Abstract

Vertebrate embryos faithfully produce bilaterally symmetric somites that give rise to repetitive body structures such as vertebrae and skeletal muscle. Body segmentation is regulated by a cyclic gene expression system, containing the Delta–Notch pathway and targets, which generates bilaterally symmetric oscillations across the Pre-Somitic Mesoderm (PSM). The position of the forming somite boundary is controlled by interaction of this oscillator with a determination front comprised of opposing gradients of FGF and retinoic acid (RA) signalling. Disruption of RA production leads to asymmetries in cyclic gene expression, but the link between RA and the oscillator is unknown. In somitogenesis, Notch signalling activates target genes through the transcription factor Suppressor of Hairless (Su(H)). Here, we report that two Su(H) genes coordinate bilaterally symmetric positioning of somite boundaries in the zebrafish embryo. Combined Su(H) gene knockdown caused defects in visceral left/right asymmetry, neurogenic lateral inhibition, and symmetrical failure of the segmentation oscillator. However, by selectively down-regulating Su(H)2 or Su(H)1 function using specific antisense morpholinos, we observed asymmetric defects in anterior or posterior somite boundaries, respectively. These morphological abnormalities were reflected by underlying asymmetric cyclic gene expression waves in the presomitic mesoderm, indicating a key role for Su(H) in coordinating the left–right symmetry of this process. Strikingly, expression of the RA-degrading enzyme cyp26a1 in the tailbud was controlled by Su(H) activity, and morpholino knockdown of cyp26a1 alone caused asymmetric cyclic dlc expression, suggesting that excess RA in the tailbud may contribute to the cyclic asymmetries. Indeed, exogenous RA was sufficient to generate asymmetric expression of all cyclic genes. Our observations indicate that one element of the Notch signalling pathway, Su(H), is required for control of RA metabolism in the tailbud and that this regulation is involved in bilateral symmetry of cyclic gene expression and somitogenesis.

Keywords: Somitogenesis; Notch signalling; Retinoic acid; Asymmetry; Segmentation; Zebrafish

Article Outline

Introduction

Materials and methods

Fish care

Cloning of two Suppressor of Hairless genes

Intron-derived riboprobes

Morpholino design and injection

Retinoic acid treatment

mRNA synthesis and injection

In situ hybridizations

Antibody staining

Results

A second active Suppressor of Hairless gene in zebrafish

Su(H)1 and Su(H)2 in neurogenesis and heart development and laterality

Su(H)1 and Su(H)2 are involved in bilateral positioning of somite boundaries

Down-regulating either Su(H) gene causes asymmetries in cyclic gene expression

Su(H) function is required for expression of the retinoic acid catabolizing enzyme-encoding gene cyp26a1, in the tailbud

Retinoic acid catabolism is required for L/R symmetric cyclic gene expression

Discussion

Two functional Su(H) genes in the zebrafish

Embryonic laterality, Notch and retinoic acid signalling

Retinoic acid and segmentation—up- or downstream?

Acknowledgements

Appendix A. Supplementary data

References

Introduction

During embryogenesis the body axis is sequentially subdivided from head to tail into blocks of tissue called somites that give rise to repetitive structures such as the vertebrae, ribs and skeletal muscle. The somites bud from the anterior most end of the unsegmented Pre-Somitic Mesoderm (PSM), forming simultaneously and symmetrically on either side of the midline. The periodicity of this process is thought to be controlled by the combined action of a ‘segmentation clock’, a genetic oscillator that regulates the temporal and spatial organization of cells within the PSM, and a ‘wavefront’ or gradient that arrests the oscillations at a determination front (Palmeirim et al., 1997, Palmeirim et al., 1998, Pourquie, 2001, Pourquie and Kusumi, 2001 and Pourquie and Tam, 2001).

The first evidence of a segmentation oscillator came from work in chick demonstrating the oscillatory mRNA expression of the basic helix–loop–helix (bHLH) gene, c-Hairy1 (Palmeirim et al., 1997). Expression initiates in the posterior-most PSM and travels anteriorly in a wave-like manner, arresting at the newly forming somite boundary (Masamizu et al., 2006). Subsequent work in mouse and zebrafish has shown that these species also possess several so-called cyclic genes with almost identical wave-like mRNA expression patterns in the PSM (Aulehla and Johnson, 1999, Aulehla et al., 2003, Holley et al., 2000, Holley et al., 2002, Jiang et al., 2000, Jouve et al., 2000 and Oates and Ho, 2002). To date, almost all identified cyclic genes are involved in the Delta/Notch intercellular signalling pathway. Importantly, somite formation and wave-like expression patterns of most cyclic genes are disrupted by mutations or perturbations in this pathway (Aulehla et al., 2003, Barrantes et al., 1999, Bessho et al., 2001b, Dunwoodie et al., 2002, Evrard et al., 1998, Holley et al., 2000, Holley et al., 2002, Jiang et al., 2000, Jouve et al., 2000, Julich et al., 2005, Koizumi et al., 2001, Morales et al., 2002, Oates and Ho, 2002, Sieger et al., 2003 and Zhang and Gridley, 1998).

Notch mediated cell-to-cell signalling is widely used by vertebrates to specify cell fate and regulate pattern formation. Binding of Notch to its ligand, Delta or Serrate, on a neighboring cell causes the intracellular domain of Notch to be proteolytically cleaved. The Notch intracellular domain (NICD) then translocates into the nucleus where it associates with the highly conserved DNA binding protein Suppressor of Hairless Su(H)/RBPJk and converts the latter from a transcriptional repressor to a transcriptional activator (reviewed in Schweisguth, 2004). Downstream target genes transcriptionally activated by the NICD–Su(H) complex include members of the Hairy family of genes, themselves known to encode transcriptional repressors, raising the possibility that a transcriptional feedback loop might be part of the oscillator mechanism (Bessho et al., 2003, Hirata et al., 2002 and Palmeirim et al., 1997). Cyclic hairy homologues have been identified in the chick and mouse PSM (Bessho et al., 2001a, Jouve et al., 2000, Leimeister et al., 2000 and Li et al., 2003), and in zebrafish, the Hairy homologues her1, her7, her11 and her12 display oscillating expression (Holley et al., 2000, Holley et al., 2002, Oates and Ho, 2002, Sawada et al., 2000, Sieger et al., 2004 and Gajewski et al., 2006). This suggests a conserved role for the Hairy family in the somitogenesis oscillator.

Evidence from different vertebrate embryos suggest that a critical event in the oscillator is periodic Notch activation, even though the molecular means of achieving this end may vary between species. Expression of the Notch ligand deltaC (dlc) is cyclic in the zebrafish, but no known Delta genes cycle in the PSM of mouse or chick. Instead, expression of the glycosyl-transferase lunatic fringe, which itself has been identified as a modulator of Notch signaling, is cyclicly transcribed in mouse and chick (Forsberg et al., 1998, McGrew et al., 1998, Morales et al., 2002 and Prince et al., 2001). To date, no oscillating lunatic fringe homologue has been identified in zebrafish (Prince et al., 2001 and Leve et al., 2001). Recent studies in mouse have also found two oscillating members of the canonical Wnt pathway (Aulehla et al., 2003 and Ishikawa et al., 2004) although no Wnt pathway gene has been found to oscillate in the zebrafish. Thus, the emerging model of the oscillator in the zebrafish is of a genetic feedback mechanism consisting of components of the Delta/Notch signaling pathway and their target genes, which causes the cells of the PSM to undergo repeated cycles of gene expression and repression. In amniotes Wnt signalling also appears to be involved (reviewed in Aulehla and Herrmann, 2004).

The determination front, initially defined by a series of elegant transplantation studies, is the position in the PSM at which cells respond to the periodic signals of the oscillator and thereby determine the position of the future segmental boundaries (reviewed in Dubrulle and Pourquie, 2004a and Aulehla and Herrmann, 2004). The size of a somite is thus given by the number of cells experiencing the transit of the determination front during one cycle of the segmentation oscillator. The A/P positioning of the determination front within the PSM is thought to be controlled by two opposing and interacting gradients of FGF and retinoic acid (RA). Highest in the posterior, a gradient of FGF signalling creates a threshold that acts as the signal to the PSM cells to arrest oscillations at the determination front (Sawada et al., 2001, Dubrulle et al., 2001 and Dubrulle and Pourquie, 2004b), and Wnt signalling may also contribute to this arrest (Aulehla et al., 2003). From the anterior, a gradient of retinoic acid (RA) can also alter the position of the determination front by antagonizing FGF signalling (Diez del Corral et al., 2003 and Moreno and Kintner, 2004). Although in Xenopus the FGF-signal transduction inhibitory phosphatase MKP3 is involved in establishing the position of the determination front (Moreno and Kintner, 2004), the mechanism of interaction of the gradients in the PSM of different species is still unclear. Control of FGF concentration in the mouse PSM is regulated by mRNA stability in cells emerging from the tailbud (Dubrulle and Pourquie, 2004b), whereas RA levels are set by the balance of RA anabolism in the somites (the source) through Raldh2, the last enzyme in the synthesis pathway, and catabolism via Cyp26a, expressed in the tailbud (the sink). The establishment of opposing FGF and RA gradients by somites and tailbud may be a general mechanism to sharpen a determination or differentiation front in a growing tissue (reviewed in (Diez del Corral et al., 2003). We note that a RA-independent mechanism involving her13.2 in the zebrafish has been proposed to control the determination front (Kawamura et al., 2005), suggesting additional complexities in the regulation of the system.

Recent reports have shown a dramatic loss of bilateral coordination of symmetric cyclic gene expression following reduction of RA production in chick, mouse and zebrafish embryos (Kawakami et al., 2005, Sirbu and Duester, 2006, Vermot et al., 2005 and Vermot and Pourquie, 2005), and in chick, an asymmetry in the final positions of somite boundaries. It was therefore proposed that the RA gradient might act additionally to, or through its role in determination front positioning, as a buffer to prevent asymmetric signalling molecules involved in visceral laterality from perturbing the bilateral synchrony of the segmentation oscillator (Brent, 2005 and Hornstein and Tabin, 2005). Importantly, in all determination front-based models (Dubrulle and Pourquie, 2002), the segmentation oscillator is downstream of, or constrained by these gradients; to date there is no evidence that the segmentation oscillator plays a role in setting the shape, amplitude, or interactions of the FGF or RA gradients.

Here we report that the function of duplicated Su(H) genes, encoding the critical transcriptional mediators of Notch signalling, are required for symmetric cyclic gene expression in zebrafish. Surprisingly, Su(H) function is required for expression of cyp26a1, the gene which encodes the RA catabolic enzyme in the tailbud, and that is in turn necessary for cyclic dlc expression. Consistent with the loss of a catabolic sink for RA in the affected embryos, increasing levels of exogenous RA induces cyclic gene expression asymmetries. These results show for the first time a direct connection between a component of the somitogenesis oscillator and the RA pathway.

Materials and methods

Fish care

Fish were kept on a 14 h light/10 h dark light cycle following standard culture methods and were staged according to (Kimmel et al., 1995). Fish strains AB or Gol were used for all experiments.

Cloning of two Suppressor of Hairless genes

Using a nesting strategy with degenerate primers, two different Su(H) transcripts were isolated from Danio rerio cDNA. All PCR was performed with the profile: 5 min at 95°C, then 30 cycles of 30 s at 95°C, 1 min at 55°C, 1 min at 72°C, then 10 min at 72°C. In the first round of amplification, SuH1: GCI CA(AG) AA(AG) (AT)(GC)I TA(CT) GGI AA(CT) GA and SuH3R: CCA I(GC)(AT) IGC ICC (AG)TC (AG)TT IAT CAT were used to generate template from embryonic cDNA for a second round using SuH2: CA(AG) (CT)TI CA(CT) AA(AG) TG(CT) GCI TT(CT) TA and SuH3R, where I denotes an Inosine residue. A 120 bp PCR product was isolated after electrophoresis, subcloned and found to contain two distinct Su(H)-related gene fragments assessed by BLAST search against the zebrafish Ensembl and NCBI databases. 5′ and 3′ RACE PCR was used to complete the cDNAs. One cDNA corresponds to the previously identified Su(H) gene (Sieger et al., 2003), termed rbpsuh from ENSDARG00000003398 on Chromosome 1 at 15.6. The newly identified cDNA is described as a novel gene ENSDARG00000052091 on Chromosome 7 at 62.0 (Ensembl v37). We have submitted our cDNA sequences to NCBI.

Intron-derived riboprobes

Genomic DNA from the Su(H)1 and Su(H)2 genes was amplified using primers designed to avoid regions of high nucleotide similarity between potential exons as follows: csl1-1 TGTTCTACGGTAATAGCGCAG; csl1-2R GTCTGCTTGTCCACTTTACGA; csl2-4 CACTACGGACAAACTGTCAAA; csl2-2R GTCTGCTTATCCACCTTACGG, yielding products of 1.5 and 1.3 kb respectively. These primer pairs were used to map the Su(H)1 and Su(H)2 genes to LG1 with a LOD of 14.0, and LG7 with a LOD of 8.1, respectively, on the LN54 radiation hybrid panel (Hukriede et al., 1999). Three introns were found in the Su(H)1 product and one in the Su(H)2 product such that overall these sequences were 66% and 93% intron-derived. After subcloning, these fragments were used to generate riboprobes for in situ hybridization (see below).

Morpholino design and injection

Antisense morpholino oligonucleotides complementary to the 5′ regions of Su(H)1 and Su(H)2, and a morpholino targeting the ATG of both Su(H) genes were designed and synthesised by Gene Tools LLC (Philomath, Oregon). Su(H)1 MO: 5′-CCG GTG TGA CAA ATA ACG CCA GGA A-3′; Su(H)2 MO1: 5′-CGC CAT CTT CCA CAA ACT CTC ACC A-3′; Su(H)2 MO2: 5′-TCC TCC TCT CCC AGA CCC TTC CAG C-3′; Su(H)1+2 MO: CAA ACT TCC CTG TCA CAA CAG G-3′. The standard control MO recommended by Gene Tools was: 5′-CCT CTT ACC TCA GTT ACA ATT TAT A-3′. The same antisense morpholino for cyp26a1 was used as previously described by Emoto et al. (2005) and with the same amount injected (1 ng): cyp26a1 MO: 5′-CGC GCA ACT GAT CGC CAA AAC GAA A-3′. Morpholinos were resuspended in distilled water, then diluted to 10 ng/μL in 1× Danieau's solution and stored at − 20°C. Prior to injection morpholinos were further diluted to the following optimized concentrations in 1× Danieau's plus 0.2 mg/mL Fast Green: Su(H)1 was targeted with Su(H)1 MO at 7 ng; Su(H)2 was targeted with 1 ng of Su(H)2 MO1 and 1 ng of Su(H)2 MO2; the Su(H)1+2 MO targeting both transcripts (equivalent to the ORF-MO of Sieger et al., 2003) was used at 7 ng; and to target both transcripts independently, Su(H)1 MO at 5 ng was combined with Su(H)2 MO1 and Su(H)2 MO2 at 0.5 ng each. The independent Su(H)2 MOs gave indistinguishable results, but with a low percentage of affected embryos. Therefore, both Su(H)2 MOs were co-injected to increase the penetrance, and are used in combination throughout the results, unless otherwise stated (see Table 1). Morpholinos were injected into the embryo at the one cell stage.

Retinoic acid treatment

All trans retinoic acid (RA, Sigma) was dissolved in DMSO to make a stock solution of 10 mM. This was then diluted in E3 medium to give a final concentration of 10^− 9, 10^− 12 or 10^− 15 M RA. Embryos were incubated in their chorions in the solution containing RA from tailbud stage to the time point at which they were fixed for analysis.

mRNA synthesis and injection

Capped RNA was synthesized from Dominant Active (DA) Suppressor of Hairless expression construct in pCS2+ vector (gift from David Wilkinson's lab, NIMR, London) using the mMessage mMachine SP6 transcription kit (Ambion). Full length coding sequence of the Su(H)1 and Su(H)2 cDNAs lacking 5′ or 3′ UTRs were subcloned into pCS2+ for expression in the embryos. One cell stage zebrafish embryos were pressure injected with the RNA diluted to 0.2–1 μg/μl in 0.1 M KCL plus 0.2 mg/mL Fast Green as tracer dye.

In situ hybridizations

In situ hybridizations were carried out as previously described (Oates and Ho, 2002), and double in situs were performed as described by Prince et al. (1998). The riboprobes were generated from plasmids as already described: her1, her7, dlc, titin (Oates and Ho, 2002). The cmcl2 and cyp26a1 plasmids were from the labs of S. Abdelilah and M. Brand respectively. Riboprobes transcribed from the Su(H) genomic clones described above were used with annealing temperatures of 50°C.

Antibody staining

10–15 somite stage embryos were fixed overnight in 4% PFA, and washed briefly in PBST, followed by 3 × 15 min in PBS containing DMSO and 0.1% Triton X-100. Embryos were then blocked at room temperature for 1 h in PBST containing BSA and 10% goat serum. The embryos were incubated overnight in primary anti-fibronectin antibody (Sigma) diluted 1:500 into block solution. Embryos were washed 4 × 15 min in PBST and then incubated in secondary antibody for 1 h at room temperature. Embryos were then washed in PBST, incubated for 15 min in PBST plus Hoechst 33342, rinsed briefly and transferred into a solution of 80% glycerol for deyolking and flat-mounted for imaging on a Zeiss confocal LSM.

Results

A second active Suppressor of Hairless gene in zebrafish

Previous work established the existence of a zebrafish homolog of the CBF1/Su(H)/Lag-1 family termed Su(H) (Sieger et al., 2003). We isolated this gene and a second homolog using a degenerate primer PCR protocol (see materials and methods). The duplicated zebrafish genes, which we term Su(H)1 (Sieger et al., 2003) and Su(H)2, are themselves highly conserved, with their proteins showing 90% amino acid identity overall (Fig. 1A). In contrast, a C-terminal serine-rich domain (45–50 aa) is highly diverged: whereas mouse and human proteins are 95% identical in this domain, Su(H)1 and Su(H)2 share only 17% identity, and only 3 residues are conserved between all four proteins. Comparison of our Su(H)1 and Su(H)2 cDNA sequences to the Ensembl v37 zebrafish genome assembly, and PCR amplification of gene-specific intron sequences demonstrates unequivocally that they are distinct genes (data not shown). Analysis of mRNA expression by in situ hybridization shows that both transcripts are maternally provided (Figs. 1B,F) and detectable throughout the embryo until around 24 h post fertilization (hpf), beyond which point both transcripts become restricted to the head (Figs. 1E,I). To control for possible cross hybridization between cDNA-derived probes and Su(H) target mRNA due to very high nucleotide similarity between the Su(H)1 and Su(H)2 transcripts, we performed additional experiments with riboprobes derived from unique intron sequences from each gene, but did not detect any differences (data not shown). These expression results confirm those of Sieger et al. (2003) for Su(H)1 and indicate that Su(H)1 and Su(H)2 are co-expressed throughout the first day of development.

Su(H)1 and Su(H)2 in neurogenesis and heart development and laterality

Finding two highly conserved Su(H) genes co-expressed throughout the embryo raises the hypothesis that they may both be required in a given cell to mediate normal levels of Notch signal transduction. Alternatively, they may be completely redundant, and only a loss of both Su(H) functions will cause a loss of Notch signalling competence. We examined the 5′ sequence of each Su(H) gene and found that the previously used antisense morpholino reagents (Sieger et al., 2003) do not distinguish between these highly similar transcripts, raising the possibility that both genes were inadvertently targeted in that study (Supplementary Figure S1). Before turning our attention to the implications of this observation for somitogenesis, we tested our hypothesis in two independent developmental contexts where Notch signalling is known to be important. We first injected morpholinos targeted to distinct regions of Su(H)1 and Su(H)2, both alone and in combination, and examined their effect on the process of lateral inhibition, a pathway well documented in zebrafish and other animals to be regulated via the Notch signalling pathway (Appel and Eisen, 1998a and Inoue et al., 1994). When either of the two Su(H) genes are down-regulated we observe that an excess of Rohon–Beard neurons are made throughout the neural axis, as identified by islet1 expression and their position in the CNS (Figs. 2B,C). When both genes are down-regulated we see a much more extensive number of neurons, including Rohon–Beard, ventral motor neurons, and also an enlargement of the trigeminal ganglia along the DV axis (Figs. 2D,F), suggesting an additive role for Su(H) genes during neurogenesis. This effect is highly similar to that seen in a mindbomb mutant, which exhibits the strongest known Notch-related neurogenic phenotype (Jiang et al., 1996 and Itoh et al., 2003). This experiment indicates that knockdown of either individual Su(H) gene does not match the phenotypic strength of a loss of Notch signalling, whereas a reduction of both Su(H) functions does.

Notch signalling has also been implicated in the early establishment of left–right asymmetry in the viscera of chick, mouse and zebrafish embryos, the outcome of which can be accurately monitored by following the situs of the developing heart (Przemeck et al., 2003; Krebs et al., 2003; Raya et al., 2003; Kawakami et al., 2005). We therefore next compared the effect of knocking down the individual Su(H) genes with that of a combined knockdown on heart development and looping. In embryos with impaired functioning of the Su(H)1 or Su(H)2 gene, 30% display defects in elongation of the heart tube (Figs. 3B,E), 50% have defects in the heart primordium migrating from the midline (Figs. 3C,F), and 20% have a normal heart as judged by position, morphology, and expression of cardiac myosin light chain (cmlc, Fig. 3D). Thus loss of only a single Su(H) gene, reveals an apparently normal situs, but causes strong defects in heart tube maturation and subsequent movements. In contrast, when both Su(H) genes are down-regulated the position of the heart becomes randomised, being found on the right hand side in 30% of the embryos (Figs. 3G–I). Combined, these experiments indicate that in neurogenesis, heart development and laterality, a strong or complete Notch phenotype is only generated when both of the zebrafish Su(H) gene functions have been targeted, and that, depending on the developmental context examined, the individual Su(H) genes can play an additive role in the process.

Su(H)1 and Su(H)2 are involved in bilateral positioning of somite boundaries

To date, mutant and morpholino-generated somitogenesis phenotypes in the Notch signalling pathway have exhibited such profound morphological and molecular defects, that more subtle abnormalities such as asymmetries in boundary formation and cyclic expression patterns may have been overlooked. For example, loss of Notch1a, DeltaD, DeltaC or Mindbomb function in zebrafish leads to chaotic, partial boundary formation, and a complete and bilaterally symmetrical loss of cyclic expression waves (van Eeden et al., 1996, Holley et al., 2000, Holley et al., 2002, Jiang et al., 2000, Gajewski et al., 2003, Itoh et al., 2003, Oates and Ho, 2002 and Oates et al., 2005). We reasoned that the additive nature of the duplicate Su(H) genes might offer a way to look at roles for Su(H) directly, and perhaps for Notch signalling in general, in the coordination of bilateral symmetry during somitogenesis.

Morpholinos were injected at the one cell stage and analysed at 24 h post fertilization (hpf) for defects in myotome boundaries marked by titin expression. Injection of the morpholino targeting Su(H)1 produced defects only in the posterior body: from approximately somite boundary 9 onwards no clear myotome boundaries could be identified (Fig. 4B) and tail outgrowth was affected. From the titin staining it could be seen that the boundaries 1–9 form symmetrically without any apparent defects, a phenotype very similar to the known Delta/Notch mutants.

When morpholinos designed to target Su(H)2 were injected singularly or combined in a lower dose, they produced defects in both anterior and posterior segment boundaries (Fig. 4C). The posterior boundaries were initially closer together and eventually the embryo stopped making boundaries altogether, concomitant with a mild tail outgrowth defect (Fig. 4C). The unusual nature of the anterior defects promoted us to examine them in more detail. Embryos stained for fibronectin accumulation clearly revealed asymmetric anterior boundary formation, as well as boundaries that did not span the entire mediolateral width of the somite (Figs. 4G–G″). In live Su(H)2 morpholino injected embryos the somite boundaries in the anterior were asymmetric or missing altogether (Figs. 4I–K), indicating that the abnormalities arose during somitogenesis, and were not the result of a later defect in myotome maturation. The asymmetries observed appeared to show no left or right sided bias within any clutch of embryos.

When morpholinos against Su(H)1 and Su(H)2 were co-injected and the embryos analysed at 24 hpf, we observed severe defects in boundaries along the entire axis of the embryo, with no consistent bias toward the left or right hand side observed, and a strong tail outgrowth phenotype (Fig. 4D). The defects are more severe than in embryos injected with Su(H)2 morpholino alone, indicating a function for the Su(H)1 gene also in the anterior trunk. These results suggest that the somite phenotypes of the Su(H) genes are not simply additive, but may reflect a redundant biochemical function, or alternatively, participation in a parallel or robust mechanism.

The difference in morphological phenotype displayed by Su(H)1 and Su(H)2 knockdown embryos indicates the specificity of the morpholino targeting and action (summarized in