Mouse embryos lacking Smad1 signals display defects in extra-embryonic tissues and germ cell formation

The earliest differentiation events during mammalian development reflect the necessity of the embryo to survive in the uterine environment and involve the formation of extra-embryonic tissues, which have a crucial role in supporting development but do not contribute to the resulting animal. At early post-implantation stages, the mouse embryo comprises three cell lineages: the extra-embryonic ectoderm, which gives rise to the placenta and the ectodermal component of the chorion, the primitive endoderm, which forms the endoderm of the parietal and visceral yolk sacs, and the epiblast, which gives rise to the embryo proper. The process of gastrulation converts the epiblast into the three definitive germ layers of the embryo, as well as generating the extra-embryonic mesoderm that participates in the formation of the visceral yolk sac (VYS), amnion, chorion and allantois. Thus, the development of the embryo within the uterine environment requires the coordinate growth and morphogenesis of both the embryonic and extra-embryonic tissues.

Inductive interactions between the tissue layers of the early embryo govern formation of the early body plan and control numerous processes, including cell proliferation, differentiation and migration (Beddington and Robertson, 1999). Members of the transforming growth factor β (TGFβ) family of growth factors regulate many of these aspects of post-implantation development (Goumans and Mummery, 2000). TGFβ ligands can be divided into two groups in accordance with biological and structural criteria: BMP/GDF and TGFβ/activin/nodal (Kingsley, 1994). Experiments in the mouse have underscored the diverse roles these classes of molecules play during development. Chimera analysis and loss-of-function studies have shown that nodal signaling is important in guiding the formation of both the anteroposterior (AP) and left-right axis (Schier and Shen, 2000). Loss of BMP4 function results in gastrulation defects, with many embryos failing to form sufficient mesoderm (Winnier et al., 1995). Those embryos surviving beyond gastrulation display truncated and disorganized posterior structures, including extra-embryonic mesoderm derivatives and entirely lack primordial germ cells (PGCs) (Lawson et al., 1999; Winnier et al., 1995). Similarly, loss of BMP2 function also disturbs early developmental processes and causes defective development of the amnion, chorion, and heart, implicating BMP2 in the formation of both extra-embryonic and embryonic mesodermal derivatives (Zhang and Bradley, 1996).

TGFβ ligands signal via heteromeric complexes composed of type I and type II cell-surface serine/threonine kinase receptors (Massague, 1998). In the ligand-induced complex, the type II receptors phosphorylate and activate type I receptors. Activated type I receptors in turn phosphorylate Smad family members, which constitute the only known substrates of the type I receptors (Massague, 1998). To date, nine mammalian Smads have been identified, which fall into three classes based on biological and structural criteria: pathway-specific Smads, common Smads and inhibitory Smads (Whitman, 1998). Thus, in response to TGFβ-mediated signaling the pathway-specific Smads, including Smad1, 2, 3, 5 and 8, are selectively phosphorylated at their C termini by specific type I receptors. Once phosphorylated, the pathway-specific Smads complex with the common Smad, Smad4, and translocate into the nucleus where, via interaction with DNA-binding proteins such as OAZ and FAST, they direct specific transcriptional responses (Wotton et al., 1999). Considerable biochemical and embryological evidence has shown that the pathway-specific Smads act as the intracellular mediators for distinct classes of ligands. Smad2 and Smad3 are phosphorylated by the TGFβ type I and activin type IB receptors (Macias-Silva et al., 1996; Zhang et al., 1996), whereas Smad1, Smad5 and Smad8 are activated by BMP type I receptors (Hoodless et al., 1996; Kawai et al., 2000; Nishimura et al., 1998). Moreover in Xenopus animal cap assays, Smad2 induces dorsal mesoderm in a manner similar to overexpression of activin and nodal (Graff et al., 1996; Suzuki et al., 1997), whereas Smad1 and Smad5 causes formation of ventral mesoderm, a response evoked by Bmp2 and Bmp4 (Baker and Harland, 1996; Graff et al., 1996).

Given the high degree of homology shared by members of the Smad family, it has been of interest to understand the extent to which they function interchangeably in vivo. The function of Smad proteins has been addressed in the mouse using gene targeting approaches and has shown that highly related Smad proteins can have diverse roles in the normal animal. Thus, Smad2 plays a central role in AP axis formation and Smad2-deficient embryos are abnormal at gastrulation (Heyer et al., 1999; Nomura and Li, 1998; Waldrip et al., 1998; Weinstein et al., 1998), ultimately forming only extra-embryonic mesoderm derivatives. By contrast, despite sharing 92% amino acid identity with Smad2, Smad3-deficient mice show only impaired immune functions and a predisposition to colon cancer (Yang et al., 1999b; Zhu et al., 1998). As yet, it is unclear whether the differences in phenotype are due to distinct responses mediated by either Smad or whether they are reflective of their unique expression patterns.

To date, the role of only a single BMP-specific Smad, Smad5, has been studied in the mouse. Smad5 mutants die by 10.5 dpc and display multiple embryonic and extra-embryonic phenotypes, including gut, cardiac and laterality defects, as well as increased mesenchymal apoptosis, and lack of forebrain tissue (Chang et al., 1999; Chang et al., 2000; Yang et al., 1999b). These defects are consistent with a generalized abrogation of BMP signaling; however, the relatively late onset of these phenotypes, in comparison with those observed for Bmp2 and Bmp4 mutations, suggests that other BMP-regulated Smads such as Smad1 or Smad8 must contribute in transducing BMP signals during early embryogenesis.

In the present study, we have explored the role of Smad1, a second member of the BMP signaling pathway, in the mouse. Similar to Smad5 mutants, Smad1^−/− embryos die at 10.5 dpc. However the phenotypes of the two mutations are largely non-overlapping. Thus, unlike Smad5^−/− embryos, loss of Smad1 does not prevent the formation and patterning of the embryo proper. Rather, from early post-gastrulation stages onwards, Smad1 mutants exhibit pronounced defects in the morphogenesis and proliferation of extra-embryonic tissues, which collectively result in embryonic lethality. These defects include overgrowth of the posterior visceral endoderm (VE), as well as the extra-embryonic ectoderm and mesoderm of the chorion. Surprisingly, loss of Smad1 has the opposite growth effect on the allantois, leading to a dramatic reduction in the size and patterning of this tissue and concomitant failure to form the umbilical connection to the placenta. In addition Smad1-deficient embryos display a marked reduction in the number of primordial germ cells (PGCs) at 8.5 dpc. Thus, our data demonstrate that Smad1 pathways play a pivotal role in controlling the proliferation and morphogenesis of the extra-embryonic tissues during early mammalian development.

A 129/SvJ mouse genomic library (Stratagene) was screened with a full-length Smad1 cDNA clone and three independent λ clones containing the first coding exon were identified. To generate the Smad1^Robm1 construct, a PGK-neo selection cassette was inserted between a 2.8 kb PstI/BamHI 5′ sub-fragment and a 5.1 kb 3′HindIII sub-fragment. Fifteen micrograms of ClaI linearized vector was electroporated into 2×10⁷ CCE ES cells (Robertson et al., 1986). DNA from drug-resistant clones was digested with EcoRI and screened by Southern blot hybridization using two external probes: a 0.8 kb SalI/PstI 5′ probe and a 0.6 kb HindIII/EcoRI 3′ probe. Four of 600 G418-resistant clones were correctly targeted, one of which was injected into C57Bl/6J blastocysts. Germline chimeras were crossed to MF1 females (Harlan) and the resulting heterozygous offspring intercrossed or further bred to MF1 females.

The Smad1^RobPC allele was generated by inserting a loxP and EcoRI site into the StuI site of the 4.5 kb EcoRI/BamHI 5′ arm. The 3′ arm was constructed by inserting a loxP-flanked PGK-hygro cassette into the MluI site of the 5.8 kb BamHI/HindIII fragment. A 3′ MCI-tk cassette was added. Drug-resistant ES-cell clones were screened by Southern blot using a 3′ external probe (as above) and an internal hygro probe. Three of 700 clones were correctly targeted and two were used to produce germline chimeras. Heterozygous Smad1^RobPC mice were mated to mice carrying a Protamine-Cre transgene (O’Gorman et al., 1997). Heterozygous males carrying the transgene were mated to MF1 females. Because Cre-mediated recombination leads to three potential configurations of the Smad1^RobPC locus, the progeny were screened by Southern blot analysis. EcoRI- and BamHI-digested tail DNA samples were probed with the 3′ external probe, hygro probe and 5′ internal probe to identify animals carrying the Smad1^Robcn allele.

For both Smad1 alleles, embryos from 9.5 dpc onwards were genotyped by Southern blot analysis of yolk-sac DNA. Younger embryos derived from Smad1^Robm1 heterozygous intercrosses were genotyped by PCR with the Smad1-specific primers 5′- GTCGAACAGTGTACGCCAAGC-3′ and 5′-CCTGTTTCCACCC-AAGGAGTC-3′, and with neo-specific primers 5′-CGTCTGATGCC-GCCGTGTTCC-3′ and 5′-CTTCGCCCAATAGCAGCCAGTCC-3′.

Total RNA was extracted from individual 9.5 dpc embryos using the TRIzol method (Gibco/BRL). Fifteen micrograms of total RNA were size fractionated on a 1.3% agarose-formaldehyde gel, transferred onto GeneScreen Plus (NEN Research Products), and probed with a ³²P random-primed 700 bp EcoRI/HindIII fragment, comprised mostly of Smad1 3′ UTR. For Western blot, embryos or COS cells, were lysed in sample buffer (0.0625 M Tris-HCL, pH 6.8, 2.0% SDS, 5.0% 2-mercaptoethanol, 0.002% Bromophenol Blue, 10.0% glycerol) and sonicated. Lysates (20 μg) were fractionated by SDS-PAGE using 12% separating gels, transferred onto nitrocellulose and membranes incubated overnight with an anti-Smad1 antibody (Upstate, 1:500) that recognized amino acids 147-258 of human Smad1. The secondary antibody was a bioatinylated goat anti-rabbit IgG (ICN, 1:150,000). Finally, filters were incubated with streptavidin/HRP conjugate (Phar Mingen, 1:1000) and HRP detected using an ECL kit (Amersham, IL).

Embryos were fixed in 4% paraformaldehyde (PFA) in phosphate-buffered saline (PBS) at 4°C overnight, followed by dehydration through a graded ethanol series. Embryos were cleared in xylene, embedded in Fibrowax (BDH Laboratory Supplies, England) and sectioned at 8μm. Sections were collected on Tespa-treated glass slides and stained with Hematoxylin and eosin using standard procedures.

Whole-mount in situ hybridization was performed as described (Wilkinson, 1992). Antisense probes specific for AFP (Waldrip et al., 1998), Sox2 (Conlon et al., 1994), T (Wilkinson et al., 1990), eomesodermin (Russ et al., 2000), Bmp4 (Jones et al., 1991), Fgf8 (Crossley and Martin, 1995) and Fgf4 (Wilkinson et al., 1988) were used as described. Embryos were post-fixed in 4% PFA, photographed, dehydrated in ethanol, cleared in xylenes, embedded in paraffin and sectioned at 8 μm. Sections were dewaxed using standard procedures, mounted in Cytoseal and photographed using Normarski optics.

Radioactive in situ hybridization was performed as described (Jones et al., 1991). The probes used include a full-length Smad1 (Waldrip et al., 1998) and the proline-rich linker domain of Smad5 (Meersseman et al., 1997). The Smad8 probe is a 437 bp clone that corresponds to the linker region (kindly provided by G. Chu). Sections were photographed using a Leitz DMR microscope and Fugichrome Velvia color slide film.

Primordial germ cells were visualized and counted according to Lawson et al. (Lawson et al., 1999). Genotype was assessed by PCR analysis of yolk-sac DNA.

Blastocysts, recovered from Smad1^Robm1 intercrosses were injected with 10-15 wild-type KT4 ES cells (Tremblay et al., 2000). After transfer into pseudopregnant foster females, the manipulated embryos were recovered at 7.5 dpc, fixed and stained for β-galactosidase activity.

Two independent targeting constructs were designed to mutate the first coding exon of Smad1. One vector replaces most of this exon and 30 bp of 3′ flanking sequence with a neomycin-resistance cassette (Fig. 1A). Because homologous recombination leads to deletion of the initiation codon and sequences encoding most of the MH1 domain, the resulting mutation is predicted to generate a null allele. A single correctly targeted ES cell clone was used to generate chimeras that transmitted the mutation, termed Smad1^Robm1, to progeny. A second null allele was generated by in vivo Cre-mediated excision of a conditional allele, Smad1^RobPC (Fig. 1B). The conditional allele harbors an EcoRI and loxP site inserted 5′ and a loxP-flanked hygromycin-resistance cassette inserted 3′ of the first coding exon. Two independently targeted ES-cell clones were used to generate chimeras. Offspring carrying the Smad1^RobPC allele were mated to mice carrying a Protamine-Cre transgene that directs Cre-recombinase expression specifically to the male germline (O’Gorman et al., 1997). Thus, germline recombination results in one of several configurations of the Smad1 locus, including the null allele termed Smad1^Robcn (Fig. 1B). The Smad1^Robcn allele removes all of the first coding exon and approximately 2 kb of 5′ and 3′ flanking genomic sequences, but leaves a single loxP and EcoRI site in the locus (Fig. 1B).

Northern blot analysis of RNA isolated from homozygous mutant embryos demonstrated that the Smad1^Robm1 and Smad1^Robcn mutations result in the production of a truncated transcript (Fig. 1E). To determine if the remaining transcript encodes a protein, Western blot analysis was performed on individual 9.5 dpc embryos with an antibody that recognizes the C-terminal domain of Smad1 (Fig. 1F). Although a band of approximately 55 kDa was detected in wild-type and heterozygous samples, neither a wild-type nor a smaller form of Smad1 protein were present in homozygous mutant embryos, confirming that both alleles produce null mutations.

Although Smad1 heterozygous mice were normal and fertile, no live born Smad1^Robm1 or Smad1^Robcn homozygous mice were produced from heterozygous matings (Table 1), and both alleles result in identical embryonic lethal phenotypes. Timed matings from heterozygous intercrosses were used to determine the onset of lethality. With death described as the presence of necrotic cells and lack of a beating heart, Smad1-null embryos were dead or dying by 10.5 dpc and failed to establish a functional placental connection. Because other mutations with defects in placentation also die at a similar developmental stage (Gurtner et al., 1995; Kwee et al., 1995; Luo et al., 1997; Yang et al., 1995), the lethality of the Smad1 mutation is likely attributable to this defect.

The first morphological abnormalities in Smad1-deficient embryos coincide with the onset of gastrulation. Although gastrulation appears to initiate normally, as evidenced by the formation of a primitive streak and emergence of nascent mesoderm, characteristic ruffles are present in the VE overlying the distal extra-embryonic region adjacent to the primitive streak (Fig. 2A,B,D). Despite the grossly abnormal shape of the posterior VE, the individual cells within the ruffles display the columnar/cuboidal morphology characteristic of extra-embryonic VE and also maintain proper cell-cell adhesion.

The second visible abnormality is a profound distortion of the embryonic region. Although wild-type embryos are bilaterally symmetric along the proximodistal (PD) axis, whole-mount and histological analyses reveals that this symmetry is rarely maintained in Smad1-mutant embryos. The epiblast and newly formed mesoderm display an abnormal twisting that is evident by 7.0 dpc (Fig. 2C,E). In spite of these profound morphological defects, Smad1-deficient embryos proceed through gastrulation and are represented at Mendelian ratios at 9.5 dpc (Table 1). Although the VE invariably remains ruffled at 8.0-8.5 dpc, the emerging AP axis is symmetrical, headfolds and somites form, gross VYS morphology and accompanying mesoderm are normal, and the node is clearly defined (Fig. 2I,J).

At the onset of gastrulation, epiblast cells move towards and ingress through the primitive streak to form mesoderm. Fate-mapping studies have shown that precursors of the extra-embryonic mesoderm are located in the proximal portion of the epiblast and are the first cell population to traverse the streak (Lawson et al., 1991). Extra-embryonic mesoderm emerges as a bulge of tissue that displaces the overlying extra-embryonic ectoderm. This bulge of cells, together with a slower growing anterior bulge, cavitates to form the posterior and anterior amniotic folds. These folds eventually fuse, forming a physical barrier between the embryo proper and the extra-embryonic tissues. A second consequence of this proximal migration of extra-embryonic mesoderm is the displacement of the extra-embryonic ectoderm, the precursor of the chorion that overlies the epiblast. These coordinated morphogenetic movements lead to the emergence of the chorion as a distinct anteriorly tilting disk that is displaced proximally towards the ectoplacental cone (EPC) and eventually fuses with the ventral surface of the developing placenta. In Smad1 mutants, this process is abnormal. While the amniotic folds form and the extra-embryonic mesoderm migrates to the line the VE, the chorion is erratically folded within the exocoelomic cavity (Fig. 2J). Despite the incorrect placement and folding of the chorion at this stage, the composition of this tissue is histologically normal and the chorionic ectoderm is lined with mesoderm (Fig. 2K).

The allantois emerges as a bud of extra-embryonic mesoderm that is continuous with the most posterior aspect of the primitive streak. As the bud expands and projects into the exocoelomic cavity, the proximal portion enlarges via mitosis and cavitation to produce a loosely scattered mesothelium capable of chorioallantoic fusion (Downs, 2000; Ellington, 1985), while the distal cells at the base of the allantois remain densely packed and continuous with the streak. In comparison with wild-type littermates at 7.5-8.5 dpc, the allantois of Smad1-mutant embryos fails to elaborate and remains as a compact mass of extra-embryonic mesoderm continuous with the streak (Fig. 2F-M). Unlike the twisting of the mutant epiblast, the growth of the allantois never recovers and thus by 9.5 dpc, three types of allantois-like structures have developed (Fig. 3). The mutant allantois can develop into a balloon-like structure containing peripheral blood-filled vessels (Fig. 3A) or a dense structure filled with compact mesenchyme and small blood-filled sacs (Fig. 3B). In a fraction of the mutants, the allantoic rudiment occasionally forms a tenuous connection with the chorionic surface of the placenta, but fails to efficiently invade (Fig. 3D).

Other defects observed in a subset of Smad1-deficient embryos at 9.5-10.0 dpc include a distended pericardium, delays in ventral closure and incomplete embryonic turning (data not shown). Some of these defects, in particular incomplete turning, may be secondary as the embryos are not tethered to the placenta. Furthermore, Smad1-deficient embryos are often irregularly positioned within the yolk-sac and are smaller than their wild-type counterparts, most probably reflecting a nutritive defect resulting from failure to establish a maternal connection.

To more carefully analyze the tissue defects in Smad1 mutant embryos, we used a panel of molecular markers. To establish whether the defects observed in early mesodermal populations originate from defects in the morphogenesis of the primitive streak, Brachyury (T) expression was assessed. In wild-type embryos, T marks the entire PD axis of the streak and nascent mesoderm (Wilkinson et al., 1990; Fig. 4A). By contrast, Smad1 mutants exhibited discontinuous and less intense T expression (Fig. 4B,C). Fgf8, a marker of the primitive streak and the base of the allantois, was also examined (Crossley and Martin, 1995). Although expression levels in the streak were comparable with those in wild-type littermates, posterior views of 7.0 dpc mutants demonstrates that the streak is irregularly twisted (Fig. 4D). In many mutants, we noted that the proximal-most domain of Fgf8 expression lay within the VE folds, presumably demarcating the base of the allantois (Fig. 4D,E). To confirm this possibility, we assessed Fgf3 expression, a marker for extra-embryonic mesoderm (Wilkinson et al., 1988). As with Fgf8, cells within the VE folds express Fgf3 confirming that extra-embryonic mesoderm migrates into the pockets of VE (Fig. 4F,G).

Reciprocal interactions between the VE and underlying tissues define gene expression profiles and thus the identity of both tissues. Alpha foetoprotein (AFP) is initially expressed throughout the VE overlying the embryonic region, but is excluded from the extra-embryonic region, owing to an inhibitory influence from the extra-embryonic ectoderm (Dziadek and Adamson, 1978). Thus, the proximal edge of AFP expression normally coincides with the distal edge of the extra-embryonic ectoderm (Fig. 5A). In Smad1-mutant embryos, AFP expression is normal in the VE overlying the embryonic region. However, AFP expression is absent in anterior aspects of the extra-embryonic region (Fig. 5A). Failure to activate AFP indicates either an intrinsic defect within the anterior VE or, more likely, that AFP induction is repressed by the presence of ectopic extra-embryonic ectoderm (Fig. 5A′).

To distinguish between these possibilities, the expression of two extra-embryonic ectodermal markers, eomesodermin (eom) and Bmp4 were examined. At early headfold stages, eom expression is detected throughout the ectoderm of the proximally migrating chorionic disk and the anterior streak (Ciruna and Rossant, 1999; Hancock et al., 1999) (Fig. 5B). Similarly, Bmp4 is expressed in the ectodermal layer of the chorion, as well as in the extra-embryonic mesodermal derivatives (Lawson et al., 1999) (Fig. 5C). In early headfold mutant embryos, both markers are observed throughout the extra-embryonic region, but are excluded from the region occupied by and immediately adjacent to the VE folds (Fig. 5B,C). These results support the hypothesis that the exclusion of AFP expression from the VE overlying the extra-embryonic regions of mutant embryos reflects the abnormal migration and/or overproliferation of the extra-embryonic ectoderm.

To determine the extent of embryonic lineage formation as well as the fate of extra-embryonic ectoderm at later stages, we examined Sox2 expression in late headfold embryos. Sox2 is normally expressed in the neuroectoderm, definitive endoderm and throughout the extra-embryonic ectoderm (Wood and Episkopou, 1999). In Smad1 mutants, while the embryonic expression pattern is normal, the extra-embryonic expression pattern is highly aberrant. Unlike wild-type embryos, a Sox2-positive projection extends from the posterior proximal edge of the embryonic region towards the EPC (Fig. 5D,E,E′). Proximal sections of Sox2-labeled embryos also reveals the presence of a copious amount of erratically folded chorionic ectoderm (Fig. 5E′′). Collectively, our analyses show that loss of Smad1 severely perturbs the growth and morphogenesis of the chorion.

BMP signals can be transmitted by Smad1, Smad5 and Smad8 (Chen et al., 1997; Hoodless et al., 1996; Kawai et al., 2000; Macias-Silva et al., 1998). To establish any correlation between tissue abnormalities seen in Smad1 mutants and the pattern of expression of BMP-transducing Smads, we compared the expression pattern of all three genes from early stages onwards. At 6.5 dpc Smad1 is highly expressed in the VE, and at much lower levels in the early epiblast (Fig. 6A). After the initiation of gastrulation, Smad1 is expressed at high levels throughout the embryonic and extra-embryonic mesoderm and VE, with lower levels detected in epiblast tissue and chorion (Fig. 6B). By comparison, Smad5 is expressed at low levels throughout the pre-gastrulation embryos but is markedly upregulated by 7.25 dpc in the entire embryonic and extra-embryonic mesoderm and epiblast (Fig. 6D-E)(Chang et al., 1999). Smad8 is not expressed above background levels at 6.5 dpc (Fig. 6G). By 7.25 dpc, Smad8 is expressed at the highest levels in the VE, amnion and allantois with lower levels apparent in the posterior streak and chorion (Fig. 6H). Thus, all three BMP-transducing Smads are co-expressed in the extra-embryonic mesoderm populations at 7.25 dpc, while only Smad1 and Smad8 are abundantly expressed in the VE. By 8.5 dpc, all three genes are uniformly and ubiquitously expressed in the embryo (Fig. 6C,F,I). These results demonstrate that while Smad1 is uniquely expressed in the early VE, by the onset of gastrulation all tissues express multiple BMP-transducing Smads.

Embryonic stem (ES) cells introduced into blastocysts predominantly colonize the epiblast (Beddington and Robertson, 1989). Thus, in chimeric embryos, ES cell derivatives are confined to the embryo, while the extra-embryonic lineages are colonized exclusively by the host blastocyst. This developmental bias has routinely been used to discern the role of specific molecules in the epiblast versus extra-embryonic tissues. We therefore generated chimeric embryos to determine if the defects observed in Smad1-deficient embryos are due solely to the requirement for Smad1 in the VE and extra-embryonic ectoderm or reflect a requirement for Smad1-mediated BMP signaling in the epiblast and its derivatives. lacZ-marked wild-type cells (Tremblay et al., 2000) were injected into blastocysts derived from Smad1^Robm1 heterozygous intercrosses. In this experiment, 25% of the chimeras are expected to be composed of Smad1-deficient primitive endoderm and extra-embryonic ectoderm. Chimeric embryos were recovered for analysis at 7.5 dpc. Six out of 28 (21%) displayed both a ruffled posterior visceral endoderm and a twisted epiblast phenotype (Fig. 7A,B). Of these, three were colonized predominately by lacZ-positive derivatives. This result suggests that the VE defect and the distortion of the egg cylinder can be attributed to the lack of Smad1 in the extra-embryonic lineages.

Cell lineage analysis has shown that at 6.0 dpc the precursors of the primordial germ cells (PGCs) lie in the extreme proximal epiblast adjacent to the extra-embryonic ectoderm (Lawson and Hage, 1994). This precursor population gives rise to descendants not only in the germline, but also in the extra-embryonic mesoderm, including allantois. Recent genetic and embryological experiments show that BMP signals emanating from the extra-embryonic ectoderm act directly on the proximal epiblast, and are required for the formation of PGCs and allantois (Lawson and Hage, 1994; Ying et al., 2000; Yoshimizu et al., 2001). Because of the biochemical evidence that places Smad1 downstream of BMP signals and because of the abnormal allantois morphogenesis in Smad1 null mutants, we chose to investigate the potential involvement of Smad1 in PGC development.

Allocation to the germ cell lineage is thought to occur shortly before PGCs can first be identified as a cluster of strongly alkaline phosphatase (AP)-positive cells at the base of the allantois (Ginsburg et al., 1990; Lawson and Hage, 1994). Around 8.5 dpc, the PGCs intercalate into the hindgut epithelium and migrate anteriorly. We collected Smad1-deficient embryos at 9.0-9.5 dpc and assayed for PGC formation by staining for endogenous AP activity. While wild-type and heterozygous littermates between 13 and 26 somites all had similar numbers of PGCs (n=80-250), the number observed in mutant embryos was significantly reduced (Fig. 8A-C). Of the 14 similarly staged mutants, six out of 14 (43%) had no PGCs, five out of 14 (36%) had fewer than ten and three out of 14 (21%) had less than 25. These results suggest that loss of Smad1 affects either the initial specification of the PGC precursor population in the proximal epiblast or survival and/or proliferation of the allocated PGC population. Further support for the role of Smad1 in the early specification of PGCs comes from the analysis of PGC numbers at the late headfold (LHF)-stage, when early PGCs are visible as a cluster at the base of the allantois. Consistent with the deficit of PGCs in later embryos, HF-stage Smad1^−/− either lack or contain less than three PGCs (data not shown). An alternative explanation is that the severe posterior streak defects in Smad1-null embryos compromise local signals within the extra-embryonic lineages that may play a role in the allocation of the PGC precursors to the germline or affect their immediate survival. To control for this possibility, we examined PGC formation in Smad2-deficient embryos. Smad2 mutants embryos do not form a recognizable primitive streak, and by 8.0 dpc the entire epiblast transforms into extra-embryonic mesoderm (Waldrip et al., 1998). In spite of these profound cellular disturbances, we found that the majority of Smad2-deficient embryos (20/25) collected at 8.5 dpc contained PGCs. Several Smad2 mutants formed abundant PGCs, many of which were concentrated in darkly stained clusters (Fig. 8D,E). The remaining Smad2 mutants (5/25) exhibited a more severe phenotype, and completely lacked AP-positive cells (data not shown). Collectively, these data suggest that the formation of PGCs from the proximal epiblast can proceed even in the absence of normal gastrulation movements. Therefore, as opposed to a secondary consequence of aberrant gastrulation, the deficit of PGCs in Smad1 mutant embryos most probably reflects a direct requirement for Smad1 in transducing extra-embryonic ectoderm-derived BMP signals

Members of the TGFβ superfamily of extracellular signaling molecules and their cognate receptors play a variety of roles during vertebrate embryogenesis. We show that Smad1, a downstream component of the BMP signaling pathway, is essential for coordinating the growth and elaboration of extra-embryonic tissues of the mouse embryo (Goumans and Mummery, 2000). This finding is consistent with the observation that multiple BMP ligands are expressed in both embryonic and extra-embryonic tissues from postimplantation stages onwards.

Smad1-mutant embryos are readily identified as early as 7.0 dpc, owing to the localized overgrowth of the VE adjacent to the posterior embryonic/extra-embryonic boundary. Despite the localized tissue disturbance, Smad1 transcripts are uniformly expressed throughout the VE (Fig. 6A-C). Intriguingly, Bmp2 is asymmetrically expressed at the onset of gastrulation (Ying and Zhao, 2001), with the highest levels in the VE overlying the proximal regions of the primitive streak. However, the VE defects seen in Smad1^−/− embryos cannot be attributed to the attenuation of BMP2 signaling alone, because embryos lacking BMP2 do not exhibit this phenotype (Zhang and Bradley, 1996). Bmp4 and Bmp8b are expressed in tissues adjacent to the affected VE population, but neither molecule nor other components of the BMP signaling pathway are known to be asymmetrically distributed. In contrast cerberus-like (Cer1), an antagonist of both BMP and Nodal signaling (Piccolo et al., 1999) is confined to the VE overlying the anterior side of the embryo. While Cer1 is not essential for development (Belo et al., 2000; Shawlot et al., 2000; Simpson et al., 1999), the anterior expression of Cer1 as well as that of other antagonists, such as dickkopf1and lefty1 (Glinka et al., 1998; Meno et al., 1999), suggests a mechanism that limits the levels of activated Smad1 anteriorly. Given the expression patterns of the antagonists and the BMP ligands during early gastrulation, it is likely that the uniform expression of Smad1 in the VE is converted to a posterior activity gradient. Our data is consistent with this Smad1 activity playing an important role in negatively regulating posterior VE growth.

The overgrowth of the VE in Smad1-deficient embryos is causally associated with a pronounced twisting of the epiblast and mesoderm tissue along the PD axis. Thus, we find that neither defect is rescued in chimeric embryos in which the epiblast derivatives are largely wild type, suggesting that this phenotype results from the lack of Smad1 in extra-embryonic tissues, most likely the VE where Smad1 is highly expressed (Fig. 6A-C). Remarkably, the morphogenetic disturbances are relatively transient, as judged by whole-mount, histological and molecular criteria, and all mutant embryos regain bilateral symmetry by 8.0 dpc. Interestingly the correction of the phenotype is coincident with formation of the definitive endoderm, the cell population that displaces the VE. Thus, one likely explanation for the recovery of bilateral symmetry is that the replacement of the VE by the emerging definitive endoderm, which is specified normally in Smad1^−/− embryos, relieves the mechanical restrictions imposed by the VE. Interestingly, loss of function of other transcription factors expressed in the VE result in altered epiblast morphology (Acampora et al., 1995; Ang et al., 1996; Ang and Rossant, 1994; Shawlot and Behringer, 1995). In these cases, abnormal morphogenesis of VE causes a constriction at the embryonic/extra-embryonic boundary. Chimera analysis confirms that these defects reside in the extra-embryonic tissues, further supporting the idea that defects in the VE secondarily impact epiblast morphogenesis (Dufort et al., 1998; Rhinn et al., 1998; Shawlot et al., 1999). Importantly, unlike Smad1 mutants, these mutations are more severe and disturb the formation of the primary embryonic axis and are often associated with defects in the definitive endoderm.

In Smad1 mutants, T expression is severely perturbed during early/mid-streak stages, but is present at normal levels by 7.5- 8.0 dpc (Fig. 4A-C; data not shown). T is normally uniformly expressed in nascent mesoderm forming along the entire length of the primitive streak. However in the absence of Smad1, we find reduced and intermittent T expression. Strikingly, at these early stages, other streak and mesoderm markers, including Fgf8, Fgf3 and eom are also expressed in an irregular pattern, indicative of the disrupted PD axis noted above, but in contrast to T, these markers are expressed at wild-type levels (Fig. 4D; data not shown). One possibility for the selective reduction of T expression is that Smad1 is involved in the activation of early T expression. Indeed, in T mutants and T chimeras allantois formation is impaired, and resembles what we describe here for Smad1-deficient embryos (Rashbass et al., 1991; Wilson et al., 1993). A recent report has suggested that SIP1, a novel zinc-finger/homeodomain protein, can interact with both activated Smads and target gene promoters. Interestingly, SIP1 binding motifs have been identified in the Xenopus T homolog (Verschueren et al., 1999), providing evidence in support of a Smad1-mediated pathway regulating T transcription.

Previous gene targeting experiments have underscored the importance of BMP signaling in the formation of the allantois. Embryos that lack Bmp4 form a rudimentary allantois (Lawson et al., 1999), while in Bmp2 mutants, allantois formation is slightly delayed (Zhang and Bradley, 1996). Embryos that lack Bmp8b, a member of the 60A subgroup of BMP ligands, have a shortened allantois attributed to developmental delay (Ying et al., 2000). Other members of the 60A subgroup are not uniquely important for allantois formation, but in Bmp5/7 double mutant embryos the allantois is reduced in size, developmentally delayed and often unable to fuse with the chorion (Solloway and Robertson, 1999). Interestingly, despite the robust co-expression of Smad1, Smad5 and Smad8 in the allantois (Fig. 6) and adjacent tissues, loss of either Smad1 or Smad5 results in distinct allantois phenotypes. Smad5-deficient embryos show a marked delay in allantois development, and in severely affected mutants the allantois fails to fuse with the chorion and may become displaced onto the amnion (Chang et al., 1999; Yang et al., 1999a). Loss of Smad1 has a more pronounced impact on allantois formation. Thus, at 8.0 dpc, while control littermates have produced an extended allantois composed of differentiated cell populations (Fig. 2H,L,M) in Smad1-deficient embryos the allantois forms a compact mass of cells that is continuous with the primitive streak (Fig. 2F,G,I-K). By 9.5 dpc, the Smad1^−/− allantois develops into an unpatterned mass of tissue containing vascular endothelium and red blood cells (Fig. 3). Taken together, these results suggest that although allantois-like tissue is specified during gastrulation, the allantois rudiment that forms in Smad1 mutants is defective in proliferation and patterning. These results are provocative because neither the Smad1 or the Smad5 loss-of-function mutation fully recapitulate the allantois phenotype of any one BMP mutant, suggesting that a single BMP activates multiple Smads. Furthermore, these distinct phenotypes suggest that Smad1 and Smad5 play discrete roles in the development of the allantois; Smad5 signals govern the timing of development and migration, while Smad1 signals direct proliferation and patterning.

Elegant cell marking experiments have demonstrated that the progenitors of the allantois and the primordial germ cells are located in the proximal region of the epiblast, close to the embryonic/extra-embryonic junction (Lawson and Hage, 1994). BMP4 and BMP8b, synthesized in the extra-embryonic ectoderm, are required for the generation of this precursor population (Lawson et al., 1999; Ying et al., 2000), and recent embryological experiments suggest that the BMP ligands act directly on the epiblast (Lawson et al., 1999). Consistent with these findings, we show that Smad1 is expressed at low levels within the epiblast (Fig. 6A), and that Smad1 mutants have drastically reduced numbers of PGCs. The simplest explanation for the marked impact on PGC formation in Smad1 mutants is that cells of the proximal epiblast are severely impaired in their ability to transduce extra-embryonic ectoderm produced BMP4/BMP8b signals. Consequently, very few to no cells are specified as PGC precursors. Furthermore, the presence of PGCs in about 50% of Smad1 mutants suggests that the requirement for Smad1 is not absolute, and that Smad5, which is also expressed at low levels in the epiblast, may partially compensate for the absence of Smad1 in some cells. Whether loss of Smad5 also impairs PGC development has yet to be reported.

A two-step model for the production of PGCs has been proposed, based on the estimation that allocation to the germ cell lineage occurs around 7.2 dpc, after the precursor population emerges from the posterior primitive streak in the nascent extra-embryonic mesoderm (Lawson et al., 1999). In this model, BMP4/BMP8b produced by the extra-embryonic ectoderm specifies a population of PGC and allantois progenitor cells in the proximal epiblast. A second signal originating in or adjacent to the proximal streak then instructs these precursor cells into either an allantois or PGC lineage. The abundant expression of Smad1 in the posterior streak, extra-embryonic mesoderm and VE, suggests that Smad1 may be a component of this putative ‘second signal’, responding to locally high levels of BMP2 produced by the posterior VE, or BMP4 synthesized in the newly formed extra-embryonic mesoderm. To refine our understanding of the tissue-specific requirements for Smad1 that underlie these events, a future goal will be to exploit the conditional allele to selectively remove Smad1 from different tissue compartments of the early embryo.

A unique aspect of the Smad1 phenotype, not shared by other mutations in BMP pathway genes, is our finding that Smad1 is required for growth and morphogenesis of the chorion. The chorion, composed of both extra-embryonic ectoderm and mesoderm, is an essential component of the mature placenta. Coincident with gastrulation, the extra-embryonic ectoderm and an accompanying layer of extra-embryonic mesoderm, move proximally towards the EPC. This proximal movement and regulated growth appears to be an intrinsic property of the chorion, as in embryos that lack most epiblast derivatives or are defective in allantois morphogenesis, chorion development proceeds normally (Chang et al., 1999; Gurtner et al., 1995; Kwee et al., 1995; Waldrip et al., 1998; Yang et al., 1995; Yang et al., 1999a). Thus, the chorion defects observed in Smad1 mutant embryos are likely to be due to a requirement for Smad1 activity in this tissue. Interestingly, as is the case with the posterior VE, the chorion represents a second site in the embryo where loss of Smad1 results in massive overgrowth of cells, implicating Smad1 pathways in regulating the growth of these specific tissues. Because Smad1 is expressed in both the mesodermal and ectodermal components of the chorion, it remains to be determined whether Smad1 signals in the extra-embryonic ectoderm regulate the growth of the extra-embryonic mesoderm or vice versa.

In addition to appropriate specification and differentiation of distinct cell types, the coordinated growth and morphogenesis of various tissue layers is also an important feature of the developing embryo. BMP signaling pathways have been shown to participate in controlling the growth of the embryo. In the present study, we have uncovered a role for Smad1 in regulating the formation and proliferation of the chorion, allantois and VE, as well as PGCs. Interestingly, Smad1 acts to inhibit growth of the chorion and VE, while Smad1 activities promote growth of the allantois and formation of PGCs. Recent reports demonstrate the widespread expression of Smad1 during later embryogenesis (Dick et al., 1998; Huang et al., 2000) and BMP signaling pathways make important contributions to developing organ systems such as the eyes, lungs and kidneys (Bellusci et al., 1996; Dudley et al., 1999; Dudley et al., 1995; Weaver et al., 1999). Future experiments should aim to explore the unique contribution of Smad1 to these later processes, using the conditional allele to bypass the early extra-embryonic defects.

A defect in primorial germ cell development was recently reported by Chang and Matzuk (

We thank Pamela Hoodless and Jeff Wrana for the Smad1 cDNA and hSmad1 expression vector, Liz Bikoff, Ross Waldrip and Audrey Jackson for isolation and initial characterization of genomic clones, Juliana Brown for assistance with the Robm1 construct, Gerry Chu for the Smad8 probe, Debbie Pelusi for assistance with genotyping, Jane Brennan, Rob Godin and Matt Deardorff for comments on the manuscript, and Patti Lewko and Joe Rocca for expert animal care. K. D. T. and N. R. D. were supported by postdoctoral fellowship grants from the NICHD. This work was supported by a grant from the NIH to E. J. R.