lunatic fringe regulates delta-notch induction of hypochord in zebrafish



Cells that occupy the midline of vertebrate embryos arise from a common population of precursors. Several lines of evidence indicate that Delta-Notch signaling regulates specification of midline precursors for different fates. We show that zebrafish midline precursors transiently express lunatic fringe, which encodes a glycosyltransferase that modifies Notch activity in response to its ligands, and that lunatic fringe function is required for Delta-Notch–mediated induction of hypochord cells at the lateral borders of the midline precursor domain. Development Dynamics 228:281–286, 2003. © 2003 Wiley-Liss, Inc.


Vertebrate embryos are patterned by signals that arise from the Spemann–Mangold organizer, a discrete group of dorsally located cells. In addition to directing the development of cells located at a distance, the organizer gives rise to cells that occupy the embryonic midline in all three germ layers. These layers include floor plate of the ventral neural tube, mesodermal notochord, 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). In anamniote embryos hypochord also occupies the midline, lying between notochord and dorsal aorta (Lofberg and Collazo, 1997; Cleaver et al., 2000; Eriksson and Lofberg, 2000; Latimer et al., 2002).

The close association of midline precursor cells within the Spemann–Mangold organizer raised the possibility that cell–cell signaling regulates their specification for different fates. Consistent with this, the function of Delta-like ligands and Notch receptors, which mediate many instances of cell–cell signaling throughout invertebrate and vertebrate development, are important for midline fate specification. For example, zebrafish embryos mutant for deltaA (dla) or mind bomb (mib), which encodes a ubiquitin ligase necessary for efficient Notch signaling (Itoh et al., 2003), had deficits of floor plate and hypochord cells and an excess of notochord cells (Appel et al., 1999). By contrast, overexpression of dla or the constitutively active intracellular domain of Notch (NICD) blocked notochord development and appeared to promote formation of hypochord and floor plate (Appel et al., 1999; Latimer et al., 2002). NICD expression had a similar affect on frog embryos, inhibiting notochord and expanding floor plate (Lopez et al., 2003). Notch signaling may also regulate midline fate specification in mice as embryos that were homozygous mutant for presenilin 1 and presenilin 2, which encode factors necessary for Notch signaling, lacked floor plate (Donoviel et al., 1999). However, mouse embryos mutant for Delta-like1 appeared to have excess floor plate and fewer notochord cells (Przemeck et al., 2003). The precise mechanisms by which Delta-Notch signaling regulates specification of organizer cells for different fates and whether these mechanisms are fundamentally different in amniote and anamniote embryos are problems that remain unresolved.

By using fate mapping and gene expression analyses in zebrafish, we produced evidence that hypochord cells arise from the lateral edges of the organizer, or embryonic shield, and migrate toward the midline during gastrulation (Latimer et al., 2002). Initially, hypochord precursors appeared to express no tail (ntl), which marks notochord precursors, and were adjacent to paraxial mesoderm cells, which express deltaC (dlc) and deltaD (dld). At later stages of gastrulation, hypochord precursors did not express ntl. Embryos in which dlc and dld functions were eliminated, by combination of mutation and antisense morpholino oligonucleotide injection, did not develop hypochord. One interpretation of these observations is that paraxial mesoderm induces, by means of DeltaC and DeltaD, a subset of neighboring ntl+ midline precursors cells to down-regulate ntl and develop as hypochord.

Delta-Notch–mediated induction of hypochord at the edge of the midline precursor domain might be similar to other examples of Notch signaling between cell groups. For example, in flies Notch activity in response to ligands expressed by distinct groups of neighboring cells is important for formation of the dorsoventral wing margin (Diaz-Benjumea and Cohen, 1995; Doherty et al., 1996; de Celis and Bray, 1997) and leg segments (de Celis et al., 1998; Bishop et al., 1999; Rauskolb and Irvine, 1999). In vertebrates, Notch signaling helps form borders between the iterated somites of the paraxial mesoderm (Hrabe de Angelis et al., 1997; Jen et al., 1999; Jiang et al., 2000; Holley et al., 2002; Sato et al., 2002; Dale et al., 2003). In each of these cases, Notch activity apparently is modulated by Fringe proteins (Fleming et al., 1997; Panin et al., 1997; de Celis et al., 1998; Aulehla and Johnson, 1999; Bishop et al., 1999; Rauskolb et al., 1999; Rauskolb and Irvine, 1999; Dale et al., 2003), which are glycosyltransferases that modify the extracellular domain of Notch (Bruckner et al., 2000; Hicks et al., 2000; Moloney et al., 2000). This modification makes Notch either more or less responsive to its ligands (Bruckner et al., 2000; Hicks et al., 2000; Moloney et al., 2000). In particular, at the wing margin Fringe potentiates Notch activity in response to Delta and inhibits Notch signaling in response to Serrate (Fleming et al., 1997; Panin et al., 1997). Thus, Fringe appears to refine the number of cells in which Notch signaling is active, which may help organize borders between distinct cell groups.

We hypothesized that Delta-Notch–mediated induction of hypochord cells at the border of the midline precursor domain requires Fringe activity. We show here that dorsal marginal cells of zebrafish embryos transiently express lunatic fringe (lfng) during gastrulation and that lfng+ cells border dlc+/dld+ cells. Injection of lfng antisense morpholino oligonucleotides (MO) blocked formation of hypochord but not floor plate cells, similar to reduction of notch1a, dlc, and dld functions. Finally, lfng MO enhanced the hypochord phenotype of aei/dld mutant embryos. We conclude that Lfng potentiates Notch activity in a subset of midline precursors in response to Delta ligands.


Dorsal Midline Cells Express lunatic fringe During Gastrulation

We used degenerate primers and polymerase chain reaction (PCR) to clone cDNA fragments of two fringe homologs, lunatic fringe (lfng) and radical fringe (rfng). Sequencing revealed that the lfng clone is identical to a gene described previously (Leve et al., 2001; Prince et al., 2001). To determine whether these genes might influence Notch-mediated midline cell fate specification, we examined their expression by in situ RNA hybridization. We did not detect rfng expression in dorsal midline cells of gastrula-stage embryos and so focused the remainder of our work on lfng.

lfng transcripts were not evident in four-cell stage or high stage (3.3 hours postfertilization [hpf]) embryos, which indicates that lfng mRNA is not maternally contributed. Sphere stage (4 hpf) embryos expressed lfng RNA in a punctate pattern throughout the blastoderm (Fig. 1A). By dome stage (4.3 hpf), lfng+ cells were concentrated at the blastoderm margin (Fig. 1B). The first clear asymmetry in lfng expression appeared at approximately 40% epiboly (5 hpf), when deep cells at the prospective dorsal margin expressed elevated levels of lfng (Fig. 1C). Double labeling revealed that these cells also expressed goosecoid (gsc) (Fig. 1L), indicating they were fated to develop as prechordal mesoderm. At the same time, a discrete cap of expression appeared in animal pole cells (data not shown). This same general pattern was maintained through approximately 50% epiboly, with punctate distribution at the embryonic margin and expression in deep cells at the dorsal margin and in an animal pole domain (Fig. 1D). At 55% epiboly, dlc and dld expression, which marks prospective paraxial mesoderm, appeared to be complementary to lfng expression in animal cap and dorsal margin cells (Fig. 1E). At 60% epiboly, the deep dorsal expression separated from the margin (Fig. 1F), suggesting that prechordal mesoderm maintained lfng expression as it moved toward the animal pole. Consistent with this, double labeling showed that lfng+ marginal cells were anterior to ntl+ cells, which have notochord fate (Fig. 1M). dlc- and dld-expressing cells appeared to be directly adjacent to lfng+ dorsal cells (Fig. 1G). Additionally, noninvoluting forerunner cells at the dorsal margin expressed lfng (Fig. 1F,G).

Figure 1.

Dorsal margin cells transiently express lfng during gastrulation. A,B: Lateral views; C–K: dorsal views; and L–O: sagittal sections of blastula and gastrula stage embryos processed for in situ RNA hybridization. A: Sphere stage (4 hours postfertilization [hpf]). lfng expression was punctate throughout the blastoderm. B: Dome stage (4.3 hpf). Arrows indicate lfng+ cells near the embryonic margin. C,D: Arrows indicate lfng+ cells concentrated at the dorsal margin at 40% epiboly (5 hpf) and 50% epiboly (5.3 hpf). E: At 55% epiboly, lfng+ cells occupied a more narrow area of the dorsal margin (arrows) and appeared to be adjacent to cells that expressed dlc and dld (red). F: At 60% epiboly (6 hpf). Arrows indicate lfng+ cells located just above the dorsal margin, and arrowheads mark lfng+ dorsal forerunner cells. G: Dorsal lfng+ cells appeared to be adjacent to dlc- and dld-expressing cells. H: At 70% epiboly (8 hpf). Arrows mark lfng+ cells at the dorsal margin. I,J: At 85% epiboly (8.6 hpf). Embryos continued to express lfng at the dorsal margin (arrows), in cells that also expressed ntl (I) and that were adjacent to dlc/dld+ cells (J). K: At yolk plug closure (YPC) stage (10 hpf). Arrow indicates lfng expression at dorsal margin. L: At 40% epiboly. Dorsal margin cells (arrow) coexpressed gsc (red) and lfng (blue). M: At 60% epiboly lfng+ cells (blue, arrow) occupied dorsal mesendoderm, just anterior to ntl+ cells (red). N: At 85% epiboly. Dorsal margin cells (arrow) coexpressed lfng (blue) and ntl (red). O: YPC stage. lfng+ cells were deep in the nascent tail bud (arrow) and in ectoderm (ec). Scale bar = 80 μm in A–C, 160 μm in D–K, 40 μm in L–O.

By 70% epiboly, lfng expression reappeared at the dorsal margin (Fig. 1H). Marginal expression was maintained as epiboly progressed (Fig. 1I), suggesting that cells transiently expressed lfng as they involuted. These cells coexpressed the notochord marker ntl (Fig. 1I) and bordered dlc/dld+ paraxial mesoderm (Fig. 1J). At yolk plug closure (YPC) stage, cells deep in the nascent tail bud expressed lfng (Fig. 1K,O). These data show that midline cells express lfng before and during gastrulation, consistent with the possibility that lfng regulates midline cell fate specification.

lfng Is Required for Hypochord but not Floor Plate or Notochord Development

To test the hypothesis that lfng regulates midline cell fate specification, we injected one- to two-cell stage embryos with antisense morpholino oligonucleotides (lfng MO) designed to block lfng translation. When assayed at 24–28 hpf, we found that lfng MO-injected embryos had deficits of hypochord cells, marked by col2a1 RNA expression (Fig. 2B), and that the penetrance of the hypochord defect was directly dependent upon the amount of lfng MO injected (Table 1). By contrast, floor plate and notochord cell numbers appeared unaffected in embryos injected with lfng MO. Hypochord, floor plate, and notochord developed normally in embryos injected with comparable amounts of a standard control MO and a control lfri MO containing mismatches at four nucleotides of the recognition sequence (Fig. 2A; Table 1 and data not shown).

Figure 2.

lfng is required for hypochord development. A: Lateral view, dorsal up, of 24 hours postfertilization (hpf) embryo injected with lfng 4 ng of mismatch control morpholino oligonucleotides (MO). Floor plate (fp), notochord (nc), and hypochord (hc) developed normally. B: A 24 hpf embryo injected with approximately 4 ng of lfng MO had a deficit of hypochord cells. Floor plate and notochord appeared normal. C: Dorsal view, anterior up, of yolk plug closure (YPC) stage embryo injected with standard control MO and probed for her4 expression, which marks presumptive hypochord precursors (arrows), and pax2, to provide an internal hybridization control (asterisks). D:her4 expression was absent (arrows) in embryo injected with approximately 4 ng of lfng MO, whereas pax2 expression was normal (asterisks). E–G:ntl, twhh, and myoD expression in lfng MO injected embryos. Scale bar in G = 20 μm in A,B, 160 μm in C–G.

Table 1. lfri Function Is Required for Hypochord Development
MOAmount (ng)# reduced hypochord/# injected
lfri112/44 (27%)
lfri4111/187 (59%)
lfri839/51 (76%)
Standard control40/120 (0%)
notch1a857/57 (100%)

We previously produced evidence that prospective hypochord cells express her4 during gastrulation and that the Notch signaling pathway promotes her4 expression (Latimer et al., 2002). Thus, we examined her4 expression at the YPC stage in injected embryos. Embryos injected with lfng MO but not with control MOs had deficits of her4+ cells (Fig. 2C,D; Table 2). By contrast, the expression patterns of ntl, which marks notochord, twhh, which marks floor plate, and myoD, which marks adaxial mesoderm, were normal in all injected embryos (Fig. 2E–G; Table 2). We conclude that lfng function is required for hypochord specification but is not required for the development of other cell types that arise in close proximity to hypochord.

Table 2. Expression of Hypochord, Notochord, Floor Plate, and Adaxial Cell Markers in lfng MO Injected Embryos
MOAmount (ng)# reduced her4/# injected# altered ntl/# injected# altered twhh/# injected# altered myoD/# injected
lfri471/82 (86%)   
lfri8 0/34 (0%)0/17 (0%)0/42 (0%)
Standard control40/120 (0%)   
lfri 4 nt mismatch control50/24 (0%)0/23 (0%)0/27 (0%)0/26 (0%)

Hypochord Development Requires notch1a Function

The hypochord phenotype of embryos injected with lfng MO was similar to that of embryos that had reduced levels of delta gene functions (Latimer et al., 2002). This finding is consistent with the idea that Lfng potentiates Notch-mediated specification of midline precursors for hypochord fate in response to Delta ligands. If so, reduction of Notch activity should phenocopy reduction of Lfng and Delta functions. Previous data showed that midline precursor cells express notch genes, including notch1a (Bierkamp and Campos-Ortega, 1993; Westin and Lardelli, 1997; Latimer et al., 2002). Thus, to test our prediction, we injected embryos with notch1a MO and assayed col2a1 expression at 26–28 hpf. All injected embryos had significant deficits of hypochord cells (Fig. 3A; Table 1). Thus, transient lfng expression might potentiate Notch activity in a subset of midline precursors, directing them toward hypochord fate.

Figure 3.

lfng potentiates Delta-Notch signaling in hypochord development. Lateral views, dorsal up, of 26–28 hours postfertilization embryos. A: Wild-type embryo injected with notch1a morpholino oligonucleotides (MO). Hypochord (hc) was absent, whereas floor plate (fp) and notochord (nc) appeared normal. B:aei-/- embryo injected with approximately 1 ng of lfng MO. Scale bar = 20 μm.

lfng Potentiates Notch Activity in Response to Delta to Promote Hypochord Development

Together with our previous work (Latimer et al., 2002), our data show that delta, notch, and lfng functions promote hypochord development. This finding is consistent with work from flies that indicated Fringe potentiates Notch activity in response to Delta (Panin et al., 1997). In zebrafish, delta genes have partially redundant functions in hypochord development, whereby loss of a single delta gene function results in incompletely penetrant, partial reduction of hypochord cell number (Latimer et al., 2002). We reasoned that, if Lfng causes Notch to signal more effectively in response to Delta for hypochord specification, then reduction of Lfng activity should enhance the hypochord phenotype of embryos that lack the function of a single delta gene. To test this, we injected approximately 1 ng of lfng MO into embryos produced by matings of aeiAR33/+ adults, which have a loss-of-function mutation of dld (Holley et al., 2000). A comparable amount of lfng MO injected into wild-type embryos produced an incompletely penetrant hypochord phenotype (27%, Table 1), which consisted of small gaps in the hypochord (data not shown). Additionally, 1 ng of lfng MO did not produce somite defects (data not shown). Thus, we were able to distinguish between lfng MO-injected aei-/- embryos and homozygous wild-type and heterozygous embryos at 24–26 hpf by the aei-/- somite phenotype. 100% of aei-/- embryos injected with 1 ng of lfng MO had a hypochord cell deficit (Fig. 3B), averaging 11.6 hypochord cells in the trunk (n = 10). By contrast, we previously showed that only 36% of aei-/- embryos had a hypochord phenotype and that affected embryos average 28 trunk hypochord cells (Latimer et al., 2002). Thus, partial reduction of lfng function enhanced the penetrance and severity of the aei-/- hypochord phenotype, consistent with the idea that Lfri potentiates Delta-Notch signaling activity.

In summary, our data show that dorsal margin cells transiently express lfng during gastrulation and that the loss of lfng function produced a deficit of hypochord cells, similar to the loss of notch and delta functions. Additionally, lfng MO injection enhanced the hypochord phenotype of embryos that lacked dld function. We propose that Lfri potentiates Notch activity in midline precursors that are adjacent to DeltaC/DeltaD+ paraxial mesoderm cells, thus regulating specification of a precise number of hypochord cells at the lateral borders of the midline precursor domain.



Embryos were produced by pair matings of fish raised in the Vanderbilt University Zebrafish Facility, raised at 28.5°C and staged according to hours post fertilization and morphologic criteria (Kimmel et al., 1995). We used the aeiAR33 mutant allele of dld (Holley et al., 2000).

fng cDNA Cloning

We used Block Maker ( and CODEHOP ( to design degenerate fng PCR primers with the following sequences: 5′-CCCGCCAGGCCCTGWSNTGYAARATG-3′ and 5′-GGCTCCTCCGGTGGCRAACCARAA-3′. Total RNA was obtained from 26 hpf embryos by using Tripure (Roche Applied Science), which was used to synthesize cDNA using an AMV Reverse Transcriptase kit (Roche Applied Science). PCR products were cloned into pPCR-Script Amp SK(+) (Strategene). Restriction digestion and sequencing revealed two distinct clones among the products, which most closely matched lunatic fringe and radical fringe sequences. A full-length lunatic fringe cDNA was constructed by using 5′ and 3′ rapid amplification of cDNA ends method (Clontech).

In Situ RNA Hybridization

Previously described RNA probes included dlc and dld (Haddon et al., 1998), ntl (Schulte-Merker et al., 1994), twhh (Ekker et al., 1995), myoD (Weinberg et al., 1996), col2a1 (Yan et al., 1995), pax2 (Krauss et al., 1991), gsc (Stachel et al., 1993), and her4 (Latimer et al., 2002). In situ RNA hybridization was performed as described previously (Hauptmann and Gerster, 2000). Embryos for sectioning were embedded in 1.5% agar/5% sucrose and frozen in 2-methyl-butane chilled by immersion in liquid nitrogen. Sections (10 μm) were obtained by using a cryostat microtome.

Morpholino Injections

Morpholino oligonucleotides (Gene Tools, LLC) were rehydrated with 1× Danieau solution. Working stocks were diluted in injection buffer (0.1 M KCl, 0.25% Phenol Red). A total of 1–2 nl or 3–4 nl were injected into the yolk, directly below the center of the blastodisk, of one- to two-cell stage embryos. The following antisense morpholino oligonucleotides were used: lfng (5′-CTTTTCCGCGATATGTTTTCAACA-3′), lfng 4 nucleotide mismatch control (5′-CTTTGCCGCTATATGTGTTAAACA-3′), notch1a (5′-CACCAAGAAACGGTTCATAACTCCG-3′), standard control (5′-CCTCTTACCTCAGTTACAATTTATA-3′).


Thanks to Hae-Chul Park and David Mawdsley for comments on the manuscript and the Vanderbilt University zebrafish facility staff for fish care. A.J.L. was funded by a NIH Developmental Biology training grant.