Early transcriptional divergence underlies cell fate bias in bovine embryos

Abstract

Developmental plasticity, or the ability of early embryonic cells to contribute to multiple lineages, is traditionally considered equal among sister blastomeres during early cleavage. However, divergence may occur earlier than expected. We performed single-cell RNA sequencing of bovine embryos from the 2- to 8-cell stages to examine transcriptional asymmetry. While gene expression was uniform at the 2-cell stage, variability increased at the 4-cell stage and became pronounced by the 8-cell stage. At this stage, blastomeres showed heterogeneity in MAPK pathway genes (e.g., RAC1, MAPK14) and the trophectoderm marker CDX2. These differences were associated with blastomere size; as larger blastomeres exhibited molecular and functional features associated with an increased propensity to generate trophectoderm cells. Thus, sister blastomeres exhibited progressive transcriptional divergence prior to compaction in bovine embryos, and these early differences may influence subsequent lineage trajectories.

Single-cell RNA-seq of bovine embryos reveals that transcriptional asymmetry arises as early as the 4-cell stage and becomes pronounced by the 8-cell stage, reflecting lineage bias before compaction.

Introduction

Early mammalian embryos exhibit a high degree of developmental plasticity, whereby individual blastomeres retain the capacity to generate both embryonic and extra-embryonic lineages. This flexibility reflects the transient totipotent state during the early cleavage stages, in which each blastomere can theoretically give rise to a complete organism. This plasticity has traditionally been considered equivalent between sister cells during early cleavage stages^1,2, with this property maintained until the morula stage, when compaction and positional cues trigger the initial lineage segregation^3,4. However, accumulating evidence suggests that developmental potential is restricted earlier than previously assumed, with molecular asymmetries and lineage biases arising during the early cleavage stages. Developmental plasticity is tightly regulated by pluripotency networks and extrinsic signals, such as the fibroblast growth factor and Wnt pathways⁵. Therefore, symmetry-breaking processes are possibly initiated well before apparent morphological distinctions.

In mice, developmental bias has been reported as early as the 2-cell stage, with one blastomere contributing more to the inner cell mass (ICM) and the other favoring the trophectodermal fate^6,7. These early tendencies are accompanied by differences in gene expression, cell cycle dynamics, and epigenetic states, even before overt polarity and spatial organization^8,9. Single-cell RNA-sequencing (scRNA-seq) studies revealed that transcriptional asymmetries between sister blastomeres emerge during cleavage in mouse embryos. These asymmetries, reported as early as the 2-cell stage in some studies, are associated with early cell fate bias, possibly representing the earliest molecular signatures of lineage segregation^8,10,11. Although the robustness and reproducibility of such early transcriptomic differences have been questioned¹², these reports suggest that symmetry breaking at the molecular level precedes and influences the initial cell fate decisions in mammals.

Whether a similar early divergence occurs in non-rodent mammals developmentally analogous to humans remains unclear. Bovine embryos offer a biologically relevant model to evaluate this, as they closely resemble human embryos in terms of cleavage dynamics, timing of zygotic genome activation, and gradual segregation of embryonic and extraembryonic lineages^13–17. These features make cattle valuable complementary models to rodents for early cell fate studies.

To date, a previous study in bovine has analysed inter-blastomere transcriptional variation at a single time point, limiting our understanding of the ways in which asymmetries evolve across developmental stages in individual embryos¹⁸. Only a few studies, even in mouse embryos, have attempted to track the transcriptional heterogeneity between blastomeres in a stage-resolved embryo-specific manner^10,11 and even fewer have extended this analysis beyond the early cleavage stages.

To address the above-mentioned limitations, we performed scRNA-seq of the individual blastomeres of bovine embryos at the zygote, 2-cell, 4-cell, and 8-cell stages. By capturing cell-specific transcriptomic profiles across successive cleavage divisions, we aimed to determine when the transcriptional divergence between sister blastomeres first emerges and whether these differences correspond to the onset of lineage bias. We found that divergence begins at the 4-cell stage and becomes pronounced at the 8-cell stage, coinciding with early signs of lineage bias. Our findings provide insights into the progressive restriction of developmental plasticity before the morula stage in a non-rodent mammalian model, enhancing our understanding of the timing of and mechanisms underlying early cell fate decisions.

Results

Progressive divergence in gene expression between sister blastomeres from the 4-cell stage

We successfully generated SMART-seq libraries from 90 individual blastomeres derived from ten zygotes (1-cell), ten 2-cell, five 4-cell, and five 8-cell stage embryos. Each library set represents a complete group of sister blastomeres from the same embryo, allowing precise comparisons of gene expression variability within and between embryos across developmental stages (Fig. 1A). We performed principal component analysis (PCA) and hierarchical clustering to assess the transcriptomic variability across developmental stages. Blastomeres of different stages (1-, 2-, 4-, and 8-cell stages) formed distinct stage-specific clusters (Fig. 1B). At the 2-cell stage, gene expression profiles were highly similar between embryos and sister blastomeres, indicating minimal variability. However, transcriptomic divergence increased as the development progressed. By the 4-cell stage, variability was observed both within and between embryos, which became more pronounced at the 8-cell stage.

Fig. 1. Transcriptomic profiling of individual blastomeres across early developmental stages.

A Schematic showing the isolation of individual blastomeres from bovine embryos at the 1-cell (n = 10), 2-cell (n = 10), 4-cell (n = 5), and 8-cell (n = 5) stages. All blastomeres from each embryo were collected (total 90 cells), enabling within- and between-embryo comparisons using scRNA-seq. B Principal component analysis (PCA) of individual blastomeres isolated from the 1-, 2-, 4-, and 8-cell stage embryos. Blastomeres of different developmental stages are plotted according to the first two principal components. C Hierarchical clustering of individual blastomeres based on the global gene expression profiles. Each branch represents an individual blastomere, with blastomeres derived from the same embryo indicated by the same color. D Conceptual illustration of gene expression variance (total sum of squares, SS) partitioned into within-embryo (SSwe) and between-embryo (SSbe) components for genes with high SSwe/SS ratios. The upper panel illustrates high SSwe with low SSbe, while the lower panel shows low SSwe with high SSbe. Colors indicate relative expression levels. E Distribution of SSwe / SS values of genes at the 2-, 4-, and 8-cell stages. SS represents the total variance in gene expression across all blastomeres, whereas SSwe represents the variance specifically between sister blastomeres within the same embryo. SSwe/SS ratio indicates the proportion of total variance attributable to within-embryo variability, with values near 0 indicating symmetric expression, and those near 1 indicating asymmetric expression between sister blastomeres.

Cluster analysis supported these observations (Fig. 1C). Although sister blastomeres clustered tightly at the 2-cell stage, such clustering was less pronounced at the 4-cell stage. By the 8-cell stage, blastomeres of the same embryo did not cluster together. Instead, each blastomere formed a separate branch, indicating substantial intercellular heterogeneity.

This progressive loss of embryo-level clustering was quantitatively captured by the SSwe/SS metric (Fig. 1D and Supplementary Data 3). SSwe/SS values exhibited a stage-dependent rightward shift in their distribution across cleavage stages. Compared with the 2-cell stage, the peak of the SSwe/SS distribution shifted substantially toward higher values at the 4-cell stage, indicating the onset of transcriptional divergence between sister blastomeres. This rightward shift progressed further at the 8-cell stage, consistent with CV-based analyses of inter-blastomere transcriptional variability (Fig. 1E).

These results suggest that transcriptional variability between sister blastomeres is initially low but increases with successive cleavage division, supporting the notion that molecular divergence begins as early as the 4-cell stage.

Enrichment of MAPK-associated signaling pathways among genes showing expression variability at the 8-cell stage

Subsequently, genes and gene sets exhibiting high inter-blastomere variability at the 8-cell stage were identified based on their coefficients of variation (CV). Genes with CV ≥ 0.8 were selected, with the additional criterion that a transcript abundance of TPM ≥ 1 was detected in at least two blastomeres across all embryos. Hierarchical clustering analysis of these genes segregated blastomeres into two major clusters. Notably, in every embryo, at least one blastomere was assigned to each cluster, indicating pronounced transcriptional heterogeneity among sister blastomeres (Fig. 2A). Genes enriched in the right-hand cluster were predominantly associated with MAPK, Wnt, cellular senescence and Rap1 related pathways (Supplementary Fig. 1 and Supplementary Data 5). Consistent with this observation, KEGG pathway analysis of the CV ≥ 0.8 gene set identified the MAPK signaling pathway as the top enriched pathway, while Rap1, cellular senescence, and Wnt signaling pathways were also enriched (Fig. 2B and Supplementary Data 6). By contrast, comparable pathway-level enrichment was not evident at the 4-cell stage (Supplementary Fig. 2 and Supplementary Data 6). Notably, KEGG analysis of genes with high SSwe/SS values (≥0.85) yielded a similar pattern (Supplementary Fig. 3 and Supplementary Data 6), indicating that these pathways were consistently associated with inter-blastomere variability at 8-cell stage. At 8-cell stage, gene overlap was observed among the MAPK, Rap1 and Wnt pathways, with shared components such as RAC1 highlighting functional interconnectivity among these signaling cascades (Fig. 2C and Supplementary Data 6). Among genes showing particularly high inter-blastomere variability and associations with these pathways were RAC1, HRAS, and MAPK14^19,20, all of which have been implicated in trophoblast differentiation, as well as CDX2, a canonical trophoblast marker (Supplementary Data 6). These genes also exhibited high SSwe/SS values (>0.85) (Supplementary Data 6), indicating that their variability primarily reflected differences between blastomeres within the same embryo rather than variability between embryos. Stage-specific expression analysis revealed distinct temporal patterns for these genes. CDX2 expression was first detected at the 2-cell stage, with inter-blastomere variability emerging at the 4-cell stage and persisting through the 8-cell stage (Fig. 2D). In contrast, HRAS, MAPK14, and RAC1 were expressed from the 1-cell stage and showed relatively uniform expression at the 2-cell stage; however, expression differences between sister blastomeres became increasingly pronounced from the 4-cell to the 8-cell stage, with RAC1 exhibiting a distinct bimodal expression pattern at the 4- and 8-cell stages (Fig. 2D). To validate the inter-blastomere variability observed in the RNA-seq data, expression levels of CDX2 and MAPK14 were examined among sister blastomeres within individual 8-cell embryos using quantitative real-time PCR. Both genes exhibited clear intra-embryo variation, with expression levels differing by approximately two- to threefold between the lowest- and highest-expressing blastomeres within the same embryo (Supplementary Fig. 4A).

Fig. 2. Assessment of transcriptional asymmetry between sister blastomeres.

A Heatmap showing hierarchical clustering of genes with high inter-blastomere variability at the 8-cell stage. A total of 1615 genes were selected by filtering for transcripts with TPM ≥ 1 in at least two blastomeres in all embryos and a mean coefficient of variation (CV) ≥ 0.8 across embryos. Columns represent individual blastomeres, grouped by embryo (E01–E05, color-coded). Genes displayed on the right correspond to cluster 2 and are associated with MAPK, Wnt, Rap1, and cellular senescence–related pathways; enrichment analysis for cluster 2 is shown in Supplementary Fig. 1. B Dot plot summarizing KEGG pathway over-representation analysis (ORA) of the highly variable gene set. Gene ratio indicates the proportion of input genes associated with each KEGG pathway. C Network representation of enriched signaling pathways, highlighting extensive gene overlap among pathways. Nodes represent pathways, and edges indicate shared genes. D Comparison of expression levels of representative trophectoderm-associated genes (CDX2, HRAS, MAPK14, and RAC1) between sister blastomeres at the 8-cell stage exhibiting high inter-blastomere variability (CV). Gene expression was analyzed from the 1-cell to 8-cell stages: 10 embryos at the 1-cell stage (E1–E10), two sister blastomeres from each of 10 embryos at the 2-cell stage (E1–E10), four sister blastomeres from each of five embryos at the 4-cell stage (E1–E5), and eight sister blastomeres from each of five embryos at the 8-cell stage (E1–E5). Each dot represents expression in a single blastomere. Gray shading on the right indicates the density distribution of expression levels.

Stratification of sister blastomeres by CDX2 expression reveals MAPK-associated transcriptional divergence

In mouse embryos, sister blastomeres with higher Cdx2 expression are more likely to contribute to the trophectoderm lineage²¹. We first stratified 8-cell-stage bovine blastomeres into high and low CDX2-expressing groups and performed PCA. High CDX2-expressing blastomeres were clustered tightly along the first principal component (PC1; Fig. 3A). A total of 365 differentially expressed genes (DEGs) were identified, all upregulated in high CDX2-expressing blastomeres relative to the low CDX2-expressing group (Supplementary Data 7). KEGG pathway enrichment analysis of DEGs using over-representation analysis (ORA) revealed that the MAPK signaling pathway was the most significantly enriched, followed by the Ras signaling and Wnt signaling pathways (Fig. 3B and Supplementary Data 7). Hierarchical clustering of CDX2-associated DEGs further revealed that, even among sister blastomeres derived from the same embryo, blastomeres were segregated into two major clusters based on divergent expression patterns of MAPK-related genes. In this analysis, blastomeres with high CDX2 expression tended to exhibit elevated expression of these MAPK-associated genes (Fig. 3C). Moreover, correlation analysis revealed that many MAPK-related genes showed strong co-expression with CDX2 at the 8-cell stage (Pearson’s r ≥ 0.5). Notably, this correlation was weak at the 2-cell stage but increased from the 4-cell stage onward (Fig. 3D, E), indicating the progressive strengthening of the association between MAPK signaling and CDX2 expression during cleavage-stage development. This relationship was further independently validated by real-time PCR analysis, which confirmed a positive correlation between CDX2 and MAPK14 at 8-cell stage (Supplementary Fig.4B).

Fig. 3. Comparative analysis of cleavage-stage blastomeres stratified by CDX2 expression.

A Principal component analysis (PCA) of individual 8-cell-stage blastomeres based on the 2000 genes showing the highest variance among 17,530 genes expressed at TPM ≥ 1 in at least one blastomere. Blastomeres were stratified into CDX2-high and CDX2-low groups according to CDX2 expression levels, with a threshold of 2.0 TPM. B Dot plot summarizing KEGG pathway over-representation analysis (ORA) of differentially expressed genes (DEGs; n = 363) between CDX2-high and CDX2-low blastomeres, showing the top 20 enriched pathways. C Heatmap showing hierarchical clustering of DEGs. Genes belonging to the MAPK signaling pathway and CDX2 are indicated on the right side of the heatmap. D Line plots illustrating stage-specific changes in pairwise Pearson correlation coefficients (r) among DEGs, calculated using Z_{ij} values derived from log₂(TPM + 1) expression after removal of between-embryo variance. DEG pairs were classified into four correlation patterns based on the stage at which r reached a maximum or minimum (4-cell maximum, 8-cell maximum, 4-cell minimum, or 8-cell minimum), and temporal changes in r are shown for each pattern. Colored lines highlight DEG pairs involving CDX2 and MAPK signaling–related genes, whereas black lines represent other DEG pairs. (E) Scatter plots showing correlations between CDX2 expression and the top eight MAPK signaling–related genes exhibiting the strongest correlations with CDX2 at the 8-cell stage, as identified in panel D Pearson correlation coefficients (r) are shown in each panel. Sample sizes: 2-cell stage (10 embryos, n = 20 blastomeres), 4-cell stage (5 embryos, n = 20 blastomeres), and 8-cell stage (5 embryos, n = 40 blastomeres).

Size-associated molecular and functional differences among blastomeres

In mice, it has been reported that at the 8-cell stage, larger blastomeres exhibit higher Cdx2 level and are more likely to contribute to the trophectoderm²². Therefore, we investigated the molecular and functional differences in lineage differentiation associated with blastomere size. At the 4-cell stage, the mean blastomere diameter was 70.3 µm (CV = 6.8%), and at the 8-cell stage it was 54.7 µm (CV = 7.0%), with size variability observed in all embryos. At both stages, the largest blastomeres were significantly larger than the smallest (4-cell: 76.8 vs. 66.0 µm; 8-cell: 60.3 vs. 49.3 µm), indicating asymmetric cleavage during early embryonic divisions (Fig. 4A).

Fig. 4. Relationship between blastomere size and gene expression asymmetry in 4- and 8-cell bovine embryos.

A Diameter measurements of individual blastomeres from five embryos (E1–E5) at the 4-cell and 8-cell stages, in which blastomeres were ranked by diameter (µm) to identify the smallest (S) and largest (L) cells within each embryo. Each dot represents a single blastomere (four per embryo at the 4-cell stage and eight per embryo at the 8-cell stage), color-coded by embryo of origin. Horizontal bars indicate mean diameters for each embryo (left panels) and for the S and L groups (right panels; n = 5 embryos each). B Scaled transcript levels (Z-scores of log₂(TPM + 1)) derived from single-cell RNA-seq of CDX2, HRAS, MAPK14, and RAC1 in the smallest (S) and largest (L) blastomeres at the 4-cell and 8-cell stages. Each dot represents an individual blastomere, color-coded by embryo of origin (n = 5 embryos per group). P values were calculated using paired Wilcoxon tests; p < 0.05 was considered statistically significant.

We compared the expression of the TE-associated genes shown in Fig. 2D between the smallest and largest blastomeres. RAC1 expression was significantly higher in the largest blastomeres at both the 4- and 8-cell stages. At the 8-cell stage, the other TE-associated genes tended to be higher in the largest blastomeres, as supported by a bootstrap analysis (Supplementary Table 2). These trends were further confirmed by real-time PCR, which showed higher expression of CDX2 and MAPK14 in the largest blastomeres compared with the smallest ones (Supplementary Fig. 4C). In the DEG analysis comparing the smallest and largest blastomeres, RAC1 was also identified, and FRAT1, a gene associated with Wnt signaling, was among the top-ranked genes (Supplementary Data 8).

In light of these size-associated transcriptional differences, we next examined nuclear translocation of YAP, a key regulator of TE differentiation²³. When the nuclear-to-cytoplasmic (N/C) YAP signal was compared between the smallest and largest blastomeres, the N/C distribution in the largest blastomeres was shifted to the right relative to that in the smallest blastomeres, indicating enhanced nuclear localization of YAP in the largest blastomeres (Fig. 5A, B).

Fig. 5. Functional asymmetries between the largest and smallest blastomeres.

A Representative confocal images of YAP immunofluorescence in day 5 morulae derived from individually cultured largest or smallest blastomeres isolated at the 8-cell stage, along with an intact control embryo (zona removed at the 8-cell stage without blastomere removal). YAP (green), DNA (Hoechst 33342, blue). White arrowheads indicate cells lacking nuclear YAP localisation. Scale bar, 20 µm. B Density plots of YAP nuclear-to-cytoplasmic (N/C) ratios in embryos derived from the largest or smallest blastomeres. The dashed line indicates an N/C threshold of 2.0. C Time-lapse images of embryos derived from individually cultured largest or smallest blastomeres. Images are shown at the following time points, from left to right: frames 1, 97, 157, 246, 275, 286, 387, and 418; the upper and lower panels represent the same time points. The colored bars overlaid on the images indicate key developmental events: purple marks cell division (Division), green marks the onset of compaction (Compaction onset), orange marks the onset of cavitation (Cavitation oneset), and blue marks full cavitation (Cavitation full). All images are shown at the same magnification; the scale bar (100 μm) is shown in the bottom-right panel. The full time-lapse sequences corresponding to this panel are provided in Supplementary Movies 1 and 2. D Timing of compaction onset, cavitation onset, and full cavitation in embryos derived from individually cultured smallest or largest blastomeres. Dots represent individual embryos; bars indicate mean. Different letters indicate statistically significant differences between groups (p < 0.05, unpaired Wilcoxon test), whereas groups sharing the same letter are not significantly different. D Timing of compaction onset, cavitation onset, and full cavitation in embryos derived from individually cultured smallest or largest blastomeres. Dots represent individual embryos; bars indicate mean. Different letters indicate statistically significant differences between groups (p < 0.05, unpaired Wilcoxon test), whereas groups sharing the same letter are not significantly different. Sample sizes: compaction onset (S, n = 59; L, n = 62), cavitation onset (S, n = 37; L, n = 46), and full cavitation (S, n = 21; L, n = 27). E Blastomere number at compaction onset between smallest (n = 18) and largest blastomeres (n = 20). Horizontal bars indicate mean values. P values were calculated using a paired Wilcoxon test. F Representative immunofluorescence images of day 8 blastocysts derived from individually cultured largest or smallest blastomeres. DNA (Hoechst 33342, gray), CDX2 (green), SOX2 (red). Scale bars, 50 µm. G Total cell number, lineage-specific cell counts, and the proportion of CDX2-positive cells (including CDX2⁺/SOX2⁻ and CDX2⁺/SOX2⁺ cells) relative to total cell number in blastocysts derived from smallest (n = 5) or largest (n = 8) blastomeres. Cell type categories include CDX2⁺/SOX2^–, CDX2^–/SOX2⁺, CDX2^+/SOX2⁺, and CDX2^–/SOX2^–. Bars represent mean ± SD. In this figure, + denotes positive staining and − denotes negative staining. P values were calculated using a Wilcoxon test; p < 0.05 was considered statistically significant. H Schematic summary illustrating that embryos derived from the largest blastomeres undergo one additional cleavage before compaction, compact earlier, exhibit higher nuclear YAP localization, and show a greater capacity to cavitate—despite having a similar timing of cavitation—compared with embryos derived from the smallest blastomeres.

In developmental kinetics, the onset of compaction occurred earlier in the largest embryos than in the smallest ones, whereas no difference was observed in the timing of cavitation (Fig. 5C, D and Supplementary Movie 1 and 2). Notably, despite the earlier onset of compaction, the number of blastomeres at the time of compaction was significantly higher in the largest embryos, indicating that they underwent one additional round of cleavage before compaction (Fig. 5E).

Following these kinetic differences, cavitation occurred more readily and was more stably maintained in the largest blastomeres than in the smallest blastomeres (Fig. 5C and Table 1). Cavitated embryos derived from the largest blastomeres tended to have a higher total cell number, whereas no significant differences were observed in the numbers of CDX2-positive or SOX2-positive cells, nor in the proportion of CDX2-positive cells relative to total cell number (Fig. 5F, G). In addition, S- and L-derived embryos that formed and maintained a cavity showed no differences in the expression of inner cell mass– or trophectoderm-associated genes (Supplementary Fig. 5).

Table 1.

Comparison of cavitation outcomes between the smallest and largest blastomeres up to 192 hpi

Group	Cultured blastomeres (n)§	Compaction, n(%)	Cavitation observed, n(%)†	Cavitation maintained, n(%)‡
Smallest	20	18(90)	9(45)	2(10)
Largest	20	20(100)	16(80)*	9(45)*

§ Data are paired within embryos (n = 20; one S and one L per embryo).

† Cavitation detected at least once during culture up to 192 hpi.

‡ Cavitation present at 192 hpi (maintained until 192 hpi).

* P < 0.05 vs Smallest (exact McNemar test; observed p = 0.039, maintained p = 0.016; two-sided).

Together, these findings support a model in which larger embryos generate more clonal blastomeres prior to compaction, display TE-associated molecular features, and exhibit an increased propensity for cavitation compared with the smallest embryos (Fig. 5H).

The largest blastomeres preferentially contribute to trophectoderm expansion in the blastocyst

Because single blastomeres isolated at the 8-cell stage lack sufficient cell numbers to reliably support blastocyst formation and lineage expansion, in contrast to blastomeres isolated at earlier stages^22,24, we generated reconstructed embryos from the smallest and largest blastomeres to functionally assess how blastomere size influences trophectoderm and inner cell mass contributions at the blastocyst stage (Fig. 6A). Both rec-S and rec-L embryos exhibited cavitation in all cases. However, post-cavitation expansion, as assessed by the increase in embryo area over time, was attenuated in rec-S embryos compared with rec-L embryos (Fig. 6B, C and Supplementary Movie 3 and 4). Consistent with this difference, rec-L embryos had a significantly higher total cell number and an increase in CDX2⁺/SOX2⁻ cells. In addition, CDX2⁻/SOX2⁺ cells, representing the inner cell mass, showed a tendency toward higher numbers in rec-L embryos. Consequently, the proportion of CDX2-positive cells relative to total cell number did not differ significantly between groups (Fig. 6D, E).

Fig. 6. Developmental outcomes of reconstructed embryos derived exclusively from the smallest or largest blastomeres of 8-cell bovine embryos.

A From each individual 8-cell embryo, blastomeres were ranked by relative diameter, and the smallest and largest blastomeres were separately isolated and aggregated to generate reconstructed 8-cell embryos enriched for relatively small (rec-S) or relatively large (rec-L) blastomeres. Reconstructed embryos derived from the same original embryo were cultured to the blastocyst stage to assess developmental potential and lineage composition. B Representative time-lapse images showing the developmental progression of reconstructed embryos derived from the smallest and largest blastomeres. Embryos were monitored for 6 day post-imaging. The numbers above each image indicate the frame number from the start of recording. All images are shown at the same magnification; the scale bar (100 μm) is shown in the bottom-right panel. The corresponding time-lapse sequences are provided in Supplementary Movies 3 and 4. C Temporal changes in embryo area (µm²) in reconstructed embryos. Panels E1–E3 show representative paired rec-S (blue) and rec-L (pink) embryos derived from three original embryos. D Representative immunofluorescence images of blastocysts derived from reconstructed embryos. DNA (Hoechst 33342, gray), SOX2 (green), and CDX2 (red) staining illustrate lineage allocation patterns in blastocysts derived from smallest-blastomere and largest-blastomere reconstructions. Scale bars, 100 μm. E Quantification of total cell number, lineage-specific cell counts (CDX2⁺/SOX2⁻, CDX2⁻/SOX2⁺, CDX2⁺/SOX2⁺, and CDX2⁻/SOX2⁻), and the proportion of CDX2-positive cells (CDX2⁺/SOX2⁻ and CDX2⁺/SOX2⁺) relative to total cell number in blastocysts produced from smallest- or largest-blastomere reconstructions. Each dot represents an individual blastocyst (n = 12 per group). Bars indicate mean ± SD. P values were calculated using a paired Wilcoxon test. In this figure, + denotes positive staining and − denotes negative staining.

Discussion

Early cleavage stage embryos are traditionally considered as collections of equivalent blastomeres, each retaining the capacity to equally contribute to the embryonic and extraembryonic lineages until the onset of compaction^3,4. However, studies on mouse embryos have challenged this view by suggesting that early cleavage divisions can generate asymmetries in gene expression, cell behavior, and developmental potential^{8,10,11,25,26}. Whether a similar early divergence occurs in non-rodent mammals remains unclear.

Here, scRNA-seq analysis revealed that transcriptional divergence between sister blastomeres was detectable at the 4-cell stage but became more pronounced at the 8-cell stage in bovine embryos. Notably, genes associated with the MAPK signaling pathway exhibited substantial inter-blastomere variability at the 8-cell stage. Given the role of Ras–MAPK signaling in regulating trophectoderm (TE) specification²⁰, this early transcriptional asymmetry is consistent with the onset of lineage bias prior to morphological polarisation, such as compaction. In addition to MAPK-associated genes, components of the Wnt signaling pathway also exhibited inter-blastomere variability at the 8-cell stage, indicating that heterogeneity across multiple signaling pathways implicated in TE lineage competence and early cell-state regulation is already present during early cleavage-stage development^27–29.

Importantly, these MAPK- and Wnt-associated genes (Fig. 4B and Supplementary Fig.6) were already expressed at the zygote (1-cell) stage, prior to major embryonic genome activation (EGA), which occurs at the 8-cell stage in bovine embryos³⁰. This suggests that the observed variability likely originates from differential inheritance or depletion of maternal transcripts rather than de novo transcription driven by EGA. Together, these findings raise the possibility that early asymmetries in the distribution of maternal factors contribute to the subsequent emergence of lineage bias.

In addition to transcriptional divergence, large blastomeres at the 8-cell stage showed integrated molecular and functional profiles that may predispose them toward TE lineage outcomes. These findings conceptually align with previous findings in mouse embryos, where blastomere size correlate with elevated Cdx2 levels and a higher probability of contributing to the TE lineage³¹. Interestingly, the largest (L) blastomeres displayed shorter cell-cycle durations prior to compaction compared with the smallest blastomeres (Fig. 5E, H). In mouse embryos at 8- and16-cell stages, outer cells that give rise to the TE are known to exhibit shorter cell cycles than inner cells that contribute to the inner cell mass (ICM)³². Accordingly, the variation in cell-cycle duration observed among blastomeres in bovine embryos may represent a conserved feature of early lineage bias across mammals.

Despite the emergence of early molecular asymmetries, developmental plasticity was not immediately lost, as large blastomeres, although more prone to generating TE cells, still retained the capacity to contribute to both TE and ICM regardless of their initial size. This suggests that early lineage bias does not inevitably compromise blastomere bipotency. This supports the concept of mosaic plasticity, wherein early asymmetries bias developmental trajectories without deterministically fixing the cell fate^8,11.

The presence of early transcriptional and functional asymmetries in bovine embryos, which exhibit developmental dynamics more comparable to those of humans than to those of rodents, suggests that symmetry-breaking processes are broadly conserved across mammals. In this context, previous studies provide important context for interpreting these findings. Lavagi et al. (2018) reported blastomere-to-blastomere variability in lineage-associated genes, including CDX2, at the 8-cell stage, although they did not interpret this heterogeneity as the onset of lineage bifurcation¹⁸. More recently, Hu et al. (2024) demonstrated that the first unequivocal transcriptional separation between the ICM and TE lineages emerges at the early blastocyst stage³³. Together, these studies indicate that bovine lineage segregation becomes transcriptionally defined relatively late. Rather than contradicting these reports, our findings instead suggest that early transcriptional divergence functions as a priming phase that biases developmental trajectories prior to definitive lineage separation between the ICM and TE lineages. This pattern aligns with conceptual models of symmetry-breaking, in which subtle asymmetries—generated by stochastic partitioning during cleavage divisions and reinforced by transcriptional circuits—gradually bias cells toward distinct developmental trajectories before overt lineage segregation occurs³⁴.

The present findings have potential implications for assisted reproductive technologies in humans and livestock, as lineage bias is already emerging and developmental plasticity is not uniform among blastomeres even before overt lineage segregation. This raises the possibility that interventions at the cleavage stage, such as blastomere biopsy, may inadvertently disrupt subsequent lineage differentiation^35–37.

In summary, sister blastomeres exhibited progressive transcriptional divergence prior to compaction in bovine embryos—paralleling observations in rodent models—and these early differences may influence subsequent lineage trajectories. By demonstrating this phenomenon in a non-rodent species with developmental dynamics more comparable to those of humans, our study provides a valuable framework for understanding early cell-fate regulation and its implication for stem cell biology, as well as advancing assisted reproductive technologies.

Methods

Ethics statement

Bovine ovaries were obtained from a local slaughterhouse. The samples were obtained from adult female cattle. As no live animals were used for experimental procedures and tissues were collected post-mortem from animals processed for commercial purposes, ethical approval was not required. All procedures complied with institutional guidelines of the Tokyo University of Agriculture and Technology. We have complied with all relevant ethical regulations for animal use.

Reagents

All reagents were purchased from Sigma-Aldrich (St. Louis, MO, USA), unless otherwise stated.

Oocyte collection

Ovaries were obtained from the Japanese Black or crossbred (Holstein × Japanese Black) cows at a local abattoir and transported to the laboratory at the Tokyo University of Agriculture and Technology. Upon arrival, the ovaries were rinsed with and maintained in 0.9% saline solution (Nippon Zenyaku Kogyo, Fukushima, Japan) at 38.5 °C. Cumulus–oocyte complexes (COCs) were aspirated from 2–6-mm antral follicles using a 19 G needle attached to a 10-mL syringe.

In vitro maturation

In vitro maturation of COCs was performed using TCM-199 with 25 mM HEPES (Gibco, Thermo Fisher Scientific, MA, USA) supplemented with 5% calf serum (CS; Gibco) and 0.1IU/mL recombinant human follicle-stimulating hormone (Follistim; MSD, Tokyo, Japan)³⁸. After washing twice with the maturation medium, 50 COCs were cultured in 500 µL of the same medium in 4-well culture plates (Thermo Fisher Scientific) and overlaid with mineral oil (FUJIFILM Wako Pure Chemical Corporation, Osaka, Japan) at 38.5 °C in a humidified atmosphere of 5% CO₂ in air for 22 h.

In vitro fertilization

In vitro fertilization was performed as previously described³⁸. Frozen semen of a Japanese Black bull was thawed at 37 °C in a water bath and layered into 3 mL of 90% Percoll solution, followed by centrifugation at 740 × g for 10 min. The pellet was resuspended in the Brackett–Oliphant (BO) medium³⁹ containing 20 mM hypotaurine, 10 µg/mL heparin (heparin sodium injection; 5000 U/5 mL; Mochida Pharmaceutical, Tokyo, Japan), and 20 mg/mL bovine serum albumin (BSA; crystallised and lyophilised) and washed via centrifugation at 540 × g for 5 min. The final sperm concentration was adjusted to 3 × 10⁶/mL in the BO medium supplemented with 20 mg/mL BSA. In vitro fertilization was performed in 100-µL droplets covered with mineral oil in 35-mm culture dishes (Nunc), each containing 20 COCs. After washing twice with the BO medium containing 10 mg/mL BSA, the COCs were co-incubated with sperms at 38.5 °C in 5% CO₂ in air for 6 h.

In vitro culture

After fertilisation, the COCs were denuded of the remaining cumulus cells and sperms via gentle pipetting in the CR1aa medium supplemented with amino acids and 5% CS. Embryos were cultured in groups of 50 in 500 µL of the same medium in 4-well plates and overlaid with mineral oil. In vitro culture was performed at 38.5 °C under low oxygen conditions (5% O₂, 5% CO₂, and 90% N₂).

Blastomere isolation

Blastomeres were separated from the 2-, 4-, and 8-cell stage embryos 28, 36, and 48hpi, respectively⁴⁰. Only 2-cell stage embryos at 28 hpi were used for subsequent 4- and 8-cell stage embryo collection to ensure normal first cleavage and developmental synchrony. Embryos were selected based on normal cleavage patterns as determined by time-lapse imaging (see Time-lapse imaging section), and embryos showing abnormal cleavage or cytoplasmic fragmentation were excluded. Zona pellucida was removed via treatment with 0.25% pronase (actinase E; Kaken Pharmaceutical, Tokyo, Japan) dissolved in phosphate-buffered saline (Gibco). The embryos were further transferred to the CR1aa medium containing 10% CS and mechanically dissociated into individual blastomeres via gentle pipetting using a glass capillary.

Single-blastomere RNA-sequencing

For transcriptomic analysis, 90 samples, including 10 zygotes (1-cell), 20 blastomeres from 10 2-cell embryos, 20 blastomeres from five 4-cell embryos, and 40 blastomeres from five 8-cell embryos, were collected. Blastomeres were isolated as described above. Before sampling, each blastomere was imaged to record its diameter using the ImageJ software (NIH), and sister blastomeres from the same embryo were tracked for within-embryo comparisons.

The overall protocol was adapted from our previously described method for single-cell analysis⁴¹, with some modifications. Total RNA extraction, reverse transcription, and cDNA library preparation were performed using the SMART-Seq HT PLUS Kit (Takara Bio, Shiga, Japan). cDNA quality was assessed using the Agilent 2200 TapeStation system with the High Sensitivity D5000 ScreenTape (Agilent Technologies, CA, USA). Sequencing was conducted on the NovaSeq 6000 platform (Illumina, CA, USA) using 150 bp paired-end reads.

Adapter trimming was performed using Trim Galore, and sequence quality was assessed using FastQC. The reads were aligned to the bovine reference genome (ARS-UCD1.2/bosTau9) using STAR, and transcript quantification was performed using RSEM, and gene expression levels were normalized to transcripts per million (TPM). Raw read counts and TPM values are provided in Supplementary Data 1 and 2, respectively. Count data were variance-stabilized using the variance stabilizing transformation (VST) implemented in DESeq2 (ver. 1.48.1). Principal component analysis (PCA) was performed using the prcomp function in R, based on VST-transformed and z-score–scaled data, and visualized with ggplot2. Hierarchical clustering and heatmaps were generated using pvclust and ComplexHeatmap (ver. 2.24.1), respectively.Differentially expressed genes (DEGs) were identified using DESeq2. Log₂ fold changes, P-values, and false discovery rates (FDRs) were estimated using the DESeq function, with multiple testing correction performed according to the Benjamini–Hochberg method. Shrinkage of log₂ fold-change values was applied using the lfcShrink function with the ashr method. Genes with an FDR < 0.1 and an absolute log₂ fold change ≥ 1 were defined as DEGs. KEGG pathway enrichment analysis was performed using over-representation analysis (ORA) on gene sets identified in each analysis. Enrichment was assessed using the enrichKEGG function in clusterProfiler (ver. 4.16.0; default parameters). In selected analyses, enrichment was independently validated using the gost function in gprofiler2 (ver. 0.2.4; user-defined threshold = 0.05, correction method = g_SCS). Enrichment results were visualized using ggplot2. Gene–gene association networks within enriched KEGG pathways were constructed using igraph (ver. 2.1.4) and visualized with ggraph (ver. 2.2.2).

Assessment of transcriptional asymmetry

To assess the transcriptional asymmetry within embryos, gene-level variability was quantified as total sum of squares (SS) and partitioned into between-embryo (SSbe) and within-embryo (SSwe) components, as previously described¹⁰ (Fig. 1D). The ratio SSwe/SS, calculated from log₂(TPM + 1)–transformed expression values, was used as the primary metric to capture transcriptional heterogeneity among sister blastomeres within individual embryos. In addition, Z_{ij} values were computed following the same framework. To facilitate the identification of genes exhibiting marked inter-blastomere variability, the coefficient of variation (CV) was additionally calculated across blastomeres and used as a complementary metric for downstream analyses. Analyses were performed using genes that met minimum expression criteria to ensure reliable variability estimates: for the 2- and 4-cell stages, genes were required to have TPM ≥ 1 in at least one blastomere in each embryo, whereas for the 8-cell stage, genes were required to have TPM ≥ 1 in at least two blastomeres within an embryo. The SS, SSbe, SSwe and CV values used in these analyses are provided in Supplementary Data 3. Highly variable genes were defined by restricting the analysis to sufficiently expressed transcripts (TPM ≥ 1 in at least two blastomeres in all embryos) and applying a stringent CV threshold (CV ≥ 0.8), which corresponds to the upper range of expression variability. The resulting gene set is provided in Supplementary Data 4.

Classification of blastomeres based on CDX2 expression

At the 8-cell stage, individual blastomeres were stratified into CDX2-high and CDX2-low groups based on CDX2 transcript abundance. Blastomeres with CDX2 expression ≥ 2.0 TPM were classified as CDX2-high, whereas those with CDX2 expression < 2.0 TPM were classified as CDX2-low. For correlation analyses, gene expression values were standardized to Z-scores (Zij) across blastomeres prior to calculation of Pearson correlation coefficients.

Time-lapse imaging

Each blastomere was individually cultured in a microwell of the LinKID micro25 dish (Dai Nippon Printing, Tokyo, Japan) containing 125 µL of the CR1aa medium supplemented with 10% CS and overlaid with mineral oil (FUJIFILM Wako Pure Chemical Corporation). Time-lapse imaging was performed using a real-time embryo culture observation system (CCM-MULTI and CCM-iBIS-L_W10; Astec, Fukuoka, Japan), with images captured every 15 min using a 10× objective lens^38,41, Tsuji et al., 2025. Embryos were monitored up to 192 h post-insemination (hpi). Retrospective analysis was performed to determine the timing of compaction onset, cavitation onset, and full cavitation, defined as follows. Compaction onset was defined as the time at which blastomeres first lost visible cell–cell boundaries; cavitation onset, as the first appearance of a fluid-filled cavity; and full cavitation, as the stage when the blastocoel expanded to occupy most of the embryo volume and was enclosed by a continuous TE cell layer.

Immunofluorescence staining

Embryos were fixed in 4% paraformaldehyde (PFA) in PBS at room temperature for 60 min, followed by three washes in 0.1% polyvinyl alcohol (PVA)–PBS for 10 min each at room temperature. Embryos were then permeabilized in 0.2% Triton X-100–PBS at room temperature for 20 min and washed three times in 0.1% PVA–PBS for 10 min each. Blocking was performed using Blocking One (Nacalai Tesque, Kyoto, Japan) diluted fivefold in PBS at room temperature for 20 min. Embryos were subsequently incubated overnight at 4 °C with primary antibodies diluted in Blocking One diluted 20-fold in PBST (0.1% Tween 20–PBS). The following primary antibodies were used: rabbit monoclonal anti-SOX2 antibody (ab92494; Abcam, Cambridge, UK) diluted 1:100, mouse monoclonal anti-CDX2 antibody (MU392A-5UC; BioGenex, Fremont, CA, USA) diluted 1:100, and mouse monoclonal anti-YAP1 antibody (WH0010413M1; Sigma-Aldrich, St. Louis, MO, USA) diluted 1:200. After primary antibody incubation, embryos were washed three times in 0.1% PVA–PBS for 10 min each at room temperature and incubated with the appropriate secondary antibodies at room temperature for 60 min. For SOX2 and CDX2 staining, Alexa Fluor™ 555 goat anti-rabbit IgG (A21428; Invitrogen, Thermo Fisher Scientific, MA, USA) diluted 1:400 and Alexa Fluor™ 488 anti-mouse IgG diluted 1:200 were used. For YAP1 staining, Alexa Fluor™ 488 goat anti-mouse IgG diluted 1:200 was used. Following secondary antibody incubation, embryos were washed twice in 0.1% PVA–PBS for 10 min each. Nuclear counterstaining was performed with Hoechst 33342 diluted 1:1000 in 0.1% PVA–PBS at room temperature for 10 min, followed by a final wash in 0.1% PVA–PBS for 5 min. Embryos were mounted on glass slides using VECTASHIELD® Mounting Medium (Vector Laboratories, CA, USA). Coverslips were supported by eight pillars made from a mixture of Vaseline (FUJIFILM Wako Pure Chemical Corporation, Osaka, Japan) and liquid paraffin, and sealed with nail polish. Fluorescence images were acquired using a confocal laser scanning microscope (AX R; Nikon, Tokyo, Japan). Cell numbers were manually counted using ImageJ (NIH, Bethesda, MD, USA) based on Hoechst-stained nuclei. CDX2- and SOX2-positive cells were identified according to nuclear fluorescence signals, and double-positive cells were classified accordingly. Quantification of YAP nuclear-to-cytoplasmic (N/C) intensity ratios was performed using ImageJ. Nuclear regions of interest were defined based on Hoechst staining, and cytoplasmic regions were delineated using Voronoi segmentation. Mean fluorescence intensities were measured after background subtraction, and N/C ratios were calculated for each cell.

Statistics and reproducibility

Statistical analyses were performed primarily using embryo-matched comparisons for analyses involving smallest (S) and largest (L) blastomeres derived from the same embryo. For continuous variables, paired two-sided Wilcoxon signed-rank tests were used unless otherwise stated. Comparisons between reconstructed embryos derived from the same original embryo (rec-S vs. rec-L) were also analyzed using paired two-sided Wilcoxon signed-rank tests. For analyses not based on embryo-matched samples, unpaired two-sided Wilcoxon rank-sum tests were applied. For categorical developmental outcomes (Table 1), paired proportions were compared using an exact McNemar test (two-sided). All statistical analyses were performed using R (version 4.3.1). Statistical significance was set at P < 0.05. Statistical analyses specific to RNA-seq data are described separately in the RNA-seq analysis section.

Author contributions

S.S. and H.K. conceptualized the study. H.K., P.F., U.S. and R.M. conducted the experiments. S.S., H.K., D.M. and T.Y. conducted data analysis. M.K. and A.K. provided support for RNA-seq analysis. H.T (Haruhisa Tsuji). provided support for time-lapse imaging. S.S. and H.K. wrote the original draft of the manuscript. S.S., H.K., D.M. and T.Y. prepared the figures. S.S., H.K., D.M. and T.Y. reviewed and edited the manuscript. S.S. supervised the project and secured funding.

Peer review

Peer review information

Communications Biology thanks Janet Rossant, Peter L. Pfeffer and Michele Boiani for their contribution to the peer review of this work. Primary Handling Editors: Manuel Breuer and George Inglis. A peer review file is available.

Data availability

The RNA-seq data generated in this study have been deposited in the Gene Expression Omnibus (GEO) under accession number GSE301333. Source data for Figs. 1E, 2D, and 3A are provided in Supplementary Data 3, 9, and 1, respectively. Source data for Figs. 3D–E, 4, 5C, 5D–E, 5G, 6C, and 6E are provided in Supplementary Data 9.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s42003-026-10198-9.