Normal fates and states of specification of different regions in the axolotl gastrula

A fate map shows what each region of an embryo will become in normal development and a specification map shows what each region will become if it is cultured in isolation. If the two maps are identical at the stage in question the embryo is described as a mosaic. If the two maps differ then the embryo is described as regulative and it is supposed that the difference arises because interactions between the parts at a later developmental stage are necessary for their correct specification.

In this paper we report fate mapping and isolation experiments on early gastrulae of the axolotl, Ambystoma mexicanum.

Conceptually, such studies are not new. The classical fate maps for both urodele and anuran embryos were published by Vogt (1929) and a map for the axolotl specifically by Pasteels (1942). Isolation studies on early gastrulae of the axolotl and other species were performed by Holtfreter (1938a, b) and many similar studies have been carried out since then. However a quantum jump in technology has taken place since this earlier work and we feel we have a duty to repeat the most important classical experiments using modern techniques to find whether the interpretations presented in embryological textbooks (e.g. Slack, 1983) are correct or not.

All amphibian embryo fate maps until 1978 were established by the method of vital staining. This is not entirely satisfactory since the dyes have a tendency to spread and fade as development proceeds. In the last few years a number of new cell lineage labels have been introduced (reviewed Slack, 1984a) which cannot be passed between cells and which allow visualization of a single labelled cell surrounded by unlabelled ones, or vice versa. In our experiments we have used orthotopic grafts from embryos uniformly labelled with fluoresceinated-lysinated-dextran (FLDx; Gimlich & Gerhart, 1984) to unlabelled hosts.

All amphibian embryo isolation experiments until 1984 involved in vitro culture of the explants until terminal differentiation had occurred followed by examination of histological sections and scoring for the presence of various differentiated cell types. This is unsatisfactory because the observed events are at the end of a long chain of causality with respect to the events of interest, i.e. specification of regions in the gastrula. It is very unlikely that cells in the early gastrula are being specified as future neurons, fibroblasts, myoblasts or erythrocytes, and much more likely that their commitments at this stage are to geographical regions rather than particular histological cell types. Now however it is possible using a variety of biochemical and immunological techniques to observe the behaviour of explants long before terminal differentiation occurs. In this study we have used the synthesis of high molecular weight glycoproteins as criteria since a previous study showed considerable regional specificity of synthesis patterns of this class of molecule during neurulation (Slack, 1984b).

The results show that a certain revision of the classical fate map is necessary, particularly in relation to the boundary between prospective neural plate and epidermis. In fact the results are rather similar to those previously obtained in this laboratory for Xenopus laevis (Smith & Slack, 1983; Slack, Dale & Smith, 1984) suggesting that urodele and anuran fate maps are not as different as has sometimes been thought. The isolation studies show that the gastrula is less mosaic than that of Xenopus; three different zones can be distinguished giving ectodermal, mesodermal and endodermal type development. Variations in ion balance or addition of particular substances to the medium do not respecify the explants, as has sometimes been claimed (Barth & Barth, 1974; Løvtrup & Perris, 1983), but treatment with vegetalizing factor (Tiedemann, 1976) can mesodermalize ectoderm explants.

Eggs of the axolotl (Ambystoma mexicanurn) were obtained and fertilized in vitro as described by Mohun, Tilly, Mohun & Slack (1980). Embryos were decapsulated and demembranated with fine watchmaker’s forceps. Stages are according to Bordzilovskaya & Detlaff (1979).

Embryos to be used as donors were injected before first cleavage with 70 nl Fluorescein–Lysine–Dextran (FLDx: 100 μg/ml in water) prepared as described by Gimlich & Gerhart (1984). Injections were carried out by using a Burleigh Inchworm to drive a 10 μl pressure-tight syringe connected to a liquid-filled glass micropipette. During and after the injection embryos were kept in 5 % NAM with 5 % w/v Ficoll so as to lower the hydrostatic pressure in the perivitelline fluid (Kirschner & Hara, 1980). NAM is ‘Normal Amphibian Medium’ (Slack, 1984b).

Operations were carried out on stage-10 embryos (onset of gastrulation) using tungsten needles and hair loops. Embryos were transferred to 5 cm plastic petri dishes coated with a 2 mm thick layer of 2 % agar (Difco Noble agar). The dishes contained 50 % NAM in which the calcium concentration was lowered to 120 μM, and to which 1 % w/w Ficoll was added. Although wound healing is relatively slow in low calcium saline, this medium prevents the explants from becoming deformed or curling up (Nakatsuji & Johnson, 1984) and makes manipulation and insertion of the graft easier. After having received the graft the host embryos were put into wells made in the agar and allowed to develop to stage 23–26. Approximately two hours after operation the medium was replaced by 5 % NAM to facilitate normal gastrulation. Note that in dilutions of NAM, only the salts are diluted, the phosphate and antibiotic remaining full strength.

Embryos were fixed in 4 % paraformaldehyde in 70 % PBSA for 18–24 hs at 4 °C, rinsed in 70 % PBSA, dehydrated in an ethanol /butanol series, embedded in paraffin wax (m.p. 56 °C) and sectioned at 10 μm. Rehydrated sections were stained with DAPI (4,6-diamidino-2-phenyl indole, 1 μg/ml) for 10 minutes, rinsed in tap water, dehydrated through ethanols, cleared in xylene and mounted in DPX. The sections were scored using a Zeiss photomicroscope with epifluorescence optics, labelled areas being assessed visually on every tenth section through the length of the specimen. The accuracy of this method was checked by making camera-lucida drawings on graph paper and counting squares and it was found to be accurate to about 10 %. Results will be presented in terms of a normalized anteroposterior distance. So for example ‘0·4–0·6’ means the region from 40–60 % embryo length from the snout.

Parts of early gastrulae were excised with electrolytically sharpened tungsten needles and cultured as explants in NAM until control embryos reached stage 13–14. They were then cut into small pieces to facilitate access of the label, transferred to plastic scintillation vial inserts and incubated with 50μCi tritiated mannose in approximately 0·3 ml NAM until controls reached stage 23–26. After incubation the tissue was rinsed twice with NAM, drained and stored frozen at − 70 °C. There is some relocation of ³H from mannose to amino acids during the labelling period but the principal bands visible on the gels are known to be glycoproteins because they bind specifically to lentil lectin Sepharose.

In cases were the animal pole explants were treated with a substance or with a medium other than NAM, this treatment was given in the initial incubation, i.e. while control embryos were undergoing gastrulation.

Vegetalizing factor was supplied by Prof. H. Tiedemann. It was Fraction E-5 diluted 1:1 with gamma globulin and was wrapped in sandwiches composed of two explants from the animal pole.

Ficoll (Ficoll-400) Pharmacia.

DAPI (4,6-diaminidino-2-phenyl indole dihydrochloride) Boehringer.

2,6-[³H]mannose (>30Ci/mmol) Radiochemical Centre Ltd., Amersham.

Cyclic AMP, free acid, Sigma.

dextran sulphate (M_r 500000) Pharmacia.

micrococcal nuclease (NFCP) Worthington.

DNase (DPFF) Worthington.

RNase (RASE) Worthington.

PMSF (phenyl methyl sulphonyl fluoride) Sigma.

Each sample of tissue was homogenized with 100 μl MNS (0-2M-sucrose, 5mM-Tris-HCl pH 8·8,1 mM CaCh, 25 μg/ml micrococcal nuclease) in a 2 ml Dounce homogenizer with B pestle, rehomogenized with 10/fl DR (lmg/ml DNase, 0·5mg/ml RNase 0·5M-Tris Cl pH 7·0) and rehomogenized again with 0·4ml HMP (0·2M-sucrose, 10 mM-sodium phosphate pH7·5, ImM-CaCl₂, 1 mM-MgCh, 2mM-PMSF). The homogenate was spun at low speed (45 seconds at Mark 1, Gallenkamp bench centrifuge) to remove yolk granules. The supernatant was decanted, the yolk pellet resuspended in another 0·5 ml HMP and spun at low speed once more. Both supernatants were pooled, 0-1 ml was removed for protein estimation by the Folin reaction and to the remaining 0·9 ml supernatant 0·1 ml 10% SDS was added, which dissolves all structures apart from melanin granules. This was then spun for 5 minutes in a microfuge (Beckmann-8700g). For gel electrophoresis the samples were concentrated by dialysing against 0·25 % SDS in 3 mM-Tris-HC1 pH 6-8, lyophilizing and dissolving in 50 μl of 5% β-mer cap toethanol, 10% sucrose and 0·002 % bromophenol blue. This was boiled for 1 minute and microfuged for 4 minutes. 5 μl of the supernant was kept for counting and the remainder stored at − 70 °C until being used for electrophoresis. Both pellets contained little radioactivity (2–3 % of that in the soluble fraction) and are not considered further.

Portions of each sample containing equal counts (6000–15000 c.p.m.), but not exceeding 100 μg protein per track, were loaded onto 4-8% SDS polyacrylamide gradient gels and electrophoresed at constant current until the bromophenol blue reached the bottom. Gels were fixed in 30% methanol and 10% acetic acid, stained with 0·05% Coomassie blue, and fluorographed by the method of Bonner & Laskey (1974).

FLDx-labelled embryos were examined to confirm that all parts became labelled following in j ection. From this it can be concluded that all labelled tissue in the grafted embryos is derived from the injected donor, and all unlabelled tissue from the host.

Orthotopic grafts with labelled donor tissue onto unlabelled recipients were carried out as illustrated in Fig. 1. Both donor and hosts were early gastrulae (stage 10). Healing of the graft was generally good but not all host embryos developed normally. In particular embryos with a dorsal marginal zone (DMZ) graft were prone to abnormal gastrulation apparently because invagination in the graft was not integrated with that of the surrounding tissue. Other embryos sometimes developed wrinkles and deformities of the grafted area which in certain cases led to a severely abnormal appearance. Only embryos that showed a normal external morphology were used for further analysis. Likewise, after examination of the sections, data were taken only from embryos with an internally normal gross anatomy, although slight deviations were tolerated such as small areas of thickened epidermis in the case of AP grafts and minor asymmetries of the axial system in embryos with a DMZ graft. About ten embryos at the head-extension stage (stage 23–26) were examined in each group (AP, DMZ, VMZ, LMZ graft).

Transverse sections through each of the four types of grafted embryo are shown in Fig. 2 to show their general appearance. The detailed results are presented in Figs 3–6 in terms of the average proportion of each structure labelled at each anteroposterior level of the body, and a reconstruction of an archetypal case of each type of graft is shown in Fig. 7.

Ten recipients of AP grafts were examined for distribution of the label throughout the embryo. Most of the label was located in the ventrolateral epidermis of the anterior half (Figs 2,3 and 7 A) endodermal and mesodermal structures being completely unlabelled. In two embryos some label was also found in the anterior-most parts of the forebrain whereas the other eight showed no labelled neural tissue at all. The proportion of labelled neural structures was much less than expected on the basis of the existing fate map of the axolotl gastrula (Pasteels, 1942). According to this the upper boundary of the prospective neural plate extends mediolaterally from the animal pole to the lateral marginal zone, which implies that roughly half of each graft would consist of presumptive neural tissue. Our results indicate that this is not the case. Therefore we suspect that the upper boundary of the prospective neural plate rather runs about halfway between the animal pole and the DMZ, as in Xenopus and other Anura (Keller, 1975).

A certain amount of labelled debris was found in the ventrolateral part of the endoderm in the region 0·1–0·7 along the anteroposterior axis (see Fig. 2F). This debris is likely to have been sloughed off from the edges of the graft into the blastocoel which moves ventrally during gastrulation. Debris can easily be distinguished from tissue by examining the sections under DAPI fluorescence to visualize the cell nuclei.

Taken together our results indicate that the animal pole region gives rise only to epidermis. The two cases in which a small amount of neural tissue is derived from it as well is likely to reflect some variability in accuracy of the operation rather than differences in fate maps between individual embryos.

Ten embryos with a DMZ graft were examined. Most of the label was found in the head and pharynx region (Figs 4, 7B) in particular in the anterior pharyngeal endoderm. It should however be emphasized that pharyngeal endoderm, head mesoderm and prechordal mesoderm form a continuous mass without any clear demarcations between them (Adelmann, 1932). We have called all this tissue ‘endoderm’ since that is quantitatively its major component, but some mesoderm is also included in this region.

Of the structures that arose from the graft the notochord was proportionally the most-heavily labelled. In three embryos the anterior 80% of its length was completely labelled, in the others most of the label was found in its anterior part, gradually declining in the posterior direction. On average the notochord was heavily labelled in the region 0·2–0·9 (Fig. 4). The proportion of labelled somitic mesoderm was moderate, on average from 18% in its anterior region to about 5 % in the middle and posterior regions. But in the three embryos with the heavily labelled notochords, mentioned above, up to 40 % of the anterior somitic mesoderm was labelled. In two of these embryos a small amount of anterior lateroventral mesoderm adjacent to the somitic mesoderm was also labelled.

Summarizing we can say that the DMZ, apart from contributing mainly to anterior endoderm, gives rise to axial mesoderm along the whole length of the body, in particular to notochord and to a lesser extent to somitic mesoderm.

In this group nine embryos were examined. Most of the label was confined to the posterior half of the body (Figs 5, 7C). The lateroventral mesoderm was proportionally the most-heavily labelled with more than 15 % of labelled tissue in the region 0·6–0·8. Label extended throughout a considerable arc of tissue with extensive mixing of labelled and unlabelled cells. So as in Xenopus, the VMZ undergoes substantial dorsal convergence during gastrulation. The posterior endoderm was also heavily labelled with a peak around 0·8.

In three embryos a very small amount (less than 4 % in any region) of somitic mesoderm in the trunk was also labelled. Three other embryos showed a small area of labelled epidermis just anterior to the blastopore. It is likely that in these cases the upper edge of the graft also contained some ectoderm. That this occasionally happens is not surprising since the marginal zone of presumptive mesoderm is narrowest at the ventral side of the gastrula. No label was found in the notochord of any embryo in this group.

Summarizing we can say that a ventral marginal zone graft (VMZ-graft), gives rise to lateroventral mesoderm and posterior endoderm. The former is somewhat anterior to the latter as we would expect from the separation of germ layers which occurs at the ventral blastopore lip during gastrulation.

Nine embryos were studied in this group. Each received one graft to the left or right side taken from the same side of the donor. Although the grafts were unilateral the results are still given as proportions of the whole transverse section, so the figures would be about double if they reflected the contributions from both sides.

The total label was around 5 % of section area from 0·15–0·9 along the anteroposterior axis (Figs 2,6). This is made up of substantial contributions to the somites, the lateral plate mesoderm and the endoderm. The contribution to the notochord came from a single case which may perhaps have been a slightly misplaced graft. The peak contribution to somites is around 0·3, to the lateral plate around 0·5–0·6 and to the endoderm around 0·8, indicating a similar separation between endoderm and mesoderm during invagination as apparent for the VMZ.

In terms both of dorsoventral and anteroposterior contributions the LMZ can thus be regarded as intermediate between DMZ and VMZ.

The same four types of explant, and also explants from the vegetal pole, were cultured in NAM and labelled with pHJmannose as described in Materials and Methods. After labelling the explants were processed for gel electrophoresis and the patterns of glycoprotein synthesis visualized by fluorography. These patterns were compared with each other and with explants from neurulae which were dissected from stage-14 embryos and also labelled until controls reached stage 23–26. It has previously been shown that regions of the neurula show distinctive differences in the synthesis of high molecular weight glycoproteins and that certain bands can serve as markers for epidermis, notochord and total mesoderm. In Fig. 8 tracks 6–10 these controls show that only epidermis makes epimucin (track 7), only notochord makes S2·2 and S3·2 (track 8) and notochord and dorsal mesoderm make more S2 and S6 than other regions.

The first problem was to determine the robustness of this procedure as a test of specification. Many workers refuse to believe in a ‘neutral medium’ which does not have any influence on the developmental pathway of the tissue. However the present procedure allows not only a comparison between in vitro and in vivo behaviour, but also enables every component of the medium to be varied to ensure that it is not having an instructive effect.

In preliminary experiments we found that total radioactive incorporation of [¹⁴C]glucose into TCA-insoluble material was unaffected by omission of K⁺ or Mg²⁺, use of NaCl at × 0·5 or × 1 normal strength and pH of 7·5 or 8·5. Omission of Ca²⁺ caused the explants to disaggregate, and NaCl at × 0·25 or × 1·5 normal depressed incorporation. Examination of the glycoprotein patterns on 4-8 % gels of animal pole explants labelled with pHJmannose showed that the pattern looked the same following all these regimes, even those in which the total incorporation was depressed. We conclude that NAM, and also similar amphibian salines of × 0·5–1 isotonicity such as Holtfreter, Niu-Twitty, de Boer, or Ringer solutions are indeed neutral culture media.

The AP explants always synthesized epimucin visible as a major band (Fig. 8 track 1). Epimucin is an epidermis-specific marker in the neurula and its formation shows that extensive amounts of epidermis arise in these explants. However the explants do not show exactly the same behaviour as neurula epidermis in that they make more of all the other species. This was noted in a previous study (Slack, 1984c) in which it was shown by electron microscopy that, for incubation periods used here, only the outer cell layer differentiates into epidermis.

DMZ explants did not make epimucin but did make amounts of S2 and S6 comparable to neurula notochord. They also made two notochord-specific species S2·2 and S3·2 (Fig. 8 track 2).

LMZ explants showed in all experiments exactly the same synthesis pattern as DMZ explants, including the presence of the S2-2 and 3-2 notochord markers (Fig. 8 track 4). This represents a major disjunction between normal fate and specification since the LMZ did not contribute to notochord in the fate mapping experiments except in a single case.

The VMZ explants showed more variable behaviour in some experiments resembling the other marginal zone explants (as in Fig. 8 track 3) and in others resembling the AP explants. This suggests that what we call ventral marginal zone lies near the junction of two differently specified regions whose boundary may not be completely fixed if mesoderm induction is still going on at the time of explantation. Hence apparently similar pieces of tissue from different batches of embryos may show different behaviours.

Explants of yolky tissue from around the vegetal pole synthesized little in the way of discrete glycoprotein species, most of the radioactivity being in high relative molecular mass poly disperse material (Fig. 8 track 5). The same behaviour is shown by the yolk mass from neurulae (track 10). It is worth emphasizing that the yolk is not metabolically inactive in other respects. When labelled with [³⁵S]methionine it shows numerous labelled bands, which are in fact then the same as for all other regions of the neurula.

In summary, the five types of explant which have been examined show three types of behaviour which may be described as epidermal, notochordal and yolk type in terms of neurula-stage tissue types. It seems reasonable to identify these behaviours with specification for each of the three germ layers: ectoderm, mesoderm and endoderm.

Certain pure chemical substances have been claimed to have a vegetalizing or mesodermalizing effect when applied to isolated ectoderm. Among these are lithium chloride (Masui, 1960, 1962), cyclic nucleotides and negatively charged polymers (Løvtrup & Perris, 1983), as well as inorganic ions (Barth & Barth, 1974). We felt that we should reexamine these claims now that a more objective test of specification is available. Animal pole explants were incubated in various concentrations of LiCl (substituting for NaCl in NAM), of cyclic AMP and of dextran sulphate. They were treated while host embryos were traversing stages 10–14 (gastrulation) and then labelled in exactly the same way as the explants described above. The lithium proved lethal at concentrations of 110 and 55 mM but apart from this none of the treatments made any difference to specification (Fig. 9), all the surviving AP explants behaved like controls and not like marginal zone explants.

According to the work of Nieuwkoop and others (Nieuwkoop, 1969,1973) the first interaction in amphibian development is an induction of an annular mesodermal rudiment from the animal hemisphere under the influence of the vegetal hemisphere. We have made several attempts to duplicate this result by showing a mesodermal glycoprotein synthesis pattern in combinations of animal pole with vegetal pole explants. Our conclusion is that the induction is demonstrable using ectoderm from stage 10 or stage 7, but only just. Evidently it is not possible using this experimental design to suppress epidermal differentiation entirely in the animal pole component. So the combinations always make epimucin. Also the different synthesis patterns described above are largely due to quantitative differences in S2 and S6 and such differences will be less apparent in combinations which are part epidermal and part mesoderm. The only conclusive proof of mesoder-malization is the identification of the notochord specific bands S2·2 and S3·2, and this is not easy against a considerable polydisperse background following sugar labelling. Our results show that in some cases these bands were visible in the AP–VP combinations (Fig. 10) and in others they were not. We feel that this is evidence in favour of the reality of mesodermal induction but it is not decisive.

We have also studied animal pole explants which were treated with ‘vegetalizing factor’ supplied by H. Tiedemann, see Materials and Methods. Tissue from stage-7 and stage-10 embryos was used and in this case the epimucin synthesis was suppressed partially or completely, and the pattern of synthesis was altered to resemble that shown by explants of marginal zone (Fig. 10). Evidently a concentrated factor can produce a more complete transformation than the vegetal pole tissue itself.

The marking experiments presented here are not intended to construct a complete new fate map for the axolotl but rather to map certain regions which are of particular interest in connection with experiments on induction. This is why the animal pole region and parts of the marginal zone were selected for study.

The term ‘marginal zone’ refers to that region of the gastrula destined to in-vaginate actively during gastrulation. It is not of course possible to see the extent of the marginal zone at stage 10 and the results of the fate mapping experiments suggest that our grafts overlapped the vegetal boundary of the marginal zone and included some prospective endoderm.

The use of a high molecular weight lineage labels has several advantages over the previous method of vital staining. The label is sharp and unambiguous and does not fade or spread from cell to cell. Like previous experiments on Xenopus (Jacobson, 1982, 1983; Smith & Slack, 1983) the results show more mixing of labelled with unlabelled cells than was suspected from vital staining. However despite this there is a clear topographic mapping from the head-extension stage back onto the early gastrula stage showing that the movements of gastrulation and neurulation are coherent, with cell mixing occurring only in the short range.

Although all of the cases are included in Figs 3–6, when we analyse the results we ascribe a different significance to the low percentage contributions depending on whether they represent a small contribution in every case, or a large contribution in a minority of cases. The latter results are of course more likely to result from slight variations in the grafting procedure than from variations in fate map between individual embryos. So we conclude that in normal development the region of the animal hemisphere indicated in Fig. 1 contributes only to epidermis, this being located anteroventrally at later stages, and that there is no contribution either to neural plate or to mesoderm.

We conclude that dorsal, lateral and ventral marginal zone map respectively to notochord, somite and lateroventral parts of the mesodermal mantle of later stages.

The results show that a very considerable dorsal convergence occurs, so a VMZ explant of about 50° partial circumference in the stage-10 embryo expands to populate the whole lateral plate/blood island region which occupies about 270° of the circumference at stage 23–26. Likewise a 50° LMZ piece expands to about 90° whereas a 50° DMZ piece contracts to about 30°. The fact that these angles add up to more than 360° indicates the degree of overlap between prospective regions due to interpenetration of cells during the dorsal convergence movements.

If a particular tissue type, such as somite, is considered at a particular transverse section plane, then it is clear that much is unaccounted for. In other words the ordinates of Figs 4–6 do not add up to 100 %. The reasons for this are probably threefold. Firstly there are gaps between the regions mapped totalling perhaps 160° around the circumference. Secondly, there may be prospective somite in the host to the animal pole side of the grafts, although this is perhaps less likely for the ventral grafts among which three cases overlapped the prospective epidermis. Thirdly, it is possible that some cells are lost as a result of damage at the cut edges of both graft and host.

Dorsoventral levels of the marginal zone thus map to different dorsoventral levels of the mesodermal mantle, but they also map to different anteroposterior levels. This is clearly apparent from the Figs 3–6 ‘whole embryo’ histograms: the DMZ populates the anterior half of the body, the LMZ populates most of the body length except the extreme ends and the VMZ populates the posterior half. In this regard there is also a clear slippage between the labelled mesoderm and endoderm: in the DMZ grafts the endoderm is anterior to the mesoderm, and in the LMZ and VMZ grafts the endoderm is posterior to the mesoderm. This is entirely to be expected on the basis of our understanding of the gastrulation movements. The dorsal invagination is an inpushing of both endoderm and mesoderm and results in the formation of the archenteron cavity. It can be visualized as similar to the deformation of a rubber ball by pushing in at one point. However the lateroventral invagination is an anterodorsal migration of mesoderm only between the other germ layers. Since the prospective mesoderm is at least partly on the surface at stage 10 (Smith & Malacinski, 1983) there must be a rupture between endoderm and mesoderm along the line of the lateroventral blastopore lip, and this is in fact clearly visible when axolotl embryos are dissected during late gastrula or early neurula stages.

Perhaps the most remarkable aspect of the results is how similar they are to those from comparable experiments on Xenopus (Smith & Slack, 1983; Slack et al. 1984 and unpublished results). In recent years various authors have highlighted different aspects of the fate maps of anurans and urodeles (e.g. Brun & Garsen, 1984; Smith & Malacinski, 1983) but the topographic mapping of the regions considered here by orthotopic grafting is essentially identical.

The principal respects in which our results are at variance with previous urodele fate maps (Vogt, 1929; Pasteels, 1942; Nakamura, Hayashi & Asashima, 1978) are twofold. Firstly the prospective boundary between epidermis and neural plate cannot run near the animal pole as usually shown. Although we have not attempted to map the neural plate we suspect that the prospective regions will be found in a supraequatorial position on the dorsal side, as in Xenopus and other anura (Keller, 1975). Secondly there is the marked projection of dorsal marginal zone to anterior mantle and ventral marginal zone to posterior mantle. This should perhaps be expected on the basis of our normal understanding of the gastrulation movements but is not emphasized in the published fate maps in which the future anteroposterior axis is considered to correspond approximately with the animal-vegetal axis of the gastrula.

Previous studies of specification have involved long-term culture followed by the scoring of the mixtures of differentiated cells which arise in the explants as ‘ectodermal’, ‘mesodermal’ etc. Although this method has given us much useful information it is subject to various objections: it is difficult to be confident that the culture medium is truly neutral; the long culture period allows secondary and tertiary interactions to occur which may obscure the original specification; explants always contain undifferentiated tissue; and, finally results are normally presented in a form which ignores the proportions of tissues within the explants.

The use of early biosynthetic properties circumvents all these problems: we can ensure the neutrality of the medium; we look only at early events; and the variations within and between explants are automatically averaged when they are homogenized. On the other hand it must be admitted that the regional markers which we presently know about do not allow as clear and qualitative a discrimination between the regions as we would like. The neurula tissues which were studied as controls gave the same results as those reported previously, viz epidermis makes epimucin, all mesodermal explants make a lot of S2 and S6, notochord makes S2·2 and S3·2, and the yolk mass makes mainly high molecular weight poly disperse material. No unique marker, or even special quantitative behaviour was observed in the neural plate. It should be emphasized that in these studies we make no claims about the absolute amounts of each molecular species since we do not know the specific activities of mannose and other precursor substances in vivo. All we claim is that the distribution of label across the bands is characteristic of the different regions.

The results show that three types of behaviour are displayed by the isolated gastrula tissues, and that these approximate to epidermis, notochord and yolk mass in the neurula although in no case show the specific features as clearly as the dissected neurula tissues. It seems reasonable to identify these states of specification with the traditional germ layers: ectoderm, mesoderm and endoderm respectively.

Despite the theoretical shortcomings of the histological method for assessing specification, our results are in fact the same as the classic histological studies of Holtfreter (1938a; Holtfreter & Hamburger, 1955) insofar as he found that the whole marginal zone except for the extreme ventral part could give rise to notochord in explants. Similar results were obtained by Koebke (1977) and Slack (unpublished). In other words the portion of the circumference which gives rise to notochord in isolates (perhaps 270°) greatly exceeds the area which does so in normal development (perhaps 30°). This situation is not found in the early anuran gastrula where the specification and fates of different parts of the marginal zone are at least similar if not identical (Holtfreter, 1938b; Slack & Forman, 1980).

The most plausible explanation is perhaps that in the axolotl at stage 10 there exists a mesodermal rudiment as an annulus of cells around most or all of the equator which has arisen as a result of mesodermal induction and which is not internally regionalized. The difference between dorsal and ventral lies in some slight bias or gradient, perhaps inherited from the fertilized egg, which guarantees that the dorsal extremum will be the first to become determined as a notochord rudiment and that this will then inhibit the appearance of dorsal centres elsewhere. This is the serial diversion theory of Cooke (1982, 1983). The ability of axolotl marginal zone to dorsalize VMZ explants from Xenopus does indeed die off in a graded fashion from dorsal to ventral rather than being sharply localized in the prospective notochord region (Slack & Forman, 1980).

The animal pole, or ectoderm, explants are the usual test tissues for experiments on mesodermal induction. Previous study of the biosynthetic behaviour of these explants (Slack, 1984b) showed no mesoderm-type behaviour on culture in NAM. The present study has made us more confident in the reality of induction by showing that the ectoderm is unaffected by a variety of alterations in the medium and by LiCl, cAMP and dextran sulphate all of which have been claimed from time to time to have mesodermalizing effects (Masui, 1961; Ogi, 1961; Englander & Johnen, 1967; Løvtrup&Perris, 1983; Barth & Barth, 1974). This means that the ectodermal specification of the tissue is quite robust and not easily altered by non-specific stimuli.

By contrast the vegetalizing factor of Tiedemann (Born et al. 1972; Tiedemann, 1976) had a profound effect and we therefore think that this factor should be taken more seriously by the scientific community than it has been in the past. Animal–vegetal combinations following the design of Nieuwkoop (1969) and Nakamura, Takasaki & Ishihara (1971) also show mesodermal induction, although not in all experimental series. We feel here that the negative results are probably due to the fact that only a small proportion of the ectoderm becomes mesoder-malized and that the notochord-specific bands cannot be seen against the considerable poly disperse background obtained after labelling with ³H-sugars.

Our general conclusion from both sets of experiments is that the results obtained by workers in the interwar period are remarkably good considering the technical limitations under which the work was done. The use of more discriminating techniques allows us to make corrections on matters of detail but in general the classical account of early amphibian development remains valid.

We thank Jim Smith and Les Dale for preparing the FLDx.