ABSTRACT
During limb development, the mesenchymal cells in restricted areas of limb bud, anterior necrotic zone, posterior necrotic zone, opaque zone and interdigital necrotic zones, are eliminated by programmed cell death. The transcripts of bone morphogenetic protein (Bmp)-2 and -4 were first detected in the areas where cell death was observed, then showed overlapping expression with the programmed cell death zones except the opaque zone. To investigate the function of BMP-2 and BMP-4 during limb pattern formation, the dominant negative form of BMP receptor was overexpressed in chick leg bud via a replication-competent retrovirus to block the endogenous BMP-2/-4 signaling pathway. This resulted in excess web formation at the anterior and posterior regions of limb buds in addition to marked suppression of the regression of webbing at the interdigital regions. Significant reductions in the number of apoptotic cells in these three necrotic zones were found in the limb buds which received the virus carrying dominant negative BMP receptor. This indicates that extra tissue formation is due to suppression of programmed cell death in the three necrotic zones. Moreover, BMP-2/-4 protein induced apoptosis of mesenchymal cells isolated from the interdigital region in vitro. Other TGFβ family proteins as TGFβ1 and Activin did not show this effect. These results suggest that BMP-2 and BMP-4 are the apoptotic signal molecules of the programmed cell death process in the chick limb buds.
INTRODUCTION
Programmed cell death (PCD) is an important process for eliminating unnecessary tissues during embryonic development for proper morphogenesis at genetically determined stages. PCD occurs in morphogenetic processes such as limb pattern formation, establishment of muscle-nerve connection, establishment of the sex-specific genital system and metamorphosis of amphibians (see review Hinchliffe, 1981).
The occurrence of PCD is a remarkable feature of limb development in the amniotes (see review Hinchliffe, 1981). The mesenchymal cells in the anterior and posterior marginal parts of chick limb buds, the anterior necrotic zone (ANZ) and posterior necrotic zone (PNZ), are eliminated by PCD at day 4-6 of development. PCD in the interdigital necrotic zone (INZ) occurs in the mesenchymal areas separating the chondrifying digits of the developing footplate in 8-day chick embryos. The INZ has been found in all amniote species with non-webbed digits indicating the presence of common mechanism for PCD. Treatment of chick embryos with bromodeoxyuridine suppressed PCD in the INZ and resulted in webs between digits remaining, indicating that PCD plays a role in shaping limb morphology (Tone et al., 1983).
If the mesenchymal tissue of the prospective PNZ at stage (st.) 17 (Hamburger and Hamilton, 1951) is transplanted to the flank, the cells die on schedule after an appropriate interval (Saunders et al., 1962). However, transplantation of the same tissue to the dorsal surface of the wing bud results in sup-pression of PCD (ibid.). In addition, removal of the surface ectoderm of the INZ area suppresses PCD in the INZ and results in extra digit formation (Hurle and Ganan, 1987). These observations imply that tissue interactions and programmed cell death in the INZ mesenchymal cells at developmentally programmed stages is crucial for proper PCD in the limb bud.
Several diffusible intercellular signaling molecules such as mullerian inhibitory substance (MIS), BMP-4 and Reaper were shown to mediate PCD in the regression of mullerian duct (Behringer et al., 1994), PCD in the neural crest cells from rhombomeres 3 and 5 (Graham et al., 1994), and PCD during Drosophila embryogenesis (White et al., 1994), respectively. In the developing limb buds, the regions of expression of bone morphogenetic protein (Bmp)-2 and -4 overlap with the PCD area (this paper), implying the involvement of these signaling molecules in PCD in the limb bud.
BMP-2 and BMP-4 are vertebrate homologs of Decapentaplegic in Drosophila, have 98% amino acid sequence identity and were shown to have invariant functions in many biological processes for tissue interactions (Kingsley, 1994a,b, and references therein) through a specific receptor complex (Koenig et al., 1994; Suzuki et al., 1994; Yamaji et al., 1994).
The BMP-2/-4 receptor complex is composed of type I and type II receptors both carrying serine/threonine kinase activity in their cytoplasmic domains (Koenig et al., 1994; Liu et al., 1995). The type I receptor/ligand complex associates with the type II receptor whose protein kinase activity is constitutively active. This association allows the type II receptor kinase to phosphorylate GS domain of the type I receptor activating its protein kinase activity. Unlike TGFβ or Activin receptor, the type I receptor of BMP-2/-4 can bind the ligand with high affinity without coexpression of type II receptor (Koenig et al., 1994; Liu et al., 1995). Several lines of evidence indicate that the mutant forms of Activin or BMP-2/4 receptors lacking the cytoplasmic domain behave as dominant negative molecules when overexpressed (Hemmati-Brivanlou and Melton, 1992; Graff et al., 1994; Suzuki et al., 1994). This would be due to the interruption of the endogenous signaling pathway by forming abortive complexes with the endogenous type II receptor(s) in vivo.
To elucidate the roles of BMP-2/-4 during limb development, we introduced the dominant negative form of BMP-2/-4 receptor type Ia (dnBMPR-Ia; Suzuki et al., 1994) into chick leg buds via a replication-competent retrovirus system to interrupt the signaling pathway. Overexpression of this dominant negative BMP-2/-4 receptor inhibited PCD in the ANZ, PNZ and INZs. In addition, we demonstrated that the BMP-2/-4 proteins behave as mediators of cell death of the mesenchymal cells derived from the interdigital necrotic zone in vitro.
MATERIALS AND METHODS
Whole-mount in situ hybridization
Whole-mount in situ hybridization analysis with DIG-labeled RNA probes was performed as described by Yokouchi et al., 1995). The chicken Bmp-4 and Bmp-2 cDNAs used as probes were isolated from a day-11 chicken embryonic cDNA library by screening with Xenopus probes (A. Kuroiwa, unpublished data).
Construction of the recombinant virus vector
pDS5 (Yokouchi et al., 1995), which is a variant of pDS3 (Iba et al., 1988) was used for the recombinant virus construction. The BspI-BamHI fragment containing the truncated mouse BMP receptor-Ia (ΔBMPRIa) cDNA that produces a receptor protein lacking the entire cytoplasmic domain (Fig. 1B) was inserted into pCITE-3 cleaved with NcoI and BamHI. pCITE-3 was constructed by replacing the ApaI-PstI fragment of pCITE-1 (Novagen) with the ApaI-PstI fragment from pCITE-2 (Novagen). The EcoRI-XbaI fragment of ΔBMPRIa-pCITE-3 was inserted into the EcoRV site of pDS5 after blunting with T4 DNA polymerase. SalI digest of this construct was ligated with the DNA fragment carrying essential genes for virus production, and ligated product was transfected to chicken embryonic fibroblasts as described previously (Iba et al., 1988).
Chick strain and virus production
All animals were obtained as fertile eggs from the supplier. For injection with A-subgroup retrovirus, specific pathogen-free White Leghorn eggs (obtained from line M, Nisseiken Ltd.) were used. Recombinant viral particles were produced as described by Iba et al. (1984, 1988). Injection of the viral particles into virus-free White Leghorn embryos (Nisseiken) was performed as described previously (Yokouchi et al., 1995). As a control, we infected embryos with the virus carrying no exogenous gene sequences, in the same way.
Skeletal staining
The legs of day-7.5 embryos were stained with Alcian blue as described by Yokouchi et al. (1995). Briefly, embryos were dissected and fixed in 10% formaldehyde overnight. Embryos were immersed in 70% ethanol/1% HCl for 3 hours, stained with 0.1% Alcian blue/ 70% ethanol/1% HCl for 3 hours, washed with 70% ethanol/1% HCl overnight and then dehydrated. The legs of day-10 embryos were stained with hematoxylin.
Detection of cell death
Vital staining with Nile blue sulfate was performed as described by Tone et al. (1983). Briefly, embryos were stained with 0.01% Nile blue sulfate/PBS for 20 minutes at 37°C, and washed for 3 hours in PBS. The TUNEL method (Gavrieli et al., 1992) for detecting fragmented DNAs which appear following apoptosis on paraffin sections was performed using an ApopTag-FITC kit (Oncor) according to the manufacturer’s protocol.
Cell culture
Micromass culture of the mesenchymal cells isolated from the third interdigital region of chick leg buds at stage 29 was performed in F-12 medium (Nissui) (Aono and Ide, 1988) with minor modifications. In this experiment, 2×105 cells were plated onto 24-well tissue culture plates coated with rat type I collagen (Becton Dickinson). After 24 hours, the cultures were fixed in 4% paraformaldehyde/PBS, processed by the TUNEL method using a Cell Death Detection Kit-ALP (Boehringer) or stained with DAPI. For in vitro virus infection, the mesenchymal cells from ID3 of st.28 embryos were isolated and cultured for 3 hours as described above. Then the culture medium was replaced with 70 μl of medium containing control (no insert virus) or dnBMPR-1a virus (2-5×106 virus/ml) and incubated for 4 hours. After changing the medium to F12 containing 5 ng/ml of bFGF, cells were incubated for 20 hours. Subsequently, the medium was changed to control or BMP-2 containing medium and cultured for a further 24 hours. By this method, nearly complete infection was observed using a monoclonal antibody against the gag protein (Potts et al., 1987). Recombinant human bFGF was supplied by KAKEN. Recombinant Xenopus BMP-2 and Xenopus BMP-4 were kind gifts from TAKEDA, and recombinant human Activin was from AJINOMOTO. Recombinant human TGFβ1 was purchased from R&D systems.
Removal of dorsal ectoderm on leg buds
Fertilized egg at 5.5 days (st. 28) was windowed, and a few drops of 0.1% Nile Blue/PBS was dropped on the leg bud. The lightly stained dorsal ectoderm on interdigit 3 was peeled with the sharpened tungsten needle. Then the egg was sealed and incubated for 12 hours.
RESULTS
Introduction of the truncated BMP receptor induces the ectopic web formation
To elucidate the functions of BMP-2 and BMP-4 during limb development, we attempted to interrupt BMP-2/-4 signaling by overexpression of the truncated form of the BMP receptor using the virus vector in the leg bud. We constructed a recombinant virus carrying the truncated form of BMP receptor-Ia (dnBMPR-Ia) (Fig. 1C), which is composed of the ligand binding domain and the transmembrane domain (Fig. 1A,B). It has already been reported that the introduction of the same form of the receptor successfully inhibited signaling of the endogenous BMP-4 in the early Xenopus embryo (Suzuki et al., 1994). We injected the recombinant viral particle suspension into the prospective right leg field of chick embryos at stage 10 (Fig. 1D) and allowed the embryos to develop to appropriate stages. By this procedure, the foreign gene contained in the recombinant virus was expressed in the mesenchymal cells of nearly the entire limb bud (Fig. 1E).
We analyzed the phenotypes at day 7.5 and day 10. On the left side limbs, the interdigital webs began to regress, as observed in normal development, at day 7.5 (Fig. 2A). In contrast, on the infected right side, ectopic mesenchymal bulges were formed at a region anterior to digit 1, and at a region posterior to digit 4 at high efficiency (Fig. 2B,D asterisks, Table 1). Moreover, regression of the web was sup-pressed (arrowheads) and the digits were more separated from each other because of the presence of extra intervening tissues. At day 10, on the control side, complete regression of webs was observed in all of the interdigital regions (Fig. 2E). In contrast, on the infected side, webs that should have regressed normally remained in the interdigital region (Fig. 2F-H, Table 1). In embryos infected with the control virus, these morphological changes were not observed (data not shown, Table 1). The ectopic bulges formed at day 7.5 (Fig. 2B,D) and which remained as webs at day 10 did not contain any cartilaginous tissue (Fig. 2G). Thus, interruption of BMP-2/-4 signaling resulted in extra tissue formation in the leg buds. In addition, formation of the second and third phalanges was inhibited at high frequency (Fig. 2F-H). This indicates that BMP-2/-4 have different functions in cartilage formation in the digital region. It is possible that persistent webbing is a secondary consequence of a shortening of the digits. However this is unlikely because webbing was observed even in the specimen in which no significant reduction of digits was detected (Fig. 2F digit 3 and 4).
The truncated BMP receptor causes suppression of programmed cell death
As shown above, regions where the ectopic bulges and webs were formed by overexpression of the dnBMPR-Ia correlated well with the regions where mesenchymal tissue was eliminated by PCD. The same phenotype was seen in the BrdU-applied limb bud, in which PCD has been suppressed (Tone et al., 1983). To analyze whether PCD was suppressed by over-expression of dnBMPR-Ia, the infected embryos were allowed to develop to day 7.5 and the dead cells were visualized by Nile blue vital staining (Tone et al., 1983). On the control side, large numbers of dead cells were detected in the ANZ, PNZ and INZs (Fig. 3A,C,E). On the infected side, no dead cells were detected within the ectopic mesenchymal bulges that formed in the ANZ and PNZ (Fig. 3B,E) and the number of dead cells in the INZs was reduced by half (Fig. 3B,D). We then examined the apoptotic cells in the INZs by the TUNEL method (Gavrieli et al., 1992). In the INZs of the control side, many labeled cells were found in the central part of the INZs (Fig. 3G,I). On the infected side, the number of labeled cells was reduced by half (Fig. 3H,J). No suppression of apoptosis was observed when the control virus was used (data not shown). These results demonstrate that dnBMPR-Ia suppresses apoptosis in the leg bud, strongly supporting the hypothesis that BMP-2/-4 mediates apoptotic signals.
Bmp-2/-4 expression and PCD in the leg bud
The results shown above indicate that BMP-2/-4 function as the apoptotic signal during limb development. We next analyzed the relationship between the spatial and temporal expression profiles of Bmp-2 and Bmp-4, and progression of apoptosis in the leg buds.
Apoptotic cells in the ANZ were detected from st. 23 (day 3.5) to st. 32 (day 7.5; blue dots in Fig. 4A-E). Apoptotic cells in the PNZ appeared at st. 29 (day 6) and was present to st. 32 (Fig. 4C-E). At st. 23, the Bmp-4 transcripts were distributed in the mesoderm of the anterior marginal region and that in the posterior proximal region (Fig. 4G,P). From st. 25 (day 5) to st. 32, the expression domains overlapped with the ANZ and the PNZ (Fig. 4H-K).
From st. 31(day 7) to st. 33 (day 8), apoptotic cells appeared in the INZs. Each INZ is composed of two subregions, a proximal and a distal region and the order of appearance of which is strictly regulated (Pautou, 1975). At st. 31(Fig. 4D), the first apoptotic zone was detected in the proximomedian part of interdigit (ID)3 and the second zone in the distal part of ID1 6 hours later. Around st. 32 (Fig. 4E), the third and fourth zones were found in the proximomedian parts of ID2 and of ID1, and the fifth zone was in the distal part of ID2. At st. 33 (Fig. 4F), their domains expanded to encompass whole of the interdigital region.
During this period, the mesenchymal expression of Bmp-4 preceded the appearance of apoptosis in the proximal interdigital region. At st. 29, Bmp-4 transcripts were first detected in the proximal ID3 (Fig. 4I, asterisk), 24 hours prior to the beginning of cell death in the proximal ID3 at st. 31 (Fig. 4D). At st. 30, the Bmp-4 transcripts were detected in the proximal ID1 and ID2 (Fig. 4J, asterisks), 24 hours prior to the appearance of cell death in the same regions at st. 32 (Fig. 4E. asterisks). In the course of PCD progression, the expression of Bmp-4 in the proximomedial IDs, where apoptosis is pro gressing, was diminished. The Bmp-4 expression continued in the mesenchymal cells surrounding the phalanges that are adjacent to the dying cells (Fig. 4K-L). At this stage, the Bmp- 4 transcripts were expressed in the mesoderm (Fig. 4Q).
Bmp-2 was expressed in all of the interdigital region at st. 29 (Fig. 4N) and continued at st. 32, the expression domain became biased distally (Fig. 4O). The expression of Bmp-2 was weaker than that of Bmp-4. Thus, the expression domains of Bmp-4 and Bmp-2 precede PCD appearance and then overlap to PCD regions in the leg buds. These observations strongly support BMP-2/-4 mediation of apoptosis of the mesenchymal cells in the developing leg bud in vivo.
Factors influencing Bmp-4 expression in the limb buds
In the tooth bud, ectodermal BMP-4 activate mesenchymal Bmp-4 expression (Vanio et al., 1993). Moreover, mesenchymal expression of homeobox containing genes, Msx1 and Msx2 were affected by ectodermal BMP-4 (Vanio et al., 1993). In the interdigital regions of the leg bud, the same genes are also expressed (Ros et al., 1994) suggesting the same regulatory cascade is conserved in the limb buds.
We analyzed whether Bmp-4, Msx1 and Msx2 expression in the leg buds were affected by the infection with the dnBMPR- Ia virus. On the control side, Bmp-4 was expressed in the mesenchymal cells surrounding the phalanges in a ‘V’ shape because of the disappearance of Bmp-4 transcripts in the apoptotic regions at st. 32 (Fig. 5A). In contrast, Bmp-4 was expressed in both the mesenchymal cells surrounding the phalanges and the medial region of the interdigits on the infected side (Fig. 5B). This apparent ectopic medial Bmp-4 signal was due to continuous expression of Bmp-4 in the mesenchymal cells that was rescued from PCD pathway by the dnBMPR-Ia. The expression pattern of Msx1 and Msx2 was not affected by infection by the dnBMPR-Ia virus (data not shown). These results suggest that dnBMPR-Ia did not interrupt the autoregulatory circuit of Bmp-4 expression, while it did suppress apoptosis.
Removal of the dorsal ectoderm of the interdigital region at st. 28 induces repression of apoptosis in the underlying mesenchyme in that region (Hurle and Ganan, 1986). This implies the involvement of the ectodermal signal in the expression of Bmp-4 in the interdigital mesenchyme. To examine this possibility, we removed the dorsal ectoderm of ID3 of the leg bud at st. 28. At this stage, Bmp-4 has just begun to be expressed in ID3 mesenchyme beneath the dorsal and ventral ectoderm. At 12 hours after the operation, in the control side, the expression domain of Bmp-4 in ID3 expanded to central region (Fig. 5C,E). In contrast, in the operated side, dorsal expression of Bmp-4 disappeared (Fig. 5D,F). This result suggests that the ectodermal signal is involved in the interdigital cell death through controlling the expression of Bmp-4.
Difference in Bmp-4 expression between chick and duck limb buds
To ascertain whether Bmp-4 is responsible for the difference in the death pattern between chick leg and duck leg, we analyzed the spatiotemporal expression profile of Bmp-4 in the duck leg bud.
In the chick leg bud at st. 29, Bmp-4 was first expressed in the proximal part of ID3 (Fig. 4I). In contrast, in the duck leg bud at the equivalent limb development stage, Bmp-4 transcripts were not detected (Fig. 6D). In the chick leg bud at st. 31, Bmp-4 was expressed in all of the proximal interdigital regions in a ‘V’ shape (Fig. 4K). In the duck leg bud at the equivalent stage, Bmp-4 expression pattern was similar to that of chick except for the most proximal part of the V (Fig. 6E arrows). At this stage, cell death in the distal region of ID1 was detectable (Fig. 6A arrowhead). In the chick leg bud at st. 33, Bmp-4 was expressed in the interdigital region except the distal and medial part where apoptosis is going on (Fig. 4L). In the duck leg bud at the equivalent stage, Bmp-4 was expressed in the whole of the interdigit region except the distal part where cell death is progressing (Fig. 6B,F). At st. 35, when the distal apoptosis was completed in the duck leg (Fig. 6C), Bmp-4 continued to be expressed in all of the webs (Fig. 6G). The later expression profile is inconsistent with the correlation of Bmp-4 expression and cell death in the interdigital regions in leg bud. We will discuss this inconsistency later.
BMP-2/-4 proteins can induce apoptosis of the mesenchymal cells from the presumptive necrotic zone in vitro
To demonstrate directly that BMP-2/-4 are the apoptotic signals in the limb bud, we tested whether these proteins can induce apoptosis of the mesenchymal cells from the presumptive necrotic zone of the leg bud in vitro. We chose the mesenchymal cells derived from interdigit 3 at stage 29, which are programmed to die 24 hours later in vivo (Fig. 4C,D). Prior to this test, we established the conditions necessary to maintain these cells in vitro for 24 hours by adding 5 ng/ml of bFGF to the F-12 medium and by using collagen-coated dishes with no addition of serum (Fig. 7A). Under these conditions, we tested the effects of BMP proteins on the mesenchymal cells after 24 hours culture. Following addition of more than 10 ng/ml of BMP-2, reduction in viability was observed (Fig. 7B). After addition of more than 30 ng/ml of BMP-2, apoptotic cells were detected (Fig. 7F-H). BMP-4 also reduced the viability of the cells in the same range as BMP-2 (Fig. 7I), and this effect was attributed to apoptosis (result not shown). In contrast, other members of the TGFβ superfamily such as TGFβ1 and Activin showed no effect on cell viability (Fig. 7I).
In order to determine that dnBMPR-1a virus infected cells are not responsive to BMP, mesenchymal cells from ID3 were isolated and infected with the recombinant virus in vitro and analyzed for BMP sensitivity. The number of living cells was decreased after treatment with BMP-2 in control virus infected culture (Fig. 7J). In contrast, the number of living cells was not significantly changed in dnBMPR1a infected culture (Fig. 7J). This indicates that cells expressing dnBMPR-1a are not responsive to BMP.
These results suggest that BMP-2/-4 are the specific factors that mediate the apoptosis of the interdigital mesenchymal cells.
DISCUSSION
Expression of Bmp-2/-4 starts at a very early stage of embryonic development (Winnier et al., 1995) and these genes are also expressed and have roles in many tissue interactions at later stages of development (Kingsley, 1994a,b). As Bmp- 2/-4 play crucial roles from the beginning of embryonic development, disruption of this system causes early embryonic lethality (Mishina et al., 1995; Winnier et al., 1995). Use of local introduction of dominant negative type receptor for BMP allowed us to elucidate the function of BMP-2/-4 at later stages by mimicking a regional loss-of-function mutation.
Presence of BMP receptor type Ia and lb mRNA in the limb bud was confirmed by RT-PCR (results not shown). It was very difficult to visualize their mRNA by in situ hybridization indicating that these transcripts are not abundant in the limb bud. In contrast, transcription from the recombinant virus in the infected limb was very easily detected by in situ hybridization. In addition, we have shown IRES sequence efficiently enhances the translation of given mRNA in the infected limb mesenchymal cells (Yokouchi et al., 1995). These results indicate that the amount of truncated BMP receptor type Ia protein in infected limb mesenchymal cells could be in great excess compared to the endogenous normal type I receptor and sufficient to inactivate endogenous type II receptor forming an abortive complex as in the case of Xenopus embryos (Suzuki et al., 1994).
BMP-2 and BMP-4 are apoptotic factors in the developing limb bud
Introduction of the dominant negative BMP receptor resulted in formation of ectopic bulges both in the anterior and posterior sides of the leg bud and formation of ectopic webs in the interdigital regions. The regions in which extra tissues were formed correspond to those in which the mesenchymal tissues are eliminated by PCD in normal development. The programmed cell death in the ANZ, PNZ and INZs of the leg bud was suppressed by the dnBMPR-Ia. These results suggest that BMP- 2/-4 mediate apoptosis in the necrotic zone of the developing limb bud. This is supported by the evidence that the expression of Bmp-2/-4 and the distribution of the apoptotic cells is closely related, as discussed below (Fig. 4). Finally, we demonstrated that BMP-2 and BMP-4 proteins can induce apoptosis of the mesenchymal cells of the leg bud in vitro. These observations indicate that BMP-2 and BMP-4 are the apoptotic factors in the programmed cell death zone of developing limb bud. Recently Zou and Niswander (1996) demonstrated that over-expression of the dominant negative form of BMPR type Ib also induced suppression of PCD in limb bud. Their observations are similar to ours, however there are several differences and discussion of the differences provide more information on the function of BMPs and their receptors during PCD, as discussed below.
Mechanism of regulation of PCD in the leg bud by BMP-4
We observed the level of expression of Bmp-4 was higher than that of Bmp-2 in the INZs of the limb buds (Fig. 4I,K,N,O). In addition, BMP-4 protein can reduce the viability of the web mesenchymal cells (Fig. 7I) in the same dose range as BMP- 2 in vitro. This suggests that BMP-4 is a major apoptotic signal in the leg bud. We also found that the expression of Bmp-4 precedes apoptosis and continues in the same regions as apoptosis in the ANZ and PNZ (Fig. 4G). This result suggests that BMP-4 is necessary for both initiation and maintenance of the apoptotic state in the ANZ /PNZ in an autocrine manner. In contrast, during the apoptotic stage in the INZ, the spatiotemporal Bmp-4 expression pattern is different from the ANZ/PNZ. The expression of Bmp-4 precedes the proximomedial apoptosis in the interdigital regions (Fig. 4I,J), and was diminished in that region when the apoptotic cells appeared (Fig. 4K). These observations suggest that BMP-4 is necessary only for determination of apoptosis of the proximal INZ in an autocrine manner. BMP-4 may be involved in the maintenance of apoptosis of the proximal INZs in a paracrine manner because Bmp-4 expression continued in the regions adjacent to the proximal medial apoptosis in the INZs (Fig. 4K,L).
It was reported that PCD in ID3 was inhibited if the overlying ectoderm has been removed (Hurle and Ganan, 1986). In this work we showed that this manipulation eliminated the Bmp-4 expression in the interdigital mesenchymal cells (Fig. 5D). This result suggests that initiation or maintenance of mesodermal Bmp-4 expression in the INZs is somehow mediated by the signal(s) from the overlying ectoderm. Considering the role of BMP-4 in PCD, this observation also gives the solution for the cascade of above phenomenon.
BMP-4 and web formation
Different avian legs have a variety of web pattern which depends on the cell death pattern in the interdigital region of the leg bud. In the chick leg bud, all of the interdigital regions suffer cell death so that no web is formed. In contrast, in the duck leg bud, the first interdigital region and only the distalmost parts of the second and third interdigital regions are removed by cell death and result in the formation of the webs in interdigit 2 and 3. As shown in this study, the expression of Bmp-4 in the duck leg bud was not detected at st. 29 and began to express later than that of chick. Moreover, the proximal medial expression was not detected at st. 31. These spatiotemporal differences in Bmp-4 expression are consistent with the difference of cell death pattern between chick and duck. These results suggest that the difference in the expression pattern of Bmp-4 is essential for the formation of the interspecies differences in cell death pattern in the leg bud. In addition, curiously, Bmp-4 was expressed in the web mesenchyme during the later stages nevertheless this region does not die. Perhaps, the component of the intracellular signal transducing system changes in later stages and BMP-2/-4 signal is used for controlling other biological systems, i.e. the phalangeal chondrogenesis, rather than cell death.
Heterochrony and morphological diversification
Considering the relationship between timing of Bmp-4 expression and progression of cell death in the interdigital zone, initiation of Bmp-4 expression seems to be the rate-limiting step for PCD. BMP-4 signaling seems to be required for the phalangeal chondrogenesis because introduction of both dnBMPR-Ia (this study) and dnBMPR-Ib (Zou and Niswander, 1996) resulted in phalangeal defects. As shown in Fig. 6E-G, Bmp-4 is also expressed in the mesenchymal cells surrounding phalanges of both chick and duck leg bud indicating that BMP- 2/-4 have a role in phalangeal chondrogenesis. These indicate that the Bmp-4 gene has a cis regulatory element responsible for the phalangeal and ID expression, in both chick and duck, for proper development of the phalanges. As shown in Fig. 4I,J, Bmp-4 expression in chick ID2 and 3 begins earlier than in duck ID2 and 3 (Fig. 6D). This indicates that the cis regulatory element responsible for timing for gene expression is changed in chick and this allows expression of Bmp-4 earlier than duck. This time point would be crucial for the interdigital mesenchymal cells to choose between proliferation or differentiation. Perhaps the presence of BMP-4 at this stage opens the pathway for cell death to the interdigital mesenchymal cells in such a situation. If this is true, it is a good example of local heterochrony for gene expression being a driving force for interspecies morphological diversification (Gould, 1977).
Autoregulatory signaling pathway and death signaling pathway of BMP-2/-4
The presence of an autoregulatory loop of Bmp-4 expression in several tissues has been demonstrated and the involvement of Msx gene expression in this loop has also been shown (Vanio et al., 1993; Zou and Niswander, 1996). However, interruption of BMP-2/-4 signaling in the limb bud in this study altered neither Bmp-4 nor Msx gene expression in the mesenchymal cells surrounding the phalanges. Zou and Niswander reported that interruption of BMP-4 signaling by dnBMPR-Ib represses both Bmp-4 and Msx2 (Zou and Niswander, 1996). There are two points of difference between our experiment and theirs. The first point is that we used BMPR-Ia and they used BMPR-Ib. The second is that in our case, the truncated form of BMPR-Ia does not have any cytoplasmic domain whereas in the case of BMPR-Ib a single amino acid change has been introduced in the kinase domain of the cytoplasmic region. The second point seems to be more crucial. Not so much has been reported on the factor(s) that interact with the cytoplasmic domain of BMPR, however several such molecules have been reported for the TGFβ receptor complex, which has similar receptor structure to the BMPR complex. Association of TRIP- 1 to the TGFβ receptor is limited only when the receptor complex is formed then phosphorylated by S/T kinase in type I receptor (Chen et al., 1995). On the other hand, p21(RAS) farnesyltransferase alpha subunit associates only with type I receptor (Kawabata et al., 1995; Wang et al., 1996). In addition, it has been demonstrated that a certain domain in the cytoplasmic region is responsible for TGFβ dependent growth inhibition among many responses to TGFβ (Saitoh et al., 1996). These factors which interact with different domains of the receptor complex would be responsible for intracellular multiple signal transducing pathways. It is easily to imagine that BMPR complex also interacts with different intracellular factors responsible for individual signaling pathways. Complex formation of BMPR-II with kinase negative type Ib receptor which still has a cytoplasmic domain, blocks both the cell death signaling pathway and signaling for Bmp-4 and Msx regulatory loop. The same truncated receptor used in our study (BMPR-Ia) blocks ventral mesoderm formation when introduced in early Xenopus embryos (Suzuki et al., 1994). This effect was not alleviated by addition of BMP-4, indicating that the effect is unlikely to be due to depletion of extracellular ligand but is due to interruption of BMP-2/-4 signal by forming abortive receptor complex (Suzuki et al., 1994). The same situation would also be true in limb mesenchymal cells infected by dnBMPR-Ia virus. So it is possible that the ligand binding extracellular domain and transmembrane domain of BMPR-I are enough to interrupt the cell death signal by forming the abortive complex with BMPR-II, whereas the Bmp-4 and Msx regulatory loop is still activated by this form of BMPR complex. In summary, different results between two laboratories suggest that the intracellular signal transduction pathway for cell death and for regulating Bmp-4 and Msx expression through BMPR complex is different in the limb bud.
Death signaling pathway
In the limb bud, the apical ectodermal ridge (AER) is the main source of FGF -2, -4, -8, and functional assays have shown that these FGFs can substitute for the AER to maintain limb outgrowth (Niswander and Martin, 1993; Fallon et al., 1994). At later stages of limb development, the AER undergoes flattening, loses its activity and FGF expression decreases significantly (Niswander and Martin, 1992; Dono and Zeller, 1994). At the same time, Bmp-2/-4 expression increases in the ID mes enchyme then apoptosis occurs. In the early limb bud, BMP-4 was demonstrated to be antagonistic to FGF activity (Niswander and Martin, 1993). It is possible that the balance of FGF signal(s) and BMP signal(s) is also important for the balance of cell proliferation and differentiation in the later stages of limb development. Recently, the BMP-4 signal was reported to result in phosphorylation of MAPKK-p38 through activation of novel MAPKKK-TAK1 (Yamaguchi et al., 1995). Signal transduction to p38 also occurs in the apoptotic process in PC12 cells induced by withdrawal of the trophic factor NGF (Xia et al., 1995) whose receptor also possesses tyrosine kinase activity in the cytoplasmic domain similarly to the FGF receptor. Similar pathways may balance death or viability status in the limb bud.
ACKNOWLEGMENTS
We thank KAKEN for FGF-2, AJINOMOTO for Activin and TAKEDA for Xenopus BMP-2 and BMP-4. The hybridoma producing monoclonal antibody against P19 gag protein was obtained from the Developmental Studies Hybridoma Bank maintained by the Department of Pharmacology and Molecular Sciences, John Hopkins University School of Medicine, and the Department of Biological Sciences, University of Iowa. This study was supported by a grant- in-aid from the Ministry of Education, Science and Culture of Japan and Naito Foundation.