Differential Screening for Dorsal-Specific Zygotic cDNAs
In Xenopus embryos, it is possible to increase or decrease the amount of organizer tissue experimentally. LiCl treatment of 32-cell embryos results in an expanded Spemann organizer comprising a ring spanning the entire marginal zone at the gastrula stage (Kao and Elinson, 1988
) that expresses high levels of gsc
transcripts (Cho et al., 1991
). At later stages, these “dorsalized” embryos consist mainly of head structures, including radial eyes and cement glands. Irradiation of the vegetal pole of fertilized eggs with ultraviolet light (UV) results in “ventralized” embryos that lack an organizer (Stewart and Gerhart, 1990
) and gsc
expression (Cho et al., 1991
To isolate dorsal-specific genes, duplicate filters of an unamplified dorsal lip cDNA library (Blumberg et al., 1991
) were hybridized with probes synthesized either from dorsalized or ventralized gastrula mRNAs. To enrich for zygotically expressed genes, the probes were subtracted with maternal (8-cell embryo) mRNA. Screening of 25,000 plaques (see Experimental Procedures) yielded six independent groups of cDNAs enriched in LiCl-treated embryos.
To identify cDNAs of interest, further screening was performed in three ways with the longest clone of each group. First, the ability of its sense RNA to induce secondary axes was explored by microinjection into ventral blastomeres. Second, whole-mount in situ hybridization (Harland, 1991
) using a mixture of embryos of different stages was used to identify clones expressed in areas of known organizing activity. Third, the activation of candidate genes by organizer-specific homeobox products was tested by whole-mount in situ hybridization of embryos microinjected with gsc
mRNA at concentrations known to induce formation of secondary axes. In initial microinjection studies (data not shown), only one clone showed weak axis-forming activity, and while it was first expressed in the dorsal lip, at later stages it was expressed in the endoderm. More importantly for this study on organizer target genes, this clone (designated endodermin
) was not activated by microinjected gsc
mRNA; the properties of endodermin
will be presented elsewhere (Y. S., H. S., B. L., and E. M. D. R., unpublished data). In situ hybridization identified only one group, consisting of three clones (which eventually proved not to be full length), that was exclusively expressed in cells with Spemann’s organizer activity. The gene encoding these cDNAs, chordin
, was found to be activated by gsc
and was chosen for further study.
chordin Is Expressed in Head, Trunk, and Tail Organizer Regions
shows the spatial and temporal pattern of expression of Xenopus chordin
. By in situ hybridization, chordin
transcripts are first detected 1 hr before gastrulation (stage 9.5; Nieuwkoop and Faber, 1967
) in nuclei scattered in the dorsal marginal zone (). When the dorsal lip is first detected at stage 10.25, the chordin
transcript is found exclusively on the forming lip (), and cytoplasmic staining can also be observed. Once the dorsal lip is fully formed (stage 10.75), chordin
expression is intense both in superficial and involuted cells (). By the midgastrula stage, when a circular blastopore is formed (stage 11), most cells expressing chordin
have involuted (explaining the more diffuse appearance of the chordin
signal), except for cells in the dorsal half of the edge of the lip itself ().
chordin Is Expressed in Regions with Head, Trunk, and Tail Organizer Activity
At early neurula (stage 13), strong chordin
expression is detected in the prechordal plate (head mesoderm) and the notochord (). At the early tailbud stage (stage 26), chordin
is transiently expressed in the forebrain, fading from the prechordal plate and anterior notochord but remaining in the posterior notochord and tailbud hinge (). Later on (stage 33, corresponding to 42 hr of development), the chordin
signal is detected exclusively in the tailbud (). Closer examination reveals that the expression is localized in a specific region in the tailbud, the chordoneural hinge(). This is of interest because transplantation experiments have shown that the chordoneural hinge retains organizer activity at this stage (Gont et al., 1993
). Expression continues in the tip of the tail in swimming tadpoles (stage 42, 72 hr after fertilization; ). In embryos dorsalized by LiCl treatment, chordin
expression is enhanced, forming a ring that spans the entire marginal zone (compare ); this explains why the gene was isolated in the differential screen. Although shows only external views of embryos, all observations mentioned above were confirmed in embryos rendered transparent by clearing solution and in histological sections (data not shown).
Taken together, these descriptive studies show that chordin
is expressed initially in the dorsal lip and then in tissues derived from the organizer. The expression in the dorsal lip and prechordal plate overlaps in part with that of the homeobox gene gsc
. The later expression of chordin
in the notochord and chordoneural hinge does not coincide with that of gsc
but does overlap with the expression of another homeobox gene, Xnot2
(Gont et al., 1993
). These data led to the hypothesis that the chordin
gene might be positively regulated (directly or indirectly) by both homeodomain proteins.
chordin Is Activated by gsc and Xnot2
To test this hypothesis, synthetic gsc or Xnot2 mRNAs were microinjected radially into all blastomeres of 4-cell embryos and hybridized with a chordin probe at the gastrula stage. Both gsc mRNA (compare ) and Xnot2 mRNA () were able to induce ectopic patches of chordin mRNA in the ventral and lateral marginal zone. We also tested whether chordin could be activated in embryos lacking an organizer. In UV-treated embryos, chordin expression is eliminated (), in keeping with the dorsal character of this gene. When UV-treated embryos were injected diagonally into two blastomeres at the 4-cell stage, two patches of chordin expression could be induced both by gsc () and Xnot2 (data not shown) mRNAs.
chordin, but Not noggin, Expression Is Activated by gsc and Xnot2 Homeobox-Containing mRNAs
The activation by these homeobox genes appears to be specific for chordin
. Radial injection of gsc
mRNA did not cause ectopic expression of the organizer-specific gene noggin
(). In addition to providing a convenient negative control, the inability of gsc
to activate noggin
in this assay suggests that noggin
is unlikely to mediate the noncell-autonomous effects of gsc
. As a further control, a biologically active homeobox mRNA of the Antennapedia
; Wright et al., 1989
) was injected and found not to activate chordin
expression (even at 5-fold higher concentrations than those used for gsc
mRNA; ). An unrelated control synthetic mRNA (human prolactin; Amaya et al., 1991
) also failed to activate chordin
We conclude that expression of chordin, but not that of noggin, can be activated by the gsc and Xnot2 homeobox gene products. The expression patterns of these transcription factors partially overlap with that of chordin; the effects of Xlim-1 and XFKH1, which also overlap in expression with chordin, were not tested in this study. While the results do not address the issue of whether this activation is direct or requires additional intermediate steps, they suggest that chordin may function downstream of dorsal transcription factors in the organizer.
chordin Induction by Activin Requires De Novo Protein Synthesis
Activin is a potent inducer of dorsal mesoderm (Smith, 1993
). Several genes can be activated by this growth factor even in the absence of de novo protein synthesis. Such Xenopus primary response genes include Mix-1
(Smith et al., 1991
(Cho et al.,1991
; Tadano et al., 1993
(Taira et al., 1992
(Dirksen and Jamrich, 1992
), and Xnot
(von Dassow et al., 1993
). Since gsc
can induce ectopic expression of chordin
mRNA, it was of interest to test whether chordin
is also a primary response gene to mesoderm induction or whether it is activated subsequent to the expression of organizer-specific homeobox genes.
compares the temporal expression patterns of gsc and chordin by Northern blot analysis. gsc has a small amount of maternal transcripts, and zygotic ones become detectable at stage 9, 2 hr before gastrulation starts. In contrast, chordin expression is not detectable until stage 9.5, 1 hr before gastrulation starts. Thus, while gsc transcripts accumulate shortly after zygotic transcription starts at midblastula, those of chordin accumulate 1 hr later.
chordin Induction by Activin Requires De Novo Protein Synthesis
To test whether chordin
is a primary or a secondary response gene, animal cap explants were incubated with activin in the presence or absence of cycloheximide (CHX), which inhibited protein synthesis by 95% (see Experimental Procedures). chordin
was induced by activin, but this induction was significantly decreased by CHX (). In contrast, gsc
induction by activin was somewhat increased by CHX, in agreement with previous observations (Tadano et al., 1993
). As noggin is an organizer-specific secreted factor, it was important, in the wider context of this study, to determine whether noggin
is a primary response gene. As shown in , noggin
transcripts were induced by activin in the presence of CHX (even to a higher level than in its absence), indicating that noggin
, like gsc
, is a primary response gene.
Together with the time course, the animal cap studies indicate that the induction of chordin by activin treatment involves intermediate steps requiring de novo protein synthesis. Thus, the induction mechanism of chordin differs from that of other organizer-specific genes described to date, including noggin.
chordin Encodes a Novel Putative Secreted Protein
The three chordin cDNAs isolated in the initial screen were partial clones resulting from reverse transcriptase priming within an A-rich stretch within the coding sequence. These clones failed to reveal any biological activity in extensive microinjection experiments performed before sequence information was available. We rescreened the cDNA library, and a full-length 3.8 kb chordin cDNA was isolated and sequenced. The deduced protein sequence of the longest open reading frame is shown in . chordin encodes a large protein (predicted molecular mass of 105 kDa) of 941 amino acids. Hydropathy analysis showed a single hydrophobic segment comprising the 19 amino-terminal amino acids, followed by a putative signal sequence cleavage site. The presence of a signal peptide and the lack of possible transmembrane segments suggest that chordin is a secreted protein. There are four possible N-glycosylation (NXS/T) sites. When compared with its own sequence by dot matrix alignment, the chordin protein was found to contain four internal repeats of 58–74 residues (). Each repeat contains ten Cys residues at conserved positions as well as four other conserved amino acids, which are boxed in .
Xenopus chordin Encodes a Putative Secreted Protein
When the Cys-rich repeats were used to search the BLAST network data bases, it was found that similar Cys-rich repeats are present in several extracellular proteins. The conservation is restricted primarily to the spacing of Cys residues (). Thrombospondin 1 and 2 and α1 procollagen types I and III are extracellular matrix proteins that contain a single Cys-rich domain near the amino terminus (Bornstein, 1992
). Interestingly, these proteins are trimeric, and the Cys-rich domains may be involved in their multimerization (Bornstein, 1992
). von Willebrand factor, a protein that facilitates adhesion of platelets during blood clotting, contains two Cys-rich domains at its carboxyl terminus (Hunt and Barker, 1987
). Since the Cys-rich repeats are found in extracellular proteins and chordin contains such repeats at the both termini, the presence of these structural motifs supports the view that chordin may be a secreted protein.
Outside of the Cys-rich repeats, the rest (681 amino acids) of the chordin protein does not have significant homology to any sequences in the data bases. We conclude from these sequence comparisons that chordin mRNA encodes a novel putative secreted protein. Because chordin is specifically expressed in regions of the embryo that have organizer activity, we next tested whether this molecule is active in inductive signaling.
chordin mRNA induces Secondary Axes
For phenotypic analysis, the chordin
cDNA was subcloned into an expression vector (Amaya et al., 1991
) and RNA synthesized with SP6 polymerase was injected into a single blastomere of Xenopus embryos. chordin
mRNA induced secondary axes at substantial frequencies when injected into ventrovegetal blastomeres (59% at the 8-cell stage, n = 46; 37% at the 32-cell stage, n = 27). When injected into dorsal or animal (top) blastomeres, a high proportion of dorsalized embryos resulted. Embryos injected with a control mRNA encoding an unrelated secreted protein (human prolactin in the same vector; Amaya et al., 1991
) were unaffected.
shows a typical secondary axis induced by a single ventral injection of chordin
mRNA at the 8-cell stage. Immunostaining with a notochord marker showed that this secondary axis () contained a notochord and lacked anterior structures such as eyes and cement glands, but had auditory vesicles, implying that the axis extended anteriorly at least as far as the hindbrain. Of 37 twinned embryos (injected at the 8-cell stage) stained with this antibody, 64% had a notochord in the secondary axis and 54% had secondary auditory vesicles. In Xenopus, the absence of a differentiated notochord in experimentally manipulated embryos is not uncommon (Steinbeisser et al., 1993
). Many axes were similar to those induced by ectopic expression of gsc
or activin (Steinbeisser et al., 1993
), but differences were also noted. In particular, embryos with double tails were found in 11% of the twinned embryos (n = 223), suggesting that chordin
can induce tail-organizing activity in some cases; this phenotype has not been observed in secondary axes induced by gsc
(e.g., Cho et al., 1991
; Steinbeisser et al., 1993
). show one such embryo with double tails in which the secondary tail does not express the notochord marker. Embryos were scored as having double tails only when the primary and secondary axes were separate throughout their entire length; in general, the secondary tailbud formed in the ventral side, directly opposite to the primary axis (). Twinned embryos with duplicated anterior structures were observed at a low frequency (15% with double cement glands and 4.5% with extra eyes; ). On the other hand, enlarged head structures, such as cement glands, were commonly observed in dorsalized embryos caused by chordin
mRNA injections into dorsal or animal blastomeres ().
chordin mRNA Induces Secondary Axes
To examine the fate adopted by injected and uninjected cells, single blastomeres were injected at the 32-cell stage with a mixture of chordin and β-galactosidase (lineage tracer) mRNAs. It appears that chordin mRNA has both long- and short-range effects. shows an embryo injected into the A4 blastomere, in which the injected cells remained in the ectoderm and did not contribute to the secondary axis. When a ventromarginal blastomere, C4, was injected, almost all embryos with secondary axes displayed labeling in part of the secondary axis itself (data not shown, but see below), suggesting that the chordin-injected cells had their fate changed into that of Spemann’s organizer.
Dose-Dependent Axial Rescue by chordin mRNA
chordin Rescues Complete Axes in Ventralized Embryos
To test further the axis-forming activity of chordin
, a UV-rescue assay was used. Due to its high level of sensitivity, this method is frequently favored in Xenopus embryology (Smith and Harland, 1991
; Steinbeisser et al., 1993
; Thomsen and Melton, 1993
). shows ventralized embryos resulting from UV treatment. They lack axial structures as indicated by a dorsoanterior index (DAI; Kao and Elinson, 1988
) of 0.1. (In this scale, a DAI of 0 corresponds to embryos with no axis, and a value of 5 to a normal embryo.) When 75 pg of synthetic chordin mRNA was injected into a single vegetal blastomere of UV-treated embryos at the 8-cell stage, substantial rescue of trunk and tail structures occurred (), although the embryos still lacked the most anterior head structures (DAI = 2.1). These phenotypes were similar to those observed in rescue experiments with gsc
or activin mRNA (Steinbeisser et al., 1993
). When 150 pg of chordin
mRNA was injected (), the entire axis was rescued (DAI = 4.2), including eyes and cement glands. When the amount of chordin
RNA was doubled (), embryos with exaggerated dorsoanterior structures, such as multiple cement glands (indicated by arrowheads), resulted (DAI = 6.2).
When UV-treated embryos were microinjected into a blastomere of the C-tier region with a mixture of chordin and β-galactosidase mRNAs as a lineage tracer, the labeled chordin-injected cells were located in the dorsal axis (n = 19), usually in anterior regions, or in endodermal cells (n = 5) (data not shown). Histological analysis showed that the injected cells contribute preferentially to dorsal tissues, i.e., notochord and somites (). This implies that expression of chordin changes the fate of the injected cells into organizer-like tissue. Most of the dorsal axis, however, was recruited from uninjected cells, including most of the somite and all of the neural tissue shown in . The noncell-autonomous effects of chordin on neighboring cells are best illustrated by the embryo shown in , in which only a small sector of notochord was derived from injected cells, while most of the notochord (as well as the rest of the dorsal axis) was recruited from uninjected cells. Some rescued axes, particularly at low chordin mRNA concentrations, lacked a notochord and the somites were fused. In such cases (), the injected cells were found in the somite, principally in the midline underlying the neural tube, indicating that chordin can rescue axial structures even in the absence of notochord tissue.
We conclude that chordin mRNA can completely rescue axis formation in ventralized embryos. The injected cells preferentially give rise to organizer derivatives and are able to recruit neighboring cells to form multiple dorsal tissues, including notochord.
chordin Modifies Mesoderm Induction
To test whether chordin
mRNA has mesoderm-inducing activity, embryos were injected in the animal pole and animal cap explants were prepared at midblastula. Animal caps injected with control prolactin or with chordin
mRNA failed to elongate and (as determined by histological analysis) consisted of atypical epidermis and lacked mesodermal tissues (). Thus, chordin
lacks mesoderm induction activity per se. However, microinjection of chordin
mRNA can dorsalize ventral marginal zone (VMZ) explants (), which have received the inductive signals that lead to formation of ventral mesoderm. A class of molecules, called competence modifiers, which cannot induce mesoderm on their own but can regulate the response of embryonic cells to induction, has been described (Moon and Christian, 1992
chordin RNA Dorsalizes Ventral Mesoderm
We next tested whether chordin can dorsalize mesoderm induced by basic fibroblast growth factor (bFGF; Slack, 1991
). When animal caps were injected with either control prolactin or chordin
mRNA and treated with bFGF (), the control caps formed ventral mesoderm (blood and mesothelium), while chordin
-injected caps formed dorsal mesoderm (notochord and muscle) as well as dorsoanterior ectodermal inductions such as cement glands and blocks of neural tissue.
We conclude that while chordin mRNA alone is unable to induce mesodermal tissues in animal caps, it is able to synergize with bFGF, leading to the dorsalization of ventral mesoderm.