A program of successive gene expression in mouse one-cell embryos

Highlights

• •A gene expression program, iEGA, initiates within 4 h of mouse fertilization
• •Upregulated genes are normatively spliced, protein coded, and soon downregulated
• •iEGA genes predict cancer-associated pathways and transcription regulators
• •Inhibiting predicted transcription regulators acutely disrupts iEGA and development

Summary

At the moment of union in fertilization, sperm and oocyte are transcriptionally silent. The ensuing onset of embryonic transcription (embryonic genome activation [EGA]) is critical for development, yet its timing and profile remain elusive in any vertebrate species. We here dissect transcription during EGA by high-resolution single-cell RNA sequencing of precisely synchronized mouse one-cell embryos. This reveals a program of embryonic gene expression (immediate EGA [iEGA]) initiating within 4 h of fertilization. Expression during iEGA produces canonically spliced transcripts, occurs substantially from the maternal genome, and is mostly downregulated at the two-cell stage. Transcribed genes predict regulation by transcription factors (TFs) associated with cancer, including c-Myc. Blocking c-Myc or other predicted regulatory TF activities disrupts iEGA and induces acute developmental arrest. These findings illuminate intracellular mechanisms that regulate the onset of mammalian development and hold promise for the study of cancer.

Introduction

In fertilization, two heterotypic cells (the gametes, sperm and oocyte) combine to cause formation of a totipotent one-cell embryo.¹ This is a foundational developmental event that coincides with embryonic genome activation (EGA), in which transcription in the new embryo initiates from gamete-derived genomes that had been transcriptionally silent until fertilization.^2,3 EGA occurs in a milieu of complex biochemical and physical changes, many unique, within nascent one-cell embryos.^3,4,5,6 Phospho-relays with multiple targets, including the cytostatic factor Emi2, precipitate meiotic cell cycle progression^3,6,7,8 and are concurrent with specialized parental chromatin remodeling,^3,9 which presumptively regulates EGA. The major sperm nucleoprotein protamine¹⁰ is removed by oocyte-derived nucleoplasmin activity and replaced by maternal histones prior to S phase, which begins ∼8 h after fertilization in the mouse.^3,11 The new embryo undergoes atypical patterns of histone modification^12,13 and chromosome organisation,^14,15,16 but conclusions differ regarding the extent to which chromatin structure is inherited from the gametes^15,17 or assembled anew.^14,16 Parental genomes become bounded by pronuclear membranes that, in the mouse, are visible ∼4.5 h after fertilization and remain for ∼10 h until the first mitotic prometaphase.^3,4,5The dynamics of mouse EGA have been inferred from injected reporter gene expression,¹⁸ bromodeoxyuridine (BrdU) labeling,¹⁹ cDNA library construction,²⁰ microarray analysis,^21,22,23 and RNA sequencing (RNA-seq)^24,25 to initiate as an event referred to as “minor” EGA in late one-cell embryos, followed by “major” EGA at the two-cell stage. However, these studies have often relied on embryos for which the time of fertilization in vivo was indeterminate even though oocytes are fertilizable for more than 12 h after ovulation,²⁶ the time of coitus and duration of sperm passage and fusion at the fertilization site are unknown,²⁷ and one-cell embryo morphology and time since fertilization are not reliably correlated.²⁸ Studies have sometimes used hundreds or thousands of embryos,^{20,21,22,24,29} potentially smoothing signals and obscuring biologically relevant differences.³⁰

These challenges preclude the degree of inter-embryo synchrony necessary for accurate transcriptome profiling in one-cell embryos. Moreover, the models they produce do not account for how maternal factor activity required for early development is regulated in the absence of endogenous transcription or address the cue that instigates gene expression, which can evidently be provided either in vivo or in distinct environments in vitro. Hints that transcription initiates at the early one-cell stage may also have been restricted by skewed library preparation protocols that potentially reflect mRNA polyadenylation (recruitment) rather than de novo gene expression.^{21,25,31,32,33}

 There are conflicting views about whether gene expression in one-cell embryos produces spliced mRNAs, with evidence of efficient canonical splicing^32,34,35 and the suggestion that it does not occur.²⁹

We recently employed polyadenylation-independent single-cell RNA-seq (scRNA-seq) of human embryos to address some of these issues, revealing that gene expression initiates at the one-cell stage.³⁴

 However, the paucity of available healthy one-cell human embryos for research hampers characterization. It precludes precise synchronization and time course profiling and confounds orthogonal validation, including functional corroboration of gene expression. It has therefore not been possible to determine precisely when EGA initiates in human one-cell embryos or whether it initiates stochastically, as a monotonic burst, or as a succession. Indeed, no time courses within the one-cell stage have been reported in any species, and the onset of EGA has been ascribed to the earliest time point at which upregulation has been determined, not necessarily the earliest point at which it occurs. The degree to which the onset of EGA is conserved between species is unknown, and models of EGA are thus incomplete. To address this, we set out to delineate EGA based on multi-platform, single-cell transcriptome time course profiling and characterization of precisely staged mouse one-cell embryos.

Results

Mouse fertilization rapidly triggers an embryonic transcription program

Synchronous one-cell mouse embryos were produced by precisely timed intracytoplasmic sperm injection (ICSI) of mature, metaphase II (mII) oocytes and collected at 2-h intervals for analysis by scRNA-seq. Precise embryo synchronization minimized noise to uncover hitherto inaccessible information about gene expression at the onset of development, and the scRNA-seq protocol avoided poly(A) capture and its attendant potential for library bias.

 We performed scRNA-seq on Mus musculus domesticus F2 hybrid (F2) and genetically distinctive, M. m. domesticus × M. m. castaneus (B6cast) embryos that developed efficiently in vitro (Figure S1A) and treated F2-B6cast scRNA-seq as one dataset to account for strain-specific effects and achieve higher statistical power. This gave an average read depth (±SEM) of 30.4 ± 1.3 million per mII oocyte or one-cell embryo. Unsupervised t-distributed stochastic neighbor embedding (t-SNE) visualization allocated each cell in the F2-B6cast time course to its appropriate position, producing clusters at each time point (false discovery rate [FDR] < 0.05; Figure 1A). An analogous time course series of independent F2 embryos subjected to microarray analysis also appropriately allocated each oocyte and embryo to its corresponding time point, corroborating scRNA-seq (FDR < 0.05; Figures 1A and 1B). The F2-B6cast scRNA-seq series comprised 4,067 differentially expressed genes (DEGs) across the 12-h time course in one-cell embryos relative to mature mII oocytes (FDR < 0.05). Of the DEGs detected by scRNA-seq, 368 (55.2%) changed in the same direction as the 667 DEGs detected by microarrays (FDR < 0.05 vs. FDR < 0.05), representing a strong Pearson correlation (r) of 0.78 (Figures 1C and S1B). Unsupervised gene-by-gene analysis of the F2-B6cast scRNA-seq series (FDR < 0.05) assigned embryos to corresponding respective positions on the time course, visualized in the heatmap…