The maternal-to-zygotic transition revisited

Early animal development is controlled by maternal mRNAs and proteins that are loaded into the egg during oogenesis. These implement fundamental molecular and cellular processes, as well as the specification of initial cell fates and pattern. As development proceeds, control is handed over from maternally provided gene products to those synthesized from the zygotic genome. This maternal-to-zygotic transition (MZT) is characterized by multiple events, the most striking among which are the elimination of a large subset of maternal gene products and the onset of transcription from the zygote's genome (Box 1).

The MZT has been referred to previously as a ‘play in two acts’ (Tadros and Lipshitz, 2009). Since then, technological advances have dramatically increased our understanding of the MZT. Although it remains a two-act play, many scenes and players have been added, and the plot has become more complex and better understood. Here, we describe how the gene expression program of embryos is shaped during the MZT. First, we describe the scale and dynamics of maternal control, as well as the processes that influence the maternal gene expression program. These include the regulation of maternal mRNA stability and translation, as well as the modulation of protein stability and functions by post-translational modifications. Then, we discuss how the zygotic gene expression program is activated. This includes roles for cell cycle length, transcriptional repressors and activators, and changes in chromatin organization. Finally, we describe how and why the maternal and zygotic gene expression programs are intertwined with each other to ensure successful handover of developmental control from one generation to the next.

All animal oocytes are loaded with mRNA species representing a large fraction of the protein-coding capacity of their genome (summarized in Table 1). Estimates from global analyses suggest that the maternal transcripts represent a range of the protein-coding transcriptome, from about one-third in the mouse (Wang et al., 2004) and Caenorhabditis elegans (Baugh et al., 2003; Stoeckius et al., 2014), to three-quarters in Drosophila melanogaster (De Renzis et al., 2007; Lécuyer et al., 2007; Tadros et al., 2007; Thomsen et al., 2010), the echinoderm Strongylocentrotus purpuratus (Wei et al., 2006) and the zebrafish, Danio rerio (Aanes et al., 2011; Harvey et al., 2013). During the MZT, a subset of the maternal transcripts is cleared from the embryo (Boxes 2 and 3). Estimates range from about one-quarter of the maternal transcripts in zebrafish (Aanes et al., 2011; Bazzini et al., 2012; Harvey et al., 2013; Mishima and Tomari, 2016), one-third in C. elegans and mouse (Baugh et al., 2003; Hamatani et al., 2004), to about two-thirds in Drosophila (Thomsen et al., 2010).

All animal embryos undergo phased maternal mRNA clearance; the initial phase is directed by maternally provided gene products whereas the later phases require zygotically synthesized products. The scale and dynamics of these phases varies across species, with respect to both time and developmental stage (Fig. 1; Table 1). Furthermore, technical differences among studies affect quantification (Boxes 2 and 3). Finally, in animals that set aside primordial germ cells in the early embryo, the MZT differs in soma versus germline (Box 4).

In Drosophila, an early maternally directed wave of decay is triggered upon egg activation and does not require fertilization; this is followed by one or more waves that require transcription of the zygotic genome and are therefore zygotically directed (Bashirullah et al., 2001; Bashirullah et al., 1999; Tadros et al., 2007; Tadros et al., 2003). About 25% of the cleared transcripts are degraded strictly by the maternal machinery, 35% strictly through the zygotic machinery, while 40% show mixed decay effected by both maternal and zygotic mechanisms (Fig. 1A) (De Renzis et al., 2007; Tadros et al., 2007; Thomsen et al., 2010). In C. elegans, about 30% of the maternal mRNAs are targeted for degradation by the one-cell stage, whereas another 30% are degraded starting at the four-cell stage, coinciding with the activation of zygotic transcription (Fig. 1B) (Baugh et al., 2003; Stoeckius et al., 2014).

In the echinoderm S. purpuratus, there appear to be three waves of maternal transcript degradation, but the role of maternal versus zygotic gene products in directing clearance is unknown (Fig. 1C) (Tu et al., 2014). In zebrafish, of the cleared maternal transcripts, about 60% degrade between the one-cell and 16-cell stages, while the remaining 40% degrade subsequently, coinciding with zygotic transcription (Aanes et al., 2011; Mathavan et al., 2005; Mishima and Tomari, 2016). Expression profiling in zebrafish lacking TATA-binding protein (Ferg et al., 2007) or treated with α-amanitin (Mishima and Tomari, 2016), and therefore unable to undergo zygotic genome activation, have shown that the latter wave of clearance requires zygotic transcription (Fig. 1D). In Xenopus tropicalis, about one-third of the maternal transcripts are cleared before the major wave of zygotic genome activation (ZGA) (Tan et al., 2013), while 15% are cleared after ZGA (Fig. 1E) (Graindorge et al., 2006).

In the mouse, the onset of the earliest wave of maternal transcript degradation is triggered by completion of meiosis (i.e. before fertilization and thus not pictured in Fig. 1F); a second wave of degradation is triggered by fertilization; and a third, zygotically directed, wave follows the major activation of the zygotic genome after the two-cell stage (Hamatani et al., 2004; Svoboda et al., 2015). Gene expression profiling in human embryos suggests that they too undergo at least two waves of maternal transcript clearance (Fig. 1G) (Dobson et al., 2004; Zhang et al., 2009b).

As maternal mRNAs can be stable for days, weeks, months or years in oocytes but are then degraded rapidly in a matter of hours or, at most, days, understanding this component of the MZT requires elucidation of the mechanisms underlying both long-term transcript stabilization and rapid transcript destabilization. These processes are regulated by RNA-binding proteins, small non-coding RNAs, RNA modifications and ‘codon optimality’ (Fig. 2).

RNA-binding proteins (RBPs) regulate either stabilization of stored maternal mRNAs in oocytes or their destabilization during the MZT. The particular RBPs that mediate these processes appear to vary across species (Table 1). However, in general, there have not been any systematic studies aimed at assessing whether RBPs identified as key regulators in one species exert similar effects in others.

In X. laevis, the Y-box RBP FRGY2 binds without sequence specificity to maternal mRNAs and functions to translationally repress and stabilize them (Bouvet and Wolffe, 1994; Matsumoto et al., 1996). Similarly, the zebrafish Y-box protein, Ybx1, globally represses maternal mRNA translation in oocytes (Sun et al., 2018), and in mouse oocytes, the Y-box protein MSY2 stabilizes maternal transcripts (Medvedev et al., 2011; Yu et al., 2001; Yu et al., 2002). Here, upon oocyte maturation, MSY2 is phosphorylated by the CDC2A (CDK1) kinase and MSY-bound mRNAs are released (Medvedev et al., 2008). In C. elegans and Drosophila, Y-box proteins are expressed during oogenesis but a role in transcript stabilization has not yet been identified (Arnold et al., 2014; Boag et al., 2005; Mansfield et al., 2002). Instead, homologs of the dead-box helicase Dhh1 (CGH-1 in C. elegans, ME31B in Drosophila) have been implicated in stabilization and translational repression of maternal mRNAs (Arnold et al., 2014; Boag et al., 2005; Wang et al., 2017).

The role of RBPs as trans-acting factors that regulate and coordinate maternal transcript degradation has been particularly well studied in Drosophila embryos (Fig. 2A). Two RBPs, Smaug, a SAM-domain RBP (Chen et al., 2014; Tadros et al., 2007) and Brain tumor (BRAT), a TRIM-NHL RBP (Laver et al., 2015a), regulate the maternally directed wave of decay. Subsequent, zygotically directed decay can be separated into two phases. The first is driven by miRNAs (Benoit et al., 2009; Bushati et al., 2008) (discussed below). The second is mediated by Brain tumor (BRAT) and possibly also Pumilio (PUM), a PUF-domain RBP (Gerber et al., 2006; Laver et al., 2015a,b; Thomsen et al., 2010). The smaug mRNA itself is translationally repressed during oogenesis by PUM and additional unknown factors; repression is relieved upon egg activation by the Pan gu (PNG) serine/threonine kinase (Fig. 3A) (Tadros et al., 2007). Smaug binds to hairpin structures in its target mRNAs (Aviv et al., 2003; Chen et al., 2014; Dahanukar et al., 1999; Semotok et al., 2008; Smibert et al., 1999) and can trigger their degradation by recruiting the CCR4-NOT-deadenylase complex (Fig. 2A) (Semotok et al., 2005). Smaug also represses target mRNA translation via recruitment of Argonaute 1 and the eIF4E-binding protein Cup (Nelson et al., 2004; Pinder and Smibert, 2013). The majority of the targets of Smaug are both repressed and degraded, while smaller subsets are degraded but not repressed, or vice versa (Chen et al., 2014; Nelson et al., 2007; Semotok et al., 2005; Semotok et al., 2008). PUM and BRAT bind to single-stranded motifs in their target transcripts (Laver et al., 2015a; Loedige et al., 2014). BRAT directs degradation of hundreds of maternal transcripts via both the maternal and the zygotic pathways (Laver et al., 2015a,b), whereas PUM appears to participate predominantly in zygotically directed clearance (Gerber et al., 2006; Laver et al., 2015a,b). In the maternal wave of decay, the targets of BRAT do not significantly overlap with those of Smaug (Laver et al., 2015a), suggesting that these two RBPs act independently of each other on distinct subsets of the maternal transcriptome. Interestingly, the smaug mRNA is itself targeted for degradation at the end of the MZT by BRAT, and possibly other factors, including PUM (Fig. 3A) (Laver et al., 2015a,b). Smaug, BRAT and PUM all promote transcript deadenylation (Newton et al., 2015; Semotok et al., 2005; Temme et al., 2010; Weidmann et al., 2014). Thus, these RBPs act as specificity factors that recruit a conserved degradation machinery to clear specific subsets of maternal mRNAs.

Another class of RBPs implicated in maternal mRNA clearance comprises the AU-rich-element binding proteins (ARE-BPs). ARE-BPs have a role in transcript destabilization during the MZT in C. elegans (D'Agostino et al., 2006; Gallo et al., 2008; Schubert et al., 2000), zebrafish (Rabani et al., 2017) and X. laevis (Detivaud et al., 2003; Graindorge et al., 2006; Paillard et al., 1998; Voeltz and Steitz, 1998), and possibly also mouse (Fig. 2A) (Ramos et al., 2004). Although a motif resembling AREs is enriched in Drosophila mRNAs that are destabilized during the MZT (De Renzis et al., 2007), a role for ARE-BPs has not yet been identified in Drosophila. In Xenopus, ARE-BPs appear to function together with the embryonic deadenylation element binding protein (EDEN-BP, also known as CUGBP1) (Graindorge et al., 2008), which directs maternal transcript clearance via the recruitment of the poly(A) ribonuclease (PARN) (Moraes et al., 2006). In zebrafish, a recent study has implicated hnRNPA1, an RBP that binds to the 3′UTR of maternal transcripts through an AGGGA motif, in clearance of a subset of maternal mRNAs after ZGA (Despic et al., 2017).

In the mouse, the oocyte-expressed protein BTG4, although not itself an RBP, links eIF4E to the CCR4-NOT-deadenylase complex to trigger degradation of maternal mRNAs that are actively translated (Fig. 2A) (Yu et al., 2016). Embryos from female Btg4^−/− mice produce oocytes that complete meiosis but fail to clear maternal transcripts. Reminiscent of the case for Drosophila smaug mutants, zygotically transcribed genes are downregulated and the embryos fail to develop to the two-cell stage (Yu et al., 2016). Furthermore, also reminiscent of smaug mRNA translation at egg activation being dependent on the PNG kinase (Tadros et al., 2007), translation of BTG4 protein is triggered by activation of the MAP kinases ERK1 and ERK2 during meiotic maturation (Yu et al., 2016).

Small non-coding RNAs that have been implicated in the MZT include microRNAs (miRNAs), which are processed from primary transcripts produced by miRNA-encoding genes (reviewed by Gebert and MacRae, 2019), PIWI-interacting RNAs (piRNAs; PIWI, P-element induced wimpy testis), which are derived from transposable elements resident in the genome (reviewed by Czech et al., 2018), and endogenous small interfering RNAs (endo-siRNAs), which are derived from double-stranded hybrids of mRNAs or transposable elements (reviewed by Piatek and Werner, 2014) (Table 1).

The most well studied are miRNAs, which play a crucial role in directing the zygotic wave of maternal transcript decay in multiple species (Fig. 2B). In zebrafish, the miR-430 family of miRNAs is expressed during early ZGA and is required for translational repression, deadenylation and decay of about 40% of the maternal transcripts that are cleared during the MZT (Fig. 2B) (Bazzini et al., 2012; Giraldez et al., 2006). An orthologous mechanism exists during the X. laevis MZT through its miR-430 ortholog, miR-427. As in zebrafish, miR-427 is expressed zygotically and mediates the deadenylation and clearance of maternal transcripts (Lund et al., 2009). An analogous pathway exists in Drosophila, mediated by zygotic transcription of miRNAs from the miR-309 cluster, which are required for the degradation of 14% of transcripts undergoing the late wave of decay (Fig. 2B) (Bushati et al., 2008). Zygotic expression of more than 70 species of miRNAs, including the miR-309 family, depends on the Smaug RBP (Benoit et al., 2009; Luo et al., 2016). In smaug mutants, failure to produce these miRNAs results in stabilization of a large number of maternal transcripts that are normally cleared in the zygotic wave of decay (Benoit et al., 2009; Luo et al., 2016). Thus, Smaug functions directly to clear maternal mRNAs prior to ZGA and indirectly to clear maternal transcripts after ZGA (Fig. 3A). The Drosophila zygotic miRNAs do not bear sequence homology to vertebrate miR-430/miR-427, suggesting that microRNA targeting of the maternal transcriptome during the MZT may have arisen independently in the vertebrate and invertebrate lineages.

piRNAs have also been implicated in maternal transcript decay in Drosophila (Barckmann et al., 2015; Dufourt et al., 2017; Rouget et al., 2010) where hundreds of maternal mRNAs have been reported to co-purify with PIWI, Argonaute 3 and Aubergine, and to be targeted by the piRNA pathway for degradation (Barckmann et al., 2015).

Endo-siRNAs (‘22G-RNAs’) have been shown to clear a subset of maternal mRNAs from early C. elegans embryos (Stoeckius et al., 2014). Although miRNAs and endo-siRNAs are present, miRNA activity appears to be suppressed in mouse oocytes and early embryos, and removal of DGCR8, which is essential for miRNA biogenesis, has no effect on preimplantation development (Ohnishi et al., 2010; Suh et al., 2010; Tang et al., 2007). It is possible that endo-siRNAs serve as regulators of maternal mRNA clearance (Fig. 2B) (Lykke-Andersen et al., 2008); however, conditional removal of AGO2 from mouse oocytes results in failure to complete meiosis, preventing analyses in early embryos (Stein et al., 2015).

Over the past few years, it has become clear that chemical modifications of mRNAs either during or after their synthesis can play a role in determining their fate. N⁶-methylation of adenosine (m⁶A), for example, can cause a structural switch in RNA that controls accessibility to protein-binding motifs (Liu et al., 2015). In zebrafish embryos, more than one-third of maternal mRNA species are methylated on adenosine (Zhao et al., 2017). The m⁶A- binding protein YTHDF2 is required for clearance of almost half of the maternal mRNAs that are degraded early in the MZT, as well as over one-quarter of those that are degraded later (Fig. 2C) (Zhao et al., 2017). Maternal YTHDF2 is also required for clearance of a subset of the transcriptome of the mouse oocyte (Ivanova et al., 2017). However, at least in zebrafish, not all methylated RNAs are destabilized, and not all transcripts that are degraded in a YTHDF2-dependent manner are methylated (Zhao et al., 2017). The exact mechanisms of regulation and the interplay between m⁶A-directed decay, and other pathways of transcript clearance, remain to be elucidated.

Another form of mRNA modification during the vertebrate MZT is terminal uridylation of the poly(A) tail of an mRNA (Chang et al., 2018). In zebrafish, Xenopus and mouse, terminal uridylation of the transcriptome increases greatly during the MZT and promotes rapid decay of a subset of transcripts with short poly(A) tails (Fig. 2C) (Chang et al., 2018).

In unicellular organisms, ‘optimal’ codons are defined as those with high cognate tRNA abundance, whereas non-optimal codons for the same amino acid have low cognate tRNA levels (Hanson and Coller, 2018). In yeast, the DEAD-box protein Dhh1p preferentially targets transcripts enriched for non-optimal codons by associating with slow-moving ribosomes and directing these transcripts for degradation (Presnyak et al., 2015; Radhakrishnan et al., 2016). Studies in zebrafish have found preferential degradation during the MZT of transcripts enriched for non-optimal codons (Bazzini et al., 2016; Mishima and Tomari, 2016), a process that may also occur in mouse, Xenopus and D. melanogaster (Bazzini et al., 2016) (Fig. 2D). It is not yet known whether proteins homologous to Dhh1p (DDX6 in mouse and fish, and ME31B in Drosophila) function in this process.

During Drosophila egg activation (i.e. at the beginning of its MZT), translational efficiency increases for up to 50% of maternal mRNAs and decreases for up to 20%, as measured by ribosome footprinting (Kronja et al., 2014b). The translation of maternal transcripts has also been profiled in mouse, where 20% of the maternal transcriptome is significantly depleted from polysomes during oocyte maturation, whereas 17% of the maternal transcriptome is significantly enriched (Chen et al., 2011). There is a direct correlation between poly(A) tail length and translational efficiency in early embryos of zebrafish, X. laevis and Drosophila (Eichhorn et al., 2016; Subtelny et al., 2014). These correlations diminish upon ZGA and are not found in non-embryonic tissue types across several species (Subtelny et al., 2014). Thus, the poly(A) tail may play a role in the regulation of translation primarily in transcriptionally silent stages.

The mechanisms of cytoplasmic polyadenylation and activation of maternal mRNA translation were first shown in X. laevis oocytes. Upon maturation, the Aurora kinase triggers binding of cytoplasmic polyadenylation element binding protein (CPEB) to cytoplasmic polyadenylation specificity factor (CPSF) and the mRNA. Together, they displace the PARN deadenylase and recruit the GLD2 poly(A) polymerase (Barnard et al., 1993; Cao and Richter, 2002; Groisman et al., 2002; Hake and Richter, 1994; Kim and Richter, 2006; Stebbins-Boaz et al., 1999). This results in poly(A) tail lengthening and translational activation. After fertilization, a subset of maternal transcripts undergoes CPEB-mediated polyadenylation in the embryo prior to zygotic transcription, likely regulating the first wave of translational activation (Collart et al., 2014). However, the mechanisms regulating polyadenylation in the X. tropicalis embryo have yet to be determined. The CPEB protein family also coordinates cytoplasmic polyadenylation and translation in zebrafish, where more than 40% of maternal mRNAs undergo cytoplasmic polyadenylation in early embryos (Winata et al., 2018). The translation of these transcripts increases during the MZT, whereas there is a decrease in translation of transcripts lacking elongated poly(A) tails (Winata et al., 2018). CPEB1 also functions during mouse oocyte maturation; ERK1/2 phosphorylation of CPEB1 causes polyadenylation-mediated translation of BTG4 and other essential maternal proteins, such as the cell cycle regulator, Cyclin B and the DAZL RBP (Sha et al., 2017). DAZL, in turn, directs translational activation during oocyte maturation (Chen et al., 2011; Fukuda et al., 2018). Autoregulation of its own mRNA by DAZL establishes a positive feedback loop together with CPEB1, which ensures the progression of meiosis and proper translational activation (Sousa Martins et al., 2016).

In early Drosophila embryos, the Wispy (GLD2) cytoplasmic poly(A) polymerase directs polyadenylation of maternal mRNAs and is essential for the MZT (Benoit et al., 2008; Cui et al., 2008; Cui et al., 2013; Salles et al., 1994). However, relative changes in poly(A) tail length and translational efficiency are retained even in the absence of Wispy. Thus, the current hypothesis is that selective shortening, rather than lengthening of poly(A) tails serves as the major determinant of maternal mRNA translation at this stage (Eichhorn et al., 2016). Furthermore, in Drosophila, PNG triggers polyadenylation and translation at the beginning of the MZT (Fig. 3A) (Tadros et al., 2007; Vardy and Orr-Weaver, 2007). More than 60% of the maternal mRNAs that undergo either increases or decreases in translation depend on the PNG kinase for these processes (Kronja et al., 2014b). The Smaug RBP serves as a major mediator of PNG-dependent repression, with most of this repression attributable to poly(A) tail shortening (Eichhorn et al., 2016). However, cytoplasmic polyadenylation is not sufficient to rescue translation in png mutants, presumably because PNG must also relieve repression by specific RBPs such as PUM (Tadros et al., 2007; Vardy and Orr-Weaver, 2007).

Cytoplasmic polyadenylation and deadenylation also play key roles in regulation of mRNA stability and translation in C. elegans (Fig. 3B) (Millonigg et al., 2014; Nousch et al., 2017; Nousch et al., 2013; Nousch et al., 2014). During oocyte maturation and the MZT, the translational status of hundreds of transcripts is coordinated by LIN-41, a TRIM-NHL RBP related to Drosophila BRAT. Acting antagonistically to LIN-41 are the RBPs, OMA-1 and OMA-2, which are related to the Tristetraprolin/TIS-11 ARE-BP (Tsukamoto et al., 2017). LIN-41 and the OMAs also interact with both the GLD-2 poly(A) polymerase and the CCR4-NOT-deadenylase (Tsukamoto et al., 2017). Combinatorial action of LIN-41, OMAs and GLD-2 in distinct RNP complexes may direct a repression-to-activation switch for specific target transcripts, which in turn encode RBPs that function in early embryogenesis (Tsukamoto et al., 2017).

In summary, during the MZT, cytoplasmic polyadenylation and deadenylation affect both the stability and translation of mRNAs. While the machinery for these processes is conserved, the particular RBPs that confer specificity on subsets of transcripts vary across species. Likewise, whereas protein kinases play a key role in triggering translational derepression of maternal mRNAs, the identity of these kinases varies across species.

Despite the documented importance of post-transcriptional regulation of maternal transcripts at the level of mRNA translation and stability, global transcriptome and proteome analyses prior to and during the MZT have revealed that post-translational regulation of protein stability and function also play a major role in Drosophila (Becker et al., 2018; Casas-Vila et al., 2017; Kronja et al., 2014a) and C. elegans (Stoeckius et al., 2014). For example, in C. elegans more than one-quarter of the transcriptome is downregulated at least twofold in early embryos, whereas only 5% of the proteome shows a similar decrease (Stoeckius et al., 2014). These results highlight the importance of an additional level of regulation at the post-translational level, through modifications that affect maternal protein stability and function, which is essential for the progression of oocyte maturation and the MZT (Liu et al., 2018a).

One highly conserved mechanism of post-translational regulation that combines phosphorylation and ubiquitin-dependent proteolysis occurs at the beginning of the MZT (reviewed by Pesin and Orr-Weaver, 2008). Most animals are arrested during meiosis by the presence of Cyclin B. Release from meiotic arrest is achieved by the anaphase-promoting complex (APC), a cyclin-dependent E3 ubiquitin ligase complex that is phosphorylated and activated upon fertilization to target Cyclin B for degradation.

About 30% of the oocyte proteome is estimated to change in phosphorylation state during Drosophila egg activation (Krauchunas et al., 2012). The PNG kinase complex, which plays a key role in post-transcriptional regulation of maternal mRNAs, undergoes such a change (Fig. 3A) (Hara et al., 2017). In Drosophila oocytes, the regulatory subunit, GNU, is phosphorylated by CYCB/CDK1, which blocks the binding of GNU to the kinase subunit. Degradation of CYCB by the APC during meiotic completion triggers dephosphorylation of GNU, which then binds to and activates the PNG kinase subunit. Subsequently, PNG phosphorylates GNU, leading to GNU degradation and the inactivation of the complex (Hara et al., 2017). This temporally coordinated activation and self-inactivation tightly restricts the activity of this kinase to a short window during the early MZT, ensuring precise post-translational control of its phosphorylation targets. The PNG kinase is essential for translation of the Smaug RBP in the early embryo (discussed above). However, Smaug is rapidly cleared a couple of hours later, at the end of the MZT (Benoit et al., 2009). Indeed, towards the end of the MZT, not only Smaug but also additional global post-transcriptional regulators, such as ME31B, Cup and TRAL, undergo among the largest decreases in abundance in the entire proteome (Sysoev et al., 2016; Wang et al., 2017). Although the mechanisms that direct clearance of these proteins are not yet known, they represent a striking example of how post-translational regulation has a major impact on post-transcriptional regulation during the MZT.

Another example of post-translational regulation comes from C. elegans (Fig. 3B). In early embryos, maternally supplied OMA-1 and OMA-2 are phosphorylated by the DYRK kinase MBK-2, which allows binding to the transcription factor TAF-4, thus sequestering it in the cytoplasm and preventing premature ZGA (Guven-Ozkan et al., 2008). By the four-cell stage, phosphorylated OMA-1 and OMA-2 are ubiquitylated by accumulating levels of the Cullin-RING E3 ubiquitin ligase (CRL), leading to degradation of the OMAs and release of TAF-4, which is then imported into the nucleus where it directs zygotic transcription (Guven-Ozkan et al., 2008). Post-translational regulation is important for inactivation and clearance of additional post-transcriptional regulators during the C. elegans MZT. The LIN-41, GLD-1 and CPB-3 RBPs are all cleared by the SCF^SEL-10 E3 ubiquitin ligase (Kisielnicka et al., 2018; Spike et al., 2018). However, the signaling pathways that trigger inactivation and clearance of these RBPs differ: the CDK1 kinase pathway clears LIN-41, while the MAP kinase pathway clears GLD-1 (Spike et al., 2018).

In the mouse pre-implantation embryo, massive degradation of maternal proteins by autophagy is triggered by fertilization and is required for progression of early embryogenesis (Tsukamoto et al., 2008). While proteomic studies have identified the ubiquitin-proteasome pathway to be important during pre-implantation development as a whole (Zhang et al., 2009a), the first specific role for this pathway during the MZT has only recently been characterized. The TAB1 kinase inhibits the action of the NFκB transcription factor by retaining it in the cytoplasm. In early embryos, the E3 ubiquitin ligase RnF114 directs degradation of TAB1, permitting translocation of NFκB into the nucleus and, thus, activation the NFκB pathway, which is essential for development (Yang et al., 2017). Indeed, either knockdown of RnF114 or persistence of TAB1 protein prevents development beyond the two-cell stage (Yang et al., 2017).

The second act of the MZT is the onset of zygotic transcription. Historically, two transcriptional waves have been distinguished: a minor wave that occurs during the cleavage divisions; and a major wave that, in many species, coincides with the lengthening of the cell cycle. Genome-wide gene expression analysis, however, has revealed that ZGA does not consist of two distinct waves, but instead reflects a period over which transcription is gradually activated (Fig. 1) (Aanes et al., 2011; Collart et al., 2014; Harvey et al., 2013; Heyn et al., 2014; Lott et al., 2011; Owens et al., 2016; Pauli et al., 2012; Sandler and Stathopoulos, 2016; Tan et al., 2013; White et al., 2017).

In terms of the embryonic mitotic cell cycles, human, mouse and sea urchin start transcription the earliest, with the first zygotic transcripts being detected at the one-cell stage (Abe et al., 2015; Aoki et al., 1997; Gildor and Ben-Tabou de-Leon, 2015; Hamatani et al., 2004; Materna et al., 2010; Yan et al., 2013) (Table 2). In C. elegans, Xenopus, zebrafish and Drosophila, the first zygotic transcripts are detected in cell cycles 2, 3, 6 and 8, respectively (Chan et al., 2018 preprint; Collart et al., 2014; De Renzis et al., 2007; Edgar et al., 1994; Hadzhiev et al., 2019; Heyn et al., 2014; Hilbert et al., 2018 preprint; Kimelman et al., 1987; Kwasnieski et al., 2019 preprint; Lécuyer et al., 2007; Lott et al., 2011; Mathavan et al., 2005; Owens et al., 2016; Paranjpe et al., 2013; Seydoux and Fire, 1994; Skirkanich et al., 2011; Tan et al., 2013; Yanai et al., 2011; Yang et al., 2002), although there is some genetic evidence that transcription may begin earlier in Drosophila (Ali-Murthy et al., 2013). In terms of absolute time, the very rapid cleavage cycles of the early Drosophila, Xenopus and zebrafish embryos mean that, in these species, ZGA begins several hours earlier than in the mouse (Fig. 1).

During genome activation, a significant fraction of all genes in the genome is transcribed: 5% in X. tropicalis (Collart et al., 2014; Tan et al., 2013), ∼10% in C. elegans (Baugh et al., 2003), ∼20% in mouse (Hamatani et al., 2004), ∼25% in zebrafish (Aanes et al., 2011; Harvey et al., 2013; Lee et al., 2013) and ∼35% in Drosophila (De Renzis et al., 2007; Kwasnieski et al., 2019 preprint; Lécuyer et al., 2007; Lott et al., 2011) (Table 2). Although different sets of genes are activated in different species (Heyn et al., 2014), the proteins these genes encode are often enriched for transcription factors and other developmental regulators, such as microRNAs (Collart et al., 2014; De Renzis et al., 2007; Lee et al., 2013). Some of these are important for the degradation of maternally loaded RNAs (Bushati et al., 2008; Giraldez et al., 2006; Lund et al., 2009), and others for the transcription of subsequently activated genes (Collart et al., 2014). For most genes, the onset of transcription during ZGA is stochastic, which means that not all cells start to activate that gene at the same time (Boettiger and Levine, 2009; Stapel et al., 2017). This initially results in large cell-to-cell differences in gene expression levels, which could – in principle – hamper development (Lagha et al., 2013). Uniform patterns of gene expression, however, are reached by spatial averaging, such as in the Drosophila syncytium (Little et al., 2013), or temporal averaging, as in zebrafish (Stapel et al., 2017), thus overcoming this problem.

Although some of the genes that are activated during ZGA are strictly zygotic (i.e. not maternally provided), many others are loaded maternally and then re-expressed in the embryo. In Drosophila, for example, only about one-third of the zygotic transcripts are purely zygotic (De Renzis et al., 2007) and in zebrafish 25% (Lee et al., 2013) (Table 2). In its simplest form, zygotic transcription of genes that are also maternally loaded provides an opportunity to reinforce the expression of specific genes. In Drosophila and zebrafish, however, many genes encode distinct maternal and zygotic mRNA isoforms through differential promoter usage, splicing and/or polyadenylation site usage (Aanes et al., 2013; Atallah and Lott, 2018; Haberle et al., 2014). In these cases, transcription does not just reinforce gene expression, it also changes the characteristics of transcripts and therefore their regulation. Functionally, maternal mRNA degradation and zygotic transcription are used to generate localized patterns of expression (Combs and Eisen, 2013; De Renzis et al., 2007; Lécuyer et al., 2007; Vopalensky et al., 2018).

Next, we explore the effect of cell cycle length, transcriptional repressors, transcriptional activators, and chromatin accessibility on the change from transcriptional repression to transcriptional activation (Table 2). We end by proposing a mechanism for the onset of transcription that takes all of these aspects into account (Fig. 4).

The early stages of development in insects, amphibians and fish are characterized by rapid cleavage divisions. Because DNA replication generally interferes with transcription (Rothe et al., 1992; Shermoen and O'Farrell, 1991), rapid cell cycles may reduce transcriptional output during the early stages of development. Indeed, lengthening the cell cycle causes a premature onset of transcription in X. laevis (Collart et al., 2013; Kimelman et al., 1987), suggesting that genome activation may be a consequence of cell cycle lengthening. However, similar experiments in zebrafish and Drosophila have not caused premature transcription (Farrell and O'Farrell, 2013; McCleland and O'Farrell, 2008; Zhang et al., 2014a). In fact, in Drosophila embryos, cell cycle lengthening depends on the onset of zygotic transcription (Blythe and Wieschaus, 2015; Pritchard and Schubiger, 1996) arguing against a dependence of transcription on cell cycle lengthening in these species. Thus, it remains unclear to what extent the lengthening of the cell cycle influences the onset of transcription during embryogenesis. What is clear, is that cell cycle length affects the length and number of transcripts that are produced in Drosophila and zebrafish embryos (Edgar and Schubiger, 1986; Hadzhiev et al., 2019; Rothe et al., 1992; Dalle Nogare et al., 2009). Indeed, in these species, the first genes that are expressed during ZGA are often short and lack introns (Heyn et al., 2014; Kwasnieski et al., 2019; Rothe et al., 1992). Thus, cell cycle length has an effect on the length and number of transcripts that are produced, but it remains enigmatic whether it has a direct effect on the onset of transcription.

It should be noted that a recent study in Drosophila has shown that the presence of ‘aborted’ truncated transcripts with intronic sequences from long genes is extensive during ZGA (Kwasnieski et al., 2019 preprint). Moreover, truncation of long transcripts can be precisely regulated and, if translated, their gene products could have developmental functions (Sandler et al., 2018). For example, the Sex lethal (SXL) RNA-binding protein directs formation of a shortened mRNA isoform of short gastrulation (sog), which encodes a dominant-negative SOG protein that may prevent TGF-β signaling during this stage of the MZT (Sandler et al., 2018). The generality of regulated and functional truncation of transcripts encoded by long genes remains to be determined.

Many proteins are maternally loaded to support embryonic development during the early cleavage stages. Among these are proteins that repress transcription (Table 2). In X. laevis and Drosophila, for example, sequence-specific repressors have been identified that inhibit the expression of subsets of genes (Brown and Wu, 1993; Dunican et al., 2008; Pritchard and Schubiger, 1996; Ruzov, 2004; Ruzov et al., 2009). In Drosophila, the transcription factor Tramtrack (TTK) is maternally loaded and represses the transcription of the segmentation gene fushi tarazu (ftz) (Brown and Wu, 1993; Pritchard and Schubiger, 1996). Reducing the amount of TTK or the number of TTK-binding sites results in premature ftz transcription, while increasing the amount of TTK has the opposite effect. Interestingly, the Smaug RBP is required for destruction of maternal ttk mRNA, thus providing a link between clearance of maternal mRNAs and ZGA (Benoit et al., 2009).

In addition to specific repressors, histones have been identified as more general repressors of transcription in X. laevis and zebrafish (Almouzni and Wolffe, 1995; Amodeo et al., 2015; Joseph et al., 2017). Histones are present in large excess in early embryos (Adamson and Woodland, 1974; Anderson and Lengyel, 1980); they bind with high affinity and little sequence specificity to DNA (Campos and Reinberg, 2009), and, when bound in the form of nucleosomes, they block access of the transcriptional machinery to DNA (Lorch et al., 1987; Workman and Kingston, 1998). In zebrafish and Drosophila embryos, the concentration of soluble histones in the nucleus drops in the approach to genome activation (Joseph et al., 2017; Shindo and Amodeo, 2019). Moreover, experiments in zebrafish have revealed that the drop in the concentration of soluble histones in the nucleus provides an opportunity for the transcriptional machinery to successfully compete for DNA access and activate transcription (Joseph et al., 2017; Pálfy et al., 2017). Developmental changes in the early embryo provide at least two mechanisms by which the concentration of histones may be reduced during embryogenesis. First, the exponential increase in DNA during the rapid cleavage divisions might titrate out histones. A role for DNA content in regulating the onset of transcription is supported by experiments in which increasing DNA content results in premature onset of transcription in X. laevis, zebrafish and Drosophila (Dekens et al., 2003; Jevtić and Levy, 2017; Lu et al., 2009; Newport and Kirschner, 1982; Prioleau et al., 1994). However, the quantification of both histones and DNA content has shown that, at least in zebrafish, the increase in DNA content is not sufficient to titrate the levels of soluble histones significantly (Joseph et al., 2017). Another possible explanation for the decrease in nuclear histone concentration involves a change in the nuclear import dynamics of histones. This could be a consequence of the dilution of import machinery caused by the increasing number of nuclei (Shindo and Amodeo, 2019) or of the marked increase in the ratio of nuclear over cytoplasmic volume during the cleavage stages (Joseph et al., 2017). The latter has been suggested to limit the capacity of the nucleus to concentrate proteins (Kim and Elbaum, 2013a,b; Kopito and Elbaum, 2007; Kopito and Elbaum, 2009), changing the distribution of histones between nucleus and cytoplasm. A role for import dynamics is supported by the observation that an experimentally induced increase in nuclear volume results in earlier genome activation in X. laevis (Jevtić and Levy, 2015; Jevtić and Levy, 2017). We conclude that both specific and general repressors play a role in changing the balance from transcriptional repression to transcriptional activation during ZGA.

The general absence of transcription during early embryogenesis, as well as the gene-specific onset of transcription that follows, are affected by the availability of the transcriptional machinery. As described above, in C. elegans, the general TAF-4 transcription factor is unavailable for transcription during the early stages of development (Guven-Ozkan et al., 2008), because the protein is sequestered by the binding of OMA-1 and OMA-2. Not until their phosphorylation at the four-cell stage, is TAF-4 released and able to translocate into the nucleus, permitting the onset of transcription (Guven-Ozkan et al., 2008). Similarly, in X. laevis, the concentration of the general TBP transcription factor is limiting before ZGA (Veenstra et al., 1999). Protein expression profiling has revealed that TBP levels increase and reach sufficiently high levels for genome activation as a result of translation of maternally stored TBP transcripts (Veenstra et al., 1999). Thus, the lack of general transcription factors in the early stages of development contributes to the general absence of transcriptional activity.

Following this period, the gene-specific onset of transcription requires gene-specific transcription factors. The transcription factors that activate zygotically expressed genes have been identified in several species, including Drosophila (Zelda), zebrafish (Pou5f3, Sox19b, Nanog), human (OCT4, DUX4) and mouse (Dppa2, Dppa4, Nfy, Dux) (De Iaco et al., 2019; De Iaco et al., 2017; Eckersley-Maslin et al., 2019; Gao et al., 2018; Harrison et al., 2011; Hendrickson et al., 2017; Lee et al., 2013; Leichsenring et al., 2013; Liang et al., 2008; Lu et al., 2016; Nien et al., 2011). A recent review provides details of their mechanisms of action (Schulz and Harrison, 2018). In Drosophila and zebrafish, RNA encoding these factors is maternally loaded and their protein levels increase during the early cell cycles due to translation (Harrison et al., 2010; Harrison et al., 2011; Lee et al., 2013; Nien et al., 2011). In zebrafish, it has been shown that the levels of such gene-specific transcription factors affect the time at which transcription starts (Joseph et al., 2017). In this context, it is interesting to note that, in Drosophila and zebrafish, zygotic transcription begins in two distinct areas in the nucleus (Blythe and Wieschaus, 2016; Chan et al., 2018 preprint; Chen et al., 2013; Hadzhiev et al., 2019; Hilbert et al., 2018 preprint; Hug et al., 2017), which may cause a local increase in the concentration of transcription factors, thereby facilitating transcription.

Chromatin regulates the accessibility of the genome for DNA-binding proteins. Thus, chromatin accessibility is key to transcriptional regulation. In its simplest form, chromatin consists of DNA wrapped around octamers of the core histone proteins joined by a histone linker protein. DNA accessibility is regulated through DNA methylation, nucleosome positioning and the stability of nucleosomes, which in turn is regulated by histone modifications and the presence of histone variants. Here, we discuss changes in global and local chromatin accessibility that accompany genome activation, and their role in activating transcription.

Overall, chromatin structure is relatively open during the transcriptionally inactive period of the MZT. Human and mouse genomes, for example, undergo global DNA demethylation upon fertilization (Guo et al., 2014; Li et al., 2018; Peat et al., 2014; Santos et al., 2002; Shen et al., 2014), but this has not been observed in other species. More generally, the open chromatin structure of early embryos is characterized by highly dispersed chromatin (Ahmed et al., 2010; Popken et al., 2014), the absence of heterochromatin domains (Ancelin et al., 2016; Laue et al., 2019; Mutlu et al., 2018), the absence of topologically associated domains (‘TADs’) (Du et al., 2017; Hug et al., 2017; Kaaij et al., 2018; Ke et al., 2017; Ogiyama et al., 2018), high levels of histone acetylation (Adenot et al., 1997; Li et al., 2014) and high chromatin mobility (Bošković et al., 2014). During ZGA, the genome becomes more compact, which may be related to the replacement of embryonic linker histone variants with somatic variants around this time (Pérez-Montero et al., 2013; Saeki et al., 2005).

In contrast, local genome accessibility increases during genome activation in Drosophila, zebrafish, mouse and human (Blythe and Wieschaus, 2016; Gao et al., 2018; Li et al., 2018; Liu et al., 2018b; Lu et al., 2016; Wu et al., 2016; Wu et al., 2018). This is accompanied by the appearance of specific histone modifications. In X. laevis embryos, for example, Histone 3 Arginine 8 (H3R8) methylation poises genes for expression (Blythe et al., 2010). Similarly, Histone 3 Lysine 4 trimethylation (H3K4me3), a mark that is generally associated with active transcription, appears on promoters during genome activation in Drosophila (Chen et al., 2013; Li et al., 2014), X. tropicalis (Hontelez et al., 2015; Lindeman et al., 2011), zebrafish (Vastenhouw et al., 2010; Zhang et al., 2014b) and mouse (Dahl et al., 2016; Zhang et al., 2016) embryos, often prior to transcription. Although in most cases the relevance of H3K4me3 has not yet been investigated directly, studies in mouse have revealed that it is required for the onset of transcription (Aoshima et al., 2015; Dahl et al., 2016; Liu et al., 2016; Zhang et al., 2016), which supports the observation that Brg1, which is required for H3K4 methylation, is necessary for ZGA (Bultman et al., 2006). Finally, acetylation of Histone 3 Lysine 27 (H3K27Ac) precedes ZGA in zebrafish (Chan et al., 2018 preprint; Sato et al., 2019 preprint; Zhang et al., 2018), and is required for the transcription of – at least – miR-430, which is one of the first genes to be transcribed. Although there is clearly a role for local histone modifications in ZGA, no local DNA methylation changes have been observed that coincide with the onset of transcription (Jiang et al., 2013; Kaaij et al., 2016; Potok et al., 2013), arguing against a role for DNA methylation in regulating ZGA. DNA methylation patterns, however, do play a role in the regulation of transcription during embryogenesis. In zebrafish, for example, hypermethylation at enhancers predicts transcription factor binding and enhancer activity (Kaaij et al., 2016; Liu et al., 2018b), whereas low levels of DNA methylation at promoters predict H3K4me3 and promoter activity (Andersen et al., 2012; Liu et al., 2018b). Finally, ZGA coincides with a significant increase in both the repressive histone modification H3K27me3 (Akkers et al., 2009; Hontelez et al., 2015; Li et al., 2014; Lindeman et al., 2011; Liu et al., 2016; van Heeringen et al., 2014; Vastenhouw et al., 2010) and DNA methylation (Potok et al., 2013), which may help to ensure the onset of gene-specific transcription (Potok et al., 2013; Zenk et al., 2017).

Taken together, chromatin structure is relatively open during the early stages of embryogenesis, which likely supports the large-scale transcriptional reprogramming that takes place during ZGA (Fig. 4). Indeed, the removal of mouse LSD1, which results in premature heterochromatin formation, interferes with developmental progression (Ancelin et al., 2016). Thus, the absence of transcriptional activity is likely not due to a repressive chromatin structure. Approaching ZGA, the genome becomes more compacted overall, while local accessibility increases. At the same time, at least in zebrafish and X. laevis, the concentration of non-DNA-bound histones drops (Amodeo et al., 2015; Joseph et al., 2017), which allows increasing levels of transcription factors to successfully compete for DNA binding (Joseph et al., 2017; Pálfy et al., 2017). Similarly, the loss of specific repressors allows for transcription of specific genes. Although the increase in local accessibility facilitates the binding of transcription factors, accessibility, in turn, often depends on the binding of specific transcription factors (Gao et al., 2018; Liu et al., 2018b; Lu et al., 2016; Schulz et al., 2015; Sun et al., 2015; Veil et al., 2019). Thus, the loss of repressors, the accumulation of activators and local changes in chromatin accessibility together prime the genome for activation.

We now have a reasonably thorough descriptive and mechanistic understanding of the two ‘acts’ of the MZT. Maternally loaded gene products are required to control early development, while the regulated onset of transcription is required for developmental progression: if transcription is inhibited, embryos arrest prior to gastrulation (Kane et al., 1996; Lee et al., 2013; Zamir et al., 1997). It remains mysterious, however, why such a large fraction of the gene products encoded by the maternal genome is loaded into the egg and then largely eliminated. One clue as to why maternal mRNAs and proteins are cleared may come from the fact that maternal and zygotic transcripts from the same gene often represent different mRNA isoforms (Aanes et al., 2013; Atallah and Lott, 2018; Haberle et al., 2014). Where the isoforms differ in untranslated regions, these would impose differential regulation of transcript stability, translation and localization. Where they differ in the open reading frame, there may also be differences in maternal versus zygotic protein structure and function. Another clue derives from the observation that zygotically re-expressed genes in Drosophila are enriched for ones whose zygotic expression is patterned (De Renzis et al., 2007). This implies that degradation of ubiquitous maternal transcripts followed by patterned zygotic re-expression is a mechanism that imposes spatial regulation of processes that underlie differential cell behavior and fate in the developing embryo.

Another possible function of degradation of maternal mRNAs and proteins is to supply nutrients to the developing embryo. Evidence for this comes from studies in the Drosophila embryo, where maternally supplied dNTPs are insufficient for completion of the DNA replication and nuclear divisions that occur prior to large-scale ZGA, thus requiring de novo dNTP synthesis from NTPs (Song et al., 2017). Calculations using previous measurements in oocytes of Xenopus and S. purpuratus have revealed similar dNTP shortages for the cleavage divisions (Song et al., 2017). Thus, degradation of maternally loaded mRNAs may provide a source of nucleotides to support DNA replication. Consistent with a requirement for NTPs, rather than a specific set of mRNAs, the maternally loaded transcriptome shows more variability across species than that of most other developmental stages in Drosophila and zebrafish (Atallah and Lott, 2018; Domazet-Lošo and Tautz, 2010; Kalinka et al., 2010). Furthermore, in Drosophila, the maternal transcripts that are degraded during the MZT are much less conserved than those that remain after the onset of ZGA (Atallah and Lott, 2018).

Mechanistically, the regulation of maternal gene products and the activation of zygotic transcription are highly interrelated. For example, in zebrafish, transcription factors that increase in abundance during the early stages of the MZT, because of translation of maternal mRNAs, are required for ZGA (Lee et al., 2013). Similarly, in Drosophila, the Zelda transcription factor, which is essential for early ZGA, increases in abundance early in the MZT through translation of its cognate maternal mRNA (Nien et al., 2011). Additional proteins that increase in abundance show enrichment for functions in chromatin organization (Kronja et al., 2014b) and therefore might play a role in ZGA. Clearance and/or repression of maternal gene products appears also to be a prerequisite for ZGA. In C. elegans, for example, degradation of OMA-1 and OMA-2 are required to permit nuclear translocation of TAF-4 and, thus, ZGA (Guven-Ozkan et al., 2008). Likewise, mutants that fail to clear and/or repress a significant fraction of the maternal transcriptome either fail to undergo, or suffer a severe delay in, ZGA (e.g. Drosophila smaug, mouse Btg4) (Benoit et al., 2009; Luo et al., 2016; Yu et al., 2016). Finally, genome activation is itself required for the clearance of a large subset of maternal RNAs. For example, in zebrafish, X. laevis and Drosophila, ZGA produces sets of miRNAs that feed back to destabilize subsets of maternal mRNAs (Bushati et al., 2008; Giraldez et al., 2006; Lund et al., 2009). If ZGA does not occur (e.g. in Drosophila smaug mutants or in zebrafish if transcription is chemically inhibited) then the miRNAs are not synthesized and their target transcripts persist (Benoit et al., 2009; De Renzis et al., 2007; Lee et al., 2013).

We thank members of the Vastenhouw lab, as well as Jan Brugues, for critically reading the manuscript; Edlyn Wu for help with Fig. 4; and the reviewers for helpful feedback.