Chimeric analysis of fibroblast growth factor receptor-1 (Fgfr1) function: a role for FGFR1 in morphogenetic movement through the primitive streak

Recently, much attention has been directed toward possible roles for FGF signaling in early vertebrate development, especially with respect to the patterning of early mesodermal and neural lineages. Studies in Xenopus laevis have provided compelling evidence that FGF signaling is directly involved in both of these processes. Exogenous basic FGF, when added to animal cap explants, is able to induce mesoderm from tissue normally fated to become ectoderm (Kimelman and Kirschner, 1987; Paterno et al., 1989; Slack et al., 1989). The introduction of a dominant negative FGF receptor into Xenopus embryos at various stages of development has demonstrated that FGF signaling is required for the expression of several early mesodermal markers, for the induction of posterior and ventral mesoderm, and for proper mesodermal maintenance during gastrulation (Amaya et al., 1991, 1993; Kroll and Amaya, 1996). Also, overexpression of embryonic FGF (eFGF) in Xenopus embryos has recently been shown to both upregulate and anteriorly extend the expression domains of posterior homeobox genes (Pownall et al., 1996). These and other recent findings which demonstrate that FGFs can induce posterior neural differentiation suggest that FGF signaling may be involved in the proper patterning of the developing anteriorposterior axis (Cox and Hemmati-Brivanlou, 1995; Kengaku and Okamoto, 1995; Lamb and Harland, 1995).

In mouse embryogenesis, mesoderm induction and patterning are thought to occur at gastrulation. The beginning of gastrulation is manifested by the formation of the primitive streak at the posterior end of the embryo. In a process of highly integrated cell and tissue movements, cells delaminate through the primitive streak and emerge between the epiblast and visceral endoderm to form the mesodermal germ layer. Fate-mapping of the mouse epiblast suggests that cells emerging from different anterior-posterior levels of the streak adopt different fates (Tam, 1989; Lawson et al., 1991; Beddington et al., 1992; Parameswaran and Tam, 1995). FGFs 3, 4, 5 and 8 are expressed within the primitive streak in spatial and temporal patterns which would be consistent with a role for FGF signaling in mesoderm induction or in regulating processes of fate determination (Wilkinson et al., 1988; Haub and Goldfarb, 1991; Niswander and Martin, 1992; Hebert et al., 1991).

Mutational analyses of Fgf3 and Fgf5 have failed to show defects in early gastrulation (Mansour et al., 1993; Hebert et al., 1994). The apparent ‘expendable’ nature of FGF3 and FGF5 function in early embryonic development is likely due to redundancy in FGF signaling as there is cross-specificity in ligandreceptor interactions (Ornitz et al., 1996), and local coexpression of other FGFs may therefore compensate for the early losses of function (Givol and Yayon, 1992). Targeted disruption of Fgf4 results in abortive postimplantation development (Feldman et al., 1995). While this implicates a role for FGF4 in the survival and growth of the inner cell mass during postimplantation development, lethality occurs too early to address the role of FGF4 signaling in mesoderm induction and patterning.

Analysis of FGF function at the level of the receptor has proved more informative. Fgfr1, for example, is first expressed throughout the primitive ectoderm. In mid-streak staged embryos, Fgfr1 expression is concentrated in the posterior mesoderm lateral to the primitive streak and is maintained in the migrating mesodermal wings; headfold stage embryos show strong expression in both neurectoderm and the developing paraxial and pre-somitic mesoderm (Orr-Urtreger et al., 1991; Yamaguchi et al., 1992). Thus, Fgfr1 expression is consistent with a role for FGFs in mesoderm induction, neural induction, and somite formation and patterning. Targeted mutation of Fgfr1 results in embryonic lethality between day 7.5 and 9.5 of development, with defects first manifesting themselves at the onset of gastrulation (Yamaguchi et al., 1994; Deng et al., 1994). Retarded development of homozygous Fgfr1 mutant embryos suggests an early mitogenic role for FGFR1, and thickening of the posterior streak suggested that Fgfr1 mutant cells were defective in migrating from the posterior primitive streak. Paraxial mesoderm of Fgfr1 mutant embryos was much reduced, yet axial mesoderm was still present and possibly expanded, suggesting that FGFR1 may be involved in the allocation of progenitor cells among alternative pathways of mesoderm differentiation (Yamaguchi et al., 1994; Deng et al., 1994). Although it was clear that FGFR1 may play a role in mesodermal patterning and perhaps cell migration away from the posterior streak, the phenotype was complex and it was difficult to distinguish primary defects associated with the Fgfr1 mutation from secondary defects resulting from grossly abnormal morphogenetic movements at gastrulation.

To further elucidate the primary modes of FGFR1 function, a chimeric analysis has been performed. A population of homozygous fgfr1^Δtmk embryonic stem (ES) cells was established and aggregated with wild-type (WT) diploid morulae to generate chimeric embryos. Careful analysis of the behaviour and distribution of fgfr1^Δtmk/fgfr1^Δtmk cells within chimeric embryos revealed that defects in mesodermal patterning may be secondary to an initial deficiency in the ability of mutant cells to traverse the primitive streak. The chimeric analysis also revealed that Fgfr1 mutant cells which have accumulated in the streak display an inherent tendency to form secondary neural tubes. A role for FGFR1 in regulating the morphogenetic movements which are thought to mediate cell fate determination and lineage restriction at gastrulation, is discussed.

The fgfr1^Δtmk allele used in this study has been previously described (Yamaguchi et al., 1994). Briefly, exons 8-14 of the Fgfr1 locus, which encode the transmembrane domain and most of the catalytic kinase domain, were replaced by homologous recombination with PGKneo; this should effectively eliminate catalytically active isoforms of the receptor. A second, independently derived mutation at the Fgfr1 locus which truncated all major isoforms of the gene at the second extracellular immunoglobulin domain resulted in a similar mutant phenotype to that of fgfr1^Δtmk/fgfr1^Δtmk embryos (Deng et al., 1994). Therefore, it is believed that both Fgfr1 mutations are effectively null alleles.

To isolate ES cell lines homozygous for the fgfr1^Δtmk mutation and marked with a ubiquitously expressed lacZ marker, fgfr1^Δtmk/+ mice (129 background) were first crossed with ROSA26 mice of a mixed background (Friedrich and Soriano, 1991). F₁ progeny were genotyped for the Fgfr1 locus by PCR (Yamaguchi et al., 1994). fgfr1^Δtmk/+; ROSA26 lacZ +/− mice were intercrossed, and the day that vaginal plugs were detected was designated embryonic day 0.5 (E0.5). Blastocysts were flushed, with M2 medium, from the uterine horns of pregnant females at E3.5 (as described by Hogan et al., 1994), and zona pellucidae were removed by treatment with acid Tyrode’s solution. Blastocysts were transferred to individual 10 mm wells containing a preformed feeder layer of mitotically inactivated fibroblasts and 1 ml of culture medium (Dulbecco’s modified Eagle’s medium [DMEM] supplemented with 100 μM non-essential amino acids, 1 mM sodium pyruvate, 1 μM β-mercaptoethanol, 2 mM L-glutamine, 15% fetal calf serum and 1× leukaemia inhibitory factor). Embryos were cultured at 37°C, 5% CO₂ for 4-5 days to allow for inner cell mass (ICM) outgrowth. ICM outgrowths were then disaggregated into small, multi-cellular clumps by mouth pipetting with a series of finely drawn out Pasteur pipettes. Disaggregated outgrowths were cultured further until ES cell colonies developed. Colonies were picked and subcultured to establish cell lines. Cell lines were genotyped for the Fgfr1 locus by Southern analysis (Yamaguchi et al., 1994) and the presence of the ROSA26 lacZ transgene was determined by β-galactosidase staining of ES cell colonies.

To assess the developmental potential of the ES cell lines generated, individual lines were aggregated with two tetraploid CD-1 embryos (Nagy et al., 1993). Previous studies have demonstrated that tetraploid cells do not contribute to fetal tissues of tetraploid embryo ↔ ES cell chimeras, therefore these chimeric embryos will be completely ES cell-derived (Nagy et al., 1993). Tetraploid CD-1 embryos were produced by the electrofusion of embryos at the two cell stage. Aggregates were transferred into the uteri of CD-1 foster mothers, chimeric embryos were dissected at early to mid-gestational stages, and developmental potential of each cell line assessed by gross phenotypic observation.

Diploid chimeric embryos were generated by aggregating an 8-10 cell clump of either fgfr1^Δtmk/+ or fgfr1^Δtmk/fgfr1^Δtmk ES cells with CD-1 8-cell embryos, using the standard morulae aggregation technique (Nagy et al., 1993). Aggregates were transferred into the uteri of CD-1 foster mothers, chimeric embryos were dissected at early to mid-gestational stages, and the contribution of ES cells and CD-1 cells to the embryo was determined by whole-mount β-galactosidase staining.

Embryos (or ES cells) were rinsed in 100 mM sodium phosphate pH 7.3, and then fixed in 0.2% gluteraldehyde, 2 mM MgCl₂, 5 mM EGTA, 100 mM sodium phosphate pH 7.3 at room temperature for 15-30 minutes depending on size of embryo (or 5 minutes for ES cells). Embryos were then washed 3 times in wash buffer at room temperature (0.02% NP-40, 0.01% deoxycholate, 2 mM MgCl₂, 100 mM sodium phosphate pH 7.3) for 15 minutes each (ES cells were washed 3× 5 minutes). Embryos or ES cells were stained in 1 mg/ml X-gal, 5 mM K3Fe(CN)₆, 5 mM K4Fe(CN)₆, 0.02% NP-40, 0.01% deoxy-cholate, 2 mM MgCl₂, 100 mM sodium phosphate pH7.3 at 37°C overnight. Samples were then rinsed with wash buffer, rinsed in PBS, and postfixed overnight in 3.7% formaldehyde at 4°C.

After whole-mount β-galactosidase staining, chimeric embryos were photographed, dehydrated through an ethanol series and embedded in paraffin. Paraffin blocks were sectioned at 5 μm, sections were mounted onto glass slides, dewaxed and counterstained with Nuclear Fast Red.

Whole-mount in situ hybridization was performed as described by Conlon and Rossant (1992). For sectioning of whole-mount stained embryos, specimens were postfixed in 3.7% formaldehyde overnight at 4°C. Embryos were embedded in paraffin, sectioned at 10 μm, mounted onto glass slides, dewaxed, and photographed using Nomarski optics.

The probes used for the whole-mount in situ hybridization studies were as follows: Brachyury (Herrmann, 1991), Shh (Echelard et al., 1993), Sox-1 and Sox-2 (unpublished, obtained from Robin LovellBadge), Pax3 (Goulding et al., 1991), Pax6 (Walther and Gruss, 1991), mox-1 (Candia et al., 1992), twist (Wolf et al., 1991), and snail (Smith et al., 1992). The Ncam probe (obtained from Cynthia Faust) consisted of a 1 kb cDNA fragment spanning exons 3-9 (which contains nearly all IgG domains), cloned into the BamHI site of pBS+ (Stratagene) phagemid vector.

ES cell lines were generated from the ICMs of blastocysts isolated from crosses of fgfr1^Δtmk/+; ROSA26 lacZ+/− animals (Fig. 1A). lacZ expression via the ROSA26 transgene is ubiquitous throughout early to mid gestational development (Friedrich and Soriano, 1991), and is therefore a useful marker of mutant cells in chimeric embryos.

From approximately 90 blastocysts cultured, 9 ES cell lines were recovered. It was expected that threequarters of the isolated cell lines would be positive for lacZ expression (as determined by β-galactosidase staining), and that a quarter of the cell lines would be homozygous for the Fgfr1 mutation. Southern blot analysis revealed that seven ES cell lines were homozygous for fgfr1^Δtmk, two cell lines were fgfr1^Δtmk/+, and that no +/+ cell lines were recovered (Fig. 1B,C). Therefore, in deriving ES cell lines from primary blastocyst cultures, there appeared to be a positive selection for fgfr1^Δtmk/fgfr1^Δtmk cell lines. This selection for homozygous fgfr1^Δtmk cells was statistically significant (P<0.005; Fig. 1C).

129 fgfr1^Δtmk/+ mice were initially crossed with ROSA26 mice of a mixed background before generating ES cell lines, therefore the fgfr1^Δtmk allele was carried on a 129-derived chromosome. Since isolation of ES cells is more efficient on a 129 background (McWhir et al., 1996 and references therein), Fgfr1 may be linked to a 129-specific polymorphism which promotes the maintenance, and hence derivation of fgfr1^Δtmk/fgfr1^Δtmk ES cell lines. Another explanation for the positive selection of homozygous fgfr1^Δtmk lines is that FGFR1 signaling may promote differentiation from a pluripotent state. In this case, fgfr1^Δtmk/fgfr1^Δtmk ES progenitors would have a selective advantage in maintaining pluripotency and hence in being isolated as an undifferentiated cell line.

fgfr1^Δtmk/fgfr1^Δtmk ↔ tetraploid CD-1 chimeras were generated using cell lines B2, C5B, C15 and A3, in order to assess the developmental potential of these cell lines. Tetraploid chimeras were examined at E8.5 and E9.5 of development. At E8.5, chimeras were developmentally delayed in comparison to similarly staged WT embryos. These embryos showed abnormal thickening of the posterior primitive streak, poorly developed head folds, an absence of somites, and ruffling of the extraembryonic endoderm. At E9.5, chimeras were poorly developed and showed an expanded midline, an absence of somites, small malformed head folds, and kinking of the posterior neural tube. All of these properties are characteristic of homozygous fgfr1^Δtmk embryos (Yamaguchi et al., 1994), indicating that the developmental potential of the fgfr1^Δtmk/fgfr1^Δtmk cell lines had not been altered. Furthermore, it has been shown that the trophectoderm and primitive endoderm lineages of tetraploid embryo ↔ ES cell chimeras are strictly embryo-derived (Nagy et al., 1993). Since fgfr1^Δtmk/fgfr1^Δtmk ↔ tetraploid CD-1 chimeras phenocopy homozygous fgfr1^Δtmk embryos, then defects associated with the fgfr1^Δtmk mutation are not caused by abnormal development of these tissues. Therefore, all defects observed in the Fgfr1 homozygous mutant embryos can be explained by the abnormal development, morphogenesis and patterning of the extraembryonic mesoderm and definitive germ layers.

Cell lines B2 and C5B (which expressed the ROSA26 lacZ transgene) were used in subsequent diploid chimeric analyses. Since both cell lines produced similar results, data for these analyses have been pooled. Tetraploid aggregations were also performed with ES cell line C16 (fgfr1^Δtmk/+, lacZ positive) in order to test its pluripotency. Normal C16 derived embryos were recovered at E13.5, and therefore this fgfr1^Δtmk/+ cell line was used as a positive control in subsequent diploid chimeric analyses.

Diploid chimeric embryos were generated by aggregating approximately eight fgfr1^Δtmk/+ or fgfr1^Δtmk/fgfr1^Δtmk ES cells with wild-type (WT) CD-1 embryos. At E9.5, fgfr1^Δtmk/+ ↔ CD-1 chimeras showed great variations in the extent of ES cell contribution to the embryo, with values ranging from <5% to 100% ES cell derived, as estimated by β-galactosidase staining (Fig. 2A,C,E). A similar range in ES contribution was observed in E9.5 fgfr1^Δtmk/fgfr1^Δtmk ↔ CD-1 chimeras (Fig. 2B,D,F). This argues against a broad requirement for FGFR1 in early mitogenesis, since a major impairment in growth and proliferation would have been expected to reduce the overall contribution of fgfr1^Δtmk/fgfr1^Δtmk cells to chimeric embryos.

Some phenotypically normal fgfr1^Δtmk/fgfr1^Δtmk ↔ CD-1 chimeras were recovered at E10.5 (Fig. 3), indicating that WT cells could rescue the embryonic lethality and gastrulation defects associated with homozygous fgfr1^Δtmk embryos. Viable chimeric embryos were also recovered at E16.5, however a complete analysis at late gestation was beyond the scope of this study. When the overall mutant cell contribution exceeded about 50%, fgfr1^Δtmk/fgfr1^Δtmk ↔ CD-1 chimeric embryos developed gross phenotypic defects which ranged from apparent posterior neural tube duplications and ruffling of the dorsal ectoderm/neuroepithelium to posterior truncations, abnormal heart development, and failure in anterior neural tube closure (Fig. 3C-G). Similar defects were not observed in fgfr1^Δtmk/+ ↔ CD-1 chimeras (data not shown). Chimeric embryos that were completely fgfr1^Δtmk/fgfr1^Δtmk cell derived phenocopied homozygous fgfr1^Δtmk embryos (Fig. 2F).

Whole-mount β-galactosidase staining of fgfr1^Δtmk/fgfr1^Δtmk ↔ CD-1 chimeric embryos revealed striking patterns in mutant cell dispersal: fgfr1^Δtmk/fgfr1^Δtmk cells were deficient in contributing to the heart, and showed a strong tendency to accumulate within the tail of chimeric embryos (Figs 2D, 3A-C). To further investigate the distribution and behaviour of mutant cells within a chimeric background, 22 E9.5-10.5 fgfr1^Δtmk/fgfr1^Δtmk ↔ CD-1 chimeric embryos, which showed varying degrees of ES-derived contribution, were histologically sectioned. All chimeras chosen appeared phenotypically normal, with the possible exception of posterior neural tube duplications. Fig. 4 depicts the level of ES cell contribution to various tissues of individual fgfr1^Δtmk/+ or fgfr1^Δtmk/fgfr1^Δtmk↔ CD-1 chimeric embryos. The anterior neural tube was used as a ‘standard’ for assessing the original ES-derived cell contribution to the epiblast since this tissue does not ingress through the primitive streak.

In fgfr1^Δtmk/fgfr1^Δtmk ↔ CD-1 chimeric embryos, mutant cells contributed extensively to ectodermally derived tissues without causing gross morphological defects. The neural tube, surface ectoderm, and neural crest derivatives such as the branchial arches and dorsal root ganglia, were equally well populated by mutant cells (Figs 4, 5). Interestingly, fgfr1^Δtmk/fgfr1^Δtmk cells showed a tendency to aggregate into distinct patches in the neuroepithelium and nasal placodes (Fig. 3E). The level of mutant cell contribution to the limb bud mesenchyme and lateral mesoderm of fgfr1^Δtmk/fgfr1^Δtmk ↔ CD-1 chimeras was similar to that found in the neural tube, but fgfr1^Δtmk/fgfr1^Δtmk cells appeared to accumulate in the mesenchyme of the tail (Figs 3A,B, 4).

In contrast, mutant cells were consistently under-represented in anterior mesoderm and endodermal lineages of fgfr1^Δtmk/fgfr1^Δtmk ↔ CD-1 chimeric embryos (Fig. 4). Few mutant cells were observed in the cephalic, heart and somitic mesoderm populations (Fig. 5), and mutant cells were almost completely absent from the gut of fgfr1^Δtmk/fgfr1^Δtmk ↔ CD-1 chimeras (Fig. 4). fgfr1^Δtmk/fgfr1^Δtmk cells were also under-represented in the notochord of these chimeras, and contributed only to notochord of the posterior trunk and tail. Similar deficiencies in ES-derived contribution were not observed in fgfr1^Δtmk/+ ↔ CD-1 chimeras, although there was a slight bias against fgfr1^Δtmk/+ cell contribution to the endoderm (Fig. 4). This might indicate a dose-dependent requirement for FGFR1 in endoderm formation, or it may reflect an inherent bias against ES-derived contribution to the endoderm of ES cell ↔ embryo chimeras (see Discussion).

Several explanations exist for the observed deficiencies in fgfr1^Δtmk/fgfr1^Δtmk cell contribution to the anterior mesoderm and endoderm of chimeric embryos. Mutant cells may have initially colonized these lineages, but either failed to proliferate, or underwent cell death. Alternatively, mutant cells may have been precluded from contributing to these lineages at the time of their inception at gastrulation. To differentiate between these possibilities, chimeric embryos were examined at earlier stages of development.

Whole-mount β-galactosidase staining of early streak-stage fgfr1^Δtmk/+ ↔ CD-1 chimeric embryos demonstrated a random distribution of ES-derived cells. β-galactosidase staining appeared more intense in posterior-lateral portions of the embryo because of the extra cell layer formed by the advancing mesodermal wings (Fig. 6A,B). In contrast, fgfr1^Δtmk/fgfr1^Δtmk↔ CD-1 chimeras showed striking patterns in mutant cell distribution: fgfr1^Δtmk/fgfr1^Δtmk cells accumulated along the primitive streak and were under-represented in the mesodermal wings, node and pre-chordal mesoderm (Fig. 6C,D). 20 E7.5 fgfr1^Δtmk/fgfr1^Δtmk ↔ CD-1 chimeric embryos which showed a large range in ES-derived cell contribution (from <5% to 95%) were chosen for histological analysis.

Serial sections confirmed that fgfr1^Δtmk/fgfr1^Δtmk cells were accumulating within the ectoderm and posterior mesenchyme of the primitive streak (Fig. 7). These fgfr1^Δtmk/fgfr1^Δtmk accumulations could be observed even in those chimeras which showed only minimal (<5%) ES-derived cell contribution (Fig. 7A-C). In very strong fgfr1^Δtmk/fgfr1^Δtmk ↔ CD-1 chimeras, patches of WT cells appeared to be selectively delaminating from an almost completely mutant epiblast layer (Fig. 7K,L). This suggests that fgfr1^Δtmk/fgfr1^Δtmk cells were at a competitive disadvantage in traversing the primitive streak.

Although mutant cells contributed strongly to the epiblast of chimeric embryos, there was a striking deficit of fgfr1^Δtmk/fgfr1^Δtmk cells in the mesodermal wings (Fig. 7D-F, G-I). This defect was not absolute; as the fgfr1^Δtmk/fgfr1^Δtmk cell contribution increased, more mutant mesoderm cells were observed (Fig. 7G-I, J-L, M-O). fgfr1^Δtmk/fgfr1^Δtmk mesodermal cells tended to accumulate along the posterior streak (Fig. 7I,L), giving the impression of the thickened streak characteristic of homozygous fgfr1^Δtmk embryos. This mutant mesodermal population also appeared less able to migrate laterally around the embryo, since the leading edges of the mesodermal wings were composed largely of WT cells (Fig. 7K,O). Even in chimeras with over 95% fgfr1^Δtmk/fgfr1^Δtmk cell contribution, WT cells formed the leading edge of the mesodermal wings; mutant cells accumulated within ‘pockets’ formed by abnormal buckling of the epiblast layer (Fig. 7O).

Headfold to early somite-stage fgfr1^Δtmk/fgfr1^Δtmk ↔ CD-1 chimeric embryos showed accumulations of mutant cells within the primitive streak and allantois (Fig. 6E,F). A deficiency in mutant cell contribution to the paraxial and presomitic mesoderm was also observed; similar trends were not observed in the fgfr1^Δtmk/+ ↔ CD-1 chimeras (data not shown). 11 E8.5 fgfr1^Δtmk/fgfr1^Δtmk ↔ CD-1 chimeras with varying degrees of mutant cell contribution (<5 to 70%) were histologically analyzed, and all showed a deficit of fgfr1^Δtmk/fgfr1^Δtmk cells in the cephalic mesenchyme, heart primordia and paraxial mesoderm populations (Fig. 8). Even chimeras which had very high fgfr1^Δtmk/fgfr1^Δtmk cell contribution to ectodermal lineages (60–70%) showed a near absence of mutant cells from anterior mesoderm (Fig. 8D). These chimeras appeared morphologically abnormal with kinked and misshapen neural folds.

fgfr1^Δtmk/fgfr1^Δtmk ↔ CD-1 chimeric embryos with minimal (<5%) ES-derived cell contribution showed accumulations of mutant cells within the epiblast layer at the posterior streak (Fig. 8A-C). Mutant cells accumulated as a distinct ‘block’ of columnar epithelial cells, with WT cells forming mesenchyme on both sides of this accumulation (Fig. 8B). Also, in chimeras with high levels of ESderived cell contribution, patches of WT cells could be seen delaminating from an almost entirely mutant epiblast layer (Fig. 8F,G). This suggests that fgfr1^Δtmk/fgfr1^Δtmk cells are deficient in undergoing the epithelial to mesenchymal transition (EMT) which occurs at the streak. Despite apparent deficiencies in traversing the streak, fgfr1^Δtmk/fgfr1^Δtmk cells contributed strongly to the mesoderm of more posterior regions of the embryo and mutant cells were well represented in lateral mesoderm. All fgfr1^Δtmk/fgfr1^Δtmk ↔ CD-1 chimeras examined demonstrated strong ESderived cell contribution to the allantois; the level of mutant cell contribution was highest at the base of the allantois and decreased distally (Fig. 6E,F). The remainder of the extra-embryonic mesoderm showed a striking deficiency in mutant cell colonization. fgfr1^Δtmk/fgfr1^Δtmk cells were occasionally observed in the blood islands in the mesodermal component of the yolk sac, but only in blood islands situated along the posterior midline (not shown).

At E9.5-10.5, fgfr1^Δtmk/fgfr1^Δtmk ↔ CD-1 chimeras showed caudal accumulations of mutant cells. Histological sections through the tails of such chimeras revealed that fgfr1^Δtmk/fgfr1^Δtmk cells appeared to form secondary neural tubes which were entirely mutant cell derived (Figs 5G, 9). 19 of the 22 fgfr1^Δtmk/fgfr1^Δtmk ↔ CD-1 chimeras examined had developed these secondary structures (Fig. 4); of the 3 chimeras which did not, 2 displayed only minimal ES-derived cell contribution.

To rule out the possibility of β-galactosidase stain seeping into adjacent WT cells, thus masking their contribution to secondary neural tubes, an aggregation was performed between ROSA26 embryos and the C15 ES cell line (fgfr1^Δtmk/fgfr1^Δtmk; lacZ negative). In this case, secondary neural tubes did not show β-galactosidase staining, confirming that ectopic neural tubes were completely fgfr1^Δtmk/fgfr1^Δtmk cell derived (data not shown).

Secondary neural tubes ranged in size from small structures which appeared to ‘pinch off’ from the primary neural tube (NT) in distal portions of the tail, to structures comparable in size to the primary NT, extending caudally from the hind limb bud level. The larger, secondary neural tubes originated from entirely fgfr1^Δtmk/fgfr1^Δtmk cell-derived expansions of the neural plate (Fig. 9), and extended laterally from the central neural groove. These ectopic neural plates were excluded from primary NT closure, and folded separately into secondary neural tubes (Fig. 9).

To verify the neural identity of these ectopic tubes, whole-mount RNA in situs were performed using Sox-1, Sox-2 and Ncam probes. Sox-1 and Sox-2 are SRY related genes; Sox-1 expression is restricted to neurectodermal lineages, and Sox-2 is initially expressed throughout the ectoderm and is later restricted to neuroepithelia (Collignon et al., 1996). Both Sox-1 (Fig. 10A) and Sox-2 (data not shown) were expressed in the primary and secondary neural tubes. NCAM, a neural cell adhesion molecule, was also expressed in both secondary and primary neural tubes, and showed normal expression in the surrounding somitic mesoderm (Fig. 10B). In situs were also performed using mox-1, twist, snail and Brachyury probes to ensure that the secondary neural tubes demonstrated no elements of mesodermal identity; mesoderm marker expression was not observed (not shown). Secondary neural tubes were often surrounded by ill-defined fgfr1^Δtmk/fgfr1^Δtmk mesenchymal tissue (Fig. 9); the pattern of mox-1 expression in the tails of chimeric embryos suggests that this poorly organized mesenchymal tissue has paraxial mesoderm identity (Fig. 10E). Lack of Brachyury staining associated with the secondary NT verified histological analyses which showed that secondary notochords were not associated with the ectopic neural tubes. Dorsal-ventral patterning of the secondary NT was also examined with Pax3 and Pax6, and shh probes. As expected, the secondary neural tubes, which lacked notochords, did not show floor plate expression of shh (data not shown). Pax3 is normally expressed in the dermamyotome and dorsal neural tube; Pax6 expression is restricted to lateral walls of the neural tube. Both Pax6 and Pax3 appeared to be expressed in their proper domains (Fig. 10C and D, respectively).

Chimeric analysis has furthered our understanding of the developmental processes affected by the fgfr1^Δtmk mutation. fgfr1^Δtmk/fgfr1^Δtmk ↔ tetraploid CD-1 aggregations demonstrated unequivocally that defects associated with the mutated Fgfr1 allele are intrinsic to definitive embryonic germ layers and to the extra-embryonic mesoderm. In fgfr1^Δtmk/fgfr1^Δtmk↔ diploid CD-1 chimeras, mutant cells contributed strongly to ectodermal lineages, and this argues against an early role for FGFR1 signaling in the development of the ectoderm and its derivatives. Furthermore, since the epiblasts of fgfr1^Δtmk/+ and fgfr1^Δtmk/fgfr1^Δtmk ↔ CD-1 chimeras showed similar ranges in ES-derived cell contribution, this argues against a strong mitogenic role for FGFR1 in early development (although a specific role in mesoderm proliferation cannot be completely discounted).

This study has also provided further insight into the role of FGFR1 in the morphogenesis and patterning of mesoderm at gastrulation. fgfr1^Δtmk/fgfr1^Δtmk ↔ CD-1 chimeras demonstrated strong deficiencies in the ability of mutant cells to populate not only the paraxial mesoderm (as already observed in the mutant analyses), but also the extra-embryonic, cephalic, heart and axial mesoderm, plus the endodermal lineages of E9.5-10.5 chimeric embryos. Some mutant cells were found in all lineages, and this confirms previous studies into the differentiation capacity of FGFR1 deficient ES cell lines as teratomas, which suggested that FGFR1 is not required in a cell-autonomous manner for cell fate specification (Deng et al., 1994). Rather, the chimeric analysis suggests that FGFR1 is required for correct morphogenetic movement of mesodermal progenitor cells through the primitive streak at gastrulation. In fgfr1^Δtmk/fgfr1^Δtmk ↔ CD-1 chimeras, fgfr1^Δtmk/fgfr1^Δtmk cells are deficient at traversing the primitive streak. Fgfr1 mutant cells accumulate along the posterior midline and consequently fail to populate the mesodermal wings which have been identified as the source of presumptive extra-embryonic, heart and cephalic mesoderm (Parameswaran and Tam, 1995).

Axial and paraxial mesoderm are thought to be produced from stem-cell populations residing within the node and anterior portions of the primitive streak (Beddington, 1994; Tam and Beddington, 1987; Lawson et al., 1991; Nicolas et al., 1996). These progenitor populations form early in gastrulation, after extra-embryonic, cephalic and heart mesoderm progenitors have exited the streak (Lawson et al., 1991). Since fgfr1^Δtmk/fgfr1^Δtmk cells appear to be defective at delaminating from the primitive streak, and since chimeric embryos show deficiencies in mutant cell colonization of the node, prechordal and early paraxial mesoderm populations, we suggest that fgfr1^Δtmk/fgfr1^Δtmk cells were out-competed by WT cells in populating the axial and paraxial progenitor populations early in gastrulation. Interestingly, the resurgence of mutant cells within the posterior notochord may indicate later recruitment of fgfr1^Δtmk/fgfr1^Δtmk cells to the axial progenitor pool over the course of gastrulation.

Definitive endoderm is also derived from epiblast cells which have delaminated at or near the anterior end of the early primitive streak (Lawson et al., 1991). Since fgfr1^Δtmk/fgfr1^Δtmk cells are defective in traversing the streak, it is perhaps not surprising that there was a near absence of mutant cells in the gut of fgfr1^Δtmk/fgfr1^Δtmk ↔ CD-1 chimeric embryos. Slight deficiencies in ES-derived contribution to the endoderm were also observed in fgfr1^Δtmk/+ ↔ CD-1 chimeric embryos. This selection against fgfr1^Δtmk/+ cell contribution to the endodermal lineage may reflect an inherent bias in the distribution of 129 ES-progenitors within 129 ES cell ↔ CD-1 chimeras. The results of other chimeric analyses in our lab have shown that such biases against ESderived cell contribution to endoderm exist independently of the Fgfr1 genotype (unpublished lab observations).

In contrast to the anterior mesoderm and endodermal populations, the limb bud, lateral mesoderm and allantois of chimeric embryos were well colonized by fgfr1^Δtmk/fgfr1^Δtmk cells. Thus, mutant cells were ultimately able to traverse the posterior streak. This skewed distribution of fgfr1^Δtmk/fgfr1^Δtmk mesoderm might indicate a differential requirement for FGFR1 at various A-P levels of the primitive streak. Fgf4 and Fgf5 are expressed predominantly within the anterior two-thirds of the egg cylinder and streak (Haub and Goldfarb, 1991; Niswander and Martin, 1992; Hebert et al., 1991) and could therefore be responsible for anterior streak-specific FGFR1 signaling. However, fgfr1^Δtmk/fgfr1^Δtmk cells were still observed accumulating within the epiblast and mesenchyme of the posterior streak. Thus, the fact that more mutant cells successfully traversed this portion of the streak may simply reflect the ‘competitive’ nature of the chimeric analysis: over the course of gastrulation, WT cells preferentially exited and fgfr1^Δtmk/fgfr1^Δtmk cells progressively accumulated within the streak. Eventually, Fgfr1 mutant cells constituted the majority of the mesodermal progenitor population, and were therefore forced to contribute to ‘posterior’ mesoderm lineages.

This competitive nature of the chimeric analysis might also explain differences observed in the behaviour and patterning of fgfr1^Δtmk/fgfr1^Δtmk mesodermal progenitors between homozygous mutant and chimeric embryos. Although cephalic, heart and axial mesoderm were produced in homozygous fgfr1^Δtmk embryos, these lineages were poorly colonized by mutant cells in fgfr1^Δtmk/fgfr1^Δtmk ↔ CD-1 chimeras where WT cells demonstrated a competitive advantage in traversing the streak. Therefore, while the homozygous mutant analyses indicate a specific role for FGFR1 in the formation of paraxial mesoderm, we argue that the primary defect associated with the fgfr1^Δtmk mutation is a general deficiency in the ability of epiblast cells to traverse the primitive streak. Although secondary roles for Fgfr1 in cell survival or proliferation cannot be discounted, they alone do not seem sufficient to explain the absence of fgfr1^Δtmk/fgfr1^Δtmk cells from anterior mesoderm and endodermal lineages at the time of their inception.

In addition to showing deficiencies in EMT at gastrulation, Fgfr1 mutant cells that managed to progress through the primitive streak tended to accumulate along the posterior midline. This aberrant morphogenetic movement of Fgfr1 mutant cells in fgfr1^Δtmk/fgfr1^Δtmk ↔ CD-1 chimeras could be the result of defects in cell adhesion and/or cell migration. Cellextracellular matrix (ECM) interactions are necessary for cell migration, and mouse embryos homozygous for mutations in fibronectin (a glycoprotein component of the ECM) and in focal adhesion kinase (a non-receptor protein tyrosine kinase thought to mediate integrin signaling) both show phenotypes remarkably similar to homozygous fgfr1^Δtmk embryos (George et al., 1993; Furuta et al., 1995). Interestingly, chimeric analyses of T (Brachyury) gene function have also shown accumulations of T/T cells at the posterior end of the embryo. T/T cells were able to traverse the streak, but the nascent T/T meso-dermal cells accumulated beneath the primitive streak thus suggesting a role for T in regulating the morphogenetic behaviour or adhesive properties of nascent mesoderm (Wilson et al., 1993, 1995). Although the similarity between T and Fgfr1 mutant cell behaviour in chimeric embryos lends credence to the Brachyury-FGF regulatory pathway which has been established in Xenopus (Isaacs et al., 1994), homozygous fgfr1^Δtmk embryos continue to express T and no changes in T expression were observed in fgfr1^Δtmk/fgfr1^Δtmk ↔ CD-1 chimeras.

A role for FGFR signaling in cell migration has been established by mutational analyses in other organisms (ReichmanFried et al., 1994; DeVore et al., 1995). In particular, recent mutations of the Drosophila FGF-R2 gene, heartless (htl), have demonstrated phenotypes strikingly similar to those observed in this study: invaginated mesodermal cells remain aggregated along the ventral midline and fail to migrate in a dorsolateral direction (Gisselbrecht et al., 1996; Beiman et al., 1996). Because these muscle progenitors fail to acquire positionspecific inductive cues, heartless embryos show a reduction in cardiac, visceral and dorsal somatic muscle fates. However, htl mesodermal precursors remain competent to receive inductive signals, and mesodermal fates can be rescued by ectopically expressed decapentaplegic protein (Gisselbrecht et al., 1996; Beiman et al., 1996). This is analogous to the results obtained in our Fgfr1 chimeric analysis which demonstrate that mutant Fgfr1 cells can still make all types of mesoderm, but fail to migrate properly through the primitive streak. Therefore, there may be a conserved role for FGFR signaling in the morphogenesis of mesoderm formation at gastrulation.

Defects in traversing the primitive streak could also involve altered adhesive properties of fgfr1^Δtmk/fgfr1^Δtmk cells. Mutant cells which accumulated within the streak of fgfr1^Δtmk/fgfr1^Δtmk ↔ CD-1 chimeras maintained a columnar epithelial morphology, indicating that fgfr1^Δtmk/fgfr1^Δtmk cells were failing to fully undergo an epithelial to mesenchymal transition (EMT). EMT events are thought to be regulated by a family of calcium dependent cell adhesion molecules (the cadherins). E-cadherin, for example, is expressed in all cells of the early egg cylinder but is down-regulated at gastrulation within the primitive streak and nascent mesodermal populations (Damjanov et al., 1986). Burdsal et al. (1993) have shown that function perturbing antibodies against E-cadherin can force an EMT in cultured epiblast tissue: epiblast cells lose cell-cell contacts, flatten and assume a mesenchymal morphology. Therefore, there appears to be a causal relationship between loss of cadherin function and EMT. FGFR1 may regulate the expression or function of these adhesion molecules. Indeed, exogenous FGF has been shown to cause mesenchymal transformation of cultured epithelial cell lines (Boyer et al., 1992), and this FGF induced EMT has been associated with the cellular redistribution of Ecadherin. There is also some evidence that cell adhesion molecules may act upstream of FGF signaling by binding to and activating FGF receptors (reviewed by Green et al., 1996). Further study will be needed to determine the relationship between FGFR1 signaling and regulation of cell adhesion at gastrulation.

Secondary neural tubes, which are composed entirely of Fgfr1 mutant cells, form within the posterior trunk and tail of fgfr1^Δtmk/fgfr1^Δtmk ↔ CD-1 chimeric embryos. They find their origins from expanded, fgfr1^Δtmk/fgfr1^Δtmk cell-derived neural plates which extend laterally from the neural groove and which presumably arise from the accumulation of mutant columnar epithelial cells observed earlier in the streak. Interestingly, these fgfr1^Δtmk/fgfr1^Δtmk cells adopt a neural over an epidermal fate. Studies in Xenopus have led to the proposal that neuralization of embryonic cells occurs when cells do not receive other inducing signals telling them to form epidermis, mesoderm, or endoderm (for review see Hemmati-Brivanlou and Melton, 1997). The formation of ectopic neural tissue by fgfr1^Δtmk/fgfr1^Δtmk cells may therefore demonstrate a similar neuronal default state for murine embryonic cells.

The formation of secondary neural tubes may be a common manifestation of disrupted morphogenesis at gastrulation: duplicated neural tubes have been observed in a number of spontaneously occurring mouse mutations which show defects in primitive streak formation and axis elongation. These mutations include fused, vestigial tail (vt), and rib vertebrae (as reviewed by Cogliatti, 1986; Gruneberg and Wickramaratne, 1974; Theiler and Varnum, 1985). Recently, vt has been shown to be a hypomorphic allele of Wnt-3a (Greco et al., 1996). Wnt3a null mutant embryos lack caudal somites, have disrupted notochord and fail to form a tail bud (Takada et al., 1994). In addition, secondary neural tubes are formed, but, unlike Fgfr1 mutants, these tubes form from cells that have successfully traversed the streak (Yoshikawa et al., 1997). This has led Yoshikawa et. al. (1997) to propose that Wnt-3a regulates the choice of cell fate between neural and paraxial mesodermal cell lineages.

Interestingly, FGFR1 does not appear to be directly required for cell fate specification, yet mutations affecting this pathway can result in similar neural anomalies. Hence, while FGFR1 and Wnt-3a may be playing different roles in the movement and specification of mesodermal progenitors at the primitive streak, disturbances in both systems result in similar phenotypic abnormalities. This study therefore underscores both the complexity of interactions which regulate mesoderm formation, and the importance of normal morphogenetic movements in the processes of patterning and fate determination.

We are grateful to Drs Robin Lovell-Badge and Cynthia Faust for providing the Sox-1 and Sox-2, and the Ncam cDNA probes, respectively. We also thank Drs J. Klingensmith, J. Partanen, and J. Pearce for critically reading the manuscript. J. R. is a Terry Fox Research Scientist of the National Cancer Institute of Canada and an International Research Scholar of the Howard Hughes Medical Institute, andB. G. C. is a Scholar of the Natural Sciences and Engineering Research Council of Canada. This work was supported by a program project from the NCIC with funds from the Terry Fox Foundation.