ABSTRACT
The T (Brachyury) deletion in mouse is responsible for defective primitive streak and notochord morphogenesis, leading to a failure of the axis to elongate properly posterior to the forelimb bud. T/T embryonic stem (ES) cells colonise wild-type embryos, but in chimeras at 10.5 days post coitum (dpc) onwards they are found predominantly in the distal tail, while trunk paraxial and lateral mesoderm are deficient in T/T cells (Wilson, V., Rashbass, P. and Beddington, R. S. P. (1992) Development 117, 1321-1331). To determine the origin of this abnormal tissue distribution, we have isolated T/T and control T/+ ES cell clones which express lacZ constitutively using a gene trap strategy. Visualisation of T/T cell distribution in chimeric embryos throughout gastrulation up to 10.5 dpc shows that a progressive buildup of T/T cells in the primitive streak during gastrulation leads to their incorporation into the tailbud. These observations make it likely that one role of the T gene product is to act during gastrulation to alter cell surface (probably adhesion) properties as cells pass through the primitive streak.
As the chimeric tail elongates at 10.5 dpc, abnormal morphology in the most distal portion becomes apparent. Comparison of T expression in the developing tailbud with the sites of accumulation of T/T cells in chimeras shows that T/T cells collect in sites where T would normally be expressed. T expression becomes internalised in the tailbud following posterior neuropore closure while, in abnormal chimeric tails, T/T cells remain on the surface of the distal tail. We conclude that prevention of posterior neuropore closure by the wedge of T/T cells remaining in the primitive streak after gastrulation is one source of the abnormal tail phenotypes observed.
Accumulation of T/T cells in the node and anterior streak during gastrulation results in the preferential incorporation of T/T cells into the ventral portion of the neural tube and axial mesoderm. The latter forms compact blocks which are often fused with the ventral neural tube, reminiscent of the notochordal defects seen in intact mutants. Such fusions may be attributed to cell-autonomous changes in cell adhesion, possibly related to those observed at earlier stages in the primitive streak.
INTRODUCTION
The T gene is deleted in mouse Brachyury mutants (Herrmann et al., 1990) causing a reduction in tail length in hemizygotes and death of homozygous embryos on the 11th day of gestation (Chesley, 1935; Gluecksohn-Schonheimer, 1944; Gruneberg, 1958). The T gene encodes a putative transcription factor (Kispert and Herrmann, 1993) and both its mRNA and protein are present in the primitive streak from the onset of gastrulation (Wilkinson et al., 1990; Herrmann, 1991; Kispert and Herrmann, 1994). Expression persists in the streak and later in the tailbud for the entire period of axis formation and axis elongation (6.5-12.5 dpc; Kispert and Herrmann, 1994). It is also expressed in the node and notochord. The pattern of T expression appears to be essentially identical in vertebrates where its homologue has been studied (Smith et al., 1991; Schulte-Merker et al., 1992). However, the phenotype of homozygous mutant mouse embryos does not obviously correlate with an essential function for T during the early stages of gastrulation, since development rostral to the forelimb bud appears grossly normal. Only in more caudal trunk regions and in later embryos is the notochord missing and other mesodermal derivatives deficient or defective (Herrmann, 1992; Beddington et al., 1992; Rashbass et al., 1994). One explanation for this would be that T is only required once all tissues rostral to the forelimb have been laid down (i.e. after the 1- to 2-somite stage). The demonstration that wild-type T protein is only required from the 9th day of gestation for continuation of its own expression (Herrmann, 1991) and that of certain other genes (Wnt-3a, Wnt-5a (McMahon, 1992), Evx-1 (Dush and Martin, 1992)) transcribed in the streak (Rashbass et al., 1994) is consistent with such an hypothesis. However, the chimeric studies reported in this paper indicate that T is necessary for normal morphogenetic movements during earlier stages of gastrulation.
We have shown previously that abnormalities characteristic of T/T embryos occur in chimeras made between T/T mutant cells and wild-type ones (Rashbass et al., 1991). Using a glucose phosphate isomerase isozyme variant as a genetic marker to distinguish mutant cells, we found that T/T cells in 9.5-11.5 dpc chimeras were unevenly distributed along the embryonic craniocaudal axis (Wilson et al., 1993; Beddington et al., 1992). Mutant cells predominated in caudal regions and were relatively sparse in mesodermal derivatives compared to non-ingressing ectodermal tissues. This led to the hypothesis that mutant cells ingressing through the streak were defective compared to wild-type cells (with which they were in competition) in their ability to move away from the midline and to populate the mesoderm at a normal rate. As a result, mutant cells accumulate in the region responsible for axial elongation, and eventually inhibit it. By virtue of a transgenic single cell marker introduced into mutant cells, we can now verify that T/T cells start to amass in the primitive streak during gastrulation. These results demonstrate that wild-type T protein does have a function during the earlier stages of gastrulation. They suggest that at least one of its early roles is to regulate the morphogenetic behaviour of nascent mesoderm by altering cell autonomous properties, probably related to cell adhesion.
MATERIALS AND METHODS
Vectors
The gene trap vectors pGT1,2 and 3 contain the βgeo reporter (Friedrich and Soriano, 1991), flanked by the En-2 splice acceptor and SV40 polyadenylation signal from pGT4.5 (Gossler et al., 1989). The En-2 exon sequence was modified to generate three derivatives which can form fusion proteins in one of each of the three reading frames. Equal quantities of each vector were mixed together for electroporation.
Isolation of gene trap ES cell lines ubiquitously expressing lacZ
ES cells were maintained as described previously (Wilson et al., 1993). 108 ES cells (either ES line BTBR6 (T/T) or BTBR7 (T/+)) were electroporated with a mixture of gene trap vectors pGT1,2 and 3 using standard conditions (Hill and Wurst, 1993) and selected in medium containing G418 (200 μg/ml, Gibco). After 10 days of selection, 200-250 macroscopic colonies from each electroporation were picked into 96-well plates (Nunc) and grown for 3 days in the absence of drug selection. Each clone was then passaged into duplicate wells in separate 96-well plates and grown overnight. One plate was stained with X-gal (Sigma) as described (Beddington et al., 1989), except that fixation was performed for 5-10 minutes at 4°C, and clones showing staining in all cells, including peripheral differ-entiated cells, identified. Fifteen clones from each of the two electro-porated cell lines, exhibiting apparently ubiquitous staining, were expanded from the duplicate 96-well plates. After expansion, four T/T and five T/+ clones were tested in chimeras for ubiquitous reporter gene expression.
Construction of chimeras
ES cells were injected into C57BL/6 or C57BL/10 × CBA blastocysts as described previously (Rashbass et al., 1990) and the embryos transferred to pseudopregnant recipient C57BL/10 × CBA F1 females. Aggregation of ES cells with 8-cell embryos were performed in 5 cm bacteriological dishes (Sterilin) in hanging drops of M16 medium (Hogan et al., 1986) containing 10% FCS. 3-4 clumps, each comprising 3-4 ES cells, were added to each drop containing a single 8-cell MF1 embryo, denuded of its zona pellucida using Acid Tyrodes solution (Hogan et al., 1986). ES cells and embryos were co-cultured in hanging drops for 4 hours at 37°C in a humidified atmosphere with 6% CO2 in air. Subsequently the dishes were inverted, the drops covered with paraffin oil (Boots, UK, Ltd) and incubated overnight. Embryos that had developed to the blastocyst stage were transferred to pseudopregnant recipients.
Recovery and staining of chimeras
Potential chimeras were dissected from the uterus 5-8 days following transfer. Fixation and X-gal staining of embryos was as described else-where (Beddington et al., 1989). Embryos up to 8.5 dpc were fixed for 10-20 minutes, whereas later stages were fixed for 20-30 minutes. The duration of staining varied according to the ES cell clone used and ranged from 1 hour to 24 hours. Embryos showing positive staining were processed for wax histology as described in Beddington (1994), and 7 μm serial sections were cut (Bright 6030 Microtome). Once dewaxed, sections were mounted in DPX mountant (BDH, Ltd.) and photographed (Kodak Ektachrome 64T film) in a Zeiss Axiophot microscope using differential interference contrast optics.
Whole-mount in situ hybridisation
The T mRNA probe was synthesised as described (Wilson et al., 1993) and in situ hybridisation was performed according to Wilkinson (1992).
RESULTS
Isolation and validation of ubiquitously expressing lacZ ES cell clones
Following electroporation and drug selection, 210 T/T ES cell clones and 229 T/+ clones were tested for bacterial β-galactosidase activity in 96-well plates. Approximately half of the clones of each genotype were positive (β-gal+) and of these, 21 (17.8%) T/T clones and 22 (17.7%) T/+ clones showed blue staining in all cells in the well (Table 1). Four T/T and 5 T/+ strongly staining clones were selected for testing in chimeras (Table 2). Like the parent cell line (Wilson et al., 1993), the four T/T clones gave rise to chimeras showing a range of abnormalities (Table 3) depending on the extent of mutant cell contribution (Fig. 1A,B). A high contribution of T/T cells (greater than about 60%) produced detectable abnormalities by early somite stages reminiscent of the intact T/T phenotype. Chimeras with a low T/T cell contribution exhibited tail and allantoic deformities, most evident from approximately 10.0 dpc. Although carrying independent gene trap insertions, all four T/T clones behaved similarly in chimeras (Table 3), indicating that the hemizygous transgene was developmentally neutral. Having demonstrated that lacZ was expressed in all embryonic tissues (see below), three of these clones (GM 6.6, 6.12 and 6.15) were used for subsequent phenotypic analysis. Clone GM7.3 which exhibited high levels of β-galactosidase activity was selected as a heterozygous control. Like its parent (BTBR7), which gave rise to more normal chimeras which had only minor tail defects at later (10.5-11.5 dpc) stages, GM7.3↔+/+ chimeras at 9.5 dpc were normal in phenotype (Fig. 1C, Table 3).
Generation of β-galactosidase positive (β-gal+) ES cell clones by electroporation of GT1,2 and 3

Gross morphology of chimeras made with marked ES cells compared to those derived from parent cell line

Chimeras from marked T/T and T/+ ES cell clones. GM6.15↔+/+(A), GM6.6↔+/+(B) and GM7.3↔+/+(C) chimeras dissected at an equivalent age of 10.5 dpc (A,B) and 9.5 dpc (C). Arrowheads in A and B indicate abnormally truncated or divided tail. (D,E). Transverse sections through a GM6.15↔+/+ chimera with high level ES cell contribution, showing βgal+ cells in all tissues. (D) Hindbrain level. (E) Forelimb level. N, neural tube; OV, otic vesicle; S, somite; G, gut; F, forelimb bud. Scale bar: 1.67 mm in A and B; 1.0 mm in C; 330 μm in D and E.
Chimeras from marked T/T and T/+ ES cell clones. GM6.15↔+/+(A), GM6.6↔+/+(B) and GM7.3↔+/+(C) chimeras dissected at an equivalent age of 10.5 dpc (A,B) and 9.5 dpc (C). Arrowheads in A and B indicate abnormally truncated or divided tail. (D,E). Transverse sections through a GM6.15↔+/+ chimera with high level ES cell contribution, showing βgal+ cells in all tissues. (D) Hindbrain level. (E) Forelimb level. N, neural tube; OV, otic vesicle; S, somite; G, gut; F, forelimb bud. Scale bar: 1.67 mm in A and B; 1.0 mm in C; 330 μm in D and E.
7.5-10.5 dpc chimeras containing the greatest ES cell contribution were sectioned transversely so that any tissue or site failing to express lacZ could be identified. Serial sections of GM6.6, GM6.12, GM6.15 and GM7.3 chimeras demonstrated that bacterial β-galactosidase activity was present in all tissues of the embryo (Fig. 1D,E; Table 2). All clones also produced stained descendants in extraembryonic mesoderm derivatives. Trophectoderm and primitive endoderm derivatives were not included in the analysis since they are seldom colonised by ES cell descendants (Beddington and Robertson, 1989), and the T gene is never expressed in these tissues (Wilkinson et al., 1990). Within the embryo, no difference was observed in the staining intensity of positive cells located in cranial as opposed to caudal regions. Therefore, the gene trap strategy has proved an efficient means of generating ES cell clones containing a constitutive, ubiquitous and neutral single cell marker with which to compare the behaviour of T/T and control T/+ cells in midgestation chimeras. While embryos containing a high proportion of ES cell descendants validate the marker system those with a contribution of 40% or less are the most informative regarding the aberrant behaviour of mutant cells. The subsequent phenotypic analysis is based on results from at least 2 of the independent T/T marked clones for any given stage and is derived from chimeras containing 40% or less donor cells.
T/T cells show abnormal behaviour during the latter stages of gastrulation
Up to full-length primitive streak stages the distribution of marked T/T cells in the epiblast appears random and there is no evidence for specific exclusion of mutant cells from particular regions. Likewise, no epiblast or ectodermal domains are consistently devoid of mutant cells at head fold and early somite stages (Fig. 2A). However, by these stages, accumulation of T/T cells along the length of the primitive streak is apparent (12/16 chimeras; Fig. 2B,C). Large numbers of T/T cells are particularly evident in the node region (5/16 chimeras), whereas lateral and paraxial mesoderm are relatively devoid of mutant cells (Fig. 2C,D). In sections, T/T cells appear to congegrate in the midline immediately ventral to the primitive streak (mesodermal layer) but there are also concentrations of mutant cells in the epiblast constituent of the streak (Fig. 2E).
T/T cells accumulate in the primitive streak. (A) Uniform distribution of T/T cells in epiblast of an early headfold stage high level chimera, dorsal view; anterior faces right. (B) Accumulation of T/T cells in primitive streak of a second headfold stage embryo, viewed from lateral side; anterior faces right. Regions other than primitive streak contain a lower proportion of T/T cells. Vertical line indicates the extent of the primitive streak. (C) Posterior view of an 8.5 dpc embryo in which T/T cells preferentially colonise primitive streak, node (white arrowhead) and the base of the allantois (black arrowhead). (D) Dorsal view of a 10-somite embryo (9.0 dpc) with T/T cells clustered in the prospective tailbud. (E) Transverse section through primitive streak of embryo shown in D. T/T cells are clustered in the midline of both the ectodermal and mesodermal component of the primitive streak. (F) 9.5 dpc GM7.3↔+/+ embryo showing random distribution of T/+ cells, except for some accumulation in the tailbud. (G) Transverse section through embryo shown in F. T/+ cells are seen in cranial (upper left) and caudal (lower right) ectoderm and mesoderm. Paired horizontal lines in D and F indicate the planes of section in E and G, respectively. Scale bar: 180 μm in A; 170 μm in B and C; 620 μm in D; 210 μm in E; 400 μm in G.
T/T cells accumulate in the primitive streak. (A) Uniform distribution of T/T cells in epiblast of an early headfold stage high level chimera, dorsal view; anterior faces right. (B) Accumulation of T/T cells in primitive streak of a second headfold stage embryo, viewed from lateral side; anterior faces right. Regions other than primitive streak contain a lower proportion of T/T cells. Vertical line indicates the extent of the primitive streak. (C) Posterior view of an 8.5 dpc embryo in which T/T cells preferentially colonise primitive streak, node (white arrowhead) and the base of the allantois (black arrowhead). (D) Dorsal view of a 10-somite embryo (9.0 dpc) with T/T cells clustered in the prospective tailbud. (E) Transverse section through primitive streak of embryo shown in D. T/T cells are clustered in the midline of both the ectodermal and mesodermal component of the primitive streak. (F) 9.5 dpc GM7.3↔+/+ embryo showing random distribution of T/+ cells, except for some accumulation in the tailbud. (G) Transverse section through embryo shown in F. T/+ cells are seen in cranial (upper left) and caudal (lower right) ectoderm and mesoderm. Paired horizontal lines in D and F indicate the planes of section in E and G, respectively. Scale bar: 180 μm in A; 170 μm in B and C; 620 μm in D; 210 μm in E; 400 μm in G.
Accumulation of T/T cells also occurs at the base of the allantois (Fig. 2C; 11/16 chimeras) and, in some chimeras, abnormal clumps of blue cells are seen on the surface of the allantois, a feature reminiscent of intact homozygous Brachyury embryos. Sections from 3 embryos show some isolated stained cells in the distal portion of the allantois but these are relatively rare. There was no evidence of aberrant morphology within the wild-type component of chimeric allantoides.
T/+↔+/+ chimeras show no sign of tissue- or region-specific differences in levels of chimerism up to early somite stages (data not shown).
Disruption of tail formation in chimeras
In our previous electrophoretic analysis of 9.5-11.5 dpc chimeras, we demonstrated an increase in T/T cell contribution to caudal regions compared with more rostral domains (Wilson et al., 1993). Seeing the precise distribution of all ES cell progeny, by virtue of a single cell marker, makes this bias even more evident, both at the gross morphological level (Fig. 1A,B) and in histological preparations (Fig. 3). This is particularly true in chimeras with an otherwise low T/T contribution. The following observations are based on the analysis of serial sections of 7 T/T↔+/+ chimeras at 9.5 dpc. From 9.0-10.5 dpc, mutant cells are particularly prevalent in the tailbud itself and in the midline ‘mesoderm’ immediately rostral to it (Table 4; Fig. 3C-E), i.e. in the region normally occupied by the notochord, or immediately adjacent to it. Mutant cells in this region appear to adhere closely to one another indicating that accumulation may be due to inappropriate cell:cell adhesion. When wild-type notochord is present only very occasionally are blue cells observed within it. Usually, the T/T midline mesodermal cells do not mix with wild-type notochord (Fig. 3F-H) but instead form short stretches of compacted mutant tissue which often appear to fuse and become incorporated into areas of mutant ventral neurectoderm (Fig. 3F-H). The distribution of T/T cells in the neural tube is not uniform. Mutant cells predominate in the ventral region (including the future floorplate) and in the midline of the caudalmost neurectoderm. In addition, a strip of mutant cells is seen extending from the origin of the allantois (on the ventral surface of the tail) back to the distal tip of the tail (Figs 3C-H, 4D,F). In contrast, the paraxial mesoderm and hindgut contain very few T/T cells (Figs 3D-H, 4D,F).
Successively more rostral sections through a 10.5 dpc GM6.15↔+/+ chimeric tail. (A) T/T cells located both internally and on the surface of the distal tail. (B-D) T/T cells predominate in axial ectoderm and mesoderm. White arrowhead in D marks the distal end of the hindgut. (E-H) Collections of T/T cells are found in midline neuroectoderm, mesoderm and ventral surface ectoderm. More proximally, T/T cells are located in the ventral third of the closing neural folds (F-H). (F) Compacted blocks of T/T tissue (arrows) beneath the neural tube fuse proximally (G, left hand block; H, right hand block) with the chimeric ventral neural tube. A normal wild-type notochord is present in F-H (empty arrowheads). Black arrowheads mark the position of a ventral strip of midline ectoderm T/T cells extending from the base of the allantois to the distal tip of the tail. Distally (C,D), these form an abnormal furrow along the ventral edge of the tail. Scale bar: 200 μm.
Successively more rostral sections through a 10.5 dpc GM6.15↔+/+ chimeric tail. (A) T/T cells located both internally and on the surface of the distal tail. (B-D) T/T cells predominate in axial ectoderm and mesoderm. White arrowhead in D marks the distal end of the hindgut. (E-H) Collections of T/T cells are found in midline neuroectoderm, mesoderm and ventral surface ectoderm. More proximally, T/T cells are located in the ventral third of the closing neural folds (F-H). (F) Compacted blocks of T/T tissue (arrows) beneath the neural tube fuse proximally (G, left hand block; H, right hand block) with the chimeric ventral neural tube. A normal wild-type notochord is present in F-H (empty arrowheads). Black arrowheads mark the position of a ventral strip of midline ectoderm T/T cells extending from the base of the allantois to the distal tip of the tail. Distally (C,D), these form an abnormal furrow along the ventral edge of the tail. Scale bar: 200 μm.
Mutant cells accumulate in sites of T expression
To examine how the sites of accumulation of T/T cells in chimeras correspond to the normal domains of T expression during tail development, the distribution of T transcripts was re-examined in the cell population responsible for axial elongation from the start of tailbud formation (∼8.75 dpc) to the stage at which the posterior neuropore closes (10.0 dpc). At 8.75 dpc, T expression is restricted to a caudal domain, comprising both ectoderm and mesoderm layers and flanked by the neural folds (Fig. 4A). Thus, part of the population of cells expressing T remains on the dorsal surface of the embryo and not until after posterior neuropore closure, which occurs at about the 30-somite stage (Copp, 1982), does the entire population of cells expressing T become internalised. (Fig. 4C,E). Presumably, internalisation is normally achieved by rostrocaudal closure of the lateral neural folds and adjoining caudal ectoderm. Consequently, the posterior streak is internalised first and the node last (Fig. 5A).
Comparison of T expression with sites of accumulation of T/T cells as primitive streak is replaced by tailbud. (A,C,E). T expression detected by in situ hybridisation to T mRNA. (A). 8.5 dpc embryo, dorsal view; T is expressed on the surface of the primitive streak. (C). 9.5 dpc embryo, lateral view. T expression is enclosed ventrally by surface ectoderm but remains exposed dorsally by the posterior neuropore. (E). Lateral view showing internal T expression once the posterior neuropore has closed at 10.5 dpc. (B,D,F). GM6.15↔+/+ chimeras. (B) 8.5 dpc chimera at an equivalent stage to that of the embryo in A. (D) 10.0 dpc chimeric tail (ventrolateral view) showing ventral midline ectoderm cells and divided tail. (F) 10.5 dpc chimera at equivalent stage to that in E. T/T cells remain located externally in the distal tail. Empty arrowheads: posterior neuropore. Black arrowhead: open posterior neuropore with bifurcation of T/T cell aggregation. Empty arrow: T/T cells located on surface. Scale bar: 210 μm in A-C and E; 310 μm in D; 290 μm in F.
Comparison of T expression with sites of accumulation of T/T cells as primitive streak is replaced by tailbud. (A,C,E). T expression detected by in situ hybridisation to T mRNA. (A). 8.5 dpc embryo, dorsal view; T is expressed on the surface of the primitive streak. (C). 9.5 dpc embryo, lateral view. T expression is enclosed ventrally by surface ectoderm but remains exposed dorsally by the posterior neuropore. (E). Lateral view showing internal T expression once the posterior neuropore has closed at 10.5 dpc. (B,D,F). GM6.15↔+/+ chimeras. (B) 8.5 dpc chimera at an equivalent stage to that of the embryo in A. (D) 10.0 dpc chimeric tail (ventrolateral view) showing ventral midline ectoderm cells and divided tail. (F) 10.5 dpc chimera at equivalent stage to that in E. T/T cells remain located externally in the distal tail. Empty arrowheads: posterior neuropore. Black arrowhead: open posterior neuropore with bifurcation of T/T cell aggregation. Empty arrow: T/T cells located on surface. Scale bar: 210 μm in A-C and E; 310 μm in D; 290 μm in F.
Model of progressive accumulation of T/T cells in distal tail. (A) T expression in primitive streak, tailbud, node and notochord. (B) Random distribution of T/+ cells in chimeras. (C) Accumulation of T/T cells in the primitive streak and allantois between 7.5 and 8.5 dpc leads to a mass of T/T cells in the ectodermal and mesodermal components of the tailbud at 9.5 dpc. This results in a physical block to posterior neuropore closure, which would normally occur between 9.5 dpc and 10.5 dpc. Consequently, at 10.5 dpc T/T cells are present on the outer surface of the distal tail, having been enclosed only partly by surface ectoderm ventrally and neuroectoderm dorsally.
Model of progressive accumulation of T/T cells in distal tail. (A) T expression in primitive streak, tailbud, node and notochord. (B) Random distribution of T/+ cells in chimeras. (C) Accumulation of T/T cells in the primitive streak and allantois between 7.5 and 8.5 dpc leads to a mass of T/T cells in the ectodermal and mesodermal components of the tailbud at 9.5 dpc. This results in a physical block to posterior neuropore closure, which would normally occur between 9.5 dpc and 10.5 dpc. Consequently, at 10.5 dpc T/T cells are present on the outer surface of the distal tail, having been enclosed only partly by surface ectoderm ventrally and neuroectoderm dorsally.
It is striking that the sites of maximum T/T cell accumulation in chimeras correspond to the normal sites of T gene expression (Fig. 4A,C,E). As a result posterior neuropore closure is often prevented due to the mass of mutant cells interposed between the posterior lateral neural folds (Fig. 4B,D,F). Such a wedge of mutant cells may be responsible for the forked tail tips frequently observed in T/T↔+/+ chimeras (eg. Fig.1B, second row, second embryo from left). While the distribution of ES cell descendants in T/+↔+/+ chimeras is much more uniform along the rostrocaudal axis (Figs 1C, 2F), several 9.5 dpc chimeras show elevated numbers of T/+ cells in the same parts of the distal tail as those populated by T/T cells (Table 4). This is also associated with tail defects although less severe than those observed in T/T chimeras (Wilson et al., 1993).
DISCUSSION
Validation of the gene trap marker
The frequency of ubiquitously staining ES cell clones in vitro (approximately 20% of all β-gal+ clones; Table 1), which corresponded to ubiquitous enzyme activity in mid-gestation chimeras for all clones tested (Table 2), indicates, as do other gene trap screens (Friedrich and Soriano, 1991; Korn et al., 1993) that the frequency of ubiquitously expressed genes accessible to the gene trap vector may be relatively high. While this makes a gene trap strategy attractive for introducing a cell marker the mutagenic nature of the vector insertion has to be considered. Although potential dominant lethal mutations (identified by germ line transmission of ES cell genotype without gene trap vector sequence) appear to be rare (e.g. 1/28 lines tested; Friedrich and Soriano, 1991), it is important to confirm neutrality of hemizygous insertions by demonstrating that independently marked clones behave identically in the embryo. Here we show that the progeny of three independent β-gal+T/T clones exhibit the same overall colonisation profile in chimeras as their unmarked parent (Wilson et al., 1993) and that this differs from the colonisation pattern of a β-gal+T/+ cell line and its parent (Wilson et al., 1993). Consequently, hemizygosity for these insertions of lacZ into ubiquitously expressed endogenous genes does not seem to have perturbed their developmental or morphogenetic potential.
Nascent mesodermal T/T cells exhibit a cell autonomous morphogenetic defect
In this section, the abnormal craniocaudal distribution of T/T cells, as opposed to defective notochord differentiation, will be discussed. Our results show that the characteristic bias of T/T cells to colonise the caudal tail is not due to exclusion of mutant cells from regions of the epiblast destined to give rise to cranial and trunk structures (Fig. 2A). Mutant cells can populate all regions of the epiblast and thus will be represented in all prospective tissue domains defined by fate maps (Beddington, 1981, 1982; Lawson and Pedersen, 1992; Tam and Beddington, 1987).
However, the chimeras make it clear that the imbalance between rostral and caudal chimerism has its inception during gastrulation and is rooted in the abnormal accumulation of T/T cells in the node, primitive streak and regions immediately ventral to the streak (Fig. 2B-E). Accumulation is first clearly discernible at the headfold stage but probably results from abnormal cell behaviour somewhat earlier. However, this chimeric analysis did not identify abnormal mesoderm movement during the earliest phase of gastrulation. Therefore, either the most rostral mesoderm cells do not require T for their normal exit from the primitive streak, or this assay is not sufficiently sensitive to detect earlier abnormalities. The most rostral mesoderm may be derived from those cells that express goosecoid and downregulate T (Schulte-Merker et al., 1994) and in lower vertebrates move by active migration rather than convergent extension characteristic of more caudal mesoderm (Niehrs et al., 1993).
The accumulation of mutant cells in the site of mesoderm production, and hence the region responsible for axial elongation, continues throughout late gastrulation and persists in the tailbud when this structure becomes the source of additional caudal mesoderm. As a consequence, there is a deficit of mutant cells in definitive mesodermal derivatives. Given the apparently normal differentiation of mutant cells in the gut and all rostral embryonic mesodermal derivatives (with the exception of the noto-chord) in chimeras containing a high percentage of T/T cells, and the demonstration that T/T tissue can give rise to mature endoderm and mesoderm derivatives when grafted to an ectopic site (Ephrussi, 1935; Fujimoto and Yanigasawa, 1979; Beddington et al., 1992), the accumulation of mutant cells in the streak and tailbud cannot be due to an inability to undergo appropriate mesodermal or endodermal terminal cytodifferentiation. Instead the defect seems to stem from a morphogenetic failure preventing nascent mesoderm from moving away from the streak or tailbud. The tendency of internal T/T cells to clump together (Fig. 3), instead of intermixing with wild-type cells as their T/+ counterparts do (Fig. 2G), suggests that the abnormality in mesoderm is one of adhesion rather than migration. Further-more, the obvious aggregation of mutant cells in the streak demarcated by a rather sharp boundary also points to altered cell adhesion being the primary defect (Fig. 2B-D). The presence of relatively large numbers of mutant cells in the ectodermal layer of the streak would argue that adhesive changes regulated by T occur during the epithelial to mesenchymal transition. Since wild-type cells are deployed normally the lack of T protein appears to have a cell autonomous effect, probably mediated by cell surface proteins involved in cell adhesion. This would explain why mutant cells accumulate precisely where the T gene is normally expressed (e.g. Fig. 4).
Adhesive and migratory changes in cells isolated from T/T embryos in vitro have been reported by Yanagisawa and Fujimoto (1977), and Hashimoto et al. (1987). Aggregates of midgestation embryonic T/T cells in rotation culture were smaller than those formed by wild-type cells (Yanagisawa and Fujimoto, 1977), although the significance of these results is unclear since a range of both morphologically normal and abnormal T/T tissues were present in the assay. Mesodermal cells from 8-9 dpc T/T embryos cultured on an extracellular matrix substratum migrate slightly more slowly than their wild-type counterparts and, although the number of embryos used in the analysis was small (Hashimoto et al., 1987), these results are consistent with altered adhesion or migration of T/T mesodermal cells. In addition, Yanagisawa et al. (1981) have proposed that the reduced mesoderm/ectoderm ratio of T/T mutants stems from a failure of morphogenetic movement during gastrulation. The chimeric assay described here extends these studies and identifies the affected population as those cells moving through the primitive streak from midgastrulation stages.
It is interesting that embryos lacking the protein integrin α5 (Yang et al., 1993) closely resemble homozygous Brachyury embryos. This raises the possibility that α5 integrin regulation may be either directly or indirectly downstream of T and, at least in part, be responsible for the cell autonomous abnor-malities in cell adhesion and/or movement that we observe in chimeras. If this is the case, we would predict that chimeras made between wild-type and α5-/α5- cells would exhibit the same anomalous mutant cell behaviour as seen with T/T cells.
Abnormal notochord development
In zebrafish, it has been shown that homozygous no tail (ntl) embryos, which lack the zebrafish homologue of the T gene, form axial mesoderm (except in the very caudal extremity of the embryo) but that this fails to differentiate into mature notochord (Halpern et al., 1993). Transplantation of ntl cells into wild-type embryos does not rescue this defect (they fail to colonise notchordal tissue) and in reciprocal transplants donor wild-type cells can still form notochord in spite of a mutant environment. Consequently, the failure of ntl mutant cells to differentiate into notochord appears to be a cell autonomous defect. However, the axial mesoderm that does form in ntl mutants can induce floorplate from the ventral neural tube indicating that the specification of floorplate seems to precede definitive notochord differentiation. Similar results have been found in T/T mouse embryos where floorplate differentiates in regions of the trunk where normal notochord is missing (Rashbass et al., 1994).
Sections of chimeras show that occasionally T/T cells can be incorporated into stretches of wild-type notochord. However, it is more common for short (<100 μm) lengths of an additional ‘notochord’ composed entirely of mutant cells to form on one or both sides of a central wild-type notochord in the tail (Fig. 3F-H). Frequently, fusions are seen between this mutant ‘notochord’ tissue and the mutant ventral neural tube (Fig. 3G,H). Like the experiments in zebrafish, these observations imply a cell autonomous defect in T/T mutant cells, which compromises their differentiation into mature notochord, making intercalation with wild-type notochord cells an exception.
Chesley (1935) and Grüneberg (1958) both report notochord/neural tube fusions in intact heterozygous and homozygous Brachyury embryos and attribute this behaviour to altered cell surface properties. Such an affinity between mutant axial mesoderm and ventral neural tube may derive from the common origin of these tissues. In both zebrafish (Kimmel and Warga, 1986) and Xenopus (Dale and Slack, 1987), single blastomeres from the 128-cell and 32-cell stage respectively give rise to progeny in both the notochord and floorplate. More impressively, single cells in Hensen’s node of the chick (Selleck and Stern, 1991) and in the anterior streak of the mouse (Lawson et al., 1991) can contribute descendants to both floorplate and notochord. In the chimeras described here, it is clear that within the neurectoderm T/T cells preferentially colonise the ventral neural tube (Fig. 3 and data not shown). This is presumably because mutant cells accumulate in the node and, consequently, will predominate in the region from which the ventral neural tube is derived (Lawson et al., 1991). Mutant cells that find themselves in the axial mesoderm but are blocked in their differentiation may, therefore, closely resemble ventral neurectoderm cells resulting in fusions between the two cell populations.
Insights into tail development
The progressive accumulation of T/T cells in the caudal aspect of chimeras demonstrates that there is a physical continuity between the primitive streak (including the node) of the egg cylinder and the tail bud of much more advanced embryos. Furthermore, the distribution of mutant cells and the constant size of the caudal domain of T gene expression in wild-type embryos (Herrmann, 1991; Kispert, 1994) give little indication that streak regression occurs in the mouse. In contrast, the streak and node appear to be incorporated into the tailbud, becoming fully internalised once posterior neuropore closure has occurred (Fig. 5). The accumulation of T/T cells prevents this closure in chimeras and thus descendants of the streak and node are left exposed on the dorsal amniotic surface of the tail. In intact T/T mutants, neural tube closure occurs later than in wild-type littermates (Wilson and Beddington, unpublished observations), and delayed posterior neuropore closure in heterozygotes is documented by Gruneberg (1958), where transverse sections of the caudal regions of intact T/+ and T/t mutants show that the posterior neuropore remains open up to a more rostral level than in wild-type littermates.
In Xenopus, cell lineage experiments indicate that cells in the late gastrula dorsal blastopore lip are the progenitors of the tailbud hinge region at later stages (Gont et al., 1993). In the mouse, transplantation experiments have shown that tailbud mesoderm and 8.5 dpc primitive streak cells have equivalent histogenetic potential and that orthotopically grafted 8.5 dpc primitive streak cells give rise to descendants in the tailbud 24 hours later as well as to trunk paraxial mesoderm and other tissues (Tam and Tan, 1992). DiI labelling of both the node and primitive streak at 8.5 dpc also shows that both contain cells that contribute to the tailbud at 9.5 dpc, but the node only gives rise to notochord while the primitive streak descendants are responsible for generating paraxial mesoderm (Wilson and Beddington, unpublished data). Thus, the transition from gastrulation via the primitive streak to axis elongation by supply of new tissue from the tailbud appears to involve a constant population of cells, which retain their topographical relationship with respect to prospective fates and become internalised during posterior neuropore closure. Consequently, the gradual accumulation of T/T cells in the streak will eventually result in a physical block to axial elongation at tailbud stages and this may be part of the reason for the deficiency of caudal mesoderm observed in Brachyury mutant embryos.
ACKNOWLEDGEMENTS
We are grateful to Ruth Arkell and Sally Dunwoodie for helpful comments on the manuscript, and Louise Anderson for technical assistance. This work was supported in part by BBSRC grant AT337/615. R. S. P. B. is a Howard Hughes Medical Institute International Scholar.