The type I serine/threonine kinase receptor ActRIA (ALK2) is required for gastrulation of the mouse embryo

Gastrulation of the mouse embryo involves extensive cell proliferation, differentiation and movement in the initially bilaminar embryonic egg cylinder, which leads to the formation of the interstitial mesoderm layer and the establishment of the embryonic body axes (Hogan et al., 1994; Tam and Behringer, 1997). In mice, gastrulation initiates at about 6.5 days post coitum (E6.5) with the formation of the primitive streak at the prospective posterior region. As gastrulation proceeds, the primitive streak extends distally, allowing neighboring primitive ectoderm (epiblast) cells to ingress through the streak to form the mesoderm layer. Although the formation of the primitive streak and various types of mesodermal tissues has been well characterized morphologically, the underlying cellular and molecular mechanisms are still poorly understood.

Studies of mesoderm induction in Xenopus embryos have identified several members of the TGF-β family including activins, Vg-1, BMP4 and nodal-related proteins (Xnr), as potent mesoderm-inducing factors (reviewed by Smith, 1993; Harland, 1994; Slack, 1994; Kessler and Melton, 1995; Jones et al., 1995). While BMP4 induces the ventral type of mesoderm, activin, Vg1 and Xnr (Xnr-1, Xnr-2 and Xnr-4) can induce dorsal mesoderm. Moreover, both activin and Xnr can specify different types of mesoderm in a dose-dependent manner, by induction of ventral mesoderm at low concentrations and dorsal mesoderm at high concentrations (Green and Smith, 1990; Green et al., 1992; Jones et al., 1995). Coordination of the action of different signals is probably crucial for the formation and patterning of various mesodermal tissues during gastrulation.

Genetic studies of the function of TGF-β family factors in mammalian development by gene targeting in mice have demonstrated that zygotic activins A and B are not essential for gastrulation, since embryos lacking both activin βA and βB genes develop to term with no gross defects in mesoderm formation and patterning (Matzuk et al., 1995). BMP4 and nodal, however, are essential for early mouse development (Winnier et al., 1995; Zhou et al., 1993; Conlon et al., 1994), consistent with their functions in Xenopus embryos. BMP4 is required for epiblast cell proliferation at the egg cylinder stage, and for the formation of extraembryonic mesoderm and posterior structures, whereas nodal function has been implicated in primitive streak formation (Conlon et al., 1994). TGF-β family members signal through membrane-bound heteromeric complexes of type I and type II serine/threonine kinase receptors and the downstream Smad family proteins (reviewed by Massagué, 1998). Upon binding to a ligand, the type II receptors phosphorylate and activate associated type I receptors, which in turn transduce the signal by phosphorylating pathway-specific Smad proteins such as Smad1 or Smad2. The phosphorylated Smad proteins are then translocated into the nucleus in association with Smad4, a common signal mediator of the TGF-β family signaling pathways, and activate transcription of downstream target genes. So far about a dozen of type I and type II receptors and at least ten members of the Smad family have been identified. However, the specificity of ligand-receptor-Smad cascades in mammalian development remains poorly defined.

ActRIA (also known as ALK-2, R1 or Tsk-7L) is a type I receptor that can bind either activin or BMP2/4 in conjunction with the corresponding type II receptors (He et al., 1993, Attisano et al., 1993; ten Dijke et al., 1993, 1994). ActRIA is widely expressed in midgestation embryos and in various tissues including brain, heart and lung in adult mice (Verschueren et al., 1995; He et al., 1993). ActRIA may also mediate the signaling of Mullerian inhibiting substance (MIS), since its transcripts colocalize with the type II receptors for MIS in the fetal Mullerian duct (Teixeira et al., 1996). Recent studies in Xenopus have shown that a constitutively active form of ALK2 induces the expression of ventral mesodermal markers, reminiscent of BMP2/4 activities in mesoderm formation, while expression of ALK4, an activin-specific type I receptor, is capable of transducing the activin signal to induce dorsal mesoderm (Chang et al., 1997; Armes and Smith, 1997). Moreover, overexpression of ALK2 appears to antagonize the dorsalizing effects of ALK4 (Armes and Smith, 1997). Recently, we have demonstrated that the ActRIB is required for the formation of an organized egg cylinder and the primitive streak during early mouse development (Gu et al., 1998).

To define the function of ActRIA in mammalian development, we disrupted the mouse ActRIA gene by homologous recombination. Our analysis of both embryos and ES cell lines homozygous for the mutation indicates that ActRIA functions in both embryonic and extraembryonic tissues to regulate mouse embryonic development.

Decidua or embryos were fixed in Bouin’s fixative, dehydrated and embedded in paraffin as described by Kaufman (1992). Serial sections were cut at 7 μm and stained with hematoxylin and eosin.

For detecting ActRIA transcripts, two ActRIA probes, one for the extracellular domain and the other for the 3′-untranslated region, were used in in situ hybridization analysis. The extracellular domain fragment spanning the region between nucleotides 504 and 742 (mustsk 7L, accession L 15436) was PCR amplified, inserted into pGEM-T and transcribed as antisense RNA from T7 promoter. The 3′-untranslated region between nucleotides 2034 and 2330 in a BesYI-Agel fragment was inserted into pGEM-4Z and transcribed as antisense RNA from the T7 promoter. In situ hybridization with sectioned embryos was performed basically as described by Wilkinson (1992). Slides were exposed for 18-20 days and counterstained with hematoxylin.

Whole-mount in situ hybridization with digoxigenin-labeled or fluorescein-labeled antisense RNA probes for Brachyury (T) (Wilkinson et al., 1990), Hesx-1 (Thomas and Beddington, 1996), Bmp4 (Winnier et al., 1995) and HNF4 (Duncan et al., 1994) was performed as described (Wilkinson, 1992). For double in situ hybridization, both digoxigenin- and fluorescein-labeled probes were used simultaneously. Alkaline phosphatase (AP)-conjugated anti-digoxigenin or anti-fluorescein antibody (Boehringer Mannheim) were used for enzyme reactions. After the first color reaction using NBT/BCIP, anti-digoxigenin antibody was heat-inactivated and replaced by anti-fluorescein antibody. The second color reaction was done using INT/BCIP (Boehringer Mannheim).

An antibody (875) was raised against a KLH-conjugated synthetic rat ActRIA peptide (amino acids 109-121, T-K-G-K-S-F-P-G-S-Q-N-F-C, which differs from the mouse sequence by an S- to-T substitution at position 117. This unique sequence, which differs from other members of the type I receptor family, was selected from the extracellular region near the transmembrane domain because of its favorable antigenecity profile analyzed by MAC Vector 5.0 (Campbell, CA). The antibodies were raised in chickens, and the IgY fraction purified with Eggstract (Promega, Madison, WI). The specificity of the antibody was documented by immunohistochemical localization of the antibody to the Mullerian duct of the 15 day rat embryo and colocalization with antibodies to the MIS type II receptor (Masiakos et al., unpublished data). In addition, western analysis of cell lysates from MIS responsive cells showed bands of appropriate molecular weight for the type I receptor. Embryos at E6.5, E7.0 or E7.5 were fixed with 4% paraformaldehyde inside decidua for 6 hours or overnight at 4°C, dehydrated, embedded in paraffin and cut at 7 μm thickness. Immunostaining was performed according to the protocol described by Hogan et al. (1994). The primary antibodies were used at a dilution of 1:200, while FITC-conjugated rabbit anti-chicken IgG (1:500) (from ICN) was used as the secondary antibody. The preimmune serum was tested for background staining.

Mouse genomic DNA for the ActRIA gene was obtained by screening a 129/Sv genomic library (Stratagene, Lambda FIX II) with rat ActRIA cDNA probes (He et al., 1993). The DNA fragments from four isolated clones range in size from 13 kb to 18 kb. An 11.5 kb genomic DNA was subcloned into the XhoI site of the pGEM-11z vector in which the SalI site was eliminated. A SalI linker was inserted into the KpnI site in exon 4, encoding the transmembrane domain. A PGK-neo-poly(A) cassette (the XhoI-SalI fragment from pKJ-1) was then inserted into the SalI site in the same transcriptional orientation as ActRIA to disrupt exon 4. The final targeting vector contained 7.5 kb and 4.0 kb genomic DNA 5′ and 3′ to the neo gene cassette, respectively (see Fig. 2A).

The J1 ES cells were transfected with a linearized targeting vector by electroporation and selected in G418-containing medium as described (Li et al., 1992). The genotype of G418^r clones was analyzed by Southern blot hybridization using a 0.6 kb XhoI-EcoRI genomic fragment as a probe (see Fig. 2A). One positive recombinant clone was obtained, from which chimeric mice were derived by microinjection of ES cells into blastocysts of BALB/c mice. Male chimeric mice transmitting the mutant allele through their germline were bred to either 129/Sv females to establish an inbred strain or to BALB/c to obtain heterozygous F₁ progeny.

To establish lacZ-marked ActRIA-deficient ES cell lines, the Rosa-26 mice (Friedrich and Soriano 1991) which contain a lacZ transgene, were crossed to ActRIA^tm1/+ to obtain ActRIA^tm1/+ Rosa-26^+/− offspring. Delayed blastocysts were produced by intercrosses of ActRIA^tm1/+ Rosa-26^+/− mice and ES cell lines were isolated from cultured inner cell mass cells as described (Robertson 1987). The genotypes of isolated ES cell lines for both the ActRIA mutation and the Rosa-26 transgene were determined by Southern blot analysis (Fig. 2C) and by X-gal staining (data not shown).

One type of chimeric embryos were generated by microinjection of 10-15 ActRIA^tm1/tm1 mutant ES cells into wild-type blastocysts or aggregation of mutant ES cells with wild-type 8-cell-stage uncompacted morula (Nagy and Rossant, 1993). The second type of chimeras was generated by injecting about 15 wild-type R/S4 ES cells (Gu et al., 1998) into blastocysts produced by intercrosses of ActRIA^tm1/+ heterozygous mice. Injected embryos were then transferred into the uteri of pseudopregnant foster mothers for reimplantation. Usually, the reimplanted blastocysts are delayed in development by approximately one day. Embryos were dissected at different stages, fixed and stained with X-gal as described (Hogan et al., 1994). After X-gal staining, embryos were re-fixed in Bouin’s solution or 4% paraformaldehyde, followed by dehydration, embedding and sectioning as described (Hogan et al., 1994). The ratio of the blue cells derived from injected ES cells to the non-blue cells derived from host blastocysts was estimated by examining serial sections of each embryo. The genotype of the host blastocysts was determined by Southern blot analysis of DNA isolated from the parietal yolk sac.

Wild-type and ActRIA-deficient ES cells were induced to differentiate by culturing ES cells in suspension as embryoid bodies (Robertson, 1987). For histological analysis, embryoid bodies were fixed in 4% paraformaldehyde for 3 hours at room temperature, dehydrated and embedded as described earlier. Samples were sectioned at 4 μm and stained with hematoxylin and eosin.

Previous analysis of RNA prepared from individual germ layers by reverse transcription polymerase chain reaction (RT-PCR) has indicated that ActRIA is expressed in the visceral endoderm at E6.5 and, subsequently, in both visceral endoderm and mesoderm at E7.5 (Roelen et al., 1994). In this study, we analyzed the expression pattern of ActRIA in embryos from E6.5 to E8.5 by immunofluorescent staining and in situ hybridization. We found that, at E6.5 and E7.0, ActRIA protein was confined to the apical surface of the visceral endoderm, while the expression in other cells of the embryo was similar to the background (Fig. 1A,B). At E7.5, ActRIA expression was detected in extraembryonic ectoderm, chorion, amnion and allantois, in addition to the visceral endoderm (Fig. 1C). However, the amount of ActRIA protein was very low in the embryonic regions. In E8.5 embryos, ActRIA expression was detected in the head mesenchyme and in the heart (Fig. 1D,F). We also performed in situ hybridization and found that ActRIA transcripts were expressed at very low levels before gastrulation (data not shown). At E7.5, ActRIA transcripts were detected in the visceral endoderm, chorion, amnion, allantois, and embryonic ectoderm (data not shown). By E8.5, ActRIA transcripts were detected in the head mesenchyme, endocardium and extraembryonic tissues such as allantois (Fig. 1E,G), consistent with the immunostaining results.

To mutate the ActRIA gene in mouse embryonic stem (ES) cells, a targeting vector was constructed in which a neomycin-resistance gene cassette was inserted into the fourth exon encoding the transmembrane domain (Fig. 2A). The linearized targeting vector was transfected into ES cells by electroporation. Of 400 G418-resistant colonies screened by Southern blot analysis using a 3′ external probe, one clone showed correct homologous recombination between the vector DNA and the endogenous ActRIA locus (Fig. 2B). The mutant allele was termed ActRIA^tm1(for disruption of the trans-membrane domain). Based on the genomic structure of the gene (Schmitt et al., 1995), it was predicted that any alternative splicing skipping the disrupted exon would cause a frame shift in the kinase domain, thus the inactivation of the receptor. In addition, levels of ActRIA transcripts were expected to be significantly reduced, since the neo gene cassette was inserted in the same transcriptional orientation as the ActRIA gene. While we cannot exclude the possibility that truncated ActRIA might be produced, ActRIA^tm1probably represents a severe loss-of-function mutation.

Mice heterozygous for the mutation (ActRIA^tm1/+) were morphologically normal and fertile. Genotyping of newborn progeny from heterozygous crosses by Southern blot analysis showed that 32% of them were wild type, 68% heterozygous and none were homozygotes (ActRIA^tm1/tm1) (Fig. 2B; Table 1), indicating that the disruption of ActRIA is recessive embryonic lethal. In addition, ActRIA^tm1/tm1 mice were never recovered from continuous breeding of heterozygous parents for more than three years, indicating that the mutant phenotype co-segregates with and is therefore resulted from the ActRIA mutation.

To determine when mutant embryos die in utero, embryos from ActRIA^tm1/+intercrosses were dissected and analyzed at different days of gestation (Table 1). At E7.5, 29 out of 38 embryos were normal and nine were abnormal. Genotyping by PCR revealed that the normal embryos were either wild type or heterozygous for ActRIA mutation, while all the abnormal embryos were ActRIA^tm1/tm1mutants. The ActRIA^tm1/tm1embryos were much smaller than their normal littermates and lacked a morphologically discernible primitive streak (Fig. 3A). At E8.5, the mutant embryos were growth retarded and disorganized (Fig. 3B). At E9.5, about one quarter of decidua contained resorption sites and no homozygous mutants were recovered. These results suggest that the development of ActRIA-deficient embryos was arrested prior to or during gastrulation.

To determine the defects in ActRIA^tm1/tm1 embryos, histological analysis of normal and mutant embryos was performed. No apparent abnormality was observed in 27 embryos analyzed at the egg cylinder stage (E6.0-6.5) (Table 1), suggesting that the mutant embryos developed normally before the initiation of gastrulation. At E7.0-7.5, about 23% (11 out of 47) of embryos showed various morphological defects and were significantly smaller than their normal littermates. In the normal embryos, the primitive streak, the mesodermal layer and extraembryonic structures such as amnion, chorion and allantois are well formed (Fig. 4A,D). In the mutant embryos, however, a morphologically distinguishable primitive streak or the mesodermal layer was absent (Fig. 4B,E). The epiblast in some regions of the embryonic cylinder lost its usual columnar shape and was thickened (Fig. 4B). A constriction near the embryonic/extraembryonic junction could be observed in most mutant embryos (Fig. 4B,C). In the extraembryonic region, the extraembryonic ectoderm failed to form an organized epithelial layer, and no allantois, chorion, amnion or amniotic folds were observed (Fig. 4B,C). In addition, the proximal visceral endoderm cells were less vacuolated and abnormal as compared with the characteristic columnar shape of the visceral endoderm in normal embryos. The parietal endoderm layer in mutant embryos appeared to be normal except that the cell density was higher than that in normal embryos. Analysis of mutant embryos at E8.0-8.5 revealed that some mesoderm-like cells were formed but disorganized (Fig. 4C,F). The embryonic ectoderm was folded, resulting in disorganization of the egg cylinder. In most E8.5 mutant embryos, a large number of dead cells were found inside the proamniotic cavity, indicating tissue degeneration (Fig. 4C).

To further characterize the gastrulation defects, expression of several early markers was analyzed in ActRIA-deficient embryos. Brachyury (T) is one of the earliest mesoderm markers. In normal embryos, T expression is first detected at E6.0 in the proximal rim of the epiblast before the initiation of gastrulation, and then in the primitive streak formed in the prospective posterior region of the epiblast (E6.5-7.0), and later in the head process and the notochord as well as in the primitive streak (after E7.5) (Wilkinson et al., 1990; Fig. 5A,B). However, T expression in E7.5 embryos with characteristic ActRIA^tm1/tm1 mutant morphology was either absent (2 out of 9) or greatly reduced (7 out of 9) (Fig. 5A).

When T expression was detected in the mutant embryos, the T transcripts were confined to a small area near the embryonic/extraembryonic junction. T expression in mutant embryos at E8.5 showed a similar pattern (Fig. 5B). The restricted T expression in mutant embryos suggests that the initiation of gastrulation and primitive streak formation probably occurred, but gastrulation failed to progress further.

Since the anteroposterior (A-P) axis is established at the egg-cylinder stage prior to the formation of the primitive streak, it is important to test whether the A-P axis is established in the ActRIA^tm1/tm1 mutant embryos. We examined the expression pattern of T and an anterior marker Hesx1 simultaneously by double in situ hybridization. Hesx1 is first detected in the anterior visceral endoderm (AVE) in the wild-type embryos at the pre- and early streak stages, and then in the adjacent prospective neuroectoderm after the primitive streak is formed (Thomas and Beddington, 1996) (Fig. 5B). In the mutant embryos, Hesx1 transcripts were detected on the opposite side of the embryo where T expression was detected, indicating that the anteroposterior axis was established in the mutant embryos (Fig. 5B).

HNF4 is a transcription factor that is expressed specifically in the visceral endoderm during early embryogenesis (Duncan et al., 1994). It has been shown that HNF4 is required for the differentiation and function of the visceral endoderm during early development. Disruption of the HNF4 gene results in reduction or elimination of expression of multiple genes in the visceral endoderm, and developmental arrest of the mutant embryos at the early gastrulation stage (Chen et al., 1994; Duncan et al., 1997). The morphological defects of ActRIA^tm1/tm1mutant embryos were similar to, though appeared to be more severe than, those of HNF4-deficient embryos. HNF-4 transcripts were first detected in the primitive endoderm at the blastocyst stage and then in the visceral endoderm in the wild-type embryos (Duncan et al., 1994, Fig. 5C). In contrast, HNF-4 expression in the visceral endoderm was drastically reduced in E7.5 and E8.5 ActRIA^tm1/tm1embryos (Fig. 5C), suggesting that visceral endoderm differentiation might be impaired.

We also analyzed the expression pattern of Bmp4 in the ActRIA^tm1/tm1mutant embryos. In normal embryos, Bmp4 is initially expressed around the distal part of the extraembryonic ectoderm before gastrulation and later in the posterior region of the embryo as well as in the extraembryonic tissues such as allantois and amnion (Jones et al., 1991; Winnier et al., 1994; Fig. 5D). In the mutant embryos, Bmp4 transcripts were detected in the extraembryonic ectoderm near the epiblast, suggesting that the early Bmp4 expression pattern was not affected in ActRIA^tm1/tm1embryos (Fig. 5D). The BMP4 expression pattern was also consistent with the developmental arrest of ActRIA^tm1/tm1 embryos at the early gastrulation stage.

ActRIA-deficient ES cells were genetically marked with a lacZ transgene inherited from Rosa 26 mice (Friedrich and Soriano, 1991) and used to test whether ActRIA mutation could affect cell proliferation and differentiation in a cell autonomous manner. Since the Rosa 26 lacZ transgene is expressed ubiquitously during early embryogenesis (Gu et al., 1998), it can serve as a cell lineage marker to trace tissue distribution of the mutant cells in developing embryos. The lacZ⁺ ES cell lines were established from inner cell mass cells of delayed blastocysts produced by intercrosses of ActRIA^tm1/+Rosa-26^+/-mice. Of ten lines established, two were confirmed to be lacZ⁺ActRIA^tm1/tm1 (Fig. 2C). Both cell lines were used to generate chimeric embryos and similar results were obtained.

We generated chimeric embryos either by injection of 10-15 ActRIA^tm1/tm1 ES cells into each wild-type blastocyst or by aggregation of ActRIA^tm1/tm1 ES cells with wild-type embryos. Chimeric embryos were dissected at embryonic stages from E8.5 to E11.5 and stained with X-gal. Comparing with non-injected littermates, most chimeras with low or medium contribution of ActRIA^tm1/tm1 ES were morphologically normal. Although about two thirds of chimeric embryos with strong X-gal staining appeared to be morphologically normal from E8.5 to E10.5 (Fig. 6A,B), the remaining one third showed various abnormalities, including smaller body size, delay in development (e.g. not turned at E9.5), abnormal head structures, and shortening of the posterior and trunk regions (Fig. 6C). Most chimeras with high contribution of ES cells were growth retarded and died at E11.5 (data not shown).

Previous studies have shown that ES cells injected into blastocysts contribute predominantly to the embryonic lineage (Beddington and Robertson, 1989). As expected, the ActRIA^tm1/tm1 cells contributed extensively to almost all tissues in chimeric embryos, but were absent in the extraembryonic ectoderm-derived tissues, the parietal yolk sac and the endoderm component of the visceral yolk sac. Histological analysis of several most extensively stained E8.5 chimeric embryos revealed that ActRIA^tm1/tm1 cells contributed to embryonic tissues of all three definitive germ layers, including mesoderm-derived tissues such as the cephalic mesenchyme, heart, somites and the mesoderm component of the visceral yolk sac (Fig. 6D-F). By counting the number of mutant and wild-type cells in serial sections of several normal E8.5 chimeric embryos, we estimated that mutant cells accounted for about 80% of the cells in chimeric embryos. However, most abnormal chimeras at E9.5 and E10.5 had a higher contribution of mutant cells (data not shown). These results indicate that a very extensive contribution of ActRIA^tm1/tm1cells in chimeric embryos could disrupt normal development (Fig. 6C), whereas the presence of sufficient number of wild-type cells could rescue such defects.

The detection of ActRIA protein in the extraembryonic cells prior to and during gastrulation suggests that the function of ActRIA in those cells is probably essential for gastrulation. This hypothesis is supported by the observation that wild-type extraembryonic tissues could support chimeras with extensive contribution of ActRIA^tm1/tm1cells to develop normally at the early gastrulation stage. To further test this hypothesis, a second type of chimeric embryos were generated by injecting wild-type R/S4 ES cells (Gu et al., 1998), containing the Rosa 26 lacZ transgene, into blastocysts produced by mating male and female ActRIA^tm1/+mice. If the ActRIA function is required in the extraembryonic cells, chimeras deriving from ActRIA^tm1/tm1blastocysts would be expected to show gastrulation defects similar to those seen in ActRIA^tm1/tm1mutant embryos.

Chimeras in which embryonic tissues consisting predominantly of wild-type cells were generated by injection of 15 wild-type R/S4 ES cells into each blastocyst (Gu et al., 1998). Of 24 chimeric embryos dissected at E9.5, 19 were morphologically normal, one was very small with disorganized embryonic structure and four were partially or completely degenerated. Of 30 chimeras dissected at E8.5, eight were growth retarded. Southern analysis of DNA isolated from the parietal yolk sac, which was derived entirely from the donor blastocysts, showed that the abnormal E8.5 and E9.5 chimeras were derived from ActRIA^tm1/tm1 blastocysts, and all 41 E8.5 and E9.5 normal chimeras were derived from either wild-type or heterozygous blastocysts. X-gal staining demonstrated a very extensive contribution of lacZ⁺ wild-type cells in all of the growth-retarded chimeras (Fig. 7B).

Histological analysis of the abnormal chimeras at E8.5 revealed that the embryonic development was much delayed and impaired even though the lacZ⁺ wild-type cells colonized almost the entire embryonic ectoderm (Fig. 7C,D). Some abnormal chimeras (n=2) showed similar defects as seen in ActRIA^tm1/tm1 embryos, including abnormal morphology of the visceral endoderm, thickening of the epiblast and constriction at the embryonic/extraembryonic junction (Fig. 7C), suggesting that mesoderm formation was disrupted or delayed. Other chimeras (n=6) developed further to the stage equivalent to E7.5 wild-type embryos and formed mesodermal tissues, chorion and amnion-like structures (Fig. 7D). However, the formation of the extraembryonic mesoderm was delayed. In particular, the visceral yolk sac failed to expand and the allantois was not formed (Fig. 7D). Together, these results indicate that the early gastrulation defect of ActRIA^tm1/tm1 embryos results from the loss of function of ActRIA in the extraembryonic cells. The development of some chimeras (6 out of 8) to a more advanced stage than ActRIA^tm1/tm1 embryos suggests that wild-type ES cells could partially rescue the early gastrulation defects.

When cultured in suspension, ES cells form aggregates known as embryoid bodies and undergo differentiation (Robertson, 1987). Since cellular differentiation and gene expression in embryoid bodies seem to parallel the developmental process of early development, ES cells have been used as an in vitro model for studying early developmental events (Coucouvanis and Martin, 1995; Duncan et al., 1997; Sirard et al., 1998). We also studied the differentiation of ActRIA^tm1/tm1ES cells in vitro. During the first 7-8 days of culture, both wild-type and homozygous mutant cells formed embryoid bodies with a well organized inner layer of columnar ectoderm-like cells surrounded by an outer layer of vacuolated endoderm cells, resembling the columnar visceral endoderm (Fig. 8A-C). There was no discernible difference between the wild-type and mutant embryoid bodies up to this stage. During further differentiation in culture, many of the wild-type embryoid bodies grew larger and formed beating cardiac muscle (data not shown) and large cysts equivalent to the visceral yolk sac (Fig. 8D,F). However, the ActRIA^tm1/tm1embryoid bodies were much smaller in size and formed neither beating muscle cells nor cystic structures (Fig. 8E,G). Histological analysis of wild-type embryoid bodies at day 17 revealed the presence of mesoderm-derived mesenchyme cells between the endoderm and ectoderm layers (Fig. 8D). In contrast, only two layers of cells, the primitive endoderm and the ectoderm, separated by an abnormally thick layer of basement membrane, were found in the ActRIA^tm1/tm1embryoid bodies (Fig. 8E). These results suggest that mesoderm formation in the mutant embryoid bodies was disrupted.

Members of the TGF-β superfamily have been shown to play important roles during early development of vertebrate animals. Genetic studies by gene targeting in mice have demonstrated that nodal, BMP4 (Zhou et al., 1993; Conlon et al., 1994; Winnier et al., 1995), two TGF-β family type I receptors, Bmpr (ALK3) and ActRIB (ALK4) (Mishina et al., 1995; Gu et al., 1998), and two Smad proteins, Smad2 and Smad4 (Nomura and Li, 1998; Waldrip et al., 1998; Weinstein et al., 1998; Sirard et al., 1998; Yang et al., 1998), are essential for early mouse development. In this study, we demonstrate that yet another type I receptor ActRIA (ALK2) is also required during early development. The apparent non-redundant function of the three type I receptors indicates the complexity of TGF-β signaling in regulation of morphogenesis during early mouse development.

The development of ActRIA^tm1/tm1embryos appeared to be normal before gastrulation, but was arrested shortly after the initiation of gastrulation. The ActRIA^tm1/tm1mutant embryos displayed multiple defects including constriction at the embryonic-extraembryonic junction, thickening of the embryonic ectoderm, delayed and severely impaired mesoderm formation, and the lack of extraembryonic structures such as amniotic fold or allantois. In addition, the morphology of the proximal visceral endoderm cells was abnormal in most ActRIA^tm1/tm1embryos.

The defects in ActRIA^tm1/tm1embryos are distinctive and less severe as compared to mouse embryos lacking either Bmpr or ActRIB receptors. In Bmpr^−/−or ActRIB^−/− embryos, the formation of the primitive streak and mesodermal cells is completely blocked. While Bmpr probably functions as the type I receptor for BMP4 to regulate epiblast cell proliferation and mesoderm formation (Mishina et al. 1995; Winnier et al., 1995), ActRIB appears to mediate the nodal or an activin-like signal that is essential for egg cylinder organization and primitive streak formation (Gu et al., 1998). In contrast, the formation and growth of the ActRIA^tm1/tm1egg cylinder were essentially normal, and the A-P axis was established as shown by Hesx1 and T expression on the opposite sides of the mutant embryos (Fig. 5). The expression of T and the presence of mesodermal cells suggest that the primitive streak was initially formed in most ActRIA^tm1/tm1embryos. Our results thus indicate that the function of ActRIA is not required for the egg cylinder formation and growth nor for the establishment of A-P axis, but becomes essential shortly after the initiation of gastrulation.

In mice, fate mapping studies have shown that the earliest mesoderm forms in the posterior region of the primitive streak and migrates proximally into the extraembryonic region to form yolk sac mesoderm, while the embryonic mesoderm is formed by epiblast cells ingressing through the primitive streak and migrating laterally (Lawson et al., 1991; Tam and Behringer, 1997). Disruption of primitive streak formation would presumably block all mesoderm formation. Since T expression was restricted to the posterior region in the ActRIA^tm1/tm1embryo, it suggested that the primitive streak was initially formed, but failed to elongate subsequently. Previous morphogenetic studies have shown that the primitive streak elongates through intercalation of newly recruited epiblast cells to regions proximal to the anterior aspect of the primitive streak (Lawson et al., 1991). Cell-cell interactions between the visceral endoderm and the epiblast may play a role in this process. The morphological defects in ActRIA^tm1/tm1embryos, such as thickening of the epiblast epithelial layer and impaired mesoderm formation, therefore, may indicate a failure of the epiblast cells to intercalate into the early primitive streak. Alternatively, the epiblast cells may fail to ingress through the streak to form mesoderm, or the nascent mesoderm may fail to proliferate or migrate.

To understand the primary defect that causes developmental arrest of the ActRIA^tm1/tm1 embryos at the early gastrulation stage, it is critical to determine in which germ layer ActRIA carries out its function. Through chimera analysis, we found that mutant ActRIA^tm1/tm1 ES cells could contribute extensively to most embryonic tissues during gastrulation when the extraembryonic tissues were derived from wild-type blastocysts. The contribution of ActRIA^tm1/tm1 ES cells to various types of mesoderm tissues in chimeras (Fig. 6) indicates that ActRIA-mediated signaling is indirectly involved in mesoderm formation. In contrast, when the extraembryonic tissues were derived from ActRIA^tm1/tm1 blastocysts, the chimeric embryos were arrested at the early to mid-gastrulation stages even when the embryonic ectoderm was colonized extensively by the wild-type cells. It should be noted, however, that most chimeras developed to a slightly more advanced stage than ActRIA^tm1/tm1 embryos as shown by the presence of well-organized embryonic mesoderm and the formation of chorion- and amnion-like structures (Fig. 8D). This result suggests that ActRIA may also function in the epiblast during gastrulation. Together, our results indicate that ActRIA functions primarily, but not exclusively, in the extraembryonic cells to regulate early gastrulation.

How the ActRIA-mediated signaling pathway regulates gastrulation at the molecular and cellular level remains essentially unknown. We speculate the following possibilities. First, the ActRIA signaling pathway may regulate expression of genes that are required for visceral endoderm differentiation and function. We showed that expression of a visceral-endoderm-specific gene, HNF4, was dramatically reduced in the ActRIA^tm1/tm1 embryos. Similar to the HNF4^−/− mutants, ActRIA^tm1/tm1 embryos are also arrested at the early streak stage with reduced T expression and delayed mesoderm formation, though the defects in ActRIA^tm1/tm1 embryos appear to be more severe. It would be interesting to investigate whether HNF4 is a direct downstream target gene of the ActRIA signaling pathway and what other genes are regulated by ActRIA in the visceral endoderm. Secondly, ActRIA may regulate expression of extracellular matrix proteins or their receptors in the visceral endoderm, which may play an important role in mediating interactions between the visceral endoderm and the overlaying epiblast during gastrulation. Thirdly, the ActRIA-mediated signal may induce the synthesis and/or secretion of a secondary signal from the extraembryonic cells, which can in turn act upon the epiblast or newly formed mesoderm to promote gastrulation.

Several other key TGF-β family signal mediators including ActRIB, Smad2 and Smad4, have been shown to have functions in the extraembryonic cells during early mouse development. While ActRIB and Smad2 function in both epiblast and extraembryonic cells, their function in extraembryonic cells is required for the formation of the embryonic rather than the extraembryonic mesoderm (Gu et al., 1998; Nomura and Li, 1998; Waldrip et al., 1998). Similar to ActRIA, the Smad4 function was shown to be required for the differentiation and function of the visceral endoderm, and inactivation of Smad4 in the extraembryonic cells blocked embryonic growth and mesoderm formation (Sirard et al., 1998; Yang et al., 1998). It is possible that ActRIA and Smad4 may function in the same signaling pathway during early mouse development.

Which TGF-β family molecule signals through ActRIA? Since ActRIA is expressed on the apical surface of the visceral endoderm before and during gastrulation, it seems possible that ActRIA may mediate signals coming from the maternal decidual cells. It has been shown that activins (both βA and βB) and BMP4 are expressed in the decidual cells surrounding the embryo at E6.5 and E7.5 (Albano et al., 1994; Feijen et al., 1994; Jones et al., 1991), therefore, can presumably diffuse into the yolk sac cavity to activate ActRIA in the visceral endoderm. The expression of ActRIA also overlaps extensively with that of BMP4 in the amnion and allantois at E7.5 (Fig. 1; Jones et al., 1991). Remarkably, chimeras derived from ActRIA^tm1/tm1blastocysts had a reduced amount of the extraembryonic mesoderm (Fig. 8C,D) and chimeras derived from wild-type embryos with extensive contribution of ActRIA^tm1/tm1ES cells showed late gastrulation defects (Fig. 6C), reminiscent of BMP4-deficient embryos which could develop to the gastrulation stage (Winnier et al., 1995). Whether ActRIA mediates signaling of BMP4 or other BMPs during mouse development requires further investigation.

The extraembryonic and posterior mesoderm in the mouse embryo is considered to be equivalent to the ‘ventral’ type of mesoderm. In this regard, it appears that ActRIA plays an important role in ventral type mesoderm formation. In contrast, it has been shown that the function of ActRIB and Smad2 is essential for the formation of the primitive streak and the embryonic (or ‘dorsal’ type) mesoderm (Gu et al., 1998; Nomura and Li, 1998; Waldrip et al., 1998; Weinstein et al., 1998). Although it is difficult to directly compare the mouse mutant phenotypes resulting from loss-of-function mutations with those of Xenopus embryos overexpressing the receptors (Armes and Smith, 1997; Chang et al., 1997), the opposite effects of ActRIA and ActRIB mutations on the formation of ventral versus dorsal mesoderm in mice and frog indicate that the function of ActRIA and ActRIB in gastrulation is likely to be conserved in vertebrates.

We thank Drs R. Beddington, J. Darnell, B. Herrmann and B. Hogan for providing in situ probes, Y. Mishina and R. Behringer for communicating results before publication, and B. Simpson for help with initial analysis of the mutant phenotype. We are grateful to Drs M. Nomura, C. Mummery and K. Defize for critical comments on the manuscript, and members of our laboratories for discussion. This work was supported by grants from Bristol-Myers/Squibb and NIH GM52106 to E. L., NIH HD1332 to P. K. D, National Research Service Award to Z. G., Claude E. Walsh Research fellowship to E. M. R. and a Wellcome International Traveling Research Fellowship to J. S.