Cleavage of the BMP-4 Antagonist Chordin by Zebrafish Tolloid

See allHide authors and affiliations

Science  12 Dec 1997:
Vol. 278, Issue 5345, pp. 1937-1940
DOI: 10.1126/science.278.5345.1937


Dorsoventral patterning of vertebrate andDrosophila embryos requires bone morphogenetic proteins (BMPs) and antagonists of BMP activity. The Drosophila genetolloid encodes a metalloprotease similar to BMP-1 that interacts genetically with decapentaplegic, theDrosophila homolog of vertebrate BMP-2/4. Zebrafish embryos overexpressing a zebrafish homolog of tolloid were shown to resemble loss-of-function mutations in chordino, the zebrafish homolog of the Xenopus BMP-4 antagonist Chordin. Furthermore, Chordin was degraded by COS cells expressing Tolloid. These data suggest that Tolloid antagonizes Chordin activity by proteolytically cleaving Chordin. A conserved function for zebrafish and Drosophila Tolloid during embryogenesis is proposed.

The establishment of pattern during embryogenesis requires an extensive hierarchy of cell-cell signaling (1). The activity of members of the BMP family is implicated in ventral cell fate determination in vertebrates. Whereas overexpression of BMP-4 results in an expansion of ventral tissues such as blood and pronephros at the expense of dorsal tissues like notochord and somites (2-5), injection of a dominant negative BMP receptor blocks BMP signaling and dorsalizes Xenopus and zebrafish embryos (4-6). Conversely, the Xenopusgenes chordin and noggin are expressed in the organizer and have been implicated in the establishment of dorsal cell fates (7, 8). Chordin and Noggin bind directly to BMP-4, thereby blocking interaction with the BMP receptor—suggesting that dorsoventral patterning of the vertebrate embryo is partially accomplished through the antagonism of ventralizing BMP signals (9, 10). In agreement, mutations in the gene encoding the zebrafish Chordin homolog chordino ventralize zebrafish embryos (3, 5, 11).

A similar antagonistic relation is required for the establishment of dorsoventral pattern in the Drosophila embryo betweendecapentaplegic (dpp) and short gastrulation (sog) (12), which encode theDrosophila homologs of vertebrate BMP-2/4 and Chordin, respectively (13). Genetic analysis has revealed five other zygotically active genes that appear to be required for this process (14). One such gene, tolloid, encodes a member of the astacin family of metalloproteases similar to BMP-1 that appears to enhance the activity of dpp (12,15-17). However, BMP-1 is identical to the enzyme procollagen C–protease (18). It thus remains unclear if vertebrate homologs of Drosophila Tolloid have orthologous patterning functions during development and, if so, how they might interact with members of the BMP family to enhance their activities during embryogenesis and bone formation.

We isolated a zebrafish member of the astacin metalloprotease gene family that we named tolloid (tld), because it shares high homology with Drosophila and mammalian Tolloid (13, 19, 20) (Fig.1A). Consistent with an early function for the gene, tld transcripts are detected throughout the early gastrula stage embryo (21, 22). Toward the end of gastrulation expression becomes restricted, accumulating both dorsally and ventrally around the closing blastopore (Fig. 1B). Expression is also detected in the ectoderm flanking the anterior neural plate at this stage (Fig. 1, B and C). At the 10-somite stagetld mRNA is expressed in the developing tailbud and in cells flanking the midbrain and hindbrain that presumably correspond to migrating cranial neural crest (Fig. 1D).

Figure 1

Predicted protein structure and expression pattern of zebrafish tld. (A) Schematic representation of zebrafish Tolloid showing the relative organization of the metalloprotease, CUB (C1 to C5) and EGF (E1 and 2) domains; the signal sequence is indicated in black, and a region of unknown function NH2-terminal to the metalloprotease domain is indicated in gray. Numbers beneath the various domains indicate the percentage amino acid identity between zebrafish (Z), murine (M), andDrosophila (D) Tolloid within these domains. The overall amino acid identity between the proteins is indicated at right. Letters above the schematic zebrafish Tolloid represent the wild-type and the protease mutant amino acid sequence in the zinc-binding region of the metalloprotease domain; amino acids in red are thought to be essential for zinc binding (16). Abbreviations for the amino acid residues are as follows: E, Glu; F, Phe; G, Gly; H, His; I, Ile; L, Leu; P, Pro; V, Val; and W, Trp. (B and C) Ninety-five to 100% epiboly stage embryos hybridized with an antisense RNA to tld. At the vegetal pole, tld expression is localized to the region surrounding the closing blastopore (arrow) with greater expression evident ventrally than dorsally. At the animal pole, tld expression is detected flanking the anlage of the anterior neural plate (arrowheads). (D) Embryos at the 10-somite stage hybridized with an antisense RNA to tld. Strong expression is detected in the tailbud, and weaker expression is detected anteriorly in the cranial neural crest (arrowheads). Expression in the tailbud and neural crest persists until later stages when expression is also detected in cells of the hematopoietic system (22). Embryos are viewed animal pole up, dorsal right (B), from the animal pole with dorsal right (C), and anterior up, dorsal right (D).

To assay possible functions of zebrafish Tolloid (Tld), we injected synthetic tld mRNA into zebrafish embryos (23) (Fig. 2 and Table1). Morphological differences betweentld-injected embryos and control siblings first become evident at the beginning of somitogenesis when injected embryos display a broader and flatter tailbud (Fig. 2, A and D). At 24 hours, injected embryos are characterized by an expanded yolk extension, disrupted tail development, and an increase in hematopoietic cells (Fig. 2, B and E); head and eyes are smaller in injected embryos. After 48 hours of development, injected embryos are generally retarded relative to uninjected controls (Fig. 2, C and F). The increase in the size of the yolk extension is still apparent at this stage as is the expansion of the hematopoietic lineage, and the ventral tail fin is bifurcated (Fig. 2, C and F).

Figure 2

The morphological and molecular phenotypes of embryos injected with synthetic tld mRNA. Uninjected control (A to C) andtld-injected embryos (D to F); stages are 1-somite (A and D), 24 hours (B and E), and 48 hours (C and F). Embryos injected with synthetic tld mRNA display a broader and flatter tailbud at the 1-somite stage (arrowheads in D). At 24 hours, injected embryos are characterized by an expanded yolk extension (asterisk in E), disruption of tail development, and an increase in the size of the pool of hematopoietic cells (arrowhead in E). After 48 hours, the increase in the size of the yolk extension is still apparent as is the expansion of the hematopoietic lineage (arrowhead in F). Also seen at this stage is a bifurcation of the ventral tail fin (compare insets in C and F). Uninjected control (G, I, and K) andtld-injected embryos (H,J, and L) at 70% epiboly (G to J) and 12-somite (K and L) stages showing the expression pattern ofeve1 (G to J) and gata1 (K and L). The ventral marker eve1 is expanded dorsally at 70% epiboly in injected embryos relative to uninjected controls; arrowheads in (I) and (J) mark the boundary between eve1-expressing and -nonexpressing cells. The expansion of the hematopoietic lineage seen morphologically at 24 and 48 hours in injected embryos is anticipated by an increase in the expression of gata1, an early marker of presumptive hemopoietic cells. Embryos are viewed animal pole up, dorsal right (A, D, G, and H), anterior left, dorsal up (B, C, E, and F), and in animal view, dorsal right (I and J); embryos are viewed ventrally onto the tailbud, dorsal up (K and L), and insets in (C) and (F) are viewed from the dorsal aspect.

Table 1

Percentages of phenotypes observed after injection of synthetic mRNAs.

View this table:

The eve1 gene is expressed in ventrolateral cells of zebrafish gastrulae (24) (Fig. 2, G and I). Expression ofeve1 is expanded dorsally at midgastrulation stages intld-injected embryos (Fig. 2, G to J). The expansion of the hematopoietic lineage at 24 hours in injected embryos is anticipated by an increase in the expression of gata1, an early marker of presumptive blood cells (25) (Fig. 2, K and L). Furthermore, the expression of Pax2/pax[zf-b], a marker of the ventrally derived pronephros, is also expanded after tldinjection (22, 26). The morphological and molecular phenotype of tld-injected embryos is similar to those caused by injection of low concentrations of BMP-4 mRNA and characteristic of embryos homozygous for the ventralizing mutationchordino, except that chordino embryos frequently display disruption of the notochord posteriorly, which was not detected in tld-injected embryos (3, 5,11).

Strong antimorphic alleles of Drosophila tolloid carry mutations that map to the metalloprotease domain, whereas nonsense mutations resulting in truncation of Drosophila Tolloid immediately COOH-terminal to the metalloprotease domain act as strong null alleles (27). Consistent with a conserved function for zebrafish Tld, overexpression of mRNA encoding a Tld mutated in the zinc-binding motif of its metalloprotease domain, TldMut, acts antimorphically, resulting in a low frequency of dorsalization (Fig. 1A and Table 1) (22). Zebrafish embryos overexpressing a truncated zebrafish Tld, TldMetallo, develop indistinguishably from wild-type siblings showing that, as inDrosophila, the complement-Uegf-BMP-1 (CUB) and epidermal growth factor (EGF) domains are required for function of zebrafish Tld (Table 1).

To examine at what level Tld acts in the ventralization pathway, we coexpressed Tld and dorsalizing proteins. Overexpression of a truncated BMP receptor effectively blocked BMP signaling and produced phenotypes similar to those of several recently isolated dorsalizing zebrafish mutations (4, 5, 28) (Table1). Embryos injected with both truncated BMP receptor (DN-BR) andtld mRNAs were phenotypically identical to embryos injected with DN-BR alone, even at a concentration of tld in excess of that needed to produce ventralized embryos (Table 1). Overexpression of noggin(nog) or chordin (chd) dorsalizesXenopus embryos by antagonizing the interaction between the ventralizing signal BMP-4 and its receptor (9,10), a phenotype also seen by overexpression of these proteins in zebrafish embryos (Table 1). Similar to results with DN-BR, nog/tld-coinjected embryos were phenotypically indistinguishable from embryos injected with nog alone. In contrast, coinjection of a twofold excess of tld mRNA with chordin mRNA resulted in a significant decrease in the percentage of dorsalized embryos and a concomitant increase in the percentage of ventralized embryos (Fig. 2, E and F, and Table 1). Increasing the concentration of tldmRNA to a fourfold excess almost completely antagonized the dorsalizing effect of chd (Table 1). These results suggest that Tld acts at a level above the reception of the BMP signal. Furthermore, because Noggin and Chordin function similarly at the biochemical level (9, 10), these data suggest a specific antagonism of Chordin by Tld.

The homology between zebrafish Tld and procollagen C–protease on the one hand and Chordin and procollagen on the other suggested that in the coinjection experiments, zebrafish Tld might function to antagonize Chordin function by proteolytically cleaving Chordin (7,16). To test this hypothesis, we examined COS cells overexpressing zebrafish Tld or TldMut for their ability to cleave Chordin (29) (Fig. 3). Epitope-tagged Xenopus Chordin (Chd-MYC) (Fig. 3, lane 1) was unaffected by incubation with mock-transfected COS cells (lanes 2 to 5). Chd-MYC was, likewise, unaffected by incubations with COS cells transfected with TldMut (lanes 10 to 13). Conversely, COS cells expressing wild-type Tld progressively degraded Chd-MYC (lanes 6 to 9). Specificity of the affect of Tolloid on Chd-MYC was demonstrated because a second exogenously added protein (RXRα-LBD) was unaffected by incubation with COS cells alone or COS cells transfected with either Tolloid construct.

Figure 3

Chordin is cleaved by COS cells expressing wild-type but not mutated zebrafish Tolloid. Immunoblot analysis of culture media from transfected COS cells (lanes 6 to 13) or mock-transfected COS cells (lane 2 to 5) incubated with c-Myc epitope-tagged Chd and purified retinoid X receptor α–ligand binding domain (RXRα-LBD); lane 1 indicates the presence of anti–c-Myc and anti–RXRα-LBD immunoreactivity in the substrate mix before they were added to the transfected cells. Chd-MYC is cleaved after incubation with cells transfected with a construct expressing wild-type Tolloid (compare lane 6 with lanes 8 and 9) but not with cells expressing a form of Tolloid mutated in the zinc-binding domain (TolloidMut, lanes 10 to 13) or mock-transfected COS cells (lanes 2 to 5). RXRα-LBD immunoreactivity is unaffected by incubation with COS cells alone or COS cells transfected with either Tolloid construct.

The results of the transfection experiments are consistent with the ability of tld overexpression to phenocopy loss-of-function mutations in chordino and with the relation betweentld and chd as determined by coinjection. We propose that overexpression of Tld ventralizes the zebrafish embryo by releasing BMP-2/4–like proteins from inactive complexes with Chordin through proteolytic cleavage of Chordin. The expression of zebrafishtld during gastrulation argues for a similar function for endogenous Tld in establishing dorsoventral pattern in the zebrafish embryo. In light of the relation between Drosophila tld,dpp, and sog and their protein homologies with Tld, BMP-2/4, and Chordin, respectively, we suggest that this function has been conserved during vertebrate and invertebrate evolution.

  • * These authors contributed equally to this work.

  • To whom correspondence should be addressed. E-mail: uwe{at}


View Abstract

Stay Connected to Science

Navigate This Article