Single-cell RNA-seq and single-cell bisulfite-sequencing reveal insights into yak preimplantation embryogenesis

Abstract

Extensive epigenetic reprogramming occurs during preimplantation embryonic development. However, the impact of DNA methylation in plateau yak preimplantation embryos and how epigenetic reprogramming contributes to transcriptional regulatory networks are unclear. In this study, we quantified gene expression and DNA methylation in oocytes and a series of yak embryos at different developmental stages and at single-cell resolution using single-cell bisulfite-sequencing and RNA-seq. We characterized embryonic genome activation and maternal transcript degradation and mapped epigenetic reprogramming events critical for embryonic development. Through cross-species transcriptome analysis, we identified 31 conserved maternal hub genes and 39 conserved zygotic hub genes, including SIN3A, PRC1, HDAC1/2, and HSPD1. Notably, by combining single-cell DNA methylation and transcriptome analysis, we identified 43 candidate methylation driver genes, such as AURKA, NUSAP1, CENPF, and PLK1, that may be associated with embryonic development. Finally, using functional approaches, we further determined that the epigenetic modifications associated with the histone deacetylases HDAC1/2 are essential for embryonic development and that the deubiquitinating enzyme USP7 may affect embryonic development by regulating DNA methylation. Our data represent an extensive resource on the transcriptional dynamics of yak embryonic development and DNA methylation remodeling, and provide new insights into strategies for the conservation of germplasm resources, as well as a better understanding of mammalian early embryonic development that can be applied to investigate the causes of early developmental disorders.

Preimplantation embryonic development is a complex process accompanied by a series of key biological events, including fertilization, cleavage (1), and the maternal-to-zygotic transition, which occurs in coordination with zygotic genome activation (ZGA) (2, 3, 4), followed by the first cell fate determination and lineage-specific differentiation (5, 6). The timing of ZGA is species-specific, despite the fact that the mechanisms controlling the start of mammalian ZGA are assumed to be highly conserved (7). During the major ZGA periods, mouse embryos are at the two-cell stage (8), human (9), goat (10), in vivo bovine (11), and pig (12) embryos are at the 4 to 8 cell stage, in vitro bovine and rabbit embryos are at the 8 to 16 cell stage (13, 14, 15), and sheep embryos are at the 16-cell to morula stage (16). These processes are critically dependent on a precisely orchestrated gene expression program regulated by various signaling molecular interactions. Normal embryonic genome activation (EGA) is necessary for continued development, whereas abnormal activation of the embryonic genome leads to embryonic developmental failure (17, 18, 19). However, expression patterns and networks that regulate EGA during early embryo development are poorly characterized in mammals, particularly in yaks. Investigation of these mechanisms will be important for understanding early embryonic mortality in humans and agricultural animals.

DNA methylation, which is an important epigenetic modification controlling mammalian embryogenesis (20), induces and sustains gene expression, transposon silencing, genomic imprinting, X chromosome inactivation, and cell differentiation (21, 22, 23). In preimplantation embryonic development, DNA methylation is highly dynamic, whereas somatic cells have relatively stable levels (24). DNA methylation changes during early embryonic development have been reported for several species (25, 26). For instance, during reprogramming of human preimplantation embryonic development, strong methylation occurs after global demethylation (27) and a similar pattern was discovered in sheep and monkey embryos (28, 29). This precisely controlled reprogramming process is necessary to avoid developmental defects or embryonic death (6, 30), and has been identified as a major cause of death in cloned embryos (31, 32, 33), although the underlying molecular mechanisms are unclear.

Yaks (Bos grunniens) are found at high altitudes (3000–5500 m above sea level) in alpine pastoral areas of the Tibetan plateau and Himalayas. Yak can adapt to harsh ecological conditions such as alpine cold, low oxygen, and intense radiation, making it a valuable species that provides meat, milk, and transportation for local pastoralists (34). However, the development rates of oocytes and in vitro embryos are low (35).

In this study, we explored the complex regulatory mechanisms of embryonic developmental in yak by constructing DNA methylation and transcriptome profiles of early embryonic development using single-cell RNA-seq and bisulfite (BS)-seq. We revealed regulatory networks associated with EGA, and further identified histone deacetylase (HDAC)1/HDAC2 and USP7 as key genes for yak development. Thus, our study provides a foundation for a better understanding of the developmental patterns of mammalian early embryos and exploring the complex mechanisms of early developmental disorders. This information is not only critical for establishing a system to evaluate the quality of embryos, but also improving the efficiency of in vitro embryo production and somatic cell nuclear transfer.

Results

Single-cell RNA-seq profiles of yak at different developmental stages in vitro

To obtain single-cell transcriptional profiles of yak preimplantation development, we mapped the transcriptional profiles of MII oocytes and preimplantation embryos at five developmental stages in vitro using Smart-seq2 (n = 3 per repetition, 5–8 embryos or oocytes per sample per repetition; Fig. 1A); we generated approximately 90 Gb data from 15 samples. After filtering, the clean reads ratio of 15 samples was consistently >91%. On average, 40.055 million to 50.028 million clean reads and 5.989 to 7.488 billion clean bases were produced (Fig. S1A, Table S1). The proportions of Ns, polyA, polyT, adapter-polluted and discarded primer-polluted reads were 0.02%, 0.06%, 0.26%, 0.9%, and 0.09%, respectively (Fig. S1B, Table S1). The distribution of the four nucleotide bases was favorable, with no AT or GC separation phenomena (Fig. S2). All samples had base quality values >20 (Fig. S3), and the percentage of bases with sequencing quality values >30 (error rate <0.1%) in cleans reads accounted for more than 90% of the total bases (Fig. S1C, Table S1). Between 91.17% and 95.30% of total reads mapped to the reference genome, with each sample having only 2.9% to 3.84% of reads multiplied by the mapping to the genome multimap rate (Table S2). Most uniquely aligned sequences were distributed in the exons of reference genome genes (Fig. S1D). Gene expression pattern analysis and gene expression level saturation curves demonstrated accurate quantification of genes at this data volume (Figs. S1, E and F and S4), thus demonstrating the high-quality of the sequencing data.

Figure 1.

Transcriptional profiles of yak in vitro MII oocytes and preimplantation embryos.A, representative images of transcriptional profiling. The scale bar represents 100 μm. B, total number of genes in each sample. C, venn diagram of differentially expressed genes at five stages. D, heatmap of correlation: The values in the grid represent correlation coefficients. E, heatmap after hierarchical clustering of the samples. The intensity of the red color reflects the degree of increased expression. F, principal component analysis (PCA). The same symbol indicates repetition of the sample.

The average number of expressed genes was 14,413 to 18,041. The number of genes expressed in MII oocytes was significantly lower relative to the 4-cell, 8-cell, and morula embryo stage (Fig. 1B). As shown in Figure 1C, fragments per kilobase of exon per million mapped fragments (FPKM) <0.1 was considered to indicate an unexpressed gene. In total, 2569 genes were specifically expressed in the morula, indicating that these genes have a stage-specific role in development. Furthermore, 7954 genes were detected in the embryo, but not in the oocyte.

The Pearson correlation coefficients between the three biological replicates of the five samples in this study were highly reproducible, with an average R > 0.81 (Fig. 1D), and the same developmental stages were almost clustered together. The greatest changes in gene expression occurring at the oocyte and morula stages (Fig. 1E), which may be explained by maternal-to-zygotic transition. Similar expression patterns within and between stages were also supported by principal component analysis (PCA) (Fig. 1F). Interestingly, one of the 2-cell embryos was clustered with a 4-cell embryo (Fig. 1E), indicating that the two embryos had comparable transcriptomes at these two stages. The relatively small distance between these two embryos in the PCA analysis also supported this (Fig. 1F).

Gene expression patterns in the preimplantation development stages of yak in vitro

To identify the gene regulatory network during embryogenesis, we used weighted gene coexpression network analysis (WGCNA) to reveal the dynamic gene expression patterns of yak preimplantation development in vitro (Fig. 2, A and B). Ten gene modules were identified (Table S3), of which the MEblack, MEdarkgrey, MElightgreen, and MEbrown modules contain a set of coexpressed genes that tended to be highly expressed at a given period and can be applied to represent core gene regulatory functions. Notably, a sequence of core genetic networks have been identified by functional analysis of the genes in these modules at specific stages: from the cell cycle of the oocyte to actin regulation and protein processing in the endoplasmic reticulum of 2-, 4-, and 8-cell embryos to RNA transcription and spliceosome and translation (Fig. 2C). This coordinated change in functional pathways illustrated the embryonic development process.

Figure 2.

Gene expression patterns during preimplantation developmental stages.A–C, WGCNA of each sample of yak. A, cluster dendrogram showing 10 modules corresponding to 10 colors. Height (Y-axis) indicates the degree of correlation. B, heatmap of correlations (and corresponding p-values) between coexpressed modules and various periods. C, functional terms of stage-specific genes. D, data have been partitioned into 6 clusters using k-means. E, heatmap of embryonically expressed genes. F, heatmap of maternally expressed genes. WGCNA, weighted gene coexpression network analysis.

To further analyze the dynamic gene expression patterns, we then performed k-means clustering and obtained six gene clusters (Fig. 2D; Table S4). Among them, clusters 1, 4, and 5, represent genes with increased expression levels during preimplantation embryonic development, indicating that they were from EGA. The genes in these clusters included genes essential for development, such as GATA6, DPPA2/4, and NANOG (Fig. 2E). The second cluster 2 represents genes that were highly expressed in the oocyte that showed low expression at the morula stage and included the developmentally important genes ZP2, OOSP2, and HIFOO (Fig. 2F). The third expression pattern, cluster 6, contained genes with expression levels that first increased and then decreased. The increased expression occurred mainly at the 2-cell and decreased at the morula, indicating that these genes, such as PAX9, MXD1, and WWOX, are only involved in EGA. The last cluster, 3, represents genes that maintained relatively constant expression levels at all stages, These genes included ATP5MC3, ATP1A1, RPL7A, and RPL27A, which are involved in energy metabolism and signal transduction, suggesting an essential role throughout preimplantation development.

Transcriptional patterns during embryonic genome activation

We detected a total of 44 to 7588 differentially expressed genes (DEGs) between any two consecutive developmental stages (Fig. 3A). The most DEGs (n = 7588) were identified at the 8-cell stage embryo to morula stage, indicating that the yak embryo genome was activated during this transition and further confirming that the main EGA stage of yak embryos occurs at this stage of the development process. A total of 5297 DEGs were detected in the MII oocyte to 2-cell stage embryo, which may be caused by the degradation of maternal transcripts, indicating that the MII oocyte to 2-cell embryo stage may play a minor role in EGA.

Figure 3.

Transcriptional patterns during the period of embryonic genome activation.A, number of DEGs between any two consecutive stages. B and C, gene ontology classification of DEGs. The “Percentage of genes” in the GO graph refers to the proportion of upregulated differentially expressed genes (DEGs) annotated to GO entries among the all upregulated DEGs (in red) or the proportion of downregulated DEGs annotated to GO entries among all downregulated DEGs (in green). D, significant functional pathways revealed from upregulated or downregulated genes between 2-cell and MII oocytes stages. E, enrichment plots from GSEA. F and G, significant functional pathways revealed from upregulated (F) or downregulated (G) genes between morula and 8-cell stages. GO, gene ontology; GSEA, gene set enrichment analysis.

The most significantly enriched gene ontology (GO) terms among the DEGs were related to developmental processes (Figs. 3, B and C and S5, A and B). The KEGG (Kyoto Encyclopedia of Genes and Genomes) pathway analyses revealed that the most enriched pathways at each developmental stage. At the MII oocyte to 2-cell embryo stage, the 3115 upregulated genes were involved in biological processes such as endocytosis, protein processing in endoplasmic reticulum, spliceosome, RNA transport, and regulation of actin cytoskeleton. It was predicted that only minor EGA occurs in this period, and a fraction of zygotic mRNAs are activated and translated. In contrast, 2182 downregulated genes were enriched in ubiquitin-mediated proteolysis, oocyte meiosis, and possibly related to the expression and degradation of maternal mRNA (Fig. 3D). Gene set enrichment analysis (GSEA) indicated that genes associated with maternal transcription were enriched for terms associated with negative regulation of gene expression (enrichment score (ES) = 0.58) and chromatin (ES = 0.59), suggesting that maternal transcripts shaped the epigenome in preparation for EGA (Fig. 3E). The transition from the 8-cell embryo to the morula was mainly enriched in ubiquitin-mediated proteolysis, regulation of the actin cytoskeleton, and the MAPK, Hippo and Wnt signaling pathway, which are closely related to embryonic development, while the most significant enrichment was in purine metabolism (p < 0.01), and the highest number of genes were enriched in pathways in cancer (Fig. 3, F and G). GSEA indicated that zygotic genes associated with EGA were enriched in terms associated with ribosome biogenesis (ES = 0.51), suggesting substantial protein synthesis at this stage (Fig. 3E). In addition, differential transcription factor predictions suggested that their differential expression may be required for EGA regulation (Fig. S5, C and E).

Cross-species transcriptome analysis of mammalian preimplantation embryogenesis

To investigate the conservation and differences in stage-specific coexpression of module genes among multiple species, we compared the gene list of yak with that of other species (human, mouse, and bovine) by multispecies WGCNA. As shown in Figure 4, A–C, we found that the gene network of yak MII oocytes had a significant overlap of 514 genes with the module of bovine MII oocytes (p < 10⁻⁵⁴), 296 genes with the module of human MII oocytes (p < 10⁻⁸), and 459 genes with mouse MII oocytes (p < 10⁻³). This similarity in stage-specific network programs indicated that these four mammalian species share some of the core maternal genes. Cross-species analysis showed that yak embryos were more similar to cattle than mouse or human, and the least similar to mouse embryos. Genes in the yak 2/4/8-cell stage specificity module overlapped significantly with those in the bovine 2-cell (p < 10⁻¹⁹) and 4-cell (p < 10⁻³⁷) stage specificity modules, the human 1-cell (p < 10⁻¹⁶) stage specificity module, and the mouse 1-cell (p < 10⁻¹³) and 4-cell (p < 10^−1.33) stage specificity modules (Fig. 4, A–C). Genes in the yak morula stage specificity module overlapped highly significantly with 503 genes in bovine blastocyst (p < 10⁻⁵⁵), with 68 genes in the bovine 16-cell stage (p < 10⁻⁷), and with 73 genes in the bovine morula stage (p < 10^−1.98), with 1020 genes in the human 8-cell stage (p < 10⁻⁷) and 693 genes in the human morula stage (p < 10⁻⁸), and with 421 genes in the mouse 8-cell (p < 10⁻⁷) and 467 genes in the mouse morula stage (p < 10⁻⁶). These data suggested that the four species shared a common core transcriptional program at the morula and blastocyst stage.

Figure 4.

Heatmaps of overlapping genes in stage-specific modules across multiple species.A–C, from left to right, the joint analysis heatmaps of yak and cattle, yak and human, and yak and mouse. Each cell includes the number of intersecting genes and the p-value (−log10). D, venn diagram of conserved maternal genes across species. E, maternal hub genes. F, venn diagram of conserved zygotic genes across species. G, zygotic hub genes.

To further investigate genes that may have key regulatory roles in these species (Fig. 4, D and F), we identified maternal (Fig. 4E) and zygotic (Fig. 4G) hub genes based on WGCNA measures of gene connectivity within modules, that is, module membership values (e.g. kME >0.7, p < 0.05), such as UHRF1, PRC1, ASF1B, SIN3A, and LOC618599 (maternal hub genes); HDAC1/2, DPPA2, HSPD1, and MRTO4 (zygotic hub genes). These results implied that these WGCNA hub genes may be key regulatory genes, acting as “master regulators” during preimplantation development.

Genome-wide methylation dynamics during preimplantation development in yak

To investigate the effects of DNA methylation on preimplantation embryos, we carried out single-cell BS-seq on yak 2-cell to morula stage embryos. The Pearson correlation coefficients between the samples (4 stages in triplicate = 12 cells) were highly repeatable at 0.55< R <0.89 (Fig. 5A), and most samples from the same stage were grouped together, with the exception of four samples from the 2-, 4-, 8-cell, and morula stages (4-cell-3 and 8-cell-3; 2-cell-2 and morula-2), which were removed from the subsequent comparative analysis (Fig. S5D). After mapping these reads to the genome, we obtained an average of 78,425,061 mapped reads with a total of 246,812,373 methylation sites (≥1× ) (Tables S5 and S6). This study showed that DNA methylation levels were higher at the 2-cell embryos, with a rapid decline in demethylation occurring at the 4-cell embryos, reaching minimum methylation at the 8-cell stage, followed by further remethylation events leading to high methylation levels at the morula (Fig. 5, B and C). CpG islands (CGI) methylation levels showed a similar trend (Fig. 5D). To visualize the distribution of methylation sites in relation to genomic location, we used the graphical tool Circos (36) to generate a methylation map by chromosome distribution. The proportions of CG, CHG, and CHH (where H = A, C or T) contexts were similar in each sample, with CG the most abundant, and CHH maintained at lower methylation levels overall (Fig. 5, E–G). In addition, little DNA methylation was found in the mitochondrial DNA.

Figure 5.

Genome-wide methylation dynamics during preimplantation embryonic development in the yak.A, correlation analysis of methylation levels for all samples. B, whole genome DNA methylation profiles for all samples. C, mapping of methylation level distribution in the whole gene 2-kb window for all samples. D, distribution of methylation levels of CGI in all samples. E–G (E–G) are the methylation levels and gene densities of CG, CHG, and CHH, respectively. Circos plot of chromosome methylation levels: from outer to inner circles are, stain map scale, a (for Yak2cell), b (for Yak4cell), c (for Yak8cell), d (for Yakmorula). H, IGV visualization of methylation kinetics in the X chromosome genomic region for all samples. I, methylation levels and gene density of the X chromosome. Circos plot of chromosome methylation levels: from outer to inner circles are, stain map scale, a (for Yak2cell), b (for Yak4cell), c (for Yak8cell), and d (for Yakmorula). Expression levels of all genes are in the black line and XIST genes are in the red line. CGI is indicated by the blue line. J, IGV visualization of DNA methylation levels (BS2cell01-BSmorula3) and transcriptome expression levels (YMII01-YMorula03) of the XIST gene in all samples. CGI, CpG islands; IGV, integrative genomics viewer.

Specific chromosomes showed varying degrees of methylation at different stages. For instance, X chromosome methylation was high at the 2-cell stage, gradually decreasing due to demethylation at the 4-cell embryo, and reaching a minimum level at the 8-cell embryo (Fig. 5H). We also observed that hypermethylated regions corresponding to overall gene expression levels were relatively low (Fig. 5I). We found that XIST transcripts were mainly expressed during the morula stage in early embryos (Fig. 5J).

Clustering analysis of genomic regions and differential loci for dynamic methylation changes

Analysis of DNA methylation in several functional regions showed consistent hypermethylation levels in the 3′-UTR, exon, and intron (Fig. 6A). The methylation pattern of the promoter and intergenic regions was identical to that of the overall genome. Methylation patterns at CG, CHG, and CHH in different cytosine contexts of embryos at various stages were also similar to that of the whole-genome (Fig. 6B). Subsequently, we found that DNA methylation along the gene body, as well as the regions 15-kb upstream and downstream of all annotated genes showed the same trend for different developmental stages. Overall, there were high levels of CG and CHG methylation in the gene body region, decreasing to the lowest levels at the transcription start site (TSS) and transcription end site (TES). The methylation level remained stable from the TES to 15-kb downstream. In contrast, CHH showed a trend of sharply decreasing methylation in the gene body region, which increased sharply near the TES. The methylation level from the TES to the downstream 15-kb was similarly stable.

Figure 6.

Clustering analysis of functional region methylation annotations and differential loci.A, distribution of methylation levels in different genomic functional regions. B, distribution of methylation levels of CG, CHG, and CHH in genomic regions in different cytosine contexts. C, clustering diagram of methylation levels of CpG loci in all samples. Each row and column represents a CpG site and cell sample respectively, and the right scale values are methylation levels (0–100).

Clustering analysis showed that the methylation status of a large number of CpG loci remained low at all phases of preimplantation embryonic development. In comparison, only a few CpG sites remained hypermethylated at these stages. Notably, abrupt methylation alterations (sudden hypomethylation or hypermethylation) were detected in the 2-cell stage, which were later reversed to their initial levels (Fig. 6C).

Unique characteristics of differentially methylated regions during EGA in the yak

EGA failure can cause embryonic developmental arrest. We used differentially methylated region (DMR) analysis to examine the genome methylation status between 8-cell embryos and morulas. We discovered approximately 79,375 DNA regions correlating with DMRs (including 78,948 hypermethylated regions and 427 hypomethylated regions) with a length range of 51 to 10.7 kbp (Table S7). The DMR length distribution was narrower for hypomethylated types than for hypermethylated types and focused in the 51 to 249 bp range (Figs. 7A and S5H). Of the DMRs, 63.9% were located mainly in introns (Fig. 7B). Among the hypomethylated regions, the exon and 5′-UTR of RPS6KA5 had a 4.05-kbp DMR. Notably, a 13.9-kbp CGI region of Chr12 contained more hypermethylation information than other DMRs. In addition, among the hypermethylated regions, USP7 contained a 4.2-kbp methylation-rich DMR, and the promoter, exon, and intron of LOC104969611 contained a 10.5-kbp DMR. The most enriched DMRs were found on Chr1, followed by a high enrichment of DMRs in the 3′ terminal region of the X chromosome (Fig. 7C).

Figure 7.

Differential methylation analysis at the 8-cell embryos stage and the morula stage.A, length of DMRs: hyper is high methylation and hypo is low methylation. B, functional annotation of DMRs. C, position of DMRs in the overall genome. D, GO functional enrichment analysis. Scatter plots are used to show the functionally significant pathways recognized from hyper (E) or hypo (F) DMRs between the morula and 8-cell stages. DMRs, differentially methylated region; GO, gene ontology.

GO analysis indicated that DMR-related genes (DMGs) were significantly enriched in the regions associated with ion/protein/nucleotide/ATP binding and cation/metal ion transmembrane transporter activity (q < 0.05) (Fig. 7D, Tables S8 and S9), indicating transcriptional activation of zygotic genes, possibly due to changes in DNA methylation. Besides, KEGG pathway analysis indicated that DMGs were significantly enriched in various signaling pathways (q < 0.05), such as metabolic pathways, the MAPK, PI3K-AKT, calcium and Hippo signaling pathways, regulation of the actin cytoskeleton, and endocytosis (Fig. 7, E and F, Table S10).

Between the 2- and 4-cell stages, we identified 14,919 DNA regions (2269 hypermethylated regions and 12,650 hypomethylated regions) corresponding to DMRs (Table S7), with a narrower DMR length distribution for hypermethylated types than for hypomethylated types (Figs. S5F and S6, A and B). GO analysis revealed that DMGs were enriched in binding, development, and transporter-related processes and were significantly enriched in ion transporter proteins (q < 0.05) (Fig. S6, E and F and Tables S8 and S9). Additionally, KEGG pathway analysis indicated that DMGs were enriched in pathways such as metabolic pathways, the MAPK, PI3K-AKT, and calcium signaling pathways, phosphatidylinositol signaling system, regulation of actin cytoskeleton, endocytosis, and protein processing in the endoplasmic reticulum (Fig. S6, C and D, Table S10). We also identified 12,097 DNA regions corresponding to DMRs (9354 hypomethylated DMRs and 2743 hypermethylated DMRs) at the 4- to 8-cell stages, located mainly in introns (Fig. S7, Table S7).

Correlation of dynamic changes in transcriptome and methylationome

To explore how DNA methylation contributes to the transcriptional regulatory network, we performed a combined multiomics analysis. Scatter plots showed a weak negative correlation between promoter methylation levels and expression of the corresponding genes at the 2-, 4-, 8-cell, and morula embryo stages (Fig. 8A).

Figure 8.

Correlation of dynamic changes in transcriptome and methylationome.A, scatter plot showing the correlation between gene expression levels and corresponding promoter methylation levels in early embryos. From left to right, 2-cell, 4-cell, 8-cell, and morula stages. Blue numbers are correlation values. B, linear plots of the proportion of methylation and nonmethylation gene of DEGs in 8 cell and morula stage. C, linear plots of the proportion of methylation and nonmethylation gene of differentially expressed genes at the 4-cell and 8-cell stage. D, DNA methylation distribution of genomic regions in different cytosine contexts for genes with different expression levels. Gene expression levels are divided into six groups from high to low, first is the highest expression; the ordinate is the methylation level; the abscissa is the gene body and the upstream and downstream 2 Kb domains. From left to right, 2-cell, 4-cell, 8-cell, and morula stages. E, violin plot of promoter methylation level distribution of DEGs. F, the line graphs show the methylation level of the repeat elements and the expression level of the repeat elements, respectively. G, venn diagram of DEGs and DMRs in adjacent consecutive periods based on upregulated and downregulated genes, from left to right, Yak4cell-versus-Yak2cell, Yak8cell-versus-Yak4cell, YakMorula-versus-Yak8cell; network diagram for differentially methylated regions of genes and DEGs that are negatively correlated in the EGA period, that is, methylation-driven genes. H, heat map of promoter methylation levels and expression levels of DMGs at different stages. I, heatmap of methylation (methylation) and expression levels (FPKM) of 26 imprinted genes. Blue and red gene names represent paternal and maternal genes, respectively. DEGs, differentially expressed genes; DMG, DMR-related genes; DMRs, differentially methylated regions; EGA, embryonic genome activation; FPKM, fragments per kilobase of exon per million mapped fragments.

Expression analysis of methylated and nonmethylated DEGs. For the DEGs expressed at low levels, the numbers of methylated and nonmethylated genes were approximately equal at the 8-cell embryo, while for the DEGs expressed at high levels, the proportion of methylated genes was higher. The opposite trend was observed at the morula stage (Fig. 8B). For the DEGs between 4-cell and 8-cell stages, the proportion of methylated genes was higher (Fig. 8C).

Based on the transcriptome sequencing, the genes were classified into five groups of first_quintile—none representing high to low expression levels (Fig. 8D). For different developmental stages, the same trend was observed in the promoter regions of genes, with gene expression levels correlated negatively with its corresponding DNA methylation levels. In terms of the global methylation levels, the DNA methylation levels at CG, CHG, and CHH in the 2-kb upstream of TSS first increased and then gradually decreased, reaching the lowest value at the TSS site. The whole genebody showed a high degree of methylation, with gene expression levels correlated positively with its corresponding DNA methylation. Methylation of CG and CHG remained stable after TES, while CHH showed a stable pattern of hypermethylation.

The distribution of methylation levels of promoters of DEGs at consecutive stages showed promoter methylation at the 2-cell embryos was higher than that at the 4-cell embryos, and promoter methylation at the morula was lower than that at the 8-cell embryos (Fig. 8E). In addition, the correlation between repeat element methylation and its corresponding expression was weak (Fig. 8F).

The Venn diagram in Figure 8G showed the association of the DEGs and DMGs during the EGA period. We found that downregulated DEGs with concomitant upregulation of methylation levels during the EGA period or upregulated in DEGs with concomitant downregulation of methylation levels served as the methylation driver genes with negative correlations between gene expression levels and methylation levels during the EGA period (Table S11). In addition, key methylation-driven genes, such as UHRF1, AURKA, PLK1, TOP2B, CENPF, NUSAP1, and CDC25B, during the EGA period were further identified by network analysis.

We further investigated the dynamics of gene expression and promoter methylation of DNA methyltransferase and demethylase. TET3 and TET2 were highly expressed in oocytes and 2-cell and began to decline at the 8-cell (Fig. 8H and Table S12). Interestingly, promoter hypomethylation was observed in 2-cell and 4-cell, suggesting that TET-mediated DNA demethylation events continue after fertilization. In addition, TET1 expression started at the morula stage, and its promoter methylation correlated negatively with its expression level, suggesting that TET1 is regulated by DNA methylation. On the other hand, DNA methyltransferase-1 was highly expressed in oocytes and gradually declined after fertilization, reaching its lowest level at the morula stage, which may be essential for early de novo methylation. In addition, UHRF1 promoter methylation correlated negatively with its expression level, suggesting that UHRF1 is regulated by DNA methylation.

Genomic imprinting is a phenomenon of parent-specific gene expression. In this study, we identified 26 imprinted genes that were annotated in the analyzed genome. We discovered two different patterns of imprinted gene-specific methylation. The first comprised 15 genes, including CD81 and SDHD, with promoter methylation that correlated negatively with the expression of their reporter alleles. The other comprised 11 genes, including IGF2, ZIM2, MEST, PHLDA2, and TSSC4, with methylation that correlated positively with the known allele expression pattern (Fig. 8I and Table S13).

Validation of DNA methylation and gene expression data

To validate the methylation pattern of DMRs identified by single-cell BS sequence analysis, we verified the methylation CpG sites in the promoter regions of six candidate genes (PHLDA2, TFRC, RAB25, CDKN1C, ANO1, and UHRF1) by BS-seq and sequencing analysis (Fig. 9, A–F). PHLDA2 methylation levels did not differ significantly among samples (Fig. 9A). RAB25 methylation levels in morula stage were significantly higher than those at the 4-cell (p < 0.01) and 8-cell (p < 0.01) stages (Fig. 9B). ANO1 had the lowest methylation at the 8-cell stage (Fig. 9C). CDKN1C methylation levels in the morula stage were significantly higher than that at the 8-cell stage (p < 0.05) (Fig. 9E). The concordance between these data suggested that the DMG identified by genome-wide BS-seq were credible.

Figure 9.

Validation of DNA methylation and gene expression data.A–F, methylation levels of six DMR-related genes validated by bisulfite sequencing PCR analysis between 4-cell, 8-cell, and morula. G–K, validation of transcriptome sequencing results by qRT-PCR. Circles on each horizontal chain represent CpG sites in the study sequence, whereas white and black circles represent nonmethylated and methylated CpG, respectively. Each horizontal strand represents a sequenced clone. The circles in the BS-PCR results indicate DNA methylation levels automatically generated by the BiQ Analyzer. In each sample, the number of methylation sites divided by the total number of sites indicates the level of gene methylation. BS-PCR and qRT-PCR were repeated three times for each sample. Different letters indicate samples that differ significantly (p < 0.05). BS, bisulfite; DMR, differentially methylated region; qRT-PCR, quantitative real-time PCR.

To validate the single-cell RNA-seq data, we selected several candidate DEGs (HDAC2, KLF4, NANOG, OOSP2, and CEP78) for quantitative real-time PCR (qRT-PCR) analysis, with H₂A as the reference gene (Fig. 9, G–K). HDAC2 expression was comparatively low in the MII oocyte stage and relatively high in the morula stage (Fig. 9G), while OOSP2 expression was relatively high expressed in oocytes and low in 2-cell embryos (Fig. 9J). Overall, the qRT-PCR and single-cell RNA-seq data were consistent.

Effects of HDAC1/2 and USP7 on the development of yak embryos

To explore the impact of DMG as an epigenetic modification-related and the key conserved HDAC1/HDAC2 and USP7 genes on yak development, we first examined the expression levels of various yak tissues by qRT-PCR. Significant HDAC2 expression was detected in ovarian tissues, while HDAC1 expression was low (Fig. 10, A and B). HDAC2 was also highly expressed at the morula (Fig. 9G). A high level of USP7 expression was found in the spleen and at the morula stage (Fig. 10, C and D). Treatment of morula embryos with the HDAC inhibitor FK228 (50 μM) (37, 38) affected blastocyst development in vitro (Fig. 10E), with significantly lower blastocyst development rate at day 7 in the FK228 group (26.0% ± 0.01) than that in the dimethyl sulfoxide (DMSO) group (35.5% ± 0.01) (p < 0.05) (Table.1), indicating that inhibition of HDAC1/2 expression decreased blastocyst formation. The counts of total blastomeres, trophectoderm (TE) cells, and inner cell mass (ICM) cells in the FK228 group were significantly less than those in the DMSO group (p < 0.05) (Fig. 10F). In addition, immunofluorescence staining revealed reduced CDX2 expression level in the FK228 group (p < 0.01) (Fig. 10, H and I), indicating that HDAC1/2 inhibition decreased the quality of blastocysts.

Figure 10.

Effects of HDAC1/2 and USP7 on the development of yak embryos.A and B, expression analysis of HDAC1/2 in different tissues of yak. C, expression analysis of USP7 in different tissues of yak. D, expression profile of USP7 in preimplantation embryonic development in the yak. E, representative photos in the bright fields on day 5 and day 7 after fertilization of yak embryos treated with DMSO or FK228. The scale bar represents 100 μm. F and G, scatter plots show the total cell number, ICM cell number, and TE cell number of blastocysts in the FK228-treated and P5091-treated groups and the corresponding control groups, respectively. H and I, immunofluorescence (IF) staining of CDX2 in yak blastocysts after DMSO or FK228 treatment. The scale bar represents 50 μm. J and K, IF of CDX2 in yak blastocysts after DMSO or P5091 treatment. The scale bar represents 50 μm. L and M, IF of 5 mC and 5hmC in yak blastocysts after DMSO or P5091 treatment. The scale bar represents 50 μm. (n = 3; 10–20 embryos per group per replicate). Data shown as Means ± SEM. ∗ p < 0.05, ∗∗ p < 0.01. DMSO, dimethyl sulfoxide; HDAC, histone deacetylases; ICM, inner cell mass; TE, trophectoderm.

Table 1.

Effect of HDAC1/2 inhibition on the development of yak embryos in vitro

Group	No. of zygote (%)	No. of cleaved embryos (%)	No. of morula on 5 days (%) per 2-cell	No. of blastocysts on 7 days (%) per 2-cell
NC	62	45 (72.7 ± 0.04)a	18 (39.8 ± 0.02)a	16 (35.5 ± 0.01)a
FK228	64	49 (76.5 ± 0.03)a	20 (40.5 ± 0.04)a	13 (26.0 ± 0.01)b

^a,bValues with different superscripts are significantly different (p < 0.05).

Treatment of morula embryos with the USP7 inhibitor P5091 (35 μM) affected blastocyst development in vitro, with a significantly lower blastocyst development rate at day 7 in the P5091 group (20.1% ± 1.2) than that in the DMSO group (38.2% ± 2.3) (p < 0.05) (Table 2), suggesting that inhibiting the function of USP7 reduced the rate of blastocyst development. The counts of total blastomeres, TE cells, and ICM cells in the P5091 group were significantly less than those in the DMSO group (p < 0.05) (Fig. 10G). In addition, immunofluorescence staining showed that the expression level of CDX2 was decreased (p < 0.01) in the P5091-treated group (Fig. 10, J and K), indicating that inhibition of USP7 function reduced the quality of blastocysts. Moreover, 5 mC and 5hmC were detected by further immunofluorescence staining, which showed that the level of DNA methylation in the P5091-treated group was increased (p < 0.05), and the P5091 inhibitor affected DNA methylation (Fig. 10, L and M). Taken together, these data demonstrated that HDAC1/2 and USP7 are critical for yak preimplantation embryonic development.

Table 2.

Effect of USP7 inhibition on the development of yak embryos in vitro

Group	No. of zygote (%)	No. of cleaved embryos (%)	No. of morula on 5 days (%) per 2-cell	No. of blastocysts on 7 days (%) per 2-cell
NC	67	55 (81.3 ± 2.8)a	30 (52.2 ± 3.1)a	22 (38.2 ± 2.3)a
P5091	69	52 (74.6 ± 2.8)a	27 (51.1 ± 2.0)a	11 (20.1 ± 1.2)b

^a,bValues with different superscripts are significantly different (p < 0.05).

Discussion

In this study, we performed single-cell RNA-seq and BS-seq analysis to explore the role of the major EGA periods in the gene expression patterns and regulatory networks of embryonic development, and related transcriptional regulators in the yak.

Unsupervised hierarchical clustering and PCA revealed (Fig. 1, E and F) the existence of two separate stages of yak preimplantation embryogenesis, represented by the transitions of the oocyte to the 2-cell stage and of the 8-cell embryo to the morula stage. The greatest changes in gene expression also occurred in the transition from the 8-cell to the morula stage, revealing that the major EGA in yaks occurred during this period. Notably, a large number of DEGs were also present from the oocyte to the 2-cell, indicating that this represents a minor wave of EGA. The major EGA period in mouse embryos occurs at the 2-cell stage (9), in human, goat, and pig embryos at the 4 to 8 cell stage (9, 10, 12), and in bovine and rabbit embryos at the 8 to 16 cell stage (13, 14, 15). The minor EGA period in mouse embryos at the 1-cell stage (39) and in bovine embryos at the 2 to 4 cell stage (40), further suggesting that the timing of EGA is species-specific. In contrast, based on transcriptome sequencing, Zi et al. concluded that EGA occurs at the 4–8-cell stage in yak, a discrepancy that may be due to differences in sequencing methods (41), and a similar observation was made in goat embryos (42).

GO analysis of DEGs during the major EGA period revealed that the genes were mainly enriched in metabolic process and binding and catalytic activity, terms that were also highly enriched during the embryonic development of goats (10). According to the KEGG analysis, the DEGs in this period were involved mainly with the Hippo, Wnt, MAPK, and ubiquitin-regulated protein hydrolysis signaling pathways, many of which are also highly enriched in goat and cattle embryos (43, 44). The Hippo signaling pathway is essential for lineage formation in preimplantation embryos and is highly conserved in mammals. Abnormal Hippo signaling can block ICM and TE lineage differentiation, causing preimplantation or post implantation embryonic arrest (45). Meanwhile, the Wnt signaling pathway affects stem cell self-renewal and development (46).

We then compared the complete transcriptomes of preimplantation embryos of four mammalian species (yak, human, mouse, and bovine) to investigate whether the development of functional modules is conserved among species. Through this cross-species analysis, we identified several hub genes as potentially key regulatory genes at the embryonic stage. While the majority of these genes were poorly described, several of them had been implicated in significant roles in earlier investigations. For example, the maternal hub gene UHRF1 determines the quality of mouse oocytes (47). In mice, sin3a regulates the first lineage differentiation through Hdac1 (48), and sin3a also affects oocyte maturation by regulating histone deacetylation (49). Asf1 is a highly conserved histone chaperone, and the loss of Asf1a or Asf1b in mouse embryos significantly reduced the developmental rate of the morula and blastocyst (50).

Zygotic hub genes included the histone variant H2AFZ, which impacts the structure of the chromatin and promotes expression of pluripotency genes. In mouse embryos, homozygous deletion of H2afz leads to abnormal endocytic clusters and embryonic death (51). In mouse and bovine embryos, HDAC1/2 is essential for ZGA to create the proper transcriptional repression and active state (37). Our study in yak embryos also found that inhibition of HDAC1/2 can reduce the rate of blastocyst development and affect the quality of blastocysts. In addition, DPPA2 and DPPA4 are key genes that regulate ZGA and Dux in mice (52). Klf4 glutamylation was necessary for mouse cell reprogramming and embryogenesis (53). Inactivation of Hspd1, which encodes the mitochondrial Hsp60 chaperone protein, leads to failure of early embryonic development in mice (54). Overall, further functional analysis of these conserved hub genes in multiple species of embryo development will help clarify the mechanisms of mammalian early embryonic development, and provide a basis for decreasing early pregnancy loss in domestic animals.

As an important epigenetic regulatory mechanism (20, 55), DNA methylation has been shown to affect yak embryogenesis (56); however, details of the methylation changes in plateau yaks are unknown. Using single-cell BS-seq, we identified two marked changes in methylation levels in yak preimplantation embryos. DNA methylation levels reduced from the 2-cell to the 8-cell due to demethylation, and then increased from the 8-cell to the morula through remethylation. Our results suggest that methylation during yak embryonic development is dynamically balanced by genome-wide methylation reprogramming, and similar results have been reported in human, monkey, and bovine embryos (25, 27, 29). The 8-cell stage, which had the lowest methylation level, is accompanied by the start of transcription and expression of many embryonic genes. Due to differences in species, comparisons of the changes in methylation in preimplantation bovine, sheep, monkey, and human embryos are characterized differently in terms of developmental timing (25, 27, 28, 29). However, all these species exhibit extremely low methylation during the major EGA stages, suggesting that changes in methylation are necessary for embryonic transcriptional activation.

EGA is critical for embryonic development, and failure can cause embryonic developmental arrest, although the specific regulatory mechanisms are unknown. By comparing the DMG in 8-cell and morula, we found that methylation changes were significantly related to metal ion transporter protein activity. Transferrin has been reported to contribute to the mechanism by which mouse embryos break the block in the ZGA stage and develop into blastocysts (57). In sheep embryos, the change in DNA methylation was also significantly related to ion transporter activity (28). Notably, methylation changes were significantly related to multiple signaling pathways, such as metabolic pathways, the MAPK, PI3K-AKT, calcium and Hippo signaling pathways, regulation of the actin cytoskeleton, and ubiquitin-mediated protein hydrolysis, which also implies translation changes occur in the protease and the protein degradation process during implantation, possibly due to DNA remethylation. In the hypermethylated region of the DMRs, we observed that a deubiquitinating enzyme, USP7, is enriched with methylation information and it has been shown that USP7 can regulate overall DNA methylation by deubiquitinating DNA methyltransferase-1 and Uhrf1 (58, 59). At the same time, USP7 is also a DEG during the EGA period, and it is highly expressed in morula. Our studies in yak embryos have found that inhibition of USP7 may reduce the blastocyst development rates by affecting the DNA methylation levels. These results will contribute to a better understanding of mammalian EGA events, with important implications for improving the efficiency of in vitro-produced embryo development using techniques such as in vitro fertilization and embryo cloning.

Through a combined multiomics analysis, we identified key methylation driver genes that correlated negatively with gene expression levels and methylation levels during the EGA period, most of which were poorly characterized, with only a few thought to be important for embryonic development. For example, in mouse embryos, abnormalities in Aurka, which is required for chromosome segregation, result in abnormal oogenesis (60). Similarly, NuSAP deletion causes abnormal spindle assembly, leading to early mouse embryonic death (61). CENPF knockdown results in impaired 2- and 4-cell development in mice (62). In bovine embryos, CENPF knockdown leads to a dramatic reduction in the developmental capacity after EGA (63). Furthermore, PLK1 is involved in the homologous recombination process and maintains the integrity of the genome in early mammalian embryonic cells (64). DNA methylation is strictly controlled by regulating the expression of key genes in EGA and is conserved in the regulation of early mammalian embryonic development (20, 65, 66). Our results could also provide a valuable resource for studying epigenetic reprogramming events in preimplantation embryonic development in other mammals.

Conclusions

Our comprehensive analysis using single-cell RNA-seq and single-cell BS-seq techniques reveals a dynamic relationship between epigenetic reprogramming and transcriptional regulatory networks in the complex regulatory mechanisms of embryonic developmental arrest in yak and maps important candidate gene networks regulating the EGA period. HDAC1/2 and USP7 were identified as key regulators of yak preimplantation embryonic development. Our data contribute to a better understanding of EGA events and lay the foundation for the expansion of improved yak herds and protecting the germplasm resources. Our findings also provide a basis for establishing an embryo quality evaluation system and understanding the mechanism of pregnancy failure in mammals, as well as an important reference for in vitro production of embryos using techniques such as in vitro fertilization and cloning.

Experimental procedures

Animals and ethics statement

Animal experiments were conducted in accordance with the Guide for the Care and Use of Laboratory Animals (Ministry of Science and Technology of China, 2006) and approved by the Animal Care and Use Committee of Northwest A&F University. The ovaries of slaughtered yaks were obtained from Datong Abattoir in Qinghai, China (67). Frozen-thawed yak semen was obtained from Xining Animal Disease Control Center (Qinghai, China).

Materials

All chemicals and reagents, unless otherwise stated, were purchased from Sigma-Aldrich. Disposable, sterile plastic ware was from Corning (Corning).

Oocyte harvesting, in vitro maturation (IVM), fertilization, and culture

Oocytes were collected and placed in a Dewar flask with sterile saline at 21 to 24 °C for transportation to the laboratory within 5 h. Cumulus-oocyte complexes (COCs) were aspirated from 2 to 8 mm antral follicles with a 10-mL syringe. The COCs surrounded by >3 layers of cumulus cells and with a homogenous cytoplasm were selected, washed with Brackett and Oliphant (BO)-Wash (IVF Bioscience) and BO-IVM medium (IVF Bioscience), and cultured in BO-IVM medium at 38.5 °C in a humidified incubator under 5% CO₂ for 20 h (68).

For in vitro fertilization, COCs (25, 26, 27, 28, 29, 30) were first added to 100 μl microdrops of BO medium. After thawing, yak semen was separated by swim up in BO medium, and then centrifuged ( ×2) at 500g for 10 min. Approximately 20 μl (1 × 10⁶ sperm/ml) resuspended sperm was added to each microdrop. After 20 h, the cumulus cells were completely dispersed in PBS by repeated pipetting and then treated with 0.25% hyaluronidase for 4 min. After washing ( ×2) in synthetic oviductal fluid with amino acids solution supplemented with 8 mg/ml bovine serum albumin, zygotes were cultured in BO-IVC medium (IVF Bioscience) (68). For the FK228 treatment experiment, yak morula was treated with FK228 (Selleck, 50 μM) for 48 h. For the P5091 treatment experiment, yak morula was treated with P5091 (Selleck, 35 μM) for 48 h. At the same time, the same dose of DMSO was used as the control group.

Mature oocytes and embryos were washed ( ×3) with RNase-free PBS-polyvinyl alcohol solution to remove RNase and ions, and so on. single-cell RNA-seq samples added to cell lysates (Annoroad), and single-cell BS-seq samples were added to Bis-lysates (Annoroad);(5 replicates/group). The lysed cells were stored at −80 °C and prior to sequencing at Annoroad.

RNA isolation, library construction, and sequencing

For single-cell RNA-seq, single-cell samples were amplified directly by Smart-seq2 to obtain amplification product complementary DNA and further library construction of eligible amplification products (44). Fragmentation of the complementary DNA samples was performed by the Bioruptor Sonication System (Diagenode Inc). After end-repair, addition of base A, addition of splice, various index markers, amplification products were obtained by PCR. At each step, the reactions were purified by magnetic beads to obtain the final library. The library fragment lengths and effective concentrations (>10 nM) were checked, and the quality-checked libraries were analyzed on the HiSeq sequencing platform (Illumina) with the PE150 paired-end sequencing program. The sequenced reads were then mapped to the Bos taurus reference genome (ARS-UCD1.2) with TopHat2 (v2.0.12, https://ccb.jhu.edu/software/tophat/index.shtml) software. The levels of gene expression were normalized by taking into account FPKM values. Genes with FPKM >0.1 were retained for downstream analysis, including Pearson correlation, PCA, and cluster analysis using R version 3.6.0 (http://www.r-project.org/). GSEA was carried out with MSigDB (v7.2, https://data.broadinstitute.org/gsea-msigdb/msigdb/release/7.2/). DESeq2 (http://bioconductor.org/packages/release/bioc/vignettes/DESeq2/inst/doc/DESeq2.html) was used to analyze the DEGs for which screening refers to fold-change in expression and q value (p-value after correction) as relevant indicators. DEGs were defined by q < 0.05 and fold-change ≥2. Gene function analysis website DAVID was used for GO analysis of DEGs associated with major biological functions (69). Signaling pathways associated with these genes were analyzed by comparison of the DAVID results with the KEGG database. Significance was set at p < 0.05.

Cross-species transcriptome comparative analysis

WGCNA was conducted as described previously (10). Based on publicly available transcriptional datasets from human, mouse (9), and bovine (70) oocytes and preimplantation embryos, we investigated whether functional modules are conserved among species. By comparing the gene module of yak with those in three other datasets. Overlapping genes were counted in any two modules from different species, and Fisher’s exact test was performed. We identified hub genes for each module by examining genes with high module membership in a module. We imported hub genes into the STRING database (71) and used the Cytoscape (https://www.bytesin.com/software/Cytoscape/) software to build the protein-protein interaction network (72).

Single-cell BS-seq test

For single-cell BS-seq, single-cell samples were BS-treated to complete C-T conversion using the EZ-DNA Methylation-Direct Kit (D5021, Zymo Research) (28). After first and second strand synthesis, PCR amplification was performed using purified DNA. Index sequences were introduced during the amplification process, and the product was purified to obtain a library. The fragment length distribution and effective concentration of the libraries were then evaluated, and the quality-controlled single-cell BS-seq libraries were subjected to 12-bp paired-end sequencing on the HiSeq2500 platform.

The down-machine data were first filtered to obtain high-quality data. The error rate of each base was estimated as Phred score, and the filtered data were analyzed in balance. The sequencing data were then compared with the B. taurus genome (ARS-UCD1.2) using Bismark tool (73). Data C base methylation information were acquired and analyzed for subsequent methylation information.

Genome-wide methylation level analysis

We assessed the reproducibility between groups and compared the methylation differences using a sliding window approach (2-kb) to calculate genome-wide methylation levels and Pearson correlation coefficient for all samples. The distribution of methylation levels and the number of CGI in functional regions at different phases of embryonic development were examined to describe methylation patterns and regulatory status (74). Long interspersed nuclear elements (LINE), short interspersed nuclear elements (SINE), and long terminal repeats (LTR) were labeled with RepeatMasker. Therefore, promoters were defined as 1000-bp upstream of the gene TSS.

Clustering and difference analysis

The CpG sites common to all embryos were identified, and the methylation levels of each site were determined. Following that, the CpG sites identified were clustered for methylation levels. Differentially methylated cytosines and DMR were identified using Bioconductor package DSS (75), and a strict Wald test was performed based on the β-binomial distribution model. The cytosine sites with a q-value < 0.05 and a difference in methylation level >0.2 between the comparison groups were defined as differentially methylated cytosines ; regions with p < 0.05 and methylation level difference >0.2 between the comparison groups and containing ≥3 CpG sites were defined as DMRs. Identify and quantify the positional distribution of DMRs in each chromosome. The length of the different levels of methylation of DMRs was then compared at consecutive developmental stages. In addition, the DMRs were structurally annotated using annotated information on the functional elements from the reference genome, and genes with intersections with DMRs were considered to be DMGs.

Correlation of gene expression and DNA methylation

The log2 gene expression levels (FPKM) of detectable genes (FPKM >0.1) and the promoter DNA methylation level and other genomic characteristics of each corresponding expressed gene were calculated using R and mapped.

Determination of DNA methylation by BS analysis

Methylation levels in 4- and 8-cell stage embryos and morula embryos were analyzed using the BS method. Samples were treated with NaHSO₃ using the EZ-DNA methylation direct kit (Zymo Research) (68, 76). PCR products were generated using the primers for PHLDA2, RAB25, ANO1, TFRC, CDKN1C, and UHRF1 (Table S14), as well as DNA samples and Tag DNA polymerase. PCR products were then purified, and cloned into the pMD18-T vector (TaKaRa), and 3 to 5 clones from each independent series of amplifications were sequenced. For each sample, three independent amplification experiments were performed. Using the BiQ Analyzer, DNA methylation levels were expressed in terms of the number of CpG-methylated sites compared to the total number of CpG-methylated sites (77).

Reverse transcription and qRT-PCR

Total RNA was cleaved from embryos (n = 15 per pool) with the Cells-to-Signal kit (Ambion Co) and TRIzol reagent (TaKaRa) (68). The mRNA was transcribed by Takara reverse transcription kit (TaKaRa). qRT-PCR was performed on an ABI StepOnePlus PCR system (Applied Biosystems) using the primers shown in Table S15. The results were normalized to the H₂A mRNA levels.

Immunofluorescence staining of embryos

Immunofluorescence staining of embryos was conducted as described previously (78). After permeabilization and blocking, embryos were incubated with primary antibodies overnight at 4 °C. After repeated washing, the samples were incubated with the corresponding secondary antibodies for 1.5 h. Embryos were then washed and sealed with 4,6-diamidino-2-phenylindole for 10 min and sealing tablets (78). Staining was repeated three times with a group of 10 to 15 embryos per replicate. Intensities were measured as described previously (79, 80). The number of ICM cells in blastocysts was estimated by the difference between the total number of cells determined by 4,6-diamidino-2-phenylindole staining and the number of TE cells determined by CDX2 staining.

Statistical analysis

All the experiments were repeated in triplicate and data are presented as the mean ± standard error of the mean (SEM). Significant differences in the level of gene expression between groups were determined by one-way ANOVA and the least squares difference tests using SPSS v18.0. p < 0.05 was set as the threshold for statistical significance.

Data availability

Data supporting the original contributions presented in the study are included in the article and supplementary materials.

Supporting information

This article contains supporting information.

Conflict of interest

The authors declare that they have no conflicts of interest with the contents of this article.

Reviewed by members of the JBC Editorial Board. Edited by Brian D. Strahl