The combination of DNA methylome and transcriptome revealed the intergenerational inheritance on the influence of advanced maternal age

2.1 AMA-related abnormal changes in the parental DNA methylome

To investigate the influence of AMA on the DNA methylome of parents and offspring, we applied reduced-representation bisulphite sequencing (RRBS) to umbilical cord blood (UCB) samples and corresponding parental peripheral blood (PPB) samples from nuclear families. Those samples were further classified into AMA and Young groups based on maternal age (AMA: maternal age of 36–43 years; Young: maternal age of 22–29 years) (Figure 1A). There were no significant differences in maternal BMI or any clinical features of offspring between the AMA and Young groups (see Table S1 for more details). The genome-wide copy number variations (CNVs) were normal in all samples (Figure S1A). The genome was then split into 200 bp bins, and only bins containing more than three CpG sites and conserved in 80% of the samples in each group were used for downstream analysis (Figure S1B).

Principal component analysis (PCA) and hierarchical clustering analysis showed no obvious separation of the AMA and Young groups in either maternal or paternal samples (Figure 1B and Figure S2A,B). The global patterns of DNA methylation levels were also similar among samples (Figure S2C,D). The mean value was approximately 75% in each group, and it was slightly lower in the AMA group than in the Young group for both maternal and paternal samples, but the difference did not reach statistical significance (Figure 1C,D). Meanwhile, the overall DNA methylation level around the gene body was similar among the four groups (Figure 1D). Interestingly, it was notable that the average methylation levels of all four kinds of retrotransposons (including LTR, LINE, SINE and SVA) were significantly lower in AMA-mother than in Young-mother, while no differences were observed in the father group (Figure 1E and Figure S2F).

We next performed intergroup comparisons to elucidate the definitive influence of AMA, and identified 516 and 464 AMA-related differentially methylated regions (DMRs) in the mother and father groups, respectively (|df|> = 15% and q value < .05; for mothers: hyper DMRs (45.7%), n = 236; hypo DMRs (54.3%), n = 280; for fathers: hyper DMRs (44.4%), n = 206; hypo DMRs (55.6%), n = 258) (Figure 1F and Figure S3A–F; see Table S3 for more details). These DMRs were broadly distributed on the genome scale and primarily located outside of the promoter regions (Figure S3J,K). Further functional genomic annotation revealed that a large percentage of DMRs were located in SINEs (37.4%–43.4%), LTRs (11.4%–19.9%), LINEs (10.7%–16.1%), as well as CpG islands (4.7%–7.8%) (Figure S4A). However, mother-DMRs and father-DMRs were quite distinct in terms of their exact genomic positions, with only 14 hyper-DMRs and 19 hypo-DMRs in common (Figure 1G). Searching for the nearest genes of DMRs in both longevity and aging databases^{34, 35} revealed that 44 genes in the mother group and 20 genes in the father group had been reported to be associated with aging (Figure 1H,I). The nearest genes of two mother-DMRs, RPS6KA5 and ETS2, were previously defined as senescence-associated secretory phenotype (SASP) genes³⁶ (Figure 1J), while no overlap existed between the nearest genes of father-DMRs and SASP genes (Figure 1J). Gene Ontology (GO) enrichment analysis revealed that the nearest genes of mother-DMRs were enriched in processes involving organ morphogenesis, histone modification, phospholipid metabolism, and leukocyte differentiation. However, the nearest genes of father-DMRs were highly enriched in wound healing, cell-substrate adhesion, and calcium ion transmembrane transport (Figure S4B). Both mother-DMRs and father-DMRs were enriched in the processes associated with neuron differentiation (nearest genes: CCND1, SPOCK1), protein homotetramerization (nearest gene: USP16) and glycosylation (nearest genes: GAL3ST1) (Figure 1G and Figure S4B; see Table S4 for details). Together, the above results indicated that aging-related alterations in the parental DNA methylomes were present in AMA pregnancy and were more conspicuous in the maternal part.

2.2 AMA induced DNA methylation changes in offspring

Then, we analysed the DNA methylomes of UCB samples to evaluate the epigenetic changes in AMA offspring. Similar to what was observed in PPB, PCA and hierarchical clustering analyses showed no obvious separation of the AMA and Young groups (Figure 2A and Figure S2A). The mean value showed no significant intergroup difference (Young: 75.31%; AMA: 74.99%; p value = .44; Figure 2B), and the average DNA methylation pattern on the genome-wide scale or specifically around gene bodies was comparable between the AMA and Young groups. (Figure 2C and Figure S2C,D,G). Nevertheless, the average DNA methylation levels of CpG islands significantly decreased in the AMA group (Young: 26.52%; AMA: 26.50%; p value = .035; Figure 2D). Intergroup comparison analysis further identified 182 hyper DMRs and 269 hypo DMRs (|df|> = 15% and q value <.05) (Figure 2E, Figure S3G–I), which were broadly distributed on all chromosomes and mainly in non-promoter regions (Figure 2F and Figure S3J). Once again, we observed a larger percentage of hypo DMRs than hyper DMRs (40.4% vs. 59.6%, Figure S3H) and a high proportion of hyper/hypo DMRs located in SINE (36.8% and 43.4%), LTR (11.5% and 19.2%), LINE (15.2% and 18.1%) (Figure S4A), as described for the parental sample.

Among the nearest genes of offspring-DMRs, 25 genes were listed in the longevity or aging databases (Figure 2G). Remarkably, the SASP gene RPS6KA5 was also observed among them, with a nearby hypo DMR located in the intron 1 region (chr14: 91033401–91033600) (Figure 2H). As shown in Figure 1J, the same region was also identified as AMA-related hypo DMR in the mother group. GO analysis suggested that disturbance of those DMRs might influence various processes associated with vesicle-mediated transport, autophagy, electron transport chain, eye development, and so on (Figure S4C). Importantly, quite a few GO terms were also found in the enrichment analysis of the mother group, including processes involving neuron differentiation, regulation of GTPase signal transduction, metabolism, and histone modification (Figure S4B,C). Integrated analysis between parental DMRs and offspring-DMRs further uncovered overlapping common DMRs that shared the same trend between offspring and parents (MK_common DMRs: n = 64; FK_common DMRs: n = 58, FMK_common DMRs: n = 13) (Figure 2I). Taken together, our findings indicated that AMA would also disturb the DNA methylome of the offspring. These changes bore many remarkable similarities to those in the maternal DNA methylome.

2.3 AMA altered the maternal and neonatal transcriptomes

In addition to examining the DNA methylome, we also performed mRNA-seq of maternal peripheral blood and UCB samples to explore the influence of AMA on gene expression (Table S1 and Figure S5A). Similar to the observations in the DNA methylome, correlational analysis and PCA based on the transcriptomes of all samples showed obvious separation of the mother and offspring groups but only slight separation between the AMA and Young groups (Figure S5B,C). Intergroup comparison analysis identified 741 and 3157 differentially expressed genes (DEGs) in the mothers and offspring, respectively (Figure 3A,B, Figure S5D,E; p < .05 and fold change > = 1.5; for mothers: Up-regulated DEGs: n = 386 and Down-regulated DEGs: n = 355; for offspring: Up-regulated DEGs: n = 1128 and Down-regulated DEGs: n = 2029; see Table S5 for details). A total of 206 DEGs in the offspring group and 43 DEGs in the mother group were recorded in the longevity or aging databases (Figure S5E). GO analysis showed that both offspring-DEGs and mother-DEGs were involved in neutrophil activation, mitotic nuclear division, and neuron development. Offspring-DEGs also focused on those processes related to embryo and placenta development, protein modification, regulation of RNA processing and translation, and mitochondrial function, while mother-DEGs were highly enriched in mesenchyme development, regulation of inflammatory response, chromosome segregation, and cell−cell adhesion (Figure S5F, see Table S6 for details).

A total of 109 common AMA-DEGs shared the same trend between the mother and offspring groups (Figure 3C, common Up-regulated DEGs: n = 39 and common down-regulated DEGs: n = 70). GMFG, as a SASP gene within this set, was upregulated in the AMA group (Figure 3C). GO enrichment analysis suggested that these DEGs might be involved in the processes of cytokine production, cell mitosis, protein modification, neutrophil activation, defence response to gram−negative bacterium and the NF–KB signalling pathway (Figure 3D; see Table S7 for details). Overall, the above results revealed that the fluctuation in the transcriptome reflected the aging-like effect of AMA on both offspring and mothers, which shared many similarities with DMRs in related biological processes. The results also presented considerable similarities in the transcriptional alterations between the mother and offspring.

We then performed an integrated analysis between DEGs and DMRs for either mother or offspring to identify DEGs potentially rooted in the DNA methylation change of nearby DMRs (offspring-DEGs, n = 58; mother-DEGs, n = 13) (Figure 3E–G). Specifically, SLC28A3, a gene encoding a nucleoside transporter, was significantly downregulated in both the offspring and mothers of the AMA group, along with an increased DNA methylation level in DMRs located in intron 1(Figure 3E–H). The reduced expression of SLC28A3 was further validated by real-time fluorescence quantitative polymerase chain reaction (qRT-PCR) (Figure 3I). This result suggested the highly interconnected changes in the transcriptome and DNA methylome in the AMA group.

2.4 Specific alterations in AMA offspring presented intergenerational correlation

Although common changes between parents and offspring have been observed in both the transcriptome and DNA methylome, how close the connection between the changes in the offspring and parents is remains uncertain. In fact, the CD24 (CD24 molecule), one of the common DEGs and encoding a glycosylphosphatidylinositol-linked cell surface protein tightly correlated with cell pluripotency,³⁷ was not only significantly downregulated in both AMA groups, but also showed strong linear corrections between paired mother and offspring (Figure S8A), which were validated using qRT-PCR, too (Figure 4A,B). This observation hinted that part of alterations in offspring might be directly correlated with the parental changes induced by AMA. Thus, benefiting from the definite parentage in our cohort, we calculated the Spearman correlation coefficients between parents and offspring for both offspring-DMRs and offspring-DEGs to identify the intergenerationally correlated DMRs (R > = 0.600 and p value < 0.05; Figures S6 and S7) and DEGs (R > = 0.600 and p value < .05; Figure S8A). Among these, 18 and 14 DMRs, respectively belonged to common mother-offspring and father-offspring DMRs shown in Figure 2I (23% and 19%), while 48 DEGs were previously identified as common DEGs showed in Figure 3C (44%; Up-regulated DEGs: n = 11; Down-regulated DEGs: n = 37; Figure 4C). This result suggested that a considerable proportion of AMA-related DMRs and DEGs observed in offspring might be directly inherited from parents.

The parental characters were passed on to offspring mainly through the gamete and embryo, and previous studies have outlined highly diverse molecular dynamics for different genomic elements and various genes during gametogenesis and pre-implantation development.^{12, 38-40} To investigate the DNA methylation pattern of those intergenerationally correlated DMRs during the epigenetic reprogramming process, we here reanalysed the published single-cell chromatin overall omic-scale landscape sequencing (scCOOL-seq) data of human pre-implantation embryonic development⁴¹ and depicted the dynamic DNA methylome patterns for 91 intergenerationally correlated DMRs detected in the data (Figure 4D and Table S8). Those DMRs were further classified into seven clusters using hierarchical clustering, and the dynamic pattern in each cluster showed strong development-stage specificity (MeE1-MeE7, Figure 4D). It indicated that DNA methylation patterns of those DMRs were strictly regulated by demethylation and remethylation mechanism during embryonic development, rather than just keep out of the epigenetic reprogramming process.

Meanwhile, we also downloaded single-cell transcriptional data of human oogenesis and pre-implantation embryos to profile the expression dynamics of 48 intergenerationally correlated common DEGs.^{38, 40} Among the 10 patterns defined in follicle generation, genes with the RO2 pattern drastically increased from secondary follicle to antral follicle, a phenomenon that might play important role in follicular lumen formation. Genes with the RO4 and RO5 patterns showed the opposite pattern and striking changes in the preovulatory stage. In particular, Suzhen Yuan and his colleagues found that the upregulation of EGR1 (RO4 pattern) participated in granulosa cell apoptosis and follicle atresia during ovarian aging.⁴² In addition, RO3, RO6, and RO9 patterns presented drastic fluctuations in gene expression from the primordial follicle stage to the primary follicle stage (Figure 4E). Six distinct clusters of gene expression patterns were identified in preimplantation embryos (Figure 4F). The dramatical increase in expression after the morula stage implied that genes with the RE1 pattern might be associated with the cell differentiation of blastocyst. The specific high expression after the 4-cell stage evinced the character of zygotic genes after major ZGA for genes with the RE2 and RE6 patterns. In contrast that of RE2 gene, the high expression levels of genes with the RE3 pattern decreased sharply after the 4-cell stage, implying that those genes might be potential maternal-effect genes that were largely degraded during the maternal-to-zygotic transition (MZT). It is worth noting that the expression of CD24 both increased sharply during the transition from the antral follicle to the preovulatory follicle during folliculogenesis, echoing the early report about its important role in the regulation of ovulation.⁴³ During the development of early embryos, the expression of CD24 was increased in the periods of ZGA and the transition from morula to blastocyst, (Figure 4F), consistent with its high expression in the following villus trophoblast cell and the key roles in mediating immune tolerance at the foetal-maternal interface.⁴⁴ Together, these results revealed that the DNA methylation patterns of intergenerationally correlated DMRs and the expression patterns of those intergenerationally correlated DEGs varied during oocyte maturation and preimplantation embryonic development, and suggested the potential impact of AMA on multiple key processes in these two periods.

2.5 AMA-related transcriptional alterations existed in AMA oocyte and blastocyst

Several previous studies have discussed the influence of AMA on human MII oocytes or embryos after cryopreservation.^45-47 We then compared offspring-DEGs and mother-DEGs with two published blastocyst-DEG lists^{26, 46} and one MII oocyte-DEG list,⁴⁷ and revealed a series of overlapping DEGs (Figure 5A, see Table S9 for details). Among common AMA-DEGs between mother and offspring groups, two upregulated genes and two downregulated genes were observed in the oocyte, while 18 genes were downregulated in the blastocyst. Meanwhile, 11 of those 22 genes (HTRA3, FCGR1A, BST2, MED12L, SLC12A1, MLNR, CNTNAP1, NPIPB4, CAPN3, DNAH10, and LOC101929076) were previously identified as mother-offspring intergenerationally correlated DEGs (Figure 5A and Figure S8A). In particular, the HTRA3 (high temperature requirement factor A3), which encodes a serine protease and has been reported to negatively regulate trophoblast invasion,⁴⁸ has a reduced expression in the AMA mother, offspring and blastocyst groups (Figure 5A–C). In folliculogenesis, the expression of HTRA3 increased sharply during the transition from the antral follicle to the preovulatory follicle, consistent with the previous report about the important role in the regulation of ovulation and luteinization.⁴⁹ In the development of early embryos, the expression of HTRA3 was downregulated soon after fertilization (Figure 5D). qRT-PCR further validated the significant downregulation in the AMA group and linear corrections between mother and offspring for HTRA3 (Figure 5E,F). Among the four DEGs observed in oocyte, MED12L, one mother-offspring intergenerationally correlated DEGs downregulated in the AMA group, showed a trend toward downregulation in both mothers and offspring in qRT-PCR validation, but statistical significance was reached only in the mother group (Figure S8B,C). In conclusion, these results supported the view that some of the intergenerationally inherited alterations observed in the transcriptome of offspring directly came from the disturbance of AMA in oocytes or preimplantation embryos (Figure 5G,H).

4.1 Ethics

All samples were collected after patients signed informed consent and in accordance with ethical standards. This research was approved by the Reproductive Study Ethics Committee of Peking University Third Hospital (approval number scheme: 2016SZ-015).

4.2 Sample collection and treatment

Thirty nuclear families with foetuses (20 parents-offspring families and 10 mother-offspring families) were recruited in this research and were divided into the AMA group and the Young group based on maternal age. The mean age of AMA mothers was 39 ± 1.22 years, while that of young mothers was 29.8 ± 2.59 years. The BMI was not significantly different between AMA and Young mothers. Women with polycystic ovarian syndrome (PCOS), diabetes, hypertension, hyperthyroidism, hypothyroidism, systemic lupus erythematosus, or pregnancy complications such as preeclampsia, gestational diabetes mellitus, severe metabolic syndrome, or intrauterine infection were not eligible. Families with premature babies or infants who suffered from chromosomal abnormalities, any birth defects, or the Apgar score below 7 were also excluded. All women delivered by caesarean section at term (37 to 42 weeks). Approximately 4 ml of UCB and parental peripheral blood samples were collected on the day of delivery and treated within 2 h. Parental peripheral blood samples were collected before delivery in order to avoid any effects of blood transfusion during delivery. Of the 4 ml samples, 1 ml was used for DNA extraction and 1 ml was used for RNA extraction immediately, and the rest was stored at −80°C for backup.

4.3 DNA extraction

QIAamp^® Blood Mini Kit (Qiagen Cat# 51104) was used to collect genomic DNA (gDNA) from the blood samples following the manufacturer's instructions. The gDNA was evaluated by a NanoDrop 300 ultraviolet spectrophotometer (ALLSHENG#AS-11020-00) to ensure an A260/A280 value ranged between 1.8 and 2.0, and stored in −80°C.

4.4 RNA extraction

QIAamp RNA Blood Mini Kit (Qiagen Cat# 52304) was used to extract total RNA from the whole blood sample. The RNA was evaluated a NanoDrop 300 ultraviolet spectrophotometer (ALLSHENG#AS-11020-00) and Agilent 2100 Bioanalyzer (Agilent# G2939BA). Qualified RNA samples were referred to as samples with an A260/A280 value of approximately 2.0 and an RNA integrity number (RIN) of no less than 7. A total of 500 ng of total RNA was reverse transcribed to obtain cDNA using the PrimeScript^™ RT reagent Kit (TaKaRa Cat# RR047A). The remaining RNA sample was applied for library construction for mRNA-seq.

4.5 Reduced representation bisulphite sequencing library construction and sequencing

RRBS was performed as follows: The digestion of 3 ng unmethylated lambda DNA (Thermo Scientific, Cat# SD0021) and 500 ng of gDNA in 40 μl mixture systems was performed by adding 5 μl FastDigest MspI and 5 μl 10x FastDigest buffer (Thermo Scientific Cat# FD0544). Then, the end-repair and adapter ligation of the digested DNA fragments were handled with the NEBNext Ultra DNA Library Prep Kit (NEB Cat# E7370), following the recommended instructions in the manual. Specifically, NEBNext methylated adapter (15 μM; NEB Cat# E7535) was used to prevent the change in adaptor sequence during bisulphite conversion treatment. After the digestion of U bases by 3 μl USER enzyme (NEB Cat# E6610A), size selection of DNA fragments was performed through agarose gel electrophoresis (2% TAE gel) and excision of gel slices containing targeted 200–700 bp DNA, which were extracted by a gel DNA recovery kit (VISTECH Cat# DC2005). Next, a MethylCode bisulphite conversion kit (Thermo Scientific Cat# MECOV-50) was used for bisulphite conversion according to the manuals. The converted fragments were amplified and tagged with specific barcode sequences by 12 cycles of PCR using Kapa HiFi U+ Master Mix (Kapa Biosystems Cat# KK2801). Finally, the product was purified by two rounds of clean-up using 0.8X Agencourt AMPure XP Beads (Beckman Cat# A63881), followed by quality testing in a Qubit 3.0 Fluorometer (Thermo Scientific Cat# Q33216). After the detection of exact fragment distributions and molar concentrations with a Fragment Analyzer^™ Automated CE System (Analysis Kit: Cat# DNF-474-0500) and a Library Quant Kit for Illumina (NEB Cat# E7630L), the qualified DNA methylation libraries were sequenced on the Illumina NovaSeq platform with the PE150 strategy.

4.6 mRNA-seq library construction and sequencing

mRNA libraries were generated using the NEBNext^® UltraTM RNA Library Prep Kit for Illumina^® (NEB Cat# E7530L) according to the manufacturer's recommendations. In short, poly(A) mRNA was isolated from approximately 500 ng total RNA with NEBNext Magnetic Oligo d(T)25 Beads provided in NEBNext Poly(A) mRNA Magnetic Isolation Module (NEB Cat #E7490). After fragmentation, priming, and the subsequent first-strand and second-strand cDNA synthesis, double-stranded cDNA was obtained and purified by 1.8X AMPure XP Beads (Beckman Cat# A63881). Then, the end preparation and adapter ligation of double-stranded DNA fragments were performed, and the hairpin loop structure within the adaptor was cut by incubating the product with USER Enzyme (NEB Cat# M5505L). Finally, purified adaptor-ligated DNA fragments were amplified and tagged with specific barcode sequences by PCR. The final mRNA libraries were assessed and sequenced as previously described in the RRBS protocol.

4.7 Real-time quantitative reverse transcription polymerase chain reaction

qRT-PCR for reversed cDNA was performed with PowerUp^™ SYBR^™ Green (Thermo Fisher Cat# A25742) in the QuantStudio 3 Real-Time PCR system as follows: 95°C for 5 min, followed by 40 cycles of 95°C for 30 s, 60°C 40 s and 72°C 1 min, followed by 72°C 5 min. The delta-delta-Ct (^ΔΔCt) algorithm was performed to calculate relative gene expression.  Each experiment was performed three times, and ACTB was used as the control. The primers are listed below:

SCD (F: ACGCTTGTGCCCTGGTATTT/R: GCACCACAGCATATCGCAAG)

CD24 (F: GCTCCTACCCACGCAGATTTA/R: GACCACGAAGAGACTGGCTG)

SLC28A3 (F: TGTCAGCACCTGCGTCAT/R: CCTGCCATTCCACTCCC)

HTRA3 (F: CTGTGTTGTTGCTGGGTCAC/R: GTGTTCTGTAGGGCGAAGGG)

MED12L (F: CTCCCTCAGTATCCAGGGCT/R: CTGCTGCAAAGGCATCTGTG)

ACTB (F: CATGTACGTTGCTATCCAGGC/R: CTCCTTAATGTCACGCACGAT)

4.8 Data downloading and processing

The single-cell gene expression matrix for human oocytes during folliculogenesis and pre-implantation embryo was downloaded from our previously published datasets.^{38, 40} The expression level was estimated using the fragments per kilobase million (FPKM). The scCOOL-seq data of human preimplantation embryos were downloaded from previously published datasets,⁴¹ and processed as previously described. Briefly, Bismark software (version 0.23.0)⁷⁰ was applied to align qualified reads to the Homo sapiens reference genome (human GRCh38/hg38) with the parameter “–paired-end and –non_directional.” Those unmapped reads were then re-aligned with single-end and non-directional model. After removing the PCR duplicates, the methylation levels of CpG sites were extracted using the function bismark_methylation_extractor" in Bismark software (version 0.23.0) for downstream analysis. The aging and longevity genes list were downloaded from the Aging Atlas³⁵ and LongevityMap (Build 3).³⁴

4.9 Fundamental analysis for RRBS data

All 150 bp bisulphite sequencing paired-end reads were trimmed to delete adaptors, bases of substandard quality (Q < 20), and the reads shorter than 36 bases using TrimGalore software (https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/; version 0.6.6) and Cutadapt (version 1.18), with the parameters“ –quality 20 –phred33 –stringency 3 –length 36 –rrbs –paired –trim1.” For bisulphite conversion rate evaluation, phage λ genome was used as an extra reference using the function “bismark_genome_preparation” in Bismark software (version 0.23.0).⁷⁰ Bismark with the parameter “bowtie2" was used to map the clean reads to the spiked-in phage λ genome, and the bisulphite conversion rate was determined by the ratio of the number of unmethylated Cs to the total number of Cs detected. Sixty samples with a bisulphite conversion rate of more than 99% were retained for downstream analysis. Subsequently, Bismark with the parameter “bowtie2” was performed to align clean reads to the H. sapiens reference genome (human GRCh38/hg38). Only the uniquely mapping readings with less than 2% mismatch were retained. Then, the number of reads supporting Cs and supporting Ts in each CpG site was counted using the function “bismark_methylation_extractor” in Bismark with the parameter ” –paired-end –no_overlap." Finally, the coverage files recording the methylation state of CpG sites were inputted into the R package methylKit (version 1.10.0) for further analysis, and only CpG sites on autosomes with more than 5-fold read coverage were retained. The genome was tiled into consecutive 200 bp windows, and the 200 bp bins covered more than three CpG sites and existed in at least eight samples per group (AMA-offspring, AMA-mother, AMA-father, Young-offspring, Young-mother, and Young-father) were retained for downstream analysis. Detailed information including sequencing depth and bisulphite conversion rate, the number of covered CpG sites and retained bins, is provided in Table S2.

4.10 Copy number variation analysis

For each RRBS library, the software “readCounter” in the HMMcopy suite following the R package “HMMcopy” (version 1.26.0) was applied to calculate the CNVs at 1 Mb resolution based on mapped RRBS reads sorted by SAMtools (version 1.3.1). The CNVs were plotted by the R function “points” and “plot.”

4.11 Global DNA methylation level estimation and differentially methylated regions identification

The DNA methylation level of any retained 200 bp bin was calculated using the R package methylKit (version 1.10.0)⁷¹ as the ratio of the total count of Cs and the total count of Cs and Ts bases within that bin. Based on this, the methylation level of each sample was calculated by averaging the DNA methylation levels of all bins. Intergroup comparisons between the AMA and Young groups were also performed using the R package methylKit (version 1.10.0),⁷¹ and DMRs was referred as 200 bp bins with q-values no more than .05 and the mean methylation difference thresholds no less than 15%.

4.12 DNA methylation pattern around the genic region

The -15 kb upstream of the TSS and 15 kb downstream of the transcription end site (TES) of each gene were separately split into nonoverlapping 100 bp windows, while the gene body range from TSS to TES was equally divided into 100 fractions. The average DNA methylation levels within every window or faction were calculated, and the mean value of each genomic locations type was then computed to profile the global methylation pattern around the genic region for every sample. In addition, the mean value of all samples in the sample group was used to profile the global methylation pattern around the genic region for each group. The methylation pattern was visualized by the R function “plot.”

4.13 DNA methylation levels within various genomic elements

The coordinate files of the known genomic elements were acquired from the UCSC Genome Browser, including Low_complexity elements, CpG island (CGI), gene body, 5′-UTR, 3′-UTR, three types of promoters (high-CpG-density promoters [HCP], intermediate-CpG-density promoters [ICP], and low-CpG-density promoters [LCP]) as previously defined,^{72, 73} and repetitive elements classified into seven categories: long terminal repeat (LTR), long interspersed elements (LINE), short interspersed elements (SINE), Retroposon_SVA, transposon, satellite, and microsatellite. The coordinate files of the imprinting control region (ICRs) were obtained from the paper of Hamada et al.⁷⁴ Metastable epialleles (MEs) were obtained from the paper of Noah J. Kessler. et al.⁷⁵ To evaluate the average DNA methylation level of each type of genomic element, only regions covering more than three CpG sites were reserved, and the mean methylation level of all retained CpG sites within a specific region was defined as the DNA methylation level of the corresponding region.

4.14 DNA methylation pattern of selected DMRs in preimplantation embryos

To calculate the methylation level of each intergenerationally correlated DMR mentioned in Figures S6 and S7, each DMR was expanded with additional 150 bp in both the upstream and downstream directions. Then, each single-cell DNA methylation data recording methylation level of every CpG sites was used as input, and expanded DMRs covering more than three CpGs in DNA methylome data of early embryo development was reserved. The DNA methylation level of corresponding DMRs was calculated by averaging the methylation level of all retained CpG sites within a specific expanded DMR. Next, samples with at least one DMRs of not NA value were reserved and merged into single file. The average methylation value of every sample in different stages was calculated respectively, and defined as the methylation level of the target expanded DMR in each stage. Ward hierarchical clustering for each intergenerationally correlated DMR was performed using the function “hclust” in R package stats (version 3.6.0) based on the scaled average DNA methylation level in different stages of early embryo development.

4.15 RNA-seq data analysis

The quality of the raw fastq data was first assessed by the FastQC tool (version 0.11.9), and then processed by the software TrimGalore with the parameter of “–quality 20 –stringency 3 –length 36 –paired” to remove the adapter, ploy-N, inferior-quality bases and reads less than 36 bases. Clean reads were then aligned to the H. sapiens reference genome (human GRCh38/hg38) using STAR software (version 2.7.8a) with default parameters. The uniquely aligned reads were subsequently counted by featureCounts software (version 1.6.3).⁷⁶ Finally, DEseq2 (version 1.24.0) was applied to generate the normalized count matrix for offspring or mother samples. Detailed information on all RNA-seq libraries is listed in Table S2.

4.16 Identification of differentially expressed genes

The intergroup differential expression analysis for either offspring or mother was performed using the R package DESeq2 (1.24.1).⁷⁷ Genes meeting the criteria of fold change values more than 1.5 or less than 0.67 and p values less than .05 were defined as DEGs.

4.17 Principal component analysis and hierarchical clustering analysis

The average methylation value of the 200 bp bins around the entire genome was used for the hierarchical clustering analysis and PCA to evaluate the global DNA methylation similarity of offspring and parental samples, with the function “clusterSamples” in R package methylKit (version 1.10.0)⁷¹ and the function “pca” in R package pcaMethods (version 1.76.0),⁷⁸ respectively. PCA was also performed to evaluate the transcriptome profile of offspring and mother samples with the function “pca” in the R package pcaMethods (version 1.76.0), using the normalized counts matrix for gene expression generated by DESeq2 (version 1.24.0). The distance matrices among RNA libraries were calculated using function dist with default parameters in the R package stats (version 3.6.0) and were visualized using the R package pheatmap (version 1.0.12). The classification of the dynamic pattern of DNA methylation during the development of preimplantation embryos and the dynamic pattern of gene expression during folliculogenesis and preimplantation embryos was based on the hierarchical clustering analysis, employing the function “dist” in the R package stats (version 3.6.0) and the function “hcluster” in the R package amap (version 08–18) with the parameter “method = pearson.”

4.18 Correlation analysis

The mean value of DNA methylation level of each remaining 200 bp bins was calculated for each group (AMA-offspring, AMA-mother, AMA-father, Young-offspring, Young-mother, Young-father), and the correlation coefficients and confidence intervals between AMA and Young groups were calculated by the function “stat_cor” in the R package ggpubr (version 0.4.0) with default parameters.⁷⁹ The Spearman's correlation coefficients and confidence intervals between parents and offspring samples were calculated using the function “cor.test” and “cor” in R package stats (version 3.6.0) for either the methylation level of selected DMRs or the expression level of selected DEGs. Only DMRs or DEGs with a p value threshold of .05 and correlation Spearman's correlation coefficients (R value) greater than 0.6 were defined as candidate intergenerational related DMRs or DEGs.

4.19 Genomic functional annotation

The nearest genes, genomic features and the distance to the transcriptional start site (TSS) of DMRs were annotated using the R package ChIPseeker (version 1.20.0) and H. sapiens annotation package org.Hs.eg.db (version 3.8.2).⁸⁰ The promoter was identified as the region from 3 kb downstream to 3 kb upstream of the TSS. The coordinate files of genomic elements were downloaded from the UCSC Genome Browser as mentioned in “DNA methylation levels in specific genomic elements.” The function “foverlaps” in R package data.table (version 1.14.0) was applied to identified the DMRs with at least 1-bp overlap with any genomic elements.

4.20 Gene Ontology enrichment analysis

GO enrichment analysis for biological processes was performed to assess the potential biological functions of selected DMRs and DEGs, using the “enrichGO” function in the R/Bioconductor package “clusterProfiler (3.8.1)” packages

4.21 Statistical analysis

An unpaired two-tailed t-test in GraphPad Prism (Version 9.2.0) was applied to determine the significance of differences in the intergroup comparison of other clinical features for either parents or progenies. Fisher's exact test in SPSS (version 26.0) was used to determine the significance of differences in the comparative analysis of sexes and modes of birth. The statistical significance of Gene Ontology (GO) term enrichment analysis was determined using a hypergeometric test in the R package clusterProfiler, and the p value was adjusted by the multiple test adjustment (Benjamini–Hochberg, BH).

For bar-dot plots of gene expression determined by qRT-PCR, the significance of differences between two groups was determined by unpaired two-tailed t-test in GraphPad Prism (Version 9.2.0). For box-dot plots of DNA methylation, the significance of differences between two groups was determined by the Wilcoxon rank-sum test. The p value for the comparison among multiple groups was determined by the Kruskal–Wallis test. For a column graph of the proportion of bins with significantly differential DNA methylation levels for paternal or maternal samples, the chi-squared test was applied to calculate the significance of differences between all 200 bp bin groups and correlated AMA-DMRs of offspring groups. Unless otherwise stated, Spearman-based correlation tests were used to determine the correlation coefficient and p value between any two groups (ns: p > = .05; ∗p < .05; ∗∗p < .01; ∗∗∗p < .001).