(C) PLOS One [1]. This unaltered content originally appeared in journals.plosone.org.
Licensed under Creative Commons Attribution (CC BY) license.
url:https://journals.plos.org/plosone/s/licenses-and-copyright

------------



Patterning with clocks and genetic cascades: Segmentation and regionalization of vertebrate versus insect body plans

['Margarete Diaz-Cuadros', 'Department Of Genetics', 'Harvard Medical School', 'Boston', 'Massachusetts', 'United States Of America', 'Department Of Pathology', 'Brigham', 'Women S Hospital', 'Olivier Pourquié']

Date: 2022-02 

In the previous section, we described the spatiotemporal dynamics of gene expression and signaling pathway activities during regionalization and segmentation of the AP axis in insects and vertebrates. Here, we describe the mechanisms that mediate these expression patterns.

As gap gene expressions provide positional information for downstream pair-rule genes, pair-rule gene expressions, likewise, have been suggested to provide positional information for downstream segment polarity genes in a yet another round of the French Flag model [ 146 ]. However, it was recently suggested that pair-rule gene expression concentrations are not read out in an ON–OFF fashion, but rather the temporal sequence of some of pair-rule gene expressions (mediated by the posterior-to-anterior shifts of gap genes, discussed in section “Toward reconciliation of simultaneous and sequential models of patterning”) encode the positional information for other pair-rule genes as well as segment polarity genes [ 14 ].

After this initial phase of stripe specifications, the full expression pattern of pair-rule genes is then stabilized and/or refined and goes into a phase of frequency doubling. These late effects seem to depend on a rewiring of the pair-rule network, mediated by late-acting enhancers. Each of these late-acting enhancers drives the full pair-rule pattern (7 stripes in Drosophila) in enhancer reporter assays, and, hence, are sometimes called “7-stripe” or “zebra” elements [ 141 ] ( Fig 4B ). This rewiring of the pair-rule network (and possibly other genes involved in AP and dorsoventral patterning in the early Drosophila embryo) seems to be mediated by timing factors encoded by 2 pioneer factors: Zelda and Opa, where Zelda is responsible for activating the early network and Opa for the late network [ 142 – 144 ]. Indeed, in opa mutants, the frequency doubling of pair-rule genes is lost in Drosophila [ 142 ]. Dichaete was also recently suggested to mediate the transition from early to late patterning [ 13 ]. Interestingly, cad (potentially along with zelda), dichaete, and opa were found to be activated sequentially in the Drosophila embryo, reflecting a similar sequential activation in space and time in the Tribolium embryo [ 13 , 145 ].

Reaction–diffusion models have been implicated in the formation of various patterns during development, most of which are periodic [ 154 , 155 ]. The formation of nonperiodic patterns within the reaction–diffusion framework would require the employment of large number of diffusing molecules and/or signaling pathways, a condition that is hard to realize in most embryonic tissues.

Reaction–diffusion is a pattern formation mechanism in which a set of diffusible molecules interact and spontaneously pattern an initially homogeneous tissue [ 148 – 151 ]. The simplest and most popular instance of the reaction–diffusion model is the “local autoactivation and lateral inhibition” mechanism [ 147 , 152 ], where a slowly diffusing molecule (say, A; see Fig 5 ) can activate itself as well as a fast-diffusing molecule (say, B). Molecule B in its turn can inhibit A’s activity. Due to the ability of molecule A to self-activate, random fluctuations in A’s activity are occasionally amplified, forming localized domains of high A concentration in space, activating in their turn, corresponding domains of B activity. Due to the high diffusivity of molecule B, its activity domains then inhibit nearby A activity domains. This eventually leads to the formation of periodic patterns of A and B concentrations ( Fig 5 ). The shapes of these activity domains (whether dotted or striped) and their characteristics (e.g., spacing between activity domains) are determined by system parameters and boundary conditions. The “local autoactivation and lateral inhibition” model, however, is one instance of the reaction–diffusion model, and it has been shown recently that the condition of differential diffusivities of molecules A and B can be relaxed if systems with more than 2 components are utilized [ 149 , 153 ].

The simple periodic pattern of pair-rule genes tempted computational modelers to postulate that they are elegantly regulated by a reaction–diffusion mechanism ( Box 2 ; Fig 5 ; [ 136 , 137 ]). However, experimental evidence showed that pair-rule genes are mainly regulated (rather inelegantly, [ 138 ]) by another round of the French Flag model. In this scheme, each pair-rule stripe (or pair of stripes) is mediated by a separate “stripe-specific enhancer.” Each stripe-specific enhancer is regulated in a dose-dependent manner by the upstream gradients of gap genes and various maternal factors (e.g., Bcd and Cad), whose identities are different depending on the specific location of the stripe along the AP axis of the embryo [ 139 , 140 ] ( Fig 4B ) (see [ 12 , 119 ] for more detailed discussion of pair-rule gene regulation in Drosophila).

(A) Maternal Dm-Hb gradient acts as a master regulator of gap genes in Drosophila. Progressive reduction of maternal Dm-Hb gradient (in various mutant backgrounds) results in progressive shifts of gap gene domains toward anterior. (B) Pair-rule stripes (shown for Dm-eve) are specified in a stripe-specific fashion in Drosophila. Left: Each one or pair of Dm-eve stripes are specified by a specific enhancer that receives inputs from upstream gap genes (shown repressive gap inputs for the 3+7 and 4+6 enhancers; strong repression shown in sold lines; weak repression in dashed lines). The full 7-stripe pattern is then stabilized by a 7-stripe (or zebra) enhancer. Right: Shown how the regulatory logics of 3+7 and 4+6 Dm-eve enhancers translates upstream gap gene expressions (shown are those of Dm-Kr and Dm-kni) into stripe pairs [ 140 ]. (C) A sketch of the genetic wiring of the Tribolium segmentation clock, composed of the 3 primary pair-rule genes: Tc-eve, Tc-run, and Tc-odd, wired into a negative feedback loop. Note that this is a parsimonious wiring explaining observed gene expression dynamics in WT and knockdown phenotypes. Actual wiring might differ from the one shown, especially that pair-rule genes are known to act as repressors rather than activators. (D) Experimental evidence of segmentation clock wiring in Tribolium. (E) A sketch of the genetic wiring of the Tribolium gap gene cascade. Note that this is a parsimonious wiring explaining observed gene expression dynamics in WT and knockdown phenotypes. Actual wiring might differ from the one shown, especially that most gap genes are known to act as repressors rather than activators. (F) Experimental evidence that gap genes are wired into a genetic cascade in Tribolium: repressing a single gap gene results in the up-regulation of upstream genes in the cascade and down-regulation of downstream genes. (G) Experimental evidence that the Wnt/Cad gradient (shown in gray) acts as a speed regulator of the segmentation clock in Tribolium. In Tc-lgs RNAi embryos: Tc-cad gradient is reduced (i.e., its peak concentration is lower than in WT) and shifted toward posterior; concomitantly, Tc-eve oscillation frequency is reduced and Tc-eve waves are shifted toward posterior. In Tc-pan RNAi embryos: Tc-cad gradient is reduced, shifted toward anterior, and stretched; concomitantly, Tc-eve oscillation frequency is reduced and Tc-eve waves are shifted toward anterior and stretched. In Tc-zen RNAi embryos: Tc-cad gradient has the same peak level and slope as in WT, but just shifted toward anterior; concomitantly, Tc-eve waves are shifted toward anterior without any sign of spatial stretch or time dilation. (H) Further evidence that gap genes are wired into a genetic cascade in Tribolium. Upon reinducing the leading gap gene in the cascade (Tc-hb) using a transgenic line where Tc-hb minigene is placed downstream of a heat-shock promoter, the whole gap gene sequence is reinduced in the SAZ. (I) A possible model for how the speed of the pair-rule clock or the gap gene cascade is modulated by a Wnt/Cad gradient in Tribolium: Wnt/Cad activates the pair-rule clock and/or gap gene cascade, but represses a multistable gene regulatory network. The gradual switch between the 2 gene networks results in the gradual slowing down of pair-rule oscillations and/or gap gene sequential activation. Throughout the figure, for all embryo schematics: anterior to the left and posterior to the right. RNAi, RNA interference; WT, wild-type.

(A) Top panel: In the FF model, different concentrations of a morphogen gradient (shown in gray) activate different cellular states (shown in different colors). The FF model can pattern nonelongating tissues (bottom left) via a nonregressing gradient (shown in gray), as well as elongating tissues (bottom right) if a retracting short-range gradient (shown in black) activate a slowly decaying morphogen (gray). (B) In a similar fashion, the FF model can generate periodic patterns if a complex regulatory logic of periodically expressed genes (shown in brown) is employed or if an intermediate step of nonperiodic patterning is introduced. FF, French Flag.

We note here that the French Flag model can also mediate periodic patterning, if a patterning gene (like the one shown in brown in Fig 3B ) is repeatedly activated then inactivated by a continuous range of morphogen gradient concentrations ( Fig 3B ). This scheme, however, requires patterning genes to be under the control of complex cis-regulatory logics. Alternatively, such a scheme can be mediated by an intermediate stage of nonperiodic patterning, a mechanism implicated in regulating pair-rule genes in Drosophila.

The French Flag model, however, can be modified to mediate patterning using a regressing short-range gradient ( Fig 3A , right) [ 135 ]. In this scheme, a regressing short-range gradient (shown in black in Fig 3A , right) activates a gene whose product is of negligible decay rate (shown in gray in Fig 3A , right). In this way, a long-range gradient will span the full length of the tissue by the end of the tissue elongation phase. Such long-range gradient can then mediate patterning via a typical French Flag mechanism ( Fig 3A , right).

In the earliest formulation of the French Flag model, the morphogen gradient was assumed to be static (i.e., does not change in shape or concentration in space and time), suitable for partitioning a nonelongating tissue ( Fig 3A , bottom left). In this scheme, long-range morphogen gradients are required to pattern large tissues. This is, however, hard to achieve in cases where tissue sizes are much larger than the range of influence of typical signaling pathways. Large tissues, in fact, are usually patterned during tissue growth. In such cases, a signaling center tethered at the elongating end of the tissue results in a regressing short-range gradient.

The French Flag model is one of the earliest models of pattern formation in development [ 130 ]. In this model, a morphogen gradient spans a tissue, and different morphogen concentrations specify different fates along the gradient (represented by different colors in Fig 3A (top)), where each fate is typically specified by the activity of 1 or more genes. Consequently, borders between consecutive gene expression domains along the tissue are specified by a series of thresholds of the morphogen gradient (e.g., T1 and T2 in Fig 3A , bottom left).

The French Flag model, in which different concentrations of a morphogen gradient turn on or off different genes ( Box 1 ; Fig 3 ) [ 130 ], has been the prime theoretical framework for how gap gene expressions are initialized in the early Drosophila embryo. Modulating the maternal expression of Hb alters the positioning of gap gene expression domains, in a fashion consistent with a French Flag model in which maternal Hb acts as a master morphogen gradient ( Fig 4A ) [ 125 , 126 , 131 ]. Later in time, cross-regulatory interactions between gap genes themselves refine their final expression patterns [ 34 ]. However, recent data, while confirming the primary role of the Hb gradient in regulating gap gene expression, indicate that gap gene regulation is much more dynamic [ 50 , 52 , 132 ] and follows a temporal mode of patterning similar to that observed in short-germ insects [ 9 , 51 ]. We discuss this point in more detail in the section “Toward reconciliation of simultaneous and sequential models of patterning.”

Mechanisms of segmentation in vertebrates and short-germ insects

The predominant view is that segmentation takes place in vertebrates and short-germ insects through a clock and wavefront mechanism (Box 3; Fig 6B, bottom right panel). This model was first proposed by Cooke and Zeeman in 1976 to be the underlying mechanism of vertebrate segmentation [156]. The basis of this model as originally formulated was that a catastrophe leading to abrupt changes in cellular properties takes place in the anterior PSM and underlies somite formation. The periodicity of this catastrophe is controlled by an oscillator that interacts with a slowly regressing maturation front, also known as the wavefront [156]. When a specific phase of the oscillator hits the wavefront, the catastrophe is triggered and results in somite individualization. Experimental evidence for the clock and wavefront model was not available until more than 20 years after its initial publication, when oscillations in the expression of the transcription factor cHairy1 were discovered in the chicken PSM [4], and later on, oscillations in the expression of pair-rule genes eve and odd were discovered in the Tribolium SAZ [6,7]. Since then, molecular characterization of the clock and wavefront components has lent credibility to the model.

Box 3. The speed regulation model Model The speed regulation model is a synthesis of various patterning schemes that all rely on a single core mechanism [3,174]: the ability of a morphogen gradient to modulate the speed of a temporal sequence, either periodic or nonperiodic. In the nonperiodic version of the speed regulation model, each cell within a tissue has the capacity to transit through successive states (shown in different colors in Fig 6A, top; each state is defined by the expression of 1 or several genes), where the speed of state transitions is regulated by a molecular factor (shown in gray at the top of Fig 6A, and henceforth called a “speed regulator”). At very low or zero concentration of the speed regulator, state transitions become so slow that states are indefinitely stabilized (Fig 6A, top left). If a group of cells is subject to a gradient of the speed regulator (Fig 6A, bottom left), all cells go through successive states, but with slower and slower speed as we go from higher to lower values of the speed gradient. This gives the appearance that cellular states propagate as waves in the high-to-low direction of the gradient. Note that such waves do not require diffusion or cell–cell communication and, hence, are called “kinematic” or “pseudo-waves” [175–180]. We call this version of the model “gradient-based speed regulation,” which is well suited for patterning nonelongating tissues (Fig 6A, bottom left). The speed regulation model can also pattern elongating tissues if the gradient is retracting as a wavefront (henceforth called “wavefront-based speed regulation” model; Fig 6A, bottom right). In the same fashion, the speed regulation can partition a tissue into periodic structures if the sequential gene activation process is simply replaced with a clock (Fig 6B). Note that in the wavefront-based version of the speed regulation, if the wavefront is in the form of a tapered gradient (which can be seen as superposition of the gradient-based and the wavefront-based modes of the speed regulation model), kinematic waves will propagate from high to low concentrations of the gradient (opposite to the direction of wavefront retraction) as the tissue is elongating. The speed regulation model in development Special cases of the speed regulation model have been previously implicated in patterning various embryonic tissues. A nonperiodic version of the wavefront-based speed regulation model (Fig 6A, bottom right) was termed the “progress zone” model [27,181] and was (arguably [182,183]) suggested to pattern the vertebrate limb bud and was again termed the “temporal regulation” model and implicated in Drosophila neurogenesis [19,20]. The clock and wavefront model [156], originally suggested as an underlying mechanism of vertebrate somitogenesis, is, in fact, a periodic version of the wavefront-based speed regulation model (Fig 6B, bottom right). Careful inspection of the axis elongation phases of vertebrate and insect segmentation indicated that segmentation genes in these species are expressed in waves that propagate opposite to the direction of wavefront retraction [4]. This observation was reconciled with the clock and wavefront model by assuming that the wavefront is mediated by a tapered concentration gradient that modulates the frequency of clock oscillations (a model developed by Julian Lewis in [4]), which is a superposition of the gradient-based and wavefront-based modes of a periodic speed regulation model (Fig 6B). Indeed, FGF signaling was to found to fit the proposed criteria for a speed regulator, as its activity level was found to modulate the speed of the segmentation clock in an in vitro assay [184]. Similar oscillatory waves were observed during the blastoderm stage of Tribolium segmentation [6]. These waves were hypothesized to be mediated by a static frequency gradient, which is a periodic version of a gradient-based speed regulation model (Fig 6B, bottom left) [6]. A similar model was used to explain the oscillatory waves observed in PSM ex vivo cultures [185]. Nonperiodic versions of these models were then suggested to mediate gap gene regulation in Tribolium [3]. Finally, a nonperiodic version of the gradient-based speed regulation model (Fig 6A, left) was coined the “temporal/spatial patterning” model and was implicated in patterning the vertebrate neural tube [21] (although it is not clear if the underlying mechanism is truly a speed regulation model or a dynamic version of the French Flag model; see [10] for a discussion).

PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 6. The SR model (see also The SR model (see also Box 3 ). (A) Top panel: In nonperiodic SR model, a speed regulator (shown in gray) modulates the speed of cellular state transitions (each cellular state is shown in a different color) in a dose-dependent fashion (top panel, right). At a very low or zero concentration of the speed regulator, cellular state transitions are arrested (top panel, left). (A) Bottom panel: The SR model can operate in a gradient-based mode to pattern nonelongating tissues (left) or in a wavefront-based mode to pattern elongating tissues (right). (B) If the processes driving cellular state transitions is periodic (i.e., driven by a clock, which expression is shown in brown; top panel), the SR model can generate periodic patterns in both elongating and nonelongating tissues (via a wavefront-based and gradient-based modes of the model, respectively). SR, speed regulation. https://doi.org/10.1371/journal.pgen.1009812.g006

The clock. Irrespective of the details of the clock and wavefront model, which are still actively debated, it is clear that a molecular oscillator is at work in the PSM of vertebrates and SAZ of short-germ insects. Such oscillations in cyclic gene expression are thought to be generated by negative feedback loops with delays [157] (Fig 7A). In vertebrates, for instance, bHLH Hes/Her transcriptional repressors can inhibit their own promoters [158]. Once induced, accumulation of HES/HER protein leads to transcriptional silencing until such time as the proteins are degraded and transcription can resume once more. Synchronization of individually oscillating cells in vertebrates is then mediated by cell–cell coupling through Delta–Notch interactions. Inhibiting Delta–Notch signaling disrupts traveling waves and leads to a salt and pepper expression pattern of Hes/Her genes in both mouse and zebrafish [159–161]. In some (but not all) arthropods, Notch signaling has been shown to be involved in segmentation [61,63,66–68]. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 7. Mechanisms of segmentation and regionalization in vertebrates. (A) Delayed negative feedback loop giving rise to oscillations in Hes/Her expression. In the absence of transcriptional repression by autoinhibition (1), Hes/Her genes are activated and mRNA transcribed (2). This leads to HES/HER protein translation (3) and accumulation. After a time delay associated with gene expression steps, HES/HER proteins reach sufficient levels to bind the Hes/Her regulatory regions and inhibit transcription (4). Autoinhibition is relieved by HES/HER degradation (5) and the cycle begins again. (B) Mechanism of FGF gradient formation. Only progenitor cells (pink) actively express the fgf8 ligand. Once cells ingress into the PSM (purple), they cease to transcribe fgf8. Progressive degradation of fgf8 mRNA and protein leads to gradient formation as cells acquire more anterior positions within the paraxial mesoderm. (C) Genomic organization of the HoxA cluster. HoxA1-13 genes are arranged colinearly within the cluster in the 3′ to 5′ orientation. Chromatin opening and gene expression start at the 3′ end and proceed in the 3′ to 5′ direction. Genes are colored according to the vertebral identities they specify. The 2 TADs (3′ and 5′) are shown as gray triangles. Anterior Hox genes are activated by Wnt signaling (red), central Hox genes by Wnt/Cdx (green), and posterior Hox genes by Gdf11/TGFβ (blue). PSM, presomitic mesoderm; TAD, topologically associated domain. https://doi.org/10.1371/journal.pgen.1009812.g007 In amniotes, oscillations of targets of the FGF and Wnt signaling pathways are thought to arise from similar time-delayed negative feedback loops. FGF and Wnt ligands provide constant pathway activation in the posterior PSM, resulting in the induction of target genes like Dusp or Axin2, respectively. As these genes are involved in feedback inhibition (i.e., dual-specificity phosphatase (DUSP) dephosphorylates mitogen-activated protein (MAP) kinases; AXIN2 is a component of the β-catenin destruction complex), their expression shuts down pathway activity, and, hence, their own transcription [75,161]. However, negative feedback loops cannot fully account for the mechanism of segmentation clock oscillations, as overexpressing constitutively active β-catenin does not impair Wnt or Notch oscillations in the mouse PSM [162]. Thus, oscillations might be generated by noncanonical regulation of Wnt and FGF targets. In mouse embryonic stem cells, desynchronized oscillations of many genes with the same period as the mouse segmentation clock have been observed [163]. This suggests that oscillatory gene expression dynamics might be more widespread than currently appreciated. In Tribolium, 3 primary pair-rule genes (eve, run, and odd) are thought (arguably, [12,113]) to be wired into a genetic clock that produces their oscillatory expressions [60] (Fig 4C). This hypothesis is supported by RNA interference (RNAi) knockdown experiments: Knocking down eve results in the down-regulation of both run and odd, knocking down run leads to the down-regulation of odd and the overexpression of eve, and knocking down odd leads to the overexpression of both eve and run [60] (Fig 4D). These experiments suggest that Tribolium eve, runt, and odd are wired into a negative feedback loop in the same order as they appear in the SAZ (Fig 4C). Knocking down 1 or more of the primary pair-rule genes has been found to have a similar (but not identical) disruptive effect on the overall pair-rule pattern in other short-germ insects as well [31,32,164,165]. So far, there is no evidence that Notch signaling is involved in segmentation in Tribolium [67,68].

The wavefront. The segmentation clock must interact with the determination front (or wavefront) to establish segment boundaries (Box 3). In vertebrates, this determination front is positioned by gradients of FGF, Wnt, and RA signaling [166]. Experimentally manipulating these gradients artificially displaces the wavefront. For instance, transient FGF inhibition shifts the wavefront posteriorly and results in larger somites [91,167]. These signaling gradients are formed by a combination of localized ligand production, progressive ligand decay, and ligand diffusion. In the case of FGF signaling, Fgf8 mRNA is actively expressed at high levels in the posterior progenitor zone [92]. However, once cells are specified as mesodermal and ingress into the PSM, they cease to express this gene (Fig 7B). As the embryo continues to elongate posteriorly and cells progressively acquire more anterior positions, dilution and degradation of existing Fgf8 mRNA molecules give rise to the posterior–anterior gradient. FGF8 protein decay has not been directly measured, but also certainly plays a role in establishing the gradient. In addition to this decay mechanism, FGF ligand diffusion takes place in the PSM and contributes to gradient formation [94]. A similar mechanism probably underlies the Wnt3a gradient, as this signaling molecule is only transcribed in the tail bud and posterior PSM [73]. Given that Cdx genes are targets of both Wnt and FGF signaling in vertebrates, a parallel posterior-to-anterior Cdx gradient is formed downstream of these pathways [168–170]. Progressive Cdx mRNA and protein degradation also contributes to Cdx gradient formation [171]. In the case of the anterior-to-posterior RA gradient, the RA-synthesizing enzyme Raldh2 is expressed in the somites and anterior PSM, whereas the RA-degrading enzyme Cyp26A1 is expressed in the tail bud [172,173]. This gradient is thus formed by a source-and-sink mechanism coupled with high RA diffusivity. In Tribolium, the wavefront is thought to be encoded by Wnt and/or Cad gradients, since their expressions overlap pair-rule oscillations in the SAZ [8,13] (see cad expression in Fig 2B; compare with pair-rule gene expressions in Fig 2A). Once out of the domain of Wnt/Cad expression, pair-rule oscillations are fixed into stripes. The expressions of opa and dichaete were recently shown to form staggered wavefronts along with Wnt/Cad [13]. Whereas opa was suggested to mediate the frequency doubling of pair-rule stripes upon their exit from the SAZ (see below), it is not clear if opa and/or dichaete are involved in slowing down and arresting pair-rule waves intro stripes, nor how they interact with the Wnt/Cad gradient. Posterior Wnt activity in Tribolium is mediated by both wnt1 and wnt8 ligands, which are expressed posteriorly in the embryo [85–87]. Wnt then activates cad resulting in a parallel Cad gradient that itself activates wnt1, forming a positive feedback loop [8,85]. In the Tribolium blastoderm, the Wnt/Cad gradient is first established by a maternal counter gradient of the Wnt negative regulator axn and is additionally modified by the repressing effect of zen and otd counter gradients [8,85]. Wnt/Cad gradient formation and retraction during the germ-band stage is not well studied, but in principle could be mediated through RNA/protein decay, similar to FGF gradient formation in the vertebrate PSM.

Segment specification and polarity. In vertebrates, the determination front set up by Wnt and FGF signaling pathways functions to endow cells with the competence to respond to the clock signal and establish the future somite prepattern. Upon stimulation by the segmentation clock, cells in the determination front transiently stabilize Notch signaling in the form of a bilaterally symmetric stripe of Notch activation that can be visualized by Notch intracellular domain (NICD) expression [186]. Notch activity results in the TBX6-dependent expression of Mesp1/2, the master regulators of the segmental program [187]. Following segment specification, Mesp1/2 become restricted to the rostral part of the future somite, whereas Delta-like1 (Dll1) and NICD are restricted to the caudal part [77,188]. This results in the establishment of somite rostral–caudal polarity. Mesp1/2 expression is eventually down-regulated through a negative feedback loop involving its target Ripply2 and Tbx6 [189,190]. Somite polarity is subsequently maintained by the transcription factors Tbx18 (rostral) and Uncx4.1 (caudal), which functionally antagonize each other [191]. In Tribolium, as the expression waves of the primary pair-rule genes eve, run, and odd propagate out of the SAZ, they slow down and eventually refine their expression patterns and undergo frequency doubling [3,6,13,60]. In addition, primary pair-rule genes regulate the secondary pair-rule genes paired (prd) and sloppy-paired (slp) at the border of the SAZ [60]. The later phase of pair-rule frequency doubling has been suggested to be mediated by opa, which is expressed along with Wnt/Cad and dichaete in a staggered wavefront [13]. The refined expressions of the Tribolium primary and secondary pair-rule genes constitute a combinatorial code that divides each clock-mediated pair-rule wave into 2 parasegments and defines their polarities. The anterior of odd-numbered parasegments is demarcated by the expression of segmental eve stripes and prd, whereas the anterior of even-numbered parasegments is demarcated by the expressions of eve, run, odd, and prd. The posterior of all parasegments is demarcated by slp expression. This combinatorial code eventually regulates the segment polarity genes en and wg at the anterior and posterior of each parasegment, respectively [80], in a fashion similar to segment polarity gene regulation in Drosophila [192].

How do clock oscillations organize into traveling waves across vertebrate and short-germ insect embryos? When cHairy1 oscillations were first discovered in the chick PSM, they were shown to be expressed in periodic waves that sweep the PSM from posterior to anterior, progressively slowing down and becoming narrower as they approach the determination front [4,193,194], an observation not accounted for in the original formulation of the clock and wavefront model. These waves do not require diffusion or cell–cell communication, and, hence, are called “kinematic” or “pseudo-waves” [4,195] (although the involvement of cell–cell communication might still be necessary to ensure synchronous oscillations in neighboring cells [159]). A simple model for how such waves are produced is to assume that the regressing wavefront modulates the frequency of the segmentation clock in a dose-dependent fashion, acting, consequently, as a regressing frequency gradient (Box 3) [4,179]. The mechanisms underlying this frequency gradient are currently not well understood. Although there is some evidence suggesting that traveling waves are initiated and controlled by FGF through a posterior-to-anterior phase delay [196], traveling waves can still take place in the absence of an FGF gradient [197]. Interestingly, different cyclic genes display distinct frequency profiles. For instance, in mice, the Wnt target Axin2 displays rapid waves that slow down abruptly, whereas waves of the Notch target Lfng travel more slowly [76]. The functional importance of the gradual slowing down of clock oscillations is unclear, since, theoretically, the performance of the clock and wavefront model is basically the same whether oscillations are arrested suddenly in a catastrophic event (as in the original formulation of the model) or through a gradual slowdown [8,179]. However, during the blastoderm stage, Tribolium pair-rule genes were shown to be expressed in waves in the absence of axis elongation or a regressing wavefront (Fig 2A), a fact that cannot be explained by the original formulation of the clock and wavefront, but rather by the action of a nonregressing frequency gradient [6]. This highlights the importance of the gradual slowdown of clock oscillations in certain scenarios. Oscillations of the segmentation clock, therefore, can be translated into periodic spatial patterns either by a regressing wavefront or by a nonregressing frequency gradient, a fact that is elucidated in a unified model for time-based patterning called the “speed regulation” model (Box 3) [3,174]. The model describes a core regulatory mechanism in which a gradient of a molecular factor (called the speed regulator) regulates the speed of the segmentation clock (or any temporal process). A nonregressing gradient of this speed regulator can by itself induce oscillatory waves, and, hence, pattern nonelongating tissues like insect blastoderms (Fig 6B, left bottom panel; Box 3), whereas a regressing gradient of the speed regulator can pattern elongating tissues like insect germ-bands and vertebrate PSMs (Fig 6B, bottom right panel; Box 3). Interestingly, the formation of the prechordal and head mesoderm in vertebrates is associated with 2 cycles of oscillatory gene expression that take place in the epiblast without a regressing wavefront and might be conceptually similar to pair-rule pulses in the Tribolium blastoderm [198]. Hence, a single model can explain both blastoderm and germ-band segmentation in short-germ insects [3]. The Wnt/Cad gradient has been suggested to act as a speed regulator for the pair-rule oscillator in Tribolium [3,8]. Experimentally manipulating the Wnt/Cad gradient in several genetic backgrounds led to stereotypical changes in the spatiotemporal dynamics of pair-rule waves [8] (Fig 4G). In most cases, reducing the intensity of the Wnt/Cad gradient led to slower pair-rule wave dynamics, shifting the Wnt/Cad gradient anteriorly or posteriorly led to a concomitant anterior or posterior shift of pair-rule waves, respectively, and stretching the Wnt/Cad gradient led to a stretch in pair-rule waves (Fig 4G). Using computational modeling, these observations were shown to be consistent with the speed regulation model [3,8]. However, these conclusions are based on correlations between the Wnt/Cad gradient and pair-rule wave dynamics, and it is still unclear whether other factors are also involved (like opa and dichaete [13]) or what molecular mechanisms underlie such speed modulation of segmentation clock oscillations (more on that below). Most of these observations are made during the early phase of segmentation in the Tribolium blastoderm, where other maternal and zygotic gradients exists (e.g., otd and zen) that could be directly involved in regulating pair-rule gene regulation. Moreover, it is unclear if the same model is applicable to later stages of segmentation in the germ-band. However, in axn RNAi embryos, the cad gradient was shown not to regress in the Tribolium germ-band. Concomitantly, gap gene waves (discussed in section “Mechanisms of regionalization in short-germ insects”), which are nonperiodic version of pair-rule waves, were observed to continue propagating and shrinking without being arrested into stable expression domains in this genetic background, as predicted by speed regulation model [3]. This suggests that regionalization, and potentially segmentation, rely on the same core mechanism in both the blastoderm and the germ-band. In vertebrates, it has recently been reported that pharmacological inhibition of FGF leads to a dose-dependent lengthening of the oscillatory period for the Lfng reporter LuVeLu in mouse PSM explants [184]. Furthermore, treatment of mouse and chicken PSM with Wnt inhibitors can alter the frequency profile of traveling waves [199]. These studies suggest that gradients of FGF and Wnt signaling might act as speed/frequency regulators in vertebrate embryos similar to how Wnt/Cad gradients have been suggested to act in short-germ insects [8]. In line with the role of Wnt/FGF signaling in frequency modulation in amniotes, sustained oscillations were observed in mouse PSM explants cultured under experimental conditions that maintained a uniform level of Wnt and FGF activities, without any sign of variability in frequency [197]. On the other hand, explants under normal conditions, in which Wnt and FGF gradient are established, a gradient of oscillation frequencies is observed [185]. Interestingly, however, in the absence of Wnt and FGF gradients, oscillatory waves are still observed in mouse PSM explants [197]. These waves, however, show no sign of progressive reduction in length as seen in explants cultured under normal conditions. This indicates that in the absence of a frequency gradient (presumably mediated by Wnt/FGF signaling gradient), cells still self-organize into spatial waves, potentially through the emergence of phase differences between neighboring cells. But if the frequency of the segmentation clock is modulated along the AP axis via a speed regulator gradient, how could this be achieved at the molecular level? In a recently devised model of AP patterning in Tribolium [3,200], segmentation genes where proposed to be wired into 2 different gene regulatory networks: a clock network and a multistable gene network (Fig 4I). If a Wnt/Cad speed regulator activated the clock network but repressed the multistable network, a gradual slowdown in oscillation will result along the speed regulator gradient (Fig 4I). A similar model was suggested for gap gene regulation in Tribolium (replacing the clock with a genetic cascade). In line with this model, reactivating the gap gene hb all over the embryo resulted in 2 distinct responses [201]. Within the Wnt/Cad expression (defining the SAZ), the gap genetic cascade was reset, and the whole gap gene sequence was reproduced. Anterior to the SAZ, the already formed gap genes expressions were erased [10] (Fig 4H). The 2 different responses of the gap gene network to the same perturbation within and outside of the SAZ indicates that gap genes are regulated by 2 different genetic programs: one within the SAZ and one outside and were suggested to be mediated by 2 different groups of enhancers active at the 2 regions [3,10]. This is in line with the observation that the gap gene Kr in Drosophila is regulated by 2 enhancers, one initiates its expression and the other maintains it [52]. Interestingly, open chromatin at early acting enhancers in the Drosophila embryo has been shown to be mediated by the pioneer factor Zelda and at late-acting enhancers by the pioneer factor Opa [143,144]. One could imagine then a scenario in which a gradual switch between early and late-acting enhancers in Tribolium to be mediate by opposing gradients of opa [13] and zelda [145], where the expressions of opa and/or zelda are regulated by (or interact with) the Wnt/Cad gradient. Interestingly, in vertebrates, it was shown that some enhancers and/or genetic programs mediate the initiation of segmentation clock waves posteriorly, and others mediate their anterior expressions [202–204]. In an alternative hypothesis, the period of the segmentation clock could be spatially modulated by a protein production time delay that increases along AP axis [205]. In yet another hypothesis, the period of the segmentation clock is regulated by the intercellular coupling delay. In vertebrates, individual PSM cells form a system of phase-coupled oscillators whose synchronization is mediated by Delta–Notch signaling [159–161]. The strength of this coupling has been proposed to modulate the period of collective oscillations [206,207], and, therefore, a gradient of coupling strength could mediate the observed frequency gradient. In zebrafish, disrupting Notch signaling either genetically or pharmacologically leads to a moderate increase in somitogenesis and clock period [207]. In Lfng mutant mice, the coupling delay is shortened, and oscillatory period is concomitantly decreased in intact PSM tissue but not isolated PSM cells [206]. Consistently, primary mouse PSM cells and in vitro–derived human PSM cells both maintain their oscillatory period when cultured at low densities such that no cell–cell signaling can take place [184,206]. The HES7 oscillatory period is unchanged even when isolated cells are treated with Notch inhibitors or obtained from Lfng mutant mice [184,206]. These results indicate that the segmentation clock pacemaker acts in a cell autonomous way, but the collective period can be modulated in the tissue context through cell–cell coupling.

[END]

[1] Url: https://journals.plos.org/plosgenetics/article?id=10.1371/journal.pgen.1009812

(C) Plos One. "Accelerating the publication of peer-reviewed science."
Licensed under Creative Commons Attribution (CC BY 4.0)
URL: https://creativecommons.org/licenses/by/4.0/


via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/