SEARCH

SEARCH BY CITATION

Keywords:

  • zebrafish;
  • polarity;
  • Hh;
  • neurulation;
  • morphogenesis

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES
  9. Supporting Information

We investigated the role of hedgehog (Hh) signalling on zebrafish neurulation, focusing on the intimate relationship between neurogenesis and morphogenesis during the neural keel stage. Through the analyses of Hh loss- and gain-of-function phenotypes, we found that Hh signalling controls the neural keel morphogenesis. To investigate underlying mechanisms, we examined cellular elongation polarity in the neural keel of Hh loss- and gain-of-function phenotypes and compared this with the deficient phenotype of a planar cell polarity (PCP) molecule, Trilobite/Strabismus. We found that Hh signalling controls cell elongation polarity of the neuroepithelium at least in part by means of PCP pathway; however, its effects are not strong enough per se to affect keel morphogenesis; instead Hh signalling mainly controls keel morphogenesis by means of affecting both medial and lateral neurogenesis. We devised a method for precise evaluation of neurogenesis in loss- and gain-of-Hh phenotypes that compensates for its delay caused by disturbed morphogenesis. We present a model that Hh signalling exerts level-dependent and binary-opposite effects on medial neurogenesis, whose modification to explain lateral neurogenesis reveals regional differences of underlying mechanisms between the two proneural domains. Such differences seem to be created in part by regional effector signalling; the effects of high Hh-signalling on medial neurogenesis can be reversed in accordance to medial Tri/Stbm level, in a polarity independent manner. Developmental Dynamics 235:978–997, 2006. © 2006 Wiley-Liss, Inc.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES
  9. Supporting Information

Secondary Neurulation

Neurulation is a morphogenetic process in vertebrate embryos that follows gastrulation and leads to the formation of neural tube (Schmitz et al.,1993). Primary neurons are born before morphogenesis of the neural anlage completes. The final divisions of primary motoneuron progenitor cells start to be observed at division 16, i.e., 10–12 hr postfertilization (hpf) stage (Myers et al.,1986; Kimmel et al.,1994). This is the time when gastrulation is completed and epiblast cells mediolaterally intercalate with each other and form either a single- or multilayered thin neuroectoderm, the neural plate. Two modes of neurulation are known that convert the neural plate into a tube. In many vertebrates, the neural tube is formed by folding of the plate (primary neurulation). In contrast, neurulation in teleost fish differs from primary neurulation in several respects, where the neurocoel, the lumen of the neural tube, is formed secondarily by cavitation of the neural anlage (secondary neurulation; Schmitz et al.,1993). An interesting aspect in secondary neurulation in contrast to primary is the unique sinking-down movement of the medial thickening as the two lateral halves of the neural plate converge toward the midline, to form the neural keel (Schmitz et al.,1993). As in primary neurulation, the mediolateral extent of the neural plate correlates with the dorsoventral extent of the neural tube (Hartenstein,1989; Papan and Campos-Ortega,1994). During the early neurulation stage, progeny from a single clone undergo morphogenetic behaviors, separating themselves along the anterior–posterior (AP) axis, and cell cycles are found to be coordinated (Kimmel et al.,1994). This finding indicates that neurogenesis and morphogenesis of the neural anlage may be closely linked.

Four Steps of Primary Neurogenesis

In zebrafish, three classes of primary neurons are formed in the neural plate. Primary motoneurons (PMNs, Fig. 1A), interneurons, and Rohon–Beard mechanosensory neurons (RBs, Fig. 1A) are formed in medial, intermediate, and lateral longitudinal proneural columns, respectively. The mediolateral positions of these three proneural columns is spatially prepatterned and interrupted by inter-proneural domains, suggested to serve as a reservoir for later neurogenesis (Bally-Cuif and Hammerschmidt,2003; Hans et al.,2004; Bae et al.,2005). Primary neurogenesis within the proneural domains can be subdivided into four processes, by analogy with neural induction (Wilson and Edlund,2001): selection, specification, determination, and differentiation (Fig. 1B). The selection process is regulated by lateral inhibition by means of NOTCH signalling that singles out a population of cells for neurons from equivalent progenitors in the proneural domains. Selected progenitor cells for neurons are allowed to express proneural genes, i.e., neurogenin1 (neurog1; Blader et al.,1997) and enter the second step, a labile phase called specification, and become biased toward neuron. At this step, the commitment to a neuron fate can be still reversed by signals that repress a neuron fate, i.e., her4, encoding a zebrafish basic helix-loop-helix protein of the HAIRY-Enhancer of split (E(SPL)) family, a target of NOTCH signalling (Take et al.,1999). Importantly, the specification marker, neurog1 is also under control of local cues, such as Hh signalling (Blader et al.,1997). The third step establishes the commitment of progenitors to a neuron fate even in the presence of signals that repress a neuron fate, allowing expression of early pan-neuron markers. Elavl3 and its family members are the earliest neuron markers, whose mRNA expression are seen in the mother progenitor cells in cycle (Marusich et al.,1994; Kim et al.,1996) and ensure progenitor cells to exit the cell cycle upon the next cell division. The last phase, differentiation into a specific functional type of neurons, accompanies expression of several new genes and requires by far more time than the aforementioned three steps for completion. The differentiation phase proceeds under the strong influence of regional differences (mediolaterally in the neural plate/keel stages). To focus on the local differences in primary neurogenesis, it would be convenient to compare the same set of gene expressions, each representing a particular phase of the primary neurogenesis. In this study, we followed two markers, one for neuronal determination (establishment of neuronal commitment) and another for the onset of early neuronal differentiation, elv3, and islet1 (isl1, a LIM homeobox gene; Korzh et al.,1993; Inoue et al.,1994), respectively.

thumbnail image

Figure 1. Shh overexpression affects gastrulation and neurulation. A: Medial and lateral proneural columns in the neural keel. Dorsal view (upper) and transverse section (lower) of the developing neuroectoderm to form the neural keel in the future spinal cord region (11.33–14 hpf, 4–10 somite stage). Medial proneural domains (two orange longitudinal columns in the middle of the dorsal view, upper) in which primary motoneurons (pink cells, isl1-positive) are formed, lie intervened by a single row of medial floor plate cells direct dorsal to the axial mesoderm (notochord, yellow) and appear as a medial thickenings in transverse section (lower). Both Shh and Twhh are expressed in the medial floor plate. Adaxial mesodermal cells (adM, yellow green) that give rise to future slow muscle lie adjacent to notochord and express basic helix–loop–helix (bHLH) myogenic factor, myoD, which is under positive control of Hh signalling. Lateral-most proneural columns (blue longitudinal stripes) produce Rohon–Beard sensory neurons (isl1-positive, blue cells) and foxd3-positive premigratory neural crest cells and appear as lateral thickenings in the transverse section (lower). As development progress, cells in the neuroectoderm converge toward the midline (two facing horizontal arrows) and bilateral thickenings fuse with medial structure. Mediolateral extent of the neural keel corresponds to ventrodorsal extent in the neural tube. Normal progress of this morphogenetic process requires at least in part a planar cell polarity gene, trilobite/strabismus. B: Four steps of primary neurogenesis: selection, specification, determination, and differentiation. The term commitment refers to both specification (reversible to selected state) and irreversible determination steps; hence, we call the latter process as the establishment of neuronal commitment as well. elv3 and isl1 mRNA expressions mark the establishment of neuronal commitment (determination) and the onset of differentiation process, respectively. elv3 mRNA expression persists after the onset of isl1 expression within the temporal range of our analyses during the neural keel stage. C: Gastrulae from wild-type (WT, left column) and shh mRNA-injected WT embryos (right column) were fixed at the 75%, 80%, and 90% epiboly stages (top, middle, and bottom, respectively) and overstained with ntl and elv3 to illuminate the deep cells. The animal pole is up, and the vegetal pole is down. Margin of the germ ring proceeds from animal (top) to vegetal (bottom). shh mRNA injection retarded dorsalward convergence movements at the close vicinity of the dorsal midline (black arrowhead), whereas ventral and lateral margin of the germ ring appear to move normally toward the vegetal pole (white arrowhead). Note that, in shh embryos, the notochord is received on one side of the embryo. D: Uninjected control embryos (WT) at the 9-somite stage (left) and shh embryos at the 4- to 5-somite stage (right). Both embryos are obtained from the same crossing and fixed at the same incubation time and processed to isl1(blue)/myoD(red) in situ hybridization followed by overstaining of β-catenin immunohistochemistry (dark brown). shh embryos resemble “trilobite”, with short anteroposterior axis and wider mediolateral extent, with a large opening of deep cells covered with a sheet of enveloping layer (EVL). Anteriormost isl1 staining in the hatching gland rudiment (polster) underlies the forebrain in uninjected controls (left); however, in shh embryos, polster was observed underneath the prechordal plate (right). E: The same pair of embryos in D were shown in the transverse sections at the level of the fourth and fifth somite. In shh embryos (lower), two halves of neural anlage did not come to fuse medially, instead, they were tethered with a single layer of EVL (black arrowhead), in good contrast to the uninjected control (upper). Nuclei were counterstained with hematoxylin (blue). F: Transverse sections at the 25-hours postfertilization (hpf) stage of uninjected control embryo (F1) and shh embryos (F2–4). Embryos are stained with isl1 (blue) and counterstained with hematoxylin/truidine blue for nuclei (light blue). F2–4 are all obtained from the same shh embryo and set out from the anterior to posterior direction. The apical and basal surfaces of the neural tube are delineated with white and brown dotted lines, respectively. G: Asymmetric shh mRNA distribution caused curved body axis toward much wider side (affected side), suggesting convergence and concomitant extension movements as an underlying mechanism to shape the neural anlage. In C and D, different focuses of planes are merged for presentation. Orientation in D and G, anterior is up; in E and F, dorsal is up; epi, epidermis; not., notochord; MLD, mediolateral distance. Scale bars = 200 μm in C, 50 μm in E,F.

Download figure to PowerPoint

Hh Ligands Are Ventralizing Morphogens

In the context of temporal and spatial patterning of primary neurogenesis, ventral motoneurons have been the most extensively studied in zebrafish. In chick and mouse midline tissues, axial mesodermal tissue of notochord and floor plate cells in the neuroectoderm secrete Shh that establishes distinct precursor domains, including those for motoneurons, in a concentration-dependent manner along the dorsoventral axis in the ventral spinal cord (Jessell,2000). Three hedgehog genes are found in zebrafish: shh (sonic hedgehog), twhh (tiggy winkle hedgehog), and ehh (echidna hedgehog). Consistent with a central role of Hh signalling in the pattern formation of ventral neuron classes in chick and mouse, loss-of-function analyses revealed that Hh signalling is required for induction of PMNs in zebrafish (Beattie et al.,1997; Lewis and Eisen,2001). In contrast, gain-of-function analyses showed that Hh signalling is not sufficient to induce supernumerary PMNs (Cheesman et al.,2004), although overexpression of the Hh regulated transcription factors Nkx6.1 and Olig2 is sufficient for the induction of supernumerary PMNs (Park et al.,2002; Cheesman et al.,2004). These observations raise the possibility that zebrafish PMNs may be induced by a different mechanism from that responsible in chick and mouse, where induction of motoneurons occurs after neural tube closure. For this reason, we focused our analyses on the early neurogenesis in the spinal cord at neural keel stage (4–10 somite stages), when morphogenesis of the neural anlage and neurogenesis proceed concomitantly. In this study, we demonstrate that neural keel morphogenesis regulates the progression of trunk neurogenesis and vice versa. First, we show that either Hh gain- or loss-of-function affects cell elongation polarity in the neural keel. Unlike the deficient phenotype in planar cell polarity (PCP) pathway, polarity defects in gain- and loss-of-Hh signalling were not strong enough to cause deficient morphogenesis. We show that retarded keel morphogenesis in gain- and loss-of-Hh phenotypes are caused through its effects not on cell polarity but on neurogenesis. Because neurogenesis and morphogenesis affect each other, we developed a reliable method to evaluate the progression of neurogenesis, mediolateral (MLD) compensation. This new method (MLD compensation) disclosed several key new mechanisms of Hh function on medial and lateral neurogenesis. We propose models wherein the effects of Hh signalling are determined by location (medial or lateral) and fates (neuron or neural crest) of the target cells and can be reversed according to its signalling level. In the end, we show an example of regional effector signalling that modifies the effects of Hhs on medial neurogenesis.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES
  9. Supporting Information

Increased Hh Signalling Affects Cell Movements

To establish a basis for our analysis of Hh function, we first analyzed gain-of-function phenotypes by injecting shh mRNA into one-cell stage wild-type (WT) embryos. Unexpectedly shh-injected embryos showed severe defects in late gastrulation movements (Fig. 1C). In many cases, only one side of the dorsal deep cells near, not just at, the dorsal midline, slowed dorsalward convergence movement and vegetalward movement of marginal germ ring (Fig. 1C, black arrowheads). Ventral and lateral deep cells at the margin of the germ ring moved normally toward the vegetal pole (Fig. 1C, white arrowheads). As a result, at the end of the gastrulation stage a large opening appeared in the deep cells directly lateral to the midline and separated completely two longitudinal medial proneural domains, while the notochordal primordium extended along one side of it. A bifurcated neural anlage along the midline is evident in shh embryos in early segmentation stages; embryos are anteroposteriorly short and mediolaterally wide, resembling “trilobite” (Fig. 1D). However, despite the abnormal morphology, both neurogenesis (isl1 expression) and segmentation (myoD) proceeds around the opening through neurulation. In shh-injected embryos the epithelial-like enveloping layer (EVL) that covers the deep cells seemed to complete epiboly normally, because EVL covered the whole embryo, as if tethering two halves of the developing neural anlage (Fig. 1E). Surprisingly, both halves of the neural plate with or without the midline structure, notochord, later developed into two independent neural tubes with neuralcoel-like cavitation, being in contact at the basal wall of the tubes (Fig. 1, F2). At the point of bifurcation, two tubes initially fused dorsally at the basal sides (Fig. 1, F3). As the two tubes come into fusion, they share the apical side (lumen) dorsally and form an inverted Y-shaped apical surface, zippering down toward the ventral side to fuse completely (Fig. 1, F4).

In addition, injection of shh mRNA into the two-cell stage embryos sometimes led to a curved body axis, instead of a straight midline, bent toward the wider side of the neural keel, presumably due to asymmetric distribution of the injected mRNA (Fig. 1G). This observation indicates that the convergence movements toward midline and concomitant anteroposterior extension movements serve as underlying forces that shape the neural keel as reported previously (Kimmel et al.,1994; Concha and Adams,1998). Taken together, our results suggest a unique possibility that Hh signalling targets morphogenetic cellular movements during early development from gastrulation through neurulation. Therefore, to explore the role of Hh signalling in the neural anlage morphogenesis, we next analyzed gain- and loss-of-Hh function phenotypes and studied the cellular basis of the morphogenesis with a special focus on elongation polarity of the individual cells in the transverse and horizontal planes.

Hh Signalling Controls Planar Polarity of the Cell Elongation in the Neural Keel

In the early Xenopus neural plate, cells lateral to the midline have medially biased lamelliform protrusions that make them move medially by means of mediolateral intercalations (Ezin et al.,2003). Although in zebrafish the lumen of the neural tube is formed secondarily by cavitation of the neural anlage, similar convergence movements toward the midline continue in the neurulation process as the neural plate thickens medially and sinks down to form the neural keel (Schmitz et al.,1993; Concha and Adams,1998). To ask whether Hh signalling affects converging morphogenetic movements of the neural anlage, we first investigated the polarity of cell elongation in the horizontal plane (planar polarity) and examined its relationship with the neural keel morphogenesis. For the analysis of planar elongation polarity, embryos were stained with isl1 for differentiating medial and lateral neurons and anti–β-catenin protein for cell contour and examined in dorsal views for deviation of angles between the line perpendicular to the midline and the major axes that go through two foci of the fitted ellipses of freehand individual cell outlines (Fig. 2A). To evaluate the effects on neural keel morphogenesis, we measured mediolateral distance (MLD) between the midline and the lateral-most edge of the neuroectoderm as a mean value of five equally spaced measurements along the length of the trunk region, as identified by myoD-positive adaxial mesoderm (inset of Fig. 2F).

thumbnail image

Figure 2. Hh signalling controls elongation polarity of the cells in the neural keel. Embryos fixed between the 4- to 10-somite stages were stained with isl1 (blue)/myoD (red) in situ hybridization followed by β-catenin immunohistochemistry (IHC; overstained in brown). A: Planar cell elongation polarity was analyzed in the dorsal view of the embryos in flat preparation (removed from the yolk cells) at the focal plane that cover the whole neural keel from the medial to the lateralmost regions. An ellipse was fitted to an individual cell, and the angle (θ) between each major axis of the cell and the dorsal midline was measured (left and middle panels). The cumulative percentage of cells was calculated in the neural keel as a function of planar cell elongation polarity estimated by θ (right panel). MO-tri embryos (green line) are the most affected phenotype, far away from the line for uninjected controls (wild-type [WT], gray line). All the other phenotypes lie between uninjected WT controls and MO-tri embryos: shh (pink line), smu homozygotes (blue), MO-tri/shh embryos (yellow orange), and hsp;shh embryos (brown line). B: Transverse sections of WT embryos (5- to 8-somite stages) are shown in the middle panel. Embryos were stained with isl1 (dark blue)/myoD (red; lost during embedding) followed by β-catenin immunohistochemistry and counterstaining of the nuclei with truidine blue (turquoise). Apicobasal cell polarity analyses were used with both left and right wings of the keel. The right half of the data were flipped horizontally, pooled with the left, and shown in the orientation of the left wing (left). Average orientations of apicobasal axes relative to the midline are 29.7° (5 som), 64.3° (6 som), 21.5° (7 som), and 13.0° (8 som). Examples of freehand outlines used for the analyses are shown on the right panel. The intervals of concentric half-rings centered at the medial floor plate are 10 μm. The neural keel sinks at the rate of 20 μm/ hr. Lateral isl1-RB neurons are found outside of the outermost ring with a radius of 60 μm. C,D: Angular deviation (circular standard deviation) of planar (C) and apicobasal (D) cell elongation polarity. From left to right: WT (gray), smu (blue), shh (pink), MO-tri/shh (yellow orange), hsp;shh (brown), and MO-tri (green). E: Summary of the circular homogeneity test between two groups. Results from planar polarity are shown in the upper half (black) and apicobasal in the lower half (white background). Asterisks are plotted at the intersection of orthogonally arranged pairs of groups, if Watson's two-sample test for circular homogeneity at the significance level (α) of 0.05 rejected a null hypothesis that distribution of circular data between two groups are identical. F: Mediolateral distance (MLD) becomes narrower (smaller) as embryos develop. MLD is calculated as a mean length (μm) of five equally spaced measurements on either side of embryos along the anterior–posterior length of the trunk region, as shown in the left photo with five white bars. Measurements of MLDs from various stages during the neural keel stage (11–14 hpf) were collected with WT, smu homozygotes, shh mRNA-injected embryos, MO-tri/shh embryos, hsp;shh embryos and MO-tri embryos. To focus on the MLD-difference from uninjected WT controls, MLD data from each embryo were subtracted with the mean value of uninjected control WT embryos at the identical developmental stage with the same number of somites. Subtracted MLD values (ΔWT in μm) for each group were pooled and shown as a dot plot with mean (thick line) and standard deviation (SD, thin lines for upper and lower limits) on the direct right. P values from statistical analysis by Tukey's multiple comparison test after one-way analysis of variance are shown with symbols: ***, P < 0.001; *, P < 0.05; and n.s. (not significant), P > 0.05. Note that MLDs in hsp;shh embryos showed no significant difference from WT embryos, despite their defects in both planar and apicobasal cell elongation polarities. Sample size: wild-type (n = 88); smu (n = 38); shh mRNA (n = 87); MO-tri/shh (n = 34); hsp;shh (n = 30) MO-tri/stbm (n = 76).

Download figure to PowerPoint

To verify our analyses of planar polarity, we first examined the deficient phenotype in PCP pathway upon injection of antisense morpholino against trilobite/ strabismus (MO-tri; Park and Moon,2002). As reported, MO-tri injection triggered mediolateral expansion of both the developing neuroectoderm and the paraxial mesoderm directly beneath it. Average MLD values during the neural keel stage (11–14 hpf) in MO-tri embryos were significantly wider, by 63.5 ± 34.3 μm (mean; thick bar next to dot plots in Fig. 2F) ± SD (thin upper and lower bars), than those in uninjected controls (178% of the uninjected counterparts; P < 0.001 in Tukey's multiple comparison test after analysis of variance [ANOVA]; Fig. 2F). In MO-tri embryos a large angular deviation of ± 33.8° was observed, significantly different from WT (Fig. 2, right panel in A and C6; Watson's test for circular homogeneity [α = 0.05], summarized in Fig. 2E).

Next, we examined the effects of overexpression of Hh ligands on planar polarity of cell elongation. In all cases, the effective overexpression of Hh ligands was verified by the lateral expansion of myoD-positive adaxial mesodermal cells (Hammerschmidt et al.,1996). As suggested from the severe defects in normal neurulation (Fig. 1C), shh mRNA-injected embryos showed significant defects in planar elongation polarity with a larger angular deviation of ± 25.9° than ± 19.3° in uninjected control embryos (Fig. 2, right panel in A and C3). MLD values in shh embryos were wider by 39.3 ± 29.8 μm (mean ± SD) than in uninjected controls (138% of the uninjected counterparts, Fig. 2F; P < 0.001 in Tukey's multiple comparison test after ANOVA). These results suggest that Hh ligands affect planar polarity during the neural keel stage.

To directly test for a requirement of Hh signalling in the planar elongation polarity during neural keel stage, we analyzed the loss-of-function Hh mutant smu. Homozygous smu mutants showed almost the same degree of defects in planar polarity as its gain-of-function counterpart of Hh signalling with angular deviation of ± 26.4° for smu homozygotes against ± 19.3° in uninjected controls (Fig. 2, right panel in A and C2; significant difference in Watson's test for circular homogeneity (α = 0.05), summarized in Fig. 2E). Consistent with the defects in planar polarity, in smu homozygotes, convergence movements of the keel were severely affected with significantly wider MLD values than WT (wider by 46.3 ± 29.3 μm, mean ± SD; 146% of WT counterparts at the same somite stage; Fig. 2F; P < 0.001 in Tukey's multiple comparison test after ANOVA).

These observations suggest one interesting possibility that Hh ligands may function as extracellular polarizing signals that direct cells along the mediolateral axis over the ranges, as suggested with Wnt11 and Wnt5a in zebrafish PCP pathway (Heisenberg et al.,2000; Kilian et al.,2003). If Shh functions as such a signal, excess amounts of extracellular polarizing signals should rescue the planar polarity defects in MO-tri embryos by compensating the decreased level of PCP signalling. To test this idea, we co-injected shh mRNA with MO-tri into WT embryos. Although the planar cell polarity defects in MO-tri embryos were not completely rescued, the angular deviation was reduced to a level equivalent to embryos with gain- or loss-of-Hh signalling (± 25.4° in MO-tri/shh embryos against ± 33.8° in MO-tri embryos, Fig. 2, right panel in A and C4), with no significant difference between MO-tri/shh embryos and shh embryos and with significant difference between MO-tri/shh embryos and MO-tri embryos in Watson's test for circular homogeneity (α = 0.05, summarized in Fig. 2E). Defects in convergence movements of the neural keel in MO-tri embryos were also partially rescued after co-injection of shh mRNA to a level equivalent to shh mRNA-injected embryos (no significant difference [P > 0.05] between MLDs from MO-tri/shh and shh embryos in Tukey's multiple comparison test after ANOVA) with narrower MLD values (47.7 ± 32.4 μm wider than and 150% of the uninjected controls) than MO-tri embryos (P < 0.05 between MLDs from MO-tri/shh and MO-tri embryos in Tukey's test in Fig. 2F). Thus, this result demonstrated that Hh ligands function as extracellular polarizing signals during gastrulation through early neurulation stages. Rescued cell elongation polarity in MO-tri embryos upon shh mRNA injection suggests Hh signalling acts at least in part by means of PCP pathway, because the effects on the polarity phenotype were not additive, although we cannot exclude the possibility that Hh can act by means of an independent pathway from PCP signalling.

The severe defects in convergence movements and epiboly of the deep cells in shh mRNA-injected embryos (Fig. 1C), at the vicinity of the dorsal midline during gastrulation, make it obscure whether Hh ligands directly affected the neural keel neuroepithelium with planar elongation polarity. To ensure overexpression of Shh later than gastrulation stage, we used GAL4-UAS transgenic lines, where transactivation of the shh gene is under the control of heatshock-inducible hsp70 promoter (Scheer et al.,2002). We applied two times of heatshocks to embryos from a cross between hsp70::gal4/+ and UAS:shh/+ at the 60% and 90% epiboly stage (as shown in Fig. 3B), which resulted in misexpression of Shh in a quarter population (later referred to as hsp;shh) of the offspring from the cross. With this condition of heatshock, all the offspring from the cross completed gastrulation quite normally without any morphological defects until the end of the neural keel stage, in well contrast to shh mRNA-injected embryos. In hsp;shh embryos, the planar polarity was similarly affected (angular deviation ± 29.5°) as in shh mRNA-injected embryos (Fig. 2A, right panel and C5; Fig. 2E, no significant difference between shh mRNA-injected embryos and hsp;shh embryos), thus further confirming the direct effects of Hh ligands on neuroepithelium during the neural keel stage. However, oddly enough, MLD values from hsp;shh embryos were completely normal (2.2 ± 16.9 μm wider than and 102% of the WT counterparts; Fig. 2F, no significant difference [P > 0.05] between MLDs from WT and hsp;shh in Tukey's test after ANOVA). For this reason, we examined individual cell polarity in the neuroepithelium literally from a different angle, in the transverse plane of the neural keel for apicobasal polarity.

thumbnail image

Figure 3. Late hedgehog (Hh) overexpression acts on Rohon–Beard mechanosensory neurons (RB) neurogenesis with a time lag. A: Transverse sections of embryos from the 25–30 hours postfertilization (hpf) stage. Uninjected control wild-type (WT) embryos (left column), hsp;shh embryos (middle), and shh embryos (right) are in situ stained with isl1 (upper row) or lim3 (bottom row). Dorsal is up, and ventral is down. isl1 expression is seen in both dorsal RB neurons and ventral motoneurons (primary and secondary). In contrast, lim3 expression is limited to ventral neurons (both motoneurons and some ventral interneurons). Hh gain-of-function embryos showed dorsal expansion of ventral neuron fates. B: RB neurogenesis can be subdivided into two phases at 18 hpf stage according to the rate of formation. The first and second phases appear to coincide with the 16th and 17th cell division. Two heatshocks (HS) to transactivate the shh gene in hsp;shh embryos resulted in the emergence of supernumerary isl1-RBs no sooner than 15.5 hpf (the 13-somite stage). C: (Upper) Confocal microscopic analyses of mitotic index with phosphorylated histone H3 protein (H3P) show broad waves of mitotic phases due to loose synchronization of the cycles. Each cell cycle ends with the completion of the cell division (M15, M16, and M17; filled region on the right side of each rhomboid). The results were a mixture of different levels of trunk neural anlage along both anteroposterior and ventrodorsal axes. Embryos from 12–18 hpf were pulsed with bromodeoxyuridine (BrdU) then immediately fixed every 30–60 min (total 23 embryos), processed with immunohistochemistry for BrdU and H3P, and H3P-positve/BrdU-negative cells were counted and shown. As to the mitotic index, no significant difference was observed among regions (data not shown, P > 0.05 ANOVA). (Bottom) Models for two possibilities to explain 8.5 hr of time lag between the first pulse of HS and the emergence of supernumerary isl1-RBs. The periods of 15th and 16th cell cycle are shown as rectangular frames. D: RB markers other than isl1 (isl2, tlx3b, and HuC-IHC) confirmed the formation of supernumerary RBs after conditional Shh overexpression. E: HNK-1 immunohistochemistry for WT (left) and hsp;shh (right) embryos to examine morphological features of RB neurons. hsp;shh embryos showed severe defects in fasciculation of dorsal lateral fasciculus (DLF); however, both central and peripheral axons of RB neurons appeared normal. F: Examples of freehand outlines of dorsal view of WT (upper) and hsp;shh embryos (bottom) at the plane of focus for isl1-RBs. Anterior is left, and posterior is to the right. The numbers on the left side indicate the numbers of isl1-RBs on either left or right side of the embryos. Scale bar = 50 μm in A.

Download figure to PowerPoint

Hh Signalling Controls Apicobasal Polarity of the Cell Elongation in the Neural Keel

Neuroepithelium during the neural keel stage is not only in the horizontal but also in the transverse plane (apicobasally) well polarized, such that individual cells in one wing of the keel are oriented to elongate in parallel to the apicobasal axis in a transverse plane, irrespective of the position of the cells (Fig. 2B). In the anterior trunk regions as the keel sinks down to form the rod, at a constant rate of ca. 20 μm/hr, the axes of polarized neuroepithelial cells in the transverse plane approaches the horizontal plane and become oriented perpendicular to the future dorsoventral axis of the neural tube. Through neural anlage morphogenesis (5–8 somite stages), apicobasal elongation polarity is maintained, with the average angular deviation of ± 21.4° in the anterior trunk region (4 embryos, 215 cells).

We first analyzed apicobasal polarity in MO-tri embryos that show severe planar polarity defects. Interestingly, both planar and apicobasal polarity of cell elongation were severely affected in MO-tri embryos with large angular deviations up to ± 38.1° (Fig. 2, D6; Fig. 2E, significant difference between uninjected controls [WT] and MO-tri embryos in Watson's test for circular homogeneity [α = 0.05]). As is the case with this, gain-of-function Hh phenotypes in hsp;shh embryos showed not only disturbed planar but also disturbed apicobasal polarity with angular deviations of ± 30.4° (Fig. 2, D5, and Fig. 2E, significant difference between WT and hsp;shh in Watson's test for circular homogeneity (α = 0.05)). In good contrast to this finding, GAL4/UAS-mediated gain-of-function Hh phenotype, both shh mRNA-injected gain-of-function and loss-of-function Hh mutant smu showed normal apicobasal polarity with angular deviations of ± 24.4° and ± 20.4°, respectively (Fig. 2, D2 and D3; Fig. 2E, no significant differences to WT in Watson's test for circular homogeneity). Although our results above do not provide any information on the possible interrelationship between planar and apicobasal polarity, shh mRNA-injected and smu phenotype suggest two polarities can be independently regulated.

To test whether Hh ligands act as extracellular polarizing signals to orient individual cells of the neural keel along the apical–basal axis, we again examined whether Shh rescues the apicobasal polarity defects in MO-tri embryos. Although apicobasal polarity of cell elongation in MO-tri/shh embryos was not completely rescued, severe polarity defects in MO-tri embryos were significantly rescued with reduced angular deviations of ± 23.8° against ± 38.1° in MO-tri embryos (Fig. 2, D4, and Fig. 2E, significant difference between MO-tri/shh and MO-tri embryos and no significant difference between MO-tri/shh embryos and uninjected controls [WT] in Watson's test for circular homogeneity [α = 0.05]). Thus, as seen with planar polarity of the cell elongation, Hh ligands act as extracellular polarizing signals in the control of apicobasal polarity of the neural keel.

Taken together with the results on planar and apicobasal polarity, our results suggest that Hh ligands act as polarizing signals that control planar and apicobasal neuroepithelium elongation during the neural keel stages. However, one of the observation above clearly denies the possibility that Hh ligands control the neural keel morphogenesis by means of their control over the polarity; in hsp;shh embryos the keel morphogenesis is completely normal (Fig. 2F), despite clear defects in both planar and apicobasal polarity (Fig. 2E). It would be simple if reduced control over the polarity to orient individual cells along each axis could explain the slowed keel morphogenesis in Hh phenotypes. Indeed, this seems to be the case with the results with deficient phenotype of planar cell polarity (PCP) pathway as seen in MO-tri embryos, because the rescue in planar and apicobasal polarity led a rescue in the keel morphogenesis (Fig. 2F). These results, thus, indicate the presence of other targets whereby Hh signalling exerts its effects on the neural keel morphogenesis.

Hh Overexpression Induces Extra RBs and Motoneurons

To obtain a definitive answer for two questions, why late Hh overexpression in hsp;shh embryos showed no morphological defects during the neural keel stages and how early-Hh signalling affects the keel morphogenesis by means of a cascade other than cell polarity, we reoriented our study by asking whether the later morphogenesis of the neural anlage in hsp;shh embryos was affected through segmentation stages.

Ventral neurons.

Because loss-of-function analyses revealed that Hh signalling is required for induction of PMNs in zebrafish (Beattie et al.,1997; Lewis and Eisen,2001) and overexpression of Hh-regulated transcription factors Nkx6.1 and Olig2 is sufficient for the induction of supernumerary PMNs (Park et al.,2002; Cheesman et al.,2004), we first examined the ventral spinal cord of shh mRNA-injected embryos and hsp;shh embryos. With hsp;shh embryos, we added two times of heatshocks during the gastrulation stage at 60% (7 hpf) and 90% epiboly (9 hpf) stages to ensure robust Shh overexpression (Fig. 3B). Consistent with the idea that Hh signalling ventralizes the neural tube, both in shh- and hsp;shh embryos, we observed significant dorsal expansion of ventral neurons marked with either isl1 or lim3, a LIM homeobox gene whose expression is limited to the ventral neurons (Glasgow et al.,1997; Fig. 3A). At these stages, the neural tube of hsp;shh embryos showed a large cross-sectional area in the transverse plane, in average 180% of WT embryos (2,317 ± 154 μm2 [n = 12] against 1,286 ± 61 μm2 in WT [n = 18]; P < 0.0001), suggesting clear defects in the normal tube morphogenesis. From the 10-somite stage (14 hpf) on, however, another population of later ventral neurons other than PMNs, secondary motoneurons, begin to be observed and made it difficult to evaluate the effects of Hh overexpression on PMNs on the basis of cell counting. Therefore, to keep track of the neurogenesis from the plate through the tube stages, we chose another neuron class, the dorsal mechanosensory primary neuron (RB neurons) as an index to evaluate the progression of the trunk neurogenesis. Although to date any possible role of Hh signalling in the dorsal part of the trunk neurogenesis has not been elucidated, misexpression of enormous amount of Shh protein might affect, most probably suppress, the dorsal neurogenesis by ventralizing the neural tube.

RB neurogenesis.

We first established a temporal map of RB neuron formation by following isl1 gene expression in the lateral proneural domains as a marker for early RB differentiation (Fig. 3B). isl1 expression is seen from as early as the neural plate stage of 10 hpf on, when presumptive primary motoneurons are induced to form medially. In wild-type embryos, we found RB neurogenesis can be subdivided into two phases according to the rate of induction (Fig. 3B). In the first phase of RB formation, isl1-positive RBs follow a linear increase at a constant rate of 8.9 ± 0.3 cells per side every hour until the end of the 18-somite stage (18 hpf, r2 = 0.7922, n = 104; Fig. 3B). The initial isl1-RB number at the 4-somite stage (11.3 hpf), approximately 30 cells per side, doubles until the 10-somite stage (14 hpf), and finally 94 isl1-RBs per side are formed at the 18-somite stage (18 hpf). In the second phase, from the 18-somite stage (18 hpf) until the end of segmentation (24–26 hpf), RB formation slowed at one-tenth of the initial rate (0.88 ± 0.3 isl1-RBs per side every hour) and 98.7 ± 8.0 isl1-RBs per side are found in the end (n = 72). Consistent with the observation that a single lineage extend themselves into a long, discontinuous strings of cells (Kimmel et al.,1994), an application of a single pulse of 5-bromo-2′-deoxy-uridine (BrdU) during the stages from 75% epiboly to the 14-somite stage, to label proliferating progenitors in S-phase, yielded uniform distribution of BrdU-positive isl1-RBs along the length of neural tube (Supplementary Figure S1, which can be viewed at http://www.interscience.wiley.com/jpages/1058-8388/suppmat). This finding indicates that new RB neurons are formed in between the old RBs and not in an unidirectional manner from anterior to posterior. Confocal stack analysis of the number of mitotic cells in the neural anlage, by immunohistochemistry with phosphorylated histone H3 protein (H3P), revealed that the first and the second phase of RB neurogenesis seem to correlate with division 16th and 17th (Kimmel et al.,1994; Geldmacher-Voss et al.,2003), respectively (Fig. 3C, upper).

Next, we analyzed the late phenotype in the neural tube in hsp;shh embryos. During the neural keel stage when no morphological defects were found in hsp;shh embryos, the number of isl1-RBs was normal until the end of the 12-somite stage (15–15.5 hpf; Fig. 3B). However, at the 13-somite stage (15.5 hpf) hsp;shh embryos showed increased number of isl1-RBs, in average 126% of the WT counterpart (87.3 ± 21.9 cells/side [n = 8] against 69.2 cells/side in WT, Fig. 3B). The increased isl1-RBs were consistently observed later than this stage. In the late segmentation stages (24–26 hpf), an average of 146% more isl1-RBs were observed in hsp;shh embryos (144.5 ± 3.9 cells/side [n = 6] against 98.7 ± 1.6 cells/side in WT), and the difference is significant (P < 0.0001; Fig. 3B). The shortened anteroposterior axis of the neural tube (Fig. 3F) indicates defects in convergence and extension movements that are necessary for the tube morphogenesis (Kimmel et al.,1994). BrdU pulse experiments showed that supernumerary isl1-RBs in hsp;shh embryos were mainly induced in the posterior regions of the tube (Supplementary Figure S1), suggesting a causal relationship between ectopic extra neurogenesis and morphogenetic defects of the tube.

Although supernumerary isl1-positive dorsal neurons in hsp;shh embryos have a strong resemblance to RB neurons in the position along the dorsoventral axis, these extra isl1 cells might be other neuron population than RB neurons, i.e., ectopic ventral neurons induced by misexpression of ventralizing signals. Therefore, we examined the expression of other markers that are known to mark RB neurons: isl2, tlx3b (Langenau et al.,2002) transcripts, and HuC protein expression (Fig. 3D). In WT embryos, isl1 and tlx3b gene expression mark the identical population of RB neurons, whereas isl2-RBs are limited to a part of them, starting to be observed from 18 hpf and rapidly growing in proportion in isl1+/tlx3b+-RBs (Supplementary Figure S2). In the late segmentation stages of hsp;shh embryos (20–25 hpf), the number of dorsal neurons with all three markers, isl2, tlx3b, and HuC, significantly increased (P < 0.0001 for all the three in two-tailed unpaired t-test; Fig. 3D). We next examined whether supernumerary-induced RBs retain morphological features of typical RBs by HNK-1 immunohistochemistry (Metcalfe et al.,1990), especially for two aspects: central longitudinal axons that extend into dorsal longitudinal fasciculus (DLF) and peripheral axons that exit neural tube dorsally and branch profusely underneath the skin (Bernhardt et al.,1990). Although hsp;shh embryos showed severe defects in fasciculation of DLF, both HNK-1–positive central and peripheral axons appeared normal (Fig. 3E). Thus, we conclude that the supernumerary dorsal neurons in hsp;shh embryos are RB neurons.

Do Hhs promote or suppress neuronal commitment?

Taken together, our results above seem to suggest neurogenesis is the target and the missing link of Hh signalling to control the neural anlage morphogenesis. However, one question remains with hsp;shh embryos; why does it take a long time lag of 8.5 hr after the heatshock (at 7 hpf) before extra isl1-RBs appear (at 15.5 hpf, Fig. 3B,C)? As shown in our previous report, heatshock-induced functional GAL4 protein is found to be expressed 1.5 hr after the heatshock (Scheer et al.,2002). The lateral expansion of myoD in adaxial mesoderm, a known target of Hh ligands, confirmed that functional Shh protein was induced no later than the three-somite stage (11 hpf). We still have 4.5 hr that require an explanation (Fig. 3C). Two possibilities can be drawn. In the first possibility (Fig. 3C, case 1), Hh signalling promotes neuronal commitment (either neuronal specification or determination process) of the progenitor cells, allowing cells to express the early pan-neuron marker elv3, and induces cell cycle exit no later than 11 hpf. If normally the initial differentiation, marked with isl1 expression, begins 4.5 hr after the onset of elv3 expression, the first wave of extra isl1-RBs would be observed at 15.5 hpf; thus, this model would explain the lag of 4.5 hr. In the second possibility (Fig. 3C, case 2), contrary to the first, Hh signalling works against neuronal commitment and promotes proliferation. In this model, a progenitor is not committed to RB neurons at the timing when usually it should be and forced further to go through another round of cell cycle before it is allowed to express elv3 and isl1 to give rise to RB neurons. This prediction will be supported, when one considers that the time period for one cell cycle at this developmental stage (the 16th cell cycle) is reported to be ca. 4 hr (Kimmel et al.,1994), which does not so much differ from the time lag of 4.5 hr that we need an explanation. In this second model, the time required for elv3-positive committed progenitors to gain isl1 expression for initiation of neuronal differentiation should be far less than 4.5 hr, well in contrast to the first model.

Thus, to understand the mechanisms whereby Hh signalling exerts its effects on RB neurogenesis, it is critical to analyze the timing of the two key steps during neurogenesis: elv3 and isl1 expression, representing the establishment of neuronal commitment, and initiation of neuronal differentiation, respectively.

MLD Is a Reliable Clock for the Neural Keel

To resolve neurogenesis both spatially and temporally into selected key steps, i.e., the establishment of neuronal commitment (elv3 expression) and the initiation of neuronal differentiation (isl1 expression), we chose the neural keel stage (4–10 somite stages). Since during the neural keel stage, unlike the neural tube stage, most of the trunk neurogenesis can be captured in one horizontal plane as spatially separated proneural stripes over the entire length of the neural anlage. Because either increased or decreased Hh-signalling affects both the keel morphogenesis (Fig. 2F) and neurogenesis (Fig. 3), we analyzed both processes as a function of time.

Plots from various stages of WT embryos showed a linear correlation between the neurogenetic (elv3 and isl1 expression) and morphogenetic index, mediolateral distance of the neural keel (MLD, inset of Fig. 2F), in both the medial and lateral columns that produces later ventral and dorsal neurons, respectively (Fig. 4). Furthermore, as the absolute age of the embryo become older with increasing number of somites, the mediolateral extent of the keel becomes narrower and a greater neurogenetic index was observed both medially and laterally. These observations suggest that neurogenesis is not only a function of the absolute age of the embryo but also can be a function of the neural keel morphogenesis.

thumbnail image

Figure 4. Mediolateral distance (MLD) is a reliable clock for the neural keel. Establishment of neuronal commitment (neuronal determination) is marked by elv3 and the subsequent onset of neuronal differentiation by isl1. A–D: The number of either elv3-positive (left) or isl1-positive (right) cells per side is plotted as a function of MLD of each embryo. Cell count in the medial (upper) and lateral (lower) proneural domains are separately shown. Color codes reflect the absolute developmental age of embryos estimated from the number of somite, as shown at the bottom. As the embryos get older and have more elv3 and isl1 cells, MLD becomes narrower in both medial and lateral proneural domains. Although MO-tri embryos are in the 8-somite stage, the number of neurons is fewer than those normally expected from uninjected controls with the same somite number. Plots from MO-tri embryos (gray triangles) appear to fit well into the regression line toward younger region of uninjected controls. E: A nomograph is presented for MLD compensation that converts MLD values into the time in hours postfertilization (hpf) specific to the development of neuroectoderm during the neural keel stages (4- to 10-somite stages in wild-type [WT]). The X-axis shows average MLD value of each embryos in length (μm), and the Y-axis shows the absolute age of embryos estimated from the number of the somite in time (hpf). We applied a one-phase exponential decay model with a fixed plateau of the curve (b in the formula in E) at 10 hpf as a time of establishment of the neural plate. The span of y (a in the formula in E; time range for neurogenesis) was fixed to 10 hr based on the observation with RB neurogenesis shown in Figure 3B. The best fit value (r2 = 0.6116) was obtained when a rate constant (k in the formula in E) is 0.01553 (n = 88). F,F′: In MO-tri embryos, lateral neuronal commitment (elv3 expression) is indistinguishable from WT when scaled in MLD time (F′); in contrast, the case in somite time (F). The regression lines (solid line) with 95% confidence intervals (region in between two dotted lines) are calculated from results with uninjected wild-type and plotted as a reference in both. F″: Each cell count data from F and F′ was subtracted with the mean calculated from uninjected WT controls at the identical stage estimated by either somite number or MLD, respectively, and expressed as the difference from uninjected controls (ΔWT).

Download figure to PowerPoint

To test this idea, we examined PCP pathway deficient phenotype in MO-tri embryos for the relationship between neurogenetic and morphogenetic indices. Because MO-tri embryos show a wider neural keel than uninjected WT embryos (Fig. 2F), if morphogenesis and neurogenesis show a linear correlation and neurogenesis itself is not impaired, proportionally lower neurogenetic index should be found in MO-tri embryos. In MO-tri embryos, we observed a normal correlation between their age (in hours postfertilization) and their somite number, which was comparable to uninjected controls, despite the widened trunk morphology. However, in support of the prediction above, the neurogenetic index was reduced both medially and laterally compared with uninjected embryos with the same somite number (Fig. 4F). This decreased neurogenetic index does not appear to be due to cell death, because we did not observe a significant increase in the number of TUNEL (terminal deoxynucleotidyl transferase–mediated deoxyuridinetriphosphate nick end-labeling) -positive apoptotic cells in the neural keel (data not shown). When the neurogenetic index was plotted with reference to the MLD of the keel (Fig. 4A–D), MO-tri embryos showed an indistinguishable degree of correlation between morphogenesis and neurogenesis from uninjected WT embryos corresponding to younger somite-stage uninjected control embryos, despite their average absolute age. These observations show that somitogenesis in the mesoderm and neurogenesis in the ectoderm can proceed independently.

This observation allowed us to develop a reliable method (MLD compensation) to analyze the neurogenetic index (elv3- and isl1-positive cells) in embryos with morphogenetic disorder (i.e., shh embryos), which enables us to discriminate “slowed, but normal” and “decreased and abnormal” neurogenesis with the same neurogenetic index. To calculate the developmental age of the neural keel from MLD values, we applied a one-phase exponential decay model with a fixed plateau of the curve (“b” in the formula; Fig. 4E) at 10 hpf as a time of establishment of the neural plate and the maximum time period of neurogenesis (“a” in the formula) fixed to 10 hr based on our observation with isl1-positive RB neurons (Fig. 3B).

By using this method, the developmental age of the neuroectoderm in hours postfertilization can be estimated without any information about the number of somites. One example is shown (Fig. 4 FF′). MO-tri embryos have fewer lateral elv3 neurons than uninjected counterparts at the same somite stages (solid line with a 95% confidence interval in Fig. 4F; left plots in Fig. 4F″), and this finding suggests abnormal neurogenesis. When embryos are staged according to the MLD values, it turned out that in MO-tri embryos establishment of neuronal commitment marked with elv3 expression is indistinguishable from uninjected control WT embryos (solid line with a 95% confidence interval in Fig. 4F′; right plots in Fig. 4F″), although progression of commitment was slowed.

Thus, MLD serves as a reliable “clock” when we estimate the effects of a factor that acts on both neurogenesis and morphogenesis. The MLD-based time axis enables us to compensate for the effects on trunk neurogenesis caused (secondarily) by effects on morphogenesis, which would be mixed up if neurogenesis were monitored by absolute embryonic age rather than MLD.

Hh Level Controls Medially Both Neuronal Determination and Differentiation

We first examined medial neurogenesis of gain- and loss-of-Hh function for the selected two key steps of neurogenesis, i.e., the establishment of neuronal commitment (elv3 expression) and the initiation of neuronal differentiation (isl1 expression) with the help of time defined by the MLD value.

Wild-type.

In WT embryos during the neural keel stage (10–14 hpf; 4–10 somite stages), both elv3 and isl1 cells followed similar linear progress, with two to three positive cells per side newly formed every hour (data not shown). Because the expression of elv3 persist after isl1 expression begins, two to three elv3-positive committed cells per side are differentiating every hour into isl1-positive PMNs. During the keel stage, an average 76.4% of elv3-committed cells are directed to PMN fate, being allowed to express one of the early differentiation markers, i.e., isl1.

smu mutant.

Hh signalling is required for primary motoneuron induction in zebrafish (Lewis and Eisen,2001). To confirm this observation in our MLD time-based evaluation, we first analyzed loss-of-function Hh mutant, smu (Varga et al.,2001). The cell count data hereafter with either elv3- or isl1-expression from groups of embryos were presented as the difference from uninjected WT controls at the identical age of embryos estimated by MLD time, as shown in Figure 4F″. Homozygous smu mutants exhibited reduced establishment of neuronal commitment (neuronal determination process; 22.8 ± 7.4 elv3 cells against 31.3 cells in wild-type counterparts, i.e., 70% commitment compared with wild-type; n = 20; P < 0.001; Fig. 5A). As reported with smu homozygotes (Varga et al.,2001), a significantly reduced number of isl1-PMNs was observed (7.7 ± 2.4 cells against 26.4 cells in wild-type counterparts at the stage of 11.3 hr in MLD time, i.e., 29% of normal PMNs; n =18; P < 0.001; Fig. 5B inset), and this finding significantly reduced the proportion of isl1-PMNs over elv3-committed cells (an index of PMN differentiation) to a rate of 35.3%, half the rate of 76.4% in wild-types (Fig. 5B; P < 0.005, Ryan's multiple comparison test of proportions). This observation is consistent with the reported requirement of Hh signalling on primary motoneuron induction, confirming our method to evaluate neurogenesis, based on the time compensated for the progression of the neural keel morphogenesis (MLD).

thumbnail image

Figure 5. Hedgehog (Hh) level controls medial neurogenesis and exerts binary-opposite effects. Various stages of embryos from different groups (wild-type [WT], smu, shh mRNA-injected, twhh mRNA-injected, and hsp;shh embryos) were staged in mediolateral distance (MLD) time and examined for medial neurogenesis during the neural keel stage (11–14 hours postfertilization [hpf]) with regard to elv3 and isl1 expression. Medial elv3-positive and isl1-positive cell count data from each group were subtracted with the mean cell number of uninjected WT controls of the identical MLD time. A,B: Subtracted cell counting results (ΔWT, cells/side) from different stages during the neural keel stage were pooled for each group and shown as a dot plot (A for elv3-positive cells, and B (inset) for isl1-positive cells) with mean (thick line) and standard deviation (SD, thin lines for upper and lower limits) on the right. P values from statistical analysis by Tukey's multiple comparison test after one-way analysis of variance are shown with symbols: ***, P < 0.001; **, P < 0.01; *, P < 0.05, and n.s. (not significant), P > 0.05. Note that both the Hh loss-of-function mutant smu and mRNA-injected Hh gain-of-function showed significant difference to uninjected WT controls. However, hsp;shh embryos showed no significant difference from WT embryos, which suggests a sequence of level-dependent Hh effects on neuronal progenitors can induce increased numbers of elv3 cells in the end compared with uninjected WT controls, as discussed in the text. The primary motoneuron (PMN) differentiation index (ratio of isl1 cells to elv3 neurons) is calculated from mean values of elv3 and isl1 cells and shown as a bar chart. For application of Ryan's multiple comparison test of proportions, calculated PMN differentiation indices were rounded to the nearest integer, and P values at the significant level of 0.05 were shown with symbols: ***, P < 0.001; **, P < 0.01. C: Level-dependent and binary-opposite model of medial Hh function. Particularly, this model insists the increased number of elv3 cells after early Hh overexpression (by mRNA injections of either shh or twhh) is a result of initial suppression of the establishment of neuronal commitment (at the transition of specification to determination) caused by a high Hh level and later permissive role on the same process under influence of a decaying (intermediate) Hh level. Selec., selection; Spec., specification; Det., determination (the establishment of commitment); Diff., differentiation (onset of differentiation).

Download figure to PowerPoint

Gain-of-function Hh phenotypes.

Next, we examined gain-of-function phenotypes of Hh signalling with two different strategies: early overexpression by mRNA injection and late overexpression by GAL4/UAS system. shh mRNA injection led to an increase in the number of elv3 cells with an average of 36.1 ± 7.0 cells (n = 51; 114% compared with uninjected counterparts; P < 0.05; Fig. 5A). mRNA overexpression of another Hh ligand in zebrafish, twhh, also showed significantly increased elv3 cells (145% of uninjected counterparts, n = 40; P <0.001; Fig. 5A). These values are in line with the proposed action of Hh ligands as a ventralizing morphogen in mouse and chicken (Jessell,2000) and also the action in zebrafish, where overexpression of Hh-regulated downstream transcription factors induced extra PMNs (Park et al.,2002; Cheesman et al.,2004). However, contrary to the accepted idea, normal or fewer numbers of isl1 PMNs were found after shh mRNA injection (21.9 ± 4.7 cells; 82% of uninjected embryos; n = 35; P < 0.001; Fig. 5B inset) than in uninjected counterparts (at the same MLD time of 11.4 hr, 26.6 cells), as observed with twhh mRNA injection (23.6 ± 4.9 isl1-PMNs against 27.8 cells in uninjected embryos; 85% isl1-PMNs compared with uninjected; P < 0.001; Fig. 5B, inset; n = 64). PMN differentiation rate (ratio of isl1-PMNs to elv3-committed cells) for overexpression of shh-mRNA were not significantly different from WT (P > 0.05, Ryan's multiple comparison test of proportions). However, mRNA injection of another Hh ligand Twhh significantly reduced the same rate to 46.4% against normal PMN differentiation ratio of 76.4% in uninjected embryos (Fig. 5B; P < 0.005, Ryan's multiple comparison test of proportions). One may argue that these observations might be due to the fact that in zebrafish mRNA injection of Hh ligands is an insufficient strategy to achieve overexpression of Hh ligands.

Therefore, to ensure robust overexpression of Hh ligands, we examined hsp;shh embryos (GAL4/UAS system) by adding two times of heatshock at 60% and 90% epiboly stages during gastrulation. Contrary to our prediction, hsp;shh embryos showed normal establishment of neuronal commitment (35.9 ± 10.5 elv3 cells against 34.0 cells in wild-type at the same MLD time of 11.8 hr, i.e., 106% of wild-type; n = 18; P > 0.05; Fig. 5A). In good contrast to this normal neuronal commitment, an even stronger reduction in the number of isl1-PMNs was observed in hsp;shh embryos than Hh mRNA injections (12.3 ± 3.6 cells against 27.3 cells in wild-types; 45% isl1-PMNs compared with wild-type counterparts, n =12; Fig. 5B, inset), and this finding severely retarded PMN differentiation to a rate of 34.1%, a lower level than smu homozygotes (35.5%), against a normal rate of 76.4% in wild-types (P < 0.001, Ryan's multiple comparison test of proportions).

Level-dependent and binary-opposite model of Hh action.

These results with hsp;shh embryos challenge us with several questions. First, as to the establishment of neuronal commitment (elv3 expression), it is contradictory that no effects (no increased elv3 cells) were observed in hsp;shh embryos unlike mRNA-injected embryos, if one insists it is a consequence of robust Hh action. The same misexpressed Shh caused an even stronger defect in the PMN differentiation ratio (isl1 cells over elv3 cells) in hsp;shh embryos than Hh mRNA-injected embryos. Second, we observed increased isl1 ventral neurons in the later neural tube stages with both strategies of Hh overexpression (Fig. 3A), an opposite phenotype that we observed here during the keel stage (Fig. 5B inset). These two questions, however, can be reasonably explained if we simply suppose that both too low and too high of Hh level acts restrictively on both the commitment establishment (elv3) and the onset of neuronal differentiation (isl1), and permissive effects on both processes lies in the middle (Fig. 5C).

The initial cue came furnished from our quantitative analysis on the effects of Hh overexpression on the late RB neurogenesis (Fig. 3). We proposed two mutually exclusive models to explain the time lag of 4.5 hr between the completion of Shh misexpression and the induction of supernumerary isl1-RBs. One model assumes that Hh signalling acts to promote neuronal commitment (elv3) and push progenitor cells out of the cell cycle; another assumes the opposite, which would be discriminated from each other by examining the time lag between the onset of elv3 and isl1. As described in detail in the next section (Fig. 6), in both wild-type and Hh gain-of-function embryos, lateral elv3 cells undergo almost concomitant differentiation into isl1-RBs, supporting the second model (Fig. 3C, case 2) that assumes a restrictive effect of Hh overexpression in the establishment of the neuronal commitment (elv3 expression). We propose that the initial high Hh levels take a restrictive and that the later decaying intermediate Hh level takes a permissive role for establishment of neuronal commitment (elv3 expression; Fig. 5C, right panel).

thumbnail image

Figure 6. Hedgehog (Hh) level controls Rohon–Beard mechanosensory neurons (RBs) and neural crest formation. A–C: The number of lateral neurons (either elv3-positive [A] or isl1-positive [B inset]) and trunk neural crest [NC] cells [C], foxd3, rectangular domain shown in the photo at the bottom) are counted for groups of embryos (wild-type [WT], MO-tri, smu, shh, twhh, and hsp;shh) and analyzed in mediolateral distance (MLD) time. As in Figure 5, cell count results were expressed as the difference from the uninjected WT controls of the identical MLD stage (ΔWT, cells/side). P values from statistical analysis by Tukey's multiple comparison test after one-way analysis of variance (A, B inset, and C) are shown with symbols: ***, P < 0.001; **, P < 0.01; *, P < 0.05; and n.s. (not significant) P > 0.05. RB differentiation index (ratio of lateral isl1 cells to lateral elv3 neurons) are calculated from mean values of elv3 and isl1 cells and shown as a bar chart. For application of Ryan's multiple comparison test of proportions, calculated ratios were rounded to the nearest integer, and P values at the significant level of 0.05 are shown with symbols: *, P < 0.05; n.s. (not significant) P > 0.05. Note that the loss-of-function Hh mutant smu showed a decreased RB differentiation index, demonstrating the requirement of Hh signalling in the lateral proneural domains. C: The loss-of-function Hh mutant smu showed significantly increased numbers of foxd3-positive trunk neural crest cells (***, P < 0.001 in Tukey's test after one-way analysis of variance). D: Models of Hh function on lateral neurogenesis and neural crest formation. Hh signalling adds fine adjustment on the established neural fates (neuron vs neural crest). Selec, selection; Spec, specification; Det, determination (the establishment of commitment); Diff, differentiation (onset of differentiation).

Download figure to PowerPoint

With this model, we will explain the first question above as follows. The absence of Hh effects on elv3 expression in hsp;shh embryos (Fig. 5A) can be understood when one assumes that progenitor cells were under proliferation as a consequence of restrictive effects on neuronal commitment (elv3), brought about by a high Hh level at the decision point during the cell cycle. After cell division, daughter progenitors do not exit and stay in the cycle and are now, during the decisive period of the cell cycle, exposed to a decaying Hh level to a permissive range to be committed as neurons and allowed to express elv3 (Fig. 5C). Thus, a sequence of level-dependent Hh effects on progenitors can induce more elv3 cells in the end compared with the normal WT situation. The same sequence should be followed with overexpression of Hh ligands by mRNA injections. The only difference from hsp;shh embryos is the cell cycle targeted by the high Hh level to put progenitors on the course for another round of proliferation. In our estimation of the cell cycle (Fig. 3C), late Hh overexpression triggered by heatshocks at 60% and 90% epiboly stages in hsp;shh embryos acted most probably on cycle 15. Thus, the first cell cycle that is affected to go through another round of cell cycle by Hh mRNA injections has to be earlier than cycle 15 and presumably cycle 14.

As to the second question about contradictory observation with ventral neurons after Hh overexpression, i.e., initially decreased isl1-PMNs (Fig. 5B inset) and later increased ventral neurons (Fig. 3A), we assume the same system where both lower and excess Hh levels act restrictively and intermediate Hh levels act permissively. Because neuronal determination marked with elv3 expression is an irreversible process, an excess number of elv3 cells, induced by a sequence of Hh level-dependent effects on neuronal commitment, remains to express elv3. In contrast, the initiation of PMN differentiation (isl1) is inhibited by a high Hh level, the very same one that forces progenitors directed for another round of cell cycle. Through this process, an excess number of elv3-positive but not differentiating neurons (negative for isl1) will accumulate in the neural anlage, which we assume are directed later to an another class of ventral motoneurons/interneurons, most probably to secondary motoneurons, in the later neural tube stage under the guidance of decaying intermediate Hh level with permissive function. As to the loss-of-function phenotype with smu mutants, we can take the effects of a lower Hh level as restrictive both on the establishment of neuronal commitment (elv3) and the initiation of neuronal differentiation (isl1), in the absence of the proof to the contrary.

The proposed model in Figure 5C seems to explain well the phenotypes with medial neurogenesis in the neural keel. One of our initial goals is to know the local differences among different proneural domains in the mechanisms of primary neurogenesis. To test whether this model of Hh function applies to the lateral neurogenesis and to know whether there is a local difference in the mechanisms of neurogenesis, we next analyzed RB neurogenesis in the neural keel.

Hh Level Controls RB and Neural Crest Formation

For medial neurogenesis, we proposed a model of Hh action, level-dependent, and binary-opposite effects on the two key steps of neurogenesis: establishment of the neuronal commitment (elv3) and the initiation of differentiation (isl1). Here, we analyze RB neurogenesis with a special focus on to what extent the model we proposed for medial neurogenesis (Fig. 5C) explains lateral neurogenesis. In this context, lateral neurogenesis in the neural keel has a specific feature that would deserve an attention. RB neurons are intermingled with premigratory trunk neural crest (NC) cells in the lateral-most proneural columns, whose fate is favorably determined by DELTA/NOTCH signalling at the expense of RB neurons (Cornell and Eisen,2000,2002). Therefore, we followed the expression of a transcription factor, forkhead box D3 (foxd3), which becomes expressed as early as the 90% epiboly stage (9 hpf) in premigratory NCs (Odenthal and Nusslein-Volhard,1998). To restrict our analysis to trunk NC cells, we followed foxd3 expression within the posterior half of the trunk region marked with myoD-positive adaxial mesoderm (Fig. 6, bottom). We found that foxd3-NC cells also follow the rules that we found with medial and lateral neurogenesis (Fig. 4E); NC formation is not a function of somitogenesis but of MLD (data not shown).

Wild-type.

In wild-type embryos during the neural keel stage, expression of elv3 and isl1 in presumptive RBs follow a similar linear increase over time and the proportion of isl1 cells over elv3-positive committed cells (an index of RB differentiation) is 94.3% (data not shown). This high differentiation rate, in good contrast to the medial PMN differentiation (76.4%), indicates the majority of lateral elv3 cells undergo almost concomitant differentiation into isl1-RBs, the observation that led us to establish the model of Hh function in the medial neurogenesis (Fig. 5C). In a similar way, the number of trunk foxd3-NC cells also followed a linear progress; however, trunk NC cells are formed at a rate of 29.0 ± 3.1 cells per side every hour, more than three times higher rate than RB neurogenesis (n = 47; data not shown).

MO-tri and smu mutant.

We next examined MO-tri embryos for RB neurogenesis and NC formation to ask whether we see a hidden function of Tri/Stbm beyond its effects on keel morphogenesis upon compensation by use of MLD time. As shown already in Figure 4F″, the establishment of neuronal commitment (elv3 cells) in MO-tri embryos are normal and fit well with uninjected control embryos upon MLD compensation (Fig. 6A). RB differentiation index of 102.4% in MO-tri embryos are also normal and comparable to the rate of 94.3% in uninjected embryos (Fig. 6B). Thus, retarded keel morphogenesis caused by severe defects in PCP pathway, together with defects in cell elongation polarity (either planar or apicobasal; Fig. 2), does not appear to affect RB neurogenesis, at least during the neural keel stage. NC cells marked with foxd3 expression in MO-tri embryos were comparable to WT (Fig. 6C; P > 0.05 in Tukey's multiple comparison test after ANOVA).

Unique phenotypes were observed with loss-of-function Hh mutant, smu. The establishment of lateral neuron commitment (elv3) was normal (32.6 ± 5.5 cells against 28.1 cells in wild-type counterparts, i.e., 116% compared with wild-type; n = 18; Fig. 6A; P > 0.05 in Tukey's test), as we observed with MO-tri embryos. However, unlike MO-tri embryos, in smu mutants, a significantly reduced (65% of WT, n = 18) number of isl1-RBs was observed (26.1 ± 4.7 cells against 40.0 cells in WT counterparts at 11.3 hr in MLD time, Fig. 6B inset, P < 0.001 in Tukey's test). This finding resulted in significantly reduced RB differentiation index to a rate of 75.8%, compared with the rate of 94.3% in wild-types (Fig. 6B), demonstrating for the first time the requirement of Hh signalling in the normal lateral neuron differentiation, although the requirement is limited to a small population. MLD compensation disclosed another important phenotype in smu homozygotes that has been missed. We found 141.7% more foxd3-NCs in smu homozygotes than WT at the same MLD time (on average 80.4 ± 25.5 NCs in smu against 56.7 NCs in WT at 11.6 hpf, Fig. 6C), which would be evaluated as normal (105.1% of WT) when the embryos were staged by the somite number (in average 80.4 ± 25.5 NCs in smu against 76.5 NCs in WT at 12.6 hpf).

These results reveal a local difference in the establishment of neuronal commitment (elv3), where in the lateral domain decreased Hh signalling is not restrictive but clearly permissive unlike in the medial region (Fig. 6D, upper half). In contrast, the onset of neuronal differentiation (isl1) is, similarly in both medial and lateral regions, restrictively affected by reduced Hh signalling (Fig. 6D). These facts present a general requirement of Hh signalling on the onset of neuronal differentiation. The increased number of foxd3 cells in smu embryos suggests a restrictive role of Hh signalling in the wild-type situation on the neural crest formation.

Gain-of-function Hh phenotypes.

Turning now to gain-of-function Hh phenotypes, we further examine the level-dependent and binary-opposite model of Hh action on the lateral proneural domains in the neural keel. Early Hh overexpression by injecting either shh or twhh mRNA-induced supernumerary elv3 cells in the lateral proneural domains (in average 139% or 119% of uninjected control, n = 57 and n = 40, respectively; Fig. 6A, significant differences to WT [P < 0.001] in Tukey's test), in good contrast to the normal elv3 cells with the late Hh overexpression in hsp;shh embryos (107.2% of WT at the average age of 11.8 hpf in MLD time, n = 20; Fig. 6A, no significant difference to WT [P > 0.05]). These results coincide well with the medial Hh function on the establishment of neuronal commitment (elv3). We thus confirm that the supernumerary neuronal commitment (elv3) after Hh misexpression in both medial and lateral proneural domains can be explained by means of a sequence of binary-opposite effects: the initial restrictive action of a high Hh level on the neuronal commitment followed by permissive action on the same process under control of intermediate (decaying) Hh signalling (Fig. 6D).

We next asked whether lateral onset of neuronal differentiation (isl1) is suppressed by both early and late Hh overexpression as we observed medially (Fig. 5B inset). The index of RB differentiation, proportion of isl1-RBs over total committed progenitors (elv3), revealed no significant differences from uninjected WT controls after both early and late Hh misexpression (104.3% for shh embryos [n = 40 for isl1, n = 57 for elv3], 98.7% for twhh embryos [n = 64, 40], 96.6% for hsp;shh embryos [n = 14, 20] against 94.3% in uninjected WT embryos [n = 57, 41]; Fig. 6B, P > 0.05 in Tukey's test after ANOVA), unlike the situation medially. This normal RB differentiation index after Hh misexpression indicates that the supernumerary elv3-committed cells undergo almost concomitant differentiation into isl1-RBs as is the case with WT embryos. MLD compensation thus disclosed the second specific feature in the lateral RB neurogenesis, where high Hh signalling does not interfere with the onset of neuronal differentiation (isl1), following the first characteristic that Hh signalling is not required for and show no defects upon depletion with the establishment of neuronal commitment (elv3), unlike medially (Fig. 6D).

We next investigated whether Hh overexpression affects NC induction. From the de-repression phenotype in NC formation with loss-of-function Hh mutant smu, it was suggested that Hh signalling normally acts restrictively on the neural crest formation in wild-type situation (Fig. 6C). Therefore, to ask whether increased levels of Hh signalling promote or suppress NC formation, we overexpressed either Shh or Twhh by mRNA injections and analyzed NC formation during the neural keel stage, with the help of MLD compensation. Both shh and twhh mRNA injections resulted in almost normal number of foxd3-NC induction (in average 86% and 94% of uninjected controls at the stages in MLD time of 11.7 hpf and 11.5 hpf, n = 52 and n = 44, respectively; Figure 6C, no significant differences to uninjected control WT [P > 0.05]). Thus, both endogenous and high levels of Hh signalling showed the same phenotype, indicating that Hh levels higher than a certain threshold level, lying between the levels of WT and smu homozygotes, permit NC induction at a normal rate (Fig. 6D, lower half).

Taken together, our model of Hh function on the primary neurogenesis, giving a special prominence to the level-dependent and opposite-binary effects, explains both medial and lateral neurogenesis with key modifications that reflect local differences between two proneural domains. However, these medial and lateral models of Hh function are induced from particular phenotypes and need to be tested whether predictions derived from these models fit to the phenotypes other than those depicted here. During the course of our experiments along this line, we examined the above mentioned MO-tri/shh embryos (Fig. 2) for the neurogenesis phenotype. We found unexpectedly and describe next that the effects of Hhs on the onset of neuronal differentiation can be medially, but not laterally, reversed upon reduced Tri/Stbm level.

Medial PCP Pathway Reverses Hh Function on the Neuronal Differentiation Into Opposite in a Polarity-Independent Manner

Throughout neurulation, zebrafish neuroepithelium remains essentially pseudostratified in the transverse plane (Fig. 2B) and neuroepithelial cells divide apically (Geldmacher-Voss et al.,2003), while postmitotic neurons are found in contact with the basal side of the neural keel (Figs. 2B, 7D, white arrowhead). In a sense that this geometric pattern has never been broken, no “ectopic” neurons were found in gain- and loss-of-Hh phenotypes through the above-mentioned comparison between medial and lateral neurogenesis. On the other hand, we cannot exclude the possibility that supernumerary (RB) neurons might have been induced outside the proneural domains. However, our models of Hh action in both the medial and lateral neurogenesis strongly suggest that Hh acts, not on the initial stage, but on the later half of the sequence of neurogenesis: the establishment of neuronal commitment (elv3 expression) and the initiation of differentiation (isl1 expression). Therefore, ectopic induction of neurons, if any, would be negatively regulated by mechanisms independent from Hh signalling.

thumbnail image

Figure 7. Medial Tri/Stbm mediated planar cell polarity (PCP) pathway reverses hedgehog (Hh) function on the medial neurogenesis into opposite in a polarity-independent manner. A–C: The number of neurons (either elv3-positive or isl1-positive) are counted during the neural keel stage (4- to 10-somite stages) and analyzed in mediolateral distance (MLD) time. As in Figures 5 and 6, results are shown as the difference from the uninjected controls of the identical MLD stages (ΔWT [wild-type], cells/side) and pooled for each group of embryos of WT, MO-tri, shh mRNA-injected, and MO-tri/shh embryos. P values from statistical analysis by Tukey's multiple comparison test after one-way analysis of variance (A,B) are shown with symbols: ***, P < 0.001; **, P < 0.01; and *, P < 0.05. Note that in shh mRNA-injected embryos, the increase from uninjected controls for elv3 (A, *P < 0.05) and decrease in isl1 cells (B, **P < 0.01) turned into the opposite upon co-injection with MO-tri, which are clearly reflected in motoneuron (MN) differentiation index (ratio of isl1 medial cells to elv3 medial neurons, C). D–F: Ectopic medial neurons are repressed by Tri/Stbm-mediated PCP pathway in a polarity-independent manner. D: In uninjected controls, medial isl1-PMNs are found (white arrowhead) at the bottom (basal side) of the neural keel intervened by floor plate cells adjacent to the notochord (dotted circle) and usually never seen at the apical side. E: However, in MO-tri embryos, ectopic isl1-medial neurons (black arrowhead) are found at the apical surface, together with the normal PMN at the basal position (white arrowhead). F,F′: This effect seems to be by means of a polarity-independent pathway, because ectopic medial neurons were not suppressed (black arrowhead; dorsal views at the focal plane of the line) in embryos with disturbed cell polarity (MO-tri/shh embryos) to a level comparable to shh embryos with no ectopic medial neurons. G: (Left) Medial Tri/Stbm-mediated PCP pathway reverses the function of a high Hh level at the two steps of medial neurogenesis: the establishment of commitment (at the transition from specification to determination) and the onset of neuronal differentiation (at the transition from determination to differentiation). As to the function of a low Hh level, analysis is pending (indicated by dotted line). (Right) Schematic drawings to present how reversed Hh function upon reduced Tri/Stbm-mediated PCP pathway caused supernumerary medial isl1 neurons. elv3 expression is shown as blank green ellipses and isl1 expression as violet-filled ellipses. P, permissive; R, restrictive. Scale bars = 50 μm for D–F, 100 μm for F′,F″.

Download figure to PowerPoint

Ectopic medial neurons.

We here show one such example of ectopic neurons in MO-tri embryos with retarded PCP signalling, where isl1-positive medial cells were unusually found in contact with the apical surface of the keel (Fig. 7E, black arrowhead), in addition to the normal isl1 medial neurons at the basal side of the keel (Fig. 7E, white arrowhead). The ectopic apical cells were also observed with medial elv3 expression, confirming that the apically formed isl1 cells contain neurons (data not shown). In MO-tri embryos, medial establishment of neuronal commitment (elv3) and PMN differentiation index (the proportion of isl1 PMN over elv3 committed cells) are normal (94.4% of uninjected control for elv3 cells; PMN differentiation index of 81.2% against 76.4% in uninjected controls; Fig. 7A–C). As we have shown, overexpression of Shh in the background of MO-tri embryos partially rescued defects in both planar and apicobasal cell elongation polarity (Fig. 2E). If ectopic apical medial isl1 neurons were induced by means of disturbed cell polarity, we could expect to see a rescued phenotype with suppression of ectopic apical isl1 medial neuron formation. To test this idea, we combined MO-tri and shh mRNA injections. Although both planar and apicobasal cell elongation polarity was partially rescued to a level comparable to shh embryos, as summarized in Figure 2E, ectopic apical isl1 medial neurons were still observed (Fig. 7F, black arrowhead; up to 30 ectopic isl1 cells in Fig. 7F′). These results clearly demonstrate that PCP signalling, independent from their effects on cell elongation polarity, prevents the formation of ectopic apical neurons medially.

Medial Tri/Stbm level determines Hh function.

We further examined MO-tri/shh embryos for elv3 and isl1-expression, to focus on the effects on neurogenesis. In good contrast to a single injection of shh mRNA, the establishment of neuronal commitment (elv3) in MO-tri/shh embryos was significantly reduced compared with shh embryos (Fig. 7A; P < 0.001 in Tukey's test after ANOVA; 91% elv3 cells of uninjected controls in MO-tri/shh embryos against 114% in shh embryos). Furthermore, in MO-tri/shh embryos, more medial isl1 neurons are found than in a single injection of shh mRNA (P < 0.001 in Tukey's test after ANOVA, Fig. 7B), and this finding led to a clearly increased neuron differentiation index (the proportion of isl1 over elv3 cells) to 96.9% against the ratio of 60.7% in shh mRNA-injected embryos (Fig. 7C, P < 0.001 in Ryan's multiple comparison test of proportions). These two observations in MO-tri/shh embryos indicate that a high Hh signalling level did not restrict but allowed both the establishment of neuronal commitment (elv3) and subsequent onset of differentiation (isl1), an opposite phenotype to that we observed above (Fig. 5C).

We thus propose a model, wherein medial functions of high Hh signalling on the neuronal commitment and differentiation are determined by Tri/Stbm level (Fig. 7G): a high Hh level is restrictive in normal Tri/Stbm level; in contrast, the same Hh- level turns into a permissive role under reduced Tri/Stbm level. This effector signalling cascade that Tri/Stbm are involved in to determine Hh function appears to be independent from cell elongation polarity, because both MO-tri/shh and shh embryos showed comparable degree of defects in cell elongation polarity (Fig. 2E). Our observation that the lateral Hh function on neurogenesis was not modified by Tri/Stbm level (data not shown) confirms that Shh protein expression in the neural keel was not affected by the Tri/Stbm level, which argues against the possibility that Shh protein was not overexpressed under reduced Tri/Stbm function.

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES
  9. Supporting Information

We found gain- and loss-of-Hh signalling affect normal neural keel morphogenesis. Through the comparison and combination with a PCP-deficient phenotype with reduced Tri/Stbm level, we found that Hh ligands act as extracellular polarizing signals, at least in part by means of PCP pathway. However, defects in the neural keel morphogenesis triggered by loss or gain of Hh function was not simply explained by disturbed cell polarity. Conditional late Hh overexpression revealed neurogenesis as a target of Hhs to affect keel morphogenesis. With the help of a newly devised method, i.e., MLD compensation, to uncover the effects on neurogenesis among retarded morphologies, a unique model was developed to explain the function of Hh signalling that argues level-dependent binary-opposite effects, where modification of the medial model to explain the lateral aspects is a direct reflection of local differences between the two proneural domains. Finally, we found that the level of medial Tri/Stbm determines the alternative Hh effects of permissive or restrictive on the trunk neurogenesis.

Hh Ligands as Candidates of Extracellular Polarizing Factors

Our results on the role of Hh signalling on zebrafish neurulation in comparison with the PCP phenotypes by means of Tri/Stbm are summarized in Figure 8, with a special focus on the interrelationship between primary neurogenesis and the neural keel morphogenesis. Although both Hh signalling and the Tri/Stbm-mediated PCP pathway control neural keel morphogenesis, the underlying mechanisms that produce the similar phenotypes are quite different. Unlike PCP molecules, Hh signalling exerts its effects on the neural keel morphogenesis by means of neurogenesis, although Hh ligands act as extracellular polarizing signals at least in part by means of Tri/Stbm-mediated PCP pathway. We also suggest that the Tri/Stbm-mediated PCP pathway includes polarity-independent cascades that prevent ectopic neurons from being generated in the neural keel and modify medial Hh function on neurogenesis.

thumbnail image

Figure 8. Summary of the role of hedgehog (Hh) signalling on zebrafish neurulation in comparison with the planar cell polarity (PCP) pathways by means of Tri/Stbm. Thick arrows indicate the major effects of each component, either permissive or restrictive. Thin arrows indicate the minor but novel findings in this study (discussed in the text). [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

Download figure to PowerPoint

PCP pathway in zebrafish has been studied mainly in two aspects: convergence and extension (CE) movements during gastrulation and caudal migration of the facial (nVII) motor neurons. tri/stbm (Jessen et al.,2002), prickle1 (pk1; Carreira-Barbosa et al.,2003), and landlocked (llk)/scribble1 (scrb1; Wada et al.,2005) are required for both CE movements and neuronal migration, whereas other PCP molecules, knypek/glypican4/6, silberblick (slb)/wnt11, and pipetail (ppt)/wnt5a (Heisenberg et al.,2000; Topczewski et al.,2001; Kilian et al.,2003), are exclusively involved in CE movements during gastrulation. Of interest, CE and neuronal migration seem to be regulated by different mechanisms. Tri/Stbm, which plays a central role in both CE and neuronal migration, is suggested to act on neuronal migration by means of a Dishevelled-independent pathway distinct from noncanonical Wnt signalling (Jessen et al.,2002).

Various examples of planar polarized morphogenetic behaviors suggest that long-range gradients of extracellular signals direct cells to become asymmetric through the PCP signalling pathway (Lecuit,2005). However, the nature of the upstream polarizing signals are not fully understood; in zebrafish, slb/wnt11 and ppt/wnt5a (Heisenberg et al.,2000; Kilian et al.,2003) are shown to play such a role. In this context, our results in this study should be worth being tested further, where Hh ligands act as extracellular polarizing signals on developing neuroepithelium, at least in part by means of PCP-signalling pathway (Fig. 2). It will be also informative to know whether Hh ligands act through Frizzled Wnt-receptor.

Another interesting question is whether Hh signalling acts directly in vivo in the lateral neuroectoderm, or whether the observed effects are indirectly the result of other relay signals. A key regulator of Hh signalling is the Hh receptor Patched (Ptc) that not only acts to transduce Hh signals but also limits the diffusion or movement of the Hh protein from its source (Lewis et al.,1999). Both zebrafish patched genes patched1 and patched2 are transcribed in the medial part of neuroectoderm as early as the neural plate stage and transcription of both ptc genes is under the positive control of Hh ligands. However, no significant level of transcripts for either ptc gene has been reported within the lateral neuroectoderm. In this context, it will be interesting to determine how far Hh ligands can travel along the mediolateral axis of the developing neuroectoderm of the zebrafish. Importantly, the multimeric form of lipid-modified Shh is soluble, in contrast to its monomer membrane-associated form, and very potent, thus being suggested to diffuse far from its site of synthesis to provide long-range patterning information (Zeng et al.,2001).

One question still remains open as to how decreased or increased neurogenesis can affect the neural keel morphogenesis. We speculate that committed progenitor cells into neuronal fates have different cell-surface adhesion properties from uncommitted progenitors. Hh-induced ectopic “committed cells” among proliferating progenitor cells in both medial and lateral proneural domains, thus, might perturb normal morphogenetic movements within the neuroepithelium.

We will also need to analyze further the loss-of-function Hh signalling mutant smu to confirm the requirement of Hh signalling in cellular movements during gastrulation. Although the offspring from the cross between smu heterozygotes all complete gastrulation quite normally (data not shown), smu mutant embryos that we analyzed are zygotic mutation. It is highly probable that maternal contribution of the smoothened gene compensates the loss of zygotic gene function. Indeed, the maternal scrb1 PCP molecule is shown to be required for CE movements during gastrulation (Wada et al.,2005).

Hh Signalling Participates in the Fine Adjustment of Neural Fates Determined by Lateral Inhibition by Means of DELTA/NOTCH Pathway

Some aspects of our results with gain-of-Hh phenotypes show similarity to the loss-of-function phenotypes of DELTA/NOTCH signalling, suggesting a possibility that Hh might weaken DELTA/NOTCH signalling. However, Hh phenotypes are not explained by this possibility. For example, embryos carrying a mutation in deltaA also show supernumerary RBs (Cornell and Eisen,2000) as we observed to a much lesser extent with gain-of-function Hh phenotypes (Fig. 3). Although deltaA mutants also showed a concomitant decrease in the trunk NC cells, which indicates an instructive function of DELTA/NOTCH signalling in the NC induction (Cornell and Eisen,2000,2002), this is not the case with our Hh gain-of-function phenotype (Fig. 6). Although both normal and high Hh levels act to limit NC formation to a given level, as exemplified in the de-repression phenotype of the Hh mutant smu with supernumerary NCs, however, they did not repress NC formation. This finding argues against the aforementioned possibility that supernumerary RBs in gain of Hh phenotypes are induced at the expense of NCs, by interfering with lateral inhibition.

Another example to support the idea that Hh signalling takes a part in fine adjustment of neural fates determined by means of DELTA/NOTCH pathway without affecting the DELTA/NOTCH pathway is shown in the medial neurogenesis. DELTA/NOTCH signalling inhibits PMN fates and promotes secondary motoneuronal fates (Appel and Eisen,1998), and deltaA mutant embryos have excess early-specified neurons (PMNs and RBs) and fewer late-specified neurons (secondary motoneurons and some classes of interneurons) and radial glia (Appel et al.,2001). If Hh signalling acted for promotion of the DELTA/NOTCH pathway to limit PMN induction and induced excess secondary motoneurons in the later stages, we should observe decreased numbers of elv3-positive neuronally committed cells, concomitant with decreased onset of neuronal differentiation (isl1). However, our results showed increased establishment of neuronal commitment (elv3; Fig. 5A), arguing against the supposition above. It is, therefore, unlikely that Hh signalling exerts its effects on neurogenesis by means of DELTA/NOTCH signalling. We thus conclude that Hh signalling adds fine adjustments to the neural fates within the frame of a blueprint for neural development established by the DELTA/NOTCH pathway that appears to play a role in wider temporal patterning of neurogenesis.

Epiboly Might Be Normal in shh Embryos

The embryonic shield in zebrafish, a thickening of the germ ring, homologous to Hensen's node in the chick and Spemann's organizer in amphibians, has been shown to be center of Hh signalling (Corbit et al.,2005). Two Hh ligands in zebrafish, shh and twhh genes, are expressed zygotically as early as from the 50–60% epiboly stages in the embryonic shield (Lewis and Eisen,2001). However, their expression pattern is slightly different from each other. At the 90% epiboly (9 hpf) stage, shh is expressed in the primordium of both notochord and floor plate; in contrast, twhh expression is limited in floor plate (Beattie et al.,1997). This finding means that the shh gene is expressed both in epiblast (outer layer of the two layers of deep cells) and hypoblast (inner layer); in contrast, twhh is limited in epiblast. To see whether Twhh plays a role in mid-gastrulation movements, we also examined twhh mRNA-injected embryos; however, unlike the case with Shh, twhh embryos completed gastrulation quite normally without any morphological disorder (data not shown). Together with the knowledge that epiboly is driven by a radial intercalation of cells in epiblast and hypoblast plays no role in this process (Kane et al.,2005), it is highly probable that morphogenetic disorders in shh-injected gastrulae (Fig. 1C) are not caused by the defects in epiboly but due to the disturbed cell behavior in the layer of hypoblast, for example, convergence movements. In this context, it is noteworthy that zebrafish PCP pathway mutants show severe defects in convergence movements; however, epiboly is normal (McFarland et al.,2005).

Hh Level or Its Effector Signalling Pathway May Explain Differences of the Hh Effects Among Species and Tissues

Hh ligands are extracellular signalling molecules that act not only as morphogens but also control the proliferation of cells (Ingham and McMahon,2001). Although there is almost no argument with this global sense of Hh function, however, in individual cases among species and different types of tissues, the effects of Hh signalling on cell proliferation can be totally opposite. For example, in the zebrafish retinal neurogenesis, Shh promotes cell cycle exit of retinal precursor cells and promote differentiation of retinal ganglion cells (RGC; Neumann and Nuesslein-Volhard,2000; Stadler et al.,2004; Masai et al.,2005), whereas in mouse Shh inhibits the same two processes (Wang et al.,2005). Not only among species but also among tissues, the effects of Hh signalling can be opposite. As we have shown in zebrafish here, a high level of Hh signalling acts restrictively on cell cycle exit of trunk neural progenitor cells and force them to undergo another round of cell cycle, which is an opposite effect on zebrafish retinal neurogenesis (Neumann and Nuesslein-Volhard,2000; Stadler et al.,2004; Masai et al.,2005). Our two findings that Hh signalling exerts binary-opposite effects in its level-dependent manner and that medial polarity-independent Tri/Stbm pathway acts as an effector on the effects of high Hh-signalling on the trunk neurogenesis may provide a clue on this issue.

EXPERIMENTAL PROCEDURES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES
  9. Supporting Information

Wild-Type and Mutant Zebrafish

Embryos were raised at 28.5°C on a 14–10 hr light–dark cycle and staged according to hours postfertilization and number of somites (Kimmel et al.,1995). smu homozygotes are obtained from the cross between heterozygotes of smub641 (Varga et al.,2001) and sorted by morphological features of ventrally curled tails, U-shaped somites, and partial cyclopia.

Transgenic Lines

An activator line, hsp70::Gal4 (heterozygote; Scheer et al.,2002); and an effector line, UAS:shh (heterozygote), were used in this study. For the construction of UAS:shh effector DNA construct, 1.7-kbp EcoRI shh fragment from pHH (from P. Ingham; Krauss et al.,1993) was filled-in and inserted into SmaI/BglII (filled in) site from pBSUASE1bII notch:intra plasmid (Scheer et al.,2002). For the establishment of the UAS effector lines, each DNA construct was injected into yolk cells direct beneath blastomere of one-cell stage wild-type zebrafish embryos at the concentration of 15 ng/μl. Injected, putative founder fish (G0) were crossed with wild-type fish, and their progeny (F1) were allowed to develop up to the 50% epiboly stage, and genomic DNA from every 50 F1 embryos was pooled and screened with PCR. Primers used for screening are as follows: UASdir2 (5′-CCA TCG CGT CTC AGC CTC ACT-3′) and shh-6 (5′-GCG CGT TAT CTT GCC CTC GTA T-3′). As to PCR-positive G0 fishes, crosses were again made with wild-type fish and their progeny (F1) were raised to adulthood. Fin clips were performed to identify positive F1 insertions.

As to crosses between the hsp70::Gal4 driver line and the UAS:shh effector line, heat shocks were applied at various times of development from the 40% epiboly stage to groups of approximately 25 embryos in zebrafish Ringer's solution in a 0.5-ml PCR tube, using a water bath precisely maintained at 40°C for a certain time period between 20 and 30 min. Transactivation of shh gene is first detected 1.5 hr after HS treatment, consistent with our previous observation (Scheer et al.,2002). As to overexpression of shh, two heat shocks at 60% and 90% epiboly were found to result in the severest effects on the patterning of isl1-positive neurons in the neural tube and this condition was exclusively used in this study.

mRNA Injections

Fragments that contain the full length of either shh or twhh were obtained by digesting pshh (Krauss et al.,1993) with EcoRI or ptwhh (Ekker et al.,1995) with Asp718/BamHI, respectively, and inserted into the EcoRI and StuI site of the pCS2+ vector, yielding pCS2+shh and pCS2+twhh, respectively. Capped mRNAs for shh and twhh were transcribed from linearized DNA templates with SP6 RNA polymerase in vitro transcription kit (MEGAscript, Ambion) with m7G(5′)ppp(5′)G cap analog according to the manufacturer's instructions. Capped mRNAs for shh and twhh were injected at a concentration of 300 ng/μl into yolk cells direct beneath the blastomere at one-cell stage.

tri/stbm Antisense Morpholino

The sequence of antisense morpholino against strabismus is 5′-GTA CTG CGA CTC GTT ATC CAT GTC-3′ as described in Park and Moon,2002 (Gene Tools). A 50 μM solution of antisense morpholino was injected into one-cell stage embryos.

In Situ RNA Hybridization

Whole-mount in situ hybridization was performed as described by Bierkamp and Campos-Ortega (Bierkamp and Campos-Ortega,1993). Briefly, digoxigenin (DIG) -labeled RNA probes were used in combination with anti–DIG-alkaline phosphatase (AP) Fab fragments (1:6,000; Roche) and staining was developed in the presence of NBT/BCIP. For two-color staining, two different either DIG- or fluorescein-labeled RNA probes were used in hybridization at 65°C, and after several cycles of washing with increasing stringency, anti–DIG-AP (1:6,000; Roche) was first added and blue staining was developed in nitroblue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate (NBT/BCIP) solution. Color development by means of AP was stopped by adding 0.1 M glycine solution (pH 2.2) and then anti–fluorescein-AP Fab fragments (1:1,000; Roche) was added and red staining was developed in FastRed solution (Roche). Antisense RNA probes used in this study are islet1, islet2, lim3, tlx3b, elavl3, myoD, and foxd3.

Counting Cells and Image and Statistical Analyses

After RNA in situ hybridization, embryos were removed from yolk cells and flat-mounted for photographs and cell count. All counting was conducted from the dorsal view, with either ×40 or oil immersion ×63 objective lens along the complete length of trunk and tail regions of spinal cord marked by myoD-adaxial mesoderm cells. Each left and right side was separately counted and treated as an independent sample. Each side was counted at least twice and averaged and processed for statistical analyses. To reveal any possible differences among multiple groups (Figs. 2, 5, 6, 7), sampled data were first examined whether they followed a Gaussian distribution and then tested whether means among groups are significantly different by one-way ANOVA (Prism3.0, GraphPad Software, Inc.). When groups showed differences (α = 0.05), Tukey's multiple comparison test was applied to see where the differences were. In some cases, frequently observed with elv3-positive medial neurons, cytoplasmic staining of markers had ill-defined borders that made large variation among counted results. In that case, counting was thought to be the most precise approximation between the minimal and maximal evaluation. Image analyses were conducted by ImageJ software (NIH, USA). Confocal image stacks were acquired by LSM410 (Carl Zeiss, Germany). Semi-thin transverse or sagittal sections (15 μm) of Durcupan-embedded embryos were made by a glass knife with LKB ULTROTOME III.

To analyze cell elongation polarity in horizontal (planar) and transverse (apicobasal) planes, freehand outlines of individual cells were drawn and the angle between the major axis of each cell and midline was measured. Statistical analyses of the axial data of cell elongation polarity were used as described by Concha and Adams (Concha and Adams,1998) and calculated by a free program, “R”: A language and environment for statistical computing (R development Core Team). Briefly, angles were measured on both sides of the neural keel and data were first analyzed in Watson's uniformity test to see whether they were random or not (α = 0.05), where random data were not further analyzed. Nonrandom data were further tested as to whether they follow a circular counterpart of Gaussian distribution in linear data, von Mises distribution, at the significance level of 0.05. Data that could be fit into von Mises distribution were regarded as directional, and the angular deviation (circular standard deviation) was calculated to see the effects of Hh and PCP pathways. Comparison of two directional circular data was conducted by Watson's two-sample test of circular homogeneity, with mean direction clustered around a given direction. Circular data were presented in “rose diagrams,” grouped into 10° bins.

Acknowledgements

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES
  9. Supporting Information

We dedicate this manuscript to the memory of Professor Dr. José A. Campos-Ortega, a great mentor, an inspiring leader of our research group, and whose warm father figure we miss. Special thanks to Dr. Laure Bally-Cuif for critical reading, ideas of isl1/elv3 ratio, smu, and whose encouragement was invaluable during the course of this work. We also thank the following individuals for their assistance: Dr. Elisabeth Knust for the breakthrough idea to combine shh and MO-tri; Dr. Uwe Strähle for constructive critical opinions; Dr. Rolf Karlstrom for critical reading and the idea to follow cell elongation polarity; Dr. Richard Adams for sharing the initial idea of MLD compensation and the idea to analyze MO-tri; Dr. André Quinkertz for his initial contribution to this work; and Christel Schenkel and Thomas Wagner for expert technical assistance.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. RESULTS
  5. DISCUSSION
  6. EXPERIMENTAL PROCEDURES
  7. Acknowledgements
  8. REFERENCES
  9. Supporting Information

The Supplementary Material referred to in this article can be viewed at http://www.interscience.wiley.com/jpages/1058-8388/suppmat

FilenameFormatSizeDescription
jws-dvdy.20720.dat1.tif10911KSupporting Information file jws-dvdy.20720.dat1.tif
jws-dvdy.20720.dat2.tif5049KSupporting Information file jws-dvdy.20720.dat2.tif

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.