A reciprocal translocation radically reshapes sex-linked inheritance in the common frog

Abstract

X and Y chromosomes can diverge when rearrangements block recombination between them. Here we present the first genomic view of a reciprocal translocation that causes two physically unconnected pairs of chromosomes to be coinherited as sex chromosomes. In a population of the common frog (Rana temporaria), both pairs of X and Y chromosomes show extensive sequence differentiation, but not degeneration of the Ys. A new method based on gene trees shows both chromosomes are sex-linked. Further, the gene trees from the two Y chromosomes have identical topologies, showing they have been coinherited since the reciprocal translocation occurred. Reciprocal translocations can thus reshape sex linkage on a much grander scale than do inversions, the type of rearrangement that is much better known in sex chromosome evolution, and they can greatly amplify the power of sexually antagonistic selection to drive genomic rearrangement. Two other populations show evidence of yet other rearrangements, suggesting that this species has unprecedented structural polymorphism in its sex chromosomes.

Introduction

Sex chromosomes occupy a unique niche in the genome: their sequences can become specialized to enhance either female fitness or male fitness. Specialization requires the evolution of sex linkage, which partly or completely blocks recombination between the X and Y (or Z and W) chromosomes, allowing these sex chromosomes to diverge genetically. The mechanism that is most frequently hypothesized to cause these changes in linkage is a chromosome rearrangement that becomes fixed on the X or Y (Bergero & Charlesworth 2008; Pennell et al. 2015; Charlesworth 2017).

Inversions are the best-known type of sex chromosome rearrangement. If an inversion occurs on either the X or Y that spans the boundary between the pseudoautosomal region (or PAR, where recombination between X and Y occurs) and the sex-determining region (or SDR, where X-Y recombination does not occur), then the SDR is enlarged. This outcome has been observed in many taxa (Bachtrog et al. 2014), including the human lineage (Lahn & Page 1999). A second type of rearrangement leading to sex linkage is a fusion between a sex chromosome and an autosome. This situation is cytologically conspicuous because it causes one sex to have an odd number and the other sex an even number of chromosomes. A recent review found 166 examples of sex chromosome-autosome fusions in the fishes and squamate reptiles (Pennell et al. 2015). Fusions have the potential to affect the sex linkage of a larger number of genes than do inversions on an existing sex chromosome: a fusion immediately causes all of the genes of the former autosome to become partly or completely sex-linked. This effect is most dramatic in species in which recombination is greatly or completely suppressed in one sex, since the entire fused chromosome then becomes fully sex-linked. This can lead to extensive degeneration of the neo-Y chromosomes, as seen in several Drosophila species (Bachtrog 2008; Satomura & Tamura 2016; Mahajan et al. 2018).

The third and least understood type of sex chromosome rearrangement is a reciprocal translocation. Only a handful of examples have been reported: in several flowering plants (Howell et al. 2009; Wiens & Barlow 1975; Grabowska-Joachimiak et al. 2015; Grabowska-Joachimiak et al. 2011), a termite (Luykx & Syren 1981), two frogs (Ryuzaki et al. 1999; Yuan et al. 2018), and — most famously — the platypus (Grützner et al. 2004; Veyrunes et al. 2008). This type of translocation may often go undetected as it is only cytologically visible during pairing in meiosis.

A second reason why reciprocal translocations involving sex chromosomes may rarely be seen is because they can be highly deleterious (White 1973). As shown in Figure 1, this type of translocation produces a pair of physically unconnected Y chromosomes and a pair of unconnected Xs. If the four chromosomes assort independently into two pairs, many gametes are produced that do not have both Xs or both Ys. These are aneuploids that have duplications and deletions of large chromosome segments, which will often cause the gamete or zygote to die. On the other hand, the two Xs may naturally segregate together in meiosis, and likewise the two Ys, in what is termed the “alternate” pattern of segregation (Griffiths 2000). Alternate segregation thus avoids the fitness cost resulting from aneuploidy.

Figure 1. The meiotic pairing of translocated sex chromosomes.

Top: Schematic of a reciprocal translocation between an arm of the Y chromosome and an arm of an autosome. The male-determining locus is shown by the diamond. Left and right arms are denoted with subscripts L and R. Bottom: The four resultant chromosomes form a quadrivalent during meiosis. The two translocated chromosomes are cotransmitted as Y chromosomes. The ancestral X (X_Anc) and the untranslocated autosome (X_Neo) are cotransmitted as X chromosomes. The greater the distance between a site and the sex-determining region, the greater the probability that recombination occurs between it and the SDR. This results in a pattern of decreasing differentiation between the Xs and the Ys with greater distance from the SDR, shown schematically by the numbers between the homologues.

In either event – selection against aneuploid gametes, or alternate segregation – the result is that the two unlinked Xs are transmitted together, as are the two unlinked Ys. To date, no reciprocal translocation involving a sex chromosome has been studied genomically and so their consequences for sex-linked inheritance and sequence evolution are not known.

Sexually antagonistic selection (or SAS) is thought to be a key force driving the evolution of sex linkage (Charlesworth & Charlesworth 1980; Bull 1983; Otto et al. 2011). The strength of selection favoring a new rearrangement is proportional to how much the recombination between the locus under SAS and the SDR is decreased by the rearrangement. Inversions that capture loci in the pseudoautosomal region are expected to have a relatively modest effect on recombination because the loci involved are already partially sex-linked. In contrast, reciprocal translocations between sex chromosomes and autosomes may have a much greater potential to enable SAS to reshape the genome. All the loci on the autosome, which were formerly unlinked to sex determination, are now partially or fully sex linked. This can dramatically decrease their recombination with the SDR, maximizing the capacity of SAS to establish the translocation. Further, because a reciprocal translocation affects so many genes, the number of potential targets of SAS that can act as drivers is much larger than with inversions. These effects will be magnified when the heterogametic sex (males in XY species) has very low recombination across the genome, as seen in a number of taxa (Sardell & Kirkpatrick 2018).

The common frog, Rana temporaria, offers a compelling opportunity to study the genomic consequences of sex chromosome reciprocal translocations. A linkage map based on microsatellites suggests that the population in Ammarnäs (Sweden) may have a reciprocal translocation between Chromosome 1, which is the ancestral sex chromosome, and Chromosome 2, which is typically an autosome (Rodrigues et al. 2016). The evidence is that both chromosomes show sex linkage, yet the total number of chromosomes is unchanged, which suggests there is not a fusion. If the reciprocal translocation hypothesis is correct, it would result in the unusual situation shown in Figure 1: the cosegregation of two unlinked X chromosomes and two unlinked Y chromosomes. In contrast, the linkage map for another Swedish population, Tvedöra, shows that Chromosome 1 is again sex linked, but Chromosome 2 is not.

As in most anurans, crossing over in males is restricted to the very ends of the chromosomes, so recombination rates are extremely low along almost the entire lengths of all chromosomes (Cano et al. 2011; Guerrero et al. 2012; Brelsford et al. 2016; Jeffries et al. 2018). Thus the reciprocal translocation in Ammarnäs may have immediately caused most loci on Chromosome 2 to become very tightly sex linked. In this regard, the putative translocation in R. temporaria may have similar evolutionary consequences as the fusions between sex chromosomes and autosomes in the Drosophila species discussed earlier.

The situation in R. temporaria poses several questions. Do genomic data confirm the hypothesis of a reciprocal translocation? If so, is there evidence that the two unlinked pairs of chromosomes have been coinherited for a substantial period of evolutionary time? How has the new sex linkage sculpted patterns of molecular variation on the transposed Chromosome 2? Do either or both Y chromosomes show evidence of degeneration, as seen in Drosophila Ys?

To tackle these questions, we analyzed patterns of molecular variation in R. temporaria sampled from three populations in Fennoscandia (Figure 2). The very large size of its genome (4.5Gb; Gregory 2018)) and the lack of a reference genome preclude population-scale analyses using whole genome sequencing. We therefore devised a strategy based on sequencing the transcriptome. The results confirm there is a reciprocal translocation between Chromosomes 1 and 2 in Ammarnäs. Both pairs of X and Y chromosomes show extensive divergence of their coding sequences, but neither of the two Y chromosomes shows signs of major degeneration (e.g. loss of gene expression). We used a novel method based on gene trees to show that both Xs and both Ys have been coinherited since the translocation occurred. We developed a coalescent model for translocated sex chromosomes. It reveals the genetic patterns suggest the translocation is not very young (that is, it is more than 2N_e generations old), and they localize the breakpoint to near the center of Chromosome 2. Neither of the other two populations we sampled show evidence for the reciprocal translocation. One of them does, however, show evidence suggestive of a different rearrangement involving Chromosome 8b and Chromosome 1.

Figure 2.

The three collecting localities in Fennoscandia.

Materials and Methods

Wild adult frogs were sampled during the breeding season from the three populations in Fennoscandia: Tvedöra (55°50’N / 13°55’E), Kilpisjärvi (69°1’ N / 20°53’ E), and Ammarnäs (65°54’ N / 16°18’ E). We caught five mating pairs from Tvedöra and five mating pairs from Ammarnäs. In Kilpisjärvi, we collected two mating pairs, two single females who had recently spawned, and three single males for a total of four females and five males. Individual mating pairs were kept overnight in 11 L plastic boxes with grass tufts and half-filled with pond water, allowing them to lay a clutch. Within 12 to 24 hours of mating, we anesthetized each member of the mating pair using MS-222, and removed the gonads, liver, and brain before severing the brain stem.

Tissues from field-sampled frogs were stored in RNAlater. RNA was extracted and the mRNA libraries were prepared by Microsynth (Balgach, Switzerland). Libraries were sequenced on the NextSeq500 platform, producing 75bp paired-end reads.

Quality control and transcriptome assembly

We assessed the quality of our sequencing data using FastQC v10.5 (http://www.bioinformatics.babraham.ac.uk/projects/fastqc). Trimmomatic 0.36 (Bolger et al. 2014) was used to remove adapters and to trim leading and trailing bases with a Phred quality less than four. We also trimmed sequences immediately after the average Phred score over four base pairs dropped below 15, and excluded pairs if either read was shorter than 36 bp after trimming. After filtering, we retained an average of 21.75 million reads per tissue sample (Table S1).

We built a de novo transcriptome using reads from the brain, liver, and gonad from a single male and a single female from each of the three populations. The transcriptome was assembled with Trinity v.2.3.0 using default parameters (Grabherr et al. 2011). We extracted the open reading frames from the transcripts using TransDecoder 3.0.1 (Haas et al. 2013), and filtered the assembly for transcripts exceeding 300 bp. Reads were then aligned to the transcriptome using Bowtie2 v.2.2.6 (Langmead & Salzberg 2012) and transcript counts were extracted using RSEM v.1.3.0 (Li & Dewey 2011). We then selected the most highly expressed isoform of each Trinity-assembled “gene,” and extracted counts for transcripts expressed at a minimum of one transcript per million (TPM) reads in at least one tissue per sex.

Mapping transcripts

We mapped our transcripts to the genome of the African clawed frog, Xenopus tropicalis, which is the most closely related anuran with a high quality reference (Hellsten et al. 2010). Although the two species are diverged by some 200 My (Brelsford et al. 2013), frog genomes are thought to be largely syntenic (Sun et al. 2015). Transcripts were reciprocally blasted to peptide sequences from X. tropicalis. One-to-one orthologs were identified when the single best match between transcript and peptide had an e-value less than 10⁻¹⁰, an overlap of at least 70% of both transcript and peptide, and a minimum identity of 40%. These criteria identified 9,904 one-to-one orthologs in R. temporaria and X. tropicalis, which we then used to assign a genomic location. Using this method, if genes on the Rana sex chromosomes evolve more rapidly, they may be detected as orthologs less frequently and so be mapped less frequently than genes on autosomes. We used randomization tests to determine if any chromosome had fewer orthologs than expected. Importantly, we detect no deficit of orthologs mapped to Chromosomes 1 and 2, our putative sex chromosomes, but do detect a deficit on Chromosomes 7 and 9. Details about the tests and their results are given in Table S2.

Our mapping strategy comes with two caveats. The first is that we are viewing our results in terms of the X. tropicalis genome coordinates, not those of hypothesized rearrangements, so care must be taken in the interpretation. Second, unidentified rearrangements fixed in the Rana and Xenopus lineages are a potential source of error. Three rearrangements are known: the homologues of Chromosomes 4, 7, and 8 in X. tropicalis are each represented by two smaller chromosomes in the R. temporaria genome (Brelsford et al. 2016). Comparisons between the X. tropicalis and a draft genome of R. temporaria detected no other large-scale rearrangements between chromosomes, and little evidence for large-scale rearrangements within Chromosomes 1 and 2 (D. Jeffries, personal communication).

Calling variants

We called variants using the pipeline outlined in Quinn et al. (2013) and Wright et al. (2015). We first removed duplicate reads from trimmed files using FastUniq v.1.1 (Xu et al. 2012). These reads were then aligned using the 2-pass alignment method in STAR 2.4.2a (Dobin 2013) with default parameters. Samtools v.1.3.1 (Li et al. 2009) was run with the probabilistic alignment disabled. We used VarScan.v2.3.9 (Koboldt et al. 2009) with a minimum variant allele frequency of 0.10, and a minimum threshold to call a homozygote of 0.85. We required a minimum of 10 reads per site and a Phred quality score greater than 20. For all analyses, we considered only transcripts that had a minimum of 100 sites and that were present in all individuals in the population of interest. Information on the number of transcripts and variants detected in each population is in Table S3.

Differentiation between X and Y chromosomes

We used two statistics to identify sex chromosomes and to quantify differentiation between the Xs and Ys. First, we calculated F_ST between males and females, which is a powerful tool for detecting differentiation between slightly diverged sex chromosomes (Rodrigues & Dufresnes 2017). The maximum possible differentiation occurs when the X and Y are fixed for different alleles, in which case F_ST between the sexes is 1/3. We calculated F_ST using PopGenome (Pfeifer et al. 2014). Trends on chromosomes were visualized using LOESS regression implemented in ggplot2 (Wickham 2009).

To identify regions of the genome in which F_ST is significantly elevated, we calculated the average F_ST in nonoverlapping sliding windows of 25 transcripts. This resulted in 115 autosomal windows in Ammarnäs, 151 autosomal windows in Kilpisjärvi, and 162 autosomal windows in Tvedöra. We devised a method using these values to calculate the 99% confidence intervals for the average values of F_ST in the windows. We recognize regions of sex chromosomes as highly differentiated when the LOESS regression falls outside of the confidence interval for their population. Those regions are shown by the shaded regions in Figures 3 and 6. Appendix S1 shows the LOESS regressions and confidence intervals for all chromosomes in all populations.

Figure 3. Divergence between the translocated sex chromosomes in Ammarnäs.

Chromosome 1 is the ancestral sex chromosome, and Chromosome 2 is the autosome with which it has undergone a reciprocal translocation. Top panels: F_ST for individual transcripts (green circles), with LOESS spline regression (blue curve). Shaded areas show highly differentiated regions (defined in the Methods). The position of the putative sex-determining locus Dmrt1 is shown by the red diamond on Chromosome 1. The horizontal bars show the chromosome regions with gene trees shown in Figure 5. Positions are in Mb as defined by the X. tropicalis reference genome. Bottom panels: The expected values of F_ST predicted from the coalescent model for a Y-autosome reciprocal translocation that is T = 10 time units old. Position on both chromosomes is given in absolute distance from the SDR measured in units of ρ_m = 4 N_e r_m (see the Methods).

Figure 6.

Divergence between males and females in Kilpisjärvi and Tvedöra. Highly differentiated regions occur on Chromosomes 1 and 8b in Kilpisjärvi, but only on Chromosome 1 in Tvedöra.

The second statistic that we used to detect genetic differentiation between the sex chromosomes is sex-biased heterozygosity (SBH), defined as

$$SBH={\log }_{10}\left(H_M/H_{\text{F}}\right),$$

where H_M and H_F are the heterozygosities (that is, the average fraction of sites that are heterozygous) in males and females. H_M is calculated as:

$$H_{\text{M}}=\frac{h_{\text{M}}+1}{t_{\text{M}}+1},$$

where h_M is the average number of heterozygous sites in a male and t_M is the total number of sites. To calculate H_M, we add 1 to those values to avoid numerical issues when either h_M or t_M is equal to 0. Female heterozygosity (H_F) is calculated in an analogous way.

Unlike F_ST, this statistic can detect regions of both high and low differentiation. On autosomes, we do not expect sex-biased heterozygosity (SBH = 0). On sex chromosomes, however, SBH can deviate from zero for several reasons. If the sequence between the X and Y chromosomes are only slightly diverged, and homologous X-derived reads and Y-derived reads are successfully mapped to the same transcript, then there will be higher heterozygosity in males (SBH > 0). Where sequences have diverged substantially from the reference sequence on the Y but not the X, some Y-derived reads may not map, leading to an excess of heterozygosity in females (SBH < 0). Finally, deletions on the Y will also cause females to have higher heterozygosity than males (SBH < 0).

We also calculated several additional descriptive statistics (d_xy, π_female, and π_male) using PopGenome (Pfeifer et al. 2014). The results for all chromosomes in all three populations are shown in Appendix S1.

Coalescent Models

To develop expectations for patterns of molecular variation on rearranged sex chromosomes, we constructed analytic coalescent models by extending those in Kirkpatrick et al. (2010). An ancestral population with recombining sex chromosomes is assumed to be at equilibrium under the Standard Neutral Model. At a time T in the past, a reciprocal translocation occurs between an autosome and the PAR region of either the ancestral X or Y chromosome, yielding a karyotype like that shown in Figure 1. The model gives analytic expressions for the expected coalescent times between pairs of genes sampled from a site on the X chromosomes, a pair sampled from a site on the Y chromosomes, and a gene sampled from an X and another sampled from the Y. These expressions provide expectations for neutral polymorphism (π) on the Xs and Ys, and the divergence (d_xy and F_ST) between them, at any position along the chromosomes. The parameters are T, the age of the translocation (in units of N_e generations), and ρ_m = 4 N_e r_m, the distance of a site from the SDR (where N_e is the effective population size and r_m is the recombination rate in males). Full details are given in Appendix S3. An example of the results is shown in Figure 3.

Gene trees

We devised a new method to detect sex-linked regions that relies on gene trees. We scanned the genome for regions of chromosomes whose gene trees include a monophyletic clade in which each male is represented once and only once. This pattern suggests that chromosome segment lies in the sex-determining region (Figure 4).

Figure 4. Schematic of a gene tree for the SDR.

Shown are chromosomes sampled from two males and two females. A new Y chromosome arises by the spread of a masculinizing mutation (the square). Once the new Y is established, all of its homologs that lack the masculinizing mutation become Xs by default (shown by “A ➔ X”). The Y-linked regions form a monophyletic clade in which each male is represented once and only once (circled by a dashed line). This is an “XY consistent clade”. The X chromosomes may or may not form a monophyletic clade, and in this example they do not.

To obtain the gene trees, we first mapped our R. temporaria transcripts to the Xenopus tropicalis reference, and assigned “N” to all other sites in the genome. SNPs were included only if they were represented by a minimum of 10 reads in all individuals. Each chromosome was then phased using Beagle v.4.1 (Browning & Browning 2007). The phased sequences were extracted using a pipeline consisting of vcftools v.0.1.15 (Danecek et al. 2011), Picard v.1.41 (http://broadinstitute.github.io/picard/), and GATK v.3.8.0 (McKenna et al. 2010). We segmented each chromosome into 20 Mb windows (in reference to the X. tropicalis assembly) using the splitter function in EMBOSS v.6.6.0 (Rice et al. 2000). Because transcripts are not distributed evenly across the genome, windows vary in the number of base pairs per window (Table S4).

Gene trees for each window were estimated using RAxML v.8.2.11 (Stamatakis 2014). Trees were viewed using FigTree v.1.4.3 (http://tree.bio.ed.ac.uk/software/figtree/).

Analysis of d_N/d_S ratios

To determine if X chromosomes show atypical patterns of evolution in their coding regions (such as the “faster-X effect” (Charlesworth et al. 1987; Meisel & Connallon 2013; Vicoso & Charlesworth 2009)), we assembled transcripts de novo from a single female from each population. We used the same reciprocal blast strategy outlined above to assign one-to-one orthologs with Xenopus tropicalis to a chromosomal location. MUSCLE v.3.8.31 (Edgar 2004) was used to construct a protein-protein alignment between orthologous X. tropicalis and R. temporaria sequences. We then used that alignment as guide for aligning the coding sequence in PRANK v.1.40603 (Loytynoja 2014). We calculated d_N, d_S, and d_N/d_S using the yn00 model of PAML v.4.9e (Yang 2007), removing genes with either a high (> 2) or low (< 0.001) d_S value. In Ammarnäs, we compared Chromosomes 1 and 2 with the autosomes (Chromosomes 3–10). In Kilpisjärvi and Tvedöra, we compared Chromosome 1 with the autosomes (Chromosomes 2–10) and Chromosome 2 with the remaining autosomes (3–10). Significance was assessed using Wilcoxon tests.

Sex-biased gene expression

In order to assess gene expression, reads were aligned to the transcriptome using Bowtie2 v.2.2.6 (Langmead & Salzberg 2012) and transcript counts were extracted using RSEM v.1.3.0 (Li & Dewey 2011). We assessed the degree of sex bias in the expression of each gene by calculating the ratio of normalized reads in males and females. We normalized the expression with EdgeR using the weighted trimmed mean of M-values, which is a scaling factor for library sizes that minimizes the log-fold change between samples (McCarthy et al. 2012; Robinson et al. 2010). To evaluate statistical significance, we used the exactTest function in EdgeR (McCarthy et al. 2012; Robinson et al. 2010). Transcripts were considered male-biased or female-biased if they had a significant p-value and if there was more than a two-fold bias in expression. To test the robustness of our results, we also generated pseudoalignments using Kallisto v.0.43.0, which produced pseudocounts that were then analyzed for gene expression in Sleuth (Bray et al. 2016). Expression analyses from RSEM and Kallisto produced similar results (Table S5), therefore the text and Figure S1 and Figure S2 present results obtained from RSEM. Less than 1% of genes were significantly sex-biased in brain and liver, and so only data from the gonads are presented here.

We used two approaches to determine if the sex chromosomes are enriched for genes with sex-biased expression. The first approach asks whether a greater fraction of genes on the sex chromosomes have sex-biased expression. We tested for statistical significance using the randomization tests. Specifically, we determined the number of transcripts assigned to each chromosome in a given population. For each chromosome, we randomly sampled the same number of transcripts from the full set of assigned transcripts, and recorded the number that were sex-biased. This procedure was repeated 10⁴ times to create a null distribution. Significance was assessed based on where the observed fraction of male-biased or female-biased genes fell on the null distribution. In the second approach, we directly assessed the ratio of gene expression between the sexes. We computed the log₂ of the ratios of expression in females to males on Chromosomes 1 and 2 and compared these to the ratios on the autosomes using Wilcoxon tests.

Results

Chromosome 1 shows classic genetic signatures of sex linkage in all three populations. The X and Y chromosomes are most differentiated in Ammarnäs, where both F_ST and sex-biased heterozygosity are elevated along most of the chromosome, a region of about 125 Mb (Figure 3, Table 1, and Appendix S1). The region is centered on Dmrt1, which is the putative male-determining locus in R. temporaria (Ma et al. 2016; Rodrigues et al. 2017). This pattern is expected: previous coalescent models (Kirkpatrick et al. 2010) and new models described below show that differentiation between sex chromosomes is expected to be greatest in the sex determining region (SDR). In the pseudoautosomal region (PAR), differentiation between the X and Y declines with increasing distance from the SDR. (In our data, F_ST is actually slightly greater at the right end of Chromosome 1 than at Dmrt1. The difference is not statistically significant, however, and we interpret the entire region between 100 Mb and 200 Mb as having approximately equally high differentiation.) To the left of DMRT1, F_ST slowly declines towards the telomere. These patterns strongly suggests that recombination between the X and Y is greatly reduced relative to autosomes, but it is not entirely suppressed (Kirkpatrick & Guerrero 2014).

Table 1. Divergence between males and females on sex chromosomes and autosomes.

Values in bold are significant at p < 10⁻⁴ after Bonferroni correction of Wilcoxon tests.

	FST			SBH
Population	Chr 1	Chr 2	Auto	Chr 1	Chr 2	Auto
Ammarnäs	0.12	0.10	0	0.18	0.14	−0.03
Kilpisjärvi	0.06	0.01	0	0.07	−0.02	−0.02
Tvedöra	0.03	0	0	0.03	−0.02	−0.03

Chromosome 2 in Ammarnäs also has elevated F_ST and sex-biased heterozygosity along much of its length (Figure 3, Table 1, and Appendix S1). This is consistent with the hypothesis of a reciprocal translocation with Chromosome 1. The large (75 Mb) region that shows differentiation suggests there is little recombination between the X-linked and Y-linked arms of the former autosome. The coalescent models described below show that a peak in F_ST is expected at the translocation breakpoint, which (from Figure 3) suggests the break lies close to the center of Chromosome 2. The slow but definite decline in F_ST to the left and right of the peak show that there is a very low but positive rate of recombination between the neo-X and neo-Y that the reciprocal translocation formed. None of the other chromosomes show significant differentiation between males and females (Appendix S1).

Gene trees of translocated sex chromosomes

We developed a novel approach for detecting genomic regions that are sex linked that is sketched in Figure 4. Consider an autosome that acquires a masculinizing mutation and then spreads to become a new Y chromosome. All copies of the SDR on this new Y descend from the mutation, and so they form a monophyletic clade in the gene tree for this chromosomal region. (The X chromosomes, by contrast, can form a paraphyletic group, as shown in Figure 4.) Thus in a sample of sex chromosomes from both sexes, the Y chromosomes will be appear as a monophyletic clade in which each male is represented once and only once. We refer to gene trees that meet this monophyly condition as “XY consistent”. In the pseudoautosomal region (PAR), by contrast, recombination between X and Y chromosomes will disrupt the monophyly of sequences that are Y-linked in the current generation.

We used this logic to scan the genome for the SDR region. We computationally phased chromosomes sampled from males and females, then constructed gene trees from nonoverlapping 20Mb windows. Candidate windows inside the SDR were identified by gene trees that were XY consistent. This topology can, however, arise by chance in a region of the genome that is not sex linked. We therefore performed coalescent simulations to find the probability of a false positive under the null hypothesis of no sex linkage. For the sample size in our study (five males and five females from each population), the probability of a monophyletic clade that includes each male once and only once is less than 0.01. XY consistency is therefore strong evidence of sex linkage.

Following a reciprocal translocation, the two Y chromosomes are expected to be transmitted as if they are physically linked. It then follows that if there has not been recombination between the SDR and the translocation breakpoint, gene trees for fully sex-linked regions on both Chromosome 1 and Chromosome 2 will show XY consistency. Further, the clades of the gene trees on the two chromosomes that are XY consistent will have identical topologies. This prediction is a very strong test of the reciprocal translocation hypothesis.

In Ammarnäs, four windows on Chromosome 1 and five windows on Chromosome 2 do indeed show XY consistency, a pattern found only once out of 48 windows on the other chromosomes (Appendix S2). (On Chromosome 1, the windows showing XY consistency do not include Dmrt1, but that could result from phasing errors, and the actual segment that is XY consistent may be much larger.) Further, Figure 5 shows that the topology of gene trees that are XY consistent in the 80–100 Mb region of Chromosome 1 is shared with a gene tree in the 120–140 Mb region of Chromosome 2. This result strongly supports the hypothesis of a reciprocal translocation between these two chromosomes. Further, it suggests that the central region of Chromosome 2 has had very little (if any) recombination with the SDR, and the two unlinked Y chromosomes have been coinherited since the translocation occurred.

Figure 5. Gene trees for the SDRs on Chromosomes 1 and 2.

The two XY consistent clades are circled by a dashed line, and the numbers give the identities of the males corresponding to those tips. The XY consistent clades on Chromosomes 1 and 2 have identical topologies, as predicted for a reciprocal translocation. Tips sampled from males are in green and blue; tips sampled from females are in gray. The scale is the maximum likelihood estimate for the number of substitutions per site. These gene trees are from the chromosome regions shown by horizontal bars in Figure 3.

Coalescent models of translocated sex chromosomes

To strengthen our inferences from the data, we developed coalescent models for translocated sex chromosomes. The models allow for a reciprocal translocation of any age, and allow for any pattern of recombination in the pseudoautosomal region (PAR) of both the ancestral and the neo-sex chromosomes. Full details are given in Appendix S3.

The models lead to several insights. In the case of a reciprocal translocation between the X chromosome and an autosome, immediately after the translocation becomes fixed, F_ST between males and females jumps from 0 to a value of ¼ across the entire length of the translocated autosome. All else equal, molecular diversity (π) is expected to be much lower in females than males. Different patterns are expected for a young Y-autosome reciprocal translocation. F_ST between the sexes jumps only to a value of 1/8, and molecular diversity is expected to be very similar in males and females. Thus genomic data may be able to distinguish between the two types of reciprocal translocations when they are young.

After the reciprocal translocation is established, patterns of diversity within and differentiation between the X and Y evolve towards an equilibrium that is approached closely after about 2N_e generations. At sites in the SDR that are fixed for different bases on the X and Y, F_ST between males and females attains a value of 1/3. In the PAR, F_ST declines with increasing distance from the SDR. Regions of both the ancestral sex chromosomes and the translocated autosome that recombine very little with the SDR (specifically, where 4N_e r is not much greater than 1) will show persistent divergence between males and females. F_ST (and other measures of divergence) will fall to levels typical of autosomes at roughly 10N_e r morgans away from the SDR. Once patterns of molecular variation approach this equilibrium pattern, the data can only say that the age of the rearrangement is at least 2N_e generations.

This pattern of decline in F_ST is expected on the translocated autosome as well as the ancestral sex chromosome. The explanation can be visualized in Figure 1, where the reciprocal translocation breakpoint looks like a street intersection. Proceeding from the SDR to the intersection, and then along any of the three other streets that lead away from the SDR, the probability increases with distance from the intersection that a crossover will occur between that point and the SDR. A crossover erodes the differentiation between the X and Y at all points further from the SDR than the crossover. Thus F_ST on the translocated autosome peaks at the breakpoint and declines to either side. The value of F_ST expected at the autosome’s breakpoint is equal to that of the breakpoint on the ancestral sex chromosome, which in turn depends on its distance from the SDR. Further results from the models are given in the Appendix S3.

Figure 3 compares the patterns in F_ST along Chromosomes 1 and 2 in the Ammarnäs data with those predicted from the coalescent models. Several conclusions can be tentatively drawn. F_ST is not uniform across Chromosome 2, but rather peaks near the center. This pattern localizes the breakpoint to the center region of Chromosome 2. F_ST is elevated over almost the entire lengths of Chromosomes 1 and 2, but it declines with distance from Dmrt1 on the left arm of Chromosome 1 and with distance from the center of Chromosome 2. This suggests that both pairs of X and Y chromosomes do recombine, but only at extremely low rates. Based on the coalescent models, we very roughly estimate the map length of each of the two Y chromosomes to be only about 2ρ, or 12/N_e centimorgans. While we do not have estimates of N_e for these populations, other widespread species of ranid frogs have effective population sizes in the hundreds or thousands (Hoffman et al. 2004; Phillipsen et al. 2011), suggesting these Y chromosomes are much less than a centimorgan long. This is several orders of magnitude less recombination than is seen on the autosomes and on the X chromosome in females (Rodrigues et al. 2013; Brelsford et al. 2016), and is consistent with data from sex chromosomes in hylid frogs (Guerrero et al. 2012).

We elected not to fit the models quantitatively to the data. Issues such as mapping the transcriptome to X. tropicalis (as described earlier) almost certainly would introduce large biases into the estimates. Further, the models show that our power to estimate the translocation’s age and other key parameters declines rapidly with the age of the translocation. The patterns in the data suggest the translocation is not sufficiently young to extract accurate estimates (Appendices S1 and S3).

Unique sex chromosome arrangements in two other populations

We compared these results from Ammarnäs with two other populations: Kilpisjärvi, to the north, and Tvedöra, to the south (Figure 2). In Kilpisjärvi, males and females are again genetically differentiated on Chromosome 1. F_ST and sex-biased heterozygosity are elevated on Chromosome 1 relative to the autosomes (Figure 6 and Table 1). The peak of differentiation lies near Dmrt1, the putative male-determining locus. Furthermore, gene trees on Chromosome 1 from 100–140Mb are XY-consistent (Appendix S2), while no other regions of the genome produce XY-consistent gene trees.

We also find an unexpected peak of differentiation unique to Kilpisjärvi. F_ST is elevated in the region between 56 and 72 Mb on the Xenopus Chromosome 8, which corresponds to the left end of Chromosome 8b in R. temporaria (Figure 6). (Chromosome 8 in X. tropicalis is homologous to two smaller chromosomes, 8a and 8b, in R. temporaria (Brelsford et al. 2016)). We also detect an increase in sex-biased heterozygosity in this region (Appendix S1).

These patterns suggest there may be a different rearrangement fixed in Kilpisjärvi. It could be a reciprocal translocation between Chromosome 1 and a small segment of Chromosome 8b. Alternatively, the data may reflect a nonreciprocal translocation of the left end of Chromosome 8b to Chromosome 1, or a fusion between Chromosomes 8b and 1. These hypotheses could be tested with additional cytological and/or linkage mapping data. (We do not think that the appearance of these population-specific arrangements are the results of assembly errors because those would be expected to affect all populations.) In any event, the relatively small regions of elevated F_ST on both chromosomes (Figure 6) suggests that recombination between X and Y is greater in Kilpisjärvi than in Ammarnäs.

The third population we studied is in Tvedöra. Recall that a previous linkage mapping study found that only Chromosome 1 is sex linked here (Rodrigues et al. 2016). Consistent with those data, we detect elevated F_ST and sex-biased heterozygosity on Chromosome 1 (Figure 6 and Table 1), but not on any other chromosomes. The peak of differentiation is smaller in Tvedöra than in Kilpisjärvi and Ammarnäs, and we did not detect any XY-consistent gene trees (Appendix S2). Further, the peak is offset from Dmrt1. This might suggest that sex in this population is determined by a duplicate of Dmrt1, as in Xenopus laevis (Yoshimoto et al. 2008), or it could be the result of bioinformatic errors. The Tvedöra population therefore appears to have yet a third sex chromosome karyotype, with higher recombination in the PAR than either Ammarnäs or Kilpisjärvi. This is consistent with previous studies based on microsatellites that found little or no differentiation between the sex chromosomes in Tvedöra (Rodrigues et al. 2014; Ma et al. 2016).

Differentiation but not degeneration of the Ys

In some taxa, sex chromosomes show distinctive patterns of sex-biased gene expression (Parsch & Ellegren 2013; Vicoso et al. 2013; Yoshida et al. 2014; Grath & Parsch 2016). Degeneration of the Y chromosome can lead to loss of expression of the genes it carries, and consequently an increase of female-biased expression on the sex chromosomes. Alternatively, mutations that affect coding genes on the Y chromosome can increase male expression, resulting in an increase of male-biased expression on the sex chromosomes. We therefore compared the proportion of sex-biased genes on the sex chromosomes relative to the autosomes.

Chromosome 1 shows slight but significant enrichment in female-biased genes in all populations, and is not enriched in male-biased genes in any population (Figure S1). Chromosome 2 has the second highest proportion of female-biased genes in all populations (Figure S1). This enrichment is small but statistically significant in Ammarnäs and Tvedöra, and not significant in Kilpisjärvi. This result is expected in Ammarnäs, where Chromosome 2 is a sex chromosome and shows evidence of differentiation between the X and Y, but it is unexpected in Tvedöra.

We also compared the average ratio of expression in females and males of genes on the sex chromosomes to genes on the autosomes. Chromosome 1 has a slightly but significantly higher female to male expression ratio than the other chromosomes in all three populations. Chromosome 2 shows a small but significant increase in sex-biased expression in both Ammarnäs and Tvedöra, but not Kilpisjärvi (Figure S2). The observation that Chromosome 2 shows sex bias even in Tvedöra (where it is an autosome) suggests the hypothesis that the sex-bias may have been an ancestral condition that favored the evolution of sex linkage in Ammarnäs.

X chromosomes of some taxa have elevated d_N/d_S ratios compared to autosomes (Meisel & Connallon 2013; Mank et al. 2010). This pattern can result from strong positive selection favoring the X to adapt to females, or from relaxed purifying selection that results from the lower effective population sizes of sex chromosomes relative to autosomes (Charlesworth et al. 1987; Vicoso & Charlesworth 2009). To examine whether such a “faster-X effect” occurred in any of our populations, we examined the d_N/d_S ratio on female-specific assemblies from each population, using Xenopus tropicalis as an outgroup. In none of these three populations is the d_N/d_S ratio significantly different between the sex chromosomes and the autosomes (Figure S3).

In sum, there is evidence for subtle divergence in the expression of genes on the X and Y chromosomes. There is no sign of accelerated accumulation of nonsynonymous mutations on the Y, which would elevate the d_N/d_S ratio. The positive values for sex-biased heterozygosity on the sex chromosomes are consistent with fixed substitutions on the Y chromosomes, but not with many large deletions or loss-of-function mutations (see Materials and Methods). Together, these results suggest the ancestral and neo-X and Y chromosomes have diverged in their coding sequences, but that the Ys have not degenerated extensively.

Discussion

Our results provide compelling support for the earlier suggestion from a linkage study (Rodrigues et al. 2016) that there is a reciprocal translocation between the ancestral sex chromosome (Chromosome 1) and an autosome (Chromosome 2) in Ammarnäs that is absent from two other populations, Kilpisjärvi and Tvedöra. Analysis of molecular variation shows that that two pairs of sex chromosomes are cotransmitted even though they are not physically linked. This inference, which could not be drawn from cytological data, is supported by patterns of divergence between males and females, and is confirmed by a new method we developed to detect sex-linked regions based on gene trees. Further, gene trees for the pair of putative Y chromosomes share an identical topology, showing they have been cotransmitted since the reciprocal translocation occurred. The transcriptome data indicate that both X chromosomes are differentiated in their coding sequence from their Y chromosome partners. There is not, however, evidence of large-scale degeneration of the neo-Y chromosome.

Comparison of the data with results from coalescent models show that the two Xs continue to recombine with the two Ys, but at an extremely low rate: the map length of each Y chromosome is very roughly 12/N_e cM. This rate is orders of magnitude lower than the autosomes, which continue to recombine in females at typical rates, and is far smaller than what could be detected with a linkage map. A similar situation is known from hylid frogs (Guerrero et al. 2012).

In the Kilpisjärvi population, Chromosome 1 is likewise a sex chromosome, but there is no evidence of rearrangements between it and Chromosome 2. Instead, the data suggest there may be translocation or fusion between Chromosomes 1 and 8b that has resulted in a small region of differentiation between the sex-linked segment of Chromosome 8b. In the third population, Tvedöra, Chromosome 1 is again the sex chromosome, and there is no evidence of rearrangements with or sex linkage of any other chromosomes. Thus each of the three populations appears to have a unique sex chromosome karyotype. This degree of structural polymorphism in sex chromosomes seems to be without precedent in the literature.

While Chromosome 1 functions as a sex chromosome in all three populations, the pattern of differentiation between the X and Y differs strikingly. In Ammarnäs, most of the chromosome shows strong differentiation. In Kilpisjärvi and Tvedöra, the differentiated regions are much smaller. These contrasting patterns likely result from differences in the frequency of recombination (Kirkpatrick et al. 2010). One hypothesis for variation in recombination involves sex reversal (Perrin 2009). In females of Rana temporaria (and other frogs), recombination rates are relatively uniform along the chromosomes (Cano et al. 2011; Brelsford et al. 2016; Rodrigues et al. 2013). In males, by contrast, crossing over is restricted to small regions very close to each telomere, and there is little or no recombination over most of chromosomes (Brelsford et al. 2016). This effectively blocks recombination between the X and Y over most of their length. But sex-reversed XY females have been found in some populations, and they show female-typical patterns of recombination (Rodrigues et al. 2018). Recombination in sex-reversed females could therefore result in very rare recombination events between the X and Y.

The differences we see between populations in the degree of differentiation along Chromosome 1 could result if sex reversal is more common in Tvedöra and Kilpisjärvi than in Ammarnäs. Populations in Sweden vary in the degree of sexual differentiation at metamorphosis. Individuals from Ammarnäs have a higher degree of gonadal differentiation at metamorphosis than those from Tvedöra (Rodrigues et al. 2015). Populations with delayed sexual differentiation may have a greater frequency of sex reversal, which could account for the differences we observe among the three populations in the degree of differentiation between their Xs and Ys.

We find that Chromosome 1 is slightly enriched for female-biased genes in all three populations, which is expected because it is a sex chromosome in these populations. Interestingly, Chromosome 2 is slightly enriched in genes with female-biased expression in both Tvedöra and Ammarnäs, despite being only sex-linked in Ammarnäs. Enrichment of female-biased expression is therefore surprising in Tvedöra. One hypothesis for this observation is that by chance Chromosome 2 happens to carry an excess of loci with sex-biased expression. This situation could result if there are many loci on that chromosome that experience sexually antagonistic selection (SAS), which can lead to the evolution of sex-biased expression (Grath & Parsch 2016; Parsch & Ellegren 2013; Cheng & Kirkpatrick 2016; Mank 2017). SAS acting on Chromosome 2 could also be the driving force for the fixation of the reciprocal translocation in Ammarnäs (discussed below).

An alternative hypothesis is that Chromosome 2 has a history of sex linkage. Sex-biased gene expression can evolve rapidly in neo-sex chromosomes (Yoshida et al. 2014). Perhaps Chromosome 2 was ancestrally sex linked, and has not lost the female-biased expression that its loci evolved during that time. However, an analysis of sex chromosome turnover in ranid frogs did not find that Chromosome 2 was the ancestral sex chromosome (Jeffries et al. 2018).

When SAS acts on the alleles at a locus, chromosome rearrangements are favored that cause that locus to become sex-linked (Charlesworth & Charlesworth 1980). The fitness advantage to the rearrangement is proportional to two factors: how much recombination is decreased between the loci under SAS and the SDR, and the strength of SAS. Reciprocal translocations involving sex chromosomes may be more prone than other types of rearrangements to become fixed by sexually antagonistic selection. A reciprocal translocation can immediately cause both arms of an autosome to become sex linked (Figure 1). In taxa like frogs, the potential for a fitness benefit to the reciprocal translocation is amplified by the extremely low recombination rates in males: both arms of the autosome become very tightly linked to the SDR. Thus any locus under SAS on any autosome has the potential to drive fixation of a reciprocal translocation with the sex chromosome. We have no evidence regarding sexually antagonistic selection in the common frog, but the very strong sex linkage of two large chromosomes suggests this may be a species worth studying in that regard.

We noted in the introduction that only a handful of examples are known of reciprocal translocations between sex chromosomes and autosomes. A sex chromosome reciprocal translocation also occurs in Rana tagoi, a congener of the common frog. These are independent translocations (Chromosomes 8 and 9 co-segregate as sex chromosomes in R. tagoi (Ryuzaki et al. 1999). It is possible that this type of rearrangement is more common in ranid frogs than many other taxa because males, which are translocation heterozygotes, suffer little or no reduction in fitness since the highly suppressed recombination in males reduces their chance of aneuploidy. This idea has been proposed to explain the reciprocal translocations that have fixed in several lineages of evening primrose (Oenothera) (Rauwolf et al. 2011).

Regardless of what evolutionary force drives the evolution of sex chromosome reciprocal translocations in R. temporaria, suppressed recombination in males largely isolates both Y chromosomes from genetic exchanges with the two X chromosomes. If suppression was complete, we would expect degeneration of the Ys that is seen in many other taxa (Bachtrog 2013). Our data, however, do not show evidence of extreme degeneration in the two Ys: they do not seem have large deletions, their loci continue to be expressed, and d_N/d_S ratios for their genes are comparable to the values seen on the autosomes. The maintenance of functional loci on the Y chromosomes are likely the result of the very low ongoing rates of recombination between the X and Y that are suggested by our data and are also seen in hylid frogs (Guerrero et al. 2012). Indeed, very low but positive recombination may explain why the ancient sex chromosomes found in some taxa have not degenerated, in striking contrast to those in mammals, birds, and Drosophila (Bachtrog et al. 2014).