An autoradiographic analysis of tissue potency in different regions of the embryonic ectoderm during gastrulation in the mouse

There is now strong evidence in rodents that all the foetal primordia are descended from a single population of cells, the embryonic ectoderm (Levak-Svajger & Svajger, 1974; Gardner & Papioannou, 1975; Diwan & Stevens, 1976; Rossant, Gardner & Alexandre, 1978; Gardner & Rossant, 1979; Beddington, 1981). During gastrulation the single epithelial sheet of embryonic ectoderm is converted into a highly complicated form, made up of a variety of tissue types and embodying the basic design of the foetus. This means that the key to foetal organization must lie in the orderly allocation of tissue primordia within the embryonic ectoderm. The analysis of chimaeras produced by orthotopic grafting of labelled embryonic ectoderm, at the late primitivestreak stage in the mouse, demonstrates that the allocation of tissues during gastrulation follows a regular pattern (Beddington, 1981; see below), of, in other words, that a fate map can be constructed. There are at least three possible explanations for such a predictable topography of presumptive tissues:

• (i) Populations of embryonic ectoderm cells may have acquired differences in their developmental potential such that their observed patterns of differentiation in the embryo simply represent the fulfilment of their potential.

• (ii) Embryonic ectoderm cells throughout the embryo may enjoy the same potential but be subject to localized influences in the egg cylinder which effect the selection of one developmental pathway, appropriate to a particular location, at the expense of other possible ones.

• (iii) Embryonic ectoderm cells may be heterogeneous in their developmental potential, the regionalization of particular patterns of differentiation resulting from the selection of cells with potential appropriate to that region.

In order to distinguish between these three possibilities it is necessary to analyse the potential of embryonic ectoderm cells from different regions of the egg cylinder and to compare this with their normal fate. The first two explanations can be tested at the tissue level whereas only a clonal analysis can discriminate between the second and third possibilities. This paper describes a series of experiments designed to discriminate between the first two possibilities.

In vitro chimaeras have been used to analyse tissue fate and tissue potential in the 8th day mouse egg cylinder. The normal fate of posterior embryonic ectoderm was assessed following orthotopic injection. This extended previous work on the developmental fate of anterior and distal embryonic ectoderm (Beddington, 1981). In addition, the potency of anterior, distal and posterior embryonic ectoderm was tested in heterotropic grafts. In these experiments it was possible to establish whether the differentiation of donor tissue in a heterotopic site conforms to the colonization pattern characteristic of its new location or whether it shows more identity with the normal fate of the donor cells.

Tritiated thymidine-labelled embryonic ectoderm from three defined regions of the 8th day embryonic egg cylinder was injected, either orthotopically or heterotopically, into synchronous unlabelled host embryos. These injected embryos were grown in culture for 36 h. Subsequently, the patterns of donor cell colonisation were analysed autoradiographically. Embryos entirely labelled with [³H]thymidine, to the same extent as donor embryonic ectoderm tissue, were also grown in culture for 36 h and processed for autoradiography (labelled controls). These embryos provided a standard for the dilution of label in the different foetal tissues over the culture period and also served as a control for any inadvertant variation in the autoradiographic procedure.

All host and donor embryos were obtained from random-bred CFLP mice on the morning of the 8th day of gestation. The embryos were dissected in PBI medium (Whittingham & Wales, 1969) containing 10% (v/v) foetal calf serum (FCS) instead of bovine serum albumin. Most of the parietal yolk sac, except for the region adjoining the ectoplacental cone, was removed.

The roller system and the media used for culturing injected embryos and labelled controls have been described in detail previously, and the efficiency of the culture system together with the justification for using [³H]thymidine as a single cell marker have also been discussed elsewhere (Beddington, 1981).

After removal from culture injected embryos and labelled controls were compared with embryos of the same age which had been maintained in utero until the end of the culture period (in vivo controls). This comparison included both gross physiological and morphological characteristics (Table 1). Any abnormal injected embryos were not included in the autoradiographic analysis of chimaeras.

After removal of Reichert’s membrane, 8th day embryos were placed in 30 mm bacteriological dishes (Sterilin) containing α-medium (Flow Laboratories), to which had been added 10 μCi/ml of [³H]thymidine (Radiochemicals, Amersham) made up to a specific activity of 10·5 Ci/mM. The embryos were cultured in this medium for 2 h at 37 °C in an atmosphere of 5 % CO₂ in air. During this period all the embryonic ectoderm nuclei become densely labelled with [³H]thymidine (Beddington, 1981). After labelling the embryos were washed for 15 min in 3 changes of PBI plus 10% FCS supplemented with 10 μM/ml of unlabelled thymidine.

One series of orthotopic injections was undertaken :

• (1) (i)

  Injection of posterior embryonic ectoderm into the posterior region.

  Six different heterotopic injections were carried out :

• (2) (i)

  Heterotopic injections into the anterior region

  • (a) injection of distal embryonic ectoderm ;

  • (b) injection of posterior embryonic ectoderm.

• (ii)

  Heterotopic injections into the distal region;

  • (a) injection of anterior embryonic ectoderm;

  • (b) injection of posterior embryonic ectoderm.

• (iii)

  Heterotopic injections into the posterior region :

  • (a) injection of anterior embryonic ectoderm ;

  • (b) injection of distal embryonic ectoderm.

Figure 1 shows the location of the 3 regions used as a source of donor cells and as the sites of injection.

After washing, labelled embryos were dissected with glass needles to isolate the appropriate fragment of embryonic ectoderm. Anterior embryonic ectoderm was removed from beneath the foregut invagination and following removal of excess lateral ectoderm consisted of a small rectangular piece of tissue (approximately 70 x 50 μm). Distal embryonic ectoderm was removed from the tip of the cylinder and again most of the lateral ectoderm was cut away to leave a small square piece of tissue (approximately 70 x 70 μm). Posterior embryonic ectoderm was dissected from the caudal end of the primitive Streak, just beneath the origin of the amnion, and trimmed to a size similar to that of anterior embryonic ectoderm. Usually, contaminating endoderm and mesoderm could be removed with glass needles but in some cases it was necessary to subject the fragments to enzyme treatment, incubating them for. 10min at 4 °C in a mixture of 0·5% trypsin and 2·5% pancreatin (Difco) in calciummagnesium-free Tyrode saline at pH 7·7 (Levak-Svajger, Svajger & Skreb, 1969). The ectoderm fragments were dissociated into small clumps of a size suitable for injection using a very fine hand-drawn Pasteur pipette. These clumps, consisting of about 20 cells (Beddington, 1981), were transferred to a micromanipulation chamber and injected into the selected region of a host embryo using a micromanipulator assembly (Leitz) as described elsewhere (Beddington, 1981).

All injected embryos which were normal on recovery from culture, and at least one labelled control in each experiment, were embedded in paraffin wax (m.p. 56 °C) and serially sectioned at 5 μm. Wherever possible a labelled control was included in the same wax block as injected embryos so that it would be immediately obvious if the autoradiography processing was defective.

The wax sections were hydrated before being incubated in a 5 % solution of trichloroacetic acid at 4 °C for 30 min. The slides were washed thoroughly in running water and then placed in distilled water and transferred to a dark room. Here they were covered with AR 10 fine-grain autoradiographic stripping film (Kodak Ltd,) and left to expose at 4 °C for 3 weeks. They were developed using D-19 developer (Kodak Ltd), and fixed in Kodafix solution (Kodak Ltd). Once the slides were dry they were scanned in a light microscope to determine whether or not the autoradiography was satisfactory. They were subsequently stained with haemalum and eosin, or haemalum alone, mounted in DPX (BDH Chemicals Ltd.) and scanned carefully in a light microscope.

Injected embryos were only considered to be chimaeric if :

• (1) They contained at least three labelled cells.

• (2) Grains over a single nucleus were found in at least two consecutive sections.

• (3) Labelled cells showed an equivalent grain density to that found over the nuclei of the same tissue in labelled controls (cells were considered to be dead if their nuclei were very densely labelled).

• (4) Labelled cells were well integrated into embryonic tissues and were not present as discrete unincorporated lumps.

The colonization patterns of donor tissue were recorded both in terms of the particular tissues colonized and also according to the regional location of the chimaerism. For this purpose the embryo was subdivided into three regions: the head region (anterior to the first somite and including the heart; the trunk region (that part which included all the somites) ; the caudal region (posterior to the last somite). The number of colonizing cells in each chimaera was estimated by counting every labelled nucleus in alternate sections. Since the sections were 5 μm in width it was considered unlikely that a single nucleus would span more than two sections. Indeed, the method of counting probably underestimates the actual number of colonising cells due to the low energy emission of tritium particles and hence their short penetration distance (Rogers, 1973). A labelled nucleus more than 2 μm from the film may fail to cause silver grain formation.

The developmental characteristics recorded for embryos from the seven classes of injected embryos are shown in Table 1. These may be compared with the same characteristics recorded for in vivo controls, which are also represented in Table 1. Although there is some variation between the seven classes of injected embryos in terms of the percentage of embryos in each class which acquire certain characteristics, such as the fusion of cranial folds or signs of axial rotation, there is no indication that any particular injection consistently produces adverse effects on subsequent development. If the somite numbers of each injected embryo class are compared separately with those recorded for in vivo controls no significant differences are found (í-test ; P ≥ 0·05). Similarly, the somite numbers of individual classes of injected embryos do not differ significantly from one another (z-test; P≥ 0·05). The number of normal injected embryos from each class which produced satisfactory autoradiographs is shown in Table 2. Exclusion of embryos from autoradiographic screening was the result of whole batches of slides having to be discarded due to defective processing. Table 2 also shows the number of chimaeras detected in each class as well as the number of embryos containing only dead donor cells or unincorporated lumps of donor tissue.

All eighteen of the normal embryos injected with posterior ectoderm provided satisfactory autoradiographs. The rate of chimaerism among these embryos is high with fifteen embryos containing colonizing donor cells (83·3%). One additional embryo (5·6 %) contained only dead cells of donor origin (Table 2). The pattern of colonization was very consistent among these embryos with the distribution of labelled cells being restricted to embryonic and extraembryonic mesodermal tissues (Table 3). Most of the chimaeras contained some dead cells of donor origin although no unincorporated lumps were detected. All of the chimaeras except one contained labelled cells in the loose mesoderm of the caudal region and in nine of these embryos (nos. 2, 4-9 and 13) the endothelial lining of the blood vessels, either arteries or veins, had been colonized in this region (Table 3). Three embryos were chimaeric in somite mesoderm, but only the most recently formed or penultimate somite had been colonized. The allantois was chimaeric in seven embryos (nos. 8-13 and 15) and in all but one of these chimaeras (no. 15) labelled cells in the allantois were contiguous with those in the posterior embryonic mesoderm. The distribution of donor cells was, in general, quite diffuse, labelled cells being interspersed among host cells with seldom more than three donor cells lying directly adjacent to one another. However, in blood vessels donor cells showed greater coherence in their distribution.

Thirty-eight normal injected embryos were recovered from culture following the injection of distal ectoderm into the anterior region. Satisfactory autoradiographs were obtained for 25 of these embryos. Two embryos contained only heavily labelled dead cells and a further two embryos had unincorporated lumps of labelled cells in the amniotic cavity (Table 2). Eleven (44%) of the remaining embryos were classified as chimaeras (Table 2). The distribution of colonizing donor cells in the chimaeras is shown in Table 4. Only three tissues had been colonized: head surface ectoderm, head neurectoderm and heart mesoderm. Most of the embryos exhibited surface ectoderm chimaerism (chimaeras nos. 1-7). Four chimaeras (nos. 6-9) had been colonized in the neurectoderm and of these two contained labelled cells in the adjacent surface ectoderm (nos. 6 and 7). Three embryos were chimaeric in the epimyocardial component of the heart (nos. 5, 10 and 11) and one of these (no. 5) had a very small separate patch of colonizing cells (7 cells) in the surface ectoderm. In all cases the colonizing cells were found in coherent patches suggesting that the cells remained together after grafting and that there was little cell mixing with host tissue. Six of the chimaeras contained additional dead donor cells either adhering to the amnion or to the surface of the embryo.

The injection of posterior ectoderm into the anterior region generated fourteen normal embryos all of which provided satisfactory autoradiographs. Nine embryos qualified as chimaeras (64·3 %) and one other embryo showed the presence of dead cells in the amniotic cavity (Table 2). From the distribution of donor cells in the chimaeras (Table 5; Fig. 2B-D) it is clear that chimaerism predominates in the surface ectoderm (nos. 1-8). Four of these embryos also showed colonisation of the anterior extreme of the foregut (nos. 1-4) and in all four cases this labelled region was continuous with the labelled surface ectoderm and, therefore, was considered to be in the ectodermal component of the foregut, the stomodeum (Fig. 2D). One embryo (no. 11) contained donor tissue in the epimyocardium of the heart. Once again labelled colonizing cells occurred only in discrete coherent patches and there was little indication of mixing with host cells. Five of the chimaeras contained dead donor cells in addition to colonizing cells.

Eighteen normal embryos, which had received injections of anterior ectoderm into the distal region, were processed satisfactorily for autoradiography. Twelve of these embryos proved to be chimaeric (66-7 %) and a further three embryos showed evidence of dead donor cells in the amniotic cavity (Table 2). The distribution of colonizing cells is shown in Table 6. In all the chimaeras it was only neurectoderm tissue which had been colonized. In nine of the chimaeras colonization was restricted to the trunk region (nos. 1-9) but in one case (no. 10) labelled cells were found extending from the trunk region into the posterior part of the embryo. In the two other chimaeras (no. 11 and 12) only caudal neurectoderm had been colonized. Labelled cells were not scattered throughout the neural tube or neural plate but were present as distinct patches within the neurectoderm. Dead donor cells were present in six of the chimaeras.

Twenty-five normal embryos were obtained from injections of posterior ectoderm into the distal tip of the egg cylinder. Two of these were lost during embedding but the remaining twenty-three provided satisfactory autoradiographs. Fourteen of these embryos (60-9%) were chimaeric and a further two embryos contained dead cells in the amniotic cavity (Table 2). Colonizing donor cells were present only in mesodermal tissues (Table 7). The majority of chimaeras had been colonized in the trunk loose mesoderm (nos. 1-10) and seven of these showed donor cells lining the caudal artery (nos. 1-7) and one (no. 10) was also chimaeric in somites on one side of the neural axis. Four other embryos were colonized exclusively in somites, two showing bilateral colonization (nos. 12 and 13) and two only unilateral colonization (nos. 11 and 14). Chimaerism was restricted to the trunk region. In blood vessels the donor cells tended to occur as coherent patches whereas in the loose mesoderm and somites, although patches were apparent, the donor cells were intermingled with host cells. Nine chimaeras contained an additional population of densely labelled dead cells.

Thirty normal embryos, which had received injections of anterior ectoderm into the posterior aspect of the primitive streak, were processed for autoradiography. Due to slipping of the autoradiographic film only sixteen of these could be screened for chimaerism. Eight embryos were classified as chimaeras (50 %). A further six embryos contained only dead labelled cells and one additional embryo had a lump of unincorporated donor tissue attached to the caudal vein (Table 2). Two of the chimaeras (nos. 1 and 2) were colonized only in the surface ectoderm of the caudal region (Table 8; Fig. 2A). One chimaera (no. 3) contained donor cells in both the caudal surface ectoderm and adjacent mesoderm. Four embryos (nos. 4-7) were colonized only in the caudal loose mesoderm and the last chimaera (no. 8) was colonized only in the allantois. Donor cells in the surface ectoderm occurred as coherent patches whereas in the mesoderm and the allantois they had a more scattered distribution. Chimaerism was restricted to the posterior region of the embryo except in one case where extraembryonic mesoderm had been colonized. Densely labelled dead cells were evident in the majority of chimaeras, both within the embryonic tissues and in the amniotic cavity.

Twenty normal embryos, injected with distal ectoderm, provided satisfactory autoradiographs. Thirteen of these embryos were chimaeric (65 %) and a further two embryos contained only dehd cells (Table 2). All except three chimaeras (nos. 11, 12 and 13) contained donor cells in the caudal loose mesoderm (Table 9). The allantois was colonized in all but two of the chimaeras (nos. 8 and 13). Three embryos had been colonized in the caudal vein (nos. 1-3) and one of these also had donor cells in the endothelial lining of the caudal artery (no. 1). Chimaera no. 4 had labelled cells lining the caudal artery in the trunk region. One embryo (no. 13) had been colonized in the mesodermal component of the amnion adjacent to the lateral limiting sulcus. Labelled cells showed a scattered distribution in the allantois and loose mesenchyme indicating that some mixing with host cells had occurred. The colonization of blood vessels tended to be rather sparse but in one chimaera (no. 4) nineteen cells were found lining the caudal artery in the trunk region and they occurred as a coherent patch. Nine of the embryos contained additional dead donor cells and one embryo (no. 6) had a lump of unincorporated donor tissue stuck to the mesoderm layer of the yolk sac.

In a previous study (Beddington, 1981) the developmental fate of anterior and distal embryonic ectoderm was analysed using an orthotopic grafting technique. An identical procedure has been used here to investigate the fate of posterior embryonic ectoderm. The overall pattern of tissue fate which emerges is as follows (Table 10):

(i) anterior embryonic ectoderm gives rise to predominantly definitive ectoderm derivatives : surface ectoderm and neurectoderm ;

(ii) distal embryonic ectoderm generates definitive gut endoderm, notochord and embryonic mesodermal derivatives, including somites ;

(iii) posterior embryonic ectoderm contributes exclusively to mesodermal tissues, both embryonic and extraembryonic.

It is clear, therefore, that different regions of the embryonic ectoderm give rise to different tissues. Such regularity in tissue allocation is not unexpected considering the well-documented fate maps, constructed during gastrulation, for embryos from other animal classes. What is, perhaps, significant is that these rudiments of a fate map for the mouse primitive-streak-stage embryo fit the general scheme of presumptive tissue topography found throughout the chordate phylum at an equivalent developmental stage (Pasteels, 1937; Nicolet, 1971). It remains to be seen whether such a consistent pattern of tissue fate reflects a common method of tissue allocation, or instead, is simply a ‘preview’ of similarities in subsequent morphogenetic mechanisms and tissue interactions essential to the initial development of the organ rudiments.

The heterotopic series of injections provides no evidence for rigid mosaicism in the embryonic ectoderm tissue during the later stages of gastrulation. Both distal and posterior embryonic ectoderm readily colonize surface ectoderm when placed in an anterior site (Table 10) despite the fact that this tissue is never colonized following their orthotopic injection (Table 10; Beddington, 1981). Similarly, embryonic ectoderm from all three regions shows the capacity to form a variety of mesodermal derivatives in heterotopic sites. However, although anterior embryonic ectoderm will colonise both embryonic and extraembryonic mesoderm when relocated in the posterior region, it does show a preference for definitive ectoderm differentiation wherever it is placed in the embryo (Table 10). Furthermore, neither anterior nor posterior embryonic ectoderm gave rise to gut endoderm or notochord in any experiment. Endoderm and notochord colonization was obtained only after orthotopic distal grafts (Beddington, 1981).

The ability of embryonic ectoderm outside the presumptive definitive ectoderm area (the anterior region) to form definitive ectoderm derivatives is consistent with studies on other vertebrate embryos. For instance, if the blastopore lip from an amphibian gastrula is transplanted to the presumptive ectoderm area it will form epidermal and neural structures (see Holtfreter & Hamburger, 1955). Chorioallantoic grafts of different regions of the chick blastoderm, at the head process stage, demonstrate that almost the entire blastoderm, and almost certainly only the epiblast component of it, retains the ability to form epidermis (see Waddington, 1952). In addition, experiments on primary induction in both amphibian (Spemann, 1938) and avian (Waddington, 1952; Gallera, 1971) embryos demonstrate that regions of the embryo normally destined to form other tissues are competent to produce neural structures. A comparable pluripotency is found during gastrulation with respect to mesodermal differentiation. Both amphibian (Holtfreter & Hamburger, 1955) and avian (Waddington, 1952) prospective definitive ectoderm can form any sort of mesodermal tissue. Therefore, with regard to epidermal, neural and mesodermal differentiation the fate maps in amphibian and avian grastrulae cannot be explained by any pre-existing mosaicism of developmental potential in different prospective tissues. The same would seem to be true of the primitivestreak-stage mouse embryo.

However, one cannot ignore the tendency of anterior embryonic ectoderm to form definitive ectoderm derivatives in heterotopic sites. There is no ready explanation for this behaviour, although it has been found in chick embryos, at a similar stage, that grafts of prospective definitive ectoderm show a certain reluctance to invaginate and that, in the absence of invagination, only neural structures are formed (Waddington & Taylor, 1937; Abercrombie, 1937). A similar failure to invaginate may account for the neural differentiation of anterior embryonic ectoderm in the distal region and its colonization of surface ectoderm in the posterior region. One might speculate that tissue fate becomes stabilized first in the anterior region and that this stabilization is associated with a resistance to any breakdown in epithelial organization. Some support for this idea comes from recent work on experimental teratomas. Pre-primitivestreak rat embryonic ectoderm rapidly loses its epithelial organization when grafted beneath the kidney capsule, whereas headfold-stage ectoderm, while producing some mesoderm, shows a greater tendency to maintain its epithelial structure (Svajger, Levak-Svajger, Kostovic-Knezevic & Bradamante, 1981),

Chimaerism in the gut and notochord occurs only following distal orthotopic grafts. This might suggest that the potential to generate gut and notochord is restricted to distal embryonic ectoderm. Such an interpretation would be consistent with experiments in birds. In chorioallantoic grafts of the chick embryo at the head-process stage only the anterior part of the primitive streak will give rise to notochord, although the ability to form gut is more widespread (Waddington, 1952). Furthermore, studies on primary induction, using either [³H]thymidine or the quail nucleolar marker to distinguish the grafted Hensen’s node, demonstrate that the gut and notochord in induced secondary axes are invariably of graft origin (Gallera & Nicolet, 1969; Hornbruch, Summerbell & Wolpert, 1979). This indicates that in avian embryos the differentiation of notochord, and to a lesser extent gut, is associated specifically with the anterior part of the primitive streak, or Hensen’s node. However, grafts of mouse distal embryonic ectoderm, presumed to be equivalent to Hensen’s node, show no inclination for autonomous differentiation, as judged by their failure to produce gut and notochord in heterotopic sites. This could be explained by the small size of the grafts ( ∼ 20 cells) compared with those used in chick experiment. A more important discrepancy is that anterior embryonic ectoderm readily produced definitive endoderm derivatives in experimental teratomas (R. S. P. Beddington, in preparation). Therefore the formation of gut and notochord in the mouse embryo cannot simply be attributed to a strict mosaicism in developmental potential. Certain influences or morphogenetic requirements peculiar to the distal region must also be important. However, while it is clear that distal embryonic ectoderm can respond to these cues, anterior and posterior embryonic ectoderm apparently cannot. One must conclude that the formation of gut and notochord in the embryo involves a rather precise interplay between the competence to form these two tissues and the ability to respond to certain cues in the distal region.

The results demonstrate that 8th day embryonic ectoderm does not behave as a mosaic tissue consisting of strictly demarcated areas, each destined to form a particular tissue by virtue of its appropriate restriction in developmental potential. This is contrary to the conclusions drawn by Snow (1981) based on experiments in which particular fractions of the embryonic egg cylinder, usually consisting of all three germ layers, differentiated in an apparently autonomous fashion when isolated in vitro. This was considered to be indicative of mosaic development. However, the autonomous differentiation of the fractions cannot be interpreted as a reflection of mosaicism in the embryonic ectoderm since most isolates also contained mesoderm and all of them had an endodermal component. The presence of other tissues is very likely to affect the development of embryonic ectoderm. For example, it is well known that the type of neural structure formed by amphibian ectoderm is dependent on the inducer and not an inherent property of the ectoderm (see Holtfreter & Hamburger, 1971; Toivonen & Saxen, 1968). Therefore, although regionalization may be apparent in explants where the germ layer relationships are not disturbed this does not mean that there is necessarily any regionalization in developmental potential within the embryonic ectoderm.

The distribution of labelled cells in the various tissues of the chimaeras, from both orthotopic and heterotopic injections, provides some indication of the patterns of growth within those tissues at the onset of organogenesis. For example, labelled cells were always found in well defined patches in the surface ectoderm, neurectoderm, notochord, gut and endothelial lining of blood vessels. In contrast, the distribution of labelled cells in certain mesodermal derivatives, such as lateral plate mesenchyme and the allantois, was mote diffuse.

Patterns of growth and the degree of cell mixing during development have been studied extensively in mouse chimaeras and mosaics (West, 1978). Autoradiographic analysis of aggregation chimaeras at the blastocyst stage has shown that there is little or no cell mixing during cleavage and blastocyst formation (Garner & McLaren, 1974). However, if only a few cells (2-3 cells) from a 4th day embryo are injected into the blastocyst their progeny can be detected in very small samples of every adult tissue examined (Ford, Evans & Gardner, 1975). In addition, the injection of single 5th day embryonic ectoderm cells generally results in chimaerism throughout the foetus and also in the exttaembryonic mesoderm (Gardner & Rossant, 1979). As prospective foetal tissues appear to have a definite topographical arrangement at the late primitive-streak stage this must mean either that there is considerable cell mixing in the embryonic ectoderm prior to gastrulation or that growth in the egg cylinder occurs such that descendant clones tend to be aligned with, and extend along, the length of the embryonic axis. If this were not the case, one would expect the clonal descendants of a primitive ectoderm cell injected at the blastocyst stage to be contained within just one or two prospective areas, and, therefore, to give rise only to a limited number of foetal or adult tissues. Instead it seems that progeny of a single primitive ectoderm cell must be distributed so that they are present in all prospective areas, ranging from the anterior region of the egg cylinder (prospective surface ectoderm and neurectoderm) to the posterior end of the primitive streak (prospective embryonic and extraembryonic mesoderm).

It is interesting that growth in the surface ectoderm and neurectoderm appears to be coherent at the onset of organogenesis (see above). The analysis of patch size in the epidermis of adult mouse chimaeras, using electrophoretic allozyme variants, reveals that fragments of skin less than 1 mm² in area nearly always contain cells of both host and donor origin (lannacone, Gardner & Harris, 1978). This indicates that the final patch size in the epidermis is very small, and therefore, extensive cell mingling must occur after the initial delamination of the surface ectoderm. Similarly, the estimated coherent clone size in the retinal pigment epithelia, a neurectoderm derivative, in both mouse chimaeras and mouse X-inactivation mosaics, on the 13th day of gestation is extremely small, constituting no more than one .or two cells (West, 1978). Although there is evidence that the distribution of cells is not entirely random at this stage, since descendant clones of one phenotype tend to be clustered into sectors (Sanyal & Zeilmaker, 1977), the distribution once again points to considerable cell mixing among neurectoderm cells prior to the foundation of the retina pigment epithelia.

In conclusion, it appears that clonal growth during mouse development must alternate between being coherent and involving extensive cell mingling. During cleavage and blastocyst formation growth is coherent but after implantation there must be cell mixing, at least in the embryonic ectoderm. At the onset of organogenesis much of the earliest differentiation and morphogenesis of the germ layers appears to be associated with coherent growth but, once again, this seems to be followed by a phase of cell mixing before the final allocation of the different organ primordia.

I would like to thank Professor R. L. Gardner, Dr V. E. Papaioannou and Miss G. Porter-Goff for invaluable discussion. This work was supported by a Medical Research Council Studentship.