Placental vascularisation requires the AP-1 component Fra1

AP-1 (activator protein 1) is a dimeric transcription factor composed of the products of the Jun and Fos proto-oncogenes (Angel and Karin, 1991). The Jun proteins (Jun, Junb and Jund) form either homo-or heterodimers with members of the Fos and ATF protein families, whereas the Fos proteins (Fos, Fosb, Fra1 and Fra2) cannot associate with each other or with ATF proteins, but form stable heterodimers with any of the Jun proteins. AP-1 is crucially involved in a multitude of cellular processes, including development and differentiation, cell proliferation, apoptosis, oncogenic transformation, and the response to genotoxic agents (Angel and Karin, 1991; Devary et al., 1992; Angel and Herrlich, 1994; Schreiber et al., 1995; Karin et al., 1997). AP-1 activity is rapidly induced by a vast number of extracellular stimuli, such as growth factors, cytokines, tumour promoters, carcinogens and specific oncogenes, and AP-1 is thought to play a central role in changing the pattern of gene expression in response to extracellular signals. These extracellular signals activate AP-1 via the ERK and JNK signal-transduction cascades through increased expression, as well as phosphorylation, of pre-existing and newly synthesised AP-1 subunits, which modulate the DNA-binding and transactivation functions of Jun and Fos proteins (Karin, 1995; Karin and Hunter, 1995; Leppä and Bohmann, 1999).

Like all other Fos and Jun genes, Fra1 (Fos-related antigen 1; also termed Fosl1) is an immediate-early gene (Cohen and Curran, 1988; Cohen et al., 1989). The DNA-binding specificity of Fra1/Jun heterodimers is indistinguishable from that of Fos/Jun heterodimers on several AP-1 binding sites (Cohen et al., 1989). In contrast to Fos, Fra1 lacks a transactivation domain, and the entire Fra1 protein (e.g. when fused to the DNA-binding domain of Gal4) fails to activate transcription (Suzuki et al., 1991; Wisdom and Verma, 1993; Bergers et al., 1995). Therefore, Fra1 can either increase or decrease total AP-1 activity depending on the status of the other Fos and Jun proteins in the cell, and has been proposed to function as a negative-feedback regulator of AP-1 (Suzuki et al., 1991; Groskopf and Linzer, 1994; Bergers et al., 1995; Welter et al., 1995; Yoshioka et al., 1995; Schreiber et al., 1997). The oncogenic potential of Fra1 is significantly weaker than that of Fos. Nevertheless, overexpression of Fra1 in established rat fibroblasts leads to anchorage-independent growth and tumour development in nude mice (Bergers et al., 1995). Furthermore, neoplastic transformation of rat thyroid cells requires induction of Fra1 and Junb (Vallone et al., 1997). Fra1 expression is subject to positive control by AP-1 in several cell types (Grigoriadis et al., 1993; Brüsselbach et al., 1995; Bergers et al., 1995; Schreiber et al., 1997; Matsuo et al., 2000). Interestingly, the basal and AP-1-induced expression of Fra1 depends primarily on regulatory sequences in the first intron, which contains three AP-1 binding sites (separated by 8 and 7 bp; Bergers et al., 1995).

Numerous in vitro studies have suggested that the different AP-1 dimers may act as tissue-specific and signal-specific transcriptional activators. Indeed, each individual targeted deletion of an AP-1 gene reported so far leads to specific phenotypes, indicating that the different AP-1 subunits, although highly homologous, are not fully redundant in vivo. Mice lacking Jund are viable, but exhibit reduced postnatal growth and multiple defects in male reproductive functions (Thepot et al., 2000), whereas deletion of Jun or Junb leads to embryonic lethality (Hilberg et al., 1993; Johnson et al., 1993; Schorpp-Kistner et al., 1999). Lack of Jun causes liver and heart defects, resulting in embryonic lethality at E12.5 (Hilberg et al., 1993; Eferl et al., 1999), whereas Junb^−/−foetuses die between E8.5 and E10.0, owing to multiple defects in extra-embryonic tissues, such as the placental labyrinth (Schorpp-Kistner et al., 1999). In contrast, mice lacking Fos or Fosb are viable and fertile. In one study, nurturing defects were detected in Fosb^−/−mice (Brown et al., 1996), whereas this phenotype was not observed in a second study (Gruda et al., 1996). Mice lacking Fos develop osteoporosis, owing to a complete differentiation block in bone-resorbing osteoclasts (Wang et al., 1992; Johnson et al., 1992; Grigoriadis et al., 1994; Matsuo et al., 2000). The biological function of Fra1 in vivo is not yet known, although it is a transcriptional target of Fos in osteoclasts and macrophages (Matsuo et al., 2000). Therefore, it is possible that deletion of Fra1 might phenocopy, at least in part, the effects of the Fos deletion. To address these questions, we have disrupted the Fra1 gene in embryonic stem (ES) cells and mice. Foetuses lacking Fra1 were found to be severely growth retarded and died between E10.0 and E10.5. Mutant embryos revealed a highly abnormal yolk sac, and the placenta lacked a properly vascularised labyrinth layer. Fra1 expression was detected in the yolk sac and the labyrinth layer using the lacZ reporter gene integrated into the targeted allele. Injection of Fra1^−/−ES cells into tetraploid wild-type blastocysts, which form exclusively extra-embryonic tissues, rescued the embryonic lethality and gave rise to Fra1^−/−foetuses that were no longer growth retarded and survived up to 2 days after birth, suggesting that the observed defects in the extra-embryonic compartment were causal to the lethality.

The mouse (strain 129/Sv) Fra1 gene has previously been cloned and characterised (Schreiber et al., 1997). To generate a targeting construct for homologous recombination, a 603 bp BsrBI-NheI fragment of Fra1 consisting of the last 66 nucleotides of exon 3, the entire third intron and the first 165 nucleotides of exon 4 was replaced by a pGNA cassette (Le Mouellic et al., 1992), fusing the lacZ gene of pGNA in-frame with the coding sequence of exon 3 of Fra1. The pGNA cassette introduces a neomycin phosphotransferase gene driven by a PyEC F9.1 enhancer/RSV LTR promoter, which allows for positive selection in ES cells with G418. The targeting construct (pGNA/Fra1) had, as a 5′ arm of homology, a 5170 bp NsiI-BsrBI fragment of Fra1 containing part of intron 1, exon 2, intron 2 and part of exon 3, whereas the 3′ arm was comprised of a 985 bp NheI-NdeI fragment of exon 4. The introduced deletions in exons 3 and 4 removed the basic region of the DNA-binding domain (except the first three amino acids) and the entire leucine-zipper dimerisation domain of Fra1. In addition, the last splice acceptor site of the gene was deleted to prevent aberrant splicing across the introduced pGNA sequences. A herpes simplex virus thymidine kinase-cassette (HSV-TK) was added to the construct for negative selection against random integration (Mansour et al., 1988), which could be positioned to the long or short arm of homology after linearisation with NotI or NsiI, respectively (see Fig. 1B).

The ES cell lines used in this study, D3 (Doetschman et al., 1985) and R1 (Nagy et al., 1993), were cultured on feeder cells in the presence of LIF as described by Wang et al. (1992). For electroporation, 10⁷ trypsinised ES cells were suspended in 800 μl PBS, mixed with 10-20 μg of linearised targeting vector (pGNA/Fra1), and an electric pulse of 260 V at 500 μF was applied with a Gene Pulser (BioRad, Munich). Cells were selected in the presence of 0.3 mg/ml G418 and 2 μM gancyclovir to enrich for clones that had undergone homologous recombination at the Fra1 locus. Screening of colonies by PCR was performed as described (Wang et al., 1992, 1994) with two sets of nested primers: fra1Y (5′-TGGGGTGGGATTTGAGACGG-3′; 3′ of exon 4 of Fra1 outside the sequence of the targeting vector) and neo3 (5′-GTCATCTCACCTTGCTCCTGC-3′; neo^R gene), and a set of primers positioned inside of primers fra1Y and neo3, i.e. fra1X (5′-CTAAAGCCCACTGAACCGCC-3′; 3′ of exon 4) and neo2 (5′-CGCCTTCTTGACGAGTTCTTCTGAG-3′; neo^R gene). Colonies positive for the diagnostic approx. 2.0 kb PCR fragment were further analysed by Southern blotting (see below).

Genomic DNA was isolated from ES cells, embryos, yolk sacs and tail biopsies, as described by Hilberg et al. (1993). Genotyping was performed by PCR and/or Southern blot analysis, which allowed us to distinguish the endogenous and targeted alleles of Fra1 based on different lengths of amplification products or restriction fragments, respectively. Southern blot analysis was performed using a probe located 3′ of exon 4 outside of the targeting vector (approx. 750 bp NdeI/SacI fragment; Fig. 1) and genomic DNA which had been digested with ApaI (Fig. 1D), SacI or SalI/NotI, and separated by gel electrophoresis (0.8% agarose/TAE). The following primers were used for genotyping by PCR: Pr1, 5′-GGGCTTTGTTGGCATAGTA-GATTG-3′ (derived from intron 2 of Fra1); Pr2, 5′-AGCTCCTTTC-TTCGGTTTCTGC-3′ (exon 3 of Fra1; deleted in the targeted allele); and Pr3, 5′-AAGCGCCATTCGCCATTCAG-3′ (nucleotides 179-159 of lacZ). Primers Pr1 and Pr2 amplify a 567 bp fragment of the endogenous allele, whereas primers Pr1 and Pr3 amplify a 699 bp fragment of the targeted allele. The Ubi-Junb and H2K^b-Fra1-LTR transgenes used in genetic rescue experiments were genotyped as described in Schorpp et al. (1996) and Jochum et al. (2000), respectively. In matings involving the H2K^b-Fra1-LTR transgene, genotyping of the endogenous and targeted alleles of Fra1 was performed by Southern blot analysis.

Targeted ES cells were injected into C57BL/6 blastocysts that were subsequently transferred into the uteri of pseudopregnant recipients.

Tetraploid pre-implantation embryos were generated by electrofusion of the two blastomeres of two-cell stage embryos as described (Nagy et al., 1993; Nagy and Rossant, 1993). The generation of chimaeras by aggregation of diploid with tetraploid morulas, or by injection of ES cells into tetraploid blastocysts was performed as described by Nagy and Rossant (1993) and Wang et al. (1997). To derive Fra1^−/− ES cells, E3.5 blastocysts were isolated from Fra1^+/− intercrosses (strain 129/Sv), cultured on feeder layers, and ES cells established as described (Hogan et al., 1994). Out of 80 blastocysts explanted, five ES cell lines were established (one Fra1^+/+, three Fra1^+/− and one Fra1^−/−). Fra1^−/− ES cells were used for injection into diploid or tetraploid blastocysts at passage 5.

Various tissues of adult chimaeric mice were dissected and minced in distilled water. Samples were lysed by three cycles of freeze-thawing and were subjected to glucose phosphate isomerase (GPI) isozyme analysis as described previously (Hilberg et al., 1993). The proportion of ES-cell- and host-derived cells in the chimaeric tissues was estimated from the ratio of the GPI-1A versus GPI-1B isozyme activity after visualisation in a coupled enzymatic assay.

Tissue samples and mouse embryos were fixed in 4% paraformaldehyde, embedded in paraffin wax, cut and counterstained with Haematoxylin and Eosin as described by Aguzzi et al. (1990). Sense and antisense cRNA probes for in situ hybridisations were in vitro transcribed in the presence of ³⁵S-rUTP as described (Aguzzi et al., 1990). The following probes were used: 4311 (Lescisin et al., 1988), Flt1 and Flk1 (Breier et al., 1995). For BrdU-labelling, pregnant mice of heterozygous timed matings were injected intraperitoneally with 60 μg of BrdU (Sigma) per gram bodyweight at day 9.5 of pregnancy. 4 hours after injection, mice were sacrificed and the decidual swellings isolated and fixed in 4% paraformaldehyde at 4°C overnight. Paraffin sections were prepared and incubated with an α-BrdU mouse monoclonal antibody (Calbiochem) at a dilution of 1:200. Thereafter, an ABC staining procedure (Vector) was performed according to the manufacturer’s instructions. To determine the percentage of proliferating cells in various tissues, more than 100 stained and unstained nuclei of several sections were counted. Embryos, placentas and decidual swellings were stained with X-gal following fixation in 0.2% glutaraldehyde/100 mM sodium phosphate, as described by Bonnerot et al. (1987). After staining, tissues were fixed in 4% paraformaldehyde, embedded in paraffin wax, cut and counterstained with Eosin.

Primary MEFs were isolated and immortalised according to the 3T3 protocol (Todaro and Green, 1963). The yolk sac was used as source of genomic DNA for genotyping. Each primary fibroblast culture was isolated from a single E9.5 embryo of 129/Sv or 129×BL/6 genetic background, and each 3T3 fibroblast line was immortalised from an individual primary culture. Fibroblasts were cultured in Dulbecco’s modified Eagle’s medium (DMEM) containing 10% foetal calf serum (37°C, 100% humidity and 5% CO₂).

The targeting strategy used for disruption of the Fra1 gene is shown in Fig. 1. Following homologous recombination, the lacZ gene was fused in-frame to the coding sequence of exon 3 of Fra1 (Fig. 1C). In addition, part of exon 3, intron 3 and part of exon 4 were deleted, removing almost all of the basic DNA-binding domain, the entire leucine-zipper dimerisation domain and the last splice acceptor site of Fra1 (Fig. 1C). Thus, the introduced changes should lead to a null mutation. The HSV thymidine kinase gene (HSV-TK) was used for negative selection against random integration. Following electroporation of ES cells, targeted clones were identified by genomic Southern blot analysis (Fig. 1D), and were obtained at a high frequency in D3 and R1 ES cells (one targeted clone per 3 and 3.6 double resistant transfectants, respectively).

Interestingly, this high frequency was observed only when the HSV-TK cassette was present on the short arm of homology (following linearisation with NsiI; Fig. 1B). When the HSV-TK cassette was present on the long arm (following linearisation with NotI; Fig. 1B), the targeting frequency was approx. 2.5-fold lower, presumably due to lower enrichment following gangcyclovir selection (data not shown). Neither additional random integrations of the targeting construct nor rearrangements of the targeted allele were detected (data not shown). Several D3-derived and R1-derived Fra1^+/− ES cell clones were microinjected into C57/BL6 blastocysts. Male chimaeras from one R1 clone (Rfr-3) and from one D3 clone (Dfr-52) transmitted the mutated allele to their offspring when crossed to C57BL/6 or 129/Sv females. All following studies were performed on mice derived from each of the two independent clones, and in both genetic backgrounds (129/Sv and 129×BL/6). Offspring were genotyped by PCR or Southern blot analysis (Fig. 1D,E).

Heterozygous males and females were fertile and phenotypically indistinguishable from wild-type littermates. However, no homozygous mutants were obtained from heterozygous intercrosses, indicating that mice lacking Fra1 die during embryogenesis (Table 1). To identify the stage of lethality, embryos from timed matings between heteroyzgotes were analysed at different stages of gestation. Fra1^−/−embryos were obtained with the expected Mendelian frequency at E8.5, E9.5 and E10.0. At E10.5, the number of viable mutant embryos was drastically reduced (1 out of 40; χ²=11.72; P<0.01). Viable mutant embryos were never observed at E11.5 or later (Table 1). Thus, foetuses lacking Fra1 died in utero around E10 of gestation. This was observed for mice derived from both ES cell clones, and with both mixed 129×BL/6 and pure 129/Sv genetic backgrounds. To confirm that the deletion of Fra1 alone caused the observed lethality, a genetic rescue experiment was performed by crossing Fra1^+/− mice with transgenic mice ubiquitously expressing the mouse Fra1 gene under the control of the H2K^b promoter (Jochum et al., 2000). Upon intercrossing, the embryonic lethality was rescued, and viable mice lacking the endogenous Fra1 gene, but carrying the transgene, were obtained at almost Mendelian frequency, demonstrating that the lethality of Fra1^−/−mice is due to the absence of a functional Fra1 protein (Table 2).

The most prominent phenotypic alteration of Fra1^−/−foetuses is their small size, which first becomes apparent at E8.5 and was observed for all mutant embryos at E9.5 and E10.0 (Fig. 2). To determine whether primarily a growth defect or developmental retardation is responsible for this small size, the developmental stage of mutant embryos recovered at E9.0-E10.0 was determined using a series of morphological criteria, e.g. somite number, allantois, heart, hindbrain, otic system, forelimb buds and flexion (Brown, 1990). Mutant foetuses recovered at E9.5 were the size of E8.5 foetuses, but the developmental stage was typical of E9.0-E9.5 foetuses (Table 3; Fig. 2A,C). All major organ systems, the body axis and properly organised somites were present, and the process of turning was completed in mutant embryos (Fig. 2C,D). Histological analysis of E9.5 and E10.0 Fra1^−/−embryos revealed no obvious structural anomalies (Fig. 2D). Importantly, the hearts of mutant embryos were beating and exhibited the normal looped morphology characteristic of this stage of development (data not shown). Except for the occurrence of a dilated pericardium in several mutants (Fig. 2C), no other morphological alterations were detected. Histological analysis demonstrated the presence of clearly distinguishable myocardial and endocardial layers, which were well connected, and normal trabeculation (Fig. 2E,F). Furthermore, foetal erythroblasts were present in blood vessels of both the yolk sac and the embryo itself (Fig. 2F and see Fig. 5). Taken together, no severe abnormalities in development or early organogenesis that might be causal to the lethality at this early stage or to the growth retardation could be detected in the embryo proper. The lacZ reporter gene was used to monitor expression of the targeted Fra1 allele by staining Fra1^+/− embryos for β-galactosidase activity. Expression of lacZ was absent in E9.5 embryos, but was widespread in E10.5 and E11.5 embryos and apparently not restricted to specific organ systems (Fig. 2G-L). Expression was significantly weaker in E10.5 than in E11.5 embryos (Fig. 2I-L).

Embryos lacking Fra1 are severely growth-retarded before they die. To explore the possibility of a proliferation defect of mutant cells in vivo, we measured the number of S-phase cells by 5-bromo-2′-deoxyuridine (BrdU) immunohistochemistry. Pregnant mice from Fra1^+/− intercrosses were injected with BrdU at E9.5, and offspring were analysed 4 h after injection. No significant difference in the number of BrdU-positive nuclei was observed between mutants (48.7±3.6%) and controls (51±3.9%) in the neural tube (Fig. 3A). Similar results were obtained in the placenta and other tissues (data not shown).

To further analyse the proliferation and developmental potential of cells lacking Fra1, Fra1^−/−ES cells were derived from E3.5 blastocysts isolated from Fra1^+/− intercrosses (129/Sv genetic background), and injected into wild-type C57BL/6 blastocysts to generate chimaeric mice. Most chimaeras showed high ES cell contribution, as judged by agouti coat colour, and transmitted the mutant allele to their offspring, indicating that cells lacking Fra1 contributed to male and female gametogenesis. The degree of ES cell contribution to various organs of 3 adult chimaeric mice was determined by glucose phosphate isomerase isozyme and Southern blot analysis. Except in the pancreas and calvaria, Fra1^−/−cells contributed efficiently to most organs, indicating a normal proliferation capacity of these cells in vivo, even when in direct competition with wild-type cells (Fig. 3B).

Next, primary fibroblasts were isolated from E9.5 embryos derived from Fra1^+/− intercrosses, and cell lines were immortalised according to the standard 3T3 protocol (Todaro and Green, 1963). The immortalisation curve of wild-type, heterozygous and mutant cells is shown in Fig. 3C. During the first 5-8 passages (i.e. the first 20-30 days), cells of all three genotypes divided at a significant rate, but subsequently proliferation slowed down until cells entered a senescent state marked by morphological alterations and a cessation of cell proliferation. The crisis of wild-type and mutant fibroblasts lasted approx. 4-5 weeks after which part of the population resumed proliferation and continued to grow as stable cell lines. There was no significant difference between wild-type, Fra1^+/− and Fra1^−/−cells in the initial proliferation rate of pre-crisis primary fibroblasts, the onset and duration of crisis, and the post-crisis proliferation rate of immortalised cells, indicating that Fra1 is not essential for fibroblast proliferation and immortalisation in vitro.

Next, the proliferation of Fra1^−/−ES cells was analysed. Proliferation curves of Fra1^−/−and wild-type ES cells were comparable, indicating that the absence of Fra1 did not affect the in vitro proliferation capacity of ES cells (Fig. 3D). Most importantly, rescued Fra1^−/−foetuses generated via tetraploid blastocyst injection (see below) were of normal size, demonstrating that the growth retardation of mutant embryos was not due to a cell-autonomous proliferation defect.

The absence of severe developmental defects in mutant foetuses together with the observed growth retardation suggests that placental defects may be responsible for the lethality. At E9.0-E9.5, the labyrinth layer of the placenta develops, in which extensive intermingling occurs between maternal blood sinuses and foetal blood vessels (Cross et al., 1994). The first step in the development of the placenta is the fusion of the allantois to the chorion; this process appeared to be unaffected in mutant embryos (Table 3). Thereafter, allantoic vessels invade into the chorionic plate, which is then converted into the labyrinth layer of the definitive placenta. In mutant embryos, this invasion process was drastically reduced, with no embryonic blood vessels found in the rather compact labyrinth layer (Fig. 4E). The vascular endothelial cells remained mostly restricted to the chorioallantoic region (Fig. 4E). The expression of Flk1, a marker for the endothelial cells in the labyrinth (Breier et al., 1995) was almost absent in mutant labyrinth layers, whereas significant expression was detected in the underlying allantoic mesoderm (Fig. 4B,F). In situ hybridisation with gene 4311 and Flt1, which are marker genes for the spongiotrophoblast layer of the placenta, revealed no significant differences between mutants and controls (Fig. 4). Similarly, no defects were apparent in the chorio-allantoic and giant cell layers of the placenta (Fig. 4A,E). Consistent with these observations, analysis of E12.5 and E16.5 Fra1^+/− placentas by X-gal staining revealed expression of Fra1 specifically in the labyrinth layer (Fig. 4J,L).

In addition, mutant embryos exhibited an abnormal yolk sac (Fig. 5). The parietal yolk sac consists of the Reichert’s membrane lined by parietal endoderm cells and of an intermittent layer of trophoblast giant cells situated between the maternal decidual cells and the Reichert’s membrane. In controls, this single layer of trophoblast cells forms a network of channels in close proximity to the Reichert’s membrane that distributes maternal blood throughout the parietal wall of the yolk sac (Fig. 5A). In mutant embryos this network appeared highly disorganised, with trophectoderm cells arranged in several layers and large maternal blood sinuses, which were not in close proximity to the Reichert’s membrane and the parietal endoderm cells (Fig. 5B). Whereas a continuous Reichert’s membrane with normal parietal endoderm cells was found, the visceral yolk sac of mutant embryos was considerably separated from the parietal yolk sac (Fig. 5B). Since these layers are not physically linked, the retraction of the visceral yolk sac might be due to reduced hydrostatic pressure in the amniotic cavity. Fra1 expression was analysed by X-gal staining and was found in trophoblast giant cells, parietal endoderm cells and the visceral endoderm, but not in the mesoderm of the visceral yolk sac (Fig. 5G). Histological analysis of the visceral yolk sac revealed the presence of endodermal and mesodermal layers; however, the visceral endoderm appeared more compact and unstructured compared with controls (Fig. 5C-F). Although blood vessels were present and contained foetal erythrocytes, they were reduced in numbers, and some of them were abnormally enlarged, suggesting that the blood supply to the foetus is impaired (Fig. 5B,D).

As a definitive test of whether the lethality of mutant foetuses is due to defects in the extra-embryonic compartment, Fra1^−/−ES cells were injected into wild-type tetraploid blastocysts (Wang et al., 1997). Tetraploid cells can contribute efficiently to most extra-embryonic tissues but not to the embryo proper (Nagy et al., 1990, 1993; James et al., 1995), and thus should complement defects in the extra-embryonic compartment of mutant foetuses. Following tetraploid blastocyst injection, mutant foetuses were isolated at E13.5, E16.5, E18.5 and P1 (Table 4). In these ‘ES foetuses’, the foetus was exclusively formed by Fra1^−/−cells, whereas the placenta contained predominantly wild-type cells, and the yolk sac consisted of both wild-type and mutant cells (Fig. 6A). Rescued Fra1^−/−foetuses were also obtained by aggregating morulae derived from Fra1^+/− intercrosses with wild-type tetraploid morulae (Table 4). At the indicated times of isolation, all rescued mutant foetuses were viable, appeared phenotypically normal and were not growth retarded, indicating that the lethality and growth retardation of Fra1^−/−foetuses can be rescued by providing wild-type extra-embryonic tissues (Table 4; Fig. 6).

Since there is good evidence from Fra1 transgenic mice for a role of Fra1 in bone cell development (Matsuo et al., 2000; Jochum et al., 2000), skeletons of E18.5 Fra1^−/−rescued foetuses were analysed (Fig. 6B). These had ossification centres of normal size and distribution, and chondrocytes, bone-forming osteoblasts and TRAP-positive multinucleated osteoclasts were present in Fra1-deficient bones (Fig. 6B-D). Thus, normal bone cell differentiation does apparently not require Fra1, at least up to E18.5.

Four Fra1^−/−newborns obtained by Caesarean section or natural delivery survived up to 2 days after birth. Histological analysis revealed no abnormalities of organs derived from rescued E18.5 Fra1^−/−foetuses, including skin (Fig. 6E), brain, heart, cardiac outflow tract, liver, intestine, spleen, kidneys and adrenal glands (data not shown). In addition, no defects were observed in organs to which Fra1^−/−ES cells inefficiently contributed in chimaeric mice (Fig. 3B) such as calvaria, thymus and pancreas. The peripheral blood of E18.5 Fra1^−/−foetuses contained cells of all haemopoietic lineages. However, the lungs of Fra1^−/−rescued foetuses were immature, as indicated by wide and cell-dense alveolar septa, which might be a cause of the early postnatal lethality (Fig. 6F). A similar lung immaturity was present in age-matched foetuses obtained by injection of wild-type ES cells into wild-type tetraploid blastocysts, indicating that the lung defect is likely to be due to the technical procedure (Fig. 6G; compare with Fig. 6H). Thus, the early postnatal lethality of Fra1^−/−rescued mice may be attributable to the experimental approach rather than the lack of Fra1.

The results reported here provide the first genetic evidence that Fra1 plays an essential role in the placentation process. Embryos lacking Fra1 die between E10.0 and E10.5, when the chorio-allantoic placenta has become the principal route of nourishment for the embryo (Cross et al., 1994; Copp, 1995). Morphological analysis suggests that extra-embryonic defects, particularly the failure to establish a functional placenta, are the cause of this lethality. Fra1^−/−placentas lack a properly vascularised labyrinth layer, although the fusion of the allantois to the chorion, which initiates the formation of the labyrinth, is not affected. However, the subsequent differentiation of the allantoic mesoderm into vascular channels is impaired in Fra1^−/−foetuses and leads only to the formation of large vessels in the chorionic plate, which do not invade and sprout into the labyrinth trophoblast. As a consequence, embryonic blood vessels do not intermingle closely with maternal blood sinuses, presumably resulting in an inefficient exchange of nutrients and gas between the maternal and embryonic vascular systems. In agreement with these extra-embryonic defects, Fra1 was found to be expressed in extra-embryonic tissues such as the placental labyrinth layer and the parietal and visceral endoderm, and trophoblast giant cells of the yolk sac, whereas expression was not detectable in the embryo proper before E10.5.

One of the most striking features of Fra1^−/−embryos is their severe growth retardation. Interestingly, this retardation does not seem to be due to a cell-autonomous proliferation defect. Primary fibroblasts and ES cells isolated from mutant embryos exhibit normal proliferation rates in vitro. BrdU labelling and generation of chimaeras via blastocyst injection of Fra1^−/−ES cells did not reveal reduced proliferation capacity of mutant cells in vivo, even when in direct competition with wild-type cells. Most importantly, rescued Fra1^−/−embryos generated via tetraploid blastocyst injection were not growth retarded. These data demonstrate that Fra1^−/−embryos grow to normal size if provided with wild-type extra-embryonic tissues, and indicate that cells lacking Fra1 have a normal proliferative capacity in vitro and in vivo. Thus, the observed growth retardation of Fra1^−/−foetuses is most likely due to the embryo ‘starving to death’ in utero caused by an inadequate supply of nutrients and oxygen, or by poisoning resulting from a lack of outward exchange via the extra-embryonic foeto-maternal exchange organs.

Importantly, the defect in the placental labyrinth alone may not fully explain the severe growth retardation of Fra1^−/−foetuses. A number of other mouse mutants with lethality at the same stage and very similar placentation defects exhibit only mild or no growth retardation, such as embryos lacking Mash2 (Ascl2), Gata2, Gata3 or Ets2 (Guillemot et al., 1994; Ma et al., 1997; Yamamoto et al., 1997). Even some mutant foetuses in which the allantois does not fuse to the chorion are less growth retarded than foetuses lacking Fra1 (Stott et al., 1993; Li et al., 1992; Gurtner at al., 1995; Yang et al., 1995). However, cessation of blood flow within the yolk sac plexus frequently causes foetal growth retardation, such as in embryos lacking HNF4 or trombomodulin (Chen et al., 1994; Healy et al., 1995). In addition, defects in the yolk-sac blood circulation are often associated with the dilatation of the pericardium as an indication of osmotic imbalance within the embryo. Indeed, a large proportion of Fra1 mutant embryos exhibited an enlarged pericardium. Thus, in addition to the placental defects, structural and functional defects in the yolk sac may also contribute to the severe growth retardation and embryonic lethality of Fra1^−/−embryos.

To verify that extra-embryonic defects are causal to the lethality, Fra1^−/−ES cells have been generated and injected into tetraploid wild-type blastocysts (Wang et al., 1997). Since tetraploid cells can efficiently contribute to extra-embryonic tissues but not to the embryo proper, this technique (Wang et al., 1997) has allowed the provision of Fra1-deficient embryos with a wild-type placenta. Unlike Fra1^−/−embryos derived from heterozygote intercrosses, Fra1^−/−foetuses rescued by tetraploid blastocyst injection did not die at E10.0, but completed embryonic development and survived up to 2 days after birth. Thus, the embryonic lethality was fully rescued, demonstrating that the presence of functional Fra1 in the embryo proper is not required for survival during embryogenesis, and that defects in extra-embryonic tissues are most likely to represent the primary cause of lethality. All tetraploid rescued mice died within two days of birth, presumably owing to a pronounced lung immaturity. Importantly, a similar lung immaturity was present in age-matched foetuses obtained by tetraploid blastocyst injection of wild-type ES cells, and is frequently observed with most wild-type or mutant ES cell lines (Wang et al., 1997). Although the early postnatal lethality of Fra1^−/−rescued mice is thus presumably due to the experimental approach, rather than the lack of Fra1, it is possible that this early lethality masks additional important functions of Fra1 in postnatal development.

Interestingly, deletion of Junb, another gene of the AP-1 family, leads to embryonic lethality between E8.5 and E10.0 with a phenotype remarkably similar to that of Fra1^−/−foetuses (Schorpp-Kistner et al., 1999). Some Junb^−/−embryos die around E8.5 and exhibit yolk-sac defects, as well as additional extra-embryonic defects that are more severe than in the Fra1 mutant. Junb^−/−embryos that escape these initial defects die around E10.0 and, like Fra1^−/−embryos, are severely growth retarded, exhibit defects in the yolk sac and a non-vascularised placental labyrinth. These strikingly similar phenotypes suggest that Fra1 and Junb might have overlapping functions during extra-embryonic development, an idea supported by the observations that the temporal and spatial expression patterns of Junb and Fra1 during embryogenesis are partly overlapping (Schorpp-Kistner et al., 1999). Moreover, Junb/Fra1 dimers can regulate the tissue-specific expression of AP-1 target genes (Vallone et al., 1997), and Fra1 is a potential transcriptional target of Junb. It has been found that Junb can repress transactivation of some promoter constructs with single AP-1 binding sites, but activate analogous reporter constructs containing multimeric AP-1 sites (Deng and Karin, 1993). Intriguingly, a major DNA element regulating Fra1 expression, which is located in the first intron, consists of a cluster of 3 AP-1 sites (Bergers et al., 1995), which might be a target for activation by Junb. The embryonic lethality of Junb^−/−embryos can be rescued by crosses with transgenic mice ubiquitously expressing Junb under the control of the human ubiquitin C promoter (Ubi-Junb transgene; Schorpp et al., 1996; Schorpp-Kistner et al., 1999). Since the phenotypes caused by Junb or Fra1 deficiency are very similar, we attempted to rescue the lethality of Fra1 mutants by using the Ubi-Junb transgene. Interestingly, out of 284 offspring, 4 Fra1^−/−mice carrying the Ubi-Junb transgene were obtained, whereas 102 offspring were Fra1^+/+ and 178 were heterozygous; 57.5% of the Fra1^+/+ and Fra1^+/− mice carried the transgene. The four rescued Fra1^−/−mice were healthy, phenotypically normal and had a normal lifespan. Thus, deregulated expression of Junb can complement at low frequency for the lack of Fra1, further suggesting that Fra1 and Junb have overlapping functions during embryonic development.

The extra-embryonic defects of Fra1^−/−foetuses could be due to defective tissue remodelling and/or cell-cell and cell-matrix interaction. The expression of several extracellular matrix (ECM) proteinases appears to be regulated in part by AP-1 (Angel and Karin, 1991; Hennigan et al., 1994; Schorpp-Kistner et al., 1999). Some of these proteases are expressed during early embryogenesis and play an important role in the invasive processes involved in establishing foeto-maternal interactions, such as MMP-9 and uPA (Birkedal-Hansen et al., 1993; Groskopf and Linzer, 1994). Similar to Junb (Schorpp-Kistner et al., 1999), deletion of Fra1 might alter AP-1 activity and consequently expression of ECM proteases, which in turn might lead to a lack of invasive processes required for the formation of the labyrinth layer, and to the observed aberrant structure of the parietal and visceral yolk sac of mutant conceptuses. Alternatively, defects in vasculogenesis or angiogenesis could explain the dilated yolk-sac blood vessels, as well as the lack of invasion and sprouting of vascular channels into the placental labyrinth. Targeted mutagenesis has confirmed the crucial roles of VEGF and its receptors Flk1 (Kdr) and Flt1 in these processes (Fong et al., 1995; Shalaby et al., 1995; Carmeliet et al., 1996; Ferrara et al., 1996). In Fra1^−/−foetuses, a vascularisation defect leads to a non-vascularised labyrinth layer as demonstrated by the lack of Flk1 expression. Large, unusually shaped foetal blood vessels are present in the chorionic plate, but are unable to invade and sprout into the labyrinth trophoblasts. This defect, as well as the defect in yolk-sac vascularisation closely resembles the defects in Junb^−/−extra-embryonic tissues (Schorpp-Kistner et al., 1999), which may involve the same molecular mechanism.

Perhaps the most surprising result of the present study is that deletion of Fra1 has significantly more deleterious effects than inactivation of Fos, the ‘prototype’ Fos family member. Mice lacking Fos are osteopetrotic, but most survive to adulthood (Wang et al., 1992; Johnson et al, 1992; Grigoriadis et al., 1994). Since Fra1 lacks a transactivation domain (Wisdom and Verma, 1993; Bergers et al., 1995), it is presumed to have only a subset of the functions of the potent transcriptional activator Fos. Why then does Fos not compensate for the lack of Fra1? Fos is expressed in most extra-embryonic tissues, including the spongiotrophoblast and the labyrinth layer of the placenta, visceral endoderm and mesoderm, and amnion, at levels that are several-fold higher than in the embryo proper (Müller et al., 1983). Thus, Fos is apparently expressed in the right tissues; however, it is conceivable that the onset of expression is too late to rescue the lethality of Fra1^−/−embryos. Significant expression of Fos in the placenta was detected at E9.5, whereas extra-embryonic membranes were only analysed at E12.5 and later (Müller et al., 1983). Alternatively, a crucial function for Fra1 in embryonic development might be transcriptional inhibition rather than activation. Because Fos, unlike Fra1, has a potent transactivation domain, a putative inhibitory role of Fra1 could not be compensated for by Fos.

In addition, Fra1 is a transcriptional target of Fos (Bergers et al., 1995; Schreiber et al., 1997; Matsuo et al., 2000). However, Fra1 expression during embryonic development must be, at least in part, Fos independent, as Fos^−/−mice develop to term, survive to adulthood and do not have the same embryonic lethal phenotype as Fra1^−/−foetuses. In agreement with this idea, an early Fos-dependent and a late Fos-independent phase of Fra1 expression were observed in 3T3 fibroblasts upon serum stimulation (Schreiber et al., 1997). In contrast, Fos^−/−osteoclast precursors are viewed as ‘virtual double knockouts’ of Fos and Fra1, as Fra1 expression is largely Fos dependent in these cells (Matsuo et al., 2000). Therefore, endogenous Fra1 does not compensate for the lack of Fos in Fos^−/−osteoclasts, although ectopically expressed Fra1 can fully rescue the Fos^−/−osteopetrotic phenotype and osteoclast differentiation block (Matsuo et al., 2000). Furthermore, Fra1 potentiates osteoclastogenesis (Owens et al., 1999). It was therefore of interest to determine whether Fra1 is necessary for osteoclast development. Using E18.5 rescued Fra1^−/−embryos obtained by tetraploid blastocyst injection, we could show the presence of multinucleated osteoclasts and the formation of normal bone marrow cavities in the absence of Fra1. Thus, unlike Fos, Fra1 is not essential for osteoclast formation, at least in newborn mice. In the future, it will be interesting to determine, by tissue-specific conditional inactivation of the Fra1 gene in bone and endothelial cells, whether lack of Fra1 leads to skeletal and vascular defects in adult mice.

We are grateful to H.-C. Theussl for tetraploid blastocyst injection, and to M. Schorpp-Kistner, K. Matsuo and the members of the Wagner group for many helpful discussions and advice. We thank W. Risau and F. Guillemot for kindly providing probes for in situ hybridization, and D. Barlow, J.-P. David, A. Fleischmann, K. Matsuo, E. Passegue, M. Schorpp-Kistner and M. Sibilia for critical and helpful comments on the manuscript. This work was supported by the Austrian Federal Ministry of Science, Transport and the Arts (S0074-MOB).