To map a genomic locus responsible for the sterility of the its mutant line, we performed screening with simple sequence-length polymorphism (SSLP) markers (see Materials and Methods). The causal genomic region was mapped on a 5.5 Mbp region on chromosome 23 containing 77 annotated open reading frames, including that of the sycp2 gene (S1 Table). We have previously reported that its spermatocytes show aggregation of the Sycp3 protein in the nucleus (see below) [25]. Because this phenotype resembles that of Sycp2 mutant spermatocytes in mice [22], we first examined the coding sequence of the sycp2 gene; however, no mutation was detected. We next examined the expression of sycp2 mRNA in its mutant and wild-type testes by RT-qPCR (Fig 1A). The expression of sycp2 in its mutant testes was only 25.9% of that in wild-type testes (p<0.01). Thus, the expression of sycp2 mRNA was significantly reduced in its mutant testes.

To examine whether the decreased expression of sycp2 is responsible for the its phenotypes, we generated sycp2 knockout zebrafish by TALEN mutagenesis targeting exon 18 (S1 Fig). We isolated three deletion mutations: two frameshift mutations (14- and 16-bp deletions; sycp2^Δ14 and sycp2^Δ16, respectively) and a mutation without a frameshift (3-bp deletion; sycp2^Δ3). As the phenotypes of the two frameshift mutant fish were identical at cytological levels (S6 Fig, S7 Fig), the sycp2^Δ14 allele was used as the sycp2 knockout allele (sycp2^-) in this study, unless otherwise noted. Notably, homozygote mutant females were obtained for the sycp2^Δ3 allele, but not for the sycp2^Δ14 and sycp2^Δ16 alleles, after intercrossing of F₁ heterozygous fish. This is consistent with our previous observation that no females appear among its mutant fish (see below) [25]. Therefore, subsequent analyses of the sycp2 mutant were performed only with male zebrafish. We confirmed the depletion of sycp2 mRNA expression in sycp2^-/- testes by RT-qPCR (Fig 1A) and of the Sycp2 protein by Western blotting (S5 Fig), showing that the frameshift mutation resulted in a null phenotype for the sycp2^Δ14 allele presumably through nonsense-mediated mRNA decay, a eukaryotic pathway to eliminate transcripts with a premature stop codon (see below) [28]. We examined the fertility of sycp2^-/- male fish after allowing natural mating with wild-type female fish. No fertilized eggs were identified after examining 825 eggs from six crosses between sycp2^-/- male and wild-type female fish, while 66.4% of eggs (249 of 375 eggs) from four crosses between wild-type male and female fish developed normally into gastrulae. Consistent with this result, histological analysis demonstrated that neither spermatids nor spermatozoa were present in sycp2^-/- testes (Fig 1B; sycp2^-/-). We obtained fertilized eggs from the mating of sycp2^Δ3/Δ3 males and wild-type females (82.1%, 348 of 424 eggs from four crosses). Therefore, it can be ruled out that the sterility and meiotic phenotypes (see below) observed in sycp2^-/- fish are caused by off-target effects on another locus.

Next, we examined whether the its mutation can complement the sterility caused by the sycp2^- mutation. After mating of its heterozygous mutant fish (sycp2^its/IM; IM is an inbred wild-type strain derived from India; see Materials and Methods) and sycp2^+/- fish, we obtained 67 offspring, among which 17 fish were sycp2^its/-. Histological analysis of the sycp2^its/- testes showed similar phenotypes as those observed in sycp2^-/- and sycp2^its/its testes, where neither spermatids nor spermatozoa, but accumulations of spermatocytes were observed (Fig 1B). After mating five sycp2^its/- males with wild-type females, no fertilized eggs were observed among 1785 eggs examined, while 31.4% of embryos from a sycp2^IM/- sibling cross (123 of 391 eggs) and 94.9% of embryos from a wild-type sibling cross (227 of 239 eggs) were fertilized and developed normally. Thus, the sycp2 knockout allele could not complement the its phenotypes. This result suggests that meiotic defects in its testes are caused by reduced expression of Sycp2.

To identify the potential mutation responsible for low sycp2 expression, we examined the sequence of sycp2 cDNA prepared from its and wild-type testes. After sequencing of several sycp2 cDNAs from its individuals, we found an insertion of 5 bp at the splice site between exons 8 and 9 in addition to the wild-type sequence (Fig 2A and S2 Fig). This 5-bp insertion was not detected in sycp2 cDNA prepared from wild-type testes. This finding implies that an abnormal splicing event occurs in its mutants. Therefore, we examined the genomic sequences of sycp2 intron 8 and identified a T-to-A substitution of the 127^th residue in the its but not in the wild-type genome (Fig 2B). This substitution generates a potential aberrant splice site upstream of the canonical splice site of exons 8 and 9. We performed RT-PCR with a primer pair flanking the exon 8–9 junction of sycp2 to determine how much of the sycp2 mRNA was spliced at the abnormal site. Strikingly, most of the its mRNA was spliced at the aberrant site in intron 8 (Fig 2C), and only ~15% of sycp2 mRNA was found to be spliced at the canonical site in its at exons 8–9 after sequencing of sycp2 cDNA (S2 Table).

Finally, to verify that this abnormal splicing was caused by the T-to-A substitution and not by a trans-factor in the its line, we carried out a mini-gene splicing assay by introducing mini-gene constructs into wild-type zebrafish (S3 Fig and Fig 2D, see Materials and Methods). RT-PCR of the splice site sequence transcribed from the mini-gene construct revealed that the T-to-A substitution in its sycp2 intron 8 was sufficient to introduce aberrant splicing in the wild-type context (Fig 2D). Notably, the 5-bp sequence inserted in the sycp2 cDNA of its contains a premature terminal TAG codon (Fig 2A). In eukaryotes, mRNAs that contain premature stop codons are often eliminated by nonsense-mediated mRNA decay [28]. Thus, the low level of sycp2 expression in its testes potentially results from the insertion of a premature termination codon by aberrant exon 8–9 splicing, corroborating the above assumption. Altogether, these results suggest that its phenotypes are primarily derived from mis-splicing resulting from a T-to-A substitution in the sycp2 intron 8.

As mentioned above, no females were observed among sycp2^-/- knockout fish, as previously reported for its mutant fish [25]. This phenotype was markedly different from spo11 mutant zebrafish, which are deficient in meiotic DSB formation; spo11^-/- females generate eggs capable of fertilization, but the resulting embryos develop abnormally, most likely due to aneuploidy [26]. To confirm the sexual phenotypes of sycp2 knockout zebrafish and to compare them with those of spo11 knockout zebrafish, we analyzed genotypes of male and female offspring from the intercrossing of each line (Table 1). To determine the sexual phenotypes, we examined the morphology of gonads dissected from 8- to 9-week-old offspring (see Materials and Methods). We observed homozygote mutant males in offspring from both sycp2^+/- and spo11^+/- intercrossing (31% and 17% of all male offspring, respectively). While 24% (11 of 46 fish) of female offspring from the spo11^+/- intercrossing appeared to be homozygous mutant, no homozygous mutant female appeared among offspring from the sycp2^+/- intercrossing. In wild-type zebrafish, all juveniles develop gonads with immature oocytes regardless of their definitive sex, and the individuals in which these immature oocytes degenerate become males [29–31]. Therefore, zebrafish mutants depleted of oocytes develop as males that are mostly infertile [32–35]. In sycp2^-/- juvenile gonads, we observed previtellogenic dictyate oocytes (late stage IB oocytes [36]) in four among five individual fish at 28 days postfertilization (S4 Fig). Therefore, the absence of females in sycp2^-/- fish (Table 1) implies that oocytes were eliminated at the juvenile age or later, possibly due to checkpoint activation in oogenesis. As a result, all sycp2 mutant fish develop as males.

SYCP2 is a conserved lateral element component of the SC in metazoans and is essential for SC formation in mice [18,22]. We previously reported abnormal aggregation of Sycp3 in its spermatocytes [25], suggesting that zebrafish Sycp2 has a conserved function in axis and SC formation. To further understand Sycp2 functions in SC assembly, wild-type and sycp2 mutant spermatocyte chromosomal spreads were immunostained with antibodies against the SC components Sycp3, Sycp2 and Sycp1 (Fig 3 and S6 Fig). In wild-type spermatocytes, Sycp3 appeared on chromatin prior to Sycp2 and Sycp1, as a few bright foci at preleptonema (Fig 3A-i). Sycp3 began to be extended as short fragments upon the appearance of Sycp2 signals at leptonema (Fig 3A-ii). At zygonema, Sycp1 began to be observed as short fragments, following the appearance of Sycp3 and Sycp2 signals (Fig 3A-iii). All the Sycp3, Sycp2, and Sycp1 fragments extended during the progression of zygonema (Fig 3A-iv), and 25 pairs of homologous chromosomes were stained for these three proteins along the entire length at pachynema (Fig 3A-v).

In sycp2^-/- spermatocytes, Sycp3 was detected only as a few bright aggregates similar to those in wild-type preleptotene nuclei (Fig 3B-i), suggesting that Sycp2 is required for the localization and extension of Sycp3 along the axis in zebrafish spermatocytes. Furthermore, despite the absence of Sycp2 and Sycp3 fragments, sycp2^-/- spermatocytes showed aberrant filaments of Sycp1 that were not homogeneous in length, and their numbers were drastically decreased compared to those in wild-type nuclei (Fig 3B-i and S6 Fig). Thus, sycp2^-/- spermatocytes are arrested at zygotene-like stages, with aberrant axis morphology and SC. In the majority of sycp2^its/its spermatocytes, Sycp3 was observed as aggregates, and Sycp2 was hardly detected, as observed in sycp2^-/- nuclei (Fig 3B-ii to 3B-iv). However, we noticed that a subpopulation of sycp2^its/its spermatocytes showed Sycp1 staining patterns that were intermediate between those of wild-type and sycp2^-/- nuclei, showing more Sycp1 filaments than sycp2^-/- nuclei (Fig 3B-i to 3B-iv, S6 Fig). In a limited population of sycp2^its/its spermatocytes, we also observed nuclei that showed pachytene-like appearances with regard to the staining patterns of Sycp3, Sycp2 and Sycp1 (Fig 3B-v). Since such nuclei were found in only one male out of 4 individual sycp2^its/its males at a rate of 3.8% of Sycp1-positive nuclei (5 among 133 nuclei), this heterogeneity of the sycp2^its/its phenotype among cells and individuals might have been derived from the stochastic nature of a splicing event caused by the its mutation. Thus, the intermediate phenotypes of sycp2^its/its between the wild-type and sycp2^-/- phenotypes indicate that sycp2^its is a hypomorphic allele. Taken together, these results demonstrate that Sycp2 is essential for SC formation in zebrafish.

In wild-type zebrafish spermatocytes, the SC emanates from telomeres (S7 Fig) [26,27]. As Sycp1 filaments were abnormal in number and length in sycp2 mutant zebrafish spermatocytes (Fig 3B and S6 Fig), we next sought to determine whether these aberrant Sycp1 filaments were associated with telomeres using a telomere-targeting polyamide (see Materials and Methods) [37]. For this purpose, the proportions of Sycp1 filaments colocalizing with telomere foci at chromosomal ends were quantitated in wild-type and sycp2 mutant spermatocytes (Fig 4A and 4B). In wild-type spermatocytes, most Sycp1 filaments colocalized with telomere foci at either end in zygonema (Fig 4A-i and 4B: 96%±4.9 (SD)) and at both ends in pachynema (Fig 4A-ii and 4B: 97%±4.1). In contrast, in sycp2^-/- spermatocytes, only ~25% of Sycp1 filaments colocalized with telomeres at chromosomal ends (Fig 4A-iv and 4B). These results indicate that in the absence of Sycp2, the majority of Sycp1 filaments extend at interstitial sites apart from telomeres. Notably, these aberrant Sycp1 filaments were costained with the meiotic cohesin Rad21l1, suggesting that Sycp1 could be recruited onto chromatin in the absence of Sycp2 (S8 Fig). In sycp2^its/its spermatocytes, the number of Sycp1 filaments colocalized with telomeres was reduced (76%±18.5) compared to that in wild-type spermatocytes (Fig 4A-v and 4B), supporting the above assumption that sycp2^its is a hypomorphic allele. Altogether, these data indicate that Sycp2 promotes the localization of Sycp1 at telomeres.

When we examined telomere foci, we noticed that a greater number of telomere foci were present in sycp2 mutant spermatocytes than in wild-type spermatocytes (Fig 4A). Since telomeres are located at both ends of the 25 paired zebrafish homologous chromosomes, 50 telomere foci per nucleus are expected to be observed after the completion of pairing in normal pachytene spermatocytes, whereas the presence of more than 50 is assumed to indicate a pairing defect in zygotene-like spermatocytes in the absence of Sycp2. To examine this idea, the numbers of telomeres were quantified and compared in wild-type, sycp2^its/its, and sycp2^-/- spermatocytes, which exhibit Sycp1 filaments (Fig 4C). In wild-type nuclei, the average number of telomere foci per nucleus was 51±4.8, indicating that almost all 25 chromosome pairs were paired at the chromosomal ends. Nuclei with fewer than 50 telomere foci were at early zygonema, in which telomere foci are observed as a cluster, causing an underestimation of the number of foci. In sycp2^its/its and sycp2^-/- zygotene-like spermatocytes, the average numbers of telomere foci per nucleus were 69±9.0 and 94±16, respectively, and were significantly higher than those in wild-type nuclei (p<0.0001). These results indicate that chromosome pairing is severely impaired in the absence of Sycp2 and that this phenotype is moderate in sycp2^its/its spermatocytes.

Next, we further examined whether pairing at homologous sites is defective in sycp2 mutant spermatocytes by using a BAC probe that recognizes an ~68 kbp region on chromosome 5 [26]. After costaining of Sycp1 with the BAC probe, pairing of BAC foci were examined in Sycp1-positive nuclei (Fig 5). When BAC signals were observed as one or two foci with an intervening distance of <3 μm in a nucleus, they were considered to be paired. In wild-type fish, the majority of nuclei were stained with paired BAC foci (88%; Fig 5A-i and 5B), while 12% of nuclei were stained with two unpaired foci (Fig 5A-ii and 5B). Thus, the BAC-stained locus was mostly paired in wild-type zygotene and pachytene spermatocytes. Strikingly, the proportion of Sycp1-positive nuclei with paired BAC foci was reduced to 51% and 8% in sycp2^its/its and sycp2^-/-spermatocytes, respectively (Fig 5A-iii, 5A-iv and 5B). Instead, the majority of Sycp1-positive nuclei were stained with unpaired BAC foci separated from each other (an intervening distance of >3 μm) in these genotypes (Fig 5A-v, 5A-vi and 5B). Some of such unpaired foci were observed as three or four foci, in a small fraction of sycp2 mutant nuclei (Fig 5A-vii and 5B). Therefore, homologous pairing at the BAC-stained locus appeared to be affected in sycp2 mutant spermatocytes. This phenotype was more pronounced in sycp2^-/- spermatocytes, in which 92% of nuclei were stained with unpaired BAC foci (Fig 5A-vi, 5A-vii and 5B). Altogether, these results showed that homologous pairing is severely impaired in the absence of Sycp2.

Sycp2 composes the chromosomal axis, which is thought to be a site for meiotic DSB formation and repair. To examine DSB formation in zebrafish, we generated a guinea pig antiserum for zebrafish Dmc1 (S5 Fig), an evolutionarily conserved meiotic protein that binds to single-stranded DNA generated at DSB sites [38,12,13]. It should be noted that our anti-Dmc1 antiserum potentially recognizes another recombinase, Rad51, since they share high similarity at the amino acid sequence level (see Materials and Methods). In wild-type spermatocytes, many bright Dmc1/Rad51 foci localized on Sycp3-stained axes in close proximity to telomere foci from leptonema to zygonema (Fig 6A-i to 6A-iii, S9 Fig). To determine whether Dmc1/Rad51 foci were formed in a Spo11-dependent manner in zebrafish spermatocytes, a spo11 knockout zebrafish line was generated by CRISPR-Cas9 mutagenesis (S10 Fig). In contrast to the case in wild-type spermatocytes, Dmc1/Rad51 signal was rarely detected in spo11^-/- spermatocytes at the leptotene-/early zygotene-like stage (Fig 6A-v). These results are consistent with those of a recent study showing that Rad51, a marker of DSBs, forms foci adjacent to telomeres in a Spo11-dependent manner [26].

Next, we examined meiotic DSB formation in sycp2^-/- spermatocytes by Dmc1/Rad51 staining (Fig 6-vi). As sycp2^-/- spermatocytes cannot be staged by Sycp3 patterns, nuclei with short Sycp1 filaments (early zygotene-like) were compared to wild-type nuclei at early zygonema. Remarkably, Dmc1/Rad51 foci were rarely detected in early zygotene-like sycp2^-/- nuclei (Fig 6A-vi). These nuclei were in striking contrast to wild-type nuclei of early zygonema, in which bright signals of Dmc1/Rad51 were observed (Fig 6A-ii). This result is similar to what was observed in spo11^-/- spermatocytes (Fig 6A-v). To further compare the loss of Dmc1/Rad51 foci in sycp2^-/- and spo11^-/- spermatocytes, we quantified Dmc1/Rad51 signals (Fig 6B). Because many Dmc1/Rad51 foci cluster around the telomere bouquet in zebrafish spermatocytes, counting numbers of discrete Dmc1/Rad51 foci is difficult. Thus, the total area stained with Dmc1/Rad51 foci in each nucleus was measured to estimate the numbers of Dmc1/Rad51 foci per spermatocyte. As a result, both sycp2^-/- and spo11^-/- spermatocytes showed a significant decrease in Dmc1/Rad51 signals compared to wild-type spermatocytes at early zygonema (p<0.0001, Fig 6B). In contrast, we could not observe a significant difference between Dmc1/Rad51 signals of sycp2^-/- and spo11^-/- spermatocytes (p = 0.3176). Therefore, the formation of Dmc1/Rad51 foci was impaired in the absence of Sycp2 at a similar level to that in spo11^-/- spermatocytes.

These results suggest that meiotic DSBs are not formed or that these recombinase proteins are unable to localize on DSB sites in the absence of Sycp2. To examine these two possibilities, we examined formation of Dmc1/Rad51 foci in sycp2^-/- spermatocytes, after exogenous induction of DSBs by γ-ray irradiation (S11 Fig). Spermatocyte chromosomal spreads were prepared after 10 Gy γ-ray irradiation of wild-type, spo11^-/- and sycp2^-/- zebrafish, and were stained for Dmc1/Rad51, Sycp3 and Sycp1. Non-irradiated siblings of each genotype were also processed in the same manner as controls. We observed a significant increase of Dmc1/Rad51 signals (average of total area stained by Dmc1/Rad51 foci per nucleus) in irradiated spo11^-/- spermatocytes (0.50±0.64) compared to non-irradiated control spermatocytes (1.39±1.02), although Dmc1/Rad51 signals were not completely recovered to wild-type leptotene or early zygotene-levels. Notably, we observed a similar increase of Dmc1/Rad51 signal in irradiated sycp2^-/- spermatocytes (0.52±0.54) compared to non-irradiated control spermatocytes (1.25±0.68). Since formation of Dmc1/Rad51 foci was partly recovered by exogeneous induction of DSBs, this result supports the idea that Sycp2 is involved in meiotic DSB formation.

We also examined the formation of RPA foci in sycp2^-/- spermatocytes by immunostaining spermatocyte spreads with an anti-human RPA antibody (Fig 7A). RPA is a single-stranded DNA-binding protein that is required for the recruitment of DMC1/RAD51 in mice [39]. In wild-type spermatocytes, RPA foci appeared at leptonema (Fig 7A-i), became bright foci or short stretches in zygonema (Fig 7A-ii and 7iii), and disappeared in pachynema (Fig 7A-iv). However, similar to Dmc1/Rad51 foci, RPA foci were rarely detected in sycp2^-/- spermatocytes (Fig 7A-v and 7A-vi).

We further examined the phosphorylation of the histone H2A variant H2AX (γH2AX), which occurs in response to DNA DSBs, in wild-type and sycp2^-/- spermatocytes (Fig 7B). In wild-type spermatocytes, γH2AX signals appeared from leptonema to early zygonema (Fig 7B-i) and diminished towards pachynema (Fig 7B-ii). The signals were observed as scattered patches, mostly in subnuclear regions where Sycp3 fragments clustered. This observation is consistent with the localization of Dmc1/Rad51 (S9 Fig) and Rad51 [26] on Sycp3-stained axes. Next, we compared γH2AX signals in sycp2^-/- spermatocytes that exhibited short Sycp1 fragments with those in wild-type spermatocytes at early zygonema. Strikingly, the γH2AX signals were drastically reduced in sycp2^-/- spermatocytes compared to wild-type spermatocytes (Fig 7B-iii). Some sycp2^-/- spermatocytes were stained with γH2AX in patches spread over nuclei rather than as clusters in a subnuclear domain.

Taken together, these results indicate that the early steps of meiotic recombination are defective in the absence of Sycp2.

In zebrafish, all juveniles develop gonads with immature oocytes regardless of their definitive sex, and individuals in which these immature oocytes degenerate become males [29–31], and oocyte-depleted mutants develop as males that are mostly infertile [32–35]. For instance, zebrafish homozygous for mutations in Piwi-interacting RNA pathway genes, such as ziwi (zebrafish piwil1) and hen1, are all phenotypically males [32,40]. Zebrafish with homozygous mutations in the fancl and brca2 genes, which code DNA repair proteins, also fail to develop ovaries during sexual differentiation and become sterile males [36,41]. In this study, we observed similar male bias in sycp2^-/- fish (Table 1). This implies that oocytes could be eliminated during gonad development in sycp2^-/- fish. Notably, such male bias is not observed in zebrafish homozygous for spo11 mutations, and mutant females generate eggs capable of fertilization (Table 1) [26]. Interestingly, in mice, both Sycp2 and Spo11 mutant females show less severe phenotypes than those in mutant males [22,42]. In addition, we observed previtellogenic dictyate oocytes in four of the five sycp2 mutants at 28 days postfertilization (S4 Fig). Oocytes of sycp2 mutants seem to develop slightly later than those of fancl and brca2 mutants [36,41]. Meiotic checkpoint and/or cell elimination mechanisms may be different between mice and zebrafish and among the mutated genes, as well as between sexes. The checkpoint may be rather active in oogenesis in sycp2 mutant zebrafish than in spo11 mutants. Since p53-mediated germ cell apoptosis is known to induce sex reversal by compromising the survival of zebrafish oocytes [36], a future interest will be to identify differences in such checkpoint activation among these meiotic mutant zebrafish.

In this study, we developed anti-Sycp2 antibodies to understand the spatiotemporal processes of all SC components in wild-type and sycp2 mutant spermatocytes (see a model in S12 Fig). Interestingly, Sycp2 appeared to be essential for the integration of Sycp3 into axial elements, since Sycp3 persistently formed aggregates in sycp2 mutant spermatocytes that were reminiscent of those observed in wild-type preleptonema, when Sycp2 had yet to form the axis (Fig 3). In mice, SYCP2 and SYCP3 form antiparallel heterotetramers [19], and the interaction of the C-terminal coiled-coil domains of SYCP2 and SYCP3 is required for the axis association of SYCP3 [22]. Thus, the SYCP2-SYCP3 interaction seems to play a conserved role in the formation of axial elements in vertebrates. We observed abnormal formation and localization of Sycp1 filaments in sycp2 mutant spermatocytes (Fig 3 and Fig 4), indicating that Sycp2 is also important for proper loading of Sycp1 onto meiotic chromatin in zebrafish, as reported in Sycp2 mutant mice [22]. This phenotype was different from spo11 knockout zebrafish, in which Sycp1 filaments were not observed in the majority of spermatocytes [26]. The colocalization of Sycp1 filaments and paired BAC foci was observed in a few sycp2^-/- spermatocytes (Fig 6A-iv and 6B). Therefore, at least some aberrant Sycp1 filaments seem to localize on such partly paired regions of homologous chromosomes in the absence of Sycp2. We also observed costaining of aberrant Sycp1 filaments in sycp2^-/- spermatocytes with stretches of Rad21l1 that seem to exist, at least in part, even in the absence of Sycp2 (S8 Fig). It was difficult to determine whether these Sycp1 filaments were loaded between sister chromatids, at our image resolution. In the case that the aberrant Sycp1 filaments were formed between sister chromatids, as observed in a hypomorphic Stag3 mutant mouse, in which the levels of the cohesin subunit REC8 are partly reduced [43], Sycp2 might function to prevent formation of the synaptonemal complex between sister chromatids.

Meiotic chromosomal axes have been thought to serve as scaffolds for meiotic DSB proteins from observations in Saccharomyces cerevisiae. In S. cerevisiae, mutations of the axis proteins Hop1 and Red1 severely impair the formation and repair of meiotic DSBs [2–6]. In addition, Red1 functions in homologous pairing, likely by potentiating Hop1-promoted pairing [44,45]. In mammals, HORMAD1 is likely to be a counterpart of budding yeast Hop1 since it localizes on the axis prior to DSB formation and since meiotic DSBs are reduced in Hormad1 knockout mice [10,46–48]. The recent work of West et al. noted that Red1 and SYCP2/SYCP3 share conserved biochemical properties as axis core proteins and that an N-terminal domain of mouse SYCP2 binds to the HORMA domain [19]. The putative HORMA-binding domain is also conserved at the sequence level in zebrafish Sycp2 (S13 Fig). In this study, we generated sycp2 null mutant zebrafish and showed that Sycp2 is required for Dmc1/Rad51 and RPA focus formation and γH2AX signals during meiotic prophase I (Fig 6 and Fig 7). A partial recovery of Dmc1/Rad51 focus formation after induction of exogenous DSBs implies that Sycp2 is required for meiotic DSB formation (S11 Fig). Furthermore, chromosome pairing at telomeres and a homologous locus were severely impaired in the absence of Sycp2 (Fig 4 and Fig 5). Our observation that DSB marker signals and homologous pairing were largely diminished in sycp2 mutant zebrafish confirmed in vivo that Sycp2 has critical functions in early meiotic recombination, as reported for budding yeast Red1. Sycp2 is a large protein that could potentially work as a scaffold for meiotic proteins. Further identification of Sycp2 interactors will provide information to understand how Sycp2 coordinates multiple meiotic events, such as telomere clustering, DSB formation and repair, and synapsis.

In this study, we showed the localization of Dmc1/Rad51 foci near telomeres in zebrafish spermatocytes (Fig 6, S9 Fig). This observation is consistent with previous observations of Rad51 localization, indicating that DSB formation occurs near telomeres in zebrafish spermatocytes [26,49]. Interestingly, meiotic DSBs occur frequently in human males at genomic regions close to telomeres [50]. In many mammals, including rodents and primates, the meiosis-specific histone methyltransferase PRDM9 is known as a major determinant of hotspots [51–55]. Since PRDM9 binds to a specific DNA motif, its DNA-binding property is thought to play a key role in hotspot determination. Since zebrafish do not have a functional prdm9 ortholog [56], it is possible that the telomere-proximal bias of DSBs rather than a PRDM9-mediated mechanism primarily underlies hotspots. Our observations that DSB signals were largely diminished in the absence of Sycp2 (Fig 6 and Fig 7) suggest that the formation of the axis structure could play a key role in the regulation of meiotic DSBs. Therefore, zebrafish will be an ideal model with which to approach this new aspect of DSB control based on chromatin structures at subtelomeric regions in human males.

All experiments were conducted in accordance with Rules for Animal Experiments at the National Institute of Genetics, Research Organization of Information and Systems. The experimental plan using zebrafish is approved by the National Institute of Genetics official ethics committee (Approval Number 27–12, 28–13, 29–13, 30–14, 31–18).

Zebrafish (Danio rerio) were maintained under standard conditions as described in The Zebrafish Book [57]. The wild-type India line (see ZFIN at http://zfin.org/action/genotype/view/ZDB-GENO-980210-28) has been maintained in our laboratory for more than 20 generations. The AB* line (see ZFIN at http://zfin.org/action/genotype/view/ZDB-GENO-960809-7) was kindly provided by Prof. U. Strähle in 2010 and has been maintained in our laboratory. The its fish have been described previously: its mutant fish were derived from the wild-type Tübingen line and then crossed with the IM strain (wild-type inbred) [58] to obtain heterozygous mutant fish [25].

Knockout lines of sycp2 were generated by TALEN mutagenesis targeting exon 18 with the following sequences: 5’-TGATGAAGCCAGCTCT-3’ (left TALEN) and 5’-GTGTCTCTCCTTCTTT-3’ (right TALEN). TALEN modules targeting these sequences were cloned into pCS2TAL3DD and pCS2TAL3RR vectors using the Golden Gate assembly method [59,60]. TALEN mRNAs were prepared with an mMESSAGE mMACHINE SP6 kit (Ambion). Injection with ~1 ng of mRNA was performed on wild-type India embryos at the 1- to 4-cell stages. Founder males were mated with India females, and the F₁ siblings were screened by genotyping. Homozygous sycp2 knockout fish were obtained by mating F₁ siblings carrying either 14-bp or 16-bp deletions at the TALEN target site (S1 Fig).

The spo11 knockout line was generated by CRISPR-Cas9 mutagenesis based on published protocols [61,62]. Template DNA for single-guide RNA (sgRNA) synthesis was prepared by amplification with a primer specific to spo11 exon 5, a universal reverse primer (S3 Table) and T4 DNA polymerase. After purification of the template DNA, sgRNA was transcribed in vitro with a MEGAscript T7 kit (Ambion) and purified with a MEGA clean-up kit (Ambion). Wild-type AB* embryos were injected at the 1- or 2-cell stage with 2.3 nl of a mixture of 10 pmol/μl Cas9 NLS protein (abm) and 100 ng/μl spo11 sgRNA. Founders were backcrossed with AB* fish, and the F₁ siblings were screened by genotyping. Homozygous spo11 knockout fish were obtained by mating F₁ siblings carrying a +1 frameshift mutation in exon 5 (S10 Fig).

The fish were genotyped after extracting genomic DNA from caudal fin clips. PCR was performed with GoTaq Green Master Mix (Promega) using specific primers for each site (S3 Table). The PCR products were examined by either agarose or acrylamide gel electrophoresis. Sequencing of the genomic sycp2 locus was performed by amplifying exons 8 to 11 of sycp2 with KOD -Multi & Epi- (Toyobo). Some PCR products were treated with ExoSAP-IT PCR Product Cleanup Reagent and were analyzed by general Sanger sequencing. Other PCR products were cloned into the pGEM-T vector and then sequenced with T7 or SP6 primers.

For mapping of the its mutation, 682 sterile its mutant fish were genotyped as described above using primers to detect SSLPs between the Tübingen and the IM lines, and the causal genomic region was mapped between the z20895 and z7550 SSLP markers.

To assess the fertility of sycp2^-/- zebrafish, six sycp2^-/- and four wild-type siblings were individually mated with wild-type females. For the complementation test, five sycp2^its/- siblings, one sycp2^IM/- sibling and one wild-type sibling were individually mated with wild-type females. After the eggs were collected, all eggs were incubated at 28°C for 6 hours, and the gastrulae and unfertilized eggs were counted.

To examine sexual phenotypes of sycp2^+/- or spo11^+/- mutant fish, offspring were obtained from mass mating of heterozygote fish. Offspring fish were dissected at 8 to 9-week old, and the morphology of the gonads was examined with a Nikon ECLIPSE TE200-S microscope. To facilitate the observation of gonad morphology, dissected gonads were placed on a slide with a drop of Leibovitz L-15 media (Sigma) and flattened with a coverslip. Gonads were classified into either ovaries with oocytes or testes with lobule structures and/or with spermatocytes, spermatids, or sperm cells (S14 Fig). Fin clips of offspring fish were used for genotyping.

Total RNA was extracted from fin clips or paired testes with RNAiso Plus (Takara) and treated with TURBO DNase (Ambion). Testis cDNA was prepared from 1 μg of total RNA with a PrimeScript RT-PCR kit (Takara) using oligo dT primers. qPCR was performed with LightCycler 480 SYBR Green I Master on a LightCycler 480 system (Roche) with the default settings, except that annealing was performed at 58°C. For the cloning of sycp2 cDNA and the analysis of sycp2 splicing at exons 8–9, PCR was performed using KOD -Multi & Epi- (Toyobo) on cDNA from five testes each of sycp2^+/+ and sycp2^its/its males. All primers used in this study are listed in S3 Table.

Adult testes and bodies of juvenile fish at 28 days postfertilization were fixed in Bouin's solution and 4% paraformaldehyde in PBS, respectively, at 4°C overnight. The fixed samples were dehydrated in an ethanol series (70%, 80%, 90% and 100%), methyl benzoate, and Lemosol (Wako) and embedded in paraffin. Hematoxylin-eosin (HE) staining was performed with 5-μm-thick sections.

Two mini-gene constructs containing either wild-type or its sycp2 intron 8 with flanking exon 8 and 9 sequences were constructed (S3 Fig). The genomic sycp2 sequence spanning exons 8 to 9 was amplified from wild-type and its homozygote genomic DNA by PCR using primers with restriction sites (S3 Table). The PCR products were cloned into pT2AL200R150G [63] using HindIII and BamHI sites. The purified plasmids and transposase mRNA were injected into wild-type embryos at the 1- to 4-cell stages as previously reported [63]. After injection, the expression of each mini-gene construct was verified by fluorescence signals and by RT-PCR of the reporter EGFP (S3 Fig).

The antibodies used in this study are listed in S2 Table. Polyclonal antibodies specific for zebrafish Sycp2,Dmc1/Rad51 and Rad21l1 were generated in this study (S5 Fig, S8 Fig). sycp2 cDNA coding amino acid residues 1071–1569 (C499) or 1238–1569 (C332) was cloned into the pET-21a vector. The recombinant Sycp2 proteins were expressed in Escherichia coli Rosetta-gami 2 (DE3) and purified by Ni-NTA agarose (Qiagen). The purified proteins were separated on a 10% acrylamide gel, and a band corresponding to each Sycp2 recombinant protein was cut out after Coomassie brilliant blue staining. The gel sections were washed in water and homogenized in PBS. After isolation of the Sycp2 recombinant proteins, 500 μg each of the recombinant proteins C499 and C332 was used to immunize guinea pigs and rats, respectively (Evebioscience). After the fifth injection of antigen, guinea pig and rat anti-Sycp2 antisera were recovered and affinity-purified using CNBr Sepharose beads (GE Healthcare) conjugated with the recombinant Sycp2 protein C499. To generate anti-Dmc1 antiserum, a Dmc1 cDNA coding amino acid residues 8–220 (NP_001018618.1) was cloned into the pQE-30 vector (Qiagen). The recombinant Dmc1 protein was expressed in E. coli M13 and purified by Ni-NTA agarose (Qiagen). The purified Dmc1 proteins were used to immunize guinea pigs (BioGate). Unpurified anti-Dmc1 antiserum was used in this study. It should be noted that the protein sequence of the Dmc1 antigen is 43% identical to the zebrafish Rad51 protein. Thus, our anti-Dmc1 antiserum potentially recognizes Rad51. An anti-Rad21l1 antiserum was generated by immunizing mice with the full-length zebrafish Rad21l1 protein. The full-length cDNA of Rad21l1 was amplified from AB* testis cDNAs and cloned into the pET-28a vector using the NEBuilder HiFi DNA Assembly (NEB).

Spermatocyte chromosomal spreads were prepared from adult testes by a dry-down method [64] adapted for zebrafish. Briefly, a pair of testes was dissected from anesthetized fish and cut into small pieces in PBS. The cells were dissociated by gentle pipetting, and the cell suspension was transferred into a new tube, leaving behind any nondissociated clumps. After washing in PBS, the cells were incubated for 8 min in hypotonic buffer (30 mM Tris-HCl pH 8.2, 50 mM sucrose, 17 mM sodium citrate, 5 mM EDTA, 0.5 mM dithiothreitol, 1x cOmplete Protease Inhibitor Cocktail (Roche)). After washing in PBS, the nuclear pellets were recovered in PBS. The nuclear suspension was mixed with a 2x volume of 100 mM sucrose, and ~10⁵ nuclei were spread on a slide dipped in a solution of PBS, 1% paraformaldehyde, and 0.05% Triton X-100. The spread slides were incubated in a humid chamber for 1 hour and then half-dried. After washing twice in 0.1% Tween 20, the slides were kept at -80°C or immediately used for immunostaining.

Blocking of the spread slides was performed with a solution of PBS, 5% skimmed milk, and 5% goat and/or donkey serum. The slides were incubated at room temperature overnight with primary antibodies and at 37°C for 1 hour with secondary antibodies. All antibodies and their dilutions at the time of use are listed in S4 Table. After the last washing step in PBS containing 10 ng/ml DAPI, the slides were mounted using VECTASHIELD Antifade Mounting Medium (Vector Labs). For telomere staining, TEN buffer (100 mM NaCl, 20 mM Tris-HCl pH 7.4, 1 mM EDTA) was used instead of PBS, and incubation with secondary antibodies was performed with 0.5 μg/ml DAPI and 15 nM TRed-HPTH59-A (telomere-targeting polyamide; HiPep Laboratories) [37] at room temperature for 3–4 hours. Note that TE buffer (20 mM Tris-HCl pH 7.4, 1 mM EDTA) was used for the wash step before incubation with the secondary antibody/DAPI/telomere-targeting polyamide mix, as salt prevents binding of the telomere-targeting polyamide.

BAC probe DNA and the Cot-1 DNA were prepared as previously described using BAC clone CH211-31P3 and salmon sperm DNA (Wako), respectively [26]. The BAC probe staining protocol was adapted from [26]. First, immunostaining of chromosomal spread was performed as described above. After immunostaining, the slides were postfixed with 400 μl of 1% paraformaldehyde in PBS for 15 min at room temperature. The slides were then washed in PBS and dehydrated by placing in 70%, 85% and twice in 100% ethanol. Dried slides were incubated with 10 μl of probe mixed with 25 μg of Cot-1 at 73°C for 3 min. After coming to temperature at ~40°C, slides were incubated in a humid chamber at 37°C overnight. Washing steps were performed as described previously [26]. The distance between BAC foci was quantified as previously described [26]. Two foci with an intervening distance of <3 μm were considered to be paired.

Testes from 8–14 fish were homogenized in 5 ml of RIPA buffer. After freezing at -80°C and sonication, the soluble fraction was recovered by centrifugation at 20,000 x g for 15 min at 4°C. The concentrations of the protein extracts were determined with a Protein Assay Lowry Kit (Nacalai). Immunoprecipitation was performed from ~2.5 mg of testis proteins of each genotype with either 25 μg of guinea pig anti-Sycp2 antibody or 25 μg of normal guinea pig IgG. The antigen-antibody complex was recovered with 50 μμl of protein A Sepharose beads (GE Healthcare). The immunoprecipitated samples were eluted in 40 μl of 2x sample buffer at 95°C for 10 min and migrated on a 7.5% Mini-PROTEAN TGX precast protein gel (Bio-Rad) at 200 V for 30 min. After transfer to an Immobilon-P membrane (Millipore) at 100 V for 70 min, immunoblotting was performed with a rat anti-Sycp2 antibody and a goat anti-rat HRP-conjugated antibody (Santacruz) using Can Get Signal Immunoreaction Enhancer Solution (Toyobo). The signals were developed with Western BLoT Ultra Sensitive HRP Substrate (Takara) and captured with an ImageQuant LAS 4000 Mini (GE Healthcare).

Histological images were captured with an Olympus BX51 microscope equipped with a Keyence VB7010 camera. Images of the embryos were captured with a Leica MX16 FA fluorescence microscope equipped with a Leica DFC310 FX camera. Cytological images were captured with a DeltaVision PersonalDV-TM fluorescence microscope with softWoRx software (GE Healthcare). All cytological images were processed using OMERO (OME) [65] and Fiji [66]. For telomere counting (Fig 4) and Dmc1/Rad51 imaging (Fig 6, S9 Fig and S10 Fig), stacks of 10 images along the z-axis (section spacing: 0.2 μm) were deconvoluted with softWoRx and projected with Fiji. The Sycp1 fragments and their colocalization with telomere foci stained by the telomere-targeting polyamide were manually counted in each nucleus. Punctate Sycp1 signals were not counted as Sycp1 fragments. To calculate the Dmc1/Rad51 focus-stained area from binary images, the same threshold was applied to all images in the same experiment to eliminate background signals. All quantifications were performed by a Fiji macro with manually selected DAPI-positive ROIs (available upon request).

Adult zebrafish were placed in a plastic box, with water depth of 2 cm. Animals were exposed to γ-rays for 11 minutes 42 seconds at 0.855 Gy/min (total exposure 10 Gy) in a Cs-137 Gammacell 40 Exactor (MDS Nordion). Irradiated fish were anesthetized 30 minutes after irradiation, and spermatocyte chromosomal spreads were prepared as indicated above. For each genotype, non-irradiated siblings were processed as controls.