An in situ transgenic enzyme marker for the midgestation mouse embryo and the visualization of inner cell mass clones during early organogenesis

No descendant preimplantation clone has ever been visualized in the embryonic portion of the early mammalian conceptus. Consequently, our appreciation of the pluripotency of individual inner cell mass (ICM) cells or epiblast cells rests on electrophoretic analysis of homogenized tissues (Gardner, 1985a). The inability to follow the fate of single cells in situ and to observe the distribution of their progeny in the embryo is a serious deficit to our understanding of cell deployment during gastrulation and early organogenesis. While the widespread tissue colonization reported for single epiblast cells is indicative of extensive cell mixing during early postimplantation development (Beddington, 1983), it is not clear whether mixing is a transient episode preceding allocation of cells to different primordia or if it is sustained within nascent definitive tissues. Only the analysis of descendant clones in the intact conceptus, or in histological preparations, will resolve this question.

Over the last ten years new cell markers have been developed for chimaera studies in the mouse (Ponder, 1987). However, all of these have certain shortcomings. Two nuclear markers now exist. One exploits differences between two species of mouse in highly repetitive DNA sequences which can be visualized using in situ hybridization (Rossant et al. 1983; 1986). However, there is some evidence in these interspecific chimaeras, at least in the central nervous system, that cells from different species mix less freely than intraspecific combinations, resulting in unusually large patch sizes in the cerebellum and elsewhere (Goldwitz, 1986). The second nuclear marker is intraspecific and makes use of a transgenic mouse strain containing approximately 1000 copies of an exogenous β-globin gene integrated as a concatamer (Lo, 1986; Varmuza et al. 1988). Analyses of chimaeras using this marker have been reported (Thomson & Solter, 1988; Clarke et al. 1988) but reliable detection of a single concatemer in every nucleus is affected by nuclear size and cell packing density and, therefore, reqùires different sectioning conditions, for different tissues. Furthermore, as emphasized by Gardner (1985b), resolution of cell type and position in complex organs necessitates a cytoplasmic marker to define the perimeter of cells. In addition, cytoplasmic markers which can be used on wholemount material can greatly facilitate the analysis of chimaeras since this will permit the observation of clones both in the intact conceptus and in subsequently sectioned material. Unfortunately, all cytoplasmic markers so far described, which are usually based on null or thermolabile variants of known enzymes, have proved to be applicable only to a limited repertoire of tissue types where wild-type enzyme activity is high (West, 1984; Ponder, 1987). For this reason, none are adequate for marking all cell types during the early stages of fetal development.

This paper describes a cytoplasmic marker suitable for studying cell deployment during early organogenesis in the mouse. The marker was produced by making a transgenic strain of mouse (designated Tg(Act-lac Z)-l ) carrying the E. coli lac Z gene under the control of the rat β-actin promoter. Single ICM cells from this strain were introduced into blastocysts and preliminary observations on their descendant clones in midgestation (10th day) chimaeric embryos are presented.

The recovery and manipulation of all embryos was carried out in PB1 medium (Whittingham & Wales, 1969) containing 10% (w/v) fetal calf serum (FCS) instead of bovine serum albumin. For DNA injections, this medium contained 5μg ml^-1 of Cytochalasin B (Sigma). Eggs were incubated before and after injection in microdrops of standard ovum culture medium (Biggers et al. 1971) pre-equilibrated under oil at 37°C in an atmosphere of 5 % CO₂. Blastocysts before and after injection were incubated under similar conditions but in a medium (Stanners et al. 1971) modified as described previously (Gardner et al. 1985).

Outbred PO (Pathology, Oxford) mice were used in all experiments. Fertilized eggs were obtained from females maintained on a light regime of 14 h light/10h dark, the midpoint of the dark period being 19.00 h. Pseudopregnant recipient females were maintained on a similar light regime except that the midpoint of the dark period was midnight. The recovery and injection of eggs was essentially identical to that described elsewhere (Hogan et al. 1986) except that the DNA injection pipette was attached to tubing filled with paraffin oil (Boots, UK, Ltd) and controlled by a de Fonbrune suction and force pump. Injections were carried out in hanging drops in a Puliv chamber using a fixed-stage microscope (Ergoval, Zeiss-Jena).

A 4·3 kb BglII/ScaI fragment, containing the E. coli lac Z gene fused to the rat β-actin promoter (350bp), was isolated from pIRV-Neo-Act-lac Z as illustrated in Fig. 1. This was extracted from a 0·5% agarose gel, purified using a silica matrix (Geneclean) and redissolved in 5 nim-Tris, 0·5 mw-EDTA (pH 8·0) to a final concentration of approximately 8ngμl^-1.

High molecular weight DNA was isolated from tail biopsies of putative transgenic offspring (Hogan et al. 1986), digested with various restriction enzymes (Fig. 2) and run on a 0-5 % agarose gel. DNA was transferred over 4h to nylon filters (Hybond, Amersham) and covalently linked to the filter with ultraviolet light (5 min). The filters were prehybridized for 2 h at 65°C with 5 × SSC, 5 × Denhardt’s solution, 0·5 % SDS and 100 μg ml^-1 herring sperm DNA. Subsequently, they were incubated overnight at 65°C in the same mix containing a nick-translated 3·1 kb ³²P-labclled flam HI fragment of pIRV-Neo-Act-lac Z (Fig. 1). Southern blots were incubated at the same time with a nick-translated mouse α-globin

(Nishioka & Leder, 1979) probe (2·2 kb HinfI-SacI fragment). For the nick translation mixture 2·5 ng ml^-1 DNase I, 50 μCi [α³²P]dCTP, 10 μM each of dGTP, dATP and dTTP, 50mm-Tris-HCl (pH7 · 4), 10mm-MgSO₄, Imm-DTT, and 50 μg ml⁻¹ BSA were added together for 3 min at room temperature before adding 5 units of E. coli DNA polymerase I. After incubation for 2h at 14°C the nick-translated DNA was de-salted on a Sephadex G-50 spun column and added to the hybridization mixture. The filters were washed twice following hybridization in 2 x SSC (15 min each at 65°C), in 2 × SSC + 0 · 1 % SDS (for 20min at 65°C), and the final high stringency wash in 0 · 1 × SSC (for 10min at 65°C). The filters were exposed to preflashed Kodak XAR film or Fuji X-ray film in cassettes, containing two intensifier screens, at −70°C. The number of copies of transgene integrated was determined on developed films by scanning densitometry using a Gelman DCD-16 computing densitometer with automatic integration.

Ear punches from putative transgenic mice and their offspring were routinely stained for E. coli β -galactosidase activity. The tissue was washed in 0 · 1 m-phosphate buffer (pH7-4) and fixed for 15 min in 0 · 2 % glutaraldehyde (Gurr) in the same buffer containing 2mm-MgCl₂ and 5mm-EGTA. It was washed for 1 h in three changes of 01 m-phosphate buffer (pH 7 · 4) containing 2 mm-MgCl₂, 0 · 01 % (w/v) sodium desoxycholate and 0 · 02% (w/v) Nonidet P-40. Staining was carried out at 37 °C in a solution of the above buffer containing X-Gal at a final concentration of 1 mg ml^-1. The X-Gal was dissolved in dimethylformamide (40 mg ml^-1) and to this was added buffer, which in addition to MgCl₂ and detergent also contained 5 mm-K₃Fe(CN)₆, 5 mm-K₄Fe(CN)₆.6H₂O and 1 mm-spermidine hydrochloride. Ear tissue was left to stain for up to 48h. All reagents, unless specified, were obtained from Sigma.

Control, hemizygous Tg(Act-lac Z)-1 and chimaeric con-ceptuses were stained intact as whole mounts using identical conditions. Controls and hemizygotes were recovered and stained on the 8th, 9th and 10th days of gestation whereas only 9th and 10th day potential chimaeras were stained.

After staining, embryos were briefly fixed in Carnoys’ fluid (15 min), dehydrated and embedded in paraffin wax (m.p. 56 °C). They were serially sectioned at 8 μm and some were counterstained with eosin before mounting in DPX and viewed in a light microscope (Dialux 20EB; Leitz).

Ninth day conceptuses, derived from Tg(Act-lac Z)-1 males mated to PO females, were dissected from the decidua into PB1 + 10% FCS. Reichert’s membrane was removed. In all but one experiment the embryos were subsequently divided into anterior and posterior halves using fine glass needles (Beddington, 1987). The anterior halves were stained for f- galactosidase activity as described above. The posterior halves were labelled with [³H]thymidine according to a procedure described elsewhere (Tam & Beddington, 1987). After 3h in labelling medium, small clumps of posterior ectoderm were prepared from those embryos that had been identified as Tg( Act-lac Z)-1/ + from X-Gal staining. In one experiment, clumps were prepared from all embryos because no parallel X-Gal staining was undertaken. These clumps of labelled ectoderm cells were injected into host unlabelled PO 9th day embryos (Tam & Beddington, 1987). Injections were made either into the primitive streak region or immediately anterior to the heart. Embryos that had received grafts were cultured for 24 h in roller bottles in 50 % rat serum diluted in Dulbecco’s modification of Eagle’s medium (Flow Laboratories; Beddington, 1987) before being fixed and stained for β-galactosidase activity (see above).

After histochemical staining, embryos were refixed in Carnoy’s and processed for routine histology, sectioned at 6 μm, and subjected to autoradiography as described by Beddington (1981), except that the trichloroacetic acid treatment was omitted. Following development, after 4 weeks exposure at 4°C, the slides were scanned and labelled nuclei counted in every second section. The coincidence, or otherwise, of blue cytoplasm and nuclear silver grains was recorded for each section.

Blastocysts were recovered from matings between hemizygous transgenic mice and PO females, on the 4th day of gestation. The zona pellucida was removed in acidified Tyrode’s solution (Nicolson et al. 1975) and, because blastocysts do not appear to express lac Z and cannot be preselected (R. Beddington, unpublished observations), all the blastocysts were subjected to routine immunosurgery (see Nichols & Gardner, 1984). Isolated ICMs were disaggregated into single cells by the method described by Gardner et al. (1985) and the isolated cells cultured in α medium before injection. Host blastocysts were obtained from PO females on the 4th day of gestation.

Blastocyst injections were carried out using the two-instrument method described by Babinet (1980). Manipulations were performed in hanging drops in a Puliv chamber using an Ergoval microscope. Injected blastocysts were cultured for up to 1 h before being transferred to recipients that were on the third day of pseudopregnancy.

201 PO eggs, in three separate experiments, were injected with the 4 · 3 kb fragment containing the rat β-actin promoter fused to the lac Z gene. Of these 99 eggs survived and 68 were transferred to three recipients which subsequently became pregnant. 16 mice were bom but 4 were eaten by their mother before DNA could be extracted for analysis. The remaining 12 offspring were tested for bacterial β-galactosidase activity in ear punches and by Southern blot analysis of total genomic DNA. One neonate (R10) had large patches of blue cells in ear tissue stained with X-Gal. This animal and one other (R12) contained integrated transgenes (Fig. 2).

From densitometry measurements, R12 contained two copies of the transgene but there was evidence (not shown) of rearrangement of the injected DNA. Neither this animal nor its offspring, either during fetal life or after birth, showed any exogenous β galactosidase activity that could be detected histochemically. R12 will not be considered further in this paper.

R10 contained six copies of the transgene integrated in the usual head-to-tail concatamer formation found in many transgenic strains (Palmiter & Brinster, 1986). The transgene was transmitted to F_t progeny at a frequency of approximately 11 · 5 % (Table 1). If this low frequency of transmission were due to lethal effects of the transgene one would have expected a smaller litter size than that recorded (Table 1). It is more likely that the low transmission frequency stemmed from mosaicism in R10, which was consistent with the patchy histochemical staining observed in ear tissue.

15 F₁ progeny from R10 mated to PO females were examined for the coincident expression of β-galactosi-dase in ear tissue and the presence of the transgene in genomic DNA. Three Tg(Act-lac Z)-1 hemizygotes were identified histochemically and only these three mice contained the transgene by Southern blot analysis. Subsequently, among offspring from F₁ matings (both intercrossed and outcrossed) 10 animals out of 89 tested (11 · 2%) have failed to show staining despite the presence of the transgene (Table 1). Within the limits of Southern blot analysis, there is no evidence for rearrangement of the transgene and methylation differences are not pronounced (unpublished observations). There has been no instance of finding apparent enzyme activity in animals that do not carry the lac Z gene.

No homozygous F₂ progeny have been obtained from intercrossing F₁ hemizygous mice. Densitometry measurements on Southern blots, comparing the mouse a-giobin and transgenic copy numbers in F₂ progeny, identified likely homozygous offspring although there was not a clear-cut bimodal distribution. However, test breeding all transgenic F₂ progeny from 2 litters, both putative hemizygotes and homozygotes identified by Southern blot analysis, by outcrossing to PO animals (16 F₂ transgenics tested; 370 offspring) did not produce any litters in which all offspring expressed lac Z and, therefore, failed to confirm the presence of homozygotes. Consequently, we assume that homozygotes die in utero, although the stage at which they are dying is not known. Litter size is within the normal range at least up to the 14th day of gestation but is somewhat reduced by birth (Table 1).

In addition, the transmission frequency of the transgene is less than expected. Screening for enzyme activity within the first week after birth showed that in Tg(Act-lac Z)-1 × PO matings 154 out of 546 (28 · 2%) offspring inherited E. coli β -galactosidase activity. Even if this percentage is corrected for the known discrepancy between presence of the transgene and detectable enzyme activity (11 · 2 % ; see above), this still leaves a transmission frequency of only 39 · 4%, as opposed to the expected 50%. The corrected value obtained from 291 offspring from hemizygous intercrosses is 49 · 1%, as opposed to the expected 66 · 7% (accepting that there will be no liveborn homozygotes). Both males and females appear to transmit lac Z at a reduced rate. This suggests that the insertion of the transgene may not be neutral either during gametogenesis or in mature gametes.

Hemizygous conceptuses from the 8th, 9th and 10th days of gestation have been analysed for expression of lac Z in embryonic and extraembryonic tissues. Conceptuses were derived from matings between hemizygous males and PO females, except for one 10th day litter where the transgene was inherited from the mother.

Staining in the four embryos analysed is restricted to the epiblast, embryonic and extraembryonic mesoderm, cells in the embryonic endoderm layer lying along the midline in the region of the head process and the occasional cell overlying the primitive streak (Fig. 3A). Staining in the epiblast is punctate, as if the enzyme is localized to intracellular vesicles. Mitotic cells show a more generalized cytoplasmic staining. Mesoderm and presumptive definitive endoderm cells appear more strongly stained than the epiblast. The punctate staining makes it difficult to evaluate the precise percentage of positively stained cells in these tissues but within any one section at least 80 % of the cells contain a blue spot. Primitive endoderm and trophectoderm derivatives are uniformly negative and indistinguishable from controls. Control embryos have no blue staining in any tissue.

The general distribution of lac Z expression remains the were stained blue (Table 2). However, the high density of grains over 11 unstained cells not only masked much of the cytoplasm but also indicated that they were probably dead cells. The failure to detect blue staining in the other 3 cells may be an extreme example of the fading of staining during autoradiography. The intensity of blue staining in all donor cells was much reduced compared to sections which had not been subjected to autoradiography. Only 2 cells were found, both in the same embryo, that were stained blue but did not have grains over their nuclei. These cells may not have been labelled with sufficient [³H]thymidine before grafting (Tam & Beddington, 1987).

The injection of a single ICM cell into a blastocyst, and its subsequent participation in development, permits not only an analysis of the potential of individual cells but also serves as a very sensitive means of assessing whether or not constitutive expression of lac Z affects development. In these experiments, it was not possible to distinguish between hemizygous and wild-type ICM cells before injection. Therefore, only some blastocysts must have been injected with a transgenic cell. The frequency of normal development and formation of chimaeras, detected histochemically, is shown in Table 3. One of these chimaeras was recovered on the 9th day (early-somite-stage) and the other two on the 10th day. More extensive analysis of clonal chimaeras will be presented elsewhere. The object of this paper is to demonstrate the potential use of this transgenic marker, not to supply a detailed analysis of cell mixing and growth patterns in nascent primordia. Therefore, only unequivocal observations, which do not require substantiation by elaborate serial reconstruction, are presented below.

The 9th day, early-somite-stage chimaera (CI) contained only relatively few Tg(Act-lac Z)-1 cells (466 cells; counted in alternate sections) in the embryo and extraembryonic mesoderm (Fig. 4A). It is possible that not all donor descendants were stained in these tissues at this stage (see above), and, of course, there may have been more descendants in the primitive endoderm and trophectoderm lineages but these would have gone undetected. However, positively staining transgenic cells were distributed along the entire axis of the embryos, including representatives in the cranial neur-ectoderm and cells in the caudal and allantoic mesoderm. 88 cells lay anterior to the first somite and 130 cells were found posterior to the last (7th) somite. The distribution of the Tg(Act-lac Z)-1 cells in the embryo (Table 4) demonstrates the pluripotency of individual ICM cells. Furthermore, the spatial distribution of the donor cells indicates that cell mixing continues within nascent primordia such as the notochord (Fig. 4D).

The older chimaeras (C2 and C3) are more informative because the β-galactosidase activity is higher and detectable throughout the cytoplasm of individual cells (Fig. 4B,C). However, the extensive contribution of donor cells, particularly to the trunk region, precluded an accurate count of total donor cell contribution. In C2 approximately 50% of cells were of donor origin whereas in C3 the contribution was nearer to 20% (Fig. 4B,C).

Examination of serial sections revealed extensive chimaerism in all regions, although the heart, cranial neurectoderm and splanchnopleure were less extensively colonized than other tissues. Some degree of cell mixing was apparent in all tissues, darkly staining cells often being intermixed with wild-type unstained cells. However, in certain primordia there was evidence for limited coherent growth. For example, small patches were common in somites (Fig. 4E), neural tube, surface ectoderm and quite large coherent patches were observed in the gut, invariably seen as elongated collections of cells orientated along the long axis of the gut (Fig. 4F).

The objective of this work was to produce an ubiquitous, cytoplasmic, in situ marker, which did not compromise development and which could be used to distinguish donor and host mouse cells in the intact embryo and in histological sections. The transgenic strain produced may not wholly satisfy these aims but it does provide, for the first time, a means of studying the deployment of cells in any fetal primordium during early organogenesis.

The rat β -actin promoter was chosen because it is thought to be constitutively active in all cells, β -actin being an integral component of the cytoskeleton and cell motility (Uyemura & Spudich, 1980). The actin genes are a highly conserved gene family and, therefore, there was every reason to believe that the rat promoter (Nudel et al. 1983) would be effective within the mouse genome. With the proviso that only enzyme activity, and not mRNA synthesis, was measured in hemizygous embryos, it appears that the transgene first becomes active after implantation (no staining has been seen in putative hemizygous blastocysts; R. Beddington, unpublished observations). Initially the staining is punctate, as if the enzyme is localized to vesicles, but by the 10th day the entire cytoplasm stains blue. Punctate staining may be a feature of cells expressing the enzyme at low levels. Cleavage-stage blastomeres, with low bacterial β-galactosidase activity, show a similar punctate pattern (Ueno et al. 1987), but there was no evidence for segregation of enzyme activity with particular subcellular fractions. Therefore, the blue spots were considered to represent cytoplasmic oxidation centres which facilitate the deposition of insoluble pigment (Ueno et al. 1987).

At all developmental stages examined, staining is restricted to epiblast derivatives. The lack of activity earlier in development, and its absence in trophectoderm and primitive endoderm derivatives, is unlikely to be a function of the promoter. A larger fragment of the β -actin promoter is active in 2-cell embryos (Bonnerot et al. 1987), and the pIRV-Neo-Act-lac Z construct, introduced into blastocysts via retroviral infection, can produce positive clones in trophectoderm and endoderm cells (P. Savatier & R. Beddington, unpublished observations). Therefore, although expression in only a single transgenic strain has been studied, it seems likely that it is the integration site of the transgene that is responsible for this observed lineage restriction in lac Z expression (Palmiter & Brinster, 1986; Allen et al. 1988). If this is the case, the sequences flanking the integration site may prove informative with respect to differential gene activity in these primary tissue lineages.

Interestingly, during the latter part of gestation lac Z expression, judged by enzyme activity, is dramatically down-regulated, leading in some cases to failure to detect staining in ear punches (Table 1), and during the first few weeks after birth the number of positive cells decreases rapidly and extensively (R. Beddington, unpublished observations). However, with the exception of a few organs (such as liver) that are entirely negative, the loss of enzyme activity does not appear to correspond to any obvious pattern with respect to tissue or cell type. Inexplicable variation between animals of the same transgenic strain, in the synthesis of transgene mRNA and its pattern of inheritance, has been reported before (Palmiter et al. 1982, 1984). In the case of Tg(Act-lac Z)-1, it is not clear whether this progressive extinction of activity is a feature peculiar to an exogenous promoter, a position effect or a reflection of normal regulatory mechanisms employed to reduce levels of nonessential proteins. Clearly, an in situ marker provides a very sensitive method for monitoring this inactivation.

The apparent ubiquitous expression of lac Z in 10th day hemizygous embryos and the complete absence of staining in wild-type embryos is not sufficient grounds for assuming β-galactosidase to be a reliable cell autonomous marker. It is necessary to demonstrate that ubiquitous staining is not the product of diffusion, and that every transgenic cell expresses the marker. On a gross level, tissues such as the heart, which show weaker staining, show the same blue profile whether stained in the intact fetus or isolated by dissection before histochemistry. However, the grafts of [³H]thy-midine-labelled transgenic epiblast provide the most compelling evidence for cell autonomy since over 94 % of cells showed unequivocal coincidence of β-galactosidase activity and nuclear silver grains (Table 2). Allowing for negative staining in putative dead cells (11 cells), the parallel marking occurs in 98% of cells. It is true that these grafts all contributed predominantly to mesenchymal tissues and that diffusion may be greater in tightly packed epithelial structures. However, inspection of epithelial tissues in the clonal ICM chimaeras reveals frequent instances of negative cells surrounded by strongly staining transgenic tissue (e.g. Fig. 4E). In addition, studies using retroviruses containing the lac Z gene, producing positively staining clones either in vitro or in vivo, have found little evidence for diffusion of staining product (Price et al. 1987; Sanes et al. 1986; Turner & Cepko, 1987; P. Savatier & R. Beddington, unpublished observations). Therefore, one may be confident that every blue cell is indeed of transgenic genotype.

The introduction of a foreign gene which produces constitutive enzyme expression, albeit in the absence of substrate at the optimal pH, may be expected to compromise development. Clearly, hemizygous Tg(Act-lac Z)-1 embryos develop normally and are wholly viable and fertile when adult (Table 1). However, from monitoring the inheritance of lac Z expression and taking into account the percentage of transgenics that do not express active enzyme, it appears that both hemizygous males and females may transmit the transgene at a reduced frequency. The normal litter size in such matings (Table 1) argues against this reduced transmission frequency being explained by embryo loss in utero. Therefore, as has been reported for other transgenic strains (Palmiter et al. 1984), the transgene may be deleterious to haploid cells, compromising gametogenesis or fertilization. The failure to produce homozygous individuals is most likely due to an insertional mutation, a relatively common occurrence in DNA injection transgenics (Palmiter & Brinster, 1986), rather than to increased production of bacterial β galactosidase. Certainly, quite widespread expression of lac Z in other transgenic strains, which can be bred to homozygosity, does not appear to impair development (Allen et al. 1988; S. Darling, personal communication).

The most rigorous test for neutrality of a marker is to examine the behaviour of cells in competition with wildtype ones. For example, [³H]thymidine was shown to put labelled preimplantation cells at a selective disadvantage in chimaeras (Kelly & Rossant, 1976). The inability to identify transgenic ICM donor cells before injection, in the chimaeras produced here, makes the interpretation of the frequency of clonal ICM chimaeras (Table 4) complicated. The frequency of chimaeras from injection of single wild-type ICM cells ranges from 17 to 26% (Gardner & Lyon, 1971; Gardner & Rossant, 1979; Gardner et al. 1985). The frequency of chimaeras (7 · 5%; Table 3) appears low but considering that donor cells were obtained from outcross matings, to avoid injecting possibly deleterious homozygous transgenic cells, and that the transgene is transmitted at a reduced frequency the number of chimaeras identified by X-Gal staining is quite acceptable. More importantly the high contribution of transgenic cells seen in the 10th day chimaeras (Fig. 4B,C) strongly suggests that they are at no particular disadvantage in competition with wild-type cells. However, in both chimaeras, the contribution to the head and the heart is less than elsewhere in the fetus and, therefore, one cannot rule out that wild-type cells have some advantage in these regions.

Previous analysis of the embryo, as opposed to extra-embryonic membranes (Gardner, 1985; Cockcroft & Gardner, 1987), in chimaeras at midgestation has relied on separation of electrophoretic variants of isoenzymes. The resolution of such an analysis is entirely dependent on the precision of dissection into component tissues. Therefore, with the exception of certain specific tissues such as gut derivatives (Gardner & Rossant, 1979), somites (Gearhart & Mintz, 1972) and selected organs later in development where pigment or enzyme variants can be distinguished in situ (see West, 1984), the embryo has been treated as a single entity. Consequently, the pluripotency of single ICM cells has been inferred from the analysis of a few dissected tissues in a small number of adult clonal chimaeras (Gardner et al. 1985).

The distribution of stained cells in the clonal transgenic chimaeras demonstrates for the first time that the progeny of a single ICM cell are found in all histologically recognizable tissue primordia, extending from head to tail, in the midgestation embryo. Although not specifically identified by alkaline phosphatase staining, some large cells in the hindgut region, which could be primordial germ cells, were also stained in both 10th day chimaeras. In previous studies, test breeding has confirmed coincident germline chimaerism in adult coat colour chimaeras derived from single ICM cell injections (Gardner et al. 1985). This makes it likely that the germ fine was, indeed, also colonized in the two 10th day transgenic chimaeras.

The analysis of growth patterns in nascent primordia requires careful reconstruction to assess the extent of cell mixing. However, certain features are evident in these chimaeras from studying both whole mounts and individual histological sections. The distribution of transgenic cells along the entire axial length of the 9th and 10th day chimaeras, which includes cells in derivatives of all three germ layers, strongly suggests that there must be extensive mixing of epiblast cells during development of the egg cylinder, prior to gastrulation (Beddington, 1983). This is reminiscent of the observed intermingling seen in descendant clones prior to gastrulation in zebrafish (Kimmel & Warga, 1986). There is clearly cell mixing in the emergent notochord, the most strongly staining tissue at the early somite stage (Fig. 3B,C), of the 9th day chimaera (Cl; Fig. 4D). This is consistent with the intercalation of notochordal cells viewed in teleost fishes (Thorogood & Wood, 1987) and supports models of notochord growth invoking elongation by a combination of cell recruitment, cell rearrangement and shearing forces (Jurand, 1962; Bancroft & Bellairs, 1976; Jacobson & Gordon, 1976; Woo Youn et al. 1980).

In the 10th day chimaeras, the first signs of obvious coherent growth are apparent. Small patches are evident in individual somites, neural tube, surface ectoderm, endothelium and gut endoderm in individual histological sections. Such patches are not seen, for example, in cranial mesenchyme or the lateral mesoderm. However, serial reconstruction is required to establish the extent of contiguity both in patches and ostensibly mixed populations. As shown before (Gearhart & Mintz, 1972), somites are clearly not clonal in origin, nor is there any indication that the most recently formed somites arise from large coherent patches of presomitic mesoderm. This corresponds to the notion that there is cell mixing within the presomitic mesoderm (Tam & Beddington, 1987; Keynes & Stern, 1988). With respect to the gut, the impression of elongated clones in the midgut suggests that the growth is longitudinal not radial. In the neural tube, patches are distributed radially in transverse section indicative of radial growth at this stage. In the future, it is hoped that analysis of chimaeras, in which the donor cell contributes rather a small proportion of progeny to the chimaera, will prove very informative regarding patterns of growth during early organogenesis.

In summary, the staining in epiblast derivatives of hemizygous embryos, particularly on the 10th day, appears to be ubiquitous and cell autonomous. This offers the first opportunity for analysing in situ the developmental potential of individual pre- and postimplantation transplanted cells, and for examining the patterns of growth in chimaeric organ primordia. In addition, the migratory pathways of cells emanating from structures such as somites or neural crest can be charted. Heterotopic transplantation of marked tissue will also make it possible to assess the autonomy or otherwise of certain patterns of gene expression, for example Hox genes, implicated in pattern formation.

We would like to thank Drs P. Ingham, P. Savatier and A. J. Copp for useful criticism of the manuscript. This work was supported by the Imperial Cancer Research Fund.