After the description of the first segmentation clock gene, hairy1, in the chicken embryo (Palmeirim et al., 1997), similar oscillatory patterns of expression were identified for other genes, and in multiple organisms (Table 2). The use of genome-wide approaches identified a wide range of genes with oscillatory gene expression during somitogenesis and evidenced that the embryo clock is an intricate oscillatory genetic network, that comprises genes belonging to multiple signalling pathways, notably, Notch, Wnt and FGF (Dequéant et al., 2006; Krol et al., 2011) (Figure 1A). These include Wnt-dependent Axin2, FGF signalling pathway genes Dusp1/2/4/6, Snail1/2, Spry2/4, and Notch pathway genes of the Already defined earlier. HES family, Lfng and Nrarp, among others. Strikingly, only two genes–the Hes1 and Hes5 orthologs–were conserved in mouse, chicken and zebrafish. Otherwise, the identity of the pathway-specific oscillating genes varied considerably, evidencing evolutionary plasticity of the conserved oscillations in signalling pathway activity (Krol et al., 2011). Several studies have shown that these intercellular communication pathways cooperate during embryo body segmentation. Niwa et al. (2007), (2011) showed that the onset of Hes7 expression in the mouse tailbud is FGF-dependent, while its maintenance and propagation throughout the PSM requires Notch signalling. A gradient of nuclear Wnt-related β-catenin was shown to control key features of PSM maturation and somite formation (Aulehla et al., 2008). Notch- and Wnt-dependent gene expression oscillations are coupled in the PSM and undergo a phase shift towards the anterior PSM. Inhibition of this phase shift in an in vitro setting delayed the arrest of EC waves and impaired tissue segmentation (Sonnen et al., 2018). Identification of EC oscillatory dynamics at the protein level has lagged behind, mostly due to the lack of appropriate antibodies and because it is technically more challenging than the well-established in situ hybridization protocols for RNA detection. However, corresponding cycles of protein expression with the same periodicity have been reported and are summarized in Table 3.

Several studies have identified negative feedback regulation as a fundamental feature of EC oscillations (Hirata et al., 2002; Bessho et al., 2003; Lewis, 2003; Chen et al., 2005). Figure 1B presents a simplified view of the negative feedback regulatory mechanisms underlying HES gene expression oscillations, whereby hes oscillations are maintained by an inhibitory action of the HES protein on its own promoter. HES also inhibits delta expression and/or expression of the Notch-modulator Lunatic fringe–Lfng, which contributes to synchronized oscillations of Notch-dependent gene expression in neighbour cells. Rapid degradation of the molecular products produced, mRNAs and proteins, ensures propagation of the oscillatory behaviour (reviewed in Kageyama et al., 2012).

Using mathematical modelling, Lewis (2003), Monk (2003) and Jensen et al. (2003) independently postulated that oscillations in gene expression are influenced by delays in the various steps of the regulatory negative feedback loop. Indeed, further experimental evidence showed that the EC mechanism depends on Delayed Negative Feedback loops and that the temporal delays are introduced in multiple steps of the process. Namely:

Embryo Clock gene expression oscillations are a cell autonomous property. This was first hinted by dissecting the chicken PSM in multiple portions and observing that the overall expression pattern of hes4 (Palmeirim et al., 1997) and Lfng (Maroto et al., 2005) remained intact. The same was observed in dissected mouse PSMs (Masamizu et al., 2006). Cyclic Lfng gene expression even persisted in dissociated chicken PSM cells, but it occurred asynchronously among cells, evidencing the need for cell-cell contact to ensure synchrony and establish robust cyclic expression patterns at the tissue level (Maroto et al., 2005). In dissociated mouse PSM cells, Hes1 oscillations also occur cell-autonomously (Masamizu et al., 2006) and Webb et al. (2016) further reported that zebrafish her1 gene retains oscillatory expression in isolated tailbud cells. In this case, oscillations in individual cells presented a longer period and were less robust, compared with the intact tissue. Altogether, these results suggest that cell-cell communication is a key requirement for oscillations to be in phase within the vertebrate PSM tissue. This was corroborated by the work of Tsiairis and Aulehla (2016), that showed that a re-aggregation of mouse PSM cells is able to synchronize oscillations. Cell-autonomous EC gene expression oscillations have also been described in other cell types, such as mouse embryonic stem cells (ESC) (Kobayashi et al., 2009), individual fibroblasts (Masamizu et al., 2006) and neural progenitor cells (Shimojo et al., 2008; Bonev et al., 2012; Manning et al., 2019). These can be synchronized in vitro by the application of a serum shock or by Notch activation (Hirata et al., 2002). However, cell-specific distinct phases of EC oscillations may also play important roles in vivo. This will be discussed further below.

EC synchronization between PSM cells is required for proper somite formation. Local synchrony within the PSM tissue is achieved through Delta-Notch signalling (Jiang et al., 2000; Horikawa et al., 2006; Riedel-Kruse et al., 2007), which also functions to overcome the effect of “noise” introduced by other biological processes, such as cell division (Horikawa et al., 2006; Riedel-Kruse et al., 2007). Riedel-Kruse and colleagues found that EC synchrony in the zebrafish embryo is achieved by simultaneous (Notch-independent) activation immediately prior to gastrulation and is then maintained by Notch-dependent self-organized synchronization. The latter was elegantly shown by incubating embryos with the Notch-inhibitor DAPT until complete EC desynchronization. Then, DAPT washout alone was sufficient to completely restore both dlc oscillations and somite formation (Riedel-Kruse et al., 2007). Delaune et al. (2012) applied single cell live-imaging of her1 expression in zebrafish wild-type and mutant embryos for deltaC, deltaD and notch1a to study the role of Delta-Notch signalling in EC synchronization. In the mutants, her1 dynamics persisted in PSM neighbour cells, but in different oscillation phases (Delaune et al., 2012). Interestingly, deltaC and deltaD work together to ensure synchrony of the zebrafish segmentation clock in distinct portions of the PSM. While deltaD is responsible for the onset of the oscillations at the tailbud level, deltaC plays in role in maintaining and amplifying the oscillations in adjacent cells along the PSM tissue (Mara et al., 2007).

Soza-Ried et al. (2014) provided conclusive evidence for the role of oscillations of Notch-Delta signalling in maintaining the EC synchronized in neighbour cells for somite segmentation. Using a deltaC zebrafish mutant line, the authors were able to rescue both her1 oscillations and somite formation by applying short artificial pulses of deltaC expression, evidencing that Notch signalling is indeed maintaining cell synchrony during somitogenesis. Accordingly, longer intervals between deltaC pulses generated larger somites (Soza-Ried et al., 2014). This was confirmed by Isomura et al. (2017) who developed an optogenetics-based system to monitor Notch-Delta signalling dynamics in neighbour cells. Light-induced Dll1 pulses in sender cells were able to generate synchronized oscillations of Hes1 expression in receiver cells. Furthermore, they were able to determine the time from the induction of Dll1 to the cleavage of NICD, which was ∼50.9 min, followed by an additional ∼77 min until maximum Hes1 levels were reached (Isomura et al., 2017). More recently, Yoshioka-Kobayashi et al. (2020) used this system to show that LFNG in sender cells introduces a 15 min-delay in the transport of Dll1 protein to the cell membrane, without which HES7 oscillations are severely dampened in individual cells of the PSM. These studies corroborate the importance of delays in cell-cell communication for EC oscillations and illustrate the power of optogenetics-based tools for dissecting these intricate regulatory mechanisms.

Notch-Delta signalling was shown to require non-muscle myosin II (NM II)-dependent contractility in both signal-sending and -receiving cells (Hunter et al., 2019). Recently, our lab evaluated the importance of fibronectin (FN) extracellular matrix assembly and signalling through the integrin-ROCK-NM II axis for somite segmentation and EC oscillations in chick PSM. We found that experimental treatments targeting FN matrix assembly, cell-FN interactions and actomyosin contractility significantly perturbed somite formation and EC gene expression, highlighting the importance of the PSM tissue’s mechanical properties for EC oscillations (Gomes de Almeida et al., 2022).

Hes7 and her1 oscillations slowdown in the anterior PSM in mouse (about 1.5-fold) (Niwa et al., 2011) and in zebrafish (Giudicelli et al., 2007; Delaune et al., 2012). Shih et al. (2015) corroborated these findings using live imaging in a transgenic zebrafish line with a her1-venus reporter. They saw that the periodicity of the segmentation clock increases by 1.5-fold in the anterior PSM, comparatively to the posterior PSM. SSoroldoni et al. (2014) had previously described this as a Dynamic Wavelength effect that, together with a Doppler effect resulting from the relative motion of the anterior PSM towards the posterior end due to tissue shortening over time, explains the rhythm of embryo body segmentation. To better understand the dynamics of EC deacceleration in the anterior PSM, Shih et al. (2015) assessed her1-venus reporter expression in cells that would form either side of a somite boundary. Within the same presumptive somite, clock oscillations were arrested in a posterior-to-anterior direction, i.e., cells that were incorporated in a posterior somite boundary ceased oscillations prior to cells that were incorporated in the anterior boundary of the same somite. Moreover, the authors reported that cells at a one-somite distance are initially synchronized in the posterior PSM and, as the clock slows down in the anterior PSM, they assume opposite phases of EC expression (Shih et al., 2015).

Another important alteration that cells experience as they transition from the posterior to anterior PSM is the relative timing of Notch- and Wnt-dependent EC oscillators (Sonnen et al., 2018). In fact, Lfng (Notch) and Axin2 (Wnt) oscillated out-of-phase in the posterior PSM and were progressively coupled towards the anterior PSM, where their synchronization was critical for somite segmentation. This was shown using an ingenious microfluidics-based system, which allowed for precise manipulation of gene expression oscillations by applying temporally controlled pulses of Wnt/Notch-specific activator molecules (Sonnen et al., 2018). Recently, it was shown that the distinct levels of FGF signalling experienced in the anterior and posterior regions of the PSM could underlie the differences in EC dynamics observed in these cells (Diaz Cuadros et al., 2020; Yaman et al., 2022). Yaman et al. (2022) further reported that the posterior-to-anterior FGF gradient in the PSM, classically solely associated with the wavefront activity, is also controlling the anterior propagation of the EC oscillations (Yaman et al., 2022).

There is also evidence of significant spatiotemporal metabolic changes along the PSM axis. Microarray analysis performed in PSM and tailbud samples of zebrafish embryos revealed that cell cycle/DNA metabolic functions are enriched in the posterior PSM, while translation/oxidative metabolism is enriched in anterior PSM and somites. The authors also reported a 2-fold increase in ATP content, as well as 2.5-fold decrease of Cytochrome C oxidase activity in the posterior PSM compared to anterior tissues (Özbudak et al., 2010). A posterior-to-anterior gradient of glycolytic activity was also linked to presomitic mesoderm development in mouse (Bulusu et al., 2017) and chicken (Oginuma et al., 2017) embryos. To test the functional relevance of these metabolic differences, PSM explants were cultured in glucose- or pyruvate-supplemented medium. While explants cultured in glucose supplemented medium developed normally, pyruvate-cultured explants displayed several defects concomitantly with loss of Lfng gene expression in the posterior PSM. Using a genetically encoded sensor for pyruvate to monitor metabolic transitions during PSM differentiation in real-time, the authors reported that pyruvate levels, i.e., glycolytic activity, decreased as cells transited towards an anterior PSM-like, more differentiated, state. Consistent with these findings, chicken embryos treated with 2-deoxy-D-glucose (2DG), a competitive inhibitor of the glycolytic enzyme hexokinase, displayed severe elongation defects, even though somite formation occurred normally. On the other hand, embryos treated with sodium azide (NaN₃)—a respiration inhibitor, had impaired somite segmentation (Oginuma et al., 2017).

The time each pair of somites takes to form is species-specific and displays great variability between organisms, ranging from 30 min in zebrafish to approximately 5 h in Human (Table 1). Similarly, the expression of segmentation clock genes oscillates with a periodicity characteristic of each species, which closely matches the time of somite formation (Figure 2; Table 2). The signalling pathways that comprise these genes are conserved; however, data suggests that cyclic genes display an evolutionary plasticity, since the specific genes involved in each pathway differ in the studied organisms (Krol et al., 2011).

A curious aspect of somite formation time is that the size of the organism does not significantly influence the time a pair of somites takes to form. For instance, somites have a relatively similar time of formation in the chicken and the emu, although dimension-wise these two birds are very distinct (Nagai et al., 2011). Likewise, the time of somite formation does not depend on phylogenetic relationships, since vertebrates belonging to different phyla can have the same somitogenesis period: 60 min for medaka and the house snake (Gomez et al., 2008).

Recent work started shedding light into the molecular mechanisms underlying divergent EC periodicity among organisms. Using in vitro models to compare EC gene expression dynamics in mouse vs. human cells, two independent studies found that the near 2-fold difference in oscillation periodicity could be explained by the different speeds in biochemical reactions within human and mouse cells, in particular mRNA and/or protein decay rates (Matsuda et al., 2020a; Rayon et al., 2020). This was documented in different cell types, which suggests that global temporal scaling mechanisms are a cell-autonomous property of the organism (Rayon and Briscoe, 2021).

EC oscillations underlying the formation of somites positioned at different A-P levels of the vertebrate body axis have different periodicities. In the chicken embryo, somitogenesis and EC oscillations occur with a 90 min-periodicity for somites 15–20 (HH12-13⁺) (Palmeirim et al., 1997). However, the final 5-8 somites (HH23) form with a periodicity of 150 min, matched by correspondingly slower cycles of Lfng gene expression (Gibb et al., 2009; Tenin et al., 2010) (Figure 3). In the opposite end, knowledge on the EC in the formation of the anterior-most somites is scarce. Rodrigues et al. (2006) characterized the expression of Notch-related EC genes in somites 1–10 and reported that, while they were dynamically expressed in the PSM, they did not present somite A-P polarity, as occurs in caudal somites. The EC was proposed to already be active even earlier in development, during gastrulation. In fact, Jouve et al. (2002) reported the existence of pulses of gene expression of hairy2 and Lfng in the prospective PSM of early chicken embryos. In zebrafish gastrulation stages, the EC already oscillates with 30-min periodicity (Riedel-Kruse et al., 2007) and in mouse, with a ∼2-h period (Falk et al., 2022). What triggers the onset of the Embryo Clock, and the existence of a clear periodicity in the early developmental stages of chick development, however, remains elusive.

Differences regarding segmentation clock operation in anterior and posterior regions of the zebrafish body axis have also been described (Choorapoikayil et al., 2012; Schröter et al., 2012; Hanisch et al., 2013). Although the double-mutant and -morphant for her1 and her7 exhibit defects throughout the entire body axis (Henry et al., 2002; Oates and Ho, 2002; Lleras-Forero et al., 2018), disrupting her1 or her7 gene expression has a different impact depending on the somites that are being formed. In fact, segmentation defects in her1 mutants are restricted to the anterior trunk, while in her7 mutants somites become defective only posterior to somite 8 (Choorapoikayil et al., 2012; Schröter et al., 2012; Hanisch et al., 2013).

Throughout development, the same gene regulatory networks can be employed in different tissues to produce multiple outcomes. Besides the segmentation of the axial vertebrate body plan, EC oscillations also participate in other developmental processes, where their dynamics differs from the one presented in the PSM during somitogenesis.

Embryonic stem cells (ESC) differentiate into cell types belonging to all three germ layers–mesoderm, endoderm and ectoderm. Evidence that oscillatory gene expression played an important role in ESC differentiation was first provided by Kobayashi et al. (2009). The authors identified unsynchronized cycles of Hes1 gene expression in mouse ESCs with a periodicity of 3–5 h–longer than the 2-h Hes1 period described for other cell types from this organism (see Table 2). Under the same culture conditions, Hes1-high cells differentiated more efficiently into mesodermal cells, while Hes1-low cells into neurons, suggesting that unsynchronized gene expression oscillations might potentiate heterogeneous cell fate specification within the same population of cells (Kobayashi et al., 2009).

Oscillatory EC expression was also described in mouse neural progenitor cells (NPC). Using real-time imaging, Hes5 and Dll1 mRNA were determined to oscillate with a 2 h-periodicity (Imayoshi et al., 2013; Shimojo et al., 2008, 2016), and for Hes1 this period was 2–3 h (Shimojo et al., 2008; Imayoshi et al., 2013) (Table 2). Likewise, Hes1 and Delta1 proteins display an identical period of oscillation (Table 3; Imayoshi et al., 2013). Studies performed in mouse NPCs reiterated the importance of gene expression oscillations for cell fate determination. During neurogenesis, the NPC population is maintained due to the repression of neural fate determination factors, such as Neurogenin2, Ascl1/Mash1 and Olig2 by Hes1 and Hes5 oscillatory expression levels. As a consequence, Neurogenin2 and Ascl1 also display oscillatory mRNA expression (Shimojo et al., 2008; Imayoshi et al., 2013). Unlike somitogenesis, oscillations during neural development are asynchronous. While undergoing differentiation, neural cells impede their neighbours to differentiate into the same cell type through lateral inhibition mediated by the Notch signalling pathway. Imayoshi et al. (2013) reported that the expression level of gene oscillations plays an important role in NPC differentiation. Upon segregating NPC according to their levels of expression of Hes1, Ascl1 and Olig2 and culturing them in differentiation medium, the authors found that high or low EC expression levels dictated different differentiation outcomes. For instance, Hes1-high NPCs differentiated into an astrocyte lineage, while Hes1-low NPCs into neurons (Imayoshi et al., 2013).

Similar to what is observed during neural development, Hes1 expression is required for the maintenance of multipotency in pancreatic progenitors, and undifferentiated neighbour cells undergo a mechanism of lateral inhibition to give rise to different cell types (de Lichtenberg et al., 2018). Even though pancreas development shares common players with somitogenesis and neural development, it was unknown if they displayed an oscillatory behaviour in this tissue. Recently, Seymour and colleagues (2020) reported that Hes1 and Dll1 proteins oscillate with a 90-min periodicity in cultured mouse pancreatic explants, and that this stimulates progenitor proliferation. The periodicity is different from the average 150 min period of Hes1 oscillations in NPC, which could be explained by lower levels of Notch activation in pancreatic progenitors. Importantly, extending the Hes1 oscillation period to ∼120 min by inhibiting NICD degradation altered cell fate specification (Seymour et al., 2020).

Oscillatory gene expression was also reported during chicken limb development (Pascoal et al., 2007) (Figure 3). Oscillations of hairy2 expression were first described in the chicken PSM, with the same periodicity as somite formation–90 min (Jouve et al., 2000). To study hairy2 expression dynamics in the developing chick forelimb, Pascoal et al. (2007) microsurgically removed one limb from HH22-26 embryos in ovo and reincubated the embryo for different periods of time. hairy2 expression was then assessed in each limb pair using in situ hybridization, revealing that hairy2 has very dynamic expression in the distal limb field, that is recapitulated every 6 h. The authors then determined that the time required to form a new autopod skeletal element is 12 h, suggesting that the limb chondrogenic precursor cells undergo two cycles of hairy2 expression for the formation of each autopod segmented element (Pascoal et al., 2007). This was the first evidence that a molecular clock is operating during limb development, a process where temporal control is also fundamental. hes4 is also expressed in the distal mesenchyme of the avian limb (Vasiliauskas et al., 2003) and recent work suggests that its expression is also cyclic during limb development Bhat et al. (2019). cultured cells from chicken pre-cartilage leg mesenchyme and observed oscillations of hes4 expression with a period of 6 h, suggesting that EC periodicity is a tissue-specific property.

The cases mentioned above clearly exemplify that the EC can play very distinct roles in different cells and tissues. For the formation of segmented structures such as somites and autopod limb elements, cells need to be synchronized to aggregate and give rise to a new segment. In multipotent cells, such as ESC, NPC and pancreatic progenitors, asynchronous EC oscillations function to allow heterogeneous cell fate responses of the population to a differentiation signal, ensuring the simultaneous specification of multiple cell types required for normal development.

As described above, most attempts to modify the EC led to a complete disruption of the oscillations. However, an increasing number of experimental approaches have succeeded in altering the pace of the EC, which is key to understanding how this biological oscillator is regulated and the functional significance of its temporal dynamics (Table 5).

Following the work of by Takashima et al. (2011), Harima and colleagues generated a mouse mutant lacking only the first two introns of the Hes7 gene. This resulted in Hes7 oscillations with an 11 min-faster periodicity than the WT mice. This shorter cycle did not abolish EC oscillations nor somite formation. As predicted by the Clock and Wavefront model, since the EC presented a faster pace, this culminated in more and smaller somites (Harima et al., 2013). Recently, a similar approach was used to modify the tempo of Hes1 oscillations in NPCs (Ochi et al., 2020). Deleting all the introns of the mouse Hes1 gene accelerated expression oscillations by 13.6 min. On the other hand, by substantially increasing the primary transcript length the authors obtained exactly the opposite result, delayed EC in 13.5 min (Ochi et al., 2020). These are powerful examples of how transcriptional delays of EC genes can be modulated to tinker gene expression dynamics.

In the EC negative feedback regulatory loop, HES proteins are imported to the nucleus and repress their own transcription. Repression is lifted upon protein degradation, allowing a new transcription cycle to begin. Hence, by changing protein stability the repressive time on the gene promoter also changes, culminating in an overall alteration in the tempo of the EC. Through replacement of a Lysine for an Arginine in position 14, Hirata and colleagues were able to increase the half-life of the HES7 protein by 10 min, which led to an increase of 10,2 min in the global pace of the Embryo Clock (Hirata et al., 2004). A similar approach was applied by Kim et al. (2011) by knocking-out Nrarp, a negative effector of Notch signalling. This approach, suggested to delay NICD turnover, extended the EC pace by 5 min and resulted in the formation of fewer and defective vertebrae (Kim et al., 2011). Wiedermann et al. (2015) also accomplished to delay EC oscillations by stabilising NICD in the chick embryo, further corroborating these findings.

A wealth of knowledge on EC pace manipulation has also been provided using the zebrafish model. The Oates lab produced mutants in genes belonging to the Notch signalling pathway that displayed slower EC oscillations (Herrgen et al., 2010; Schröter and Oates, 2010). Knock-out of notch1a, mib1, and deltaD slowed the EC pace by 7%, 19%, and 23%, respectively (Herrgen et al., 2010), and mutating hes6 delayed the EC by 6.5% (Schröter and Oates, 2010). Finally, chemical inhibition of the notch signalling pathway using DAPT delayed the EC by 18% (Herrgen et al., 2010). Corresponding delays in somitogenesis periodicity and a reduced final somite number corroborated the importance of Delta-Notch coupled oscillations for timely embryo body segmentation. Accordingly, elevation of Delta-Notch signalling accelerated EC oscillations and somite formation (Liao et al., 2016). Liao et al. (2016) created fish lines with 7 (Dover) or 100 (Damascus) extra copies of deltaD. Only the Damascus mutant displayed alterations to the EC, where oscillations were 1.6 min (6.4%) faster than in the wild-type. This increased the number of trunk segments by 7,6% and, despite the dramatic overexpression of deltaD, segmentation defects were rarely observed (Liao et al., 2016).

Other intercellular communication pathways contribute to the proper timing of the Embryo Clock. Sonic hedgehog (Shh) was shown to participate in EC tempo regulation. By comparing chicken PSM explants cultured with and without a source of Shh (notochord tissue and/or SHH-expressing cells), Resende et al. (2010) showed that the absence of Shh significantly delayed both EC oscillations and somite formation. Absence of Shh signalling led to an 85% increase of the EC period, from 90 min to approximately 2 h and 45 min. Similar experimental approaches showed that Wnt pathway inhibition by CKI-7 extended the EC pace from 90 to 120 min (Gibb et al., 2009). Comparable results were further obtained in the mouse model. Here, both CKI-7 treatment and activation of Wnt signalling using LiCl delayed Hes7 oscillations (González et al., 2013). Using PSM-like tissues induced from mouse ESC, Yoshioka-Kobayashi et al. (2020) performed a high throughput chemical library screening and identified multiple small compounds capable of altering the period of Hes7 oscillations by up to 40 min (Yoshioka-Kobayashi et al., 2020). These included modulators of a wide range of cellular processes and signalling pathways and further characterization of these alterations will surely improve our knowledge on EC operation.

Recently, we showed that the fibronectin-integrin-ROCK-NM II signalling axis regulates EC dynamics in the chicken PSM. Importantly, inhibition of actomyosin-mediated contractility delayed the period of hairy1 (hes4) oscillations from 90 to 120 min (Gomes de Almeida et al., 2022), unveiling a previously unappreciated biomechanical regulation of the EC periodicity.