Species-specific roles of the Notch ligands, receptors, and targets orchestrating the signaling landscape of the segmentation clock

Somitogenesis is a hallmark feature of all vertebrates and some invertebrate species that involves the periodic formation of block-like structures called somites. Somites are transient embryonic segments that eventually establish the entire vertebral column. A highly conserved molecular oscillator called the segmentation clock underlies this periodic event and the pace of this clock regulates the pace of somite formation. Although conserved signaling pathways govern the clock in most vertebrates, the mechanisms underlying the species-specific divergence in various clock characteristics remain elusive. For example, the segmentation clock in classical model species such as zebrafish, chick, and mouse embryos tick with a periodicity of ∼30, ∼90, and ∼120 min respectively. This enables them to form the species-specific number of vertebrae during their overall timespan of somitogenesis. Here, we perform a systematic review of the species-specific features of the segmentation clock with a keen focus on mouse embryos. We perform this review using three different perspectives: Notch-responsive clock genes, ligand-receptor dynamics, and synchronization between neighboring oscillators. We further review reports that use non-classical model organisms and in vitro model systems that complement our current understanding of the segmentation clock. Our review highlights the importance of comparative developmental biology to further our understanding of this essential developmental process.


Introduction Overview of the Notch signaling pathway
The canonical Notch signaling pathway is characterized by the activation of receptors on signal-receiving cells by ligands on signal-sending cells (Figure 1; Reviewed extensively in Kopan and Ilagan, 2009;Siebel and Lendahl, 2017;Zhou et al., 2022).In principle, the interacting extracellular domain of Notch ligands consists of a di-sulfide-rich DSL (Delta in mammals; Serrate in Drosophila; Lag-2 in C. elegans) domain and multiple tandem Epidermal Growth Factor (EGF) repeats.Notch receptors consist of a Notch Extracellular Domain (NECD), a Transmembrane Domain (TM), and a Notch Intracellular Domain (NICD) that are processed and fused in the endoplasmic reticulum and Golgi.Upon ligand-receptor binding, endocytosis is initiated in the signal-sending cell that exposes a characteristic S2 cleavage site in the NECD to the ADAM (A Disintegrin and Metalloprotease) family of proteases in the extra-cellular matrix (ECM).Subsequently, an S3 cleavage in the signal-receiving cell by the γ-secretase complex composed of Presenilin (Psen) 1/2, Nicastrin, APH1, and Presenilin enhancer (Psenen) releases the NICD to translocate into the nucleus.Here, NICD binds to a DNA-binding CSL complex (CBF1/RBPj in mammals, Suppressor of Hairless in Drosophila, Lag-1 in C. elegans) and a transcriptional coactivator mastermind-like (Maml 1-3) to initiate transcription of Notch target genes.In the absence of NICD, the CSL complex acts as a transcriptional repressor.Alternatively, a Notch ligand can bind the receptor intracellularly and prevent its trafficking to the cell surface thereby inhibiting the activation of this pathway (Figure 1).Both the activation and inhibition of this pathway are necessary for context-dependent biological processes (Reviewed extensively in Henrique and Schweisguth, 2019).
Although the fundamental components of this pathway remain conserved across different species, the numbers and function of Notch receptors and ligands vary greatly across model organisms due to the differential degrees of genome duplications and loss across species (Table 1).

Modes of regulation of the Notch pathwaya brief introduction
There are several layers of Notch signaling regulation, which have been extensively reviewed in the past (Kopan and Ilagan, 2009;Henrique and Schweisguth, 2019).Here, we will briefly introduce two, which will be complemented later in the review with a more focused view in the context of species-specificity and early development.
The first layer of regulation is at the level of Notch ligandreceptor interactions, which can happen in trans (ligand and receptors present on different cells; Figure 1), or cis (ligand and receptor on the same cell; Figure 1) (Jacobsen et al., 1998;Artavanis-Tsakonas et al., 1999;Hicks et al., 2000;Sprinzak et al., 2010;Nandagopal et al., 2019), depending on the Notch ligandreceptor pair interacting.For example, in vitro culture studies using immortalized cell lines have shown that Dll1 and Dll4 can act as trans-activating or cis-inhibiting ligands (Andrawes et al., 2013;Preuße et al., 2015), whereas Dll3 has been shown to act as a cis-inhibiting ligand by physically binding to Notch1 and blocking Notch signaling both in vitro in HEK293 cells and in vivo during both X. laevis and mouse neurogenesis (Ladi et al., 2005).Dll4 has been shown to have a strong cis-inhibitory effect on Notch1 receptors which can partially override Dll1-mediated trans-activation in vitro using HeLa cells and Chinese Hamster Ovary cells (Preuße et al., 2015).Another study, which also used immortalized HeLa cells in vitro, showed that Dll3 can act in cis, and potentiate the trans-activating capabilities of Dll1 (Bochter et al., 2022).While trans and cis-regulation have been mechanistically demonstrated, it is still challenging to generalize the activating/ inhibiting effects of different ligands observed in the above studies to particular tissue-or species-specific contexts.For example, mutant mice carrying knock-in Dll4 in the endogenous Dll1 locus demonstrated that Dll1 and Dll4 are completely interchangeable during early retinal development, partially interchangeable during myogenesis, and not interchangeable during somitogenesis (Preuße et al., 2015).
The second layer of Notch signaling control arises through posttranslational modifications of Notch receptors and ligands (Figure 1).Some of these include o-glucosylation, o-fucosylation, or N-Acetylglucosamination of consensus residues in the EGF repeats that allow a cell to distinguish between different ligandreceptor interactions (Kakuda and Haltiwanger, 2017).For example, o-glucosylation and o-fucosylation of the Notch receptors are carried out by Protein O-Glucosyl Transferase (POGLUT) and Protein O-Fucosyl Transferase (POFUT) respectively in the trans-Golgi network (Shi and Stanley, 2003;Acar et al., 2008).These modifications act as a base for N-Acetylglucosamination Overview of the Delta-Notch signaling pathway depicting transactivation and cis-inhibition.Trans-activation is initiated when the DSL domain of the Delta/Serrate/Jagged ligands binds to the extracellular domain of Notch receptors.This initiates two crucial cleavagesthe S2 cleavage by the ADAM protease; and subsequently an S3 cleavage by gamma-secretase.The two cleavages result in the production of free NICD that translocates into the nucleus to initiate transcription of target genes.Notch receptors, with are also a target of Notch signaling are subject to a wide range of post-translational modifications in the trans ER-Golgi space by Protein O-Fucosyltransferase (POFUT), Protein O-glucosyltransferase (POGLUT), Lunitic fringe (Lfng), etc., which subsequently affect the type of ligand-receptor interactions and downstream signaling.Illustrations adapted and modified from (Zhou et al., 2022).
imparted by Notch signaling targets called the Fringe proteins which modulate the receptor's ability to respond to ligands like Dll1 or Jag1 (Kakuda and Haltiwanger, 2017;Bochter et al., 2022).Seemingly, both Dll1 and Dll3 also have these conserved residues, making them amenable to these modifications (Bochter et al., 2022).Expression of one of the Fringe proteins-Lunatic Fringe (Lfng) in a signal-sending cell (containing the trans-activating ligand Dll1) has been shown to attenuate Notch-activation in vitro using HEK293T and NIH3T3 cells (Bochter et al., 2022).In vitro studies using immortalized cell lines have also shown that Lfng can potentiate Notch activation by Dll1, but inhibit activation by Jag1 (Brückner et al., 2000;Yang et al., 2005;Kakuda and Haltiwanger, 2017).This is intriguingly in contrast to in vivo data during somitogenesis between E8.75-E11.5,where lfng (the only Fringe gene expressed during mouse somitogenesis (Johnston et al., 1997)) can posttranslationally inhibit NICD formation from the Notch1 receptor, thereby generating a cyclic NICD profile (Morimoto et al., 2005;Shifley et al., 2008).These results suggest that post-translational modifications imparted by Lfng on Notch receptors and ligands can be lineage-specific, and their consequences can be highly contextspecific, for example, if Lfng is present in the signal-sending/ receiving cell.
Notch signaling, and these different modes of regulation play crucial roles during animal development, but more importantly, are crucial for proper somitogenesis (Reviewed in Hubaud and Pourquié, 2014;Oates et al., 2012).In the following sections, we will review this developmental process, its history, and the complexity it involves at both the tissue and cellular levels.

Somitogenesis and the segmentation clock
The process of somitogenesis is an early developmental event occurring between the human embryo Carnegie stage 9 to stage 13 or within the first month post-fertilization (Orts-Llorca, 1981;O'Rahilly and Müller, 2001;O'Rahilly and Müller, 2010).However, knowledge about this process in human embryos has been largely lacking and most of what we have learned to date has been through classical model organisms like chick, mouse, and zebrafish.Most recently however, single-cell RNA sequencing data from early human gastrulating embryos has helped us infer critical information about human somite formation (Xiang et al., 2019;Tyser et al., 2021;Zeng et al., 2023).
Somitogenesis involves the sequential formation of transient, block-like structures called somites that eventually give rise to the bones, skeletal muscles, and dermis associated with the vertebrae.Somites differentiate from the presomitic mesoderm (PSM) tissue that spans along the anterior-posterior axis of a developing embryo.In this tissue, somites bud off of the anterior PSM, while new mesoderm cells are being generated by the mesoderm progenitors at the posterior end (or the tail bud) (Chalamalasetty et al., 2014;Row et al., 2016).The rates of these two processes at opposing ends of the PSM regulate the shrinkage of the tissue over time.Once the tissue is exhausted, somitogenesis is terminated, resulting in a speciesspecific duration of somitogenesis and somite numbers (Yoshikawa et al., 1997;Yoon and Wold, 2000;Ciruna and Rossant, 2001;Fior et al., 2012).
Pioneering work regarding the mechanisms underlying somitogenesis used chick as a model system with crucial speculations made regarding the origin of somites.One of the many views postulated the existence of a somitic 'pre-pattern' in the primitive streak that gets translated into the somite series produced by the PSM (Pasteels, 1937;Waddington, 1952).These views were backed by fate-mapping techniques in both chick and mouse that suggested the presence of stem-like cells in the primitive streak whereby their position along the A-P axis defined the future organization of the somitic tissue (Selleck and Stern, 1991;Psychoyos and Stern, 1996;Wilson and Beddington, 1996;Eloy-Trinquet and Nicolas, 2002).
More evidence for the existence of a pre-pattern came when a part of the chick PSM, if surgically inverted even as early as Hamburger-Hamilton (HH) stage 2 (15-20 somite stage), formed somites in an opposite posterior to anterior fashion (Figure 2A; Christ et al., 1974;Palmeirim et al., 1998;Menkes and Sandor, 1977).

Species
Gene References

Mouse/ Human
Notch1, Notch2, Notch3, Notch4; NOTCH2NLA/B/C (Human) Additionally, when approximately one-fourth of a chick embryo at the gastrula stage was excised, or mutant mouse embryos with a shorter axis were analyzed around E10.5, the number of somites formed remained the same, with each new somite also being shorter containing fewer cells (Waddington and Deuchar, 1953;Flint et al., 1978).Landmark experiments performed using chick and snapping turtle embryos, where variable lengths of the segmental plate were excised and cultured in vitro showed that the number of somites formed from the intact segmental plate in the embryo and the explant, combined, always remained constant-about 10-12 somites (Figure 2B; Packard, 1978;Packard, 1980).Important work from Tam and others in the 1980s showed similar compensatory associations in mouse embryos.When one of the cells at the 2cell blastula stage was surgically removed, or embryos were treated with DNA replication inhibitors resulting in smaller, half-sized embryos, but having the conserved number of somites (Figure 2C; Tam, 1981).Similar results have been reported in small-sized zebrafish embryos, which show a smaller somite size such that the somite number remains unchanged (Ishimatsu et al., 2018).Another recent study has also re-enforced the view of a prepattern by performing grafts of the primitive streak (PS) from older (HH8) to younger (HH4) chick embryos (Busby et al., 2023).While GFP + grafts from an HH4 PS to an HH4 PS appeared in more anterior (first formed) somites, grafts from an HH8 PS transplanted into HH4 only populated the posterior (later formed) somites.This revealed that the progenitor addition to the PSM is regulated by an intrinsic timer or a definitive prepattern (Busby et al., 2023).
These studies showed and strongly corroborated the existence of a molecular pre-pattern that regulates the metameric organization program of somites.To explain its molecular basis, Cooke and Zeeman in 1976 theoretically postulated the presence of the classical 'Clock-and-Wavefront' model (Cooke and Zeeman, 1976;Reviewed in Resende et al., 2014).It predicted that the formation of repeated patterns with high precision is possible if the developing tissue consists of a "clock" made up of individual phase-locked cellular oscillators that interact with a slowly regressing front generated by a signaling gradient to change the cell state (Figure 3).It was not until 1997 when the first experimental evidence for the clock surfaced in the chick embryo, and was termed the 'Segmentation clock' (Palmeirim et al., 1997).Notch targets c-hairy1/2 which belong to the Hairy/enhancer of split (Hes/her/ Hey) group of basic-helix-loop-helix (bHLH) transcription factors (TFs) were shown to oscillate in the PSM.These oscillations occurred in a spatiotemporally coordinated manner resulting in "traveling" waves sweeping in a posterior-to-anterior direction once for every somite with a periodicity of 90 min (Palmeirim et al., 1997;Pourquié and Tam, 2001), matching the time taken to form one somite in chick, a process that begins at around HH stage 7. Oscillatory Hes/ her/Hey genes are also present in mice and zebrafish that make up the clock during somitogenesis (Table 2) (Reviewed in detail in Kageyama et al., 2007;Maroto et al., 2008).These genes have orthologous and paralogous relationships to one another given the large number of gene duplications in zebrafish (Figure 4; Taylor et al., 2001), and are critical to the segmentation clock to varying degrees as we will discuss later below.Readers interested in a detailed analysis of these orthologous A molecular prepattern underlies the specification of PSM cells from the tail bud (A).Inversion of the unsegmented chick PSM results in this tissue forming somites in the opposite direction (B).An explant of chick segmental plate will form 10 somites irrespective of the site/timing of cutting (C).A smaller mouse embryo will form shorter somites, but maintain the mouse-specific somite number.Illustrations adapted and modified from (Pourquié, 2004).
Despite the elegance of the clock-and-wavefront model in explaining pattern formation during somitogenesis, certain properties of the segmentation clock make this process more complex than this proposed model.

Complexities beyond the clock-andwavefront model
Determination front: gradient or cyclic?
Classically, the clock-and-wavefront model suggests the presence of regressing FGF/WNT gradients in the PSM that act as the 'determination front' determine the somite borders and the pace of somitogenesis (Figure 3; Sawada et al., 2001;Naiche et al., 2011;Akiyama et al., 2014).However, many studies have shown oscillating FGF and Wnt signaling targets in the caudal part of the mouse E8.5-10.5 PSM such that Notch and FGF targets oscillate inphase, and out-of-phase of Wnt targets (Figure 5A; Dequéant et al., 2006;Hayashi et al., 2009).Furthermore, Notch and Wnt targets oscillate in-phase in the rostral E10.5 PSM when grown ex vivo as 2D cultures and disrupting this spatial relationship between the two pathways can disrupt somitogenesis (Sonnen et al., 2018).In addition, Hes7 has been shown to regulate the expression of FGF and Wnt target genes like Dusp6, Sprouty2, Axin2, and Snai1 during somitogenesis.For example, studies showed that Hes7 represses the expression of FGF targets, Sprouty2 and Dusp4, in the E10.5 mouse PSM (Niwa et al., 2007;Hayashi et al., 2009).One study showed that disrupting Hes7 cycling by changing gene length disrupted Axin2 and Dusp6 cyclic expression in the E10.5 PSM (Takashima et al., 2011).On the contrary, another study showed that Hes7 −/− embryos still display dynamic Wnt (Axin2) and FGF (Sprouty2) target gene oscillations (Ferjentsik et al., 2009).This raises the question as to whether or not Wnt and FGF signaling oscillate or slowly regress as a gradient, or both (Figure 5A).Therefore, other models like the 'twophase model' or the 'phase-shift model' have also been suggested that incorporate an interacting Notch, FGF/Erk, and Wnt oscillator (Reviewed in detail in Hubaud and Pourquié, 2014).Whether these expression patterns co-exist during the segmentation clock or are mutually exclusive remains to be resolved (Aulehla et al., 2003).It is possible to speculate that the wavefront is oscillatory under the control of Hes genes in the middle and anterior PSM but acts as a gradient in the more posterior tissue.
More importantly, a recent study on the zebrafish posterior PSM showed that, unlike mice, Notch-target her7 and Erk signaling oscillate out of phase (Figure 6A; Simsek et al., 2023).This species-specific difference adds an additional layer of complexity to uncovering the mechanisms underlying these phase relationships between the different signaling pathways.
Another deviation from the classic clock-and-wavefront model, that emerged in the decade following its discovery was that the clock is not constant, but rather highly dynamic at both the spatial and temporal axes.

Changing periodicity of the segmentation clock spatially along the A-P axis
Firstly, the period of the segmentation clock varies along the A-P axis of the PSM, slowing down in the anterior PSM, thereby Schematic representation of the 'Clock-and-wavefront' model that shows the existence of a clock travelling from the posterior to the anterior PSM (in grey) coupled with a regressing wavefront (in yellow) which interact precisely with another T-box TF, Tbx6 (in cyan) to specify anterior PSM cells to differentiate into a somite.S-1: unsegmented PSM that will form the next somite; S0-The somite that is being specified; S1-The most recently formed somite; S2-The earliest formed somite that has completed rostro-caudal patterning (blue-red gradient).Illustrations adapted and modified from (Pourquié, 2011).
generating waves of cyclic gene expression (Figure 5B).This phenomenon is apparent in the mouse and zebrafish PSM (Soroldoni et al., 2014;Shih et al., 2015;Liao et al., 2016;Tsiairis and Aulehla, 2016;Liao and Oates, 2017).In a 10-12 somite-stage zebrafish embryo, the her1 clock in the posterior PSM oscillates with a 1-somite periodicity (30 min), whereas with a slower 2-somite periodicity (~60 min) in the anterior PSM (Shankaran et al., 2007;Shih et al., 2015).In mice, the period increases from ~130 min in the posterior PSM to ~170 min in the anterior PSM as assayed by oscillations of another clock gene Lfng in E10.5 2D PSM explants (Tsiairis and Aulehla, 2016).Alongside a period gradient, there also exists a 'phase gradient' such that somite formation in the anterior PSM always coincides with a 2π phase-shift of the clock compared to the posterior PSM (Lauschke et al., 2012).Essentially, this means that the cells in the anterior PSM are one cycle apart from the cells present posteriorly, reiterating the passing of one cycle for every somite formed.Additionally, a newly formed somite always spans ~20% of this phase-gradient irrespective of PSM length (Figure 5B; Lauschke et al., 2012).In other words, a smaller PSM would form a "steeper" phase gradient to maintain this 2π phase shift, and as a result, 20% of this gradient results in smaller segments (Figure 5B).This presents a plausible mechanism underlying smaller-sized embryos forming the same number of (and smaller) somites.
One potential consequence of such a period/phase gradient has been hypothesized based on the existence of a "secondary" nonchanging oscillator along the length of the PSM (Goodwin and Cohen, 1969;Lauschke et al., 2012).The phase of the segmentation clock changes as cells move along the PSM, and its relative phase difference with this secondary oscillator can help give cells the necessary spatial (alongside temporal) information to differentiate into a somite.One potential candidate for this secondary oscillator in mice is Wnt signaling.As mentioned in the previous sub-section, disrupting the phase relationship between Notch and Wnt oscillations results in disorganized somitogenesis.In zebrafish, one candidate is her12 whose phase relationship with her1 also changes along the PSM (Shankaran et al., 2007).However, whether or not these candidate oscillators also display a period/phase gradient along the PSM needs to be systematically characterized.
But what is the cause underlying this gradient?Chemical-based reduction in the levels of Wnt signaling in both mouse and chick Clustal-based phylogenetic tree of mice (m) and zebrafish (z) Hes1/5/7 orthologs.Amino acid sequences of mHes7 (NP_149030.2),mHes5 (NP_ 034549.1),mHes1 (P35428) zher1 (Q90463), zher7 (Q9I9K1), zher11 (Q6W4T8), zher2 (Q90464), zher4 (Q90466), zher12 (Q6TA36), zher15 (Q7T3J0), and zher6 (Q6PBX3) were first aligned in the FASTA format using MUltiple Sequence Comparison by Log-Expectation (MUSCLE).The FASTA alignment file was input into the EMBL-EBI Simple Phylogeny tool.The tree was generated in a Clustal format with distance corrections and a UPGMA (Unweighted Pair Group Method with Arithmetic Mean) clustering method.The phylogenetic tree depicts the separation of the Hes1, Hes5, and Hes7 clusters.The symbol beside each gene shows their expression dynamics in the PSM, which is unclear specifically for her6 as it is expressed only during the 3-5 somite stage during fish segmentation.Complexities beyond the clock and wavefront (A).The regressing Wnt/FGF-based gradients from the posterior to anterior PSM specify the determination front, however, targets of these signaling pathways (green-Wnt; yellow-FGF) oscillate in the PSM, with changing phase relationships to Notch signaling targets (grey) (B).The period and the phase of cyclic gene oscillations change along the length of the PSM (C).The period of the segmentation clock changes temporally as somitogenesis proceeds (D).The anterior and posterior body segmentation are plausibly regulated by separate mechanisms.Illustrations adapted and modified from (Schröter et al., 2008;Lauschke et al., 2012).
PSM slows down the clock, forming one fewer somite compared to the untreated controls (Gibb et al., 2009).The Wnt3a ligand exists as a gradient in the PSM (Aulehla et al., 2003;2008), and as a result, reducing levels of Wnt signaling along the length of the PSM could be one potential reason for the period gradient.Perhaps, understanding the molecular and genetic mechanisms underlying the period gradient within a particular species could help us better understand how and why the clock time differs across different species.

Temporal alteration in the segmentation clock period
The period of the clock also changes as segmentation proceeds, remaining rather constant for most of the duration of somite formation, but drastically slowing down by ~1.5-fold for the last few somites in both chick and zebrafish (Figure 5C; Schröter et al., 2008;Gibb et al., 2009;Tenin et al., 2010).This layer of complexity has been less characterized during mouse segmentation.The time taken for somite formation in chick embryos between HH22-HH24 decelerates from 90 min to ~150 min, and this slowing correlates with when Wnt3a levels change from a gradient in the PSM to being confined to the tip of the tail bud to finally disappearing (Gibb et al., 2009;Tenin et al., 2010).During this time, most if not all Notch activity is also lost in the PSM and the clock terminates before all somites are formed at HH25, suggesting a clock-independent regulation of somite formation towards the later stages of development (Tenin et al., 2010).Interestingly, something similar is observed in mouse embryos at E12.5 when all Wnt activity is restricted to the caudal tip and is completely lost at E13.5 before somitogenesis ceases at E14.5, suggesting a similar temporal alteration of the clock during mouse somitogenesis (Gibb et al., 2009).These results also suggest the intriguing possibility that the slower mouse clock (compared to chick) is due to the mouse PSM intrinsically expressing lower levels of Wnt ligands/effectors compared to the chick PSM.

Separate mechanisms regulating the clock along the A-P axis
Another layer of complexity in the process of somitogenesis arises from the speculations that the mechanisms regulating this process differ during anterior and posterior body specification (Figure 5D).The switch between primary and secondary body segmentation in mice occurs at the lumbosacral junction (Gossler and Tam, 2002).
Both Dll3 −/− and Lfng −/− mice have highly disorganized thoracic and lumbar vertebrae, however, Lfng −/− mice can produce several normal sacral vertebrae before forming a disorganized sacrum and a short tail (Stauber et al., 2009;Williams et al., 2014;2016;Bochter et al., 2022).Dll3 −/− ; Lfng −/− mice produced no normal sacral Transcriptional landscape of the segmentation clock.(A) A simplified schematic of the complex regulatory landscape underlying the mouse and zebrafish segmentation clocks.(B) "Core" and "Peripheral" bHLH genes and Lfng, amongst many others, occupy this regulatory landscape.Based on mutant phenotypes, these clock genes regulate the segmentation clock to varying degrees.Zebrafish consists of many more her/hey genes as a part of its segmentation clock, as compared to mouse or chick, plausibly making the zebrafish clock more robust to genetic or environmental perturbations compared to other species.Finally, we know very little about similar core vs. peripheral clock genes in other species including higher vertebrates like humans, resulting in a knowledge gap that needs to be studied to fully understand this biological clock.
vertebrae, suggesting that while Lfng and Dll3 may work together during anterior body specification, they exhibit independent roles posteriorly.Furthermore, the deletion of an enhancer Fringe clock element 1 (FCE1) upstream of Lfng drastically reduced Lfng expression in the caudal E10.5 PSM while maintaining its anterior expression.Lfng ΔFCE1/ΔFCE1 embryos phenocopied Lfng −/− embryos during anterior segmentation including heavily disorganized and fused somites and missing ribs, but formed normal tail vertebrae (Shifley et al., 2008).qRT-PCR showed, unlike Lfng −/− , Lfng ΔFCE1/ΔFCE1 embryos still showed very low cyclic Lfng activity (Shifley et al., 2008;Williams et al., 2014).Similarly, while Dll1 −/− mouse embryos fail to form somites throughout the body axis (De Angelis et al., 1997), embryos homozygous for a dominant negative allele of Dll1 can segment, but show fused/missing and split somites only during cervical and thoracic segmentation (Cordes et al., 2004).This suggests that while robust, high-level cyclic Lfng or Notch ligand function is crucial for primary body segmentation, this necessity is no longer present posteriorly, indicating a switch in mechanisms underlying the segmentation clock.Finally, FGF4 mouse mutants also show disorganized somite specification only during the 5-6 somite stage (E8.5) but show normal somitogenesis during the 24-26 somite stage (E9.5),corroborating this complexity (Anderson et al., 2020).
As we will also discuss later below, her1 −/− zebrafish embryos, or her1-morpholino treated embryos show minor segmentation defects with diffuse somite borders only for the first 3-5 somites, while her7 knockdown resulted in severe somite border defects after a 5-somite delay (Henry et al., 2002).Furthermore, her1 −/− ; her7 −/− embryos do not form somites throughout the embryonic axis (Henry et al., 2002).Inserting her1 and her7 transgenes in single chromosomes (different from the endogenous locations) in these double-knockout embryos resulted in partially rescued posterior, but not anterior segmentation (Keseroglu et al., 2023).This again suggests separate mechanisms regulating somitogenesis along the embryonic axis.These complexities arise as a result of intercommunication between the Notch, Wnt, and FGF signaling pathways.
In the following sections, we review the species-specific Notch signaling dynamics of the segmentation clock with a focus on bHLH-TFs, ligand-receptor functions, and how these dynamics are propagated by Notch-mediated cell synchronization.Readers interested in a detailed review specifically on the zebrafish segmentation clock can refer to (Venzin and Oates, 2020).

bHLH-transcription factors as the core regulators of the segmentation clock
The Notch-responsive family of bHLH-TFs are transcriptional repressors that can bind to the E-box or N-box regions of certain promotors and suppress gene expression (Bessho et al., 2001b;Jones, 2004;Chen et al., 2005).Some of these transcription factors also bind to their own promoters to generate molecular oscillations via a negative feedback loop in the PSM (Holley et al., 2002;Bessho et al., 2003;Jones, 2004;Giudicelli et al., 2007).Mutating these binding sites on target promoters prevents this binding in vitro (Chen et al., 2005).Murine orthologs of Notch-responsive bHLH TFs c-hairy (Hes/Hey genes) include Hes1-7, Hey1-2, and HeyL, of which, Hes1, Hes5, Hes7, and Hey2 are expressed cyclically in the E8.5-E10.5 mouse PSM (Del Barco Barrantes et al., 1999;Jouve et al., 2000;Leimeister et al., 2000;Bessho et al., 2001b).Hey1 transcripts can also be detected in the mouse E9.5 embryos by in situ hybridization, but are not oscillatory (Leimeister et al., 1999).However, microarray experiments have shown that Hey1 indeed oscillates in the mouse E9.5 PSM (Krol et al., 2011).Similar to the first described expression patterns in chick embryos, mouse embryos also show autonomous traveling waves of cyclic gene expression sweeping along the PSM with a mouse-specific periodicity of 2 h (Aulehla et al., 2008).In fact, the Hes1 gene has also been shown to cycle with a conserved 2-h periodicity in other cell types like neural progenitor cells ex vivo, fibroblasts, myoblasts, and embryonic stem cells in vitro, suggesting that a species-specific ultradian rhythm of these bHLH-transcription factors is a conserved, vital feature of vertebrate development (Hirata et al., 2002;William et al., 2007;Ochi et al., 2020).This has been reviewed in detail (Carraco et al., 2022).
Interestingly, a high divergence has been observed in vivo and in silico between chick, mouse, and zebrafish when comparing oscillating genes in the PSM (Krol et al., 2011).Specifically, while the Notch, Wnt, and FGF signaling pathway targets oscillate in all three species, the cyclic expression patterns of orthologs of only two genes-Hes1 and Hes5 are conserved (Krol et al., 2011).We speculate that comparisons across more species could reveal additional conserved oscillators.As we will discuss in the following sections, a comparative analysis of the various Hes/ her genes reveal diverging core clock genes and their importance in regulating the segmentation clock.We will review, in-depth, the complex regulatory web casted by these transcription factors during somitogenesis with the anticipation of detangling this species-specific divergence.

Loss of function phenotypes reveal core versus peripheral players of the mouse segmentation clock
While Hes genes are cyclically expressed in the mouse PSM, not all of them seem to be essential for maintaining the clock.For example, Hes7 −/− E8.5 and E9.5 PSM lose all Notch-responsive cyclic activity, including that of other clock genes like Lfng and Hes1, and show misexpression of somite maturation markers resulting in disfigured fused somites (Table 3; Bessho et al., 2001a;Ferjentsik et al., 2009;Hirata et al., 2004).However, the expression of upstream components like the Delta ligands and Notch receptors is unaffected suggesting the hierarchy of Notch signaling (Bessho et al., 2001b).On the other hand, Hes1 −/− mice still show cyclic Lfng activity in the E10.5 PSM and somite formation is unaffected (Jouve et al., 2000).Lfng −/− E9.5-E10.5 PSM also show cyclic Hes7 and Notch activity, but form disfigured somites, a phenotype that is still less severe compared to Hes7 −/− mice (Ferjentsik et al., 2009).Conceptually, this puts Hes7 as a core clock gene, Lfng as a peripheral clock gene, and Hes1 as a more peripheral gene (Figure 6B).In fact, the promoter of Lfng consists of N-box sequences to which Hes7 physically binds and represses expression (Chen et al., 2005).Interestingly, DNasefootprinting assays have shown that Hes1 only binds N-box sequences (and not E-boxes) to repress transcription (Sasai et al., 1992).Yet, Hes1 does not seem to affect Lfng transcription, indicating that these Notch signaling-related DNA-protein interactions are also highly lineage-specific.This kind of core vs. peripheral function may arise as a result of functional redundancy where the loss of Hes1 is fully compensated by Hes7.However, what makes one Hes gene a core player over another remains elusive.Despite this core function of Hes7, somites are formed nonetheless in Hes7 −/− mice, albeit disfigured.This suggests the possibility of other core clock genes or the presence of signaling pathways like FGF or Wnt regulating the core clock circuit alongside Notch.Indeed, several Notch loss-of-function (LOF) mutant embryos that lose all cyclic Hes7 activity in the rostral and medial E9.5-E10.5 PSM, still show some caudal Hes7 expression, which is abrogated upon FGF inhibition (Ferjentsik et al., 2009).This suggests that FGF-induced Hes7 expression in the posterior PSM acts independently of Notch and possibly acts as a clock initiator.Whether or not this FGF-driven Hes7 is cyclic needs to be systemically characterized since one study has shown it to be non-cyclic (Ferjentsik et al., 2009), while another showed that this expression is dynamic (Niwa et al., 2007), despite both studies using in situ hybridization of the E9.5 caudal PSM.Furthermore, both Dll1 −/− and Notch1 −/− PSM at E9.5 show a caudally restricted Hey2, but not Hey1 expression, strengthening the idea of a caudally restricted Notch-independent clock centered around bHLHtranscription factors (Leimeister et al., 2000).More evidence of a peripheral role of Hey1 comes from the fact that its ubiquitous upregulated expression throughout the E9.5 PSM could still form up to ~13 somites (Feller et al., 2008).
Surprisingly, the functioning of the core gene Hes7 appears to be RBPjk-independent despite its known canonical requirement for Notch target gene expression (Figure 1).RBPjk −/− embryos, which die at E9.5, still show cyclic Hes7 activity in the PSM (Ferjentsik et al., 2009).In agreement, a recent study used live light-sheet imaging of transgenic Lfng reporter carrying mice to study the onset of oscillatory dynamics in E6.5 mouse embryos and showed two interesting observations.First, Lfng shows a huge burst in its expression followed by two to three oscillations before the onset of somite formation.Second, and more importantly, RBPjk −/− embryos also show four to five Lfng oscillations prior to the onset of somitogenesis, but form disrupted somite borders subsequently (Falk et al., 2022).Although Hes7 acts upstream of Lfng and as a result, this burst of Lfng expression does not inform us about Hes7 dynamics, these results suggest an RBPJk-independent Notch activation.On the other hand, mice embryos mutant for Psen1/2, which lack γsecretase activity do not form any somites (Huppert et al., 2005;Ferjentsik et al., 2009).These results could suggest a noncanonical Notch pathway downstream of Psen1 cleavage diverging before RBPjk function that still activates the core and peripheral clock genes to regulate the mouse segmentation clock.
her7 acts as the core clock gene during zebrafish segmentation In zebrafish, due to the high number of gene duplications (Taylor et al., 2001), there are approximately 23 her/hey genes of which 10 are expressed in the PSM (Table 2).Of these, 7 genes-her1, her7, her2, her4, her12, her15, and zhey1 oscillate (Table 2; Holley et al., 2000;Oates and Ho, 2002;Winkler et al., 2003;Sieger et al., 2004;Shankaran et al., 2007;Krol et al., 2011).One exception is her6 which is expressed in the PSM only between the 3 and 5-somite stage embryo, following which its expression is undetectable in the PSM and restricted only to newly formed somites (Pasini et al., 2004).Each of these different her genes are critical for regulating segmentation, but to different extents.
Embryos at the 5-10 somite stages in which her7 expression was disrupted either by morpholino-mediated knockdown or by its overexpression, resulted in a massive downregulation of its own transcription, that of her1 and other clock genes like deltaC, her12, and her15 (Oates and Ho, 2002;Gajewski et al., 2003;Giudicelli et al., 2007;Shankaran et al., 2007).Loss of her7 transcription also resulted in disrupted somite borders and fused somites throughout the embryonic axis after a 5-somite delay (Henry et al., 2002;Giudicelli et al., 2007).Loss of her1 expression by a genomic deletion, morpholino, or overexpression also resulted in a complete loss of its transcription, however, her7 and other cyclic gene transcripts including deltaC, her12, and her15 were attenuated but not fully lost, and were still cycling (Holley et al., 2002;Oates and Ho, 2002;Gajewski et al., 2003;Giudicelli et al., 2007;Shankaran et al., 2007).In agreement, neither her1-deletion nor her1morpholino induced any severe segmentation phenotypes, other than deformed boundaries and elongated first few somites (Henry et al., 2002;Schröter and Oates, 2010).Furthermore, these defective boundaries were only observed in ~85% of mutant embryos, suggesting incomplete penetrance which was not observed in her7 mutant embryos (Keseroglu et al., 2023).
These loss-of-function phenotypes suggest that her7 and her1 may behave as a core and peripheral clock gene respectively (Figure 6B), despite the fact that the two genes are genomically paired, are co-expressed, and show overlapping expression patterns in the PSM (Gajewski et al., 2003;Zinani et al., 2020).In addition, a combined loss of her1 and her7 led to a higher penetrance of the somitic defects by disrupting somite borders along the whole embryonic axis including the first few somites indicating a complementary function or redundancy between the two genes (Henry et al., 2002;Oates and Ho, 2002;Sieger et al., 2004).
Unlike her1 or her7, another bHLH-TF her13.2(hes6 homolog) is independent of Delta-Notch signaling and under the direct control of FGF signaling.Due to this, her13.2does not oscillate in the PSM, but is expressed in a posterior-to-anterior gradient (Kawamura et al., 2005).Potentially as a result of this differential expression, a complete loss of her13.2,either by a genomic deletion or morpholino-mediated knockdown did not result in any severe segmentation phenotypes, and instead slowed down the clock, resulting in fewer (six to seven fewer), but bigger somites (Schröter and Oates, 2010).Interestingly, the addition of a her1morpholino in a her13.2−/− background or alongside her13.2morpholinoresulted in severe disruption of segmentation along the entire embryonic axis, indicating a genetic interaction between the two genes (Schröter and Oates, 2010).Morpholino-mediated knockdown of either her4 or her6 alone resulted in a severe disruption of her1 expression in the 10-12 somite stage embryo but showed minor segmentation defects with few somitic fusions (Takke and Campos-Ortega, 1999).This is particularly surprising given that her6 is not expressed in the PSM at this stage, indicating interesting somite-PSM genetic crosstalk which has not yet been seen in mice or chick.Similar to her13.2, double morpholinoknockdown of her1 and her4 resulted in severe defects with multiple fused somites (Takke and Campos-Ortega, 1999).
Four other PSM-specific her genes, her11, her12, her15, and zhey1 are only partially important for the maintenance of the segmentation clock.Misexpression of either her12 or her15 in the 10-12 somite stage embryos disrupted the cycling of clock genes and showed diffused somite borders in only ~50% of the embryos (Shankaran et al., 2007).This low penetrance could be due to the wide range of homo/heterodimers that the Her proteins form to regulate gene expression (see below).An even lesser effect is seen when using MO-mediated knockdown of her11 or zhey1 at the 8-10 somite stage, where cyclic expression of her1/7 and other cyclic genes like deltaC is unaffected (Sieger et al., 2004).
Despite these diverging roles of individual bHLH-transcription factors, these results provide evidence that the hairy/Hes/her genes control their own expression, as well as that of other oscillating bHLH genes and other clock genes in the segmentation clock (Figures 6A, B;Bessho et al., 2001a;Ferjentsik et al., 2009;Shifley et al., 2008;Zhang and Gridley, 1998).Finally, although Lfng is expressed in the segmented somites, it does not oscillate in the zebrafish PSM (Mayburd et al., 2001;Prince et al., 2001).This emphasizes that, at least in zebrafish and mice, Hes7/her7 genes act as the core clock genes (Figure 6B).

Lunatic fringe, but not bHLH TFs, is the core driver of chick segmentation
Expression studies in chick have also shown a similar functional divergence between bHLH genes like c-hes1, c-hey1, c-hairy1, c-hey2, c-hairy2, and c-hes5, all of which show different expression pattern in the chick PSM at HH stage 9-10 ( Palmeirim et al., 1997;Jouve et al., 2000;Leimeister et al., 2000;Krol et al., 2011).
Contrary to mouse or zebrafish, in chick, c-lfng may act as a core player (Figure 6B).Overexpression of c-lfng leads to a downregulation and loss of c-lfng, c-hairy1, and c-hairy2 cyclic mRNA expression in the PSM, which phenocopied Notch inhibition in the chick embryo (Dale et al., 2003).Additionally, microRNA-125a-5p-mediated post-transcriptional regulation of c-lfng in vivo is essential for chick segmentation suggesting that a tight control of its expression in the PSM is crucial for driving the segmentation clock (Riley et al., 2013).Interestingly, mir-125a-5p is dispensable for the mouse segmentation clock (Wahi et al., 2017), strengthening the view of a more core function of c-lfng in chick compared to the Hes genes in mice.In line with c-lfng being a core clock gene during chick somitogenesis, it also plays a critical role in somite polarization.In the anterior PSM, c-lfng is expressed as a tight band in the presumptive somite border.Mis-regulation of c-lfng expression results in disrupted somite borders and compartmentalization (Dale et al., 2003).Interestingly, when a group of PSM cells at the presumptive somite border were transplanted into an un-segmented PSM region, it induced ectopic fissure and border formation in the transplanted area (Sato et al., 2002).Essentially, border cells contained positional information about border formation which was encoded by a high c-lfng expression.This information was maintained even when transplanted elsewhere in the PSM resulting in ectopic somite borders.
A core function of hes/hairy genes in the chick segmentation clock cannot be ruled out; however, the lack of studies characterizing their roles through RNAi-mediated knockdowns makes it difficult to accurately predict these functions.

Half-lives and transcriptional delays of bHLH-Transcription factors
There are several features of the mouse segmentation clock that give it its characteristic 120-min period, one of them being delays associated with transcription and protein expression (Jensen et al., 2003;Monk, 2003;Hirata et al., 2004).One of the events that can affect these delays is splicing (Takashima et al., 2011;Alpert et al., 2017).
Introduction of an intronless Hes7 transgene fused to a luciferase showed that the luminescence preceded the endogenous protein expression domain by ~21 min in E10.5 mouse PSM explants, suggesting a reduced delay in protein expression (Takashima et al., 2011).Surprisingly however, the period of the transgenic Hes7 did not significantly change, possibly due to interference from the endogenous Hes7.Mutant mice containing intronless Hes7 form highly disfigured, fused somites and a truncated body axis phenocopying Hes7 −/− mice, indicating the importance of fine delays in maintaining correct patterning (Takashima et al., 2011).In fact, transgenic mice containing intronless Hes7, and not WT Hes7 showed severe segmentation defects resembling Hes7 −/− mice (Harima et al., 2013).This suggests that intronic delay, but not copy number per se is essential to maintain proper Hes7 oscillations, and thereby somitogenesis.In contrast, another study reported kinked tails in Hes7 +/− mice, indicating that the dose of Hes7, only when it drops below WT levels, could be important for somitogenesis (Bessho et al., 2001b).
Remarkably, introducing a transgene lacking two of the three Hes7 introns did not lead to any severe segmentation defects, but formed additional cervical and upper-thoracic segments, suggesting an acceleration of the clock.Furthermore, the removal of any one intron did not affect the clock or somitogenesis at all (Harima et al., 2013).These results show that while intronic delay is crucial, this criticality is dependent upon the number of introns present in a "dose-dependent" manner.
Once the Hes7 mRNA is spliced and the Hes7 protein is made, it needs to be degraded with rapid turnover to allow for the next cycle of expression to begin.Stabilizing the mouse Hes7 protein by mutating ubiquitination sites on Hes7 prevented its degradation and increased its half-life from 22 min (WT) to ~30 min when measured in vitro (Hirata et al., 2004).In silico models showed that this small increase in Hes7 half-life severely dampened oscillations after 3-4 cycles.In agreement, knock-in mice containing this mutant Hes7 showed normal somitogenesis and cyclic Hes7 and Lfng expression until the 3-4 somite stage, following which the clock was fully disrupted resulting in defective segmentation (Hirata et al., 2004).These results show that protein turnover rates are central to maintaining proper cyclic gene expression, and therefore proper body segmentation.
In fact, a recent study used a transgenic Hes7 construct to show that the mouse Hes7 has an ~2-fold faster decay rate compared to the human HES7, which correlates with the ~2-fold faster mouse clock (Matsuda et al., 2020a).The authors also found that the human HES7 decayed at a mouse-specific rate when the transgene was expressed in the mouse PSM in vitro.Global translation inhibition in Zebrafish embryos at the 12-14somite stage revealed that the half-life of Her7 is only ~3.5 min (Giudicelli et al., 2007;Ay et al., 2013), compared to an ~20 min and ~40 min half-life of mouse Hes7 and human HES7 respectively (Matsuda et al., 2020a).These Hes/Her protein half-lives correlate with species-specific clock periods.These studies suggest that beyond the gene sequence itself, the surrounding transcriptional, translational, and degradation machinery plays a crucial role in regulating species-specific developmental timing.
However, this correlation may not extend to the chick segmentation clock, where blocking de-novo protein synthesis did not affect the dynamicity of either c-hairy1 or c-hey2 expression.This suggests a different mechanism for cyclic expression in chick (Palmeirim et al., 1997;McGrew et al., 1998;Leimeister et al., 2000).As a result, transcriptional/protein decay rates may only be one of the determinants of species-specific clock periods, and further comparative studies will be necessary to uncover the other ones.

Dimerization of bHLH-transcription factors
bHLH-TFs like the Hes/her genes have the ability to form homo/ hetero-dimerize with other bHLH genes via their helix-loop-helix structures (Jones, 2004).Altering dimerization parameters of Hes7 in silico can alter clock properties in the mouse PSM, however in vivo data for the same is lacking (Song et al., 2011).
In zebrafish however, the dimerization network of bHLH-TFs has been more extensively characterized.her genes form homo-and heterodimers in a "dimer cloud" (Figure 6A; Schröter et al., 2012), and this dimerization ability is essential for pattern formation in the fish embryo (Henry et al., 2002;Zinani et al., 2020;Zinani et al., 2022).her1 and her7 are present head-to-head on the same chromosome which enables their co-transcription and co-repression by a Her1/7 dimer.One study mutated the her1 and her7 genes on opposite sister chromatids resulting in reduced correlated gene expression dynamics and disrupted somitogenesis (Zinani et al., 2020).Her7 and Her13.2 also form heterodimers and enable autorepression of the her1 and her7, but also modulate the dimer topology within each cell crucial for the functioning of the clock (Figure 6A; Schröter et al., 2012).Her13.2 also affects the stability of the Her1/7 dimer via proteinprotein competition and its loss results in increased stability of this dimer and a slower clock (Schröter et al., 2012).On the other hand, another study reported that Her7 acts as the critical node of the dimer topology and limits the availability of Hes13.2 to dimerize with the other cyclic Her proteins like Her1, Her12, and Her15 (Trofka et al., 2012).Indeed, electrophoretic mobility shift assays (EMSA) have shown that the cis-regulatory elements of her1 contain Her12 binding sites (Brend and Holley, 2009), and that Her12 plays an important role in regulating her1 expression (Shankaran et al., 2007).
Nonetheless, these studies suggest the complex interplay between bHLH-TFs necessary for somitogenesis, a feature not as extensive in mice or chick due to the limited number of Hes/Hey genes in the two species.This brings out an intriguing question-Does an extensive network of bHLH-TFs display redundancy in a way that makes zebrafish somitogenesis more robust to environmental/genetic aberrations?her1 knockout embryos show only ~85% penetrance in segmentation defects suggesting some functional compensation by other her genes (Keseroglu et al., 2023).More comparative studies using different her mutants and across different species are necessary to address this question.
Upstream players of the Notch signaling pathway like the ligands and receptors play an important role in regulating and synchronizing this complex transcriptional landscape painted by bHLH-TFs, and the loss of these upstream components also disrupts somitogenesis.
These data suggest while Dll1 is crucial for regulating the expression of all Hes genes and Lfng, Dll3 is crucial only for Lfng and perhaps more peripheral bHLH gene expression.Potentially due to this "stronger" influence of Dll1, Dll1 −/− mice are unable to establish any rostrocaudal compartmentalization, fail to make epithelial somites, show incomplete axis formation, and eventually die by E12 (De Angelis et al., 1997;Preuße et al., 2015;Shimojo et al., 2016).On the other hand, loss of Dll3 results in less severe phenotypes where although somite borders are not visible, clear metameric units are evident, albeit highly disorganized, which eventually also differentiate into different somitic lineages (Dunwoodie et al., 2002).Furthermore, Dll3 −/− mice die postnatally within 10 days of birth (Dunwoodie et al., 2002).These phenotypes also reinforce the influence of a loss of core clock genes in Dll1 −/− embryos versus the loss of more peripheral genes in Dll3 −/− embryos.Interestingly, there are many more documented human DLL3 clinical mutations (versus DLL1) that result in a disorganized vertebral column.To date, there have been 28 reported DLL3 mutations resulting in a form of spondylocostal dysostosis (SCDO), SCDO1, which is a type of vertebral malformation (Kusumi et al., 2004;Umair et al., 2022).On the other hand, there has been only 1 DLL1 mutation causing another type of SCDO, SCDO7 (Barhoumi et al., 2019).One plausible reason for this could be that, similar to mice, mutations in DLL1 (as opposed to DLL3) are likely to be more embryonic lethal in humans, resulting in this disparity in reported cases.
All of the above studies combined show that a complex, spatiotemporally regulated medley of Notch ligands, receptors, and their targets is of critical importance in regulating the mouse segmentation clock.
Zebrafish deltaC may be more similar to the mouse Dll1 In zebrafish, deltaC (beamter; bea) and deltaD (after eight; aei) LOF mutant embryos show different segmentation phenotypes in the PSM, suggesting that they may serve separate roles in regulating the segmentation clock (Holley et al., 2000;Holley et al., 2002;Mara et al., 2007).While deltaD (aei) mutants display defective somitogenesis after 8-9 somites, deltaC (bea) mutants show disruption following just 3-4 somites indicating that, like dll1, deltaC acts as a more core regulator of segmentation (Van Eeden et al., 1996;1998;Jiang et al., 2000).In fact, aei/bea double mutants show the same segmentation phenotypes as the bea mutants, reflecting the more primary function of deltaC (Jülich et al., 2005).This is in contrast to expression patterns of cyclic genes, where in situ hybridization in bea mutants display a salt-and-pepper expression of her1 and her7 throughout the 8-12 somite stage PSM, whereas her1 expression is completely lost (except in the anterior and posterior tips) in aei mutants (Holley et al., 2000;2002).This phenotype is partially reminiscent of Hes7 expression in Dll1 −/− mouse PSM.In contradiction, one study used high-resolution single-cell imaging and fluorescent in situ hybridization to show that deltaD −/− embryos show no posterior expression of her1 in the 12-somite stage PSM (Mara et al., 2007).As a result, it may be difficult to directly associate ligand similarities between different species based on mutant phenotypes alone.Nonetheless, these studies suggest a more "synchronization" role for DeltaC such that its loss results in a random, salt-and-pepper expression, but a more "driver" role for DeltaD such that its loss results in a complete loss of cyclic gene expression (Holley et al., 2002).
Irrespective of the ligand or receptor function lost, at least a few somite cycles must elapse before the blockade of Notch signaling and severe segmentation defects.This can happen if Notch signaling regulates synchronization primarily, but not the clock directly.Alternatively, it is also possible that Notch signaling in fish is important only for the posterior clock, but not the anterior one, where a different signaling pathway is a primary driver.In fact, a recent study showed that her1 −/− ;her7 −/− fish can still show a proper segmentation profile if provided with exogenous pulses of FGF/Erk signaling inhibition between the 17-22 somite stage (Simsek et al., 2023).These results are complemented by the observation that deltaD;deltaC single and double knockout fish embryos show normal axial development, and are viable and fertile (Jülich et al., 2005), as opposed to ligand knockout mice which are embryonic lethal (Figure 7, Lower panel).

The levels of Notch signaling affect the segmentation clock properties
An increase in Notch signaling levels slows down the mouse clock While the presence of Notch signaling is necessary for the integrity of the mouse clock, the level of Notch signaling itself can also regulate the clock by changing its properties (Figure 7, Upper panel).For example, mice with a LOF mutation in the Notch regulated ankyrin repeat protein (Nrarp), a cyclic Notch target that represses NICD cleavage in the mouse PSM (Sewell et al., 2009), form ~2 fewer vertebrae compared to WT mice, and ~50% of homozygous mutant mice show a kinked tail (Kim et al., 2011).In line with the function of Nrarp, Nrarp −/− E10.5 PSM show ~1.9fold higher NICD levels.Furthermore, when WT pregnant mice were treated with LY-411575 (0.1 mg/kg; a γ-secretase inhibitor that inhibits Notch signaling) every 24 h starting at E7.5, NICD levels in E10.5 embryos dropped by 10%-20%, resulting in a slightly faster clock in ~50% embryos that formed 1 extra somite compared to the untreated controls.Remarkably, the same dose of LY-411575 treatment to mice carrying Nrarp −/− embryos also resulted in a rescue in ~50% of embryos forming one additional somite compared to the untreated mutants.This suggests that the loss of Nrarp results in an increase in NICD levels, in turn increasing the period.Interestingly however, the cyclic expression of clock genes like Hes7, Lfng, and that of NICD were unaffected in these mutant embryos.As a result, Nrarp may also be a more peripheral component, consistent with studies showing that the loss of Dll3 results in the complete loss of Nrarp expression (Sewell et al., 2009;Bochter et al., 2022).One plausible explanation for the increased period in Nrarp −/− embryos is that higher levels of Lfng may be necessary to build up to effectively repress the higher amount of NICD, which may take significantly more time, hence slowing the clock.However, increasing the half-life of NICD also slowed the clock (Wiedermann et al., 2015).This indicates that not just the levels of NICD, but the time spent by NICD in the cell also affects clock periodicity, suggesting a Lfng-independent mechanism.

An increase in Notch signaling levels speeds up the fish clock
The fish clock is also sensitive to the levels of Notch signaling, however in the direction opposite to that of mice.Transgenic zebrafish embryos containing multiple copies of deltaD produced more somites and vertebrae compared to WT embryos (Figure 7, Upper panel; Liao et al., 2016).Interestingly, the authors generate two lines, Dover with ~10-fold higher DeltaD levels in the homozygotes (copy number ~15), and Damascus with ~50-fold higher DeltaD in the heterozygotes (copy number ~150).While both the homozygous and heterozygous Dover mutants showed normal somitogenesis, none of the homozygous Damascus mutants survive.Surprisingly, the heterozygous Damascus mutants develop normally, and form an overall 2 additional trunk somites compared to WT siblings, with clock periods being ~1.5 min faster specifically in the trunk region (4-19 somite stage).In contrast, neurogenesis was affected even with the slightest increase in DeltaD levels, suggesting the robustness of somitogenesis to tolerate gene expression changes compared to other lineages.
On the other hand, DAPT treatment (another γ-secretase inhibitor that inhibits Notch signaling) or homozygous aei/des embryos show a slower segmentation clock at the 3-4 somite stage (heterozygotes do not, even though protein levels are halved (Liao et al., 2016)), resulting in larger anterior somites before the onset of segmental defects (Herrgen et al., 2010).In the presence of DAPT, Damascus mutants showed no change in the clock period (Liao et al., 2016).This suggests that while no deltaD copies (aei embryos) confer segmental defects, moderately abundant copies do not affect somitogenesis, or even embryonic development.This could be because deltaD does not oscillate in the fish PSM, and as a result, too much of it does not disrupt its endogenous expression dynamics.Overexpression of Dll1 in mouse embryos can disrupt Dll1 oscillations and thereby somitogenesis (Shimojo et al., 2016).As a result, it is possible that the introduction of multiple copies of deltaC disrupts somitogenesis.
These results reinforce that while in mice, Notch signaling primarily drives the expression of oscillatory genes, in zebrafish, Notch signaling only synchronizes and propagates her1/7 Species-specific differences in clock dynamics and Notch mutant phenotypes.Elevated levels of the Notch signaling pathway opposingly affect the mouse and zebrafish clock periods, and as a result, the final number of somites/vertebrae formed (Upper panel).Species-specific differences in the requirement of Delta ligands during embryonic development.Mouse-red; zebrafish-blue (Lower panel).. oscillations.This is supported by the 'delayed coupling theory' where reduced Notch signaling results in weaker coupling strength and higher delays in information transfer between neighboring PSM cells, thereby resulting in a collective slower clock (Morelli et al., 2009;Herrgen et al., 2010;Liao et al., 2016;Yoshioka-Kobayashi et al., 2020).Due to the lack of genetic tools, similar Notch signaling manipulative studies have not been reported yet in chick embryos.It would be interesting to characterize the effects of in ovo electroporation of c-delta1 or c-notch1/c-NICD cDNA on the chick segmentation clock time.

Notch signaling and cell-cell synchronization
Lfng acts as a synchronizer in the mouse PSM The role of Notch signaling as a synchronizer in mouse embryos has evolved over many decades, since initially Notch signaling was only considered to be essential to maintain gene oscillations in individual PSM cells, based on Notch ligand and receptor LOF mutant phenotypes.Two studies showed that when mouse E10.5 or chick PSM tissue (14-18 somite stage) is dissected into multiple fragments, they are each able to oscillate maintaining the correct clock schedule compared to an intact tissue (Maroto et al., 2005;Masamizu et al., 2006).This showed that at least at the global level, groups of cells maintained synchrony despite being cultured separately.With the advent of single-cell resolution imaging, the role of Notch signaling as a cell synchronizer alongside being a 'switch' to maintain oscillations became easier to study.Masamizu et al., 2006 showed first that dissociated E9.5-10.5 transgenic mouse tailbud cells showed asynchronous and unstable Hes1 oscillations in vitro (Masamizu et al., 2006).Another study culturing dissociated E9.5 transgenic mouse tail bud explants showed similar dampened and weak Lfng oscillations (Hubaud et al., 2017).A similar experiment in 14-18 somite stage chick embryos showed that when the PSM was dissociated and cultured in vitro in separate pools, c-Lfng oscillated asynchronously across pools as assayed by in situ hybridization, suggesting that cell-cell communication is necessary to maintain stable, synchronous oscillations (Maroto et al., 2005).However, was this property Notch-dependent?Using an in vitro culture technique of mouse E10.5 tail bud explants, Tsiairis and Aulehla, (2016) showed that DAPT treatment affected the ability of dissociated and reassembled PSM cells to self-organize into spatiotemporal wave patterns upon reaggregation in vitro (Lauschke et al., 2012;Tsiairis and Aulehla, 2016).
A major obstacle in understanding the role of Notch as a synchronizer in mouse segmentation stems from the fact that Notch signaling directly controls Hes7 or Lfng oscillations throughout the PSM.As a result, the dissociation of the role of Notch as a driver of oscillations and as a synchronizer between oscillators is a challenge.The dampening and instability of Hes7 and Lfng oscillations in Notch pathway LOF mutants or DAPT-treated embryos may be because of reduced expression of these genes rather than due to a loss of synchrony.To combat this, one study used chimeric mouse embryos consisting of WT and Dll1 −/− cells and showed that the synchronization rates in the E10.5 PSM reduced linearly with decreasing number of WT cells in the chimeras (Okubo et al., 2012).They also used chimeric mice containing WT and Lfng −/− cells to show that Lfng could regulate Hes7 oscillations noncell autonomously.Another study also used immortalized cell lines to show that Lfng can repress Dll1 (alongside NICD) cellautonomously to regulate levels of Notch signaling cellautonomously (via NICD) and non-autonomously (via Dll1) (Okubo et al., 2012;Bochter et al., 2022).These data suggest a unique role of Lfng in being responsible for maintaining synchronization during somitogenesis (Figure 6A).A recent study supports the these results through single-cell tracking of Hes7-positive and negative cells in Lfng −/− PSM that showed individual oscillators lose synchrony (Yoshioka-Kobayashi et al., 2020).

Her1/7 directly repress deltaC in zebrafish
In zebrafish Notch-LOF embryos, the first few somites develop normally and the progressive worsening of somitogenesis has been attributed to a loss of synchronization between individual oscillators.For example, in a 'resynchronization assay', zebrafish embryos treated with DAPT that start to form disfigured somites after about 5 somites, promptly re-form normal somites after DAPT withdrawal due to the "resynchronization" of the clock (Liao et al., 2016;Uriu et al., 2021).As the tissue recovers following DAPT withdrawal, the defective and normal somites intermingle transiently showing that recovery occurs at multiple levels: At the cellular level with rapid resynchronization between individual cells via Delta-Notch signaling; and the tissue level with the transfer of this resynchronization "information" throughout the PSM (Uriu et al., 2021).deltaD/aei or notch1/des heterozygous mutants fail to resynchronize effectively after DAPT treatment, whereas deltaC/bea mutants recover similarly to WT embryos, suggesting ligand-specific roles in carrying out synchronization (Mara et al., 2007).In contrast, embryos with multiple deltaD copies can resynchronize faster than their WT siblings (Liao et al., 2016).
Similar to mice, advances in single-cell resolution imaging of Her1 oscillations in vivo showed that individual cells oscillate autonomously, which begin cycling out of phase in deltaD (aie), deltaC (beamter), and notch1a (des) embryos (Delaune et al., 2012).A more recent study corroborated this by culturing the caudal PSM cells in vitro at low densities and detecting cell-autonomous, but asynchronous and noisy her1 oscillations (Webb et al., 2016).This high background noise was absent in cells belonging to an intact PSM, suggesting a role for collective tissue-level processes in increasing the precision of the clock.
Interestingly, the suggested mechanism of synchronization in zebrafish is slightly different from mice (Figure 6A).The Her1/ 7 dimer is predicted to directly repress deltaC expression levels (Horikawa et al., 2006).One study showed that the deltaC expression pattern in her1-LOF fish is only altered in the anterior PSM but not posterior, suggesting that this predicted mechanism of synchronization differs along the embryonic axis (Choorapoikayil et al., 2012).However, this remains to be systematically characterized.

Insights from non-classical model systems and future perspectives
Much of the early knowledge about body segmentation before studies in chick embryos came from arthropods like Drosophila (Davis and Patel, 1999;Pourquie, 2003).However, unlike vertebrates, genetic studies have shown that Notch signaling does not have a link to fly segmentation (Johnston and Nüsslein-Volhard, 1992).Surprisingly, homologs of the genes delta, notch, and hairy all show a segmental expression pattern during early development of another arthropod, the spider Cupiennius salei (Stollewerk et al., 2003).RNAi-mediated knockdown of delta and notch resulted in disorganized hairy segmentation, similar to vertebrates.Similar segmented expression of Notch components and their roles in embryonic segmentation have been observed in other invertebrates like the cockroach Periplanata americana and the centipede Strigamia maritima, but not in the cricket Gryllus bimaculatus (Pueyo et al., 2008;Reviewed extensively in;Kainz et al., 2011;Brena and Akam, 2013;Clark et al., 2019).This suggests that a wider sampling of invertebrate species beyond classical models like Drosophila can reveal otherwise unidentified complex ancestral segmentation mechanisms.As a result, it is essential to expand our horizon on this developmental process by using nonclassical in vivo and in vitro model systems, including other nonmodel vertebrates and potentially even invertebrate species to better understand this process.

Reptiles
Corn snake embryos form 200-300 somites with a timing of ~100 min per somite (Table 4; Gomez et al., 2008).Interestingly, the house snake which has very similar developmental characteristics to the corn snake, displays a clock of ~60 min suggesting that small genetic differences within a given genus can drastically affect clock timing (Gomez et al., 2008).This correlation is challenging to verify when only using classical models like mouse and fish, which inherently are very different.Reptiles like the green anole lizard and alligator have also been used to study somitogenesis which unlike traditional model species, do not express Lfng in the PSM.While the corn snake does, it exhibits significantly more stripes in the PSM compared to mice, suggesting a different mode of clock gene regulation (Gomez et al., 2008;Eckalbar et al., 2012).In contrast, similar to classical models, Dll1 displays oscillatory expression in the anole lizard (Eckalbar et al., 2012).Another example is Dll3, while expressed uniformly throughout the mouse PSM, is confined to the caudal-most PSM and displays static anterior bands in the anole lizard.These observations suggest possibly conserved and divergent mechanisms regulating the core versus peripheral clock players respectively in the reptilian segmentation clock.

Other teleosts
Studies on other teleost fish species like Oryzias latipes (Medaka) and Tagifuku rubripes (Japanese pufferfish) also show some interesting clock divergence from zebrafish (Danio rerio) (Elmasri et al., 2004;Gajewski et al., 2006).Medaka embryos show a clock timing of ~60 min at 26 °C, whereas zebrafish embryos tick approximately two-fold faster at 25-30 min at 26 °C suggesting similar intra-species differences like snakes (Table 4; Elmasri et al., 2004;Schröter et al., 2008).In fact, 60 min is the clock speed at the posterior PSM which slows down by almost half to ~120 min in the anterior PSM of a 3-5 somite-stage medaka embryo (Vibe, 2021).Given that the anterior speed of the medaka embryo (unlike zebrafish embryos) is very similar to the posterior PSM clock in mice, understanding why a period gradient arises in medaka could help decode the underlying mechanisms of species-specific clock times.Genetic knockout phenotypes have also shown that the Medaka her7 is functionally similar to the zebrafish her1 and her7 combined, and the Medaka her1/11 to the zebrafish her11 over her1 (Elmasri et al., 2004;Sieger et al., 2006).These phenotypes further support the idea that teleost embryos use her7 as a core clock driver and her1/11 genes as more peripheral ones.
Another advantage that Medaka has for decoding the genetics underlying species-specific clock times is that different inbred Medaka sub-species show different her7 clock periods (Seleit et al., 2023).They also have variable PSM and somite lengths, which positively correlate with clock periodicity.Crossbreeding species with different periods provides a useful approach to understand the genetic basis of clock periodicity.Accordingly, intercrossing these different Medaka species results in a wide range of meiotic recombination across the genome, yielding F2 progeny showing a wide range of clock periods (Seleit et al., 2023).The authors further show that while clock period scales with PSM size in parental fishes, these two parameters do not correlate in F2 progeny.On the other hand, somite size still scales with PSM size.This raises the possibility that there is a spectrum of somite numbers formed in the F2 embryos.However, this has not been shown.In addition, whole genome sequencing of F2 embryos revealed that the Medaka dll1 is responsible for controlling PSM size, but not somite number/border formation or clock period.This result is in contrast to Dll1 −/− mice where no somite borders are formed, indicating a potential evolutionarily divergent role of Notch ligands.
One way to combat the inability to perform inter-species crosses in mammals is to carry out in vivo or in vitro gene-swapping experiments.One study showed that the species-specific oscillatory period of Hes7 is not entirely regulated by local genomic sequences (Matsuda et al., 2020a).In particular, the human segmentation clock ticks with a periodicity of ~5 h (Chu et al., 2019;Matsuda et al., 2020b;Diaz-Cuadros et al., 2020) and when the mouse Hes7 locus was swapped with the human HES7 in vivo, it results in only an ~20-30 min extension of the mouse period when measured in E10.5 mouse PSM explants (Matsuda et al., 2020a).This genomic manipulation caused minor, but significant segmentation defects in the form of a curved spine and a kinked short tail.Possibly, mutant embryos containing the human HES7 that showed a change in the clock time of >20-30 min may have been embryonic lethal, thereby only the embryos with the minimal clock perturbation survive.Furthermore, because this gene swapping was limited to a BAC construct extending only up to the 3′ UTR of the neighboring gene Per1, one cannot exclude the possibility of cis-elements downstream of Hes7 affecting its periodicity.While there are only limited reports on how the segmentation clock and the circadian clock could interact genetically (Umair et al., 2022), the PER1 gene also displays an inphase gene oscillation phenotype with HES7 in human PSM cells in vitro (Matsuda et al., 2020b).

In vitro 2D and 3D systems
Aligning with the ideology of the importance of comparative developmental biology, recent studies have established embryonic stem cells or induced pluripotent stem cells (iPSCs) derived from various species of different sizes and gestational lengths such as marmosets, rhinoceros, cattle, rabbits (Lázaro et al., 2023) and pigs (Conrad et al., 2023) (Table 4).These species-specific PSCs display intrinsic developmental timing phenotypes when differentiated to a PSM and other lineages (Matsuda et al., 2020a;Rayon et al., 2020;Lázaro et al., 2023).The in vitro clock period observed across these different species also correlates with in vivo gestation times, body size, and biochemical reaction rates in vitro (Matsuda et al., 2020a;Conrad et al., 2023;Diaz-Cuadros et al., 2023;Lázaro et al., 2023).Therefore, carefully curated in vivo embryonic staging data can help determine how closely these in vitro systems can mimic somitogenesis.In addition, recent studies have showcased the power of both 2D and 3D mouse and human in vitro models in recapitulating hallmarks of somite formation, thereby giving us insights, especially into human development (Diaz-Cuadros and Pourquie, 2021;van den Brink et al., 2020;Budjan et al., 2022;Chu et al., 2019;Diaz-Cuadros et al., 2020;Matsuda et al., 2020b;Miao et al., 2022;Sanaki-Matsumiya et al., 2022;Veenvliet et al., 2020;Yamanaka et al., 2022).Multiple studies have shown human pluripotent stem cells (PSCs)-derived PSM cells capable of undergoing 3D morphological somite-like budding that can further our knowledge of how temporal information translates to spatial information in our species (Miao et al., 2022;Sanaki-Matsumiya et al., 2022;Yamanaka et al., 2022).

Conclusion
In this review, we summarized existing knowledge on the segmentation clock biology by leveraging the similarities and differences of the mouse segmentation clock with that of zebrafish and chick.Through insightful comparisons with non-classical model organisms and novel in vitro models, we illuminate the profound importance of these systems in advancing our understanding of the segmentation clock and its evolution.In the future, we anticipate that novel in vivo and in vitro approaches will provide complimentary knowledge to better understand vertebrate segmentation and embryonic development.

TABLE 1
The number and types of Delta ligands and Notch receptors vary across species.

TABLE 2 PSM
-specific Notch signaling-related Hes/her/Hey genes in mouse, zebrafish, and chick.

TABLE 3
Mice knockout phenotypes of PSM-specific Hes/her/Hey targets, Notch ligands, and receptors.

TABLE 4
Species-specific segmentation clock periods across various in vivo and in vitro model systems.