R-loop homeostasis at transposable elements safeguards transcriptional silencing in replication-quiescent oocytes

Abstract

R-loops, though implicated in genome stability, have poorly defined transcriptional roles due to confounding replication processes in proliferating cells. Here, by leveraging replication-free oocytes and RTACC-seq, a tailored high-specificity R-loop mapping approach for low-input samples, we identify dynamic R-loop enrichment at both gene promoters and distal transposable elements. We find that transposable element-associated R-loops accumulate in transcriptionally active stage oocytes and resolve in fully grown oocytes, coinciding with chromatin condensation and transcriptional silencing. R-loops exert opposing, location-dependent regulatory functions, with reduced transcription rate at promoters and enhanced rates at transposable elements, raising the possibility of enhancer-like activity at a subset of hybrid-associated transposable element loci. In mice, oocyte-specific deletion of Rnaseh1, a gene that encodes an RNA:DNA hybrid-specific endonuclease, leads to persistent R-loops, chromatin decondensation, and impaired transcriptional silencing in fully grown oocytes, culminating in premature ovarian failure and female subfertility. These findings characterize the mechanistic basis of R-loops as mediators of chromatin remodeling, transcriptional regulation, and germ cell developmental competence.

R-loops have poorly defined transcriptional roles. Here, the authors use high-specificity R-loop mapping to show that R-loops at transposable elements enhance transcription in growing oocytes, while RNase H1-mediated resolution in fully grown oocytes supports genome silencing and female fertility.

Introduction

R-loops are emerging as context-dependent regulators of gene expression and chromatin architecture, influencing both transcriptional activity and genome stability^1–3. These three-stranded structures form when an RNA strand hybridizes with the complementary DNA, displacing the non-template strand as a single-stranded DNA (ssDNA). R-loops predominantly arise in cis at actively transcribed loci, where negative supercoiling behind RNA polymerase promotes RNA-DNA hybridization^4–6. Their effects largely depend on their orientation relative to replication fork progression. In a co-directional orientation, R-loops can stabilize stalled forks, preventing genome instability and acting as transcription-replication couplers⁷. Additionally, R-loops regulate DNA topology by modulating supercoiling and recruiting topoisomerases, influencing transcription elongation efficiency⁸. However, in a head-on orientation, R-loops obstruct replication fork advancement, leading to replication stress and DNA damage^7,9. Beyond their role as transcriptional byproducts, trans-acting R-loops form when an RNA strand from a different locus or an exogenous source invades homologous DNA sequences, as seen in RAD51-mediated telomeric-repeat-containing RNA invasion during homology-directed repair, highlighting their role in genome maintenance^5,10,11. Given this functional pleiotropy, conventional systems struggle to disentangle the direct transcriptional consequences of R-loops from their broader impact on genome stability.

Oogenesis provides a unique model to study transcription-associated R-loop formation and regulation, as it is entirely devoid of DNA replication but characterized by extensive transcriptional activity. As oocytes transition from the non-growing stage (NGO) in early primary follicles to the growing stage (GO) in late secondary follicles, the emergence of a non-surrounded nucleolus (NSN) marks the onset of transcriptional activation and maternal transcript accumulation^12,13. In the antral stage, fully grown oocytes (FGO) undergo chromatin condensation, forming a surrounded nucleolus (SN) and initiating global transcriptional silencing^14,15. This timely transcriptional repression is crucial for oocytes to acquire meiotic and fertilization competence^16–18, as disruptions in this process can lead to developmental failure and infertility^19–21. The precise mechanisms governing this transcriptional shutdown remain unclear, raising the question of whether R-loops contribute to chromatin reorganization and genome silencing.

Growing evidence suggests that R-loops enriched at transposable elements (TEs) contribute to transcriptional activation and chromatin modulations²². TEs are classified into DNA transposons, which move without RNA intermediates, and retrotransposons, which transpose via an RNA intermediate, with retrotransposons comprising over 96% of TEs in the mouse genome²³. Among these, long terminal repeat (LTR) elements can serve as alternative promoters and/or enhancers, initiating over 15% of transcripts in mouse oocytes^24–27. LTR-driven transcription plays a critical role in gene regulation, epigenetic reprogramming, and species-specific DNA methylation patterns in FGOs and preimplantation embryos^24–26,28. In mouse oogenesis, mammalian apparent LTR retrotransposons (MaLRs) and endogenous retrovirus-K (ERVK) elements contribute most significantly to TE-initiated transcription^26,28. However, the interplay between TE-driven transcription, R-loop regulation, and epigenetic modifications remains poorly understood during oocyte development, warranting further investigation. Ribonuclease H1 (RNase H1), a key enzyme in R-loop resolution, contains an N-terminal DNA-RNA hybrid binding domain (HBD) and a C-terminal endonuclease catalytic domain²⁹, capable of cycle-independent recognition and hydrolysis of RNA in DNA-RNA hybrids, thus preventing excessive R-loop accumulation³⁰. Loss of Rnaseh1 in male mice results in pachytene-stage spermatogenesis arrest due to unresolved R-loops and transcriptional dysregulation³¹, emphasizing the importance of tightly regulated R-loop dynamics in genome stability and fertility. However, whether RNase H1-mediated R-loop resolution is required for TEs silencing and oocyte maturation remains unknown.

Precise genome-wide R-loop mapping assists in discovering loci-specific R-loop functions; conventional detection and sequencing methods have limitations in specificity, resolution, and scalability³². Widely used RNA:DNA immunoprecipitation sequencing (DRIP-seq) relies on S9.6 immunoprecipitation, but suffers from GC bias³³ and cross-reactivity with double-stranded RNA (dsRNA)³⁴. Alternative S9.6-independent methods, such as R-loop chromatin immunoprecipitation (R-ChIP), leverage the natural affinity of the HBD of catalytically inactive RNase H1 for detection, but require intracellular expression of tagged RNase H1, posing technical challenges and potential disruptions to native R-loop landscapes and normal transcriptional processes or chromatin interactions³⁵. Strand-specific k-methyloxazolone-assisted single-stranded DNA sequencing (spKAS-seq) maps ssDNA regions overlapping with R-loops but necessitates RNase H treatment to confirm specificity, as it may capture non-R-loop ssDNA structures³⁶. Recently developed R-loop identification assisted by nucleases and sequencing (RIAN-seq) specifically degrades single-stranded RNA (ssRNA), ssDNA, and double-stranded DNA (dsDNA) before library construction, but may result in signal loss and require large cell quantities, limiting its suitability for ultra-low sample sequencing. Given these constraints, a high-specificity, high-resolution, and low-input R-loop mapping method is necessary to dissect transcription-associated R-loops during oogenesis.

To circumvent this, we developed a highly specific and efficient R-loop detection and library construction approach for oocytes, leveraging a tailored small-scale Tn5-assisted chromatin cleavage with sequencing³⁷. Our results revealed a strong R-loop enrichment at TEs, particularly LTRs, rather than promoters, in parallel with transcriptional upregulation during NGO-to-GO development, while R-loop elimination in FGOs coincided with stabilized transcript levels. Furthermore, oocyte-specific Rnaseh1 knockout further demonstrated that RNase H1 depletion impairs follicular development and significantly reduces female fertility. Mechanistically, persistent, unresolved R-loops, especially at TEs, disrupted transcriptional silencing during the GO-to-FGO transition by increasing chromatin accessibility at transcription start sites (TSSs). These findings underscore the essential role of RNase H1-mediated R-loop resolution in maintaining transcriptional control independently of replication, thereby safeguarding oocyte developmental competence and ensuring female fertility.

Results

RTACC-seq enables ultra-low mapping of bona fide R-loop dynamics

R-loops are emerging regulators of transcription, yet in most proliferating systems, their formation and consequences are difficult to disentangle from replication-associated effects. Oogenesis provides a unique context to study co-transcriptional R-loop dynamics since it proceeds without DNA replication while undergoing extensive, stage-specific transcriptional remodeling. To define R-loop distribution across this developmental program, we established a modified ultra-low-input Tn5-assisted chromatin cleavage sequencing approach³⁷ for high-specificity, genome-wide R-loop profiling (RTACC-seq). Nuclear localization signal (NLS)-HBD repeats fusion sequence (NLS-3×HBD-3×HA) is generated to enhance RNA: DNA hybrid-specific detection and library construction (Fig. 1a). NLS ensures that the translated HBDs localize exclusively and efficiently to the nucleus rather than in cytoplasmic mitochondria^29,35, while HBD tandem repeats are crucial for recognizing and tightly binding to hybrids in a sequence-independent manner^10,38. The HBD-hybrid binding may promote R-loop capture efficiency by stabilizing R-loop and reducing its degradation during the cell lysis and nuclear DNA extraction processes. Furthermore, this method effectively minimized mitochondrial DNA contamination and enabled precise R-loop genome-wide mapping.

Fig. 1. Genome-wide detection of R-loops by RTACC-seq during early oogenesis.

a Schematic representation of sample collection and RTACC-seq workflow in HEK293T cells and mouse oocytes. In both systems, RNA:DNA hybrids are captured via HA antibody-guided pA/G-Tn5 tagmentation for small-scale library construction and sequencing. NGO non-growing oocyte, GO growing oocyte, FGO fully grown oocyte, NLS nuclear localization sequence, HBD RNA:DNA hybrid binding domain. b Genome-browser example illustrating RTACC-seq signal in HEK293T cell and co-localization with established R-loop mapping signals. c Metaplot of RTACC-seq signal centered on R-loop peaks (±2 kb) in GO under three matched conditions. d Box plot representing R-loop signal intensities in reads per kilobase of transcript per million mapped reads (RPKM). Center line, median; box limits, upper and lower quartiles; whiskers, 1.5 × interquartile range. RTACC-seq analyses were performed using n = 2 biologically independent RTACC-seq libraries per condition unless otherwise indicated. e Bar chart showing the total number of R-loop peaks at each oocyte stage. f R-loop redistribution across different oocyte developmental stages. g Distribution of R-loop signals across gene promoters and distal regions across stages. h Observed/expected enrichment of RTACC-seq peaks across distal region classes at each stage. i Representative snapshots from the UCSC Genome Browser at LTR-associated F13b showing hybrid-dependent RTACC-seq enrichment. Source data are provided as a Source Data file.

We first validated the performance and fidelity of RTACC-seq in HEK293T cells, where sufficient material enables rigorous cross-method benchmarking. We generated HEK293T RTACC-seq libraries across a titration of cell inputs and used these data to assess low-input reproducibility and concordance with multiple published R-loop mapping datasets in the same cell type. RTACC-seq generated from as few as 100 cells recapitulated bulk profiles across the genome (Supplementary Fig. 1a) and showed high concordance (R = 0.95) with bulk RTACC-seq across genes (Supplementary Fig. 1b), demonstrating robust signal capture at ultra-low input. Across genes, both ultra-low and bulk input RTACC-seq signals positively correlated with multiple orthogonal R-loop mapping approaches, including RChIP, DRIPc, MapR, and CUT&Tag using S9.6 or HBD-based fusions (Supplementary Fig. 1c), and co-localized with these signals at representative loci (Supplementary Fig. 1d). We additionally included PRO-seq tracks at representative loci to visualize nascent transcription, confirming that the displayed RTACC-seq signals occur in actively transcribed genomic contexts and highlighting that RTACC-seq detects R-loop enrichment not only at promoter/TSS but also at transcriptionally active distal regulatory elements, such as enhancers (Fig. 1b and Supplementary Fig. 1e). Since enhancer RNAs are often short-lived and low-abundance, they can be under-represented in steady-state RNA measurements³⁹, making nascent transcription context particularly informative for interpreting distal R-loop signals.

Having established robust performance of ultra-low input RTACC-seq in HEK293T, we next validated hybrid-dependent specificity in 100 oocytes using GO, the high-transcriptional stage for parallel controls: an experimental hybrid-binding fusion (NLS-3×HBD-3×HA), a negative control by RNase H treatment to remove hybrids (NLS-3×HBD-3×HA + RNase H), and a non-binding negative control (NLS-mCherry-3×HA). A genome-browser example showed robust signal in the HBD condition, while RNase H treatment collapsed the signal toward baseline, consistent with hybrid dependence (Supplementary Fig. 1f). Critically, enrichment was absent in the mCherry control (Supplementary Fig. 1f), arguing against the signal being driven solely by Tn5 transposase bias towards accessible chromatin. In addition, metaplot analysis revealed a strong enrichment of R-loop peaks over NLS-3×HBD-3×HA, which was markedly reduced after RNase H treatment and absent in mCherry control (Fig. 1c). Moreover, the GO RTACC-seq signal is positively correlated with NSN oocyte KAS-seq profiles¹² (Supplementary Fig. 1g). The correlation is expected to be modest since the two datasets are not completely stage-matched and the two assays capture related but not identical molecular features, with RTACC-seq detecting hybrid enrichment while KAS-seq target ssDNA exposure. In addition, GO accumulates a progressively expanding maternal RNA pool, which can compress the dynamic range of RNA-based metrics and attenuate locus-level correlations. In comparison, RTACC-seq yields more punctate enrichment tracks (Supplementary Fig. 1h), suggesting improved apparent locus definition for the R-loop signal. Collectively, these analyses demonstrate that RTACC-seq faithfully captures nascent R-loops with low-input reproducibility, cross-assay concordance in the same cell type, and hybrid-dependent specificity.

Stage-specific dynamics of R-loop formation and resolution during oogenesis

We next applied RTACC-seq to profile R-loop distribution during early oogenesis across three key developmental stages: NGOs, harvested from early secondary follicles at postnatal day 7 (PD-7, 20–30 μm), GOs, from late secondary follicles at PD-14 (40–50 μm), and FGOs, from antral follicles 44 h post-injection with 5 IU pregnant mare serum gonadotropin at PD-28 ( ~ 70 μm) (Fig. 1a and Supplementary Fig. 2a). This developmental window spans the transition from transcriptional activation during oocyte growth to transcriptional silencing in FGO. For oocyte RTACC-seq, NLS-3×HBD-3×HA mRNA was microinjected into the cytoplasm, followed by nuclear accumulation of the translated fusion protein driven by the NLS (Fig. 1a). Genome-wide quantification revealed stage-dependent R-loop regulation, with R-loop peak signals exhibiting a modest increase from NGO to GO but declining sharply in FGO (Fig. 1d, e). Notably, most of the previously undetectable R-loop peaks in NGO became enriched in GOs, while half of the initial R-loop peaks were lost (Fig. 1f). The proportion of hybrid-marked regions increased from 57 to 70% from NGO to GO, but dramatically dropped from 70% in GO to ~ 1% in FGO, accompanied by a corresponding expansion of unmarked regions from 30 to 99% (Fig. 1f). The stage-specific R-loop accumulation is likely driven by active transcription during NGO-to-GO transition, whereas the subsequent loss in FGO is consistent with chromatin remodeling and global transcriptional repression associated with oocyte maturation¹⁷. This trend was evident in both promoter and distal regions (Fig. 1g). Given the observation that R-loop peaks were biased toward distal regions (Fig. 1g), we performed observed/expected framework⁴⁰ to quantify the genomic occupancy of R-loop peaks during oocyte growth. Within distal sites, R-loop peaks were preferentially enriched at TE loci, with the strongest enrichment over LTR elements, followed by short interspersed nuclear elements (SINEs) (Fig. 1h). Moreover, an LTR-associated example locus F13b further illustrates HBD-dependent signal enrichment at distal TEs (Fig. 1i), suggesting genome-wide specificity.

Since TEs, especially those retrotransposons, can also contribute to alternative promoters/TSSs²⁸, we directly compared RTACC-seq enrichment between promoter-proximal and distal TE annotations (Supplementary Fig. 2b). Notably, R-loop peaks were consistently enriched in distal TEs relative to promoters across all three stages (Supplementary Fig. 2b), reinforcing that distal TE-associated regions represent a dominant compartment of the oocyte R-loop landscape. Meanwhile, since TE annotations are pervasive in the genome, we next asked whether this enrichment could be explained simply as a byproduct of pervasive host transcription. To address this possibility, we used an observed/expected framework to resolve RTACC-seq peak occupancy across distal no-TE regions, gene bodies, and TE classes (Supplementary Fig. 2c). This analysis again highlighted preferential enrichment at LTR beyond what would be expected from their genomic abundance, arguing against a purely co-transcriptional byproduct explanation, and instead, supporting a distal TE-biased R-loop accumulation in GO. Furthermore, we additionally stratified genes into low- (FPKM < 5, Cluster I) and high-expression (FPKM > 5, Cluster II) groups and performed a gene-body analysis, which is more stable at low-input and relies largely on uniquely mappable genic reads (Supplementary Fig. 2d, e). Despite the marked difference in RNA abundance between these groups (Supplementary Fig. 2d), the gene-body R-loop signal was not significantly different (Supplementary Fig. 2e), indicating that R-loop enrichment over intragenic TE segments does not simply scale with host-gene expression and may reflect mechanisms distinct from distal TE-associated R-loops. Collectively, these analyses reveal a spatiotemporally regulated program of R-loop and identify distal TEs, particularly LTR, as the primary genomic context for R-loop accumulation during oogenesis.

Genomic context-specific regulation of transcription by R-loops

During the NGO-to-GO transition, 18% of genes gained de novo R-loop signal, while only 6% lost it, yielding a net expansion of R-loop-marked genes (Supplementary Fig. 3a). Conversely, the GO-to-FGO transition was dominated by a widespread R-loop resolution, with ~92% of genes showing signal elimination and ~1% retaining it (Supplementary Fig. 3a). These trends mirrored the stage-wise peak turnover (Fig. 1f), consistent with R-loop accumulation coinciding with transcriptional activation in GO, followed by efficient resolution during transcription silencing in FGO. Having established the spatiotemporal distribution of oocyte R-loops, we next examined how promoter and distal R-loop signals remodel across oocyte growth and how these dynamics relate to transcriptional output.

Promoter RTACC-seq profiles across three stages resolved three major trajectory classes, including loci showing (1) a modest decrease of R-loops from NGO to GO, followed by a progressive decline into FGO, (2) maximal R-loop enrichment at GO then decline thereafter, and (3) a steady decrease in R-loops across growth (Fig. 2a). Consistent with the global loss of R-loops during GO-to-FGO, the largest divergence among promoter trajectories occurred during NGO-to-GO, when both increases and decreases were observed across loci (Fig. 2a). This bidirectional window was therefore used for a focused two-stage stratification. We obtained four promoter transition classes that separate loci with stable-high (Clusters I), stable-low (Clusters IV), lost (Clusters II), or gained (Clusters III) R-loop signal from NGO to GO (Fig. 2b–d). Since total RNA in oocytes increases continuously from NGO to GO due to large-scale maternal RNA accumulations (Supplementary Fig. 3b), absolute RNA-seq abundance is strongly influenced by this expanding background pool; we therefore used an RNA accumulation index as a normalized net RNA gain between stages to approximate the transcriptional output rate. This index quantifies the difference in mRNA levels between GO and NGO for genes in each cluster, enabling relative comparisons across loci while accounting for the progressively increasing RNA background during oocyte growth. Intriguingly, R-loop-gain Cluster III showed lower RNA accumulation than R-loop-loss Cluster II (Supplementary Fig. 3c). Consistently, transcriptome-wide correlation analysis confirmed a negative relationship between promoter-associated R-loop level and the transcriptional output during this stage (Fig. 2e). Given that this process is characterized by an overall increase in RNA levels (Supplementary Fig. 3b), this inverse relationship suggests that promoter-associated R-loops are not a simple readout of transcriptional activation and may instead be linked to reduced transcriptional efficiency or altered RNA output at a subset of loci during oocyte growth.

Fig. 2. R-loop accumulation and transcription during NGO-to-GO transition.

a Clustering of promoter-associated R-loop signal across developmental stages. Blue lines, individual loci; red lines, cluster medians. b Promoter-associated R-loop signal in NGO and GO. Center line, median; box limits, upper and lower quartiles; whiskers, 1.5 × interquartile range. n, number of promoters. Exact p values are indicated in the panel and were calculated by two-sided unpaired Student’s t tests, unless otherwise indicated. Values below numerical precision are shown as p < 1e⁻⁴. c Heatmap of promoter-associated R-loop signal changes from NGO to GO. d Representative browser views of promoter-associated R-loop signal at Tab2 (Cluster II) and Stip1 (Cluster III), with locus-specific log2 fold change in R-loop signal (top) and RNA accumulation (bottom). RNA accumulation index = FPKM_(GO-NGO)/(FPKM_NGO + 0.1). RTACC-seq tracks are representative of n = 2 biologically independent RTACC-seq libraries per condition unless otherwise indicated. e Correlation between R-loop fold change and RNA accumulation from NGO to GO (n = 7078 promoters). Red line, linear regression fit. f Clustering of distal-associated R-loop signal across stages. g Distal TE loci stratified as R-loop⁻ or R-loop⁺ in NGO and GO based on RTACC-seq enrichment. R-loop⁻ loci, RPKM < 10. h RNA-seq signal over the distal TE loci shown in (g). NGO R-loop⁻: n = 27,276; NGO R-loop⁺: n = 31,478; GO R-loop⁻: n = 23,672; GO R-loop⁺, n = 36,374 distal TE loci. Data are presented as mean ± s.e.m. unless otherwise indicated. i Distal TE-associated R-loop signal in each cluster. Boxplot elements are as in (b). n, number of distal TE loci. j Heatmap of distal TE RTACC-seq changes from NGO to GO. k Representative browser views of distal-associated R-loop signals at Btg4 (cluster II) and Dsc2 (cluster III), with locus-specific log2 fold change in R-loop signal (top) and RNA accumulation (bottom). l Overlap between promoter and distal TE R-loop-gain genes. m RNA accumulation for promoter-only, overlapping, and distal TE gene sets from (l). n, number of genes. n RNA accumulation index with increasing distance from Cluster III R-loop-gain TE. n, number of genes. Source data are provided as a Source Data file.

In parallel, clustering at distal regions revealed an overall R-loop dynamics similar to promoters across three stages, with a bidirectional change from NGO to GO, followed by a universal decline from GO to FGO (Fig. 2f). Since R-loops form co-transcriptionally, we next asked whether distal TE loci show evidence of transcription when they display RTACC-seq enrichment. We stratified distal TE loci into R-loop⁺ and R-loop⁻ groups based on RTACC-seq enrichment in both NGO and GO (Fig. 2g). RNA-seq quantification showed significantly higher RNA-seq signal at those R-loop⁺ loci than at R-loop⁻ loci in both stages (Fig. 2h), supporting the association between hybrid-enriched distal TEs and transcription in oocytes. Direct locus-by-locus coupling of R-loops with ongoing transcription is challenging in oocytes due to input limitations, so a locus-resolved validation was performed in HEK293T cells using uniquely mapped RTACC-seq and PRO-seq reads. Consistent with the oocyte RNA-seq result, R-loop⁺ distal TEs exhibited stronger PRO-seq signals than their R-loop⁻ counterparts (Supplementary Fig. 3d–f), and PRO-seq signals were positively correlated with RTACC-seq across the same TE insertions (Supplementary Fig. 3g), indicating co-occurrence of hybrids and nascent transcription at individual loci.

Functionally, distal TE-associated R-loops might behave differently from promoter R-loops. R-loop-gain Cluster III showed a subtle but non-significant higher in RNA accumulation compared with R-loop-loss Cluster II (Fig. 2i–k and Supplementary Fig. 3h), indicating that the dynamic gain of these distal TE-associated R-loops is not accompanied by a uniform increase in transcriptional output across loci during NGO-to-GO transition. This pattern suggests that the transcription of hybrid-associated TE might also be modulated by promoter-associated R-loop at nearby regions, such that effects from distal TE-associated and promoter-proximal hybrids may be integrated at the same locus, thereby diminishing the overall influence. To examine this possibility, we stratified genes according to whether they carried promoter-only, distal TE-only, or shared R-loop-gain Cluster III signatures. Indeed, 351 genes with both promoter- and TE-associated gain of R-loop during the transition (Fig. 2l) showed intermediate RNA accumulation relative to promoter-only or TE-only Cluster III genes (Fig. 2m), suggesting either combined or partially non-redundant regulatory contributions from R-loops at these two genomic contexts. Moreover, transcription gradually attenuated with increasing distance from Cluster III hybrid-enriched TE loci (Fig. 2n), suggesting a possible distance-dependent regulatory influence between TE-linked R-loops and nearby transcriptional activity.

Together, these findings argue that R-loops are not merely passive byproducts of transcription. Instead, their relationship with transcription is context-dependent, with promoter-associated R-loops showing a modest negative relationship with transcriptional output, whereas distal TE-associated R-loops are broadly associated with transcriptional activity, but only a subset appears functionally linked to transcriptional regulation.

Oocyte-specific RNase H1 deletion leads to premature ovarian failure and reduced fertility

Building on the observed R-loop dynamics and corresponding transcriptional changes, we next prompt to investigate the underlying mechanisms responsible for R-loop regulation. RNase H1, a ribonuclease that specifically hydrolyzes the RNA strand in RNA:DNA hybrid of R-loop, and is therefore essential for preventing abnormal R-loop accumulation²⁹. Immunofluorescence (IF) analysis showed that nuclear RNase H1 levels gradually increased during oocyte development after primordial follicle activation (Fig. 3a, b). To assess its physiological role, we generated oocyte-specific conditional Rnaseh1 knockout mice (Rnaseh1^oo−/−) by crossing Rnaseh1-floxed mice (Rnaseh1^fl/fl) with Gdf9-Cre transgenic mice, enabling the specific deletion of Rnaseh1 in oocytes as early as the primordial follicle stage (Supplementary Fig. 4a). Loss of RNase H1 protein in oocytes was validated by both IF (Fig. 3c) and western blot analysis (Fig. 3d), confirming the absence of RNase H1 protein in FGO from Rnaseh1^oo−/− mice. Functionally, Rnaseh1^oo−/− females displayed a severely compromised reproductive capacity, with only 1–2 pups born in in the first litter, followed by complete infertility (Fig. 3e). Ovarian histology in 5-month-old Rnaseh1^oo−/− mice showed shrunken ovaries, follicle loss, and germ cell depletion (Fig. 3f and Supplementary Fig. 4b), with significantly reduced ovarian weights (Fig. 3g). Similar degeneration was observed by 10 weeks of age, with ovaries nearly devoid of mature follicles and predominated by atretic oocytes (Fig. 3h and Supplementary Fig. 4c), indicative of premature ovarian failure (POF). Consistently, ovarian weights in Rnaseh1^oo−/− mice were significantly reduced (Fig. 3i). Collectively, these findings demonstrate that RNase H1 plays an essential role in follicle maintenance and ovarian function, and its loss in oocytes leads to partial infertility and POF.

Fig. 3. Oocyte-specific deletion of Rnaseh1 impairs ovarian function and female fertility in mice.

a IF confocal analysis of RNase H1 expression in primordial, primary, secondary, and antral follicles. Representative oocytes in the corresponding ovaries are shown in enlarged views below. Scale bars, 50 μm. b Quantification of the RNase H1 immunofluorescent signals from (a). n = 6 oocytes per stage. Unless otherwise indicated, data are presented as mean ± s.e.m. and exact p values are indicated in the panels by two-tailed unpaired Student’s t tests. c IF confocal analysis of RNase H1 expression in from Rnaseh1^fl/fl and Rnaseh1^oo−/− FGO. Scale bars, 50 μm. d Western blot analysis of RNase H1 protein levels in 100 FGOs isolated from Rnaseh1^fl/fl and Rnaseh1^oo−/− mice. DDB1 was blotted as a loading control. Molecular weight markers are shown at right. e Fertility test of female mice. Rnaseh1^fl/fl (n = 6) and Rnaseh1^oo−/− (n = 5) female mice were mated with wild-type males for 6 months, and the average number of pups per female was quantified. f–j Hematoxylin and Periodic acid Schiff (H&PAS) staining and immunohistochemistry (IHC) analysis of mouse vasa homolog (MVH) expression in 5-month-old (f), 10-week-old (h), and 3-week-old (j) Rnaseh1^fl/fl and Rnaseh1^oo−/− mouse ovaries, respectively. Scale bars, 50 μm. Bar graph comparing the ovarian weights of Rnaseh1^fl/fl and Rnaseh1^oo−/− mice at 5 months (g) and 10 weeks (i). n, number of ovaries per group. k Quantification of follicle numbers at different developmental stages in the ovaries of 3-week-old Rnaseh1^fl/fl and Rnaseh1^oo−/− mice (n = 4 ovaries per genotype). IHC analysis of cleaved caspase 3 expression in ovaries of 3-week-old Rnaseh1^fl/fl and Rnaseh1^oo−/− mice (l), and the percentage of atretic follicles per ovary surrounded by cleaved caspase 3-positive GC cells (m) in (l). n, number of ovaries per genotype. Scale bar, 50 μm. Source data are provided as a Source Data file.

To pinpoint the onset of POF in Rnaseh1^oo−/− mice, we again performed histological analyses of ovaries from 3-week-old mice. Some follicles with clear signs of degeneration in the ovary from Rnaseh1^oo−/− mice (Fig. 3j and Supplementary Fig. 4d), possibly by pyknosis or cytoplasmic fragmentation/degradation, indicating an early onset of ovarian follicle depletion⁴¹. Although some follicles in Rnaseh1^oo−/− mice were able to develop into the antral stage, the overall number of primordial and primary follicles was significantly reduced, whereas the proportion of atretic follicles was markedly increased compared to controls (Fig. 3k and Supplementary Fig. 4e). Given the importance of oocyte-granulosa cell (GC) interactions during folliculogenesis^42,43, we also examined GC health and observed increased apoptosis and reduced mitosis in 3-week-old Rnaseh1^oo−/− ovary (Fig. 3l, m and Supplementary Fig. 4f, g). These defects likely stem from impaired oocyte signaling due to RNase H1 loss. In summary, these findings suggest that RNase H1 is essential for maintaining ovarian follicle integrity. Its absence disrupts both oocyte and GC development, accelerates follicular atresia, and results in early-onset POF and ultimately contributes to female subfertility.

TEs as major foci for abnormal R-loops accumulation in Rnaseh1-deleted oocytes

RNase H1 contains a mitochondrial targeting sequence (MTS) that directs it to the mitochondria, and its knockout in mice causes embryonic lethality due to defective mitochondrial DNA (mtDNA) replication^29,44. However, IF confocal analysis in our study demonstrated that mitochondrial distribution and activity remained unaffected in RNase H1-deficient FGOs (Supplementary Fig. 4h–m), suggesting that the observed infertility was not a result of mitochondrial dysfunction. Instead, we found a significant accumulation of nuclear R-loops across the follicular stages in Rnaseh1^oo−/− oocytes (Fig. 4a, b). Since RNase H1 was predominantly localized in the oocyte nucleus rather than the cytoplasm (Fig. 3a), it likely contributes to follicle integrity via maintaining nuclear R-loop homeostasis. Under physiological circumstances, FGOs from antral follicles displayed relatively low levels of R-loops; however, RNase H1 deletion led to persistent R-loop accumulation in FGOs (Fig. 4a–d), indicating a failure in R-loop clearance during oocyte maturation. Quantification of the NSN/SN oocyte status revealed no significant difference between Rnaseh1^fl/fl and Rnaseh1^oo−/− FGOs (Supplementary Fig. 4n), indicating that oocyte-specific Rnaseh1 deletion did not affect the NSN-to-SN transition during development.

Fig. 4. Aberrant R-loop accumulation in RNase H1-deficient FGO.

a IF confocal analysis of R-loop level (S9.6) and MVH expression in 3-week-old Rnaseh1^fl/fl and Rnaseh1^oo−/− mouse ovary. Scale bar, 100 μm. b Quantification of oocyte S9.6 R-loop IF signals from (a) (n = 5). Unless otherwise indicated, data are presented as mean ± s.e.m. with exact p values by two-tailed unpaired Student’s t tests indicated in the panels. c IF confocal analysis of R-loop level (S9.6) in FGOs from 3-week-old Rnaseh1^fl/fl and Rnaseh1^oo−/− mouse ovary. Scale bar, 10 μm. d Quantification of oocyte S9.6 R-loop signals from (c) (n = 5). e R-loop peaks detected in FGOs from 3-week-old Rnaseh1^fl/fl and Rnaseh1^oo−/− mice. f R-loop peak signal (RPKM) in FGOs from 3-week-old Rnaseh1^fl/fl and Rnaseh1^oo−/− mice. Center line, median; box limits, upper and lower quartiles; whiskers extend to the most extreme data points within 1.5 × the interquartile range. g Heatmap showing the accumulation of R-loop in FGOs from 3-week-old Rnaseh1^fl/fl and Rnaseh1^oo−/− mice. h Proportion of increased R-loops in Rnaseh1^oo−/− FGOs at gene promoter or distal TEs. n, number of annotated loci. i Observed/expected ratio overlap of RTACC-seq peaks across LINEs, LTRs, and SINEs in Rnaseh1^oo−/− FGOs. Color indicates log₂(observed/expected) enrichment, and dot size indicates the enrichment ratio. j Representative browser views comparing R-loop signals between Rnaseh1^fl/fl and Rnaseh1^oo−/− FGOs. n = 2 biologically independent RTACC-seq libraries per condition. Source data are provided as a Source Data file.

To further elucidate the specific distribution of R-loops genome-wide, RTACC-seq was performed on Rnaseh1^fl/fl and Rnaseh1^oo−/− FGOs. As expected, R-loop signals were dramatically elevated in Rnaseh1^oo−/− FGOs (Fig. 4e, f). Among those 81.5% of abnormally accumulated R-loops (Fig. 4g), 35.7% mapped to gene promoters while 64.3% in distal regions (Fig. 4h). Notably, within these R-loop-enriched distal regions, 68% of the increased R-loops were associated with TEs (Fig. 4h), particularly LTR and SINE (Fig. 4i, j), suggesting that TEs are major hotspots for abnormal R-loop accumulation in the absence of RNase H1. Together, these findings highlight a nuclear-specific role of RNase H1 in clearing R-loops during oocyte maturation and suggest that unresolved R-loops at distal TE sites may underlie impaired follicle development and female subfertility.

TE-enriched R-loops can disrupt genome silencing in FGOs

Building on the observation that R-loops at distal TEs correlated with increased transcription during the NGO-to-GO transition, we next examined whether unsolved R-loops induced transcriptional dysregulation in Rnaseh1^oo−/− FGOs. During normal oocyte development, mRNA levels rise sharply from NGOs to GOs and then stabilize during the GO-to-FGO transition (Fig. 5a and Supplementary Fig. 4o), consistent with the establishment of transcriptional quiescence during the NSN-to-SN chromatin transformation, a process essential for oocyte maturation, as failure to achieve proper gene silencing has been linked to developmental incompetence and infertility^18,45. However, Rnaseh1^oo−/− oocytes displayed a continuous elevation in mRNA levels during this transition, accompanied by global transcriptional upregulation (Fig. 5a), and a sharp increase in differentially expressed genes (DEGs) (Fig. 5b), indicating that failure in transcriptional silencing likely contributes to POF and infertility.

Fig. 5. Disrupted genome silencing in RNase H1-deficient FGOs.

a Change of mRNA levels across NGO, GO, and FGO from Rnaseh1^fl/fl and Rnaseh1^oo−/− mice respectively. b Log₂ fold changes of transcripts in NGO, GO, and FGO from Rnaseh1^oo−/− mice compared to the Rnaseh1^fl/fl mice across stages. c Gene cluster analysis of gene expression pattern in NGO, GO, and FGO from Rnaseh1^fl/fl mice. Boxplot quantifying the change in mRNA levels in each cluster. n, number of genes in each cluster. Center line, median; box limits, upper and lower quartiles; whiskers extend to the most extreme data points within 1.5 × the interquartile range. d Mapping the unusual up-regulated genes in Rnaseh1^oo−/− FGOs to clusters identified in (c). e Correlation analysis between the fold change of mRNA level in Rnaseh1^oo−/− and Rnaseh1^fl/fl oocytes during GO-to-FGO transition. f Gene set enrichment analysis (GSEA) of the differentially expressed genes (DEGs) in Rnaseh1^oo−/− versus Rnaseh1^fl/fl FGO. g Heatmap comparing the mRNA (left) and R-loop (right) levels between Rnaseh1^fl/fl and Rnaseh1^oo−/− FGOs. h Comparison of R-loop (left) and mRNA (right) levels at gene promoter and TEs. Center line, median; box limits, upper and lower quartiles; whiskers extend to the most extreme data points within 1.5 × the interquartile range. i Scatter plot of transposable element expression changes in Rnaseh1^fl/fl versus Rnaseh1^oo−/− FGOs. n, number of TE loci. j Specific percentage of LTR subtypes identified in (i). ERVK Endogenous Retrovirus-K, ERVL Endogenous Retroviral-Like elements, ERV1 Endogenous Retrovirus-1, GPYSY Ty3-retrotransposons superfamily. k Heatmap showing the mRNA level (upper) and their corresponding R-loop level (lower) of LTRs identified in (i). l Gene expression levels against their distance from hybrid-enriched ERVK loci. n, number of genes in each distance bin. Data are presented as mean ± s.e.m. Exact p values are indicated and calculated by two-tailed unpaired Student’s t tests. Source data are provided as a Source Data file.

To characterize the genes abnormally upregulated in Rnaseh1^oo−/− FGOs, we analyzed transcriptional differences between Rnaseh1^fl/fl and Rnaseh1^oo−/− oocytes. Clustering analysis of Rnaseh1^fl/fl oocytes across the NGO-GO-FGO transition revealed three distinct gene expression patterns—Cluster I (2204 genes) exhibited an initial increase in expression followed by transcriptional silencing; Cluster II (3753 genes) showed an initial increase followed by stable expression, and Cluster III (4060 genes) displayed continuous upregulation throughout oocyte development (Fig. 5c). Strikingly, in Rnaseh1^oo−/− FGO, 54% of the upregulated genes (>2-fold) belonged to Cluster I, indicating that genes normally silenced at the FGO stage remained transcriptionally active due to RNase H1 depletion (Fig. 5d). In contrast, only 17% of the upregulated genes (>2-fold) fell into Cluster III (Fig. 5d), reinforcing the idea that loss of RNase H1 primarily disrupted the silencing of genes that should have been downregulated, rather than further amplifying the expression of those originally or programmed active genes. Furthermore, a negative correlation between mRNA changes in Rnaseh1^fl/fl and Rnaseh1^oo−/− FGOs indicated a different transcriptional trend (Fig. 5e). Consistently, the gene set enrichment analysis (GSEA) revealed that DEGs in Rnaseh1^oo−/− FGOs were predominantly enriched for genes typically expressed at the GO stage in the control group (Fig. 5f). Together, these findings suggest that RNase H1 depletion hinders genome silencing during GO-to-FGO transition, indicating by significant upregulation of genes that should be transcriptionally silenced for oocyte maturation.

Since we observed significant R-loop accumulation in Rnaseh1^oo−/− FGO, especially at distal TEs (Fig. 4), we were then intrigued to elucidate whether that contributed to the failure of genome silencing. Notably, genes with increased mRNA levels (>2-fold) (Fig. 5g, left) exhibited elevated R-loop formation (Fig. 5g, right), suggesting a direct link between R-loop accumulation and transcriptional derepression. This finding indicates a causal relationship wherein the loss of RNase H1 disrupts normal R-loop resolution, thereby preventing proper transcriptional silencing and leading to widespread aberrant gene activation. Notably, genes with increased mRNA levels (Fig. 5g, left) exhibited higher R-loop accumulation at distal TEs compared to promoter regions, suggesting TE-enriched R-loops, rather than promoter-proximal R-loops, serve as primary drivers of gene upregulation (Fig. 5h). Among 114 upregulated TEs (>2-fold), only 17 were LINEs, whereas 97 were LTRs (Fig. 5i). Within, 40% of the upregulated LTRs belonged to ERVK elements (Fig. 5j), which are one of the major contributors of LTR-initiated transcription in murine oocytes²⁸. In consistent, ERVK elements exhibiting heightened transcription also possessed increased R-loop levels (Fig. 5k), supporting a reciprocal relationship wherein R-loop accumulation reinforces TE transcriptional activation.

Most importantly, genes located within 5 kb of ERVK-enriched R-loops were more upregulated than those further away (Fig. 5l), aligning with the observation that hypomethylated LTRs can act as alternative promoters and drive transcription of adjacent genes⁴⁶. Additionally, a comparable increase in transcription was observed at each locus near the R-loop-enriched ERVK region (5–240 kb) in Rnaseh1^oo−/− FGO (Fig. 5l). Taken together, these findings suggest a multi-faceted process in which unscheduled ERVK-enriched R-loops interfere with gene silencing in Rnaseh1^oo−/− FGO by (1) initiating TEs transcription, (2) inducing the activation of nearby genes through TEs activity, and (3) reinforcing R-loop formation at transcriptionally active loci. In other words, unresolved R-loops at ERVKs promote TE transcription, activate nearby genes, and reinforce R-loop formation, driving widespread transcriptional derepression and impairing oocyte maturation.

TE-associated R-loops promote chromatin opening and transcriptional dysregulation in RNase H1-depletion oocytes

To investigate whether R-loop accumulation promotes gene expression by increasing chromatin accessibility, we performed an assay for transposase-accessible chromatin with sequencing (ATAC-seq) for Rnaseh1^fl/fl and Rnaseh1^oo−/− FGOs. The results revealed a significant increase in chromatin accessibility at TSSs in Rnaseh1^oo−/− FGOs (Fig. 6a), along with a marked directional concordance (81.6%) between increased ATAC-seq signal and mRNA levels (Fig. 6b), suggesting a positive relationship between chromatin accessibility and gene activation. Together with the observed transcriptional upregulation in Rnaseh1^oo−/− FGOs (Fig. 5), these findings propose that uncontrolled R-loops accumulation in the absence of RNase H1 induces excessive chromatin opening, ultimately leading to aberrant gene expression and failure to establish the transcriptional silencing required for oocyte maturation.

Fig. 6. Increased chromatin accessibility at genes enriched with R-loops.

a Metaplot (top) and heatmap (bottom) of ATAC-seq signals in Rnaseh1^fl/fl and Rnaseh1^oo−/− FGO. b Concordance between mRNA and ATAC-seq fold changes. c Overlap of chromatin-opening regions detected in Rnaseh1^fl/fl and Rnaseh1^oo−/− FGO. d Metaplot of ATAC signals of 858 overlapping genes (left) and 6165 genes with increased chromatin accessibility (right) identified in (c). e Clustering of R-loop signal at the 6165 gene promoters (left) and corresponding distal TEs (right) between Rnaseh1^fl/fl and Rnaseh1^oo−/− FGOs in (c). f Heatmap of R-loop signal (left) and corresponding ATAC-seq signal (right) at R-loop-enriched ERVK loci. g ATAC-seq signals at the promoter of R-loop-enriched ERVK elements. Center line, median; box limits, upper and lower quartiles; whiskers extend to the most extreme data points within 1.5 × the interquartile range. Unless otherwise indicated, the exact p value is calculated by two-sided unpaired Student’s t tests. n = 35 ERVK loci. h ATAC-seq signals at three annotated R-loop-enriched genes in Rnaseh1^fl/fl and Rnaseh1^oo−/− FGO. Center line, median; box limits, upper and lower quartiles; whiskers extend to the most extreme data points within 1.5 × the interquartile range. n, number of loci in each group. i Metaplot of ATAC-seq signals at promoters of genes associated with R-loop-enriched TE in Rnaseh1^fl/fl and Rnaseh1^oo−/− FGOs. j Representative snapshots from the UCSC Genome Browser comparing R-loop and ATAC-seq signals between Rnaseh1^fl/fl and Rnaseh1^oo−/− FGOs. Promoters are highlighted in light blue and TEs in light pink. n = 2 biologically independent sequencing libraries per condition. Source data are provided as a Source Data file.

A total of 7023 open chromatin regions were identified in Rnaseh1^oo−/− FGOs, of which only 858 overlapped with controls (Fig. 6c). This minimal overlap is consistent with the naturally compact chromatin structure of FGOs⁴⁵, and highlights a distinct chromatin landscape in RNase H1-deficient oocytes. Furthermore, while ATAC-seq peak intensities in overlapped regions were comparable between both groups, the non-overlapping peaks displayed significantly higher signal intensities (Fig. 6d), suggesting that RNase H1 depletion leads to excessive chromatin decompaction, likely driven by R-loop accumulation. To explore the relationship between R-loop and chromatin accessibility, we first clustered 6,165 abnormally open loci based on R-loop signals at promoters (Fig. 6e, left). Cluster I and II exhibited consistently high or low R-loop levels in both Rnaseh1^fl/fl and Rnaseh1^oo−/− FGOs, with only Cluster IV exhibiting an increased R-loop accumulation in Rnaseh1^oo−/− FGOs (Fig. 6e, left), suggesting that chromatin opening is not necessarily linked to R-loop accumulation at promoter regions directly. However, both Cluster I and II were associated with enhanced R-loop accumulation at TEs in Rnaseh1^oo−/− FGOs (Fig. 6e, right), suggesting that TE-enriched R-loops may alter chromatin structure beyond their immediate loci. To confirm this, we randomly selected R-loop-enriched ERVK elements and assessed their chromatin accessibility. As expected, these ERVK loci exhibited a significantly open chromatin state (Fig. 6f–h). More strikingly, R-loop-enriched TE-proximal gene promoter regions also exhibited significantly increased openness despite lacking local R-loop enrichment (Fig. 6i, j), indicating that R-loops at TEs are able to alter chromatin accessibility at gene promoters. Overall, these findings suggest that TE-enriched R-loops not only drive local chromatin decompaction but may also propagate accessibility changes to adjacent gene promoters, indirectly enhancing gene expression in the absence of RNase H1.

RNase H1 overexpression reinstates proper transcriptional regulation

Gene ontology analyses revealed that the active genes primarily participated in chromosome segregation, DNA methylation, and transcription initiation, underscoring their crucial roles in maintaining genomic stability and regulating gene expression during oocyte development (Fig. 7a). To investigate the impact of RNase H1 depletion on chromatin conformation and transcriptional activity in FGO, we performed a series of IF confocal analyses. Heterochromatin Protein 1α (HP1α), a well-characterized marker of constitutive heterochromatin, known for its dual role in DNA binding and recognition of dimethylated or trimethylated histone H3 at lysine 9 (H3K9me2 or H3K9me3), contributes to chromatin compaction and transcriptional silencing⁴⁷. HP1α staining in Rnaseh1^fl/fl FGO appeared to be compacted and closely colocalized with chromatin, similar to the HP1α expression pattern in the surrounding nucleolus (SN) GV oocytes⁴⁸. In contrast, we observed a significant decrease in HP1α expression in Rnaseh1^oo−/− FGO (Fig. 7b, c). This reduction in HP1α levels suggests a global loss of condensed, tightly packed heterochromatin regions, indicative of a more relaxed and transcriptionally permissive chromatin state. Heterochromatin loss is accompanied by an increase in the transcriptome, particularly the upregulation of TEs⁴⁹, aligning with our findings of elevated TE expression and enhanced chromatin accessibility at distal TE regions. Similarly, Rnaseh1^oo−/− FGO exhibited a significant increase in H3K27 acetylation (H3K27ac) deposition (Fig. 7d, e), indicative of elevated enhancer activity^50,51. In the meanwhile, H3K4me3 levels remained unaffected by RNase H1 deficiency (Supplementary Fig. 4p, q), suggesting that histone methylation associated with transcription activation is largely preserved^50–52. Importantly, this widespread gain of H3K27ac correlates with aberrant activation of TEs, especially LTR elements⁵³, suggesting that these R-loop-enriched TEs may function as enhancer-like elements whose activation further amplifies transcriptional activity in FGOs. Together, these results support the notion that RNase H1-mediated suppression of R-loops is critical for preserving chromatin integrity and preventing distal TEs activity at this developmental stage.

Fig. 7. Increased transcription and DNA damage in RNase H1-depleted FGOs.

a Gene ontology analyses of upregulated genes in RNase H1-depleted FGOs. IF confocal analysis (b) and intensity quantification (c) of heterochromatin protein 1-alpha (HP1α) levels in FGO. Scale bars, 20 μm. IF confocal analysis (d) and intensity quantification (e) of H3K27 acetylation (H3K27ac) levels in FGO. Scale bars, 20 μm. IF confocal analysis (f) and intensity quantification (g) of 5-ethynyluridine (EU) levels in FGO. Scale bars, 20 μm. IF confocal analysis (h) and intensity quantification (i) of phosphorylated histone H2AX (pH2AX) levels in FGO. Scale bars, 20 μm. IF confocal analysis (j), quantification of pS5 foci number (k), and pS5 foci diameter measurement (l) of phospho-RNA pol II CTD-Ser5 (pS5) levels in FGO. Scale bars, 50 μm. Data are presented as mean ± s.e.m. Exact p values are indicated in the panel and were calculated by two-sided unpaired Student’s t tests. Source data are provided as a Source Data file.

Moreover, the significant increase in the 5-ethynyluridine (EU) incorporation observed in Rnaseh1^oo−/− FGO marked a substantial elevation in newly transcribed RNA (Fig. 7f, g). This finding aligns with our ATAC-seq results (Fig. 5), which revealed enhanced chromatin accessibility and mRNA accumulation following RNase H1 deletion. These observations collectively indicate a failure of genome silencing, coinciding with a decondensed chromatin state that facilitates aberrant transcriptional activity. Meanwhile, RNase H1 knockout led to a marked increase in DNA damage during oogenesis (Fig. 7h, i), as evidenced by the significant elevation of phosphorylated histone H2AX (pH2AX), a well-established marker for DNA double-strand breaks. This suggests that the loss of RNase H1 disrupts genome integrity, potentially due to the accumulation of unresolved R-loops and subsequent genomic instability.

To assess transcriptional initiation, we examined phosphorylated Pol II at serine 5 (pS5) foci (Fig. 7j). Ser5 phosphorylation is associated with transcription initiation, serves as a marker of active TSSs^54,55. Rnaseh1^oo−/− FGO, displayed increased numbers but smaller pS5 foci, suggesting dispersed transcription initiation events with reduced activity per locus (Fig. 7j–l). Remarkably, reintroducing functional RNase H1 normalized pS5 foci (Fig. 7j–l), while a catalytically inactive RNase H1 mutant (RNase H1^D183N) failed to rescue the phenotype (Fig. 7j–l), emphasizing the necessity of programmed R-loop resolution.

Discussion

As oogenesis is entirely devoid of DNA replication during the NGO-to-FGO transition, our system ensures that the observed R-loops mainly reflect bona fide transcriptional events and are replication-independent, in contrast to dividing cells, providing a useful framework for elucidating the transcriptional regulatory functions of R-loops. Using a high-specificity R-loop detection approach, RTACC-seq, we observed a net gain of R-loops preferentially at distal TEs compared to gene promoters during NGO-to-GO transition, coinciding with global transcriptional activation, followed by RNase H1-mediated resolution of R-loops in GO-to-FGO transition, aligning with transcriptional silencing and chromatin condensation (Fig. 8). Notably, TE-enriched R-loops, particularly at LTRs, may further enhance transcriptional activation at proximal gene promoters, suggesting a synergistic role between distal TE transcription and associated R-loop formation in global gene activation in GOs (Fig. 8). In contrast, promoter-associated R-loops may hinder transcriptional efficiency. Overall, the spatiotemporal regulation of R-loops by RNase H1 appears to be critical for oocyte nucleus to shift from a diffuse NSN configuration in GOs to a condensed SN structure in FGOs, a process crucial for genome silencing and the acquisition of full developmental competence^12,13.

Fig. 8.

Schematic illustration of chromatin remodeling and gene expression in the growing oocyte and the fully grown oocyte under the influence of R-loops.

A key advantage of our detection strategy is the substantially enhanced affinity and specificity of our probe, achieved through the in situ expression of the engineered NLS-3×HBD-3×HA mRNA, allowing a more robust and reliable R-loop capture. Stopar and Nicholson⁵⁶ have demonstrated that a triple-HBD structure exhibits greater binding efficiency than single- and double-HBD constructs, supporting the rationale behind our probe design strategy. This approach effectively integrates sequencing adapters directly into the captured R-loop sites, ensuring efficient library preparation and amplification from ultra-low input oocyte samples. Similarly, R-loop CUT&Tag employing two HBD repeats of RNase H1, also detected R loop enrichment at promoter regions, consistent with R-ChIP³⁵ and MapR⁵⁷, but with universally enhanced signals, particularly at gene bodies¹⁰. One consideration for RNase H1-based approaches, such as R-ChIP and MapR, is that they rely on ectopic delivery or expression of a catalytically inactive hybrid-binding protein, which may not be entirely neutral with respect to native hybrid homeostasis. Prior work has noted that exogenous inactive RNase H1 may compete with endogenous RNase H1 or other hybrid-binding factors for access to these structures, thereby perturbing the native hybrid turnover and causing potential disruptions to normal transcriptional processes³⁵. Moreover, expression of inactive RNase H1 in cells can stabilize a subset of pre-existing hybrids, which may contribute to a higher accumulation of detectable R-loops^58,59. Thus, some differences between RNase H1-based mapping methods and affinity-based approaches may reflect not only the R-loop capture strategy, but also partial perturbation or stabilization of pre-existing hybrid populations. Interestingly, although our strategy shares methodological similarities with R-loop CUT&Tag¹⁰, we revealed stronger R-loop enrichment at TEs compared to gene promoters. This discrepancy may be attributed to the unique transcriptional landscape of oogenesis, where TE-initiated chimeric transcripts are highly abundant in murine oocytes⁶⁰ but are largely silenced by epigenetic modifications in somatic cells^61,62.

Given that TE transcription is developmentally regulated, it is plausible that R-loops at TEs in oocytes reflect an alternative mode of transcriptional regulation distinct from somatic cells, where promoter-associated R-loops predominate. DRIP-seq^6,32 and spKAS-seq³⁶, on the other hand, detect R-loops across the entire genic regions, with the majority at gene bodies, followed by TSSs. The preference for gene-body R-loops in these methods can be explained by the stability and persistence of elongation-associated R-loops³², whereas enhancer- and promoter-proximal R-loops, though forming at a higher frequency, are typically shorter and transient due to the DNA sequence composition and local chromatin topology^35,63. Moreover, oocyte KAS-seq¹² measures exposed ssDNA, which can reflect both the displaced ssDNA in an R-loop and ssDNA generated by transcription bubbles, whereas RTACC-seq directly enriches RNA:DNA hybrids. These assay differences motivated our emphasis on hybrid dependence and cross-assay benchmarking in the same cell type rather than over-interpreting single-locus comparisons. Our precise and sensitive R-loop detection strategy allows for a comprehensive analysis of the spatiotemporal distribution of R-loops in oocytes, particularly at TEs, which have been largely underrepresented in previous studies. By capturing R-loop dynamics with minimal perturbation, our approach offers a valuable framework for studying the functional significance of R-loops in genome regulation.

Distal TE-biased R-loop distribution revealed in this study suggests an alternative layer of gene regulation, possibly linked to LTR-mediated transcriptional activation during early germ cell growth in mouse^24–27. Although evidence for TE-associated R-loops has been reported across species, their regulatory significance remains largely unresolved. Previous reports have demonstrated R-loop enrichment at LTRs in yeast⁶⁴, at L1 LINE in human cell line², and at both LINEs and LTRs in A. thaliana seedlings and Drosophila embryos⁶⁵. These findings highlight a widespread yet diverse association between R-loops and TEs, suggesting a potential but poorly characterized role in transcriptional regulation across different species and contexts. In our oocyte system, distal TE-enriched R-loops coincide with increased chromatin accessibility and higher RNA accumulation rate at nearby loci, consistent with a model in which a subset of TE-associated hybrids accompanies transcriptionally permissive regulatory states rather than serving as a universal driver. Distal TE-derived RNAs generated at these loci may contribute to enhancer-like behaviors via R-loop formation, facilitating transcriptional activation at specific genomic loci⁶⁶. This mechanism could explain why TE transcription often coincides with major developmental transitions, such as oogenesis and zygotic genome activation (ZGA), where widespread chromatin remodeling and gene expression are required^24,60.

In contrast, changes in promoter-associated R-loops showed a modest inverse relationship with transcriptional output rate, indicating that promoter-associated R-loops are not a simple quantitative proxy for transcriptional activation in GO. Several features of GO likely attenuate locus-level correlations, including the progressively expanding maternal RNA pool, which compresses dynamic range in steady-state RNA measurements, as well as the possibility that promoter R-loops reflect heterogeneous mechanistic states, such as co-transcriptional hybrids versus paused or stalled polymerase complexes, that need not scale with net RNA accumulation. Future work will benefit from ultra-low-input nascent transcription profiling in oocytes.

Uncontrolled R-loop accumulation is widely recognized as a contributor to genomic instability, transcriptional dysregulation, cellular dysfunction, and diseases^1,9. For instance, mutations in SETX, a helicase that resolves R-loops, cause amyotrophic lateral sclerosis type 4 (ALS4), where increased helicase activity reduces R-loop accumulation, leading to increased promoter methylation and abnormal TGF-β signaling, ultimately resulting in neuronal dysfunction⁶⁷. Moreover, mutations in BRCA1/2 increase R-loop accumulation, inducing DNA damage and elevating cancer risk^8,68–70. Oocyte-specific knockouts of Setx⁷¹ and Brca2⁷² in mice have been linked to POF, yet the underlying mechanisms remain largely undefined. While these phenotypes have been primarily attributed to DNA damage and meiotic defects, an alternative explanation may involve pathological accumulation of unresolved R-loops. Our findings suggest that R-loop-induced transcriptional dysregulation and DNA damage may be a key factor contributing to POF, highlighting the importance of regulatory R-loops in maintaining oocyte development and female fertility.

R-loops preferentially localize at unmethylated CpG island promoters, suppressing DNA methylation-mediated silencing by blocking DNMTs access^4,67, and promoting transcription elongation by facilitating H3K36me3 deposition, a histone mark associated with active chromatin⁶. Additionally, in mouse embryonic stem cells, promoter-proximal R-loops recruit histone acetyltransferase complex Tip60-p400, enhancing chromatin accessibility while simultaneously blocking polycomb repressive complex 2 (PRC2)-mediated H3K27me3, thereby preventing transcriptional repression⁷³. Conversely, R-loop at antisense-transcribed RASSF1 locus recruits PRC2 to specifically silence RASSF1A promoter via H3K27me3⁷⁴. Directly coupling TE-associated R-loops to locus-resolved transcription in oocytes remains technically challenging. TE-derived reads are frequently multi-mapped across highly similar copies, making insertion-level quantification underpowered at oocyte-scale depth. Software such as TE transcripts improves TE expression estimation by modeling ambiguous reads but typically reports at repeat family/subfamily resolution rather than returning a coordinate-resolved set of insertion intervals for direct intersection with RTACC-seq peaks⁷⁵. Therefore, we present oocyte TE transcription evidence at an aggregate level and provide locus-resolved validation in HEK293T using uniquely mapped RTACC-seq and nascent transcription profiling by PRO-seq. Established PRO-seq provides nucleotide-resolution readouts of engaged transcription but traditionally requires substantial input material. Recent advances toward single-cell nascent transcription profiling⁷⁶ suggest feasible paths toward ultra-low input implementations that could be optimized for oocytes in the future.

Collectively, R-loops may influence chromatin structure and promote gene expression by modulating epigenetic landscapes through multiple mechanisms: (1) blocking the recruitment of transcriptional repressors, (2) preventing de novo DNA methylation, and/or (3) stabilizing chromatin structures that favor continued transcription. Future research should focus on dissecting the interplay between R-loop accumulation and epigenetic modifications, particularly how these mechanisms contribute to the transition from transcriptional activation during NGO-to-GO development to genome silencing in FGO and subsequent reactivation during ZGA. Ultimately, our study lays a foundation for understanding R-loop-mediated transcriptional control during oogenesis, supporting a stage-dependent dynamic interplay between R-loops and chromatin remodeling in genome regulation. Given the fundamental role of TEs in shaping species-specific gene regulatory networks, further exploration of R-loop-driven TE functions may offer more insights into gene regulation, chromatin remodeling, and developmental reprogramming.

Methods

Ethics

All procedures involving animals were approved by the Institutional Animal Care and Use Committee of Guangzhou Medical University (protocol GY2024-413) and Guangdong Second Provincial General Hospital (protocol 2021-DW-KZ-001-02) and were performed in accordance with institutional and relevant national guidelines. Experiments were conducted in female C57BL/6J mice. The ages of animals used in each experiment are provided in the corresponding Methods subsections and figure legends. Sex was incorporated into the study design by restricting experiments to females, as male animals were not relevant to the experimental objectives; therefore, no sex-based comparative analysis was performed.

Cell culture

We used the human embryonic kidney cell line HEK293T (ATCC, CRL-3216). We cultured cells in DMEM complete medium (Dulbecco’s Modified Eagle’s Medium with D-Glucose, L-Glutamine, and Pyruvate) supplemented with 10% fetal calf serum, 100 U/mL Penicillin, and 100 mg/mL Streptomycin. The HEK293T cell line tested negative for mycoplasma contamination.

Animals

All animal experiments were performed in female C57BL/6 J mice. Mice were maintained in a controlled environment of 20–22 °C, with a 12 h light/12 h dark cycle, 50-70% humidity, and food and water provided ad libitum. The conditional Rnaseh1^fl/fl mouse strain was generated as described previously. Briefly, loxP sites were inserted to flank the targeted Rnaseh1 exon 2 and 3, and the resulting Rnaseh1^fl/fl strain was crossed with Gdf9-Cre transgenic mice to achieve Cre-mediated recombination specifically in oocytes from the primordial follicle stage onward.

Oocyte collection and culture

Oocytes were isolated in 37 °C pre-warmed M2 medium (M7167; Sigma-Aldrich) at three developmental stages. Oocytes were selected by diameter under a stereomicroscope using a calibrated scale to minimize heterogeneity and improve reproducibility: P7(NGO), 20-30 µm; P14(GO), 40–50 µm; P28(FGO), ~70 µm. For FGO experiments, the 4-week-old female mice were injected with 5 IU pregnant mare serum gonadotropin (PMSG; Ningbo Sansheng Pharmaceutical) and humanely euthanized 44 h later. Only surrounded nucleolus (SN) type oocytes were used, identified by clear nucleolar morphology visualized by Hoechst 33342 staining immediately prior to sample processing. Oocytes were cultured in M16 medium (M7292; Sigma-Aldrich) covered with mineral oil (M5310; Sigma-Aldrich) at 37 °C in 5% CO₂. Two2 μM milrinone (MedChemExpress, HY-14252A) was added to the culture medium to inhibit spontaneous GVBD. GV and MII oocytes were treated with the same concentration of DMSO.

In vitro transcription and microinjections

Plasmids encoding the indicated constructs were linearized, purified by phenol chloroform extraction, and ethanol precipitated. Capped RNA was synthesized in vitro using the SP6 mMESSAGE mMACHINE® Kit (Invitrogen, AM1340) followed by polyadenylation using Poly(A) Tailing Kit (Invitrogen, AM1350), according to the manufacturer’s instructions. Transcribed mRNAs were recovered by lithium chloride precipitation and resuspended in nuclease-free water at a concentration of 500 ng/μL. All the microinjections were performed using an Eppendorf Transfer Man NK2 micromanipulator. For microinjection, denuded oocytes were incubated in M2 medium with 2 μM milrinone (Med Chem Express, HY-14252A). Approximately 5–10 pL of mRNA per oocyte was injected into the cytoplasm. Injected oocytes were washed twice in pre-warmed M16 medium and then cultured for 6 h at 37 °C in 5% CO₂ to allow protein expression and nuclear accumulation before RTACC-seq library preparation.

RTACC-seq library construction and sequencing for mouse oocytes and HEK293T cells

RTACC-seq was adapted from published work³⁷ with modifications. For mouse oocytes RTACC-seq, oocytes at different stages were microinjected with NLS-3×HBD-3×HA or NLS-mCherry-3×HA mRNA and then incubated in M2 medium with 2 μM milrinone for 6 h. Oocytes were transferred from culture medium to wash buffer (DNase/RNase-free 0.01% BSA), washed three times, and collected in DNA-low-binding tubes (100 oocytes per sample). After removing the zona pellucida, samples were lysed in DB-1 buffer (see below) at 4 °C on a Thermo Mixer (TS100) at 500 rpm for 10 min, with vortexing every 2.5 min for three times. For HEK293T cells RTACC-seq, cells were transfected with the NLS-3×HBD-3×HA-GFP plasmid using Linear Polyethylenimine (TRAN, FT401-01). After 24 h, different numbers of cells were collected for subsequent processing based on GFP expression. Cells were washed three times with wash buffer (DNase/RNase-free 0.01% BSA) and collected in DNA-low-binding tubes.

Hyperactive pAG-Tn5 Transposase (Vazyme Biotech, S614-01) complexes were prepared 20 min prior to the sample lysis step. DB-1 buffer was prepared by adding 1 μL fully dissolved 5% digitonin (Sigma, D141) to 1 mL buffer 1 (10 mM Tris-HCl pH = 7.4, 150 mM NaCl, 0.5 mM spermidine, 1 × EDTA-free Roche complete protease inhibitor). For complex assembly, 7 μL of DB-1 buffer, 0.5 μL pAG-Tn5, and 0.5 μg HA-antibody (Cell Signaling, #3724) were added to a 200 μL low-binding tube and mixed by vortex and incubated at 4 °C for 30 min.

Following lysis, 35 μL DB-1 buffer, the pA/G-Tn5 mixture, and 12.5 μL pre-warmed (37 °C) 5 × TTBL (Vazyme Biotech, TD502) were added into the resulting samples and then incubated in an Eppendorf Thermo mixer at 37 °C for 30 min at 400 rpm. Next, 2 μL of 10% SDS and 40 ng of carrier RNA were added, and samples were incubated in an Eppendorf Thermo Mixer at 55 °C for 30 min at 400 rpm. DNA was purified by 2 × AMPure beads (Beckman, A63882). Libraries were amplified using Vazyme Biotech (TD601) with the following PCR program: 72 °C for 3 min; 98 °C for 30 s; thermocycling for 16 cycles at 98 °C for 15 s, 60 °C for 30 s and 72 °C for 3 min; followed by 72 °C for 5 min. PCR products were purified using 0.4× - 1.7× AMPure beads (Beckman, A63882) for size selection and subjected to next-generation sequencing.

Immunofluorescent confocal analysis for oocytes

Oocytes were fixed in 4% paraformaldehyde (Sigma, 158127) in PBS for 30 min at room temperature, then washed in PBST (0.1% Tween and 0.01% Triton X-100 in PBS). Permeabilization was performed with 0.2% Triton X-100 (Sigma, X100) in PBS for 10 min, followed by blocking in 1% bovine serum albumin (Sigma, V900933) for 30 min. Oocytes were incubated with the primary antibody for 1 h at room temperature, washed three times in PBST, and incubated with the secondary antibody for 30 min. Finally, oocytes were mounted in an anti-fade reagent with DAPI, and images were captured using a confocal laser-scanning microscope (LEICA DMi8 and ZEISS 980), with support from the Scientific Research Center, Guangzhou Medical University. Antibodies used are listed in Supplementary Data 1.

Western blot analysis

Oocytes were lysed in β-mercaptoethanol (Gibco, 21985023) and SDS (Thermo, 15553-035) containing loading buffer, then heated at 95 °C for 10 min. Total proteins were separated by SDS-PAGE and transferred onto polyvinylidene fluoride (PVDF) membranes (Millipore, IPVH00010). Membranes were blocked in Tris-buffered saline with 0.1% Tween 20 (TBST) containing 5% defatted milk for 30 min, incubated with primary antibodies overnight at 4 °C, washed in TBST, then incubated with a horseradish peroxidase-linked secondary antibody for 1 h, followed by three TBST washes. The primary antibodies used are listed in Supplementary Data 1.

Periodic acid-Schiff (PAS) staining

Ovaries were fixed in 10% formalin (Sigma, HT5011) for 24 h, paraffin-embedded, and sectioned at 5 μm onto silane-coated slides. Sections were deparaffinized with xylene and rehydrated through a graded ethanol series. They were incubated in 0.5% periodic acid solution (Solarbio, G1280) for 5–10 min, rinsed in distilled water, then treated with Schiff’s reagent (Solarbio, G1280) for 10–15 min in the dark. After rinsing under running tap water for 5 min, sections were counterstained with Mayer’s hematoxylin for 1 min, differentiated, blued, rinsed in water, dehydrated, cleared in xylene, and mounted. PAS-stained sections were imaged using the MSHOT microscope.

Immunohistochemistry (IHC)

Ovaries were fixed in 10% formalin (Sigma, HT5011) for 24 h, paraffin-embedded, and sectioned at 5 μm onto silane-coated slides. Antigen retrieval was performed using heat-induced epitope retrieval (HIER) in citrate buffer (pH 6.0) via microwave heating for 15 min, followed by cooling at room temperature for 20 min and PBS rinsing. To minimize nonspecific binding, sections were blocked with 5% normal goat serum and 0.3% Triton X-100 in PBS for 1 h at room temperature. Primary antibodies were applied overnight at 4 °C, followed by three PBS washes. A biotin-conjugated secondary antibody was incubated for 30 min at room temperature in the dark. Signal amplification was performed using the Vectastain ABC kit (Vector Laboratories, PK-6100), and immunoreactivity was visualized with the 3,3-diaminobenzidine (DAB) peroxidase substrate kit (Vector Laboratories, SK-4100) until a brown precipitate formed.

Immunofluorescence microscopy on sections

The pretreatment steps were similar to those used in IHC. For IF, sections were probed with primary antibodies as indicated in the Results, followed by incubation with Alexa Fluor 594- or 488-conjugated secondary antibodies (Thermo, 35502, 35552, 25561). Slides were mounted with VectaShield containing DAPI (Thermo, D1306). Digital images were captured using an epifluorescence microscope (LEICA DMi8) with 10–100× objectives. Semi-quantitative fluorescence analysis was performed using the NIH ImageJ software. Antibodies used are listed in Supplementary Data 1.

TUNEL assay

TUNEL assays were performed on 10% formalin-fixed paraffin-embedded sections using the ApopTag Plus Peroxidase In Situ Apoptosis Detection Kit (Sigma, S7101) following the manufacturer’s instructions. Briefly, sections were treated with 20 μg/mL proteinase K for 15–20 min, equilibrated in equilibration buffer for 5 min, and incubated with TdT reaction mixture at 37 °C for 60 min. After the reaction, sections were treated with the stop solution for 10 min, incubated with peroxidase-conjugated anti-digoxigenin antibody for 30 min, and stained with DAB substrate solution. Finally, sections were counterstained with hematoxylin, dehydrated, and mounted for the following analysis.

MitoTracker and ΔΨm measurement

To determine mitochondrial membrane potential in oocytes, the procedure was conducted according to the instructions with MitoTracker (Beyotime, C1035) and the JC-1 ΔΨm assay kit (Yeasen, 40706ES60). Briefly, the collected GV oocytes were washed with M2 medium several times and then transferred into M2 medium containing staining solution and incubated at 37 °C in the dark for 30 min. After that, oocytes were washed with M2 medium to remove the excess staining solution. All the live oocytes were observed, and relative fluorescence intensity was calculated by confocal microscopy.

EU incorporation assay

To detect transcriptional activity, oocytes were cultured in M16 medium containing 1 mM 5-ethynyl uridine (EU) for 1 h. EU staining was performed using a Click-iT^® RNA Alexa Fluor^® 488 Imaging Kit (Life Technologies, C10329) according to the manufacturer’s instructions. The mean signals were measured across the middle of each oocyte and quantified using ImageJ software.

RNA-seq library generation and sequencing

Oocytes were collected from the indicated genotypes (10 oocytes per sample). Each sample was lysed directly using 4 μL lysis buffer (0.2% Triton X-100, RNase inhibitor, dNTPs, oligo-dT primers, and 100 pg ERCC mRNA spike-in) and immediately used for cDNA synthesis using the Smart-seq2 method as previously described. Sequencing libraries were constructed from 500 pg of amplified cDNA using the TruePrep DNA Library Prep Kit V2 for Illumina (Vazyme Biotech, TD503) according to the manufacturer’s instructions. Barcoded libraries were pooled and sequenced on the Illumina HiSeq X Ten platform using 150 bp paired-end reads.

Transposase accessible chromatin (ATAC) with high-throughput sequencing

Each sample contains 50 oocytes collected from mated female mice at 44 h after PMSG injection. Oocytes were washed with DNase/RNase-free 0.01% BSA and collected in DNA-lowering binding tubes. Lysis buffer (0.05% digitonin, 0.5% NP-40, and 0.5% Tween20) was added immediately after sample collection. Tn5 enzyme in Chromatin Profile Kit for Illumina (Novoprotein, N248) was used to fragment chromatin at 37 °C on Thermo Mixer (TS100) with 1000 rpm for 30 min and stopped by stop buffer immediately. After extraction with DNA extract beads, DNA fragments were amplified for 15 cycles using the following PCR conditions: 72 °C for 3 min; 98 °C for 30 s; and thermocycling at 98 °C for 15 s, 60 °C for 15 s and 72 °C for 8 s; following by 72 °C 2 min. After the PCR reaction, libraries were purified with the 1.2x DNA clean beads following by fragment selection with 0.5x/0.7x AMPure beads (Beckman, A63882). Barcoded libraries were pooled and sequenced on the Illumina HiSeq X Ten platform using 150 bp paired-end reads.

RNA-seq data processing

RNA-seq was performed using biological replicates for all samples. Raw reads were trimmed to remove adapters and low-quality fragments using Trim-galore (version 0.6.7)⁷⁷ and mapped to the mouse genome (mm10) using HISAT2 (version 2.2.1)⁷⁸ with default parameters. Uniquely mapped reads were subsequently assembled into transcripts guided by reference annotation—University of California at Santa Cruz (UCSC) gene models, using Cufflinks (version 2.2.1)⁷⁹. At the same time, for analysis of transposable element (TE) transcript’s count, the software TEtranscripts (version 2.2.3)⁷⁵ was used with default parameters. The level of expression of each gene was quantified as the fragments per kilobase of transcript per million mapped reads (FPKM) and was further normalized to the ERCC spike-in. Samples prepared in different batches were normalized to GV-stage oocyte samples from each batch. Only expressed genes (FPKM > 1 in at least one sample) were considered in all analyses, unless otherwise specified. Functional annotation was performed using clusterProfiler (version 4.0)⁸⁰. Statistical analyses were performed using the R software (http://www.rproject.org). Correlation analyses were performed using Pearson methods, and correlation coefficient (R) and exact P values were calculated using the cor function in R. The complete linkage hierarchical algorithm was used to cluster the genes. Quality controls of the RNA-seq results are provided in Supplementary Table 1 and Supplementary Data 2, 3.

RTACC-seq and ATAC-seq data processing

Raw reads were removed, adapters and low-quality fragments using Trim-galore (version 0.6.7)⁷⁷. The paired-end RTACC-seq reads were aligned to the mouse genome (mm10) with the parameters: -local --very-sensitive-local --phred33 --no-unal by Bowtie2 (version 2.3.5.1)⁸¹. All unmapped reads, non-uniquely mapped reads, and PCR duplicates were removed by using the Samtools (version 1.9)⁸² and Sambamba (version 0.7.1)⁸³. For downstream analysis, we normalized the read counts by computing the number of reads per kilobase of bin per million reads sequenced (RPKM) and generated bigWig files by deepTools (version 3.5.1)⁸⁴ with the parameters: bamCoverage -bs 10 --effectiveGenomeSize 2913022398 –normalize. Using RPKM using merged replicate BAM files. Each stage peak was identified by macs2 (version 2.2.7.1)⁸⁵ software with the parameters: --nomodel -t --nolambda --fe-cutoff 5 -q 0.001 -g mm --keep-dup all. And using the parameters: --nomodel -t -c --fe-cutoff 3 -p 0.001 -g mm --keep-dup all for identifying the different peaks in a fully grown oocyte. The peak annotation by the R package ChIPseeker (version 1.40.0)⁸⁶. To minimize batch and cell type variation, the RPKM values were further normalized using Z-score transformation to identify stage-specific signals with the R package Mfuzz (version 2.64.0)⁸⁷. TE targets were searched in the mm10 genome with the UCSC Genome Browser. The distances between all predicted target sites and Ensembl-annotated genes were then assigned with Bedtools closest (version 2.30.0)⁸⁸. Quality controls of the RTACC-seq and ATAC-seq results are provided in Supplementary Tables 2, 3 and Supplementary Data 4–10.

Observed/expected enrichment analyses

The analysis was performed based on a previously published study⁴⁰ with minor modifications. Tracks for the repeats were obtained from the UCSC Genome Table Browser for the mm10. RTACC-seq peaks across different oocyte growth stages were used to calculate the observed-to-expected frequency ratios for distinct genomic elements, including different transposable element (TE) classes (SINE, LINE, and LTR), gene bodies, promoters, and distal regions. For each RTACC-seq dataset, the intersection counts with the genomic elements were determined using bedtools intersect (version v2.28.0). The expected occurrence frequencies in the genome were calculated by intersecting the genomic elements with randomized genomic coordinates (matched for number, length, and chromosome ID) corresponding to each RTACC-seq peak set. The observed-to-expected ratios for these genomic elements were computed for each oocyte stage-specific RTACC-seq dataset and visualized as a heatmap using the heatmap function of the ggplot2 package in R.

Statistical analysis

Unless otherwise indicated, results are given as mean ± s.e.m. Exact p values are indicated in the panels and were calculated by two-tailed unpaired Student’s t tests. Additional statistical details, including exact sample sizes (n) and replicates, are provided in the corresponding figure legends or Source Data file.

Reporting summary

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

Author contributions

Q.Q.S. conceived, designed, and supervised the study. X.Y.T., S.Y.L., and Q.Q.S. wrote the manuscript. X.Y.T., S.Y.L., Y.Z., H.F.F., J.W., Z.L.W., Q.W.L. and X.Q.L. performed the mouse experiments. Y.W.W. and S.Y.L. performed the ATAC experiments, X.Y.T. and S.Y.L. performed the RTACC-seq experiments. S.Y.L. and Y.S.C. analyzed the high-throughput sequencing data.

Peer review

Peer review information

Nature Communications thanks Miguel Branco, who co-reviewed with Kero Guynes, and the other, anonymous, reviewers for their contribution to the peer review of this work. [A peer review file is available].

Funding

Q.Q.S. discloses support for the research of this work from the National Key Research and Development Program of China [grant numbers 2021YFA1100300, 2021YFC2700100], National Natural Science Foundation of China [grant numbers 32270894, 82101731], Guangdong Basic and Applied Basic Research Foundation [grant number 2022B1515020038], and Guangzhou Science and Technology Project [grant number 202201011005].

Data availability

The RNA-seq data have been deposited in GEO with the accession number GSE288679. The RTACC-seq data have been deposited in GEO with the accession number GSE288680. The ATAC-seq data have been deposited in GEO with the accession number GSE288634. The MapR, RChIP, DRIPc-seq, CUT&Tag (2xHBD-His-Tag, 2xHBD-GST-Tag, and S9.6), and PRO-seq data were downloaded from NCBI GEO under accession number GSE156400. 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/s41467-026-73781-y.