|Home | About | Journals | Submit | Contact Us | Français|
Molecular studies have begun to unravel the sequential cell-cell signalling events that establish the dorsal-ventral, or ‘back-to-belly’, axis of vertebrate animals. In Xenopus and zebrafish, these events start with the movement of membrane vesicles associated with dorsal determinants. This mediates the induction of mesoderm by generating gradients of growth factors. Dorsal mesoderm then becomes a signalling centre, the Spemann’s organizer, which secretes several antagonists of growth-factor signalling. Recent studies have led to new models for the regulation of cell-cell signalling during development, which may also apply to the homeostasis of adult tissues.
With its rapid embryonic development, large egg size (1-2 mm in diameter) and high numbers of embryos (1,500 per female), Xenopus provides a favourable model system for the study of vertebrate development, and it has been used extensively to probe the events in early embryogenesis. The dorsal side of the amphibian embryo contains the information for the differentiation of many different cell types. At the gastrula stage, the dorsal side of the embryo can be recognized by the presence of the dorsal blastopore lip (FIG. 1). When the gastrula embryo is bisected by ligature with a hair loop into dorsal and ventral fragments, the half that contains the dorsal blastopore lip develops into an entire embryo, whereas the ventral half remains as a ‘belly piece’ (Bauchstück) that is devoid of all axial organs1. In a complementary experiment, the dorsal lip transplanted to the ventral side of a host embryo induces the formation of a twinned embryo that contains axial structures and organs2. This dorsal-ventral difference can be traced back to the fertilization stage, with the dorsal side always forming opposite to the sperm entry point.
Region of reduced pigmentation that marks the future dorsal side of the fertilized egg.
The first external sign of asymmetry in the Xenopus egg is the appearance of an unpigmented dorsal crescent (called the grey crescent in some amphibians)3, which is caused by a rotation of the egg cortical cytoplasm that is driven by microtubules4 (FIG. 1). Dorsal determination seems to be associated with the cytoplasm that surrounds the heavy yolk platelets in the vegetal pole. When the heavy yolk and associated cytoplasm is made to flow towards the animal pole of an amphibian egg, for example by inverting the egg by 180° or by centrifugation, a twinned dorsal axis is formed5. Isolating the molecules that mediate the phenomena behind these experimental observations has been the Holy Grail of amphibian embryology. Remarkably, the general outlines of a molecular pathway that regulates dorsal development from fertilization to gastrulation are starting to emerge.
Here we review how dorsal determinants located in membrane vesicles in the vegetal pole of the embryo are transported to the dorsal side by cortical microtubules. This event correlates with the activation of the canonical Wnt signalling pathway on the dorsal side, causing the stabilization and nuclear localization of the β-Catenin protein. In turn, this leads to the generation of a gradient of signalling molecules related to Nodal in the endodermal region at the blastula stage, resulting in the induction and patterning of the mesodermal germ layer. During gastrulation, a signalling centre (Spemann’s organizer) becomes established in the dorsal mesoderm and expresses numerous organizer-specific genes, notably secreted proteins that bind to growth factors in the extracellular space and prevent them from signalling. These antagonists include molecules such as Noggin, Chordin, Cerberus, Frzb-1, Crescent and Dickkopf (DKK). One of the main conclusions from this research is that cell differentiation in the gastrula embryo is regulated by inhibitory secreted molecules. We also address in some detail the Chordin pathway, in which signalling by bone morphogenetic proteins (BMPs) bound to Chordin is regulated by the combined action of the protease Xolloid and of another BMP-binding protein called Twisted-gastrulation (xTsg).
BONE MORPHOGENETIC PROTEINS
(BMPs). Molecules of the TGF-β family that can induce bone formation and ventralize the vertebrate embryo. In zebrafish, a mutation in either BMP-2b or BMP-7 has a similar effect, suggesting that they work as heterodimers; either mutation also inhibits transcription of BMP-4.
The critical function of the dorsal determinants stored in the Xenopus egg is best illustrated by the experiments shown in FIG. 2. When fertilized eggs are irradiated at the vegetal pole with ultraviolet light, which impairs microtubule function by crosslinking GTP to tubulin, the embryo develops as a belly piece, lacking all dorsal axial organs (FIG. 2a, left). Other treatments that disrupt microtubules, such as nocodazole and low temperature, have similar outcomes4. The ventralized belly pieces develop all three germ layers, but the mesoderm is of the ventral type (lateral plate mesoderm and blood) and the ectoderm consists exclusively of epidermis6-8 (FIG. 2b). Remarkably, these ventralized embryos can be completely rescued by the microinjection of certain synthetic messenger RNAs. For example, upon a single injection of chordin mRNA into 4- to 32-cell embryos, the formation of head, trunk and tail is restored (FIG. 2a, right), and the rescued embryos contain a variety of dorsal tissues, such as central nervous system (CNS) in ectoderm, and somite, notochord and kidney in mesoderm (FIG. 2c). The most extraordinary aspect of the ultraviolet-rescue experiment is that the same result can be obtained with a large variety of mRNAs that function in very different signalling pathways (FIG. 2d). Complete rescue can be caused by the microinjection of mRNAs that encode Wnts and several downstream effectors of its pathway, such as β-Catenin and proteins that regulate β-Catenin degradation9-11, and by microinjection of a number of Nodal-related factors12 or BMP antagonists, such as Noggin and Chordin13,14. How can such diverse molecules all have the same properties in such a simple ultraviolet-rescue assay? The thesis that we develop in the following pages is that Wnts, Nodals and anti-BMP molecules represent sequential steps in a common pathway leading to dorsal axis development. The ultimate consequence of activating this pathway in the embryo would be to generate a region of low BMP signalling that is required for dorsal differentiation.
A breakthrough in the study of early dorsal axis establishment was the discovery that β-Catenin is translocated into the nuclei of cells on the dorsal side of both Xenopus and zebrafish embryos at the early blastula stage15,16. Xenopus β-Catenin has been shown to be required for dorsal axis formation by the inhibition of maternal mRNAs using antisense oligonucleotides17,18. β-Catenin protein starts to accumulate in the dorsal cytoplasm as early as the 2-cell stage and in dorsal nuclei by the 16-cell stage16. By the midblastula stage, nuclei stain positively for β-Catenin in the entire dorsal side, including regions that later give rise to endoderm, mesoderm and ectoderm; this shows that dorsal signals are present in a much wider area than was previously thought15.
Another crucial discovery has been that the cortical microtubules, which extend in parallel arrays from the sperm entry point to the dorsal side of the embryo, not only mediate the 30° rotation of the egg cortex with respect to the underlying cytoplasm, but also transport membrane vesicles located at the vegetal pole to the dorsal side19 (FIG. 3a). Importantly, these membrane vesicles are transported dorsally 90° or more towards the animal pole, at the extraordinary speed of 25-40 μm min-1 (REF. 19). The membrane vesicles are able to bind the fusion protein Dishevelled-green fluorescent protein (Dsh-GFP) and transport it to the dorsal side20. Dsh is a component of the Wnt signalling pathway and it associates with the cytoplasmic side of membranes on activation by Wnt11. When the mRNA that encodes the Frizzled-7 Wnt receptor is ablated in oocytes that have been microinjected with antisense oligonucleotides, the formation of the dorsal axis is abolished21. It is therefore attractive to speculate that Frizzled-7 may be present in the membrane vesicles that function as dorsal determinants, becoming activated by an as yet unidentified Wnt protein present inside the membrane vesicles and leading to the stabilization and nuclear localization of β-Catenin (FIG. 3a).
Transient rod-like structure derived from the endoderm, which is located beneath the notochord in vertebrate embryos.
In ultraviolet-treated embryos, the parallel microtubular arrays do not form, and nuclear β-Catenin is found only in vegetal pole nuclei15 (FIG. 3b). The membrane vesicles marked by Dsh-GFP remain in the vegetal pole20, and the expression of β-Catenin target genes, such as homeobox (siamois and xtwin) and secreted factor genes (nodal-related-3 and cerberus), remain confined to the vegetal pole8,22,23. In this vegetal location, the β-Catenin signal does not reach the marginal zone and animal cap, and is unable to trigger the differentiation of dorsal cells into mesoderm and ectoderm. When vegetal cytoplasm is injected into marginal or animal regions, however, the differentiation of dorsal tissues is restored23.
In the zebrafish, arrays of parallel microtubules are formed in the vegetal pole of the zygote that, when disrupted by ultraviolet irradiation, result in ventralization of the embryo24. Partial removal of the vegetal yolk cell mass results in severely ventralized embryos25,26. It is therefore conceivable that zebrafish vegetal cytoplasm contains vesicles similar to those that function as dorsal determinants in Xenopus. In some freshwater snails (for example, Bithynia), a small protrusion of cytoplasm of less than 1% of the egg volume, called a polar lobe, is formed during early cleavage, and its removal results in the loss of many mesodermal structures. These polar lobes are filled with intriguing small membrane vesicles of about 3 μm in diameter, which are located at the vegetal pole cortex during oogenesis27. In the mouse, genetic evidence supports the proposed role of the Wnt pathway in axis formation. Mutation of axin, an inhibitor of β-Catenin, results in duplicated axes10, and targeted gene inactivation of Wnt-3 or β-catenin results in the absence of axis formation28,29.
In conclusion, recent studies indicate that the initial dorsal asymmetry in the vertebrate embryo may be triggered by the transport of dorsal membrane vesicles capable of activating the β-Catenin pathway.
The next step in dorsal-ventral axis formation is the induction of mesoderm by the endoderm. The established view from embryological studies was that the endoderm releases two signals, one from the ventral endoderm, which induces ventral mesoderm, and a second from the dorsal endoderm (a region called the Nieuwkoop centre), which induces dorsal mesoderm (or Spemann’s organizer tissue). A third signal subsequently emanates from the Spemann’s organizer in the plane of the mesoderm to refine the initial dorsal-ventral pattern7,30 (FIG. 4). The mesoderm-inducing signal is released by endoderm after midblastula31 and can be investigated experimentally by combining explants of vegetal and animal tissue, as done initially by Nieuwkoop.
Nodal-related proteins, which belong to the transforming growth factor (TGF)-β family of growth factors, have important functions in mesoderm formation in different species. In the mouse, only one nodal gene has been identified, and the effects of its inactivation suggest a central function in the formation or maintenance of mesoderm32,33. In zebrafish, two nodal-related genes, cyclops (cyc) and squint (sqt), have been found34. In the absence of each individual gene, cyclopic embryos develop, but when both gene products are removed, embryos lack head and trunk dorsal mesoderm, as well as endoderm, and fail to express the organizer-specific homeobox gene goosecoid35.
We mentioned above that ultraviolet-induced ventralization of Xenopus embryos is completely rescued by microinjection of mRNAs encoding proteins related to Nodal12. In Xenopus, there are five mesoderm-inducing Nodal-related (Xnrs) genes (REFS 12,36; and M. Asashima, personal communication), making loss-of-function studies particularly difficult. However, the induction of mesoderm by endoderm was discovered initially in amphibians, and has been analysed in considerable detail in these animals30.
A specific inhibitor of Xnrs provided a way of testing the role of these molecules in Xenopus mesoderm induction. It was found that one of the inhibitory proteins secreted by the organizer, Cerberus37, was an antagonist of Xnrs, and that the Xnr-binding activity resided in the carboxy-terminal domain38. A construct comprising only this domain, called Cer-short, provides a valuable tool to test the role of the several Xnrs in development, as it specifically inhibits mesoderm-inducing Xnrs but not other TGF-β molecules, such as Activin, Vg1, BMP-4 and Derrière39.
When Nieuwkoop’s original mesoderm-induction experiments of combining animal and vegetal explants were repeated using Cer-short as a reagent to inhibit Xnrs, the induction of both dorsal and ventral mesoderm was blocked39. At the blastula stage, Xnrs are expressed in a dorsal to ventral gradient in endodermal cells39, which is accompanied by the preferential phosphorylation of Smad2 (a downstream effector of TGF-β signalling) on the dorsal side40. Dose-response experiments using increasing amounts of cer-short mRNA confirmed the existence of an endogenous Xnr activity gradient in endoderm39. A modified model of Xenopus early development, in which the induction of both dorsal and ventral mesoderm is mediated by a gradient of several Nodal-related signals released by the endoderm at the blastula stage, is shown in FIG. 4. In this view, the three-signal model mentioned above may be considered a two-signal one30,39.
The gradient of Xnr expression in the endoderm is thought to be activated by three maternally provided molecules: Vg1,VegT and β-Catenin (FIG. 4).Vg1, a TGF-β factor, and VegT, a T-box transcription factor, are both localized to the vegetal pole of the Xenopus oocyte and are potent inducers of endoderm41,42. Depletion of maternal VegT leads to the absence of endoderm42. In VegT-depleted embryos, Xnr transcription and mesoderm formation are severely inhibited and can be rescued by injection of Xnr mRNA43. Wild-type embryos microinjected with VegT and Vg1 have only low levels of Xnr transcription; however, when β-Catenin is also provided, it cooperates with VegT and Vg1 to achieve the high levels of Xnr expression that cause organizer induction39 (FIG. 4). A plausible explanation for the ultraviolet-induced phenotype, in which all three germ layers are present (FIG. 2b), is that, even though β-Catenin is lacking on the dorsal side, endogenous VegT and Vg1 levels in endoderm are sufficient to generate low levels of Xnr signalling, thereby mediating the induction of ventral mesoderm (FIG. 4).
In conclusion, after midblastula, the β-Catenin signal, in combination with other maternal genes, activates a dorsal-ventral gradient of several Nodal-related signals in the endoderm that, in turn, mediate the induction and patterning of the mesodermal layer.
As discussed above, nuclear β-Catenin accumulates in the entire dorsal side of the Xenopus blastula15. This stabilization of β-Catenin protein has profound effects in all three germ layers. It has been known for a long time that the dorsal ectoderm of the embryo is predisposed, when compared with ventral ectoderm, to respond to neural-inducing signals44. Ectopic expression of β-Catenin in animal-cap ectoderm is able to repress BMP-4 transcription, thus promoting neural development45. Similarly, signalling through β-Catenin represses transcription of the epidermal homeobox gene Dlx3,a Xenopus anti-neural factor46. In zebrafish, β-Catenin both induces transcription of the homeobox gene bozozok, which is related to goosecoid, and is required for the dorsal repression of BMP-2b/4 transcription in the dorsal marginal zone47,48.
In the mesoderm, even after the initial induction has taken place, β-Catenin and TGF-β signals (probably Nodal-related) are integrated at the level of promoter DNA to trigger transcription of the organizer-specific gene goosecoid49-51. In Xenopus, β-Catenin can also directly activate transcription of the homeobox genes siamois and xtwin8. These two closely related homeobox genes are transcriptional activators and therefore are different from bozozok and goosecoid in zebrafish, which are considered to be transcriptional repressors52.The goosecoid promoter in Xenopus contains two distinct DNA regulatory elements (see supplementary information online, FIG. S1). Xtwin/Siamois binds to the proximal element, thereby relaying the β-Catenin signal. The distal element contains binding sites for a mesendodermal homeodomain protein, Mixer, and for a heterodimer of Smad2 and Smad4. Transcription of mixer is activated, and nuclear translocation of the Smad2/Smad4 heterodimer is stimulated by the TGF-β/Xnr signal that, in this way, is transduced to the goosecoid promoter51. In the promoters of other TGF-β-inducible genes, the function of Mixer is carried out by the Forkhead-related protein FAST-1 (REF. 53). The Wnt and TGF-β pathways might even be integrated in the absence of separate promoter elements, as in the case of the xtwin promoter54,55 (see supplementary information online, FIG. S1).
In the endoderm, inhibition of β-Catenin signalling by ultraviolet irradiation blocks the induction ofXlHbox-8, a marker of dorsal endoderm, which can be restored by microinjection of TGF-β factors56. Cell-dissociation experiments have shown that the secretion of TGF-β factors, such as Xnrs, is required for endodermal differentiation57. In addition, the initial expression of Xnr-1, Xnr-2 and Xnr-4 in the endoderm requires the β-Catenin pathway39.
In conclusion, signals triggered by the dorsal determinants pattern all three germ layers. β-Catenin activates genes such as siamois and bozozok, and the expression of Nodal-related factors, which subsequently cooperate in several parallel pathways as effectors of dorsal development52,58,59. Detailed studies on promoters, such as that of goosecoid, are gradually providing an understanding of how these many pathways are brought together at the level of transcriptional regulation of organizer-specific genes.
At the gastrula stage, the main dorsalizing centre of the embryo is Spemann’s organizer, which is located in the dorsal mesoderm (FIG. 4). Its molecular exploration proved a productive fishing ground for the discovery of new genes (see supplementary information online, FIG. S2) and produced unexpected findings. The main surprise was that the organizer is a source of secreted antagonists that bind to growth factors in the extracellular space and prevent them from binding to their cognate receptors. These novel antagonists can be classified according to the growth factors that they inhibit (FIG. 5).
The BMP antagonists Chordin, Noggin and Follistatin do not share any common structural elements13,14,60; however, a single ventral injection of mRNAs of any of these proteins leads to the formation of twinned axes, recapitulating the effects of grafting Spemann’s organizer. When BMP antagonists are overexpressed ubiquitously, embryos become radially dorsalized, mimicking the phenotype caused by lithium chloride (LiCl) treatment during early cleavage. LiCl transforms the entire mesoderm into Spemann’s organizer and acts through the inhibition of GSK-3, a serine/threonine kinase that phosphorylates β-Catenin and targets it for degradation61,62 (FIG. 3a). So these anti-BMP activities mimic those caused by increased β-Catenin signalling. Transcription of noggin and chordin is induced radially by treatment with LiCl and both genes were cloned during the course of screens involving LiCl treatment of Xenopus embryos13,14. Conversely, chordin and noggin transcripts are greatly reduced in ultraviolet-treated embryos and, as mentioned above, when microinjected, they completely rescue the ultraviolet-induced ventralized phenotype (FIG. 2a). The expression of Chordin is negatively regulated by signals from the ventral side of the embryo that activate ventralizing homeobox genes, such as vent and vox, which repress chordin expression in ventral-lateral regions of the embryo63-65. Chordin and Noggin bind BMPs directly in the extracellular space, preventing BMPs from binding to and signalling through its cognate BMP receptor66,67. Follistatin also binds BMPs, although, unlike the other BMP antagonists, the Follistatin-BMP complex can bind to the BMP receptor but is unable to signal68.
Wnt inhibitors secreted by the organizer are of two types, Frzbs and Dkks (FIG. 5). Frzb-1 is a secreted protein that contains a domain similar to the Wnt-binding region of the Frizzled Wnt receptors, and functions by binding to Wnts and antagonizing their activity69,70. Frzb-1 is a member of a large family of related Wnt inhibitors that have been renamed secreted Frizzled-related proteins (sFRPs)11. Other members of this family are also expressed in Spemann’s organizer, namely sFRP-2 and crescent71. Another sFRP, sizzled, is expressed in the ventral side of the gastrula72. Dkk-1 was the founding member of a new class of Wnt antagonists, and contains two new cysteine-rich domains73. Dkk-1 and Frzb-1 are expressed in the deep layers of Spemann’s organizer, including the future head mesoderm and anterior endoderm. When ubiquitously over-expressed, Dkk-1 and Frzb-1 lead to embryos with enlarged heads and a shortened trunk. This phenotype is caused by an expansion of Spemann’s organizer, thought to be caused by inhibition of ventralizing Xwnt-8 signals present in the ventral-lateral marginal zone69,73,74. In combination with inhibitors of BMP signalling, Dkk-1 and Frzb-1 cooperate in the formation of head structures, and microinjection of neutralizing Dkk antibodies causes microcephaly73,75. The many Wnt antagonists expressed by Spemann’s organizer differ in biological activities. For example, overexpression of crescent mRNA causes cyclopia, whereas overexpression of frzb-1 mRNA leads to enlarged eyes71. This indicates that different Wnt antagonists bind to overlapping, but distinct sets of Wnt signals.
Cerberus, a secreted protein of 260 amino acids expressed in the anterior-most region of involuting endoderm37, is the founding member of a large family of cell-cell signalling regulators76. Microinjection of cerberus mRNA into the ventral side of the embryo leads to induction of ectopic head structures in the absence of trunk formation37. Cerberus protein is a multivalent antagonist that binds to Xnrs, Xwnt-8 and BMP-4 in the extracellular space38. These three signalling pathways are required for trunk development, and secretion of Cerberus by anterior endoderm serves to maintain a trunk-free region in the anterior of the embryo so that the head territory can develop.
Antivin/Lefty is an Activin/Nodal antagonist that has been isolated in frog, fish and mouse. It is a divergent member of the TGF-β superfamily that lacks the α-helix required for dimerization77,78. Mouse mutants lacking lefty-2 form excess mesoderm, a phenotype that is partially suppressed by heterozygosity for nodal, suggesting that the main function of Lefty-2 is to downregulate Nodal signalling79. Because transcription of antivin/lefty is induced by Nodal, it acts as a feedback inhibitor that limits the Nodal signal in time and space79,80. Interestingly, the phenotypic effects of Antivin/Lefty in zebrafish can be suppressed by overexpression of the extracellular domain of the Activin/Nodal receptor type IIB (REF. 79). This result indicates that Antivin/Lefty may block Nodal signalling by binding to TGF-β/Nodal receptors and inhibiting their activity (FIG. 5). In Xenopus, Xnr-3 (a divergent Xnr) lacks mesoderm-inducing activity but can induce neuralization instead81. Although its mechanism of action has not been determined, it may be worthwhile in future studies to explore whether Xnr-3 can also function as a competitive inhibitor of TGF-β receptors.
In conclusion, studies of Spemann’s organizer have led to the discovery of new mechanisms of regulating cell-cell signalling, in which secreted antagonists of growth factors are important in patterning the vertebrate embryo.
We now discuss in more depth the protein Chordin, as studies on this molecule have uncovered a finely regulated pathway of cell-cell signalling. Chordin is a secreted protein that contains four internal cysteine-rich domains of about 70 amino acids (designated CRs, see supplementary information online, FIG. S3) that bind BMP molecules and prevent their binding to BMP receptors66. All four CRs bind BMPs at a detectable level, but CR1 and CR3 bind more effectively than do CR2 and CR4. The affinity of the binding of BMPs to CR1 or CR3 is ten times lower than that of full-length Chordin for BMPs82. In agreement with these biochemical properties, injection of mRNA encoding CR1 or CR3 causes dorsalization in Xenopus embryos82. Although less effective than full-length Chordin, the CR1 and CR3 modules retain significant anti-BMP activity.
Genetic analyses in zebrafish, mouse and Drosophila support an important function of Chordin in dorsal-ventral patterning. In zebrafish, the strongest ventralizing mutation isolated in extensive genetic screens corresponded to the inactivation of the Chordin homologue Chordino83,84. In chordino mutants, neural plate and dorsal mesoderm are reduced, whereas epidermis and ventral mesoderm are expanded at the gastrula stage85 (FIG. 6a). The crucial role of BMP signalling in early embryonic patterning is illustrated by the fact that the dorsal-ventral mutants identified so far in zebrafish encode components of the BMP signalling pathway. Mutants in swirl/BMP-2b, somitabun/Smad5 and snail house/BMP-7 have strong dorsalized phenotypes86-88. The swirl dorsalized phenotype is epistatic to chordino in chordino;swirl double mutants85, suggesting that Chordin is a specific antagonist of BMP ventralizing activities. In mouse, chordin;noggin double mutants lack the forebrain and anterior notochord, and have a greatly reduced pharyngeal endoderm, as well as a randomized left-right axis89. These genetic studies in zebrafish and mouse show that secreted Chordin protein is required for correct patterning of the three germ layers.
In Drosophila, the product of the short-gastrulation (sog) gene90 is the functional homologue of Chordin91 and antagonizes the BMP homologues Dpp and screw, which are the zygotic dorsal-ventral morphogens of the fly embryo92-94. Loss-of-function mutations in sog result in an expansion of dorsal ectoderm at the expense of neurogenic ectoderm90 (FIG. 6b). As in the case of swirl and chordino, in double-mutant studies, dpp and screw are epistatic to sog93,95, suggesting that sog/chordin is a specific antagonist of BMP signalling.
In Drosophila, a protease called Tolloid is an integral component of the dorsal-ventral patterning system, and functions as an enhancer of DPP/BMP signals92,96. In zebrafish, the most frequently isolated dorsalized mutant, mini-fin, is a loss of function of the zebrafish tolloid gene. This mutant shows a reduction of ventral and an expansion of dorsal markers at late gastrula, and eventually develops into a viable fish that lacks the ventral tail fin97.
Using a direct biochemical approach, it was found that Tolloid and its Xenopus homologue, Xolloid, function by cleaving inactive SOG-DPP or Chordin-BMP complexes98,99 (FIG. 7a). Cleavage of these complexes permits the re-activation of BMPs, which are then able to signal again, ventralizing Xenopus explants98.Recently, the precise cleavage sites of Xolloid in Chordin have been identified100 and found to be located at conserved aspartate residues 30 amino acids downstream of CR1 and 16 amino acids downstream of CR3. Therefore, the cleavage of Chordin by Xolloid releases intact CR modules in a complex with BMPs, and this cleavage may allow previously inactive BMP molecules to signal once again.
Recent results from Drosophila and Xenopus indicate that a further molecule may participate in the Chordin/Sog, Xolloid/Tolloid, BMP/Dpp pathway. The Drosophila gene twisted-gastrulation (dtsg) encodes a secreted protein necessary for the formation of the amnioserosa, the tissue in the Drosophila embryo that requires the highest levels of Dpp/Screw activity101,102 (FIG. 6b). Therefore, dTsg is a molecule that is required to attain maximal BMP signalling activity. Careful inspection of the amino-acid sequence of dTsg revealed some sequence similarity with the CRs of Chordin, which was validated biochemically by the demonstration that dTsg can bind BMPs103.The Xenopus twisted-gastrulation homologue (xtsg) is expressed in the ventral pole of the embryo as part of the BMP-4 synexpression group, suggesting that this gene may function in the BMP pathway103,104.
A combination of co-injection experiments in Xenopus embryos and protein crosslinking103 support the biochemical pathway depicted in FIG. 7b. xTsg is expressed in the same cells that synthesize BMP-4, and it is therefore possible that a complex of both proteins is secreted by ventral cells. In addition to binding BMPs, the xTsg protein is able to bind to Chordin itself. In the presence of xTsg, Chordin is a more efficient BMP-binding protein, forming a stable ternary complex103. As Chordin/BMP complexes do not bind to BMP receptors, in this first aspect of its function xTsg would convert Chordin into a better antagonist of BMP signalling (FIG. 7b).
A group of genes that have similar expression domains in the embryo, which usually correlate with function in a common biochemical pathway.
After Xolloid cleavage, however, xTsg-BMP binary complexes are formed preferentially. These binary complexes do not interfere with binding of BMP to its receptor and replace inhibitory CR1-BMP complexes. In agreement with these biochemical experiments, dorsalization by CR1 is readily competed by xtsg mRNA, and reduction of endogenous xTsg levels by a dominantnegative form of xTsg increases the anti-BMP effects of CR1 mRNA in Xenopus embryos103. In this second function, xTsg would provide a permissive signal to allow peak BMP signalling (FIG. 7b). BMP activity would be maximal in regions that have the highest concentrations of Xolloid, Chordin proteolytic fragments and xTsg. In Drosophila, the homologous region would correspond to the amnioserosa. In regions of the embryo in which residual full-length Chordin is still present, xTsg would facilitate the binding of any BMPs that have been released by Xolloid cleavage to fresh molecules of Chordin; this mechanism could facilitate the generation of borders between embryonic territories and ensure that peak BMP signalling occurs only in a specific region of the BMP activity gradient. In future, it will be worthwhile to investigate whether additional factors, such as a co-receptor, exist in this pathway. In addition to the two biochemical functions described above, xTsg also enhances the cleavage of Chordin by Xolloid in microinjected embryos (E.M.D.R., J.L. and M.O., unpublished observations). This third function suggests that the Chordin-BMP-xTsg ternary complex may be a better substrate for Xolloid cleavage.
Studies in Drosophila add a further layer of complexity to the function of Tsg. Constructs of Sog, termed Supersogs, which contain CR1 and further amino acids downstream of the first Tolloid cleavage site show new signalling properties. When overexpressed in the wing imaginal disc, full-length SOG will inhibit the BMP factor Glass bottom boat/60A (GBB), but not DPP. However, Supersog will inhibit DPP as well105. Because Supersog is insensitive to cleavage by Tolloid105, this inhibitory specificity may be explained by the formation of trimolecular complexes of Supersog, dTSG and DPP. Supersog-like fragments have been detected in Drosophila embryos overexpressing SOG. Furthermore, in biochemical studies dTSG binds to SOG and generates Supersog-like fragments in the presence of Tolloid and DPP105. In the future, it will be of interest to investigate whether ‘Superchordin’ fragments are generated in vertebrates.
In conclusion, Twisted-gastrulation seems to have many functions in BMP signalling. In vertebrates, it can increase the binding of BMP to full-length Chordin to create a better antagonist, compete for the residual BMP binding activity of CRs, thereby promoting BMP activity, and facilitate cleavage by Xolloid. In Drosophila, TSG generates inhibitory specificities by changing the cleavage site of Tolloid on its SOG substrate. Despite the current complexity, these recent studies indicate that xTsg/dTSG are part of an extracellular regulatory pathway involving Chordin/SOG and Xolloid/Tolloid that finely regulates the levels of BMP/DPP signalling during dorsal-ventral patterning.
The CR domains of Chordin are sufficient for BMP binding, and modules of similar sequence are present in many other proteins. The similarities are found in the spacing of the cysteines, as well as in additional amino acids (BOX 1). These Chordin-like CRs may function as binding sites for TGF-β superfamily members in the extracellular matrix82,106. Some of these proteins are produced in large amounts and, indeed,procollagen I is the most abundant protein of the human body. Recently deposited extracellular procollagen might serve as a sink for TGF-β factors that could then be released when required for tissue homeostasis. When procollagen is processed by its N-terminal proteinase, the NH2-propeptide that is released still retains a trimeric structure; it is therefore tempting to propose that a further proteolytic cleavage closer to the CR domain may be required to release individual CR domains and to reactivate the bound growth factors (BOX 1). In addition, proteins related to xTsg may function together with these other CR modules to further regulate growth-factor signalling. These exciting possibilities will keep researchers occupied in the near future.
Individual cysteine-rich (CR) domains similar to those of Chordin are present in fibrillar procollagen I, procollagen II and procollagen V, Thrombospondin-1 (TSP-1), and connective tissue growth factor (CTGF) (a). Other proteins contain several CR domains, such as Nel-1, Crossveinless-2, CRIM-1 and Kielin, which contain 4, 5, 6 and 27 CR domains, respectively107-110. In the case of procollagen IIA, the CR domain is located in the N-propeptide region and directly binds to BMP-4 and to TGF-β1 (REFS 82, 106). In Xenopus assays, full-length procollagen IIA mRNA has anti-BMP activity, which requires the CR domain82; however, the monomeric collagen IIA CR module does not have detectable biological activity. After secretion, procollagen IIA forms a homotrimer and the N propeptide containing the CR remains attached to collagen fibrils in the extracellular matrix (b)106. This structure brings together three CR domains and, as in the case of the multiple CR repeats of Chordin, may result in the stabilization of the interaction between the CRs and BMP. The cleavage closest to the CR, which would release bound growth factors, is hypothetical and is indicated by a question mark. The CRIM-1 homologue of Caenorhabditis elegans also has anti-BMP activity in Xenopus embryos82. Kielin has dorsalizing activity that is consistent with moderate anti-BMP activity110. The Crossveinless-2 protein of Drosophila contains five CR repeats but, unlike the other CR-containing proteins, is required genetically to increase levels of DPP/BMP activity during formation of the cross veins of the fruit fly wing108. Thus, it seems that many CR modules are involved in growth-factor regulation in the extracellular matrix.
The past few years have brought considerable progress in our understanding of how dorsal-ventral patterning is established in Xenopus and zebrafish embryos. It appears that the dorsal displacement of membrane vesicles is a key event in stabilizing β-Catenin on the dorsal side of the embryo. This, together with other maternal factors — such as VegT and Vg1 — leads to the generation of a gradient of Nodal-related signals. High levels of Xnrs induce Spemann’s organizer in dorsal mesoderm. The organizer secretes a cocktail of growth-factor antagonists that, in turn, further refine the pattern of the three germ layers. The ventralized ultraviolet-induced phenotype can be rescued by injecting vegetal cytoplasm (containing dorsal-determining vesicles), β-catenin mRNA, Nodal-related factors, or BMP antagonists, such as Chordin and Noggin. These gene products can all be considered as parts of a continuum that, starting with the prevention of β-Catenin degradation on the dorsal side, eventually leads to the inhibition of BMP signals that would otherwise cause the ventralization of the entire embryo. Although this model is admittedly oversimplified, it is attractive because it provides an explanation for why so many different gene products rescue the ventralization caused by inhibiting the movement of dorsal determinants.
Many questions remain unanswered. Can the elusive dorsal determinant vesicles be purified? Do they contain as yet unknown maternal Wnt ligands? Do additional Nodal-related genes exist in mammals? Can other growth factor/antagonist complexes — such as those of Cerberus, Dkks or Frzbs — also be reactivated by proteolytic regulation? Do all CR modules in extracellular proteins function in the regulation of the TGF-β superfamily? Do extracellular matrix CR modules interact with Twisted-gastrulation homologues? The studies that we have reviewed here identify a plethora of regulatory molecules in the early embryo, and, as we move into the era of the genome, it will be worthwhile to test to what degree these findings will serve as paradigms for understanding homeostasis in adult tissues.
DATABASE LINKS β-Catenin | Nodal | Noggin | Chordin | Cerberus | Frzb-1 | Crescent | Dickkopf | Xolloid | Twisted-gastrulation | Dishevelled | Frizzled-7 Wnt receptor | siamois | Xtwin | Nodal-related-3 | Axin | Cyclops | Squint | Goosecoid | Vg1 | BMP-4 | Derrière | Smad2 | VegT | Dlx-3 | BMP-2 | zebrafish Goosecoid | Mixer | Smad4 | FAST-1 | XIHbox-8 | Xnr-1 | Xnr-2 | Xnr-4 | Follistatin | Vent | Vox | sFRP-2 | Sizzled | Xwnt-8 | lefty-2 | nodal | BMP-2b | Smad5 | BMP-7 | short-gastrulation | Dpp | screw | | Drosophila Tolloid | zebrafish tolloid | Dtsg | GBB | procollagen I | N-terminal proteinase | procollagen II | procollagen V | CTGF | Thrombospondin-1 | Crossveinless-2 | Kielin | APC | GSK-3
FURTHER INFORMATION Xenbase | The zebrafish information network | Axeldb | De Robertis lab homepage
ENCYCLOPEDIA OF LIFE SCIENCES Xenopus embryo: β-Catenin and dorso-ventral axis formation | BMP antagonists and neural induction
We apologize to the many colleagues whose work we were unable to discuss owing to space limitations. We thank J. Abreu, J.I. Kim and E. Pera for comments on the manuscript. J.L. is a PEW Latin American Fellow, and M.O. and O.W. are Human Frontiers Science Program Organization postdoctoral fellows. Our laboratory is supported by the NIH and the Howard Hughes Medical Institute.