Three-dimensional mapping of intact ovaries reveals the aging dynamics of the ovarian reserve Authors: Arturo D’Angelo, Daniel Franco-Barranco, Marco Musy, Juan Manuel Duran, James Sharpe, Nicholas Stroustrup, Ignacio Arganda-Carreras, Elvan Böke Published: 2026 ABSTRACT Female fertility depends on a finite pool of oocytes that depletes during aging1,2, yet the spatiotemporal dynamics of this depletion remain poorly understood. Traditional methods obscure the three-dimensional architecture of the ovary, limiting quantitative insights. Here we combine light-sheet microscopy, artificial intelligence-driven segmentation and mathematical modeling to map over 85,000 oocytes in whole ovaries across the reproductive lifespan in mouse. We find that newly activated oocytes represent a fixed fraction of the total oocyte pool despite an age-related decline in oocyte numbers. Spatial analysis revealed that oocytes are enriched along the lateral ovarian axis, and local oocyte density positively correlates with activation. We also uncover a bimodal distribution of oocyte sizes, suggesting a bottleneck during oogenesis. Finally, a differential equation-based model captures the kinetics of oocyte activation and loss. Our findings establish a quantitative framework for understanding ovarian aging and suggest that an organ-scale regulatory mechanism coordinates the age-related decline in oocyte numbers. Main Reproductive lifespan in most mammals is defined by a finite supply of immature egg cells, termed oocytes, established during fetal life and depleted with age. Oocytes reside in ovarian follicles, which consist of the oocyte and surrounding somatic cells. The human ovary at birth contains approximately 1–2 million oocytes, which decrease to ~400,000 by puberty and ~1,000 by menopause1–3. Only about 400 oocytes are ovulated over a lifetime. The dynamics of follicle activation and depletion underpin key transitions in reproductive aging, and dysregulation of these processes contributes to disorders such as primary ovarian insufficiency and premature menopause4–6. Most of our current understanding of follicle dynamics comes from two-dimensional (2D) histological analyses, which, while high-resolution, sacrifice volumetric context. As a result, how oocytes are distributed in three-dimensional (3D) space and how this spatial organization changes with age remain largely unknown. Similarly, the relationship between follicular location, growth stage and activation dynamics is still not fully elucidated. Whole-ovary imaging and automated oocyte quantification The ovarian reserve in house mice (Mus musculus) undergoes changes that parallel fluctuations observed in humans7 and exhibit age-related decline2,8. The availability of inbred mouse colonies provides a unique opportunity to map ovarian reserve numbers by age in genetically identical large populations. We therefore employed the inbred C57BL/6J (BL6) strain to investigate ovarian reserve changes during aging. A key challenge in quantifying ovarian reserve lies in the ovary’s complex, non-uniform anatomy spanning several millimeters. Historically, quantifying oocyte numbers in adult mouse ovaries has relied on manual counting from serial histological sections and subsequent normalization9. While widely used, this approach is labor-intensive, sacrifices the spatial information on the oocyte’s position within the ovarian tissue, and may introduce inaccuracies in oocyte numbers7. Recent advances in light-sheet microscopy now enable whole-organ imaging of ovaries10,11. To overcome these limitations, we performed high-resolution imaging of whole ovaries using selective plane illumination microscopy (SPIM) with tissue clearing12. We optimized both immunostaining and clearing protocols to achieve the identification of all oocytes within the ovary: we used a DDX4 antibody to label oocytes and TO-PRO-3 to counterstain nuclei in ovaries from 5-week-old mice (Fig. 1a–d and Supplementary Video 1). We selected this age to capture the maximum ovarian reserve following the first wave of oocyte activation and loss13 as it corresponds to the pubertal to early postpubertal period14. Our optimized protocol enabled identification of all oocytes in situ and discrimination of developmental stages based on size (Fig. 1d), providing a platform to investigate ovarian architecture and follicular composition in their native spatial context (Fig. 1b,c).Fig. 1Light-sheet imaging and automated analysis of oocyte size in adult mouse ovaries.a, Cartoon summarizing the dataset size and the analysis pipeline. L indicates left ovary and R indicates right ovary. b, Light-sheet (SPIM) imaging of whole-mount adult ovaries from C57BL/6J mice labeled with an antibody against DDX4, an oocyte marker. Images are shown as maximum-intensity projections across four z-sections, as indicated on the left, for improved visualization. c, Single-plane images colabeled with an anti-DDX4 antibody to mark oocytes and TO-PRO-3 dye to mark DNA. d, Enlarged regions of interest (ROIs) from c highlighting different stages of folliculogenesis. Primordial oocytes are enclosed in primordial follicles and are considered dormant as they do not grow or divide3. Once activated, follicles progress through primary, secondary and antral stages, culminating in ovulation of a fertilizable egg53. Representative images from n = 101 ovaries. e, Maximum-intensity projected SPIM images of whole-mount ovaries from 5-week-old C57BL/6J mice labeled with DDX4 and manually segmented (yellow). Representative images from n = 5 ovaries. f, The same ovary as in e, segmented using the 3D model. g, Comparison of total ovarian reserve quantification between manual segmentation (e) and model-based segmentation (f). Mouse illustration in a created in BioRender; Dangelo, A. https://biorender.com/au0pcmm (2026); eppendorf tube illustration in a created in BioRender; Dangelo, A. https://biorender.com/zn0uceq (2026).Source data Light-sheet imaging and automated analysis of oocyte size in adult mouse ovaries. a, Cartoon summarizing the dataset size and the analysis pipeline. L indicates left ovary and R indicates right ovary. b, Light-sheet (SPIM) imaging of whole-mount adult ovaries from C57BL/6J mice labeled with an antibody against DDX4, an oocyte marker. Images are shown as maximum-intensity projections across four z-sections, as indicated on the left, for improved visualization. c, Single-plane images colabeled with an anti-DDX4 antibody to mark oocytes and TO-PRO-3 dye to mark DNA. d, Enlarged regions of interest (ROIs) from c highlighting different stages of folliculogenesis. Primordial oocytes are enclosed in primordial follicles and are considered dormant as they do not grow or divide3. Once activated, follicles progress through primary, secondary and antral stages, culminating in ovulation of a fertilizable egg53. Representative images from n = 101 ovaries. e, Maximum-intensity projected SPIM images of whole-mount ovaries from 5-week-old C57BL/6J mice labeled with DDX4 and manually segmented (yellow). Representative images from n = 5 ovaries. f, The same ovary as in e, segmented using the 3D model. g, Comparison of total ovarian reserve quantification between manual segmentation (e) and model-based segmentation (f). Mouse illustration in a created in BioRender; Dangelo, A. https://biorender.com/au0pcmm (2026); eppendorf tube illustration in a created in BioRender; Dangelo, A. https://biorender.com/zn0uceq (2026). Source data We manually segmented oocytes at different stages from 3D-reconstructed SPIM image volumes, generating a training dataset for a deep learning model within the BiaPy framework15. This instance segmentation model automatically identifies and delineates individual oocytes in cleared ovarian tissue (Extended Data Fig. 1a). To assess its performance, we compared the AI-generated segmentations with expert annotations, both before and after minimal correction (Fig. 1e,f). The model achieved a 95.3% agreement with manual segmentation, increasing to 96.7% following correction, highlighting its robustness and accuracy (Extended Data Fig. 1b,c). This automated approach enabled us to reconstruct the full oocyte size distribution in adult mouse ovaries (Fig. 1g and Supplementary Video 2) and, notably, proved effective for quantifying oocytes in cleared human ovarian cortex samples as well (Extended Data Fig. 2a–c and Supplementary Video 3), highlighting the pipeline’s cross-species applicability. Together, these results demonstrate a robust, scalable platform to enable spatially resolved, high-throughput quantification of the ovarian reserve in the intact ovarian tissue. Total oocyte quantification reveals high heterogeneity between isogenic animals Ovarian reserve varies substantially among humans16, raising the possibility that oocyte numbers may also vary widely between animals. To directly address this, we analyzed eight 5-week-old BL6 female mice and quantified the total number of oocytes per animal. We observed up to a threefold variation in total oocyte number among same-aged individuals with identical genetic backgrounds, revealing substantial inter-individual variability in oocyte numbers in isogenic mice (Extended Data Fig. 3a). This variability was already present in 5-week-old animals, suggesting that it arises before sexual maturation. Analysis of left and right ovaries from each animal found little asymmetry in oocyte numbers (Extended Data Fig. 3b). These findings reveal unexpectedly high variability in ovarian reserve among genetically identical animals, emphasizing the need to account for individual differences even within isogenic populations. Oocyte reserve declines with age, but transitioning fraction remains stable Oocytes are classified according to their size and the surrounding follicular structure2,17,18. To investigate oocyte growth kinetics, we used the segmentation masks generated by our artificial intelligence (AI) model to calculate the oocyte diameters and plot their distribution in 5-week-old ovaries. This analysis revealed a sharp peak around 15 µm, followed by a long tail corresponding to larger oocytes (Fig. 1g). High-resolution confocal imaging of oocytes around this peak (12–40 µm in diameter) allowed us to distinguish dormant primordial follicles, identified by flattened squamous granulosa cells19, growing primary follicles, which contain cuboidal granulosa cells (Extended Data Fig. 3c,d), and transitioning follicles, characterized by mixed squamous and cuboidal cell morphologies20. Based on size and morphology, we classified oocytes into primordial (<16 µm), transitioning (16–20 µm) and primary (20–40 µm) stages (Extended Data Fig. 3c,d), enabling distinction between dormant and recently activated populations. Next, we investigated how the ovarian reserve changes with age. Given the substantial interanimal variability observed at 5 weeks (Extended Data Fig. 3a,b), we analyzed 56 virgin mice spanning 5 to 60 weeks of age, using at least five mice per age. Using our high-throughput AI pipeline, we segmented over 85,000 oocytes from more than 100 intact 3D-imaged ovaries (Fig. 2a,b). This large dataset revealed substantial interindividual variability in total oocyte numbers across all ages examined. By contrast, differences in oocyte numbers and ovarian volume between the left and right ovaries of the same animal were comparatively small (Extended Data Fig. 3e–i). As expected, oocyte numbers declined significantly with age (Fig. 2b), and their relative contributions to the total population shifted over time: primordial oocytes decreased from ~70% to ~55% of the pool, while growing oocytes displayed the opposite trend, increasing from ~15% to ~30% (Fig. 2c). Notably, total oocyte numbers remained similar between 10 and 22 weeks of age (Fig. 2b).Fig. 2Ovarian reserve declines with age, but the transitioning oocyte fraction remains stable.a, Maximum-intensity projected SPIM images of ovaries from 5-, 10-, 22-, 31-, 40-, 50- and 60-week-old inbred C57BL/6J mice labeled with an antibody against DDX4, an oocyte marker. The n indicates the number of mice used for each age group. Scale bar: 200 μm. b, Quantification of the number of total, primordial, transitioning and growing oocytes. Each dot represents an ovary. Ovaries from the same animal share the same color. Boxes indicate the interquartile range (IQR), the center line represents the median, and whiskers extend to the minimum and maximum values. Statistical comparisons were performed between each age and the reference age (week 5, w05) using two-sided Mann–Whitney U tests. P values were adjusted for multiple comparisons using the Benjamini–Hochberg false discovery rate (FDR) correction. FDR-adjusted P values (q values) are displayed above the corresponding brackets. Note that, due to substantial interindividual heterogeneity in oocyte numbers, the sample sizes used were not enough to resolve the decline in oocyte numbers between 10-week-old and 22-week-old animals. c, Box-and-whisker plots showing the proportion of primordial oocytes (left), transitioning oocytes (center) and growing oocytes (right) relative to the total number of oocytes. Each dot represents an ovary. Ovaries from the same animal share the same color. Boxes indicate the IQR, the center line represents the median, and whiskers extend to the minimum and maximum values. Pairwise comparisons were performed against the baseline age (w05) using two-sided Mann–Whitney U tests. P values were adjusted for multiple comparisons using the Benjamini–Hochberg FDR and are displayed as adjusted P values (q values). d, Scatter plot showing the mean number of transitioning oocytes as a function of the mean number of total oocytes per age group. Each point represents the age-specific mean, color-coded by age. Error bars indicate the standard error of the mean (s.e.m.) for both total (x axis) and transitioning (y axis) oocyte counts. The red line represents a linear regression fit. e, Distribution of oocyte sizes in the ovary throughout aging. For each age group, histograms of oocyte diameter were computed per ovary using fixed bin edges spanning 5 to 100 µm (30 bins). For each age, per-ovary histograms were averaged to obtain the mean count per diameter bin (blue line), and variability across ovaries is shown as ±1 s.d. (shaded area). Each subplot corresponds to one age group and depicts the average oocyte number as a function of oocyte diameter. Insets show a magnified view of the 40–100-µm range. Please see Extended Data Fig. 4a,b for statistical analyses. For b–e, the sample sizes were as follows: w5, n = 20 ovaries from 12 mice; w10, n = 15 ovaries from 9 mice; w22, n = 16 ovaries from 9 mice; w31, n = 18 ovaries from 9 mice; w40, n = 12 ovaries from 7 mice; w50, n = 10 ovaries from 5 mice; w60, n = 10 ovaries from 5 mice.Source data Ovarian reserve declines with age, but the transitioning oocyte fraction remains stable. a, Maximum-intensity projected SPIM images of ovaries from 5-, 10-, 22-, 31-, 40-, 50- and 60-week-old inbred C57BL/6J mice labeled with an antibody against DDX4, an oocyte marker. The n indicates the number of mice used for each age group. Scale bar: 200 μm. b, Quantification of the number of total, primordial, transitioning and growing oocytes. Each dot represents an ovary. Ovaries from the same animal share the same color. Boxes indicate the interquartile range (IQR), the center line represents the median, and whiskers extend to the minimum and maximum values. Statistical comparisons were performed between each age and the reference age (week 5, w05) using two-sided Mann–Whitney U tests. P values were adjusted for multiple comparisons using the Benjamini–Hochberg false discovery rate (FDR) correction. FDR-adjusted P values (q values) are displayed above the corresponding brackets. Note that, due to substantial interindividual heterogeneity in oocyte numbers, the sample sizes used were not enough to resolve the decline in oocyte numbers between 10-week-old and 22-week-old animals. c, Box-and-whisker plots showing the proportion of primordial oocytes (left), transitioning oocytes (center) and growing oocytes (right) relative to the total number of oocytes. Each dot represents an ovary. Ovaries from the same animal share the same color. Boxes indicate the IQR, the center line represents the median, and whiskers extend to the minimum and maximum values. Pairwise comparisons were performed against the baseline age (w05) using two-sided Mann–Whitney U tests. P values were adjusted for multiple comparisons using the Benjamini–Hochberg FDR and are displayed as adjusted P values (q values). d, Scatter plot showing the mean number of transitioning oocytes as a function of the mean number of total oocytes per age group. Each point represents the age-specific mean, color-coded by age. Error bars indicate the standard error of the mean (s.e.m.) for both total (x axis) and transitioning (y axis) oocyte counts. The red line represents a linear regression fit. e, Distribution of oocyte sizes in the ovary throughout aging. For each age group, histograms of oocyte diameter were computed per ovary using fixed bin edges spanning 5 to 100 µm (30 bins). For each age, per-ovary histograms were averaged to obtain the mean count per diameter bin (blue line), and variability across ovaries is shown as ±1 s.d. (shaded area). Each subplot corresponds to one age group and depicts the average oocyte number as a function of oocyte diameter. Insets show a magnified view of the 40–100-µm range. Please see Extended Data Fig. 4a,b for statistical analyses. For b–e, the sample sizes were as follows: w5, n = 20 ovaries from 12 mice; w10, n = 15 ovaries from 9 mice; w22, n = 16 ovaries from 9 mice; w31, n = 18 ovaries from 9 mice; w40, n = 12 ovaries from 7 mice; w50, n = 10 ovaries from 5 mice; w60, n = 10 ovaries from 5 mice. Source data Intriguingly, however, the fraction of transitioning oocytes—those exhibiting both squamous and cuboidal granulosa cell morphologies—remained stable at approximately 14% of the ovarian reserve across all ages, even when the total reserve declined by more than an order of magnitude (Fig. 2c,d). This suggests that, throughout the entire reproductive lifespan of the animal, and despite the large variation in oocyte numbers between isogenic individuals, the ovary maintains a proportionate number of oocytes primed for activation at any given time. We next asked whether suppression of ovulation could slow the loss of the ovarian reserve. Ovulation suppression has been shown to reduce chromosomal mis-segregation in aging oocytes, thereby improving their quality and delaying functional decline21. However, its effect on the quantity of remaining oocytes has not been clearly established. To test this, we compared the ovarian reserve in female mice subjected to repeated mating with that of age-matched virgin controls, as pregnancy suppresses ovulation. Mating began at 7 weeks and continued for 18 weeks, with animals undergoing multiple successful pregnancies. At 31 weeks, we quantified total oocyte numbers, along with primordial, transitioning and growing follicles. We observed no significant differences between mated and virgin mice in any of these categories (Extended Data Fig. 3j). These findings suggest that suppression of ovulation, while beneficial for oocyte quality, does not slow the rate of oocyte loss with age. Together, these findings suggest that follicle recruitment is regulated at the tissue level and operates independently of ovulatory activity. The ovarian reserve exhibits a bimodal size distribution of oocytes Once activated, oocytes undergo dramatic growth, increasing their volume by approximately 64-fold22. How this growth is regulated remains poorly understood. If oocytes were activated at a constant pace and grew at a uniform rate, we would expect their sizes within the ovary to form a smooth continuum extending up to the ovulation stage. By contrast, if activation occurs in pulses or if growth is not continuous, we would anticipate depletions or accumulations at particular size ranges, reflecting pauses. To test these possibilities, we mapped the size distribution of all oocytes within intact ovaries across ages. We found that the size distribution of oocytes is bimodal: a dominant peak representing dormant oocytes, and a smaller secondary peak at approximately 60 µm (with a three-component Gaussian mixture preferred over one or two components at P ≤ 0.001 at all ages) (Figs. 1g and 2e, Extended Data Fig. 4a,b and Supplementary Table 1). This pattern suggests that oocytes encounter a developmental bottleneck or checkpoint at around 60 µm. Remarkably, the bimodal distribution is preserved throughout the reproductive lifespan of the animal, even though the ovarian reserve diminishes by more than an order of magnitude (Fig. 2e, Extended Data Fig. 4a,b and Supplementary Table 1). This persistence indicates that the mechanisms governing the ~60-µm bottleneck are robust to aging and independent of total oocyte number. Spatial patterning of oocytes reveals lateral enrichment along the ovarian axis Oocyte distribution within the ovary follows spatial cues23, but the 3D organization of oocytes in an ovary remains poorly characterized. Leveraging our 3D-positional data, we quantified oocyte organization across the ovarian volume. We began by generating isosurface renderings of each ovary, revealing a consistent, stereotypical bean-like morphology (Fig. 3a). This visualization delineated the concave medial side of the ovary, known as the hilum, the entry point for nerves and blood vessels, which served as a reproducible anatomical landmark across samples. To systematically quantify radial organization, we fit a spline along the central longitudinal ovarian axis to serve as an internal reference line, or ‘ruler’, from which the position and angle of each oocyte could be measured. This allowed us to assess oocyte distribution patterns across animals and age groups (Fig. 3b and Extended Data Fig. 5).Fig. 3Regions with high primordial oocyte density show higher activation.a, Isosurface rendering of a 5-week-old C57BL/6J ovary in light gray. Segmented oocytes are plotted as spheres and color-coded by their diameter. The solid red line represents the spline curve used for spatial reference. 0° corresponds to the superior side of the ovary, 45–135° and 225–315° represent the medio-lateral regions, and 180° corresponds to the hilum. b, Quantification of the radial distribution of oocytes shown in a, plotted separately for primordial (left), transitioning (middle) and growing (right) oocytes. c, Quantification of the radial distribution of oocytes during aging. Each solid line represents the mean percentage of oocytes per angular bin, and shaded areas indicate ±1 s.d. across ovaries (week 5, 8 ovaries; week 10, 5 ovaries; week 22, 6 ovaries; week 31, 9 ovaries; week 40, 5 ovaries; week 50, 7 ovaries; week 60, 5 ovaries). Percentages of oocytes contained in bins of 45° are shown for primordial oocytes (left), transitioning oocytes (middle) and growing oocytes (right). d, The same as in c, but replotted on sine-transformed axes, with individual ages (colored lines) and the locally estimated scatterplot smoothing regression across all ages (black). e, Box plots showing the distribution of mean primordial cluster size per ovary across age groups. Primordial cluster size is defined as the number of primordial oocytes within a radius of 100 µm from each primordial oocyte. Each red dot represents the average primordial cluster size per ovary across different ages. The center line of each box indicates the median, the box represents the IQR (25th–75th percentile), and whiskers extend to 1.5× the IQR. f, Scatter plot showing the association between primordial cluster size and the mean number of transitioning oocytes. Each dot represents the mean number of transitioning oocytes associated with a given cluster size, color-coded by age. Error bars indicate s.d. For e and f, the sample sizes were as follows: week 5, 20 ovaries; week 10, 15 ovaries; week 22, 16 ovaries; week 31, 18 ovaries; week 40, 12 ovaries; week 50, 10 ovaries; week 60, 10 ovaries.Source data Regions with high primordial oocyte density show higher activation. a, Isosurface rendering of a 5-week-old C57BL/6J ovary in light gray. Segmented oocytes are plotted as spheres and color-coded by their diameter. The solid red line represents the spline curve used for spatial reference. 0° corresponds to the superior side of the ovary, 45–135° and 225–315° represent the medio-lateral regions, and 180° corresponds to the hilum. b, Quantification of the radial distribution of oocytes shown in a, plotted separately for primordial (left), transitioning (middle) and growing (right) oocytes. c, Quantification of the radial distribution of oocytes during aging. Each solid line represents the mean percentage of oocytes per angular bin, and shaded areas indicate ±1 s.d. across ovaries (week 5, 8 ovaries; week 10, 5 ovaries; week 22, 6 ovaries; week 31, 9 ovaries; week 40, 5 ovaries; week 50, 7 ovaries; week 60, 5 ovaries). Percentages of oocytes contained in bins of 45° are shown for primordial oocytes (left), transitioning oocytes (middle) and growing oocytes (right). d, The same as in c, but replotted on sine-transformed axes, with individual ages (colored lines) and the locally estimated scatterplot smoothing regression across all ages (black). e, Box plots showing the distribution of mean primordial cluster size per ovary across age groups. Primordial cluster size is defined as the number of primordial oocytes within a radius of 100 µm from each primordial oocyte. Each red dot represents the average primordial cluster size per ovary across different ages. The center line of each box indicates the median, the box represents the IQR (25th–75th percentile), and whiskers extend to 1.5× the IQR. f, Scatter plot showing the association between primordial cluster size and the mean number of transitioning oocytes. Each dot represents the mean number of transitioning oocytes associated with a given cluster size, color-coded by age. Error bars indicate s.d. For e and f, the sample sizes were as follows: week 5, 20 ovaries; week 10, 15 ovaries; week 22, 16 ovaries; week 31, 18 ovaries; week 40, 12 ovaries; week 50, 10 ovaries; week 60, 10 ovaries. Source data Primordial oocytes were consistently depleted from both the concave (hilum) and convex poles of the ovary, while showing consistent enrichment along the lateral mid-regions of the long axis (beta shape1 parameter 0.73 with 95% confidence interval 0.72–0.75; Supplementary Table 2). This pattern was not limited to primordial oocytes: transitioning and growing oocytes exhibited the same spatial bias, and this distribution was maintained throughout aging (Fig. 3b–d, Extended Data Fig. 6a and Supplementary Table 2). These findings align with recent observations showing that the ovarian surface epithelium is sparsely populated near the hilum24. While follicle activation has been proposed to occur stochastically25,26, local microenvironmental factors can influence whether a primordial oocyte gets activated or remains dormant27,28. To test whether neighboring-follicle density influences oocyte activation, we performed a spatial density analysis of primordial oocytes across the ovarian volume. For each primordial oocyte, we quantified the number of neighboring primordial oocytes within a 100-µm radius. We observed highly heterogeneous local densities of follicles, which decrease with age: at 5 weeks, any primordial oocyte has, on average, over 10 neighbors, whereas at 60 weeks, this number is reduced to approximately 2.5 (P < 0.0001) (Fig. 3e, Extended Data Fig. 6b and Supplementary Table 2). We did not observe persistent high-density clusters of primordial follicles at later ages, suggesting that the high local density of primordial follicles does not inhibit their activation. Instead, we found that higher local primordial oocyte densities were associated with increased numbers of transitioning oocytes (P < 0.0001) (Fig. 3f, Extended Data Fig. 6c and Supplementary Table 2), irrespective of age. To further confirm this association, we also examined the inverse relationship: whether a high local density of transitioning oocytes was associated with increased numbers of primordial oocytes. We again identified a positive association (P < 0.0001) (Extended Data Fig. 6d,e and Supplementary Table 2). These results suggest that a high density of primordial oocytes in a given region is not inhibitory toward follicle activation, as previously suggested for prepubertal mouse ovaries23. On the contrary, we find that regions with more primordial oocytes tend to have more active oocytes as well. Mathematical modeling captures the kinetics of ovarian reserve depletion during aging Current technologies do not allow in vivo longitudinal studies of the ovary; therefore, we leveraged our large, age-spanning dataset to model the dynamics of ovarian reserve depletion and infer the underlying regulatory processes. To construct this model, we classified oocytes into four developmental categories based on follicular morphology and size: primordial (diameter ≤20 µm), primary (20 µm < diameter ≤ 40 µm), secondary (40 µm < diameter ≤ 60 µm) and antral (diameter >60 µm) (Methods). These four stages were chosen because they represent major morphological milestones in folliculogenesis29. This classification enabled us to plot stage-specific oocyte distributions across all time points, effectively capturing how the size profile of the ovarian reserve changes with age (Fig. 4a).Fig. 4A mathematical model recapitulates ovarian aging dynamics.a, Violin plots showing the number of oocytes per ovary across different age groups, categorized by oocyte stage: primordial, primary, secondary and antral. Each dot represents an individual ovary. Light-gray lines indicate model fits based on varying initial conditions, while the black line represents the fit starting from the mean values per each category of week 5 animal. The sample sizes were as follows: w5, n = 20 ovaries from 12 mice; w10, n = 15 ovaries from 9 mice; w22, n = 16 ovaries from 9 mice; w31, n = 18 ovaries from 9 mice; w40, n = 12 ovaries from 7 mice; w50, n = 10 ovaries from 5 mice; w60, n = 10 ovaries from 5 mice. b, Schematic representation of the mathematical model. The model suggests that oocyte death is minimal during the primordial, primary and secondary stages but increases at the antral stage. An activation rate of 2.5% of oocytes per week is sufficient to maintain the follicular pool for at least 60 weeks.Source data A mathematical model recapitulates ovarian aging dynamics. a, Violin plots showing the number of oocytes per ovary across different age groups, categorized by oocyte stage: primordial, primary, secondary and antral. Each dot represents an individual ovary. Light-gray lines indicate model fits based on varying initial conditions, while the black line represents the fit starting from the mean values per each category of week 5 animal. The sample sizes were as follows: w5, n = 20 ovaries from 12 mice; w10, n = 15 ovaries from 9 mice; w22, n = 16 ovaries from 9 mice; w31, n = 18 ovaries from 9 mice; w40, n = 12 ovaries from 7 mice; w50, n = 10 ovaries from 5 mice; w60, n = 10 ovaries from 5 mice. b, Schematic representation of the mathematical model. The model suggests that oocyte death is minimal during the primordial, primary and secondary stages but increases at the antral stage. An activation rate of 2.5% of oocytes per week is sufficient to maintain the follicular pool for at least 60 weeks. Source data We modeled oocyte developmental trajectory over time using a system of ordinary differential equations to describe the rates of change for each transition. In this framework, follicle progression from primordial to antral stages (denoted Pi) is governed by stage-specific transition rates (ki) and death rates (di), where index i represents each step in the sequence (Supplementary Data 1). The parameters of these equations were fitted to experimental observations of oocyte numbers in each subgroup (Fig. 4a), yielding the following results: k1 (primordial → primary) = 0.025808 ± 0.005288 per week k2 (primary → secondary) = 0.217252 ± 0.040008 per week k3 (secondary → antral) = 0.337478 ± 0.056654 per week d1 (primordial death) = 0.017856 ± 0.007404 per week d2 (primary death) = 0.0000 ± 0.0000 per week d3 (secondary death) = 0.0000 ± 0.0000 per week d4 (antral death and ovulation) = 0.722362 ± 0.147043 per week All constants are expressed in units of ‘per week’. These values indicate that, each week, approximately 2.5% of primordial oocytes are activated, 21.7% of primaries transition to the secondary stage, and 33.7% of secondaries develop into antral follicles. These results establish quantitative estimates of the oocyte activation and growth dynamics in the mammalian ovary (Fig. 4a,b). Our model predicts very low death rates for primary and secondary oocytes in adult ovaries and suggests that most oocyte loss occurs at the antral stage, consistent with findings in the literature19,30 and further supporting our model’s validity. Given the high interindividual variability in the ovarian reserve size, we next tested whether the model could solve the equations starting from the initial conditions of each individual 5-week-old ovary. Notably, the model remained stable across all these starting conditions (Fig. 4a). Taken together, these analyses demonstrate that a mathematical model can recapitulate the kinetics of ovarian reserve depletion across the mouse reproductive lifespan. This framework provides quantitative estimates of follicle recruitment, growth and death rates, and establishes a basis for further refinement and hypothesis generation. Discussion The duration of a female’s fertility is determined by the initial number of her oocytes and their depletion rate8. Therefore, understanding the kinetics of the ovarian reserve is key to understanding ovarian aging. By imaging intact mouse ovaries and quantifying over 85,000 oocytes across ages, we first uncovered striking interindividual variability in oocyte numbers among genetically identical mice, an aspect largely overlooked due to the lack of quantitative, whole-organ analyses. We next found that, while the total ovarian reserve declines dramatically with age, the proportion of transitioning follicles remains stable, suggesting a homeostatic mechanism that regulates follicle activation. Our results indicate that oocyte activation is sensed at the tissue level and that the ovary can monitor the total number of oocytes across a wide distribution of ovarian reserve sizes and animal ages. The high variation in oocyte numbers among isogenic, age-matched mice is already present before sexual maturation, suggesting an origin in early development. Initial reserve size may reflect differences in primordial germ cell migration and proliferation during colonization of the genital ridge, where small differences in founding populations can establish different baselines. In addition, the extent of oocyte loss during germ cell nest breakdown and early postnatal life may vary between individuals. Notably, low-reserve ovaries also exhibit reduced total volume and fewer growing oocytes, suggesting that these early events impact the overall ovarian function. Recent whole-ovary imaging approaches have begun to overcome limitations of section-based follicle quantification31–33. In particular, Gaylord et al. combined imaging with single-cell transcriptomics to compare ovarian aging between mouse and human, providing a cross-species framework for understanding age-dependent oocyte loss34. Our total oocyte numbers and their age-related decline are in close agreement with this study, supporting the robustness of whole-mount imaging strategies for ovarian reserve assessment. However, the present work extends beyond global oocyte counts by explicitly leveraging the spatial and geometric information inherent to intact ovary imaging. By segmenting entire oocytes rather than nuclei, we are able to reconstruct oocyte size distributions, quantify local follicle density and analyze spatial organization across the full ovarian volume with age. This enables us to identify principles of ovarian aging that are not apparent from aggregate counts alone, including the maintenance of a fixed fraction of oocytes primed for activation across ages, spatial enrichment of follicles along the lateral ovarian axis, and a persistent bimodal distribution of oocyte sizes indicative of a developmental bottleneck during oogenesis. A central unresolved question in ovarian biology is why some oocytes become activated while the majority remain dormant at any given time. Our data reveal that regions with higher primordial oocyte density contain more recently activated oocytes. This observation challenges the long-standing ‘high-density inhibition’ hypothesis in mouse ovaries23. Recent spatial transcriptomics of the aging mouse ovary revealed that reproductive aging is marked by a breakdown in tissue-wide spatiotemporal coordination of folliculogenesis and niche organization, complementing our study to highlight the importance of spatial context in ovarian aging35. We further found that the size distribution of all oocytes within the ovary is bimodal, with a first peak at ~15 µm representing the pool of primordial oocytes, and a second peak at ~60 µm (Fig. 2e). The latter coincides with the onset of gonadotropin responsiveness36, raising the possibility of oocyte growth rates changing during oogenesis. Indeed, recent research suggested a burst of oocyte growth right after this stage driven by an oocyte-specific organelle37. Elucidating the mechanisms underlying this bottleneck will be important for understanding oocyte growth dynamics. Finally, the scale and resolution of our dataset allowed us to explore the kinetics of ovarian reserve depletion through mathematical modeling. In building this model, we assumed that oocyte growth and death rates remain broadly similar across aging ovaries. We acknowledge that this is an assumption that may require refinement as additional data become available. The model predicted minimal cell death at the primordial, primary and secondary stages, with the majority of oocyte loss occurring at the antral stage. This prediction is consistent with previous experimental reports19,30, thereby reinforcing the robustness of our model’s assumptions. We believe this framework establishes a foundation for generating hypotheses and refining future models of ovarian reserve dynamics. Humans and mice share the fundamental principle of a nonrenewable oocyte pool that declines with age, yet they differ in scale and timing: Humans are endowed with approximately one million oocytes at birth, losing more than half by puberty34. Mice exhibit similar depletion but with smaller absolute numbers, starting with approximately 5,000–10,000 oocytes38. Furthermore, mice are poly-ovulatory with short (4–5-day) estrous cycles, whereas humans are typically mono-ovulatory with ~28-day menstrual cycles. These differences probably influence follicle depletion dynamics. Nevertheless, our finding that the transitioning fraction remains stable may represent a universal mammalian principle of homeostatic recruitment. In both species, the vast majority of follicles remain quiescent, with only a small proportion undergoing activation at any given time to ensure reproductive longevity. Moreover, the substantial heterogeneity of oocyte numbers even among isogenic mice highlights the challenge of assessing human ovarian reserve, and the need for large cohorts to have robust conclusions. Altogether, by merging whole-organ imaging with quantitative modeling, our work reframes the ovarian reserve as an actively regulated system rather than a passive reservoir, opening new avenues for understanding the principles governing female reproductive longevity and their broader implications for tissue maintenance and aging. Methods Ethics The C57BL/6J mice used in this study were purchased from Charles River Laboratories and housed in the PRBB biomedical park, accredited by the international Association for Assessment and Accreditation of Laboratory Animal Care under specific pathogen-free conditions at 22 °C with 40–60% humidity, in a 12-h light/dark cycle, and with access to food ad libitum. All use of animals was performed according to the requirements of directive 2010/63 EU and the Guide. Ethical Committee permission to work with human ovary samples was obtained from the Comité Étic d’Investigació Clinica (CEIC)-Parc de salut MAR (Barcelona) and Comité Ético de Investigación Clinica (CEIC)-Hospital Clinic de Barcelona with approval number HCB/2023/0777. Written informed consent was obtained from both participants. Reagents and antibodies The following reagents were purchased from the indicated suppliers: normal goat serum (NGS; Merck Millipor, cat. no. S26-100ML), benzyl benzoate (Sigma, cat. no. W213802-1KG-K), benzyl alcohol (Sigma, cat. no. 402834-500ML), ToPro3 (Invitrogen, cat. no. T3605), Triton (Sigma, cat. no. X100-500ML), dimethyl sulfoxide (DMSO, Sigma, cat. no. D2438-10ML), low-melting agarose (Invitrogen, cat. no. 16520050), formaldehyde (Thermo, cat. no. 28908) and sodium azide (Sigma, cat. no. 71290). The following antibodies were purchased from the indicated suppliers: anti-DDX4 (Abcam, cat. no. AB13840) and goat anti-rabbit IgG (H + L) cross-adsorbed secondary antibody Alexa Fluor 488 (Invitrogen, cat. no. A-11008). They were both used at 1:1,000 dilution. Whole-mount ovary immunostaining Ovaries were obtained from C57BL/6J mice aged 5–60 weeks. After dissection, the ovarian bursa was manually removed, and ovaries were fixed for 1 h at room temperature (RT) in 2% paraformaldehyde in phosphate-buffered saline (PBS), then transferred to 80% methanol/DMSO at −20 °C for at least one night, as previously described31. Ovaries were then washed in 50% methanol/PBS, 25% methanol/PBS, and PBS for at least 30 min each step and permeabilized for 5 days at RT with gentle agitation. Ovaries were blocked for 1 day at RT in 10% NGS, 1% Triton X-100 and 0.05% NaN3, then incubated with primary antibody for at least 3 days, washed 4 × 30 min in 1% Triton X-100/PBS, incubated with Alexa-conjugated secondary antibody for at least 3 days at RT, and washed again 4 × 30 min in 1% Triton X-100/PBS. Human ovary cortex immunostaining Pieces of the human ovarian cortex were donated by two individuals aged 25 years. Tissue was fixed in 4% paraformaldehyde for 4 h at RT, then transferred to 80% methanol/DMSO at −20 °C for at least one overnight. Samples were permeabilized for ≥8 days in 2% Triton X-100/PBS, blocked overnight in 10% NGS, 1% Triton X-100 and 0.05% NaN3, incubated for ≥4 days with primary antibody, and then incubated for ≥4 days with Alexa-conjugated secondary antibody. Mounting ovaries in agarose block and BABB clearing The mouse ovaries or human ovarian cortex pieces were washed twice in Milli-Q water (to remove PBS salts) and transferred to a 2-ml Eppendorf tube containing 2 ml of 1% low-melting agarose. The ovary was positioned in the middle of the tube, and the tube was laid on its side on ice to accelerate solidification. After ~30 min, the agarose had solidified; the end of the tube was cut with a blade, and the agarose cylinder containing the ovary was extracted. The cylinder was further trimmed to fit the SPIM sample holder. Samples were then transferred to a 20-ml glass tube and incubated in 100% methanol for at least one overnight period; methanol was subsequently removed, and tissue clearing was performed in benzyl alcohol:benzyl benzoate (BABB; 2:1, vol/vol). SPIM microscopy Whole ovaries were labeled with an antibody against DDX4 and the dye TO-PRO-3, to label oocytes and DNA, respectively. The ovaries were imaged using a Luxendo MuViSPIM equipped with a 10× detection lens and two-side illumination setup, resulting in a resolution of a pixel size of 0.867 μm. Images were spaced 2.5 μm in Z. Excitation at 488 nm was used for detecting DDX4-positive oocytes and 642 nm for detecting TO-PRO-3-labeled nuclei. To cover the whole volume, multiple tiles were acquired. Stitching of the tiles was performed using either MosaicExplorerJ39 or the Luxendo image software. Confocal microscopy Whole ovaries labeled with an antibody against DDX4 to label oocytes and TO-PRO-3 dye to label DNA and were imaged using an SP8Stellaris system equipped with a 63× glycerol lens. We used the function tile-scan to cover a large ovary area. Oocyte recognition using BiaPy To specifically address the task of oocyte segmentation, that is, instance segmentation task, we used a bottom-up approach to train the network to predict three channels: binary foreground mask (F) instance contour map (C) and an Euclidean distance map to the center of the instance (Dc). This model can be called FCDc, following the nomenclature and the successful combination of these channels in other state-of-the-art methods40–45. After prediction, all outputs are thresholded and combined to enhance object delineation. A connected components algorithm was then applied to the combined mask to identify discrete, non-overlapping oocyte instance seeds. These seeds were subsequently used as markers in a marker-controlled watershed algorithm46, which incorporates three key components: (1) the inverted foreground probability map as the input image (defining the topographic surface), (2) the instance seeds as the marker image (indicating the initiation points of the flooding process) and (3) a binarized version of the foreground mask (B) as the constraint mask (limiting the extent of region expansion). This integrated pipeline enables the effective segmentation of individual oocyte instances. A visual representation of the process is provided in Extended Data Fig. 1a. The model has five levels with 48 filters in the initial level that get summed by 16 on each level (until 96 in the bottleneck), ELU activation functions and transposed convolutions to perform the upsampling in the decoder. We used batch normalization to regularize the model. The model was trained using a patch size of 40 × 128 × 128 and a kernel size of 3 in all convolutions. Our approach is fully integrated within the BiaPy library15, using Pytorch version 2.4. As data augmentation techniques we employed horizontal and vertical flips, elastic transformations and gridmarks47. We set aside 10% of the training samples for validation. Cross-entropy was used as a loss function, with a learning rate of 1 × 10−4, employing a cosine-decay scheduler with warm-up48, and ADAMW optimizer49. The model was trained until convergence using eight NVIDIA GeForce RTX3090 graphics processing units. We generated an online tutorial, which can be accessed at https://biapy.readthedocs.io/en/latest/tutorials/instance_seg/ovarian-reserve.html#post-analysis-scripts. Data analysis Ovarian reserve quantification After visual inspection and correction of the segmented mask, we calculated the 2D bounding box around each oocyte using MorpholibJ Fiji plugin50. From the 2D bounding box, the diameter can be obtained by calculating the square root of the product of the short and long sides of the bounding box: √(a × b). For choosing the threshold between primordial and transitioning oocytes we use the mean ± the standard deviation (s.d.) of the values measured in Extended Data Fig. 2d. We obtained values from 16.30 μm to 20.11 μm in diameter. Oocytes with diameter from 10 to 16.30 μm were classified as primordial; those with diameters of 16.30–20.11 μm were classified as transitioning; and oocytes greater than 20.11 μm were classified as growing. To account for slight differences in oocyte sizes across different ovaries, we aligned all individual primordial peaks to the average primordial peak of the entire dataset. First, we calculated the global average primordial peak across all ovaries and then aligned each individual primordial peak to this reference to minimize potential variability arising from fixation on different days. After alignment, the mean positional shift across all ovaries was effectively zero (min −3.6 µm, max 2.1 µm, s.d. 0.86 µm). For the human segmentation mask, we applied a size-based filter to exclude segments smaller than 15 µm, which correspond to segmentation artifacts. Gaussian mixture model fits We fit log2-transformed oocyte size distributions with Gaussian mixture models using the mixtools package in R. We then identified the most likely number of underlying components contributing to the measured oocyte distribution using the bootstrapped likelihood ratio statistic to evaluate the superiority of a k-component fit compared to a k + 1 component fit. Oocyte radial quantification Isosurface rendering of the ovarian surface was performed from the raw image stacks using the Vedo library51 to visualize the 3D morphology of the ovary. From the corresponding segmented masks, the center of mass of each oocyte was computed using the MorphoLibJ plugin in Fiji. A spline was then manually positioned through the ovary, connecting the center of the short axis and extending along the long axis. Due to the geometry of the ovary, this midline typically lies within a single plane, allowing a consistent definition of its orientation. Based on this reference, the angular distribution of oocytes was measured, distinguishing dorsal oocytes (located at small angles, facing the concavity of the spline) from ventral oocytes (located at angles approaching 180°). To quantify deviations from homogenous radial distributions, we took advantage of the symmetry of our data along the 0–180° axis to perform an angular transformation sin(angle/2), which maps all angles to a scale between 0 and 1, with 1 being the maximum angular distance from the 0° reference position. We then fit beta distributions to the transformed data, and identified deviations from a homogeneous angular oocyte distribution by testing the hypothesis that both beta distribution parameters were equal to one using bootstrap 95% confidence intervals. Intra- versus interindividual differences in oocyte counts To compare intraindividual differences between an individual’s two ovaries with the interindividual differences between individuals’ ovaries, we fit the linear regression model yi = βXi + ei, where yi is the number of oocytes in an ovary, Xi is the categorical covariate coding for each animals ID, β is the estimated difference in oocyte counts between individuals and ei is the residual. Quantifying relationship between oocyte local densities We quantified the relationship in local densities between transitioning and primordial oocytes by fitting a second-order polynomial using linear regression, and estimated the significance of the t-statistic for first- and second-order terms. Quantifying the effect of age on oocyte cluster sizes To quantify the effect of age on oocyte cluster sizes, we first investigated the distributional form of the count data and found it was well described by a negative binomial with an inflated number of clusters containing exactly one oocyte. The data naturally contained no oocyte clusters with zero oocytes. To model this 1-inflated data, we subtracted 1 from each cluster size, fit the transformed data using a zero-inflated negative binomial regression model using the pscl package in R, and estimated significant effects on age for both the negative binomial scale parameter, quantifying changes in the typical cluster size with age, as well as the degree of 1-inflation observed, quantifying the degree of saturation of measurements at a minimum count size of 1. Modeling the ovarian reserve The oocyte size distribution was partitioned into four diameter-based classes: primordial (10–20 µm), primary (>20–40 µm), secondary (>40–60 µm) and antral (>60 µm). Class boundaries were defined to be non-overlapping. Oocyte activation is widely accepted as a unidirectional process5. Once an oocyte initiates growth, it either continues maturing or undergoes atresia, but it does not revert to dormancy. Conceptually, this process can be represented as a sequence of transitions: primordial oocytes either progress to the primary stage or die; primary oocytes either advance to the secondary stage or die; secondary oocytes either develop into antral follicles or die; and antral follicles either ovulate or degenerate. In the static 3D dataset, oocytes morphologically transitioning from primordial to primary were observed and classified as ‘transitioning.’ However, while this transitional state can be identified at the level of individual oocytes, the corresponding transition rates into and out of this state cannot be directly measured or reliably inferred from static observations. Intermediate states for later transitions (primary→secondary and secondary→antral) were also not reliably identifiable. Therefore, we implemented a dynamical model describing transitions between the four main stages. The model infers transition and death rates from static stage counts without requiring direct observation of intermediate states. To account for natural variation in ovarian size, we normalized oocyte counts to the total organ volume (Extended Data Fig. 3h,i), which reduced the spread of the distributions and removed outliers (Extended Data Fig. 7). A set of differential equations is then solved using as initial conditions the number of oocytes present in each category at week 5 (Supplementary Data 1). The minimization is then performed using the sequential least-squares programming (SLSQP) algorithm as implemented in the SciPy library. During fitting to the experimental oocyte counts, the least-squares constrained minimization leads to zero values for the death rates for primary and secondary type oocytes (d2 and d3). As this is consistent with what is known in the literature20,33, these two parameters were fixed to zero to obtain our final results (Fig. 4a). The model equations are provided in the Supplementary Information and the code is available via GitHub at https://github.com/Boke-Lab/ovarian_reserve/. Statistics and reproducibility Statistical analyses were performed using RStudio and Python packages, including NumPy, SciPy, Pandas and StatsModels. Sample sizes were chosen empirically on the basis of published studies to ensure reliable statistical testing7,9,23,31. No statistical method was used to predetermine sample size. No data were excluded from the analyses. Mice were chosen by the animal facility personnel randomly from the relevant age group without participation of the authors. Data collection and analysis were not performed blind to the conditions of the experiments. Data distribution was assumed to be normal unless specified otherwise. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary InformationSupplementary Data 1: modeling oocyte dynamics. Reporting Summary Peer Review File Supplementary Video 1SPIM z-stack of a 5-week-old C57BL/6J ovary, labeled with anti-DDX4 (green) and TO-PRO-3 (magenta). Supplementary Video 2SPIM z-stack and 3D reconstruction of a 5-week-old C57BL/6J ovary labeled with anti-DDX4, overlaid with the corresponding segmentation mask. Supplementray Video 3SPIM z-stack of a strip of human cortex labeled with anti-DDX4 and overlaid with the corresponding segmentation mask. Supplementary Table 1Tab 1: bootstrap P values testing the hypothesis that k-component Gaussian mixture model (GMM) is superior to a k + 1 component GMM fit to the same data, with k = 1, 2 and 3, performed separately for each age group. Data associated with Extended Data Fig. 4a,b. Tab 2: the best-fit three-component GMM parameters for each age group. Data associated with Extended Data Fig. 4a,b. Supplementary Table 2Tab 1: the beta regression parameters for sine-transformed oocyte angular distributions for primordial, transitioning and growing oocytes. Data associated with Extended Data Fig. 6a. Tab 2: parameter estimates for the zero-inflated negative binomial regression for count-1 transformed data, estimating the effect of age on oocyte count distributions relative to the week five reference. Data associate with Extended Data Fig. 6b. Tab 3: second-order polynomial fits predicting the transitioning oocyte counts as a function of primordial oocytes counts, for each age. Data associated with Extended Data Fig. 6c. Tab 4: second-order polynomial fits predicting the primordial oocyte counts as a function of transitioning oocytes counts, for each age. Data associated with Extended Data Fig. 6e. Supplementary Data 1: modeling oocyte dynamics. Reporting Summary Peer Review File SPIM z-stack of a 5-week-old C57BL/6J ovary, labeled with anti-DDX4 (green) and TO-PRO-3 (magenta). SPIM z-stack and 3D reconstruction of a 5-week-old C57BL/6J ovary labeled with anti-DDX4, overlaid with the corresponding segmentation mask. SPIM z-stack of a strip of human cortex labeled with anti-DDX4 and overlaid with the corresponding segmentation mask. Tab 1: bootstrap P values testing the hypothesis that k-component Gaussian mixture model (GMM) is superior to a k + 1 component GMM fit to the same data, with k = 1, 2 and 3, performed separately for each age group. Data associated with Extended Data Fig. 4a,b. Tab 2: the best-fit three-component GMM parameters for each age group. Data associated with Extended Data Fig. 4a,b. Tab 1: the beta regression parameters for sine-transformed oocyte angular distributions for primordial, transitioning and growing oocytes. Data associated with Extended Data Fig. 6a. Tab 2: parameter estimates for the zero-inflated negative binomial regression for count-1 transformed data, estimating the effect of age on oocyte count distributions relative to the week five reference. Data associate with Extended Data Fig. 6b. Tab 3: second-order polynomial fits predicting the transitioning oocyte counts as a function of primordial oocytes counts, for each age. Data associated with Extended Data Fig. 6c. Tab 4: second-order polynomial fits predicting the primordial oocyte counts as a function of transitioning oocytes counts, for each age. Data associated with Extended Data Fig. 6e. Source data Source Data Fig. 1Statistical source data. Source Data Fig. 2Statistical source data. Source Data Fig. 3Statistical source data. Source Data Fig. 4Statistical source data. Source Data Extended Data Fig. 1Statistical source data. Source Data Extended Data Fig. 2Statistical source data. Source Data Extended Data Fig. 3Statistical source data. Source Data Extended Data Fig. 4Statistical source data. Source Data Extended Data Fig. 5Statistical source data. Source Data Extended Data Fig. 6Statistical source data. Source Data Extended Data Fig. 7Statistical source data. Statistical source data. Statistical source data. Statistical source data. Statistical source data. Statistical source data. Statistical source data. Statistical source data. Statistical source data. Statistical source data. Statistical source data. Statistical source data.