlssX expression. For example, local chromatin structure may significantly affect gene expression and dosage compensation in mammals and Drosophila is known to be acompanied by changes in chromatin structure (Straub & Becker, Reference Straub and Becker2007). The relationships between the specific sequence motifs and chromatin structure are not well understood, so it is difficult to infer what changes in or near the SlssX gene could have been advantageous, if this involved changes in local chromatin structure.
Interestingly, polymorphism in the DD44X gene closely linked to the SlssX apparently is not affected by the sweep in or near SlssX. A selective sweep is expected to influence a neutral linked locus if recombination between the selected and neutral loci is less than half the selection coefficient, s/2 (Maynard Smith & Haigh, Reference Maynard Smith and Haigh1974; Stephan et al., 1992). No recombinants were observed in genetic crosses between these two genes and recombinational distance between SlssX and DD44X should be less than 1 cM (Filatov, Reference Filatov2005a). Thus, any sweep in SlssX that is driven by a mutation with selective advantage of 1–2% or more should also affect DD44X. In fact, the selective advantage of the mutation in or near the SlssX gene might have been lower than 1%, as the methods of Kim & Stephan (Reference Kim and Stephan2002) and Przeworski (Reference Przeworski2003) provide estimates of about 0·0003% and 0·5%, respectively.
The discovery of a selective sweep in the SlssX may be a mere coincidence with the fact that this gene may be evolving dosage compensation; nevertheless, it provides a strong incentive to conduct a further study of relative expression levels of the SlssX and the spermidine synthase activity in males and females. If expression studies confirm that this gene is evolving dosage compensation, then the evidence for a recent selective sweep in the SlssX will be the first report of evolving dosage compensation ‘caught in action’.
I thank Graham Muir for correction of the manuscript. Barbara Mable and two anonymous reviewers provided useful suggestions that helped to improve the manuscript. This work was supported by a grant to D. F. from NERC.
