Main

Despite their critical role in encoding genetic information, chromosomes frequently mis-segregate during human meiosis, producing abnormalities in chromosome number—a phenomenon termed aneuploidy. Aneuploidy is the leading cause of human pregnancy loss, as well as the cause of genetic conditions such as Klinefelter, Turner and Down syndromes1,2. It is estimated that only approximately half of human conceptions survive to birth, primarily because of the abundance of aneuploidies that are inviable in early gestation5,6.

Work in humans and model organisms has established that one risk factor for aneuploidy involves variation in the number and location of meiotic crossover recombination events, especially in the female germline3,4. Notably, female meiosis initiates in fetal development, when replicated homologous chromosomes (homologues) pair and establish crossovers, which, together with cohesion between sister chromatids, hold homologues together in a ‘bivalent’ configuration. Homologues segregate (meiosis I) upon ovulation after the onset of puberty, whereas sister chromatids segregate (meiosis II) after fertilization. The physical linkages formed by meiotic crossovers help stabilize paired chromosomes during this prolonged period of female meiotic arrest7. Cohesin complexes, loaded in developing fetal oocytes, link sister chromatids and are crucial for chromosome synapsis and crossover formation8,9. Failure to form bivalents due to lack of crossovers10 or their suboptimal placement11, as well as age-related cohesin deterioration12, can lead to premature separation of sister chromatids and the related phenomenon of reverse segregation, which together represent the predominant mechanisms of maternal meiotic aneuploidy13.

Although producing sex-specific recombination maps and revealing associations with crossover phenotypes at meiosis-related genes, the largest studies of crossovers in living human families lacked aneuploid participants and only speculated about such relationships14,15. Much of current knowledge about the connection between human recombination and aneuploidy, as well as their genetic bases, thus comes from smaller samples of people living with survivable aneuploidies, limiting statistical power. By contrast, recent advances in single-cell sequencing have enabled simultaneous discovery of crossovers and aneuploidies in sperm and eggs, but are typically relegated to small numbers of gametes (in the case of oocytes) or small numbers of donors, hindering understanding of variability and potential shared genetic architecture of these phenotypes16,17,18.

Clinical genetic data from pre-implantation genetic testing (PGT) of in vitro fertilized (IVF) embryos help overcome these limitations and offer an ideal resource for characterizing aneuploidy and mapping meiotic crossovers at scale. Here we used single nucleotide polymorphism (SNP) array-based PGT data from 139,416 blastocyst-stage embryo biopsies and 22,850 sets of biological parents to (1) map recombination and aneuploidy, (2) test their relationship quantitatively and (3) discover genetic factors that modulate their incidence and features. Our analysis revealed an overlapping genetic basis of female recombination and aneuploidy formation involving common variation in key meiotic machinery. Together, our work offers a more complete view of the sources of variation in the fundamental molecular processes that generate genetic diversity while impacting human fertility.

Meiotic aneuploidy is common in embryos

Seeking insight into meiotic crossover recombination and the origins of aneuploidies, we performed retrospective analysis of data from PGT. Specifically, these data comprised SNP microarray genotyping of bulk (approximately six cells) trophectoderm biopsies from 156,828 blastocyst-stage embryos (5 days post-fertilization), as well as DNA isolated from buccal swabs or blood from both biological parents (24,788 patient–partner pairs) (Fig. 1a and Supplementary Figs. 1 and 2; Supplementary Methods). We developed a hidden Markov model (HMM), called karyoHMM, to trace the transmission of parental haplotypes to sampled embryos and thereby identify aneuploidies and crossover recombination events. Specifically, we modelled transitions between the haplotypes transmitted from the same parent as crossovers and inferred the chromosome copy number that best explained the embryo data (Fig. 1b and Supplementary Fig. 3; Supplementary Methods).

Fig. 1: Data from PGT of IVF embryos offer insight into crossover recombination and aneuploidy.

Colours indicate maternal (purple) versus paternal (blue) data features. a, Data comprise SNP microarray genotyping of trophectoderm biopsies from sibling embryos, as well as DNA from parents. b, Tracing transmission of parental haplotypes from parents to embryos reveals evidence of crossovers, as well as aneuploidies. c, Aneuploidies primarily involve gain or loss of maternal homologues and are enriched on particular chromosomes. Complex aneuploidies (more than five affected chromosomes) and genome-wide ploidy abnormalities (for example, triploidy) are excluded (Extended Data Fig. 1). d, Aneuploidies affecting maternal homologues increase with maternal age, whereas aneuploidies affecting paternal homologues exhibit no significant relationship with paternal age. e, Maternal crossovers exceed paternal crossovers. Embryos with crossover counts outside of 3 s.d. from the sex-specific mean are excluded. f, Crossover counts differ between disomic chromosomes of euploid (n = 46,856) and aneuploid (n = 34,542) embryos containing at least a single maternal crossover (two-sided Poisson GLMM), but the proportion of crossovers occurring within hotspots does not (two-sided Gamma GLMM). Error bars indicate 95% confidence intervals. Illustration in a adapted from NIH BioArt Source (https://bioart.niaid.nih.gov/bioart/209) under a Public Domain licence CC0 1.0.

Applying this method to a dataset where low-quality samples were removed (139,416 remaining embryos; Supplementary Methods), we identified 41,480 (29.8%) embryos with at least one aneuploid chromosome (92,485 aneuploid chromosomes; Extended Data Fig. 1). Trisomies exceeded monosomies (57,974 trisomies, 34,511 monosomies; ratio, 0.626; 95% confidence intervals, 0.624, 0.630; two-sided binomial test, P < 1 × 10−100), indicative of selection before blastocyst formation6. However, trisomies and monosomies of all individual autosomes and sex chromosomes were detected within the sample (Fig. 1c). Aneuploidies largely involved gain or loss of maternal versus paternal homologues (84,044 maternal:8,441 paternal; ratio, 0.909; 95% confidence intervals, 0.907, 0.911; two-sided binomial test, P < 1 × 10−100) and were strongly enriched for chromosomes 15, 16, 21 and 22, replicating previous literature19.

We also replicated the association between maternal age and the incidence of aneuploidies affecting maternal homologues (binomial generalized linear mixed model (GLMM), (\hat{\beta }) = 0.235, s.e. = 2.19 × 10−3, P < 1 × 10−100; Supplementary Table 1)13. The data were well fit by a model with a quadratic term for maternal age (Fig. 1d, Supplementary Fig. 4 and Supplementary Table 1; Supplementary Methods). Positive associations with maternal age were also significant when stratifying the phenotype to maternal meiotic aneuploidy of individual chromosomes (Supplementary Table 1). Further supporting selection against meiotic aneuploidies, per patient rates of maternal meiotic aneuploidy were inversely associated with per-cycle embryo counts, even when controlling for maternal age (binomial GLMM, (\hat{\beta }) = −0.030, s.e. = 6.88 × 10−3, P = 1.29 × 10−5). Despite the statistical power afforded by the large sample size, we observed no significant association between paternal age and aneuploidies affecting paternal homologues (binomial GLMM, (\hat{\beta }) = −1.06 × 10−3, s.e. = 0.013, P = 0.936; Fig. 1d and Supplementary Table 1), consistent with previous findings19. The absence of paternal age association also held for the sex chromosomes, where paternal meiotic aneuploidies were relatively more common (binomial GLMM, (\hat{\beta }) = 2.14 × 10−3, s.e. = 0.020, P = 0.914; Supplementary Table 1).

Aneuploid embryos possess fewer crossovers

Previous studies have shown that abnormal number or placement of crossovers confers risk for meiotic aneuploidy1,4. These include studies of survivable trisomies20,21, gametes2,16,17 and embryos16,22, which broadly demonstrated that aneuploid chromosomes are depleted of crossovers compared with corresponding disomic chromosomes.

Across 46,861 euploid embryos (and requiring at least three sibling embryos; Supplementary Methods), we identified 2,310,257 maternal- and 1,499,155 paternal-origin autosomal crossovers (3,809,412 total) at a median resolution of 99.43 kilobase pairs (kbp) (Fig. 1e). The mean counts of sex-specific crossovers per meiosis (49.30 maternal, 31.99 paternal), as well as their genomic locations (Spearman correlation (r) at 100-kbp resolution: 0.96 maternal, 0.98 paternal), were consistent with previous pedigree-based studies of living human cohorts14,15. We also observed substantial proportions of chromosomes that lack detected crossovers from a given parent (maternal = 1.67–35.56%, paternal = 7.83–51.77%), particularly among short chromosomes such as chromosomes 21 and 22 where aneuploidies are common (Extended Data Fig. 2). Acknowledging the limited resolution of the genotyping array at chromosome ends, these estimates conform with observations from living human pedigrees14.

Previous literature offers conflicting evidence about the relationship between counts of meiotic crossovers and maternal age, with some studies reporting a positive association14,15,23 and others reporting a negative association24. As those studies focused largely on living families, positive associations were interpreted typically as evidence of selection against aneuploid embryos, which possess fewer crossovers on average and increase in frequency with maternal age. Within our sample, we observed no significant association between maternal age and number of maternal crossovers (Poisson GLMM, (\hat{\beta }) = −2.62 × 10−5, s.e. = 1.68 × 10−3, P = 0.988). This observation held even when restricting analysis to euploid embryos (Poisson GLMM, (\hat{\beta }) = 5.12 × 10−4, s.e. = 1.43 × 10−3, P = 0.721), offering a point of evidence against the hypothesis that embryonic aneuploidy explains previously reported age associations with crossovers.

We used these crossover data to perform genome-wide association studies (GWAS) across four phenotypes: mean count of autosomal crossovers across euploid embryos (crossover count); fraction of crossovers within recombination hotspots based on published genetic maps (hotspot occupancy); mean timing of DNA replication at crossover sites (replication timing); and mean guanine–cytosine content ±500 bp around crossover sites (GC content; Supplementary Methods). We identified 15 unique association signals achieving genome-wide significance (P < 5 × 10−8), all of which replicated previous findings14,25 (Supplementary Table 2 and Extended Data Figs. 3–6), including a haplotype spanning RNF212 with opposing directions of association with maternal versus paternal recombination rates (lead SNP rs3816474; maternal (\hat{\beta }) = −0.089 ± 0.013 s.e., P = 1.84 × 10−11; paternal (\hat{\beta }) = 0.186 ± 0.013 s.e., P = 1.76 × 10−47; Extended Data Fig. 3). Complementing these GWAS, we performed transcriptome-wide association studies (TWAS) to associate predicted gene expression across several tissues26 with recombination phenotypes, identifying 35 unique genes significantly associated with at least one recombination phenotype (P < 3.0 × 10−6; Supplementary Table 3; Supplementary Methods). Prominent hits included the synaptonemal complex component C14orf39 (also known as SIX6OS1)27 and crossover-regulating ubiquitin ligase CCNB1IP1 (also known as HEI10)28, implying that previously reported genetic associations at these loci could be driven by non-coding regulatory mechanisms14.

To examine the relationship between crossovers and aneuploidies, we contrasted patterns of crossovers between aneuploid and euploid embryos. One technical limitation for direct detection of crossovers using genetic data from trisomic chromosomes is that crossovers can be missed when both reciprocal products of a single crossover event are transmitted to the embryo16. To overcome this concern, we instead contrasted counts of crossovers on disomic chromosomes of aneuploid embryos (with aneuploidy affecting a different chromosome) to corresponding disomic chromosomes of euploid embryos. This comparison relies on the previous observation that crossover counts positively covary across chromosomes within meiocytes29—a phenomenon that we replicated for euploid embryos within our dataset (intraclass correlation coefficient = 0.176; 95% confidence intervals, 0.11, 0.3; P < 1 × 10−100 maternal; intraclass correlation coefficient = 0.088; 95% confidence intervals, 0.05, 0.16; P < 1 × 10−100 paternal; Extended Data Fig. 7; Supplementary Methods). As input to our test, we identified 1,505,107 maternal- and 1,007,176 paternal-origin crossovers on disomic chromosomes across 34,542 embryos with at least one aneuploid chromosome (and requiring at least three sibling embryos). Using a Poisson GLMM (Supplementary Methods), we found that the number of crossovers was significantly lower on the disomic chromosomes of aneuploid embryos relative to euploid embryos ((\hat{\beta }) = 0.105 difference in marginal means ± 6.923 × 10−5 s.e.; P = 4.64 × 10−150; Fig. 1f). These results are consistent with the understanding that reduction in crossovers—and absence of crossovers, in particular10—confers risk for meiotic aneuploidy.

SMC1B variants associate with aneuploidy

Previous studies have suggested that the incidence of female meiotic aneuploidy may be individual-specific, even after accounting for maternal age30. To test this hypothesis, we fit a quasi-binomial generalized linear model (GLM) to the per patient counts of embryos affected versus unaffected with maternal meiotic-origin aneuploidy, including maternal age as a quadratic covariate (Supplementary Methods). Compared with a simulated binomial null distribution, the observed incidence of meiotic aneuploidy was significantly overdispersed across female patients, controlling for maternal age (dispersion parameter (φ) = 1.15, P < 0.01; Supplementary Fig. 5). Overdispersion was also apparent when stratifying analysis to maternal meiotic aneuploidies affecting individual chromosomes (Supplementary Table 4). These observations of overdispersion suggest a role of genetic and environmental factors beyond age in observed variation in maternal meiotic aneuploidy.

To investigate the genetic component, we scanned for variation in maternal genomes associated with the incidence of maternal meiotic aneuploidy. We implemented these association tests using a binomial GLMM, controlling for covariates including maternal age (Supplementary Methods). We first tested for cis-genetic effects on aneuploidy risk by associating incidence of aneuploidy affecting each individual chromosome with maternal genotypes restricted to that chromosome, but we identified no associations achieving genome-wide significance (P < 5 × 10−8). Proceeding to a genome-wide analysis considering maternal meiotic aneuploidies affecting any chromosome, we discovered two genome-wide significant associations (Fig. 2a and Supplementary Fig. 6). The first hit (lead SNP rs9351349, (\hat{\beta }) = 0.078, s.e. = 0.014, P = 2.93 × 10−8) lies within an intergenic region of chromosome (Chr.) 6 but did not replicate in a held-out test set comprising 15% of female patients ((\hat{\beta }) = 0.021, s.e. = 0.033, P = 0.529). The second hit (lead SNP rs6006737, (\hat{\beta }) = 0.066, s.e. = 0.012, P = 2.21 × 10−8) lies on Chr. 22 and replicated in the held-out test set ((\hat{\beta }) = 0.059, s.e. = 0.028, P = 0.033). The minor (C) allele of rs6006737 within our sample is globally common, segregating at high frequencies (gnomAD allele frequency (AF) = 0.78) in African populations but at lower frequencies in European (gnomAD AF = 0.35) and other non-African populations31. The effect is additive, whereby for a 40-year-old patient, each copy of the risk allele confers an estimated 1.65% additional average risk of aneuploidy (Fig. 2b). We also detected evidence of a small but statistically significant interaction between maternal age and genotype (likelihood ratio test, χ2(1) = 4.24, P = 0.040), indicating that the effect of genotype increases with increasing maternal age ((\hat{\beta }) = 0.026, s.e. = 0.013, P = 0.045). Notably, the size and direction of the main effect of genotype is relatively consistent for aneuploidies of all individual autosomes (Extended Data Fig. 8), suggesting general, genome-wide impacts on meiotic fidelity.

Fig. 2: Variants defining a haplotype spanning SMC1B are associated with incidence of maternal meiotic aneuploidy.

a, Results of GWAS tests of maternal meiotic aneuploidy and maternal genotype (two-sided binomial GLMM). The dotted line indicates the threshold for genome-wide significance (P = 5 × 10−8). b, Fitted relationship between maternal age and incidence of aneuploidy, stratified by maternal genotype at aneuploidy-associated lead SNP rs6006737. c. Regional Manhattan plot depicting the associated locus on Chr. 22, with points coloured based on pairwise linkage disequilibrium with the lead SNP rs6006737 (diamond). Mbp, megabase pairs.

The associated haplotype spans approximately 120 kbp, encompassing four genes: UPK3A, FAM118A, RIBC2 and SMC1B (Fig. 2c). SMC1B encodes a component of the ring-shaped cohesin complex (Fig. 3a), with integral roles in sister chromatid cohesion and homologous recombination during meiosis32,33. Smc1b-deficient mice of both sexes are sterile, and female mice exhibit meiotic abnormalities including reduction in crossovers, incomplete chromosome synapsis, age-related premature loss of sister chromatid cohesion and chromosome mis-segregation32,33. Previous work in humans demonstrated associations between a less common (gnomAD global AF = 0.06) SMC1B missense variant (rs61735519; r2 with GWAS lead SNP rs6006737 = 0.089, D′ = 0.943) and recombination phenotypes14. Although imputed with moderate accuracy (dosage r2 = 0.80), this missense variant exhibits only modest association with aneuploidy within our sample ((\hat{\beta }) = 0.112, s.e. = 0.040, P = 4.80 × 10−3). Meanwhile, the more common aneuploidy-associated haplotype tagged by GWAS lead variant rs6006737 lacks amino acid altering variation (r2 < 0.1 for all SMC1B nonsynonymous variants), motivating us to explore potential regulatory mechanisms driving the observed phenotype.

Fig. 3: The aneuploidy risk haplotype is associated with lower expression of SMC1B, driven by two independent causal signals.

a, Schematic of the meiotic cohesin complex. b, Each copy of the aneuploidy risk allele is associated with reduced expression of SMC1B in human lymphoblastoid cell lines (LCLs; n = 731) from diverse populations. Bars represent the first and third quartiles of the data, white points represent the second quartile (median) and whiskers are bound to 1.5× the interquartile range. c, Pairwise linkage disequilibrium between a set of SNPs including GWAS lead SNP rs6006737 and variants defining fine-mapped eQTL credible sets for SMC1B. d, Fine-mapped eQTL rs2272804 (credible set 1) lies within a putative promoter sequence within open chromatin, while variants defining a second credible set are distributed throughout the upstream region of SMC1B.

Associated haplotype is an SMC1B expression quantitative trait locus

Querying the GWAS lead variant (rs6006737) in data from the Genotype Tissue Expression (GTEx) Project26, we observed that the aneuploidy risk allele is associated significantly with reduced expression of SMC1B across diverse tissues. Although invaluable, GTEx largely includes participants of European ancestries, limiting resolution for fine-mapping of causal expression-altering variants. To address this limitation, we also queried the GWAS lead variant in MAGE, which includes RNA sequencing data from lymphoblastoid cell lines from 731 people from 26 globally diverse populations34. Consistent with GTEx, rs6006737 is a strong expression quantitative trait locus (eQTL) of SMC1B in MAGE ((\hat{\beta }) = −0.429, s.e. = 0.048, P = 4.68 × 10−18; Fig. 3b). Fine-mapping within MAGE decomposes the eQTL signals for SMC1B into two credible sets containing candidate causal variants (coverage = 0.95) (Fig. 3c,d). Whereas one credible set includes nine variants distributed throughout the upstream region of SMC1B, the other is defined by a single SNP (rs2272804; posterior inclusion probability > 0.99), 144 bp upstream of the SMC1B transcription start site.

The regulatory potential and accessibility of the putative promoter CpG island sequence within which rs2272804 resides is supported by published epigenomic and ATAC-seq (assay for transposase-accessible chromatin using sequencing) data from human ovaries35,36 (Fig. 3d). We further noted that the SNP lies within a predicted binding motif of ATF1—a transcription factor expressed in female germ cells37 and inferred previously to regulate paralogue SMC1A based on chromatin immunoprecipitation sequencing data38. Binding of ATF1 to the SNP-encompassing locus is also supported by high-confidence chromatin immunoprecipitation sequencing peaks in induced pluripotent stem cells (WTC11) assayed by the ENCODE Project38. By performing an electrophoretic mobility shift assay, we found that a DNA construct containing the alternative allele of rs2272804 had more than threefold lower binding affinity (dissociation constant, KD) for purified human ATF1 in vitro than a construct containing the reference allele (Student’s t-test, mean reference KD = 56.62 nM ± 4.65 s.d., mean variant KD = 173.39 nM ± 15.24 s.d., P = 2.60 × 10−4), consistent with the observed eQTL effect (Extended Data Fig. 9). Taken together, these results suggest a potential non-coding regulatory mechanism underlying the observed genetic association with maternal meiotic aneuploidy.

TWAS reveals new links to meiosis genes

Motivated by our observations at SMC1B, we next sought to examine whether other cis-regulatory effects on expression could influence aneuploidy risk. We therefore used TWAS to test whether predicted gene expression across tissues is associated with incidence of aneuploidy (Supplementary Methods). Across 16,685 protein-coding genes, we identified two hits achieving transcriptome-wide significance (P < 3 × 10−6; Extended Data Fig. 10). Although led by adjacent gene RIBC2 (P = 2.19 × 10−7), the peak on Chr. 22 includes SMC1B (P = 7.63 × 10−6), replicating our findings from GWAS and downstream functional dissection. We hypothesize that RIBC2 represents a secondary, noncausal association, whereby the same haplotype (and potentially the same causal variant39) co-regulates expression of both genes, driving their correlation (Supplementary Fig. 7). The second peak lies on Chr. 14 and is led by C14orf39 (P = 1.65 × 10−7), which encodes a component of the synaptonemal complex, which mediates synapsis, recombination and segregation of homologous chromosomes during meiosis27. Previous studies have linked rare C14orf39 variants to human infertility40,41 and demonstrated associations between common C14orf39 variants and recombination phenotypes14,25. Our results connect these findings and show that both rare and common variation influencing female fertility phenotypes can converge on the same meiosis-related genes. Although not achieving transcriptome-wide significance, a third peak, on Chr. 12, includes NCAPD2 (P = 2.16 × 10−5), which encodes a regulatory subunit of the condensin I complex, involved in chromosome condensation during both meiotic and mitotic prophase42. Together, our findings highlight the role of common non-coding cis-regulatory variation influencing expression of meiosis-related genes in modulating risk of maternal meiotic aneuploidy (Extended Data Fig. 10).

Pleiotropic effects on fertility traits

Given the relationship between crossovers and aneuploidies, we next aimed to contextualize our association findings and examine the potential shared genetic basis with other fertility-related traits. To this end, we identified the lead variant from each genome-wide significant peak in female recombination and aneuploidy GWAS and queried their associations with all recombination and aneuploidy phenotypes, as well as published GWAS of female reproductive ageing and infertility traits (that is, phenome-wide association). Our analysis revealed that the risk allele of the aneuploidy-associated lead SNP rs6006737 is also associated with lower rates of female recombination within our data ((\hat{\beta }) = −0.033, s.e. = 0.011, P = 0.002), consistent with the known role of SMC1B variation in this phenotype32. Extending to published GWAS data43,44, we observed that the aneuploidy risk allele is additionally associated with greater age at menarche ((\hat{\beta }) = 0.021, s.e. = 0.003, P = 3.82 × 10−12) and lesser age at menopause ((\hat{\beta }) = −0.047, s.e. = 0.013, P = 2.06 × 10−4) and thus a shorter female reproductive timespan (Fig. 4).

Fig. 4: Aneuploidy, recombination and female reproductive ageing traits share an overlapping genetic basis.

The lead SNP from each peak from GWAS of aneuploidy and recombination was queried for association with other fertility-related phenotypes (two-sided linear or logistic model from respective GWAS study). Darkness indicates significance of association (P value), while colour indicates direction of association. SNPs are polarized such that the aneuploidy-increasing allele is queried across all traits. Each hit is labelled based on meiosis-related candidate genes within the associated region (top) with the exception of the common 17q21.31 inversion, as well as the locus containing ACYP2 and TSPYL6, where no such candidate is apparent.

Strikingly, three of the genome-wide significant hits for female recombination rate (Supplementary Table 2) also exhibited nominal associations with aneuploidy in consistent direction. The first hit (lead SNP rs4365199; aneuploidy (\hat{\beta }) = 0.056, s.e. = 0.012, P = 5.58 × 10−6; gnomAD global AF = 0.39) comprises a 175-kbp haplotype spanning synaptonemal complex component C14orf39, consistent with our previous TWAS results. The second hit (lead SNP rs12588213; aneuploidy (\hat{\beta }) = 0.037, s.e. = 0.012, P = 1.46 × 10−3; gnomAD global AF = 0.42) comprises a 15-kbp haplotype spanning CCNB1IP1, encoding an E3 ubiquitin ligase demonstrated as essential for crossover maturation and fertility in mice28. The last hit (lead SNP rs3816474; aneuploidy (\hat{\beta }) = 0.041, s.e. = 0.014, P = 5.04 × 10−3; gnomAD global AF = 0.22) comprises a 59-kbp haplotype spanning the E3 ubiquitin ligase RNF212, encoding an essential regulator of meiotic recombination that interacts with CCNB1IP1 and helps to designate sites of crossovers versus non-crossovers45. Several of these recombination and aneuploidy-associated variants also exhibited secondary associations with ages at menarche and menopause (Fig. 4). Whereas previous studies have reported links between DNA damage response and reproductive ageing43,46,[47](https://www.nature.com/articles/s41586-025-09964-2#ref-CR47 “Stankovic, S. et al. Genetic links