Imaging and manipulating the segmentation clock

Abstract

During embryogenesis, the processes that control how cells differentiate and interact to form particular tissues and organs with precise timing and shape are of fundamental importance. One prominent example of such processes is vertebrate somitogenesis, which is governed by a molecular oscillator called the segmentation clock. The segmentation clock system is initiated in the presomitic mesoderm in which a set of genes and signaling pathways exhibit coordinated spatiotemporal dynamics to establish regularly spaced boundaries along the body axis; these boundaries provide a blueprint for the development of segment-like structures such as spines and skeletal muscles. The highly complex and dynamic nature of this in vivo event and the design principles and their regulation in both normal and abnormal embryogenesis are not fully understood. Recently, live-imaging has been used to quantitatively analyze the dynamics of selected components of the circuit, particularly in combination with well-designed experiments to perturb the system. Here, we review recent progress from studies using live imaging and manipulation, including attempts to recapitulate the segmentation clock in vitro. In combination with mathematical modeling, these techniques have become essential for disclosing novel aspects of the clock.

Introduction

The animal body is formed by a series of repetitive structures along the head-to-tail axis, such as the vertebrate spine and ribs. In vertebrates, these “segment-like” structures are derived from somites, an embryonic tissue. Somites differentiate from the presomitic mesoderm (PSM), a proliferative progenitor population located at the caudal end of the embryo during the mid-gestation period. During this differentiation process, a group of cells at the anterior end of the PSM are “pinched off” and differentiate into epithelialized somites, which are distributed on both sides of the neural tube. The development of these epithelialized somites occurs sequentially from head to tail, with regular intervals and periodicity. The periodicity is species-specific: 2 h in mice and 4–6 h in humans [1]. The mechanism underlying this spatiotemporally coordinated developmental process is now recognized as a segmentation clock, which consists of a set of cell signaling and gene regulatory networks (GRNs). The existence of this clock, which regulates the precision of this periodic event, was first postulated in a theoretical framework as the “clock and wavefront” model [2]. This model considered two interacting activities to act in the PSM: a clock (an oscillator that keeps the tempo) and a wavefront (a posteriorly moving front of rapid cell differentiation) without specifying molecular or cellular details. Several alternative models have also been proposed, including the positional information model, in which PSM cells transition between distinct states under the influence of morphogen gradients [3], and the cell cycle model, in which synchronous cell cycle is linked to the segmentation clock [4]. In 1997, expression of the hairy1 gene, a Hes/Her homologue, was found to oscillate in chicken PSM with a period of 90 min, which exactly corresponded to the rhythm of somitogenesis in chicken embryos, revealing the molecular evidence of the oscillatory clock [5]. Subsequently, a number of oscillatory genes and signaling pathways were identified in various model organisms, such as zebrafish, chicken, and mouse. Notably, such genes mainly act downstream of Notch, Fgf, and Wnt signaling pathways [6–9]. The Hes/Her gene family, which acts downstream of Notch signaling, shows cyclic expression patterns. Expression of Hes/Her genes is first observed at the posterior end of the PSM, then propagates through the tissue towards the anterior end, before gradually slowing and finally halting at the prospective somite position. This traveling wave pattern repeats every oscillation cycle as the embryo elongates posteriorly (Fig. 1). In contrast, Fgf and Wnt signaling form posterior-to-anterior gradients in the PSM (Fig. 1), in combination with antagonistic RA activity, and set the position of transition from immature PSM to somites, representing the molecular detail of wavefront [10–12]. Genes involved in Fgf and Wnt signaling also oscillate, but the phase relationship of these oscillators with Notch signaling oscillators changes around the wavefront, suggesting that the phase shift between these oscillators is the underlying mechanism of the clock–wavefront interactions [13, 14].

Fig. 1.

Schematic view of the vertebrate segmentation clock. Upper panel: Expression dynamics of Hes/Her genes and the Fgf/Wnt gradient in the PSM. Somites are periodically formed on both sides of the neural tube. Lower left panel: Hes/Her genes show oscillatory expression within the PSM. Lower right panel: the oscillation of Hes/Her genes is driven by the cell-autonomous negative feedback loop and is coupled to cell–cell interactions via Delta/Notch signaling

The Hes/Her family genes encode a bHLH transcription factor that can bind to their own promoter region and repress expression. This negative feedback loop drives the oscillation. In zebrafish, Her1 and Her7 genes show oscillatory expression with a period of 30 min in the PSM; the expression levels of these genes depend on Notch signaling input (Fig. 1), as revealed by mutants for DeltaC and DeltaD, Notch signaling ligands expressed in the PSM, and by chemical inhibition of Notch by the gamma-secretase inhibitor DAPT. In the mouse, Hes7 exhibits cyclic expression with a 2 h periodicity that is under the control of Notch and Fgf signaling [15]. Although each PSM cell harbors an autonomous oscillator, the oscillation phase is affected by biological “noise” and also shows some phase drift. Fluctuations in Hes/Her oscillation timing in the PSM have been observed; however, the clocks are mostly synchronized between neighboring cells to generate coordinated oscillation at the tissue level, indicating the existence of a phase-coupling mechanism. This synchronized oscillation is important for spatiotemporal precision when cells differentiate into somites by establishing sharp boundaries. Notch signaling has been demonstrated to contribute to this synchronization process as loss of these signal molecules results in segmentation defects [16]. In zebrafish, oscillatory DeltaC and static DeltaD collaborate to activate Notch signaling. Disruption of Notch signaling by mutation of Delta genes or by chemical inhibition leads to defects in somite formation [16, 17]. In both mutant and chemically treated embryos, Her1 expression is observed as a salt-and-pepper pattern instead of the normal striped pattern present in wild-type (WT) embryos. This phenotypic change can be interpreted as a gradual desynchronization of the oscillators during development, due to phase drift in the absence of Notch signaling [16]. However, the lack of direct observation or quantification of clock dynamics in single PSM cells precluded reaching a clear conclusion. The salt-and-pepper pattern could also be produced if the amplitude of oscillation or net expression level is affected. In the mouse, complete loss of Notch signaling disrupts Hes7 clock oscillation because the promoter activity of Hes7 is driven by Notch signaling. This has hampered investigation of the exact role of Notch signaling in the mouse. Furthermore, in the mouse and chicken, Lunatic fringe (Lfng), a glycosyltransferase that modulates Notch signaling, shows cyclic expression in the PSM downstream of Hes7, adding another level of complexity into the system in these species. A complete understanding of the mechanism that underlies this major developmental event will only be achieved by observing and quantifying the dynamic signals and expression of clock-related factors. In this review, we discuss experimental strategies for uncovering the mechanism of the segmentation clock and focus on live imaging and other manipulation strategies, and on the use of systematic analyses with mathematical simulations.

Visualizing the expression dynamics of clock genes using bioluminescent reporters

After the discovery of oscillatory expression of hairy1 in the chicken, several other genes were subsequently demonstrated to oscillate in the PSM, including orthologues of Hes/Her genes such as mouse Hes1 and Hes7 and zebrafish Her1 and Her7 [6–9]. However, despite considerable attempts to understand the nature of the clock, dissection of the mechanism that regulates its dynamics was limited because the studies were based on the observation of gene expression patterns using static analyses such as in situ hybridization and immunohistochemistry. The expression dynamics of clock genes were inferred by aligning dozens of stained tissue samples, because oscillation of Hes/Her genes in the cells within the PSM proceeds in a synchronous manner. These approaches are still valuable for studying the effects of gene ablation or chemical inhibitor treatments that affect overall gene expression patterns at the tissue level; however, they are limited in that it is not easy to obtain quantitative assessments of the phenotype, especially when phase coupling or amplitude is affected. In addition, analysis of oscillations within single cells is impossible because of the invasive aspect of the staining process. This technical limitation can now be surmounted due to the development of technologies that allow quantitative analyses of the dynamics of clock genes in real time. For example, we observed oscillation of Hes1 expression in the mouse PSM using an expression reporter based on bioluminescence [18]. In this system, firefly luciferase, which had been destabilized by the conjugation with degradation-tolerant mutated ubiquitin, was expressed under the control of the Hes1 promoter. Although bioluminescence-based imaging had been already developed for research into circadian oscillations, faster degradation kinetics in the reporter protein were necessary to monitor the ultradian dynamics of the segmentation clock. Time-lapse imaging of the reporter revealed robust and stable oscillation of Hes1 in the PSM, corresponding to the previously reported expression pattern and period. Notably, the biggest advance in using the reporter system is the measurement of Hes1 activity in a single-cell dissociation culture. We found that in contrast to the robustness of coherent oscillation in PSM tissue, dissociated PSM cells showed unstable oscillation, suggesting that cell–cell communication plays an essential role in synchronization and maintenance of stable oscillation [18]. This methodology was further extended to produce a Hes7 reporter system (Fig. 2, upper panel) [19], which we will discuss in a later section.

Fig. 2.

Comparison of mouse Hes7 bioluminescent and fluorescent reporters. Upper: Hes7 bioluminescent reporter [19]. Spatio-temporal expression profile of the luminescence, obtained with a cooled CCD camera attached to a wide-field microscope [19]. Lower: Hes7 fluorescent reporter [23]. Spatio-temporal expression profile of fluorescence obtained with a confocal laser-scanning microscope [23]. In both profiles, the x axis indicates the time, while the y axis indicates the anterior (up)-posterior (down) axis of the PSM

There are several pros and cons in using bioluminescent reporters. With respect to the advantages, first, the fast kinetics in production of a functional luciferase protein in the cells allows monitoring of fast-changing gene expression dynamics. Second, there is virtually no background signal in the samples and, because of this high sensitivity, increasing the exposure time works well for increasing the signal captured by the detector. Third, bioluminescence has a very broad linear dynamic range, as has been demonstrated in many in vitro assay systems, and therefore, it is suitable for quantification analysis. However, the detection of weak signals from tissues requires the use of specialized detectors and microscopes with high sensitivity, which are dedicated to this type of imaging analysis. Furthermore, it is difficult to monitor bioluminescence at the single-cell resolution in tissues.

Visualizing the expression dynamics of clock genes using fluorescent reporters

As an alternative to bioluminescent reporters, Aulehla et al. utilized a fluorescent protein for monitoring Lfng expression in the mouse PSM [12]; the yellow fluorescent protein Venus fused with a PEST domain (a degradation signal sequence) was driven by cyclic Lfng promoter. Using this reporter (designated as LuVeLu), they compared the oscillation dynamics of Lfng in the PSM of WT mice and mutant mice carrying a gain-of-function allele of beta-catenin, a key mediator of canonical Wnt pathway. Time-lapse imaging demonstrated that although clock oscillation continues despite the increased level of beta-catenin, the signal intensity and oscillation amplitude of the reporter are reduced in the mutant compared to WT. Importantly, detecting and quantifying the differences in oscillation amplitudes could not be easily performed using fixed and stained samples, highlighting the advantage of live imaging. Although the study revealed that it is feasible to use a fluorescent protein to analyze the oscillatory dynamics of endogenous gene expression, the spatial resolution of the obtained images was limited, probably reflecting the difficulties in collecting signals of the fluorescent reporter from live elongating PSM tissue.

A Venus-based reporter has also been exploited in zebrafish to capture clock dynamics, but rather than designing reporters that monitor the promoter activities of genes, Delaune et al. [20] generated a fusion protein reporter. In the study, they established zebrafish transgenic line carrying Her1-Venus protein reporter. Transgenic zebrafish carrying a Her1-Venus protein reporter in which Venus was fused to the C-terminus of the Her1 protein and expressed under the control of the Her1 promoter. As Her1 is a transcription factor, expression of the fusion reporter was confined to the nucleus. One of the key observations in the study was oscillation of the clock gene at the single-cell level, within live tissues. Single-cell quantification is highly important for comparison of oscillation between neighbors in a normal synchronous regime as well as in the abnormal asynchronous regime found in Notch signaling-related mutants. By use of a combination of nuclear and membrane-localized reporters, each PSM cell could be tracked and Her1 dynamics quantified over time; the phase dynamics could then be analyzed by fitting the time-series. In this way, use of a fluorescent reporter enhanced the spatio-temporal resolution of the signal detection in live tissues, enabling them to directly assess the level of synchronicity of the cells. Delaune et al. observed that Her1 oscillation was desynchronized in zebrafish with mutations of beamter (bea/detaC), deadly seven (des/notch1a), or after eight (aei/deltaD), directly confirming a long-standing speculation regarding the role of Notch signaling in phase coupling [16, 20]. Furthermore, cell tracking with the Her1 reporter revealed a novel mechanism in the segmentation clock despite major sources of biological noise, such as cell division. Mitosis was found to preferentially occur during the off phase of Her1 oscillation in PSM cells, suggesting that the effect of noise on coherent oscillation is likely reduced to some extent. Interestingly, daughter cells after cell division showed highly synchronous oscillations in both wild-type and Notch-related mutants, suggesting that clock factors might have been passed equally to the new cells at mitosis and that these factors were sufficient to maintain synchronicity for a few oscillation cycles even in the absence of Notch signaling. Importantly, Delaune et al. also reported analysis of other reporter lines that used different strategies, such as fusion of the Venus protein to the N-terminus of Her1 [20]. In the latter case, Venus expression was not oscillatory but persisted even after formation of somites, emphasizing the importance of the proper design and validation of reporter cassettes and the possibility of inappropriate result interpretation when reporter expression does not reflect endogenous dynamics.

Using another original reporter designated as Looping, Soroldoni et al. [21] linked the segmentation clock with the well-known physical phenomenon, the Doppler effect, in which the pitch of a moving sound source alters depending on whether it is approaching or moving away from an observer. The reporter used in the study was based on a BAC transgene in which the intact chromosomal locus for Her1 is restored with an insertion of Venus attached to the C-terminus of Her1. In general, reporter lines producing a functional endogenous protein can display overexpression if the construct is introduced as multiple copies. It is, therefore, necessary to minimize the gene copy number when constructing transgenic lines. Soroldoni et al. used I-SceI meganuclease to ensure low-copy number insertion into the genome during transgenesis [21]. Using their reporter line, they observed Her1 oscillation in the PSM of developing zebrafish embryos. Previously, it was thought that the oscillation period of the clock gene corresponded to the period of actual segmentation. However, Soroldoni et al. observed that the period of segmentation at the anterior PSM is actually faster than the period of oscillation in the posterior PSM [21]. Systematic measurements of reporter expression throughout the PSM for several rounds of oscillation indicated that when the PSM moves posteriorly, the anterior end of the PSM meets approaching waves of Her1, thus increasing the frequency of waves at the anterior front. For this reason, they likened this phenomenon to a Doppler effect.

To date, live imaging-based studies have been conducted mainly in the zebrafish and mouse, animal models with well-established genetic tools. Overall, the zebrafish embryo is an ideal model for in vivo fluorescence microscopy, because the embryos, which develop within eggs, are small and transparent. The relatively rapid development of zebrafish embryos allows observation of almost the entire process of somitogenesis. In contrast, imaging mammalian embryos, which normally develop in utero, is considerably more difficult, because optimized culture conditions are necessary; additionally, there is also the problem of light scattering, which may originate from the embryo itself and from the culture medium [22]. Mouse embryos are photosensitive and require fine-tuning of parameter settings during illumination and image acquisition, especially when combined with laser-scanning microscopy. Generally, the expression reporters for oscillatory clock genes do not provide a bright signal because of their short half-life, which is critical for mimicking the rapid kinetics of an endogenous gene/protein; this lack of signal brightness makes it difficult to improve spatio-temporal resolution. Although zebrafish Her1 reporters have been used for quantifying clock oscillation at the single-cell level in the context of synchronization, the GRN architecture of the segmentation clock in zebrafish and other higher organisms such as the mouse are different as described in the introduction. Thus, the development of an imaging system with a bright clock gene reporter to allow quantitative analysis of clock oscillation at the single-cell level in the mouse is still required; such a system might provide novel insights into the mammalian segmentation clock in the context of synchronization.

To achieve this goal, we established a single-cell live imaging system for mouse PSM tissue using an Hes7 reporter in combination with a newly developed fluorescent protein (Fig. 2, lower panel) [23]. In this study, we sought to improve the brightness of the Venus-based reporter for Hes7 by accelerating the maturation process, which is a bottleneck of fluorescent protein reporters for reproducing the endogenous dynamics. Eight new point mutations were introduced into Venus, resulting in a new fluorescent protein designated as Achilles that had a faster maturation rate. A fusion protein comprised of Achilles attached to the N-terminus of the Hes7 protein was used to generate a novel reporter transgenic line named Hes7-Achilles. This reporter line was found to provide brighter signals than with Venus, and we were able to demonstrate single-cell quantification of Hes7 oscillation in live intact mouse PSM (Fig. 2, lower panel). In the mouse segmentation clock, Lfng has been suggested to be involved in the phase-coupling process [24]. Although a previous investigation using a luciferase-based Hes7 reporter showed that the amplitude of Hes7 oscillation is lower in Lfng-null PSM than in the WT [13], it was unclear whether this resulted from desynchronization or attenuation of oscillation in the absence of Lfng. The improved spatio-temporal resolution obtained by the Hes7 reporter based on fluorescence enabled quantification of the instantaneous oscillation phase in each PSM cell as well as the degree of synchronization. We found that in the absence of Lfng, Hes7 oscillation became both desynchronized and attenuated at the single-cell level [23]. This phenotype cannot be explained by changes in the intrinsic oscillation (demonstrated in dissociation cultures where cell–cell communication is absent), indicating that Lfng is likely to have a role in controlling cell–cell coupling. Thanks to the improved culture conditions, as described in the next section, mixed cultures of dissociated WT and Lfng-KO PSM cells carrying the Hes7 reporter were performed, showing that the KO cells were not able to couple to the WT cells. The results of a mathematical simulation and an optogenetic sender/receiver assay of Notch signaling in the absence/presence of Lfng, indicated that Lfng controls the synchronization of Hes7 oscillation within the PSM by generating a delay in the signal-sending process of Notch signaling. Notably, establishment of single-cell analysis using a bright fluorescent reporter accelerated accurate and comprehensive quantification and expanded the range of experimental designs for assays involving quantification of oscillation within cell populations in intact tissues, single-cell isolation cultures, and mixed cultures of different genotypes [23]. In addition, we also found that in a C2C12 cell line culture model, Lfng delayed expression of Dll1 on the cell surface [23]. Although it remains unclear whether this finding also applies to actual PSM tissue, fluorescent reporters that enable visualization of Notch signaling components with fine spatio-temporal resolution might answer this question [25]. This study highlighted the importance of controlling the time delay in the cell–cell coupling process, confirming previous speculations from a mathematical model and an in vivo experiment [26, 27].

In vitro culture of primary PSM cells

To decipher the segmentation clockwork at the cellular level, it is important to establish a good culture condition for PSM cells. Lauschke et al. used the LuVeLu reporter to demonstrate scaling of the phase gradient in mouse PSM in a novel two-dimensional explant culture model, called quasi-monolayer PSM or mPSM [28]. The culture was derived from tail bud mesoderm at embryonic day 10.5, a progenitor cell population located at the posterior tip of the tail. Although careful characterization of the differences of this culture from intact culture is needed, the great benefit of this 2D model is that it facilitates very precise measurement of reporter expression and is suitable for long-term observation. The culture condition method provides stable and reproducible results, because the major problems seen in 3D imaging, such as tissue drift and scattering of light, are negligible. In addition, it allows highly sensitive signal detection with the use of high magnification lenses that have a short working distance. The 2D model was found to show LuVeLu oscillation with a period equivalent to intact PSM, a concentric traveling wave sweeping in a central-to-peripheral direction, and regression of LuVeLu activity upon boundary formation. The authors also investigated the mechanism behind segment scaling, a progressive decrease in segment size proportional to the decrease in PSM size that occurs in both intact PSM and mPSM. They found that the velocity of the traveling wave and the phase difference between mPSM cells were scaled to the length of the mPSM and segment size, suggesting that phase gradient is a crucial determinant of segment size. This study also points to an advantage of mPSM culture model over intact PSM for manipulating clock behavior by perturbing the dynamics in a systematic way, i.e., spatio-temporally defined manner; mPSM culture model exhibits its robustness when combined with chemical/signaling molecule treatments or complex experimental settings such as microfluidics. The original mPSM method employed minimal medium and allowed the gradual differentiation of PSM; Hubaud et al. showed that the culture medium could be modified to prolong the oscillatory PSM state [29]. The optimum conditions for retaining the posterior PSM state involved activating posterior PSM signals such as Fgf and Wnt while blocking retinoic acid and BMP signaling. Interestingly, under these conditions, concentric traveling waves continue more than 20 times without inducing spatiotemporal organization of the PSM, i.e. a molecular gradient of Fgf or Wnt.

In zebrafish, Her oscillation can be examined in vitro in singly isolated primary PSM cells. It has been speculated that PSM cells might have an intrinsic autonomous oscillator that can cycle even in the absence of neighboring cells. Webb et al. successfully identified primary cell culture conditions in which dissociated zebrafish PSM cells exhibited several pulses of oscillation in the presence of Fgf8b, confirming the existence of autonomous oscillation [30]. According to the delayed coupling theory of coupled oscillators in the PSM, the intrinsic period of isolated oscillators should significantly differ from the collective period of a population of coupled oscillators [26]. A previous study on mouse PSM cells obtained results consistent with this expectation as they suggested that isolated oscillators were less precise [18]. Time-lapse imaging of the Her1 reporter confirmed that individual cells showed oscillations with a longer period and of less precision [26].

Cultures of singly isolated mouse PSM cells were also attempted. Although oscillatory gene expression is dampened in dissociated mouse PSM cells, because Notch signaling is disrupted [18], Hubaud et al. found that LuVeLu oscillation resumes in the absence of Notch signaling when dissociated PSM cells are cultured on bovine serum albumin (BSA)-coated dishes [29]. PSM cells become round and static when they are cultured on BSA but elongated and motile when they are cultured on fibronectin, a standard coating substrate used for many types of cultured cells. Interestingly, Yap signaling, a mechanotransduction pathway, is more active in cells on fibronectin than in cells on BSA, suggesting that Yap signaling inhibits LuVeLu oscillation. Indeed, treatment with latrunculin A, which represses Yap signaling, restores LuVeLu oscillation in dissociated PSM cells cultured on fibronectin. This single-cell culture method is useful to examine whether a particular gene is required for cellular oscillator machinery or cell–cell coupling, because the abnormality in either process similarly disrupts synchronized oscillation in the PSM. Using this culture method, we found that whereas Hes7 oscillation is dampened and desynchronized in Lfng-null PSM, it occurs almost normally in dissociated Lfng-null PSM cells, suggesting that Lfng mainly functions in the cell–cell coupling process [23].

Manipulation of the segmentation clock

Mechanism to drive the oscillatory expression of Hes/Her genes has been explained by autonomous negative feedback loop via transcriptional repressor activity of these proteins, reminiscent of transcription-translation feedback loop of circadian oscillation. Unlike the circadian clock, the periodicity of the segmentation clock varies from species to species, and this might be probably important to couple somitogenesis to the entire embryogenesis program of each animal. One important question was what regulates the tight periodicity. A mathematical model suggested that transcriptional and translational delays in an auto-inhibition circuit determine periodicity [31]. Takashima et al. tested whether a time delay in the splicing of RNA contributed to regulation of sustained oscillations in the segmentation clock [19]. By comparing the expression dynamics of the Hes7 bioluminescent reporter in which the transcript of the reporter gene is attached to either full-length Hes7 (+ intron) or to the coding region only (-intron), we demonstrated that the presence of introns delayed the timing of reporter expression [19]. Furthermore, when all introns are removed from the Hes7 gene, Hes7 expression does not oscillate but becomes steady, indicating the importance of the delay in splicing for sustained oscillations [19]. Her/Her genes have been recognized as crucial components of the segmentation clock as they exhibit oscillatory expression with periodicity corresponding to somitogenesis period and disruption of these genes lead to absence of downstream oscillatory genes and subsequently the most severe form of segmentation defects [32–34]. However, whether these factors are genuine pacemakers of the segmentation program had still remained elusive. Inspired by the previous intron-less model, we hypothesized that if Hes7 is the fundamental clock pacemaker in the mouse segmentation clock machinery, changing time delay within the feedback process by deleting introns within the Hes7 gene should shorten the segmentation period. Since deletion of all three introns completely disrupts the oscillation, we tried to create more moderate modifications by deleting one or two introns to test the hypothesis. As a result, transgenic mice that carry Hes7 transgene lacking two introns exhibited an acceleration of Hes7 oscillation and subsequent segmentation, as revealed by increased number of cervical vertebrae [35]. Interestingly, this clear acceleration of the segmentation process can be only observed in the cervical and upper thoracic level of the spine, and then segment fusion defects occur in the more-caudal regions. Whether Hes7 is a sole pace-making component, or there are other factors involved in combination with Hes7, is still an open question. Since Wnt and Fgf signaling activity and their downstream genes also exhibit oscillation within the PSM, one possibility is that these signaling molecules might also need to be modulated simultaneously to have a stable accelerated oscillation, although this is challenging to test in the context of intact PSM as it requires simultaneously modulating multiple factors.

Another approach to manipulate the segmentation clock was done by combining mPSM with chemical treatments or complex experimental settings such as microfluidics. Using this method, Sonnen et al. revealed phase-shift encoding in the regulation of segment size using chemical entrainment in mPSM [14]. In addition to LuVeLu, which is basically a reporter for Notch signaling activity, they generated a knock-in reporter for Axin2, a bona fide downstream target of canonical Wnt signaling [11, 36], to quantify Wnt activity in the PSM. In the reporter line, either Venus-PEST or Luciferase-PEST was inserted into the endogenous Axin2 locus by connecting with a 2A self-cleaving peptide, and was designated as Axin2T2A. Simultaneous observation of LuVeLu and the luciferase version of Axin2TA enabled comparison of the spatiotemporal relationship of Notch and Wnt activities; they found that the two signals oscillated in an anti-phase manner in the posterior mPSM, in agreement with previous studies using static analysis [11, 37, 38]. However, the LuVeLu wave was found to traverse the mPSM more slowerly than that of Axin2TA, resulting in in-phase oscillation at the anterior mPSM. This phase shift during the travel of the expression wave from posterior to anterior is reminiscent of the previously shown relationship between Notch and Fgf signaling in which the activities of the two signals are in-phase posteriorly but anti-phase anteriorly [13]. To directly assess the function of the phase shift of Wnt and Notch, Sonnen et al. combined microfluidics with microscopy of reporter expression in an mPSM culture [14]. By time-controlled chemical treatment with a Notch inhibitor (DAPT) and a Wnt activator (CHIR99021; GSK3β inhibitor; Chiron), they examined the effect of chemical entrainment of the two signals either simultaneously or separately. They found that altering the endogenous phase relationship by simultaneous pulses of the two inhibitors resulted in delayed oscillation arrest and impaired formation of physical segment boundaries; these observations are evidence of a dynamic signaling process encoded by two oscillating signaling pathways for regulating segmentation in time and space.

Coordinated dynamics of clock oscillation is crucial for a proper segmentation, and Delta/Notch-based synchronization has been considered to be important in the process as revealed by genetic or transplantation experiment, DAPT treatment, and live imaging-based analysis [16, 23, 39]. Yet, to directly test this mechanism requires artificial or synthetic approaches such as entrainment of the clock by external stimulus. To this end, Soza-Ried et al. created transgenic zebrafish carrying heat shock-inducible DeltaC expression cassette [8]. Strikingly, single pulse of heat shock-induced DeltaC expression rescued formation of several segmentation in the DeltaC mutant background where normally only fragmented or irregular somites would been formed, meaning that artificial DeltaC pulse could entrain the intracellular clock. Interestingly, this entrainment-based rescue worked for longer duration with repeated cycle of pulses, and even with varied frequency which departed from natural one, enabling the modulation of segmentation periodicity.

To gain insights into the phase response of Hes oscillation in the synchronization regime, we applied similar artificial entrainment scheme to a much finer cellular context, combined with real-time monitoring system of clock phase based on live imaging. These systems include synthetic biology-based approaches to reconstitute the minimal circuit in cell lines. We inserted an optogenetic gene expression induction system into the murine myogenic cell line C2C12 [40]. The light-inducible synthetic transcription activator GAVPO was utilized to establish a controllable perturbation system for the Hes1 oscillator. Application of pulsatile light illumination caused pulsatile induction of Hes1; using this system, we showed that the Hes1 oscillator could be entrained by intracellular Hes1 input. We also successfully reconstituted entrainment of Hes1 oscillation by Dll1 input from neighboring cells, providing evidence of the dynamical information transmission through cell–cell interaction.

In vitro model for the segmentation clock

Somitogenesis occurs at the mid-gestation stage of embryogenesis, and many key morphological changes are observed in this developmental period [41]. The characterization of developmental events with live imaging at this stage is, therefore, complicated and time-consuming. In addition, live embryo-based analysis is not suitable for a high-throughput analysis frameworks such as mutant library screening or drug screening, which can provide vital information on known and unknown key regulators of GRNs. Genetic mutations in several genes including HES7 and LFNG are known to cause segmentation defects in vertebrae (SDV) [42, 43]. Clarification of the exact mechanism underlying these defects is problematic as investigation of the human segmentation clock is virtually impossible due to prohibition on culturing human embryos after day 14 of embryonic development at which the primitive streak appears. Recent advances in the development of culture systems for recapitulating developmental events such as the somite segmentation clock might circumvent this difficulty and provide major breakthroughs into mechanisms. In this section, we will discuss the recent progress and utility of novel in vitro models.

In vitro systems that recapitulate the segmentation clock have been established as an alternative strategy to live embryo-based analysis. There have been several attempts to induce paraxial mesoderm in vitro from mouse and human pluripotent stem cell (PSCs) [44–47], as this lineage is a promising source of various important derivative tissues such as dermis, skeletal muscle, vertebral components and joints [48]. In recent studies, Chal et al. recapitulated the development of muscle lineage from human induced PSCs (iPSCs) and mouse embryonic stem cells (ESCs); they induced functional PSM with high efficiency by mimicking the activation and inactivation of signaling pathways in embryonic development [49, 50]. Although they successfully induced paraxial mesoderm or PSM, which had characteristic expression of marker genes and proteins, it is not clear whether the induced cell population has a functional oscillating segmentation clock as no examination of gene expression dynamics was undertaken.

We modified a previously defined method [49] by use of floating culture of embryoid-body like aggregates, a method often used in three-dimensional organoid formation [51], to induce PSM from mouse ESCs [52]. To monitor Hes7 activity during the PSM induction process, a bioluminescent reporter was introduced into the cells. In cultured aggregates on day 4 of induction, propagating Hes7 oscillation waves from the center to the periphery of the aggregates were clearly observed and shown to have a period of 2–3 h, a comparable period to its in vivo counterpart. Notably, in the peripheral regions where the propagation wave stopped, segment-like boundaries were formed after each pulse of Hes7 oscillation. This observation suggested the formation of an antero-posterior axis and somite differentiation similar to the in vivo situation. In situ hybridization of the aggregates for PSM-specific marker genes such as Fgf8 and Dusp4 confirmed that an antero-posterior axis was formed in the peripheral-to-central direction. These overall features are reminiscent of the PSM explant culture such as mPSM [28], indicating that the culture condition allowed the cells to self-organize a PSM-like differentiation program, which is coupled to the oscillating segmentation clock. It has been suggested that the somite is a self-organizing structure and can be decoupled from the clock; support for this speculation comes from an experiment in which a chicken posterior primitive streak was grafted onto the extra-embryonic region and simultaneous formation of multiple somite-like structures was seen [53]. There is still uncertainty about which culture conditions will allow cells to coordinate a segmentation clock and somite differentiation and how the axis is formed in a self-organized manner. The in vitro system of ESC-derived PSM formation will be useful for the analysis of the coupling mechanism between the clock and somite formation. This in vitro system also allowed high-throughput drug screening of the effects of chemical application by real-time monitoring of Hes7 oscillation. A variety of compounds were tested including some that influenced Notch, Wnt, and Fgf signaling, and others selected from a chemical library [23, 52]. These analyses led to the identification of the BET-family of proteins as novel factors in the segmentation clock; BET proteins might have a role in the regulation of phase-synchronization [52]. A similar screening of the effects of small molecules on somite segmentation in live zebrafish embryos has been carried out; this screening assessed the effects of chemical treatment by an in situ hybridization-based segmentation phenotypic analysis [54]. It is possible that an in vitro model of the mouse segmentation clock might provide a comparable mammalian counterpart to the zebrafish screening system and allow both quantification of clock oscillation and phenotypic analysis.

Recently, an in vitro model of human somitogenesis based on human ESCs and iPSCs was reported [55–57]. Chu et al. used previously described methods to induce mesoendoderm, PSM, and somites from human ESCs in a step-wise manner and examined gene expression dynamics during the induction process by RNA-seq analysis [55]. They found evidence of transient activation of segmentation clock genes during the PSM to somite transition; to confirm this observation, they analyzed HES7 oscillation by introducing NanoLuc Luciferase into the endogenous locus of HES7 via CRISPR/Cas9-mediated gene targeting and confirmed that HES7 oscillates with ~ 5 h periodicity, which is consistent with earlier estimations [58, 59]. By introducing a known mutation of HES7 (R25W) via CRISPR/Cas9, they modeled spondylocostal dysostosis, a congenital disease with abnormal vertebral formation. They showed that HES7 oscillation was completely eliminated in the homozygous clone-derived PSM induced in vitro. Matsuda et al. induced PSM from human iPSCs using a similar step-wise induction method [57]. The induced human PSM was capable of differentiating into somatic mesoderm, sclerotome, and dermomyotome. By collecting the induced PSM samples in the oscillatory state and performing RNA-seq analysis, they identified ~ 200 human oscillatory genes showing varied phase-shift. By utilizing the luciferase-based HES7 reporter, they showed that HES7 oscillations were dampened in the human PSM induced from iPSCs in which SDV-related genes (HES7, LFNG, and DLL3) were mutated or from SDV patient-derived iPSC lines [57]. Diaz-Quadros et al. applied an in vitro induction strategy for mouse PSM to human PSM cells [56]. Interestingly, their culture method recapitulated the synchronization of the oscillators, a characteristic property of the PSM. They also examined the roles of YAP and FGF signaling on the control of HES7 oscillation, opening the way to perform a functional characterization of the human segmentation clock using an in vitro model. The induction of PSM cells or its derivatives does not always bring about segmentation or formation of epithelialized somites. van den Brink et al. reported that somitogenesis can be mimicked in vitro through the induction of gastruloid, three-dimensional aggregates of ESCs that exhibit germ-layer specification and axial organization [60]. They characterized the gastruloids derived from mouse ESCs and found that the key regulators of somitogenesis were expressed in the aggregates. By utilizing the LfngT2AVenus reporter, they showed that the clock oscillation waves within the gastruloids had a period of about 2 h. Strikingly, by embedding the gastruloids in a low concentration of Matrigel, they observed sequential formation of segment-like structures with antero-posterior polarity. It remains unclear whether the structures contained epithelialized somites, or whether these segment-like structures were a rigid boundary; nevertheless, this study demonstrated the possibility of recapitulating the entire process of somitogenesis, including the induction of oscillatory PSM, sequential differentiation, epithelialization of somites, and boundary formation.

As described above, the in vitro system offers a unique opportunity to analyze the segmentation clock without using live embryos. It is now possible to make hundreds of PSM-like tissues in vitro without any difficulty, and this method will be useful for high-throughput drug and gene screening. Such screening will lead to identification of novel genes responsible for synchronized oscillations. This method is also useful for analyzing species-specific differences of the segmentation clock. For example, the period of the segmentation clock is different among species, and this method offers a powerful tool to directly analyze the mechanisms for such species-specific features.

Concluding remarks

In summary, live imaging-based quantification of clock gene expression is a critical method for the dissection of the segmentation clock, and in vitro models of the clock provide a powerful analytical framework for obtaining novel insights into its mechanisms.