BMP–Smad1/9 signaling plays a critical role in regulating zebrafish PGC proliferation

Abstract

The germ cell fate in zebrafish is determined by germ plasm, whereas mammalian germ cell fate is induced by bone morphogenetic protein (BMP) signaling. It remains elusive whether BMP signaling is implicated in zebrafish germ cell development. Here, we demonstrate that BMP–Smad1/9 signaling plays a critical role in zebrafish primordial germ cell (PGC) maintenance rather than fate determination. BMP inhibition or smad1/9 knockdown reduces PGC numbers. Furthermore, we generated PGC-specific smad1/9 knockouts using a transgenic approach with PGC-specifically expressed Cas9 and ubiquitously expressed guide RNAs. Smad1/9 deficiency in PGCs leads to impaired PGC proliferation and increased apoptosis, consequently reducing PGC numbers. Transcriptome analysis revealed unchanged PGC-specific gene expression, but a marked upregulation of DNA damage response-related genes, which is validated by ectopic ATR–pChk1 activation in PGCs and PGC restoring by ATR inhibition. Collectively, these findings underscore conserved but functionally distinct roles of BMP signaling in vertebrate PGC development.

BMP–Smad signaling is essential for mammalian primordial germ cell (PGC) fate specification, but whether BMP signaling functions in the zebrafish germline remains elusive. Here they show that BMP signaling is required for PGC proliferation but not fate specification.

Introduction

The formation and transmission of germ cells are crucial for species survival and evolutionary continuity. In mammals, PGCs are induced from a subset of proximal epiblast cells around the onset of gastrulation by BMP signals from the adjacent extraembryonic ectoderm and visceral endoderm¹. Disruption of BMP signaling components, including Bmp2, Bmp4, Bmp8b, Smad1, Smad4, and Smad5, results in the complete loss or significant reduction of PGCs^1–3. Notably, BMP signaling has also been implicated in PGC specification in other animals, such as the hemichordate Ptychodera flava and the basal insect Gryllus bimaculatus, suggesting that BMP-mediated PGC induction may represent an evolutionarily conserved mechanism among metazoans^4,5. In contrast, PGC fate in some non-mammals, including Caenorhabditis elegans, Drosophila melanogaster, Xenopus laevis, and Danio rerio, is pre-determined by maternally inherited germ plasm containing mRNAs and proteins, such as vasa, ziwi, dazl^6–8. In zebrafish, four pre-PGC clusters (PGC precursors), which inherit sufficient amounts of germ plasm, are located separately in the early blastula (1k-cell stage)⁶. After specification, PGCs migrate toward the gonadal region while undergoing proliferation⁹. However, whether BMP signaling is essential for zebrafish PGC development remains unexplored.

The BMP signaling pathway is a crucial regulator of various biological processes, including development, cell differentiation, tissue homeostasis, cell apoptosis, stress responses, and disease progression^10,11. Activation of BMP–Smad signaling triggers phosphorylation of downstream effectors Smad1, Smad5, and Smad9 (also known as Smad8), which form a complex with Smad4, translocate into the nucleus, and promote or repress gene transcription in a context-dependent manner¹². BMP–Smad signaling is widely known to function as a morphogen with the gradient distribution along the ventral–dorsal axis during early zebrafish embryogenesis, promoting ventral cell fates^13–15. Previous studies found that genetic disruptions of BMP ligands, such as bmp2b and bmp7, receptor alk8, and downstream effector smad5, result in dorsalized phenotypes¹¹. In addition, the non-canonical BMP signaling pathway, also named the Smad-independent pathway, has been reported to regulate some developmental processes through MAPK, PI3K–AKT, Rho–GTPase, or Wnt pathways¹⁵.

In this study, we show that BMP–Smad signaling is activated in zebrafish PGCs during early embryogenesis and plays a critical role in maintaining the PGC population. Inhibition of BMP signaling or whole-embryo knockdown of smad1/9 reduced PGC numbers. Germline-specific knockout of smad1 or smad9 did not disrupt PGC specification or migration but impaired PGC proliferation and increased apoptosis, likely due to DNA damage response and cell cycle defects.

Results

BMP–Smad signal inhibition induces PGC developmental defects

To determine whether the BMP–Smad signaling pathway plays a role in PGC development, we first examined the expression levels of BMP–Smad factors in PGCs during embryogenesis based on published PGC RNA-seq datasets^16,17. smad5 is highly expressed in PGCs from the high stage to 24 h post-fertilization (hpf), while smad1 and smad9 show low expression before the dome stage and are then elevated (Fig. 1a). smad4 is also expressed throughout the developmental stages while smad1/5/9 are present before 24 hpf (Fig. 1a). Next, we detected the level of phosphorylated Smad1/5/9 (pSmad1/5/9), an activated form of Smad1/5/9, in PGCs by immunostaining. We observed higher levels of pSmad1/5/9 signals in PGCs on the ventral side compared with those on the dorsal side from the dome to the epiboly stages (Fig. 1b, c), consistent with the ventral-to-dorsal gradient of BMP signaling. Subsequently, PGCs migrate dorsally and form two lateral PGC clusters at the somite stages, which exhibited comparable pSmad1/5/9 signals with surrounding somatic cells within embryos (Fig. 1b, c), indicating the existence of an active BMP–Smad signaling in PGCs during early embryogenesis.

Fig. 1. Active BMP–Smad signaling in PGCs is required for PGC development.

a Expression of smad1, smad4, smad5, and smad9 in PGCs during early embryogenesis by RNA-seq^16,17. Quantification (b) of nuclear pSmad1/5/9 intensities normalized to cytoplasmic signals by pSmad1/5/9 (magenta) immunofluorescence and PGC marker Ddx4 (green) (c) in individual PGCs and somatic cells at the indicated stages and dorsal/ventral regions. PGCs were indicated by yellow arrowheads. Each dot represented one cell. Nc, number of observed cells. Error bars represented standard deviation (SD) with the mean indicated by the center line in (b). The schematics in (c) depicted the embryonic stages (animal pole view for dome stage; lateral views, dorsal to the right, for others) with PGCs shown in green. V ventral, D dorsal. d Effectiveness of BMP inhibition in embryos. Upper, experimental design for BMP inhibition using the BMP inhibitor LDN-193189; lower left panel, representative images of pSmad1/5/9 detected by immunostaining; right, bar graph showing nuclear pSmad1/5/9 intensities normalized to cytoplasmic levels in individual PGCs. Each dot represented one PGC. Nc, number of observed PGCs. Error bars represented SD with the mean indicated by the center line. e Visualization of PGCs by ddx4 ISH in DMSO- or LDN-treated embryos. The ratio of embryos with representative characteristics was indicated. f Visualization of PGCs in 3 dpf Tg(piwil1:mCherry-CAAX-nanos3 3′UTR) embryos. Upper, schematic of the transgene; lower, representative images of embryos treated with DMSO or LDN from the high stage to 1 dpf (the “Early” group). g Quantification of PGC numbers following treatment as in (f). Box plot showing PGC numbers per embryo in each group. The median is indicated by the center line. The bottom, top edges, and whiskers represent the min and max values and 1.5 times the interquartile range (IQR), respectively. Each dot represented one embryo. Ne, number of observed embryos. Statistical significance was determined by two-sided Student’s t-tests. Source data are provided as a Source Data file.

Next, to determine whether active BMP–Smad signaling is necessary for PGC development, we treated wild-type (WT) embryos with the BMP type I receptor inhibitor LDN-193189 (hereafter abbreviated as LDN), which could efficiently block phosphorylation of Smad1/5/9¹⁸. Given that BMP signaling is essential for dorsoventral patterning starting during blastulation¹⁹, we started the treatment with the low concentration from the high stage (the “Early” group in Fig. 1d) or with the various concentrations from the shield stage (the “Late” group in Fig. 1d; Supplementary Fig. 1a), to compare the dorsoventral patterning upon LDN treatment. We found that most of the embryos in the “Early” group exhibited truncated or curved tails, a relatively mild dorsalization phenotype (Supplementary Fig. 1a). However, higher concentrations of LDN induced severe body axis defects even in the “Late” group (Supplementary Fig. 1a). To further validate the BMP inhibition by LDN, we detected pSmad1/5/9 levels by immunostaining and observed a marked reduction in pSmad1/5/9 signal in both “Early” and “Late” groups with 2 μM LDN at the bud stage (Fig. 1d). Then, we collected control and LDN-treated embryos at 1 and 2 days post-fertilization (dpf) for PGC detection by whole-mount in situ hybridization (ISH) for ddx4 (also known as vasa), a germline-specific gene⁶ (Fig. 1e). The results revealed an obvious reduction in ddx4 signal in “Early” group at 1 dpf and both “Early” and “Late” groups at 2 dpf (Fig. 1e). To reliably count PGC numbers, we used Tg(piwil1:mCherry-CAAX-nanos3 3′UTR) transgenic embryos²⁰ for early LDN treatment, in which PGCs are specifically labeled by mCherry fluorescence driven by the 2.5-kb piwil1 promoter and the nanos3 3′ untranslated region (3′UTR)²¹. Microscopic observation revealed a significant decrease in PGC numbers in LDN-treated embryos in the “Early” group at 2 and 3 dpf (Fig. 1f, g). It is worth noting that when the LDN-treated embryos exhibited a severe dorsalization phenotype, the number of PGCs also showed a significant decrease (Supplementary Fig. 1b), which is consistent with previous reports suggesting that disrupting the dorsoventral axis affects PGC development²². Taken together, these results suggest that active BMP signaling is important for the PGC population in zebrafish embryos.

Knockdown of smad1/9 results in decreased PGC number

To directly suppress BMP–Smad signaling, we chose to knock down the downstream effectors smad1/5/9 in Tg(piwil1:mCherry-CAAX-nanos3 3′UTR) embryos using previously validated morpholino antisense oligonucleotides (MOs)^23–25. Triple knockdown by injecting a relatively low dose of smad1/5/9 MOs, without causing severe embryonic dorsalization, led to a remarkable PGC reduction at 2 dpf (Fig. 2a). To determine which specific smad gene is responsible for PGC population maintenance, individual genes were knocked down separately. We found that smad1 morphants, which displayed mild trunk curvature and regressed yolk stalk (Fig. 2b), as previously reported²⁶, showed a dramatic reduction in PGC number but no significant increase in mislocalized PGCs (Fig. 2b–d). In addition, immunostaining indicated that pSmad1/5/9 levels were markedly reduced in smad1 morphants, including PGCs at the 75% epiboly stage (Fig. 2e). Similarly, smad9 knockdown also led to reduced PGC number as counted at 2 dpf (Fig. 2f). In contrast, smad5 knockdown did not affect PGC number at 2 dpf (Supplementary Fig. 1c, d). However, due to the antibody limitations, we were unable to directly assess endogenous Smad5 protein levels after MO injection, and therefore cannot fully confirm the extent of smad5 knockdown. These data suggest that smad1 and smad9 are implicated in PGC population maintenance.

Fig. 2. Knockdown of smad1/5/9 leads to a reduction in PGC numbers.

a Effects of smad1/5/9 triple knockdown in Tg(piwil1:mCherry-CAAX-nanos3 3′UTR) embryos. Injected doses were 0.5 ng smad1 MO, 0.5 ng smad5 MO, and 3 ng smad9 MO. Left top, schematic of the knockdown strategy; left bottom, representative images showing embryonic morphologies and PGC numbers at 2 dpf, with red and green arrows indicating yolk stalk regression and ventral tail fin reduction, respectively; right, box plot showing PGC numbers per embryo. The median is indicated by the center line. The bottom, top edges, and whiskers represent the min and max values and 1.5 times the IQR, respectively. b Effect of smad1 knockdown alone. Left, embryonic morphology and PGC location of representative 1 dpf embryos injected with 0.5 ng control or smad1 MO; right, bar graph showing the percentage of mislocated PGCs. c Impact of smad1 knockdown on ddx4 expression in 2 dpf embryos, examined by ISH. Each embryo was injected with 0.5 ng MO at the 1-cell stage. d Impact of smad1 knockdown on PGC numbers in Tg(piwil1:mCherry-CAAX-nanos3 3′UTR) embryos at 2 dpf. Left, representative images with PGCs in red; right, box plot showing PGC counts per embryo. The median is indicated by the center line. The bottom, top edges, and whiskers represent the min and max values and 1.5 times the IQR, respectively. Each embryo was injected with 0.5 ng MO. e Effect of smad1 knockdown on pSmad1/5/9 (magenta) immunostaining. Tg(kop:gfp-CAAX-nanos3 3′UTR) embryos were injected with 0.5 ng MO at the 1-cell stage, and collected at 75% epiboly stage for measurement. Left, representative immunostaining images with PGCs labeled by GFP (green); right, quantification of nuclear pSmad1/5/9 intensities normalized to cytoplasmic signals, with each dot representing one PGC. Error bars represented SD with the mean indicated by the center line. f Effect of smad9 knockdown on PGC numbers in 2 dpf Tg(piwil1:mCherry-CAAX-nanos3 3′UTR) embryos injected with 16 ng MO at the 1-cell stage. Left, representative images with PGCs (mCherry, red); right, box plot showing PGC counts per embryo. The median is indicated by the center line. The bottom, top edges, and whiskers represent the min and max values and 1.5 times the IQR, respectively. Statistical significance was determined using two-sided Student’s t-test unless otherwise indicated (a, d–f). Two-sided Fisher’s exact test was used in (b). Nc, number of observed PGCs. Ne, number of observed embryos. Source data are provided as a Source Data file.

Generation of smad1/9 PGC-specific mutants

To further investigate the cell-autonomous functions of Smad1 and Smad9 in PGCs, we employed our recently developed CRISPR/Cas9-based double-transgenic genome editing system that specifically induces gene mutation in germline cells²⁰. To do this, we established two transgenic lines. The first transgenic line is Tg(piwil1:zCas9-nanos3 3′UTR), in which the piwil1 promoter⁷ drives the PGC/GC-specific expression of zebrafish codon-optimized Cas9 (zCas9)²⁷ in the presence of nanos3 3′UTR (Fig. 3a; Supplementary Fig. 2a). The second transgenic line ubiquitously expresses smad1 or smad9 gRNAs under the control of the U6a/b/c promoters along with gsc:tdTomato (tdTomato signals at dorsal region at 6 hpf) or cryaa:CFP (CFP signals in the lens at about 30 hpf) for easy identification of transgenic embryos (Fig. 3a; Supplementary Fig. 2b, c). Gene editing could occur in the PGCs of embryos derived from crosses between these two transgenic lines (female for zCas9 and male for gRNAs), resulting in PGC-specific mutations of the target gene (Fig. 3a; “Methods” section). TdTomato- or CFP-positive embryos were identified as PGC-specific KO (cKO) embryos, and negative siblings were used as controls (Fig. 3a; Supplementary Fig. 2b, c). Notably, to gain a more comprehensive observation of PGCs at different stages, we crossed the Tg(piwil1:zCas9-nanos3 3′UTR) line with Tg(kop:Eos-CAAX-nanos3 3′UTR) or Tg(piwil1:mCherry-CAAX-nanos3 3′UTR) for early-stage (from the dome stage to pre-2 dpf) PGC labeling or long-term (from the shield stage to adulthood) PGC labeling, respectively (Fig. 3a).

Fig. 3. smad1 is required cell-autonomously for PGC development.

a Schematic of the mating strategy used to generate smad1-cKO (or smad9-cKO) embryos (“Methods” section). PGCs were labeled with kop:Eos (green) or piwil1:mCherry (red), and cKO embryos identified by gsc:tdTomato or cryaa:CFP. b Genome editing efficiency of three smad1 gRNA target sites assessed by targeted amplicon sequencing in somatic cells and PGCs sorted from sibling and smad1-cKO embryos at 15-somite stage. NGS, next-generation sequencing. N. reads, number of aligned reads per target site. Representative images of sibling and smad1-cKO embryos at the 13-somite stage (c) and 1 dpf (d). Embryos were dorsally viewed (c), or laterally viewed with anterior to the left (d). PGCs were kop:Eos-positive (green). The gsc:tdTomato-positive (red) notochord (NT) was indicated by arrow (c). e ISH-detected ddx4 expression in sibling and smad1-cKO embryos at 2 dpf. The gonad region was enlarged for better visibility. f Quantification of PGC numbers per embryo in sibling and smad1-cKO groups at the indicated stages. g The proportion of mislocated PGCs in sibling and smad1-cKO embryos at 1 dpf. h Sex ratios of sibling and smad1-cKO adult fish at 4.5 months post-fertilization (mpf). i Rescue of PGC numbers in smad1-cKO embryos by smad1-nanos3 3′UTR mRNA injection (250 pg/embryo). Left, representative images of 2 dpf embryos with a lateral view and anterior to the left. PGCs were kop:Eos-positive (green); right, box plot showing PGC numbers per embryo. In f and i, the bottom, top edges, and whiskers represent the min and max values and 1.5 times the IQR, respectively. The median is indicated by the center line. Each dot represented one embryo. Statistical analysis was performed using two-sided Student’s t-test (f, i) or two-sided Fisher’s exact test (g, h). Ne, number of embryos analyzed; Nc, number of PGCs quantified; Nf, number of fish tested. Source data are provided as a Source Data file.

To validate the genome editing efficiency of our PGC-specific KO system, we first utilized FACS to sort mCherry-labeled PGCs in control siblings (with zCas9 in PGCs but without gRNAs), and smad1-cKO embryos at 2 dpf (Supplementary Fig. 2d). Quantitative RT-PCR (qRT-PCR) confirmed that ddx4 was markedly enriched in sorted PGCs relative to somatic cells (Supplementary Fig. 2d), indicating effective sorting. Next, we assessed the mutation efficiency at the three smad1 gRNA target sites in both PGCs and somatic cells by performing whole-genome amplification (WGA) on a limited number of FACS-sorted cells at the 15-somite stage, followed by targeted amplicon sequencing of the regions flanking these sites (“Methods” section). Sequencing results revealed that, at the gRNA1 target region, 81.10% (n = 85,644/105,598) reads were mutated in smad1-cKO PGCs, with 76.23% (n = 80,495/105,598) being frameshift indels. In contrast, only 0.26% (n = 397/152,603), 0.31% (n = 435/142,067), and 0.48% (n = 522/109,669) of reads were mutated in sibling somatic cells, sibling PGCs, and smad1-cKO somatic cells, respectively (Fig. 3b; Supplementary Fig. 2e). For the gRNA2 target region, 62.65% (n = 176,032/280,997) reads, including 44.32% (n = 124,548/280,997) frameshift indels, were found in smad1-cKO PGCs, compared with 0.16% (n = 409/261,506) in smad1-cKO somatic cells, 0.14% (n = 553/395,388) in sibling somatic cells, and 0.16% (n = 616/394,178) in sibling PGCs (Fig. 3b; Supplementary Fig. 2e). For the gRNA3 target region, 37.22% (n = 33,258/89,362) reads were mutated, including 36.40% (n = 32,532/89,362) frameshift indels, in smad1-cKO PGCs, and extremely low mutation frequencies were observed in smad1-cKO somatic cells (1.38%, n = 478/34,593), sibling somatic cells (1.39%, n = 747/53,823), and sibling PGCs (1.28%, n = 707/55,227) (Fig. 3b; Supplementary Fig. 2e). Notably, pSmad1/5/9 signaling was reduced in smad1-cKO PGCs (Supplementary Fig. 2f–h), further supporting the efficient BMP–Smad signaling disruption. Together, these results indicate that the PGC-specific knockout system works effectively. It is worth mentioning that, as previously reported²⁰, biallelic loss-of-function mutations are likely present in only a fraction of PGCs.

PGC-specific knockout of smad1/9 leads to PGC reduction

Next, we examined PGC development in smad1- or smad9-cKO embryos with PGCs labeled by kop:Eos expression at pre-2 dpf or by piwil1:mCherry expression after 2 dpf (Fig. 3a). First, we counted the number of PGCs in smad1-cKO embryos from the shield stage to 4 dpf and observed a significant reduction in PGC numbers in smad1-cKO embryos compared to siblings from the 90% epiboly stage through 4 dpf (Fig. 3c–f). The magnitude of the decline in PGC numbers ranged from 27.77% to 57.35% (Fig. 3f), which is reasonable since only a subset of PGCs is expected to carry biallelic null mutations through our transient editing system. Despite reduced PGC numbers, overall embryonic morphology, including axis formation, remained unaffected in smad1-cKO embryos at 1 and 2 dpf (Fig. 3d, e; Supplementary Fig. 2b), consistent with the extremely low mutagenesis rates observed in somatic cells of smad1-cKO embryos (Fig. 3b; Supplementary Fig. 2e). Notably, the majority of smad1-cKO PGCs migrated normally to the future gonads at 1 dpf (Fig. 3e, g). Given that PGC numbers influence sex determination in zebrafish with a higher number of PGCs promoting female development and a lower number skewing towards males^28–30, we raised these smad1-cKO embryos and their siblings to adulthood (4.5 months post-fertilization) and found that smad1-cKO fish exhibited a strong male bias in sex determination (Fig. 3h). Importantly, overexpression of synthetic smad1-nanos3 3′UTR mRNA, which could be specifically translated in PGCs, partially recovered PGC numbers in smad1-cKO embryos (Fig. 3i). Therefore, these data suggest that loss of Smad1 in PGCs would trigger a decline in PGC numbers.

Similarly, smad9-cKO embryos showed a reduction in ddx4 ISH signal at 2 dpf (Fig. 4a) and a 39.68% decrease in PGC number at 3 dpf (Fig. 4b). To further block BMP–Smad signaling in PGCs, we generated smad1&9 double-cKO mutants based on our transgenic system (“Methods” section). First, Tg(U6:smad1 gRNAs;gsc:tdTomato) and Tg(U6:smad9 gRNAs;cryaa:CFP) lines were crossed to obtain double gRNA-expressing males, which then mated with Tg(piwil1:zCas9-nanos3 3′UTR);Tg(kop:Eos-CAAX-nanos3 3′UTR) females. Since cryaa:CFP expression in the lens becomes detectable around 30 hpf, smad1&9-cKO embryos cannot be identified based on cryaa:CFP until 30 hpf. Then, we counted PGCs in smad1&9 double-cKO embryos at 2 dpf and found an even greater decrease in PGC number compared to either smad1-cKO or smad9-cKO embryos (Fig. 4c). Thus, we concluded that Smad1 and Smad9 in PGCs are critical for forming normal PGC numbers in zebrafish.

Fig. 4. PGC-specific knockout of smad9 leads to reduced PGC numbers.

a Representative ddx4 ISH images of sibling and smad9-cKO embryos. The gonad region was enlarged for better visibility. b Effect of smad9-cKO on PGC number. Left, representative images showing PGC development in 3 dpf sibling and smad9-cKO embryos. Right, box plot showing PGC numbers per embryo. The median is indicated by the center line. The bottom, top edges, and whiskers represent the min and max values and 1.5 times the IQR, respectively. c Quantification of PGC numbers in different mutants. Left, box plot showing PGC numbers per embryo in sibling, smad1-cKO, smad9-cKO, and smad1&9-cKO embryos at 2 dpf. The median is indicated by the center line. The bottom, top edges, and whiskers represent the min and max values and 1.5 times the IQR, respectively. Each dot represented one embryo. Right, representative images of PGCs in the corresponding embryos. PGCs were labeled by kop:Eos. Dotted circles marked cryaa:CFP in the lens of smad9-cKO embryos; red fluorescence indicated gsc:tdTomato in smad1-cKO embryos. NT, notochord. White arrowheads indicated cellular debris. Statistical analysis was performed using a two-sided Student’s t-test. Ne, number of embryos analyzed. Source data are provided as a Source Data file.

Deficient proliferation and apoptosis in smad1 KO PGCs

To determine whether reduced PGC numbers in PGC-specific smad1/9 KO mutants result from impaired cell proliferation and/or apoptosis, we traced kop:Eos-labeled PGCs via live imaging within smad1-cKO embryos and siblings from 50% to 90% epiboly stages (Fig. 5a). From the germ ring to 90% epiboly stages, 51.43% (n = 18/35) of imaged PGCs in siblings exhibited cell division, while only 10.53% (n = 2/19) of the imaged PGCs in smad1-cKO embryos underwent cell division (Fig. 5b). To validate defective PGC proliferation, we performed EdU incorporation in smad1-cKO embryos and siblings from the shield to bud stages and collected embryos at 2-somite stage for observation of kop:Eos-labeled PGCs (Fig. 5c). The majority (81.91%, n = 77/94) of PGCs in siblings showed EdU signals, indicating active DNA replication and cell proliferation (Fig. 5c). In contrast, significantly fewer (62.50%, n = 45/72) smad1-cKO PGCs exhibited EdU signals, suggesting deficient cell proliferation of smad1-cKO PGCs (Fig. 5c). The higher proliferation rate detected by EdU compared to live imaging (Fig. 5b, c) likely reflects the greater sensitivity of EdU to DNA synthesis. Additionally, live imaging captures fewer cells for analysis due to cell migration and embryo depth. We observed numerous cellular fragments with piwil1:mCherry fluorescence in piwil1:mCherry;smad1-cKO embryos at 2 dpf (Fig. 5d, asterisks) and piwil1:mCherry;smad9-cKO embryos (Fig. 4b, arrowheads). Live imaging of piwil1:mCherry;smad1-cKO PGCs captured the process in which PGCs shed abundant debris (Fig. 5d). Via immunostaining of active Caspase-3, PGC apoptosis was evidenced by the increased active Caspase-3 signals in both smad1-cKO and smad1&9-cKO PGCs, compared to siblings (Fig. 5e; Supplementary Fig. 3). Therefore, these results indicate that impaired PGC proliferation and increased apoptosis of PGCs most likely account for the reduced PGC numbers in smad1/9-cKO embryos.

Fig. 5. smad1-cKO PGCs exhibit defective proliferation and increased apoptosis.

a Representative tracking images of dividing PGCs in sibling and smad1-cKO embryos from germ ring to 90% epiboly stage. The kop:Eos-labeled PGCs were marked by red and white asterisks. b Ratio of dividing PGCs in embryos shown in (a). c Detection of proliferating PGCs via EdU incorporation assay. Embryos were treated with EdU from the shield to the bud stage and collected at the 2-somite stage for immunostaining with EdU and Ddx4. Top panel, schematic of assay procedure; lower left, representative images of PGCs (red). Ddx4- and EdU-positive PGCs were indicated by yellow arrowheads, and Ddx4-positive but EdU-negative PGCs by white arrows. Lower right, bar graph showing the percentage of EdU-positive and EdU-negative PGCs in sibling and smad1-cKO embryos. d Live images of smad1-cKO PGCs undergoing apoptosis. Top panel, piwil1:mCherry-labeled PGCs in the gonad region; bottom panel, time-lapse images showing PGCs undergoing apoptosis. Yellow arrowheads marked surviving PGCs, white arrows indicated apoptotic PGCs, and asterisks denoted apoptotic cell debris. e Immunostaining of active Caspase-3 (green) and mCherry (PGCs, red) in sibling and smad1-cKO embryos at 2 dpf. Left panel, representative images of PGCs in the gonad region. Active Caspase-3–positive and –negative PGCs were indicated by white arrows and yellow arrowheads, respectively. Right, bar graph showing the percentage of active Caspase-3–positive or –negative PGCs. Statistical significance was determined using a two-sided Fisher’s exact test. Nc, number of observed PGCs. Source data are provided as a Source Data file.

Cell cycle defects and DNA damage response in smad1 KO PGCs

To investigate molecular mechanisms underlying BMP–Smad signaling regulation in PGC population maintenance during early embryonic development, we performed RNA-seq of Eos-positive PGCs sorted by FACS from kop:Eos;smad1-cKO embryos and siblings, respectively, at 75% epiboly and 1 dpf stages with three biological replicates (Fig. 6a; Supplementary Fig. 4a, left). First, we checked the expression levels of PGC-specific genes such as nanos3, ddx4, ca15b, and dnd1^6,31–33 and found no significant changes in smad1-cKO PGCs (Fig. 6b). Moreover, we performed RNA-seq of sorted PGCs from smad1 MO- and control-MO-injected embryos at 75% epiboly and 1 dpf stages with two biological replicates (Supplementary Fig. 4a, right), and found that expression levels of PGC-specific genes in smad1-KD PGCs were also comparable to those in control PGCs (Supplementary Fig. 4b). Collectively, these data suggest that Smad1 is unlikely to participate cell-autonomously in zebrafish PGC fate determination.

Fig. 6. DDR activation, primarily through the ATR pathway, led to reduced PGC numbers in smad1-cKO embryos.

a Schematic illustration of PGC isolation and RNA-seq in sibling and smad1-cKO embryos. b Expression levels of PGC-specific genes in sibling and smad1-cKO PGCs. c Volcano plots showing DEGs between smad1-cKO and sibling PGCs at the indicated stages. Red and blue dots indicated upregulated and downregulated genes, respectively. d Expression of DEGs between the 1 dpf cKO and sibling PGCs at the 75% epiboly and 1 dpf stages (left heatmaps) and functional enrichment of upregulated genes in smad1-cKO PGCs at 1 dpf (right, bubble plot). e Detection of DNA damage in PGCs (Ddx4, green) via γH2A immunostaining (magenta) in sibling and smad1-cKO embryos at the 5-somite stage. Red and white dotted circles indicated γH2A-positive and -negative PGCs, respectively. Right, quantification of nuclear γH2A intensities normalized to cytoplasmic levels. Each dot represented one PGC. f Immunostaining of pChk1 (magenta) in sibling and smad1-cKO embryos at the bud stage. PGCs were labeled by Ddx4 (green). Red and white circles indicated pChk1-positive and -negative PGCs, respectively. Right, quantification of nuclear pChk1 intensities normalized to cytoplasmic levels. Each dot represented one PGC. g Representative images (left) and quantification of PGCs (right) in the gonad regions of sibling and smad1-cKO embryos treated with DMSO or ATR inhibitor (AZ20) from the sphere stage to the 1 or 2 dpf. PGCs were labeled with kop:Eos (green). Each dot represented one embryo. Error bars represented SD with the mean indicated by the center line in (e–g). Statistical significance was determined by a two-sided Student’s t-test. Nc, number of observed PGCs; Ne, number of observed embryos. Source data are provided as a Source Data file.

Next, detailed analyses identified 80 upregulated genes and 106 downregulated genes (P < 0.01, fold change (FC) > 2) at 75% epiboly stage and 335 upregulated genes and 78 downregulated genes at 1 dpf in smad1-cKO PGCs (Fig. 6c; Supplementary Data 1), indicating mild changes in the transcriptome of smad1-cKO PGCs at 75% epiboly stage (Fig. 6c, left). The upregulation of 335 genes in cKO PGCs at 1 dpf represents the largest set of differentially expressed genes (DEGs) caused by Smad1-cKO (Fig. 6c, right), whose expression was not significantly changed at the 75% epiboly stage (Fig. 6c, left; Fig. 6d, left heatmaps; Supplementary Fig. 4c). Gene ontology (GO) analysis of these 335 upregulated genes revealed that the GO terms mitotic cell cycle, chromosome segregation, mitotic checkpoint activation, and DNA damage response (DDR) were significantly enriched (Fig. 6d, right), which might be responsible for cell cycle defects and cell apoptosis of smad1-cKO PGCs at 1 dpf. Notably, we confirmed the upregulation of apoptosis-related genes in smad1-cKO PGCs at 1 dpf, but not at the 75% epiboly stage (Supplementary Fig. 4d).

Furthermore, to validate the DNA damage in smad1-cKO PGCs, we conducted γH2AX immunostaining and observed elevated γH2AX signals in smad1-cKO PGCs (Fig. 6e). Meanwhile, to exclude non-specific DNA damage that might be caused by CRISPR/Cas9 genome editing, we first analyzed DEGs in smad1-KD PGCs. We identified 69 upregulated genes and 111 downregulated genes at 75% epiboly stage, and 394 upregulated genes and 687 downregulated genes at 1 dpf in smad1-KD PGCs (Supplementary Fig. 4e). However, the overlapping DEGs between cKO and KD PGCs were quite limited, and many developmental genes in KD PGCs were also affected, which may reflect distinct consequences of global BMP signaling disruption in smad1-KD embryos compared to PGC-specific Smad1 deficiency in smad1-cKO embryos. Notably, despite limited overlaps, many genes involved in DNA damage repair and cell apoptosis, such as tp53, casp8, sesn3, foxo3b, gadd45aa, gadd45ba, tp53inp1, ccng1, and phlda3, were also upregulated in smad1-KD PGCs (Supplementary Fig. 4f; Supplementary Data 1), hinting that smad1 deficiency might lead to excessive DDR in PGCs.

Then, to further exclude the possibility that the excessive DNA damage response in cKO PGCs was caused by non-specific Cas9 activity, we performed an additional control experiment using gRNAs targeting an irrelevant gene, tyrosinase (tyr), which is silenced in PGCs during early embryogenesis. First, we generated Tg(U6:tyr gRNAs;gsc:tdTomato) line using validated gRNAs^34,35 and mated it with Tg(piwil1:zCas9-nanos3 3′UTR);Tg(kop:Eos-CAAX-nanos3 3′UTR) females (Supplementary Fig. 5a). By targeted amplicon sequencing of gRNA target regions, 29.75% (n = 547,590/1,840,038) reads at the gRNA1 target region in PGCs were mutated, and almost no mutated reads were detected at gRNA2 target region (Supplementary Fig. 5b). As expected, the PGC numbers in tyr-cKO embryos and siblings were comparable (Supplementary Fig. 5c, d). To enhance the genome editing efficiency, we injected three tyr gRNAs and commercial Cas9 protein into Tg(kop:gfp-CAAX-nanos3 3′UTR) embryos at the 1-cell stage (Supplementary Fig. 6e). We detected 28.36% (n = 85,365/300,997) mutated reads at gRNA1 target region and 37.01% (n = 112,219/303,226) mutated reads at gRNA3 target region in PGCs (Supplementary Fig. 5f). Consistently, no PGC number decrease was observed in tyr-KO embryos compared with the control embryos (Supplementary Fig. 5g, h). Together, these results strongly suggest that the reduction in PGC numbers in smad1-cKO embryos was caused by Smad deficiency.

Aberrant ATR activation was responsible for PGC reduction

Ataxia telangiectasia and Rad3-related (ATR) is the apical DNA replication stress response kinase, specifically in proliferating cells^36,37. A key role of ATR is to phosphorylate and activate the protein kinase Chk1^38,39, slowing or arresting cell cycle progression when DNA damage occurs, which then allows more time for DNA repair to prevent premature mitotic entry^40,41. However, if the DNA damage is too extensive, ATR–Chk1 activation could directly lead to cell apoptosis^42,43. To determine whether ATR–Chk1 is aberrantly activated in smad1-cKO PGCs in response to DNA damage, we assessed phosphorylated Chk1 (pChk1) level via immunostaining and found that smad1-cKO PGCs exhibited significantly elevated pChk1 signals (Fig. 6f). In contrast, pChk1 levels were not elevated in PGCs of tyr-KO control embryos (Supplementary Fig. 5i, j). Notably, we attempted to rescue PGC apoptosis using AZ20, an ATR inhibitor that has been shown to restore erythrocyte number by counteracting DNA damage-induced cell cycle arrest and apoptosis in zebrafish erythroid progenitors⁴⁴. Interestingly, AZ20 treatment could restore PGC numbers in smad1-cKO embryos at both 1 and 2 dpf (Fig. 6g, red). AZ20-treated siblings showed no obvious changes in PGC numbers (Fig. 6g, blue). Moreover, the rescue effect of AZ20 was dose-dependent, with a lower concentration resulting in a lower rescue efficiency (Supplementary Fig. 6a). Thus, these data indicate that ectopic activation of ATR–Chk1 in response to DNA damage might trigger cell proliferation arrest and apoptosis of smad1-cKO PGCs.

In addition, we tested whether another major DDR pathway is involved—the ataxia-telangiectasia mutated (ATM)–Chk2 pathway, which primarily responds to DNA double-strand breaks (DSBs). To this end, we assessed the level of phosphorylated Chk2 (pChk2), a marker for ATM activation⁴⁵, in mutant PGCs by pChk2 immunostaining. First, we validated the pChk2 antibody in zebrafish embryos treated with etoposide, a topoisomerase II inhibitor inducing DSBs⁴⁶, and KU60019, an ATM inhibitor, and detected strong pChk2 signals activated by DSBs and decreased pChk2 signals upon KU60019 (Supplementary Fig. 6b). Then, via pChk2 immunostaining, we can hardly detect ATM activation in PGCs from either smad1-cKO mutants or siblings (Supplementary Fig. 6c). Furthermore, treatment with KU60019 could not restore PGC numbers in smad1-cKO mutants (Supplementary Fig. 6d), underscoring that ATM–Chk2 signaling may not be involved in PGC reduction in smad1-cKO mutants.

Alternative splicing of DNA damage response-related genes

There were only minor changes in the transcription levels due to the loss of Smad1 in PGCs at 75% epiboly stage (Fig. 6c). Notably, increasing evidence has shown that Smads, beyond their functions in transcriptional regulation, also participate in post-transcriptional regulation across multiple species^47–49. Therefore, we investigated whether Smad1 contributes to post-transcriptional regulation by performing transcriptome-wide analysis of alternative splicing (AS) in smad1-cKO PGCs using replicate multivariate analysis of transcript splicing (rMATS)⁵⁰. There were 616 significant AS events in smad1-cKO PGCs compared to those in sibling PGCs (FDR < 0.05 and inclusion level difference >10%), including 516/616 exon skipping (ES), 19/616 intron retention (IR), 6/616 mutually exclusive exons (MXE), 32/616 alternative 5′ splice sites (A5SS) and 43/616 alternative 3′ splice sites (A3SS) (Fig. 7a; Supplementary Data 2). Then, we identified 512 protein-coding genes displaying differentially spliced transcripts (AS genes for short) according to these 616 splicing events. Furthermore, we identified 660 AS genes in smad1-KD PGCs compared with those in control PGCs (Supplementary Data 3). Remarkably, we observed a highly significant overlap (hypergeometric test, P = 2.44e-81) between the AS genes in smad1-cKO (121/512) and smad1-KD (121/660) PGCs (Fig. 7b), such as ccdc187 (coiled-coil domain containing 187) that is involved in microtubule anchoring and located in the centrosome⁵¹, mak (male germ cell associated kinase) that encodes a serine/threonine protein kinase related to cell cycle regulation⁵², tinf2 (TERF1-interacting nuclear factor 20) that encodes one subunit of shelterin complex involved in distinguishing between telomeres and DNA damage⁵³, among others (Supplementary Fig. 7a). Altered splicing of these genes was confirmed by RT-PCR (Fig. 7c; Supplementary Fig. 7b). Moreover, GO analysis of all these 1,051 AS genes revealed that the terms chromatin regulator, centrosome, DNA damage checkpoint signaling, and cell cycle were significantly enriched (Fig. 7b). Thus, these data hint that loss of Smad1 in PGCs results in altered splicing of genes related to chromatin structure, DNA damage response, and cell cycle.

Fig. 7. Genes involved in DNA repair, chromatin organization, and cell division show altered splicing patterns in smad1-cKO PGCs.

a Schematic summarizing changes in splicing patterns identified by rMATS between sibling and smad1-cKO PGCs at 75% epiboly stage. Changed events were filtered for FDR < 0.05 and absolute percent spliced in (ΔPSI) > 0.1. ES, exon skipping; IR, intron retention; A5SS, alternative 5′ splice site; A3SS, alternative 3′ splice site; MXE, mutually exclusive exon. The proportions of each alternatively spliced (AS) event were shown in parentheses. b Left, Venn diagram showing overlaps among AS genes in cKO and KD RNA-seq datasets with DEGs in PGCs. P-value was determined via a hypergeometric test. Right, enriched GO terms for all AS genes in cKO and KD PGCs. c RT-PCR validation of altered splicing of selected AS genes in smad1-cKO PGCs at 75% epiboly stage. The compositions of exons for the major products were indicated on the right, with the blue exon being alternatively spliced. d Co-immunoprecipitation (co-IP) of Smad1 with spliceosome components Snrpa and Snrpb in HEK293T cells. e Proximity ligation assay (PLA) of pSmad1/5/9 and Snrpb in 75% epiboly embryos. Top: schematic of the PLA principle. Bottom left: representative images showing PLA probe signals between pSmad1/5/9 and Snrpb in PGCs (kop:GFP, green). Bottom right, quantification of nuclear PLA signals normalized to cytoplasmic levels in PGCs. Each dot represented one PGC. Nc, total number of observed PGCs. Error bars indicated SD, with the mean indicated by the center line. Statistical significance was determined by a two-sided Student’s t-test. f Model of BMP–Smad signaling regulating zebrafish PGC population. WT PGCs proliferate and survive properly, contributing to balanced sex differentiation. smad1/9-cKO PGCs exhibit altered splicing and DNA damage response, resulting in ATR activation, reduced proliferation, and cell death, consequently inducing decreased PGC numbers and biased sex differentiation toward males. Source data are provided as a Source Data file.

To investigate whether Smad1 directly or indirectly regulates alternative splicing in PGCs, we examined the interaction between pSmad and splicing factors both in vitro and in vivo. First, via co-immunoprecipitation (co-IP) with Smad1 in HEK293T cells, we checked several small nuclear ribonucleoproteins (snRNPs) in core spliceosome components, such as small nuclear ribonucleoprotein polypeptide A (Snrpa) and small nuclear ribonucleoprotein polypeptide B and B′ (Snrpb)⁵⁴. Co-IP assay revealed a strong interaction between Smad1 and Snrpb, but only a weak or negligible interaction between Smad1 and Snrpa (Fig. 7d). Then, to validate the interaction, we performed proximity ligation assay (PLA) in zebrafish embryos. In brief, when two proteins are adjacent to each other (typically less than 40 nm), specially designed secondary antibodies with attached DNA oligonucleotides (PLA probes) could be close enough for PLA probe ligation, rolling circle amplification, and bound by fluorescently labeled probe, which creates a bright fluorescent spot at the site of interaction (Fig. 7e, upper)⁵⁵. Embryos treated with only the pSmad1/5/9 antibody were set as a negative control. Compared with PGCs in the control group, PLA signals were significantly stronger in PGCs within embryos treated with both pSmad1/5/9 and Snrpb antibodies (Fig. 7e). Therefore, these data make it evident that the pSmads might regulate RNA splicing in the nucleus of PGCs by interacting with a spliceosome component.

Discussion

The BMP–Smad signaling pathway is required for PGC induction in mammals, whereas zebrafish PGCs are specified by maternally provided germ plasm. In this study, we observed active BMP–Smad signaling in zebrafish PGCs and demonstrated that BMP–Smad1/9 signaling is dispensable for PGC fate determination and migration but required for PGC proliferation. We generated PGC-specific knockout of smad1 and/or smad9 mutants using our recently established double-transgenic strategy²⁰ to minimize somatic cell interference with PGC development, such as dorsoventral axis formation mediated by BMP signaling. Based on transcriptome analysis and PGC tracing, we conclude that a significant reduction in PGC number, but not PGC-specific gene expression changes and cell migration defects, is triggered by PGC-specific loss of Smad1/9. Notably, ATR–pChk1 is overactivated in smad1-cKO PGCs, along with cell cycle arrest and ectopic cell apoptosis, leading to PGC reduction, which can be partially rescued by inhibition of ATR–pChk1. Meanwhile, alternative splicing analysis revealed that Smad1 might regulate mRNA splicing of genes involved in chromatin regulation, cell cycle checkpoint, and DNA damage response, which is validated by direct interaction between Smad1 or pSmad1/5/9 and Snrpb (Fig. 7f). Therefore, we propose that BMP–Smad signaling is fundamental for PGC development across vertebrates but fulfills distinct roles in zebrafish.

In mice, cells in the proximal tier of the epiblast around the gastrulation stage will be induced to commit to the PGC fate by BMP signals from the extraembryonic ectoderm¹. In contrast, in zebrafish, four clusters of PGC precursors are located separately in the early blastula (1k-cell stage), two near the ventral region and two near the dorsal region⁶. Given the gradient distribution of BMP signaling along the ventral–dorsal axis within zebrafish embryos, the PGCs located on the ventral side displayed higher pSmad1/5/9 signals, while PGCs on the dorsal side showed lower pSmad1/5/9 signals from the dome to the epiboly stages (Fig. 1b, c). Then, all PGCs migrate dorsally and form two lateral PGC clusters at the somite stages²². The levels of pSmad1/5/9 were comparable in PGCs and surrounding somatic cells from somite stages to 1 dpf (Fig. 1b, c), when we observed the defective PGC proliferation during somatic stages and then PGC apoptosis after 1 dpf. Therefore, we cannot conclude whether the differential pSmad1/5/9 signals in early PGCs across the dorsoventral axis are involved in regulating PGC development based on our current results. Notably, recent lineage-tracing studies have revealed that zebrafish PGCs are indeed a heterogeneous population with differential developmental potentials⁵⁶. It will therefore be interesting to investigate involvement and the function of pSmad1/5/9 dynamics in PGCs across different developmental stages and in the context of their heterogeneity, using spatial single-cell RNA-seq or pSmad1/5/9 dynamic tracing in PGCs.

Nevertheless, although high genome editing efficiency in PGC-specific KO (Fig. 3b), several PGCs could survive and differentiate into germ cells in adulthood (Fig. 3h), which could be ascribed to the technical limitations of our study. That is, after KD or PGC-specific cKO, the target mRNAs or genes in some PGCs may escape from the translation block or gene editing, allowing these PGCs to proliferate normally. We speculate that no PGCs with biallelic null mutations in smad1/9 would survive; that is, Smad1/9 in PGCs would be essential for PGC survival. Further improvements in PGC-specific knockout strategies are warranted to achieve complete ablation of Smad1/9 and thus enable a more definitive validation of our hypothesis.

Preserving the high degree of genome integrity and stability in germ cells is crucial for reproduction and species continuity. However, PGCs encounter a high risk of DNA damage due to the following reasons. First, zebrafish early embryos undergo synchronously rapid cell proliferation (approximately 15 min per cycle during cleavage stages and 30 min per cycle during gastrulation stages)⁵⁷, comparable to fast-dividing systems such as E. coli and early Drosophila melanogaster embryo. However, the zebrafish haploid genome (1.4–1.6 Gb, GRCz11) is substantially larger than that of E. coli (4.6 Mb) and D. melanogaster (140 Mb), resulting in extremely high DNA replication stress in zebrafish embryos before gastrulation, including somatic cells and PGCs. Second, zebrafish PGCs also initiate their zygotic genome activation before migration and then go through extensive transcriptome and epigenetic reprogramming during migration to accomplish the lineage specification process¹⁶. The combination of these factors poses a significant risk of DNA damage, which increases the need for robust DNA repair in zebrafish PGCs. One interesting question is whether BMP signaling-mediated genome integrity maintenance is germ cell-specific or ubiquitous in early embryos. Notably, we also observed cell death or reduced body length in embryos injected with low doses of either triple smad1/5/9 MOs or individual MOs, without significant dorsoventral axis defects (Fig. 2b, c; Supplementary Fig. 1d). Recently, it has been reported that BMP signaling relieves DNA replication stress, promoting cardiomyocyte proliferation during heart regeneration in zebrafish⁵⁸. This ability of BMP signaling to overcome replication stress might also be conserved in mammalian cells^58,59.

Previous studies on how the BMP signaling regulates the cell cycle or DNA damage repair have typically focused on transcriptional regulation by pSmad1/5/9^60–64. In our study, we observed PGC reductions during the epiboly and somite stages, while no obvious transcriptome changes were detected at 75% epiboly stage, hinting that BMP–Smads signaling might regulate PGC proliferation and DNA damage response through mechanisms other than classical transcriptional regulation. Intriguingly, we detected abnormal splicing events in both smad1-cKO and smad1-KD PGCs. Meanwhile, it has been shown that Smad3 regulates alternative splicing by directly associating with hnRNPE1, promoting epithelial-to-mesenchymal transition and metastasis in mouse⁴⁹. Thus, we tested whether Smad1 could directly interact with splicing factors and found that it binds Snrpb. However, more detailed investigations are warranted to reveal how BMP–Smad regulates alternative splicing in zebrafish PGCs in the future. Altogether, our findings, along with the systematic PGC-specific smad1/5/9 knockout and analysis, may help pave the way for future studies of BMP–Smad signaling pathway in PGC development.

Methods

Zebrafish husbandry and strains

All zebrafish husbandry and experimental procedures were approved by the Tsinghua University Animal Care and Use Committee. The wild-type Tübingen (TU) strain was used for this study and for generating transgenic and mutant lines. Adult zebrafish were maintained in a water-circulating system at 28.5 °C under a 14-h light/10-h dark cycle. Embryos were raised at 28.5 °C in Holtfreter’s solution (0.05 g/L KCl, 0.1 g/L CaCl₂, 0.025 g/L NaHCO₃, and 3.5 g/L NaCl, pH 7.0). For imaging of PGCs, 0.003% phenyl-2-thiourea (PTU; Sigma-Aldrich, catalog No. P7629), a tyrosinase inhibitor, was added to the Holtfreter’s solution from 24 hpf onward to inhibit pigmentation.

gRNA and mRNA synthesis

For mutagenesis of tyr by CRISPR/Cas9, previously validated gRNAs were synthesized^34,35. DNA templates for transcription were generated by PCR using T7‑linked forward primers containing the target sequences and a common reverse primer complementary to the gRNA scaffold. gRNA and Primer sequences are listed in Supplementary Tables 1 and 2. PCR products were purified and used directly as templates for in vitro transcription with the MEGAscript T7 Kit (Thermo Fisher, catalog No. AM1334). After 4 h at 37 °C, reactions were treated with DNase I, and gRNAs were purified by phenol/chloroform extraction and ethanol precipitation.

For the smad1-cKO rescue experiment, an expression vector containing the smad1 coding sequence followed by the nanos3 3′UTR was constructed in the pXT7 backbone. For validating smad5 MO efficacy, a vector containing the MO-targeted exon3 of smad5 fused to gfp-CAAX was constructed using the pCS2 backbone. In both cases, the respective inserts were amplified from zebrafish cDNA and assembled into the linearized backbone using Gibson Assembly (ABclonal, catalog No. RK21020). Capped mRNAs were generated from linearized plasmids using the mMessage mMachine T7 or SP6 kit (Ambion, Thermo Fisher catalog No. AM1344; AM1340) at 37 °C for 2–3 h. DNase I treatment was applied to degrade the template DNA, and the synthesized mRNA was purified using the RNeasy Mini Kit (Qiagen, catalog No. 74104). Purified mRNA was aliquoted, diluted, and stored at −80 °C.

Morpholinos and embryo microinjection

Embryo microinjection was performed at the 1-cell stage. Injection materials and quantities used were as follows: (1) transgenesis: 10–30 pg Tol2-based vector DNA and 150 pg transposase mRNA per embryo; (2) mutagenesis by CRISPR/Cas9: 400 pg of each tyr gRNA and 1/12 dilution of commercial Cas9 NLS protein (NEB, catalog No. M0646M) per embryo; (3) mRNA overexpression and MO injection: dosages were provided in the figure legends. The sequences of the MOs were reported before and as follows: smad1 MO: 5′-AGGAAAAGAGTGAGGTGACATTCAT-3′²⁴; smad5 MO: 5′-AACAGACTAGACATGGAGGTCATAG-3′²³; smad9 MO: 5′- TGCATCGTGAAACGGGTTGATTTTA-3′²⁵; standard control MO: 5′-CCTCTTACCTCAGTTACAATTTATA-3′.

Inhibitor treatment

Inhibitors were dissolved in DMSO, and stock solutions were freshly diluted in Holtfreter’s solution before use. Embryos treated with corresponding concentrations of DMSO were used as controls.

To inhibit BMP signaling, embryos were treated with 2 μM, 5 μM, or 8 μM LDN-193189 (Selleck, catalog No. S2618) during the indicated developmental stages. To inhibit ATR–pChk1 checkpoint activation, embryos were treated with AZ20 (100 nM or 2 μM; Shanghai Yuanye Bio-Technology, catalog No. S81255) from the sphere stage, with treatment lasting until 1 or 2 dpf as indicated in the figure legends. To inhibit ATM–pChk2 checkpoint activity, embryos were treated with 2 μM KU60019 (Selleck, catalog No. S1570) from the sphere stage to 2 dpf.

To validate the effectiveness of pChk2 antibodies in detecting double-strand breaks (DSBs), embryos were treated from 3 to 6 hpf with 200 µM etoposide (MedChemExpress, catalog No. HY-13629), a topoisomerase II inhibitor that reliably induces DSBs. For rescue experiments, embryos were co-treated with 200 µM etoposide and 2 µM KU60019 during the same period.

Generation of PGC-specific knockout mutants

We employed previously established double-transgenic systems²⁰ to achieve PGC-specific knockout of smad1, smad9, or tyr in zebrafish. The smad1 and smad9 gRNA target sequences were designed using the CHOPCHOP online platform (https://chopchop.cbu.uib.no/), while the tyr gRNAs were as previously validated^34,35. For early-stage PGC analysis, Tg(piwil1:zCas9-nanos3 3′UTR);Tg(kop:Eos-CAAX-nanos3 3′UTR) females were crossed with Tg(U6:smad or tyr gRNAs;gsc:tdTomato) males. cKO embryos were selected based on tdTomato fluorescence during gastrulation, and Eos fluorescence enabled PGC visualization from the dome stage to approximately 2 dpf. For long-term PGC analysis, Tg(piwil1:zCas9-nanos3 3′UTR);Tg(piwil1:mCherry-CAAX-nanos3 3′UTR) females were crossed with Tg(U6:smad1 gRNAs;cryaa:CFP) males. mCherry expression allowed persistent PGC labeling, while CFP fluorescence in the lens (detectable from about 30 hpf) was used to distinguish cKO embryos from siblings.

Females were validated based on PGC-specific fluorescence and cas9 ISH detected in their progeny. To generate smad1&9 double knockouts, Tg(U6:smad1 gRNAs;gsc:tdTomato) and Tg(U6:smad9 gRNAs;cryaa:CFP) lines were crossed to obtain double gRNA-expressing males. These males were then mated with Tg(piwil1:zCas9-nanos3 3′UTR);Tg(kop:Eos-CAAX-nanos3 3′UTR) females to generate smad1&9-cKO embryos, which are selected by double-positive CFP and tdTomato fluorescence.

Live imaging of embryos

For live imaging of embryos during gastrulation and somite stages, the chorion was gently torn open using 1 mL syringe needles (0.45 × 16 mm RWLB) without fully detaching it from the embryo, allowing the embryo to be well-positioned and protected. Embryos were placed in custom-carved wells in solidified low-melting-point agarose in a glass-bottom confocal dish and positioned for optimal imaging. To capture dynamic PGC migration and embryo morphological changes, z-stack scanning was used to minimize loss of PGCs between frames. For embryos post-1 dpf, they were manually dechorionated (if still within the chorion), anesthetized with 0.02% tricaine, and embedded in 1% low-melting-point agarose in Holtfreter’s solution, which was covered with tricaine-supplemented Holtfreter’s solution.

Confocal and live-cell imaging were performed using a Leica TCS SP8 STED 3X or Olympus FV3000 microscope. Image processing was carried out with Imaris (Oxford Instruments) and ImageJ software. Stained tissues and phenotype images were captured in 5% methyl cellulose (Sigma-Aldrich, catalog No. M-6385) using a CCD camera under a Leica M125 C stereoscope.

PGC quantification

PGCs were quantified primarily by manual counting under a fluorescence macroscope (Olympus MVX10 equipped with a 2× Plan Apo objective, 10× eyepieces, and 6.3× zoom), complemented by confocal microscopy‑based 3D reconstruction. For manual counting, embryos were gently rotated using a fine needle while the focal plane was continuously adjusted to scan the entire embryo volume in three dimensions. This procedure allowed reliable identification and enumeration of individual PGCs at all developmental stages examined and was particularly useful in early‑stage embryos (e.g., gastrulation and early somite stages), where PGCs are deeply embedded and spatially dispersed. For later stages (e.g., 1–3 dpf), when PGCs aggregate into compact clusters within the gonadal region, embryos were additionally imaged as confocal Z‑stacks encompassing the entire PGC‑containing area. Individual PGCs were counted by systematic inspection of each optical section, and 3D reconstructions were generated to resolve closely apposed cells, providing high‑resolution spatial context for accurate quantification within dense clusters. Where both approaches were applied, PGC counts obtained from fluorescence macroscopy and confocal 3D reconstruction were highly consistent.

Immunostaining

For whole‑mount immunostaining of embryos before 1 dpf, samples were collected and fixed in 4% paraformaldehyde (PFA) overnight at 4 °C with gentle shaking. After fixation, they were washed three times in PBS for 15 min each at room temperature (RT), dehydrated through a methanol gradient (30%, 50%, 75%, 100% in PBST (0.2% Triton X-100 in PBS)), and stored in 100% methanol at −80 °C for at least 30 min. Rehydration was performed through a reverse methanol gradient. Embryonic chorions were manually removed in PBST, followed by antigen retrieval with 10 mM sodium citrate (pH 6.0) at 95 °C for 10 min. After cooling naturally to RT, embryos were washed in PBST and blocked (1% bovine serum albumin (BSA), 10% normal goat serum in PBST) for 1 h at RT. Embryos were incubated with the primary antibody diluted in blocking buffer overnight at 4 °C. Primary antibodies included rabbit anti-pSmad1/5/9 (1:200, Cell Signaling Technology, catalog No. 9511S), rabbit anti-Ddx4 (1:1000, GeneTex, catalog No. GTX128306), mouse anti-Ddx4 (1:500), rabbit anti-pChk1 (1:200, Cell Signaling Technology, catalog No. 2348), rabbit anti-pChk2 (1:200, MedChemExpress, catalog No. HY-P80799), rabbit anti-γH2AX (1:200, Cell Signaling Technology, catalog No. 2577S).

For immunostaining in embryos after 1 dpf (specifically for active Caspase‑3 detection at 2 dpf), a cryosection‑based protocol was used. Embryos were fixed, washed, dehydrated in sucrose gradients (15% and 30% sucrose-PBS), and embedded in Optimal Cutting Temperature matrix compound (SAKURA, catalog No. 4583). Embedded embryos were sectioned using a cryostat microtome (Leica CM1900) and air-dried for 30 min. Slides were washed in PBS, followed by antigen retrieval with sodium citrate, blocked in buffer, and incubated with primary antibodies overnight at 4 °C. Parafilm was used to prevent evaporation during incubation. Primary antibodies included rabbit anti-active Caspase-3 (1:200, BD Biosciences, catalog No. 559565), mouse anti-mCherry (1:200, EASYBIO, catalog No. BE2026).

After primary antibody incubation, samples (whole-mount embryos or cryosection slides) were rinsed in PBST three times for 30 min each. Next, samples were incubated with fluorochrome-conjugated secondary antibodies diluted in blocking buffer at 1:200 overnight at 4 °C. DAPI staining (1 μg/mL, Thermo Fisher) was performed for 10 min. Finally, samples were washed and mounted for imaging using a confocal microscope (Leica TCS SP8 STED 3X or Olympus FV3000).

The secondary antibodies used included Alexa Fluor 647 AffiniPure Goat anti-Rabbit IgG (H+L), Rhodamine (TRITC) AffiniPure Goat anti-Rabbit IgG (H+L), Alexa Fluor 488 AffiniPure Goat anti-Rabbit IgG (H+L), Alexa Fluor 647 AffiniPure Goat anti-Mouse IgG (H+L), Rhodamine (TRITC) AffiniPure Goat anti-Mouse IgG (H+L), and Alexa Fluor 488 AffiniPure Goat anti-Mouse IgG (H+L) (all from Jackson ImmunoResearch Labs).

EdU proliferation analysis

PGC proliferation was detected using the Cell-Light EdU Apollo643 In Vitro Kit (RiboBIO, catalog No. C10310-2) according to the manufacturer’s instructions. Briefly, manually dechorionated shield-stage embryos were incubated in Holtfreter’s solution containing 500 μM EdUTP until the bud stage. Embryos were then washed in fresh Holtfreter’s solution and incubated at 28.5 °C. At the 2-somite stage, embryos were fixed in 4% PFA overnight at 4 °C. After fixation, excess aldehyde was quenched with 2 mg/mL glycine solution for 5 min RT, followed by washes in PBS, PBST, and PBS. Embryos were stained with Apollo staining buffer for 30 min in the dark, followed by methanol and PBS washes. PGCs were labeled with Ddx4 immunostaining, and nuclei were counterstained with DAPI. The embryos were mounted for confocal microscopy (Olympus FV3000).

Proximity ligation assay

Tg(kop:gfp-CAAX-nanos3 3′UTR) embryos were collected at 75% epiboly stage and processed in accordance with the instructions of the first-day steps of the immunostaining (see “Immunostaining”), except that the blocking procedure and antibody dilution were performed using the buffers provided in the Duolink PLA kit (Sigma-Aldrich, catalog No. DUO82007 and DUO82008). Embryos were incubated overnight at 4 °C with mouse anti-Snrpb (1:200, Santa Cruz Biotechnology, catalog No. sc-374009) and rabbit anti-pSmad1/5/9 antibodies. The next day, embryos were washed twice in 1× Wash Buffer A (Sigma-Aldrich, catalog No. DUO82049) at RT for 5 min each, followed by incubation with PLA PLUS and MINUS probes (Sigma-Aldrich) diluted 1:5 in the supplied antibody diluent for 1 h at 37 °C. After two additional washes in Wash Buffer A, ligation was performed by incubating each sample with 40 μL ligation solution (1 μL ligase, 8 μL 5× ligation buffer, and 31 μL ddH₂O) at 37 °C for 30 min in a humidified chamber. Following another two washes with Wash Buffer A, amplification was carried out with 40 μL amplification solution (0.5 μL polymerase, 8 μL 5× amplification buffer, and 31.5 μL ddH₂O) for 100 min at 37 °C in a humidified chamber. Embryos were then washed twice in 1× Wash Buffer B (Sigma-Aldrich, catalog No. DUO82049) at RT for 10 min each, followed by a final wash in 0.01× Wash Buffer B for 1 min. PGCs were labeled with chicken anti-GFP antibody (1:1000, Abcam, catalog No. ab13970).

Whole-mount in situ hybridization (ISH)

Antisense RNA probes were synthesized using the DIG RNA Labeling Kit (Roche, catalog No. 11175025910) following the manufacturer’s protocol, assessed by gel electrophoresis, purified with the RNeasy Mini Kit (Qiagen), and stored at −80 °C.

ISH was performed using standard procedures. Briefly, embryos were fixed in 4% PFA, dehydrated and rehydrated through methanol gradients, and manually dechorionated. Proteinase K (1 μg/mL) was applied for 5 min to embryos prior to 2 dpf and for 10 min to embryos at 2–3 dpf, followed by re-fixation and washes. After prehybridization for 4 h at 60 °C, DIG-labeled probes (1 ng/μL for ddx4, 1.5 ng/μL for cas9) were hybridized overnight at 60 °C. Embryos were then washed, blocked, and incubated with anti-DIG-AP antibody (1:3000, Roche, catalog No. 11093274910) overnight at 4 °C. Signals were developed using BM Purple (Roche, catalog No. 11442074001), and embryos were imaged after PBS washes and post-fixation.

Cell culture, transfection, and co-immunoprecipitation (co-IP)

HEK293T cells were maintained in Dulbecco’s Modified Eagle Medium (DMEM; Gibco) supplemented with 10% fetal bovine serum (FBS; Gibco) and 1% penicillin-streptomycin (Gibco), at 37 °C in a humidified incubator with 5% CO₂.

For transfection, cells were seeded in 6 cm culture dishes such that they reached 40–60% confluency at the time of transfection on the following day. Plasmid transfection was performed using VigoFect transfection reagent (Vigorous Biotechnology, catalog No. T001) according to the manufacturer’s instructions. For co-IP assays, cells were transfected with the following plasmid combinations: (1) 3 μg of zebrafish smad1 coding sequence fused with a C-terminal 3×Flag tag (smad1−3×flag); (2) 3 μg of smad1−3×flag and 3 μg of plasmid encoding zebrafish snrpa with a C-terminal HA tag (snrpa-HA); or (3) 3 μg of smad1−3×flag and 3 μg of plasmid encoding zebrafish snrpb with a C-terminal HA tag (snrpb-HA). After 24 h, the culture medium was replaced with fresh DMEM.

Cells were harvested 48 h after transfection and washed three times with cold PBS. They were lysed on ice for 10 min in 500 μL lysis buffer (25 mM Tris-HCl, pH 7.2; 150 mM NaCl; 5 mM MgCl₂; 1% NP-40; 5% glycerol). After centrifugation at 12,000 × g for 12 min at 4 °C, the supernatant was collected. For input, 10 μL of lysate was mixed with SDS loading buffer, boiled at 100 °C for 5 min, cooled, and stored at 4 °C until further analysis. For immunoprecipitation, pre-washed Pierce Protein A/G Magnetic Agarose Beads (Thermo Fisher Scientific, catalog No. A36797) were added to the remaining lysate (≥5 μL beads per sample) and incubated overnight at 4 °C with gentle rotation. Beads were collected by brief centrifugation (1000 × g), washed 5–6 times with lysis buffer, and bound proteins were eluted in SDS loading buffer by boiling at 100 °C for 5 min. Samples were spun briefly and cooled on ice.

Proteins were resolved by 12.5% SDS-PAGE (Yeasen, catalog No. 20326ES), transferred to nitrocellulose membranes, and blocked for 1 h at RT with 5% BSA in TBST (50 mM Tris-HCl, pH 7.6; 150 mM NaCl; 0.1% Tween-20). Membranes were incubated with primary antibodies overnight at 4 °C, followed by three washes with TBST (20 min each) and incubation with HRP-conjugated secondary antibodies for 2 h at RT. After thorough washing in TBST, signals were developed using an enhanced chemiluminescence detection kit (Beyotime, catalog No. P0018S).

Primary antibodies included mouse anti-DDDDK (Flag) tag (1:10,000; MBL, catalog No. M185-3L) and mouse anti-HA tag (1:3000; MBL, catalog No. M180-3). The secondary antibody was HRP-conjugated goat anti-mouse IgG (H&L) (1:10,000; EASYBIO, catalog No. BE0102).

Fluorescence-activated cell sorting (FACS) assay

Transgenic embryos expressing fluorescent protein-labeled PGCs at desired stages were collected and dechorionated using protease treatment. Embryos were transferred to a 1.5-mL RNase-free Eppendorf tube and washed three times with Holtfreter’s solution, followed by three washes with nuclease-free PBS. For early-stage embryos (before 1 dpf), single-cell suspensions were prepared by gentle pipetting in 500 μL of pre-chilled nuclease-free PBS. For embryos after 1 dpf, a 10-min incubation in 0.25% trypsin at 28 °C preceded trituration. Samples were centrifuged at 500 × g for 5 min at 4 °C, the supernatant was discarded, and the pellet was resuspended in pre-chilled PBS or PBS with 10% fetal bovine serum to halt trypsin activity. This washing and centrifugation process was repeated 2–3 times until the pellet turned white. Cell suspensions were passed through a 35 μm mesh filter (Falcon, catalog No. 352235) to remove clumps. Cells were stained with DAPI to distinguish live and dead cells, and the suspension was sorted using a MoFlo Astrios EQ Cell Sorter.

RNA extraction

For RNA extraction from FACS-sorted PGCs or somatic cells (approximately 3000 cells) for qRT-PCR in Supplementary Fig. 2d, TRIzol reagent (Invitrogen, catalog No. 15596018) was used. Cells were lysed in 500 μL TRIzol, incubated, and then mixed with 100 μL chloroform. The aqueous phase was collected after centrifugation and mixed with isopropanol and 10 μg RNase-free glycogen (Invitrogen, catalog No. 10814010). Following incubation at −20 °C, RNA was pelleted by centrifugation, washed with 70% ethanol, air-dried, and dissolved in nuclease-free water. RNA integrity was assessed by agarose gel electrophoresis, and the concentration was measured. RNA was stored at −80 °C.

Quantitative RT-PCR

Oligo(dT)-primed first-strand cDNA was synthesized from total RNA using M-MLV reverse transcriptase (Promega, catalog No. A2791) following the manufacturer’s instructions. qRT-PCR was conducted with 3–5 ng of cDNA using TransStart Top Green qPCR SuperMix (TransGen Biotech) on a Roche LightCycler 480 II. Reactions were performed in triplicate to reduce error, and gene expression levels were normalized to β-actin using the 2 − ΔΔCT method⁶⁵. Primer sequences used for RT-PCR were listed in Supplementary Table 3.

Assessment of genome editing efficiency by targeted amplicon sequencing

Approximately 300 PGCs or somatic cells were isolated by FACS from 15-somite stage smad1-cKO and control sibling embryos. Genomic DNA was extracted and subjected to WGA using the Discover-sc Single Cell Kit (Vazyme, catalog No. N601-01) according to the manufacturer’s instructions. The amplified genome DNA was diluted and used as a template for PCR with Phanta Max Super-Fidelity DNA Polymerase (Vazyme, catalog No. P505) to amplify three target regions of smad1 gRNAs per sample (Primer sequences are listed in Supplementary Table 4). PCR products were purified by gel extraction. For each group, the three PCR products were pooled in equimolar amounts and used for library preparation with the NEBNext Ultra II DNA Library Prep Kit (NEB, catalog No. E7645S). Libraries were sequenced on an Illumina HiSeq-PE150 platform, generating 150-nt paired-end reads by Novogene.

Sequencing reads were analyzed using CRISPResso2 (v2.3.2) in “CRISPRessoPooled” mode⁶⁶, with default parameters and substitutions excluded. Notably, the parameter --min_paired_end_reads_overlap was set to 0 to retain paired-end reads even if they lacked any overlapping region. This adjustment was critical because the gRNA target sites were deliberately designed near one end of each amplicon rather than centrally, ensuring that either the forward (R1) or reverse (R2) read alone would confidently cover the target site for editing analysis. Indel frequencies at the target sites were quantified to determine genome editing efficiency. gRNA sequences were listed in Supplementary Table 1.

RNA-seq library preparation and sequencing

Full-length cDNA from fewer than 400 FACS-sorted PGCs was generated using the Single Cell Full Length mRNA-Amplification Kit (Vazyme, catalog No. N712), according to the manufacturer’s manual. Cells were lysed in 1 μL of lysis buffer with RNase inhibitor, stored at −80 °C if necessary, and reverse transcribed using an Oligo(dT) VN Primer. Sc Reverse Transcriptase added adapter sequences to the 3′ end, followed by 18 cycles of cDNA amplification. For library preparation, 1 ng of cDNA was processed with the TruePrep DNA Library Prep Kit V2 (Vazyme, catalog No. TD503), where adapters were inserted via transposase reaction and tagged fragments were amplified. After size selection (approximately 450 bp) and purification using VAHTS DNA Clean Beads (Vazyme, catalog No. N411), libraries were sequenced on the Illumina HiSeq-PE150 platform, generating 150-nt paired-end reads by Novogene.

RNA-seq data processing

RNA-seq reads were first subjected to quality control and preprocessing using fastp. This step included adapter trimming and removal of low-quality reads to ensure high-quality data for subsequent alignment. The cleaned reads were then aligned to the zebrafish reference genome (danRer11) using STAR (version 2.7.10b) with the following parameters: --twopassMode Basic --quantMode TranscriptomeSAM GeneCounts --outSAMtype BAM SortedByCoordinate --outFilterMultimapNmax 20 --alignSJoverhangMin 8 --alignSJDBoverhangMin 1 --outFilterMismatchNmax 999 --outFilterMismatchNoverReadLmax 0.04 --alignIntronMin 20 --alignIntronMax 1000000 --alignMatesGapMax 1000000 --outFilterType BySJout. Aligned reads were then counted for each gene using featureCounts (version 2.0.3). StringTie (version 2.2.1) was used to calculate TPM (Transcripts Per Million) values with the following parameters: stringtie -e -B -G Danio_rerio.GRCz11.108.gtf.

To identify differentially expressed genes, read counts were imported into DESeq2 with significance thresholds P-value < 0.01 and |log2FoldChange| > 1. Volcano plots were generated using ggplot2 (version 3.5.1) to display log2(FoldChange) vs. −log10(P-value).

Functional annotation was conducted using the Database for Annotation, Visualization and Integrated Discovery bioinformatics resource^67,68. GO terms for each functional cluster were summarized as a representative term, and the statistical significance of these terms was visualized using P-values.

Alternative splicing analysis and validation by PCR

Alternative splicing events were analyzed using rMATS (version 4.1.2) across five basic splicing patterns using two (control vs. smad1-KD) or three (control vs. smad1-cKO) biological replicates. BAM files were aligned to the zebrafish genome (GRCz11.108.gtf) with the following parameters: rmats.py --b1 control.txt --b2 smad1-cKO (or smad1-KD).txt --gtf Danio_rerio.GRCz11.108.gtf -t paired --readLength 150. rMATS (version 4.1.2) detects differential splicing by employing a hierarchical framework to model exon inclusion levels (denoted as PSI), accounting for estimation uncertainty in individual replicates and variability among replicates. Differential splicing events were filtered based on FDR ≤ 0.01 and |ΔPSI| > 0.1. Results were visualized using rmats2sashimiplot (version 3.0.0).

To validate selected alternative splicing events, full-length cDNA prepared from FACS-sorted control and smad1-cKO PGCs (see “RNA-seq library preparation and sequencing”) was used as a template for PCR amplification using Phanta Max Super-Fidelity DNA Polymerase. Amplified products were resolved on 2% agarose gels or subjected to Sanger sequencing. Primer sequences used for validation were listed in Supplementary Table 5.

Quantification, statistical analysis, and reproducibility

Fluorescence intensity measurements were performed using ImageJ (Version 1.54F). For immunostaining, the central Z-plane was selected, and nuclear regions were delineated using the DAPI signal. Cytoplasmic regions were outlined adjacent to the nucleus. Mean fluorescence intensity was calculated as total intensity (IntDen) divided by the area (Area). All experiments were performed at least twice with similar results. Specifically, experiments corresponding to Figs. 1d–g, 2f, 3h, i, 5c, 5e, 6e, f, 7e, and Supplementary Figs. 1a, b, 2d, 2f, g, 3, 5c, d, and 6b were repeated independently 2 times with comparable results. All other experiments are shown in the main and Supplementary Figs. were repeated independently 3 times or more. For RNA-seq, three biological replicates per group were used for comparisons between sibling control and smad1-cKO PGCs at each developmental stage, and two biological replicates per group were used for comparisons between PGCs from control-MO-injected and smad1-KD embryos at each stage. Statistical analyses and graphs were generated using GraphPad Prism 8, with error bars representing mean ± SD and center line displaying the mean in column graphs. Box and whisker plots showed all points from minimum to maximum, with the median indicated by the center line and whiskers displaying 1.5 times the IQR. Statistical details, including sample sizes and methods, are provided in figure legends. All Student’s t-tests were unpaired, and P-values were two-sided for both Student’s t-tests and Fisher’s exact test.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Conception and design of the research: A. Meng, X. Wu, and T. Zheng. Drafting the manuscript: A. Meng, X. Wu, and T. Zheng. T. Zheng performed most experiments. Y. Li, W. Zhang, and C. Xing contributed to the generation of Tg(piwil1:zCas9-nanos3 3’UTR). G. Li, Z. Wei, and J. Li assisted with the co-IP assay. Z. Jiang helped with the PLA experiment and participated in some discussions. R. Shah assisted with zebrafish husbandry and embryo sampling.

Peer review

Peer review information

Nature Communications thanks Andrew Nelson and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The RNA-seq data of PGCs are available from NCBI GEO under accession code GSE299653. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-70624-8.