Mesodermal subdivision along the mediolateral axis in chicken controlled by different concentrations of BMP-4

During early embryogenesis in vertebrates, one of the most dynamic events is gastrulation, which leads to the initial formation of the mesoderm (see review, Kimelman et al., 1992, and references therein). In chicken embryos, the mesoderm is derived from the cells that invaginate through the primitive streak, located along the midline. Subsequently, except for the axial-structure-forming cells, the mesodermal cells migrate laterally and anteriorly, and eventually those that are located medially differentiate into somitic mesoderm, and those located more laterally contribute to the lateral plate mesoderm (Selleck and Stern, 1991; Schoenwolf et al., 1992). Thus, as diagrammatically shown in Fig. 1, the mesoderm is subdivided in a stereotypic manner along the mediolateral (M-L) axis. This M-L axis is also considered to be a dorsoventral (D-V) axis since a flat chicken embryo is sequentially folded into a three dimensional structure so that the medial-most structure will be located in the dorsal-most region and the lateral-most one will be in the ventral-most area. It is important to unveil the mechanisms by which this spatially determined subdivision of the mesoderm along the M-L axis is established.

The present experiments were designed to see how the subsets of the mesoderm are determined by taking advantage of the flat structure of a chicken embryo, which facilitates embryological manipulations at the molecular level. When we transplanted presomitic mesoderm into the lateral plate, the graft showed char-acteristics of the lateral plate. This indicates that the presomite is not yet committed and also that the lateral plate has a cue for lateralization. We hypothesized that BMP-4, a member of the TGF β-family, which is normally expressed in the lateral plate (Roberts et al., 1995; Pourquie et al., 1996; Takahashi et al., 1996; Watanabe and LeDouarin, 1996), is a molecule responsible for this lateralization. It is known that BMP-4 is involved in the formation of the ventral mesoderm during gastrulation in Xenopus embryos (Jones et al., 1992; Fainsod et al., 1994; Harland, 1994; Schmidt et al., 1995). We investigated the effects of ectopic administration of BMP-4 on the presomitic mesoderm in chicken and observed that BMP-4 lateralized the presomitic mesoderm in a concentration-dependent manner. From these results, we propose that BMP-4 acts as a lateralizer during the mesodermal subdivision along the M-L axis.

Fertilized White Leghorn chick and Japanese quail eggs from commercial sources were used throughout this study. Microsurgery was performed on embryos at stage 14 (about 21-somite) according to Hamburger and Hamilton (HH) (Hamburger and Hamilton, 1951) (about 50 hours of incubation in a humidified atmosphere at 38.5°C). A window was made in the eggshell and India ink diluted 1:5 in Hanks solution (140 mM NaCl, 5.4 mM KCl, 5.6 mM glucose, 0.34 mM Na₂HPO₄, 10 mM Hepes, 1 mM MgCl₂, 1 mM CaCl₂ pH 7) was injected into the sub-blastodermic cavity to make the embryonic structures more visible. A slit was made in the ectoderm overlying the segmental plate with a sharpened tungsten needle. An aggregate of COS 7 cells was then transplanted onto the anterior-most region of the presomitic region. In most cases, the operated embryos were fixed either at stage 18 (E3.0) or stage 22-23 (E3.5 to E4.0). Embryos were then treated for in situ hybridization (see below).

Quail-chick chimeras (Le Douarin, 1969, 1982) were made as follows. A fragment of the somite or somatopleure was isolated from quail embryos in 1.25% pancreatin (Gibco/BRL) in phosphate-buffered saline (PBS, Ca₂₊, Mg₂₊ free) for 5 minutes at room temperature followed by the addition of fetal bovine serum to stop the reaction. The fragment was then transplanted into chicken host embryos of an equivalent developmental stage. Three types of transplantation were performed in ovo. (1) The anterior-most region of the quail segmental plate was transplanted into the somatopleural region of a chicken embryo at the same level along the anteroposterior (A-P) axis. One day after the operation, the embryos were fixed with 4% paraformaldehyde (PFA) in PBS for in situ hybridization. The specimens were then refixed with Carnoy’s fixative for Feulgen staining (Feulgen and Rossenbeck, 1924) to identify the quail cells among the chick cells in the chimera. Although the Feulgen staining following in situ hybridization decreases the clarity of the quail nuclear marker, the grafted quail cell population was still obvious (for example, see Fig. 1A-D). (2) A quail fragment of a somite of somite-stage I was transplanted between the somite (somite-stage I) and the overlying ectoderm of a chicken embryo. (3) The medial-most fragment of the somatopleural mesoderm of a chicken embryo was replaced by the equivalent portion of a quail tissue. In cases (2) and (3), the specimens were directly fixed with Carnoy’s fixative 2 days after the operation and stained by the Feulgen method. For histological sectioning, the fixed embryos were dehydrated in ethanol and xylene and embedded in paraffin, followed by serial 7 βm sectioning.

COS 7 cells were grown until 70% confluent and transfected with 1 βg of pCDM8 (In Vitrogen)-derived vector, which contains the elongation factor promoter and the mouse full-length cDNA of BMP-4 or the lacZ gene along with 6 βl of lipofectamine (GIBCO BRL) in a 35 mm culture dish for 5 hours according to the manufacturer’s instructions. Transfected cells were cultured for 24 hours and then transferred to a dish coated with 1% agar to obtain cell aggregates. A cell aggregate of approximately 100 βm diameter was used for transplantation. For dilution of BMP-4/COS, the transfected COS cells were trypsinized and dissociated into single cells and mixed at a ratio of cell number, 1:1, 1:4, 1:9 and 1:24, respectively, with the cognate COS cells that had also been dissociated. Thus, a relative dose of BMP-4 secreted by a COS cell aggregate reflects each dilution level. At least four to twelve specimens were obtained at each dilution level.

cDNA fragments for RNA probe preparation were as follows: (1) Pax 3, a 330 bp fragment of the chicken Pax 3 cDNA subcloned in pBluescript SK₊(Stratagene) (given by Dr Kuroiwa, Nagoya); (2) MyoD, an EcoRI-EcoRI fragment of 1500 bp chicken MyoD (CMD-1) subcloned in SKII (β) (gift from Dr Fujisawa, Tokyo); (3) Hox b-6, an EcoRI-EcoRI fragment of 1100 bp of chicken Hox b-6 subcloned in bluescript SK (given by Dr Kuroiwa, Nagoya); (4) a fragment of the chicken cytokeratin cDNA was obtained by RT-PCR according to the published sequences (Charlebois et al., 1990b); (5) Msx 2, an EcoRI-EcoRI fragment of 602 bp containing the 5 β-half region of the homeobox, and BamHI-EcoRI fragment of 795 bp at the 3 β-most region of the non-coding region of the quail Msx 2 gene (Takahashi and Le Douarin, 1990), were subcloned into pBluescript KS₊, respectively. The two cRNAs for Msx 2 were mixed in the hybridization buffer; (6) Pax 1: an EcoRI-EcoRI fragment of 1526 bp chicken Pax 1 subcloned into KS II (Ebensperger et al., 1995).

Degenerate oligonucleotide primers directed to the sequences MKEKS and FDGCYQN from the sequence of the mSim1 containing bHLH domain (Fan and Tessier-Lavigne, 1994; Pourquie et al., 1996) were used to isolate a 618 bp fragment of the chick Sim1 gene by RT-PCR using mRNA derived from E2 chick embryos as a template. Digoxigenin-labeled RNA probes were prepared according to the instructions of Boehringer Mannheim. Both antisense and sense probes were prepared and specific signals with antisense probe were confirmed for all of the probes.

Embryos were fixed for 2 hours in 4% paraformaldehyde in PBS, washed twice in PBT (PBS, 0.1% Tween-20) and then dehydrated by passing them successively through 25%, 50%, 75% and 100% methanol in PBT. The specimens were bleached in 6% hydrogen peroxide in methanol for 2 hours, followed by successive rehydration and washing twice in PBT. The specimens were treated with proteinase K (20 βg/ml) for 20 minutes for embryos at embryonic day 4 (E4), and 15 minutes for E2. They were refixed in 0.2% glutaraldehyde/4% paraformaldehyde in PBS for 20 minutes at room temperature. After two PBT washes, the embryos were transferred to hybridization buffer (50% formamide, 5 β SSC pH 4.5, 1% SDS, 50 βg/ml tRNA, 50 βg/ml heparin) and prehybridized for 1 hour at 70°C. Hybridization was carried out overnight at 70°C in the hybridization buffer containing a digoxigenin-labelled RNA probe. The embryos were washed three times for 30 minutes each time in 50% formamide, 5 β SSC pH 4.5, 1% SDS at 70°C, then after a quick wash in RNase buffer (0.5 M NaCl, 10 mM Tris pH 7.5, 0.1% Tween-20), they were treated with RNase A (100 βg/ml) for 30 minutes at 37°C, followed by a quick wash in RNase buffer. The specimen were next washed three times for 30 minutes each in 50% formamide, 2 β SSC pH 4.5 at 65°C, followed by washing in TBST (8 mg/ml NaCl, 0.3 mg/ml KCl, 2.5 mM Tris pH 7.6, 0.1% Tween-20). The embryos were preblocked in 10% fetal calf serum in TBST followed by incubation overnight in blocking solution containing alkaline phosphatase-conjugated anti-digoxigenin antibody (Boehringer-Mannheim) that had been preabsorbed with chicken embryo powder. After the embryos were washed extensively in TBST for at least 5 hours, they were processed to NTMT (100 mM Tris-HCl pH 9.5, 100 mM NaCl, 50 mM MgCl_2, 0.1% Tween-20). The alkaline phosphatase activity was visualized by incubating embryos in NTMT containing 0.45 mg/ml nitroblue-tetrazolium chloride (Boehringer Mannheim) and 0.175 mg/ml 5-bromo-4-chloro-3-indolyl phosphatase (Boehringer-Mannheim). After stopping the color reaction, embryos were post-fixed in 0.1% glutaraldehyde/4% paraformaldehyde in PBS for 30 minutes at 4°C to be processed for paraffin embedding.

Hybridized embryos were embedded in paraffin wax after dehydration with ethanol and xylene. Serial sections of 15-25 βm thickness were prepared and mounted in Entellan (Merck).

Embryos at the 9-day embryonic stage from which visceral organs and eyeballs had been removed were fixed in 80% ethanol, 20% acetic acid and 0.015% Alcian blue. After dehydration in 100% ethanol for 5 days, they were stained with 0.01% Alizarine red in 0.5% potassium hydroxide for 2 hours. Staining was stopped in 1% glycerol, 20% potassium hydroxide for 4 hours and cleared in 50 % glycerol.

Whether or not the differentiation of the presomitic mesoderm is irreversibly committed along the M-L (future D-V) axis when it becomes morphologically distinct from the lateral plate was determined by transplanting a piece of a quail presomite taken from a 2-day embryo (E2) into the lateral plate region of a chicken host embryo (schematically shown in Fig. 1, see Materials and Methods for details). In this experiment, we examined expression of cytokeratin mRNA as an early marker for the lateral plate (see Charlebois et al., 1990a,b for expression at earlier stages, and see Fig. 1A for E3), and Pax 3 expression for presomitic and somitic tissues (Williams and Ordahl, 1994). At present, cytokeratin expression is the only molecular marker that exclusively distinguishes the lateral plate from the presomitic region, because other molecular markers, such as expression of BMP-4, Msx 2, Hoxb-6 and Prx 1, which were used as shown later, were supportive but not conclusive for identification of the lateral plate. We also localized grafted cells by the Feulgen method, which identifies the quail nuclear marker (Le Douarin, 1969, 1982). The transplanted cells ceased to express Pax 3 (Fig. 1C,D; n=3), and instead turned on expression of cytokeratin mRNA (Fig. 1A,B; n=15). When old somitic mesoderm of somite-stage X (tenth somite anterior from the most recently formed somite; Christ and Ordahl, 1995) was transplanted into the same site, the graft continued to express Pax 3 (Fig.1 F-H; n=3) and did not turn on cytokeratin expression (Fig. 1E; n=5). We conclude, therefore, that the presomitic mesoderm is not firmly committed along the M-L axis and also that cue signals reside in the lateral region to confer the lateral mesodermal characteristics.

It has recently been shown that the BMP-4 mRNA is expressed in the lateral plate region (Roberts et al., 1995; Pourquie et al., 1996; Takahashi et al., 1996; Watanabe and LeDouarin, 1996; and also see Fig. 2A,B). Briefly summarized, expression of BMP-4 mRNA is obvious in the presumptive lateral plate mesoderm by stage 10 (=10-somite) and the expression level becomes high by stage 13 in the lateral plate, particularly at the presomite level along the A-P axis. We, therefore, asked whether BMP-4 in the lateral plate has a role in specification of the mesoderm along the M-L axis. We transplanted COS cells transfected with BMP-4 cDNA, which are known to exert specific BMP-4 activity (Takahashi et al., 1996), into the anterior-most region of the presomitic mesoderm of an E2 embryo (left diagram in Fig. 2 and also Materials and Methods for details). The molecular markers that we used to assess the effect of BMP-4/COS were expression of Pax 3 mRNA for the presomite and dermomyotome (Goulding et al., 1994; Williams and Ordahl, 1994), MyoD mRNA for the myotome (Weintraub et al., 1991; Pownall and Emerson, 1992) and Pax 1 mRNA for the sclerotome (Deutsch et al., 1988) (Fig. 2, right diagram).

When examined at E4 (2 days after the transplantation), a region where the somite would be normally formed was largely smaller with BMP-4/COS than that in the control side (Fig. 2F,H). In such a region affected by BMP-4, the formation of the somite was locally perturbed. The transplantation abolished the expression of Pax 3 (Fig. 2C,F, n=32) and Myo D (Fig. 2D, n=6), indicating that the dermomyotomal cells were absent. The signal of Pax 1 mRNA was also significantly reduced in the cells deep around the notochord (Fig. 2H; n=6). COS cells transfected in the absence of DNA or with the β-gal gene showed no such activities (Fig. 2E for β-gal/COS).

Using Pax-3 expression, the effect of BMP-4/COS was detected within 3 hours of the operation, when the implanted somite had developed to somite stage III. In the normal stage III somite, an epithelial dermomyotome has developed and expresses Pax 3. At this stage of a BMP-4/COS-implanted somite, however, neither the epithelial structure nor the Pax 3 signal was detected. (Fig. 2G, n=8). β-gal/COS did not affect the dermomyotomal differentiation (n=16, data not shown). Thus, when BMP-4/COS was transplanted into the presomitic region, response of the mesoderm to BMP-4 was clearly any notable effect (n=6 and n=5, respectively, data not shown).

We also observed the loss of formation of the dorsal root ganglion in the experimental side (Fig. 2F,H). The central nervous system was not appreciably affected. Grafting BMP-4/COS into the intermediate mesoderm or the lateral plate region resulted in no gross effect on the morphology (data not shown).

The consequence of BMP-4/COS on somite-derived morphogenesis at later stages was further investigated by examining skeletal formation at E9.0 in toto. No rib cartilage was formed at the BMP-4/COS-grafted site over a two-somite length (n=9, Fig. 2I). The vertebral cartilage was also severely disrupted at the operated site. Such an effect was not observed for β-gal/COS (n=7, Fig. 2J)

BMP-2-expressing COS cells had the same effects as BMP-4/COS, when examined with Pax 3 (n=4, data not shown). In contrast, COS cells transfected with BMP-7 cDNA (Nishimatsu et al., 1992), cells secreting lefty, a BMP family member (Meno et al., 1996) and cells producing Wnt 1 or Wnt 3a (Parr et al., 1993) showed no such activities in our Pax 3 experimental system, although these cells have been shown to exhibit specific activity of this molecule in a separate assay (Nishimatsu et al., 1992; Parkin et al., 1993; Meno et al., 1996; Kitajewski, personal communication). The effect on expression of Pax 3 and Myo D is therefore elicited specifically by BMP-4 and BMP-2.

Loss of the markers for the somitic mesoderm, particularly the dermomyotome, caused by the BMP-4/COS transplantation raised two possibilities: (1) the cells that would normally contribute to the dermomyotome died in response to BMP-4/COS, or (2) these cells survived but were transformed to another cell type. To distinguish between these possibilities, we transplanted a quail presomite in between the skin.

When the quail cells were transplanted together with BMP-4/COS (Fig. 3D), in contrast, the quail-derived cells were found as an unorganized mesenchymal cell population underneath the skin near the original transplantation site (Fig. 3E-G). Thus, a substantial fraction of the cells that were exposed to BMP-4 survived and were located at the place where they would normally reside, but they never formed the epithelial structure of the dermomyotome.

We conclude, therefore, that BMP-4 transformed the cells that would otherwise differentiate into the dermomyotome to another cell type.

To investigate the action of BMP-4 on somitic mesoderm, in addition to cytokeratin expression, an exclusive marker for the lateral plate, mRNA expressions of Msx 2, Hoxb-6 and BMP-4 were used as supportive markers for the lateral plate. Msx 2, BMP-4 and Hoxb-6 mRNAs are normally expressed in the lateral plate and along the dorsal midline both in the ectodermal and mesodermal components (Takahashi et al., 1992, 1996; and this study).

As shown in Fig. 4A,B, BMP-4/COS elicited ectopic ectoderm and somite of a chicken host embryo in order to label the dermomyotomal cells (Aoyama, 1993). When quail tissue was transplanted together with β-gal/COS as a control, all the quail cells were found in the epithelial structure of the dermomyotome (Fig. 3A-C,G expression of cytokeratin mRNA at the site of the graft (n=12). Observations of histological sections showed that the ectopic cytokeratin signal was localized in the superficial region including the mesenchyme and skin surrounding the BMP-4/COS (Fig. 4B). Induced expression of Msx 2 (n=15, Fig. 4D), BMP-4 (n=4, Fig. 4F) and Hoxb-6 (n=5, Fig. 4H) was also detected in the operated side. These patterns for Msx 2, BMP-4 and Hoxb-6 were almost complementary to that of Pax 3 as shown earlier (Fig. 2C). β-gal/COS transplantation did not induce expression of any of these markers (Fig. 4C,E,G,I, n>6 for each marker).

The induced expression of these markers was not due to the lateral plate cells having invaded the somitic area. This was shown by the fact that, when a piece of the lateral plate mesoderm was labeled by the quail-chick chimera technique together with the BMP-4/COS implantation into the presomite, the labeled cells stayed in the lateral plate where they would normally reside and did not migrate into the somite region (n=6, data not shown). The formation of the mesonephric tissue derived from the intermediate mesoderm located between the somite and the lateral plate mesoderm appears not to be grossly affected (Fig. 4B).

BMP-4 is known to induce the epidermis differentiation around the gastrulation stage, examined by the animal cap assay in early Xenopus embryo (Sasai et al., 1995; Wilson and Hemmati-Brivanlou, 1995). Although the stage of the somitic lineage focused in this study is much later than that studied in Xenopus embryos, we needed to exclude the possibility that the converted presomite by the effect of BMP-4 was the epidemis. To do this, we examined expression of the Prx 1 gene, which is normally positive exclusively in the mesoderm of the lateral plate and dermomyotome (Nohno et al., 1993). As shown in Fig. 4J,K (n=6), the presomite-derived tissues affected by BMP-4 were clearly positive for Prx 1, indicating that the affected tissue was not the epidermis but the mesodermal mesenchyme.

Taking the patterns of all these markers together, we conclude that at least a part of the somite was transformed to lateral plate mesoderm by BMP-4.

It has recently been reported by Pourquie et al. (1996) that transplantation of the cells that produce retrovirus carrying the BMP-4 gene converted the medial portion of the somite to the lateral somite rather than to the lateral plate. Expression of cSim 1, which is normally restricted to the lateral portion of the somite, was used in their analysis. We speculated that the apparent discrepancy in the resulting lateralization observed in this study and theirs reflects the difference in the effective concentration of BMP-4. To test this, we mixed BMP-4-producing COS cells with normal COS cells to give a dilution to 1/2, 1/5, 1/10 and 1/25 (see Materials and Methods) and examined their effects on expression of three markers, Pax 3, cSim 1 and cytokeratin mRNAs over a range of dilutions. More than four embryos were assessed for each marker for every dilution level of BMP-4/COS.

The fraction of the dermomyotome expressing Pax 3 as measured along the length of the mediolateral direction diminished as the concentration of BMP-4 increased (Fig. 5A-C). A reduced size of the dermomyotome revealed by Pax 3 expression was obvious at the dilution level 1/10, and the dermomyotome was almost lost at the level of 1/2 (Fig. 5A,C).

For induction of cytokeratin expression (index: cytokeratin-positive area induced ectopically/normal somitic area along the M-L line, see Fig. 5B), BMP-4 was not effective at the level 1/10. At 1/5, the index was significantly higher than the control and it was almost 1.0 at the non-diluted level (the entire mesoderm along the M-L axis was converted to express cyto-keratin mRNA; Figs 5A,C and 4B). The cytokeratin-positive area was observed in an area where the Pax 3 signal was abolished. Thus, at the threshold level of 1/5, the characteristics of the lateral plate were induced. Levels higher than this threshold affected a larger area of the presomite converting it to lateral plate.

At the dilution level of 1/25, the occupation index for cSim 1 (fraction of the cSim 1-positive dermomyotome/remaining dermomyotome) was substantially increased from the control value of 0.4, observed with β-gal/COS, indicating that a greater fraction of the somite displayed characteristics of lateral somite (Fig. 5A,C). By the dilution level 1/10, the dermomyotome was shortened and became fully positive for cSim1, and, at dilution levels higher than 1/5, the dermomyotome formation was lost. Consistent with this mediolateral conversion within the somite by low BMP-4 revealed by cSim 1 expression, Pax 3, which is normally more intensely expressed in the lateral somite, was also upregulated in regions some distance away along the A-P axis from the implanted site even with the undiluted BMP-4/COS, showing that declining concentration of BMP-4 conferred lateral somite characteristics.

These results demonstrated that there are multiple thresholds in the responsiveness to BMP-4 of the somitic mesoderm to be lateralized: a low concentration of BMP-4 results in somitic mesoderm with some lateral-somite characteristics and a high BMP-4 concentration induces a complete transformation of the somite to the lateral plate.

The somitic mesoderm, when it is formed, is morphologically distinct from more lateral mesodermal tissues. It has not heretofore been known to what extent the morphological segregation of the somitic mesoderm reflects the cellular specification and commitment. It has been reported, using a cell labeling technique (Stern et al., 1988), that the lineage of the cells located in the rostral portion of the segmental plate is restricted to the somite whereas those in the caudal portion contribute to the somites and also to the more lateral mesoderm. In this study, we have demonstrated that the presomitic mesoderm is not fully committed to the somite, since. when positioned ectopically in the lateral plate, it was respecified to the lateral plate. We have provided evidence that BMP-4, which is expressed in the lateral plate, is one of the signals involved in this mesodermal specifi-cation (Fig. 6A), and have also demonstrated that different concentrations of BMP-4 elicit distinct outputs in mesodermal subdivisions along the M-L axis.

The ventralizing activity of BMP-4 has also been reported in Xenopus embryos: the ‘fate’ of the dorsal mesoderm is changed to the ventral one by BMP-4 (Jones et al., 1992; Fainsod et al., 1994; Harland, 1994; Schmidt et al., 1995). These findings are largely consistent with our observations described in this study. However, most studies in Xenopus embryos are performed by injecting BMP mRNAs at earlier stages where cells are not yet specified. Further, manipulations of embryos at such early stages often interfere with the formation of both the D-V and A-P axes at the same time.

Here, the effect of BMP-4 was studied in a specific tissue and at a specific developmental stage. Transplantation of BMP-4-expressing cells induced a local conversion of the presomite to the lateral plate as shown by expression of several markers for the lateral plate, and also by labeling the presomite at the graft site. Lack of rib cartilages in a region corresponding to the graft site at later stages also supports this mediolateral conversion of the mesoderm. Our observation on cartilagenous formation affected by BMP-4 is also largely consistent with that reported recently by LeDouarin’s group (Monsoro-Burq et al., 1996). The results presented in this study suggest that, during normal development, BMP-4 expressed in the lateral plate is a cue that produces the lateral characteristics in the mesodermal tissues. Our results are also consistent with the reports using younger materials: when the prospective somitic mesoderm taken from the primitive streak at stage 4 (gastrulation stage) was heterotopically grafted into the prospective lateral plate region of the primitive streak, the graft contributed to the lateral plate (Garcia-Martinez and Schoenwolf, 1992). The axial tissues, Hensen’s node or notochord, are thought to provide the dorsalizing factors that are important for the formation of the axial and paraxial structures (Spemann, 1938; van Straaten and Hekking, 1991). BMP-4 administered in the somitic region in this study appears to override activities of these dorsalizing factors.

BMP-4 expression in the lateral plate mesoderm is readily detected after stage 10 (10-somite) (this study, and Watanabe and LeDouarin, 1996). Prior to this stage, BMP-4 mRNA is instead observed in the ectoderm where the prospective lateral plate mesoderm (BMP-4 negative) underlies it (Watanabe and LeDouarin, 1996). It is conceivable that the lateral ectoderm first signals the mesoderm to be specified to the lateral plate by the action of BMP-4 and that this action, in turn, produces a high concentration of BMP-4 in the lateral plate complex including the mesoderm. It has been reported, in Drosophila, that signals from the ectoderm specify the underlying mesodermal subtypes and also that decapentaplegic (dpp), a Drosophila homolog of BMP-4, mediates these interactions (Frasch, 1995).

The intermediate mesoderm, located between the somite and the lateral plate, differentiated into the mesonephric tissues including the Wolffian duct and mesonephric tubules despite BMP-4 administration. It is conceivable that the presomitic cells are responsive to BMP-4 signal by expressing BMP-4 receptors while the intermediate mesoderm is not responsive due to the lack of the receptors. The distribution pattern of BMP-4 type I and type II receptors remains to be investigated.

When BMP-4/COS was transplanted into the somite of somite-stage V, the responsiveness of the somite to BMP-4 declined. This suggests that determination of the somitic mesoderm along the M-L axis takes place around somitic-stage V, where dermomyotomal and sclerotomal differentiation becomes distinct. It has been reported that axial signals such as Sonic hedgehog emanating from the notochord (Fan and Tessier-Lavigne, 1994; Johnson et al., 1994) and signals from the neural tube (Munsterberg and Lassar, 1995; Pourquie et al., 1996) are important for these differentiations. It is conceivable that an initial lateralizing signal conveyed by BMP-4 produces a plastic polarity in the somite, and that axial signals, coming later, fix and elaborate the polarity.

We have found that different concentrations of BMP-4, locally administered in ovo, result in different mesodermal subtypes. At a low concentration, the medial component of the somite was taken over by the lateral somite (Fig. 6B), whereas a higher level converted the entire somite to the lateral plate (Fig. 6C). This suggests that, during normal embryogenesis, BMP-4 acts at different levels of concentration for the mesodermal specification along the M-L axis. A simple model is that the mesoderm positioned laterally, which is exposed to the highest level of BMP-4, is specified to the lateral plate, whereas the medial-most mesoderm is exposed to the lowest level and acquires characteristics of the medial somite (Fig. 6A). It seems that, at the relative level of BMP-4 in our experimental system above the threshold 1/5, the characteristics of the lateral plate were established, whereas the level above 1/25 conferred the lateral somite characteristics. Since BMP-4 mRNA is not detected in the somite mesoderm, in order for BMP-4 to be effective in the somite, its signal must be transmitted from the lateral plate to the somite either directly or indirectly. It has been reported that dpp acts as a gradient morphogen (Lecuit et al., 1996; Nellen et al., 1996). Distribution of the effective BMP-4 molecule in vivo remains to be investigated. It is unknown whether the lateromedial decline of the effective level of BMP-4 is simply by diffusion or is negatively controlled by unidentified inhibitors. This model is obviously oversimplified because it excludes possible effects emanating from the axial tissues such as chordin or noggin (Harland, 1994; Sasai et al., 1994), or those from the dorsal neural tube (Munsterberg and Lassar, 1995; Pourquie et al., 1996).

Noggin and chordin, expressed in the Spemann’s organizer in Xenopus, have been reported to counteract BMP-4 by a direct binding at the protein level (Piccolo et al., 1996; Zimmerman et al., 1996). Our preliminary observations also support that these molecules may act along the axial tissues (A. T. and Y. T., unpublished). Therefore, it is most likely that the subdivision of the mesoderm along the M-L axis is established by possible cooperation and counteraction between the axial-derived factors and different concentrations of BMP-4. A similar model for the mesodermal subdivision has been proposed by Pourquie et al. (1996) by showing that the M-L polarity within the somite was perturbed by retrovirus-encoded BMP-4. The present study has revealed that the BMP-4 concentration that Pourquie et al. (1996) achieved roughly corresponds to the 1/25 dilution level of our experimental system, which causes the somite to be lateralized. It has also been reported that different concentrations of BMP-4 and activin, another TGF β-family member, induced different subtypes of ectoderm (Wilson and Hemmati-Brivanlou, 1995) and mesoderm (Green et al., 1992; Gurdon et al., 1995), respectively, during gastrulation, using the animal cap assay in Xenopus embryos.

In a previous study, we observed that, when the roof plate, which is positive for BMP-4 mRNA, was transplanted in the vicinity of the host neural tube, it induced ectopic expression of Msx 2 in its overlying cells (Takahashi et al., 1992). In this study, we did not detect induction of cytokeratin mRNA expression upon the roof plate transplantation (n=8, data not shown). It remains to be studied whether this is because the roof plate secretes some unknown inhibitory factor(s) for lateral plate formation although it produces a high level of BMP-4, or whether it secretes a low concentration of BMP-4 protein, which is effective only for induction of Msx 2 and not for cytokeratin expression.

Lastly, importance of BMP-4 during early developmental processes was also demonstrated by a loss-of-function analysis in mice (Winnier et al., 1995). Although majority of homozygous null BMP-4 embryos died around gastrulation, some surviving embryos exhibited defects in the formation of the ventral-most mesoderm. A gain-of-function analysis of BMP-4, which is administered at a specific site and at a specific stage as presented in this report, is particularly useful to study the effects of the molecule on mesodermal specification and determination at relatively late stages, and also to study morphological consequences of ectopic action of BMP-4.

We are grateful to Hisato Kondoh, Yuji Yokouchi and Jeff Pollard for critical reading of the manuscript. We also thank Atsushi Kuroiwa for providing us with the Hoxb-6, Pax 3 and BMP-4 probes, Jan Kitajewski for Wnt1 and Wnt3a-producing cells, Hiroshi Hamada for lefty producing cells, Atsuko Fujisawa for the MyoD probe, Sumihare Noji and Tsutomu Nohno for the Prx 1 probe, and Haruhiko Koseki for the Pax 1 probe. We are grateful to Kazunori Hanaoka and Hirohiko Aoyama for helpful discussions. This work was supported by research grants from the Ministry of Education, Science and Culture of Japan, Research for the Future Program of JSPS (Japan), the Ciba-Geigy Foundation (Japan) for the Promotion of Science, Yamada Science Foundation, the Kitasato University Research Grant for Young Researchers, the Naito Foundation, and CREST (Core Research for Evolutional Science and Technology) of Japan Science and Technology Corporation (JST).