The differentially expressed genes regulated by Gsdf signaling
To identify the targeted genes responsible for Gsdf signaling before the initiation of spermatogenesis, RNA-seq was performed (transcriptome sequencing data were deposited in NCBI under accession number PRJNA759752).The differentially expressed genes (DEGs) were evaluated by comparing one-month-old gsdf+/+ and gsdf-/- gonads before the XY germ cells resume proliferation (Saito et al., 2007, Kanamori, 1985), with the threshold of a log2 fold change (FC) ≥ 3 and a P-value of false discovery rate (FDR) < 0.05 as measures of statistical significance. FKB506-binding protein 7 (fkbp7), a molecular chaperone located in the endoplasmic reticulum (ER) as a suppressor of ATPase activity (Garrido et al., 2019), was remarkably upregulated (log2 FC ≥ 3), and overlapped among four pair-wise comparisons of normal testes to normal ovaries (WTT vs WTO); gsdf-deficient ovaries to normal ovaries (HmO vs WTO); gsdf-deficient testes to normal testes (HmT vs WTT); and gsdf-deficient testes to gsdf-deficient ovaries (HmT vs HmO) as revealed by Venn diagram (Fig. 1a). The details are listed in Table S1. RNA transcripts of follicle-stimulating hormone receptor (fshr), microRNA 202 (miR202), a zinc-finger protein with a C-terminal domain of four cysteine and two histidine residues (zc4h2), and an immunoglobulin-like domain factor were significantly upregulated in a subset of three comparisons: WTT vs WTO, HmO vs WTO, and HmT vs HmO.
To assess the concordance of RNA-seq data on expression quantification, the relative transcripts per kilobase expression of exon model per million mapped reads (mean TPMs) and log2MeanTPMs of sexually differentiated DEGs shown in Fig. 1b, were verified by real-time PCR quantification (qPCR) (Fig. 1c, details in Table S2 and Table S3). The sexually dimorphic expression of fshr, sycp1, sycp3, and miR202 was consistent with previous reports (Gay et al., 2018, Aoki et al., 2009, Herpin et al., 2007), and was higher in the testis than in the ovary as verified by qPCR (Fig. 1C, Table S3), with primers listed in Table S4. The log2FC of sycp3 (HmO vs WTO comparison) was 2.07, whereas that of sycp1 (HmO vs WTO) was 4.3, indicating that the expression of sycp1 was more sensitive to the Gsdf signal (Fig. 1b, Table S1). The expression of sycp1 in the ovary was higher than that in the testis, resembling to the expression patterns of dynll2 and nanos3 (Fig. 1b). Nevertheless, the expression of sycp1 increased significantly in one-month-old gsdf-intact and gsdf-deficient testes (Fig. 1c), indicating the high proportion of sycp1-expressing germ cells in testes before the initiation of spermatogenesis. The expression of sycp3 was higher in normal or gsdf-deficient testes than in normal or gsdf deficient ovaries (Fig. 1b), resembling that of fkbp7 and dnaaf1 (dynein, axonemal, and assembly factor), and the structural maintenance of chromosome 1b (smc1b). nanos2 and nanos3 are related to stem-cell-like characteristics and are differentially expressed in medaka germ cells (Aoki et al., 2009). Notably, an increase of nanos3 was detected in both gsdf deficient ovaries and testes, compared to their normal counterpart gonads. In contrast to no difference between normal ovaries and gsdf deficient ovaries, nanos2 expression significantly increased in gsdf-deficient testes compared to normal testes (Fig. 1c). These results showed that the population of nanos2-expressing germ cells were sensitive to Gsdf and were mainly located in the testis, which was different from the population subsets of nanos3-expressing germ cells responsive to Gsdf in both the ovary and testis (Fig. 1c) (Zhang et al., 2021).
Meiotic recombination targeted by Gsdf signal
To assess the targeted transcriptomes and the related pathways responsive to Gsdf, Kyoto Encyclopedia of Genes and Genome (KEGG) enrichment analysis was performed to investigate common DEGs among three overlapped pair-wise comparisons (41 upregulated DEGs), or two overlapped pair-wise comparisons (44 downregulated DEGs), as revealed by Venn diagram (Fig. 1A). The KEGG-enriched pathways are displayed in Fig. 2A in a bubble map, while the full details are listed in Table S5. The peroxisome proliferator-activated receptor (PPAR) pathway is well characterized and activated by TGFβ signals (Pleniceanu et al., 2017). Glycerolipid metabolism, ECM-receptor interaction, cell adhesion molecules, and tight junction pathways were downregulated (Fig. 2A, right panel), which is consistent with the involvement of fatty acid synthesis in female-to-male sex reversal (Sakae et al., 2020). The pathway of DNA homologous recombination (HR) consisting of sycp3 (localized in chromosome18, Chr18) and sycp3 like (sycp3l) (localized in chromosome11, chr11) was significantly upregulated (Fig. 1a and Fig. 2A) (Pasquier et al., 2016). RNA alternative splicing (AS) variants of sycp3 lacking N-terminus transmembrane domains remarkably increased in gsdf-deficient gonads, which was confirmed by RT-PCR with primers flanking the region of sycp3 exon 2 and 5’ upstream adjacent phka1a (Fig. 2B, the right corner). The super-coil structure of normal Sycp3 was putatively replaced by the linear structure without N-terminus AS variants predicted by Phyre2 software (Fig. 2B, right). A high frequency of AS junction tracks was found in sycp3l and sycp1 region as shown by IGV Sashimi plot. This suggests that the active splicing machinery may promote the translation of defective Sycp1 and Sycp3l proteins, thus further interfering with DNA repair and meiotic HR events.
Malformation of Balbiani body with abnormal expression of Sycp3-Piwi in gsdf deficient oocytes
Piwi, the major component of the meiotic nuage to silence the transposon activity during meiotic HR (Yashiro et al., 2018), was expressed predominantly in mitotic spermatogonia and oogonia as well as meiotic spermatocytes and oocytes, as shown by immunofluorescence analysis (Fig. 3). The anti-Sycp3 positive signals were detected in meiotic spermatocytes and oocytes (Fig. 3a–a’; Fig. 3b–b’). Strong anti-Sycp3 signals were detected in normal oocyte Balbiani bodis (Bbs), although they declined significantly in the small Bbs of gsdf-deficient XX and XY ovaries (Fig. 3c–c’; Fig. 3d–d’). The low level of Sycp3 protein may have been due to the blocked translation of defective sycp3 itself as a variant or the instability of protein products in gsdf deficient oocytes (Fig. 3c-3d). Compared with normal oocytes (stage III oocytes in Fig. 3b), gsdf deficient oocytes have fewer Piwi nuages in the cytoplasm (stage III oocytes in Fig. 3d), suggesting that the special transposons may become active without being inhibited, thereby promoting DSBs or HR events in gsdf−/− oocytes.
Increase of HR errors in gsdf-deficient testes
Proliferative germ cells that undergo cyst-division are different between XY male and XX female (Saito et al., 2007), and are distinguishable by EdU (a thymidine analog) incorporation, while gonads without EdU treatment serve as a negative control (Figure S1). Although the spotted bars corresponding to chromosomes were obviously different between females (unevenly distributed in oogonia in Figure S2a1–a3) and males (evenly distributed in spermatogonia in Figure S2b1–b2), some spermatogonia were similar to oogonia in shape and pattern (SpgB2a in Figure S2b2–b2’ vs Og in Figure S2a3), but different from cystic spermatogonia (SpgB2b in Figure S2b1’). The identical shape may have been derived from the same cell cycle stage, similar DSBs, or HR events shared by these germ cells. Like humans (Capalbo et al., 2017, Gruhn et al., 2013), medaka germ cells in normal spermatogenesis undergo HR errors occasionally (e.g., non-exchange chromosomes or achiasmate chromosomes) to become oocyte-like (Ol) cells (Figure S2c). The evidence of a large number of Ol cells in Sissy gonads supports the hypothesis that impaired Gsdf signaling leads to high-frequency HR errors during meiosis (Zhang et al., 2021).
The effect of Gsdf on meiosis was further investigated by EdU incorporation. According to the ratio of SpgB2 per total SpgB number for three testes, each with 14 count fields EdU-positive, SpgB2a (ccOg alike) was double in gsdf-deficient testis (19.24%) compared to normal testis (10%) (Figure S3A–B). The counting results are summarized in Table S6. The separation of sister germ cells and cystic dividing cells are shown in Fig. 4.
Asynchronous cysts formation in gsdf-deficient testes
TEM observations showed that synchronous cysts containing cystic-proliferative spermatocytes (Spc) (Fig. 5a1), spermatids (Spt) and sperm (Spm) (Fig. 5a2), round and elongated spermatids (Fig. 5a3− a4), and densely packed sperm cells (Fig. 6a5-6a6) were present in the cross sections of normal testes (n = 3). However, many asynchronous cysts appeared in gsdf deficient XY testes, which showed that germ cells at different developmental stages were in the same cyst (Fig. 5b1-b4, n = 3) (Zhang et al., 2016). Spermatocytes (Spt), mitotic metaphase SpgB and sperm were present in a single cyst (Fig. 5b1); dead Sertoli cells (*Sc) and ruptured cysts (Fig. 5b2); round spermatids with loosened packaging of genomic DNA (Fig. 5b3), and one cyst containing spermatozoa with irregular packaging of genomic DNA into tiny sperm heads (low magnification in Fig. 5b4, and high magnification in Fig. 5b4’-b4”) were also found, supporting the theory that Piwi abnormality leads to the destruction of histone-protamine exchange during spermiogenesis (Gou et al., 2017).
Aneuploidy and low fertility of gsdf-deficient gametes
Meiosis in normal and gsdf deficient gametes was investigated using a cytological approach (Giemsa staining, Fig. 6). A total of 1312 prophase cells from normal and gsdf−/− testicular or ovarian biopsies were examined. The details are listed in Table S8. The defects were obvious as early as during the transition between leptotene/zygotene stages; instead of a bouquet configuration (Fig. 6a1 − 2), loosened chromosome loops were observed in gsdf-deficient testes (arrows in Fig. 6a3, enlarged in 6a3’) and ovaries (arrows in Fig. 6a7–a7’). Aneuploid chromosomes (n = 25) were found in gsdf-deficient spermatocytes (Fig. 6a4, a5), in contrast to normal chromosomes in gsdf-intact XY spermatocytes (n = 24 shown in Fig. 6a2). Incomplete synapsis formation may ultimately lead to pachytene prolongation and the failure of pachytene completion. The lack of Gsdf destroys the proportion of normal prophase stages through the early leptotene/zygotene transition, and prolongs the pachytene of gsdf−/− testis and ovary, suggesting that Gsdf may affect the initiation of meiosis and the formation of SC. The proportional distribution of prophase stages shown in Fig. 6A8.
The qualities of gsdf−/− gametes were evaluated by breeding experiments. The fertilization rate of normal XY or transgenic gsdf XX sperm and normal XX oocytes ranged from 94.4–98.9% (Table 1, Family 9–12), whereas the fertilization rates of gsdf deficient sperm (Table 1, Family 1–3) and eggs (Table 1, Family 4–7) ranged from 56.6–71.9% and 41.7–76.4%, respectively. These results indicated that aneuploid chromosomes in gsdf-deficient gamete led to a serious decline in the fertility of oocyte and sperm. The fertilization rate of gsdf-deficient XX oocytes and Sissy gsdf−/− sperm were the lowest at 39.3% and 37.5%, respectively, as shown in Table 1 (Family 8). The embryogenesis was severely damaged and delayed from one-cell fertilization throughout the entirety of embryogenesis. Two-cell division was delayed and asymmetric, and embryo development was severely hindered (Fig. 6b–d). In contrast to normal embryos, which reached stage 18 (St18 late neurula stage, 26 h after fertilization) according to the embryogenesis stages (Iwamatsu, 2004), the offspring embryos of gsdf-deficient males and normal females were delayed in St13 (early and middle gastrula) (Fig. 6d). This result indicated that the embryogenesis was damaged at the beginning of development, which may affect the development of multiple tissues. To our surprise, progeny produced from mating families between gsdf-deficient XY males (n = 3) and normal XX females developed as all XY males, while the offspring of gsdf−/− XY females (n = 4) mated with gsdf transgenic XX males were XX, with seven males and 49 females derived from three mating families. Fry fish of Family 7 (gsdf-deficient XY female bred with Tg XX male) were dissected and dmy-genotyped at hatching and showed the ratio of four XY males to eight XX females (Table 1). These results suggest that XX and XY germ cells respond differently to Gsdf signaling. gsdf-deficient X-bearing oocytes and Y-bearing sperm have more survival advantages during the processes of oogenesis and spermatogenesis, respectively.
In summary, we describe the sex chromosomes susceptible to HR errors in gsdf−/− gametogenesis from the aspects of X-Y non-disjunction and aneuploidy abnormalities. The loss of gsdf leads to defective gametogenesis, stimulates FKBP7-mTOR activity and promotes female cytoskeleton assembly through the downstream robust mRNA transcripts and splicing variants. Nevertheless, gsdf deficiency does not block all male cascades. Aberrant expression of sycp1 and sycp3 eventually leads to mis-segregation between homologous chromosomes (Fig. 6a3–a3’), which is likely to result in the occurrence of aneuploidy gametes caused by the non-disjunction of X-Y chromosomes, This can also explain the low fertility of gsdf−/− gametes and the sex bias of gsdf deletion offspring.