A fate map of the epiblast of the early chick embryo

The early stages of development in amniotes are particularly important subjects to study because it is at this time that the early body plan is laid down. However, there are no detailed fate maps available for these stages in any amniote. In the chick embryo, which is probably the best studied, one reason was the absence of a suitable staging system for the preprimitive streak embryo until 1976, when Eyal-Giladi and Kochav produced their detailed stage table. Most of the chick fate maps that have been published were produced before this staging system became available.

Now that both an accurate staging system and a better method for mapping (carbocyanine dyes) exist, we have used them to construct a detailed fate map for the preprimitive streak and early primitive streak chick embryo. The results reveal an orderly pattern of cell movements, consistent with previous observations of gross morphogenetic movements in the early blastoderm. The present results also reveal considerable overlap between different prospective areas, as has recently been described for the mouse (Lawson et al., 1991), rather than sharp dividing lines between different territories, as suggested by some older publications. Less expected was the finding that certain presumptive areas appear to move independently of others, even against the predominant gross pattern of movements.

Fertile hens’ eggs (obtained from Coppocks Poultry Farm, Carterton) were incubated at 38°C for 0–8 hours to obtain embryos between Eyal-Giladi and Kochav (1976; in Roman numbers for preprimitive streak stages) stage-X and Hamburger and Hamilton (1951; in Arabic numerals for later stages) stage-3. Embryos were explanted in Pannett and Compton (1924) saline by the technique of New (1955), with modifications (Stern and Ireland, 1981). Following marking with DiI and/or DiO (see below), the embryos were incubated for 24-36 hours at 38°C in a humid atmosphere.

The carbocyanine dyes 1,1′-dioctadecyl-3,3,3′,3′-tetramethyl indocar-bocyanine perchlorate (Molecular Probes, Inc.) (DiI) and 3,3′-dioctadecyloxacarbocyanine perchlorate (Molecular Probes Inc. Di-O-C18-(3)) (DiO) were used for labelling, following methods previously described (see Stern, 1990; Selleck and Stern, 1991). These intensely fluorescent dyes are lipophilic, become incorporated into cell membranes and are not transferred between cells (Honig and Hume, 1989; Serbedzija et al., 1990; Wetts and Fraser, 1989).

Briefly, DiI or DiO was first dissolved at 0.5% (w/v) in absolute ethanol and this diluted 1:9 with 0.3 M sucrose in distilled water at 40°C (see Serbedzija et al., 1990; Selleck and Stern, 1991). The dye was applied to the desired region using gentle air pressure, using micropipettes pulled from 50 μl capillaires (Sigma) with an Ealing vertical electrode puller. A small group of cells (5-30) could thus be labelled.

A total of 1326 injections were made into 726 embryos (600 embryos were labelled with both dyes), applied to different positions (see Table 1). Embryos were not used for subsequent analysis if after incubation: (i) they had developed abnormally; (ii) most of the embryo contained label; or (iii) dye was seen as fine granular material over the surface of the tissues. 358 of the embryos developed normally to stages 11–13. A total of 332 injections, in 200 embryos, was used for the final analysis.

To use a more accurate staging system than those published, a quantitative method was used (Fig. 1), in which the position of the anterior border of the hypoblast sheet was used as a reference for staging. An eyepiece graticule divided into 100 units was used. Using a zoom in the dissecting microscope (Nikon SMZ-2T), the visible size of the area pellucida was adjusted to fit exactly into the 100 eyepiece units, with position ‘0’ corresponding to the posterior edge of Koller’s sickle. Using our scale, the width of the sickle varies between 5 and 12 units. The position of the anterior edge of the hypoblast at the midline was recorded, as was the site of injection. A similar method was used to record the position of the site of injection relative to the width of the area pellucida. In this way, a grid was constructed from which the positions of injection sites were mapped.

According to this method, in stage XI embryos, the anterior border of the hypoblast sheet had reached a position of less than 30; stage XII was ascribed to a hypoblast sheet extending to position 30-69; stage XIII corresponds to a position in the range 70-100. Stage XIV embryos were those in which a cellular bridge visible posterior to Koller’s sickle (see Eyal-Giladi and Kochav, 1976) and/or the sickle itself has become V-shaped. Embryos with a primitive streak were used up to the stage at which the tip of the streak had reached no further than the 55 position.

This quantitative method also allowed us to assess retrospectively the original position of marked cells in those cases where the orientation of the embryo had not been assessed correctly at the time of marking. In total, the orientation had been assessed correctly in 323/358 (90%) embryos. It is interesting that even stage 2 embryos, which already have developed a primitive streak, sometimes (3/111) changed the orientation of the embryonic axis during subsequent culture. Because of this, we analyzed particularly carefully each of the 35 embryos in which the orientation had shifted from the original: in none of these did the fate of the labelled cells fall outside the boundaries of that presumptive region as determined from the remaining 323 embryos.

After incubation, the embryos were fixed in PBS (pH 7.0) containing 0.25% glutaraldehyde and 4% formaldehyde for at least 1 hour and stored in this fixative. All embryos were mounted in cavity slides and examined in toto with an epifluorescence microscope (Olympus Vanox-T, with 200 W high-pressure mercury lamp). DiI was visualized using 547 nm (green) peak excitation, where it emits at 571 nm (red); DiO was excited at 484 nm (blue) and observed by its emission at 507 nm (green). In a few cases, labelled embryos were also observed using a BioRad MRC500 confocal laser scanning microscope.

In addition to examination of the fixed embryos as whole mounts, some embryos were processed histologically to confirm the localization of the labelled cells. For this, the fluorescence had first to be photoconverted to an insoluble product by photooxidation of 3,3′-diaminobenzidine (DAB) exactly as described previously (Stern, 1990; Selleck and Stern, 1991). After this, embryos were dehydrated in a series of alcohols, embedded in fibrowax, sectioned at 10 μm, mounted on glass slides and dewaxed in xylene before being mounted in DePeX for bright-field observation.

The results obtained (based on a pattern of 873 contributions to different tissues from 332 injections analysed), are shown in Figs 2-3, and specific examples shown in Figs 4–5.

We decided to standardize the positions of the injections using a scale of orthogonal coordinates, adjusted to 100 units along each the anteroposterior axis and the left-right axis, as described in Materials and Methods. To standardize the temporal aspect of the maps, we used the position of the anterior border of the hypoblast at stages X-XIII and the anterior tip of the primitive streak at stages XIV-3 (Fig. 1). In this way, a continuous scale was obtained, spanning these stages. For the summary maps shown in Figs 2 and 6, however, embryos were grouped by the stage criteria of Eyal-Giladi and Kochav (1976).

The following description summarises the main results emerging from detailed analysis of the data, classified according to each cell type containing labelled cells.

111 injections contributed labelled cells to the gut (Figs 2A, 4D,I). The gut endoderm differs from most other presumptive areas in that it is quite compact at all stages studied. At stage X, no presumptive gut cells are found in the midline except for a very small component at the posterior end of the embryo; they form a wing shape, hinged at the posterior midline (Fig. 2A). In general, this presumptive area converges towards the midline and moves anteriorly at the stages studied. The anterior movement seems to occur in two distinct stages (Fig. 3): a first one, at stages X-XI, when the anterior border of the prospective gut tissue at the midline moves at the same rate as the anterior border of the hypoblast. The second stage occurs at stages XIII-XIV, with the midline component of the prospective gut region reaching the centre of the blastoderm at stage XIV. Between stages XII-XIII, there is no apparent anterior movement at the midline. The most lateral cells contributing to the gut converge towards the midline up to stage XIV. In general, our impression is that injections into the midline of the presumptive gut region produced labelled cells in only the dorsal part of the gut at stages 11-13; more lateral regions of the blastoderm contribute to other, lateral/ventral regions of the gut (data not shown).

Only a few (20) injections contributed cells to this region (Fig. 5A). For this reason, it is difficult to draw general conclusions about its movement. From the data obtained, it can be seen that this presumptive population tends to be restricted to the midline of the embryo at all stages studied (Fig. 2B). When comparing the three stages that contributed more than one data point (XI, XII, 2–3), a tendency for this area to move anteriorly during development can be seen.

93 injections contributed labelled cells to the chordamesoderm (Figs 2C, 4I,K). Like the presumptive gut, this region is wing-shaped at stage X, hinged at the posterior midline; the movements of the two regions are also similar and occur in two distinct stages (Fig. 3). At stages 2–3, the presumptive chor-damesodermal territory is wider than that of the gut at the same stage (Figs 2C,6).

49 injections contributed to the somitic mesoderm (Figs 2D, 4G,H,K, 5C): 40 to the medial halves (Figs 2D, 4K, 5C) and 18 to the lateral halves (Fig. 2D). Nine of the injections (two at stage X, three at stage XI, two at stage XIII and two at stage 2-3) contributed labelled cells to both halves (Fig. 4H). At stage X, the lateral somite territory is contained completely within that of the medial somite, but the overlap gradually diminishes. The separation appears to be brought about mainly by the medially directed movement of the medial-half somite population, whilst the lateral cells remain at their original position. The presumptive medial-half-somite cells, in addition, show a tendency to move anteriorly at these stages.

10 injections gave rise to labelled cells in the intermediate mesoderm (Figs 2E, 4G,H). Their positions are sparsely distributed over many regions of the blastoderm and it is therefore difficult to draw definite conclusions about their movement.

38 injections contributed to the heart (Figs 2F, 4D,H). From stage XII to stages 2-3, there is relatively little change in the position or size of the presumptive heart territory.

Cells contributing to the lateral plates (88 injections) are widely spread around the surface of the blastoderm, but still forming wing-shaped territories hinged about the posterior midline (Figs 2G, 4G,J,L, 5D). There is little indication of movement before formation of the primitive streak.

211 injections contributed cells to neural tissues (Figs 2I,J, 4A-C,E, 5B), of which 65 were restricted to the forebrain, anterior tip of the neuroepithelium (prospective olfactory region) or optic lobes (Figs 2J, 4A-C, 5B). The neuroectoderm territory, like the prospective lateral plate mesoderm, is widely distributed over the surface of the blastoderm at all stages studied and shows no particular movement, except for anterior movement of its posterior border from stage XIV, leading to absence of this presumptive cell type from the posterior-lateral region. The presumptive olfactory and optic regions form a discrete sub-region of the neuroectodermal territory, which gradually becomes localized at the midline. showing anteriorly directed movement from stage XI (Fig. 3). Several injections produced descendants in neural tissues and in lateral plate mesoderm, and the two labelled regions were continuous at the posterior end of the embryo, level with the regressing Hensen’s node at stage 11-13. A similar phenomenon has been observed in the urodele Triturus pyrrhogaster (Hama, 1978).

These tissue types are considered together for three reasons:

(1) because at early stages and in posterior regions the boundary between amnion and surface ectoderm is poorly defined, (2) because it was sometimes difficult to distinguish unambiguously the precise layer containing the labelled cells and (3) because many of the injections contributed to both tissues. 244 injections contributed to these regions (Figs 2H, 4A,C). In general, cells contributing to surface ectoderm and to extraembryonic tissues (amnion, yolk sac and its stalk, etc.) are very widely distributed over the entire surface of the prestreak blastoderm. However, at stage XIV, there is a large lateral/posterior region devoid of cells with these fates (Fig. 2H).

Six injections (one at stage XI, one at stage XII, four at stages 2-3; in Fig. 4G) gave rise to labelled progeny. In these cases, labelled cells were found laterally in anterior regions, and more medially in more posterior parts of the embryo (as described by Hama, 1978 for Triturus pyrrhogaster). In the majority of cases, the most anterior cells were located in the lateral mesoderm and gradually shifted via the intermediate mesoderm to the lateral halves of the somites at more posterior positions.

In general, many injections contributed progeny to large regions of the anteroposterior axis of the embryo. We are therefore unable to ascribe particular regions of the anteroposterior axis to specific areas in the early stages of development. This is consistent with the findings of Selleck and Stern (1991) who reported that even progeny derived from some single labelled cells spans extended regions of the embryonic axis. However, one exception is the sensory areas associated with the forebrain, where some injections only contributed progeny to this region.

We also analyzed the distribution of descendants of labelled cells in terms of whether or not they crossed the midline. The results of this analysis are shown in Table 2. In general, labelled progeny tend to be restricted to one side of the embryo. When this analysis is extended by comparing the distribution in different tissue types, it is seen that the descendants of presumptive somite cells do not follow this rule: labelled progeny tend frequently to cross the midline (χ², 1 d.f. = 7.06; P<0.01). We also compared the frequency with which progeny crossed the midline in terms of the stage at which the injection had been done. No significant differences were found.

H2>Assessment of the technique used

Fate maps of the early embryo have been constructed by many authors, both before the formation of the primitive streak (Kopsch, 1926; Gräper, 1929; Wetzel, 1929; Kopsch, 1934; Pasteels, 1937; Malan, 1953; Vakaet, 1970, 1984) and after (e.g. Rosenquist 1966; Nicolet, 1971). These maps were produced either by local killing, or marking using spots of water-soluble dyes applied through the vitelline membrane in ovo, or using carbon particles. However, these studies all suffer from technical defects. Transplantation of marked cells might have disturbed the spatial, and perhaps temporal, organization of the tissues to be mapped, while carbon or carmine particles may not always follow the cells in their movements. In addition to the disadvantages of these methods, the classical maps suffer from the problem that they were produced before a reliable staging system was available for early stages of development.

Recently, both some new marking methods (the carbocyanine dyes, DiI and DiO; see Honig and Hume, 1989; Wetts and Fraser, 1989; Serbedzija et al., 1990; Stern, 1990; Selleck and Stern, 1991) and a new staging table (Eyal-Giladi and Kochav, 1976) have been described, allowing the labelling of small groups of cells at precisely controlled stages of the very early chick blastoderm without using transplantation or particulate markers.

We opted for a quantitative method for assessing the position of sites of injection and for staging the embryos, so that data obtained from different embryos can be compared. The positions of injection sites were measured with respect to a pair of orthogonal axes (anteroposterior and mediolateral, each divided into 100). By this method, size variation along each of these axes of the area pellucida was eliminated by scaling.

To avoid complications introduced by possible left/right asymmetry of the embryo (c.f. Strehlow and Gilbert, 1993), we decided to collect data from the left half of the blastoderm; the degree of left/right asymmetry remains to be assessed in future studies. It is worth mentioning, however, that injections to the left side often gave rise to labelled cells that extended to the contralateral side of the embryo, as found by Gallera and Nicolet (1969) and Rosenquist (1966) in chick and Lawson et al. (1991) in mouse.

In some cases, the embryos were observed under fluorescence immediately after injection, to confirm the position and size of the label. After incubation, embryos were again observed by fluorescence and the germ layers containing labelled cells discriminated by focussing through a high numerical aperture objective (10×, NA=0.40 and 20×, NA=0.65). If there was doubt concerning which tissues contained labelled progeny, the embryos were photooxidised and examined in histological sections.

It is worth pointing out that these fate maps were constructed by labelling groups of cells rather than single cells. Therefore, when an injection produces progeny in more than one tissue, we cannot formally distinguish between two possibilities: (a) each presumptive area contains a mixture of cells with different presumptive fates and (b) each area contains common progenitors for these fates.

In general, the distribution of different prospective cell types in the early chick embryo agrees broadly with findings made in other vertebrate species (zebrafish, amphibian, mouse). There are a few minor differences, however. Among them, in the chick, the region contributing to mesodermal cell types appears broader than in the mouse (Lawson et al., 1991). Also, compared to the mouse and urodele fate maps (see Lawson et al., 1991 for summaries of both), the region of cells contributing to the notochord in the chick is also broader. In contrast, the region contributing to gut endoderm appears marginally broader in the mouse than in the chick.

Our maps show that specific prospective areas change shapes and positions during development. From these changes, the gross pattern of movements can be extrapolated. Overall, the predominant movements for most areas between stages X and 3 are convergence towards the midline and anterior movement along the midline, in the posterior half of the area pellucida. Different presumptive tissues undergo these movements at different times. For example, ‘axial’ tissues such as the chordal mesendoderm undergo convergence at stages X-XI and appear to move anteriorly in two steps, at stages X-XI and XIII-XIV (Fig. 3). The presumptive prechordal plate is already at the midline at stage X-XI. The posterior-medial parts of the prospective neuroectoderm and of the notochord territories appear to move anteriorly together with the tip of the primitive streak.

This overall pattern, convergence towards the midline and anterior migration, has been described previously for early stages of the chick (Patterson, 1909; Graeper, 1929; Wetzel, 1929; Pasteels, 1937; Spratt, 1946; Vakaet 1960, 1970, 1985) and other vertebrates, such as teleosts (Oppenheimer, 1936; Pasteels, 1936; Ballard, 1973, 1981, 1982, Trinkaus and Fink, 1992), amphibians (Nakamura, 1938; Pasteels, 1940; Nieuwkoop and Sutasurya, 1979) and reptiles (Pasteels, 1937). In the mouse, Lawson et al. (1991) describe a similar convergence (“anisotropic” spread between members of a clone, orientated towards the forming primitive streak). Thus, the “Polonnaise” movements that have been described in the literature can be seen as an overall trend in the epiblast. However, when individual prospective areas are followed, it is found that many depart from this general pattern.

After stage XIII, but before the appearance of the primitive streak, the anterior border of the presumptive olfactory placode and optic lobe, notochord, medial somite and prechordal plate areas reach the centre of the area pellucida. This observation agrees with Spratt’s (1946) careful study of epiblast movements. He describes a convergence of the whole epiblast to the posteromedial part, which could be likened to a fan, hinged at the centre of the area pellucida and closing before the primitive streak appears. The centre of the area pellucida does not appear to move until later (stage 3), during primitive streak elongation.

After stage XIV areas such as the heart and neuroectoderm (Figs 2F,I, 6) appear ‘T’-shaped, but are not all coincident. Vakaet’s (1984) study of cell movement seems best to explain how this shape arises from the patterns seen at earlier stages. He showed not only medial convergence towards the posterior streak, but also anteromedially directed movement just in front of the elongating streak at stage 2. Our study suggests that this movement may start even before stage XIV.

Classical fate maps of the chick embryo were drawn with the different presumptive territories separated by sharp borders (Patterson 1909; Graeper, 1929; Wetzel, 1929; Pasteels, 1937; Rosenquist 1966; Vakaet, 1984, 1985; Nicolet, 1971). However, more refined techniques have now become available. These are starting to indicate that, in general, such sharp borders do not exist. For example, in the frog, Wetts and Fraser (1989) report that the descendants of labelled ectoderm cells become mixed, albeit slowly, a process that continues through-out early development. During gastrulation, some of this mixing is driven by cell intercalation (Shih and Keller, 1992a,b; Keller et al. 1992), as is also seen in teleosts (Trinkaus and Fink 1992). In the French frog, Pleurodeles waltl, Delarue and Boucaut (1992) have demonstrated finer differences in the extent of cell mixing between deep and superficial circum-blastoporal cells: dorsally, deep and superficial cells intermix more extensively than ventrally. In the zebrafish (Kimmel and Law, 1985a,b; Kimmel and Warga, 1988; Warga and Kimmel, 1990; Ho, 1992) and preimplantation (Winkel and Pedersen, 1988) and postimplantation mouse (Lawson et al., 1991); however, there is much more extensive intermixing of cells, resulting in some indeterminacy in the patterns of descendants derived from single, identified cells. Lawson et al. (1991) summarised this with particular clarity, by concluding that “morphogenetic movements occur in the presence of extensive, although not indiscriminate, cell mixing in the epiblast, and that descendants of a single progenitor may be spread widely, and also be present in different germ layers” (p. 905).

The present experiments reveal two, apparently contradictory, trends. First, the finding that certain presumptive cell types in the epiblast undergo rather different movements to those of other presumptive regions suggests that there is indeed extensive cell mixing. Second, the finding that descendants of a small group of cells, labelled by a single dye injection, tend to contribute only to a few, more or less adjacent tissue types, suggests that this mixing is not as extensive. This is the case in most regions except the posterior midline, where many presumptive cell populations come together and overlap. One way to reconcile these observations is to suggest that cell mixing is fairly widespread, affecting large areas of the epiblast, but that it is not random and that cells do not wander to very distant sites (c.f. Lawson et al., 1991: “not indiscriminate”; see above).

Interestingly, certain presumptive cell types move in characteristic ways, sometimes against the prevailing currents. For example, the prospective lateral plate mesoderm territory does not appear to change position between stages X and 3, even though this territory overlaps with others, which do move as described above from stage X. Likewise, the presumptive optic lobe and olfactory areas seem to move differently from other surrounding cell fates. This could be taken to indicate that some prospective cell types are already specified at very early stages of development, perhaps as early as stage X, although we cannot formally rule out the possibility that morphogenetic cell movements are not well coordinated at these early stages of development. The exact degree of such specification for different cell types may vary between different vertebrate classes, giving rise to the apparent discrepancies in the findings of different authors in terms of the degree of cell mixing for different species.

Slack (1991) has argued that a comparison between fate maps and specification maps is of great value in identifying those regions that require cell interactions (‘induction’) to define the fates of the cells contained in them. In addition to the present study, fate maps of the early embryo before the formation of the primitive streak were published by Kopsch (1926), Gräper (1929), Wetzel (1929), Kopsch (1934), Pasteels (1937), Malan (1953) and Vakaet (1970, 1984, 1985). ‘Specification maps’ for the unincubated blastoderm were produced by Hoadley (1926a,b,c,d, 1927), Olivo (1928a,b), Murray and Selby (1930), Waddington (1933, 1935), Butler (1935), Dalton (1935), Rudnick (1932, 1935, 1938a,b, 1944, 1948, 1961), Hunt (1937), Spratt (1940, 1942, 1947) and Rawles (1943). These were generated by culturing pieces of blastoderm on the chorioallantoic membrane of a host embryo, in a plasma clot or on the surface of another early blastoderm. Unfortunately, these specification maps are very crude because no good staging system was available at the time, because the pieces isolated are often very large and often comprised more than one germ layer.

In addition, many reviewers (e.g. Pasteels, 1945; Romanoff, 1960; Nicolet, 1971; Balinsky, 1975) have failed to distinguish between fate maps and specification maps and have amalgamated results from the literature to construct composites that probably have little value. The maps produced by Butler (1935) and by Rudnick (1948) stand out from others because they appear to have been constructed more carefully by attempts to separate the layers. In the former case the description of the early embryos is clear enough to allow us to conclude that it refers to embryos at around stage XII.

We have therefore compared our results with those of Butler (1935) and Rudnick (1948) (Fig. 6) for stages XII and 2-3, respectively. Several areas, like neuroectoderm, eye, notochord and heart are in equivalent positions in our fate maps and in the published specification maps. The only territory that does differ is the presumptive gut. This is smaller in our fate map than in Butler’s specification map. When parts of the epiblast (including area opaca) are isolated and cultured, they become gut irrespective of their original positions. However, in normal development, only the posterior one third of the epiblast becomes gut. This indicates that cell interactions are required after stage XII for restricting the gut territory, implying that one or more inductive interactions are involved in the early development of the gut.

Staining with monoclonal antibody HNK-1 reveals a mosaic, salt-and-pepper pattern in chick embryos at stages XII-XIII (Canning and Stern, 1988). At stages XII-2, staining is graded in the posterior-to-anterior and medial-to-lateral directions, with the primitive streak containing more immunoreactive cells than more remote regions of epiblast. By immunogold labelling and ablation experiments, Stern and Canning (1990) showed that the HNK-1-positive cells at stages XII-XIII are pre-cursors of the mesendoderm of the early primitive streak. In agreement with these patterns, our fate maps, as well as those of previous authors, show that the posterior epiblast and that surrounding the streak contribute more cells to the mesendoderm than anterior and lateral regions.

The finding that the epiblast contains a mixture of HNK-1-positive and -negative cells, together with the results of immunogold lineage analysis, led to the suggestion (Stern and Canning, 1990) that precursors of the early mesendoderm are mixed with other cells in this tissue before appearance of the primitive streak. Our fate maps cannot provide an independent test of this because the method used labels groups of cells rather than single ones. Future analysis of the descendants of single epiblast cells will be required to provide further insights into this question.

Several authors have described restricted expression of other markers at early stages of chick development (stage X-3). For example the carbohydrates FC10.2 (Loveless et al. 1990) and (NAc-lac)n (Thorpe et.al. 1988), the cytoskeletal proteins vimentin and cytokeratin (Page, 1989) and the homeobox gene goosecoid (Izpisúa-Belmonte et al. 1993) are expressed in certain subsets of cells in the early embryo. In general the investigators concluded, however, that these are markers specific for cell states such as ingression, movement or ‘organizing’ properties rather than for particular fates.

Hensen’s node, the ‘organizer’ of the amniote embryo, has been shown by many studies to contain several distinct cell types: gut endoderm, prechordal plate, notochord/head process, the medial halves of the somite mesoderm and floor plate of the neural tube (e.g. Rosenquist, 1966; Selleck and Stern, 1991; Schoenwolf, 1992; Schoenwolf et al., 1992). Which cell populations of the early embryo give rise to the node? Although we have not analyzed the movements of each region in detail at intermediate stages, our data are consistent with the view that regions containing various presumptive cell types later found in the node may come together at the posterior midline very early, at about stage XII. After this, but still before primitive streak formation (about stage XIV), they move together to a position close to the centre of the blastoderm, where Hensen’s node will eventually form (Spratt, 1946).

In a recent study, Izpisúa-Belmonte et al. (1993) showed that the homeobox gene goosecoid, a marker for organizer cells, is first expressed in a small population of cells in the middle layer of the prestreak stage embryo, associated with Koller’s sickle. Fate mapping with DiI reveals them to be precursors of some of the cells of Hensen’s node, which also express this gene. However, the same study demonstrated that these early goosecoid-expressing cells have the ability to induce others also to express the gene, and the node itself contains more cells than can be accounted for from the early expressing population. Therefore, the node appears to be derived from at least two distinct populations of cells: one group, found in the middle layer around Koller’s sickle from stage X, and others, in the epiblast.

Our fate maps show the locations of the territories occupied by cell types derived from the node in the epiblast between stages X and XIV. In the future, it will be interesting to investigate the onset of inducing ability: is it associated with any particular prospective cell type, or does it require a particular combination of cell types?

This study was funded by a Wellcome Trust Prize Studentship to YH and by grants from the Human Frontier Science Program (held jointly with Drs E. M. De Robertis and P. Gruss) and the Wellcome Trust to C. D. S. We are grateful to Dr A. Stoker for allowing access to his computer, to Dr T. Cunnane for his generous help with printing of computer graphics in colour, and to Mr. G.J. Carlson for skilfull technical assistance. We are also grateful to Drs Kirstie Lawson, Julian Lewis and Jonathan Slack for their helpful comments on the manuscript.