2.2. Oocytes undergo aging‐associated transcriptome changes with increased heterogeneity at a late stage of reproductive life

We then separately analyzed datasets from oocytes and from their surrounding cumulus cells using principal component analysis (PCA) (Figure ​(Figure1c).1c). PCA of cumulus cells detected significant changes from young to middle mice along the PC2 axis, while no significant difference was detected from middle to old mice (Figure ​(Figure1c,1c, Cumulus). In contrast, PCA of oocytes detected no significant changes from young to middle mice along the PC1 axis that showed a significant difference from middle to old mice (Figure ​(Figure1c,1c, Oocyte). These results suggest that aging‐associated global changes in the transcriptome occur during relatively early life stages in cumulus cells, while oocytes largely maintain their transcriptome until they undergo global changes at a later life stage. Consistent with this idea, we detected a substantial number of differentially expressed genes (DEGs) in cumulus cells between young and middle mice (Middle down‐regulated: n = 225; up‐regulated: n = 190), as well as between young and old mice (Aged down‐regulated: n = 520; up‐regulated: n = 350) (Figure ​(Figure1d,1d, Cumulus; Figure S3a). In contrast, the number of DEGs in oocytes was small between young and middle mice (Middle down‐regulated: n = 3; Middle up‐regulated: n = 10), but it dramatically increased between young and old mice (Aged down‐regulated: n = 1,201; up‐regulated: n = 1,663) (Figure ​(Figure1d,1d, Oocyte; Figure S3b). Pairwise analysis of oocytes and their surrounding cumulus cells showed that the PC2 values of oocytes were not correlated with the PC1 values of cumulus cells (Figure ​(Figure1e),1e), suggesting that aging‐associated transcriptome changes are poorly coordinated between the oocyte and its surrounding cumulus cells in individual follicles. These results were largely reproduced by experiments with a second set of samples, which included oocytes and their cumulus cells from mice at 2, 4, 9, and 14 months old (Figure S3c–e). We also examined samples from 2‐month‐old mice without hyperovulation, and found that hyperovulation had a substantial impact on the transcriptome of cumulus cells but not on that of oocytes (Figure S3f). Collectively, these results indicate that oocytes undergo aging‐associated transcriptome changes at a late stage of reproductive life, in an uncoordinated fashion with their surrounding cumulus cells.

2.3. Distinct characteristics of aging‐associated genes between oocytes and cumulus cells

Gene ontology (GO) analysis of DEGs revealed distinct characteristics of aging‐associated changes in transcriptomes between oocytes and cumulus cells. Comparison of the datasets of old mice to those of young mice showed that in cumulus cells, genes involved in transcription regulation were predominantly down‐regulated with aging, while genes involved in DNA damage/repair and cell cycle regulation were up‐regulated (Figure ​(Figure2a).2a). In contrast, in oocytes, genes involved in DNA damage response and in utero embryonic development were down‐regulated with aging, while genes related to chromatin regulation, such as nucleosome assembly and epigenetic modification, were markedly up‐regulated (Figure ​(Figure2b).2b). Moreover, genes involved in cell cycle regulation were aberrantly regulated in the oocytes of old mice, consistent with previous reports (Hamatani et al., 2004; Pan et al., 2008). DEGs detected in the oocytes of middle mice included genes involved in oxidation–reduction processes and steroid hormone synthesis (Figure S3g).

Open in a separate window

FIGURE 2

Distinct characteristics of aging‐associated genes between oocytes and cumulus cells. (a) Representative GO terms of down‐regulated (top) and up‐regulated (bottom) enrichment between young and old cumulus cells. (b) Representative GO terms of down‐regulated (top) and up‐regulated (bottom) enrichment between young and old oocytes. (c) Regulatory networks visualizing potential key transcriptional regulators in cumulus cells associated with aging. Down‐regulated and up‐regulated top 10 nodes are colored dark blue and dark red, respectively. (d) Regulatory networks visualizing potential key transcriptional regulators in oocytes associated with aging. Down‐regulated and up‐regulated top 10 nodes were colored in dark blue and dark red, respectively. The threshold used for regulator‐target connection in oocytes was lower than in cumulus cells

To identify regulators of the aging‐associated changes in gene expression, we constructed transcriptional regulatory networks of core transcriptional regulators and their target genes (Huynh‐Thu et al., 2010). That analysis revealed a core hub of aging‐up‐regulated genes, including Luzp4, Gm15093, Tia1, Gm15127, and Gm15109, and a core hub of aging‐down‐regulated genes, including Fos, Fosb, Sik1, Zfp36, and Jun, in cumulus cells (Figure ​(Figure2c).2c). However, in oocytes, the connections of the network were relatively weak, although lowering the detection thresholds allowed us to identify a potential hub that regulates the expression of genes associated with aging (Figure ​(Figure2d).2d). Together, our results reveal the regulatory networks of gene expression programs underlying the aging of oocytes and their surrounding cumulus cells.

2.4. Pronounced effects of calorie restriction on transcriptome profiles in oocytes

We next analyzed whether transcriptome profiles in oocytes and their surrounding cumulus cells are affected by CR. We used an adult‐onset 40% feeding regimen that was reported to prevent aging‐associated oocyte aneuploidy (Selesniemi et al., 2011; Turturro et al., 1999), and obtained oocytes and their surrounding cumulus cells from 9‐month‐old (middle) mice (n = 27 oocytes from 3 mice) (Figure ​(Figure3a).3a). t‐SNE analysis demonstrated that oocytes from CR mice formed a cluster that was distinct from oocytes from ad libitum‐fed (AL) mice at any age, while the cluster of cumulus cells from CR mice was positioned adjacent to that of cumulus cells from AL mice (Figure ​(Figure3b).3b). While PCA of cumulus cells showed no clear separation between CR‐middle and AL‐middle mice along an axis that appeared to represent aging‐associated effects (Figure ​(Figure3c,3c, Cumulus, along the PC2 axis), CR tended to shift oocytes in the opposite direction to aging‐associated effects along the PC1 axis (Figure ​(Figure3c,3c, Oocyte, compare CR‐middle with AL‐middle). Moreover, oocytes showed a substantially greater number of DEGs between CR‐middle and AL‐middle mice (CR down‐regulated: n = 3,488, up‐regulated: n = 3,630), compared to cumulus cells (CR down‐regulated: n = 416, up‐regulated: n = 593; Figure ​Figure3d).3d). These results suggest that oocytes have a greater transcriptome response to CR than cumulus cells.

Open in a separate window

FIGURE 3

Pronounced effects of CR on transcriptome profiles in oocytes. (a) Design overview of CR. Note that the AL datasets used in this figure are identical to those used in Figure ​Figure1.1. (b) t‐SNE plot showing the distribution of single oocytes and their surrounding cumulus cells. Circles and triangles represent cumulus cells and oocytes, respectively. The pairs of oocytes and their surrounding cumulus cells are connected with a line. (c) PCA plot of all expressed genes in cumulus cells (left) and in oocytes (right), respectively. **p < 0.01 and ***p < 0.001 (pairwise t test with holm correction). (d) The number of DEGs between CR and AL in oocytes at middle age. (e) Representative GO terms of down‐regulated (top) and up‐regulated (bottom) enrichment between CR and AL oocytes

The GO analysis revealed that CR preferentially down‐regulated genes involved in translation, mitochondrial translation, and transcription in oocytes (Figure ​(Figure3e).3e). Interestingly, GO terms for CR‐up‐regulated genes in oocytes included cell cycle/division, DNA damage/repair, nuclear division, and chromosome segregation (Figure ​(Figure3e),3e), while their surrounding cumulus cells had up‐regulated genes involved in DNA replication and cell cycle (Figure S4a). Up‐regulated genes involved in chromosome segregation in oocytes included those encoding proteins regulating kinetochore‐microtubule attachment (Knl1, CENP‐E/H/W/F, Spc25, Mis12, Ndc80, Ska3, Bub1, HJURP, and Meikin), spindle assembly (Kif2c, Kif11, and Nek2), and chromosome cohesion (Stag1/SA1, Sgo1, Rad21, and Esco2). Construction of transcriptional regulatory networks suggested core regulators of CR‐associated changes in cumulus cells and oocytes (Figure S4b and S4c). Overall, our analysis demonstrates that CR substantially affects oocytes with transcriptome changes including the elevated transcription of genes involved in chromosome segregation.

4.2. Oocyte collection

Mice were injected for hyperovulation with equine chorionic gonadotropin (eCG, ASKA Pharmaceutical) or CARD HyperOva (KYUDO). Ovaries were collected 48 h after injection and put into M2 medium containing 200 nM 3‐isobutyl‐1‐methyl‐xanthine (IBMX, Sigma) at 37°C. Fully grown follicles were dissected to obtain cumulus–oocyte complexes. From each follicle, an oocyte and a pool of its surrounding cumulus cells were isolated with gentle pipetting. Oocytes at the GV stage and their surrounding cumulus cells were retained for further experiments. After washed in 0.1% PBA in PBS, individual oocytes and their surrounding cumulus cells were manually collected in 10 μl cell lysis buffer [10 U RNasein plus (Promega), 10% RealTime ready Cell Lysis Buffer (Roche), 0.3% NP40 (Thermo Fisher Scientific), and RNase‐free water (TAKARA)] in a 0.2 mL single tube. The collected cell samples were lysed and stored at −80 until downstream preparation.

4.3. RamDA‐seq

The sample single‐tubes were assembled into a 96‐well plate format by a tube tray and holder (Nalge Nunc). The cell lysates were dissolved at 10°C and were agitated for 2 min at 2,000 rpm using a ThermoMixer C at 4°C. The dissolved cell lysates were denatured at 70°C for 90 s. For oocyte samples, 10 µl genomic DNA digestion mix (0.5× PrimeScript Buffer, 2 U DNase I Amplification Grade, 1 µl 500,000‐fold diluted ERCC RNA Spike‐In Mix I (Thermo Fisher Scientific) in RNase‐free water) was added to 10 µl of each denatured sample, which was then incubated at 30°C for 5 min and held at 4°C. Two µl of the digested products was dispensed into 96‐well PCR plates, and 1 µl reverse transcription with random displacement amplification (RT‐RamDA) mix (2.5× PrimeScript Buffer, 0.6 pmol oligo(dT)18 (Thermo Fisher Scientific), 8 pmol 1st‐NSRs 100 ng T4 gene 32 protein (New England Biolabs), and 3× PrimeScript enzyme mix (TAKARA) in RNase‐free water) were added. The samples were incubated at 25°C for 10 min, at 30°C for 10 min, at 37°C for 30 min, at 50°C for 5 min, and then at 94°C for 5 min. For cumulus cell samples, the denatured cell lysate samples were added to 10 µl genomic DNA digestion mix including 1 µl 100,000‐fold diluted ERCC RNA Spike‐In Mix I and incubated under the same conditions as the oocyte samples. After genomic DNA digestion, 6 µl of the products was dispensed into 96‐well PCR plates and 9 µl RNase‐free water was added to adjust the RNA amount to the oocyte samples, and 1 µl of the diluted products was re‐dispensed into new 96‐well PCR plates. The samples were added to 2 µl component‐adjusting RT‐RamDA mix (4% RealTime ready Cell Lysis Buffer, 0.12% NP40, 1.45× PrimeScript Buffer, 0.16 U DNase I Amplification Grade, 0.6 pmol oligo(dT)18, 8 pmol 1st‐NSRs 100 ng T4 gene 32 protein and 1.5× PrimeScript enzyme mix in RNase‐free water). The samples were incubated under the same conditions as the oocyte samples. After RT‐RamDA, the samples were added to 2 µl second‐strand synthesis mix [2.5× NEB buffer 2 (New England Biolabs), 0.625 mM each dNTP Mixture (TaKaRa), 40 pmol 2nd‐NSRs and 0.75 U Klenow Fragment (3’‐5’ exo‐) (New England Biolabs) in RNase‐free water] and incubated at 16 for 60 min, then at 70 for 10 min. The second stranded cDNAs were purified using 15 μl AMPure XP beads (Beckman Coulter) and a handmade 96‐well magnetic stand for low volumes. The double‐stranded cDNAs were directly eluted with 3.75 μl 1.33 × diluted Tagment DNA Buffer (Illumina) and mixed well using a vortex mixer and pipetting. Sequencing libraries were generated using the Tn5 tagmentation‐based method with 1/4 volumes of the Nextera XT DNA Library Preparation Kit (Illumina) according to the manufacturer's protocol. Fourteen cycles of PCR were applied for the library DNA. After PCR, sequencing library DNA was purified using 1.2× the volume of AMPure XP beads and eluted into 24 μl TE buffer. RamDA‐seq libraries were quantified and evaluated using a MultiNA DNA‐12000 kit (Shimadzu). The sequencing was performed with Illumina NextSeq 500 High Output v2 (75 cycles; single‐end reads), and 74 individual oocytes and 72 cumulus cell samples were sequenced. Additionally, as a reproducibility experiment, 127 individual oocytes and 127 cumulus cell samples were prepared and sequenced in a same manner.

4.4. Sequence data analysis

Hisat2 v2.0.5 (Kim et al., 2015) was used to map the reads to the mouse genome (GRCm38) after trimming adaptor sequences and low‐quality bases using Fastq‐mcf v1.04.807 (Aronesty, 2011). The resulting binary alignment/map (BAM) files were sorted using samtools (Li et al., 2009), and final mapping qualities were assessed using RSeQC v2.6.4 (Wang et al., 2012). The featureCounts v1.5.1 tool of the Subread package (Liao et al., 2013) was used to generate counts of reads multi‐mapped to annotated genes using the GENCODE (vM15) annotation gtf file. We evaluated the quality control metrics to filter out low‐quality samples but our data did not contain a low‐quality sample. We manually removed one pair of samples because following data analysis indicated that the pair did not include oocytes. Using the remaining data derived from 73 individual oocytes and 71 cumulus cell samples, we performed downstream analysis.

Based on the Log₁₀ transformed RPKM values of the expression, we visualized t‐SNE plots using the Rtsne package in R (Krijthe, 2015). PCA was also conducted using R (R Core Team, 2020).

DEGs (false discovery rate, FDR <0.05) were identified with the edgeR package (Robinson et al., 2009) in R, which normalizes library sizes with the trimmed mean of M values (TMM) method, using the dataset of genes expressed in at least 5 samples. Aging‐ or CR‐associated GO were identified using DAVID v6.8 (GO term Biological Process Direct) (Huang et al., 2008).

To identify potential key transcriptional regulators, we performed regulatory network analysis with a GENIE3 package, a random forest method that computes the links between each gene and all other genes (Huynh‐Thu et al., 2010). We used raw read counts of aging‐ or dietary‐associated DEGs (>0.5 log₂‐fold coverage, FDR <0.05) as input. Only the transcriptional regulator‐target connected with above thresholds (cumulus: 0.030, oocyte: 0.015) were retained and used for the network analysis. The resulting networks were visualized using CytoScape 3.8 (Shannon et al., 2003).

4.5. Immunostaining

Oocytes that underwent nuclear envelope breakdown (NEBD) within 60 min after the induction of meiotic resumption were collected after 14–18hr after NEBD. The oocytes were fixed with a fixation buffer (1.6% paraformaldehyde in 100 mM PIPES (pH7.0), 1 mM MgCl2, 0.1% Triton‐X100) at room temperature for 30 min. After washing 4 times, the oocytes were incubated in PBT (PBS +0.1% Triton‐X100) at 4°C overnight. The oocytes were blocked with 3% BSA‐PBT at room temperature for 1 h and then incubated with human anti‐centromere antibodies (ACA, Antibodies Incorporated, 15–234) at 1:200 in 3% BSA‐PBT at 4°C overnight. The oocytes were washed four times in 3% BSA‐PBT and incubated with secondary antibodies in 3% BSA‐PBT at room temperature for 2 h. The secondary antibody used was Alexa Fluor 488 goat anti‐rabbit IgG (H+L) highly cross‐adsorbed or anti‐human IgG (H+L) cross‐adsorbed (Molecular Probes). DNA was counterstained with 5 µg/ml Hoechst33342 (Invitrogen). For imaging, a customized Zeiss LSM780 confocal microscope equipped with a 40×C‐Apochromat 1.2NA water immersion objective lens (Carl Zeiss) was used. We recorded z‐confocal sections (every 0.5 μm) of 512 × 512 pixel xy images to capture all chromosomes.

For cohesin immunostaining, we used a protocol modified from a previous study (Lee et al., 2011). Fully grown oocytes were incubated in IBMX‐free M2 medium for 5 h at 37°C. Oocytes at metaphase I were microinjected with 4 pl purified Rec8 antibody (0.4 mg/ml) (Ding et al., 2018) and incubated for 8 min at 37°C before fixation.

We thank the animal facilities of RIKEN Kobe. N.T. was supported by the RIKEN JRA program. We also thank A. Matsushima (RIKEN BDR) for assistance with the infrastructure for the data analysis. This work was supported by research grants MEXT/JSPS KAKENHI JP16H06161/JP18H05549/JP21H02407 to T.S.K., JST CREST (JPMJCR16G3 and JPMJCR1926) to I.N., JP18J00928 to T.M., and by RIKEN BDR and All‐RIKEN Research Project “Integrated life science research to challenge super aging society.”