Reconstruction of female heterogamety from admixture of XX-XY and ZZ-ZW sex-chromosome systems within a frog species

2.1 Frogs and ethics

We collected a total of 305 frogs G. rugosa from 32 populations in Kinki district, Central Japan (Table 1; Figure 2; Supporting Information Table S1). Phenotypic sex of the adult frogs was determined by dissection and observation of size and shape of gonads. Animal care and experimental procedures were conducted under approval of the Committee for Ethics in Animal Experimentation at Hiroshima University (Permit Number: G13-3).

2.2 Identification of heterogametic sex

The heterogametic sex of 14 of 32 populations was reported in one of our previous studies (Ogata et al., 2008), and in this study, we determined those of the remaining 18 populations by genotyping using two sex-linked genes, ADP/ATP translocase (AAT) and Steroidogenic Factor 1 (SF1). Genotyping of AAT was performed according to the protocol described previously by Sakisaka, Yahara, Miura, and Kasuya (2000). The genotyping of SF1 was performed as follows: Total genomic DNA was extracted from whole blood using the DNeasy blood and tissue Kit (Qiagen, NY) according to the manufacturer's instruction. PCR was performed using Ex Taq (TaKaRa, Japan) as follows: 1 μl of DNA was amplified in 25 μl reaction volume containing 12.5 μl 2 ×  buffer and 0.5 μl of each of the 12.5 μM primers at 94°C for 40 s, 60°C for 20 s and 72°C for 20 s for 35 cycles using Smart cycler (TaKaRa). Based on the deletion of 14 nucleotides in 5′ untranslated regions of SF1 on W and X chromosomes, W and X alleles were distinguished from those of the Z and Y in size by running a 3% agarose or 6% polyacrylamide gel electrophoresis (Supporting Information Figure S1). The primers used were as follows: forward 5′-CTT TCA GGA GAG CGA GCC GT-3′ and reverse 5′-GCG GTG CAG CTT GTA GTC CT-3′.

2.3 Mitochondrial DNA

The 421-base pair (bp) partial fragment of the mitochondrial cytochrome b gene was amplified according to the method described previously (Lee et al., 1999), but with a newly designed forward primer: forward 5′-TYA CCG GCC TAT TCC TAG C-3′. The nucleotide sequence was determined using an abi prism 310 genetic analyser (Applied Biosystems) according to the manufacturer's instruction. The sequences of five haplotypes, Cytb A1-3 and B1-2 have been deposited with the DDBJ (Accession nos LC030018–LC030022). The five haplotypes are classified into two haplogroups, Cytb-W and Cytb-E, based on the sequences (Supporting Information Figure S2), of which PvuII sitemaps are different and discriminate the two haplogroups by polymerase chain reaction–restriction fragment length polymorphism analysis (RFLP): The amplified 548 base pairs (bp) partial fragments of the mitochondrial cytochrome b gene were digested with PvuII and electrophoresed by 3% agarose gel (Supporting Information Figure S3). For the construction of gene trees, we used mega4 (Tamura, Dudley, Nei, & Kumar, 2007); a distance matrix based on Kimura's two-parameter method (Kimura, 1980) was calculated and clustered by the neighbour-joining method (Saitoh & Nei, 1987).

2.4 Microsatellite DNA of sex-linked SOX3

SOX3 is located on the Z, W, X and Y chromosomes of G. rugosa (Uno et al., 2008). To amplify the CA repeat microsatellite DNA on the upstream region of SOX3, we designed the following PCR primers: forward: 5′-CCT TCT GGG TTA AAC AAA TCA A-3′, reverse 1: 5′-GTG TTG TGC CCT GAG TCA T-3′ and reverse 2: 5′-GGA ATC TTT AGA GGT GTG AG-3′ (Supporting Information Figure S4). To amplify the fragments of W and X chromosomes and those of Z and Y chromosomes, we used the forward primer and the primers of reverse 1 and reverse 2, respectively. PCR was performed at 94°C for 40 s, 56°C for 30 s and 72°C for 30 s for 27 cycles and ending at 72°C for 2 min. Fragment analysis was performed using an abi prism 310 genetic analyser with GS500 or GS1000 size markers, with two kinds of fluorochrome (5Hex and 6FAM) attached to the 5′ end of the forward primers. The program genescan 2 was used for the genotyping analysis (Applied Biosystems).

2.5 Artificial crossing

Experimental crosses between populations were performed following the method described by Ohtani, Miura, Kondo, and Uchibori (1997). For reciprocal crossings, we used ZZ males and ZW females of the Neo-ZW1 subgroup from Toyono (P24) in Osaka Prefecture, and XY males and XX females from Hamamatsu in Shizuoka Prefecture (a closed circle with white outline on the map of Japan in Figure 1 and a white circle on the right bottom map in Figure 2), and Sekigahara (P2) in Gifu prefecture for the XY group (Table 1 and Figure 2). The sex of juveniles was determined by dissection and observation of size and shape of gonads. Heterozygous/homozygous state in the sex-linked loci involved was determined by PCR, using SOX3 microsatellite DNA and SF1 as described previously (sections “2.2” and “2.4”).

2.6 Genotyping of juveniles developed from egg masses in the wild

Five egg masses in total, possibly laid by the same mating pair, were collected from a tiny pond in Minakuchi (P8) in Shiga Prefecture (Figure 2). This population is located within the hybrid zone between the Neo-ZW2 subgroup and XY group. From these egg masses, a total of 87 juveniles metamorphosed. Two or three weeks after metamorphosis, we identified the sex of all the juveniles by dissection and observation of size and shape of gonads (Supporting Information Table S2). Heterozygous/homozygous state in the sex-linked loci involved was determined by PCR, using SOX3 microsatellite DNA and SF1 as described previously (sections “2.2” and “2.4”).

2.7 Population structure

We used DArTseq (Diversity Arrays Technology) for population study following the protocol described by Kilian et al. (2012) and Lambert, Skelly, and Ezaz (2016). We genotyped 94 frogs from 11 populations (P2, 3, 7, 8, 10, 12, 13, 16, 17, 23, 24, indicated with asterisks in Table 1) following prior work using DArTseq. From a total of 40,013 single nucleotide polymorphism (SNP) sites identified by DArTseq, we identified 2,974 loci (948 SNP and 2,026 presence-and-absence [PA] loci) with no missing alleles across samples and which were not sex-linked (Supporting Information Table S3). Removing sex-linked SNP loci is critical for population genomic analyses to minimize false genetic structuring (Benestan et al., 2017). To characterize genetic relationships among 94 frogs, structure (ver. 2.3; Pritchard, 2010) was run on these 2,974 loci for 10,000 Markov chain Monte Carlo cycles following 10,000 burn-in cycles, using admixture model with independent allele frequencies. Ten replications were performed for each K, in the range K = 1–5, and the optimal K was estimated using structure harvest software (Earl & Vonholdt, 2012). Additionally, we used a custom R script developed in Lambert et al. (2016), to calculate a Hamming Distance Matrix on 12,735 non-sex-linked markers (3,934 SNP loci and 8,801 presence–absence loci). These loci were selected because they were present in at least 97% of individuals and were not sex-linked. Hamming's distance represents the cumulative number of pairwise distances across all loci amongst individuals and offers an additional approach for visualizing population structure using genomic data. Finally, using the “mantel” function in the r package “ecodist,” we performed Mantel tests on the 12,735 loci to test for spatial clustering in our data. We also used principal component analyses (PCA) to assess deme clustering across this large genomic data set.

2.8 Sex-linked loci analyses

Following Lambert et al. (2016), we identified sex-linked SNP and PA loci independently for each population. Because sample sizes were relatively small (n = 2–6 per sex per population), we removed all loci with any missing data and identified loci which were perfectly sex-linked to minimize the number of spuriously identified sex-linked loci. Prior analyses have shown that 13–15 individuals per sex are necessary to minimize identification of false-positive sex-linked loci (Brelsford, Lavancy, Sermier, Rausch, & Perrin, 2017; Lambert et al., 2016). A SNP locus was considered perfectly sex-linked if all individuals of a given phenotypic sex were homozygous (e.g., XX or ZZ) and the other sex was heterozygous (e.g., XY or ZW). For PA markers, a locus was considered perfectly sex-linked if it was only sequenced in one sex. We pooled all unique sex-linked loci identified for each population to analyse the degree of sex-linkage of all loci across populations (Supporting Information Tables S4 and S5). As with the broader non-sex-linked genomic data set, we used a Hamming's distance matrix to assess population structure in the sex-linked marker data and used a Mantel test to assess whether loci were similarly sex-linked at closer spatial scales. Finally, we used a PCA to test for clustering between males and females of Neo-ZW1, Neo-ZW2, and XY populations in the sex-linked data set.

3.1 Male and/or female heterogamety

To identify the precise distributions of the Neo-ZW and XY groups with different types of sex chromosomes, we examined the heterogametic sexes of 32 populations based on genotypes of sex-linked genes. Data were collected for 18 populations from the Kinki district, including Lake Biwa in Central Japan, while data for the other 14 populations were taken from the study of Ogata et al. (2008). A summary of the heterogametic sex and geographic distributions is presented in Table 1 and Figure 2. Five populations (P1-5) in the eastern part of the Kinki district and 20 populations (P13-32) in the western part had heterogametic males (XY) and females (ZW), respectively. These populations belonged to either of the two typical groups, XY and Neo-ZW. In contrast, in the seven populations (P6-12) located between the two groups, both male and female frogs were heterogametic, showing intermingling of the two groups (in yellow, Figure 2). In the vicinity of the P8 population, we identified not only heterogametic males and females, but also homogametic males and females developed from the same egg mass (we found only five small egg masses in total, possibly laid by the same mating pair, within a tiny pool; Supporting Information Table S2), providing evidence of a successful cross between a heterogametic ZW female and XY male in the wild.

3.2 Discordance of heterogametic sex and mitochondrial haplotype

To discern the cytoplasmic origins of the two groups and the intermingling populations, we analysed the mitochondrial cytochrome b gene. Five haplotypes (A1-3 and B1-2) were identified and classified into two distinct haplogroups, Cytb-W (West, B1 and B2) and Cytb-E (East, A1–A3; Supporting Information Figure S2). Because the PvuII restriction map differed between the two haplogroups, we conducted PCR-RFLP analysis on the 32 populations to identify their haplogroups (see Materials and Methods). Fourteen Neo-ZW populations (P19-32) in the west were found to be Cytb-W type, while all five XY populations (P1-5), four Neo-ZW populations (P13-16) surrounding Lake Biwa and seven intermingling populations (P6-12) were identified as Cytb-E type (Table 1 and Figure 2). In the remaining two populations of Neo-ZW (P17 and P18), both haplogroups were detected within a single population: Six males and four females were Cytb-E while three males were Cytb-W in population 17, and two males were Cytb-E while four males and one female were Cytb-W in population 18. Based on the mitochondrial haplogroups, we reclassified the Neo-ZW group into two subgroups: Neo-ZW1 with Cytb-W in the west, and Neo-ZW2 with Cytb-E or Cytb-E and Cytb-W in the east (in red and purple, respectively; Figure 2 and Table 1); the latter being further classified into three subpopulations based on the frequency of Cytb-E: Neo-ZW2-A (100%), Neo-ZW2-B (76.9%) and Neo-ZW2-C (28.6%) (Figure 2; Table 1).

3.3 Hybrid origin of nuclear genome

An unexpected finding was that the mitochondrial DNA of the Neo-ZW2 subgroup originated completely or partly from the XY group. To investigate the nuclear genomic origin of Neo-ZW2, we performed structure analysis to determine population structure using polymorphic autosomal 2,974 loci (Supporting Information Table S3) in 94 frogs from 11 populations (numbers with asterisks in Table 1). The Delta K value was largest in k = 2: 158.35 (k = 2), 0.79 (k = 3), 2.61 (k = 4) and 10.20 (k = 5). The distribution of the alleles in the structure map (k = 2) clearly showed that the Neo-ZW1 subgroup and XY group formed distinct clusters (in red and blue, respectively; Figure 3a), and all individuals from the Neo-ZW2 and intermingling populations were assigned to both clusters with similar likelihoods (Figure 3a). A principal components analysis using 12,735 autosomal loci (3,934 SNP and 8,801 PA loci) showed no overlap between Neo-ZW1, Neo-ZW2 or XY populations and that Neo-ZW2 populations were located in PC space directly between Neo-ZW1 and XY populations (Figure 3b). A Mantel test indicated high spatial genetic structuring with substantial genetic similarity among closer populations (r = 0.69, 95% CI 0.67–0.71, p = 0.001).

3.4 Origins of sex chromosomes

To investigate the origin of sex chromosomes in the Neo-ZW2 subgroup, we identified a total of 2,358 sex-linked markers (518 SNP loci and 1,840 presence–absence loci) across populations with sex-linked loci identified in each population (Table 2; Supporting Information Tables S4 and S5). While the sample sizes for each population were low, and certainly resulted in many false positives, the proportion of male or female heterogametic loci always conformed to the known XY or ZW system of a population, indicating that our filtering identified true sex-linked loci. blastn (https://blast.ncbi.nlm.nih.gov/Blast.cgi) searches yielded no homologies with any known genes.

The PCA on 2,358 sex-linked markers indicated substantial variation between populations and sexes (Figure 4). As expected, there was no overlap between the sexes for any deme (XY, Neo-ZW1, Neo-ZW2). Particularly interesting was the remarkable PC space overlap in sex-linked loci between ZW females of the Neo-ZW2 deme and XY males of the XY deme, and there was no overlap between ZW females of Neo-ZW1 and -ZW2 demes, although ZZ males from Neo-ZW1 and -ZW2 demes showed substantial overlap in PC space. Hamming's Dissimilarity Distances showed that, across sex-linked loci, ZW females of Neo-ZW2 were least dissimilar (most similar) to XY males of XY (33% dissimilar, STD 2.0%) and ZW females of Neo-ZW1 (34% dissimilar, STD 3.0%). For comparison, ZW females of Neo-ZW1 and XY males of XY were 39% dissimilar (STD 2.0%). For the sex-linked marker data set, a Mantel test suggested that the spatial genomic structuring was significant, but substantially weaker than in the autosomal data (r = 0.21, 95% CI 0.17–24, p = 0.001).

Next, we analysed microsatellite DNA from the upstream region of the SOX3 gene (SRY-related HMG-box 3), a strong candidate for sex determination (Miura, 2017; Miura, Ohtani, Ogata, & Ezaz, 2016; Miura et al., 2009). In total, 29 alleles were identified on the Y and Z chromosomes (Table 3 and Supporting Information Table S6). Out of them, two alleles (S8 and S25) were detected both in the Y chromosome of the XY group and Z chromosomes of Neo-ZW2, but not in the Z chromosomes of Neo-ZW1, while four alleles (S4, S7, S10 and S12) were detected in the Z chromosomes of both the Neo-ZW1 and -ZW2 groups, but not in the Y chromosomes of XY.

Table 3. Allele frequencies of microsatellites of SOX3 on Z and Y chromosomes

Note

• Blue colour letters indicate haplotypes that are detected in XY groups and/or others without Neo-ZW1 group, while red letters indicate haplotypes that are detected in Neo-ZW1 and/or others without XY group. Pink letters denote haplotypes that are detected in Neo-ZW2 and intermingled populations. Green letters indicate the haplotype that is found across the XY and Neo-ZW groups. Brown letters indicate haplotypes that are found only in the intermingled populations.

On the W and X chromosomes, six alleles were identified (Table 4). Out of them, three alleles (222, 232 and 248) were detected in Neo-ZW2-A and -C and XY populations, but not in the Neo-ZW1 populations. The allele 240 was detected in all W chromosomes of both the Neo-ZW2-B and Neo-ZW1, and at very low rates in two XY populations.

Table 4. Allele frequencies of microsatellites of SOX3 on W and X chromosomes

Note

• Abbreviation of blue letters is the same as in Table 2. Red letters indicate the haplotype that is predominantly detected in Neo-ZW1 group.

3.5 Sex of unusual genotypes WY, WX, ZY and ZX in hybrids

When two populations in this species, each of which bears either the ZZ/ZW or XX/XY types of sex chromosomes, meet and hybridize, reciprocal crossings are expected: One is a crossing of ZW female × XY male, and the other is a crossing of XX female × ZZ male. In the former case, four types of unusual sex chromosome genotypes are expected. To test this, we artificially crossed the frogs to examine their sexes (totally five crossings using two ZW females and five XY males). ZY frogs were all males, and almost all WX frogs (96.6%) were females, while ZX and WY frogs were males or females (56.8% of ZX frogs and 32.3% of WY frogs were females, respectively; Supporting Information Table S7). Interestingly, in the three crosses using the ZW1 female frog (Supporting Information Table S7), more ZY and less WX progeny were born than was expected of parity (30 ZY and 10 WX frogs from a total of 75 frogs; χ² = 11.09, p = 0.011), while WY frogs were less frequent than expected in the crosses using the ZW2 female (13 WY frogs from a total of 71 frogs; χ² = 8.317, p = 0.039). On the other hand, from the four reciprocal crosses (XX × ZZ), 83.1% of XZ frogs were females and 16.9% were males (Supporting Information Table S8). In our assessment of sex of the XZ genotype in the backcrosses without W and Y chromosomes (XX females × XZ or ZX (F1) males, and XZ (F1) females × ZZ males), the sex ratio of heterozygous XZ was relatively even, that is, 50.5% males and 49.5% females (Supporting Information Table S8).

3.6 Lethality of WW and YY genotypes

The WY genotype was found to be male (67.7%) or female (32.3%) in the crosses between Neo-ZW1 and XY. This implies that a backcross of WY and XY males to ZW and/or WY females, respectively, should lead to the production of WW and YY genotypes in the subsequent generations. If degeneration of the W and Y chromosomes had already progressed, WW or YY frogs would not be viable. To examine the potential lethality of the WW and YY genotypes, we artificially crossed WY males with ZW females (six crosses), and XY males with WY females (three crosses). All WW embryos simultaneously died of the same oedema just after hatching at 6 days postfertilization (dpf; Supporting Information Figure S5; Table S9), showing progressive W chromosome degeneration. In contrast, YY embryos developed beyond the larval stage (Supporting Information Table S9). However, the number of sexually matured YY adults was significantly lower than expected (YY, WX and WY + XY = 10 (13.8%), 14 and 48; χ² = 6.49, p = 0.037), suggesting a slight underdevelopment of the YY tadpoles and frogs and thus, the proceeding of Y chromosome degeneration.