Delta-Notch signaling induces hypochord development in zebrafish

During vertebrate development dorsally located cells, known as the gastrula or Spemann-Mangold organizer, provide a source of signals that induce embryonic axis formation. After gastrulation, cells that originated within the organizer occupy the midline of all three germ layers: floor plate of ventral neural tube, mesodermal notochord ventral to neural tube and dorsal endoderm (Spemann, 1938; Schoenwolf and Sheard, 1990; Selleck and Stern, 1991; Gont et al., 1993; Catala et al., 1996; Wilson and Beddington, 1996). Also at the midline of anamniote embryos, such as axolotls, frogs and fish, are hypochord cells, which form a single row directly ventral to notochord (Lofberg and Collazo, 1997; Cleaver et al., 2000; Eriksson and Lofberg, 2000). Midline cells are an important source of signals that pattern nearby tissues subsequent to axis formation. For example, notochord and floor plate secrete Sonic hedgehog (Shh), which patterns ventral neural tube and somites (Tanabe and Jessell, 1996; Stickney et al., 2000), and hypochord cells secrete VEGF, which appears to be required for dorsal aorta formation (Cleaver and Krieg, 1998). Thus, an important step toward understanding vertebrate development is understanding the mechanisms that specify midline cells.

In zebrafish, development of midline cells requires the Delta ligand-Notch receptor signaling system. Specifically, embryos with a mutation of deltaA (dla), one of four known zebrafish Delta homologs, have a deficit of floor-plate and hypochord cells and excess notochord cells; by contrast, overexpression of dla blocked notochord development and produced excess floor-plate and hypochord cells (Appel et al., 1999). Fate mapping had shown that notochord and floor-plate precursors are close together in the zebrafish organizer (Shih and Fraser, 1995; Melby et al., 1996) and functional analyses of no tail (ntl), which encodes a T-box transcription factor homologous to mouse Brachyury, had indicated that Ntl mediates a choice between notochord and floor-plate fates (Halpern et al., 1997). As gastrula stage zebrafish embryos express Delta and Notch genes (Bierkamp and Campos-Ortega, 1993; Dornseifer et al., 1997; Westin and Lardelli, 1997; Haddon et al., 1998; Appel et al., 1999), we hypothesized that Delta-Notch signaling during gastrulation regulates specification of zebrafish midline precursors for notochord, floor-plate and hypochord fates by regulating ntl expression.

We present tests of hypochord specification. By fate mapping, we found that hypochord precursors arise from the edge of the shield, which is the zebrafish gastrula organizer. These cells are closely associated with cells that become either notochord or slow muscle but not floor plate. Presumptive hypochord precursors express notch5 and her4, a Notch target gene, during gastrulation. Initially, all her4-expressing cells express ntl, which marks midline precursors, but later they do not. This raises the possibility that hypochord precursors emerge from the ntl-positive midline precursor population. Cells that express her4 are close to cells of the paraxial mesoderm, which express deltaC (dlc) and deltaD (dld), and loss-of-function experiments show that dla, dld and dla, which gastrulating embryos uniformly express, are redundantly required for her4 expression and hypochord development. Finally, we show that unregulated Notch activity inhibits ntl expression. We propose that cells at the lateral edges of the midline precursor domain can develop either as notochord or hypochord, and that hypochord development is induced by Delta ligands expressed by neighboring paraxial mesoderm cells.

Embryos were collected from single pair matings and raised at 28.5°C in embryo medium (15 mM NaCl, 0.5 mM KCl, 1 mM CaCl₂, 1 mM MgSO₄, 0.15 mM KH₂PO₄, 0.05 mM NH₂PO₄, 0.7 mM NaHCO₃) and staged according to hours postfertilization (hpf) and morphological criteria (Kimmel et al., 1995). The aei mutant allele we used, originally designated aei^tr233 (van Eeden et al., 1996) and redesignated aei^AR33, creates a stop codon within the fifth EGF-like repeat of dld (Holley et al., 2000).

We injected one- to four-cell stage zebrafish embryos with a 2% solution of DMNB-caged fluorescein dextran (Molecular Probes) in 1× Danieau solution. Injected embryos were maintained in the dark at 28.5°C in embryo medium until shield stage. Embryos were dechorionated with watchmaker’s forceps and mounted in 2% methyl cellulose on bridged slides with the dorsal side of the embryos facing upwards. The dye was photoactivated in three to five cells using 5-10 second pulses of 365 nm light (Serbedzija et al., 1998) generated by a Photonics Micropoint Laser System focused using a 40× objective mounted on a Zeiss Axiskop. Dye activation was confirmed using epifluorescence optics. Labeled embryos were maintained at 28.5°C until 24 hpf, at which time the fates of the labeled cells were determined by examining living embryos with a compound microscope. Images were obtained using a Hammamatsu Orca CCD camera. Embryos selected for histochemical analysis were fixed in 4% paraformaldehyde.

Previously described probes include those for notch5 (Westin and Lardelli, 1997), notch1a (Bierkamp and Campos-Ortega, 1993), her4 (Takke et al., 1999), myod (Weinberg et al., 1996), ntl (Schulte-Merker et al., 1994), dld and dlc (Haddon et al., 1998), col2a1 (Yan et al., 1995), and isl1 (Appel et al., 1995). In situ RNA hybridization was performed essentially as described (Hauptmann and Gerster, 2000; Jowett, 2001). For double labeling, best results were obtained by staining with BCIP/NBT (Roche Diagnostics) first, followed by staining with BCIP/INT (Roche Diagnostics). To convert the fluorescent signal in photoactivated embryos to a blue precipitate, the embryos were incubated with alkaline phosphatase-conjugated anti-fluorescein antibody (Roche Diagnostics) at 1:10,000 dilution followed by staining with BCIP/NBT. Slow muscle cells were detected by incubating embryos with monoclonal F59 antibody (gift of Frank Stockdale) at 1:10 dilution, followed by incubation with Alexa Fluor 568 goat anti-mouse IgG conjugate (Molecular Probes) at 1:200 dilution. Myc expression was detected using monoclonal anti-Myc IgG antibody (Oncogene Research Products) at 2 μg/ml, followed by goat anti-mouse IgG (Jackson ImmunoResearch Laboratories) at 1:500 dilution, then streptavidin-conjugated peroxidase (Jackson ImmunoResearch Laboratories) at 1:1000. Peroxidase activity was used to produce a brown precipitate in the presence of Fast DAB (Sigma).

Embryos for sectioning were embedded in 1.5% agar/5% sucrose and frozen in 2-methyl-butane chilled by liquid nitrogen. Sections (10 μm) were obtained using a cryostat microtome.

Whole-mount embryos were cleared in methanol, mounted in 75% glycerol and photographed with a Spot digital camera (Diagnostic Instruments) mounted on a compound microscope. Images of sectioned material were also obtained with the Spot camera.

Morpholino antisense oligonucleotides (Gene Tools, LLC.) for dla (5′-CTTCTCTTTTCGCCGACTGATTCAT-3′) and dlc (5′-AGCACGTAATAAAACACGAGCCAT-3′) were resuspended in water. Working dilutions of 2.5-5.0 mg/ml were made using 1×Danieau solution and approximately 5 nl injected into one- to two-cell stage embryos collected from matings of wild-type or aei^AR33/+ fish using a pressure injector (Applied Scientific Instruments, Inc.). Injected embryos were fixed at yolk-plug closure, three-somite stage or 24 hpf in 4% paraformaldehyde, and were processed for in situ RNA hybridization. Synthetic mRNA encoding the intracellular domain of Xenopus laevis Notch1 fused to the Myc tag (NICDMT) (Wettstein et al., 1997) was produced from a construct kindly provided by Chris Kintner using the mMessage Machine kit from Ambion. Approximately 5 nl mRNA at 100 ng/μl was injected into a single cell of eight-cell stage embryos. To make a construct for heat-shock induction of NICDMT, we moved coding sequence from our mRNA expression plasmid and fused it to the zebrafish heat shock 70 promoter (Shoji et al., 1998) in pBluescript and injected approximately 5 nl of the resulting plasmid at 50 μg/μl in the blastodisc of one-cell zygotes. Injected embryos were raised at 28.5°C until shield stage, and subjected to three cycles of 37°C for 15 minutes followed by 28.5°C for 45 minutes. After the end of heat shock cycles, the embryos were raised at 28.5°C until yolk plug closure stage and fixed in 4% paraformaldehyde.

Previous fate maps revealed that zebrafish floor-plate and notochord precursors arise from the embryonic shield, which is centered on the dorsal margin of gastrula stage embryos (Shih and Fraser, 1995; Melby et al., 1996). However, trunk hypochord precursors have not been placed on any fate map. To identify the location of hypochord precursors, we labeled cells near the margin of shield-stage embryos by photoactivating caged fluorescein. Single labeled cells near the dorsal midline margin produced notochord, whereas clusters of labeled cells centered a few cell diameters farther from the dorsal midline margin gave rise to notochord, floor-plate and ventral neural tube cells, and, occasionally, tail hypochord (data not shown). These observations are consistent with published fate maps (Shih and Fraser, 1995; Melby et al., 1996). We labeled trunk hypochord cells only when we photoactivated clusters of approximately three to five cells at the margin about 15 cell diameters away from the dorsal midline, at the edge of the shield (Fig. 1A,B). Labeled cell clusters that produced only hypochord cells were rare. Instead, we usually observed the following cell types produced by single labeled clusters: hypochord and notochord, hypochord and muscle, hypochord, notochord and muscle, notochord and muscle, or only notochord or muscle (Table 1). We sectioned some 24 hour post fertilization (hpf) embryos that contained labeled hypochord and muscle cells and found that labeled muscle cells were always at the lateral edge of the somite (Fig. 1C), in the position of slow muscle cells (Devoto et al., 1996). We confirmed that these cells were slow muscle by labeling with F59 antibody (Fig. 1D), which identifies zebrafish slow muscle (Devoto et al., 1996). Slow muscle cells arise from adaxial cells, which are next to notochord during gastrulation and early segmentation stages (Devoto et al., 1996). We conclude that hypochord precursors are closely associated with notochord and slow muscle precursors and not with floor-plate precursors in shield stage embryos.

We have begun to follow the movements of labeled hypochord precursors during gastrulation from their starting positions at the edge of the shield to their final positions directly beneath the notochord. Within 1 hour of photoactivation, the cells of the polyclones began to be distributed in the anterior-posterior axis (data not shown). By 3 hours after photoactivation, near the end of the gastrula stage, the labeled cells formed a row next to notochord, which could be identified by morphology (Fig. 1E). Sometimes notochord cells were also labeled. Labeled hypochord cells moved below notochord by about 11-11.5 hpf (data not shown). We conclude from our fate-mapping observations that hypochord precursors originate between notochord precursors and adaxial cells, and migrate to their final positions below the notochord.

Cells bordering the posterior dorsal midline of mid and late gastrula stage embryos express notch5 and her4 (Westin and Lardelli, 1997; Takke et al., 1999) (Fig. 2A-D). The distribution of these cells is similar to the distribution of hypochord precursors as they move towards the dorsal midline (see above). To investigate the identity of cells that express notch5 and her4, we examined sectioned material from embryos hybridized with RNA probes. Sagittal and transverse sections of 9.5 hpf embryos at the end of the gastrula stage showed that deep mesendodermal cells, close to the yolk, express notch5 and her4 (Fig. 2E-G,I-L). her4- or notch5-positive cells are next to adaxial cells, marked by myod expression (Fig. 2F,G). We never observed cells that express both notch5 and myod or her4 and myod. Cells that express her4 and notch5 also were closely associated with ntl-positive midline precursors (Fig. 2H-K). Posteriorly, all her4-positive cells express ntl (Fig. 2I). More anteriorly some her4-positive cells express ntl but others do not (Fig. 2J). Similarly, some notch5-positive cells express ntl, whereas others do not (Fig. 2K). The similarity of her4 and notch5 RNA distribution indicated that the same cells might express these genes, which we confirmed by double labeling (Fig. 2L).

One interpretation of our data is that some ntl-expressing cells at the lateral edges of the midline precursor domain initiate notch5 and her4 transcription and that, later, these same cells downregulate ntl. Based on our fate map data described above, we predict that these cells are hypochord precursors. Consistent with this prospect, after gastrula stage, by 11.5 hpf, notch5- and her4-expressing cells occupy a single row ventral to notochord, where hypochord cells differentiate (Fig. 2M and data not shown).

The coincident expression of notch5 and her4 is consistent with the possibility that Notch5 activity promotes her4 expression. However, we detect her4 RNA at the same time or, perhaps, slightly earlier than notch5 RNA. Midline cells express at least two other notch genes, notch1a and notch1b (Bierkamp and Campos-Ortega, 1993; Westin and Lardelli, 1997) (Fig. 2N). This observation opens the possibility that multiple Notch receptors promote her4 transcription in presumptive hypochord precursors.

Taken together, our data indicate that hypochord precursors arise at the lateral edges of the ntl-expressing midline precursor domain and that similarly positioned notch5/her4-positive presumptive hypochord precursors downregulate ntl expression. These observations raise the possibility that hypochord cells originate as ntl-expressing midline precursors.

Gastrulation stage zebrafish embryos express three of four known Delta genes. During mid and late gastrulation stages, embryos express dla uniformly, except for a few cells near the dorsal midline margin that transiently express dla at elevated levels (Appel et al., 1999). By contrast, paraxial mesoderm, but not axial mesoderm, strongly expresses dlc and dld (Dornseifer et al., 1997; Haddon et al., 1998) (Fig. 3A-F). Double labeling experiments showed that dlc and dld-expressing cells are adjacent to ntl-positive midline precursors (Fig. 3G,H and data not shown). Additionally, we found that dlc-positive cells are next to her4-expressing cells (Fig. 3I). This raises the possibility that her4 expression is induced by DeltaC and DeltaD ligands expressed by neighboring paraxial mesoderm cells.

We used loss-of-function strategies to test the role of Delta-Notch signaling in hypochord specification. Previously, we showed that dla^dx2 mutant embryos, with incomplete penetrance and variable expressivity, had fewer hypochord cells than wild-type embryos (Appel et al., 1999). Likewise, aei^AR33 mutant embryos, which lack dld function (Holley et al., 2000), have an incompletely penetrant and variable reduction of hypochord cell number (Fig. 4B; Table 2). We do not know of a dlc mutation, thus, we injected wild-type embryos with dlc antisense morpholino oligonucleotides (MO) to interfere with dlc translation (Heasman et al., 2000; Nasevicius and Ekker, 2000). On average, injected embryos also had fewer hypochord cells than uninjected embryos (Fig. 4C; Table 2). These data raise the possibility that Delta genes are functionally redundant for hypochord development. To test this, we injected embryos produced by intercrosses of heterozygous aei^AR33 adults with approximately 25 ng dlc MO, a dose that did not produce morphological defects when injected into wild-type embryos (data not shown). At 24 hpf, injected homozygous aei^AR33 mutant embryos were identified by the somite defect that results from loss of dld function (van Eeden et al., 1996; Holley et al., 2000). Very few hypochord cells developed in homozygous aei^AR33 mutant embryos injected with dlc MO (Fig. 4D; Table 2). This result supports the idea that dlc and dld have redundant functions in hypochord development. Within the sibling embryos, we could identify two classes; approximately one-third had hypochord defects similar to wild-type embryos injected with dlc MO, whereas approximately two-thirds had hypochord phenotypes intermediate to wild-type and homozygous aei^AR33 mutant embryos injected with dlc MO (data not shown). We speculate that these two classes represent homozygous wild-type and heterozygous aei^AR33 mutant embryos, respectively. We also tested loss of dla function by MO injection. We observed only small gaps in the hypochords of some of the wild-type embryos injected with dla MO (Table 2). However, aei^AR33 mutant embryos injected with dla MO had fewer hypochord cells than uninjected aei^AR33 mutant embryos (Table 2). Therefore, dla function also may contribute to hypochord development. These data support the possibility that paraxial mesoderm expression of dld and dlc and uniform expression of dla promote hypochord development.

We have identified her4-expressing cells as hypochord precursors and Notch activity regulates her4 expression (Takke et al., 1999). To test if her4 expression by presumptive hypochord precursors requires Delta functions, we injected embryos produced by heterozygous aei^AR33 adults with dlc MO and examined them at 9.5 hpf. Of 49 embryos probed for her4, 17 (35%) had substantially fewer her4-expressing cells (Fig. 4F) and 11 (22%) had no her4-expressing cells (Fig. 4G). These data indicate that DeltaC and DeltaD ligands promote her4 expression. By contrast, all injected embryos expressed notch5 normally (Fig. 4E) indicating that notch5 expression by presumptive hypochord precursors does not require DeltaC and DeltaD. Taken together with our expression pattern data, these observations provide strong evidence that DeltaC and DeltaD, expressed by adaxial cells, induce her4 expression in neighboring midline precursors during gastrulation, specifying them for hypochord fate.

To test the specificity of our MO effects we examined primary neurogenesis. Neural plate cells express dla and dld but not dlc (Dornseifer et al., 1997; Appel and Eisen, 1998; Haddon et al., 1998). dla^dx2 mutant embryos have a large excess of primary neurons, perhaps because of the dominant negative nature of the allele (Appel et al., 1999; Appel et al., 2001). aei^AR33 mutant embryos have only a small increase in the number of primary neurons (Holley et al., 2000) (Fig. 4I). Twenty-four percent of embryos (22/90 embryos) produced by an intercross of aei^AR33/+ adults and injected with approximately 25 ng dlc MO had primary neuron phenotypes similar to uninjected aei^AR33 homozygous mutant embryos (Fig. 4J). Thus, reduction of dlc function did not appear to enhance the neural phenotype produced by loss of dld function, indicating that the dlc MO did not interfere, nonspecifically, with neurogenesis. By contrast, we found that injection of dla MO produced excess primary neurons in a dose-dependent manner when injected into wild-type embryos (data not shown). Injection of approximately 5 ng dla MO, which did not cause a strong neural phenotype in wild-type embryos, greatly enhanced the neural phenotype of aei^AR33 mutant embryos (17/74 embryos; 23%) (Fig. 4K). As the dla and dlc MO phenotypes correlate with the dla and dlc expression patterns, we conclude that the MO effects are specific to their intended targets.

One interpretation of our results is that Delta signaling, primarily from paraxial mesoderm, to midline precursors expressing Notch receptors inhibits ntl expression and promotes her4 expression, consequently instructing these cells to develop as hypochord. We tested this by expressing a constitutively active form of Notch. To perform these experiments, we found it necessary to limit the number of cells that expressed constitutively active Notch so that embryos would develop fairly normally. First, we confirmed and extended a previous observation that Notch activity promotes her4 expression (Takke et al., 1999) by using a heat shock promoter (Shoji et al., 1998) to express during gastrulation a constitutively active form of X. laevis Notch1 (Wettstein et al., 1997) fused to the Myc tag (NICDMT). This resulted in embryos that had ligand-independent Notch activity in some but not all cells. Of 57 injected embryos, 44 had numerous cells that expressed NICDMT, identified by staining for the fused Myc epitope. Many of these cells expressed her4 (data not shown). Analysis of sectioned embryos revealed that NICDMT-positive cells ectopically expressed her4 (Fig. 5A). We never saw ectopic her4-expressing cells that did not express NICDMT, indicating that Notch induction of her4 expression is cell autonomous. Next, we injected one cell of eight-cell stage embryos with synthetic mRNA encoding NICDMT, raised the embryos to the end of the gastrula stage and probed for ntl expression. Twenty-two of 116 (19%) injected embryos had midline cells that expressed NICDMT; the ntl expression domain was reduced in each of these indicating that unregulated Notch activity in midline precursor cells inhibits ntl expression (Fig. 5B,C). Tissue sections revealed that NICDMT-positive midline cells did not express ntl, whereas neighboring cells did (Fig. 5D), indicating that Notch activity cell autonomously inhibits ntl expression. Together, these observations show that Notch activity can inhibit ntl expression and promote her4 expression, consistent with our hypothesis that Delta-Notch signaling induces a subset of midline precursors to develop as hypochord.

We have shown previously that dla mutant embryos have excess notochord cells and deficits of floor-plate and hypochord cells, whereas wild-type embryos injected with dla mRNA have fewer notochord cells and excess floor-plate and hypochord cells (Appel et al., 1999). We hypothesized that Delta-Notch signaling during gastrulation mediated cell fate decisions made by midline precursors. However, because we did not know the exact spatial relationships of midline precursors, we could not determine the mechanisms by which Delta-Notch signaling mediated midline cell specification. We have considered two broad possibilities. First, midline precursors might have equivalent developmental potential. Delta-Notch signaling between neighboring cells might, in a process called lateral specification (Artavanis-Tsakonas et al., 1995; Greenwald, 1998), cause midline precursors to take different fates. An example of this type of mechanism is specification of the anchor cell and ventral uterine precursor cell during formation of the Caenorhabditis elegans gonad (Seydoux and Greenwald, 1989). In this case, we would expect zebrafish midline precursors to be intermixed. Second, midline precursors might be induced to develop differently by Delta-Notch-mediated signals from a separate cell population, similar to the way that equivalent blastomeres are specified for different fates by signaling from a nonequivalent blastomere in C. elegans (Mickey et al., 1996). We would predict that midline precursors that produce different cell types occupy distinct positions relative to the source of inductive signal. Our results from a combination of fate mapping, gene expression pattern analyses and tests of gene function lead us to favor the latter mechanism for specification of hypochord cells in zebrafish. Specifically, we propose that, during gastrulation, high levels of Delta signals from paraxial mesoderm induce neighboring midline precursors to develop as hypochord cells, whereas nearby midline precursors not exposed to high levels of Delta signals develop as notochord (Fig. 6).

Our data indicate that hypochord precursors are not intermixed with other precursors throughout the shield but occupy a distinct domain at the lateral edges of the shield. When we labeled small clusters of cells at the lateral edge of the shield, those that gave rise to hypochord often also produced notochord and/or slow muscle. Rarely, if ever, did these polyclones also comprise floor-plate and other ventral neural cells, fast muscle cells or endoderm. By contrast we never observed trunk hypochord cells in polyclones that originated at the dorsal midline, although we occasionally did see tail hypochord cells. Thus, trunk hypochord precursors apparently arise only from the lateral edges of the shield, in close association with notochord and slow muscle precursors. Because of the small number of hypochord precursors, we currently are unable to determine the precise arrangement of hypochord, notochord and slow muscle precursors. However, our data showing that hypochord precursors map to the edge of the shield, coincident with the edge of the notochord precursor domain (Shih and Fraser, 1995; Melby et al., 1996), are consistent with our hypothesis that hypochord cells are specified from midline precursors by an inductive signal from muscle precursors.

Fate mapping of axolotl and histological analyses of frog and fish embryos indicated that hypochord is derived from endoderm (Lofberg and Collazo, 1997; Cleaver et al., 2000; Eriksson and Lofberg, 2000). However, labeled hypochord cells did not appear in a fate map of zebrafish endoderm (Warga and Nusslein-Volhard, 1999). In addition, mutations of one-eyed-pinhead and casanova, which are required for endoderm development (Schier et al., 1997; Alexander et al., 1999), do not result in loss of trunk hypochord (Alexander et al., 1999) (and data not shown). Our data showing hypochord precursors are closely associated with precursors of other mesodermal cell types at the onset of gastrulation indicate that zebrafish hypochord originates from mesoderm and not endoderm.

Previous publications showed that cells that form two columns on either side of the midline of late-gastrulation stage embryos transcribe notch5 and her4 (Westin and Lardelli, 1997; Takke et al., 1999). We demonstrated here that the same cells expressed notch5 and her4, and that they were deep within the mesoderm, close to the yolk. The distribution of notch5- and her4-expressing cells during late gastrula stage was similar to distribution of photoactivated cells that went on to produce hypochord. After gastrula stage, notch5- and her4-expressing cells were ventral to notochord, in the final position of hypochord. We also found that cells that expressed notch5 and her4 during late gastrulation were next to adaxial cells, the slow muscle precursors. This is consistent with our fate-mapping results that hypochord-containing polyclones often contained slow muscle cells. notch5- and her4-expressing cells also were close to ntl-positive midline precursors, again consistent with our observations that polyclones often contained both hypochord and notochord cells. Thus, notch5- and her4-expressing cells are probably hypochord precursors.

We showed previously that gastrulating embryos express dla uniformly, except for transient upregulation at the dorsal midline (Appel et al., 1999). Thus, the pattern of dla transcription does not predict the pattern of hypochord precursor specification. We now show that paraxial mesoderm, which includes adaxial cells adjacent to midline precursors, expresses dlc and dld. Thus, adaxial cells could be the primary source of Delta signals that induce hypochord development.

Results from our loss-of-function experiments support the idea that dlc and dld signaling from adaxial cells induce hypochord development and also indicate that dla may contribute to hypochord specification. Embryos that lack dld function, by mutation, or have reduced dlc function, by MO injection, have incompletely penetrant and variably expressive hypochord phenotypes. By contrast, all embryos that lack both dlc and dlc functions have very few hypochord cells, indicating that dld and dlc are functionally redundant for hypochord specification. We showed previously that embryos having the dominant negative dla^dx2 mutant allele have fewer hypochord cells (Appel et al., 1999). This mutant allele, which is predicted to substitute tyrosine for a conserved cysteine in the second EGF repeat of the extracellular domain, may produce DeltaA protein that can interfere with the functions of DeltaD and DeltaC, perhaps by forming heterodimers. As gastrulating embryos uniformly express dla transcripts, mutant DeltaA protein could interfere with the hypochord-inducing functions of DeltaD and DeltaC. However, we found here that reduction of dla function by antisense morpholino oligonucleotide injection also reduced the number of hypochord cells in wild-type embryos and enhanced the penetrance and expressivity of the aei/dld mutant phenotype. This apparent requirement for dla activity raises the question of whether a localized source of Notch ligand can fully account for the pattern of hypochord specification. However, dla is not sufficient to specify hypochord in the absence of dlc and dld. Thus, we propose that high levels of Delta proteins produced by adaxial cells are required to induce midline precursors to develop as hypochord (Fig. 6).

How might Delta signals induce hypochord development? One key might be regulation of ntl expression. ntl mutant embryos lack notochord and rostral hypochord and have excess floor plate (Rissi et al., 1995; Halpern et al., 1997). Halpern et al. (Halpern et al., 1997) proposed that ntl regulates a midline precursor fate decision by promoting notochord and inhibiting floor-plate development. We propose that modulation of ntl expression within midline precursors by Delta-Notch signaling is required for hypochord development. In our model, ntl promotes formation of a population of midline precursors that have the potential to develop either as notochord or hypochord. Activation of Notch in a subset of precursors by Delta ligands expressed by neighboring paraxial mesoderm cells induces her4 and represses ntl expression. Consistent with this, we show here that constitutive Notch activity can cell-autonomously drive ectopic her4 expression. In the absence of Notch activity, her4 expression is not induced, as we demonstrated here, and excess midline cells express ntl, as we showed previously (Appel et al., 1999). Thus, Notch activity diverts midline precursors from notochord to hypochord fate.

As her4 is a member of the hairy-Enhancer of split gene family, which generally encode transcription repressors (reviewed by Fisher and Caudy, 1998), Notch inhibition of ntl expression could occur via direct action of Her4 on ntl regulatory elements. Notably, in the tunicate Ciona intestinalis, Notch activity promotes expression of the Brachyury gene, apparently by direct binding of Supressor of Hairless (Corbo et al., 1997; Corbo et al., 1998). Direct comparison of the regulatory DNA that controls expression of Brachyury homologs in tunicates and zebrafish may provide interesting insights to the evolution of the embryonic midline.

Thanks to Lila Solnica-Krezel and Jacek Topczewski for comments on the manuscript, to Marnie Halpern for discussions, to Frank Stockdale, John Kuwada, Michael Lardelli, José Campos-Ortega and Chris Kintner for reagents, and to the Vanderbilt zebrafish facility staff for fish care. This work was supported by NIH grant HD38118.