The maintenance of oocytes in the mammalian ovary involves extreme protein longevity

Abstract

Women are born with all of their oocytes. The oocyte proteome must be maintained with minimal damage throughout the woman's reproductive life, and hence for decades. Here we report that oocyte and ovarian proteostasis involves extreme protein longevity. Mouse ovaries had more extremely long-lived proteins than other tissues, including brain. These long-lived proteins had diverse functions, including in mitochondria, the cytoskeleton, chromatin and proteostasis. The stable proteins resided not only in oocytes but also in long-lived ovarian somatic cells. Our data suggest that mammals increase protein longevity and enhance proteostasis by chaperones and cellular antioxidants to maintain the female germline for long periods. Indeed, protein aggregation in oocytes did not increase with age and proteasome activity did not decay. However, increasing protein longevity cannot fully block female germline senescence. Large-scale proteome profiling of ~8,890 proteins revealed a decline in many long-lived proteins of the proteostasis network in the aging ovary, accompanied by massive proteome remodeling, which eventually leads to female fertility decline.

Fig. 1. Oocytes contain a large number of very long-lived proteins.

a, Fully labelled ¹³C₆-Lys females were mated and fed ¹³C₆-Lys chow until they gave birth (pulse). Upon birth, the pups were transferred to unlabelled foster mothers (chase) and continued to receive ¹²C₆-Lys chow after weaning. In total, 4,948 oocytes were collected from 92 eight-week-old pubertal female progeny and processed for bottom-up MS. b, Selected pathways enriched with long-lived proteins in oocytes. All ¹³C₆-Lys-labelled proteins detected after 8 weeks were subjected to over-representation analysis. Shown are the percentage of genes of the gene set detected as enriched with their respective adjusted P values for selected pathways. The P values are based on a hypergeometric test and have been adjusted using Benjamini–Hochberg multiple hypothesis testing. For the complete list of enriched pathways with their corresponding exact P values, see Supplementary Table 2. c, Selected biological processes and protein complexes with long-lived proteins. Shown are selected genes with corresponding protein %H values (inferred fraction of ¹³C₆-Lys, mean value over biological replicates). For a complete list of proteins and their corresponding %H values, see Supplementary Table 1. Source data

Fig. 2. Modelling protein turnover throughout ovarian development.

a, Fully ¹³C₆-Lys labelled pregnant mice were fed with ¹³C₆-Lys chow until they gave birth (pulse). The pups were subsequently raised on ¹²C₆-Lys (chase). Ovaries were collected from the female progeny at eleven time points (three animals per time point) and processed for DDA MS. b, Progeny from fully ¹³C₆-Lys labelled pregnant mice were fed with ¹³C₆-Lys until weaning. After the weaning period (3 weeks after birth), progeny were fed ¹²C₆-Lys chow. Ovaries were collected from female progeny at six time points (three animals per time point) and processed for DDA MS. c,d, Example data used for modelling protein turnover. Shown are MS1 intensities over time (linear c; log₁₀, d) scale for ¹³C₆-Lys (black and purple) and ¹²C₆-Lys (grey and pink) labelled histone H4, for two pulse lengths (pulse until birth and until weaning). Circles indicate experimental data points for three biological replicates, dashed lines are only for visual aid. e, Schematic of the experimental design and mathematical modelling. Pulse–chase protein concentration data were used to inform the model and estimate protein turnover rates. Dilution factors due to ovary growth were estimated by the model and compared to experimental ovary growth measurements for validation (indicated by a grey dashed arrow). f, Graphical illustration of the employed protein turnover model to estimate H_1/2 values in ovaries. g,j, Example (histone H4) of protein turnover model fitting. Experimental data are indicated as black dots, median and confidence ranges of model fit are indicated by the red line and shaded area, respectively. g,h, The ratio (R) between heavy (H) and light (L) labelled protein intensities in the chase (¹²C₆-Lys) from birth (g) and from weaning (h) experiment datasets are shown over time. i, Normalized protein abundance derived from DIA-MS measurements is shown over time for histone H4. j, Proportion of heavy labelled histone H4 compared with all proteins as derived from the estimated protein turnover posterior parameter distribution. Source data

Fig. 3. Mouse ovaries have a >10-fold higher fraction of extremely long-lived proteins than other post-mitotic tissues.

a, Distribution of the modelled H_1/2 values in the ovary. b, Fraction of proteins with H_1/2 > 100 days in the mouse ovary samples in this study and various mouse and rat tissues across three studies^–. c, Distributions of estimated H_1/2 values for proteins located in indicated subcellular compartments. Red and blue stars indicate significantly larger and smaller H_1/2 distributions compared with the whole modelled proteome, respectively (two-sided Kolmogorov–Smirnov test; P values: nucleus 0.0093; mitochondrion 8.84 × 10^–6). Number of proteins in each compartment indicated in parentheses. Boxplots indicate median, first quartile and third quartile, as well as minimum and maximum after outlier removal. d, Cluster analysis of inferred ¹³C₆-Lys levels in the proteins of the aging ovaries. The dendrogram on the left corresponds to the clustering of inferred percentage ¹³C₆-Lys. The leftmost bar shows the medians of inferred proteins half-lives, colouring on log₁₀ scale. The second and third bars indicate the latest time point at which the ¹³C₆-Lys pulse was detected in the data corresponding to chase from birth and weaning. The rightmost bar labels the three identified protein longevity clusters corresponding to proteins with high, intermediate and low amounts of inferred ¹³C₆-Lys content. The leftmost heatmap shows the inferred percentage ¹³C₆-Lys; time points from 6 weeks onwards were used for clustering. Central and rightmost heatmaps show experimental data of the chase ¹²C₆-Lys from birth and weaning, respectively. e, Modelled percentage of ¹³C₆-Lys labelled proteins compared to ¹²C₆-Lys labelled proteins was used to derive three protein longevity clusters (low, intermediate and high inferred ¹³C₆-Lys content). Bar plot showing number of proteins in the low, intermediate and high protein longevity clusters. f, Violin plot showing distributions of inferred half-lives of proteins in the low, intermediate and high protein longevity clusters. Boxplots indicate median, first quartile and third quartile, as well as minimum and maximum after outlier removal. Number of proteins in each cluster as indicated in e. g–i, Dot plots comparing the H_1/2 values of different proteins in the ovary (x axis) with organs and tissues (y axis) measured in ref.  (g), ref.  (h) and ref.  (i, liver; j, cricoid cartilage). Test for association between paired samples using Spearman correlation coefficient (C) was performed with P values estimated using algorithm AS 89. Proteins are colour-coded according to the protein longevity cluster they belong to, as in d. Source data

Fig. 4. The ovary contains long-lived proteins that persist throughout the lifetime of a mouse.

a, Selected pathways enriched in the high protein longevity cluster. Shown are the proportions of genes of the selected gene set detected in the high protein longevity cluster with their respective adjusted P values for selected pathways. The P values are based on the hypergeometric test and have been adjusted with Benjamini–Hochberg multiple hypothesis testing. For the complete list of enriched pathways with their corresponding exact P values, see Supplementary Table 5. b, Selected biological processes and protein complexes with long-lived proteins in ovaries. Shown are gene names with corresponding protein H_1/2 values (medians, 5% and 95% quantiles; designated as Q5 and Q95, respectively) in days. For the complete list of proteins and their corresponding H_1/2 values, see Supplementary Table 3 and Supplementary Data 2. c–f, MS1 intensities over time in log₁₀ scale for ¹³C₆-Lys (black and purple) and ¹²C₆-Lys (grey and pink) labelled cytochrome c, somatic (c); mitochondrial import receptor subunit TOMM70 (d); 28S ribosomal protein S9, mitochondrial (e); and malate dehydrogenase, mitochondrial (f) for two pulse lengths (pulse with ¹³C₆-Lys until birth and pulse with ¹³C₆-Lys until weaning). Circles indicate experimental data points for three biological replicates, dashed lines are only for visual aid. Source data

Fig. 5. Subsets of granulosa, stromal and theca cells are long-lived in the ovary.

a, Proportion of the proteins from the high protein longevity cluster with greater or less than twofold enrichment of corresponding transcripts in specific cell types in the ovaries of postnatal day 2 mice. b, Distribution of the transcripts of the proteins from the high protein longevity cluster with greater than twofold enrichment in specific cell types of the ovaries of postnatal day 2 mice. c, Dot plot showing the expression patterns of the transcripts of the proteins from the high protein longevity cluster. Size of the dot represents the proportion of the cells expressing the gene, colour denotes log₂(fold change) in the one versus all cell types differential gene expression test. d–k, NanoSIMS imaging of ovaries from 4-week-old (d–g) and 8-week-old (h–k) mice that were pulsed with ¹³C₆-Lys until weaning (3 weeks after birth), followed by a chase period with ¹²C₆-Lys. d,h, Histological sections of ovarian tissue (first column on the left). Insets show magnified areas highlighted in the histological sections (middle section of d and h). ¹³C₆/¹²C₆ ratio image for the indicated histological sections and corresponding values are given on the right-hand side of the magnified insets. e,i, Dot plots showing ¹³C₆ /¹²C₆ signal in different ovarian cell types. f,j, Dot plots showing ¹³C₆ /¹²C₆ signal of granulosa cells in the primary, secondary and antral follicles. g,k, Dot plots showing ¹³C₆ /¹²C₆ signal of theca cells in the primary, secondary and antral follicles. Numbers of analysed cells are shown in parentheses. P values were calculated using unpaired two-tailed Student’s t-test. n.s., not significant; *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001. Scale bars, 10 µm. Source data

Fig. 6. Protein aggregation does not increase in aged oocytes.

a, Representative immunofluorescence images of fully grown, germinal vesicle stage mouse oocytes from 9-week-old and 65-week-old mice stained with the ProteoStat aggresome dye. Magenta, aggresome (ProteoStat); cyan, DNA (Hoechst). b, Dot plot showing number of ProteoStat-positive structures in oocytes as shown in a. c, Dot plot showing total intensity of ProteoStat-positive structures in oocytes as shown in a. d, Representative immunofluorescence images of brain slices from 9-week-old and 65-week-old mice stained with the ProteoStat aggresome dye. Magenta, aggresome (ProteoStat); cyan, DNA (Hoechst). e, Dot plot showing number of ProteoStat-positive structures in brain slices as shown in d. f, Dot plot showing total intensity of ProteoStat-positive structures in brain slices as shown in d. g, Representative immunofluorescence images of early follicles from 9-week-old and 65-week-old mice stained with the ProteoStat aggresome dye. Magenta, aggresome (ProteoStat); cyan, DNA (Hoechst). No obvious aggresome accumulation was detected in either age group. All data from two independent experiments. Number of analysed oocytes and brain areas are in parentheses. Data are shown as mean ± s.d. P values were calculated using unpaired two-tailed Student’s t-test. n.s., not significant; ****P ≤ 0.0001. Scale bars, 10 µm. Source data

Fig. 7. Proteasomal activity does not decay in aged oocytes.

a, Time-lapse images of mouse oocytes from 9-week-old mice expressing Ub(G76V)-mClover3-T2A-mScarlet in the presence of DMSO or 10 μM MG-132. Time is given as hours after DMSO or MG-132 treatment. b, Quantification of the mean fluorescence intensity of mClover3 in oocytes as shown in a. c, Quantification of the fluorescence intensity ratio of mClover3 to mScarlet in oocytes in a. d, Time-lapse images of mouse oocytes from 9-week-old and 65-week-old mice expressing Ub(G76V)-mClover3-T2A-mScarlet. Time is given as hours after injection of the reporter mRNA. e, Quantification of the fluorescence intensity ratio of mClover3 to mScarlet in oocytes in d. f, Schematic diagram of the experiment shown in g. Ub(G76V)-mClover3 and mScarlet were expressed for 5 h in the presence of MG-132, which blocks the degradation of Ub(G76V)-mClover3. MG-132 was then washed out and oocytes were imaged in the presence of the translation inhibitor cycloheximide (CHX), which blocks the synthesis of new proteins. g, Time-lapse images of mouse oocytes from 9-week-old and 65-week-old mouse oocytes expressing Ub(G76V)-mClover3-T2A-mScarlet. Experiment was performed as shown in f. Time is given as hours after MG-132 washout and CHX wash-in. h, Line graph showing normalized fluorescence intensity ratio of mClover3 to mScarlet in oocytes in g. All data from two independent experiments. Number of analysed oocytes in parentheses. Data are shown as mean ± s.d. P values were calculated using unpaired two-tailed Student’s t-test. n.s., not significant; ****P ≤ 0.0001. Scale bars, 10 µm. Source data

Fig. 8. Proteostasis networks are lost with ovarian aging and decreased fertility.

a, Schematic overview of experimental design. Ovaries were collected from female mice at three time points (9-, 12- and 50-week-old mice; three animals per time point analysed). Samples were processed for DIA-MS. b, Cluster analysis of protein abundances during reproductive decline. Normalized DIA-MS data for 9-, 12- and 50-week-old ovaries were subject to clustering to minimize variance within clusters. Dendrogram of resulting clusters is shown on the left. Colours indicate DIA intensities centred to a mean of 0 and s.d. of 1 on per protein basis. Resulting protein abundance clusters are highlighted with different colour keys. (c-e) Pathway over-representation analysis of protein abundance cluster 2 (c), cluster 4 (d) and cluster 5 (e). f, Heatmaps showing protein abundance change for the high protein longevity cluster with assignment to protein abundance clusters. g, Pathway over-representation analysis of high protein longevity cluster proteins in protein abundance clusters 2 and 4 in the ovary. c–e and g show the proportions of genes detected in the gene set with their respective adjusted P values for the most prominent pathways. The P values are based on the hypergeometric test and have been adjusted for multiple hypothesis testing with Benjamini–Hochberg procedure. For the complete list of enriched pathways and their corresponding exact P values, see Supplementary Table 9. Source data

Extended Data Fig. 1. Generation of fully-¹³C₆-Lys-labelled FVB/N female mice.

(a) Schematic overview of complete ¹³C₆-Lys labelling of mice. Fully-¹³C₆-Lys-labelled FVB/N female mice were created by successive matings and a diet consisting exclusively of ¹³C₆-Lys feed. F0 females were fed with the ¹³C₆-Lys feed for 8 weeks, after which they were mated with wild-type FVB/N males who were kept on a standard diet. The F0 mother was further fed with the ¹³C₆-Lys feed and the F1 pups born from this mating were fed with the ¹³C₆-Lys feed until weaning. Thereafter, only the F1 offspring was further fed with the ¹³C₆-Lys feed. Upon reaching sexual maturity, the F1 ¹³C₆-Lys-fed females were mated with wild-type FVB/N males who were kept on a standard diet. The pups of the F2 generation were the first generation of mice used for experiments. Females of the F1 generation were continuously mated to produce experimental animals. Fully labelled ¹³C₆-Lys breeders were periodically substituted by the pups from the fully labelled offspring, and therefore, the F2 labelling efficiency represents the minimal labelling efficiency, as all subsequent generations had a larger fraction of ¹³C₆-Lys due to longer ¹³C₆-Lys feeding time. The percentage of ¹³C₆-Lys incorporation in blood samples in each generation is indicated in the top left corner of female mouse pictograms. (b-e) Histograms show ¹³C₆-Lys incorporation rates in (b) F0 blood sample for 1,653 lysine-containing peptides, (c) in F1 blood sample for 206 lysine-containing peptides, (d) in F1 oocyte samples for 137 lysine-containing peptides, and (e) in F2 blood sample for 162 lysine-containing peptides. Median ¹³C₆-Lys incorporation rates are indicated. Heavy isotope (¹³C₆) incorporation rates were calculated from available lysine-containing peptides showing heavy (¹³C₆) to light (¹²C₆) lysine ratios. Source numerical data are available in source data. Source data

Extended Data Fig. 2. Sample preparation for MS DIA or DDA analysis and protein turnover in oocytes.

(a) Schematic representation of sample preparation for MS DIA or DDA analysis. (b) Box plot showing the volume (µm³) of oocytes in fixed primordial follicles and at germinal vesicle stage from live and fixed samples. Boxplots indicate median, 1st quartile, 3rd quartile, as well as minimum and maximum after outlier removal. Number of oocytes indicated in parenthesis. (c) Venn diagram of ¹²C₆-Lys containing proteins and ¹³C₆-Lys containing proteins detected in mouse oocytes. (d) Rank plot showing inferred remaining fraction of ¹³C₆-Lys labelled proteins (%H) in oocytes for each quantified protein (individual proteins denoted as black circles). Selected examples are highlighted in red. (e) Bar plot showing number of proteins with %H of 0–1, 1–50, 50–90, or 90–100%. Source numerical data are available in source data. Source data

Extended Data Fig. 3. Modelling protein turnover in the ovary.

(a) Overview of the protein turnover modelling approach described in supplemental materials. (b) Graphical illustration of the 2Lys-peptide model to determine the fraction of free ¹³C₆-Lys during ovarian ageing. (c) Estimated fraction of free ¹³C₆-Lys for the 2Lys-peptide model for the chase (¹²C₆-Lys) from birth and from weaning experiments. Solid lines indicate median; shaded areas indicate 5% and 95% confidence ranges. (d-e) Experimental data and model fits of 2Lys-peptide model (d) and total ovary protein fold change (e) over chase time for the chase (¹²C₆-Lys) from birth and from weaning experiments. Dots indicate experimental data; lines and shaded areas indicate median and confidence ranges of model fits. (f) Change of total protein amount in the ovary over mouse age. Total protein amount was determined from BCA measurements and compared to changes in the volume of the ovary. Boxplots indicate the estimated fold change in total protein amount. Boxplots indicate median, 1st quartile, 3rd quartile, as well as minimum and maximum after outlier removal over 3,078 modelled proteins. Even though the volume of the ovary was not considered in the protein turnover model fitting, the estimated fold changes are in good agreement with experimentally measured total protein amount fold changes. Source numerical data are available in source data. Source data

Extended Data Fig. 4. Comparison of the H_1/2 values estimated from different models.

(a-d) Comparison of estimated H_1/2 values resulting from the protein turnover model considering free ¹³C₆-Lys pool with 2Lys-peptide based model (a) and the ‘classical’ protein turnover model not allowing reincorporation of ¹³C₆-Lys into newly synthesized proteins considering all chase time points (b), or only chase time points larger than 3 weeks (c) or 6 weeks (d), respectively. Dots indicate medians, grey lines indicate confidence ranges. All proteins with estimated H_1/2 values < 100 days are indicated as orange dots. (e-g) Comparison of previously published H_1/2 values (Rolfs et al.) and H_1/2 values calculated for the same published datasets using the protein turnover model developed in this manuscript for ovaries. Shown are comparisons for liver (e), cartilage (f) and skeletal muscles (g). In (a-g) test for association between paired samples using Spearman correlation coefficient (C) was performed with p-values (p) estimated using algorithm AS 89. p < 10⁻¹⁶ indicates approximated p-values. (h) Rank plot showing median (red dots) and confidence ranges (pink lines) of estimated H_1/2 values for all modelled proteins in ovaries. (i) Over-representation analysis of proteins with rapid turnover in ovaries. Shown are the percentage of genes detected in the gene set with their respective adjusted p-values for the most prominent pathways. The p-values are based on the hypergeometric test and have been adjusted for multiple hypothesis testing with Benjamini–Hochberg (BH) procedure. For the complete list of enriched pathways and their corresponding exact p-values, see Supplementary Table 4. Source numerical data are available in source data. Source data

Extended Data Fig. 5. Comparison of the H_1/2 values of the different proteins in the ovary with their corresponding H_1/2 values in other organs, and examples of long-lived proteins in the ovary.

(a) Distribution of H_1/2 determined for various mouse and rat tissues across three studies (Fornasiero et al.; Kluever et al.; Rolfs et al.) compared to the estimated H_1/2 distribution in mouse ovary samples in this study. Red line indicates the median H_1/2 values in the ovary, while the grey line indicates a H_1/2 value of 100. (b) Long-lived members of protein complexes. Shown are protein complexes with long-lived proteins as violin plots. Dots indicate individual proteins. Complexes are sorted by median complex H_1/2 values. (c) Fraction of long-lived oocyte proteins in the high protein longevity cluster of the ovary. (d-f) Example MS1 intensities over time in log10 scale for ¹³C₆-Lys- (black and purple) and ¹²C₆-Lys- (grey and pink) labelled peptides for two pulse lengths (pulse with ¹³C₆-Lys until birth and pulse with ¹³C₆-Lys until weaning) for glutamate dehydrogenase 1, mitochondrial (d); histone deacetylase 1 (e); and Myosin-11 (f). Circles indicate experimental data points for three biological replicates, dashed lines are only for visual aid. Source numerical data are available in source data. Source data

Extended Data Fig. 6. Changes in enrichment scores of different pathways throughout ovarian development and aging.

(a) Schematic overview of experimental design. Ovaries were collected from female mice at 8 time points (1 day, 1, 2, 3, 5, 9, 12 and 50-week-old mice; three animals per time point analysed). Samples were processed for DIA-MS. (b) Gene set enrichment analysis (GSEA) revealed up- and down-regulated pathways over mouse age. For each time point (1 day, 1, 2, 5, 9, 12 and 50 weeks) normalized DIA-MS data for 8,890 detected proteins were subject to GSEA. Normalized enrichment scores for each pathway and time point are shown as heatmap upon hierarchical clustering (dendrogram shown on the left). Each term is significantly enriched at least in one time point. Source numerical data are available in source data. Source data

Extended Data Fig. 7. Changes in abundance of different proteins belonging to different gene sets throughout ovarian development and aging.

(a-d) Normalized DIA-MS intensity data showing the abundance changes over time of the proteins in the selected pathways significantly enriched in the gene-set enrichment analysis (GSEA) of the DIA-MS data: aerobic respiration (a), oxidative phosphorylation (b), regulation of acute inflammatory response (c), and regulation of humoral immune response (d). Source numerical data are available in source data. Source data

Extended Data Fig. 8. Distribution of ovary proteins in the different protein longevity clusters and protein abundance clusters, and abundance changes over time of proteins in the mTOR signalling pathway.

(a) Percentage of ovary proteins that are detected across the three protein longevity clusters, as well as in one of the six protein abundance clusters derived from the DIA-MS data. (b) Heatmaps showing protein abundance change for low, intermediate, and high protein longevity clusters with assignment to protein abundance clusters. (c) Normalized DIA-MS intensity data showing the abundance changes over time of proteins in the mTOR signalling pathway. Source numerical data are available in source data. Source data