LvNotch signaling plays a dual role in regulating the position of the ectoderm-endoderm boundary in the sea urchin embryo

In combination with classical manipulation studies, recent molecular analyses have begun to elucidate the cellular and molecular mechanisms that pattern the sea urchin animal-vegetal (A-V) axis during early development (reviewed by Davidson et al., 1998; Logan and McClay, 1998; Wessel and Wikramanayake, 1999; Angerer and Angerer, 2000; Ettensohn and Sweet, 2000). This patterning process establishes three fundamental tissues along the A-V axis: mesoderm at the vegetal pole, endoderm overlying the mesoderm and ectoderm in the animal region of the embryo. A key event in this process is the placement of the boundary that divides the endoderm and ectoderm. This border separates the endoderm tissue that invaginates into the blastocoel during gastrulation from ectoderm tissue that remains outside to cover the embryo and larva.

Embryological studies have suggested that cell-cell interactions play an important role in specifying the boundary between the ectoderm and endoderm. Lineage studies have revealed that the ectoderm-endoderm border does not correlate with early cleavage divisions, suggesting instead that cell-cell interactions establish this boundary (Logan and McClay, 1997; Ransick and Davidson, 1998). Supporting this notion, blastomere isolation experiments have shown that interactions between animal blastomeres suppress endoderm forming potential in presumptive ectoderm cells (Henry et al., 1989). In addition, blastomere removal and transplantation studies have indicated that the micromeres, the vegetal-most cells in the 16-cell stage embryo, initiate a vegetal-to-animal wave of inductive signaling required for both normal overlying secondary mesenchyme cell (SMC) specification in the early blastula, and endoderm specification in the late blastula to early gastrula stage (Horstadius, 1973; Khaner and Wilt, 1991; Ransick and Davidson, 1993; Ransick and Davidson, 1995; Ransick and Davidson, 1998; Sweet et al., 1999; McClay et al., 2000).

Recent work has begun to reveal the molecular mechanisms that regulate the position of the ectoderm-endoderm boundary. A sea urchin BMP2/4 homolog is expressed in presumptive ectoderm in the blastula embryo, and appears to influence ectoderm-endoderm boundary position by suppressing endoderm formation within presumptive ectoderm cells (Angerer et al., 2000). In addition, β-catenin, a component of the Wnt signaling pathway (reviewed in Wodarz and Nusse, 1998), may also have a role in mediating the position of the ectoderm-endoderm boundary. Nuclear β-catenin signaling is required for several early aspects of endo-mesodermal specification in cleavage stage embryos (Wikramanayake et al., 1998; Logan et al., 1999; Huang et al., 2000; McClay et al., 2000; Vonica et al., 2000). At the late mesenchyme blastula stage, nuclear β-catenin is present specifically within the nuclei of presumptive endoderm cells bordering the presumptive ectoderm (D. R. Sherwood, PhD thesis, Duke University, Durham, NC, 1997) (Logan et al., 1999), suggesting a possible later function in regulating the position of the ectoderm-endoderm boundary.

The sea urchin homolog of the Notch receptor, LvNotch, may also have a role in mediating the position of the ectoderm-endoderm boundary. LvNotch signaling is activated within presumptive SMCs by underlying micromeres during early development, (Sherwood and McClay, 1999; Sweet et al., 1999; McClay et al., 2000), suggesting that LvNotch may be a component of the micromere-initiated vegetal-to-animal cascade of inductive signaling that influences the formation of endoderm. In addition, LvNotch protein is expressed dynamically within both presumptive ectoderm and endoderm cells in the blastula embryo (Sherwood and McClay, 1997), indicating that LvNotch could also function in these tissues to regulate the position of the ectoderm-endoderm boundary. While the Notch pathway has not yet been implicated in establishing germ-layer boundaries in other organisms, Notch signaling has been shown to mediate the formation of other types of boundaries, such as in Drosophila limb development and vertebrate somite formation (Irvine, 1999; Rawls et al., 2000). It is thus important to understand the possible role of LvNotch signaling in positioning the ectoderm-endoderm boundary in the sea urchin, both to gain a broader understanding of how Notch signaling is used for establishing boundaries, as well as elucidating the molecular mechanisms that pattern the sea urchin A-V axis.

In this study, we have investigated the role of LvNotch in mediating the position of the ectoderm-endoderm boundary using activated and dominant-negative forms of the receptor combined with lineage and mosaic analyses. We first demonstrate that activation of LvNotch signaling throughout the embryo shifts the ectoderm-endoderm boundary more animally along the A-V axis, whereas a loss or reduction of LvNotch signaling moves the boundary vegetally. Mosaic analyses of LvNotch function further show that LvNotch signaling has at least two distinct roles in specifying the position of the ectoderm-endoderm boundary. LvNotch signaling appears to function cell autonomously in the animal region of the embryo to promote endoderm formation more animally, while LvNotch signaling in cells vegetal to the ectoderm-endoderm border regulates a cell non-autonomous signal(s) that also establishes the endoderm higher along the A-V axis. Finally, we show that vegetal LvNotch signaling cell non-autonomously regulates nuclear β-catenin localization at the boundary, suggesting that vegetal LvNotch signaling may control the position of the boundary through the regulation of the Wnt pathway.

Adult Lytechinus variegatus were obtained from Jennifer Jackson (Beaufort, NC), and from Susan Decker (Hollywood, FL). Gametes were harvested and fertilized as described (Hardin et al., 1992). Embryos were cultured at 21-23°C in artificial sea water (ASW).

All LvNotch DNA constructs have been described (Sherwood and McClay, 1999), and were used as templates to generate in vitro transcribed 5′ capped mRNAs using the T3 mMessage mMachine kit (Ambion). mRNAs were passed through Microspin G-50 columns (Pharmacia) to remove free nucleotides, precipitated and resuspended in double distilled H₂O. mRNA concentrations were determined, then mixed with glycerol (40% v/v) and injected into fertilized eggs (2.0-6.0 pg/zygote) as described (Mao et al., 1996; Sherwood and McClay, 1999).

Preparation of eight-cell stage embryos for injection into single blastomeres was identical to that described above, except that eggs were fertilized in 5 mM p-aminobenzoic acid (PABA), which was washed out immediately after fertilization. Treatment with PABA softened the fertilization membrane, allowing the embryos to be freed from the membranes, which adhered to the injection dish. To lineage and identify injected blastomeres, fluorescein dextran (M_r 40×10³, Molecular Probes) was added to the mRNA/glycerol mixture at a concentration of 1.5 mg/ml. All mRNAs were injected at approximately 0.4-0.8 pg/blastomere at the eight-cell stage. Western analysis, whole-mount immunofluorescence, and the phenotypes of injected zygotes all indicated that fluorescein dextran did not affect the translation of injected mRNAs nor the phenotypes produced (see Fig. 7; data not shown). Similar to Lytechinus pictus embryos (see Henry et al., 1989), we noted that in Lytechinus variegatus, the third cleavage plane was usually slightly subequatorial. Thus, for consistency, we only injected blastomeres in eight-cell-stage embryos from batches of eggs producing embryos with slightly subequatorial third cleavage divisions. After injection, embryos were transferred to 50 mm petri dish lids coated with 1% agar, and analyzed under fluorescence at the 16-cell stage for embryos with healthy mesomere/mesomere or macromere/micromere pairs containing the mRNA/fluorescein dextran mix. These 16-cell-stage embryos were either labeled with DiI (below), or transferred to individual wells (coated with 1% agar) in a 96-well plate for culturing.

Individual mesomeres and macromeres in 16-cell-stage embryos were labeled iontophoretically with DiI (5 mg/ml in ethanol; Molecular Probes) (see Logan and McClay, 1997) on 1% agar coated dishes. The percentage of individual mesomeres and macromeres that normally contribute to the gut and ectoderm, respectively, was inferred from the lineage results of Logan and McClay (Logan and McClay, 1997). DiI- and fluorescein dextran-labeled plutei were imaged using a cooled CCD camera (Princeton Instruments) on a Leica DMRB microscope and Metamorph software.

To target LvN^act mRNA into animal halves, eggs were fertilized in 5 mM PABA and injected with LvN^act mRNA as described above. Following injection, PABA was washed out of the injection plate and embryos were cultured until the eight-cell stage. The embryos were then transferred to a 1% agar-coated dish containing hyaline extraction media (Fink and McClay, 1985). After a 2 minute incubation, embryos were transferred to an agar-coated dish containing Ca²⁺-free seawater, where animal and vegetal halves of embryos were separated using a fine glass needle. Animal halves were identified by the distinctive pattern of mesomere cleavage divisions, and cultured in 96-well plates containing ASW. Embryoids resulting from these animal caps were fixed with methanol after 48 hours of development as described (Sherwood and McClay, 1999), and stained with the Endo1 monoclonal antibody (Wessel and McClay, 1985).

Late mesenchyme blastula embryos injected with LvN^actΔANK5 and LvN^act were fixed with 2% paraformaldehyde in ASW, followed by rapid permeabilization with 100% methanol as described (Sherwood and McClay, 1999). To identify the LvN^act or LvN^actΔANK5 proteins, embryos were stained with the intracellular directed LvNotch polyclonal antibody, rabbit anti-ANK at 1:1000 dilution (Sherwood and McClay, 1997), which revealed the overexpressed intracellular domain of LvNotch, but not endogenous LvNotch. The number of cells expressing LvN^act or LvN^actΔANK5 was determined by optically sectioning stained embryos with a Zeiss 510 laser-scanning confocal microscope and counting the total number of nuclei containing these proteins. Nuclear β-catenin was visualized with a guinea pig anti-β-catenin polyclonal antibody as previously described (Logan et al., 1999). LvN^act/β-catenin double stained images were obtained by sequential confocal sectioning of double-labeled embryos, and these images were overlaid using Adobe Photoshop.

To determine whether LvNotch is involved in positioning the ectoderm-endoderm boundary, LvNotch signaling was perturbed within embryos by injecting fertilized eggs with mRNA encoding a constitutively activated (LvN^act) or dominant negative (LvN^neg) form of the receptor (see Sherwood and McClay, 1999). The activated form of LvNotch consists solely of the intracellular domain of the receptor, whereas the dominant negative form contains the extracellular and transmembrane domains of LvNotch, but lacks the intracellular domain. We have previously shown that after injection into fertilized eggs, the protein products of both constructs are expressed through the late mesenchyme blastula stage (Sherwood and McClay, 1999), the time at which molecular markers and manipulation studies suggest the sea urchin ectoderm-endoderm boundary is established (Davidson et al., 1998; Logan and McClay, 1998).

To assess whether LvNotch signaling affects the position of the ectoderm-endoderm boundary, we combined lineage analysis of individually labeled mesomeres at the 16-cell stage with injection of LvN^act and LvN^neg mRNA (Fig. 1A). The fate of mesomeres is a sensitive indicator of the position of the ectoderm-endoderm border. Lineage studies have indicated that the 16-cell stage mesomeres are usually positioned slightly animally to the ectoderm-endoderm boundary and give rise solely to ectoderm (Logan and McClay 1997). However, the mesomeres are close enough to the boundary such that approximately 16% of these cells randomly contribute progeny to both the ectoderm and the endoderm (the gut tissue in pluteus larvae). A shift in the ectoderm-endoderm boundary animally would thus be expected to increase and a vegetal shift of the border to decrease the number of mesomeres that contribute progeny to the gut.

In close agreement with previous lineage studies (Logan and McClay 1997), labeled mesomeres contributed progeny to larval gut tissue in approximately 20% of embryos injected with glycerol or with LvN^actΔANK5, a control construct similar to LvN^act but missing 14 amino acids required for signaling (Sherwood and McClay, 1999; Table 1). The injection procedure and protein overexpression thus had no effect on the lineage of mesomeres. Injection of LvN^act, however, dramatically influenced mesomere fate: individually labeled mesomeres contributed progeny to the gut in greater than 90% of injected embryos (Table 1; Fig. 1B,C). Furthermore, progeny of mesomeres that contributed to the gut in LvN^act injected embryos were typically found in all three gut compartments (n=12/14 cases where the mesomere contributed descendants to the gut). In contrast, in LvN^actΔANK5- and glycerol- (control) injected embryos, mesomere descendants found in the gut were never distributed throughout the gut. Rather, the descendants were usually confined to the hindgut (n=9/14 combined cases where the mesomere contributed progeny to the gut), which arises from endoderm cells nearest the ectoderm-endoderm boundary (Ruffins and Ettensohn, 1996).

To determine whether endogenous LvNotch signaling participates in the positioning of the ectoderm-endoderm boundary, mesomere lineages in LvN^neg-injected embryos were also examined. Expression of LvN^neg throughout the embryo had an opposite effect on mesomere fate compared with LvN^act: only one labeled mesomere contributed progeny to the gut (2% of cases; Table 1, Fig. 1D,E), and these progeny were restricted to the hindgut. Injection of mRNA encoding full-length LvNotch receptor, LvN^full, did not reduce the percentage of mesomeres contributing to gut tissue, demonstrating that the LvN^neg protein has a specific dominant negative function, rather than acting as a nonspecific sink for ligands of other signaling pathways that may also bind to the extracellular domain of Notch proteins (see Rebay et al., 1991; Wesley, 1999). Indeed, overexpression of LvN^full increased the percentage of mesomeres that contributed progeny to the gut compared with LvN^actΔANK5- and glycerol- (control) injected embryos (Table 1), indicating that full-length LvNotch overexpression activates the pathway involved in ectoderm-endoderm boundary positioning. Taken together, these results show that LvNotch influences the position of the ectoderm-endoderm boundary by promoting endoderm formation more animally in the embryo: activation of LvNotch signaling positions the ectoderm-endoderm boundary more animally along the A-V axis, whereas loss of LvNotch signaling shifts the border vegetally.

Several possible mechanisms exist by which LvNotch may regulate the position of the ectoderm-endoderm boundary. Previous work has demonstrated that the vegetal-most cells, the micromeres, initiate a sequential vegetal-to-animal wave of inductive signaling at the 16-cell stage that specifies overlying SMCs and endoderm during early development (reviewed in Davidson et al., 1998). Recent studies indicating that micromere signaling induces overlying SMC fate by activating LvNotch in adjacent presumptive SMCs (Sweet et al., 1999; McClay et al., 2000), suggests the possibility that LvNotch signaling within the presumptive SMCs is a component of the inductive signaling wave. Alternatively, LvNotch signaling could function cell autonomously within the presumptive endoderm cells to establish the ectoderm-endoderm boundary more animally. LvNotch protein is expressed at high levels along the apical domain of presumptive endoderm cells in the late blastula embryo (Sherwood and McClay, 1997), coincident with the time that the ectoderm-endoderm boundary is thought to be established (Davidson et al., 1998; Logan and McClay, 1998). Finally, it is possible that LvNotch signaling could act cell non-autonomously within the ectoderm to position the boundary. LvNotch is expressed at low levels throughout the presumptive ectoderm (Sherwood and McClay, 1997), and molecular and cellular studies have suggested that the ectoderm also has an important role in positioning the boundary (Henry et al., 1989; Angerer et al., 2000). Based on these possible distinct signaling functions for LvNotch, a mosaic analysis was performed to elucidate the mechanism(s) by which LvNotch signaling regulates the position of the ectoderm-endoderm boundary.

To address whether vegetal LvNotch signaling influences the position of the ectoderm-endoderm boundary by controlling another inductive signal, the lineage of uninjected mesomeres overlying LvN^act- and LvN^neg-injected macromere/micromere pairs was determined (Fig. 2A). mRNAs were co-injected with the lineage marker fluorescein dextran into single blastomeres at the eight-cell stage, the last stage at which we found consistent mRNA injection into single blastomeres technically possible. Injected embryos were then followed to the 16-cell stage when the asymmetric fourth cleavage division allowed embryos with injected macromere/micromere pairs to be identified. Individual mesomeres overlying injected macromeres were then labeled with DiI to follow the fate of their descendants (Fig. 2A). If vegetal LvNotch signaling regulates another inductive signal that positions the ectoderm-endoderm boundary, mesomeres overlying macromeres with increased levels of LvNotch signaling (LvN^act injected) would be expected to contribute to gut tissue at a higher frequency than controls. Conversely, mesomeres overlying macromeres with reduced LvNotch signaling (LvN^neg injected) would be predicted to contribute to the gut less frequently.

Labeled mesomeres overlying LvN^act-injected macromeres contributed to the gut in greater than 95% of embryos versus 24% in LvN^actΔANK5-injected controls (Table 2; Fig. 2B,C). Furthermore, descendants of mesomeres that overlay LvN^act-injected macromeres contributed to more gut tissue than in controls. Labeled descendants were sometimes found in all gut compartments (n=4/26 cases where mesomere contributed descendants to the gut), whereas in LvN^actΔANK5 macromere-injected embryos, mesomere descendants were typically restricted to the hindgut (n=8/10 cases where the mesomere contributed progeny to the gut) and never found in the foregut. Inhibition of LvNotch signaling in macromere descendants via injection of LvN^neg resulted in overlying mesomeres contributing cells to the gut in only 2% of cases, compared with 24% in LvN^actΔANK5- and 35% in LvN^full-injected controls (Table 2; Fig. 2D,E). These results indicate that endogenous vegetal LvNotch signaling acts cell non-autonomously to influence ectoderm-endoderm boundary position, regulating another inductive signal that promotes endoderm formation in overlying cells.

To determine whether LvNotch also functions within the animal region of the embryo to promote endoderm formation more animally, LvNotch mRNA constructs and fluorescein dextran were co-injected at the eight-cell stage, and embryos with injected mesomere pairs at the 16-cell stage were followed (Fig. 3A). Based on the known lineage of mesomeres (Logan and McClay, 1997), these constructs would be expressed primarily within the presumptive ectoderm; however, the vegetal-most extent of these clones would be situated near or at the ectoderm-endoderm boundary. We reasoned that if LvNotch signaling functions either cell non-autonomously within the ectoderm or acts cell autonomously within endoderm cells to promote endoderm formation more animally, that perturbation of LvNotch signaling within mesomere pairs would alter the frequency at which they contributed cells to the endoderm.

LvN^act-injected mesomere pairs contributed cells to the gut nearly twice as frequently as observed in LvN^actΔANK5-injected controls (64% versus 36%; Table 2; Fig. 3B,C). Descendants of LvN^act-injected mesomere pairs found in the gut also contributed more extensively to gut tissue, contributing to all gut compartments in seven of 37 cases. In four cases, mesomere descendants even formed ectopic gut tissue composed exclusively of fluorescently labeled cells (Fig. 4A-C), suggesting that LvNotch signaling may function cell autonomously within the descendants of mesomeres to promote endoderm formation. In contrast, descendants of LvN^actΔANK5-injected mesomere pairs were usually confined to the hindgut (n=12/15 cases), and never found in the foregut or in ectopic protrusions of gut tissue. Demonstrating a required role for LvNotch in promoting endoderm formation in mesomere descendants, mesomere pairs injected with LvN^neg contributed descendants to the gut 12% of the time, a significantly lower frequency compared with LvN^actΔANK5- and LvN^full-injected controls (Table 2; Fig. 3D,E). Taken together, these lineage results indicate that LvNotch signaling also functions within the animal region of the embryo to promote endoderm formation, possibly through a cell-autonomous mechanism.

Although the above experiments were suggestive of a cell autonomous mechanism for LvNotch signaling within the mesomere descendants, they did not rigorously exclude the possibility of a cell non-autonomous role for LvNotch in these cells. We, thus, tested the possibility that LvNotch signaling influences the position of the ectoderm-endoderm boundary through the control of a cell non-autonomous lateral signal by DiI labeling uninjected mesomeres that neighbor LvN^act-injected mesomere pairs (Fig. 5A). If LvNotch signaling within the animal region of the embryo promotes the formation of endoderm by regulating a lateral signal, uninjected mesomeres neighboring LvN^act-injected mesomeres would be expected to contribute descendants to the gut at a higher frequency than controls. The descendants of DiI-labeled mesomeres that neighbor LvN^act-injected mesomeres, however, contributed to the gut at a similar frequency as control embryos (Table 2; Fig. 5B,C), strongly suggesting that LvNotch signaling within the animal region does not regulate a lateral signal.

We next examined whether LvNotch promotes endoderm formation within the animal region of the embryo by controlling a cell non-autonomous signal that acts along the A-V axis. Uninjected macromeres underlying LvN^act-injected mesomere pairs were labeled with DiI (Fig. 5D), and the ectoderm contribution of these cells was determined (Fig. 5E). Previous lineage studies have indicated that individually labeled macromeres contribute cells to the ectoderm in most cases (90% or greater, see Logan and McClay, 1997). If LvNotch signaling controls a signal in the animal region that acts along the A-V axis to promote endoderm formation, macromeres underlying LvN^act-injected mesomeres should be converted to an endoderm fate, and thus contribute descendants to the ectoderm at a reduced frequency or number. The percentage of cases in which macromere descendants contributed cells to the ectoderm, however, was not significantly different between macromeres underlying control injected LvN^actΔANK5 and experimentally injected LvN^act mesomere pairs (95% versus 91%; n=18/19 and 21/23, respectively; P=1.0, Fisher’s exact test). Moreover, the mean number of ectoderm cells contributed by macromere descendants underlying LvN^actΔANK5- and LvN^act-injected mesomeres was not significantly different (27.5±3.3 versus 26.6±5.4; n=19 and 23, respectively; P=0.9, two-sample t-test), offering compelling evidence that LvNotch signaling within the animal region of the embryo does not promote endoderm formation by regulating a cell non-autonomous signal along the A-V axis.

As a direct test of whether LvNotch signaling can function cell autonomously within mesomeres to promote endoderm formation, we isolated animal caps from untreated and LvN^act-injected embryos (Fig. 6A). In untreated embryos, isolated animal caps form ciliated epithelial embryoids that are devoid of endoderm (Fig. 6B; Horstadius, 1973). Animal halves from embryos injected with LvN^act, however, formed endoderm tissue in 51% of the cases examined (n=18/35): invagination of archenteron tissue was observed by 24 hours of development, and this tissue expressed the hindgut/midgut marker, Endo1, by 48 hours of development (Fig. 6C-E). Together with the lineage experiments, these animal cap isolation results are indicative of a cell-autonomous role for LvNotch signaling within the animal region of the embryo in promoting endoderm formation.

Notch signaling has been shown in several developmental contexts to influence cell proliferation (Go et al., 1998; Johnston and Edgar, 1998; Carlesso et al., 1999; Walker et al., 1999). We therefore examined whether the vegetal or animal effects of LvNotch signaling on ectoderm-endoderm boundary positioning might include alterations in proliferation, either through an expansion of cells over or retraction of cells from the boundary. Macromere/micromere and mesomere pairs were injected with LvN^act or LvN^actΔANK5 mRNA as described above (see Figs 2A, 3A). As LvN^act and LvN^actΔANK5 localize tightly to the nucleus in all cells that express these proteins (Sherwood and McClay, 1999), we looked at whether activation of LvNotch signaling alters proliferation by examining the number and position of cells containing nuclear-localized LvN^act and LvN^actΔANK5 proteins at the late mesenchyme blastula stage (Fig. 7, Table 3). This stage was chosen since the ectoderm-endoderm boundary is thought to be established at this time (Davidson et al., 1998; Logan and McClay, 1998). Furthermore, the expression of LvN^act in injected blastomeres is lost shortly after the late blastula stage, indicating that the direct effects of altering LvNotch signaling must occur prior to or near this time. The mean number of descendants of injected macromeres and mesomere pairs expressing LvN^act, as well as the distribution of these cells along the A-V axis, however, was not significantly different from embryos containing the LvN^actΔANK5 protein (Table 3). Thus, LvNotch signaling does not appear to regulate the position of the ectoderm-endoderm boundary either vegetally or animally by influencing cell proliferation.

In the sea urchin, nuclear localized β-catenin, a component of the Wnt signaling pathway (Wodarz and Nusse, 1998), is present in presumptive endoderm cells bordering the presumptive ectoderm in the late mesenchyme blastula stage (D. R. Sherwood, PhD thesis, Duke University, Durham, NC, 1997) (Logan et al., 1999). Formation of the dorsal-ventral boundary in the Drosophila wing requires the coordination of both Wnt and Notch signaling (Diaz-Benjumea and Cohen, 1995; Rulifson and Blair, 1995; Neumann and Cohen, 1996; de Celis and Bray, 1997; Micchelli et al., 1997). Therefore, the relationship of LvNotch signaling to the localization of nuclear β-catenin at the ectoderm-endoderm boundary was analyzed to determine if the Wnt and Notch signaling pathways cooperate to mediate the position of this boundary.

We first asked whether LvNotch signaling regulates the localization of nuclear β-catenin at the ectoderm-endoderm boundary by determining β-catenin distribution in embryos injected with LvN^act and LvN^neg. Paralleling the effects on ectoderm-endoderm position, injection of LvN^act shifted nuclear localized β-catenin more animally along the A-V axis, leading to an approximate 50% increase in the volume of the embryo vegetal to nuclear β-catenin, compared with untreated controls (Fig. 8A,B,D). Furthermore, injection of LvN^neg shifted nuclear β-catenin localization lower such that the volume of the embryo vegetal to nuclear β-catenin decreased by approximately 30% (Fig. 8C,D). These results demonstrate that LvNotch signaling regulates the nuclear localization of β-catenin, and suggest that LvNotch may at least in part mediate the position of the ectoderm-endoderm boundary through the control of β-catenin localization at the border.

As experiments in the previous sections indicated that LvNotch promotes endoderm formation both cell autonomously within the animal region of the embryo and cell non-autonomously vegetally, we next asked whether both signaling functions of LvNotch regulated nuclear β-catenin localization at the boundary. LvN^act mRNA was placed into macromere/micromere and mesomere/mesomere pairs (as in Figs 2A, 3A), and the relationship of the nuclear-localized LvN^act protein was compared with nuclear β-catenin distribution in late mesenchyme blastula embryos. Injection of LvN^act into pairs of mesomeres did not shift the localization of β-catenin at the ectoderm-endoderm boundary (n=35/35 embryos), even when LvN^act directly bordered or overlapped nuclear β-catenin (Fig. 9A-C). Injection of LvN^act into vegetal macromere/micromere pairs, however, led to a clear shift in the localization of nuclear β-catenin more animally into neighboring cells lacking LvN^act (n=30/33 embryos; Fig. 9D-F). Vegetal LvNotch signaling can thus regulate nuclear β-catenin localization cell non-autonomously, and may control the position of the ectoderm-endoderm boundary by regulating a signal(s) that positions nuclear β-catenin at the border.

The results presented in this work demonstrate that LvNotch signaling influences the position of the ectoderm-endoderm boundary in the sea urchin embryo. Overexpression of a constitutively activated form of LvNotch throughout the embryo shifted the ectoderm-endoderm boundary animally, whereas expression of a dominant-negative form of LvNotch shifted the boundary vegetally. Together with our previous studies that show a role for LvNotch signaling in specifying SMC fate (Sherwood and McClay, 1999), these results demonstrate the importance of LvNotch in the overall patterning of the sea urchin A-V axis. Given that cell proliferation is not altered by perturbation of LvNotch through the mesenchyme blastula stage, whereas the SMCs (Sherwood and McClay, 1999), as well as the ectoderm-endoderm boundary (Fig. 8), are affected at this time, our results further demonstrate that LvNotch signaling alters the fates of cells along the A-V axis, rather than expanding or retracting territories through changes in cell proliferation (summarized in Fig. 10A).

Notably, when the position of the ectoderm-endoderm boundary was shifted by altered LvNotch signaling throughout the embryo, the ectoderm and endoderm tissues appeared to accommodate to the change in the amount of territory. For example, the shift in the ectoderm-endoderm territory animally by constitutively activated LvNotch did not expand the hindgut tissue (which forms from endoderm cells adjacent to the boundary) relative to other endoderm derivatives (see Fig. 1). Rather, our lineage results indicated that the patterning of the entire endoderm territory was affected. Similarly, the ectoderm territory near the ectoderm-endoderm boundary (the anal ectoderm), also was not specifically altered by shifts in the boundary (D. R. S., unpublished).

Our mosaic studies showed that LvNotch signaling acts in cells vegetal to the ectoderm-endoderm boundary to cell non-autonomously promote the position of endoderm more animally in overlying cells. Although our analysis could not precisely indicate which vegetal cells use LvNotch to send the cell non-autonomous signal, a likely possibility is the presumptive SMCs. Embryological experiments have shown that the vegetally localized micromeres initiate a sequential vegetal-to-animal cascade of inductive signaling necessary for the specification of SMCs and endoderm (Ransick and Davidson, 1995; Sweet et al., 1999; McClay et al., 2000). Significantly, micromeres appear to activate LvNotch signaling in the overlying SMC precursors to specify the SMC fate in the early blastula stage (Sweet et al., 1999; McClay et al., 2000). Thus, activation of LvNotch within the SMC precursors may couple SMC specification and the expression of a cell non-autonomous signal(s) that helps to define the animal boundary of the endoderm. Consistent with this idea, constitutive activation of LvNotch in vegetal cells converted most of these cells into SMCs and increased the non-autonomous signal; whereas inhibition of LvNotch reduced or eliminated SMC specification, and appeared to diminish or abolish the non-autonomous signal within vegetal cells (Fig. 2B-E; Table 2).

These mosaic studies also revealed one of the molecular targets of vegetal LvNotch signaling: the localization of nuclear β-catenin at the ectoderm-endoderm boundary in the late blastula embryo. Vegetal overexpression of activated LvNotch specifically shifted both the ectoderm-endoderm boundary and nuclear β-catenin localization more animally into overlying uninjected cells. Previous studies have demonstrated that vegetal nuclear β-catenin signaling is dynamic and required for several early aspects of endo-mesodermal specification (Wikramanayake et al., 1998; Logan et al., 1999; McClay et al., 2000). It is therefore possible that a later function of nuclear β-catenin within endoderm cells bordering the ectoderm is to mediate the recruitment of the animal-most endoderm cells. Supporting this notion, blocking entry of nuclear β-catenin within mesomeres, which normally give rise to ectoderm and sometimes endoderm cells at the boundary, prevents mesomere descendants from ever contributing cells to the endoderm (M. Ferkowicz and D. M., unpublished). Vegetal LvNotch signaling may therefore regulate the ectoderm-endoderm boundary through its effects on β-catenin localization at the border. Given that translocation of β-catenin to the nucleus is a downstream consequence of Wnt signaling (Wodarz and Nusse, 1998), one candidate for the cell non-autonomous signal regulated by vegetal LvNotch is a Wnt ligand. The isolation of several Wnt homologs expressed during early sea urchin development is consistent with this possibility (Ferkowicz et al., 1998), and it will be important in the future to determine which, if any, are regulated by vegetal LvNotch signaling.

Our mosaic studies further revealed that LvNotch signaling within animal cells also promotes the formation of endoderm more animally. Unlike vegetal LvNotch signaling, however, these effects appeared to be confined to the cells that contain altered LvNotch signaling. No cell non-autonomous effects on the position of the ectoderm-endoderm boundary were detected in untreated cells neighboring mesomere descendants in which LvNotch signaling was perturbed. Furthermore, ectopic endoderm tissue induced by expressing constitutively activated LvNotch in animal cells consisted solely of cells containing activated LvNotch. Animal caps, which in untreated embryos form ectodermal vesicles devoid of endoderm (Horstadius, 1973), were also induced (in approximately 50% of cases) to form endoderm derivatives by expression of constitutively activated LvNotch. Taken together, these observations offer compelling evidence that LvNotch signaling functions via a distinct, cell-autonomous mechanism in the animal region of the embryo to promote endoderm formation.

The cell autonomous nature of the effects of perturbing LvNotch signaling on endoderm formation suggests that LvNotch signaling probably functions within the presumptive endoderm in mesomere descendants. Consistent with this possibility, endogenous LvNotch is specifically expressed at high levels at the adherens junctions and along the apical domain of presumptive endoderm cells at the mesenchyme blastula stage (Sherwood and McClay, 1997), placing it in a position to interact with classical transmembrane ligands for Notch (reviewed in Kimble and Simpson, 1997) or a soluble processed ligand (Qi et al., 1999). It is important to note, however, that overexpression of dominant negative LvNotch did not eliminate endoderm formation, but rather shifted it more vegetally (see Fig. 1; Fig. 2D,E; Sherwood and McClay, 1999). Thus, cell autonomous LvNotch signaling may only be required within the more animal regions of the presumptive endoderm (i.e. close to the ectoderm-endoderm boundary). Furthermore, activation of LvNotch throughout the embryo did not extend endoderm tissue to the animal pole, suggesting that additional factors may confer or actively restrict the ability of LvNotch signaling to cell autonomously promote endoderm formation in the animal region. This could also explain why some isolated animal cap embryoids and injected mesomere pairs failed to form endoderm in response to constitutive activation of LvNotch signaling. For example, other endoderm specification factors required in cells for LvNotch to promote endoderm formation might not always have been present within the mesomere descendants. In support of this possibility, mesomere isolation experiments have indicated that there is considerable variability between different embryos and different batches of embryos in whether maternal vegetal determinants extend into mesomeres (Henry et al., 1989).

Another indication that the animal and vegetal signaling functions of LvNotch act distinctly in influencing the position of the ectoderm-endoderm boundary was the observation that unlike vegetal LvNotch signaling, activation of LvNotch within the animal region did not influence nuclear β-catenin localization at the boundary. These findings are consistent with both signaling functions of LvNotch working independently to promote endoderm formation more animally. Alternatively, it is possible that β-catenin localization at the boundary could mediate an interaction between both signaling functions of LvNotch. For example, nuclear-localized β-catenin could regulate the presentation of a ligand that activates LvNotch signaling directly at the border. The observation that activation of LvNotch within mesomeres fails to stimulate nuclear entry of β-catenin also implies that LvNotch signaling near the ectoderm-endoderm boundary promotes endoderm formation through a β-catenin-independent signaling mechanism. This is not unprecedented, as micromeres transplanted to the animal pole also induce endoderm without stimulating nuclear β-catenin entry (Logan and McClay, 1999).

This study provides the first undertaking of a mosaic analysis that examines the molecular mechanisms guiding ectoderm-endoderm boundary positioning in the sea urchin embryo, and contributes to our understanding of the signaling pathways that are likely to coordinately regulate the positioning of this boundary (summarized in Fig. 10B). It will be important in the future to develop more refined techniques to perturb Notch signaling, in order to better define when and where LvNotch functions, and to further address how LvNotch signaling is coordinated with other signaling pathways (e.g. BMP and Wnt) to position the ectoderm-endoderm boundary. Nevertheless, these experiments offer important new approaches to extend our understanding of ectoderm-endoderm boundary positioning in the sea urchin embryo, and in the case of LvNotch signaling, clearly demonstrate distinct functions for LvNotch in the animal and vegetal regions of the embryo in positioning this boundary.

We thank C. Logan, A. Ransick and A. Wikramanayake for their suggestions on lineage labeling and micromanipulation. We also thank E. Davidson, G. Hong and P. Sternberg for supporting the completion of this work. Furthermore, we are grateful to N. Sherwood, R. Fehon, J. Gross and members of the McClay laboratory for invaluable discussions and advice throughout this work. This research was supported by Sigma Xi Grants-in-Aid of Research awards to D. R. S., NIH grant HD14483 to D. R. M. and NSF equipment grant BIR-9318118.