G-quadruplex structures are key regulators of mammalian spermatogenesis

Abstract

Background

Male infertility, impacting 8–12% of couples globally, often lacks clear etiology. G-quadruplexes (G4s), noncanonical DNA structures, are implicated in genomic regulation but remain underexplored in spermatogenesis. This study investigates G4 dynamics and their roles in male fertility.

Methods

We employed antibody-based staining, cleavage under targets and tagmentation (CUT&Tag) sequencing, and a novel nanobody-based proximity labeling system (nanoG4BPL) to map G4 distribution and interacting proteins in mouse testicular cells. In vivo G4 stabilization with pyridostatin and clinical analysis of testicular tissues from patients with nonobstructive azoospermia (NOA) were conducted.

Results

G4 structures are enriched in testicular tissues, displaying stage-specific dynamics during spermatogonial differentiation, meiosis, and spermiogenesis. Genome-wide profiling revealed the dual roles of G4s in coordinating gene expression with active epigenetic marks and facilitating genome architecture via CTCF interactions. G4 stabilization disrupted double-strand break repair during meiosis, with nanoG4BPL identifying Nijmegen breakage syndrome 1 (NBS1) as a G4-interacting protein promoting phase separation for homologous recombination. Clinically, patients with NOA exhibited significantly elevated G4 levels in spermatocytes.

Conclusion

G4 structures are critical regulators of spermatogenesis, orchestrating gene expression, chromatin remodeling, and meiotic fidelity. Their dysregulation, particularly in patients with NOA, suggests a mechanistic link to male infertility, providing novel insights into its pathogenesis and highlighting potential avenues for future diagnostic or therapeutic exploration.

Supplementary Information

The online version contains supplementary material available at 10.1186/s11658-025-00839-y.

Background

Infertility affects approximately 8–12% of couples globally, with male factors contributing to nearly half of these cases [1]. Over recent decades, mounting evidence has documented a concerning rise in male infertility, exemplified by an annual increase in age-standardized prevalence of 0.291% for males between 1990 and 2017, alongside a parallel decline in sperm concentration exceeding 50% since the 1970 s [2]. Despite this growing burden, the etiology of male infertility remains elusive in a substantial proportion of cases. Prospective studies reveal that up to 75% of oligozoospermia and 60% of nonobstructive azoospermia (NOA) cases are classified as idiopathic, lacking identifiable causes despite advances in diagnostic approaches [3]. Therefore, elucidating the intrinsic and extrinsic factors regulating spermatogenesis is an urgent priority to unravel these widespread yet enigmatic conditions.

Spermatogenesis orchestrates nucleic acid dynamics across three phases: spermatogonial self-renew and differentiation, meiosis, and spermiogenesis. During differentiation, spermatogonia undergo tightly regulated chromatin remodeling [4]. Meiosis drives genome haploidization through SPO11-induced DNA double-strand breaks (DSBs) in bendable DNA regions, enabling homologous recombination (HR) and ensuring chromosomal segregation [5–8]. Finally, spermiogenesis compacts the genome by replacing histones with protamines, forming transcriptionally inert sperm nuclei. This continuum—from spermatogenic stem cells SSC renewal to protamine-mediated DNA condensation—centers on precise nucleic acid regulation, with disruptions in chromatin remodeling directly linking to infertility. Notably, noncanonical DNA structures widely exist throughout genomes and are closely related to the transcription, replication, and stability of the genome. For instance, R-loops regulate meiotic transcription and DNA double-strand break (DSB) repair during mammalian spermatogenesis, with their accumulation causing sterility in RNase H1 knockout mice[9, 10]. Similarly, Z-DNA is remodeled in prospermatogonia to prevent DSBs and promote DNA methylation [11], while cruciform DNA influences meiotic chromosome segregation [12], and DNA triplex structures anchor DNA loops to chromatid scaffolds in spermatids [13]. Therefore, in-depth exploration of the distribution and functions of noncanonical DNA structures in spermatogenesis will help identify the causes of abnormal spermatogenesis.

G-quadruplexes (G4s) are widely studied noncanonical nucleic acid structures stabilized by Hoogsteen bonding and cations that dynamically regulate genome stability, DNA replication, transcription, and chromatin remodeling across diverse biological contexts [14–17]. While their roles in cancer, neurodegeneration, and immune diseases are well documented [18–22], their spatial–temporal orchestration in spermatogenesis remains strikingly underexplored. This gap persists despite compelling evidence linking G4s to germline-specific processes. A deficiency in the G4 resolvase DHX36 helicase disrupts spermatogonial differentiation by failing to upregulate the expression of the c-kit gene. This is because the promoter region of the c-kit contains G4 structures, and DHX36 can bind to these G4 structures to promote c-kit expression [23]. Additionally, DHX36 helicase deficiency also leads to transcriptional dysregulation and causes failures in meiosis and spermiogenesis [24]. Although these studies suggest critical roles for G4s in spermatogenesis, a systematic framework to define how G4 landscapes are dynamically shaped during the continuum of spermatogenesis remains lacking. In particular, whether they functionally intersect with meiotic recombination remains elusive.

Deciphering the G4-interacting proteome represents a central challenge in elucidating their physiological and pathological mechanisms. Current methods for mapping G4–protein interactions—including in vitro pull-down assays [25], small-molecule crosslinkers [26, 27], and motif-based predictions [28]—have provided foundational insights into G4 biology. Recent advances leveraging endogenous G4-binding domains, such as DHX36, further enabled proximity labeling in live cells yet remain constrained by their dependence on specific helicase–G4 interactions [29]. Notably, the natural precision of antigen–antibody recognition has not been exploited to interrogate G4 interactomes in living cells.

In this study, we provide the first systematic characterization of DNA G4 dynamics across all stages of spermatogenesis. By integrating in vivo perturbation of G4 structural homeostasis, antibody-based staining, genome-wide profiling, and proximity labeling, we reveal stage-specific G4 landscapes and their functional interplay with epigenetic marks, meiotic recombination machinery, and chromatin remodelers. Notably, the nanobody-based proximity labeling system (nanoG4BPL) enables unbiased identification of G4-interacting proteins in live cells, overcoming limitations of existing methods. Furthermore, the discovery that NBS1 undergoes phase separation upon G4 binding uncovers a novel mechanism for concentrating DNA repair factors at G4-enriched DSB hotspots, shedding light on the spatiotemporal orchestration of meiotic recombination. Finally, the clinical findings in idiopathic infertility patients highlight G4 dysregulation as a potential mechanistic link between environmental stressors and male reproductive disorders, opening new avenues for diagnostic and therapeutic strategies.

Materials and methods

Clinical sample collection and ethical approval

Clinical samples were obtained from pathologically confirmed patients. The study received approval from the Ethics Committee of the University (approval no. KY20253837-1). Written informed consent was obtained from all participants prior to sample collection. Testicular biopsy tissues were obtained from nine patients, and full details are presented in Supplementary Table 1.

The inclusion criteria were as follows: for patients with NOA, males aged 18–50 years with azoospermia confirmed by at least two semen analyses and testicular biopsy indicating maturation arrest without evidence of obstruction; for controls: fertile males with normal spermatogenesis or obstructive azoospermia, who are expected to regain fertility after procedures such as vasectomy reversal.

The exclusion criteria were history of chemotherapy, radiotherapy, or testicular trauma within the past 6 months; presence of active systemic infection or malignancy; receipt of hormone therapy within 6 months prior to biopsy; presence of active infection or inflammatory disease; inability to provide informed consent.

Animal experiments and ethical approval

All animal experiments in this study were approved by the Experimental Animal Science and Technology Ethics Committee of the University. Male C57BL/6J mice were used for all experiments. Adult mice (8–10 weeks old) were used for most experiments, while postnatal day 9 (P9) and day 18 (P18) mice were used for meiotic studies to capture specific stages of spermatogenesis. Mice were housed under controlled conditions, including a 12-h light/dark cycle, temperature of 22 ± 2 °C, and relative humidity of 50–60%, with ad libitum access to standard chow and water. For G-quadruplex stabilization studies, adult mice were randomly assigned to treatment or control groups. The treatment group received daily intraperitoneal (i.p.) injections of pyridostatin (PDS; cat. no. GC37043, GlpBio) at a dosage of 30 mg/kg/day dissolved in sterile saline, for 14 consecutive days. The control group received an equivalent volume of sterile saline. For meiotic studies, P9 mice were administered PDS (30 mg/kg/day, i.p.) dissolved in sterile saline or vehicle (sterile saline) daily from P9 to P17 and euthanized at P18 to assess meiotic progression. Euthanasia was performed by cervical dislocation following isoflurane anesthesia.

Cell lines

The mouse spermatogonial cell line GC-1 spg (no. BFN60810509) and the mouse hepatoma cell line Hepa1-6 (no. BFN60700206) were obtained from the Shanghai Cell Bank (Shanghai, China). Both cell lines were authenticated by short tandem repeat (STR) profiling and tested negative for chlamydia and mycoplasma contamination. Cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM, Gibco) supplemented with 10% fetal bovine serum (FBS, Gibco), 100 U/mL penicillin, and 100 μg/mL streptomycin. All cells were maintained at 37 °C in a humidified incubator with 5% CO₂.

Plasmid construction

Plasmids were synthesized by Tsingke Biotech Co., Ltd. (Beijing, China). All plasmids were constructed using Gibson assembly with the pEASY^®-Basic Seamless Cloning and Assembly Kit (TransGen Biotech, cat. no. CU201). The SG4 sequence, sourced from Galli et al. [30], was used for subsequent experiments. To generate prokaryotic expression plasmids, the coding sequences of EGFP-SG4-TurboID and NBS1 were inserted between the NcoI and SpeI sites of the pGEX_3XFlag-pATn5-FL vector, yielding pGEX-EGFP-SG4-TurboID and pGEX-NBS1, respectively. For eukaryotic expression, the SG4 coding sequence was cloned between the HindIII and NotI sites of the pCMV3 vector to produce pCMV3-SG4. An EGFP tag was introduced into pCMV3-SG4 via polymerase chain reaction (PCR) amplification, generating pCMV3-EGFP-SG4. The TurboID sequence was inserted into pCMV3-EGFP-SG4 by PCR amplification to create pCMV3-EGFP-SG4-TurboID. Additionally, enhanced green fluorescent protein (EGFP) was introduced into pCMV3 via PCR amplification to construct pCMV3-EGFP-TurboID. All constructs were verified by Sanger sequencing (Tsingke Biotech Co., Ltd., Beijing, China).

Histology and immunostaining

Tissues (heart, intestine, brain, testis, liver, spleen, lung, and kidney) were fixed in 4% paraformaldehyde for 24 h at room temperature, embedded in paraffin, and sectioned at 5 μm using a Leica microtome (Leica Biosystems, Germany). For hematoxylin and eosin (HE) staining, sections were deparaffinized in xylene, rehydrated through a graded ethanol series, stained with hematoxylin for 1 min and eosin for 30 s, dehydrated, cleared in xylene, and mounted with coverslips. For immunostaining, sections were deparaffinized in xylene, rehydrated through a graded ethanol series, and subjected to antigen retrieval by boiling in citrate buffer (pH 6.0) at 100 °C for 90 s in a pressure cooker, then cooled to room temperature. Sections were blocked with 10% goat serum (Fuzhou Maixin Biotech, China) in phosphate-buffered saline (PBS) for 1 h at room temperature and incubated overnight at 4 °C with primary antibodies. After washing three times with Tris-buffered saline containing 0.05% Tween-20 (TBST), sections were incubated with secondary antibodies for 1 h at room temperature, washed, and mounted with 4′,6-diamidino-2-phenylindole (DAPI)-containing mounting medium or coverslips. For cellular immunofluorescence, cells were treated with 5 μM pyridostatin for 24 h, fixed in 4% paraformaldehyde for 15 min, permeabilized with 0.1% Triton X-100 in PBS for 8 min, and blocked with 10% goat serum in PBS for 1 h at room temperature. Cells were incubated with primary antibodies overnight at 4 °C, washed three times for 5 min with PBS, and incubated with secondary antibodies for 1 h at room temperature. Cells were mounted with DAPI-containing mounting medium. Images were acquired using an Olympus fluorescence microscope equipped with OLYVIA 4.1.1 software. Antibodies used are listed in Supplementary Table 2.

Cell treatment with HP2 and flow cytometry

Single-cell suspensions were prepared from mouse tissues (testis, liver, spleen, lung, and kidney) via mechanical dissociation followed by enzymatic digestion with 0.25% trypsin, 5 µg/mL DNase I, and 200 U/mL collagenase I at 37 °C for 30 min. HEP1-6 cells were dissociated into single-cell suspensions using 0.25% trypsin at 37 °C for 5 min. Cells were incubated with 10 µM HP2 at 37 °C for 2 h, followed by three washes with PBS. For apoptosis detection (Fig. 1), cells were stained using the Annexin-V fluorescein isothiocyanate (FITC) apoptosis detection kit (cat. no. BB-4101, BestBio) according to the manufacturer’s protocol. For G4 detection (Fig. 8), cells were fixed with 4% paraformaldehyde at room temperature for 15 min, permeabilized with 0.1% Triton X-100 for 10 min, and stained with a G4-specific probe for 30 min at 37 °C following the manufacturer’s instructions. Flow cytometry was performed for both analyses, and data were processed using FlowJo software (version 10.10.0).

Fig. 1.

G4 structures are enriched in testicular tissue and their homeostasis is critical for spermatogenesis. (A, B) Flow cytometric analysis of testicular and lung cells treated with the G4-specific probe HP2 (10 μM) or control for 2 h. Upper panels show forward/side scatter (FSC-A/SSC-A), and lower panels show Annexin V staining for apoptosis detection. (C) Quantitative analysis of Annexin V+ cells in testis and lung following HP2 treatment. Data represent mean ± standard deviation (SD) from three independent experiments; *P < 0.05; ns, not significant; ***P < 0.001. (D) Representative images of organs from mice intraperitoneally injected with PDS or saline for 14 days. (E) Organ weights in control and PDS-treated mice. Data represent mean ± SD from three independent experiments; **P < 0.001; ns, not significant. (F) Conventional flow cytometric analysis of testicular cells from control and PDS-treated mice, stained with Hoechst 33342 and DAPI to distinguish spermatogenic cell populations: SPG, spermatogonia; preL, preleptotene spermatocytes; L/Z, leptotene/zygotene spermatocytes; PD, pachytene/diplotene spermatocytes; MII, metaphase II spermatocytes; STD, spermatids. (G) Flow cytometric quantification of spermatogenic cell populations. Data represent mean ± SD from three independent experiments; ns, not significant; ***P < 0.001. (H) Quantification of different germ cell type proportions in the seminiferous tubule lumen. Twenty intact seminiferous tubules at each developmental stage were randomly selected and analyzed from three biological replicates per group. Data represent mean ± SD; ns, not significant; *P < 0.001.

Fig. 8.

Abnormal G4 accumulation in spermatogenic cells links oxidative stress to male infertility. A Immunofluorescence staining with 1H6 antibody and DAPI of testicular tissues from patients with obstructive azoospermia (OA), Y chromosome AZFc microdeletion (AZFc_Del), and idiopathic nonobstructive azoospermia (iNOA 1 and iNOA 2). Scale bar, 200 μm. B Quantification of G4 intensity from immunofluorescence images in A. Data represent mean ± SD from three independent experiments; **P < 0.01, ***P < 0.001. C Flow cytometric analysis of testicular single-cell suspensions stained with G4-specific probe HP2 and Hoechst 33342. Cells were classified into haploid, diploid, and tetraploid populations. D Quantitative analysis of G4-positive cells based on flow cytometric data in C. Data represent mean ± SD from three independent experiments; **P < 0.01, ***P < 0.001, ****P < 0.0001

Flow cytometry analysis of spermatogenic cells

Testes from adult male mice were decapsulated, and seminiferous tubules were isolated by digestion in PBS containing 200 U/mL collagenase I and 5 µg/mL DNase I at 37 °C for 15 min. After washing with PBS, tubules were further digested with 0.05% trypsin at 37 °C for 15 min to obtain single-cell suspensions. Digestion was terminated by adding DMEM supplemented with 10% FBS. Cells were filtered through a 40-µm cell strainer, stained with Hoechst 33342 (1:10,000) at 37 °C for 30 min, and resuspended in DMEM with 5% FBS. Cells were sorted using a BD FACSAria III flow cytometer (BD Biosciences, FACS Aria II) and collected in DMEM with 10% FBS. Sorted cells were centrifuged at 500 × g for 5 min at 4 °C, resuspended in PBS, and lysed in lysis buffer. Data were analyzed using FlowJo software (version 10.10.0).

STA-PUT velocity sedimentation

Testes from adult male C57BL/6 mice were decapsulated and digested in 5 ml Krebs buffer containing 0.25% trypsin, 100 μg/ml DNase I, and 250 μg/ml collagenase I at 37 °C for 21 min. Debris was removed by centrifugation at 200 × g for 5 min. Germ cells were resuspended in 15 ml Krebs buffer, centrifuged at 600×g for 5 min, washed once more, and filtered through a 200-mesh screen. The STA-PUT apparatus was assembled with a 2–4% BSA gradient in Krebs buffer (400 ml each of 2% and 4% BSA/Krebs buffer, mixed via a peristaltic pump at 60 rpm, 4 V). The single-cell suspension in 15 ml 0.5% BSA/Krebs buffer was layered onto the gradient. After 3 h of sedimentation at unit gravity, fractions were collected every 50 s (12 ml/tube, 50 tubes) using the pump at 48 rpm. Fractions were centrifuged at 600 × g for 5 min (RT) and resuspended in 1 ml Krebs buffer, and 40-μl aliquots were stained with Hoechst 33342 for fluorescence microscopy. Cell types were identified by nuclear morphology: pachytene spermatocytes and round spermatids. Remaining cells were pelleted (600 × g, 10 min, 4 °C), snap-frozen in liquid nitrogen, and stored at −80 °C. Like fractions were pooled in 15 ml Krebs buffer for downstream analysis.

G4 CUT&Tag library construction

ConA beads were equilibrated to room temperature, washed with 1 × ConA Bind Buffer, and resuspended. Mouse testicular cells (10⁶ per sample) were sorted via STA-PUT, pelleted, washed with Cell Wash Buffer, and resuspended in 90 µL buffer. Cells were mixed with 10 µL ConA beads per sample and incubated with rotation at room temperature for 10 min. Primary antibody (1:25) in Primary Antibody Buffer was added to the bead–cell complexes and incubated overnight at 4 °C. After washing, secondary antibody in Secondary Antibody Buffer was added and incubated at room temperature for 1 h, followed by two washes. For tagmentation, Dig-300 transposase dilution was added and incubated at room temperature for 1 h. The reaction was stopped with ethylenediaminetetraacetic acid (EDTA), sodium dodecyl sulfate (SDS), and Proteinase K at 70 °C for 10 min. DNA was purified using AFTMag NGS DNA Clean Beads, washed twice with 80% ethanol, and eluted in 22 µL nuclease-free water. PCR amplification was performed with 20 µL purified DNA, Gloria Nova 2× PCR Mix, and N5/N7 primers. The amplified product was purified again with beads and eluted in 22 µL 1 × Tris–EDTA (TE) buffer for sequencing.

G4 data analysis

Raw sequencing data were processed using Trim Galore (version 0.6.7, with 30 cores) to remove low-quality sequences and adapter contamination. Trimmed reads were aligned to the mouse genome (mm39) using Bowtie2 (version 2.5.1). Alignment parameters for single-end data were set to --local, --very-sensitive, --phred33, and -p 5, while paired-end data utilized --local, --very-sensitive, --no-mixed, --no-discordant, --phred33, -I 10, -X 700, and -p 5. The resulting SAM files were converted to BAM format using Samtools view (version 1.18, -b -F 0 × 04), sorted with Samtools sort (using 20 threads), and indexed with Samtools index. Blacklisted regions were excluded using Bedtools intersect (version 2.31.0) to minimize bias. Genome coverage was calculated with bamCoverage (version 3.5.0) using reads per genomic content (RPGC) or counts per million (CPM) normalization (--binSize 20, --normalizeUsing RPGC/CPM, --numberOfProcessors 50), generating BigWig files for visualization. Peak calling was performed using MACS2 callpeak (version 2.2.7.1) with parameters -f BAM, -g mm, -B, and -q 0.05 for single-end data, and -f BAMPE, -g mm, -B, and -q 0.05 for paired-end data; differential peak analysis was conducted with MACS2 bdgcmp. Peaks were annotated using ChIPseeker (version 1.38.0) and ChIPpeakAnno (version 3.18.0). Data visualization was achieved with IGV (version 2.6.3), while normalization and scaling were performed using computeMatrix (version 3.5.0, with --beforeRegionStartLength 3000, --regionBodyLength 5000, --afterRegionStartLength 3000). Heatmaps and profile plots were generated using plotHeatmap and plotProfile, sorted by sum and customized with color schemes such as blues and reds. Peak overlap was assessed with Intervene Venn, and enrichment analysis was conducted using clusterProfiler (version 3.14.3).

Chromosome spreading

Male C57BL/6 mice (8–12 weeks) were euthanized, and testes were dissected, rinsed in PBS, and stripped of the tunica albuginea. Seminiferous tubules were isolated, segmented, and treated hypotonically (30 mM Tris, 50 mM sucrose, 17 mM trisodium citrate, 5 mM EDTA, 2.5 mM dithiothreitol (DTT), 1 mM phenylmethylsulfonyl fluoride (PMSF), pH 8.2) for 30–60 min at room temperature. The tubules were dissociated in 100 mM sucrose (pH 8.2), and the cell suspension was spread onto slides precoated with fixative (1% paraformaldehyde, 0.15% Triton X-100, pH 9.2). Slides were incubated in a humid chamber for at least 3 h, air-dried, washed with 0.4% Photoflo, and stored at −80 °C for immunofluorescence staining.

Nanobody-based G4 binding protein proximity labeling system (nanoG4BPL)

GC-1 spg cells were cultured in T75 flasks and individually transfected with EGFP-SG4, EGFP-SG4-TurboID, or TurboID plasmids. Following transfection, cells were washed twice with PBS at 37 °C and incubated with 200 µM D-Biotin for varying time intervals. The biotinylation reaction was halted by rapidly transferring the cells to ice and washing them five times with prechilled PBS. For protein extraction, cells were lysed on ice for 30 min in radioimmunoprecipitation assay (RIPA) buffer containing a protease inhibitor cocktail (APExBIO Technology, catalog no. K1010), followed by centrifugation at 4 °C for 20 min. The supernatant was collected, and protein concentration was quantified using the bicinchoninic acid (BCA) assay. A portion of the protein sample was reserved for subsequent analysis, while the remainder was incubated with an appropriate amount of M-280 streptavidin magnetic beads on a horizontal rotator at room temperature for 45 min. The beads were then washed five times with 1 ml TBST buffer; each wash involved gentle mixing, brief centrifugation, and supernatant removal, with a fraction of the beads and final wash buffer retained for analysis. After removing residual supernatant, the magnetic beads were immediately frozen and stored at −80 °C for downstream experiments.

Label-free quantitative proteomics data analysis

Proteins were eluted from magnetic beads using Qinglian Biolysis buffer (Qinglian Biotech) and digested with the Qinglian MagicOmics-MMB8X kit (QLBIO). Briefly, 20 μL of protein sample was incubated with MMB beads at 37 °C for 30 min, mixed with 45 μL Binding buffer, and incubated with shaking at room temperature for 15 min. After discarding the supernatant, beads were washed three times with washing buffer, resuspended in 20 μL digestion buffer, and incubated at 37 °C for 4 h. Digestion was quenched with 5 μL Quench buffer, and samples were lyophilized. The lyophilized sample was resuspended in 10 μL mobile phase A (100% water, 0.1% formic acid) and centrifuged at 14,000 × g for 20 min at 4 °C, and 400 ng of supernatant was analyzed by liquid chromatography–mass spectrometry (LC–MS) using mobile phase B (80% acetonitrile, 0.1% formic acid; elution conditions in Supplementary Table 2). A timsTOF HT mass spectrometer (Bruker) with a Captive Spray ion source was used in data-dependent acquisition (DDA) mode, scanning m/z 100–1700 with a resolution of 60,000 at m/z 1222, ion mobility range of 0.6–1.6 cm²/(V·s), and a 1.1 s cycle time with ten parallel accumulation–serial fragmentation (PASEF) cycles. Data were searched against a Mus musculus UniProt database (55,260 proteins, downloaded 7 March 2023) using FragPipe software, excluding peptides shorter than 7 or longer than 40 amino acids.

Western blot and coomassie blue staining

Total protein was extracted from cells by washing twice with 1 mL PBS, lysing with RIPA buffer containing protease inhibitor cocktail (APExBIO Technology, catalog no. K1010) on ice for 30 min, and centrifuging at 4 °C for 20 min to collect the supernatant. Protein concentration was determined using a BCA kit (Thermo Fisher); 200 μL of BCA working reagent (A:B, 50:1) was added to protein standards and samples in a 96-well plate (triplicates) and incubated at 37 °C for 30 min, and absorbance was measured to calculate concentrations. For Western blot, 7.5% SDS-polyacrylamide gel electrophoresis (PAGE) gels (1.50 mm thick) were prepared with separating gel (4 mL separating gel solution, 4 mL buffer, 80 μL modified coagulant) and stacking gel (1.5 mL separating gel solution, 1.5 mL buffer, 20 μL modified coagulant). Proteins (15 μg) were denatured with Loading buffer at 98 °C for 10 min, cooled on ice, and electrophoresed at 200 V for 30 min until the bromophenol blue reached the gel bottom. Proteins were transferred to a methanol-activated polyvinylidene fluoride (PVDF) membrane (45 s activation) at 400 mA for 35 min at room temperature. The membrane was blocked with protein-free blocking solution (Epizyme, catalog no. PS108P) for 15 min at room temperature, incubated with primary antibody overnight at 4 °C, washed three times with 1 × TBST (10 min each), incubated with secondary antibody for 1 h, and washed again. For biotinylated protein detection, membranes were incubated with streptavidin–horseradish peroxidase (HRP) for 1 h at room temperature post-blocking, followed by the same washing steps. Chemiluminescence imaging was performed using the VILBER FUSION Solo 6S EDGE imaging system with auto-exposure mode settings. For Coomassie Blue staining, gels post-electrophoresis were washed in 50 mL deionized water, heated in a microwave for 3 min twice, and incubated in 50 mL deionized water on a shaker for 5 min. Gels were stained with 20 mL Coomassie Blue solution for 20 min with shaking until bands were visible, then destained with 100 mL deionized water (changed every 10 min) for 3 h, and imaged using a chemiluminescence imaging system.

CRISPR library screening

Hepa1-6 cells were transduced with lentiviral particles encoding Cas9 (lentiCas9-Blast, Addgene #52,962) at a multiplicity of infection (MOI) of 0.5. Lentiviral particles were generated by co-transfecting HEK293T cells with lentiCas9-Blast, psPAX2, and pMD2.G plasmids at a 4:3:1 ratio using Lipofectamine 2000 (Thermo Fisher Scientific). Transduced cells were selected with 15 μg/mL Blasticidin S for 7 days to establish a stable Cas9-expressing cell line (Hepa1-6-Cas9). A clustered regularly interspaced short palindromic repeats (CRISPR) single guide RNA (sgRNA) library targeting G4-binding protein genes (three sgRNAs per gene), cloned into lentiGuide-Puro (Addgene #52,963), was transfected into Hepa1-6-Cas9 cells (5 × 10⁵ cells per well in six-well plates) using Lipofectamine 3000 (Thermo Fisher Scientific). Knockdown efficiency of the CRISPR sgRNA library was validated by quantitative PCR (qPCR) using primers specific to the targeted genes, confirming reduced messenger RNA (mRNA) expression levels. Subsequently, cells were co-transfected with DR-GFP reporter (Addgene #26,475) and I-SceI expression plasmids (Addgene #26,477) using Lipofectamine 3000. At 48 h post-transfection, cells were harvested, washed with PBS, resuspended in PBS containing 1% FBS, and analyzed for GFP-positive cells by flow cytometry (BD FACSCanto II, BD Biosciences). A minimum of 10,000 events per sample were recorded, and data were analyzed using FlowJo software (version 10.10.0, BD Biosciences).

DNA oligonucleotides and ELISA analysis

The biotinylated DNA oligonucleotides were synthesized by Tsingke Biotech Co., Ltd. (China). The biotinylated oligonucleotides were folded in 10 mM Tris pH 7.4, 100 mM KCl, and 0.1% Tween-20, annealed by heating to 95 °C for 10 min, followed by slow cooling down to 23°C. Circular dichroism (CD) spectroscopy was performed using a Chirascan CD spectropolarimeter (Applied Photophysics). DNA oligonucleotides (hTeloG4, KitG4, MycG4, TbaG4, VegfG4, and mutant MYC) were prepared at 10 μM in 10 mM Tris–HCl, 100 mM KCl (pH 7.4). Samples were scanned in a 1-mm-pathlength cuvette over a wavelength range of 220–340 nm, with measurements taken every 1 nm and an integration time of 1 s per point. Spectra were recorded in triplicate, smoothed using Chirascan software, and plotted as molar ellipticity (deg cm² dmol⁻¹) versus wavelength (nm) using GraphPad Prism 10. For enzyme-linked immunosorbent assay (ELISA), streptavidin-coated 96-well plates (Thermo Fisher) were incubated with biotinylated oligonucleotides in 100 mM KCl, 50 mM KH₂PO₄ for 1 h at room temperature, then blocked with 3% BSA for 1 h. Plates were incubated with serially diluted protein (400 nM to 0 nM in 3% BSA) for 1 h, followed by anti-EGFP antibody (Abcam, ab290) for 1 h, and then HRP-conjugated secondary antibody (Abcam, ab6721) for 1 h. After washing, TMB substrate (Roche) was added, the reaction was stopped after 10 min, and absorbance was measured at 450 nm using a PHERAstar microplate reader (BMG Labtech). Binding affinity (K_d) was calculated by fitting data to a one-site specific binding model using GraphPad Prism 10.2.3.

SG4 protein map

MAEVELQASGGGFVQPGGSLRLSCAASGGTSGTYNMGWFRQAPGKEREFVSAISYRDNMTPYYADSVKGRFTISRDNSKNTVYLQMNSLRAEDTATYYCARYQGRLRIHQSTYWGQGTQVTVSS

Recombinant protein expression

Escherichia coli strains were streaked on Luria broth (LB) agar plates containing ampicillin (100 μg/mL final concentration; Beyotime Biotechnology, cat. No. ST008) and incubated at 37 °C overnight. A single colony was inoculated into 30 mL LB medium with ampicillin (100 μg/mL) and cultured at 37 °C, 220 rpm for 4 h. This culture was then transferred into 1970 mL prewarmed LB medium (with ampicillin, 100 μg/mL) in a 6-L flask and grown at 37 °C, 220 rpm for 3 h until the OD₆₀₀ reached 0.5–0.7. After cooling the culture for 1 h, protein expression was induced with 0.25 mM isopropyl β-d−1-thiogalactopyranoside (IPTG, final concentration), and incubation continued at 18 °C, 220 rpm overnight. Cells were harvested by centrifugation (9500 × g, 4 °C, 1 h), resuspended in 200 mL pre-chilled HEGX buffer, and supplemented with 2 mL protease inhibitor cocktail. Cells were lysed by sonication on ice (45 s per cycle, 50% duty cycle, output 7, 10–12 cycles, with cooling between cycles), and the lysate was clarified by centrifugation (16,000 × g, 4 °C, 30 min). The supernatant was loaded onto a 10 × Econo-Pac^® column prepacked with 2.5 mL Chitin Resin, prewashed twice with 20 mL HEGX buffer, and incubated at 4 °C overnight. The column was washed twice with 20 mL HEGX buffer, and protein was eluted with 65 mL HEGX buffer containing 100 mM DTT and protease inhibitor cocktail (6 mL per column), with gentle shaking at 4 °C for 48 h. The eluate was dialyzed using a Slide-A-Lyzer 10 K MWCO cassette in 2.5 L Tn5 dialysis buffer with 1.7 mM DTT at 4 °C for 1–2 h, followed by overnight dialysis in fresh buffer. The protein was concentrated using an Amicon Ultra-15 centrifugal filter (2880 × g, 4 °C, 1 h) and stored at −80 °C.

AlphaFold3 structural prediction

Structural predictions of the NBS1 protein in complex with G4 and non-G4 structured DNA were generated using AlphaFold3. The full-length human NBS1 sequence (UniProt ID: Q9R207) was used as the protein input. DNA sequences included MycG4 (5′-TGAGGGTGGGTAGGGTGGGTAA-3'), a G-quadruplex-forming sequence, and its mutated counterpart mMycG4 (5′-TGAGTGTGCGTAGTGTGTGTAA-3′). For the MycG4 system, two potassium ions (K⁺) were incorporated to stabilize the G-quadruplex structure. The AlphaFold3 pipeline was run with default parameters, utilizing multiple sequence alignments generated by MMseqs2. Five independent predictions were generated for each complex (NBS1-MycG4 and NBS1-mMycG4), and the top-ranked model was selected based on the highest predicted local distance difference test (pLDDT) score (> 90 for protein regions, > 85 for DNA interfaces). Predicted structures were visualized using PyMOL (version 2.5.0).

Molecular dynamics simulations

Molecular dynamics (MD) simulations were performed to validate AlphaFold3-predicted structures of NBS1 in complex with MycG4 and mMycG4. Simulations were conducted using GROMACS (version 2021.4) with the AMBER99SB-ILDN force field for proteins and the parmbsc1 force field for DNA. Initial coordinates were obtained from the top-ranked AlphaFold3 models. Each system was solvated in a cubic box of TIP3P water molecules with 10-Å padding, neutralized with 0.15 M NaCl, and, for the MycG4 system, supplemented with two K⁺ ions to stabilize the G-quadruplex. Systems were energy-minimized, equilibrated, and subjected to production MD simulations for 100 ns. Simulations were analyzed to confirm the stability of the complexes.

Protein and DNA preparation

The purified NBS1 protein was labeled with FITC for fluorescence detection. G4 DNA sequences, MycG4 and mMycG4, were synthesized and labeled with Cy3 for visualization. The DNA was diluted to a working concentration of 1 μM in phase separation buffer (20 mM Tris–HCl, pH 7.5, 100 mM NaCl, 1 mM MgCl₂, 5% PEG-8000). FITC-labeled NBS1 (final concentration 5 μM) was mixed with Cy3-labeled MycG4 or mMycG4 DNA (final concentration 1 μM) in a total volume of 20 μL of phase separation buffer. The mixture was incubated at room temperature for 10 min to allow droplet formation. Fluorescence images were acquired using an Olympus fluorescence microscope equipped with OLYVIA 4.1.1 software.

Testicular biopsy and cell isolation

Testicular biopsies were performed using fine-needle aspiration under local anesthesia in a clinical setting. The obtained tissue was divided: one portion was immediately fixed in 4% paraformaldehyde, embedded in paraffin, and sectioned for immunofluorescence. The remaining tissue was mechanically minced into small fragments, followed by enzymatic cell digestion and flow cytometry analysis (detailed in “Flow cytometry analysis of spermatogenic cells” section).

Statistical analysis

Data in the present study are presented as mean ± standard deviation (SD). GraphPad Prism 10.2.3.0 software (GraphPad, San Diego, CA, USA) was used for the statistical analyses. Significant differences between two groups were analyzed using the two-sided Student’s t-test and two-tailed Mann–Whitney U-test, and differences between multiple groups were measured using one-way analysis of variance followed by Bonferroni post hoc tests. A value of P < 0.05 was considered statistically significant for any differences. P-values are denoted in figures or figure legends by *P < 0.05, **P < 0.01, and ***P < 0.001. For HP2-induced apoptosis in testicular cells and secondary tissue comparisons, effect sizes (Cohen’s d) were computed using the pooled standard deviation from three independent biological replicates per group. Achieved power was then estimated for a two-sample, two-sided t-test at α = 0.05. Confidence intervals for differences in means were obtained using Welch’s method.

Results

G4 structures are enriched in testicular tissue, and their homeostasis is critical for spermatogenesis

To interrogate the function of G4 structures in testicular tissue, we treated cells from various organs with a previously developed G4-specific probe HP2 that preferentially interacts with G4 DNA structures over DNA duplexes, single-stranded DNA, RNA, and protein structures [31]. We then analyzed the cellular responses using flow cytometry. The results showed that, compared with tumor cells and cells from other organs, testicular cells exhibited more significant morphological changes and apoptosis after HP2 treatment (Fig. 1A–C, Supplementary Fig. S1A–D). These experiments indicate that testicular cells are more sensitive to disruptions in G4 structural dynamics.

We further explored the impact of G4 homeostasis imbalance on spermatogenesis through in vivo experiments. We intraperitoneally injected adult male mice with a highly specific G4 stabilizer pyridostatin (PDS) [32] for 14 days, during which no abnormal behaviors or body weight loss were observed (Supplementary Fig. S1E, F). Compared with the saline-treated control group, mice in the PDS-treated group exhibited significantly reduced testicular size and weight, while other organs, including liver, spleen, lung, and kidney, showed no apparent changes (Fig. 1D, E). Hematoxylin–eosin (H&E) staining revealed that the seminiferous tubules of PDS-treated mice were significantly narrowed, with disordered cell arrangement and a marked reduction in the number of spermatogenic cells (Supplementary Fig. S1G). To quantify changes in spermatogenic cells at different stages, we used Hoechst 33342 staining combined with flow cytometry to assess the cellular composition of mouse testes. The results showed that, compared with the control group, the PDS-treated group had a significant reduction in the number of spermatogenic cells at all stages, with the most pronounced decreases observed in spermatogonia (SPG), preleptotene spermatocytes (preL), and leptotene/zygotene spermatocytes (L/Z) (Fig. 1F, G). Immunofluorescence co-staining with SYCP3 and γH2AX, followed by quantification of germ cell populations in 20 randomly selected seminiferous tubules per stage based on established staining patterns and nuclear morphology, revealed a significant reduction in the meiotic cell population, particularly pachytene/diplotene spermatocytes, in PDS-treated testes (Fig. 1H, I) [33, 34]. We further treated mice with PDS for 12 h and 36 h and performed flow cytometry using R&D fluorescence-activated cell sorting (FACS), a more precise method we developed based on changes in RNA and DNA abundance, which indicated damage to leptotene/zygotene and pachytene/diplotene spermatocytes (Supplementary Fig. S1H, I). In summary, our findings indicate that the testis is more sensitive to G4 homeostasis imbalance compared with other organs.

G4 structures exhibit dynamic distribution at the cellular level during spermatogenesis

Elucidating the distribution dynamics of G4 structures across various spermatogenic cell types is essential for comprehending their functional roles in spermatogenesis. We initiated our investigation by conducting immunohistochemical (IHC) analysis utilizing anti-G4 antibodies, 1H6 and BG4, to delineate G4 localization patterns throughout the four distinct stages of the seminiferous epithelium. Both antibodies exhibited minimal nonspecific background staining and demonstrated differential cellular staining intensities, indicating specific labeling of testicular G4 structures and suggesting that G4 abundance varies among cells at different spermatogenic stages (Fig. 2A, Supplementary Fig. S2A). However, the staining patterns of 1H6 and BG4 exhibited distinct trends from spermatogonia to spermatids, suggesting differences in their recognition targets (Fig. 2A, Supplementary Fig. S2A). It has been reported that BG4 recognizes both DNA and RNA G4s, while 1H6 is specific to DNA G4s [35, 36]. These divergent staining patterns imply that DNA G4s and RNA G4s have different distribution dynamics during spermatogenesis.

Fig. 2.

G4 structures exhibit dynamic distribution at the cellular level during spermatogenesis. A Immunohistochemical staining of mouse testis sections with anti-G4 antibody 1H6 across different stages of the seminiferous epithelium (XI, IV–VI, XII, and V–VI). Cell types are indicated as Ser (Sertoli cells), Spg (spermatogonia), Lep (leptotene), Zyg (zygotene), Pac (pachytene), RS (round spermatids), EES (early elongating spermatids), and LES (late elongating spermatids). Scale bar, 100 μm. B Immunofluorescence co-staining of Sertoli cells identified by SOX9 with 1H6 and DAPI. Scale bar, 50 μm. C Immunofluorescence co-staining of undifferentiated spermatogonia (Undiff SPG) identified by PLZF and differentiated spermatogonia (Diff SPG) identified by c-KIT with 1H6 and DAPI. Scale bar, 50 μm. D Immunofluorescence co-staining of spermatocytes at different meiotic stages with 1H6 and SYCP3. Stages and corresponding seminiferous tubule stages are indicated as pre-L (preleptotene), L (leptotene), Z (zygotene), P (pachytene), and D (diplotene). Scale bar, 2 μm. E Immunofluorescence co-staining of spermatids at different developmental steps (I–XVI) with 1H6 and ACRV1. DAPI marks nuclei. Scale bar, 2 μm. F Schematic representation illustrating the relative staining intensity of 1H6 and BG4 antibodies in different stages of spermatogenic cells and Sertoli cells. Intensities were quantified from co-immunostaining with stage-specific markers, analyzing 20 cells per type

We further employed multicolor immunofluorescence techniques, combining the anti-G4 antibody with specific antibodies targeting distinct germ cell types, to investigate substage-specific variations in G4 structures during spermatogenesis. Additionally, we aimed to elucidate the biological significance underlying the differential staining patterns observed with the two G4 antibodies by comparing their staining profiles. We used SOX9 as a marker for Sertoli cells, PLZF to identify undifferentiated spermatogonia, and c-Kit to label differentiated spermatogonia. Spermatocytes were identified by SYCP3 staining of the chromosomal axis and classified into preleptotene, leptotene, zygotene, pachytene, and diplotene stages on the basis of chromosomal behavior. Additionally, acrosomal vesicle protein 1 (ACRV1) was used to categorize spermatids into 16 developmental steps (steps 1–16).

Immunofluorescence results of both antibodies showed that Sertoli cells exhibited the lowest G4 levels, with only faint nucleolar staining observed (Fig. 2B). However, the staining patterns of the two antibodies diverged in subsequent cell stages. In preleptotene spermatocytes, 1H6 staining showed a temporary decrease in G4 levels, followed by a resurgence in leptotene spermatocytes. Subsequently, G4 levels remained low in zygotene, pachytene, and diplotene spermatocytes (Fig. 2D). In contrast, BG4 staining showed elevated G4 signals in the leptotene stage, which persisted through zygotene and pachytene stages before gradually declining in diplotene spermatocytes (Supplementary Fig. S2D).

Upon entering the round spermatid stage, both antibodies demonstrated a decrease in G4 signal intensity. Specifically, G4 signals weakened further in the nucleoplasm while progressively intensifying in heterochromatin regions (Fig. 2E, Supplementary Fig. S2E). During the subsequent round-to-elongated spermatid transition, the staining patterns of the two antibodies diverged again. In 1H6 staining, G4 levels in round spermatids (RS) surged dramatically at step 9, reaching the highest point observed during spermatogenesis, and this peak persisted through step 12 (Fig. 2E). In contrast, BG4 staining showed only a slight increase in G4 signals during this period (Supplementary Fig. S2E). By step 13, when ACRV1 expression was restricted to the sperm head, both antibody staining results revealed that G4 signals disappeared and remained absent until step 16 (Fig. 2E).

On the basis of these staining results, we generated a profile of G4 intensity changes detected by both antibodies during spermatogenesis (Fig. 2F). As spermatogonia differentiated, G4 levels increased in staining with both antibodies, suggesting that G4 structures are involved in spermatogonial differentiation and that RNA G4 formation is also abundant at this stage. Subsequently, 1H6 staining showed elevated G4 levels in leptotene spermatocytes, followed by a decrease, implying a critical role for DNA G4 structures during the initiation of meiosis. In contrast, BG4 staining revealed persistent G4 signals through zygotene and pachytene stages, during which transcriptional activity in spermatocytes first decreases and then increases broadly, potentially accompanied by elevated RNA G4 signals.

Upon entering the round spermatid stage and during the round-to-elongated spermatid transition, RNA transcription gradually declined and eventually ceased, leading to a decrease in RNA G4 levels, consistent with BG4 staining patterns. In contrast, 1H6 staining showed a dramatic increase in G4 levels during this transition. This sharp increase in G4 signals closely aligned with the timing of the histone-to-protamine transition in spermatids. During this phase, extensive topoisomerase II-mediated reduction in genomic supercoiling occurs [37, 38], while topoisomerase II has been reported to bind G4 structures [39, 40], suggesting a potential role for G4 structures in the critical process of chromatin remodeling during this transition phase.

Stage-specific genomic distribution reveals dual functions of DNA G4 structures in gene regulation and genome organization

The dynamic distribution of G4 structures in spermatogenic cells prompted us to further investigate their genomic distribution features in various cell subpopulations using omics approaches. We employed the STA-PUT velocity sedimentation system to fractionate adult mouse testicular cells into four distinct populations: spermatogonia (SPG), spermatocytes (SPC), round spermatids (RS), and elongated spermatids (LS). On the basis of the aforementioned staining results, we concluded that the 1H6 antibody more accurately represents the distribution of DNA G4 structures. Therefore, we selected the 1H6 antibody for use with CUT&Tag-seq to systematically map the spatiotemporal dynamics of G4 structures during spermatogenesis, generating a comprehensive genome-wide G4 atlas.

Visualization of the overall data using Covplot revealed that G4 structures were broadly distributed across all chromosomes, with significant differences in signal intensity among spermatogenic stages (Supplementary Fig. S3A). Specifically, G4 signals were markedly higher in SPG, SPC, and RS cells compared with LS cells. Further comparison of G4 signals across entire transcripts showed that the transcription start site (TSS)-related regions exhibited much stronger signals than other areas (Fig. 3A). Among the four stages, spermatogonia displayed the highest signal intensity, followed by slightly higher signals in spermatocytes than in round spermatids, with elongated spermatids showing the lowest intensity (Fig. 3B).

Fig. 3.

Stage-specific genomic distribution reveals dual functions of DNA G4 structures in gene regulation and genome organization. A Heatmap showing genomic distribution of G4 peaks across different spermatogenic stages using CUT&Tag sequencing technology in mouse testicular cells sorted into four populations (SPG, SPC, RS, and LS) by STA-PUT. B Genomic metaplot showing relative signal intensity of G4 peaks across entire gene regions (from 3 kb upstream of transcription start site (TSS) to 3 kb downstream of transcription end site (TES)). C Venn diagram showing overlap of G4 peaks between different spermatogenic stages and previously published mouse embryonic stem cell (mESC) dataset GSE173103. D Pie charts showing genomic distribution of G4 peaks across different spermatogenic stages. E Genome browser tracks displaying G4 peak distribution in two representative chromosomal regions. F Heatmap showing colocalization of G4 peaks with various histone modifications (H3K4me1, H3K27ac, and H3K36me3) and RNA polymerase II (POLR2A) across different spermatogenic stages. G–H Heatmaps showing colocalization of G4 peaks with CTCF (G) and SPO11 (H) across different spermatogenic stages

Analysis of CUT&Tag sequencing data revealed a pronounced stage-specific distribution of G4 structures. We identified 15,472 G4 peaks in SPG, 16,820 in SPC, 5,187 in RS, and only 425 in LS. To validate the reliability of our experimental system, we rigorously compared our G4 CUT&Tag data with publicly available datasets from mouse embryonic stem cells (mESCs) (GSE173103) [41]. By aligning the top 3000 G4 peaks with the dataset GSE173103, we found substantial overlap across cell types: 41.2% for SPG, 41.7% for SPC, 50.8% for RS, and 38.8% for LS, reinforcing the robustness and reproducibility of our findings and also indicating that some common G4 structures exist in both spermatogenic cells and other cell types, which may play fundamental roles in cell survival (Fig. 3C). Comparison with the ENDOQUAD database [42] further revealed significant colocalization of germ cell G4 with endogenous G4 (Supplementary Fig. S3B). Gene annotation of the top 3000 G4 peaks revealed that, except in LS, G4 peaks were predominantly enriched in promoter regions, with the most significant enrichment observed in SPG and SPC (Fig. 3D). In contrast, LS exhibited a clear shift in G4 distribution toward distal intergenic regions, which is consistent with the fact that transcription has been shut down in LS cells [43] (Fig. 3D). Similarly, genome browser tracks confirmed G4 enrichment in promoters in SPG, SPC, and RS, validating consistency with GSE173103 (Fig. 3E).

To explore the mechanistic roles of G4 structures in transcriptional regulation, we integrated our data with histone modification profiles (H3K4me1, H3K4me3, H3K36me3, H3K27ac, and H3K27me3) and POLR2A profiles from the ENCODE database for mouse testis. Heatmap analysis revealed strong colocalization of G4 peaks with active epigenetic marks, including H3K4me1, H3K4me3, H3K36me3, and H3K27ac, suggesting that G4 structures may facilitate chromatin accessibility and transcription initiation, thereby influencing gene expression during spermatogenesis (Fig. 3F, Supplementary Fig. S3C). Conversely, no significant G4 signals were detected in H3K27me3-enriched (repressive) regions (Supplementary Fig. S3C). Notably, G4 peaks were highly enriched at POLR2A binding sites, indicating their potential role in promoting transcription complex assembly and regulating gene expression (Fig. 3F).

Although G4 structures are predominantly located at promoters in SPG, SPC, RS, and LS, the overlap of G4-associated genes across these stages is limited, underscoring dynamic, stage-specific regulation (Supplementary Fig. S3F). To further explore these unique patterns, we conducted Gene Ontology (GO) enrichment analysis on all G4-associated genes within each specific cell type. GO enrichment analysis of genes proximal to G4 peaks further elucidated their functional role (Fig. S3G–J). In SPG, SPC, and RS, G4-associated genes were primarily linked to biological pathways such as chromosome segregation, nuclear division, and RNA splicing. Stage-specific enrichment was also evident: SPG genes were enriched in mitotic cell cycle transition pathways; SPC genes were enriched in double-strand break repair, underscoring their role in meiotic recombination; and RS genes were enriched in flagellar organization and assembly, reflecting the demands of spermiogenesis. In LS, G4-associated genes shifted toward pathways related to cell recognition, Wnt signaling, hormone secretion, and steroid metabolism, consistent with sperm maturation processes.

We further analyzed the relationship between G4 and CTCF, a key regulator of three-dimensional genome architecture. The results showed that, in SPG, SPC, RS, and LS, G4 significantly colocalized with CTCF, suggesting that G4 structures may contribute to chromatin looping and topological organization (Fig. 3G).

The transient elevation of G4 fluorescent signals during the leptotene stage suggests a close association between G4 structures and homologous recombination (Fig. 2D). The Spo11-mediated DSBs initiate meiotic recombination. SPO11 creates these DSBs through a topoisomerase-like reaction that links a SPO11 molecule to each DNA end. DNA nicks nearby release SPO11 covalently bound to short oligonucleotides (SPO11-oligos). Thus, the sequencing data of mouse SPO11-oligos represent the hotspots of recombination [44]. To explore the relationship between G4 structures and meiotic recombination hotspots, we compared our G4 data with the sequencing data of SPO11-oligos. We were surprised to find that G4 signals colocalized to some extent with the SPO11-oligos data (Fig. 3H). Given that G4 structures are predominantly located in promoter regions (Fig. 3D), we investigated the relationship between SPO11-oligos and gene promoters, revealing no enrichment of SPO11-oligos at promoters (Fig. S3D) [44]. Similar results were obtained from END-seq data in PRDM9 WT mice, whereas PRDM9 knockout led to DSB enrichment at promoters (Supplementary Fig. S3E) [45]. These data suggest colocalization between nonpromoter G4 regions and SPO11-oligos, highlighting potential G4-mediated regulation of DSB sites.

Although SPO11-oligos are generated in leptotene spermatocytes and DSBs are repaired by the pachytene/diplotene stage, we found that G4 signals in spermatogonia and spermatozoa also colocalized with SPO11 (Fig. 3H). This suggests that, in addition to PRDM9 determining the location of hotspots, G4 structures may also play a role in determining hotspot locations. After all, studies have shown that megabase-scale similarities of DSB patterns among individuals with different PRDM9 alleles indicate that there are other factors besides PRDM9 that determine the location of hotspots [46].

DNA G4 homeostasis perturbation impairs DNA double-strand break repair

Above findings indicate that G4 structures play dual roles: (1) collaborating with active epigenetic marks to regulate stage-specific gene expression and (2) acting as chromatin structural elements to facilitate functional genome partitioning and special DNA events, such as participating in homologous recombination during meiosis. Therefore, we further investigated the impact of G4 homeostasis imbalance on meiosis.

We treated postnatal day 9 (P9) mice, which are just entering meiosis (preleptotene stage), with PDS and euthanized them at postnatal day 18 (P18) when spermatocytes have completed the first meiotic division (Fig. 4A). Testicular atrophy and reduced testicular weight were observed (Fig. 4B, C). The seminiferous tubules were narrowed, with evident damage to spermatocytes (Fig. 4D). Flow cytometry analysis similarly indicated impaired meiosis in these mice (Fig. 4E). To clarify that the above phenotypes are a result of PDS action on DNA, we treated mice with PDS and observed that the number of peaks in spermatocytes was significantly higher after PDS treatment compared with the untreated group (23,368 versus 16,820), and the normalized DNA G4 signal in the promoter regions and in the peaks shared between pre- and post-treatment PDS groups was higher than in the control group (Supplementary Fig. S4A, B).

Fig. 4.

DNA G4 homeostasis perturbation impairs DNA double-strand break repair. A Schematic timeline of experimental design for mouse treatment with pyridostatin (PDS). B Representative images of testes from control (CTRL) and PDS-treated mice at P18. C Quantification of testicular weight in control and PDS-treated mice. Data represent mean ± SD from three independent experiments; ***P < 0.001. D H&E staining of seminiferous tubules from control and PDS-treated mice at P18. E Flow cytometric quantification of spermatocytes. Data represent mean ± SD from three independent experiments; ***P < 0.001. F Chromosome spreads of spermatocytes at different meiotic stages (leptotene, zygotene, pachytene, and diplotene) from control and PDS-treated mice. Scale bar, 10 μm. G The proportion of autosomes DSB repaired ratio in CTRL and PDS spermatocytes. Data represent mean ± SD from three independent experiments; **P < 0.01. H Immunofluorescence staining of HEP1-6 cells treated with 5 μM PDS or control, showing γH2AX and –DAPI staining. Scale bar, 100 μm. I Quantitative analysis of the percentage of γH2AX-positive foci per nucleus in HEP1-6 cells over time (0, 0.5, 1, 2, 4, 8, 16, 24, and 32 h) following PDS treatment. Data represent mean ± SD from 20 nuclei analyzed per time point

Subsequent chromosome spread analysis, stained for SYCP3 and γH2AX, revealed,that in leptotene and zygotene spermatocytes, γH2AX signals were uniformly distributed across the nuclei in both PDS-treated and control groups, with no significant differences (Fig. 4F, G). This suggests that excessive G4 stabilization by PDS does not substantially affect the initial formation of DSBs during early meiosis. However, in pachytene and diplotene spermatocytes, marked differences emerged. In the control group, DSB repair synchronized with the completion of autosomal synapsis, as evidenced by the disappearance of γH2AX signals from autosomes, with staining restricted to the XY body. In contrast, PDS-treated spermatocytes retained widespread γH2AX signals on autosomes (Fig. 4F, G). This persistent γH2AX staining indicates that excessive G4 stabilization disrupts DSB repair via the homologous recombination (HR) pathway.

To further validate these findings, we used HEP1-6 cells as an in vitro model. Following PDS treatment, we observed significant accumulation of γH2AX foci in the nucleus, with the number of foci increasing over time (Fig. 4H, I, Supplementary Fig. S4C). This time-dependent increase in γH2AX foci confirmed that G4 stabilization significantly inhibits DNA repair, consistent with the in vivo observations in spermatocytes. However, while these results demonstrate that G4 stabilization impairs HR-mediated repair, the underlying mechanism remains unclear. To elucidate which specific proteins in the DNA damage and repair pathway are affected, we next sought to identify G4-interacting proteins that might mediate this process.

High-resolution mapping of G4-interacting proteins using the nanobody-based proximity labeling system reveals their role in DNA repair

Identifying the functional executors of G4 structures—G4-binding proteins—is critical for understanding their biological roles. To address limitations in current methods for studying G4–protein interactions, we developed a strategy based on the high affinity and specificity of antigen–antibody recognition. We cloned the gene encoding SG4, a camelid-derived single-domain nanobody comprising 135 amino acids [30]. The small size of SG4 minimizes steric interference, allowing access to structurally constrained G4 regions, while its nanomolar binding affinity across diverse G4 conformations enables robust interaction capture. To systematically profile G4-associated proteins, we fused SG4 with TurboID [47], a proximity-dependent biotin ligase and an EGFP tag (Supplementary Fig. S5A). This configuration permits real-time visualization of G4 dynamics through fluorescence imaging while enabling streptavidin-based enrichment of proteins interacting with or proximal to G4 structures under native chromatin conditions.

We first expressed and purified the EGFP-SG4-TurboID (EST) fusion protein using the CBD prokaryotic expression system (Fig. 5A, Supplementary Fig. S5B). To verify whether EST possesses the same affinity for G4 molecules as the SG4 nanobody via ELISA, we constructed biotinylated DNA oligonucleotides with various topologies, including parallel MycG4, Kit1G4, VegfG4, antiparallel TbaG4, and hybrid hTeloG4, as well as a non-G4 control (mutated MycG4), with their structures confirmed by circular dichroism (Supplementary Fig. S5D). These biotinylated DNAs were immobilized on streptavidin-coated ELISA plates, and the binding affinity of EST was detected using an EGFP antibody (Fig. 5B). ELISA assays demonstrated that EST exhibited nanomolar affinity for various DNA oligonucleotides: MycG4 (4.542–5.812 nM), TbaG4 (3.147–4.246 nM), VegfG4 (6.668–9.220 nM), Kit1G4 (12.98–19.60 nM), and hTeloG4 (66.22–100.5 nM), while showing no binding to the mutated MycG4 (Fig. 5C). Electrophoretic mobility shift assay (EMSA) assays also revealed the binding of EST to G4 structures (Supplementary Fig. S5C).

Fig. 5.

High-resolution mapping of G4-interacting proteins using the nanobody-based proximity labeling system reveals their role in DNA repair. A Coomassie Blue staining showing purification of EGFP-SG4-TurboID (EST) fusion protein. Lanes represent the following steps: IPTG induction (± IPTG), first and second flow-through (F.T.), wash, DTT elution (+ DTT), and SDS elution. Protein markers indicate EGFP-SG4-TurboID and Intein-CBD. B Schematic of ELISA assay system for evaluating EGFP-SG4-TurboID binding affinity. C ELISA-based binding affinity curves of EGFP-SG4-TurboID with DNA oligonucleotides (Kit1G4, MycG4, TbaG4, hTeloG4, VegfG4, mutated MycG4, no DNA, and no SG4 controls). D Immunofluorescence images of GC-1 spermatogonial cells expressing EGFP-SG4-TurboID, EGFP-SG4-TurboID without biotin (-Bio), or EGFP-SG4 without TurboID. E Schematic workflow of nanoG4BPL. F Biotin incubation time screening for nanoG4BPL. G Western blot detection of EGFP expression in protein extracts from GC-1 cells transfected with EGFP-SG4 (ES), EGFP-SG4-TurboID (EST), or TurboID alone (T). H Western blot detection of EGFP expression in magnetic beads from cells transfected with EGFP-SG4 (ES), EGFP-SG4-TurboID (EST), or TurboID alone (T). I Venn diagram showing overlap of G4-interacting proteins identified by nanoG4BPL (SG4) with the following datasets: PLGPB, LIMCAP, and CMPP. J Gene Ontology (GO) enrichment analysis of G4-associated proteins identified by nanoG4BPL

We further validated the localization and enzymatic activity of EST in the GC-1 spermatogonial cell line. Upon expression, EST localized predominantly to the nucleus as confirmed by EGFP fluorescence (Fig. 5D). Specificity was assessed using a streptavidin red fluorescent probe following biotin incubation, which showed significant colocalization of streptavidin signals with EGFP, while no red fluorescence was observed in controls lacking TurboID or biotin treatment (Fig. 5D). These results confirm the high specificity and biotin labeling activity of EST in recognizing diverse G4 structures both in vitro and in vivo. We therefore termed this system as nanobody-based G4 binding protein proximity labeling system (nanoG4BPL).

Utilizing EST, we implemented a proximity labeling workflow to profile G4-interacting proteins in live GC-1 cells (Fig. 5E). Cells were transiently transfected with EST, with controls expressing EGFP-SG4 (lacking TurboID) or TurboID alone (for diffuse labeling). We optimized the biotin incubation time to 10 min (Fig. 5F). Western blot analysis using an EGFP antibody confirmed the expression of ES and EST in the cells (Fig. 5G, Input). Detection with streptavidin-HRP indicated that both EST and TurboID alone (T) biotinylated proximal proteins, with T showing broader biotinylation (Fig. SE, Input). During proximity labeling, the fusion protein first biotinylates itself. As expected, we successfully detected EST, but not ES, in samples precipitated by M-280 streptavidin magnetic beads (Fig. 5H), demonstrating that EST effectively biotinylated itself and neighboring proteins, which were subsequently captured by the streptavidin beads.

Subsequent four-dimensional (4D) label-free mass spectrometry identified 3330, 3614, and 3577 proteins in the ES, EST, and T, respectively (Supplementary Fig. S5F). Applying stringent criteria (EST/ES ≥ 2.0 and EST/T ≥ 2.0), we precisely identified 302 proteins specifically associated with G4 structures. These included known G4-interacting proteins such as Rnaseh1, Timeless, Sirt1, and Mecp2, with significant overlap with existing G4-interacting protein databases (Fig. 5I). For instance, 11 proteins overlapped with CMPP [27], 7 with PLGPB [29], and 3 with LIMCAP [26], confirming the reliability and robustness of nanoG4BPL.

Gene Ontology (GO) enrichment analysis of the 302 G4-associated proteins revealed enrichment in canonical G4-related pathways, such as RNA splicing, RNA polymerase II activity, telomere maintenance, and histone modification, as well as DNA double-strand break (DSB) repair and homologous recombination (HR) pathways (Fig. 5J). This corroborated our earlier findings, reinforcing the critical link between G4 structures and HR-mediated DNA repair during spermatogenesis. Notably, we also identified enrichment in ribosome biogenesis and ribosomal RNA (rRNA) processing pathways, suggesting a novel role for G4 structures in rRNA metabolism. This finding aligns with recent reports demonstrating nucleolar abnormalities and disrupted rRNA metabolism in oocytes following RHX36 knockout [48].

NBS1 regulates homologous recombination repair at G4-enriched DSB hotspots

On the basis of the GO enrichment analysis of potential G4-interacting proteins identified in the nanoG4BPL and corroborated by literature review, we selected 60 potential G4-interacting proteins associated with DNA damage repair. We then constructed a CRISPR library, aiming to screen for proteins involved in meiotic homologous recombination and interacting with G4 using the DR-GFP homologous recombination reporter system (Fig. 6A, Supplementary Fig.S6A). Among the 37 proteins that promoted homologous recombination, we focused particularly on NBS1 and ataxia-telangiectasia mutated (ATM) (Fig. 6B). NBS1, a subunit of the MRN complex (Mre11-RAD50-NBS1), is known to be the first to reach the DNA ends following SPO11-induced DSBs during meiosis, acting as an adaptor protein to recruit Mre11 and RAD50 [49, 50]. In response to DSBs, the FxF/Y motif of NBS1 directly binds to ATM and is essential for retaining active ATM at sites of DNA damage [51]. Recent studies have also confirmed that cells become more sensitive to the cytotoxic effects of PDS when NBS1 is knocked out [39]. Therefore, we propose that NBS1 may be a G4-binding protein that forms foci on G4 structures, further recruiting molecules such as Mre11, RAD50, and ATM.

Fig. 6.

NBS1 regulates homologous recombination repair at G4-enriched DSB hotspots. A Schematic of CRISPR-based screening workflow. B Scatter plot depicting HR repair efficiency of 60 screened proteins detected by DR-GFP system using flow cytometry, with y-axis representing GFP positivity rate, corresponding to homologous recombination efficiency. C, D Heatmaps showing colocalization of NBS1 with G4 structures (SPC_G4) and SPO11-oligos at transcription start sites (TSS) in spermatocytes. Signal intensity is plotted within ±3 kb windows centered on TSS. E Genome browser tracks showing distribution of NBS1, G4 structures (SPC_G4), and SPO11 signals in representative regions. F Computational simulation of conformational changes in NBS1 upon binding to MycG4 and mMycG4. G Molecular dynamics simulation snapshots of NBS1 binding to MycG4 and mMycG4 at 0, 25, 50, 75, and 100 ns. H Time course plots of root mean square deviation (RMSD) over 100 ns following NBS1 binding to MycG4 and mMycG4. I Time course plots of root mean square fluctuation (RMSF) for NBS1-MycG4 and NBS1-mMycG4 complexes over 100 ns. J Time course plots of solvent-accessible surface area (SASA) following NBS1 binding to MycG4 and mMycG4

Using NBS1 chromatin immunoprecipitation sequencing (ChIP-seq) data (GSE66424) [52], we observed significant enrichment of NBS1 at transcription start sites (TSS), which correlated well with the distribution of G4 structures (Supplementary Fig. S6B). Further analysis revealed that G4 structures of spermatocytes colocalize with NBS1 (Fig. 6C). The endogenous G4 signature from the ENDOQUAD database also showed strong colocalization with NBS1 (Supplementary Fig. S6C). Additionally, we found partial colocalization of SPO11-oligos with NBS1 (Fig. 6D). Genome browser tracks further confirmed colocalization of NBS1, G4, and SPO11 (Fig. 6E). Collectively, these data indicate that NBS1 is indeed associated with G4 structures and participates in the repair of SPO11-mediated DSBs.

However, whether the NBS1 protein directly interacts with G4 structures or is recruited to G4 sites through other proteins remains to be further investigated. We used AlphaFold3 to predict the structural characteristics of NBS1 when interacting with G4 and non-G4 structured DNA molecules. In silico modeling demonstrated that, when the MycG4 DNA sequence and two potassium ions were introduced, MycG4 successfully folded into a G4 structure, while the mutated MycG4 (mMycG4) remained in a linear configuration (Supplementary Fig. S6D, E). When NBS1 was computationally docked with either MycG4 or mMycG4, both interactions induced conformational changes in NBS1, creating a defined binding pocket compared with the unbound protein state (Fig. 6F). However, structural differences between the folded MycG4 and linear mMycG4 substrates suggested distinct binding modes: the ordered MycG4 structure promoted stable NBS1 engagement, whereas the flexible mMycG4 might permit transient interactions.

Molecular dynamics simulations validated these structural predictions (Fig. 6G). The NBS1-MycG4 complex maintained greater conformational stability, exhibiting reduced positional fluctuations and solvent-accessible surface area compared with the NBS1-mMycG4 complex, which showed increased mobility and surface exposure (Fig. 6H–J). These observations provide evidence for direct NBS1 recognition of G4 structures, which appears to stabilize the protein’s conformational state.

NBS1 exhibits phase separation properties when binding to DNA G4 structures

The presence of intrinsically disordered regions (IDRs) within the NBS1 structure (Fig. 6F, Supplementary Fig. S7A) raises the possibility that G4 DNA binding could induce liquid–liquid phase separation in NBS1, potentially facilitating the assembly of DNA repair complexes at G4-enriched double-strand break (DSB) hotspots during spermatogenesis. Notably, the MRN complex has been previously reported to undergo compartmentalization and concentration at DNA damage sites [53]. Building on these observations, we plan to employ in vitro experiments to validate the direct interaction between NBS1 and G4 structures and determine whether G4 binding promotes NBS1 phase separation.

To test this hypothesis, we expressed and purified recombinant NBS1 protein in vitro using the CBD prokaryotic expression system and assessed its binding affinity for G4 DNA structures via ELISA (Fig. 7A). NBS1 exhibited robust binding to various G4 DNA topologies, including parallel KitG4, antiparallel TbaG4, and hybrid hTeloG4, while showing significantly weaker affinity for mMycG4 sequence (Fig. 7B). These results confirm the high specificity of NBS1 for G4 structures, consistent with its role in homologous recombination repair at G4-enriched genomic loci (Fig. 6C).

Fig. 7.

NBS1 exhibits phase separation properties when binding to DNA G4 structures. A Coomassie Blue staining showing purification of recombinant NBS1 protein using CBD prokaryotic expression system. Purification steps are as follows: IPTG induction (−IPTG, + IPTG), flow-through (F.T.), first and second washes (1st wash, 2nd wash), DTT elution (+ DTT), and final SDS stripping. Purified NBS1 (~ 100 kDa) and Intein-CBD (~ 35 kDa) bands are labeled. B ELISA binding assay showing NBS1 affinity for various DNA G4 structures. C Confocal microscopy images of in vitro liquid–liquid phase separation assay. Scale bar, 50 μm. D Immunofluorescence staining of U2OS cells treated with 5 μM pyridostatin (PDS) for 24 h or mock treatment. Cells are stained with DAPI, γH2AX, and NBS1. Scale bar, 20 μm

We next performed in vitro phase separation assays to evaluate whether NBS1-G4 interactions promote phase separation. FITC labeled NBS1 was incubated with Cy3-labeled MycG4 or mMycG4 DNA. In the MycG4 group, confocal microscopy revealed colocalization of NBS1 and G4 DNA within distinct, dynamic liquid-like droplets, indicative of phase-separated condensates. In contrast, no colocalization or droplet formation was observed in the mMycG4 group, highlighting the specificity of G4 DNA in driving NBS1 phase separation (Fig. 7C, Supplementary Videos 1, 2).

To investigate whether NBS1 foci formation increases following PDS-induced G4 stabilization, we treated U2OS cells with pyridostatin (PDS, 5 μM) for 24 h to stabilize G4 structures and examined NBS1 recruitment patterns. Immunofluorescence staining revealed a statistically significant elevation in NBS1 foci formation in PDS-treated cells compared with untreated controls (Fig. 7D). These results further support the notion that G4 stabilization enhances NBS1 recruitment through phase separation. The phase separation properties of NBS1–G4 interactions provide a novel molecular framework for understanding how G4 structures orchestrate the spatiotemporal organization of DNA repair machinery in meiotic cells.

Abnormal G4 accumulation in spermatogenic cells links oxidative stress to male infertility

Above results have established that dynamic regulation of G-quadruplex (G4) structures is critical for spermatogenesis. Oxidative damage induced by external factors such as smoking generates 8-oxoguanine, a lesion known to promote G4 formation [54, 55]. Such oxidative stress may lead to spermatogenic defects, contributing to male infertility [56]. Therefore, we propose that patients with idiopathic infertility may exhibit elevated G4 levels.

To test this hypothesis, we performed G4-specific staining on testicular tissues from patients with idiopathic nonobstructive azoospermia (NOA) and Y-chromosome AZFc microdeletion, comparing them with samples from obstructive azoospermia (OA) patients exhibiting normal spermatogenesis. Immunohistochemical analysis identified notably higher G4 levels in the seminiferous tubules of patients with NOA relative to OA controls (Fig. 8A, B). Patients with AZFc microdeletions also exhibited elevated G4 content, which may correlate with chromatin structural irregularities resulting from defective homologous recombination (Fig. 8A, B). For quantitative validation, we applied the G4-specific probe HP2 and the DNA dye Hoechst 33342 in flow cytometry assays using testicular single-cell suspensions derived from biopsy samples. This analysis revealed that G4 levels were significantly increased in tetraploid spermatocytes, suggesting a potential association between elevated G4 content and meiotic abnormalities in affected patients (Fig. 8C, D). These data position G4 dysregulation as a potential mechanistic link between environmental damage and male reproductive disorders, providing insights into the etiology of infertility in affected populations.

Discussion

Previous studies investigating the roles of G-quadruplexes (G4s) during spermatogenesis have primarily centered on the G4-resolving enzyme DHX36. Deletion of Dhx36 in embryonic day 15.5 gonocytes via Vasa-Cre resulted in arrested spermatogonial differentiation [23]; targeted knockout of Dhx36 in type A1 spermatogonia using a Stra8-GFPCre knock-in mouse model induced meiotic defects and disrupted spermiogenesis, ultimately causing male infertility [24]. In oocytes, conditional knockout of Dhx36 led to enlarged nucleoli, aberrant chromatin configuration, reduced chromatin accessibility, transcriptional dysregulation, and meiotic arrest [48]. Collectively, these findings implicate G4 structures as critical regulators of spermatogonial differentiation, meiotic progression in spermatocytes, and chromatin condensation in spermatozoa. However, DHX36 possesses dual functionality, as it can resolve both DNA and RNA G4 structures [48]. Consequently, the phenotypic consequences observed in Dhx36-deficient models may arise from the accumulation of either DNA G4s, RNA G4s, or a combination thereof. The specific contributions of DNA G4 structures to spermatogenic processes have yet to be systematically elucidated. Our study addresses this critical gap by systematically characterizing the dynamic changes and genomic distribution patterns of DNA G4 structures during spermatogenesis.

The sample size limits our ability to detect subtle effects in some tissues. Post hoc power analysis confirmed adequate statistical power (1.0) for our primary finding that testicular cells exhibit significantly greater sensitivity to G4 dynamics compared with other tissues, supporting the robustness of this conclusion (Supplementary Table 3). However, low statistical power was observed for liver (~ 0.37) and kidney (~ 0.35), indicating that our study may have been underpowered to detect modest changes in these tissues. Despite this limitation, the dramatically increased sensitivity of testicular cells to G4 modulation was consistently observed and forms the biological basis for our focused investigation of this tissue type.

We demonstrate that DNA G4s are enriched in testicular cells, with elevated levels specifically observed during critical phases of spermatogenesis: spermatogonial differentiation, meiotic homologous recombination in spermatocytes, and the histone-to-protamine transition in spermatids. Notably, prolonged administration of the G4-stabilizing agent PDS in mice recapitulated a Sertoli-cell-only syndrome-like phenotype, with minimal effects on other organs, underscoring the essential roles of G4s during these sensitive developmental windows.

However, a striking discrepancy emerged when staining mouse testes with antibodies 1H6 and BG4: BG4 consistently detected high G4 levels across various stages of prophase spermatocytes, whereas 1H6 signal peaked specifically at the leptotene stage, followed by a decline. Also intriguingly, during the round-to-elongated spermatid transition, 1H6 staining became more prominent again than BG4. This divergence in staining patterns likely arises from differences in antibody specificity—BG4 recognizes both DNA and RNA G4s, while 1H6 is specific to DNA G4s [35, 36].

The leptotene stage-specific peak in 1H6 signal aligns with the occurrence of extensive DSBs during homologous recombination, suggesting a role for DNA G4s in regulating DSB processing at this critical meiotic phase. Despite significant transcriptional repression during leptotene, as spermatocytes progress to pachytene/diplotene stages, widespread transcription resumes, leading to the generation of abundant RNA G4s [57], which correlates with the persistent BG4 signal. During the round-to-elongated spermatid transition, however, transcription gradually ceases, resulting in reduced RNA G4 abundance. Concurrently, dynamic changes in DNA topology, mediated by topoisomerase II, are required for chromatin remodeling [37, 38]. Notably, topoisomerase II has been shown to interact with G4 structures to resolve topological stress and PDS treatment can trap topoisomerase II [39, 40]. The elevated 1H6 signal during this transition thus supports a critical role for DNA G4s in facilitating topoisomerase II-mediated chromatin reorganization during the spermiogenesis. Our staining results reveal signal differences between BG4 and 1H6, suggesting a unique distribution of RNA G4s during spermatogenesis. Future studies could utilize BG4 antibody-mediated RNA-related sequencing approaches to further investigate the roles of RNA G4s in this process.

The observed differences in antibody specificity between 1H6 and BG4 prompted us to select 1H6 for subsequent CUT&Tag assays. This choice was critical to avoid potential confounding signals, as BG4 may cross-react with chromatin-associated RNA G4s, which could be misinterpreted as DNA G4 signals. Genome-wide profiling of DNA G4 distributions across different spermatogenic cells revealed that 40–50% of G4 signals overlapped with those reported in other cell types, indicating the conserved nature of G4 landscapes across diverse samples. Enrichment of G4s at promoters correlated with active histone marks and POLR2A occupancy, supporting their role in transcriptional regulation. Gene Ontology analysis of G4-associated loci further implicated G4s in regulating cell-specific processes such as spermatogenic differentiation and meiotic initiation.

The apparent discrepancy between immunofluorescence (IF) enrichment of G4 DNA (Fig. 2E) and the limited CUT&Tag peaks in late-stage elongating spermatids (Fig. 3A) can be attributed to the dramatic chromatin compaction during spermiogenesis. CUT&Tag efficiency is reduced in highly condensed regions, where transposase access is restricted, leading to underdetection of G4 sites [58–60]. Conversely, IF captures stable G4 structures irrespective of chromatin accessibility, providing complementary insights into their persistence.

The transient resurgence of G4 fluorescent signals in leptotene spermatocytes, detected by 1H6 antibody, coincided with SPO11-mediated DSB formation, suggesting that G4 structures may have a relationship with recombination hotspots. Partial colocalization of DNA G4 signals with SPO11-oligos was observed in spermatogonia, spermatocytes, and spermatids. The colocalizations in spermatogonia indicate that G4 signals preceded SPO11-oligo signals, implying that G4s may act as critical determinants of hotspot selection. While PRDM9 defines recombination hotspots at fine-scale resolution, megabase-scale recombination patterns are established independently of PRDM9 [46]. Meiotic recombination is spatiotemporally coupled with DNA replication at megabase scales [61], and genome-wide analyses have revealed that most metazoan replication origins contain G4-forming sequences [15]. Thus, SPO11-colocalized G4 signals may originate from replication initiation sites. These colocalized G4s persist even in elongated spermatids where DNA is tightly compacted by protamines. This persistence suggests that recombination sites colocalizing with G4s are not fully compressed. The underlying mechanisms governing this selective retention of chromatin accessibility warrant further investigation.

Treatment of mice with PDS at the onset of meiosis effectively blocked meiotic progression, as evidenced by aberrant DSB repair during the pachytene/diplotene stages. This observation prompted us to identify G4-interacting proteins involved in DSB repair. Current proximity labeling techniques applicable in live cells primarily rely on small-molecule crosslinkers capable of penetrating cell membranes or proximity labeling systems based on the G4-binding domain of DHX36 [26, 27, 29]. However, the inherent G4-stabilizing activity of small molecules and the propensity of DHX36 to bind RNA G4s posed significant limitations, necessitating the development of a novel G4 protein capture system.

Inspired by Silvia Galli’s work [30], we cloned the sequence of the G4-targeting nanobody SG4 and successfully engineered a proximity labeling system, nanoG4BPL. Employing this system, we captured a series of G4-interacting proteins. Subsequent functional validation using a CRISPR library and a DR-GFP homologous recombination reporter system revealed that NBS1 is a critical G4-interacting protein involved in homologous recombination. The colocalization of G4s, SPO11, and NBS1 suggests that G4s may serve as signaling landmarks for initiating meiotic recombination. Recent studies have demonstrated that Spo11 generates gaps ranging from 34 to several hundred base pairs in length via proximal double cleavage, with these gap signals overlapping and correlating to topoisomerase II binding sites. Furthermore, GC-rich DNA is significantly overrepresented in the double DSB fragments derived from these gaps [5]. Building upon this study, we hypothesize that G4 structures may play a critical role in the process of SPO11-mediated proximal double cleavage. Additionally, given that non-homologous gap repair leads to deletions and ectopically re-integrated double DSB fragments result in insertions, the formation of double DSBs is a critical mechanism underlying evolutionary diversity and pathogenic germline aberrations [5]. Therefore, DNA G4 sites overlapping with recombination hotspots may represent loci of aberrant genomic alterations, potentially linked to germline mutations.

The phase separation properties of NBS1–G4 interactions offer a novel paradigm for how G4s organize DNA repair machinery in meiotic cells, potentially explaining the spatiotemporal clustering of repair factors at G4-enriched genomic loci. Studies have shown that SPO11-generated breaks typically occur along chromosomal axes. Insights from yeast models suggest that this localization arises because the RMM complex, upon binding to DNA, forms phase-separated granules that subsequently recruit the SPO11 complex to execute cleavage, with the RMM complex being initially recruited to chromosomal axes [62]. Our findings reveal that NBS1 undergoes pronounced phase separation in the presence of DNA G4 structures, indicating that both SPO11-mediated DSB formation and MRN complex-mediated resection require a phase-separated environment. However, the relationship between phase-separated granules formed by the SPO11 complex and those formed by NBS1, as well as the precise mechanisms by which G4–NBS1 interactions influence the recruitment and activation of MRE11/RAD50 and ATM, remain to be elucidated. Future studies will focus on mapping the interaction interfaces using techniques such as electron microscopy, followed by site-directed mutagenesis in mice to dissect the functional consequences of these interactions. These investigations represent the next frontier in our research. Future work employing mouse models with specific G4-binding defects in NBS1 will be essential to dissect how this interaction regulates resection efficiency and the recruitment of repair machinery at G4-rich hotspots versus other genomic regions.

Clinically, our observation of elevated G4 levels in patients with idiopathic NOA and AZFc microdeletion positions G4 dysregulation as a potential mechanistic link between chromatin structure abnormalities and male infertility. We further discovered that the primary increase in G4 occurs in tetraploid spermatocytes. The correlation between G4 accumulation in spermatocytes suggests that G4 homeostasis is particularly critical during the prophase of meiosis, a period highly sensitive to genomic perturbations. Environmental toxins including heavy metals, smoking, heat exposure, and chemical pollutants generate reactive oxygen species (ROS) that preferentially oxidize guanine residues to 8-oxoG [56]. This oxidative lesion triggers the base excision repair (BER) pathway, where 8-oxoguanine DNA glycosylase 1 (OGG1) removes the damaged base, creating an abasic (AP) site. Subsequently, APE1 endonuclease processes the AP site, generating single-strand breaks that facilitate the unmasking and stabilization of latent G4-forming sequences. The acetylation of APE1 further enhances its binding affinity to G4 structures, promoting their transcriptional regulatory function [54, 55]. However, the clinical findings are based on a relatively small cohort, necessitating validation in larger populations and diverse ethnic groups. Future studies with expanded cohorts should address whether different etiologies of male infertility exhibit distinct patterns of G4 accumulation, and whether G4 levels correlate with clinical severity. Given our findings linking oxidative stress to G4 stabilization, prospective studies examining G4 dynamics before and after therapeutic interventions would provide insights into the potential for G4-targeted strategies.

Collectively, our findings highlight the critical roles of G4 structures as dynamic genomic elements, with profound implications for spermatogenesis, and further open new avenues for investigating clinical idiopathic male infertility.

Conclusions

This research establishes G-quadruplex (G4) structures as critical regulators of mammalian spermatogenesis, offering new insights into the molecular basis of male infertility. We reveal that G4s are highly enriched and dynamically distributed across spermatogonial differentiation, meiosis, and spermiogenesis. These structures coordinate gene expression by interacting with active epigenetic marks and contribute to genome organization through associations with CTCF and meiotic recombination machinery. Using a novel nanobody-based proximity labeling approach (nanoG4BPL), we identified NBS1 as a key G4-binding protein, which promotes homologous recombination at G4-rich DNA break sites through phase separation, ensuring meiotic accuracy. Perturbation of G4 stability with pyridostatin disrupted DNA repair and impaired spermatogenesis, confirming their indispensable role. Furthermore, elevated G4 levels in spermatocytes from patients with idiopathic nonobstructive azoospermia suggest a link between G4 dysregulation and infertility, potentially driven by environmental factors such as oxidative stress. These findings elucidate the roles of G4s in genomic regulation during spermatogenesis and provide a molecular framework for understanding and addressing male reproductive disorders.

Abbreviations

G4

G-quadruplex

NOA

Non-Obstructive Azoospermia

DSB

Double-strand break

HR

Homologous recombination

NanoG4BPL

Nanobody-based G4 binding protein proximity labeling system

PDS

Pyridostatin

SPG

Spermatogonia

SPC

Spermatocytes

RS

Round spermatids

LS

Elongated spermatids

PreL

Preleptotene spermatocytes

L

Leptotene spermatocytes

Z

Zygotene spermatocytes

P

Pachytene spermatocytes

D

Diplotene spermatocytes

STD

Spermatids

mESC

Mouse embryonic stem cells

CBD

Chitin-binding domain

AZF_Del

Y-chromosome azoospermia factor microdeletion

Author contributions

L.S., M.Y., S.H., W.W., Y.F., and Q.W. conceived and designed the study. L.S., M.Y., and S.H. performed the majority of the experiments, including G4-specific staining, CUT&Tag sequencing, and proximity labeling with the nanoG4BPL system. W.R., L.C., and Z.T. conducted the in vivo mouse experiments and flow cytometry analyses. Z.C., G.Y., and S.Z. contributed to the immunohistochemical and immunofluorescence staining of testicular tissues. G.H. and D.M. performed the computational analyses, including genome-wide G4 mapping and integration with epigenetic and SPO11-oligo datasets. L.Y. and L.Z. developed and optimized the nanobody-based proximity labeling system (nanoG4BPL) and conducted mass spectrometry analyses. W.M. contributed to the clinical sample collection and analysis of G4 levels in infertility patients. L.S., M.Y., S.H., W.W., Y.F., and Q.W. interpreted the data and wrote the manuscript. W.W., Y.F., and Q.W. supervised the study and secured funding. All authors reviewed and approved the final manuscript.

Funding

This work was supported by the National Natural Science Foundation of China (nos. 82220108004, 82173204, 82203633 and 82202933), the Innovation Capability Support Program of Shaanxi (2023-CX-TD-72), the Innovation Capability Enhancement Program of Shaanxi (2024TD-08), the Natural Science Basic Research Program of Shaanxi (2022JZ-62; 2024JC-YBQN-0783), the Key Research and Development Program in Shaanxi (2023-YBSF-251), the Joint Innovation Fund of Innovation Research Institute program in Xijing Hospital (LHJJ24JH19), and Xijing Hospital Healthcare Professionals Training and Development Boost Program (XJZT25CX25).

Data availability

The CUT&Tag sequencing data generated in this study have been deposited in the NCBI Sequence Read Archive (SRA) under the BioProject accession number PRJNA1360028. Other datasets generated and/or analyzed during the current study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

The human tissue samples utilized in this study were approved by the Medical Ethics Committee of the First Affiliated Hospital of the Air Force Medical University (approval no. KY20253837-1; 4 March 2025). Informed consent was obtained from all participants in accordance with the principles of the Declaration of Helsinki. For animal experiments, approval was granted by the Laboratory Animals Welfare Ethics Committee of Fourth Military Medical University, which acts in accordance with the International Council for Laboratory Animal Science (ICLAS) guidelines (approval no. 20220801; 1 August 2022). Animal experiments were carried out in strict compliance with ICLAS ethical principles.

Consent for publication

Written informed consent for publication was obtained from all participants whose individual data (including details, images, or videos) are included in this manuscript.

Competing interests

The authors declare no competing interests.